Institute of Cognitive Science Albrechtstrasse 28 49069 Osnabrück, Germany

Constraints on the Contraction of Preposition and Definite Article in German

Bachelor Thesis by Maria Cieschinger [email protected] Oct. 12th, 2006

Supervisors: Prof. Dr. Peter Bosch, Dr. Carla Umbach

Table of Contents

0. Abstract

p. 3

1. Introduction

p. 4

2. Löbner's Proposal for Definites

p. 8

2.1 Noun Types and Concepts

p. 9

2.2 Semantic and Pragmatic Definites

p. 11

2.2.1 Semantic Definites

p. 11

2.2.2 Pragmatic Definites

p. 19

2.3 Problems

p. 24

3. Contractions in Löbner's Framework

p. 28

4. A New Theory of Contractions

p. 37

4.1 Motivation

p. 38

4.2 Three Different Uses of Definite Descriptions

p. 41

4.2.1 The Generalised Use

p. 41

4.2.2 The Specific Use

p. 44

4.2.2.1 The Small-world Use

p. 44

4.2.2.2 The Contextual Use

p. 50

5. Conclusion

p. 53

6. References

p. 56

2

0. Abstract In this paper we will try to give an adequate account of preposition-article contractions in standard written German. Löbner (1985) suggests that contractions are not possible in front of pragmatic definites, but possible if the definite description is semantically definite. Of course, we have to understand the distinction between pragmatic and semantic definites that he proposes, before we can test his hypothesis. So, after presenting Löbner's theory in Section 2 we will check his claims with the help of several examples in Section 3. It will turn out that his claims with respect to contractions are correct for pragmatically used definite descriptions, but many observations with regard to semantic definites cannot be predicted correctly by his theory. This is why in Section 4 I will propose a different account of contractions. The use of either the contracted or the un-contracted form appears to be not a matter of choice, but it is dependent on the way in which the respective definite description is used. I will suggest a distinction between different uses of definite descriptions that is inspired by Donnellan's (1966) notion of the referential and the attributive use of definite descriptions. The two main categories I will propose are the specific and the generalised use. The specific use roughly corresponds to the referential use of definite descriptions, the generalising use corresponding to some extent to the attributive use in Donnellan's sense. The specific use can again be subdivided into two subcategories: The small-world and the contextual use. Contractions are obligatory if definite descriptions are used in the small-world use or generalisingly, the un-contracted forms are required by contextually used nominals. It will turn out, that such an approach seems to give an adequate account of contractions.

3

1. Introduction In some languages that make use of the definite article, we can find forms that seem to be contractions of a proposition and the definite article. In Italian, for example, the contracted forms are obligatory if the phonological environment allows a contraction and the non-contacted forms are never used in such environments. In German, on the other hand, we can find both contracted and non-contracted forms, even in sentences that have an otherwise identical structure, as in the following example.1

(1) Mary and John are watching a news report about the St.-Marien-Hospital in Osnabrück. Mary says to John: (a) Ein Freund von mir arbeitet in dem Krankenhaus. [A friend of mine works in the hospital.] (A friend of mine works in that hospital.) (b) Ein Freund von mir arbeitet im Krankenhaus. [A friend of mine works CONTR-in-the hospital.] (A friend of mine works in a hospital.)

As can be seen from the English translations, (1a) and (1b) differ in meaning. In (1a) Mary states that one of her friends works in a particular hospital, viz. in St.Marien-Hospital, whereas (1b) is considered a true statement if and only if one of her friends works in some hospital or other, but not in St.-Marien-Hospital.2 In contrast to the first case, here Mary might not even be able to answer the question which hospital it is that her friend works in and she would still have succeeded in saying something meaningful, something true iff her friend works in a hospital. This simple example already indicates that the use of preposition-article contractions cannot simply be reduced to an optional choice on the part of the speaker, as suggested by Lyons (1999: 328f). Also Heim's (1991: 488) proposal can be shown to be inadequate with the help of our example; she claims that it is obligatory to form preposition-article contractions in the presence of certain prepositions, 1

In the following I will give two translations for the German examples where necessary: The first is a word-by-word translation (in square brackets), the second is idiomatic (in parentheses). CONTR indicates the preposition-article-contraction. In cases where the idiomatic translation is sufficiently close to the German original sentence, I will sometimes use CONTR in these translations, too. 2 The exclusion of St.-Marien-Hospital here is due to the Gricean implicature; if Mary had wanted to state that her friend works in St.-Marien-Hospital, she would have used (1a) instead of (1b).

4

whereas the demonstrative determiner cannot form contractions in otherwise similar environments. In (1a) dem is not used as a demonstrative determiner and we can also find other examples that stand in opposition to Heim's suggestion.

(2) Karl sitzt in dem (*im) Auto, das er letzte Woche im Lotto gewonnen hat. (Karl is sitting in the (*CONTR-in-the) car that he won in the lottery last week.) (3) ... und dann habe ich zu der (*zur) alten Hexe gesagt, sie solle unsere Kinder endlich in Ruhe lassen. [... and then have I to the (*CONTR-to-the) old witch said, she should our children finally leave alone.] (... and then I told the old witch that she shouldn't harass our children any more.)

In neither of the two sentences do we find demonstrative determiners, nevertheless the contracted forms are not possible. Lyons' claim is too weak and Heim's explanation for the phenomenon of contractions in German is too restrictive. In this paper, we are trying to find a theory that comes close to a correct account of the data.

In the course of our investigation we will encounter many different problems that have to be solved with respect to contractions in German. Our aim is, of course, to give an adequate account of all of these problems. However, it should be noted that we are restricting our domain of research deliberately. First of all, we will not consider all possible preposition-article contractions that can be found in the German language, rather we will concentrate only on those forms that have found their way into standard German, i.e. into written language. By no means are we trying to give reasons or explanations as to why some forms are considered part of standard German and others are only present in colloquial language or in certain dialects, and we will not engage in any speculations about possible future inclusions of certain forms into standard German. So even though contractions in phrases such as umme Ecke ('round the corner), anner Wilmersdorfer Straße (at Wilmerdorfer Straße), inner Schule (in school), aufer Arbeit (at work) exhibit similar characteristics to contractions in written

5

German, we will not include such forms into our investigation. The contractions we are going to consider are only the following: im, am, ins, ans, zur, zum, beim and vom. Secondly, in German the use of the definite article in combination with given names and family names is highly dependent on the dialect of the speaker. Generally, it can be said that the definite article precedes given names frequently in Southern German dialects, while its combination with family names seems to be a phenomenon that is not restricted to Southern German dialects and can also be found in written language, albeit rather rarely. But since intuitions and judgements of acceptability differ widely with respect to different regions speakers grew up in, we will exclude this phenomenon from our analysis. A further restriction to our domain of investigation is that we will not be concerned with certain collocations that contain contractions. We will take is as given that phrases like im Vorfeld (in the run-up to something), im Nachhinein (with hindsight), im Geheimen (in private), im Stillen (inwardly), im stillen Kämmerlein (when nobody watches you) are considered collocational and hence they will not be discussed in this paper.

One final remark: The reader may have noticed that I have avoided any reference to the "inner structure" of contractions. It seems to be a generally accepted fact that contractions are constructed this way: von + dem > vom, to give just one example. Another idea would be that contractions can be ambiguous, i.e. that they can be constructed with the help of both the definite and the indefinite article. We would then get something like (von + dem) or (von + einem) > vom. In this paper, however, we will not be concerned with the inner structure of contractions or with the morphological processes that build up contractions. We will assume that the contracted and the un-contracted forms are two different words and we will try to give an adequate account for the use of either form.

