Department of Linguistics UC Berkeley Title: Conditional and Concessive Clauses in Modern Greek: A Syntactic and Semantic Description Author: Nikiforidou, Vasiliki Publication Date: 1991 Series: Dissertations, Department of Linguistics Permalink: http://escholarship.org/uc/item/98b0s8rq Keywords: Linguistics, Dissertations Copyright Information: All rights reserved unless otherwise indicated. Contact the author or original publisher for any necessary permissions. eScholarship is not the copyright owner for deposited works. Learn more at http://www.escholarship.org/help_copyright.html#reuse
eScholarship provides open access, scholarly publishing services to the University of California and delivers a dynamic research platform to scholars worldwide.
Conditional and Concessive Clauses in Modern Greek: A Syntactic and Semantic Description By
Vasiliki Nikiforidou Graduate of the University of Athens, Greece) 1983 M.A. (University of California) 1985 DISSERTATION Submitted in partial satisfaction of the requirements for the degree of DOCTOR OF PHILOSOPHY in LINGUISTICS in the GRADUATE DIVISION of the UNIVERSITY OF CALIFORNIA at BERKELEY Approved: W
•
Date
Chair
10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C onditional a n d Concessive Clauses in M odern G reek: A sy n ta c tic a n d sem antic description
C o p y rig h t c 1990 Vasiliki Nikiforidou
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Conditional a nd Concessive Clauses in M odern Greek: A sy n ta c tic and s e m a n tic description by Vasiliki N ikiforidou A b s tra c t T h is dissertation exam ines a n u m b e r of types of conditional a nd con cessive clauses in M odern Greek, focusing on the individual constructions as well as on their relationship. I a rg u e t h a t alth o u g h conditionals and concessives are related, th ey are also d istin c t constructions with special sy n ta c tic and s e m a n tic / p rag m a tic properties. T h e conditionals which I look a t a re those in tro d u c ed by an a n d na. I show t h a t na conditionals have a more lim ited d istrib u tio n th a n an condi tionals
and
that
the
c o n s tra in t governing
their distribution
is of a
s e m a n tic /p r a g m a tic n a tu re . I also show t h a t conditionals should be viewed as bi-clausal constructions whose essential properties refer to b o th the a n te c e d e n t and the consequent. The an a n d na conditional clauses are con s titu e n ts of the concessive conditionals in tro d u c ed by na, ke na and akoma ki an.
T h is s tu d y looks also a t these concessive conditionals with an eye
to d e te rm in in g the e x te n t to which th eir properties are derivable from the p ro p e rtie s of their c o n s titu e n t parts. T h e conclusion here is th a t although c o n ditionals and concessives are close se m a n tic relatives, they should still be a n alyzed as distinct constructions w ith p a rtly unp red ic ta b le properties. T h e concessives are also exam ined as a class which includes not only con cessive conditionals b u t also factive (an ke type) concessives and as conces sives, which, I argue, are neither factive n or conditional. Issues raised in th e course of describing these clauses include the in te ra c tio n
betw een
g ra m m a r
(conventionalized
m orpho-syntactic
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 2 -
principles) and general p rag m a tic principles of in terpretation, the distinc tion betw een predictability a nd m o tiv atio n and th e alleged dichotom y between sem antics a n d pragm atics.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
To my parents and Christos
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Acknow ledgem ents Infinite th a n k s are d ue to my c o m m ittee m em b e rs. Eve Sweetser has been a p a tie n t a n d supportive as well as critical a nd clearm inded chair. C h arles Fillm ore, P a u l K ay, George Lakoff an d J o h a n n a Nichols all helped w ith va lu a b le criticism a n d insights from w hich this work has greatly profited.
T h e y also gave m e lots of e n c o u ra g e m e n t a t times when it was
crucially needed a n d their h u m o r and s u p p o rt m a d e the whole thing look less impossible. T h a n k s is also due to Michele E m a n a tia n for com m ents a nd to the people of the W ednesday se m in a r a n d especially Jane Espenson, Adele G oldberg, Jean-P ierre Koenig, L a u r a Michaelis, T o n y Moy, Eric P e d e rso n a n d F re d e rik e V a n der Leek. Finally, I w a n t to th a n k Claudia B ru g m a n for h a v in g always been a s u p p o rt a n d an inspiration. I, of course, a m solely responsible for the c o n te n t of th is work. Special th a n k s are d u e to m y p a re n ts w ho did n o t w ithdraw their s u p p o rt a t a n y tim e and w h o ’ve always believed t h a t language is w orth studying, a n d to m y sister for helping p u t things in to perspective. M arin a m ad e my final s ta y in Berkeley not only possible b u t pleasant and is responsible for th e neatness of m y diagram s. Finally, Christos helped me with a lot of things b u t m ost of all he k e p t me going an d saw me through; he h as m y deepest th a n k s and appreciation. D uring early p a rts of this study, I was s u p p o rte d by the Sloan F o u n d a tio n g r a n t No 86-10-3. T h e In stitu te of Cognitive Studies a t Berkeley provided office space and has always been a stim u la tin g academ ic environ m ent.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
T ab le o f C o n te n ts
A b str a c t D ed ica tio n A c k n o w le d g e m e n ts 1. In tr o d u ctio n ................................................................................................
1
1.0 Aim s of this study; why are conditionals a n d concessives w orth investigating .....................................................................................
1
1.1 O u tlin e of the dissertation ................................................................
3
1.2 C o n stru c tio n G ra m m a r: theoretical b a c k g ro u n d ......................
8
1.3 T h e sem antics-pragm atics dichotom y ..........................................
19
1.4 P re d ic tio n and m otivation ................................................................
24
2. T h e sy n ta x an d sem a n tics o f lexical an d c la u sa l c o n sti tu e n ts o f th e co n cessiv e c o n stru ctio n s ................................................
29
2.0 In tro d u c tio n ...........................................................................................
29
2.1 K E .............................................................................................................
31
2.1.1 C o n textual or lexical "evenness"?
2.2.
.................................
31
2.1.2 Analyses of "even" operators ............................................
33
2.1.3 K E and possible focuses .....................................................
37
NA ..........................................................................................................
41
2.2.1
T h e non-conditional uses ..................................................
41
2.2.1.1 N A in main clause e n v iro n m en ts .....................
41
2.2.1.2 N A as a non-factive s u b o rd in a to r ....................
44
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.2.2 NA as a conditional m a r k e r ................................................
47
2.2.3 Sum m ary a nd Conclusions ................................................
59
2.3 AS ...............................................................................................................
62
2.3.1 Syntactic p r o p e r t i e s ...............................................................
62
2.3.2 S e m a n tic /p ra g m a tic f u n c t i o n s ..........................................
64
3. T h e N A and A S co n cessiv e c o n s tr u c tio n s ....................................
71
3.0 Introduction ............................................................................................
71
3.1 N A concessives .......................................................................................
72
3.1.1 T he "pragm atics only" a c co u n t ........................................
72
3.1.2 NA Concessives .......................................................................
75
3.1.3 The relationship betw een concessives a nd condi tionals ...................................................................................................
88
K E NA concessives............................................................................
93
3.2.1 W ord-order a n d c onstituency considerations ..............
94
3.2.2 K E NA: some s e m a n tic /p r a g m a tic properties ............
97
AS concessives ....................................................................................
103
.........................................................................
103
3.3.2 T he AS sem antics ..................................................................
107
4. C o m p o sitio n , C o m p o sitio n a lity an d M o tiv a tio n ......................
117
4.0 Introduction ............................................................................................
117
3.2.
3.3
3.3.1 AS or KI AS?
4.1 T h e NA a nd AN conditionals a n d concessives: sem antics depending on form .......................................................................................
119
4.2 T h e AS clauses: m o tiv atio n as p a r t of th e sem antic s tr u c tu r e .........................................................................................................
130
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.
C o n d itio n a ls a n d co n cessiv es as b i-clau sal c o n str u c
tio n s ........................................................................................................................
140
5.0 Introduction ............................................................................................
140
5.1 C onditional constructions: problems of analysis ........................
143
5.1.1 Some a p p roaches to conditionals ......................................
143
5.1.2 M odern G reek conditional constructions ......................
146
5.1.3 T h e conditional p a tterns .....................................................
164
5.1.4 C o n d itio n a l p a tte rn s and systemic coherence .............
169
5.2 T he concessive c o n d i t i o n a l s ..............................................................
179
T h e sem a n tics o f c o n cessiv es ...........................................................
185
6.0 Introduction ............................................................................................
185
6.1 T he factive c o n c e s s iv e s .......................................................................
186
6.2 Three types o f concessives ..............................................................
192
6.2.1 Concessiveness via scalarity ................................................
193
6.2.2 "T ru e " concession .................................................................
198
6.2.3 "D irect" concessiveness ......................................................
203
C on clu sion s and a rea s for further research ..............................
209
B ib liog ra p h y ......................................................................................................
217
6.
7.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C h a p te r 1
In tr o d u ctio n
1.0 A im s o f th is stu d y ; w hy a re c o n d itio n a ls and c o n cessiv es w o rth in v e stig a tin g T h is dissertation has two basic goals. F irst, to give an as full a description as possible of M odern G reek c o n d itio n a ls an d concessives. A n d secondly, to advance ou r u n d e rs ta n d in g of th e interrelationship of s y n ta c tic constructions with sem antics a n d p ra g m a tic s. These two goals are in te rre la te d , since it will become clear t h a t the complex of relations betw een th e forms a nd meanings of the different G reek concessive a n d conditional constructions can only be described in a fram ew ork which 1) allows us to express generalizations b o th a t the level of regular se m antic c om positionality and a t the level of u n p re d ic ta b le b u t m otivated form m eaning relationships and, 2) allows us to link b o th p rag m a tic a n d s e m a n tic inform ation directly to (classes of) syntactic constructions as well as to lexical items. C onditionals are a fairly universal c o n s tru c tio n . T hey have alw ays been constructions of special interest, j u d g in g from the a m ount of litera ture which has been devoted to th eir s e m a n tic s b o th from the p o in t of view of the logical tradition and from the p o in t of view of their discourse properties. T h e analysis of conditionals p r e s e n te d here differs from b o th those analyses which descend from the logical tra d itio n and those which see conditionals as serving uniform ly a discourse function. I will be arguing t h a t alth o u g h there m ay be such a universal fu n ction associated w ith con ditionality, conditionals should be bro k en dow n into distinct constructions
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-2w ith v a ry in g form al a nd sem antic properties.
Different languages m a y
v a ry in th e w ay they express different kin d s of conditional m eanings a n d one p u rp o se o f this thesis is to discover w h a t kin d s of associations betw een m e a n in g a n d form Greek makes in its conditional p a tte rn s . Concessive clauses, unlike th e ir "neighbor" th e conditionals, have never been stu d ie d in their own rig h t as a class1. G reek has an extrem ely rich s y s te m of ways to express concessiveness corresponding to d istinct c o n stru c tio n s w ith special properties. A s tu d y of ( a t least) the concessive c o n ditionals requires reference to conditionals w ith which th ey s h a re m a n y of th e ir properties. Conditionals a n d concessives can, therefore, be said to fo rm a n a t u r a l class. In conditionals an d concessives we have a com plex b u t in te rre la te d set of constructions, constituting an excellent te s t case for a n u m b e r of issues o f theoretical interest. T he refined d istinctions we need to m ak e in o rd e r to give an a d e quate description of such clauses will d e m o n s tra te the need to include sem antic and p ra g m a tic factors as an explicit p a r t of the s y n ta c tic c o n structional analysis.
T h is in t u r n will argue for th e need to
recognize th e construction (i.e. a pairing of m eaning a n d form ) as the re le v a n t u n it of description.
M oreover, th e kind of sem antic description
t h a t we need to give to conditionals a nd concessives directly addresses the issue o f th e sem antics versus prag m a tic s distinction a nd argues in favor of abolishing a strict dichotom y.
Finally, c o n d itional a n d concessive con
stru c tio n s raise the question of exactly how m e a n in g composes a n d p e rta in d irectly to t h e issue of compositionality. M y conclusion in this a re a will be t h a t a s tric tly compositional analysis will be very h a rd to m a in ta in .
On
1 W ith the exception of K onig’s (1988) paper.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-3th e other hand, we could very profitably ta lk a b o u t com positional m ean ing with respect to
a p a rtic u la r s tru c tu re as well as a b o u t th e sem antic
m otivation underlying p a rtic u la r constructions.
C o n s tru c tio n G ra m m a r,
th e theoretical fra m ew o rk assum ed in this dissertation, allows th e expres sion of b o th these k in d s of generalization. All of these issues will be taken u p a nd discussed m ore extensively in the following sections.
1.1
O u tlin e o f th e d iss e r ta tio n This dissertation
is concerned w ith certain types o f conditional and
concessive clauses of M o d e rn G reek. T he types of conditional sentences to be discussed are exemplified in (l)-(3) below: (1) na
imun
onasis
0a
‘if’ b e -IM P F -P -lS G
to a ^ o ra z a
Onassis F U T it b u y - IM P F - P - lS G
"If I were Onassis, I would b u y it". (2) as
meletisi
ke 0a
‘if’ s tu d y -P E R F -N P -3 S G
perasi a n d F U T p a ss-P E R F -N P -3 S G
" If he studies, he will pass". (3)
an ixe er0i o j a n is
0a ixe er0i ke i m aria
‘if’ c o m e -P E R F E C T -P -3 S G th e John F U T
c o m e -P E R F E C T -P -3 S G
a n d the M ary "If Jo h n h ad come, M a ry would have come also". Na, as2 and an a re different conditional m arkers. A lth o u g h they were all translated by
"if", th e adverbial clauses they in tro d u c e are each
2 S a and a# have many o th er functions besides their conditional one and these will be also discussed in the following chapters. T he conditional and the concessive uses are, however, the m ain focus of this study.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-4characterized by different sem antic a n d p ra g m a tic properties. P a r t of this s tu d y , therefore, will be devoted to describing these properties. T h e concessive clauses which will be ex a m in ed are exemplified in (4)(3) below: (4) onasis
n a im un
Sen 0a boru sa na to a y o ra so
Onassis "na" b e -P -lS G
N E G F U T c a n -I M P F - P -lS G subj. it buy
" E ven if I were Onassis, I w o u ld n ’t be able to b u y it". (5) ke onasis
n a im un
a n d Onassis
Sen 0a boru sa n a to ay o ra so
"na" b e -P -lS G
NEG
FUT
can-E M PF -P-lSG
subj. it
buy "Even if I were Onassis, I w o u ld n ’t be able to b u y it". (6) a k o m a ki
an
er0i
o janis
emis Sen 0 a pam e
"even" a n d if c o m e -P E R F -N P -3S G the J o h n
we
NEG
FUT
go-
PER F-N P-1PL "Even if John comes, we w o n ’t go". (7)
as vreksi "as"
emis 0a pam e
rain -P E R F -N P -3 S G
we F U T
g o - P E R F -N P -I P L
" W e ’ll go even if it rains / alth o u g h it m a y rain". (8)
pije
an ke
Sen to n kalesan
g o -P E R F -P -3S G ‘if’ ‘a n d ’ N E G him in v ite-P E R F -P -3 P L "He w ent even tho u g h th ey did n o t invite him ". Besides discovering a nd describing the sy n ta c tic a nd sem antic / prag m atic p ro p erties of these adverbial causes, th e p u rpose of this dissertation will be to investigate th e relationship of these clauses to the conditional clauses
of
(l)-(3).
I will
be
arguing
th at
a lth o u g h
concessives a n d
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 5c o n ditionals are und o u b ted ly related constructions (b oth syntactically a n d s em antically), they are also distinct in ways t h a t c a n n o t be p re d ic te d by a n y general principles. As such, th ey should be tre a te d as d istin c t con s tr u c tio n s with special properties. T h e te rm "construction" will be used in its in tuitive sense to refer to f o rm -m e a n in g pairings, m u c h in the way t h a t it has been used in t r a d i tio n a l g ram m ars. "C onstruction", however, will be also used in its m ore te c h n ic a l sense with the p a rtic u la r co n te n t t h a t it has w ithin the th eo re ti cal fra m ew o rk of C onstruction G ra m m a r. C o n s tru c tio n G r a m m a r is a d e scriptive a n d theoretical fram ew ork c u rre n tly being developed in th e Linguistics D e p a rtm e n t of the U niversity of C alifornia a t Berkeley. W ith in th is fram ew ork the term "construction" also refers to pairings of p a rtic u lar form s w ith particular sem antics. O ne of th e basic tenets of the th eo ry is t h a t th e construction is the basic u n it of description an d t h a t it encom passes sim ultaneously syntactic an d s e m a n tic /p r a g m a tic inform ation. T h is dissertation is a case s tu d y intended to illustrate this a nd oth er t e n e ts o f C onstruction G r a m m a r. O n the o th e r h a n d , it is also in te n d e d to a rg u e for some of the theoretical constructs a n d assum ptions of C o n s tru c tio n G r a m m a r. T h a t is, I will s t a r t b y assum ing t h a t s tru c tu re s like the ones exemplified in (l)-(8) are constructions, a nd therefore, t h a t their d e scrip tio n will have to include b o th syntactic an d s e m a n tic /p r a g m a tic in fo rm a tio n .
But, I will also argue t h a t in som e cases we are required to
view th e m as such. In th e rem ainder of this section, I outline th e s tru c tu re of the disser t a tio n .
In th e next section, I outline briefly som e of th e basic assu m p tio n s
a n d tools of Construction G ra m m a r. Finally, in the two subsequent sec tions, I will take up two them es which recur in the course of this study:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-6th e distinction betw een sem antics a n d p rag m a tic s, a n d th e issue of se m a n tic m otivation. In c h a p te r 2, I discuss th e syntactic a n d se m antic properties of the con stitu e n ts w hich m a k e up the concessive c onstructions. T h e im p o rta n c e of describing these in d ep e n d e n tly lies in th e fa c t t h a t this will e nable us to decide late r on how com positional th e concessive constructions are. Some of th e c o n stituents are lexical, as, for exam ple, the in tro d u c to ry connective ke.
O thers, however, are clausal, t h a t is th e y are them selves adverbial
clauses.
One such clausal c o n stitu e n t is the conditional clause introduced
by na, which p a rtic ip a te s in the na a nd the ke na concessive constructions. A n other clausal c o n s titu e n t is th e type of clause in tro d u c ed by as which also serves to in tro d u c e concessive clauses. T h is c h a p te r, therefore, will be mostly concerned w ith describing the sy n ta c tic a n d s e m a n tic /p r a g m a tic properties of these con stitu e n ts outside the concessive e n vironm ent. C h a p te r 3 exam ines three types of concessive clauses introduced by na, ke nas a nd as. T h e description here will focus on those aspects of their syntax a n d sem antics which are n o t derivable a n d p redictable from th e properties of their c o n s titu e n t p a rts and, therefore, m u s t be a tt r ib u t e d to the concessive constructions as a whole. W e can th in k of this ch a p te r, therefore, as com prising the constructional level of description.
Besides
giving an as full a description as possible, th is c h a p te r aims also to d raw a distinction betw een degrees of com positionality a n d required co n stru c tional description, as illustrated by th e na a n d ke na concessives, on the one h a n d , a nd the as concessives on the other. 3 I will also look briefly a t th e akoma Jfci an construction, especially as it relates to the other two concessive conditionals.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-7C h a p te r 4 addresses some of th e theoretical issues which arised in the previous tw o c h a p te rs. In particu la r, I discuss th e issue of th e sem antic composition of th e c o nstituents from th e p o in t of view of th e d o g m a of s trict com positionality a n d t h a t of se m a n tic m o tiv atio n . I first show t h a t the G reek d a t a resist an analysis on a "rule-to-rule" basis. I also discuss the issue of s e m a n tic pred ic ta b ility versus m o tiv a tio n w ith respect to th e as c onstruction, m o tiv atio n
for
arg u in g t h a t
in some cases, exploring th e
a given construction
sem antic
helps to w a rd s a more a d e q u ate
description of its se m a n tic properties. C h a p te r 5 will be concerned with w h a t v/e m a y th in k of as discourse properties of conditionals and concessives. T h e linguistic context to be sys tem atically e x a m in e d here will be lim ited to th e im m ediate e m b e d d in g e n v iro n m en t of these adverbial clauses which is, of course, the m ain clause t h a t th ey m odify. T his chapter, therefore, will be concerned w ith form al a nd sem antic p roperties which have to do w ith b o th the an te c e d en t a n d the consequent of conditionals an d concessive conditionals and which have to be s ta te d a t th e cross-clausal level (e.g., dependencies of the verb tense in th e a n te c e d e n t an d th e consequent). A t this level, the constructions include th e whole sentence, t h a t is b o th a n te c e d e n t and consequent. T h e discussion will be c o n d ucted in term s of th e an conditionals, th e assu m p tion being t h a t
th e kind of cross-clausal p roperties discussed here also
apply to th e o th e r , m ore restricted in d istrib u tio n , types of conditional clauses (e.g. ones in troduced by na). C h a p te r 6 will exam ine th e sem antics of th e concessive clauses as a class. T h e first section will give a brief description of the sy n ta c tic and sem antic p ro p ertie s of th e factive (an ke type) concessives. In th e rest of the c h a p te r, I arg u e t h a t based on the kinds of concessive constructions
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-8discussed in the previous c h a p te rs, we should distinguish th ree types of concessive semantics, i.e. three ways in which the " p -> ( im p lie s ) - q H aspect of conventional m eaning m ay be said to arise from the co n trib u tio n s of different pragm atic a n d se m a n tic factors. Finally, c h a p te r 7, will be th e concluding c hapter, in which, besides reviewing briefly m y results, I will touch upon unresolved questions and are a s in which th e p rese n t s tu d y could be profitably extended. A final note is in o rd e r here w ith respect to the d a t a in this thesis. I h a v e m ade an effort to use as m a n y atte ste d exam ples as possible, col lected from b o th oral a n d w ritte n discourse. These are m a rk e d in the tex t b y single quotes. I h a v e n ’t been entirely successful in using only atte ste d exam ples, b u t I have trie d to use th e m in cases w here ju d g e m e n ts of g ram m atic a lity tended to be very fine a nd controversial.
1 .2 C o n s t r u c t i o n G r a m m a r : t h e o r e t i c a l b a c k g r o u n d T h e theoretical a n d descriptive fram ew ork assum ed in this disserta tion is t h a t of C o n stru c tio n G r a m m a r. I will be assum ing this fram ew ork in asm u ch as, for exam ple, I will presum e t h a t it is necessary to provide sim ultaneously both a sy n ta c tic a nd a sem antic description of the adver bial clauses t h a t are th e topic of this dissertation. B u t I will also be a rg u ing for the fram ew ork by presenting evidence t h a t a full a nd adequate description of these clauses requires reference to sem antic a n d p rag m a tic conditions. Construction G r a m m a r is c u rre n tly being developed a t the D e p a rt m e n t of Linguistic of th e U niversity of California a t Berkeley.
Fillm ore
(1988, 1989), Lakoff (1987), Fillm ore and K ay (1987) a n d Fillmore, K a y a n d 0 ’ Connor (1988) are i m p o r ta n t case studies within the C onstruction
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-9G ra m m a r
fram ew ork,
which
also
contain
explicit
discussions
of
its
assum ptions a nd consequences. M u c h work on specific g ra m m a tic a l topics has been done w ithin the C o n s tru c tio n G r a m m a r on specific g ram m atical areas. L a m b re c h t (1986a) a n d (1986b), B ru g m a n (1988), G oldberg (1989), Fried (1990) can be considered rep re sen ta tiv e examples. I will n o t a t t e m p t to com p a re C o n stru c tio n G r a m m a r fully a n d sys tem atically w ith other, older or c u rre n t, fram ew orks of gram m atic a l description. It seems to me, how ever, t h a t a clear a nd s h a rp distinction can be draw n betw een C o n s tru c tio n G r a m m a r a nd GB in t h a t th e latte r requires th a t all g ra m m a tic a l p h e n o m e n a (or all w hich are w o rth y of description) be u ltim a te ly reducible to som e general universal principles. Construction G ra m m a r, on the o t h e r h a n d , first recognizes all the p oten tial gram m atical idiosyncracies t h a t a given language m a y present w ithout advancing a ny claims a b o u t th e universality of the rele v an t descriptive param eters. C o n stru c tio n G r a m m a r is closer to theoretical fram ew orks such as L FG (Bresnan 1979, 1982), H P S G (P ollard a n d Sag 1987) and Cognitive G r a m m a r (Langacker 1987, forthcom ing) t h a n to GB; it shares with all this latte r g roup some assu m p tio n s a b o u t the necessary associa tion of syntax an d sem antics. Y e t, besides th e differences in notation, there are also i m p o r ta n t differences in some of the fu n d a m e n ta l assum p tions about language, w hich each of these theories m akes. Cognitive G ra m m a r, for exam ple, does not recognize a form al level of sy n ta x exist ing in its own right; ra th e r, s y n ta x is an epiphenom enon, th e only cogni tively real levels being sem antics a n d phonology. C onstruction G r a m m a r, on t h e oth er h a n d , recognizes the existence of a syntactic level as such, alth o u g h this level is necessarily associated with a level of sem antic description. N either level, however, is given
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 10 priority. In this respect, C o n stru c tio n G r a m m a r is highly rem iniscent of th e a p p ro a c h ta k e n by m ost tra d itio n a l g ra m m a rs , w here labels including b o th m orp h o -sy n ta c tic and se m a n tic in fo rm a tio n , (for example, "genitive of possession" or "ablative of s e p a ra tio n " ), are com m on. C o n stru c tio n G r a m m a r, however, sees the p a irin g of sy n ta c tic a nd sem antic in fo rm a tio n as a theoretical prerequisite, w ith all its necessary implications. W ith respect to language acquisition, for exam ple, the claim is t h a t sy n ta c tic a n d sem antic aspects are acquired sim u ltaneously by the child a n d t h a t linguistic inform ation stored in th e m in d includes syntactic and se m a n tic in fo rm a tio n a t th e same site. S y n ta ctic a n d sem antic inform ation a re by definition,
therefore,
included
in th e
description of any g ra m m a tic a l
phe n o m e n o n a n d this is w h a t basically differentiates C onstruction G r a m m a r from any theory which e ith e r elim inates the sem antic side of the description or tre a ts syntax a n d s e m a n tic s as com pletely s e p a ra te and a u to n o m o u s levels of analysis. W h a t constitutes a se m antic d e scrip tio n m ay also be a m a t t e r of some controversy. I will have m ore to say on this subject, in the following section on the s e m a n tic s /p ra g m a tic s distinction.
Here, it is sufficient to
note t h a t C o n struction G r a m m a r pays equal a tte n tio n to the sem antic and sy n ta c tic aspects of the description. T h e extensive research on fram e sem antics as well as m ore recent d ev elo p m en ts in sem antic theory as, for exam ple, the idea of p rototypically s tr u c tu r e d sem antic categories (Rosch 1978, 1981 a nd others) are ta k e n into a c co u n t. T h e litera tu re on th e role of pro to ty p e s tr u c tu r e and se m a n tic fram es in the analysis of linguistic m eaning has am p ly d e m o n s tra te d th e in adequacy of the necessary-andsufficient conditions approach, a n d C o n s tru c tio n G r a m m a r incorporates th e results of such enriched s e m a n tic analysis in th e description of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-11 se m a n tic pole.
In some respects, C o n s tru c tio n G r a m m a r can be con
sidered to be an o u tg ro w th of th e sem antic research on fram es.
M ost
im p o r ta n tly for o ur purposes (since this dissertation is concerned w ith clausal constructions), fram es can be associated w ith whole constructions, as well as w ith specific lexical items.
M oreover, as b o th conditional an d
concessive clauses are them selves c o n stitu e n ts of larger bi-clausal (adver bial + m ain clause) constructions, se m antic description will need to be given for the non com positional aspects of these larger c o n structions as well; t h a t description m akes reference to p ra g m a tic factors a n d it is h a rd to conceive how it could be m ade in te rm s of se m antic features. A n im p o r ta n t difference betw een C o n s tru c tio n G r a m m a r a n d oth er c u rr e n t theories (especially GB) is t h a t th e form er p u ts no restrictions on the permissible range of d a t a for investigation.
Indeed, it seems to be a
distinctive fea tu re of C o n struction G r a m m a r t h a t it pays equal a tte n tio n to th e productive an d the idiosyncratic aspects of a language; r a t h e r th a n draw ing a s h a rp distinction betw een th e two, C o n struction G r a m m a r m a in ta in s t h a t th ere is a c o n tin u u m from th e fully productive to th e com pletely idiosyncratic, recognizing t h a t m an y linguistic p h e n o m e n a can p a r ta k e of both characterizations. All p h e n o m e n a of a given language deserve, therefore, some kin d of description in th e g r a m m a r of t h a t language. T his should n o t be rea d as denying the obvious differences t h a t exist betw een "core" and "periphery" p h e n o m e n a o r betw een universal a n d language-specific ones. It sim ply m eans t h a t "periphery" a nd language-specific p h e n o m e n a are as i m p o r ta n t as those which m ay prove to be universal. T h e a ssum ption a nd conviction underlying th is choice is t h a t only by paying a tte n tio n to some complex stru c tu re s (which in some theories would occupy the periphery) a n d to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 12 stru c tu re s which m a y be considered (sem i)idiom atic, can we arrive a t an u n d e rs ta n d in g of w h a t a complete account of a g r a m m a r of a language m ig h t require. Fillm ore and K a y (1987) devote p a rt of th e ir discussion to the issue of of idiom aticity a n d propose t h a t we should recognize an idiom aticity cline. A t one e nd of the c o n tin u u m , we find constructions w ith no specific lexical inform ation associated w ith them . A t the o th e r e nd of the continuum , there are constructions which have all their lexical in fo rm a tio n conven tionally specified a n d whose overall m eaning ha s to be independently stated; we m ay t h in k of these as the " re a l11 idioms. Between these two extrem es we find va rio u s possibilities, one of w hich is constructions con taining positions w hich m ay be filled by an y s o rt of filler, the overall in te rp re ta tio n being p red ictably uniform. T h a t is, such a construction does n o t have all of its lexical content specified; b u t th e c onstruction as a whole m u st be associated w ith a p articular constructional sem antics, which then combines relatively regularly or compositionally w ith a sem antics of any slot-fillers.
These exam ples are central to m y analysis, since p a rt of the
p o in t to be m ade w ith respect to Greek conditional a n d concessive con s tructions is t h a t th e m eaning is often associated w ith a c e rtain construc tional p a tt e r n r a th e r t h a n with a specific w ord o r words. Fillm ore an d K a y fu rth er claim t h a t th e n u m b e r of constructions which are purely s y n ta c tic is relatively small. P u r e ly syntactic construc tions would be constructions which have no specific sem antic o r pragm atic inform ation associated w ith th em and which do n o t include any specific lexical items. Such a construction in English, a n d for t h a t m a t te r in G reek as well, is the su b je c t-p red ic a te construction (this is considered to be a very general co n stru ctio n which underlies bo th th e passive a n d the active
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 13 constructions).
W hile in o ther languages it m a y be necessary to provide
very specific inform ation as to w hat kind of things can a p p e ar as su b je c ts, b oth English and Greek are p re tty m u c h free in w h a t sem antic roles th ey allow as subjects.
N o other specific in fo rm a tio n need be associated w ith
such a construction, which can therefore be considered a purely sy n ta c tic one. Such constructions, however, are relatively rare; an d this necessarily highlights th e im portance of other p a rtia lly filled a n d / o r p a rtia lly con stra in e d constructional p a tte rn s, w here s e m a n tic /p r a g m a tic restrictions play a central role. T h e notion of th e construction will figure p ro m inently in th e descrip tion of th e adverbial clauses which are th e topic of this thesis, a n d it will do so a t tw o levels. First, we have to id entify constructions (w ith special syntactic a n d sem antic properties) a t th e level of th e adverbial clause itself. These constructions can be roughly identified by the type of connec tive by which they are introduced. It is also necessary, however, t o recog— r:
lil£C
tpM
^
laigei
C uuanuC tiuua
w nnu
AM
cuuipiiac
p. i I*
uutu
^ U /\
A/I IfQ o l\ 1A1 AM/]
o n e clu. v Cl u i a i a i m
4
A T V *At M
one m a in
clause. T hese are identified b o th by th e fo rm of the in tro d u c to ry connec tive a n d by th e verbal m orphology (specifically, aspect an d tense m ark in g ) in th e a n te cedent a n d the consequent. T h a t is, w h a t identifies these biclausal constructions is not th e lexical fillers of th e verbal slots in th e two clauses, b u t ra th e r th e tense a n d aspect m orphology on the verb. As I noted above, traditional g r a m m a r s implicitly rely on th e idea t h a t g ra m m a tic a l description contains b o th sy n ta c tic a n d se m a n tic infor m ation. In m odern linguistics too, th e idea t h a t the form -m eaning pairing should be ta k e n as th e basic unit for th e linguistic description is h a rd ly new. S aussure’s (1915) definition of th e sign is a definition o f formm eaning pairing and one m ight be te m p te d to say t h a t a co n stru c tio n is
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 14 like a m o rp h em e, only bigger. T his "extension", however, is new and would go, I believe, c o n tra ry to S a u s s u re ’s im plication t h a t the sign is confined to the m orphem e level. As n o te d before, constructions can be recognized a t m ultiple levels and c o n stru c tio n s can contain o th e r c onstruc tions. As a result of this, the m eaning pole of a construction can be either a n y k in d of p rag m a tic or sem antic in fo rm a tio n ("whole fram e" k in d of in fo rm a tio n included) or, as in the case o f th e sub je c t-p red ic a te c onstruc tion discussed above, it can be a fairly a b s tr a c t specification to the effect t h a t th e sem antics of this construction is to predicate a p ro p e rty or a rela tion of t h e su b je c t referent. C on stru c tio n s are, therefore,
given a p ro m in e n t place in C o n s tru c
tion G r a m m a r. T h e y are a b s tr a c t u n its of description, consisting of an association of form al and sem antic pro p ertie s. W h a t in older tra n s fo r m a tional fram ew orks was the o u t p u t of p h ra s e s tru c tu re a nd tra n s fo rm a tional rules, in C onstruction G r a m m a r c a n be considered to correspond to descriptions of constructions, which, how ever, lack associated sem antic d escription (at least under the "in te rp re tiv e" theoretical model -Jackendoff 1972). A t the o th e r pole, G enerative S e m an tic s (cf. Lakoff 1963, 1970) took s e m a n tic units to be the relevant units o f description. C onstruction G r a m m a r , on th e oth er hand, takes th e c o n s tru c tio n per se to be the central u n it of description. It follows, t h a t in c o n tr a s t to the older tra n s fo r m a tional fram ew orks as well as to GB, C o n s tru c tio n G ra m m a r is m o n o stra tal; a construction comprises all th e s y n ta c tic inform ation w hich used to be assigned to tw o (or more) distinct levels as well as the sem antic infor m a tio n w hich in these theories re p re sen te d y et a n o th e r level. M orphem es a n d lexical items a re them selves "lexically filled" con s tru c tio n s,
characterized
by
"valence
descriptions"
which
comprise
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 15 a b b re v iate d descriptions of th eir c o m b inatorial possibilities.
A valence
description of a verb, e.g., "is a c h a ra c te riz a tio n of the linguistic entities whose presence is required (or p e rm itte d ) w hen it is used in a p a rtic u la r m eaning ..." (Fillm ore 1989:65).
Such valence descriptions co n tain the
n u m b e r of the ‘p a r t i c i p a n ts ’ w hich ac co m p a n y a given lexical item a n d the gram m atic a l fu n ctio n a n d the sem antic role of each such p a rtic ip a n t. They
m ay
also
co n tain
o th e r
specifications
of
a
form al
or
s e m a n tic /p r a g m a tic n a tu r e w here necessary. As an illustration, let us con sider the valence descriptions of th e verb afino ( = "let, allow"), w hich will figure in su bsequent discussions concerning
th e origin of th e as connec
tive. W e m u st ac co u n t for the different a rg u m e n t s tr u c tu r e s of afino as th ey ap p e ar in (9)-( 11) below: (9)
o janis
afise
tin
m aria
the-N O M J o h n -N O M let-P E R F -P -3 S G
the-A C C M ary-A C C
" Jo h n let go o f / left M ary". (10)
o janis
afise
tin
m a ria n a fiji
the-N O M Jo h n -N O M let-P E R F -P -3 S G
the-A C C
M ary -A C C ‘n a ’ leave-
P E R F -N P -3 S G "John let M a ry go". (11)
o janis
afise n a
fiji
the-N O M Jo h n -N O M let-P E R F -P -3 S G
i
m a r ia
‘n a ’ lea v e -P E R F -N P -3 S G the-
N O M M ary-N O M "Jo h n let M a ry go". Exam ples ( 9 ) - ( ll) correspond to th ree d istinct valence descriptions of th e verb, given as (9’)-(l 1’) below, in F illm ore’s (1989) "box" no tatio n .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 16 -
O')
lexeme "AFINO" ca tV
min+
1
2
Agt
Pat
^n o m
( 10 ')
^acc
lexeme "AFINO" catV
min+
1
2
Agt
—
^NOM
^ ACC
C Cont
V* 2(1)
( 11 ')
lexeme "AFINO" catV
min+
1
C
Agt
Cont
N NOM _,
V na
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 17 T h e top row in (9’), (10’) and (11’) indicates th e "lexeme" (i.e. the ab stra ctio n over all th e possible morphological realizations of a given verb) with which th e p a rtic u la r valence descriptions are associated.
T h e deci
sion to m ake this ho m o p h o n o u s with the first person singular, present tense fo rm of the verb is completely a rb itra ry , sim ply following th e gen eral tre n d for Greek to list t h a t as the d e fa u lt form of th e verb (e.g. in dictionary entries). T h e specifications in th e second row indicate respec tively t h a t th e category of th e lexeme is verbal a n d t h a t it is -(-minimal, t h a t is it represents th e m a jo r lexical category w ithin th e verbal ("V") category. Below th e specification of V, the third row of entries in (9’), (10’) and (11’) c ontains in form ation a b o u t the g ram m atic a l roles (subject, object etc.) served by the c o m plem ents of afino. "1" s ta n d s for the a rg u m e n t which is the su b je c t (note t h a t in Greek, we d o n ’t need to have a n overt N P realizing t h a t a rg u m e n t), while "2" rep resen ts the direct object. Finally, C s ta n d s for " other" complement. T h e n e x t row identifies the sem antic role associated w ith each a rg u m e n t, and any o ther sem antic inform ation which m ay be relevant to the kinds of entities which can serve as a rg u m e n ts. T h e se m a n tic role of the s u b je c t in all the valences of this verb is t h a t of "A gent", while the direct o b je c t in ( 9 ’) is the P a t ie n t. In (10’), th e " -- " in th e place of the semantic role assignm ent indicates th a t the verb in its use as a three-place predicate does n o t assign a sem antic role to its syntactic object. Som e evidence for this comes from th e possibil ity of having idiom c h u n k s filling th a t slot, t h e a rg u m e n t being th e same as t h a t given for th e d istin ctio n between "raising" a n d "equi" stru c tu re s (all the caveats which n o rm ally apply to this a r g u m e n t apply here as well; cf. M cCawley 1988). C onsider for example, (12):
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 18 (12)
afise ton kom bo n a ftasi sto xteni let-P E R F -P -3S G the k n o t ‘n a ’ re a c h -P E R F -P -3 S G
to -th e com b
"He let his back get to the wall". Finally, the sem antic role of the th ird c o m p le m e n t is to express th e "con te n t" of the act of allowing. T h e th ird row in th e valence descriptions
provides inform ation on
the g ram m atical form w hich th e constituents in s ta n tia tin g the a rg u m e n ts m u s t have. In all valences, th e su b je c t will be in sta n tia te d by a noun p h rase (N being here a sufficient conventional a b b re v iatio n for a m axim al noun-phrase, since all ph rasa l con stitu e n ts t h a t a p p e a r as com plem ents are necessarily m aximal), a n d t h a t noun phase will have case N om inative. T h e direct object in (10’) will also be realized as a no u n phrase, w hich however, will have case Accusative. T h e th ir d com plem ent in (10’) a n d (11’) will be realized as a clause ( V = m a x i m a l p h ra s e of verbal type) intro duced by na. Finally, the 2(1) index in (10’) (whose operation is illustrated by exam ple (10)), is a "co-instantiation" or "control" index which tells us t h a t the [the-ACC m aria-A C C ] represents sim ultaneously the o b jec t com p lem ent of afino and the s u b je c t com plem ent of /etryo. T he description of these valences contains in form ation which m ay be th o u g h t of as re d u n d a n t, in t h a t it m ay be p red ic ta b le from general p r in ciples. W e m ay not need to specify, for exam ple, t h a t th e case of the s u b je c t will be nom inative no r t h a t the sem antic role associated w ith th e s u b jec t will be t h a t of an agent; if there is an a g e n t a rg u m e n t, we m a y pred ic t (by the sem antic role hierarchy) t h a t it will show up as the s u b ject. These valence descriptions, therefore, can be th o u g h t of as re p re sen t ing re d u n d a n tly the o u t p u t of such general o p e ra tio n s as well.
W h a t is
im p o r ta n t is t h a t by following the (relevant) valence descriptions in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 19 parsing each of these sentences, we can decide w h ether th e y satisfy th e req u irem e n ts imposed b y the valence, in which case th e sentence will be a g ra m m a tic a l sentence of the language, or not, in which case th e se n te n c e will be u n g ram m atic a l. T h e preceding discussion w as oversimplified. M uch m ore needs to be said if we are to fully ju stify th e theoretical an d n o tatio n a l choices assum ed even by this lim ited p re s e n ta tio n of data'*.
It should be n o te d
t h a t m u ltip le valence descriptions associated w ith one lexical item are n o t a h a n d ic a p of the th e o ry , nor do th e y rep re sen t inability to c a p tu r e gen eralizations, since it is possible t o t r e a t c e rtain kinds of valence v a ria b ility (e.g. th e one represented by valences (10’) a n d (11’)) by m ea n s of lexical rules rela tin g valences to each o th e r in a regular fashion (cf. F illm o re 1989). T h is possibility will n o t be explored here for Greek. I sim ply wish to give a n idea of w h a t a valence including b o th sem antic a n d s y n ta c tic in fo rm a tio n would look like. F o r m ore com plex constructions such as th e ones to be exam ined in this d issertation, even a m inim al specification of th e associated s e m a n tic /p r a g m a tic in form ation m ay be m uch m ore com plex t h a n th e simple se m a n tic role specifications given here for a single lexical i te m ’s valence.
1.3 T h e se m a n tic s-p r a g m a tic s d ich o to m y A p o in t which will come u p repeatedly in th e following c h a p te rs is th e alleged dichotom y betw een se m a n tic s an d pragm atics. T h e analysis of th e G re e k d a ta prese n ted here ad d s to a b ro a d range of d a t a w hich sug gests t h a t we have no reason to believe t h a t such a distinction can be 4 Fillmore (1989) is the most detailed presentation of the theoretical assumptions and notational devices of Construction Grammar.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 20-
profitably m ain ta in e d . T h e s e m a n tic s /p ra g m a tic s d ichotom y is m anifested in two m ain areas of se m a n tic research. F irst, in th e description of lexical m eaning and specifically in th e assu m p tio n t h a t the m eaning of words could be ade q u a te ly described in term s of b in a ry a nd objective sem antic features (an a s su m p tio n shared by a research tra d itio n following K a tz a n d Fodor 1963). A n d secondly, in th e description o f sentence sem antics where, in line w ith
th e logical tra d itio n ,
it was assum ed t h a t tru th -conditions
e x h a u ste d th e m eaning description a t th e sentence level. T h e assu m p tio n t h a t lexical m eaning could be described in term s of features h as been a rgued against in all of th e extensive literature on frame sem antics.