Now that we have seen what we are not going to be concerned with in this paper, we should turn to what we are going to do. We will start by presenting a theory of definite NPs that was introduced by Löbner (1985). Löbner distinguishes three different types of (uses of) nouns: Sortal, relational, and functional nouns. Different uses of nouns in different linguistic environments can in turn give rise to sor6

tal, relational, and functional concepts. The meaning of the definite article in Löbner's framework is to indicate that the noun immediately following the definite article is to be interpreted as a functional concept.3 According to Löbner, the subcategorization of uses of nouns enables us to distinguish systematically between semantic and pragmatic definites. Roughly speaking, the unambiguous reference of pragmatic definites depends crucially on a particular situation, whereas semantic definites refer unambiguously independent of the particular situation. I will try to argue, however, that some of the conceptions that are essential for the theory as such are questionable or at least need more clarification. My main concerns are the applicability of the classification of uses of nouns and the role of the situation of utterance in Löbner's framework. With respect to contractions, Löbner claims that the occurrence of contracted and non-contracted preposition-article combinations corresponds directly to his distinction between semantic and pragmatic definites. So after looking at the details of his theory, we will test this hypothesis with the help of several examples and of general observations about contractions. We will see that although many of our observations can be correctly predicted by Löbner's theory, there remain some unsolved problems with regard to contractions. It will become clear that the theory does not make the correct predictions in many cases, that some of his own examples are not very convincing, and that Löbner's theory seems incapable of handling the use of epithets. We will then proceed to an analysis of contractions that is inspired by (but distinct from) Donnellan's (1966) theory of definite descriptions. Since, however, the theory suggested by Donnellan cannot be applied to contractions in general, we will need some modifications. These modifications mainly apply to the attributive use of definite descriptions and might change the original theory so much that we should probably not refer to the result as a result in the sense in3

Apparently, Löbner assumes a structure like (a) for complex definite descriptions. Heim & Kratzer (1997: 82f.), on the other hand, point out that a structure like (b) (proposed by Barbara Partee in 1976) is more appropriate for semantic analyses. (a) [[Det N] PP/relative clause/...] (b) [Det [N PP/relative clause/...]] In our analysis we will stick to the second version and we will use the term nominal (NOM) to refer to the constituent [N PP/relative clause/...]. It should be noted, though, that Löbner's terminology differs from ours; it seems that he uses the term NP to refer to nominals. Nonetheless, in the discussion of Löbner's proposal I will often use the term nominal instead.

7

tended by Donnellan any more. The two main distinctions between different uses of nouns (or of definite descriptions in general) that I want to propose are the generalised and the specific use. The generalised use roughly corresponds to Donnellan's attributive use in that the speaker makes a statement about something or other that fits the description used. Generic sentences and nominals will also be included in this category. The specific use, on the other hand, is inspired by Donnellan's notion of the referential use of definite descriptions. All nominals that are used referentially are classified as specific, but we will also consider nominals as being used specifically if they show certain syntactic characteristics. As a result of those considerations we will also find definite descriptions in this class that would be regarded attributive in Donnellan's theory; we could say that the syntactic properties can override the semantic properties of definite descriptions. Definite descriptions that belong to the category of the generalised use will obligatorily be used with contractions in the presence of the appropriate environment, while those that belong to the specific category do not exhibit a uniform distribution with respect to contractions. That is why we will again subdivide this class, viz. into the small-world and the contextual category. The small-world use necessarily occurs with the contracted forms, the contextual use uniformly requires the non-contracted forms. Let us now turn to Löbner's proposal for definite descriptions.

2. Löbner's Proposal for Definites According to Löbner (1985) the problem of preposition-article contractions in standard German has already been solved. He claims that "cliticization of the article to a preposition as a host [...] is possible if and only if the NP is semantically definite and not too complex" and that "the contracted forms are not possible in the same phrases as soon as they are used to refer 'pragmatically' in our sense" (Löbner 1985: 311f). In order to put ourselves in a position where we can sensibly decide whether this is right, we must first take a closer look at (the relevant aspects of) the theory as it is presented in Löbner (1985) and we have to get acquainted with the necessary terminology.

8

2.1 Noun Types and Concepts According to Löbner "nouns have two basic interpretations. Taken in isolation they can be considered either sortal nouns or relational nouns" (Löbner 1985: 292). As the names suggest, sortal nouns classify objects as belonging to particular sorts exhibiting certain characteristics, and relational nouns refer to objects that stand in a specific relation to other objects. Sortal nouns can logically be understood as conjunctions of (one or more) one-place predicates, i.e., as a conjunction of other sortal nouns. Relational nouns differ from sortal nouns in that they express relations, i.e. two- (or many-) place predicates, possibly in conjunction with several one-place predicates. It is not possible to substitute one-place predicates by many-place predicates and vice versa; that it why sortal nouns cannot be reduced to relational nouns or the other way round. Löbner himself uses the nouns woman and wife to exemplify the distinction between sortal and relational nouns. The meaning of woman can be understood as a conjunction of other sortal nouns such as human, adult and female. The meaning of wife differs from that of woman in that while a wife is also a woman, a wife additionally always is the wife of somebody, i.e. a wife stands in a particular relation to another object. We can express this fact with the help of a two-place predicate. A special case of relational nouns are functional nouns. They also express relations, i.e. many-place predicates, but here the relation is a function. This means that an object is related unambiguously to another object, it is a one-to-one relation. We can use the noun wife again to make this distinction clearer. If we imagine a society in which a man can be married to more than one woman at a time, the noun would be relational, i.e. it expresses a one-to-many relation. More than one person could stand in a wife-relation to a particular person. In a society that only allows monogamous marriage, however, a particular man can only have one wife at a time. In that case wife would be a functional noun, because it relates a person unambiguously to another person. Functional nouns, then, refer unambiguously in virtue of their inherent meaning, "they do not allow for more than one referent" (Löbner 1985: 293). Relational4 and sortal nouns, in contrast, can refer unambiguously only accidentally. This can happen if by coincidence only one object satisfies the one-place predicates of a sortal noun or if, for example, a 4

I will use the term relational in the narrower sense from now on. The three types of nouns are: sortal, relational and functional nouns.

9

woman in a polygamous society happens to be the only and not the, say, third wife of someone, and the one-to-many relation wife becomes a one-to-one relation due to the circumstances. Note, that the reference of the two nouns that we used as examples crucially depends on external factors. We have already seen that the classification of wife in the polygamous society as a one-to-one or a one-to-many relation differs under different circumstances. Time also plays an important role for determining the referents in our examples. As for sortal nouns, an individual might belong to a certain sort at one time but not at another. A person, let us call her Claire, might be a woman in the year 2006 (i.e. fulfil the one-place predicates human, adult and female), but she would not belong to the sort woman in 1983 when she was only three years old. It is obvious that the reference of a noun depends on various factors (such as time and location), which we will simply call a reference situation from now on. This view of course changes the logical properties of the noun types that we introduced above. Instead of one-place predicates we will now use twoplace predicates for sortal nouns, the first argument being a situation. The same changes apply to the many-place predicates for relational and functional nouns, we add a situation as the first argument of the respective relations.

The sortal-relational-functional distinction that Löbner proposes for nouns also applies to concepts. Concepts are often represented by nouns (Löbner 1985: 295), but complex definite descriptions can also represent concepts. "They can be understood as effective mental procedures with a certain input/output characteristic." (Löbner 1985: 295) But what is an effective mental procedure? It is some kind of mental process (Löbner remains rather vague here) that maps a situation and objects to objects (in the case of relational and functional concepts) or situations to objects (in the case of sortal concepts). Roughly speaking it is a procedure to evaluate the predicates that are given by the noun. (Sometimes the context of an utterance has to be consulted in order to be able to form a concept, because some (or even all) of the arguments of relational and functional nouns might not be specified explicitly.) Löbner's discussion of semantic and pragmatic definites heavily relies on his notion of sortal, relational, and functional concepts. One and the same noun can express different kinds of concepts in different situations. It should be noted 10

here that Löbner speaks rather sloppily, and often uses the term noun instead of concept, a point we will discuss in more detail in Section 2.3. In the rest of this section I will try to stick to Löbner's distinction between nouns and concepts, but we will see that, unfortunately, due to Löbner's own phrasing, this distinction will get a little blurred at times.