F illm o re ’s (1976) discussion of the
m eaning of th e word
"bachelor" shows, for exam ple, t h a t c u ltu re -d e p e n d e n t un d e rsta n d in g of m arria g e ab le age (which is not identical w ith physical puberty) is p a rt of the fra m e w ith respect to which th e word "bachelor" acquires its meaning. E x am p les can easily be m ultiplied a nd th e litera tu re is too extensive to be exhaustively cited. (Fillmore 1971, 1982, 1985, Sweetser 1987 are represen ta tiv e exam ples a n d contain a full list of references on the topic.) Recent w ork on m e ta p h o r (Lakoff a n d Johnson 1980, B ru g m a n 1983 a n d others) h as sh o w n t h a t m eta p h o ric al u n d e rs ta n d in g of one dom ain in term s of a n o th e r is often a necessary p a r t of m eaning description; m etaphorical m appings, however, are often cultu re-d e p en d e n t, as is the stru c tu rin g of the do m ain s which are m apped. T o th e e x te n t t h a t a description of w ord-m eaning m u st take into a c co u n t encyclopaedic "w orld-know ledge", a n d to the ex te n t t h a t t h a t know ledge is often relativized a n d c u ltu re-dependent, semantics c a nnot be o bjectively based a n d the division betw een sem antics an d pragm atics
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 21 a p p e ars artificial. Staying within the assum ptions of th e s ta n d a r d fea tu re analysis would require either t h a t the cu ltu ral an d p ra g m a tic in fo rm a tio n be incorporated into the sem antic fea tu re analysis of a given lexical item, or t h a t it be seen as outside the scope of linguistic description: n e ith e r of these options ap p e ars satisfactory. B u t the d a t a in this dissertation do n o t argue against th e sem anticsp rag m a tic s distinction simply b y offering su p p o rt for fram e-sem antic a n a lyses associated w ith particular lexical items. R a th e r, th ey a rg u e for p r a g m atic inform ation being directly associated w ith a given w ord or a given constructional p a tte r n (cf. K a y 1983, 1990). One m ig h t th in k of this as fram ing, with the u n d e rstanding t h a t the fram e is of a fairly a b s tr a c t n a tu re , w ithout any conventional encyclopaedic knowledge a b o u t th e con te n t. A lternatively, we m ay th in k of this p rag m a tic in fo rm a tio n as a specification to th e effect th a t a p a rtic u la r word or co n stru ctio n provides certain guidelines for the in te rp re ta tio n of w hatever specific m a te ria l accom panies it o r "fills" its slots. In th e rest of this section, I will f u rth e r e la b o rate on th e way pragm atic in fo rm a tio n is associated w ith G reek con ditional a nd concessive constructions. W hole constructional p a tte rn s will be associated in each case w ith a p a rtic u la r m eaning, a nd th a t m ea n in g is "by definition" p ra g m a tic . W ith respect to the conditional sentences, for example, I will argue t h a t they should be seen as bi-clausal constructions (composed o f an a d v e rb ial a n d a m ain clause) realized by p a rticu la r verb forms in th e a n te c e d en t a n d the consequent. O ne of such conditional form al p a tte rn s is associated w ith the m eaning "general correlation": roughly, it says som ething like "w herever th e a ntecedent happens/occurs, th e n the consequent h a p p e n s/o c c u rs as well". W h a te v e r the specific lexical c o n te n t of the two clauses, th e bi-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 22 -
clausal stru c tu re should be in te rp re te d along these lines. This a p p ro a c h to conditionality (proposed by Fillm ore 1990; cf. also A k a ts u k a 1986) is very different from the tru th -c o n d itio n a l a p p ro a c h to conditional m eaning. T h e latte r neutralizes the form al differences betw een the conditional con structions a n d associates a m a te ria l implication sem antics (or som e other; cf. c h a p te r 5) with th e ‘if’ m a rk e r. T h a t is all th ere is to the s e m a n tic s of a conditional sentence in a tru th -c o n d itio n a l fram ew ork.
A n y f u rth e r
differences are relegated to th e p r a g m a tic dom ain. In the constructional a p p ro a c h a d o p te d here, one kind of m ea n in g difference
between
conditional
constructions,
is
associated
w ith
the
different verbal forms: these form s m ay, for example, m a r k v a ria tio n betw een givenness, h y p o th e tic a lity a nd cou n te rfa c tu a lity . A n o th e r k in d of m eaning contrast is associated w ith the choice of conditional m a r k e r per se, an d to account for t h a t difference we have to refer to p ra g m a tic condi tions such as givenness, or use in m aking suggestions. In both cases, how ever, this pragm atic significance is conventionally associated w ith aspects of the constructions, a nd hence constitutes th e a ctual m eaning of a given conditional p a tte rn , unless of course we are willing to reduce m e a n in g to truth-conditions. T u rn in g next to the concessives, I will argue t h a t there are a t least th ree distinguishable kinds of concessive sem antics associated w ith distin ct constructions. All three, however, m ake reference to prag m a tic factors or conditions, a nd these factors a re in each case w h a t I ta k e to be th e con ventionally conveyed sem antics of these constructions.
One class of con
cessive constructions (the na, ke na ones), for exam ple, is associated w ith scalar semantics, a n d the p r a g m a tic n a tu r e of scales has been n o ted by m a n y analysts (Fauconnier 1973, K a y 1990 etc.). T a k in g English even as a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 23 parallel example, we n o te t h a t proper n a m e s (which surely lack lexical se m antic scalarity) are p rag m a tic ally in te rp re te d as p a r t of a scalar model in exam ples like "Even Jo h n a te three helpings", w here " Jo h n " is seen as a low point on some scale of food-consum ption. T h e sc ala r in te rp re ta tio n in these
constructions
can
be
e ith e r
associated
w ith
a
specific
word
(corresponding roughly to E ng lish "even") o r to a w hole constructional p a tt e r n . As K a y (1990) notes, "even" is c h a ra c te riz e d as having direct p r a g m a tic in te rp re ta tio n in t h a t the scalar m odel t h a t is presupposed by t h e sentence which c o n tains it c a n be th o u g h t of as c o n ta in in g inform ation w hich is p a r t of the s h a re d b a c k g ro u n d of the sp e a k e r a n d h e a re r at the tim e of the u tte ran c e . A d ire c t p ra g m a tic in te r p re ta tio n
(though not
scalar) is also associated w ith t h e Greek concessive m a r k e r as inasm uch as it codes th e s p e ak e r’s a c t of "g ra n tin g " or a d m ittin g s o m e th in g previously asserted b y the hearer. Finally, a p rag m a tic in te r p re ta tio n is also associ a te d w ith the factive ( a n ke) concessives; I a rgue t h a t they directly code "opposition" or "adversity" betw een the c o n te n t of t h e adverbial and the m a in clause. Describing the m eaning of all these connectives a n d constructional p a tte rn s , therefore, requires reference to p ragm atics. A s K a y (1990) notes w ith respect to "even", it m a y be possible to m a in ta in a distinction betw een sem antic (i.e. tru th -c o n d itio n a l) m ea n in g a n d p r a g m a tic meaning b y a ttr ib u tin g the scalar se m a n tic s of "even" to con v e n tio n al implicature. P re su m a b ly , we could do the s a m e for the o th e r concessive a n d conditional m a rk e rs
discussed here, a nd still m a in ta in
the tr a d i ti o n a l semantics-
p ra g m a tic s distinction. A lternatively, we could say t h a t since a n adequate s e m a n tic description of the constructions m a r k e d by such w ords requires reference to inherently p ra g m a tic (contextually inferred) m eaning, we
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
24
-
m ight as well conclude t h a t the tra d itio n a l s e m a n tic s-p ra g m a tic s distinc tion does n o t ap p ly . In describing the Greek conditionals a n d concessives, I will ta k e the l a t te r view, it as evident t h a t the se m a n tic s of some kinds of words o r c o n s tru c tio n a l p a tte rn s consists really o f "guidelines" on how to in te rp re t a given sentence. As I n o ted in the preceding section, C o n stru c tion G r a m m a r , w hich a d o p ts the findings of th e fra m e sem antics research, does n o t ha v e a n y theoretical c o m m itm e n t to m a in ta in in g the distinction and th e re is a growing body of litera tu re w hich argues against it. T h e p resent s tu d y sh o u ld be seen as adding a piece of d a t a which also argues in the sam e direction.
1.4 P r e d ic tio n an d m o tiv a tio n In describing conditional and concessive clauses, we will find it neces sary to refer to the constructional p a rt of th e ir m eaning, t h a t is to the p a r t of th eir se m a n tic s which is n o t p redictable from the m eaning p ro p er ties of their c o n s titu e n t parts. As such, this p a r t of th e m eaning will have to be a t t r i b u t e d to th e construction as a whole. T h e g r a m m a r , therefore, will consist of (a t least) the description of constructions w ith th e ir associated a n d form al properties. A t a nother level, however, we m a y also w a n t to talk a b o u t relations betw een two or more constructions w ith an eye to determ ining the e x te n t to which the formal a nd se m a n tic p ro p ertie s of one of these m ay be said to follow from the properties of th e other. One way of doing this is b y m eans of a lexical rule, o f the s o r t m en tio n e d in section 1.2. Lexical rules are usually m e a n t to be generalizations over a lte rnations a t the lexical level, relating valence descriptions. O n e such lexical rule for English, e.g., a n d I believe for Greek
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 25 as w ell5, is the rule of "Dative m o v e m e n t" , relating the two valences asso c ia te d w ith the verb "give" (and o th e r verbs) in "He gave the book to J o h n " versus "He gave John the b o o k ". A lth o u g h I h a v e n ’t seen a ny p r o posals for lexical rules relating c o n structions a t o th e r th a n the lexical level, it is conceivable t h a t som ething a k in to a lexical rule relates the na condi tio n a ls w ith th e na concessives. A s I will show in detail in c h a p te r 3, for m a lly na conditionals are distinguished from na concessives by a wordo r d e r difference. T he "rule" also creates a difference in m eaning b u t, a lth o u g h such rules have been claim ed to relate completely synonym ous valence descriptions, in reality th e y often do n o t (cf. th e "causativization" rule). Seen in this light, therefore, a n d w ith th e provision t h a t lexical rules co u ld be seen as relating units longer t h a n the valence of a single verb, the G re e k conditional-concessive relationship could be handled by such a rule. B u t lexical rules will not help us to talk a b o u t all possible kinds of rela tio n sh ip s between constructions.
T h e re are cases where the differences
a re exclusively sem antic in na tu re . Rules of th e kind described above are b y definition "ill-equipped" to describe these cases, since such rules usually re la te "form ally" distinguishable c onstructions. In describing, for instance, th e v ario u s clauses introduced by as (cf.
c h a p te rs 2 and 4), we find t h a t
in som e cases (though not invariably) th e difference between th e c o n stru c tio n s is prim a rily sem antic and th e re is no w ay of formally distinguishing tw o
kin d s of as clauses. If we decide to t r e a t such uses of
as as
h o m o n y m o u s, then, of course, th ere is no need to talk a bout a relation ship. If, o n th e other hand, we decide to t r e a t th em as sem antically rela te d , th e re are (at least) two w ays of analyzing the relationship: either 6 Consider examples like e8oie to trivlio tion jani, "He gave the book to John", versus ehote tu ja n i to vivlio, "He gave John the book". For Greek, we would have to further specify th at the prepositional "goal* phrase will be in the genitive case when promoted to direct object.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 26 by recognizing m ultiple distinct b u t related m eanings, or by postulating a single general a n d / o r a b s tr a c t m eaning covering all th e m eaning distinc tions. In c h a p te r 4, I a rgue t h a t we have good reasons to take as as being really polysemous, t h a t is as representing a larger se m antic category with related b u t distinguishable sub-senses. W e need some w ay, therefore, of talk in g a b o u t these sem antic con stru c tio n a l differences (or p a tte rn s of constructional polysemy) in a way t h a t c a p tu res th e com m onalities between t h e senses, as well as th e fact t h a t are properties u nique to a p a rticu la r sense. O ne w ay of doing this is by p o stu la tin g a ‘r a d ia lly ’ stru c tu re d category of the so rt proposed by Lakoff (1987) illustrated by his detailed analysis of ‘t h e r e ’ constructions6. Lakoff’s claim is t h a t radially stru c tu re d categories exist bo th in th e lexi con, w here th e y relate different senses of a given lexical item (or mor phem e), a nd in the g r a m m a r where they rep re sen t (sem antic a n d syntac tic) relationships betw een constructions. In b o th cases, th ey have the same function, nam ely to m o tiv a te correspondences betw een form and meaning. In c h a p te r 4, I propose a skeletal analysis of th e different as construc tions along these lines. In some cases the variation is only semantic, whereas in others different as clauses are differentiated by both form al and se m antic properties. Describing th e s tru c tu re of th e category is a way of noting w hich constructions are m ore closely related t h a n the o th ers and identifying such properties of a given construction as can be derived from its closest "relative". As I said in the beginning of this section, th e g r a m m a r will have to in d ep e n d e n tly specify th e constructions of a given language. W h y then ® Brughman (1988) contains a detailed description of another "lexical network" or category with constructions a t the nodes, namely the network of constructions which are headed by the lexeme HAVE.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 27 concern ourselves w ith m otivation a t all; especially since, even in the "sim plest" case where no form al differences exist be tw e en two constructions, we will have to specify w h e th e r a given se m a n tic "node" of the category is n o t only possible (i.e. m o tivated) but, also, a c tu a lly existing? I can th in k of two reasons for doing so. T h e first refers to th e linguist’s task of describing the se m a n tic s /p ra g m a tic s of a given co n stru ctio n as adequately as possible. By noticing the sem antic relationship(s) betw een the con stru c tion being described and some oth er con stru c tio n s ("m a rk e d " by the sam e le x ic al/g ra m m a tic a l
item), valuable insights m a y
be gained
into the
sem antics of th e form er. I have found this to be tru e b o th with respect to th e as concessives (which, I argue, are m o tiv a te d via a "permission" sense of as) a n d w ith respect to the na conditionals, whose "dubitative" s e m a n tics can be tra c ed to its non-conditional uses. T h e second reason has to do w ith m y intuitions as a native speaker of G reek, to th e effect t h a t these senses of, e.g., as a re indeed related. Somehow t h a t relationship will have to be described. Obviously, in this kind of task th ere w o n ’t, be total agree m e n t betw een speakers as to w h a t is related to w h a t and in describing sem antic relationships we need to refer to som e "idealized" abstra ctio n , w hich will be u navoidably affected by th e a n a ly s t’s own intuitions. Despite these difficulties, inasm uch as this kind of analysis tells us som ething a b o u t th e sp e ak e r’s linguistic com petence, it is w o r th doing. In th e discussion of the various c onstructions, therefore, I intend to consistently a ddress the issue of their m o tiv a tio n a l semantics. In some cases (for example, for as) I do so in m ore detail t h a n I do for others. T h is is because, as I s ta te d above, I believe t h a t identifying the sem antic m o tiv atio n for the p a rticu la r as construction w hich is the main concern of this
s tu d y
(nam ely
th e
concessive)
helps
considerably
in giving
an
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 28 a d e quate se m antic description of t h a t construction. In y et o th e r cases, I h a v e n ’t been able to find a s tra ig h tfo rw a rd se m a n tic m o tiv a tio n for the overall in te rp re ta tio n s of a given construction (cf. especially th e discussion of conditionals in c h a p te r 5). E ven there, how ever, I have m a d e an a tt e m p t a t providing m o tivation by discussing w h a t I called th e "systemic coherence", t h a t is th e e x te n t to w hich a given c o n s tru c tio n a l p a tt e r n can be seen as fitting into a broader sy stem of sim ilar p h e n o m e n a . I can only hope t h a t this kin d o f work m ay serve as a ste p to w a rd s th e discovery of m ore s u b s ta n tia l m otivation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 29 -
C h a p te r 2
T h e s y n ta x and sem a n tics o f lexical a n d c la u sa l co n stitu e n ts o f th e co n c e ssiv e c o n stru ctio n s
2.0 In tr o d u ctio n In this chapter, I will exam ine the sy n ta c tic a n d sem antic properties of th e c o n stitu e n ts w hich we find as com ponents of th e different types of concessive clauses. O f interest, of course, are those c o n stituents which recur across different constructions; these are b o th of th e lexical and the clausal type. T h e p u rpose here is to isolate th e sy n ta c tic a n d semantic pro perties w hich can be assigned to these c o n stitu e n ts (constructions or not) a n d to distinguish these properties from th e pro p ertie s which can only be a tt r ib u t e d to the larger (concessive) construction as a whole. Several clarifications are in order here. T h e first has to do with the distinction ‘lexical’ versus clausal which I referred to above.
From the
survey of concessive clauses given in the in tro d u c tio n it is clear th a t the recurring c o n stitu e n ts are either of the type exemplified by ke or by the ty p e exemplified by no clauses or as clauses. W ith in lexical constituents I will e x am ine the sem antics of ke outside the concessive context, trying to de te rm in e how m u ch of its sem antic c o n trib u tio n is a ttr ib u ta b le to ke itself, in oth er th a n th e concessive environm ents, a n d how m uch arises only in a concessive context. F o r the clausal constituents, we need to address questions of internal order, constituency a n d basic semantics, again o utside th e concessive con text. As we shall see, some types of clauses m ean different things and have different s y n ta x depending on w hether or n o t th ey a re p a r t of a concessive
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 30co nstruction, an d , therefore, the difference will have to be specified a t the level of the larger (concessive) construction. F o r y et oth er types of clause, however, the difference betw een the concessive a n d th e non-concessive con s tru c tio n s is m in im a l a n d a g reat deal of the form al a nd sem antic p roper ties of th e concessive c onstruction are derivable from the properties of the c o n s titu e n t clause. T o c a p tu re this difference, I shall need to m ake refer ence to bo th the form al and s e m a n tic /p r a g m a tic properties of the consti t u e n t clause as a whole (i.e., of the ‘c o n d itio n a l’ or ‘a s ’ sentences as con stru c tio n s them selves), a n d to individual c o n stitu e n ts of these clauses, and in p a rtic u la r to the connectives by which each is introduced. In p a rtic u la r for na, there have been a tt e m p t s to provide unified a n a lyses of its sem antics w hich claim t h a t na is polysem ous. O n this account, all the uses of na listed in 2.2.1 (including those which are radically different from th e conditional and concessive ones) are related. However, as I a rgue in c h a p te r 3 below, there seem to be also strong argum ents for setting a p a r t th e conditional and th e concessive uses of na from all its o th e r uses a n d to t r e a t th e m as a se p ara te class of constructions.
T h e re
are also a rg u m e n ts to th e effect th a t the conditional and concessive con stru c tio n s them selves, alth o u g h more closely rela te d th a n the others, should nevertheless be considered separate a n d distin ct constructions. However, once we distinguish each of the constructions, there are interesting generalizations which refer, for exam ple, to the na conditionals as a class, regardless of the form in which each na clause is expressed in (e.g., by w h a t p a rtic u la r tense or mood c o m b in a tio n and of the special sem antics a tta c h e d to this form). These generalizations m ay be m otivated p a rtly by the sem antics of no in other t h a n th e conditional and the con cessive contexts (i.e., by the other uses of na), b u t are by no m eans
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 31 p re d ic ta b le or d e rivable from them . T he fact rem ains, therefore, t h a t a d e ta ile d sy n ta c tic a n d sem antic analysis of na conditionals a n d concessives is necessary if we a re to ac co u n t fully for th eir syntactic-sem antic a n d p r a g m a tic behavior as a class of na uses which is different from th e rest, b u t also as distinct ty p es of constructions themselves.
2.1 K E
2 .1 .1 C o n te x tu a l or lex ica l "evenness"? T o find o u t w h a t is th e sem antics of ke w hen it is found in a condi tio n a l e n v iro n m en t, t h a t is w hen it introduces concessive clauses, we need to consider w h a t is t h e sem antics of ke in a non-conditional environm ent. K e (and its regular a llom orph before vowels kt) is hom o p h o n o u s with the c o o rd in a te c o n ju n c tio n " an d ". T h e first question, therefore, we need to address, is w hether ke can be found to express a n y th in g akin to concessive ness1 in o th e r th a n th e conditional, clausal environm ents. If such a use exists, it m ay m o tiv a te or even allow us to give a fully compositional a c c o u n t of the cases where ke is obligatorily co n stru ed as ’’even".
Exam
ples (1) a n d (2) show t h a t ke can indeed h a v e such a meaning:
1 It is too early a t th is point to try to define w h a t concessiveness means. In general, there have been two ways of thinking a b out a sentence containing, for example, the word even, regardless of w heth er the sentence is a conditional or not. One of them makes reference to an (independently existing) counter-to-expectation idea which is evoked by the proposition containing even while the other analyzes even as a scalar term . I review some of the literature on the non-conditional even in this section and some on the conditional even in the nex t chapter. In the course of describing the constructions I hope t h a t it will become clearer w hat concessiveness can and c a n n o t mean with respect to the Greek d a t a and in c h a p te r (5) I will a tte m p t a semantic characterization of concession. F o r this section, it is enough to distinguish between the "regular", conjunction meaning of ke and its concessive use, no m a tte r how we choose to analyze the latter.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 32 -
(1) ‘-vriski
kanis
find-3SG
"ynisio meli
one
p u re
s ta
x o ria ’
honey in-the villages
"C an one find p u r e h oney in the villages ?" ‘-ke
stin ABina
vriskis
‘a n d ’ in-the A th e n s
an
psaksis’
find-2SG if
look for -2SG
"You can find it even in A thens, if you look for it". (2) ‘-0a erBun oli FUT
p a rti nom izis’
come-3PL every b o d y to-the p a r t y th ink-2S G
"Will everyone ‘-ke
sto
i
m a ria
‘a n d ’ the M ary
com e to th e p a rty do you think?" 8a erBi an epim inis’ FUT
come-3SG if insist-2SG
"Even M ary will come, if you insist". T h e speaker of (1) has in m ind t h a t it in general unlikely th a t pure h oney can be found in a large city, or, if we th in k of ke in scalar term s, t h a t it is more unlikely to find it there th a n , e.g., in a village. Similarly, t h e speaker of the ke sentence in (2) assum es t h a t M a ry is more unlikely t o come th a n some
o th e r people
who h ad been invited, or t h a t we m ay
general not expect
M a ry to go
to parties.
t h u s trigger a n "even" rea d in g of ke.
in
A n a p p ro p ria te context can
O utside this context, however, the
sa m e sentence m ay have a an additive in te rp re ta tio n .
In other words, the
"even" reading of ke m a y be contextually available b u t it is also cancell a b le an d , therefore, n o t a conventionalized m eaning of ke. If for example, t h e co n tex t sentence in (1) were ‘In Salonica you can find p u re h o n e y ’, the answ er could have been th e sam e with ke having this tim e an additive (‘also’) in terp retatio n :
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 33 ‘In A th e n s as w ell/also you can find it, if you look for i t ’. T h is way of talking a b o u t things m ay be misleading in t h a t it does not ta k e into ac co u n t the in to n a tio n difference between the additive an d the "even" in te rp re ta tio n of ke. T h e "even" reading of ke requires a rising in to n a tio n on
na 0 in a n, for exam ple, w hich is lacking from
th e also
in te rp re ta tio n . T o the e x te n t t h a t this is a gram m aticized in tonational p a t t e r n ("m aria" in (2) is also c haracterized by the same rise), it is an oversimplification to say t h a t th e "even" reading is simply contextually controlled. However, I w a n t to d ra w a distinction betw een cases like (1) and (2) on th e one hand, where the "even" m eaning is n o t conventionalized, and cases where ke is in a conditional en v iro n m en t introducing a concessive clause. W hen com bined with a t least one type of conditional clause, ke is unam biguously concessive. F o r the non-clausal uses, therefore, we m ight still say t h a t ke basically m eans "and" a nd th a t the fact t h a t som etim es it gets tra n sla ted as "even" is a fact a b o u t the translation ( t h a t is, it is English which requires "even" in such cases) r a th e r than a b o u t ke (how ever see also (9), (10)).
2.1 .2 A n a ly ses o f "even" o p era to rs Supposing t h a t ke has a n "even" reading, even if only as a conversa tional im plicature, how can we characterize this meaning? In o th e r words, w h a t exactly is the sem antics of ke in exam ples like (1) and (2). No sys te m a tic study of the sem antics a nd s y n ta x of th e Greek ke has ever been done. T here have been, however, studies of other lexical items w ith similar sem antics in English an d F rench, nam ely "even" an d "meme". My purpose
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 34 here is not to review in detail all of th e analyses given so far for English a n d F re n ch b u t r a th e r to see if any of the analyses given for, e.g., the English "even", can adequately ac co u n t for the non-clausal uses of ke as well as for its use to introduce concessive clauses. T h e latte r use, and the e x te n t to which it can be shown to fall o u t from th e form er is the main focus of this stu d y . Studies of ‘e v e n ’ include Fillm ore (1965), H orn (1969), Fraser (1970), F a u c o n n ie r (1976), K a r ttu n e n a n d P e te rs (1979), K onig (1986) and K ay (1990). All of t h e m (except perhaps F illm o re ’s) involve to some degree some idea of a scale or gradience an d also some idea of a presupposition, since a sentence c o ntaining ‘even’ is ta k e n to depend on some o ther sen tence contextually p resen t either implicitly or explicitly. Fillm ore (1965) employs th e notion of "counter-to-expectation" to describe th e sem antics of "even". He says t h a t a sentence with "even" violates some ex p e cta tio n as in (3), (3) She even reads Sanskrit, where the idea conveyed is not only t h a t she reads S a n sk rit b u t also th a t this is somehow surprising. T h e "counter to expectation" idea is also p rese n t in the analyses of H orn (1969), K a r tt u n e n and P eters (1979) a n d K onig (1986). K a rttu n e n a n d P e ters say t h a t th e phrase focused b y "even" is th e "least likely" from a set of o th er possible candidates of which th e pred ic ate is true. Horn, on th e o th er hand, gives a scalar analysis for non-subject "even" (e.g., Muriel even cam paigned for Hubert), b u t proposes a non-scalar semantics of sub je c t "even" (e.g.
Even Muriel voted for H ubert), which is based on a
parallel analysis for subject and non-subject "only". Finally, Konig who talks explicitly a b o u t "even" in an "even if" clause, uses both the notion
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 35 of th e scale and t h a t of c o u n te r-to -e x p e c ta tio n to describe th e sem antics of concessive clauses:
" E v e n 11, says Konig, presupposes (a) t h a t th ere is an
a lte rn a tiv e to the focus value which satisfies th e open sentence in its scope, an d (b) t h a t th e value given in th e focus is th e least likely and therefore m ost surprising of all values u n d e r consideration in a given context'2. K a y (1990) contains detailed criticism of the analyses of Fillmore, K a r t t u n e n and Peters, Horn an d F a u c o n n ie r. K a y ’s proposal is t h a t ‘e v e n ’ (b o th
in su b je c t
and
non-su b je c t
position)
can
be
m ost adequately
analyzed as m a rk in g a sentence (the te x t p roposition-tp) which is m ore in fo rm ativ e
th an
an
expressed
or
im plied
c ontext
proposition
(cp).
‘G r e a te r inform ativeness’, is in t u r n defined as unilateral e n ta ilm e n t (the tp entails the cp, b u t n o t th e o th er w ay a ro u n d ) in a scalar model (whose dim ensions m ay be pragm atically de te rm in e d ). K a y argues convincingly t h a t F illm o re ’s a n d H o r n ’s analyses can be recast successfully in his analysis w ith o u t losing their original insights, a n d a t the sam e tim e a c co u n t for cases which violated th e predictions of the original analyses. T h u s, he shows t h a t in F illm o re ’s analysis, th e "counter-expectational" notion is neither a sufficient nor a necessary condition for a felicitous "even"
sentence, while, arguing a gainst H orn, he shows t h a t subject
"even" is also scalar. Finally, in response to F auconnier, who sees "even" as a necessarily end-of-scale o p e ra to r, K a y argues t h a t end-of-scaleness is n o t necessarily involved a n d should ra th e r be seen as a special case of his "m ore inform ative" idea. 2 Anscombre and D ucrot (1983) also use the notion of the scale in talking about the sem antics of "meme" (=e»ren). They, however, see scales as a special "argum entative" dim ension of language, which is distinct from the "logical order". I am not in a position to evalu ate this analysis, which should be nevertheless m entioned as another example of a scalar analysis for "even".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 36 T h e exam ples w ith non-clausal ke seem also to a c com m odate a scalar analysis well. Consider for e x am ple (4)-(7): (4) lsto
telos simbaBise
ke
tin m arB a’
to -th e end like-P E R F -P -3S G even th e M a r t h a "In th e end, he liked even M a rth a " . (5) ‘ke
stin
arxi
itan
Siskolo’
even to-the beginning be-P -3SG
hard
"Even in the beginning it was h a rd " . (6) x tipisa
ke
to kuSuni ala kanis 8en aku se
r in g - P E R F - P - lS G even the bell
b u t nobody
NEG
h e a r-P E R F -P -
3SG "I even rang the bell b u t n o b o d y hea rd ". (7) ‘ke t a x ris tu je n a pia filise
ke
ton p e tr o ’
a n d th e C hristm as p a rt. kiss-P E R F -P -3S G even th e P e te r "A nd a t C hristm as he kissed even P e te r " . W e can th in k of all instances of ke in (4)-(7) as presupposing some s o rt of scale against which the c o n s titu e n t focused by ke occupies a less likely (or more inform ative, in K a y ’s term s) position th a n some oth er u n d e rsto o d a rg u m e n t. Clearly, all the scales set up in (4)-(7) are com pletely prag m a tic ally d eterm ined, i.e., the term s evoking the scale do not have a priori scalar sem antics. However, as I argue in c h a p te r 3, th ere d oesn’t seem to be a ny s tra ig h tfo rw a rd w ay in which th e ‘m ore in fo rm a tiv e ’ tp analysis can be applied to the clausal ke, t h a t is to the ke w hich, together w ith a condi tional m arker, serves to introduce an "even if" type of clause. T here are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 37 e xam ples of such sentences to which th e "m ore inform ative" idea (and, in fact, a n y analysis which reduces the c o u n te r-to -e x p e c ta tio n idea to a scalar (end-of-scale or not) phenom enon, e.g. F a u c o n n ie r ’s) does not seem to apply in any obvious way. F o r such cases, I claim, we seem to be obliged to m a k e reference to some notion of a nc o u n te r-to expectation" frame which is evoked by th e sentence containing ke. Since the clausal uses of ke to in tro d u ce concessive sentences is th e m ain focus of this dissertation, I will discuss this m ore fully in C h a p te r (3), w here I look a t the semantics at a c o n structional level.
2 .1 .3 K E an d p o ssib le fo cu ses Leaving aside the sem antics of non-clausal ke, we now come to its dis trib u tio n a l properties.
Following M cCawley (1988:611), K ay (1989) and
others, I will speak of th e "focus of ke” as th e e le m en t which is contrasted w ith its alternatives, a n d "scope" the sentence in which the substitution of alternatives is carried o u t (cf. also C hom sky 1972:99-102 for a related idea of focus and its relation to presupposition). In exam ples (1) and (2) above, ke picks as its focus an adverbial P P which a c ts as an a d ju n c t and a NP which is the s u b je c t of the sentence respectively. In (8) below, the focus of ke is a N P which is the o bject (a m orphological accusative): (8)
ita n be-3SG
toso Gimomenos p u so
a ngry
xtipise ke t o n
t h a t hit-3SG ‘k e ’ th e
ja n i John
"He was so a n g ry t h a t he h it even Jo h n ". N orm ally ke precedes its focus. E x am ples (9) a n d (10), however, are exceptions to this rule: (9) ‘irGe
s to n "yamo
ke efere
ke 8 o r o ’
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 38 com e-P-3SG to -th e w edding a nd bring-P-3SG even present "He cam e to th e w edding and he even b r o u g h t a present". (10) ‘e k atse
ke k a p nise ke to tsiy a ra k i t u ’
sit-P-3SG a n d sm oke-P-3SG even the cigarette
his
"He stayed a n d he even sm oked his cigarette". E xam ples (9) a n d (10) represent w hat we m a y call the double ke con struction in w hich th e second ke obligatorily has an "even" reading. In (9) a n d (10) ke ap p e ars in the m iddle of the verb p hrase which is its focus. It is clear t h a t w h a t is being co n tra ste d in (9) and (10) is the whole VP an d n o t ju s t the the last constituents, t h a t is the d irect objects.
T h e idea in
(9) is t h a t we d i d n ’t expect h i m / h e r to come a t all, m u ch less to bring a present a nd in (10) t h a t h e /s h e was not expected to s ta y a t all, m uch less to m ake them selves so obviously a t home. This is also su p p o rte d by the fact t h a t the prosodic pe a k in these sentences is on th e verb (here efere a nd kapnise) and n o t on th e direct objects.
In this type of construction,
where there is a t least one co n stitu e n t following the verb (it m ig h t be an intransitive verb followed by some kind of adverbial phrase), so t h a t in a way there is a position available for ke, this second ke only has an "even" reading. W here the V P , however, consists only of a verb, it seems clear t h a t the generalization is t h a t whenever the conjunction in te rp re ta tio n is neces sary for the s tr u c tu r e of th e sentence, th en the "even" in te rp re ta tio n is sim ply unavailable. Consider, for example, (11), where ke can only be in te rp re te d as " an d " : (11) irGe
ke efaye
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 39 com e-P-3SG a n d eat-P -3S G
"He cam e a n d he ate".
In such cases
th e "even" in te rp re ta tio n requires the presence of other lexical items w hich w ould u nam biguously code an "even" m eaning. Kiolas is th e lexical item required in a positive context, kan in a negative: (12) ir8e
ke efaye kiolas/*ke
com e-P-3SG a n d eat-P-3SG even "He came an d he even ate (stayed for dinner)". (13) ir0e
ke Ben efaye
come-P-3SG a n d
k a n /* k e
N E G eat-P-3SG
even
"He came a nd he d id n ’t even eat". Since ke is the only lexical item p a rticip a tin g in the adverbial conces sive clauses, I will n o t exam ine th e sem antics and distrib u tio n of these o th e r items. F o r ke, we m a y conclude th a t a lth o u g h it does n o t seem to be conventionally associated with an "even" m eaning (except p e rh a p s in some environm ents), it can be used with such a m ea n in g in certain contexts. T h e final th in g we need to address is th e relationship betw een akoma a n d ke. T he facts a b o u t the d istribution a n d semantics of akoma, w hich som e g ra m m a ria n s (c.f. T z a rtz an o s 1963:131-2) analyze as also m eaning "even", can be s u m m a riz ed as follows: (A) W h e n akoma precedes ke or follows b o th ke and the focused p h r a s e 3, th e m eaning becomes unam biguously "even". T hus, while (1), as I said, can have a purely a d d itiv e /"a lso " m eaning, w ith the a d d itio n of akom a (a k o m a ke stin aBina...), the in te rp re ta tio n of the focused c o n s titu e n t 3 A ko m a shows floating effects, which I am not going to examine here. T he positions I describe in (A) can be considered the neutral ones, while end-of-sentence position is also possible. As expected, there is considerable variation in acceptability from speaker to speaker.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 40 becomes obligatorily scalar.
A k o m a ke can app e ar in all th e sam e environ
m en ts t h a t ke (in th e m eaning "even") can. (B) akoma m u st be accom panied by ke, if it is going to have an "even" reading. By itself, it can only h a v e a tem p o ral "yet" o r "still11 in te rp re ta tion. W e can illustrate this w ith ex a m p le (14): (14)
boris
akom a
(ke)
can-2SG a k o m a
and
na
fijis
an
0es
su b j. leave-2SG if
w ant-2S G
a. "You can still go if you w a n t" (w ith o u t th e ke) b. "You can even go if you w a n t " (with the ke) G iven these, t h e question arises as to w h a t exactly is the m eaning of ke a n d th e m ea n in g of akoma a n d w h a t is the division of labor between th em . W h e n I consider issues of com position and com positionality later on, I will ta k e the position t h a t a lth o u g h the function of ke to introduce con cessive clauses is clearly m o tiv a te d (given its im plicature properties), there is also a d istinction to be d ra w n b etw een this function, w here it obliga torily m eans "even", and th e cases discussed in this section. T here is a f u rth e r
difference
between
th e
k in d
of
"even"
m eaning
that
non
conditional ke expresses a n d t h a t expressed by the ke which introduces a clause (only the form er can be alw ays th o u g h t of as scalar).
A k o m a ke
expresses "evenness" obligatorily, b o th in clausal a n d non-clausal environ m ents. O n th e one han d , we h a v e ke, w ith a contextually available "even" reading, and, on t h e other, we ha v e akom a which by itself c a n n o t have an "even" reading a t all. The tw o of th e m together a dd up to a conventional way of expressing "evenness". F ig u rin g o u t the sem antic processes which led a "still" an d an "also" to com e to m ea n "even" m ay be an interesting question itself, alth o u g h not a question to be discussed here.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 41 2.2. NA
2 .2 .1 T h e n o n - c o n d i t i o n a l u s e s
2 .2 .1 .1 N A in m a i n c l a u s e e n v i r o n m e n t s T h e m o rp h e m e na is m ulti-functional in M odern Greek. Its use is so frequent a nd so varied t h a t there is no one single stu d y which encom passes all the different uses. M a ckridge (1985), contains the m ost com plete list of th e uses, a lthough, as th e a u th o r adm its, no a tte m p t a t a system atic inves tigation is m ade. Joseph a n d P h ilip p a k i (1987) and Hesse (1980) also con ta in lists of the uses as does T z a rtz a n o s (1963). Since this d isse rta tio n is m ostly concerned with th e concessive na, I am only going to look in detail a t one use of na, which, I argue, is the only one directly related to the concessive use. T h a t use is the conditional. However, in view of the fact t h a t some of th e p ro perties of the conditional na can be tra c ed to its sem antics in com pletely different environm ents, I will tr y to address briefly the o th e r uses o f na as well. I c a n n o t hope to do justice or even list exhaustively all of th e o th e r uses (the reader is referred to the works m en tioned above). I sim ply in te n d to list some of n a ’s basic uses, which will help clarify the range of its d istrib u tio n and semantics. A t this point, I am n o t m aking an y claims a b o u t w h e th e r there is a single polysemous na or w h e th e r we are dealing w ith com plete hom onym y.
Later, I will m ake
some suggestions in this direction. T h e first group of uses is th e na in a m ain clause e n vironm ent. The basic use here is th e e m p lo y m e n t of na to express the im perative, obliga torily in the t h ir d person a nd a lte rn a tin g with a monolexemic form in the second:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 42 (15) n a fijis /
fije
amcsos
‘n a ’ go-2SG / go-2SG-IM P im m ediately "Leave im m ediately!" (16) n a
fiji
amesos
‘n a ’ go-3SG im m ediately "I dem a n d t h a t h e /s h e leave im m ediately". In the rest of its uses in m ain clauses, na expresses various modalities like an o p tativ e ("w ish") as in (17) or, as in example (18), a "deliberative question" (M ackridge 1985:283): (17) n a
isuna b ro sta
‘n a ’ be-2SG th ere (18) n a
er0o
n a ’ c om e-lS G
ki
"I wish you h a d been
th ere !"
e"yo and
I
F ro m exam ples (15)-( 18),
"Shall / Should I come too ?" it is
obvious t h a t the so-called m ain-clause
uses of na are less "m ain-clause like" t h a n a sentence containing a verb not preceded by na. In fact, given t h a t M odern Greek does not have any non-finite c o m p le m e n tatio n , we could m ak e an a rg u m e n t similar to t h a t m ade for L atin by R. Lakoff (1968) to the effect th a t in all these examples, there could be a m ain verb understood to which the na clause serves as a com plem ent. T h is is obvious for the im perative cases, a nd conveyed by the translation in (17). (18) could be equally well translated as "Do you w a n t me to come too?". A parenthesis is in order here to talk briefly a bout th e morphology of the Greek verb. W e need to differentiate between the different verb forms which follow na in th e conditional a nd concessive clauses.
For this, I am
going to use M a ck rid g e ’s categorization of the verb form s which captures
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 43 th e c o n tra sts in
the verb m orp h o lo g y . T he description here will be
oversimplified, not taking into a c c o u n t verbs which lack certain aspectual form s. I will also not address th e question of exactly w h a t p a r t of th e e n d ings expresses aspect and w h a t tense. T h e table below is tak e n from Mackridge (1985:103) and illustrates t h e form s which c ontrast, using th e verb "to love" in the 1st person sin g u lar. (V erb endings are underlined.) T a b le I Aspect Imperfective T ense N on-past Past
T h e perfect, aspects
by
P e rfe c tiv e
a"yapo a 7 ap u sa
a"yapiso a ^ a p is a
Perfect exo a^apisi ixa a 7 apisi
past a n d n o n - p a s t, is distinguished from th e o th er
a different ending,
but
the p a s t/n o n
p a st
d istin ctio n
is
expressed by verb "to have" (in n o n -p ast exo and p a s t ixa form s). T he perfective non-past, ayapiso, is t h e only form which ca n n o t occur by itself b u t requires the presence of a p a rticle . N a and as (which will be discussed m o re fully below) are such particles. 0 a , the future m ark e r, is a n o th e r. T h e perfective non-past form m u s t be preceded by one of these (an d th a t is w hy it is sometimes called a bo u n d -fo rm ), while these m ark e rs can also "govern" any of th e o ther form s in tab le I. T h e discussion of w h e th e r M o d e rn Greek has a m orphologically dis t in c t su b ju n c tiv e (Joseph an d P h ilip p a k i 1987:179-180), or w h e th e r there is only a sem antic distinction b e tw e e n indicative and su b ju n c tiv e (Mackridge 1985:104 a n d ch.9), hinges on th e analysis th a t one gives to this fact a b o u t th e distribution of the perfective non-past. M ackridge’s p o in t is t h a t since
th e
particles
are
an
inseparable
p a rt
of
the
verb
form,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 44 su b ju n c tiv e n e s s does n o t inhere in a p a rtic u la r v e rb ending, b u t is a func tion of th e su b ju n c tiv e m arkers, and, therefore of th e s y n ta c tic /s e m a n tic co n te x t. T his, according to him, does not w a r r a n t th e existence of a m or phological distinction betw een indicative a n d s u b ju n c tiv e . On th e other h a n d , Joseph a n d P h ilip p a k i (1987:179-180) a rg u e t h a t since th ere is a t least a n in d e p e n d e n t difference in the ty p e of neg a tio n , we should recog nize a d istin ctio n betw een the su b junctive (m a r k e d form ally by the p a rti cles na a n d as a n d th e negation) a nd th e indicative.
T a k in g p a r t in this
d e b a te is beyond th e scope of this dissertation. W h e n I address the condi tional use of th e na -verb form, I will sim ply assum e t h a t the na verb form is indeed prim a rily characterized b y irrealis sem antics, although in some cases (cf. (22)) this generalization b rea k s dow n.
2 .2 .1 .2 N A as a n o n -fa c tiv e su b o r d in a to r I now proceed to listing the rest of the uses o f na, and in p a rticu la r its so-called s u b o rd in a te uses. Na m ay in tro d u c e a clause which acts as the s u b je c t of a v e rb or as a com plem ent of a verb, an adjective, a preposition or a c o n ju n c tio n . N a can also introduce clauses w hich depend on nouns (form ing a s o rt of "infinitival" relative clause). I will use the term "depen d ent" in (20) as a cover term to refer to b o th c o m p le m e n t an d modifier uses of na. Finally, na serves to introduce a d v e rb ial clauses. Below I give one ex a m p le for each of these categories of usage4: (19) S ubject: 4 In the examples th a t follow, P E R F stands for perfective, EMPF for imperfective and Perfect will ju s t be perfect. P is past and N P non-past. F inally, where I d o n ’t m ark the P E R F or IM P F , this means th a t the particular verb does not show this distinction morphologically.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 45 -
bori
na
fiji
c a n-IM P F -3S G
n a g o -P E R F -N P -3 S G
"It is possible t h a t he
will Ieave/He m a y leave". (20)
"D ependent"
(20a) of a verb: 0elo
na
fiji
w a n t- lS G
na
g o -P E R F -N P -3 S G
"I w a n t h im /h e r to go"
(20b) of a noun: Sen NEG
ine
an0ropos be-3SG
n a ton
person
ebistevese na
him
tru s t- D 4P F -N P -2S G
"He is not a person to be tru s te d " . (20c) of an adjective: panda alw ays
pro0im os
n a voi0isi
eager/w illing
na
h e lp -P E R F -N P -3 S G
"Always eager to help". (20d) of a preposition: 0a p e rim en a m e FUT
m exri
w a it- I P L
until
na na
efevje leave-IM PF-P-3SG
"W e would have w aited till h e /s h e left". (20e) of a conjunction: ir0es
prin
come-P-2SG
na before
ksimerosi na
d a w n -P E R F -N P -3 S G
"You came before daw n (‘= i t d a w n e d ’)".