2.2 Semantic and Pragmatic Definites Now that we have introduced Löbner's basic terminology, we can proceed to his two different uses of definite descriptions as semantic and pragmatic definites. Semantic definites represent functional concepts and the nominal refers unambiguously independently of a particular situation. Pragmatic definites crucially depend on the context of an utterance for unambiguous reference and they have head nouns that are used relationally or sortal. We will begin with the semantic definites.

2.2.1 Semantic Definites Let us first consider functional concepts that have a situation as their only argument, FC1s as Löbner calls them. Proper names paradigmatically belong to this class of functional concepts. Löbner claims that "within a certain range of situations, proper names refer unambiguously to certain objects. They constitute constant functional concepts, as their value does not vary with their possible arguments." (Löbner 1985: 299) Remember that the possible arguments for FC1s are situations. What the above statement means, then, is that there are (smaller or larger) sets of situations in which a particular proper name refers unambiguously. With regard to personal names, the set of situations in which such a name can be used felicitously is rather small, especially in the case of bare first names. Other proper names, such as geographical names or dates for example, have a much larger range of situations in which they yield one and the same value. Phrases that contain proper names are also considered FC1s; examples from Löbner's text would be the year 1984, the opera Rigoletto or the number zero. Although sortal nouns are used here for disambiguation, the phrases are interpreted as functional concepts, and hence as semantic definites, because of the proper names that are

11

used.5 Löbner also subsumes under the category of proper names NPs with an abstract sortal head noun that is specified explicitly by a subordinate clause:

(1) the rumour that Reagan is going to resign (2) the idea to have pizza now

Unfortunately, he does not give an explanation for this view (and it is very questionable whether such phrases should be regarded as proper names), but in the discussion of Löbner's theory of contractions below we will stick to the classification of such phrases as semantic definites. The last group of cases that can be considered as proper names is much less controversial. Examples for this group are the World Trade Center in New York, the Tower in London or the Louvre in Paris. Here we have sortal head nouns, but since the objects that are referred to are unambiguous in a certain locally restricted domain of situations they can acquire the status of proper names. A similar line of reasoning lies behind the classification of certain uses of other sortal nouns as FC1s, though not as proper names. In a locally restricted domain, say, a particular neighbourhood, we can uses phrases that have sortal head nouns such as the pub, the supermarket or the playground unambiguously if they perform a certain unique role in that domain. Because of the unique role of these objects in a particular set of situations, the above examples give rise to functional concepts and are semantic definites. Another important group of FC1s are certain complex definite descriptions. Sortal or relational head nouns that are combined with superlatives, ordinals or adjectives such as last, only, next, same or other can form functional concepts. The head nouns provide a set of alternatives due to their non-functional status and the adjectives single out a particular member of that set. So in these cases we again get a one-to-one mapping from (a set of) situations to objects, i.e. we have semantic definites.

5

Löbner also claims that "definiteness [...] is not a matter of overall reference, but only a matter of the link provided by the noun following the article immediately." (p. 303) However, the link provided, e.g., by the sortal noun year in the definite NP the year 1984 does not seem to provide a one-to-one link, i.e. a functional link. It is not really clear to me how Löbner's conception of definiteness in fact applies to the examples cited above, but a detailed discussion of this question would lead us into a direction that is far off the topic of this paper.

12

Functional concepts that have both a situation and an object as their arguments (FC2s) are slightly more difficult to analyse in Löbner's framework.6 It should be noted that FC2s always have functional nouns as heads (Löbner 1985: 302); relational head nouns, as in some FC1s, do not seem to give rise to FC2s. (Sortal head nouns are not possible either, because they only have a single argument and hence cannot be used to build FC2s.) As we have seen above, FC2 nouns yield unambiguous values for their respective arguments. Löbner also uses the (more intuitive) term 'linking' instead of 'returning a value for the given arguments' in his discussion of FC2s. In his words: "FC2s provide one-to-one (i.e. unambiguous, functional) links from arguments to their values. The argument in turn, will be linked to the situation referred to [...]" (Löbner 1985: 301). We can imagine this idea as some kind of link-chain in which the value of a noun that gives rise to an FC2concept is linked unambiguously to an object and that object is then linked to the situational argument. (Of course this is a simplistic picture and it is a little unclear how FC3 nouns could be integrated into this picture, but it might be more intuitive than functions for some readers.) It is important to notice that the link between the argument and its value needs to be a one-to-one link, whereas the link between the situation and the argument can also be a one-to-many link. If all links are functional, the overall reference of the NP will be determined, but if the chain of links is somehow interrupted (e.g. by a one-to-many link between the situation and the argument as in the father of one of my friends) the reference of that NP will be indefinite, even though the definite article is used and the NP is considered definite. We have to distinguish between two kinds of definiteness here: The definiteness of the head noun and the determinedness of the whole NP (Löbner 1985: 303). Löbner is only concerned with the first kind, which is provided by the link between the argument and its value alone. If and only if that link is unambiguous (independently of the situation), then the NP will be semantically definite in Löbner's terms. The cases where all arguments are explicitly given are not that interesting. We simply check whether the argument and its value stand in a functional rela6

Löbner does not include functional three-place predicates in his detailed analysis of semantic definites. Distance, difference, or relationship are examples for the admittedly small group of functional nouns that need two objects as arguments in addition to the situational argument (FC3 nouns). In principle, they should show the same characteristics as FC2 nouns, but we are not going to explore them in more detail.

13

tion. If the arguments are not given, however, Löbner's theory seems to provide significant insight into linguistic phenomena such as the "associative anaphoric use" and the "introductory situational use" (Löbner 1985: 303). In introductory statements like (3) This is the clutch.

the implicit argument (most probably a car) is provided by the "immediate physical environment" (Löbner 1985: 304), it can thus be regarded as an implicit deictic7 argument. According to Löbner, here clutch represents an FC2 and requires an object as its argument (additional to the situational argument) and a suitable argument is provided by the situation. Even though the argument is dependent on a particular situation, the link between the argument and its value (the clutch) is an unambiguous link and the NP is a semantic definite. According to Löbner, only functional nouns can be used in introductory statements (Löbner 1985: 304), because only they represent general functional concepts. Löbner has adopted the notion of an introductory statement from Hawkins (1978) (Löbner 1985: 303). Whether or not this notion is sensible, Löbner's prediction that only nouns that are used functionally can occur in such statements does not seem to be correct. Consider the following examples:

(4) A veterinary has been called to the zoo. He wants to know which tiger it is that he is supposed to examine. The attendant answers: This is the tiger. (5) A group of children are visiting the zoo. Their teacher says: This is a tiger.

As I understand Löbner, only nouns that are always used functionally, and therefore represent general functional concepts, can occur in introductory statements. In (4), however, it is not at all obvious that the noun tiger represents a general functional concept, even less an FC2 with an implicit deictic argument. We could, of course, argue that tiger in (4) is a functional noun, because the referent of that 7

I use the term deictic in Löbner's sense here, i.e., the implicit argument refers to an object that is somehow present in the utterance situation, as opposed to objects that are introduced verbally in a text or a conversation.