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 46 (21) S u bordinate clause na
se
na
ixe
you-ACC
Si
0
a
ekane
m eyali
s e e-P E R F E C T -P -3 S G
fasaria
FUT
m ake-P -3S G
big
fuss "If he had seen you, he would have m ade a big fuss". T h e list above is n o t exhaustive since there are m a n y m ore construc tions in M odern G reek which have th e na construction as one of th e ir con stituents. All I have trie d to do here is give a sam ple of w h a t I consider to be the main categories of usage for na. Na is not alw ays in an irrealis environm ent.
In relation to examples
like (2 0 a), it is w o rth pointing o u t t h a t na m ost often introduces th e com plem ent of a verb of necessity, ability or volition (cf. also M ackridge 1985:285, T z a rtz an o s 1963:185) a nd in this sense it does c o n tra s t w ith the com plem ents of factive verbs, usually introduced by oti (cf. 23). However, as ( 2 2 ) shows, this is by no m eans a necessary condition on the d istrib u tion of na. ( 2 2 ) ton
i8 a
him see-P -lS G (23) siniSitopiisa realize-P-lSG
na
fevji na
leave-IM PF-N P-3SG
ksa fn ik a suddenly
oti
"I saw him leav in g 11.
fevji th at
leave-IM PF-N P -3SG
"I suddenly realized t h a t he is leaving". T he examples above show also t h a t the verb form w hich follows na can be any of these illustrated in table I. As we shall see, this is a property of the na construction which is preserved in its su b o rd in a te clause use. Exam ple (21) above exemplifies the use of na to in tro d u c e a condi tional. N a by itself can introduce a concessive clause as well, a n d p a rt of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 47 c h a p te r 3 will be devoted to this la tte r co n stru ctio n a nd its relationship to th e conditional. F o r the rest of this section, I will investigate th e p roper ties of th e na conditional since this is one of the constituents of the con cessive construction.
2.2.2 N A as a c o n d itio n a l m arker. T his use of na involves a p a rtic u la r c o nstruction, different from the others in t h a t na this time introduces a n adverbial r a th e r th a n a comple m e n t clause. E xam ples (24)-(26) illustrate f u rth e r th e use of na to in tro duce a conditional clause: (24) n a
bi
ksafnika
a
0
jini me-yali fasaria
n a e n te r-P E R F -N P -3 S G suddenly
FUT
h a p p e n -P E R F -N P -3 S G
big
fuss "If h e /s h e enters suddenly, there will be a big fuss". (25) n a na
im un
onasis
b e -P -lS G
0a
Onassis
to
FUT
ayroraza it
b u y -IM P F -P -lS G
"If I were Onassis, I would buy it". (26) n a na
xate
ftasi
8
io lefta n o rite ra
a rriv e -P E R F E C T -P -2 P L tw o
0
a tu s proftenate
m in u te s
earlier
FUT
th e m catch-
IM P F -P -2P L "If you h ad arrived two m inutes earlier, you would have cau g h t them ". Form ally, na conditionals have th e na verb form in the beginning of th e clause, w ith the other c o nstituents following. A fter the verb various orders are possible. Crucially, however, for o u r purposes, (cf. C h a p te r 3), th e verb form preceded by na is always in the beginning of the conditional clause. T h e only constituents t h a t can come betw een na and the verb are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 48 the negation m o rp h e m e a nd a pronom inal o b je c t (cf. ex. (21)). A fter the verb phrase, various o rd ers of constituents a re possible, som e of which m ay be considered less m a rk e d th a n others. A p a r t from the difference t h a t I’ll describe in chapter 3 (which is cru cial to th e resulting in te rp re ta tio n ), every o rd er of con stitu e n ts which is possible in the conditional is also possible in th e concessive. T h e following word o rd e r p a tt e r n s a p p e a r to be the com m onest ("subject" refers to a NP s u b je c t as opposed to j u s t inflection m ark e d on the verb, a n d ( ) indicate optionality). (i) n a co p u la (subject) p redicate (cf. (25)) (ii) n a v e rb -in tran s. (subject) (adverbial) (cf. (24)) (iii) n a v e rb -tra n s itiv e (subject) object (adverbial) o b jec t (subject) (adverbial) T h e na conditional, however, is more restricted in use t h a n th e ordi n a ry c o n ditional m a rk e r an ( = if )- All of the exam ples above are also pos sible w ith an in the place of na). However, alongside (24)-(26), we have (27)-(29) which are u n g ra m m a tic a l (although th ey are again gram m atic a l w ith an): (27)* n a na
meletisis
toso
s tu d y - P E R F - N P - 2 S G
sklira
8a
perasis
t h a t / s o hard
F U T pa ss-P E R F -N P -2 S G
"If you s tu d y t h a t h a rd , you will pass". (28) *na
pijenis
stin
na g o -IM P F -N P -2S G
aGina
to -th e
pare
ke m ena
A thens take-EMP-2SG
and me
"If you are going to A thens, tak e me also / as well".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 49 (29) ??na na
stilo
8
e m a ta 0a
ta
s e n d - P E R F - N P - l S G parcels
taxiSromisi o FUT
them
jan is m ail-P E R F -N P -3 S G
t h e Jo h n "If I send (any) parcels (as we assum e I am), Jo h n will mail th e m " . O n th e basis of (27)-(29), (an d other examples presented below), I suggest t h a t a general c o n stra in t governing th e distrib u tio n of na condi tio n a ls is t h a t they can only be used in cases w here the speaker h as doubts a b o u t the t r u t h of the protasis. W e can call na conditionals "d u b ita tiv e". In the course of th e discussion, I will try to qualify this notion further. E x a m p le (27) contains the anaphoric elem ent "toso" (—so, th a t) .
In
o ne in te rp re ta tio n , "toso" m ay be referring back to a previous s ta te m e n t b y th e in te rlo c u to r to th e effect t h a t h e /sh e is actually stu dying hard. In th is co n te x t a n d u n d e r the reading where the speaker accepts th e protasis as true, th e na conditional is u n gram m atical. "Toso" m a y also refer back to some previously specified (either by the speaker or by the interlocutor) degree of required studying, which, however, th e speaker is not viewing as very likely to be achieved by the addressee ("If you s tu d y t h a t hard, w hich I d o u b t, you will pass"). U nder th e la tte r in te rp re ta tio n , the exam ple w ith na is considerably im proved. E x a m p le (28) is u n g ra m m a tic a l with na against the b a c k ground of t h e in te rlo c u to r having j u s t said t h a t he is going to A thens. T h e speaker o f (28) accepts this as a given prem ise in the conversation and on the basis o f this m akes the suggestion expressed in the apodosis. E xam ple (29), on th e oth er h a n d , is slightly h a rd e r to explain. It is the use of th e first per son in the protasis which renders th e use of na difficult; this is because the use of the first person tends to indicate (though not in any absolute waycf. th e discussion below) th a t the speaker considers it likely t h a t the case
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 50 of his having to send parcels m ay indeed arise (since, after all, he should know). In A k a t s u k a ’s (1985:625) terms, na conditionals, then, c a n n o t be used to express w h a t she calls "surprise conditionals" ( A k a ts u k a ’s d a ta come from English a n d Japanese). I do not wish to a rgue w ith A k a t s u k a ’s claim t h a t this type of conditional represents y e t a n o th e r p o in t on the irrealis c o n tin u u m , because it takes tim e for newly learned inform ation to be assim ilated into o n e ’s established body of know ledge ( A k a ts u k a 1985:625). M y p o in t is sim ply t h a t in examples like (28) above a n d (30) below, which A k a t s u k a calls "su rp rise /re g re t" conditionals, th e speaker regards the pro tasis as h a v in g been contextually established a n d for his purposes as being given. T hese a re the cases where na conditionals are com pletely excluded. (30) - isxirizete
oti
ine
o
pio
tim ios
cla im -IM P F -N P -3 S G th a t be-3SG the m ost
honest
"He claims t h a t he is the m ost honest". - * n a /a n n a /if
ine
toso
be-N P-3SG
timios
th at
ja t i
honest
Sen plironi w hy
NEG
pa y -IM P F -N P -
3SG ta the
xrei tu d ebts his
"If he is so honest, why isn’t he paying his d e b ts ?" H a im a n (1978, 1986) argues th a t conditional protases are always the topics o r givens of their sentences a nd as such they are neither challenged nor denied by m aterial in the apodosis; ra th e r, they are "(pre)-supposed" to be tr u e and thus constitute the fram ew ork o r sta rtin g p o in t from which th e sentence proceeds. A k a ts u k a (1986) on th e o th e r hand, argues t h a t no
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 51 conditional protases a re givens a nd if they are topics, they are contrastive topics (i.e.
"as for X" topics) ra th e r th an th e m a tic topics ("speaking of
X"). Even in exam ples like (29) or (30), where "if" can be p a ra p h ra se d by "since",
A k a ts u k a
argues
that
there is an
epistemological
difference
betw een "since" a nd "if", since the latter m ay code only indirectly accessi ble
inform ation
th ro u g h
external
evidence
(linguistic
com m unication
included). T hus, while (31) is acceptable with "if" in the p a rtic u la r con text, it is not acceptable if th e speaker has j u s t gone near the window and seen for himself t h a t it is indeed raining: (31) -It has s ta rte d raining. -If it is raining, let’s s ta y in. H a im a n ’s claim is also argued against in Sweetser (1990:5.3.2), where, however, a distinction is being m ade between given and non-given condi tionals. Sweetser notes t h a t given conditionals in English are all epistemic an d speech-act conditionals5, while it seems impossible to get a "given th a t/s in c e " reading for a c o n te n t (if X (cause), th e n Y (effect)) conditional. If th ere is indeed a general correlation between "given" protases and epistemic a nd speech-act readings, then this would explain w hy speech-act an d epistemic conditionals ten d to be ung ram m atic a l with na. If the con te n t of the a n te c e d en t is contextually established as true, th e n it cannot really be questioned- be a tru e hypothetical; ra th e r, it serves as th e back ground for the p e rform ance of the speech-act of the consequent or for draw ing a conclusion. As we have seen, exam ples (28), (29) a n d (30), which do not welcome na (unless they are in te rp re te d as casting doubt), 5 In Sweetser’s terms, an epistemic conditional has the semantic stru ctu re of ‘If I know th a t X is true, then I conclude th a t Y is true’. A speech-act conditional, on the other hand, is of the form ‘If X is the case, then I perform the speech-act expressed in the apodosis’.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 52 -
a re speech-act conditionals6. E x a m p le (32) below is an epistemic, an d , as pred ic te d , it is also b a d w ith na (although perfect with an): (32) —t a
fota
ine
a n ik ta
the lights b e -N P -3P L open —* n a /a n ine
a n ik t a
"T h e lights are on".
ine
n a / a n be-N P -3P L open
mesa
be-N P -3S G
in
"If the lights are on, he is in". T h e G reek d a ta argue stro n g ly for a " d u b ita tiv e /n o n - d u b ita tiv e " , "g iv e n /n o n -g iv e n " distinction since "given",
" n o n -dubitative"
na conditionals are excluded from
contexts.
F u rth e rm o re ,
evidence
against
H a i m a n ’s claim comes from th e observation t h a t ju d g e m e n ts of g r a m m a ti c a l l y change depending on w h e th e r specific c ontextual knowledge favors a d u b ita tiv e or no n -d u b itativ e, a likely or non-likely in te rp re ta tio n . C on sider, for example, (33): (33) ‘n a e r 0 i na
stin
o ra tu
com e-P E R F -N P -3S G
0
a p r o la v u m e ’
to -th e tim e his F U T
m ak e it-P E R F -N P -
1PL "If he comes on tim e, w e’ll m a k e it". F o r this exam ple to be g ra m m a tic a l a t all, it has to be the case t h a t the speaker has no way of know ing if th e person referred to will be on tim e or n o t and furth e rm o re he has reason to believe t h a t it is unlikely t h a t he will show up on time; the m ore unlikely, th e b e tte r it gets. In fact, the 6 Van der Auwera (1986) also discusses this type of conditionals. His claim, however, is th a t we should not distinguish between conditional speech acts and speech acts abou t conditionals. T he description here does not hinge on th a t, since I am only making reference to given vs. non-given conditionals which m ay or may not be speech acts.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 53 -
shared know ledge betw een s p e a k e r and h e a re r in this case h a p p e n s to be t h a t th e person in question is h a rd ly ever on time. E xam ple (33), however, is com pletely i n a p p r o p r ia te in a context w here the speaker knows t h a t the person is a lre ad y on his way a n d has no reason to believe t h a t th ere will be a n y th in g to d e ta in him (in which case we could p a ra p h ra se (33) with "Since h e ’ll be here o n time, w e ’ll m ake it".) Sim ilarly example (29) above im proves c o nsiderably if w h a t th e speaker h a d in m ind is som ething like (34): (34) If th e rem o te a n d unlikely possibility arises in which I’ll have to send parcels, th en he can m ail th e m for me. This s e m a n t ic /p r a g m a t ic condition reg a rd in g na conditionals has to be s ta te d in te rm s o f the " d u bitative" vs. "n on-dubitative" distinction r a th e r t h a n in term s of the "given/non-given" one. As we have seen, with respect to (27), ( p a r t of) the protasis m ay be "given" and yet the sentence w ith na m ay b e good if it is rea d as casting d o u b t on the c o n te n t of the a n te c e d e n t.
We
could
say,
therefore,
that
na
conditionals
are
d u b ita tiv e /c o u n te r f a c tu a l in a broad sense, covering English exam ples like (35) a nd (36): (35) H ad he been late, we would have missed the show. (36) W e re /s h o u ld he be late, we would miss th e show. T h e c o u n te rfa c tu a l in te r p re ta tio n can be easily seen as the one e n d of the d u b ita tiv e c o n tin u u m , with th e "given" a n d accepted as true (e.g. (28)) seen as the o th e r. N a is allowed under th e d u b ita tiv e an d the countefactu al (cf. (26)) in te rp re ta tio n s b u t ruled o u t in the cases where th e
speaker
has no d o u b t w h atsoever a b o u t th e t r u th of the protasis. T hese in te r p re ta tio n s a re speaker (or hearer) imposed r a th e r th a n irrevocably evoked by p a rtic u la r examples. E xam ples (27) and (29), as I
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 54 -
said above, get b e tte r u n d e r a d u b ita tiv e interpretation, alth o u g h the a n a phoric elem ent in the form er a n d the first person singular in the latte r te n d to rule th e m o u t o u t of context. Similarly, (37) and (38), regular sp e e rh -a c t conditionals, are u n g r a m m a tic a l if the speaker is sim ply repeat ing the c o n te n t of the a n te c e d e n t as established background. T h e asterisks in the na version refer to this reading. If, however, the intended reading is s o m e th in g like "If, by a n y chance, you are the last one to leave..." (for (37)), a n d "If, by any chance, you get hungry..." (for (38)), th e n the use of na becomes acceptable: (37) * n a / a n n a /a n klise
ise
o telefteos p u 0a fiji
be-N P-2SG th e last
t h a t F U T leave-P E R F -N P-3S G
t a fota
t u r n off-IMP the lights "If you are the last one to leave, t u r n off the lights". (38) * n a / a n pinasis
exi
fai sto psijio
n a / a n get h u n g r y -P E R F - N P - 2 S G have-N P-3SG food in-the fridge "If you get h ungry, th e re is food in the fridge". Should an exam ple like (30) be considered counterevidence to th e generalization proposed for na conditionals? T h e na conditional in the con te x t of (30) is s ta rre d despite th e fact t h a t the overall in te rp re ta tio n seems to be t h a t the sp e ak e r does n o t believe t h a t the person referred to is indeed honest. If this is the case, th en examples like (30) would be excep tions to the generalization t h a t na likes d u b ita tiv e /c o u n te rfa c tu a l con texts. However, I would like to a rgue t h a t (30) (and exam ples like it) are n o t really an exception to the rule. T h e speaker of the conditional in (30) is still accepting the c o n te n t of th e protasis as established a n d uses it as
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 55 th e b a c k g ro u n d for perform ing th e speech-act of the apodosis. It is this aspect of the m eaning which rules o u t the use of no. T h e "I d o n ’t really believe p" p a r t of the m eaning is derivative an d should be seen as an im plicature
arising
from
such
exam ples
in
cases
where
real-world
knowledge points to a conflict betw een th e c o n te n t of th e antecedent and t h a t of the consequent. As an im plicature, it can be cancelled w ithout any o d d ity and th e question in the consequent can be rea d as a genuine request for inform ation ra th e r t h a n a challenge: "If he is so honest, which I have no reason to do u b t, why isn ’t he p a y in g his debts"?. Such examples do not, therefore, constitute counterevidence to the generalization pro posed for na n or do they w a r ra n t th e existence of a special construction to a ccount for such cases. In other words, I see no reason to distinguish (30) a n d the like from an example like (39): (39) -- pijeno
stin
aGina
go -IM P F -N P -lS G to-the —* n a /a n
pijenis
A th e n s
" I ’m going to Athens".
stin aGina j a t i n a m in erGo mazi su
n a / a n go-IM PF-N P-2S G to -th e A thens w hy subj. N E G com e-PERFN P - 1 SG with you "If you a re going to A thens, w hy d o n ’t I come with you?" An interesting issue arising in the c o n te x t of this discussion is the interaction betw een the tense of the verb a n d the dub ita tiv e constraint. It is clear t h a t th e co n stra in t on th e d istrib u tio n of na conditionals cannot be sta te d in term s of the tense of th e verb in th e conditional clause, since the sam e exam ple can improve o r get worse depending on the sp e ak e r’s intentions or knowledge a n d th e degree to w hich he considers something to be likely or true. However, since t h e tense p a tte rn s themselves in a condi tional
en v iro n m en t
have
conventional
in te rp re ta tio n s
associated
with
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 56 th em , we m ay expect these m eanings to in te r a c t w ith th e general meaning of a na conditional. So, for exam ple, the p lu p erfec t ( = P E R F E C T PA S T ) in the a n te c e d e n t is norm ally associated w ith a c ounterfactual in te rp re ta tion (cf.chapter 5); the protasis of such a c o n d itional is taken to presup pose as fa c t its c o n te n t with reversed p o la rity in th e p a st tense (in the term s used here in the P E R F E C T IV E P A S T ) . 7 N a clauses with the P E R F E C T P A S T in the antecedent are norm ally alw ays acceptable: (40) na ix an er0i na
stin
ora
c o m e -P E R F E C T -P -3 P L
tus 0a tus ixes to -th e
tim e
8
i
their
FUT
them
see-
PE R F E C T -P -2S G "If th e y h a d come on time, you would have seen th em ". T h is is because c e rtain ty about th eir n o t com ing can be seen as the nega tive end p o in t of a likely c o ntinuum , w hich sanctions the use of na. P re d ic tab ly , however, when the p luperfect is e m b e d d e d in a "given" con te x t which cancels counterfactuality (the protasis of (41) does not convey t h a t they d i d n ’t come on time), th e use of a na conditional is ruled out: 7 Comrie (1986) argues against the pluperfect being conventionally associated with counterfactuality, at least for English and suggests th a t alleged counterfactuals in other languages should be also reexamined. He cites examples like (i), where the possibility of w hether the butler did it or not is at the very least left open: (i) If the b u tle r had done it, we would have found ju s t the clues th a t we did in fact find. Based on such cases, Comrie suggests t h a t co unterfactu ality is an implicature rather than the m eaning of the conditional pluperfect. However, he goes on to say th a t it is interesting to speculate on why counterfactuals should be a stronger implicature with conditionals th a t have past time reference. Com rie’s is certainly a valid point. However, u ntil chapter 5 where I actually argue th a t the pluperfect has a conventional inte rp re tatio n in certain conditional contexts, ! am simply going to assume th a t like English, Greek has a t least a strongly preferred counterfactual in terpretation associated with the pluperfect form, leaving as an open question the issue of w hat status exactly we should a ttr ib u te to this meaning (cf. also Fillm ore 1989 for a tre a tm e n t of the pluperfect conditional as a distinct construction with co unterfactual meaning).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 57 (41)-ixan er0i
stin
o r a tus
c o m e - P E R F E C T - P - 3 P L to -th e time their "T h e y had com e on tim e". - * n a / a n ixan e r 0 i
stin o ra tu s
jati
8 en
tus i 8 a
n a / a n c o m e - P E R F E C T - P - 3 P L to-the tim e th e ir w h y NEG
them
s e e -P E R F -P -lS G "If th ey ha d com e on tim e, how come I d id n ’t see them ?" T hings a re even less clear w ith o th er verb form s. T h e IM P E R F E C T IV E P A S T for exam ple, m a y be associated w ith b o th hypothetical and c o u n terfactual m ea n in g as in (42), (43) respectively (cf. c h a p te r (5)). (43) can also be read as a hypothetical: (42) an s
evlepe
o p a te r a s su 0a se skotone
a n you see-IM P F-P -3S G the fa th e r your F U T you kill-IM PF-P-3SG "If your f a th e r saw you, he would kill you". (43) ‘an ton ipostirizan
stis ekloyes 0a
e v je n e ’
an him s u p p o rt-IM P F -P -3 P L to-the elections F U T get electedIM PF-P-3SG "If they s u p p o rte d him in th e elections, he would be elected". "If they h a d su p p o rte d h im in th e elections, h e would have been elected". Both (42) and (43) are also possible with na since h y p o th e tic a l a n d coun terfactual m eanings are, as we have seen, com patible w ith it. Consider, for example, (44): (44) n a s
evlepe
o p a te r a s
su
0a
se
s k o to n e
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 58 -
na you see-IM PF-P-3SG the fath e r y our
FUT
you kill-IM PF-P-
3SG "If y o u r fath e r saw you, he would kill you". E x a m p le (44) is, however, m ore a p p ro p r ia te in a context where the sp e ak e r know s t h a t the father is nowhere near a n d therefore it is highly unlikely t h a t he will show up. O n the o th e r h a n d , it is highly in ap p ro p ri ate in a c o n te x t where th e fath e r is sta n d in g in full view of the speaker r e a d y to intervene in t h a t case, if the speaker w ished to admonish a last w a rn in g to the child (to stop doing w h atever she is doing), he /sh e would ha v e to use a n an conditional w ith the sam e verb form. Finally, a n d in relation to (44), consider an e x am ple like (45): (45)‘na se Si n a you
o
p a te ras su
8
a
se
see-P E R F -N O -3S G the fath e r
sk o to si’ your
F U T you-ACC kill-
PE R F-N P-3SG "If y our fa th e r sees you, he’ll kill you". E x a m p le (45) features the P E R F E C T IV E N O N - P A S T in the antecedent, a verb form which, if anything, is not associated with counterfactuality®. D epending on the verb form in the consequent, th e perfective non-past in th e a n te c e d e n t m a y m ean different things and in this respect it m ay be m ore difficult to describe w h a t such a tense can m ean in some conditional e n v iro n m e n t th a n sim ply to identify th e non-occurring meanings. T h e p o in t is t h a t in view of the non-occurring m eanings, this is the verb form 8 Both th e perfective, non-past and the imperfective, p a s t express hypotheticality. As is the case, however, in many languages, the p ast forms are more hypothetical in t h a t a p ast antecedent is seen as less likely to occur than a non-past one, even if the content of b o th refers to present or future situations (cf. also Mackridge 1985). The fact remains, however, t h a t unlike the p ast forms, the non-past ones are never associated with counterfactuality.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 59 th a t , w h e n in a na conditional, we should e xpect the greater variation in ju d g e m e n ts . As I pointed o u t with respect to (33) a nd (27), (29) above, these a re the cases where a speaker has to h a v e in m ind (or the hearer im agine) some special contextual conditions. S u c h exam ples bring home a p o in t forcefully m ade by Fillmore (l9 8 9 :c h .3 a n d elsewhere). A na conditional w ith a n o n -p ast verb form is only good if th e re are special contextual c onditions holding which exclude a possible given, likely or certain reading for th e protasis; on the other h a n d , th e h e a re r of such sentences will be forced to imagine t h a t such con ditions d o indeed hold, even if h e /s h e d o e sn ’t h a v e direct access to w hat th e s p e a k e r has in m ind. One way of describing this is by recognizing th a t c o n s tru c tio n a l
m eaning
(in
this
case,
confinement
to
d u b i ta ti v e /c o u n t e r f a c t u a l contexts which c haracterizes th e na construc tion) m a y be superim posed on the m eaning of th e n o n -p ast verb form in a co n ditional context. Even if the m eaning of the verb-form itself m ight lead us to a p refe rre d "likely" or "very possibly" m eaning, the sem antics of na overrules this in te rp re ta tio n .
2 .2 .3
S u m m a r y a n d C o n clu sio n s
I h a v e a rg u e d in the preceding section, t h a t th e use of na conditionals in M o d e rn G reek is governed by a c o n s tra in t w hich limits their distribu tion to d u b ita tiv e a nd counterfactual env iro n m en ts. T h e significance of this c o n s tr a in t lies in the fact t h a t a full description of na conditionals requires
reference
to
pragm atic
factors;
these
factors
are sometimes
verifiable by b o th speaker and hearer ( t h a t is th ey b o th know by the p reced in g c o n te x t t h a t the protasis is given), while a t o th e r times the s p e ak e r uses a no conditional to p resent th e protasis as highly unlikely
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 60 an d the h e a re r is forced to conclude t h a t this is indeed the case. D escribing na conditionals, therefore, requires a un it of description w hich a t th e same time encom passes g r a m m a tic a l/s y n ta c tic in fo rm a tio n and
s e m a n tic /p r a g m a tic
inform ation.
T h is
u n it
is the
con stru c tio n .
Notice t h a t in the case of na conditionals we m a y have to recognize tw o levels of description: one a t w hich all na constructions are unified by this c o n s tra in t, i.e. by this s e m a n t ic /p r a g m a t ic feature against " ce rta in ty " . A n d a n o th e r one a t which, they each re p re s e n t distinct constructions w ith p a rtic u la r sem antic a nd p r a g m a tic properties.
A na conditional w ith the
p lu p e rfe c t in the protasis (and always an a p p ro p ria te verb fo rm in the apodosis- cf. c h a p te r 5) is c o u n te rfa c tu a l, while a na protasis w ith a n o n p a s t v e rb form isn’t (the protasis is p resented as unlikely b u t n o t neces sarily as non-true). T h is type of work is in the tra d itio n of a growing body of w ork which pays a tt e n ti o n to th e s e m a n tic /p r a g m a tic inform ation a tta c h e d to a p a r tic u la r fo rm a nd considers it a p ro p er p a r t of linguistic description. T h is te n d e n c y is evidenced in F illm o re ’s (1983, a n d 1989) work on g ra m m a tic a l con stru c tio n s containing s y n ta c tic /s e m a n tic descriptions of p a rtic u la r con ditional p a tt e r n s in English), in th e Fillm ore, K ay an d O ’C o n n o r (1989) analysis of th e "let alone" co n stru ctio n a n d K a y ’s (1990) s tu d y of "even". It is also evidenced in LakofFs (1987:462-585) detailed analysis of th e English constructions introduced by "here" an d "there" a nd in his recast ing of th e coordinate stru c tu re c o n s tr a in t (Lakoff 1986) in fram e-sem antic term s. L a m b re c h t 1986a a n d 1986b studies the p rag m a tic m o tiv atio n for p a rtic u la r clause types in colloquial F re n c h a nd in G e rm an binom ial expressions respectively. B ru g m a n (1988) is an exhaustive sy n ta c tic a n d s e m a n tic /p r a g m a tic description of the English constructions h e a d e d by
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 61 "h a v e " . Finally, Sw eetser (in p rep a ra tio n ) a rg u e s t h a t change p redicate c o n structions
a re
associated
with
tw o
d istin c t
in te rp re ta tio n s
which
c orre sp o n d to role a n d individual readings (in th e sense o f F a u c o n n ie r 1985). T his list is by no m eans exhaustive. I h a v e already m entioned work on conditionals in which th e discourse e n v ir o n m e n t plays a crucial role (cf. A k a t s u k a a nd C om rie). T h e position t h a t I a m a dvocating though and w hich is re p re sen te d by the works listed above, d e p a rts so m ew hat from th e discourse tra d itio n (as this is, for exam ple, represented in H opper 1985) b y p aying sim u lta n e o u s a tte n tio n to fo rm a n d m eaning a n d by recognizing t h a t p a rtic u la r constructions m ay be conventionally associated w ith a certain m eaning. Finallly, to w h a t e x te n t is th e se m a n tic s of na in a conditional e n v iro n m e n t m o tiv a te d by its sem antics in o t h e r environm ents? Since, I ha v e only done a sy stem atic stu d y of no as a conditional, a n y observation here is speculative.
It seem s worth pointing o u t, however, t h a t in its use
to introduce a v e rb c o m p le m e n t (cf. ( 2 0 a), na usually introduces the com p le m e n t of verbs of necessity, ability or volition, b u t, crucially, never the c o m p le m e n t of a factive verb. Moreover, in its use in a m ain clause e n v iro n m en t, na is used to express m odalities like the op tativ e , questions or im peratives (cf. (15)-(18)) b u t never a d e c larative s ta te m e n t. T h e use of na, therefore, to in tro d u c e a conditional and, fu rth e r, a conditional which is governed by a specific co n stra in t against "givenness" does app e ar to be a m o tiv a te d extension. T h is is n o t m e a n t to imply t h a t a description of th e conditional con s tru c tio n as such is r e d u n d a n t; from th e discussion above, it is clear th a t a t the very least we need to m ake reference to the special use (and m ea n ing) of th e tenses in a conditional clause a n d to th e w ay t h a t these
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 62 in te ra c t w ith th e s e rn a n tic /p r a g m a tic function of a na conditional. On the other h a n d , it is useful to recognize t h a t the choice of na (as opposed to any o th e r from the available com plem entizers, e.g., oti or pos) to express this kin d of c o n ditional is n o t to ta lly u n m o tiv a te d and a rb itra ry .
In the
course of this d issertation, I hope to show t h a t a t least som e of o ur choices of connectives for in tro d u c in g a concessive clause are m o tiv ate d to a lesser or g re a te r e x te n t by the m eanings t h a t a connective (or s u b o rd in a to r) m ay have in o th e r contexts. Moreover, th e kin d of m eaning t h a t a connective has in these o th e r contexts m ay m o tiv a te the kind of concession t h a t it expresses. T h is is w h a t I suggest is tru e of na in its conditional use and, as I a rgue in c h a p te r (3), of as in its use to introduce a concessive.
2.3 A S
2.3.1 S y n t a c t ic p ro p ertie s T h e preverbal particle as also serves to introduce a clause. T h e gen eral consensus a m ong trad itio n al g ra m m a rs a n d analysts is t h a t the clauses in tro d u c ed by as are more like m ain clauses t h a n clauses in tro duced by na. W h e re a s bo th na a n d as serve to introduce im perative type clauses, as, unlike na, c a n n o t be used as a s u b o rd in a to r to introduce the co m p le m e n t o f verbs, nouns etc.
(cf. 2.2.1). Exam ples of as sentences are
listed in (46) to (50) below: (46) as mi fam e
to ra
as N E G e a t- P E R F - N P - lP L now (47) ‘as e rS un
" L e t’s n o t eat now".
oti o ra 0e!un’
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 63 as c o m e -P E R F -N P -3 P L w h a t tim e
w a n t-N P -3 P L
"L et th em come w henever th ey w a n t" . (48)
‘as exi vreksi
9e m u oso
as ra in - P E R F E C T - N P -3 S G
lip a m e ’
god m y while be a w a y -IM P F -P -IP L
"If only it has rained ,God, while we were away". (49) ‘as m u
y rap si ena y r a m a ke 0a t a
as I-G E N
w rite one
lette r a n d
kan o n iso ’
F U T them a rra n g e -P E R F -N P -
1 SG
"Let him w rite me a letter, a n d I ’ll a rra n g e things". (50)
8
elo
na
ton ynoriso
w a n t - N P - lS G subj. him
ki as ine
vlakas
m e e t- P E R F - N P - lS G and as be-N P-3SG
s tu p id "I w a n t to m eet him even if he is s tu p id " . As far as the m o rp h o lo g ic a l/sy n ta c tic facts go, na a nd as are indistin guishable. As, like na, can govern a ny of the verb form s listed in table I, a n d like na clauses, clauses w ith as are n egated by m i(n ) (cf. (46)). This last fe a tu re according to Joseph a nd P h ilip p a k i (1987:179) is one of the form al ways of characterizing th e su b junctive. In 2.2.2 I noted t h a t besides the negative m orphem e and an object p ro n o u n , th ere is no oth er c o n s titu e n t which can come between na a nd the following verb form. T his is t r u e of as as well; exam ple (51) w ith a p ro noun is g ra m m a tic a l b u t (52) w ith a n o th e r co n stitu e n t between as a n d the verb is not: (51) as (min)
ton
8
o
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 64 as N E G
him
s e e -P E R F -N P -lS G
"I s h o u ld n ’t see him ". (52) *as avrio
80
tin m aria
as tom orrow s e e -P E R F -N P -lS G the M ary "I could see M a ry tom orrow ". Finally, w ith respect to the possible w ord o r d e r p a tte rn s (i.e. order of c o n stituents following th e verb), as has all th e possibilities available to it t h a t na has (cf.
2 .2 .2
).
2 .3 .2 S e m a n t i c / p r a g m a t i c f u n c t i o n s M ackridge (1985:298-299) claims t h a t as serves to introduce clauses which m ay have tw o m ain functions: the h o r ta t iv e or permissive on the one h a n d a nd the concessive on the oth er (Since he says th a t there are only tw o m ain functions, presum ably the h o r ta tiv e and the permissive are m ore closely rela te d to each other a n d belong to th e sam e class).
These
are exemplified in (53)-(55) below: (53) as p am e
ki emis
as g o - P E R F -N P -I P L and we " L e t’s go too". (54) as min erQun
afu Sen 0elun
as NEG c o m e -P E R F -N P -3 P L since N E G
w a n t-N P -3 P L
"Let th em n o t come since they d o n ’t w a n t to ". (55) ela
sto
p a rti m as ki as feris
to n a n d ra
su
come-IM P to -th e p a rty our and as b rin g -N P -2 S G the husband your "C om e to o u r p a r t y even if you bring y o u r h u s b a n d " .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 65 E xam ple (53) exemplifies, according to M ackridge, the h o rta tiv e use, (54) the permissive and (55) th e concessive. In its hortative sense, M a c k ridge suggests, as expresses an " in ju n c tio n " on the p a r t of the sp e ak e r which is rath e r m ore a wish or a desire t h a n if na were used in the sam e c ontext. This observation disagrees w ith th e Joseph and P h ilip p a k i t r e a t m e n t of as, who claim t h a t this p a rticle is the m a in periphrastic suppletive form of the im perative (which h as m onolexem ic forms only for th e second person singular a n d second person p lu ra l). If by imperative th e y refer to th e prototypical m eaning associated w ith im peratives, nam ely "giving an o rd er" , then the form w ith na (cf. exam ples (15), (16)) is c e rtain ly m ore im p e rativ e th an th e form w ith as, w hich, as M ackridge suggests, te n d s to be m ilder. i4s clauses are p rim a rily suggestions which in c o n te x t m a y be in te rp re te d as mild or indirect orders. C o m p a re for example (56) (w ith a m onolexemic im perative), (57) w ith th e na form an d (58) with as: (56) ela
amesos
come-IM P im m ediately (57) n a er0i
"C om e im m ediately!"
amesos
n a co m e -P E R F -N P -3 S G
im m e d ia te ly
"(I dem and t h a t ) he come
im m ediately". (58) as erGi
amesos
as com e-P E R F -N P -3 S G im m e d ia te ly " W h y doesn’t he come rig h t now?" I’ll suggest below t h a t as has a t least tw o more d istinct functions t h a n those proposed by M ackridge, n a m e ly th e conditional a n d its use to express a wish. W ith respect to M a c k r id g e ’s rem a rk s, however, it is w o rth stressing
th a t
as
does
seem
to
have
a
distinct
permissive
sense
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 66 -
synchronically, a lth o u g h this is in general d isputed an d m ost g r a m m a r s do n o t list such a m ea n in g (cf. T z a rtz a n o s 1963).
A s is historically derived
from th e singular im p e ra tiv e form of the verb afino "let, allow" (Im p e ra tive ase >
as - cf. T z a rtz a n o s 1963, K ria ra s 1973). However, as th e syn
tactic difference b etw een (59) a n d (60) (with the lexical verb in the im p e rativ e ) shows, as is com pletely gram m aticized: (59) as fiji
o jan is
as lea v e -P E R F -N P -3 S G th e Jo h n " J o h n s h o u ld / m a y go". (60) as(e) ton ja n i n a
fiji
let the John s u b j. lea v e -P E R F -N P -3 S G "Let (allow) J o h n to go". In (60),
as(e) is still a three-place predicate, case-m arking its o b jec t
" J o h n " . T h e as of (59), on th e o th e r h and, is governing th e whole clause and is necessarily followed by the verb.
I agree w ith M ackridge t h a t we
need to recognize a d is tin c t perm ission sense for as synchronically as well, r a th e r t h a n su bsum ing b o th m eanings u n d e r a single one. T h e la tte r analysis a p p ears to be p ro b le m atic in several respects. If such a m eaning exists, w h a t is its n a tu re ?
Is it an abstra ctio n over suggestion and pe rm is
sion a n d , if so, w h a t kin d of com m on features are th ere b etw een sugges tion a n d permission?. If, on the o th e r han d , we identify it w ith either suggestion or perm ission, we would have to derive th e o th e r in te r p re ta tio n in som e so rt of principled way, which would allow suggestion to be in te r p rete d as permission in som e contexts b u t not, e.g., as an ord er. E x a m p le (54), for instance, could be rea d b o th as a suggestion and as g r a n tin g per mission (if, for exam ple, th e speaker is a fath e r talking a b o u t his children).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 67 It c a n n o t, however, be interpreted as an order.
F u rth e rm o re , I suggest in
c h a p te r 4 t h a t th e conditional and the concessive uses m ak e m uch more sense a nd some of their properties fall o u t, if th e y are each ta k e n as being m o tiv a te d from d istin c t permissive a n d suggestion uses. C ontextually, a n as clause m ay be also in te rp re te d as a wish, as for exam ple in (61) (cf. also (48) above). Since this kind of m eaning is not alw ays relatable to a suggestion, we m a y need to recognize it as a distinct function: (61) as vreksi
0e m u
as ra in -P E R F -N P -3 S G god m y "If only it would rain, god!" Finally, as clauses can particip a te in s tru c tu re s like these illustrated in (62)-(65): (62) ‘as kani
tin 8ulia tu ke 0a pai
as do -N P -3 S G the
job
his a nd
k a la ’ F U T g o -P E R F -N P -3 S G well
"If he does his job, h e ’ll be fine". (63) ‘as eyrafe
ena y r a m a ti vSomaSa
as w rite-IM P F -P -3S G one
letter th e
week
ke Sen 0a to n k s e x n a y a n ’ a nd
NEG
F U T him fo rg e t-IM P F -N P -3 P L
"If he h ad been writing one letter a week, th ey w o u ld n ’t have forgot ten
him ". (64) as er0is
ke 0a ta
pum e
as com e-P E R F -N P -2 S G a nd
F U T th e m s a y -P E R F - N P - I P L
"If you come, we’ll t a lk ./ If you come, y o u ’ll regret it".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 68 (65) ‘as ixe xtipisi
to kuSuni ke 0a to m eta n io n e ’
as rin g -P E R F E C T -P -3 S G the bell
and
F U T it re g re t-IM P F -P -
3SG "If he h a d rung the bell, he would have regretted it". Sentences (62)-(65) are conditionals. T h e as clause in such cases expresses a condition for the second clause in (62), a counterfactual s itu a tio n in (63) and (65) while, (64) d e pending on w h e th e r or n o t we choose th e idiom atic in te rp re ta tio n of th e phrase n8a t a purne11, can either express a sim ple contingency or a conditional t h r e a t (the idea being t h a t th e s p e ak e r does not in fact w a n t the h e arer to come).
T h e tw o in te rp re ta tio n s are for
mally indistinguishable b u t only in th e case where the in tended m ea n in g is a th re a t or a n egative suggestion can we have polarity items. C o m p a re , for instance, (66) to (67): (66) as pis as
t i p o t a / * k a ti
sa y -P E R F -N P -2 S G
ke 9a to metaniosis
a n y th in g /s o m e th in g
and
FUT
it
regret-
P E R F -N P -2 S G "If you say anything, you will regret it". (67) as pis
k a ti/* tip o ta
ke 0a se sinxoresun
as say-P E R f-N P -2S G s o m e th in g /a n y th in g and
F U T you forgive-
P E R F -N P -3 P L "If you say som ething, t h e y ’ 11 forgive you". T h a t (62)-(65) form conditional p a tte rn s is also evidenced by th e fac t t h a t as we m ay expect, there are tense dependencies between th e a n te c e d e n t and the consequent. T his will be discussed in m ore detail in c h a p te r 5. F o r now, I can j u s t give one exam ple to illustrate this.
C o m p a re (65) above
with (68):
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 69 -
(68) *as ixe xtipisi
to k u 8 u n i ke 0a to m etaniosi
as rin g -P E R F E C T -P -3 S G th e bell
and
F U T it reg re t-P E R F -
N P -3S G E x a m p le (68) illustrates sim ply t h a t a P A S T v e rb form in th e a n tecedent, c o nstrains the verb form in th e consequent, requiring a p a st verb form th e re as well. A lot more m ay be said a b o u t as as a conditional. F o r one thing, a full s tu d y would have to specify w h a t types of c o n ditional m eanings m ay be expressed by this kin d of a n te c e d en t. In general, we m ay expect t h a t th e possibilities of as as a co n d itio n a l m ark e r will be som ehow constrained by its general m eaning of expressing a suggestion in non- conditional e n v iro n m en ts (cf. also c h a p te r 4). So, for exam ple, w hereas (69) is accept able, (70) under the reading w hich expresses the s p e a k e r ’s d o u b t as to the place w here the hearer lives, is not. B oth (69) a n d (70) are of course fine w ith a regular an conditional (cf. 2.2.2): (69) as eyrafes
esi to miso ke 0a
e y ra fa
as w rite-IM PF-P-2SG you th e half and F U T
w rite -IM P F -N P -1 S G
ki eyo to ipolipo a nd I
the rest
"If you had w ritten h a lf (of it), I ’d have w ritte n the rest". (70) *as emenes
eSo ke 0a ikseres
as live-IM PF-P-2SG here a n d
afti tin ta v e r n a
F U T know -P -2S G
this the ta v e rn
"If you lived here you would know this ta v e r n " . So (62)-(65), in addition to the conditionality, m a y be seen as preserving p a rts of th e semantics of suggestion as well. In (64) a nd (66) we have w h a t we m ay call negative suggestions, t h a t is things t h a t th e speaker does not
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 70 (or did not) w a n t the hearer to do (or have done). Still, it seems possible to trace this k in d of m eaning back to a suggestion one, via a Searlean type of ac co u n t of in d ire ct speech acts (Searle 1981: c h a p te r 2). T he p o in t is sim ply t h a t som e m o tiv atio n al sem antics is easily im agined for the condi tional uses of as.