14

noun performs a certain unique function in a locally restricted domain, but this interpretation does not correspond to Löbner's prediction: In (4) the head noun is not "a functional noun in its own right" (Löbner 1985: 304). As for sentence (5), the indefinite article is used, so tiger can obviously not be used functionally, and hence should not be possible in an introductory statement like (5). In anaphoric associative uses the implicit argument is usually provided by the linguistic context of an utterance. Consider the following example:

(6) Fritz read a review about an interesting book he wanted to buy. But when he arrived at the bookstore he had forgotten the title.

Title is usually used as a functional noun (a title is always the title of something and that something can only have one title) and the missing argument (an interesting book) is given in the preceding sentence. The link obviously is functional, so the noun title represents an FC2. This analysis is theoretically appealing because it does not need to assume discourse referents introduced by an interesting book and we can do away with speculations about the nature of the relation between a book and its title according to Löbner. "The crucial condition [for the use of associative anaphors, von mir eingefügt] is that the head noun [...] provides a general twoplace functional concept for which there is an appropriate argument in the immediate linguistic context." (Löbner 1985: 304)8

Let us now turn to the last class of semantic definites: Configurational uses of definite descriptions. Configurations are abstract situations that essentially consist of objects and certain relations between those objects. Such an abstract situation contains only those objects that are explicitly mentioned in a sentence and the objects are linked to each other via relations. These relations in turn are functional, i.e., one-to-one relations, because they unambiguously relate the objects to each other. Consider the following example from Löbner (Löbner 1985: 304):

8

It remains an open question in Löbner's paper, however, how we can recognise 'appropriate' arguments. Do they have to appear in the sentence directly preceding the anaphoric NP? Does the lexical entry of title somehow specify what kinds of arguments it can take? Songs, books, plays, newspapers?

15

(7) He was the son of a poor farmer.

The relational noun son does not provide a one-to-one link from its arguments to its value, or, to put it differently, sentence (7) does not imply that the poor farmer has only one son. Nevertheless, the definite article is used here and according to Löbner the definite article indicates that the following noun is to be interpreted as a functional concept (Löbner 1985: 314). So how do we get to an interpretation of son as an FC2? Löbner suggests that statements like (7) refer to an abstract situation, to a configuration, instead of referring to a particular situation. In the sentence at hand only one son and one father are explicitly mentioned and, therefore, the configuration has only those two objects as elements. So, the link provided by son unambiguously relates the son and the father and is thus a functional link and the head noun son represents a functional concept.9 That is why the definite article can be used in this kind of "generic" statements.10 It is obvious that the relations between objects that are provided by relational or functional nouns play a crucial role in configurational uses of definite descriptions. This explains why sortal nouns do not occur in statements that make use of abstract situations: They do not provide the necessary links between the elements of a configuration. Sentences like (8) where the relational noun is replaced by a sortal noun are unacceptable. (8) *He was the man of a poor farmer.11

9

It could be argued that here the link is a one-to-one link only accidentally and not due to the inherent meaning of the noun. The relation provided by son is inherently a one-to-many relation, not a functional relation. So, the noun son is obviously used functionally here, and not relationally. Interestingly, Löbner claims that both relational and functional nouns can occur in configurational statements and that the relational nouns give rise to functional concepts due to the structure of the abstract situation. The distinction between nouns and concepts seems to get rather fuzzy at this point. We will return to this point in Section 2.3 below. 10 The term generic is used by Löbner only to indicate that a statement refers to an abstract instead of a particular situation and should (probably) not be understood as generic in the sense that sentences like The whale is a mammal. are generic. We will discuss generic statements and generic nominals in Section 4.2.1. 11 It could be argued that statements like (8) are unacceptable because of the uninterpretability of a phrase like man of x. This interpretation can rather easily be rephrased in Löbner's terms, however: As we saw in Section 2.1, sortal nouns cannot express relations, and man is obviously used as a sortal noun in sentence (8). That is why something like man of x (which is supposed to express a relation) is not interpretable.

16

So far we have only been concerned with the object-argument and not with the situation argument of relationally or functionally used nouns. First we need to distinguish between an abstract situation and a "real" or "actual" situation (Löbner 1985: 306). An abstract situation is a configuration in the sense explicated above and the real situation is provided by the situational argument of a relational or functional noun. But how can we connect configurations to actual situations? Löbner claims the following. "Sentences involving [...] abstract situations convey the notion that the real situation referred to is of that type [...] configurational definites generate in a first step an abstract situation [...] which then must be anchored in one of several possible ways in the actual situation referred to." (Löbner 1985: 306) Anchoring in this sense means that the abstract situation must be a subsituation of the real situation, i.e. the real situation has to be such that all objects that are present in a configuration and all their relations to each other must also be given in the real situation. The relation between the abstract and the real situation is, of course, indefinite. The configuration is only one possible subsituation of the actual situation, that is why Löbner claims that it can be anchored in the real situation in "one of several possible ways". As was the case with FC2s above, also here the definiteness expressed by the definite article only applies to the one-to-one link between the argument of the noun immediately following the article and its value, and not to the determinedness of the reference of a complete phrase or sentence. Löbner also wants to include certain sentences that do not contain the definite article into the class of configurationally used NPs. He claims that statements like (9) and (10) generate abstract situations.

(9) She goes to church. (10) He came after lunch.

It is not really clear to me, however, in what sense church and lunch are nouns that can provide the necessary relations in a configuration. With regard to (9) we could say that the sortal noun church gives rise to a functional concept (an FC1, to be more precise), because it performs a certain unique function in a locally restricted domain of situations. I am not sure, though, whether this view actually conforms with the intuitive understanding of (9). The sentence is true iff "she"

17

goes to some church or other on a regular basis (in order to attend church service). It is not necessary that the object denotated by church has a certain unique function in a locally restricted domain of situations. We could easily imagine a street full of churches and that "she" goes to a different church in that street every Sunday. Here, none of the churches performs a unique function, they all have the same function, viz. that of being a place where mass is held. Another line of reasoning would be to point out that the noun church in (9) is used to refer to church service or mass, and not to a building. Similar problems seem to arise, though. In a configuration only those objects that were explicitly mentioned are consider, i.e., she and church in our example. But what could be the functional link that relates those two objects? Even if we accept that church service or mass are used functionally here, it is not really clear to me what kind of relation would be established by the head noun. Sentence (10) appears to be even less convincing. Lunch is certainly not relational or functional; so, again, we seem to have a sortal noun that is supposed to generate a configuration. We could, of course, argue that the noun lunch has an implicit argument (a particular date most probably, e.g. May 15th 1974), but this does not conform to Löbner's idea that only those objects are part of the configuration that have been introduced via explicit mention. And, again, even if we assumed that the noun was used functionally in (10), what kind of relation could be established by lunch in our example? The only other object that is explicitly mentioned is "he", so the relation would have to be an unambiguous relation between "he" and a lunch. What relation could this be? We could say that the verb provides a relation between the two objects (Löbner 1985: 315ff), but it is doubtful that a verb like come really provides two distinct thematic roles which "he" and lunch could play. We could, of course, claim that there exists some kind of verbal concept like come-after that could provide the necessary thematic roles that would then relate the two objects by a one-to-many link. But this solution seems rather ad-hoc and does not correspond to the syntactic structure of the sentence or to the usual syntactic characteristics of the verb to come. And is it really desirable to strain the lexicon that much? If we suggest a verbal concept like come-after, we would also have to propose something like come-before, come-between-x-and-y and maybe even more concepts. This seems rather redundant to me.

18

Let us summarise the discussion of semantic definites in Löbner's own words before we move on to pragmatic definites. In definite descriptions that are semantic definites the head noun is "always a functional concept, either by virtue of its lexical meaning plus, occasionally, additional general constraints12, or due to the restriction to limited abstract configurations within which relational concepts can be used unambiguously" (Löbner 1985: 307).

2.2.2 Pragmatic Definites Semantic definites represent functional concepts (and thus refer unambiguously) independently of a particular situation. Pragmatic definites, on the other hand, depend on the utterance situation for unambiguous reference and Löbner claims that they have sortal or relational head nouns, never functional head nouns. Löbner distinguishes three types of pragmatic definites: Endophoric, anaphoric and deictic definites.

Endophoric definites are nominals that contain a disambiguating attribute. This attribute can be a relative clause, or an adverbial or prepositional phrase, but in the following discussion we will restrict ourselves to relative clauses. For the endophoric use of a definite description it is essential that the relative clause provides an unambiguous link between the object referred to with the help of the head noun and other objects, i.e. it gives rise to a functional concept. Usually, the functional concept depends on the thematic roles that are given by the verb, and does not, as in the case of semantic definites, depend on the inherent meaning of the head noun, locally restricted domains of situations or abstract situations. Löbner claims that the thematic roles of a verb directly correspond to functional concepts (Löbner 1985: 315), that it is the inherent meaning of the verb that provides a one-toone link between objects and it depends on the particular situation whether the functional concepts refer unambiguously (Löbner 1985: 317). But also locational predicates or the time or purpose of an action can provide unambiguous links and thus give rise to a functional concept (Löbner 1985: 316). The idea behind this claim is that, with respect to verbs, there usually is one agent of an action and one recipient, so the relation expressed by a verb is functional. A similar argument can 12

The additional constraints that can help to form functional concepts consist of locally restricted domains of situations in phrases like the pub or certain adjectival attributes like first or next.

19

be given for locational predicates, for example. An object can normally be found in exactly one place, and that same place, in turn, cannot be occupied by another object, hence the functional link between an object and a location.

Let me exemplify the endophoric use of a definite description with an example:

(11) The party John went to last night was boring.

John is linked to a situation and the phrase party John went to last night relates parties to John via a kind of go-to-relation, i.e. it is a relational concept. If we further assume that people usually go to only one party on a particular evening (indicated by the phrase last night), then the nominal phrase actually constitutes a functional concept, relating John unambiguously to one party. Note, that it is the thematic roles of the verb that provide the functional interpretation of the definite description: One person usually goes to one place or event (at a particular time). Intuitively, one could propose a classification of endophoric definites as FC1s and hence as semantic definites. According to Löbner, however, this is not possible due to the inner structure of the NP party John went to last night.13 We can paraphrase (11) in the following way.

(12) John went to a party last night and the party was boring.

Here, party cannot be considered semantically definite in the sense discussed above. We have to consider the context of the utterance (i.e. the preceding sentence) in order to determine whether or not the sortal head noun party expresses a sortal or a functional concept. It is neither the inherent meaning of the noun or local restrictions of situations, nor an abstract situation that accounts for the functional interpretation of the head noun. So, in that sense we can say that endophoric uses of definite descriptions depend on a particular situation of utterance for unambiguous reference and should therefore be regarded as pragmatic definites.

13

Löbner also gives a second argument against the classification on endophoric NPs as semantic definites, viz. that the definite article only applies to the immediately following noun and not to the phrase as a whole. This seems a little circular, though. It is Löbner's proposal that suggests such a meaning of the definite article, so I do not think that this is a valid argument at this point.

20

Anaphoric definites differ from endophoric uses of definite descriptions in that they do not contain an explicit specification that gives rise to a functional concept (Löbner 1985: 317). Löbner restricts his discussion to direct anaphors and we will follow him with respect to this restriction. Direct anaphors refer to an object that has previously been introduced into the discourse. Let us see what Löbner has to say about the introduction of discourse referents and the structure of the "universe of discourse" (Löbner 1985: 317). A universe of discourse can be regarded as a network which is "a complex abstract situation, in which every constituent – object or event – plays a unique role" (Löbner 1985: 317). As we saw above, in an abstract situation the members of that configuration are related to each other via functional relations. Remember that only those objects that are explicitly mentioned figure in an abstract situation and that this is the reason for the functionality of the links between the objects. The same applies to a discourse-network. The constituents, i.e. the nodes of the network, are introduced via explicit specification and the relations between the constituents are one-to-one links. Additionally, each node is provided with a certain amount of predicative information.

Let us now come to the insertion of nodes into the network. According to Löbner, the first nodes that constitute the universe of discourse are objects or locations that are given by the utterance situation. New nodes will be integrated into this universe by means of explicit mention. Löbner claims that the "introductory situational use" of definite descriptions as in (3) (This is the clutch.) can be used to establish new nodes in the network. Also FC2s with semantically definite arguments are capable of first mention use (Löbner 1985: 303), and, of course, nouns that are preceded by the indefinite article. But there are many more possibilities to create new nodes. Most endophoric definites are capable of first mention use and so are FC2s with an implicit anaphoric argument ("associative" anaphors). Configurational uses of definite descriptions as well as FC2s with indefinite arguments can also introduce new nodes. All of these are nodes that refer to objects, but events are also constituents of the universe of discourse. They are created by the use of verbs and verbs can also contribute predicative information for objectnodes (Löbner 1985: 317). The relations between the nodes are established either by the inherent meaning of semantic definites, configurational uses of NPs and

21

additional general constraints, or by the thematic roles of verbs or locational or temporal attributes. To return to anaphoric definites, a nominal phrase can be used anaphorically if (and only if) it refers to a node already present in the universe of discourse. Since the nodes carry predicative information, the sortal or relational content of the anaphoric phrase can be reduced to a minimum, but they have to be sufficient to single out a particular discourse referent. Sometimes such additional information can be used to further specify a given discourse referent and will then be added to the predicative information already contained in the respective node or new links between the constituents will be established. To sum up, the sortal or relational head nouns of anaphoric NPs are pragmatic definites, because the unambiguous reference of these nouns crucially depends on the discourse-network, i.e. it depends on a particular situation of utterance.

We saw that anaphoric definites refer to previously introduced discourse referents. Deictic definites, in contrast, refer to objects that are present in the situation of utterance. At first sight, they appear to be pragmatic definites par excellence, but not all uses of deictic definites are pragmatic in Löbner's sense. The "immediate situation use" of deictic NPs is semantically definite. Löbner gives the following examples for this use.

(13) Beware of the dog! (14) Harry, mind the table!

In (13) the noun dog represents a functional concept, viz. the dog that belongs to the house, so the NP is a semantic definite. The implicit deictic argument is provided by the context of the utterance, but the NP itself is an FC2 and should be analysed analogously to (semantically definite) FC2s with implicit arguments that we have discussed above. This seems reasonable, but in my opinion Löbner's second example is less convincing. Here, table is supposed to represent a functional concept, too. It refers to "the table that always stands where it stands" (Löbner 1985: 310) and therefore, according to Löbner, gives rise to a functional concept. It is not really clear to me, however, how this notion of a functional concept corresponds to his earlier characterization of FCs. Furthermore, I do not think it is

22

necessary that the table always stands at a particular place if (14) is uttered. Harry's wife could have rearranged the furniture while Harry was asleep and she might use a sentence like (14) to stop him from running into a table that usually stood two feet further to the right. In that case Löbner's explanation for the functionality of the concept expressed by table is not applicable. Or what about sentences like (15), (16) and (17)?