T o sum m arize, therefore, we have seen t h a t clauses introduced by as are
prim arily
associated
w ith
expressing
a
suggestion
about
w hat
s h o u l d / c a n / m a y etc. be done (or could ha v e been d o n e in the case of p a st verb tenses) a n d also perm ission. T h e y m a y be also in te rp re te d as wishes a n d finally th e y m a y serve as the a n te c e d e n ts of conditional sentences whose tw o p a r ts a re "conjoined" by ke. T h is m a y n o t be a com plete or fully detailed d escription of all the uses of a n as clause, b u t for the p u r pose of describing the concessive use, it sh o u ld be enough. In c h a p te r 3, I will discuss th e concessive use of the as clauses (cf. exam ple (50) above), arguing t h a t a lth o u g h we need to recognize a s e p a r a te as concessive con s tru c tio n ,
the
use
of
as to
introduce
a
concessive clause
is highly
m o tiv ate d , if we recognize t h a t one of its non-concessive uses is to express permission.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 71 -
C h a p te r 3
T h e N A and A S co n c e s s iv e c o n str u c tio n s
3.0 In tr o d u ctio n In c h a p te r 2, I ex a m in ed som e of the properties of the lexical an d clausal c o n stituents of concessive clauses. In particular, I discussed the properties of ke in a non-clausal en v iro n m en t and the properties o f two types of clauses, those in tro d u c ed by na and as. U ndoubtedly, th e r e is a g reat deal m ore to say a b o u t these constituents. All I can hope to have done
here
is
to
have
p o in te d
out
some
of
their
sy n ta c tic
and
s e m a n tic /p r a g m a tic pro p ertie s w hich will necessarily figure in the d e s c rip tion of concessive constructions. In this C h a p te r, I will exam ine th e constructional properties o f the concessive clause a n d also look a t the properties of their internal c o n sti tuents in a larger context. T h e issue of compositionality, will be t a k e n up in C h a p te r 4, b u t as I have alre ad y noted in the preceding c h a p te r, we should expect t h a t n o t all pro p ertie s of the concessive clauses will fall o u t from th e properties of their in te rn al constituents. T his fact is easily accom m odated by a descriptive a n d theoretical fram ew ork like t h a t of C onstruction G r a m m a r, since
description of the constructions a t every
level is th e basic t e n a n t of th e theory. T h e limiting case in which all o f the properties of the larger constructions follow from the properties of its con s titu en ts is also a c co m m o d a te d n a tu ra lly . Im portantly, however, full co m positionality in C o n stru c tio n G r a m m a r is not a theoretical b u t a n e m p ir i cal m a tte r , ju d g ed on a case-by-case basis. T he description of these p ro perties, therefore, will be done w ith a n eye to determ ining the e x te n t to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 72 -
which they follow or d o n ’t follow fiom the properties of the c o n s titu e n t parts. I will look a t b o t h th e sy n ta c tic and the sem antic properties of these concessive clauses. O n the sy n ta c tic side, I will tak e up questions of consti tuency,
which
in
th e
preceding
c h a p te r I j u s t
presupposed
w ith o u t
justification. I will also consider questions of internal order a t the con structional level. O n th e se m a n tic side, I will look a t the general sem antic and pragm atic pro p ertie s of the constructions. T he c h a p te r will be o rg a n ized by construction: t h a t is, I will first examine the clauses in tro d u c ed by na, then the clauses in tro d u c e d by ke na and finally the clauses introduced by as. T he description in th is c h a p te r will concentrate on the clausal level. Cross-clausal
relationships
and
dependencies
( th a t
is,
dependencies
between the concessive a n d th e m a in clause) will be taken up in C h a p te r 5. Also, in this c h a p te r, th e question of w hat exactly we m ean by conces siveness will be left a t an in tuitive level and also be taken up in a later chapter.
3.1 N A c o n c e s s iv e s
3.1.1 T h e " p r a g m a t i c s o n l y " a c c o u n t I will begin this section by outlining an approach to the relationship between conditionals a n d concessives which is presented in Konig 1986 and Bennett 1982. T h e K onig a n d B e n n e tt analyses are based on a tru th conditional a p p roach to conditionals an d a purely p rag m a tic view of the conditional-concessive conditional relationship and t h a t is w hy I will be referring to these a ccounts as th e pragm atic accounts.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 73 K o n ig ’s general point is th a t concessives introduced by connectives like th e English "even if" should be considered a specific type of condi tional.
Based on th e G reek data, I a m going to argue t h a t a lth o u g h con
cessives of the "even if" type share certain properties w ith a regular condi tional sentence, th e y nevertheless have p ro p erties which m u st be a ttr i b u ted to the "even if" construction as such. K o n ig ’s a rg u m e n ts can be su m m a riz ed as follows: F irst, evidence for the concessive as a conditional su b ty p e is a d d u c ed by th e fact t h a t in m an y languages, connectives of the "even if" type are derived from condi tional connectives. Even in languages like G e rm a n , w here the concessive conditional a n d the conditional are in tro d u c ed by different connectives, it used to be the case t h a t the concessive frequently had a conditional use in earlier periods. G reek is no exception in this respect; b o th the na a n d the an m ark e rs ha v e conditional a n d concessive uses 1 and to th e e x te n t t h a t such polysemy is suggestive, it is indeed an a rg u m e n t t h a t conditionals and concessive (conditional) are indeed related. Secondly, Konig claims, the classification of "even if" constructions as a p a rticu la r ty p e of conditional is s u p p o rte d by sem antic facts as well: "All sem antic properties t h a t differentiate "even if" conditionals from ord in a ry conditionals can be shown to be d ue to the c o n trib u tio n th a t "even" m akes to the m eaning of such constructions" (Konig, 1986:232). As I will show, th e G reek d a ta c o n stitu te direct counterevidence to this claim. T he following sum m arizes K o n ig ’s ac co u n t of th e c o n trib u tio n of "even"
and
th e
derivation
of
concessives
from
conditionals:
T r u th
1 No study has been done to determine which use was historically prior but there is no reason not to assume t h a t in accordance with the historical developments in other languages, these markers had first a (purely) conditional meaning. In any case, the argument about the relationship of conditionals and concessives is not affected by our lack of knowledge of the historical facts.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 74 conditionally, exam ples (1) an d (2) below are the sam e, th e ir m eaning difference being of a purely p ra g m a tic nature: (1) If P e te r comes, I will not stay. (2) E v e n if P e te r comes, I will not sta y . T h e difference is due to an in te rp re tiv e principle w hich Geis a n d Zwicky (1971:562) have called "conditional perfection" a n d f o rm u la te d as in (3)
(cf. also F auconnier
1985:114
for a description
of th e
sam e
phenom enon): (3) A sentence of th e form X - - > Z invites an inference of th e form - X ->
- Z.
K onig claims t h a t this inference is like a G ricean generalized implicatu re
and
th at
despite claims
to
th e
c o n tra ry
(m ainly
by
Levinson
1983:145ff), it can be given a s tra ig h tfo rw a rd G ricean account. T h is d e b a te is b e y o n d the scope of this dissertation. W h a t is of in te rest is how this inference relates to the sem antics of the concessives. A se n tence of the form "if p, q" , Konig says, an d its m ore categorical c o u n te r p a r t "q (any way)" can be assum ed to form a scale: < q (anyw ay), if p,q > m ax im of q u a n tity , the
. By the
assertion of the weaker s ta te m e n t "if p, q" will
im plicate "-q (anyw ay)" and th u s to the inference t h a t p is a necessary as well as sufficient condition for q. (T his account is also fo u n d in Cornulier 1983). Concessive conditionals, on the o th e r h a n d , exclude "conditional per fection" from the set of admissible inferences by entailing o r presupposing t h a t th e conditional relationship holds for a whole series of a ntecedents. W herever, therefore,
the protasis of a given conditional c o n tains an
expression t h a t m ark s an extrem e p o in t on a scale - w hich licenses the inference th a t the conditional relationship holds for all o th e r values on the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 75 sam e scale, and th u s for a series of a n teced en ts - conditional perfection is cancelled and the co n d itio n a l can be in te rp re te d as a concessive. According to K onig, expressions which can have this cancelling effect include the following: (i)
all focus particles (e.g. j u s t) w hich evaluate their focus value as r a n k ing low on a scale.
(ii)
all expressions specifying e x tre m e values in a p a rtic u la r propositional schem a (e.g. n o t d r in k a dro p , d rin k a whole bottle).
(iii) all superlatives a n d pseudo-superlatives. (iv) free-choice quantifiers like "any". T h e view t h a t th e difference in m eaning betw een conditionals and concessives is not tru th -c o n d itio n a l, b u t it is instead due to the presence of such words, and in p a rtic u la r of "even", is also held by B ennett (1982): "So far as tru th -c o n d itio n s a re concerned, there is no class of "even if" conditionals'*. B e n n e t t ’s p a p e r is also explicitly concerned with unifying the meaning of "even" in non-conditional en v iro n m en ts with the m eaning of of "even" in an "even if" e n v iro n m en t.
3.1.2 N A C oncessives Let us now t u r n to th e G reek d a ta . T he prediction m ade by the Konig and B ennett analyses would be t h a t the a ddition of "even" or of any word or expression from th e list above (in th e Konig account) would be sufficient for pro d u cin g concessive m eaning.
In this context, "conces
sive" a nd "concessiveness" will be used to refer to the cases where the con te n t of the a ntecedent a n d t h a t of th e consequent are seen to be in some
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 76 sort of opposition or adversity relationship. T h is in te rp re ta tio n m ay arise in m ore th an one way; this discussion, however, will be deferred until c h a p te r (6), w here I will investigate m ore th o roughly th e different ways in which p m ay be said to imply -q.
T h e (non tru th -c o n d itio n a l) difference
betw een conditionals a n d concessives lies in th e cancellation by these expressions of th e "conditional perfection" im p licature which would o th e r wise arise in m o st contexts. T h e G reek concessive clauses defy these p red ictions in a t least tw o ways: first, th ey m a k e a distinction between the "even" word a n d th e o th e r classes of "end-of-scale" words by showing t h a t for th e latter m ere a d d itio n to a conditional clause is n o t a sufficient con dition for giving rise to concessiveness. Secondly, they show th a t the addi tion of all such expressions (including "even") is n o t a necessary condition in ord er for the clause to have concessive m eaning. In 2.2.2, I ex a m in ed the use of na conditionals in M odern Greek. C on sider for exam ple, (4) (exam ple (25) in 2.2.2): (4)
n a im un n a b e -P -lS G
onasis
0a to
Onassis
FUT
a^yoraza it
b u y - IM P F - P - lS G
"If I were Onassis, I would buy it". According to the Konig story, since "Onassis" is a pseudosuperlative contained in th e protasis of a conditional, sentence (4) should also have a concessive reading available as well. Consider, for exam ple, (5) where the negation in the consequent creates an a p p ro p r ia te "incom patibility" con tex t for the protasis an d the
apodosis (if one is Onassis, one is norm ally
expected to be able to b u y anything): (5) ?? n a
im un
onasis
8en
0a
to
a y o ra z a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 77 n a b e - N P - lS G Onassis "If
NEG
FUT
I were Onassis, I w o u ld n ’t (be able to)
it b u y - IM P F - P - lS G buy it".
T h e ?? indicate t h a t (5) is pragm atically odd because it is still obliga torily (purely) conditional. T h e m ere presence of the pseudosuperlative is not eno ugh to m a k e it com patible w ith the negative consequent. O n the o th e r han d , the English exam ple (6), Konig claims, is a per fect concessive (Konig 1986:ex. (23)): (6) If I were Rockefeller, I would not be able to pay for this. However, even in (6), in order to get the concessive inte rp re ta tio n , the sentence has to be read w ith special em phasis (rising intonation) on "Rockfeller" ( w h a t H aim an (1986:223) calls "squeal" intonation. "Squeal", H aim an claims, is a special diacritic). Intonation, in tu rn , is not som ething th a t can be ta k e n into account when composing the m ea n in g of th e clause and, therefore, in this respect the Konig and B e n n e tt accounts m ay be insufficient even for English. T u rn in g
again to Greek, reading exam ple (5) w ith emphasis on
"Onassis", c e rtain ly improves it. However, only two of m y fifteen infor m an ts re p o rte d t h a t they m ight actually say it this way. F o r the rest, the n a tu ra l way to get a concessive reading involves preposing the crucial word, as in (7). (7)
onasis
na
im un
8en
0a borusa
n a to
a"yoraso
"Even if I were Onassis, I w o u ld n ’t be able to b u y it". W e o b ta in sim ilar ju d g e m e n ts if we try to get a concessive from a conditional by a dding oth er a p p ro p ria te words from K o n ig ’s list, except for ke ( = even). Consider, (8) to (10) where again ?? s ta n d s for p ra g m a ti cally odd:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 78 -
(8) (a)?? n a
pji
na
m ia
oulia 0a
7
d r in k - P E R F -N P - 3 S G
one
meBisi drop
FUT
get
d ru n k -
P E R F -3 S G "If he d rin k s a d ro p , h e ’ii get d ru n k " . (b) m ia
7 0
ulia
na
pji
0a
me0isi
"Even if he d rin k s one drop, h e ’ll get d r u n k " . (9) (a) ?? n a na
ita n
o
pio
xondros
be-P-3SG th e m ost
fa t
0a
te tia
Sen
x ria z o ta n
NEG
FUT
anOropos
tu
person
kosm u
of-the world
Sieta
need-IM PF-P-3S G
such
diet
"If he were the f a t t e s t person in the world, he w o u ld n ’t need such a diet". (b) o
pio xondros an0ropos tu kosmu na itan Sen 0a
x riazo tan te tia Sieta. "Even if he were th e f a tte s t person in the world, he w o u ld n ’t need such a diet". (10) (a)?? n a na
ixe er0i
m ono
C O M E -P E R F -P -3 S G
o
janis Sen 0a xorusam e only
the
Jo h n
NEG
FUT
fit-
IM P F-P -1P L "If j u s t J o h n h a d come, we w o u ld n ’t have fit". (b) m ono
o ja n is
n a ixe
er0i Sen 0a
xorusam e
"Even if j u s t J o h n ha d come, we w o u ld n ’t have fit". In (8), we have " a n expression specifying an ex tre m e value", in (9) a real (i.e. m ark e d as such) su p e rlative and in (10) "just" w hich m a r k s its focus as being low on a scale. In none of these examples, however, does the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 79 concessive read in g arise auto m atically . Now consider exam ples (ll)-(13): (11)
‘stin aOina n a to -th e
A th e n s
pas na
Sen
vriskis
g o -P E R F -N P -2 S G
Sulia’ NEG
find-IM PF-N P-2SG
jo b "Even if you go to A T H EN S, you w o n ’t get a j o b 1’ . (12) tin Eleni n a th e
silavun
Sen 0 a
Helen n a a rre s t-P E R F -N P -3 P L
mafiun NEG
FUT
tip o ta lea rn -P E R F -N P -
3PL nothing "Even if th ey a rre s t HELEN, th e y w o n ’t learn anything". (13) ‘o iSravlikos th e p lu m b e r ta
na
min na
ixe er0i
NEG
0a
c o m e -P E R F -P -3 S G
FUT
k a ta f e r n a m e ’
th e m m a n a g e -IM P F -P -IP L "Even if the P L U M B E R h a d n ’t come, we would have m anaged". In (11) w h a t is preposed 2 is a prep o sitio n a l com plem ent, in (12) the d ire c t object, in (13) the subject. C rucially, however, in none of these exam ples is the preposed phrase an o u t-o f-c o n te x t scalar word, i.e. a word t h a t could in some imaginable sense fit in K o n ig ’s list (cf. Nikiforidou 1988). Nevertheless, placing these c o n s titu e n ts in first position results in th eir becoming scalar term s and acquiring e x tre m e values on the given scales (e.g. "A thens" in (11) ranks high on a scale of likely places to get a jo b , i.e. the m eaning is something like "even if you go as far as A thens...". 2 It should be clear t h a t I am not using the terms "preposed" or "preposing" in a transformational sense. All they serve is to contrast this type of sentence with the sentence where na and the verb form appear in the beginning
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 80 "Helen" in (12) ra n k s high on th e scale of people who if arre ste d , are likely to let th e police know w h a t hap p e n ed - "even if they go so far as to arrest Helen..."etc.). W h e th e r these c o nstituents have end-of-scale m eanings or w h e th e r they sim ply denote som e extrem e (tow ards a n end-of-scale) value does n o t seem to be im p o r ta n t. In fact, we could t r e a t th e m eanings th a t these term s acquire as relative, in the way t h a t K a y (1990) analyzes "even" - see 2.1.2. W e could say, t h a t is, t h a t this ty p e of concessive clause in Greek, w ith th e "crucial" c o nstituent preceding th e verb m arks its c o n te n t as being m ore inform ative th a n a n implied (series of) other propositions. T h e overall se m a n tic im pact of sentences like (7)-(10) and (11)-(13) is t h a t since the consequent will hold for th e m ore inform ative (and o ften least likely) of th e antecedents, it will also hold for any other a n te c e d e n t "less inform ative"
(or more likely). (In 3.2, however, I am
going to suggest t h a t K a y ’s a c c o u n t runs into problem s w ith a n o th e r type of concessive construction). In the na concessives featuring a preposed con s titu en t, the series of a n te c e d en ts which are set up by the protasis, consist in propositions whose only c hanging p a rt is the p a r t corresponding to the focused constituent. In o ther w ords, w h a t is in focus is not the whole VP b u t only the c o n s titu e n t in th e beginning of th e clause. T his is indicated in the English tra n s la tio n by th e capital letters which m ark the focus. W ith respect to this construction, it m ight be argued t h a t preposing the c o n stitu e n t is n othing b u t an o th er p ragm atic device which Greek employs in order to get the concessive reading. Even if this is th e case, I am going to argue below t h a t this does not m ake the G reek na concessives any less problem atic for the tru th -c o n d itio n a l/p ra g m a tic story. F or now, I ju s t w a n t to point o u t t h a t th e re has not been any system atic correlation recognized in Greek betw een th e first position in a sentence and a focused
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 81 c onstituent. Joseph a n d P hilip p a k i (1987:98-99) n ote t h a t, if a n ything, the clearest correlation b e tw e en position a nd focus holds for the sentence-final position of direct o b je c ts a n d case-marked in direct objects: if these consti tu en ts are "m oved"
to final position, they have to be in te rp re te d as
e m p h a tic focus. M a ck rid g e (1985:234-239) also notes t h a t th e focus, if any, tends to be placed a t t h e end. A final th in g we need to clarify is w hat e x actly counts as preposing. I m entioned in 2.2.2, t h a t th e only element, besides negation, w hich can come between na a n d th e following verb form is an o b jec t personal pro noun.
This, however, does n o t count as preposing, so, for example, (14)
can only be in te rp re te d as a conditional (and, therefore, w ith this p a rtic u lar apodosis it is p ra g m a tic a lly odd): (14)?? n a na
tu
eleyes
he-G E N
kalim era
0a se
sa y -IM P F -P -2S G
good
evrize
m orning
FUT
you
swear-IM P F-P "If you said good m o rn in g to him, he would sw ear a t you". I have also e n c o u n te re d examples where m o re t h a n one c o n stituent is preposed. In this case, th e resulting in te rp re ta tio n is again scalar.
Con
sider, for example, (15) a n d (16): (15) ‘kalim era
me
to n kalitero
good-morning w ith th e
best
tropo n a way
tis na
pis she-G E N
say-P E R F -
NP-2SG 0a
se
FUT
vrisi’ you
sw e a r-P E R F -N P -3 S G
"Even if you say good m orning to her in the nicest way, she will swear a t y ou".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 82 -
(16)lo
janis m e to
th e
John
golf
w ith th e
na GOLF
er0i na
Sen
xoram e’
c o m e -P E R F -N P -3 S G
NEG
fit-
IM P F -N P -1 P L "Even if J o h n w ith his Golf comes, we d o n ’t fit". Regardless of w h e th e r we ta k e th e c o n stitu e n t s tr u c tu r e in (15) to be [ [V NP]
PP] or [V N P PP] it is obvious t h a t the direct o b jec t and the
prepositional a d ju n c t do not m a k e u p a constituent. T h e sam e is true for (16) w here th e preposed c o n s titu e n ts are this tim e the s u b je c t and a prepositional
a d ju n c t.
In a n y s o rt of tra n sfo rm a tio n a l o r
m ovem ent
account, trying to relate the tw o types of stru ctu res, this fact would present a serious problem . The
in te rp re ta tio n
in these
constructions,
w ith
two
c o n stituents
ap p earin g before the verb, is analogous to the cases where only one consti tu e n t is a t th e beginning. B oth c o n stitu e n ts acquire a scalar (or m ore informative) m eaning. In (16), for exam ple, the speaker is saying t h a t even if J O H N comes (who in this c o n te x t has to be in te rp re te d as c ontrasting w ith e.g. a heavier person) a nd even if he brings the Golf (as opposed to, e.g.,the Mini), we would not fit^. T his is similar to the cases which Fillmore, K ay and O ’C onnor (1988) call m ultiple foci sentences w ith English "let alone" as in (17) (Fillm ore, K a y and O ’C onnor ex.75): (17) Y ou couldn’t get a poor m a n to wash y our car for $2 let alone a rich m a n to wax your tru c k for Si.
3 Example (18) could have a reading where there is only one focus, under which "John" and the "Golf" would make up one constituent. This is the "complex NP" reading in which "John with the Golf" serves to distinguish "John" from other individuals without, e.g., cars and which would be identical to the cases with only one focused constituents. The problematic case for the transformational analysis is the double focus reading.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 83 T h e na clause w ith tw’o c o n s titu e n ts in the beginning corresponds to the first p a r t (i.e. the pre-"let alone" p a rt) of (17), show ing this "piling effect" in the in te rp re ta tio n . E vidence t h a t w h a t is a c tu a lly in focus in this type of na clause is (are) the preposed constituent(s) (which, in each case, are s e ttin g up a relevant scale), can be fo u n d in exam ples like (18) and (19) w h e re the focus (or foci) a re m a d e explicit by the presence of a n o th e r c o n tra s tin g focus (foci): (18)
uiski n a w hiskey
pji
na
Sen
0a
meGisi
d r in k - P E R F -N P - 3 S G
NEG
FUT
get
d ru n k -
P E R F-N P-3SG poli ligotero bira m uch
less
beer
"Even if he d rinks W H ISK E Y , he w o n ’t get d ru n k , m u c h less (if he d rin k s ) beer". (19)
uiski to proi w hiskey
na
th e m orning
pji
Sen 0a
me0isi
n a d r in k - P E R F -N P - 3 S G
NEG
F U T get
drunk-P E R F -N P poli ligotero bira much, less
beer
to the
vraSi evening
" E v e n if he drinks W H IS K E Y in the M O R N IN G , he w o n ’t get d ru n k , m u c h less beer in the evening". O n e final case it m ay be useful to look a t is th e case where th e pro tasis of t h e na conditional contains only an in tra n sitiv e verb with no overt s u b je c t p h rase present (person a n d
n u m b e r will always of course be
m ark e d on th e verb). Consider for exam ple (20):
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 84 -
(20)
‘n a na
fonaze
0a
ton
sc re a m -IM P F -P -3 S G
akuyam e’
FUT
him
h e a r-IM P F -P -IP L
"If he (had) scream ed, we would (have) h e a r(d ) him ". E xam ple (20) has a conditional in te rp re ta tio n . Also, however, (and it seems to me p red ic ta b ly so), w ith the right (squeal) intonation, it can also have a concessive reading, as for exam ple in (21): (21) n a na
FONAZE
ben
sc re a m -IM P F -P -3 S G
0a
ton
NEG
aku-yame FUT
him
hear-IM PF -P-
IPL "Even if he h a d scream ed, we w o u ld n ’t have he a rd him ". In such cases, g r a n te d th e intonation difference, the transfer between a conditional a n d a concessive in te rp re ta tio n is relatively free, because, I subm it, in such cases th e re is no o ther "preposeable" constituent; the focused (by the in to n a tio n ) c o n stitu e n t is the V P itself and its the verb which sets up a scale. E xam ples (22) and (23) also illustrate the same point: (22) ‘na
ixe klapsi
0a
ton
n a c ry -P E R F E C T -P A S T -3 S G
singinuse’ FUT
him
m ove-IM PF-P-3SG
"If she had cried, she w ould have moved h im " . (23)
na na
IX E K L A P S I
ben
c ry -P E R F E C T -P A S T -3 S G
0a
to n
N EG
singinuse FUT
him
move-
IM PF-P-3SG "Even if she had cried, she w o u ld n ’t have m oved him ". (I am changing th e po larity of the m ain clause in ord er to give to the sen tence the pragm atically m ost plausible in te rp re ta tio n .
T h e point is th a t
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 85 even if I h a d n ’t, th e sentence with this kind of intonation would have to be interpreted as evoking incom patibility betw een the antecedent and the consequent) In th e na concessive construction, therefore, it is the constructional p a tte rn associated w ith a p a rticu lar w ord o rd e r which picks the focus (or foci) a n d forces an "even" reading despite th e fact t h a t no "even" is present. S chem atically, we could represent th e construction as in (24): (24)
X
na
V (Y)
, Z
scalar X is the v a riable sta nding for the fro n te d constituent(s), V is the verb form which follows na, Y is the variable s ta n d in g for w hatever consti tuents follow (w hich m ay be 0) and Z is the v a riable sta n d in g for the con sequent. O n th e sem antic side, we will have to specify a t least th a t X m u st be interpreted as scalar. In unificational term s, this m eans t h a t if X is a priori scalar, th e n this feature unifies w ith th e general m eaning of the con struction. If X is not scalar, then scalarity is inherited from the construc tion. This p a rtic u la r sem antic p ro p e rty can only be a ttr ib u te d to the specific form in (24). It is not a fact a b o u t th e sem antics of th e individual parts t h a t X sh o u ld have this p a rtic u la r m eaning, since, as we have seen, it is often the case t h a t X becomes scalar only b y virtue of its appearing in this particular position, in this p a rtic u la r c onstruction.
Even if (granted
the intonation difference) we take a na clause w ith no constituent o ther than the verb in th e beginning to be am b ig u o u s betw een an "even" and a conditional reading, the fact rem ains t h a t (24) w ith the fronted consti tu en t can only h a v e an "even" reading. This is N O T a general fact a b out conditionals, it is a fact a b out na conditionals. Consider (25)-(27) which
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 86 f e a tu re t h e "regular" conditional m a rk e r an (cf. also 2.2.2): (25) o n a s is
an
on a ssis
im un
8a
if b e -P -lS G
to
FUT
a 7 o ra z a it
b u y - IM P F - P - lS G
"If I w ere Onassis, I would b u y it". ??onasis an im un
Sen 0a
to
a 7 o ra z a
"?If I were Onassis, I w ou ld n ’t buy it". (26) k a lim e r a
an
g o o d -m o rn in g
tis if
pis
she-G E N
ola
0a
s a y -P E R F - N P - 2 S G
pane everything
FUT
g o - P E R F -N P -3 P L .L P k a la well "If y o u say good m orning to her, e v e ry th in g will be fine". ??kalim era
a n tis
pis
0a
se
vrisi
good m o rn in g if she-G EN s a y -P E R F -N P -2 S G
FUT
you
swear-
PE R F-N P-3SG "?If you say good m orning to her, s h e ’ll sw ear a t you". (27)
to th e
poSilato an bicycle
ftasi
if stin
pari
(*8en) 0a
ta k e -P E R F -N P -3 S G
(N E G )
FUT
ora tu
a rriv e -P E R F -N P -3 S G to-the tim e his "If he takes the bicycle, h e ’ll (*won’t) arrive in tim e". E x a m p le s (25) an d (26) d e m o n s tra te t h a t to the extent t h a t only a concessive in te rp re ta tio n is notionally available, the sentence is a t best very o d d . T hings get even clearer when, as in (27), the preposed phase is no t a n a priori scalar term , in which case th e "even" reading is impossible
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 87 to get (this is in dicated by
the s ta r in th e negation m orphem e. T he
impossible reading to g e t is the one in which t h e b a c k g ro u n d u n d e rs ta n d ing is t h a t taking the bicycle would help one arrive in time). E xam ples (25)-(27) show t h a t word order is not enough to produce an "even if" reading. E x a m p le (28) below shows t h a t e m p h a tic stress on the focus alone can n o t give rise to concessiveness either: (28) ?? an im un O N A 3 IS Ben 0a to a 7 oraza "Even if I were Onassis, I w o u ld n ’t b u y it". F inally, concessives co n ta in in g the an conditional differ from na concessives in t h a t ke is u n a m b ig u o u sly concessive in th e la tte r form er. T o get a concessive-only reading o u t of an
b u t not in the
an conditional,
the
clause has to be in tro d u c ed by akoma k e /k i (cf.2.1.3). C om pare, for e x a m ple, (29) to (30): (29) ki an erGun
Sen 0a pam e
ke if c o m e -P E R F -N P -3 P L N E G F U T g o - P E R F - N P - I P L (a) "And if they com e, we w on’t go". (b)"Even if they com e, we w o n ’t go". (30) a k o m a ki an e r0 u n Sen 0a pam e "Even if they come, we w on’t go". T h e re
is an
in to n a tio n
difference corresponding to
the two different
in te rp re ta tio n s of (29) (m a jo r stress on an for the "and" reading, m a jo r stress on erQun for the "even" reading). T h e fac t rem ains, however, t h a t th e ke of the ke...na concessives can only m e a n "even" a n d is, therefore, "constructionalized" in a way t h a t the k e /k i o f an an concessive is not. T h e latter, as we have seen, has both "an d " a n d "even" readings still. Again, co n tra ry to K o n ig ’s predictions, no obvious generalizations seem to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 88 -
be available reg a rd in g the relationship of conditionals and concessives and a description is required on a construction basis. T his conclusion is fu rth e r en h a n ce d by th e different focus strategies em ployed in the two types of concessive clauses. As we have seen, in th e k e ...n a construction ke a n d its focus m u s t be a d ja c e n t and in th e beginning o f the clause. In contrast, in the akom a ki an concessive, th e connective is alw ays imm ediately followed by the verb p h ra s e a nd focus differences are sim ply m arked by intonation, as in (31): (31) a k o m a ke a n ferun to fortryo Sen x o ram e a k o m a ke if b rin g -P E R F -N P -3 P L th e tr u c k N E G fit-IM P F -N P -IP L "Even if th ey bring the truck, we w o n ’t fit". (32)*akom a ke to forti-yo an ferun Sen x oram e D epending on w h e th e r the stress is on ferun o r on fortiyo, two different scales a re set up corresponding to th e tw o different focuses (bring not bring, tru c k vs. car). As (32), however, shows, the option
vs., e.g.,
of "moving"
the focus next to akoma ke is not available, a lth o u g h (cf. (25)-(27)) there is no general restriction against m oving a ny c o n s titu e n t before the an condi tional m arker.
3.1 .3 T h e rela tio n sh ip b etw een c o n cessiv es an d co n d itio n a ls In C h a p te r (2), I looked a t th e sem antics of na as a conditional. In the section above, I exam ined na clauses with a n "even if" reading, w h a t for the tim e being I call concessives. In this section, I will look a t the rela tion of the conditional and the concessive
construction, outside and
beyond the tru th -c o n d itio n a l predictions.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 89 In th e preceding section, I argued t h a t one of th e inevitable conclu sions we m u s t d ra w is t h a t na conditionals a n d na concessives m u st be considered to be two distinct constructions. W h a t I w a n t to discuss now is w hy the question of conflating th e m would have arisen in the first place. It seem s clear t h a t both
na conditionals and
na concessives are
related "uses" of a na clause, a n d both very distinct from the o ther uses of na listed in 2.2.1. In term s of distribution, th e y are the only uses of na where this form is not governed by a lexical governor (as we have seen, na can be governed by a verb, a noun, an adjective, a preposition and a con ju n ction). In its conditional a n d concessive environm ents,
na is again
s u b o rd in a te , b u t this tim e the governor is a m ain clause.
M oreover, in
c h a p te r (5), I will show th a t certain types of cross-clausal dependencies (between th e a n te c e d en t and th e consequent) o b ta in both in the condi tional a n d th e concessive constructions. T his again goes to show t h a t con ditional a n d concessive (conditionals) are closely related constructions. Sem antically, both conditionals and concessive conditionals involve an a n te c e d e n t a n d a consequent an d a relation betw een th e two. Beyond this general d escription, however, we also need to describe this relation in each case (cf. c h a p te r (5)). Finally, historical evidence of th e sort addressed by Konig (1986) a n d H aim an (1986), which shows th a t conditional connec tives in several languages develop into concessive (i.e. "even" type) connec tives, also argues for a close sem antic relationship between conditionals an d concessive conditionals of this sort. In C h a p t e r (2), I showed t h a t the use of th e na conditional is res tricted
to d u b ita tiv e /c o u n te rfa c tu a l contexts.
It follows t h a t in cases
where th e protasis is contextually given a n d its t r u t h is established na con ditionals are ruled out. So (33) with a na (b u t n o t w ith an an) conditional
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 90is u n g ram m atic a l: (33) - erxete
o
com e-IM P F -N P -3S G
ja n is
avrio
the John
tom orrow
"John is com ing tom orrow ". - a n /* n a an / * n a
erxete
pes tu
n a se
com e-IM P F -N P -3S G tell-IM P h im
voiGisi su bj.you h e lp -P E R F -
N P -3SG "If h e ’s coming, tell him to help you". In the preceding section I pointed o u t som e of th e form al a n d s e m a n tic characteristics of th e na concessives, arg u in g t h a t in some respects they c a n n o t b u t be considered a different c o n stru c tio n from th e no conditional. Now I w a n t to discuss one aspect in which th e se m a n tic s of na concessives m ay be (at least) m o tiv ate d . As I said in 3.1.3., th e sem antic effect of t h e construction in (24) is to force a scalar in te r p re ta tio n of the c o n s titu e n t which fills the first slot in th e a n te c e d en t clause. It is clear t h a t in m a n y cases th e scale which is set up is com pletely c o n text-dependent, and in th is sense we m ig h t m ore a p p ro p riate ly ta lk a b o u t the s e m a n tic -p ra g m a tic effect of the construction or rath e r, of th e prag m a tic s being p a r t of th e sem antics of the c o n struc tion. T h e first c o n s titu e n t, by v irtue of being in the position where it is, acquires a scalar reading and this sets up a w h o le series of a n te c e d en t pos sibilities of which the an te c e d en t which c o n ta in s it is in te rp re te d as the th e least likely of all th e others.4 4 Kay 1990 derives this counter-to-expectation idea t h a t characterizes many sentences containing "even" from the fact th a t, normally, asserting something which is further along (or more informative) on a scale violates a (quantity) conversational implicature arising from the assertion of a lower (or less informative) point on the same scale. Asserting th a t "she reads Latin", for example, conversationally implicates th at "she doesn't read Sanskrit" (if the background assum ption is t h a t Sanskrit is harder than Latin and therefore further along th e more informative scale). Asserting t h a t "she (even) reads Sanscrit"
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 91 T h is s e m a n tic /p r a g m a tic function of th e na concessive construction, w hereby one is asserting t h a t the c onsequent will hold for any of the a n te c e d e n ts w h o se focused c o n stituents are lower on th e relevant scale th a n th e focused constituent which actually a p p e a r s in the ante c e d en t, is, I claim , in c o m p a tib le with "givenness" (A lternatively, we m ay th in k of this as scalar sem antics correlating w ith
"non-givenness").
If "given"
along the lines o f (33) above is ta k e n to refer to som ething w hich is en tirely e sta b lish e d or accepted as tru e in c o n te x t, th en it is obvious w hy a na concessive c a n n o t be in terpreted as "given". T h e p ra g m a tic force of this c o n s tru c tio n is exactly to m ake a novel c o n trib u tio n to the conversa tion by offering a t least one piece of in fo rm a tio n (i.e. the one represented by the focused co n stitu e n t) w hich is new (regardless of w h e th e r we analyze this c o n tr ib u tio n as more inform ative, less likely, m ore extrem e, etc.). Consider, for e x a m p le (34)-(37): (34)a. -o jan is ben akui the J o h n
poli kala
N E G hear-3SG very well
"Jo h n c a n n o t hear very well", b.-kufos n a ine
prepi
na
akusi
d eaf n a be-N P-3SG m ust-IM P su b j. h e a r-P E R F -N P -3 S G "Even if he is DEAF, he should h e a r (the w a y we are shouting)." (35)a. -o ja n is ine
kufos
the J o h n be-3SG deaf " Jo h n is d e a f" . violates this implicature. For the purposes of the present discussion, the point is not how this interpretation arises b u t simply t h a t it is available.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 92 b.-*kufos n a ine prepi na a kusi "Even if he is D EA F, he sh o u ld be able to hear". (36)a. - 0 a er0i o
antiproeSros
F U T come-3SG the vice-president " T h e vice-president will com e", b.-o proeSros n a er0i th e president
b en sozomaste
n a c o m e -P E R F -N P -3 S G
N E G save-M D -IM P F-N P-
1PL "Even if the P R E S ID E N T (himself) comes, it w o n ’t save us". (37)a. -0 a er0i o
proeSros
F U T come-3SG th e p re s id e n t " T h e president will com e". b.-*o proeSros na er0i Sen sozom aste " E ven if the president comes, it w o n ’t save us." T h e p o in t of examples (34)-(37) is sim ply t h a t the sentences with the fronted constituents are b ad (or inap p ro p ria te ) as replies to s ta te m e n ts which establish as tru e or given th e c o n te n t of the protasis. So for exam ple (3 4 b )(= 3 5 b ), the na concessive construction, is not a good answer to (35a) b u t it is a good answer to (34a) w here its protasis can be seen as providing new inform ation. T his c o n s tr a in t is d ue to the p ra g m a tic force associated with this construction which in tu rn arises from th e in te rp re ta tio n t h a t the construction is forcing on its first constituent. "Non-givenness" in th e case of na concessives is, as I said, a corrolary of scalarity. Na conditionals, on the o th e r han d , are also excluded from cases w here the content of the protasis is completely given (cf. c h a p te r 2,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 93 exam ple 39), even if, in th e case of na conditionals, non-givenness is a by p ro d u c t of the general c o n s tra in t in term s of the d u b ita tiv e /c o u n te rfa c tu a l contexts. In a ny case, na conditionals are perfectly com patible w ith na concessives in t h a t th e y b o th correlate with non-givenness, a n d in exactly th is respect, it m a y be sa id t h a t this s e m a n tic /p ra g m a tic aspect of na con cessives is doubly m o tiv a te d .
3 .2 . K E N A concessives This section will e x am ine th e type of dependent concessive clause introduced by ke na. T h e description will focus m ainly on tw o issues:
the
sy n ta x (and especially th e c o n s titu e n t structure) of these constructions and some of the s e m a n tic /p r a g m a tic properties of th e ke na concessives in rela tion to the na concessive construction described in the previous section. T h e obvious difference betw een the na and the ke na clauses is t h a t in the la tte r we have w h a t we m a y call an explicit, lexical "even" (i.e. ke). P a r t of the point in this section will be to show t h a t not all th e differences between the two con stru c tio n s can be a ttr ib u te d to the presence of ke, at least not in any obvious way. Examples of ke na concessives are listed in (38)-(40) below: (38) ke n a er8i
i m a r i a Sen 0 a er8i
even n a co m e -P E R F -N P -3 S G the M ary
o janis NEG
FU T
c o m e -P E R F -N P -
3SG the John "Even if M ary comes, J o h n w o n ’t (come)". (39) ‘ke noritera n a e rx o ta n even
earlier
Sen 0a prolavenam e’
n a com e-IM P F -P -3S G NEG
F U T be in tim e-IM P F -P -
1PL
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 94 -
"Even if he h a d come earlier, we w o u ld n ’t have m ade it". (40) ‘ke n a min ixate xtipisi
to k u 8 u n i 9a sas a k u y a m e ’
even n a N E G rin g -P E R F E C T -P -2 P L th e bell
F U T you hear-
IM P F -P -1 P L "Even if you h a d n ’t rung the bell, we would have heard you".
3 .2 .1
W o rd -o rd er and c o n stitu e n c y co n sid era tio n s Exam ples (41)-(44) illustrate th e various possibilities in w ord-order
(an d focus) p a tte rn s in the ke na concessives: (41) ke
i m aria n a er0i
Sen im aste
a rk e ti
even the M ary na c o m e -P E R F -N P -3 S G N E G b e -N P -IP L enough "Even if M A R Y comes, we are n o t enough". (42) ke
to kalitero tis forem a n a foresi
even the best
her dress
fenete
apesia
n a w e a r-P E R F -N P -3 S G look-EMPF-NP-
3SG terrible "Even if she wears her B E ST dress, she looks terrible".'*
5 We may note here an interesting distinction first observed by Brugman (1986). Consider an example like (i): (i) ke tin jineka sto kokino aftokinito na ixe 5i even the woman in-the red car na see-PERFECT-P-3SG Sen 0a boruse n a t i n anaynorisi NEG
FUT can-IMPF-P-3SG subj. she-ACC recognise
"Even if he had seen the woman in the RED car, he wouldn’t have been able to identify her". In (i) the prosodic focus is "red"; however, what "women in cars") and t h a t is what Brugman calls the only possible way of rendering the reading example (i). Example (ii) can only mean t h a t what
is actually being contrasted here are entities (namely th e "contrast focus". Note t h a t in the Greek examples where the contrasted elements are women, is like in is being contrasted is cars:
(ii) ke sto kokino aftokinito na ixe 8i tin jin eka Sen 0a boruse na tin anaynorisi The generalization for Greek, therefore, should be t h a t in cases where the contrast and the prosodic focus are not identical, ke should be immediately next to the contrast focus.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 95 (43) ke
a sk im a n a tis fer0is
even b a dly
Sen 0a 0imosi
n a her b e h a v e -P E R F -N P -3 S G N E G
FUT
get angry-
P E R F-N P-3SG "Even if you behave BADLY to her, she w o n ’t get m a d " . (44) ‘ke n a er0i
to "yrama
even n a com e-P E R F -N P -3 S G th e Sen 8 a maQume NEG
oli
letter
tin isto ria ’
F U T l e a r n -P E R F - N P -I P L whole th e story
"Even if the letter COM ES, we w o n ’t learn the whole story". T h e first question we need to address is w h a t is th e s tru c tu re in this ty p e of clause (i.e. in the a n te c e d e n t clause). T h e ke w ord has a fixed posi tion in t h e clause ( th a t is, it c a n n o t a p p e a r in any o th e r t h a n th e first position in the sentence), so t h a t, for exam ple, (43’) is u n g ram m atic a l: (43’) *na tis fer0is ke askim a Sen 0 a 0imosi If an in te rp re ta tio n is available for such non-initial ke, it has to be an " a ls o /a n d " in te rp re ta tio n , cf. (45): (45) n a er0i
ke o
jan is Sen 0a xoresum e
n a com e-P E R F -N P -3S G a n d th e J o h n
NEG
F U T fit-P E R F -N P -
1PL "If J o h n comes also, we w o n ’t fit". W e m ay ask, therefore, w h e th e r ke a n d its focus m ake up a c o n stituent. T h e re is really no conclusive test t h a t m ay be applied here b u t some indi cations m a y come from the observation t h a t alongside (41)- (44) we have also sentences like (46) (see also (15), (16)) w ith double foci which obvi ously do n o t m ake up a single c o n stitu e n t:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 96 (46) ‘ke olokliri brizola mesti mesi even whole
steak
tis
im eras n a fai
in-the m iddle d .a rt. d a y -G E N n a e a t- P E R F -N P -
3SG (pali) 0a p in a i’ still
F U T be hun g ry -IM P F -N P -3 S G
"E ven if he eats a W H O L E steak in the M ID DLE of the day, h e ’ll (still) be h u n g ry ". T h e p o in t here is sim ply t h a t since the direct o b je c t and the tim e a d v e r bial do n o t represent one c o n stituent, a n d since they are b o th in focus, even if we assum ed t h a t ke a nd its focus(es) m a d e up a c o n stituent, it would be h a rd to know w h a t to call such a node.® Finally, w ith respect to exam ple (44) we m a y ask w h e th e r ke a nd na m ake u p a single constituent. T his is the way t h a t ke na is tre a te d in m ost tra d itio n a l gram m ars (cf. for exam ple T z a rtz a n o s 1963), th e idea being t h a t it represents a complex concessive connective w ith o u t necessarily analyzable parts. However, if we tak e (41)-(43)
on the one h a n d , a nd
(44)
on the other, to be instances of the sam e c o nstruction, there is no reason to suppose t h a t we have a ke na con stitu e n t. Sem antically, all ke is doing in (44) is picking as its focus th e whole V P or t h e whole clause and, th ere fore, as is the general rule with this co n stru c tio n it has to be im m ediately 0 This situation (see also (15) and (16) with the "ke-less" construction) is similar to t h a t presented by other focus constructions. For example, Fillmore, Kay and O ’Connor (1988) in their description of the let alone construction note t h a t the things t h a t can be conjoined by let alone need not be syntactic constituents. Consider, for example (i): (i) A poor man wouldn’t wash, let alone a rich man wax, your car for )2, let alone your truck for $1. Given t h a t let alone is also a focus construction, this may not be so surprising. On the other hand, not all focus constructions can feature more than one focus, unless these foci make up a constituent (cf. for example Prince 1981 for a description of topicaliiation and Yiddish movement as other focus constructions). The it-cleft, for instance, cannot focus more th a n one constituent (Mt was Yoahiko on Tuesday t h a t I saw- cf. also McCawley (1988 58-60). It was Yoshiko in the car t h a t I saw, is grammatical, I think, only if "in the car" is a modifier of "Yoshiko").