(15) Mind the gap! (16) Beware of the bugs: Fire ants can kill Americans.14

(14) and (15) have a very similar structure, so the head noun in (15) should also express a functional concept. It is not really obvious, though, why gap should be regarded as a functional noun here. Does gap in (15) express a functional concept, because it is the gap between the platform and the door of the tube, i.e., because it stands in a certain relation to other object (as does dog to the house in (13))? Or because it is the gap that always is where it is, similar to Löbner's explanation for the functionality of table in (14)?15 It is not clear to me, either, in what sense bugs in sentence (16) could represent a functional concept. Functional concepts usually express relations, but what kind of relation could be represented by bugs in a sentence like (16)? Do the bugs belong to something or some place, i.e., are they bugs of something or other? An explanation along the lines of that for (14) does not seem very probable either. It seems that Löbner's account of what he calls the "immediate situation use" needs further clarification if we want to be able to handle sentences like (15) and (16) appropriately in his framework. There is a second use of deictic definites, which is pragmatically definite, because the unambiguous reference of the head nouns depends on a particular situation: The "visible situation use" (Löbner 1985: 309). "This use is possible if one and only one object of the kind belongs to the situation of utterance, if that object is visible (at least for the addressee) and known to him as such." (Löbner 1985: 310) Löbner proposes not only to include nouns that denote visible objects

14

from http://abcnews.go.com/GMA/Health/story?id=2143091&page=1

15

The speaker of (15) probably assumes that the hearer can recognise the functionality of the head noun, otherwise she would most likely use a different description. But if the hearer knows that the gap is where it always is, then what use would there be in uttering (15)?

23

into the category of "visible situation use", but also referents of nouns that are perceived by other organs than the eyes: smell and noise would be examples for such nouns. But also here I think that Löbner's theory needs further explication. (17) Pass me the table!

Imagine that Mary uses an orange box as a table. The noun table then represents a functional concept, because the object table refers to functions uniquely as a table (cf. Löbner 1985: 293). Statement (17) is clearly deictic and, hence, the noun is regarded as being pragmatically definite. The problem is that functional nouns are not possible in pragmatic definites (Löbner 1985: 307) according to Löbner.

2.3 Problems Now that I have presented Löbner's proposal for definites, I want to make a few remarks before we proceed to the application of this theory to contractions in German. I already hinted at some of the difficulties that I have with Löbner's ideas. These were of a rather specific nature, but there are also two general problems to the theory: The distinction between sortal, relational and functional nouns on the one hand and the role of situations on the other. As we saw in the above discussion, the three types of uses of nouns have distinct characteristics. In Table 1 I briefly summarised the most important aspects.

24

Type of Use of Characteristics Noun Sortal nouns

- conjunction of one-place predicates - classify objects - can represent functional concepts in restricted domains of situations - often used as head nouns in pragmatic definites

Relational nouns

- conjunction of one-place and many-place predicates, i.e., they specify one-to-many relations between objects - can generate abstract situations - often used as head nouns in pragmatic definites

Functional nouns

- conjunction of one-place and many-place predicates, they specify one-to-one, i.e., functional relations between objects - identify referents - refer unambiguously independent of utterance situation - represent functional concepts - cannot be used as head nouns in pragmatic definites - are usually not preceded by indefinite article

Table 1: Characteristics of different types of uses of nouns in Löbner's framework.

Sortal nouns are a conjunction of one-place predicates and classify objects (Löbner 1985: 293), they can represent functional concepts in restricted domains of situations (Löbner 1985: 299f) and are often used as head nouns in pragmatic definites. Relational nouns contain sortal information, i.e. one-place predicates, and additionally specify one-to-many relations between objects (many-place predicates), they often occur in sentences that generate abstract situations and in pragmatic definites. Functional nouns are similar to relational nouns in that they provide relations between objects in addition to sortal information, but here the relations are functional, i.e. one-to-one relations. They refer unambiguously independently of a particular situation of utterance due to their inherent meaning, they give rise to functional concepts (Löbner 1985: 296), always identify a referent instead of just classifying it (Löbner 1985: 293) and they cannot appear as head nouns in pragmatic definites (Löbner 1985: 307). Functional nouns also exhibit another characteristic which has not been mentioned yet: "They do not allow the 25

indefinite article as long as it can be presupposed that they have a referent" (Löbner 1985: 297). Roughly speaking, they can appear with the indefinite article only if it is not certain that they actually refer to something, e.g. in questions about the existence of some object. This appears to be a very problematic view. All count nouns can be preceded by the indefinite article. In particular those that Löbner regards as "the most perspicuous cases" (Löbner 1985: 294) of functional nouns, viz. nouns that refer to unique parts of objects like roof or head. Murderer is also considered an obviously functional noun by Löbner (Löbner 1985: 294). To me, the following sentences are perfectly acceptable.

(18) He fell off a roof yesterday and broke his left leg. (19) The other day, I found a head in the trash bin. (20) A murderer was arrested at the airport yesterday.

(18), at a first glance, looks like a sentence that generates an abstract situation which must then be anchored in the real situation. But what kind of relation should hold between "he" and a roof? The noun roof obviously does not provide a relation to "he". We could propose that the thematic roles that are provided by the verb establish a fell-off-relation. But then the distinction between sentences like (18) and endophoric definites would get very fuzzy. With respect to endophoric definites, Löbner argued that they are pragmatic definites mainly on the ground that the functional relations are not provided by the noun, but by the linguistic context, i.e. the verb, and that the unambiguous reference of the thematic roles depends on the particular situation of utterance. Similar thoughts can be invoked by (19), but the same problems arise. Sentence (20) is also problematic, because Löbner considers the head noun murderer an obvious case of a noun that is used functionally (Löbner 1985: 294), nevertheless the indefinite article can be used. Similar to (18) and (19) we could claim that (20) generates an abstract situation, but also similar to (18) and (19), it is not clear how or what kind of relations could be established between the objects that are explicitly mentioned in sentence (20). But, again, this argument might not be valid, because the head nouns in all the

26

above examples are not preceded by the definite article and Löbner only considered configurations that are constructed by means of definite descriptions.16 A different approach to sentences like the above could be to regard the head nouns as sortal nouns. Löbner concedes that nouns cannot be subdivided in a strict way and therefore prefers to talk of concepts instead (Löbner 1985: 295). However, I have the feeling that he does not stick to his own assumptions. If it depends on the rest of a sentence whether or not a noun is functional, then he should not use phrases like "[Definite descriptions] with a functional head noun obviously yield functional concepts" (Löbner 1985: 296) or "Pragmatic definites [...] have sortal or non-functional relational head nouns" (Löbner 1985: 307). It seems that a great deal of his account depends on the division of nouns into three distinct classes, but his argument seems to be somehow circular if this division really depends on the linguistic environment of a noun. I am not claiming, though, that Löbner's hypothesis about the meaning of the definite article is inappropriate. It might well be that the meaning of the definite article is to indicate that the following noun should be interpreted as a functional concept (Löbner 1985: 319). It is his methodology that I am not really satisfied with.

Let us now turn to the second difficulty that I have with Löbner's theory, the role of situations. If I understand Löbner correctly, the situational argument of sortal, relational or functional nouns refers to the situation of utterance. Locations, objects and relations between the objects that are present are somehow contained in the situational argument as well as the time and place of an utterance and information about the speaker and the audience (and possibly many more aspects, too). The situation of utterance plays a crucial role for the distinction of semantic and pragmatic definites. In the case of "'semantic definites' the referent of the definite is established independently of the immediate situation or context of utterance. [...] 'Pragmatically definite' NPs, on the other hand, are essentially dependent on spe16

I think that the same complications arise in a sentence like (21) where the definite article is used. (21) John went to the doctor.

Also in this case, the noun doctor does not provide a relation between "John" and a doctor, which is necessary for the generation of an abstract situation. The verb might establish such a relation, but then we run into the same problems as above.