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 97 to th e left of its focus. In fact, the very existence of exam ples like (41)(43), where there is an intervening c o n s titu e n t between ke a n d na, argues a g a in s t their being one c o n stituent. F u r th e r m o r e , there is no reason to assum e th a t na an d the following verb phrase do not m ak e u p a consti t u e n t (which would be th e consequent of having a s tru c tu re like [[ke na] [VP...]]; the only c o n stituent t h a t can come betw een na a nd th e following V P is the negation and an object personal pronoun, and these, as we have seen in 2.2.2, are exactly the sam e c o n s titu e n ts which can come betw een na a n d the rest of the V P in the cases w here na acts as a c om plem entizer a n d where, therefore, there is every reason to believe t h a t na introduces a c o n s titu e n t (e.g., Qelo na f iy o "I w a n t to go").
3 .2 .2 K E N A: som e 9 em a n tic/p ra g m a tic p rop erties T h e difference between a na and a ke na concessive is t h a t the la tte r contains ke. T h e examples so far show t h a t there are no u n p red ic ta b le differences between the tw o constructions. W h a t the na concessive does by w ord order, the ke na construction does by having ke m ark a focus as well as by requiring initial ke a n d focus order; gram m ar, in o th e r words, plays a role here also. Both constructions serve to p u t one co n stitu e n t into focus se ttin g up a series of possible a n te c e d en ts (which m ay of course be j u s t two). In w hat follows, I will discuss tw o properties of the ke na concessive which, a t first glance, differentiate it from th e na clauses a n d c a n n o t be a ttr ib u te d in any obvious way to the presence of ke.
C onsider exam ples
(47), (48): (47) -avrio
erxete
o
petros
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 98 tom orrow com e-IM P F -N P -3S G the P e te r " P e te r is coming to m o rro w ". -ke n a erxete
(pali) e x u m e
na
kanum e
even n a com e-IM PF -N P-3S G (still) h a v e -N P -I P L su b j. d o -N P -IP L tin pio poli 8ulia moni m as th e m ost
work alone w e-G E N
"Even if he is coming, we still have to do m ost of th e work ourselves". (48) - ‘tin
ixe xtipisi
i z esti’
she-A CC h it- P E R F E C T - P - 3 S G the h e a t "She had suffered a h e a t s tro k e ". - ‘ke n a tin
ixe xtipisi
i zesti eprepe
even n a she-ACC h it- P E R F E C T - P - 3 S G the h e a t m ust-P -3S G na
prospatjisi
na
e r0 i’
subj. try -P E R F -N P -3 S G su b j. com e -P E R F -N P -3 S G "Even if she had suffered a h e a t stroke, she should have tried to com e". If we take ke to have the sem antics of o th e r scalar term s along th e lines of th e analyses given for "even" in c h a p te r (2), th en th e use of th e ke na sentences in (47), (48) is a little bit surprising. In all th e o th er exam ples (cf. (4l)-(44)), ke can be an a ly z e d as focusing a c o n s titu e n t a n d th u s as m arking a proposition which is m ore inform ative (in K a y ’s term s) or less likely t h a n some other assum ed or expressed s ta te m e n t in the context. In these cases, however, the sp e ak e r of th e ke na concessive is only repeating inform ation
which
is
a lre a d y
given
in
the
context.
The
sem a n tic s/p ra g m a tic s of such clauses is to convey to th e addressee t h a t the speaker is willing to g r a n t
the t r u t h
or the correctness of th e
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 99 a ddressee’s s ta te m e n t for the purpose of being able to a ssert th e conse quent. W e may call this use "speech-act" use (cf. also th e discussion on as in th e next section). W ith respect to th e se m antic m otivation, it is clear t h a t "givenness" is n o t p a r t of the o r d in a r y sem antics of ke in its use to m e a n "even". It is also not p a rt of the se m antics of a na conditional (cf. 2.2.2) or a na con cessive. In c h a p te r (4), w hen I consider the issue of m o tiv a tio n , I argue t h a t although this use of ke na is certainly d istinct from th e clearly scalar cases, it is also rela ta b le to or m o tiv a te d by th e scalar s e m a n tic s of ke na in its "regular" use. However, the scalarity which is involved in examples like (47) and (48) is of a different n a tu re , involving the s p e a k e r ’s degree of c o m m itm en t to th e c o n te n t of th e consequent. W e m a y th in k of this as scalarity a t the level of th e illocutionary force (or sp e ech-act level), ra th e r th a n scalarity in th e s u b je c t m a tte r . (47), (48) a n d th e like are, therefore, not scalar in the sam e way t h a t (41)-(44) are, a lth o u g h th e y still share this general feature of in co m p a tib ility between the c o n te n t of the protasis and t h a t of the apodosis. W e m ay p a ra p h ra se w hat th e s p e a k e r of (47) says, for example, with th e following: "although the fact t h a t P e te r is coming m ay lead one to e x pect t h a t the work will be divided, th is is n o t actually so".7 Note t h a t as we m ig h t actually expect w ith such exam ples, there is no difference in m eaning correlating w ith the difference in focus (i.e. there are no different scales set up because there a re no scales or, rath e r, no scales of the sam e kind). So, for example, (50) a n d (51) in th e context of 7 If we take the speech-act use of ke na to involve no scalarity whatsoever, one way of "reconciling" the two functions of ke na might be to think of this in terms of what T ra u g o t t (1988, 1989) calls "pragmatic strengthening", i.e. th e process by which a previously contextuatly triggered meaning of a word is "promoted" to the sta tu s of an independently available (conventional?) meaning. For ke (or "even"), for example, we would w ant to say t h a t the counter-to-expectation or unlikely frame which is often arising from the violation of the implicature of the scalar semantics (cf. footnote 3), in these examples functions as the meaning proper itsetf. This, however, is little more th a n a description of the d a ta and not a real argument as to which analysis is correct.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 100 (49) are s y n o n y m o u s a nd both are a p p ro p r ia te answers to (44): (49) ine
p a m b lu to s
be -N P -3 S G very rich "He is very rich". (50) ke p a m b lu to s n a ine
prepi
na
sevete
tu s alus
even very rich n a be-N P-3SG m u s t- IM P - N P subj. respect-IM P F -N P 3SG the oth ers "Even if he is v e ry rich, he should show some respect for the o thers". (51) ke n a ine p a m b lu to s prepi na sevete t u s alus
E x a m p le s (52), (53) below exem plify f u rth e r the difference betw een th e na a n d th e ke na concessive constructions a n d the fact th a t the form er b u t not the la tte r is restricted to "non-given" contexts: (52) liyo a p o a fto n a ixe pji little
0a ixe peBani
from th is n a d r in k - P E R F E C T -P - 3 S G F U T d ie - P E R F E C T -P -
3SG "Even if he h a d d ru n k a little of this he would have died". (53) ke
liyo apo afto n a ixe pji
8 a ixe peBani
even little fro m this na d r in k - P E R F E C T -P - 3 S G F U T d ie -P E R F E C T P-3SG O ne
m ig h t
suppose
sentences
(52)
and
(53)
were
com pletely
synonym ous, b o th having the two readings w hich the English tra n s la tio n has: (a) He d i d n ’t d rin k any of the stuff, (b) He d ra n k a lot or all of it. If we t u r n th e c o u n te rfa c tu a l into th e corresponding negative sta te m e n t, then, in th e term s of Horn (1985), re a d in g (a) would correspond to the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 101 m etalinguistic reading of the negation: "He d i d n ’t d rin k a little, in fact he d i d n ’t d r in k any". However, whereas (53) does have b o th readings, (52) is only a p p ro p r ia te in a co n tex t w here "he d id n ’t d rin k a n y ". T h e difference in th e tw o readings c a n n o t be a ttr ib u te d to a difference in focus and in fact P a u l K a y (p.c.) has convinced me th a t the assum ed c o n tr a s te d propo sition is the sam e for bo th interpretations, nam ely t h a t he d r a n k a lot. T h e difference can be described as follows: F o r the "he d r a n k a lot" in te r p re ta tio n this assum ed contrasted proposition (i.e. th e c o n te x t propo sition) is set in the a c tu a l world, which is the world n orm ally co n tra stin g to th e w orld of a cou n te rfa c tu a l antecedent.
F o r this in te r p re ta tio n , the
c o n te x t proposition can be som ething like w h a t we have in (i) below ("cp" m a r k s th e context proposition, " tp " the text proposition): (i)
—pe0ane ja ti ipje olokliri tin b o tilja
(cp)
"He died because he d r a n k th a t whole b o ttle". —ke li-yo apo afto na ixe pji 0a ixe peBani (tp) "Even if he had d r u n k a little of it, he would have died". F o r th e "he d id n ’t d rink any" in te rp re ta tio n , however, th e c o n te x t propo sition is n o t set in the a ctual world b u t in a n o th e r hyp o th etic a l world: (ii) ~ 8 o k s a
to 0eo Sen ipje ka0olu
" T h a n k God, he d i d n ’t d rin k any". an to ixe pji 0a ixe pe0ani
(cp)
"If he h ad d ru n k it, he would
have died".
—ke li 7 0 apo afto n a ixe pji 0a ixe pe0ani (tp) "Even if he had d r u n k a little, he would have died". In (ii), the cp is neither set in th e actual world (in which
he d i d n ’t drink
any) nor in the hypothetical world of the c o u n terfactu al a n te c e d e n t of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 102 tp (in which he d r a n k a little). R a th e r, it is set in yet a n o th e r hypothetical world in which he d r a n k it all. According to th e above, therefore, (52) c a n n o t be an a p p ro p ria te answer to (54): (54) pe0ane
jati
ipje
olokliri tin b o tilja
die-P E R F -P -3 S G because drink-P E R F -P -3S G whole
th e bottle
"He died b e c ause he d r a n k th e whole bottle". This difference derives, I claim, from the "given-non-given" distinction which was discussed above and which distinguishes th e na from the ke na concessive con stru c tio n . I argued above t h a t sim ple na concessives are non-givens. It follows t h a t (52) cannot be used in a co n te x t where "he actually d ra n k a little": (52’) — ipje
liyo
d rin k -P E R F -P -3 S G
eftixos a little
fo rtu n a te ly
"He d r a n k a little bit, fortunately". —* liyo a
apo a fto n a ixe pji little
fro m
this
na
0a ixe
pe0ani
d r in k -P E R F E C T -P -3 S G
FUT
die-
P E R F E C T -P -3 S G "Even if he h a d d r u n k a little, he could have died". On the other h a n d , (54) entails th a t "he d ra n k a little" a n d this is w h a t rules out (52) as a n a p p ro p ria te answer to (54). T h e "explanation", in other words, is th e s a m e one which accounts for (52’), except t h a t in this case w hat counts as "given" is an entailm ent r a t h e r th a n th e "surface" sentence itself.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 103 T h e ke of th e ke na concessive is u n d o u b te d ly related to th e n o n c o n d itio n a l ke. Syntactically, they can both a tta c h to the sam e k in d of c o n s titu e n ts. Sem antically, except for the cases where this special speecha c t re a d in g seems to be the only one available for the conditional ke (cf. (47),
(48)),
conditional
and
non-conditional
receive t h e sam e kind of analysis.
ke can s tr a ig h tf o rw a r d ly
However, w ith the non-conditional ke,
th e concessive reading was c o n te x tu a lly triggered (and therefore, c o n te x t u ally cancellable as well), w hereas for th e ke in th e conditional e n v ir o n m e n t th e only possible reading is a concessive one. T h a t is, when a p p e a r in g in th e c o n stru c tio n in (55), ke can only be concessive: (55)
ke (X) n a V (Y) , Z (consequent) In th e na a n d ke na types of concessive clauses, we have w h a t Lakoff
(1987:507-509) calls m inimally different constructions. B oth are focus co n stru c tio n s w ith very similar s e m a n tic a n d p ragm atic functions. H ow ever, th ere are also differences which m u s t be explicitly specified in th e d e scrip tion of each of these separately, if we w a n t these descriptions to be a n a d e q u a te guide to the use of each®.
3.3
A S co n cessiv es
3.3.1 A S or K I AS? T h e issue I will exam ine in th is section is w hether the concessive clauses in question are in troduced by as or by ki as. T h e question arises w ith respect to alternations like th e ones exemplified in (56)-(57): 8 Yamaguchi (1989) comes to a similar conclusion with respect to certain types of Japanese concessive clauses.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 104 -
(56) 8en ixe perasi
sto
8iayonism a
N E G p a s s - P E R F E C T -P -3 S G to-the ki a s /a s
lei
exam
etsi
"an d " a s /a s s a y -IM P F -N P -3 S G
so
"He h a d n ’t passed the exam , even if he says so". (57)
??ki a s /a s
lei
etsi Sen ixe perasi
"and" a s /a s say -IM P F -N P -3 S G so
s to
NEG pass-P E R F E C T -P -3S G
to -th e Siayonism a "Even if he says so, he h a d n ’t passed the exam ". M ackridge (1985:298) claims t h a t in its concessive use as is alw ays im m ediately preceded by k i (the allom orph of ke before vowels). However, his exam ples are all like (56), w here the concessive clause follows th e m ain clause. He fails to notice, therefore, th a t in cases w here th e clause in tro duced by as precedes, in tro d u c in g the clause w ith ki as is distin ctly odd, if n o t un g ram m atic a l. T z a rtz a n o s (1963) also notes t h a t in its concessive use as is often preceded by k i ( w ith o u t specifying the e n v iro n m e n ts ) b u t is also careful to note t h a t as and k i as are synonymous. T h e question, therefore, is w h e th e r as or ki as is th e c o n v e ntional way of introducing this type of concessive clause. In view of e x am ples like (56), w here bo th ki as a n d as are possible introducers, a n d of (57) w here ki as is n o t ap p ro p riate , M ack rid g e ’s claim is obviously w rong. W h e th e r the con cessive is introduced by as or by k i as, the resulting m ea n in g is th e sam e. T h e difference is t h a t if as (as opposed to ki as) in tro d u c e s the clause, th en th e as clause appears to be s ta n d in g on its own as a se n tence (m uch like th e non-concessive clauses introduced
by
as), a n d
ex a m p le
(56), for
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 105 instance, is read w ith a b re a k in th e middle. Consider also (58)-(59): (58) as
vreksi
II 0a
pam e
as rain -P E R F -N P -3 S G
FUT
g o - P E R F -X P -I P L
"Even if it rains, w e’ll (still) go". (59) 0a pame II as vreksi "W e’ll go even if it rains". In contrast, w hen th e concessive is in tro d u c ed by ki as, the sentence is read as a com plex sentence containing a reg u la r su b o rd in a te ( b u t cf. next p a ragraph) clause.
T h e generalization on th e distrib u tio n of as and
ki as is th a t th e y are inte rch a n g e a b le when th e concessive clause which th ey introduce follows b u t ki as is odd when the concessive comes first. Consider (57) above a n d also (60), (61): (60) ‘i sakorafa prye
poli k a la stus
ayones
the S akorafa g o -P E R F -P -3 S G very well to -th e ki a s /a s
m in
pire
games
to xriso’
" and" a s /a s N E G ta k e -P E R F -P -3 S G the gold "Sakorafa did very well in the games, even if she d i d n ’t get the gold (m edal)". (61) ??ki a s /a s m in pire to xriso i sakorafa p iy e poli kala stus ayones "Even if she d i d n ’t get th e gold, Sakorafa did very well in the gam es". This in t u r n suggests t h a t ke still retains some of its co n ju n c tio n status. One way of looking a t this, is to say t h a t ki as m ay be on its way to becoming a g ram m aticized concessive connective (like, a nd p e rh a p s under the influence of, ke na) b u t it h a s n ’t y e t generalized its use to all
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 106 -
environm ents. We m ay speculate t h a t this preference for the en v iro n m en t w here the concessive clause follows has, as I said, something to do w ith ke coming from "and" a n d / o r w ith th e ob se rv a tio n t h a t in general (not j u s t in Greek- cf. Konig 1986), this position seem s to favor or even force a con cessive reading (so if a concessive is d o u b ly m ark e d anywhere, we would expect it to be in this position).
C onsider, for example, (62), (63), w ith
the "regular" conditional an: (62) 0a
pam e
an
xionisi
F U T g o -P E R F -N P -IP L
if
sn o w -P E R F -N P -3S G
"W e’ll go (even) if it snows". (63) an xionisi 0a pame "If it snows, w e’ll go". In (62), th e speaker m ay be expressing e ith e r a real com m itm ent to going (" W e ’ll go anyway, even if it snow s"), or a regular ("if and only if") condi tional.
(63) on the oth er han d , seems to have only a purely conditional
reading (cf.also English "I will go if it rains" vs. "If it rains, I will go"). Finally, consider examples (64)-(67): (64) 8ulevi
oli tin im era (ki) as ine
arostos
work-EMPF-NP-3SG whole th e d ay (ke) as be-NP-3SG sick "He works all day even i f/a lth o u g h he is sick". (65) ‘Sulevi oli tin im era (ki) as ine k e a ro s to s ’ (66) 8en kim i0ika
ka0olu to vraSi
NE G sle e p -P E R F -P -lS G a t all (ki) as
pira
th e evening
farm ako
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 107 (ki) as t a k e - P E R F - N P - lS G m edicine (ki) as ta k e - P E R F - P - lS G
m edicine
"I d id n ’t sleep a t all at n ig h t even if /a lth o u g h I took medicine". (67) ‘8en kimiBika to vra8i (ki) as p ira k e f a r m a k o ’ C o m p a re d to the ke na concessives w here ke occupies a fixed (first) posi tion in th e sentence, an d , as I show ed in th e preceding sections, is u n a m b i guously concessive, th e (second) ke in (65), (67) is n o t p a r t of the as con stru c tio n .
F o r one thing, th e "even" rea d in g of ke in these cases can be
cancelled. In an exam ple like 8e n kimiQika kaQolu to vrabi (ki) as pira k e farm ako k e xapia "I d id n ’t sleep a t all a t n ig h t a lth o u g h I took (both) m edicine and pills", ke (in bold face) fu n ctio n s as "an d " in the context of th e tw o ke construction ( = " b o t h ...a n d " ) .
As I suggest in the following
section, there is a difference in m eaning b e tw e en th e as or ki as clause w ith o u t this (second) ke a nd w ith it, a difference w hich is predictable given the m eaning of ke and the m eaning I propose for the as concessives. T he p o in t here is sim ply t h a t this second ke sh o u ld n o t be confused with th e ke in ki as.
3.3.2 T h e A S sem a n tics In this section, I will investigate the s e m a n tic /p r a g m a tic properties of the as concessives, arguing t h a t a t least in one respect, as concessives are different from both ke na (i.e.
even if concessives) a n d an ke ( even though)
concessive clauses. Consider, for exam ple, (68), (69): (68) ke na
fiji
o ja n is eyo 8a
even na leave-P E R F -N P -3S G the J o h n
m ino I
F U T sta y -P E R F -N P -
1SG
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 108 -
"Even if Jo h n leaves, I ’ll sta y ". (69) an ke 6a if even
fiji
o janis e yo 0a m ino
F U T lea v e -P E R F -N P -3 S G the J o h n
I
F U T s ta y -
PER F-N P-1SG "Even th ough J o h n will leave, I will sta y ". In (68) we have th e ke na c o nstruction, in (69) an "even th ough" ty p e of concessive introduced by an ke. In th e n e x t c h a p te r, I will talk m ore a b o u t this kind of concessive a n d the problem s it poses for a com positional analysis of these s tru c tu re s (notice t h a t " a n d if" comes to m ean "even if" b u t "if a n d " comes to m ea n "even th o u g h "). W h a t is relevant to th e present discussion is t h a t from (68) we can d ra w no e n ta ilm e n ts as to t h e tr u t h of th e antecedent.
E x a m p le (68), as is also the case with th e
corresponding conditionals of this form , does n o t convey a n y th in g as to w hether J o h n will or will n o t leave. In c o n tra s t to the conditional, h o w ever, (68) does entail the consequent, in o th e r words, the speaker conveys th a t in e ither case h e /s h e will stay. O n the o th e r h a n d , (69) clearly e n ta ils or presupposes the tr u t h of the a n te c e d en t. E n ta ilm e n t seems to be th e right notion here since, for one thing, t h a t " Jo h n will leave" is n o t preserved u n d e r negation. T his has been the basis of a tra d itio n al division between th e so-called concessive conditionals a n d the "pure" or fa c tu a l concessives (cf. Konig 1986) a n d an a rg u m e n t for the form er being j u s t an o th e r (pragm atic) variety of conditionals. I have p a rtially addressed th is question in 3.2. T h e point of th e following discussion is to see if and h o w as concessives fit into this bin ary (factual vs. non-factual) d istinction. Consider (70)-(72): (70) ‘as
exi
kali foni 8en
b o ru m e
na
tin
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 109 as ha v e -N P -3 S G good voice afisume
na
N E G c a n - P E R F - N P - I P L subj. her
tra y u S isi’
l e t - P E R F - N P - I P L subj. sin g -P E R F -N P -3 S G "W e c a n n o t let h e r sing even if she HAS a good voice". (71) ‘as
exane
tin psixrem ia tu 8en
as lose-IM PF-P-3SG th e na
ton
subj. him
cool
his
epre p e
N E G m u st-IM P F -P -3 S G
voiGisis’ he lp -P E R F -N P -2 S G
"You s h o u ld n ’t have helped him even if he DID K E E P losing his calm ". (72) ‘as m in as
NEG
ixe erGi
sto pa rti m u eyo Ga p ao
c o m e -P E R F E C T -P -3 S G to -th e p a r t y m y
I
F U T go-
P E R F -N P -1 S G sto to -th e
8iko t u ’ his
"I will go to his p a rty even if he D ID N ’T come to m ine". W h a t the sp e aker is doing in (70)-(72), by using an as clause, is saying som ething like the following: "I will g r a n t (you) (the t r u t h of) w h a t was ju s t said; nevertheless I am still going to s a y / t h i n k / d o w h a t I was going to anyw ay". Notice t h a t in the English tra n sla tio n s of (70)-(72), "even if" is am biguous betw een a "true" (i.e. scalar) in te r p re ta tio n of "even" and a speech-act in te rp re ta tio n (as in exam ples (47), (48) above).
T h a t is we
m ay in te rp re t th e translation of (70), for exam ple, as c o n tributing by the use of the "even" clause a contrasting focus (in this case "good voice"), "higher up" or m ore informative on the rele v an t scale th a n some other
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 110 focus in a n assum ed proposition.
O r, alternatively, we m ay in te rp re t (70)
in the w ay we interpreted (47), (48) , as not m ak in g any novel c o n trib u tion by th e use of th e "even if" clause, b u t sim ply as accepting as the basis for a f u rth e r assertion som ething which has already been c o n te x tu ally established. On
th e o th er hand,
Greek
as clauses can only have the la tte r
in te rp re ta tio n , i.e. w h a t in 3.2.2 I called th e speech-act use. T h is is w h a t a n as concessive codes and therefore a m ore a p p ro p r ia te tra n sla tio n of all th e as concessives so far m ight be to replace th e whole concessive clause b y "even so". T h e re a re a few things to note here. If we ta k e th e as clauses to m ean som ething roughly like "I’ll g r a n t you w h a t you said, nevertheless I still m a in ta in X ", then we may expect t h a t an as concessive is in ap p ro p ria te in a context where th e speaker w a n ts to prese n t the c o n te n t of the m ain clause as following anyw ay by virtue of asserting t h a t it holds for an extrem e value. So, for example, (70) is an a p p ro p r ia te response to (73) b u t n o t to (74): (73)
exi
kali foni
nomizo
h ave-N P -3S G good voice th in k - IM P F -N P -lS G "She ha s a good voice, I th in k " . (74)
exi
m etria foni
nomizo
h ave-N P -3S G mediocre voice th in k - IM P F -N P -lS G "She has a mediocre voice, I th in k " . O n the o th e r hand, an "as...ke" clause com bines the as a n d th e ke m ea n ings in a predictable sort of way. W e could p a ra p h ra se exam ple (75) for instance as " I’ll g r a n t you w h at you said a n d I ’ll g r a n t you even m ore
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- Ill -
th an th a t" : (75) as
exi
ke
kali foni Sen
b o ru m e
as ha v e -N P -3 S G even good voice
na
tin
N E G c a n - I M P F - N P - I P L subj.
her afisume
na
trayuSisi
l e t - P E R F - N P - I P L subj. sing-PE R F-N P-3SG "I’ll g ra n t you t h a t she even has a good voice. W e still c a n ’t let h e r sing". Predictably, (75) is a good response to (74) b u t n o t to (73). In all (70)-(72), however, as well as in (75), th e " I’ll g ra n t" or th e "let’s say or agree" p a r t of the meaning is there.
T h e re is a difference in
meaning, in o th e r words, between (75) a n d (76), a difference which m ay be best described in term s of a different fram ing. In (76), w ith ke na, the speaker is fra m in g th e concessive as a conditional, saying am ong other things t h a t h e /s h e does not in fact know w h e th e r th e person in question has a good voice. In (75), with as, on th e o th e r h a n d , th e fram ing of the concessive is d one via the speech act force associated w ith as. One of th e implications of this fram ing is the conveyance by th e speaker of the idea th a t w h e th e r th e c o n te n t of the as clause is tru e o r not does not m atte r. W h a t m a tte rs is t h a t for the purposes of th e conversation the speaker will accept it as true. (76) ke kali foni n a exi Sen borume n a tin afisume na trayuSisi This difference is partially reflected in the f a c t t h a t ke...
na conces-
sives b u t n o t as concessives exhibit tense dependencies betw een antecedent and consequent. C o m p a re for example (72) above (with a P E R F E C T ,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 112 P A S T form in the as clause a nd a f u tu re in the m a in clause) w ith (77): (77) ke even
na
ixe erOi
i m a ria 8en
na
c o m e -P E R F E C T -P -3 S G
th e
M a ry
NEG
FUT
come-
P E R F E C T -P -3 S G 0a ixe e r0 i/ FUT
*0a
er0i
o ja n is
c o m e -P E R F E C T -P -3 S G
FUT
c o m e -P E R F -N P -3 S G
the
John "E ven if M ary h a d come, Jo h n w o u ld n ’t h a v e /* w o n ’t com e". In (77), th e perfect p a st form in the a n te c e d e n t requires a p a st verb form in th e consequent (the perfect p a st as in this exam ple, or the im perfective past). Interestingly, th e only case w here (77) is g ra m m a tic a l w ith a n o n p a s t v e rb in the consequent is when we give it w h a t I ’ve been calling a speech-act reading; t h a t is (77) w ith the f u tu r e in th e apodosis is accept able as a n answer to (78): (78) i m aria ixe er0i
sto
p a rti tu
omos
the M ary c o m e -P E R F E C T -P -3 S G to -th e p a rty his nevertheless "M a ry had come to his p a rty , nevertheless". If, as I suggested, the as clauses have this special kind of concessive sem antics associated w ith th e m , th e n it seems clear t h a t the distinction betw een factual a n d non-factual concessives does n o t apply. T h e speaker of an as concessive does n o t say e ither t h a t "I d o n ’t know if p" or t h a t "I know t h a t p"; he sim ply sta te s t h a t for th e purposes of th e conversation, we m ay accept or agree t h a t it is true. T h is p a rtic u la r p ra g m a tic function w hich is really the sem antics of the as concessives, m a y be tra c ed to its non-concessive meaning; In 2.3.2, I noted t h a t one function of the non-concessive as is to introduce a clause
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 113 w hich expresses perm ission. Depending on the tense of th e as clause, this m ay be in te rp re te d as d irect granting of perm ission on the p a rt of th e sp e aker or as a s ta te m e n t to the effect t h a t th ere was nothing preventing th e person(s) referred to from doing so m e th in g .
Consider, for instance,
(79), (80): (79) as
er9i
as c o m e -P E R F -N P _ 3 S G (80) as
"He m a y com e".
e rx o ta n
as co m e-IM P F -P -3 S G "He could h a v e come (there was n o th in g p rev e n tin g him )". In all th e exam ples w ith the non-concessive as th e perm ission expressed by th e speaker h as to do w ith th e "real-w orld” . T h e speech-act (of perm is sion) conveyed by such exam ples is a speech-act whose intended effect is to influence the real w orld developm ents, or, in th e case of a p ast verb form, to express w h a t could have been done o r h a p p e n e d . Following Sweetser (1990:49-75), we m a y call this meaning th e c o n te n t m eaning of as and we m ay p a ra p h r a s e it in general with "I allow (you) to do X". T here m ay a lot m ore to say a b o u t the sem antics of perm ission, if we w anted to d e te r mine w h a t is it t h a t such a fram e includes. S earle’s (1984:14-15) discussion of directives, for exam ple, m ay be relevant to d e te rm in in g th e presupposi tions (sincerity conditions), propositional c o n te n t a n d general illocutionary force in these cases. B ut, for my purposes, w h a t is im p o r ta n t is sim ply t h a t in such cases th e speech act is o p e ra tin g in (or affecting) the world, or, as Searle p u ts it, the direction of fit is "from w orld to words" (i.e. the purpose of th e sp e aker is to get the real world situ a tio n to m atch his words).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 114 -
O n th e o th er ha n d , one way to th in k of th e concessive m eaning of as is to view it as a perform ance of an act of perm ission, b u t this tim e in th e speech a c t world. T h e speaker of an as concessive is no longer saying "I allow you to do X" b u t sim ply "I allow you to say X or I g rant you X" for th e p urposes of the conversation an d specifically for the purpose of per form ing th e assertion (or w hatever speech act) which is expressed by th e o th e r clause. T h e co n te n t of th e as clause, in o t h e r words, is no longer a pplying to the r e a l/c o n te n t world b u t to th e speech-act world or the w orld of o u r conversation.
It is a sp eech-act of permission perform ed or
o p e ra tin g in the speech-act world. T h e non-applicability of t r u t h a n d falsity ju d g e m e n ts to th e as p a r t of th e concessive construction stems, I suggest, from their operating in the speech-act world. So, for exam ple, it is w eird to follow up the u tte ra n c e of an as concessive (e.g. (81)) with a s ta te m e n t like (82): (81) as
vreksi
0a
p am e
as ra in -P E R F -N P -3 S G F U T g o - P E R F - N P - I P L "W e ’ll go even if it rains". (82) ben
ine
ali0ia oti 0a vreksi
N E G be-N P-3SG t r u t h
t h a t F U T rain -P E R F -N P -3 S G
" I t ’s not tru e t h a t it will rain". G ra n tin g permission presupposes th a t perm ission ha s been asked for. It is this aspect of the m eaning of c o n te n t as which is preserved in its use in th e world of the conversation a n d accounts for the fact th a t the c o n te n t of an as concessive is already on th e table a n d u n d e r discussion.
As conces
sives are offered as b a c kground to the assertion of th e main clause, and a t least in this respect their sem antics is m o tiv a te d in being directly relatable
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 115 to their non-concessive use. W h a t does n o t follow from th eir c o n te n t sem antics, is t h a t speech act as also codes concession; t h a t is the speech a c t m eaning of as also includes a c o u n ter-to-expectation fram e, a require m e n t which has to be satisfied by the c o n te n t of the two clauses. C onsider for exam ple (83)-(85): (83) 0a
vreksi
avrio
F U T ra in -P E R F -N P -3 S G tom orrow (84) as
vreksi
piknik 0a
"It will rain tom orrow ".
pam e
as rain -P E R F -N P -3 S G picnic
F U T g o - P E R F -N P -I P L
"Even if it rains, we’ll go on the picnic". (85) ??as vreksi piknik ben 0 a pam e "Even if it rains, we w o n ’t go on th e picnic". E xam ple (85) shows t h a t we c a n n o t have the speech a c t use of as followed by a context w hich u n d e r norm al assum ptions would violate the counterto-expectation requirem ent. .4s is not unique in having both c o n te n t and speech a c t meanings. As Sweetser (1990) argues, m a n y lexical a nd g ram m atic a l items show this polysemy between c o n te n t a nd speech a c t m eanings.
T he m odal verbs,
c onjunctions like "an d " a nd "because" and conditional m arkers are am ong such items. C onsider,
for exam ple the different m eanings of "m a y ",
exemplified in (86), (87) (Sweetser 1990:70): (86) He m ay go. (87) He m ay be a university professor b u t he sure is d u m b . (86) exemplifies th e r o o t/c o n te n t m eaning of "may" (I allow him to go) while (87) exemplifies the speech-act m eaning (I allow into our conversa tional world th e s ta te m e n t t h a t he is a university professor). Predictably,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 116 (87) presupposes t h a t the s ta te m e n t t h a t he is a university professor has been a lre a d y m a d e or is somehow con te x tu ally given. In c h a p te r 4, I will pursue such parallelism s be tw e en as a n d o th e r lex ical item s a b i t fu rth e r, suggesting t h a t this polysem y is m ore system atic th a n j u s t looking a t th e sem antics of as w ould lead us to believe.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 117 -
C h a p te r 4
C o m p o s itio n , C o m p o s itio n a lity a n d M o t iv a t io n .
4.0 In tr o d u c tio n T h is c h a p te r will investigate the degree to which the se m a n tic p ro p e r ties of th e concessive clauses discussed in th e previous c h a p te r fall o u t of th e se m a n tic properties of their c o n s titu e n t p a rts . I will not in tro d u c e any new d a t a b u t sim ply consider th e d a t a a lre ad y discussed w ith respect to th e th re e notions appearing in th e title of this c hapter. A lrea d y in the previous c h a p te r, I ha v e h in te d a t some of the difficulties t h a t an analysis adhering to s tric t com positionality would run into. Here, I would like to e x p a n d on t h a t a n d also address th e question of com position from a constructional p o in t of view which recognizes the con s tr u c tio n as the relevant u n it of description.
T h e co m p o n e n t p a rts of a
given construction will be c o n structions them selves w ith p a rtic u la r se m a n tic a n d p ra g m a tic properties to begin with. T h e Greek d a ta argue against th e d octrine of strict com positionality a n d , in p a rticu la r, against the idea of sem antic composition t h a t relies on th e "rule-to-rule" hypothesis. T w o aspects of this view of th e in te raction betw een sy ntax and sem antics do n o t seem to be easily acco m m o d a te d by th e G reek concessives. One of th e m is the idea t h a t th e se m antic composi tion of a syntactically com plex expression is accom plished by iterating local o perations in specific s u b p a rts of the c onstruction.
T h e o th er refers
to th e belief t h a t sem antically in te rp re te d c o n s titu e n ts are first s itu a te d in c o n te x t a n d th en acquire their contextualized in te rp re ta tio n .
T h e Greek
concessive clauses resist this k in d of analysis in tw o respects. F irst, because
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 118 the relevant m eaning of th e c o n s titu e n t parts is often not available unless these p a rts are in a p a rtic u la r g ra m m a tic a l e n vironm ent.
A n d , secondly,
because there are se m a n tic p ro p e rtie s of the constructions which are not derivable from the se m a n tic p ro p ertie s of their p a rts .
S tric t composi
tionality, as assum ed in th e ru le -to-rule hypothesis, denies th e existence of expressio n s/c o n stru ctio n s/id io m s which can be bo th com positional and a rb itra ry at th e sam e tim e a n d th u s misses the fact t h a t th ere m ay be generalizations to be e x tra c te d even from partially com positional expres sions. Since it is possible t h a t technically there m ay be a w a y of describ ing such facts while m a in ta in in g the compositional dogm a, or even, as I suspect is more likely, claim t h a t this range of d a ta is o utside the scope of linguistic analysis p roper, it m a y be more accurate to say t h a t they argue against the spirit if n o t th e lette r of such an a p p ro a c h
to sem antic
analysis. P a rtia l com positionality also pertains to the d istinction m ade by Lakoff (1987:
147-48 a n d elsewhere) between prediction a n d m otivation.
Langacker (1987: c h a p te r 12) d ra w s w h a t I take to be a parallel distinction between (full) com positionality a n d analyzability. L a n gacker w ants to dis tinguish further two kinds of analyzability. One kind refers to a composite stru c tu re [C] including the c o n te n t of structures [A] and [B]; the other, refers to a person specifically ascribing to [C] the co n te n ts of [A] a nd [B]. This, according to L angacker, corresponds roughly to a difference between conscious and unconscious processing (Langacker 1987:457). However, in talking
a bout
the
sem antics
of
a
p articular
co n stru c tio n
as
being
m otivated, I w o n ’t ta k e into a c c o u n t this distinction, a ssum ing th a t speak ers in general recognize the presence of particular co m p o n e n ts in a com po site stru c tu re (w ith o u t im plying
t h a t they system atically
analyze the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 119 c o n trib u tio n of each), despite of the fact t h a t the m eaning of the whole is, more often t h a n not, more t h a n the sum of the m eanings of the com ponents. T h e re is a n o th e r aspect to the com positionality dogm a as represented by the school of M ontague sem antics (cf. Dowty, W all and P eters 1981), Ladusaw (1979)). According to this notion of sem antic composition, given the m eanings of the c o n s titu e n t parts, the m eaning of the whole can be derived via a fixed m a th e m a tic a l function. A lthough there m ay be such a function on th e basis of which composite m eaning is calculated, it c a n n o t b u t suffer on intuitiveness, due to th e kind of units which p a rticip a te in the m eaning-com posing function. P racticioners of this kind of analysis characteristically assum e an overall theory of m eaning which excludes all kinds of cognitive s tru c tu re s a n d relationships.
F o r conditionals, this
tra n sla tes into tru th -c o n d itio n a l m eaning exhausting th e meaning descrip tion, whereas I would like to argue t h a t reference to sem antic fram es (which by definition include knowledge stru c tu re ) seems to be a necessary p a r t of the description of the co n stra in ts on the d istrib u tio n of such sen tences.
4.1 T h e N A
and A N
co n d itio n a ls and concessives: s em a n tics
d ep en d in g on form I will s t a r t by considering the na concessives, which as I have shown in c h a p te r 3, come in tw o varieties, exemplified in (1) a n d (2) below: (1) 8eka xronia n a perasun 8en 0a to ksexasi ten years n a p a ss-P E R F -N P -3 P L N E G F U T it forget-P E R F -N P -3S G
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 120 "E ven if ten years go by, he w o n ’t forget it". (2) ke 8eka x ro n ia n a perasun Sen 8a to ksexasi "Even if ten years go by, he w o n ’t forget it". E x a m p le (l) fea tu re s the na concessive, example (2) has th e ke
na type of
clause. E x a m p le s (1) a n d (2) are concessive conditionals a n d according to th e t r u t h conditional analyses I outlined in the previous c h a p te r (Konig 1986, B e n n e tt 1982), th ey are sem antically indistinguishable from "regu lar" conditionals, since m eaning is e quated with t r u t h conditions. The difference in m e a n in g between (1) a n d (2) and the c o rresponding condi tionals is relegated to the realm of pragm atics, arising from th e cancella tion of "conditional perfection" (cf. c h a p te r 3) due to the presence of a w ord such as ke or a n y other end-of-scale term in the co n d itio n a l clause. However, as we have seen, in the type of th e concessive clauses exemplified in (1) a n d (2) concessiveness does n o t arise autom atically. T h e concessive reading requires t h a t the relevant p h rase or word appears in a fixed posi tion in th e sentence, namely the first position. So, for exam ple, (3) is pri m arily a conditional while (4) is a concessive: (3) n a ixe kafsona avrio Sen 0a pije n a m e n a have-P -3S G h e a t wave to m o rro w
NEG F U T g o - IM P F - P - lP L
"If there was going to be a h e a t wave tom orrow , we w o u ld n ’t go". (4) kafsona n a ixe avrio
8a pije n a m e
h e a t wave n a have-P-3SG to m o rro w F U T g o -IM P F -P -IP L "Even if th ere was going to be a h e a t wave tom orrow , we w ould (still) go". G ra n te d the va ria tio n in ju d g em e n ts w ith respect to (3) (some subjects reported t h a t w ith exaggerated stress the relevant phrase, exam ples like
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 121 (3) m ig h t be read also as concessives), th e fact rem ains t h a t (4), even w ith o u t ke, can only have a concessive reading. Moreover, the wrord or phrase which a p p e a rs in first position in such stru c tu re s does not have to have an a priori end-of-scale or sim ply scalar word. A n y word in this posi tion will require the co n stru c tio n of a p rag m a tic scale as, for exam ple, (5): (5) to vivlio n a Siavasum e pali Sen 0a k a ta lav u m e tip o ta the book n a r e a d - P E R F - N P - I P L still N E G
F U T u n d e rs ta n d -P E R F -
N P -1PL a n y th in g "Even if we read th e B O O K , we still w o n ’t u n d e rs ta n d a n y th in g " . T h e point w ith such exam ples is t h a t
the use of the w ord "book"
does not norm ally evoke a scalar model in the in te rp re ta tio n of the sen tence on w hich reading the book" would represent a point. Y et in this constru ction "book" evokes a scalar in te rp re ta tio n . W e c a n n o t equate this with some so rt of topicalization or an y oth er focus construction since the preposed p h rase is n o t simply in focus or topicalized, b u t requires a scalar (as opposed to a ny o ther) in te rp re ta tio n .
This is a case, therefore, where
we have a p a rtic u la r m eaning associated with a construction. Once the p a tte rn is recognized as such it can be claimed to have general an d p ro ductive p ro p e rtie s 1. As I said in c h a p te r 3, we can simply represent the preposed p hrase with a variable a nd know t h a t th e filler of this slot will be given a scalar in te rp re ta tio n regardless of its in h ere n t sem antics.
1 Fillmore (1989) m akes the same point with respect to the English ditransitive construction (V N N), which can be superimposed on certain verbs increasing their valence and forcing particu la r in te rpre tations which cannot b u t be associated with the construction. G oldberg (1989) explores in detail the semantics of this arg u m e n t structure, arguing th a t specific semantic constraints should be directly associated with the ditransitive structure r a th e r tha n with the individual verbs which participate in the construction.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 122 W ith in the constructional p a tte r n , m eaning can be considered to be fully compositional. A n y term in t h a t first position will unify w ith the se m a n tic s of the c onstruction, unless it has some sem antic fe a tu re or specification which clashes w ith the c o n structional semantics. (However, I h a v e n ’t been able to th in k of such a w ord in o rd er to test this). Notice t h a t this is very different from saying t h a t th ere are two words "book" or tw o m eanings of the w ord "book", one w ith non-scalar and the o th e r with scalar semantics; instead, th e claim is t h a t scalarity c onstru ction.