27

cial situations and contexts for the non-ambiguity [...] of a referent." (Löbner 1985: 298)17 First of all, the classification of a noun as relational or functional does depend on the situation of utterance. I gave an example for this (the noun wife) in Section 2.1. It might be the case that a noun has one of the interpretations in a lot of different situations of utterance that share, say, the time and space aspects, and differ with respect to all other situational information. Nevertheless, a phrase like "independently of the particular situation referred to" (Löbner 1985: 299) seems inappropriate. This becomes even more evident, if we consider simple FC1s (cf. Section 2.2.1). According to Löbner, sortal nouns, for example, can give rise to functional concepts in a restricted domain of situations. Also here, the applicability of the definite article does depend on the situation of utterance. And, conversely, it is possible to imagine quite a few (if we like even rather similar) situations in which an NP like the girl (which should be pragmatically definite) refers unambiguously. The claims Löbner makes about the role of the situation of utterance are either too strict or to loose. But they surely are not very welldefined.

3. Contractions in Löbner's Framework In Section 1 we already saw that both contracted and non-contracted form can be used in German and we are looking for a theory that makes the correct predictions about the conditions for the use of either form. Löbner proposes a solution that draws on his distinction between semantic and pragmatic definites. In this section we will test Löbner's hypothesis with the help of different observations and examples, assuming, of course, that the theory presented above is free of any unclarities, i.e. we will ignore the doubts mentioned earlier. Löbner claims the following: "Cliticization of the article to a preposition as a host [...] is possible if and only if the NP is semantically definite and not too complex [...] Contracted forms are particularly frequent in configurational uses [and ...] the contracted forms are not possible [...] as soon as they are used to refer "pragmatically" in our sense to specific objects" (Löbner 1985: 311f). Let us first concentrate on those uses of definite descriptions, which do not allow contracted forms, i.e. pragmatic definites. 17

It is unclear what Löbner means by "context" here (the linguistic environment? the universe of discourse? or are situation and context identical?), but we will ignore this vagueness.

28

(1) Ich habe mir dieses Buch von dem (*vom) Mann, der nebenan wohnt ausgeliehen. (I borrowed this book from the (*CONTR-from the) man, who lives next door.) (2) Paul hat sich ein Haus gekauft. Zu dem (*Zum) Haus gehört ein riesiger Garten. [Paul has himself a house bought. To the (*CONTR-to-the) house belongs a huge garden.] (Paul bought a house. With the house comes a huge garden.) (3) Pointing to a shop: In dem (*Im) Laden kann man Wein kaufen. [In the (*CONTR-in-the) shop can one wine buy.] (They sell wine in the shop.)

In the examples above the nominals that are interesting for our discussion are endophoric, anaphoric, and deictic definites, respectively. It is indeed the case that the contracted forms cannot be used in pragmatic definites, Löbner's theory makes correct predictions. His claims about contractions in semantic definites, however, are much less persuasive. First of all, the "if and only if" in the first sentence of the citation is too strong. The implication only works in one direction: If a contraction is used, then the following NP is a semantic definite, but if the following NP is semantically definite it is not a necessary consequence that contractions can be used. We can find many examples for semantic definites that cannot be preceded by a contracted form (or that are at least preferably used with the un-contracted forms) if the referents of the respective nominals are not known to some extent to the audience.

(4) Fritz hat gestern eine Rezension über ein interessantes Buch gelesen, das er sich heute kaufen wollte. Von dem (?Vom) Titel hatte er sich allerdings nur den ersten Buchstaben gemerkt. (Yesterday, Fritz read a review about an interesting book that he wanted to buy today. Unfortunately, he could only remember the first letter of the (?CONTR-of-the) title.)

29

(5) Bei dem (?Beim) Patienten Müller muss man aufpassen, er ist cholerisch und gewalttätig. (You have to be careful with the (?CONTR-with-the) patient Müller, he is choleric and violent.) (6) Karl hat diese Informationen von dem (?vom) Bürgermeister einer kleinen walisischen Stadt bekommen. (Karl got this information from the (?CONTR-from-the) mayor of a small town in Wales.)

Sentence (4) is a slightly modified version of one of Löbner's own examples for FC2s with implicit anaphoric arguments (cf. Löbner 1985: 304). The definite description is semantically definite, nevertheless the non-contracted form is preferred. In (5) the phrase corresponding to the patient Müller is also a semantic definite, it is an FC1 which has a usage that resembles that of a proper name (cf. Löbner 1985: 299). But also here the contracted form does not seem to be preferred if the referent of the phrase is not known to some extent. To clarify this point, consider the following different situations in which (5) could be uttered:

(5a) A nurse starts working in a hospital. She does not know any of the patients yet and is given a list that contains information about each patient. The head nurse explains the list and points out that the patient Müller is a somewhat problematic patient. (5b) Two nurses work in the same hospital and they know the patients. They talk about their experiences with different patients. When they come to Mr. Müller one of the nurses utters (5).

In the second situation the contracted form will most probably be preferred, while the un-contracted form will be used in a situation like (5a). In both cases the definite description is used as a semantic definite, and should therefore exhibit the same characteristics according to Löbner. It is not clear, how the difference in the use of the contracted and the un-contracted form can be accounted for in Löbner's framework.18 Similar problems arise in sentence (6). The definite description the mayor of a small town in Wales is, again, one of Löbner's own examples (Löbner 18

In Section 4 I will propose an account of contractions that gives an explanation for the difference between the use of the definite description in situations like (5a) and (5b).

30

1985: 302). The phrase expresses an FC2 with an indefinite argument and like in (5) the contraction appears to be unacceptable in certain contexts, at least to some speakers of German. If both the speaker and her audience know the person referred to to some extent, the contraction will probably be used, and the uncontracted form is usually preferred if the definite description is e.g. used to introduce a new discourse referent that was not known until then. So, also in this example we seem to use either the contracted or the un-contracted form, depending on the knowledge of the speakers about the referents of the respective definite description. As stated above already, Löbner's proposal does not seem to be able to account for this difference in use. Secondly, it should be noted that the formulation "cliticization [...] is possible" (Löbner 1985: 311) is inappropriate. It implies an optional choice on the part of the speaker between the contracted and the non-contracted form. However, it is usually not a matter of choice whether a contraction can be used or not.19 We have already seen examples for the unacceptability of contractions, but we can also find instances of NPs that can only occur with contracted forms.

(7) Vom (*Von dem) Nachdenken bekommt Paula immer Kopfschmerzen. [CONTR-from-the (*from the) thinking hard gets Paula always a head ache.] (Thinking hard always gives Paula a headache.) (8) Im (*In dem) letzten Jahr ist meine Nichte in die Schule gekommen. [CONTR-in-the (*in the) last year has my niece into school got.] (My niece started attending school last year.) (9) Im (*In dem) Irak herrscht Bürgerkrieg. [CONTR-in-the (*in the) Iraq there is civil war.] (There is civil war in Iraq.) (10) Ludwig Erhard war im (*in dem) Deutschland der Nachkriegszeit ein bekannter Politiker. [Ludwig Erhard was CONTR-in-the (*in the) post-war Germany a wellknown politician.] (Ludwig Erhard was a well-known politician in post-war Germany.) 19

There are rather few definite descriptions that appear to always have the same meaning, irrespective of use of the contracted or the non-contracted form. In the next section I will suggest that this apparent indifference to the contraction is in fact due to the possible over-riding of certain constraints.