W h a t is required is
is a p ro p erty of the
t h a t the relevant constituent(s)
are
found w ithin the configuration in I: I. ([(ke) X a a V ” ] [ V ” ]| Y e t, not all conditionals associate scalar sem antics w ith w ord order. C onditionals introduced by an require explicit m ark in g w ith akoma ke. E x a m p le (6) cannot be read as a concessive a nd is th u s p ragm atically odd, w hereas (7) is a regular concessive: (6) ?? plusii an itan
Sen 0a
rich if be-P-3PL
to a y o ra z a n
NEG
FUT
it buy -IM P F -P -3 P L
"If th ey were rich, th ey w o u ld n ’t be able to buy it". (7) a k o m a ke an itan plusii
Sen 0a to a y o ra z a n
"Even if they were rich, th e y w o u ld n ’t be able to buy it". M oreover, as (7) shows, different conditionals em ploy different focusing strategies. While in th e na a n d ke na concessives th e focused c o n s titu e n t always comes first, in the akoma ki an concessive conditionals th e verb directly follows the akom a ke connective a nd focused constituents are sim ply m a rk e d by in to n a tio n as in (8) vs. (9):
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 123 (8) ‘a k o m a ki an "yrapsis
oli tin 8iatrivi
still and if w rite -P E R F -N P -2 S G
whole th e dissertation
prepi na minis na teliosis ke ta t ip i k a ’ m ust-IM P subj. sta y -P E R F -N P -2 S G subj. finish an d th e formalities "Even if you write the whole dissertation, you have to s ta y a n d finish up with the formalities also". (9) *akom a ke oli tin 8iatrivi an yrapsis prepi na minis n a teliosis ke ta tip ik a Such facts suggest strongly t h a t conditionals and concessive condi tionals have to be tre a te d as distinct constructions. T h e claim t h a t the difference can be a ttr ib u te d solely to prag m a tic s and t h a t it is the purely compositional p ro d u ct of a word like "even" or other end-of-scale term s is no t easily m aintained. Concessiveness seems to be associated with p a rticu lar p a tte rn s which differ from conditional to conditional a n d as such they are an inherent p a r t of the gram m ar of such clauses.
Leaving aside the w ord-order differences, I now come to the semantics of ke an d the o th e r concessive connectives considering th e m from a compo sitional point of view. In ch a p te r (2) I suggested th a t ke m a y occasionally be interpreted as "even" in a non-conditional en v iro n m en t a lthough this m eaning should be r a th e r tre a te d as an im plicature th an as a conventional interpretation.
In the preceding chapter, I showed t h a t ke in the ke na
construction is unam biguously concessive, while in c o m bination with the an conditional it is not. T h e relevant co n tra sts are exemplified in (10)-(12) below:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 124 (10) ke n a fijis to ra Sen prolavenis to leoforio a n d n a lea v e -P E R F -N P -2 S G
NEG c a tc h -IM P F -N P -2 S G th e bus
"Even if you leave now, you w o n ’t catch the b u s ". (11) ki an fijis t o r a
Sen prolavenis to leoforio
"A nd if you leave now, you w o n ’t catch the bu s". "Even if you leave now, you w on’t catch the b u s " . (12) a k o m a ki an fijis to ra
Sen prolavenis to leoforio
"Even if you leave now, you w o n ’t catch the b u s ". T h e crucial ex a m p le is (11) which, although in tro d u c e d by ke, can have both a n "and" a n d a n "even" in te rp re ta tio n . Unless we ta k e the k e ’s in (10) a n d (11) to be hom onym ous (which goes a g a in s t all intuition), it seems inevitable t h a t some reference m u st be m a d e to th e constructional environm ent. W e need to know th a t it is not j u s t ke composing with a conditional b u t t h a t it is ke composing w ith an an o r a n a conditional in order to know which are th e available sem antic in te rp re ta tio n s .
These
observations, coupled with the behavior of ke in o t h e r concessive construc tions m ake a s tra ig h tfo rw a rd compositional analysis very problem atic. I h aven’t yet discussed the factive concessives (cf. n e x t ch a p te r), b u t the point a b o u t com positionality can be m ade here. C onsider, for exam ple (13): (13) an ke Sen to n kalesan if a n d N E G
ir0e
h im inv ite-P E R F -P -3 P L c o m e -P E R F -P -3 S G
"Even th o u g h th ey did not invite him, he cam e". As (13) illustrates, ke and an, depending on conventionalized o rd e r (since we cannot m ak e this order system atically produce th e m eaning difference), produce radically different m eaning. In (11) and (12) we have an "even if"
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 125 clause while in (13) th e tr u t h of the protasis is entailed. These facts sug gest t h a t alth o u g h th e p a rticip a tio n of ke in such sentences m a y be m o tiv a te d , it is also conventionalized w ith respect to th e co n stru c tio n in which it is found; sim ply saying t h a t ke m eans "even" will not necessarily pred ic t the overall m ea n in g of a given concessive clause. T h e only generalization which seems to be e x tractable is t h a t ke in these biclausal s tr u c tu r e s is m a rk in g concessiveness, a lth o u g h its exact c o n trib u tio n m a y differ from construction to construction. T h u s besides t h e ke na, akom a k i an an d an ke c onstructions, we have ke a p p e a rin g in t h e type of concessive clause exemplified in (14). The concessive clause in (14) is also a d istin c t construction w ith special properties. T h e p o in t here is sim ply th a t it is also m a rk e d by ke: (14) ke pu xtipise
k a nis Sen jirise n a ton kitaksi
ke t h a t h it- P E R F -P -3 S G
nobody N E G tu rn -P E R F -P -3 S G s u b j. him
look-P-3SG " W h a t if he was injured! Nobody tu rn e d to look".
In ch a p te r 2, I a rgued th a t describing th e constraint on the d is trib u tion of na conditionals requires reference to a "non-given" condition which is contextually d e te rm in e d a n d which can be relaxed or tig h ten e d d e p e n d ing on the s p e a k e r’s intended
m eaning
and
the addressee’s imposed
in te rp re ta tio n . N a conditionals, like an conditionals, can express various different m eanings d e p e n d in g on the verb form in the a n te c e d e n t a n d the consequent,
but
in
all
cases
this
general
restriction
to
d u b ita tiv e /c o u n te r f a c tu a l contexts is superim posed on a n y o th er in te r p re tation.
We m ay th in k of this as a fram e (in th e sense o f Fillm ore 1976,
1982, 1985) evoked by na conditionals. T h e n a tu r e of this fram e is a little
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 126 b it different from th e ones described by Fillmore in t h a t d o u b t a b o u t the t r u t h of a c e rtain s itu a tio n is not knowledge generally s h a re d a n d con s ta n tly evoked by use of a word or a c onstruction, b u t r a th e r it is knowledge d e te rm in e d by the im m ediate c o n te x t a n d / o r a se t of limited assum ptions sh a re d by th e speaker a n d the addressee.
In this sense, it
m a y be m ore a c c u r a te to say t h a t a na clause tells you how to fram e a s itu a tio n 2.
N evertheless, it is knowledge of a p rag m a tic
n a tu re , not
analyzable in te rm s of sem antic features, and since it is evoked every tim e a na conditional is used, this is yet another case where the p ra g m a tic s are p a r t of th e sem antics of a given expression (in ou r case, construction). On the basis of th e word order differences, I argued t h a t na condi tionals a n d
na concessives are distinct constructions. T h e w ord order
difference is n o t in a n y way compositional, since it c a n n o t be taken into account in a m a th e m a tic a l, m eaning-com posing function. It can be argued to be a t best m o tiv a te d , if it is taken as highlighting a p a rtic u la r consti t u e n t and, therefore, as producing focus c ontrast. On the o t h e r hand, as I said above, th ere is no w ay of predicting t h a t this p a rtic u la r focusing will result in a scalar in te rp re ta tio n for th e preposed c o n stitu e n t. I have also argued, however, t h a t th e scalar inte rp re ta tio n associated w ith such stru c tures correlates w ith a n d m otivates "non- givenness" insofar as the pro tasis in these co n stru c tio n s specifies some point which is f u rth e r along on some scale th a n a p o in t specified in some other, c o n te x tu ally given or assumed proposition. In this respect, th e speaker is m aking a novel (non 2 The same distinction is made in Kay (1987), who discusses the inte raction between inferences w arra n te d by the g ra m m a r of the text and the "parsimony principle" (which characterises the ideal reader). Roughly speaking, Kay expresses this as follows: By choosing a scenario which fits the relevant frame (here the d ub itative schema), we achieve a matching at the sem antic level of something which is already required by the (gram m ar of) the text.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 127 -
given) co n trib u tio n . We could say, therefore, t h a t the pragm atics of a na conditional, t h a t is the pragm atic c o n s tra in t conventionally associated w ith it, composes.
I think of this as p ra g m a tic com positionality in the
sense t h a t it is th e pragm atics of a na conditional which determ ines w hat things it likes to go with or p a rtic ip a te in. T h is way of talking a b o u t things, however, requires th a t we take seriously th e idea t h a t the compos ing p a rts are constructions which m ay bring with them some special sem antics and pragm atics and t h a t prag m a tic s m ay also m o tiv ate pa rtly the kinds of m eanings t h a t will arise or th e kinds of m eanings a given con stru c tio n will unify with. T h e question of m otivational sem antics arises also w ith respect to w h a t I called th e speech-act use of ke...na, exemplified here in (16): (16)— o janis erxete
avrio
th e Jo h n com e-IM PF-N P-3SG tom orrow "Jo h n is coming tom orrow ". — ke n a erxete
pali exum e
n a k a n u m e oli tin 8ulia
ke n a com e-IM PF-N P-3SG still h a v e - N P - l P L subj. d o - N P - IP L whole the work "Even if he is coming, we still have to do all th e work". In discussing such examples in th e previous c h a p te r, I suggested t h a t they do not involve scalarity in the sam e way t h a t the "regular", non-speechact uses do. Here, I would like to e laborate on t h a t further.
T h e scalar
model evoked by the non-speech-act uses is the same in this case as well, except t h a t it is now in the backg ro u n d . Consider (16) as p a r t of the dialogue in (17):
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 128 (17) — exume na k a n u m e oli tin 5ulia j a to p a rti "W e have to do all this w ork for th e p a rty " . — m a erxete o jan is avrio "B ut John is coming to m o rro w ". — ke na erxete pali e x u m e n a k a n u m e oli tin 8ulia "Even if he is coming, we still have to do all the w ork". T h e background u n d e rs ta n d in g
in this conversation involves a scalar
model very much like th e one evoked by the content use of th e s a m e ke sentence: having someone else a ro u n d entails having to do less w ork th a n if noone is around to help. T h is model is not, however, directly evoked by th e sentence containing ke b u t is r a th e r p a r t of the b a c k g ro u n d reasoning. W h a t the sentence w ith ke expresses in this case is t h a t despite th e conclu sions t h a t the scalar m odel would lead one to draw in this cases, the speaker still com m itted to the validity of the proposition expressed in the apodosis.
In this case, th e speech act is an assertion b u t it could also be
a n order as in (18), or a n y th in g else: (18) -- ‘ine mono j a m ia m e ra m a m a ’ be-NP-3sG only for one d a y m om "Mom, it is only for one d a y ". --‘ke j a mia m era n a ine
pijene n a rotisis ton p a te ra s u ’
and for one day n a be -N P -3 S G go-EMPER su b j.a s k -P E R F -N P -2 S G the father your "Even if it is for one day, go ask y o u r father". Insofar as the sentence with ke (including the consequent) is m ore e n lig h t ening for the addressee as to the c o m m itm e n t or the degree of insistence of the speaker towards the speech act of the consequent, we m ig h t still say
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 129 t h a t ke still m a r k s a m ore inform ative proposition in the sense of K ay (1990). T echnically, however, I d o n ’t see a n y s tra ig h tfo rw a rd way of applying K a y ’s analysis to this type of exam ple. Since "even" in this analysis always tak e s wide scope, both the te x t proposition a n d the con te x t proposition in this kind of construction w ould have to be in a condi tional form.
In the non-speech-act use of th e sentence, this would not
present an y p ro b le m since the context (less inform ative) proposition could be som ething like "If J o h n is n o t coming, w e ’ll have to do all the work ourselves". Y et, in th e case of the speech-act use, th e context proposition is not a conditional. W e m ay arg u e , therefore, t h a t the use of ke...na in this speech- act context is m o tiv a te d w ith respect to the scalar sem antics of this construc tion in o th er contexts: th e c o n te n t use involves scalarity in th e subject m a tte r as opposed to th e speech-act use which involves scalarity at the level of the cog n itiv e /c o n v ersa tio n al reasoning. Still, we have to distin guish between na an d k e...n a concessives, since only the latte r allow this kind of speech-act use. Yet, nothing in the se m a n tic description of these constructions, would p re d ic t this difference a n d , therefore, th e difference would have to be specified. Having said th a t, however, we m ay also argue t h a t the fact t h a t it is ke...n a which m ay have e ither in te rp re ta tio n is less a rb itra ry th a n if it were na. T he na concessives rely on word o rd e r to sig nal th e scalar in te r p re ta tio n , whereas the k e ...n a clauses have also an explicit m arking, nam ely ke. W e m ight expect, therefore, t h a t it would be harder for a c o n stru c tio n which uses a fixed o rd er to give rise to scalar m eaning a t first place to be also used in cases w here scalarity, depending on how one th in k s a b o u t it, is either cancelled or p u t on a different basis. K e ...n a ’s, on th e oth er h a n d , are characterized by both a word order
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 130 p a tte r n a n d a n explicit lexical ite m to which th e am b ig u ity can be a t t r i buted.
4.2 T h e A S clauses: m o t iv a t io n as p a rt o f t h e se m a n tic s t r u c tu r e In th e preceding chapters, I looked a t clauses introduced by as w hich again have m ore t h a n one function. In (19)-(22) I list the types of as clauses which will figure in this discussion: (19) as fam e t o r a
oxi a r y o t e r a
as e a t- P E R F - N P - I P L
n o t later
"L et’s e a t now, n o t later". (20) as er8i
to ra an Qeli
as c o m e -P E R F -N P -3 S G now if w a n t-X P -3 S G "He m ay come now, if he w a n ts ". (21) as kanis t a m a 0 im a ta su
ke 0a se afiso
as d o-N P -2S G the lessons y o u r an d F U T you le t-P E R F -N P -lS G na
vjis
ekso
subj. go-P E R F -N P -2S G o u t "If you do your hom ew ork, I ’ll let you go o u t ( = H o w a bout doing y our hom ew ork a nd I ’ll let you out)". (22) as fonazi
eyo 8en p a o
as scream -EM PF-NP-3SG
I
N E G g o -P E R F -N P -lS G
"I a m not going, even i f /th o u g h he is screaming". In (19) as is introducing a suggestion, in (20) as is expressing perm ission ((20) can be also read as a suggestion), while (21) a nd (22) exemplify the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 131 -
co n d itio n al and th e concessive uses of as respectively. In c hapter 3, I suggested th a t it is th e permission sense w hich is directly related to the concessive. I have also suggested t h a t the condi tional use is related to a n d p a rtly constrained by the use of as to in tro duce a suggestion. T h e claim I w a n t to advance is t h a t in th e case of as, rela tin g the conditional to the non-conditional use and th e concessive to th e nonconcessive provides considerable insights to th e kind of c o n d itio n a l and concessive meanings expressed by as an d goes some way to w a rd s p re d ic t ing their s e m a n tic /p r a g m a tic properties. I will first look briefly a t th e con ditional uses and th en go on w ith the concessives. As conditionals are m ore constrained th a n an conditionals. G ood and ba d examples of as sentences are listed in (23)-(27): (23) ??as emenes
eSo
ke 8a ikseres afti tin ta v e rn a
as live-IM PF-P-2SG here and F U T know -IM P F -P -2S G th is th e tav e rn "?If you would only live here, you would know this ta v e rn ". (24)??as ixes pio polla lefta
ke 0a erxosun m azi mu
as have-P-2SG m ore m oney
and F U T com e-IM P F -P -2S G w ith me
"?If you would j u s t have m ore m oney, you could come w ith m e". (25) as ixame epenSisi
stin IBM
ke
0a im astan
as in v e s t-P E R F E C T - P - IP L to-the IBM a n d F U T
b e -P -I P L
plusii tora rich
now
"If we had invested in IBM, we would be rich now".
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 132 (26) *as kseris y a lik a ke 0a se proslavun as know -N P-2SG French a nd F U T you h ire -P E R F -N P -3 P L "If you know F rench, t h e y ’ll hire y o u ". (27) as ma0is
y a lik a
as lea rn -P E R F -N P -2 S G
ke 0a se p roslavun French a n d F U T you hire-P E R F -N P -3 P L
"If you learn F rench, th e y ’ll hire y ou". A precondition for as conditionals is t h a t these s tru c tu re s have to be coor dinated, th a t is the presence of ke is necessary. T h is condition ch a ra c te r izes the as conditional construction and has to be specified. A further general condition, however, seem s to be traceable to their function to introduce a suggestion.
Suggestions a n d orders are distinct
speech acts; an order is necessarily addressed to an individual distinct from the speaker, while a suggestion can include the speaker in the set of people the suggestion is addressed to. W h a t suggestions and orders share is th a t am ong the felicity conditions for th e pe rfo rm a n c e of such speech acts is the belief of the speaker t h a t the person(s) to whom an order or a suggestion is addressed or is referring to is(are) in a position to carry o u t w hat is suggested/ordered. If I suggest t h a t you go, I am presupposing th a t you are able to do it, and so on. It is this p a r t of the semantics of as which is preserved in its use to introduce a conditional. Consider (26) and (27) as a m inim al pair; (26) containing the verb "know" is ruled o u t for the same reason t h a t "Know French!" or "I suggest t h a t you know French" (in its directive, not in the assertion reading) are ruled out. (27), on the other han d , with the verb "learn" is fine. T h e difference in (23)-(25) can be also accounted for in th e same way: (25) is acceptable since invest ing in IBM is som ething t h a t we could have d one if we chose to. Exam ples
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 133 (23) an d (24), on th e oth er h a n d , are o d d w ith an as conditional, unless th ey are in te r p re te d as a ttr ib u tin g responsibility to th e addressee for living where h e /s h e does and for not having e n o u g h m oney respectively. If t h e conditional use of as is related to th e suggestion m eaning, th e concessive use can be directly m o tiv ate d t h r o u g h the permission sense (cf. exam ple (20)). T h e claim is t h a t an ex a m p le like (29) is related to (28): (28) as fiji as lea v e -P E R F -N P -3 S G "He m a y leave ( = 1 allow him to leave)". (29) as fiji
e yo
0a m ino
as lea v e -P E R F -N P -3 S G I
FU T sta y -P E R F -N P -lS G
"I’ll s ta y even if he leaves ( = 1 g ra n t you t h a t he m a y leave. Still, I will s ta y )". In c h a p te r 3, I suggested t h a t a n a c cu ra te p a r a p h r a s e for (29) would have to be so m e th in g like "I g ra n t t h a t w h a t y o u said- in th is case t h a t he m ay leave- m a y be true. Nevertheless, I still m a i n ta i n w h a t I did, despite w h a t you th in k are adverse circum stances".
I ha v e also suggested t h a t this
m eaning m a y be ta k e n as a direct extension of th e permissive sense in th e speech a c t world: "I allow you to do X" is parallel to "I allow you to say X" or "I allow X as a premise to our c o n v e rsa tio n ". T h e fact t h a t tr u t h conditional ju d g e m e n ts are not relevant w hen faced w ith as clauses is precisely because th e y o p e ra te at the speech-act level. T h e sam e is of course tru e of t h e spe ec h -a ct use of ke...na (cf. c h a p te r 3).
T h e fact t h a t th e as clause in these exam ples is alw ays
"given" or on the table also falls o u t from this kind o f special speech-act semantics; I can only g ra n t or allow into th e con v e rsa tio n something which
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 134 -
has been alre ad y said or suggested, as perm ission in th e real world is n o r m ally preceded by a request for perm ission.
T h is kind of sem antics is
bu ilt into a n as clause in a way t h a t it is not b u ilt into a k e...na clause. O u t of context, the ke...na clause of (17) is a m b ig u o u s betw een a c o n te n t a n d a speech-act reading, t h a t is th e scalar m odel can be either in the s u b je c t m a t te r of the sentence itself or it can be p a r t of the conversational b a c k g ro u n d . A n as concessive, on th e o th e r h a n d , lexicalizes th e speech ac t reading. t h a t as shows is n o t unique. C o n t e n t/ perm is
T h e kind of polysem y sive senses a nd speech-act
senses are also m an ifested by the verb
boro
"m ay, c a n ". Unlike "m ay", however, w hich is also characterized by the sam e
kin d
of polysemy
(cf. Sweetser
1990),
boro gram m aticizes
the
speech-act a n d th e epistemic readings, since u n d e r these in te rp re ta tio n s it can only be an impersonal, th ird person singular. Consider (30)-(35): (30) He m a y go. (31) He m a y be a university professor b u t I a m n o t sure. (32) He m a y be a university professor b u t he sure is dum b. (33) boris
na
fijis
m ay -IM P F -N P -2 S G subj. lea v e -P E R F -N P -2 S G "You m a y leave". (34) bori
na
ise
ka0ij itis
ala 8en ime siyuri
may-EM PERS subj. be-N P-2SG professor b u t N E G b e -N P -lS G sure "You m ay be a professor b u t I a m not sure". (35) bori
na
ise kaGijitis ala
may-EM PERS
Sen su fenete
subj. be-N P-2SG professor b u t N E G you look-M -NP-
3SG
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 135 "You m ay be a professor b u t you d o n ’t look it". In the case of "m a y ", content, epistemic and sp e ech-act readings are n o t distinguished by gram m atic a l differences. In the case of boro(i), however, epistemic a nd speech-act readings are m a rk e d by different syntax; bori in these cases is an impersonal verb and it can only a p p e a r in the third per son singular with the su b junctive clause acting as th e subject. Following H o r n ’s (1985) a rg u m e n t concerning the a m b ig u ity of negation, we may say t h a t such facts s u p p o rt a lexical polysemy analysis for bori r a th e r than one which tre a ts this verb as pragm atically a m biguous betw een
o ntent and
n o n -c o n te n t readings. T hey also s u p p o rt a lexical polysem y analysis for "m ay", a n d by analogy for the other m odals also, as Sweetser (1986, 1990) proposes. If a language has a distinct form (or a form al difference) for a p a rticu la r m eaning, th en we m ay ta k e this as an a rg u m e n t for lexical polysemy in languages which do not m ake th e form al distinction. T h e m etalinguistic reading of the negation is n o t form ally m ark e d in any language, the speech-act and epistemic m eanings of such verbs are. Things get a bit hairy when one s ta r ts considering w h a t counts as a formal difference or a different form. T h u s H o r n ’s a rg u m e n t refers to the observation t h a t no language seems to have a d istin c t m orphem e or word for
the
m etalinguistic
negation,
a lthough
th e
m etalinguistic
reading
imposes special requirem ents on th e s y n ta x of th e sentence in which it is contained. One of these, for example, is t h a t it c a n n o t be followed by "but": (36) *It is not a rare p henom ena b u t it is a rare phen o m en o n . I tak e bori (of the epistemic an d speech-act readings) as being relevant to H o rn ’s kind of a rg u m e n t because I consider t h a t it in sta n tia te s both a dis tinct lexical item (inasmuch as it is no longer declensible) a nd a different
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 136 construction. T h e form er assum ption, however, m ay be a m a tte r of some dispute. T h e clearest a rg u m e n t for distinguishing betw een lexical polysemy and p r a g m a tic a m b ig u ity is of course t h a t the la tte r b u t n o t the form er should be com pletely predictable. Thus, we m ay expect t h a t a n y language with a n e g a tio n m o rp h e m e will be able to use it w ith a m etalinguistic reading. T h e polysem y of may, bori etc., on the o th e r han d , is not pred ic t able in th e sam e w ay and neither is the polysem y of as. Given the perm is sion sense, for exam ple, we c a n n o t p redict t h a t concession will also be there as an available m eaning nor t h a t th e extension will ta k e place in all domains; m a y a n d bori have epistemic readings b u t it is not clear t h a t as has too (a t least, I h a v e n ’t been able to find a n y clearly epistemic e x a m ples).
T h e se m a n tic extensions characterizing these items are m otivated
b u t not predictable. As is m ore like "may" in n o t m aking an y form al distinction between content a n d speech-act readings. Yet, as I suggested, it could be argued to belong to this b ro a d e r system of system atic lexical polysemy. This could m ake its se m antics m otivated in a m ore general w ay too, insofar as it belongs to a system of sim ilar polysemies.
N ote t h a t bo th
bori and
"may", like as, h a v e also suggestion m eanings also, w hich would m ake the parallelism even m ore striking. As is also derived from a verb m eaning "let, allow" a n d we m ay argue t h a t permission is still th e central m eaning of the category, b o th as the one historically prio r a n d as the one th ro u g h which we c a n m o tiv a te b o th suggestion a n d concession. In se ttin g up the sem antic category of as, we need to recognize a t least four d istin c t m eanings. Of these four, perm ission and concession, on the one h a n d , a n d suggestion a n d conditionality on th e other, are directly
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 137 rela te d to each o th er. T h e kind of relationship m a y differ in each case too; th e perm ission-concession relationship involves tra n sfe r of do m ain s (con t e n t to speech-act) a nd Sweetser (1990) argues t h a t this kin d of relation is m eta p h o ric al, w hereas th e as in its conditional use is p a rtly co n stra in e d by th e suggestion se m antics b u t conditionality itself m ay come from the kind of co n stru c tio n t h a t as is found in (as p ke q). Finally, the perm ission to suggestion rela tio n sh ip can be described in th e following way: given th a t as is derived from the im perative form of afino "let, allow ", we can talk a b o u t the m ea n in g c hange as involving sentences like the following: (37) as to n ja n i n a fijL as the J o h n su b j. leave-P E R F -N P-3S G "Let J o h n go". (38) as fij i o ja n is as le a v e -P E R F -N P -3 S G the John "(I suggest t h a t ) J o h n leaves". In (37) th e s p e ak e r is m aking a request for letting a person go. In (38) on th e o th er h a n d , th e request is tow ards "allowing" som ething to happen; in b o th cases th e h e a re r can be th o u g h t of as th e one giving perm ission, in th e case of th e suggestion reading, however, the request for permission concerns an action r a t h e r th a n a person. In o th e r words, from the p o in t of view of the speaker, one way to render the suggestion felicitous would be th e
p e rm iss io n /a g re e m e n t
of
the
hearer.
D ia g ra m m a tic a lly ,
we
can
represent th e se m antic s tru c tu re of as as follows (leaving aside th e "wish" m eaning). T h e arrow s in the diagram rep re sen t the se m antic m otivation links betw een th e different senses.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 138 -
T h e A S category o f c o n s t r u c t io n s
Permission
Suggestion
C oncessive
Conditional
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 139 Ideally, po stu la tin g this kind of se m a n tic relationships should be a ccom panied by a historical s tu d y of the o rd e r of developm ent of each m eaning, confirming the proposed sem antic relationships. This, however, is beyond the scope of this stu d y . All I can a rg u e for here is t h a t recognizing some s tr u c tu r e to th e as category renders t h e various functions of as m uch less a rb itra ry . T h e r e is no w ay of predicting t h a t as should have a concessive m e a n ing as well and, therefore, th e existence o f an as concessive construction has to be in d ep endently specified.
However, th e non-concessive sem antics
of as m akes the concessive m eaning highly m otivated. Saying this, how ever, requires a th eo ry of m eaning which recognizes polysem y relationships and dom ains of m eaning w ith special properties, like th e speech-act dom ain. It seems useful to distinguish betw een this kind of m inim al reference th a t h as to be m ad e to th e concessive use o f as from cases like the na con stru c tio n where concessiveness is a tta c h e d to a word o rder p a tte rn a nd m o tivation is not so easily discernible. E ven if we m o tiv ate the preposing as a focusing strateg y , no obvious ex p la n a tio n suggests itself as to why the focused c o n stitu e n t should acquire a scalar reading. In short, w ith respect to these types of G reek concessives, we need to talk a b o u t bo th facts which, lacking a sy stem atic explanation, should be considered p a rtly idiomatic, and facts which, although analyzable a n d clearly m otivated, are not strictly com positional in the sense discussed in the introduction. W e need, in oth er words, to recognize a c o n tin u u m of compositionality w hich is n o t possible w ith in
a theoretical model of
sem antics where th e only two poles are s tric t com positionality and co m plete idiom aticity.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 140 -
C h a p te r 5
C on ditio nals and concessives as bi-clausal constructions
5.0 I n tr o d u ctio n T h is chapter will exam ine w h a t we m ay think of as discourse p r o p e r ties of conditionals and concessives or, alternatively, conditional a n d con cessive constructions as a whole. So far, in discussing na a nd as c o ndition als a n d concessives, I have b e e n essentially discussing p ro p erties which c oncerned and could be s ta te d in te rm s of the antecedent.
T hese p ro p e r
ties characterized all na and as clauses, regardless of the e m b e d d in g con text. The
im m ediate em b e d d in g
c o n te x t for an adverbial su b o rd in a te
clause is the m ain clause t h a t it d e p e n d s on. T he focus of this c h a p te r, therefore, will be the description of properties which have to do w ith both a n te c e d e n t and consequent a n d as such have to be sta te d a t th e crossclausal level. T h e discussion of conditionals will be centered on the an conditionals, which a re the "regular”, " g ard e n -v a rie ty " conditionals of M o d e rn Greek. Unlike na and as conditionals, th e an conditional m arker is n o t c h a ra c te r ized by a n y special constraints a n d its distribution is b ro ader t h a n t h a t of the o th e r two; every conditional clause introduced by no or as can be also in tro d u c ed by an, b u t th e reverse, as we have seen, is not tru e . T h e kind of facts to be discussed in this c h a p te r is relevant to all conditional clauses a nd, therefore, the discussion will b e held in term s of one ty p e of condi tional clause which can be considered representative of these properties and
which, as I said, will be
th e
an conditional. T he cross-clausal
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 141 -
dependencies an d associated m eaning s tru c tu re s can be assum ed to hold for
the
oth er
kinds
of conditional
clauses
as
well, unless they
are
p re e m p te d by th e general co n stra in ts governing na a nd as which I have already discussed. In dealing w ith conditionals, th ere have generally been two m ain approaches. One of th e m tre a ts th e conditional a nd the m ain clause separately, assum ing t h a t their se m antic c o n trib u tio n can be separately assessed a n d evaluated. T h e o th e r a p p ro a c h m a in ta in s t h a t th e m eaning of a conditional lies in the relationship betw een th e a n te c e d en t a nd the con sequent a n d t h a t s e p a r a te assessment n o t only does not m ak e m uch sense b u t is often impossible. T h is is an oversimplified s ta te m e n t an d th ere are m an y differences a m o n g analyses which subscribe to one or the o th e r of these traditions. T h e com m on d e n o m in a to r can, however, be found in the distinction noted above. T h e analysis of G reek conditionals t h a t I will present here recognizes t h a t the m eaning of a conditional sentence lies in th e relationship between the two clauses and, fu rth e rm o re , t h a t a p a rtic u la r conditional m eaning is a tta ch e d to a p a rtic u la r formal p a tte rn , rep resented by the verb form s in the ante c e d en t a n d the consequent. W h a t I will be arguing is t h a t certain Greek conditional sentences s u p p o rt strongly th e idea t h a t conditionals should be viewed as bi-clausal constructions w ith special formal and sem antic properties. T he second p a r t of this c h a p te r will look m ore system atically a t the conditional-concessive conditional relationship. In the preceding chapters, I showed t h a t there is m ore th a n one way of form ing a concessive o u t of a conditional, a nd t h a t we have to tak e in to a c co u n t th e type of conditional clause w hich is p a rtic ip a tin g in a concessive c o nstruction. C e rta in formal
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 142 -
a n d se m a n tic p ro p ertie s are inseparably linked to th e concessive clauses as such a nd have to be considered as p a rt of th e g r a m m a r p r o p e r of this type of a d v e rb ial clause. In this c hapter, I will try to describe som e of the p ro perties which ch a ra c te riz e the concessive co n d itio n a ls as a class. These p ro p erties concern the s e m a n tic s /p ra g m a tic s of all concessive conditionals a nd require reference to th e consequent as well. C o n s tru c tio n a l differences of th e so rt observed in the previous c h a p te rs will be neutralized in this c o n te x t a n d the discussion will again be m ad e in te r m s of one type of con cessive clause, since the properties described here a re c om m on to all con cessive conditionals. T h e general question t h a t I will be try in g to answer in this section is th e following: Given a c e rtain c o n ditional sentence—in our term s, a certain v e rb a l p a tt e r n w ith associated m e a n in g — w h a t is the effect of a d d in g a w ord like "even" to the conditional clause? C a n we a b s tr a c t any com m onalities am ong the different concessive conditional p a tte rn s which arise? A s in all th e preceding chapters, I will also t r y to a ddress the issue of m o tiv atio n . Some of the conditional sentences I will be describing have p roperties which a re clearly constructional in t h a t th e y are n o t in a ny way p red ic ta b le from the properties of th eir c o m p o n e n t p a rts. On the oth er h an d , a closer look a t the sem antics of these c o m p o n e n ts (in this case, a t th e sem antics of the aspect and tense forms involved) m akes the overall p icture less a r b i t r a r y th a n it would have otherw ise seem ed, showing t h a t the association of p a rtic u la r forms with p a rtic u la r m eaning, although not com pletely predictable, is highly constrained.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 143 -
5.1 C o n d itio n a l co n stru ctio n s: p ro b lem s o f a n alysis
5.1.1 S o m e a p p r o a c h e s to con d ition a ls In th e logical tra d itio n , it is generally believed t h a t th e connection betw een the p a n d the q of a conditional sentence is not p a r t of the m ea n ing. C o n d itio n a ls 1 like ( l) below have been th e kind of exam ple used to d e m o n s tra te th e correctness of this position: (1) If this is A thens, th en I am the Pope. Still, logicians have also noted t h a t in any fo rm u la of the form "if p, then q", th ere is som e kind of connection betw een p a n d q, although, due p a rtly to the difficulty of describing the e xact n a tu r e of this connection, th e question h a s been relegated to p ragm atics; the relationship between th e p a n d the q is not, in o th er words, p a r t of the m eaning proper of a conditional sentence. As far as I can tell, this conclusion holds regardless of the analysis t h a t one gives to "if". T h u s Grice (1975) gives a m ate ria l implication analysis to an "if p, th en q" sta te m e n t a n d m a in ta in s t h a t the oddity of this analysis in cases where the t r u t h or p ro b ab ility of q clearly depends on t h a t o f p is due to th e violation of th e m ax im of quality.
Stalnaker
(1975) proposes a possible worlds solution to th e problem of conditionality, addressing
some
of
the
functional approach; in his
c o u nterintuitive
im plications
of
the
tru th -
approach, w hich is based on K r ip k e ’s (1963)
1 Conditionals of the kind exemplified in (1) have sometimes been called counterfactuals. Since I w ant to reserve this term for conditional p attern s with specific formal properties, I will not call them th a t. As Fillmore (p.c.) points out, counterfactuality in these cases is not associated with the verbal forms in the antecedent and the consequent, since the same verb forms could have a non-counterfactual meaning. R ather, in the specific example, counterfactuality is pragm atically triggered, arising from the " b la ta n t falsehood" of the consequent (cf. A k a ts u k a 1986).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 144 m odal logic, "if p, th e n q" is true in this world if q is tru e in the nearest world to this one in which p is true.
Yet, th e pro b le m of the connection
betw een p and q is still a pragm atic one, according to Stalnaker: "The second issue—the one t h a t has d o m inated recent discussions of contraryto-fact conditionals—is th e pragm atic pro b le m of counterfactuals. This problem derives from th e belief, which I sh a re w ith m o st philosophers writing a b o u t th is topic, th a t the form al p ro p ertie s of the conditional function, together w ith all the facts, m ay n o t be sufficient for determ ining the t r u t h value o f a counterfactual; t h a t is, different t r u t h valuations of conditional sta te m e n ts m ay be consistent w ith a single valuation of all non-conditional s ta te m e n ts .
T h e task set by th e p ro b le m is to find and
defend criteria for choosing am ong these different valuations.... These cri teria are prag m a tic a n d n o t sem antic". (S ta ln a k e r 1975:165-166).^ This position has been inherited by linguists with otherwise widely different stands. T h u s Geis and Zwicky’s (1971) "conditional perfection" (cf. c h a p te r 3) is also a p rag m atic inference. M cC aw ley (1981) m aintains t h a t the p,q relation is p ragm atic in n a tu re a n d so does Haim an (1978) and G a z d a r (1979) (who subscribes to S ta ln a k e r ’s analysis of "if"). Finally, D u d m a n (1984, 1984a) claims t h a t the p and th e q of a conditional clause are separately g e n e ra te d and independently pa rse d a n d t h a t this is the only way of explaining th e choice of the verb form s in a conditional like "If you called th e police right away, th e children are safe now". In all these analyses,
th e sem antics of a conditional does n o t include the
2 Adams (1990) has yet another approach to the problem of the p, q connection, according to which a conditional inference is valid if, roughly, it is impossible for its premises to be probable while its conclusion is im probable (or, if the improbability of the conclusion does not exceed the sum of the im probabilities of the premises). Finally, Lewis (1976) holds th a t a v a r ia n t of S taln a k er’s approach applies to counterfactuals.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 145 -
d escription of th e relationship between th e a n te c e d e n t a nd the consequent. A k a ts u k a (1986) is explicitly in d isa g re em e n t with this position. A ccording to her, the connection between p a n d q is an integral p a r t of th e "if p, th e n q" con stru c tio n ’s linguistic m eaning. T h a t is, each condi tional
sentence
shares
an
abstract,
"correla tio n /c o rre sp o n d e n c e
between
g ra m m a tic a l p
and
m eaning
sim ilar
to
q n. W h a t is co n te x tu ally
dete rm in e d is the specific n a tu re of the " co rrelation/correspondence" in each conditional sentence. According to A k a tsu k a, even the indicative conditionals of th e ty p e exemplified in (1) m anifest a connection betw een the p and the q p a rts of th e sentence, which has to do with the e v a luative ju d g e m e n t of th e speaker: th e degree of ab su rd ity in p correlates w i t h / corresponds to th e degree of a b s u rd ity in q. So, for example, in (1) the speaker is asserting som ething like "Y our claim t h a t this city is A th e n s is ju st as ab su rd as saying t h a t I a m the Pope". A k a tsu k a claims t h a t her analysis fits well w ith the observation t h a t the q of indicative counterfactuals m u s t be a clear falsehood. T h e analysis of the Greek conditional p a tte r n s in the next section agrees w ith A k a t s u k a ’s position in th a t it tak e s th e antecedent-consequent connection to be p a r t of the meaning of a conditional sentence.
However,
there is also a difference in th a t the description of the Greek conditionals will be m aking reference to the specific verbal form s which a p p e a r in th e protasis a n d th e apodosis, claiming t h a t th e m ea n in g of a conditional p a t tern is inalienably linked to its form. To a con stru c tio n g ra m m a ria n this is hardly surprising, although it m ay be surprising to analysts trying to unify conditionals un d e r a truth-conditional or o th e r m eaning.
T he analysis of
conditionals given here, will be, therefore, in th e spirit of F illm ore’s (1983,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 146 1989:chapter 20) analysis of the English conditional p a tt e r n s , always con necting a given c o n d itional p a tt e r n w ith one or more m e a n in g stru c tu re s. A lthough I will be
disagreeing w ith th e specific fo rm -m e an in g m appings
t h a t they po stu la te , it should be pointed o u t t h a t the sa m e a p p ro a c h to conditional sentences is also ta k e n by m ost trad itio n al g r a m m a r s . T z a rtz a nos (1963) can be again considered representative in th is respect, c o n ta in ing as it does th e m o st detailed discussion of conditional p a tte r n s . I have no ob jec tio n to A k a t s u k a ’s claim t h a t there is a general m e a n ing, a b s tr a c t e nough to cover all conditional sentences. I also agree with her claim t h a t th e pro to ty p ic a l m eaning of "if p"
is to express the
sp e aker’s " u n c e rta in ty to w a rd s/u n c o n tro lla b ility of p", w hich seems to be a further elucidation of th e com m on c haracterization of c o n ditionals as "hypothetical". I believe, however, t h a t there is a lot m ore to be said a b o u t the conditional p a tte rn s . In p a rticu la r, the Greek conditionals argue strongly for more specific form -m eaning pairings, since it is th e case t h a t the choice of th e verb form of p, for exam ple, m ay dep en d on the specific meaning associated w ith th e conditional p a tte r n as a whole.
5.1.2 M od ern G reek c o n d itio n a l c o n str u c tio n s T he organization of this section will be as follows: first, I will discuss a p a rticu la r conditional p a tte rn , arguing t h a t a t least in this case we have to m ake reference to constructional m eaning, taking into a c co u n t bo th antecedent a nd consequent. T h e n , I will describe some o th e r conditional p a tte rn s which I will also tak e as representing fo rm -m eaning pairs. T he stru c tu re of the a rg u m e n t here is t h a t since for a t least one case we need to refer to constructional m eaning a nd the ante c e d en t-co n se q u e n t relation ship, we m ay have reason to believe t h a t o th e r conditionals are also
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 147 c h a ra c te riz e d by this kind of constructional m eaning, even in the cases w here th e evidence, a lthough by no m eans con tra d ic to ry , m ay allow o th e r analyses as well. T h e conditional p a tte rn s th a t I will discuss do not by any m ea n s e x h a u s t th e conditional constructions of G reek a n d should, th ere fore, be seen as only a p a r t of w h a t a full description of conditionals sh o u ld u ltim a te ly include. A s in the o th e r cases, however, I a m also going to address the issue of m o tiv a tio n , w hich in the co n tex t of th e p resent discussion translates into deciding w hich of the linguistic choices of verbal forms can be considered to be m o tiv a te d (by the sem antics of these form s in o th e r environm ents) a n d to w h a t extent. T h e hope is t h a t this discussion will fu rth e r clarify th e m e a n in g s tru c tu re of such sentences and help p in p o in t the boundaries be tw e en constructional a n d com positional m eaning.
T h e first conditional p a tte r n to be discussed is exemplified in (2)-(4) below: (2) ‘a n p jo
koka kola
if d r in k - P E R F - N P - lS G
revom e’
coke
b u r p -I M P F - N P - lS G
"If (whenever) I drink coke, I b u r p " . (3) a n vreksi
vrexi
if rain -P E R F -N P -3 S G
ja ta
k a la
rain -IM P F -N P -3 S G for the good
"If (whenever) it rains, it po u rs". (4) ‘an er0i
stin a 0 in a
m en i
if c o m c -P E R F -N P -3 S G to-the A th e n s
mazi m as’ stay-EM PF-NP-3SG with us
"If (whenever) he comes to A thens, he stays with us".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 148 T h e se m antic relationship between the a n te c e d e n t a n d the consequent in exam ples (2)-(4) can be th o u g h t of as one of "general correlation" (as is also
indicated
by
the
translation):
w henever
p
h a p p e n s/h o ld s,
q
h a p p e n s /h o ld s as well. Fillmore (1989) describes a sim ilar conditional p a t tern for English, as in (5): (5) If he wakes u p early, she wakes up early. T h e G reek conditional sentences above em ploy all the PER FEC TIV E], N O N -P A S T fo rm of the verb in the ante c e d en t a n d th e I M P E R F E C T IV E , N O N -P A S T in th e consequent. T he overall m ea n in g associated with this p a tte rn is t h a t of general correlation. The
crucial
thing
here is the choice of th e verb
form
in th e
antecedent, t h a t is, the use of the Perfective n o n -p ast. I will argue t h a t a t least this choice is to some degree constructional, t h a t is, it c a n n o t b u t be a ttr ib u te d to th e whole conditional p a tte r n w ith its associated meaning. T he a rg u m e n t refers crucially to a distinction m anifested by an and otan clauses when th e y are used to express a p a rtic u la r m eaning. Consider (2)(4) above alongside (6)-(14): (6) *an
pino
koka kola
if d r in k - IM P F -N P - lS G
revom e
coke
bu rp -E M P F -N P -lS G
"If I d rink coke, I burp". (7) *an
vrexi
vrexi
if rain -IM P F -N P -3 S G
ja ta
k a la
rain-IM PF-N P -3S G
for the good
"If it rains, it pours". (8)*an
erxete
stin aBina
meni
if c o m e-IM P F -N P -3S G to-the A thens
m azi mas s ta y -IM P F -N P -3 S G w ith us
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 149 "If he comes to A thens, he s ta y s with us". (9)*otan
vreksi
vrexi
when rain -P E R F -N P -3 S G
j a ta kala
rain -IM P F -N P -3 S G for the good
"W hen it rains, it pours". (10) o ta n
vrexi
vrexi
when rain-IM P F -N P -3S G
j a ta kala
rain -IM P F -N P -3 S G for th e good
"W hen it rains, it pours". (11) ??otan
pjo
k o k a k ola
when d r in k - P E R F -N P - lS G
revome
coke
b u r p -I M P F - N P - lS G
" W hen I d rin k coke, I b u r p " . (12) otan
pino
k o k a kola
when d rin k -IM P F -N P -lS G
revome coke
b u r p -I M P F - N P - lS G
"W hen I d rin k coke, I b u r p " . (13) an
pinis
k o k a kola
if drink-IM P F -N P -2S G
8ose
coke
m u ke m ena
give-IM PER I-G E N a nd me
"If you are drinking coke, give me some too". (14) otan er0i
o janis
0a
pam e
when com e-P E R F -N P -3S G t h e J o h n
FUT
g o - P E R F -N P -I P L
"W hen John comes, w e’ll go". Examples (6)-(8) show t h a t the Im perfective, non-past is u n g ra m m a tic a l in the protasis under the general c o rre la tio n reading. As (13) shows, this is not a general co n stra in t against th is form in a conditional protasis, b u t rather, a c onstraint in the case w here the consequent contains an Im per fective, non-past as well a nd the resulting m eaning is t h a t of general corre lation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 150 These ju d g e m e n ts are reversed when the adverbial clause is i n tr o duc e d by otan "w hen". As the c o n tr a s t betw een (9) a n d (10), a n d (11) a n d (12) shows, the otan clauses require the Im perfective, n o n -p a s t in th e a n te c e d e n t in the reading "w henever p, th en q". E xa m p le (14) shows t h a t , once again, the restriction is n o t a general one a g a in s t the perfective form s in th e otan a n tecedent. W h e n th e apodosis has a f u tu re verb form a n d th e overall m eaning changes, a perfective is g ra m m a tic a l. F o r som e exam ples, as for instance (12), the version with the perfective is less b a d th a n th e corresponding an clause w ith a n imperfective. I d o n ’t know w hy this should be so. T h e fact rem ains, however, t h a t betw een the otan perfective a n d the otan im perfective clauses (always u n d e r the rele v an t reading) th ere is a clear preference for th e latter. In the next section, I will discuss which of these linguistic choices of verb forms for the expression of th e "general correlation" m eaning is m ore m o tiv ate d and why. Here, I j u s t w a n t to p o in t o u t t h a t the an vs. otan distinction argues for a con stru c tio n al analysis, t h a t is for an analysis which m akes reference to the form al properties of the a n te c e d e n t a n d th e consequent (and
to the
associated
meaning).
Given
th at
th e overall
in te rp re ta tio n is very sim ilar in the an a n d th e otan clauses a n d t h a t a n ’s a n d o t a n ’s behave very sim ilarly in oth er respects, th e different verb form s a p p earing in the protasis u n d e r th e general correlation reading will have to be described by reference to th e presence of an vs. otan a n d by refer ence
to
th e
relevant
in te r p re ta tio n .
T his
in te rp re ta tio n ,
in
tu rn ,
is
uniquely associated with th e v e rb a l forms of b o th a n te c e d en t a n d conse q u e n t3.
A n ’s an d o t a n ’s4 are b o th a p p ro p ria te su p p o rtin g e n v iro n m e n ts
3 Yu (1983) claims th a t tense p a tte r n s are always the sam e in conditional and temporal clauses (and some other types of adverbial clauses as well). In view of the facts presented here, we are obviously in disagreem ent.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 151 -
for the perfective, no n -p ast. As I noted in c h a p te r (2) (2.2.1), this verbal form is called b o u n d because it c a n n o t occur w ith o u t one of th e s u p p o r t ing particles (na, Qa, as). In an clauses an d otan clauses (and c e rta in o th e r types of adverbial clauses to be discussed in th e next section), th e perfec tive n on-past is also acceptable. So com pare (4) a n d (14) above to (15), (16) below: (15) 6a y r a p s i /* y r a p s i FUT
t a m aG im ata tu sto alio Som atio
w r ite - P E R F -N P -3 S G
the lessons
his
to-the o th e r ro o m
"He will do his hom ew ork in th e o ther room ". (16) epiSi 0a p a r i / * p a r i because
FUT
f a rm a k o
na m in
t a k e - P E R F -N P - 3 S G
fai
m edicine
na
NEG
eat-
P E R F -N P -3 S G "B ecause/since he will ta k e medicine, he s h o u ld n ’t e a t" . T z a rtz an o s (1963:65) is one o f the tra d itio n a l g ra m m a rs w hich recog nizes the existence of this conditional p a tte rn a n d describes it b y reference to both the tense form s in the a n te c e d e n t and th e consequent (th e condi tional of the "indefinite rep e titio n "). He fails, however, to notice th e co n tr a s t with the c o rresponding otan p a tte rn . M ackridge (1985:302) also gives such an exam ple b u t subsum es it u n d e r his "factual" conditionals ca te g o ry w ith o u t noting a n y special constructional requirem ents.
* Otan is derived from ote "when"-!- on. This morphological and historical relatedness is not of course unique to Greek. M any languages manifest such synchronic and diachronic relations between their "if" an d their "when" m arkers (cf. T r a u g o tt 1985) and in some languages (e.g. G erm an) there is not even a distinct lexical item. T he fact t h a t in other languages, e.g. English, "if" and "when" clauses are very similar w ith respect to their formal properties as well (cf. Fillm ore 1989 for a description of these parallel constructions), makes the Greek d a t a a p p e ar even more strange.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 152 -
G eneral correlations can be also s itu a te d in th e past.
In this case,
b oth the a n te c e d e n t a n d the consequent require th e Im perfective p a st, as in (17), (18): (17) an
ixe
stin tsepi tu m ia Sraxm i tin e8ine
if ha v e -IM P F -P -3 S G to-the p o c k e t his one d r a c h m a it give-IM P F-P 3SG "If (whenever) he h a d a d ra c h m a in his pocket, he gave it aw ay". (18) ‘an
ekleje
i aSelfi tu
if cry -IM P F -P -3 S G the sister
ekleje
ki a fto s ’
his cry -IM P F -P -3 S G an d him
"If (whenever) his sister cried, he cried too".
T h e n e x t conditional p a tte rn t h a t I will discuss is exemplified in (19)(21)
below:
(19) an o ja n is p n iy o ta n
8en 0a ixe zisi ke i iSia5
if the J o h n drow n-IM P F -P -3S G N E G F U T liv e -P E R F E C T -P -3 S G and the herself "If J o h n h a d drow ned, she w o u ld n ’t have lived herself". (20) an o ja n is ixe pniji
Sen 0a ixe zisi ke i iSia
if the J o h n d ro w n -P E R F E C T -P -3 S G
N E G F U T liv e -P E R F E C T -P -
3SG and th e herself
5 In all the conditional examples in the preceding chapters, I have glossed 0 a w ith F U T ( = F u t u r e m a rk e r) and I will continue to do so in this c h a pte r. It should be obvious th a t depending on the following verb form, 0 a can express any th in g from future to past time reference and various modalities. So F U T is no thing b u t a glossing convenience and the description of the verb forms in the antecedent and the consequent will be making reference to the whole verb complex, including the 0a particle.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 153 "If Jo h n had drow ned, she w o u ld n ’t have lived herself". (21) an o janis ixe pniji
Sen 0a zuse ke i i8ia
if the John d ro w n -P E R F E C T -P -3 S G N E G
FUT
Iive-L\1PF-P-3SG
and the herself "If Jo h n had drow ned, she w o u ld n ’t have lived herself". W h a t the conditionals in (19)-(21) have in c om m on is t h a t they are all past counterfactuals, t h a t is they presuppose as tru e th e content of the ante c e d en t clause w ith reverse
polarity in th e p a st tense. T here are
several more things, however, to be said w’ith respect to such patterns. T he conditional p a tte rn s of (20) a n d (21), w ith th e P E R F E C T P A S T in th e antecedent, are p a s t c o u n terfactuals in th e sense described above, regardless of th e verb form in th e consequent which can be either the FU T + Im p e rfe ctiv e P a s t or F U T + P e r f e c t P ast. T h e conditional p a tte rn in (19), w ith the Im perfective P a s t in the a n te c e d en t clause, is also a counterfactual. D epending on w h e th e r the verb phrase of the antecedent is sta tiv e or not, there m ay be some different entailm ents, yet, in both cases the sen tence can be a rg u e d to be a c ounterfactual. C o m p a re , for example, (19) above with (22): (22) an o janis itan or0o8oksos
0a ixe jini proeSros
if the John be-P-3SG orth o d o x
F U T b e c o m e -P E R F E C T -P -3 S G
president "If
John
w e r e /h a d
been
an
orthodox,
he
would
have
become
president". Since the verb p h rase in th e a n te c e d en t of (22) denotes a s ta te rather th a n a point event, we can expect t h a t this s ta te will be either construed as having its relevance ended in the p ast or as h aving it continued up to the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 154 -
present. T h u s (22) m ay have tw o possible readings: (i) Jo h n was n o t an o rth o d o x a t the tim e of the election b u t he now is, or (ii) Jo h n was n o t an o rth o d o x then a n d still isn’t. B oth of these readings, however, qualify as past c o u n te rfa c tu a l in the sense o u tlin e d above a n d these differences are r a th e r d ue to the sem antic c h a ra c te r of sta te s t h a n to a constructional difference. E xa m p le (23), on the o th e r h a n d , illustrates the fact t h a t w hen the o rder of the relevant tenses is reversed, t h a t is w hen th e Perfect p a s t is in the ante c e d en t a nd the Im perfective P a s t in the consequent (com pare w ith (19) a nd (22)), past in te rp re ta tio n is no longer imposed on the v e rb of the apodosis. T his is again indicated by t h e two tra n sla tio n s of (23): (23) an ixe er0i o jan is
0a p ije n a m e ki emis
if c o m e -P E R F E C T -P -3 S G th e J o h n
F U T g o -IM P F -P -IP L and we
"If Jo h n h a d come, we would have g o n e /w o u ld go also". T h is is in disagreem ent w ith w h a t T z a rtz a n o s (1963:64) reports. He claims t h a t in an exam ple like (24), th e tim e reference of the consequent is obligatorily in the past. Y et, since b o th th e (a) a n d the (b) c o n tinuations are possible, this is clearly not true: (24) an ixe lavi
engeros to
if rec e iv e-P E R F E C T -P -3S G 0a FUT
tile 7 rafim a
in tim e t h e
tele g ra m
pijene go-IM PF-P-3SG
"If he had received the telegram in tim e, he would go/w ould have gone". (a) ala 8en to pire but NEG
ke Sen pije
it tak e -P E R F -P -3 S G a n d
N E G go-PE R F -P -3S G
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 155 "B u t he d id n ’t receive it an d he d id n ’t go". (b) ala Sen to pire but
NEG
ke Sen
8a
it ta k e -P E R F -P -3 S G a n d
pai NEG
FUT
g o - P E R F -P -
3SG "B u t he d id n ’t receive it a nd he w o n ’t go". As I n o ted before, a n o n -past in te r p re ta tio n is not available to th e I m p e r fective p a s t of (19) an d (22). C o u n te rfa c tu a lity in th e p a st appears, therefore, to be inalienably linked to the P E R F E C T P A S T form . If th ere is a perfect p a s t in the con sequent (as in (19)), th en a protasis w ith an Im perfective p a s t expresses p a st co u n te rfa c tu a lity . C o u n te rfa c tu a lity in the p a st is not, however, a necessary c o n c o m ita n t of th e Im perfective P a s t.
So alongside (19) a n d
(22) consider (25)-(26): (25) a n ton ipostirizan
stis eklojes 0a evjene
if him s u p p o rt-IM P F -P -3 P L to -th e elections F U T be e le c te d -IM P F -P 3SG "If th e y had s u p p o rte d him in th e elections, he would have been elected". "If th ey s u p p o rte d (were to s u p p o rt) h im in the elections he w ould be elected". (26) an to n a y a p u se 0a tin p a n d re v o ta n if him love-IM PF-P-3SG
F U T him m arry -IM P F -P -3 S G
"If she loved him, he would m a rry her". "If she h ad loved him, he would have m arrie d her".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 156 Exam ples (25) a n d (26) have the Imperfective P a s t in the a n te c e d e n t and the F U T -f Im perfective P a s t form in the consequent. Y et the sta tiv e vs. non-stative distin ctio n forces us to m ake a distinction here as well. Exam ple (26), w ith a s ta tiv e verb, has two possible in te rp re ta tio n s as is indicated by th e tra n s la tio n : One as a present c o u n te rfa c tu a l ("she d o e sn ’t love him ") and one as a p a s t counterfactual ("she d id n ’t love love him "). In the latte r reading, (26) is synonym ous w ith the corresponding p a s t counterfactual p a tt e r n s w ith th e P e rfect P ast. T h e conditional in (25) on the other hand, m a y be said to be three ways am biguous, depending on the s p e ak e r’s location in tim e w ith respect to the tim e reference of th e antecedent and th e consequent. It can refer to p a st or p resent situations, like (26), b u t it can also be in te rp re te d as referring to a situ a tio n in w hich the time of the u tte r a n c e precedes the tim e of th e events described in th e antecedent. It is possible to have a stative verb in the Im perfective past and still have the overall in te r p re ta tio n be one th a t refers to fu tu re events.
For
this, however, to be the case th e verb has to be in te rp re te d as expressing a change of state r a t h e r t h a n a continuing state and not all stative verbs will allow this in te r p re ta tio n . "Love", e.g., in (26) c a n n o t be in te rp re te d as "come to love" a n d this rules o u t th e future reading. (27), on the o th e r hand, can be in te rp re te d as referring to the future, if "live" is in te rp re te d as "move" or "com e to sta y " : (27) an
emene
e5o k o n d a 0a tin proslam vanan
if live-IM PF-P-3SG here
near
F U T her hire-IM P F -P -3P L
"If she lived (cam e to live) som ew here near, they would hire her". W h a t kind of se m a n tic s should we a ttr ib u te then to the Im perfective Past, F U T + Im p e rfe c tiv e P a s t p a tte rn ? I suggest t h a t it is sufficient to say
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 157 t h a t the only c o n s tr a in t on this p a tte rn (the s ta tiv e /n o n -s ta tiv e distinction having to be in d ep e n d e n tly stated) can be expressed by saying t h a t the "epistemic stance" o f the speaker has to be o ne in which they associate them self w ith th e w orld in which the pro ta sis does not hold. T he te rm "epistemic
stance"
is taken
from Fillm ore
(1990) an d
refers to
the
s p e a k e r’s a s s u m p tio n s a b o u t the ac tu a lity of th e c o n te n t of the protasis. No relation needs, however, to be specified betw een the tem poral positions of the p a n d th e q o n the one hand, and of the s p e ak e r on the other, since, as we have seen, in the Imperfective p a s t p a tt e r n the s p e a k e r’s tem poral location can all precede, be sim ultaneous or follow th e tem poral position of the p a n d th e q. In th e (strictly) counterfactual readings, it is im m ediately obvious why the sp e aker associates him /herself w ith t h e world of -p, since b o th the present a nd t h e p ast counterfactual readings of (25), for example, assert t h a t p did n o t happen or does n o t hold. In w h a t sense is the "future" reading also a counterfactual? T h e f u tu r e reading of (25) can be said to be c o u n te rfa c tu a l in t h a t th e speaker, in using an Imperfective past, is ta k in g a p a rtic u la r perspective w ith respect to the co n ten t of p; t h a t perspective is o ne which implies t h a t as th e world stands cu rre n tly w hat p expresses is n o t going to happen and t h a t it would require some change to m ake it happen. In other words, bringing a b o u t p would somehow go a g a in s t the normal course of events. C om pare, for exam ple (28) to (29): (28) an e rx o ta n
o janis
0a pijenam e ki emis
if c o m e-IM P F -P -3S G the John
F U T g o -IM P F -P -IP L a nd we
"If J o h n came, we would also go".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 158 (29) an erfli
o ja n is
9a p am e ki emis
if c o m e -P E R F -N P -3 S G th e J o h n
FUT
g o - P E R F - N P - I P L a n d we
"If J o h n comes, w e ’ll also go". W e m ig h t tra n s la te (28) even m ore a c cu ra te ly as "If J o h n were to come..." which implies t h a t as things s ta n d now, it is less likely t h a t he will come t h a n t h a t he w o n ’t. T h e speaker of (28) on th e o th e r h a n d , w ith the P e r fective n o n -p ast (a p a t t e r n to be discussed below'), a lth o u g h also referring to a fu tu re s itu a tio n , does n o t m ak e th e sam e a ssum ptions a b o u t the epistem ic s ta tu s of th e antecedent; ra th e r, (29) is clearly hyp o th etic a l in t h a t th e speaker is equally disposed to w a rd s p a n d -p. F r o m a c o n s tru c tio n a l p o in t of view, we could say t h a t th e relevant readings of th e Im perfective p a s t are p icked w ithin the context of the ver bal form in th e o t h e r clause. W hich in te r p re ta tio n we choose is of course constrained by general requirem ents of coherence an d th e general m eaning of a conditional c o n stru c tio n ("T h e re is som e so rt of connection, causal or not, betw een the p a n d the q"). In an e x a m p le like (19), given t h a t the consequent
imposes
a
past
tim e
reference,
th e n
th e
c o n te n t of the
ante c e d en t, if it is to be somehow in te rp re te d as causing the events in the consequent, will h a v e to be in te rp re te d w ith p a st tim e reference as well. In (23) on th e o th e r h a n d , where the tenses are reversed, the a n te c e d e n t imposes p a s t tim e reference (it is an e v e n t done an d finished in th e past) b u t th e v e rb fo rm of the consequent c a n be in te rp re te d in tw o ways because an event com pleted in th e pa st could have effect on some o th er p a s t event or on e v e n ts to ta k e place in t h e fu tu re . T h e permissible com bi nations, however, as well as their range o f m eanings a p p e a r to be p a r t of th e g r a m m a r of every language. W e h a v e seen t h a t Greek, for exam ple, has
a
past
c o u n te rfa c tu a l
reading
for
its
Im perfective
past,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 159 F U T + Im p e rf e c tiv e p a st c o nstruction, w hereas English does not extend its p a tt e r n
with
the corresponding sem antics in t h a t direction. In o th e r
words, (30) c a n n o t be in te rp re te d as a p a st cou n te rfa c tu a l: (30) If she loved him , he w ould m arry her.
A n o th e r conditional p a t t e r n is exemplified in (31)-(33): (31) an vreksi
5en 0a pao
if ra in -P E R F -N P -3 S G
NEG
F U T g o - P E R F - N P - lS G
"If it rains, I w o n ’t go". (32) ‘an er0is
0a p e ra su m e
if c o m e -P E R F -N P -2 S G
k a la ’
F U T p a s s - P E R F - N P - I P L well
"If you come, w e ’ll have a good tim e". (33) an
xtipisi
to k u 8 u n i 0a aniksun afti
if rin g -P E R F -N P -3 S G th e bell
F U T o p e n -P E R F -N P -3 P L they
"If the bell rings, t h e y ’ll answ er". Exam ples (31)-(33) have th e P E R F E C T IV E N O N - P A S T in the a n te c e d en t and th e fu tu re m ark e r w ith the P E R F E C T IV E N O N - P A S T in the conse quent. T h e m eaning associated with this p a t t e r n is the expression of a prediction which is som ehow dep e n d en t or c o n tin g e n t on the fulfillment of the c o n te n t of th e a n te c e d en t. T he perfective form in the antecedent has fu tu re m eaning (although no 0a occurs in t h a t clause) a n d again it seems reasonable to assum e t h a t this in te rp re ta tio n is c ontextually triggered by the verb form of the consequent. In th e "general correlation" p a tt e r n which was discussed first, th e same Perfective n o n -p a s t form expressed a regularly recurring event a n d w h a t triggered t h a t inte rp re ta tio n was the Im perfective n o n -p ast of th e consequent. Unlike th e "general correlation"
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 160 p a tte rn ,
however,
where
the
"if"
an d
the
"when"
clauses
behave
differently, in the f u tu r a te prediction p a tt e r n "if" and "when" clauses (like the corresponding English ones—cf. Fillm ore 1989) feature the sam e verb form. So com pare (34) to (32): (34) o ta n er8is
0a
pe ra su m e
w hen c o m e -P E R F -N P -2 S G
k a la
F U T p a s s - P E R F - N P - I P L well
" W h e n you come, w e’ll have a good tim e".
F inally, consider (35)-(38): (35) a n
ir9an
stin
o r a tus 0a
i8an tin eleni
if co m e -P E R F -P -3 P L to -th e tim e th eir F U T se e-P E R F -P -3 P L th e Helen "If th ey came on time, th ey saw Helen". (36) a n evrekse
sta
nisia 0a fiyan
i turistes
if ra in -P E R F -P -3 S G to -th e islands F U T lea ve-P E R F -P -3P L th e t o u r ists "If it rained in the islands, th e tourists left". (37) an eftase
stin ora tu
if arriv e-P E R F -P -3 S G
pire
taksi
to -th e tim e his ta k e -P E R F -P -3 S G taxi
"If he arrived on time, he took a taxi". (38) ‘an i ava ayorase
k e n u rjia
"yuna
if th e Ava b u y -P E R F -P -3 S G new kerSise
to
fur
laxio’
w in -P E R F -P -3 S G th e lottery "If A v a bo u g h t a new fur, she won the lo ttery " .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 161 E xam ples (35)-(38) all have th e P E R F E C T IV E P A S T form in the protasis. (35) an d (36) w ith th e F U T + P e rf e c tiv e p a s t in the consequent express a causal relationship betw een th e p and the q which denote events t h a t took place in the past. T hese exam ples can be also read as epistemics (in the sense of Sweetser 1990) inasm uch as they can be in te rp re te d as meaning som ething like "If p took place th en we m ay conclude t h a t q took place as well", a lth o u g h c o n te n t a n d epistemic readings are h a rd to tease out. In (37) a n d (38), however, where th e causal relationship is reversed (if any thing, it is th e c o n te n t of th e consequent which tem p o rally and logically precedes t h a t of th e a n tecedent), the epistemic reading is the only avail able one. W h a t has to be accounted for is the difference in th e verbal form of the consequent. In all (35)-(38) the time of the u tte ra n c e is later th a n the time of the events in the a n te c e d en t an d th e consequent. In (35) and (36), however, p tem p o rally and logically precedes q and this configuration is expressed by the Perfective p a s t in the a ntecedent, followed by the F U T m arker a n d a n o th e r Perfective P a s t in the consequent. In (37) and (38) on the o th er h a n d , it is the q which precedes a n d is the cause of p and this situation is expressed by th e perfective p a s t followed by a (simple) perfec tive p a s t in th e consequent. These p a tte rn s are n o t interchangeable as (39) and (40) show: (39)*an ir0an
stin
o r a tus
i8an tin eleni
if c o m e -P E R F -P -3 P L to-the time their s e e-P E R F -P -3 P L the Helen "If they cam e on tim e, they saw Helen". (40) *an eftase
stin o ra tu
0a pire taksi
if a rriv e -P E R F -P -3 S G to-the time his F U T ta k e -P E R F -P -3 S G taxi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 162 -
"If he arrived on tim e, he took a taxi". Greek, therefore, chooses to gram m aticize a difference between p a st epistemic a nd non-epistem ic conditionals which English, (as has been already indicated by th e translations) doesn’t. A gain, this seems to suggest th a t every such p a tt e r n ha s to be taken in its ow n r ig h t a n d the associated meaning s tip u la te d on a case-by-case basis. T h e re are m a n y m ore conditional p a tte rn s w hich w o n ’t be discussed here.
A full s tu d y of conditionals should d e te rm in e , am ong o th e r things,
the occurring c o m binations of verbal forms o u t of all the potential ones which arise given the m orphological distinctions t h a t G reek m akes in the verbal system. Fillm ore (1990) proposes t h a t a general c o n s tra in t govern ing the com patibility of v erbal forms in the a n te c e d e n t a n d the consequent in English is t h a t the "epistem ic stance" of p has to be the sam e as the "epistemic stance" of q. A lthough, as we have seen, G reek an d English differ in the way they m ay gram m aticize different conditional meanings, I hav e n ’t seen the "epistem ic stance" condition being violated by any of the Greek examples. So, for exam ple, the reason w hy (41) is ruled o u t is because the P e rfect P a s t in the antecedent requires a counterfactual stance, while th e fu tu re in the apodosis requires a hypothetical one (cf. (31)-(33)): (41) *an ixe aniksi
to -yrama 0a
if o p e n -P E R F E C T -P -3 S G the letter
skotoGi F U T be k illed -P E R F -N P -3 S G
T h e m ain pu rp o se of this section has been to a rg u e t h a t conditionals should be viewed as bi-clausal constructions with associated m eaning.
As
I said before, this is th e a p p ro a c h taken by m ost trad itio n al g ram m ars, even if implicitly. However, the categories of conditional sentences t h a t these g ram m ars recognize are much broader t h a n the ones I proposed
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 163 here, lum ping to g e th e r m a n y p a tte rn s w hich, I h a v e argued, are distinct. T h u s T z a rtz a n o s (1963) recognizes four ty p e s of conditionals: (i) th e Mfac tu al" ones, (ii) the " counterfactuals", (iii) those w hich express "a simple t h o u g h t of the sp e a k e r", a n d (iv) exam ples co rre sp o n d in g to w h a t I called " fu tu ra te
pred ic tio n "
and
"general
c o rre la tio n "
(which
are
grouped
tog eth e r u n d e r a n o th e r , nameless category). G iven the "an" vs. th e "otan" discrepancy discussed in the beginning of this section, we have strong rea son to believe t h a t these are d istinct p a tte r n s .
Sim ilarly, there is no obvi
ous reason as to w hy (42) a nd (43) below sh o u ld be in separate categories. A ccording to T z a rtz a n o s (1963:63-64), (42) is a n instance of the "simple t h o u g h t of the sp eaker" category while (43) belongs to the counterfactuals: (42) 0a plutizes
an pijenes sto parisi
F U T get rich-IM P F -P -2S G if g o -IM P F -P -2 S G to -th e Paris "You w o u ld /c o u ld become rich if you w e n t to P a ris". (43) a n e p e rn a if ta k e - I M P F - P - lS G
ali jin e k a 0a e p e rn a oth er wife F U T
ke prika ta k e - I M P F - P - l S G
a nd dowry
"If I took a n o th e r wife, I would tak e d o w ry as well". T his p a tt e r n has b o t h p ast c o u n te rfa c tu a l a n d n o n - p a s t readings (cf.
the
discussion above) w hich are equally available to (42) a nd (43). M ackridge (1985) claims also t h a t "as far as verb form s a re concerned, conditional sentences m ay be divided loosely into " fa c tu a l" a n d "co u n te rfa ctu a l"". W ith in each group, fu rth e r subdivision is ba sed on tim e reference, e.g. factuals are subdivided into present, past, f u tu r e fac tu als etc.. T h e analysis of the conditional p a tte rn s I p re s e n te d here suggests t h a t tim e reference of th e ante c e d en t a n d / o r th e consequent is one of the categorizing factors b u t it is n o t the only one. T h e tem poral sequence of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 164 -
the p a n d th e q (cf. (35)-(38)), th e te m p o ra l location of th e speaker with respect to t h a t of the p a n d th e q (w ith lack of a deictic center, as in the general correlation p a tte rn , being also a possibility), and hyp o th etic a lity vs. counterfactuality® are oth er crucial p a ra m ete rs. A full s tu d y of condi tional p a tte r n s m ay reveal some m o re. A factor which I h a v e n ’t discussed yet, b u t in te n d to m ention in th e n e x t section, is th e speech-act s ta tu s of the consequent.
5.1.3 T h e co n d ition a l p a tte r n s In this section, I sum m arize th e conditional constructions t h a t were identified so far. Once again, it s h o u ld be pointed o u t t h a t this s tu d y is not com plete. Some m ore p a tte r n s will em erge from the discussion in the next section as well. W h a t I w a n t to do here is simply give an idea of w h a t a condensed description of these p a tte rn s m ight look like from a C o n stru c tio n G ra m m a r perspective which looks a t these sentences as biclausal stru c tu re s with p a rtic u la r sem antics. T h e m eaning labels should be th o u g h t of as mnemonic indexes r a t h e r th a n as an a c cu ra te description of the sem antics. For this I will use the kind of "box" n o ta tio n used in Fillm ore (1989) in his description of the English conditional p a tte rn s, which pairs the formal bi-clausal p a tt e r n with a sem antic form ula. To m ake cross-reference a little easier, in table I, I list the c onstructions by the n a m e th e y are given here to g e th e r with th e n u m b er of the exam ples in th e previous section which are in s ta n tia tio n s of these constructions. T h e ® T za rtza n o s and Mackridge use the te rm "factual" conditionals to refer to w hat I think of as h y p o t h e t i c a l I w ant to reserve the term "factual" for a special category of conditionals with "peculiar" properties (cf. section 5.1.4).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 165 nam e on th e left of th e arrow is the (mnem onic) n a m e of the construction, the n u m b e rs on th e right are the num bers of the exam ples illustrating it. Table I 1. "G eneral C o rre la tio n " — > (2), (3), (4) 2. " F u tu r e P r e d ic tio n " — > (31), (32), (33) 3. " P a s t C o u n te rf a c tu a l" — > (19), (22) 4. " P a s t C o u n te rf a c tu a l" — > (20), (21), (24) 5. " P a st, P re se n t C o u n te rf a c tu a l" — > (26), (27) 6. " P a st, P re se n t, F u t u r e C o u n te rf a c tu a l" — > (25), (28) 7. " P a s t C o n tin g e n c y " — > (35), (36) 8. " P a s t E p iste m ic "— > (37), (38)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 166 -
1.
"General Correlation"
Antecedent
Consequent
Perfective Non-Past
Imperfective Non-Past
2.
"Future Prediction"
Antecedent
Consequent
Perfective Non-Past
FUT+Perfective Non-Past
3.
"Past Counterfactual"
Antecedent
Consequent
Imperfective Past
FUT + Perfect Past
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 167 -
4.
"Past Counterfactual"
Antecedent
Consequent _ _ FUT +
Perfect Past
Perfect Past Imperfective Past
5.
"Past, Present Counterfactual"
Antecedent Imperfective Past (stative)
6.
Consequent FUT+Imperfective Past
"Past, Present, Future Counterfactual"
Antecedent
Consequent
Imperfective Past (non-stative)
FUT+ Imperfective Past
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 168 -
7.
"Past Contingency"
Antecedent
Consequent
Perfective Past
FUT+Perfective Past
8.
"Past Epistemic"
Antecedent
Consequent
Perfective Past
Perfective Past
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 169 -
5.1.4 C onditional p a tte rn s and system ic coherence T h is section will be devoted to looking in more detail a t th e condi tional p a tte r n s already discussed. T h e p u rp o se of the discussion here is to place th e observed p a tte rn s in a b r o a d e r co n te x t a nd determ ine to w h a t e x te n t th eir sem antics is m o tiv ate d b y th e sem antics of their c o m p o n e n t p a rts a n d specifically by the se m antics of th e verb forms with w hich they are associated. T h e first p a tte r n I discussed was t h a t of the "general c o rrelation", as exemplified in (44) below: (44) an klapsi
o jan ak is
klei
if c ry -P E R F -N P -3 S G the little J o h n
ke
i
aSelfula tu
cry-IM P F-N P-3SG a nd th e sister
his "If (whenever) little John cries, his sister cries also". In co n tra st, th e "equivalent" otan ( = w h e n ) clauses employ the Im perfec tive (as opposed to th e Perfective) n o n -p a s t for the expression of th e gen eral correlation meaning: (45) o ta n klei
o jan ak is
klei ke i abelfula tu
w hen cry-IM P F -N P -3S G the little J o h n cry-IM PF-N P-3SG a n d the sis ter his "W hen little John cries, his sister cries also". T h e first thing to observe is t h a t th e an ("if") clauses seem to follow a m ore general p a tte r n for adverbial clauses which do allow the P erfective non-past. As I noted in chapter (2) a n d in section 5.1.2 above, this form is necessarily "bound" in t h a t it has to be preceded by one of the verbal p a r ticles na, 0 a or as. T h e only oth er e n v iro n m e n ts in which it can a p p e a r by itself are th e types of adverbial clauses listed in (46) and (47).
These
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 170 clauses, like t h e an and the otan ones, can "support" a Perfective non past: (46) afu
p jo
k oka
kola
revom e
afte r d r in k - P E R F - N P - lS G coke
b u r p -I M P F - N P - lS G
"A fter I d r in k coke, I bu rp ". (47) p rin pjo
koka kola
before d r in k - P E R F - N P - lS G
revome
coke
b u r p -I M P F - N P - lS G
"Before I d rin k coke, I bu rp ". Exam ple (47) is pe rh a p s p ragm atically odd in t h a t it suggests some so rt of psychological conditioning, b u t t h a t is beside th e point. In both (46) and (47) th e m ea n in g which is expressed can again be described as one of "gen eral correlation"
between the c o n te n t of the adverbial and the m ain
clause: every tim e, b e fo re /a fte r I d rin k coke, I bu rp . Like the an clauses, the afu and t h e prin clauses express this m eaning with a Perfective form. T h e sam e sentences with the (48) *afu
p in o
Imperfective are ungram m atical:
koka kola
after d r in k - IM P F -N P - lS G (49) *prin pino
revom e
coke
koka kola
before d r in k - IM P F -N P - lS G
coke
b u rp -IM P F -N P -lS G revome b u r p -I M P F - N P - lS G
Unlike the an a nd the otan clauses, however, where bo th the Perfective non-past a n d the Imperfective n o n -past are possible verb forms in the antecedent (w ith different resulting readings), the Imperfective n o n -past is completely excluded from the prin clauses: (50) prin
*erxete/er0i
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 171 -
before c o m e -IM P F -N P -3 S G /c o m e -P E R F -N P -3 S G k a 0 a riz u m e /0 a k a O a risu m e /ix a m e ka0arisi c le a n - I M P F - N P - I P L /
FUT
c le a n - P E R F - N P - l P L / c le a n - P E R F E C T - P -
1PL Before he c o m e s /c a m e , we c lean/w ill c le a n /h a d cleaned up". Depending on the v e rb form in the m ain clause, (50) can be re a d as a "general correlation" (w ith the Im perfective non-past), as a f u tu r e prom ise ( w ith th e F U T + P e r f e c tiv e n o n -p ast) or as a description of p a s t events (with the Perfect p a st). In every case, however, the Im perfective n o n -p a s t is excluded from th e s u b o rd in a te clause. T he Im perfective n o n -p a st is acceptable in th e afu clauses b u t it forces a "since" re a d in g of the adverbial clause. T h u s (51) w ith th e Im p e r fective non-past can only be re a d as a "since" clause, (52) with th e Perfec tive non-past (cf. also (46) above) can only be in te rp re te d as " a fte r" , while (53) and (54) w ith th e Im perfective p a st and the Perfective p a s t respec tively, can have b o th readings: (51) afu erxete
0a
ksekinisum e
ki emis
since c o m e-IM P F -N P -3S G F U T s ta r t- P E R F - N P - lP L and we "Since he is on his way, we’ll s ta r t also". (52) afu
er0i
0a
ksekinisum e
ki emis
after co m e -P E R F -N P -3 S G F U T s ta r t - P E R F - N P - l P L a nd we "A fter he comes, w e ’ll s ta r t too". (53) afu
erxotan
pijene
"afu" com e-IM PF -P-3SG
ke tin evlepe
go-IM PF-P-3SG a n d her see-IM PF-P-3SG
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 17*2 -
"Since he was coming (anyw ay), he used to go and see he r". "A fter he came, he would go a n d see her". (54) afu
ir0e
pije
"afu" com e-P E R F -P -3S G
ke tin
i8e
g o -P E R F -P -3 S G a n d her se e-P E R F -P -3 S G
"Since he came, he w e n t to see h e r" . " A fter he came, he w e n t to see he r". O ne ten ta tiv e generalization we m a y e x tra c t, therefore, is t h a t the Im perfective non-past seems to c o rre la te w ith factivity. T his is clear for the
afu sentences where the use of this v erbal form forces a factual
in te rp re ta tio n . If this generalization is tru e , it would also explain the incom patibility of this verbal form w ith th e prin clauses, since the s itu a tion depicted by a prin ("before") clause is by definition not realized^. T hings are less clear when w'e consider the an clauses with an Im p e r fective non-past. Yet, here as well th ere a p p e ars to be a correlation betw een Imperfective non-pasts a n d fa c tu a l or "given" readings of the p ro tasis. Consider, for example, (55) a n d (56): (55) an
ksekinai
o janis as fi^ um e
ki emis
if sta rt-IM P F -N P -3 S G the J o h n as le a v e -P E R F -N P -lP L an d we "If (since) Jo h n is 'a r t i n g o u t now, let us leave also". (56) an fevyum e
tora ja t i
Sen
vazis
if lea v e -IM P F -N P -IP L now w hy
NEG
to palto su p u t o n-IM P F -N P -2S G the
coat your 7 T here are languages, as e.g. C hagga (B antu family), where the "before" clauses contain an actual negative m orphem e. T he literal tran slatio n of a clause like "before he d ra n k ", for example, would be "before he d id n ’t d rink ".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 173 "If (since) we are a b o u t to leave, w hy d o n ’t you p u t on your coat?" As is indicated by th e tra n sla tio n , there is a strong factual, "given" flavor associated w ith (55) a n d (56). O u t of c o n te x t, th ey bo th suggest t h a t the c o n te n t of th e pro ta sis has alre ad y been "on the floor", discussed and accepted as the prem ise for fu rth e r conversation; both (55) a nd (56) would be in ap p ro p ria te discourse opening s ta te m e n ts . T o the extent t h a t na con ditionals con stitu te a test for factivity (cf. c h a p te r 2), we m ay note th a t, as expected, the Im perfective n o n -p ast c a n n o t a p p e ar in a na conditional antecedent: (57) *na
erxete
as fiyum e
ki emis
n a com e-IM P F -N P -3S G as le a v e -P E R F -N P -I P L a nd we (58) *na
vrexi
Sen 0a pam e
n a rain -IM P F -N P -3 S G
N E G F U T g o - P E R F -N P -lP L
W e m ay also note t h a t if Im perfective non-pasts correlate w ith fac tu al readings, it should follow t h a t th ey also correlate with epistemic and speech-act conditionals.
P rotases with this verbal form tend to be the
antecedents of speech-act and epistemic conditionals (I hav e n ’t been able to find any conditional with an Im perfective non-past which is neither epistemic nor speech-act). Again, this would be a predictable correlation to the e x te n t th a t, as said above, such c o nditionals tend also to be "given" (by virtue of their se m a n tic s /p ra g m a tic s ).
T h u s (55) and (56) above are
speech-act conditionals. Exam ple (59) below is a n epistemic: (59) an
vrexi
eSo
pera
if rain-EM PF-NP-3SG here
vrexi
ke s ta nisia
ra in -IM P F -N P -3 S G and to -th e islands
"If it is raining here, it is raining ( = i t m u s t be raining) in the islands as well".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 174 By way of conclusion, we m a y say t h a t th e an clauses side w ith the m a jo rity of the o th e r a d verbial clauses in th e choice of the Perfective non-past to express a general co rrelation. In all the o th er types of adver bial clauses discussed here (and which are th e clauses t h a t can s u p p o rt this verbal form) the Im perfective n o n -p a s t is e ith e r u n g ram m atic a l or forces a factual reading. T his is still not a full e x p la n a tio n in sem antic term s of the formal properties of these clauses. T h e preceding discussion, however, serves to em bed th e an clauses in a larger context, seeing to w h a t extent th ey can be considered regular or irregular. W h a t I w a n te d to show is that, within the paradigm of the relevant clauses, th e otan clauses are the ones which behave strangely. T h e p o in t once again
is t h a t s ta tin g this kind of
discrepancy betw een the an a n d the otan clauses requires reference to the relevant reading associated w ith th e bi-clausal stru c tu re . A complete sem antic e x p lanation would have to include, among other things, an account of the sem antics of im perfectivity and an account of exactly w h at aspect of this sem antics it is which forces a factual or "given" in te rp re ta tio n in these otherwise (roughly) irrealis contexts.
It
would also have to explain th e in te rac tio n betw een th e otan semantics and the semantics of an im perfective form which m akes it possible for this form to appear in the a n te c e d e n t of a "general correlation" otan clause w ithout giving rise to any factual im plications. Speculating a bit fu rth e r on this c o n tra s t, we m ay say the following. A general characterization of th e im perfective sem antics proposed by Comrie (1976) is t h a t im perfectivity m ark s "explicit reference to the inter nal tem poral s tr u c tu r e of a situ a tio n , viewing a situ a tio n from w ithin". Imperfectivity is fu rth e r subdivided into a n u m b e r of distinct categories
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 175 o ne of which
is "hab itu al" aspect. Different languages have of course
different ways of lexicalizing th e distinctions. I w on’t a t t e m p t to justify C o m rie ’s definition b u t simply a ssum e t h a t it is correct a t least in its essentials. M o d e rn G reek does n o t lexicalize imperfective aspect differently from h a b itu a l aspect; I t a k e the la t te r to be th e relevant asp ec tu al c h a ra c te riz a tio n of t h e "general correlation" m eaning: every tim e t h a t X also Y. W h e re does this leave us th e n w ith respect to th e an vs. otan con trast? W e m ay say th a t for the
an clauses, h a b itu a lity comes from th e for
m al (im perfective) m ark in g of
th e m ain clause (the consequent)
an d the
p e rfe c tiv e/im p e rfe c tiv e opposition is "saved" for the expression of factuality or "givenness" vs. n o n -fa c tu a lity /"n o n -g iv e n n e ss". Between th e perfec tive a n d the im perfective, the im perfective seems to be a m ore m o tivated choice for the expression of "givenness", if th e speaker, in tak in g some th in g as given o r true, can be seen as tak in g a n inside perspective on ( and th u s in a way "a d o p tin g ") the s itu a tio n .
O n the o th er h a n d , th e otan
clauses, a lready fu rth e r along on th e "factuality" co n tin u u m , do n o t seem to
m ak e
a
f a c tu a l/n o n -f a c tu a l
distinction.
W h a te v e r
th e
distinction
betw een (60) a n d (61), it is not one of factuality: (60) o ta n
ksekinane
w hen s ta r t- I M P F -N P - 3 P L
ibopiise mas n o tify -IM P E R us
"W hen they s ta r t o u t, let us k n o w ". (61) o t a n ksekinisun when s ta r t- P E R F - N P - 3 P L
iSopiise mas n o tify -IM P E R us
It should be m ade clear t h a t n othing I said m itigates th e need of describing the an and th e otan clauses a t th e constructional level, since no th in g of w h a t I said here is general enough to have predictive power.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 176 Yet, e m b e d d in g a n y constructional properties into a larger context serves to m a k e these p ro p ertie s look less a rb itra ry t h a n they w ould otherwise do.
T h e next issue I w a n t to address arises in connection to examples like (62), (63): (62) a n ton
ixe 8i
i
m aria
if him s e e -P E R F E C T -P -3 S G the M ary 0a ton
ixe 8i
F U T him
ke
o janis
s e e -P E R F E C T -P -3 S G a nd the John
"If M ary h a d seen him , John would have seen him also". (63) ‘a n
ixe pai
i
aSelfi su
sto parti
if g o - P E R F E C T - P - 3 S G
the sister your
0a ixe pai
filos tis’
ke o
to -th e p a r ty
F U T g o -P E R F E C T -P -3 S G a nd th e boyfriend her "If your sister h a d gone to the p a rty , her boyfriend would have gone also". Sentences (62) a n d (63) are instances of the p a s t c o u n te rfa c tu a l construc tion discussed in the previous section, with b o th the a n te c e d e n t and the consequent c o n taining th e Perfect past. T he issue arises w hen we observe t h a t these exam ples,
besides having the expected p a s t counterfactual
in te rp re ta tio n , have also a reading in which th e P erfect p a s t is not coun terfactual. T h a t is b o th (62) and (63) m a y be rea d as expressing some sort of reasoning a b o u t p a s t events (I will e la b o rate some m ore on t h a t below). I believe th a t th e English translations could be am b ig u o u s in the same way. tutes
T h e question we have to address, therefore, is w h e th e r this consti counterevidence
to
the
claim
th at
the
PERFECT
PA ST,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 177 F U T + P E R F E C T P A S T construction is associated with counterfactuality. W hat
I suggest
is th a t
it
does
not
c o n stitu te
counterevidence
inasm uch as we can show th a t th e P erfect p a s t of the non-counterfactual reading is a "real" Perfect past (pluperfect) a n d , therefore, distinct from the P e rfe c t p a s t of the counterfactual c o n stru c tio n . It is in fact the case t h a t the n o n -co u n te rfa ctu a l reading of (62) a n d (63) requires t h a t the P e r fect p a s t
be
in te rp re te d
as a true
(non-conditional) pluperfect with
a p p ro p r ia te tem p o ral and deictic properties. If th e hearer of (62) a n d (63) is able to go th ro u g h the relevant co n te x t a n d find t h a t such conditions are satisfied, th e n (62) and (63) will be in te r p re te d as non-counterfactuals. Such a c o n te x t is the one given in (64), w here t h e P erfect p a st is n o t only in te rp re te d as "past" b u t as a tru e "perfect" as well, due to the presence of the "before" clause: (64) Sen
kse ru m e
an
N E G k n o w -N P -IP L prin
erBun
o janis
ixe fiji
if the John
lea v e -P E R F E C T -P -3 S G
i ali
before com e -P E R F -N P -3 P L
the o thers
"We d o n ’t know if John had left before th e o thers came". ala a n
ixe fiji
Sen 0a
tus iSe/ixe Si
b u t if lea v e -P E R F E C T -P -3 S G NEG
F U T th e m
see-IM PF-P-3SG /see-
P E R F E C T -P -3 S G "But if he ha d left, he couldn’t have seen th e m " . Some evidence for there actually being tw o distinct Perfect pasts (or, perhaps, tw o distin ct meanings of the P e rfe c t P a s t) comes from the obser vation t h a t the counterfactual m eaning a p p e a rs to be restricted to the c onditional-m ain
clause environm ent.
That
is
(65) below
cannot
be
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 178 in te rp re te d as "I w o n d e r if he d i d n ’t come to the p a rty ": (65) a n arotiem e
an
ixe er0i
sto parti
w o n d e r -I M P F - N P -lS G if com e-P E R F E C T -P -3 S G to -th e p a r ty "I w onder if he h a d com e to the p a rty " . R e lated to the above, is the issue of the "given" conditionals which, as we said above, te n d to have epistemic or speech-act m eanings (Sweetser 1990). As I showed in c h a p te r 3, it is exactly when a conditional is "given" t h a t n o rm a l tense dependencies between the ante c e d en t a nd a n d the con sequent tend to break dow n. So (67) is gram m atical only in a "given" con te x t such as t h a t of (66): (66) o janis ixe er0i
sto p a rti su
the Jo h n c o m e -P E R F E C T -P -3 S G to-the p a rty your "Jo h n had come to y o u r p a rty " . (67) an
ixe er0i
sto p a rti m u 0a pao ki e-yo sto Siko tu
if c o m e -P E R F E C T -P -3 S G to -th e p a rty my F U T g o - P E R F -N P -lS G a nd I to -th e his "If he had come to m y p a rty , th en I will go to his". It m ay be t h a t we have to consider such conditionals as a class of their own, since the "given" ( = s i n c e ) reading of th e protasis in a way can cels the conditionality, th u s m aking the "if" clause in d ep e n d e n t of the m ain.
W e may also observe, however, t h a t Fillm ore’s (1990) general con
stra in t on the well-formedness of conditionals, seems to hold in these cases as well. T h e "epistemic stance" of the p and the q in such exam ples is still the same, except t h a t now it is an "actual" epistemic stance.
T h e speaker
in this case is associating h im /h e r s e lf with the world of p and the world of
9-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 179 -
In th e case of (64), w here th e context provided the a p p ro p ria te condi tions for the in te rp re ta tio n of th e Perfect p a s t as a real pluperfect, th e resulting in te rp re ta tio n becam e of "h y p o th e tic a l stance" a n d the overall m eaning was epistemic. In (66) a n d (67), w here the protasis was e s ta blished by the context a n d t h e in te r p re ta tio n could be a rg u e d to be t h a t of a speech-act conditional (alth o u g h n o t exclusively so, since "his coming to m y p a rty " is b o th the cause for m y going to his and the "cause" of m y asserting so), the "norm al" in co m p a tib ility betw een a P e rfe c t past a nd a fu tu re was overridden. In c h a p te r (3), all th e ke na exam ples which exhi bited such freedom of choice w ith regard to the tense forms in the a n te c e d en t
and
the
consequent
were
"given",
speech-act
examples
(inasm uch as (67) is). It would be interesting if it tu rn e d o u t t h a t in every case w here the conventionalized m eaning associated with a conditional p a tte rn changes or norm al tense dependencies b rea k down, th e resulting in te rp re ta tio n is either epistemic or speech-act. to consider w hether the rele v an t condition
F u rth e r, it is interesting for the relaxing of tense
requirem ents should be s ta te d in term s of "givenness" or in term s of the speech-act and epistemic m eanings, if in fact these two options are e m p iri cally distinguishable.
5.2 T h e co n cessiv e co n d itio n a ls In this section, I will look briefly a t the concessive conditionals as biclausal constructions. The general question I w a n t to address is the effect t h a t a word like "even", or in G reek ke, has on a conditional construction if it is add e d to the protasis. W h a t we are interested in here is the in te rac tion between a concessive conditional protasis a n d the sem antics of the whole construction, rath e r t h a n form al a nd sem antic differences w hich
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 180 -
have to do w ith th e different concessive clauses. In th e previous sections, I identified several distinct co n d itio n a l p a t terns. In relation to these, consider the n ext set of examples: (68) a k o m a
ki a n
p jo
nero
revome
" a k o m a " a n d if d r in k - P E R F - N P - lS G w ater
b u rp -IM P F -N P -lS G
"E ven if I d r in k w ater, I b u r p " . (69) a k o m a
ki a n pji
nero
0a
refti
" ak o m a " a n d if d r in k - P E R F -N P - 3 S G w ater
FUT
burp-P E R F -N P -
3SG "Even if he d rin k s w ater, he will b u r p " . (70) a k o m a ki an ton ipostirizan
stis eklojes Sen 0a evjene
" a k o m a " a n d if him s u p p o rt- IM P F - P - 3 P L to-the elections N E G F U T be elected-fM PF -P -3SG "Even if th ey h ad s u p p o r t e d /s u p p o r te d
him in the elections,
he w o u ld n ’t have b e e n /w o u ld n 't be elected". (71) a k o m a ki an ixe afisi
ksekliSota
a k o m a a n d if le t-P E R F E C T -P -3 S G Sen 0a NEG
ixan bi FUT
unlocked
i kleftes
e n te r -P E R F E C T - P - 3 P L th e thieves
"Even if he h a d left (the house) unlocked, the thieves w o u ld n ’t have e n te red " . Exam ple (68) contains the v erbal form s which in sta n tia te th e "general correlation" conditional p a tte rn , (69) contains those associated w ith th e "future p rediction" one. T h e te m p o ra l relationship between th e a n te c e d e n t and
th e
consequent
in
(68)
and
(69)
is still
the
sam e
as
in
th e
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 181 corresponding conditional p a tte rn s : the p and th e q are still tem porally u n a n c h o re d (w ith respect to th e speaker) in (68) (with p tem porally preceding q), while (69) is still referring to f u tu r e events with respect to the tim e of th e u tte ran c e . However, (68) and (69) are no longer condition als in t h a t th e relationship betw een the p a n d the q no longer holds. E x a m p le (68) can no longer be described as expressing a general correla tion, nor (69) as expressing a p rediction w hich is c o ntingent on the fullfilm ent of th e c o n te n t of the protasis. R a th e r, w h a t th e y b o th assert is t h a t q holds in any case. T h e in te rp re ta tio n is similar in (70) and 71). In b o th cases, the rela tive positions of the p a n d the q a n d the tim e of the u tte ra n c e rem ain unch a n g ed from the corresponding conditional p a tte rn s , yet the condi tional relationship is no longer present. (70) asserts t h a t "he w o uldn’t have been elected or w o u ld n ’t be elected anyw ay" while (71) asserts th a t "the thieves c o u ld n ’t have entered no m a t t e r w hat". I will not go th ro u g h all the p a tt e r n s since t h e effect of the concessive e n v iro n m e n t ap p e ars to be consistent t h ro u g h o u t in j u s t the respect described above. In c o n tra st to th e conditionals, the concessive condition als ascribe a special s ta tu s to th e consequent. F r o m a truth-conditional point of view, in which conditionality is not p a r t of th e m eaning of a con ditional, the effect of the "even" w ord can be described as cancelling the prag m a tic im plicature t h a t q is ac tu a lly conditional on p (cf. the Geis and Zwicky description of conditional perfection in c h a p te r 3). F ro m a con stru c tio n a l point of view, we would have to say t h a t th e concessive word or constructional p a tte rn overrides the conditional aspect of meaning®. 8 I am not sure exactly how this would work in term s of unification. T he simplistic view would be to say th a t since "even" may be described as coding a relationship between scalarly arranged propositions, with "even" coding the m ost unlikely/inform ative one,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 182 -
T here is m ore to the semantics of the concessive clauses th a n the assertion of the consequent. Some suggestions in th a t direction
will be
m ade in the next c h a p te r w here I will a tt e m p t a m ore complete c h a ra c te ri zation of the concessive sem antics. The final question I w ant to address in this section is how we should describe this p a rtic u la r aspect of the conces sive semantics, na m e ly th e assertion of q. Konig (1986) d raw s a distinction between sentences where the focus of "even" (and o th e r w ords like it) is th e whole "if" clause a n d sentences where "even" focuses only one constituent. In the case where "even" focuses th e whole clause, the consequent is entailed, otherwise not. C o m pare, for example, (72) to (73): (72) Even if it rains, w e’ll go on a picnic. (73) Even if he paid me a million dollars, I w o u ld n ’t do it. In (73), intonation can disa m b ig u a te between th e reading where the whole V P is in focus a n d the reading where w h a t is in focus is "a million dol lars". T he claim is t h a t only un d e r the form er reading the consequent is entailed. I agree with Konig t h a t there is a difference between (72) a nd (73) b u t it is not clear to me t h a t entailm ent vs. im plicature is the right w a y of describing this difference. In the case where th e whole clause (VP) is in focus, the contrasting a lte rn a tiv e under the most com m on in te rp re ta tio n is the hypothetical world of -p; p and -p "exhaust" th e possibilities in th e relevant
universe
of
discourse
and
this
certainly
reinforces
th e
there should be a conflict between th a t meaning and conditionality (since "even" ca n be thought of as adding more than one antecedent possibilities). However, given t h a t so many languages resolve this conflict successfully (by having concessive conditionals), it seems th a t it should be resolved in a unificational description as well.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 183 -
in te rp re ta tio n t h a t the consequent necessarily holds. W hen only one con s ti t u e n t is in focus, the scalar model evoked is usually one t h a t has more th a n two values (although not necessarily). T h e focused c o n stituent is in te rp re te d as expressing a relatively extrem e, unlikely or more inform a tive p o in t on th a t scalar model b u t it leaves open th e possibility th a t there m ay be a nother m ore extreme point, for which, if true, the consequent w o u ld n ’t necessarily follow. It seems to m e, however, t h a t the se m antic force of these concessive conditionals
is really
to
assert the
consequent
(Sweetser (1990) a n d
Fillm ore (1990) m ake the same point w ith respect to English concessive conditionals). It m ay be possible to follow up a concessive clause with a contex t like th a t of (74), b u t in t h a t case the overall effect in the interpre ta tio n is t h a t of "playful" language: (74) a k o m a
ki an m u
eSinan ena e k a to m irio
"akom a" and if me 8en 0a to NEG
give-IM PF-P-3PL one million
ekana
F U T it do -IM P F -P -lS G
"Even if they gave me one million , I w o u ld n ’t do it". a la an mu eSinan
8io
0a to skeftom un
b u t if me give-IM PF-P-3PL two
F U T it th in k -IM P F -P -lS G
"B u t if they gave me two, I would consider it". Following the distinction above, exam ple (75) w ith the whole V P in focus is even more awkw ard: (75) a k o m a
ki an vreksi
0a pam e
"akom a" and if rain-PE R F -N P-3S G F U T g o - P E R F -N P -lP L
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 184 "Even if it rains, w e’ 11 go". ala an xionisi
8en
0a
b u t if sn o w -P E R F -N P -3 S G
pam e NEG
FUT g o -P E R F -N P -IP L
"B u t if it snows, we w o n ’t go". T h e whole question of w h e th e r the consequent is e n ta ile d in this type of clause hinges on w h e th e r exam ples like (74) a n d (75) express some sort of contradiction. I w o n ’t pursu e this question f u rth e r.
W h a t is i m p o r ta n t
is t h a t the overall s e m a n tic s of a concessive conditional includes the asser tion of the consequent. W h e th e r this assertion is a n e n ta ilm e n t or not rem ains an open question. A nd perhaps, from th e p o in t of view of C on stru c tio n G r a m m a r, w here th ere is no sharp line b e tw e en sem antics a nd pragm atics, the answ er to t h a t question may n o t be all t h a t im p o rta n t.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 185 -
C h a p te r 6
T h e sem antics o f concessives
6 .0 I n t r o d u c t i o n T h is c h a p te r will look in some m o re de ta il a t th e sem antics of the concessive clauses which have been ex a m in ed in th e previous chapters. T h e focus here will be on th e s e m a n tic s /p ra g m a tic s p e r se a n d once again, c o n structional differences betw een distin ct types of concessive clauses will be n e utralized if these clauses can be show n to belong to the same se m a n tic class. In section 1 o f this c hapter, I will look briefly a t the constructional properties of factive concessives, which, so far, have only been m entioned in passing. T h e t e r m "factive" or "factual" concessive refers to the ty p e of adverbial clause introduced by connectives like "alth o u g h " and
"even
th ough" in w hich the tr u t h of the pro ta sis is presupposed. W ith in this class as well, G re e k has more types of connectives t h a n English a nd more distinct types of clauses. T h e discussion here will focus on the com m onali ties of these a d v e rb ial constructions r a t h e r t h a n th e differences, centering on th e k in d of clause introduced by an ke w hich is m orphologically related to the connectives introducing concessive conditionals. In th e rest of th e chapter, I will address th e issue of the sem antics of the concessive clauses in general, including b o t h fac tu al a n d non-factual concessives. W h a t I will argue is t h a t from th e survey of concessive clauses which were e x a m in ed in th e previous c h a p te rs, we need to recognize a t least th re e types of concessive sem antics. T h e distinctive com m on feature betw een factual a n d non-factual concessives (and y e t a th ird category
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 186 repre sen te d by th e speech-act concessives) is t h a t som ehow th e co n te n t of th e adv e rb ial clause is taken to be in opposition to th e c o n te n t of the m ain clause. W h a t I suggest is t h a t d epending on th e type of the conces sive co n stru ctio n , this shared sem antic feature is "arrived a t" in different ways
or
comes from
different underlying sem antics.
E xa m in in g
the
different kinds of underlying sem antics allows us to qualify in m ore detail th e n o tio n of "opposition" which seems to c h a racterize concessives as a class.
6.1 T h e fa c tiv e con cessiv es T h e type of clause to be discussed here is exemplified in (1) and (2) below: (1) ‘an ke ita n
ke
i
if a n d be-P-3SG and to p a rti pije
peBera
tu
the m other-in-law his
k a la ’
th e p a r t y go-P E R F -P -3S G well "Even though his m other-in-law was there, the p a r t y w e n t well". (2) p arolo
pu efiyan noris
eftasan
a lth o u g h leave-P ER F -P-3PL early pio
a rriv e -P E R F -P -3 P L
a r y a apo olus
m ore late
th a n everybody
"A lth o u g h they left early, they arrived late r th a n everybody else". I will postpone the discussion of th e sem antics of these clauses till the next section. Here, we can note roughly t h a t the c o n te n t of the clause in tro d u c ed by an ke in (1) and parolo pu in (2) is perceived as being in some s o rt of opposition or as being in co m p a tib le w ith th e c o n te n t of the
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 187 -
m ain clause: "Y ou m ig h t expect t h a t if his m o th e r in law were there the p a rty would be a failure b u t, nevertheless, it w ent well". T h e adverbial clause in (1) is in tro d u c e d by an ke (lit. "if a n d /e v e n " ) a n d , as I argued in c h ap ter 4, this is a com pletely gram m aticized way of expressing "even though" (as opposed to the opposite order ki an). In (2), we have a different factive connective, parolo pu (lit. "fo r-a ll-th at"). In view of the fact t h a t English " alth o u g h " also contains "all", th ere m a y be an interest ing sem antic re g u la rity to be observed with respect to the sem antic developm ent of these connectives. This, however, is b e y o n d the scope of the discussion here. A s I said in c h ap ter 4, it w ould be h a rd e r to find m otivation for th e sem antics of an ke, whose m eaning a p p e a r s to depend on gram m aticized order. T he sem antic a spect of these clauses which differentiates th em from concessive conditionals (cf. c h ap ter 3), is th a t the c o n te n t of th e adverbial clause is presupposed to be true. Unlike concessive conditionals, factive concessives assert b o th th e p and the q. In this respect, factive concessives resemble other types of adverbial clauses such as preposed
"because"
clauses and afu clauses: (3) epiSi efiye
noris
Sen
prolave
because leave-P E R F -P -3S G early NEG have tim e -P E R F -P -3 S G na
tis
Si
subj th em se e-P E R F -N P -3S G "Because he left early, he d id n ’t have time to see th e m " . (4) afu
ise eSo
as
piasum e
Sulia
since be-N P-2SG here ‘let’ ‘g e t’- P E R F -N P -I P L
w o rk
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 188 "Since you are here, let’s get to w ork". F a c tiv e concessives, like the clauses above, are negated by Sen. This, how ever, does n o t distinguish th e m from th e an conditionals which also use th e Sen negation. (As I noted in c h a p te r 2, th e d istrib u tio n of negation is conditioned by th e na a n d as particles w hich trigger the mi(n) negation, e v e ry th in g else being n e g a te d b y Sen). Finally, as expected, unlike conces sive conditionals (the "speech-act" ones n o t included), there are no tense dependencies1 of the so rt discussed in th e previous c hapter, between the a d v erb ial and th e m a in clause. Consider, for exam ple, (5) and (6): (5) a n ke ixan er8i
sto p a rti
if a n d c o m e -P E R F E C T -P -3 P L Sen 0a NEG
pao FU T
mu
e'yo
to-the p a r t y m y I
sto Siko tus
g o -P E R F -N P to -th e their
"Even though th ey h a d come to my p a rty , I w o n ’t go to theirs". (6) ‘a n ke 0a p a n e if a n d
FU T
ekSromi
Sen
a y o ra s a n sleeping b a g ’
g o -P E R F -N P -3 P L excursion
NEG
b u y -P E R F -P -3 P L
sleeping bag "Even though th e y will go on an excursion, th e y d i d n ’t b u y a sleeping bag". O ne interesting fac t a b o u t factive concessives, a nd in p articular a b o u t th e an ke ones, is exemplified in (7) a nd (8) below:
1 Following the discussion of conditionals in the previous chapter, it should be clear t h a t the te rm "tense dependencies" refers to the conventional interpretations associated with a given conditional p a tte r n (antecedent + consequent), which may differ from the "norm al" meaning of a particu la r verb form outside the conditional environment.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 189 (7) an ke xoris
tilus Sen stenoxoriete
if a n d w ith o u t friends N E G
kaBolu
be sad -IM P F -N P -3 S G
a t all
"Even th o u g h (he is) w ith o u t friends, he is not worried a t all". (8)
an ke
m ikrulis t a
if a n d
‘little ’ th e m
k a ta fern i
m ia
x a ra
m a n a g e -IM P F -N P -3 S G
"perfectly"
" E ven th o u g h (he is) young, he m an a g e s perfectly". W h a t is "exceptional" a b o u t (7) an d (8) is t h a t the adverbial clause does n o t co n tain an o v e rt copula o r o th e r verb. This is interesting since G reek as a rule does n o t allow pred ic atio n w ith o u t verbal su p p o rt, except in cases w here special c o n structions are involved, as, e.g., (9): (9) kalos o keros sim era good th e w e ather to d a y " T h e w e a th er is good to d ay ". T h a t (9) represents a pred ic atio n r a th e r t h a n a modification s tru c tu re is evidenced by th e fact t h a t kalos does not a p p e a r in betw een the article o a n d the noun keros. O n the o th e r h a n d , omission of the copula requires t h a t the c o n stituents a p p e a r in th e o rder t h a t they do in (9) (and t h a t som e o th e r conditions of p ra g m a tic n a tu r e , w hich I w on’t discuss here, are also satisfied).
E xa m p le (10), for instance, is u n g ram m atic a l even th o u g h
it would be a good order if there were an intervening copula verb: (10) * o keros kalos
sim era
It is possible t h a t the concessive clauses are a s e p a ra te kind of phenom enon, a nd are p ro b ab ly m o tiv a te d by a different reason from the m ain-clause instances of non-verbal predication.
T h e con stitu e n ts t h a t
can app e ar w ith o u t verbal s u p p o rt w ith an ke do not have to be w h a t F illm ore (1989) calls non-verbal predicates;
non-verbal predicates are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 190 defined as those p h ra s a l types which can com bine w ith the copula to form a predication s tr u c tu r e a nd m ay be a ny of n o m in a l, adjectival, preposi tional or particip ia l type. T hus, besides (7) a n d (8) which contain an adjectival a nd a prepositional non-verbal p redicate respectively, consider (11) (with a nom inal) an d (12) (with a participle): (11) ‘an ke m usikos
8en a y a p a s
if a n d m usician N E G
tis ales tex n e s’
love-IM PF-N P-2SG th e oth er a rts
"Even though (you are) a musician, you d o n ’t love th e other a rts". (12) an ke a y a p im e n i
je n ik a
if a n d love-PAJRT-NOM-PL
exun
ke t a p ro v lim a ta tus
generally h a v e -N P -3 P L a n d the
p rob
lem theirs "Even though th e y are on good term s in general, th ey have their p rob lems too". I ta k e the s tru c tu re s in (7), (8) and (11), (12) to represent "reduced" factive concessive clauses, where the understood verb is the copula. As (11) shows, the en tity of which the non-verbal predicate
is predicated m ay be
co-referential w ith the addressee and as such it would have been expressed by a second person gram m atic a l form. In this respect a t least, th e reduced concessive clauses differ from their main-clause c o u n te rp a rts, where the deleted copula seems to be restricted to th ird person singular, present tense. C on stru c tio n s exemplified in (13), on the o th e r hand, m ay be an instance of yet a n o th e r phenom enon: (13) ‘dinete
orea
a n ke kapos x t ip i ta ’
dress-IM PF -N P-3S G nicely if an d som ew hat "loudly" "She dresses nicely, if som ew hat loudly".
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 191 T h e "deleted" v erb in this case has to be an identical copy of the overt verb of the clause, a n d such constructions have th e feeling of a coordinate r a th e r th a n a su b o rd in a te stru c tu re . In the English translation, I used th e "if" construction which does not contain a full clause b u t ra th e r a sentence fra g m en t. T h a t this is a some w h a t weird use of "if" is evidenced by th e fa c t t h a t Greek uses a factive concessive in th e sam e case (omission of th e v e rb is impossible with a n y of the conditional or concessive conditional m arkers), a n d indeed the m ea n ing appears to be factual. T h a t we m ay need a special construction for English is also evidenced by the fact t h a t , unlike the Greek construction, the sentence fra g m e n t c a n n o t be replaced b y a full clause (which is not true if "if" is replaced by "even if"). T h u s , co m p a re (13) above to English (14) and Greek (15) below: (14) * She dresses nicely, if she dresses loudly". (15)
dinete orea
an
ke dinete kapos
dress-IM PF -N P-3S G nicely
if a n d
x tip ita dress-EM PF-NP-3SG som ew hat
"loudly" "She dresses nicely, even if/even th o u g h she dresses so m e w h at loudly". I take the Greek non-verbal an ke constructions to be essentially synonymous to th eir full-fleshed c o u n te rp a rts, since I c a n n o t th in k of a context where we can use one b u t not th e o th e r.
Sweetser (1990), on the
other hand, seems to suggest t h a t these f ra g m e n t constructions are devoted to expressing concessiveness in th e speech-act dom ain.
I will not
argue one way or th e o th e r a b o u t the English d a ta , since I h a v e n ’t done any systematic investigation. F or the Greek an ke clauses, however, both full and fragm ental, I suggest t h a t although th e y can be p u t to speech-act uses ("I insist t h a t Y, even if I a d m it X"), th e ir p rim ary function is to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 192 -
code direct opposition between the c o n te n t of t h e adverbial a nd the m ain clause or betw een the sentence f ra g m e n t a n d th e rest of the clause. E x a m ple (13), for instance, can be used as an answ er to the question "How does she dress?", w ith o u t an y feeling of "g ra n tin g " o r "adm itting" to w a rd s the c o n te n t of th e sentence fragm ent or of "insisting" tow ards the c o n te n t of th e m ain assertion.
6.2
T h r e e ty p e s of co n cessiv es In this section, I will look in m ore detail a t the semantics of the con
cessive clauses which were exam ined in previous chapters. I have been using th e te r m "concessive clause" as a cover te r m for adverbial clauses whose c o n te n t is somehow perceived as being in opposition w ith th e con t e n t of th e m a in clause. This s e m a n t ic /p r a g m a t ic property w hich a p p e a rs to define the concessive class of clauses h a s been noted by m a n y analysts. Konig (1986) form ulates it as p—> - q ( th a t is t h e content of th e adverbial clause n o rm a lly implies the negation of th e m ain clause), a n d so does Fillmore
(1990)
who
describes
concessive
conditionals as
counter-to-
expectation. W h a t I argue for in this section is t h a t , although th e "p—> - q " feature is und o u b ted ly present in all kinds of concessives th a t I discussed here, we should distinguish th ree kinds of concessive semantics which in G reek a re associated w ith distinct c onstructions.
I will try to show t h a t
while the c o n te n t of p "going against" o r being adverse to t h a t of q is always p a r t of the interpretation, this aspect of t h e semantics is differently m otivated in different concessive constructions.
T h e m apping is n o t one-
to-one since more th an one construction m a y s h a re a particular sem antics. W hat
I
suggest
is
that
we
need
to
recognize
three
kinds
of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 193 "concessiveness", i.e. three ways in which p m a y be said to imply -q:
(i)
concessiveness via scalarity, (ii) "true" concessiveness a n d (iii) direct con cessiveness.
The
following sections explain
and
c o n tra st these three
categories.
6.2.1 C o n c essiv en ess v ia scalarity T h e first k in d of concessive sem antics to be discussed is associated with the concessive conditional constructions of (16)-(18): (16) o j a n is the
na
John
erGi
na
Sen 0a t a k a ta f e r u m e
com e-P E R F -N P -3S G
NEG
FUT
th e m
manage-
FUT
th e m
manage-
FUT
th e m
m anage-
PER F-N P-1PL "Even if J O H N comes, we w o n ’t m anage". (17) ke o janis a n d th e Jo h n
na
erOi
Sen 0a ta k a ta f e r u m e
na co m e -P E R F -N P -3 S G N E G
PER F-N P-1PL "Even if J O H N comes, we w o n ’t m anage". (18) ke n a
er0i
o jan is Sen 0a ta k a ta f e r u m e
and n a com e-P E R F -N P -3 S G the J o h n N E G P E R F -N P -1 P L "Even if Jo h n CO M ES, we w o n ’t m anage".
B o th the na (of (16)) and the ke na (of (17),(18)) constructions have been discussed in detail in c h a p te r (3), w here I p o in te d o u t the differences a nd the similarities between th e two. T h e akoma ki an concessive condi tional in (19) was also discussed in relation w ith th e na and ke na con structions:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 194 -
(19) a k o m a ki an er0i ‘a k o m a ’ a n d
8en 0 a ta k a ta fe ru m e
if c o m e-P E R F -N P -3S G
NEG
FUT
th e m
m anage-
PER F-N P-1PL "Even if J o h n comes, we w on’t m anage". All th e constructions in (16)-(19) are associated w ith scalar sem antics. Even if th e focused c o n stitu e n t would n o t otherw ise presuppose for its in te rp re ta tio n som e so rt of p rag m a tic scale, su c h a scale is imposed by the construction.
W h a t I suggest here is t h a t th e cou n te r-to -e x p e c ta tio n
feature in this case is inalienably linked to th e scalar p ro p erty . In w h a t follows, I will t r y to m ak e th e connection m ore explicit. T h e na concessive construction in (16) serves, as I showed in ch a p te r 3, to p u t one o r m ore co n stitu e n ts into focus (by assigning th e m to first position) a n d , furth e rm o re , to force a scalar in te r p re ta tio n of th e preposed constituent(s). T h is m eaning is associated w ith the c o n structional p a tte rn itself (cf. c h a p te r 4) an d is superim posed on th e relevant phrase regardless of its in h erent sem antics (e.g., a proper nam e like " Jo h n " in (16)). W h a t we m ean by scalar in te rp re ta tio n is t h a t th e re is a t least one m ore assumed proposition set up or imagined, w ith respect to which the na clause is in te rp re te d as " further advanced" or in K a y ’s (1990) term s m ore informative.
T h e r e m ay be m o re th a n one such propositions (i.e. m ore
th an one c o n tra s tin g points on t h a t p ragm atic scale) b u t this is beside the point. As I said in th e previous ch a p te r (5.2), th e re te n d s to be a correla tion betw een tw o-point scalar models a n d focused V P ’s a n d
between
"more th a n tw o points" scales a n d o ther ty p es of focused constituents, although this is by no m eans a n absolute correlation. T h a t is, a focused VP can be (b u t need not be) ta k e n as focusing on th e tru th -v a lu e of the clause as a whole, an d there being only tw o tru th-values, while o th e r
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 195 co n stituents d o n ’t have this possibility. T hus, "Jo h n " in (16) represents a m ore inform ative point on a scale of, e.g., c o m p e te n t people, th a n o th e r individuals. T h e proposition a bout "Jo h n " a n d th e propositions a b o u t the other individuals are p a rt of the assu m e d b a c k g ro u n d knowledge required for th e in te rp re ta tio n of this sentence. O ne difference betw een th e na an d the ke na constructions is th a t the la tte r allows th e V P to be the focused c o n stitu e n t (e.g. (18)), i.e. the c o n s titu e n t w hich is changing value in these oth er co n tra stin g worlds. Finally, th e akoma ki an concessive conditional allows either th e V P or any other c o n s titu e n t to b e in focus, b u t it does so by in tonation, since the only possible w ord-order p a tt e r n is th e one exemplified in (19). W h a t (16) - (19) have in com m on, therefore, is th a t they all require a t least one ba c k g ro u n d proposition a g a in s t which th e concessive conditional clause is e v a luated and in te rp re te d 2. F o r (16), one such alternative w orld m ight be the one represented by th e a n te c e d e n t o f (20); in t h a t world, th e content o f the m ain clause m ay be said to hold: (20) If M ary comes (who is m uch less c o m p e te n t t h a n John), we w o n ’t m ake it. (20), on the other hand, implicates by q u a n ti ty t h a t if "John" or a ny other m ore com petent person comes o ur chances of m aking improve.
it m a y
(16) contravenes this im plicature a n d thus gives rise to the
counter-to-expectation meaning associated w ith th e construction^. 2 T o be precise, the scalar model in such cases is set up by the whole conditional sentence a n d the points on its two axes represent alternative values of some constituent of the antecedent and the consequent. 3 Sweetser (1990) analyzes English "even if" conditionals as involving "an in herent feeling of scale" as well as coding a "despite" relationship. Here, I suggest an analysis which makes the second of these properties derivative from the first.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 196 T h e m o st plausible alternative world set u p by th e pro ta sis of (18) is the world of -p. In co n te x t utte ran c e of a sentence like (21), (21) If J o h n d o e s n ’t come, we w o n ’t m anage, can be said to im plicate t h a t "if he comes, we will m an a g e " (cf. th e description o f conditional perfection in c h a p te r (3)). Once again, (18) directly c o n trav en es this implicature. A very sim ilar se m antic analysis can be given to th e akom a k i an construction exem plified in (19). As we saw in ch a p te r (3), th e akom a ki an concessives do n o t allow preposing of th e focused phrase, b u t distinguish w hat is in focus by in to n a tio n . D epending on the focused co n stitu e n t, different scalar m odels (with different num bers of possible values) m a y appear as the presu p p o sed b a c k g ro u n d . As K a y
(1990) notes in his analysis of no n -co n d itio n a l
"even",
differences in p o larity , m odality and mood often h a v e to be ne u tra liz ed in the process of con stru c tin g a presupposed c o n te x t p roposition fro m a given (explicit) u tte ra n c e . W e can observe t h a t th e s a m e th in g is t r u e for cases where English "even" or Greek ke introduce a concessive conditional. T h u s alongside (16)-(19), consider an example like (22): (22) ke n a to n and
na
Sis
m in
him se e-P E R F -N P -2S G
anaferis t ip o t a NEG
m e n tio n - P E R F - N P - 2 S G a ny
thing "Even if you see h im , d o n ’t m ention a n y th in g " . In an in tu itiv e way, the ante c e d en t o f (22) repre sen ts a precondition for the felicitous u tte r a n c e of th e consequent, in th e sam e way t h a t it does in the corresponding "pure" conditional. W h a t does th e scalar model a nd the propositions t h a t it generates look like in this case? O bviously they cannot be expressed in the sam e form t h a t th e a c tu a l sentence is in, since
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 197 in th e w orld of -p, the reason for u tte rin g the consequent is no longer present. H ow ever, scalarity is involved in such exam ples as well, a nd t h e presence of ke requires t h a t some scalar model is b u ilt for the i n te r p r e ta tion of t h e sentence. T he form of these propositions is d istinct from th e form of t h e se n tence itself. F o r (22), an assum ed b a c k g ro u n d proposition could be s o m e th in g like (23), (23) If you d o n ’t r u n into someone, you d o n ’t ha v e the chance to say a n y th in g . which in t u r n im plicates (24) (the knowledge involved here is generally shared w orld-know ledge ra th e r t h a n specific knowledge shared by th e speaker a n d th e addresse): (24) If you r u n in to th em , you do have th e chance to say som ething. W h a t (22) says th e n (and in this sense, is more inform ative), is t h a t "even if you do r u n in to him, you should ignore y our chance a nd not say a n y thing". In s u m m a r y , therefore, we can say t h a t the "p—> - q " p ro p e rty for these concessive conditionals is b o u n d to the scalar pro p erty .
These p a r
ticular c o n s tru c tio n a l p a tte rn s (with or w ith o u t an explicit ke) set u p scalar b a c k g ro u n d s for the in te rp re ta tio n of these sentences, a n d scales, as shown by v a rio u s analysts, have certain im plicational properties. T h e con tra v e n tio n
of a n
im plicature gives rise to th e coun te r-to -e x p e c ta tio n
sem antics a n d we can say, therefore, t h a t w h a t ke does in this case is very similar to its s e m a n tic /p r a g m a tic function in a non-conditional e n v iron m en t (cf. c h a p te r 2). In the next type of concessive sem antics to be dis cussed, it c a n be a rg u e d th a t the scalar p r o p e rty plays a role there as well (at least for one construction), alth o u g h t h a t second aspect of concessive ness is clearly d istin c t from th e stra ig h tfo rw ard scalar cases discussed
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 198 above.
6.2.2. "True" concession T h e second type of concessive sem antics is associated w ith exam ples like (25) a n d (26): (25) (a) m a m a ine m ono jia e n a vraSi m om be-N P-3SG only for one night "Mom, it is only for one night" (b) ke j a e n a
vraSi
and for one
n a ine prepi n a rotisis
night
‘n a ’ be-N P-3SG
must-EM PERS subj. ask-
P E R F -N P -2 S G ton p a te r a su the fa th e r your "Even if (even though) it is only for one n ight, you (still) have to ask your father". (26) (a) 0a vreksi malon avrio F U T rain -P E R F -N P -3 S G
p ro b a b ly tom orrow
"It will p ro b ab ly rain tom orrow ". (b) as vreksi
piknik 0a p am e
‘as’ rain -P E R F -N P -3 S G
picnic F U T g o - P E R F -N P -I P L
"Even if it rains, w e’ll (still) go on a picnic". E xam ples (25) a nd (26) are instances of tw o d istinct constructions dis cussed in detail in the previous chapters. T h e sentence in (25b) is a n instance of the ke na construction, whereas (26b) is a n instance of the as construction. Form ally, as shown in th e previous chapters, th e y are very
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 199 -
distinct. Sem antically, however, they share a p a rtic u la r ty p e of concessive m eaning which differentiates th e m from the scalar constructions of th e previous section a n d from the factive constructions of the following one. I refer to this type of concessive sem antics as "true" concession, because it is closer to the real m ea n in g of th e words "concede" a n d "concession". T h e sem antic force of these constructions is to convey t h a t th e speaker is "conceding" or g r a n tin g the c o n te n t of th e adverbial clause, nevertheless m aintaining th e consequent. As I argued in C h a p te r 4, th e constructions in (25) a n d (26) arrive a t the comm on in te r p re ta tio n via distinct m o tiv atio n s or d istin ct sem antic paths. T h e ke na
construction can be seen as e xtending its basic scalar
sem antics to the d o m ain of speech-acts (note t h a t only ke na can be used with this sem antics, as opposed to the na concessive construction which cannot - cf. ch.(3)).
A scalar model can be said to be set up or presu p
posed by (26a), which is, however, not th e a c tu a l concessive clause b u t p a rt of the b a c k g ro u n d context: "If the going aw ay is only for one night", m ay be said to e n ta il t h a t "nobody would s a y /in s is t t h a t one should tell one’s father" (p—> - q ) . T his is presupposing an assum ed conditional along the lines of "If it is for m ore th a n one night, th e n it w ould be norm al to insist t h a t the fa th e r will have to know" (-p—> - q ) . T h e speaker of the ke na clause, is co n tra v en in g the first of these conditional sta te m e n ts by conceding p an d nevertheless insisting t h a t -q. O ne w ay to th in k of t h a t is as a m etalinguistic use of ke na. T h e use of a ke na clause in such con texts serves (in a w ay) to "correct" the addressee’s scalar model and the speech-acts which are based a n d depend on t h a t model, a n d replace it by the speaker’s one.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 200 T h is type of concessive sem antics is distinct from th e s tr a ig h tf o r wardly scala r one of ke na, in t h a t the concessive clause is n o t coding directly a "m ore in fo rm a tiv e 11 distinct proposition.
T h e ke na clause in
th e context of (25) is repeating som ething already "given" by th e preced ing discourse, a n d it can only be said to be m ore in fo rm ativ e in t h a t it is tak e n as th e b a c k g ro u n d for the s p e a k e r’s asserting t h a t h i s / h e r c o m m it m en t to q is g rea ter t h a n the interlocutor m ig h t have reason to assum e it was.
T h e s p e a k e r’s c o m m itm en t to q despite his conceding p is w h a t
sem antically "unites" the ke na with the as construction. T h e as concessive, however, is different from the ke na one in t h a t it directly codes this speech-act of conceding in th e sem antics. In th e ke na example discussed above, the co n tex t m a d e it clear t h a t t h e p was already "given" o r established a n d th at, therefore, the in te rp re ta tio n we a tt r ib u t e to ke na h a d to be slightiy modified to a c co m m o d a te th e co n te x tu al influence. T h e as clause of (26), on the o th e r hand, codes directly and as p a r t of its m eaning the fact t h a t p is given a nd th a t th e s p e ak e r concedes it for the purpose of asserting q.
T he "p—> - q " im plication in this case,
comes, I suggest, from the "granting" sem antics of th e as clause, since "granting" gives positive (if possibly re lu c ta n t) s u p p o rt to p; you m ight expect t h a t if I g r a n t you p it is because I intend to den o u n ce q (the preceding discourse having established t h a t q is som ehow adverse to p). Yet, I go on to assert q despite m y g ranting p and t h a t is w here conces siveness comes from. An as concessive, therefore, derived from a lexical afino "let, allow", has the " g ra n t, allow (into the conversation)" sem antics associated with it directly in a way t h a t th e ke na concessives d o n ’t; o u t of context, a ke na clause m a y be am biguous between a real scalar reading a n d a speech-act
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 201 re a d in g of conceding.
T h e re is one final p o in t to b e clarified w ith respect to these " tr u ly conceding" constructions. In describing th e examples above, I have said rep e a te d ly t h a t the speaker is conceding th e c o n te n t of the a d v e rb ial clause in o rd er to reaffirm his c o m m itm e n t to q. Q, however, does n o t need to b e identical to th e s p e a k e r ’s c o m m itm en t, as this is established by co n te x t. T h u s , alongside, (25) a n d (26) w here the sp e aker of th e conces sives has p resu m a b ly alre ad y d e m a n d e d t h a t the father be asked (in (25)) a n d asserted t h a t a picnic is a b o u t to ta k e place (in (26)), consider (27): (27) — 0a p a m e v o lta FUT
e
g o - P E R F -N P -I P L w a lk excl.
"W e ’ re going for a walk, a r e n ’t we ?" — 0a
vreksi
om os
m alon
F U T r a in -P E R F -N P -3 S G
b u t probably
"B ut, it is probably going to rain " . -- as / ke n a ‘a s’/e v e n
if
vreksi
0a p a ru m e
ra in -P E R F -N P -3 S G
ombreles FUT
t a k e - P E R F - N P - 1PL
um brellas "Even if it rains, we can ta k e u m brellas".
W h a t (27) illustrates is t h a t th e sp e a k e r’s c o m m itm e n t is to w a rd s "going for a walk"; q, in this case, is ta k e n as representing a s ta t e m e n t consistent w ith a nd "supportive" of the s p e a k e r’s u ltim a te c o m m itm e n t, a lth o u g h it is n o t directly expressing th e c o m m itm en t proper. T h is is a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 202 -
fe a tu re on which the conceding constructions discussed herein differ from the " w h e th e r or n o t" c o n s tru c tio n 4 of (28).
In th e c o n te x t of (27), (28)
below is u n g ra m m a tic a l (although it can, of course, be used to m ean sim ply t h a t "w e’ll ta k e um brellas even if it d oesn’t rain", w ith no im plica tions a b o u t a walk) w hereas (29) is fine: (28) * vreksi Sen vreksi ra in -P E R F -N P -3 S G
0a p a ru m e
om breles
N E G rain -P E R F -N P -3 S G F U T ta k e -P E R F -N P -
1PL um brellas "W h e th e r it rains or not, w e’ll ta k e um brellas". (29) vreksi Sen vreksi ra in -P E R F -N P -3 S G
0 a p am e volta NEG
rain -P E R F -N P -3 S G F U T
go -P E R F -N P -
1PL picnic " W h e th e r it rains or not, w e’ll go for a w alk". Similarly, the factive concessives which will be discussed next, if p u t in to this kin d of context, p u t some special requirem ents on w h a t the m ain clause can or c a n n o t be (cf. the discussion in th e next section). T h u s (30), in th e co n te x t of (27), is also odd: (30) * an ke 0a vreksi if
and
FUT
0a p a ru m e ombreles rain -P E R F -N P -3 S G
FUT
take-P E R F -N P -IP L
um brellas "Even th o u g h it will rain, w e’ll ta k e um brellas".
* This construction deserves special description which, however, will not be done here. It is interesting to note th a t it looks as though it conjoins tw o main clauses (note the use of th e 8