31

In all of the above cases the contracted forms are obligatory. It is not easy to decide whether nominalised verbs like Nachdenken (thinking hard) in (7) are semantic or pragmatic definites. I would tend to regard them as semantic definites, though. In sentence (7), at least, an abstract situation seems to be generated which has to be anchored in the real situation. It appears to be generally the case that nominalised verbs cannot be preceded by un-contracted forms, so the contraction is not only possible but even obligatory here. The NP corresponding to last year in (8) is a semantic definite (cf. Löbner 1985: 301) and the contraction is, again, mandatory. Some, but not all, proper names that refer to regions or monuments are used with the definite article in German. If such a proper name, i.e. such a semantic definite, is preceded by an appropriate preposition, the contracted form has to be used, as can be seen in sentence (9). Interestingly, even those proper names that usually are not preceded by the definite article require contracted forms as soon as they are modified in a certain way. If the modified noun phrase yields a sortal concept in Löbner's sense the contractions are obligatory, as can be seen in (10). Unmodified proper names of this kind do not allow contracted forms.

(11) Ludwig Erhard war in (*im) Deutschland ein bekannter Politiker. (Ludwig Erhard was a well-known politician in (*CONTR-in-the) Ger many.)

I said that modified nominals can yield sortal concepts in Löbner's sense. This means that a phrase like Deutschland der Nachkriegszeit (post-war Germany) is not used as a proper name, but it refers to a certain kind (of whatever it is that the proper name usually refers to). The modified NOM in (19) seems to imply a set of "different kinds of Germany", including post- and pre-war Germany, Southern Germany, Germany in the 1840ies, etc, and the definite description is used to refer to one of the members of that set with the help of the sortal information that is provided by the respective modifying phrase, der Nachkriegszeit (post-war) in our example. Whether or not such sortal concepts give rise to FCs that are then either pragmatic or semantic definites shall not interest us at this point, the reason being that, if we strictly follow Löbner, it should not be possible that phrases that contain proper names can form sortal concepts in the first place. In Section 2.2.1 we

32

saw that such phrases are considered as being "very close to proper names" (Löbner 1985: 299) in Löbner's framework and that they are not regarded as sortal concepts, but as FC1s. I am stressing this point, because, as I already pointed out, there is linguistic evidence that proper names and certain modified versions of proper names behave differently.

(12a) Deutschland, Süddeutschland (Germany, South-Germany) (12b) Deutschland der Nachkriegszeit, südliches Deutschland, wiedervereinigtes Deutschland (post-war Germany, Southern Germany, reunified Germany)

The nouns in (12a) are used as proper names and do not occur with contracted forms. Those in (12b), on the other hand, refer to a certain kind (of Germany) and require contracted forms.

Let us briefly examine a few examples that Löbner gives for cases of semantic definites that are preceded by contractions (Löbner 1985: 312).

(13) im Mai, zur Halbzeit [CONTR-in-the May, CONTR-at-the halftime] (in May, at half-time) (14) Er hat beim Poker 2 Pfund verloren. [He has CONTR-at-the poker two pounds lost.] (He lost two pounds playing poker.)

Löbner does not explain his examples any further, so let us assume that he considers them as obvious cases of semantic definites. We begin with the phrase zur Halbzeit. I would like to point out that Halbzeit is not an inherently functional noun, it can be used to refer to either the first or the second half-time, so the noun does not express a one-to-one relation. Furthermore, I want to suggest that the phrase zur Halbzeit is an idiom and hence should not be regarded as an example for semantic definites. If zur Halbzeit is used without the adjectives erste (first) or zweite (second) it refers to the half-time break, i.e., it is synonymous to zur Halbzeitpause. In that case, the noun would in 33

fact express a functional concept. Nonetheless, zur Halbzeit seems to be a rather misleading example for a semantic definite, because the functionality of the concept expressed by the noun is crucially dependent on the meaning of Halbzeitpause, and does not depend on the inherent meaning of Halbzeit. With regard to im Mai I guess that Löbner considers it an obvious case of a semantic definite because it requires a year as an argument and that there is always exactly one May for each year.20 I do not think, however, that Mai (or any other month) necessarily needs an argument. Consider the following sentences.

(15) Mein Freund hat im Mai Geburtstag. (My boyfriend's birthday is in May.) (16) Cäsar starb im März. (Cesar died in March.)

In both cases the contracted form is obligatory, so the respective phrases should be semantically definite. But why? Neither noun requires a year as an argument. My boyfriend's birthday is in May every year. As for (16), the sentence is perfectly acceptable without any need for (direct or indirect) reference to the year 44 BC. In my opinion, the uses of May and March, respectively, should again be regarded as sortal concepts and it is not at all obvious that they are clear cases of semantic definites. If that was correct, a recourse to abstract situations would not explain the semantic definiteness of (15) or (16) either. The nouns are sortal and hence cannot provide a relation to the subjects. And if we assume that the verbs provide the needed relations, we run into the same problem as in Section 2.3 with (14) and (15), viz. that the distinction between semantic and pragmatic definites would get rather blurred. With respect to (14), it is a clear case of obligation of the contracted form. The non-contracted forms are unacceptable in conjunction with the noun Poker. This, again, shows that the choice between contracted and non-contracted forms is not optional at all.

20

I chose this interpretation, because regarding Mai as a proper names seems implausible and I could not come up with another explanation for the classification of May as semantic definite.

34

I do not have much to say about Löbner's claim that contractions occur frequently in configurational uses. In my opinion, this formulation is so weak that not much is in fact conveyed by it. And I do not want to repeat my arguments against abstract situations or against the view that nouns like hospital or doctor are nonsortal yet again.

Before we proceed to a new (and hopefully more adequate) theory of contractions, let us briefly return to pragmatic definites and the impossibility of contractions in such uses. Consider the following example from Section 1 repeated here for convenience.

(17) ... und dann habe ich zu der (*zur) alten Hexe gesagt, sie solle unsere Kinder endlich in Ruhe lassen. [... and then have I to the (*CONTR-to-the) old witch said, she should our children finally leave alone.] (... and then I told the old witch that she shouldn't harass our children any more.)

Epithets are always used as anaphoric NPs21 and cannot be preceded by contracted forms. As pointed out above, Löbner's claims about contractions and pragmatic definites seem accurate. Epithets are used anaphorically, are pragmatic definites and the prediction that they do not occur with contractions seems to be correct. I do not see a way, though, to give a satisfying account of epithets in general within Löbner's framework. Let us, in a few words, recapitulate how the anaphoric link can be established according to Löbner. The universe of discourse is a network that is built up in the course of a discourse. It is an abstract situation, so only those objects, events and relations are part of the network that are explicitly mentioned. The

21

I could not come up with examples for the deictic or the endophoric use of epithets (but maybe there are some??). In sentences like (18) Do you see the old witch over there? (19) The old witch, who always scared our children, died at last. the definite description appears to be used deictically and endophorically, respectively, in Löbner's sense. However, such sentences seem to be acceptable only if the hearer already has some information about the referent of the NP, and then I would regard the NP as an anaphoric expression.

35

predicative information that is associated with each node is provided by the inherent meaning of the nodes or by explicit specification. An NP which is used anaphorically refers back to nodes in the universe of discourse and "it is sufficient to employ some distinctive sortal information in order to refer to the node in question" (Löbner 1985: 318). With respect to epithets, however, it seems questionable whether the information that they supply can ever be sufficient to single out a particular node in the network. Let us try to find out what kind of information epithets do provide. If a phrase like the old witch is used as an epithet it obviously does not have its literal meaning and I would suggest that nouns that can be used as epithets have an additional lexical meaning.22 To simplify matters, let us assume that the old witch has a literal meaning and a non-literal meaning roughly corresponding to the mean and hateable woman. Interestingly, an epithet does not really provide (possibly additional) sortal information about the referent, rather it expresses the attitude of the speaker towards the referent. In a sentence like (17), for example, the person referred to by the old witch is not necessarily a mean and hateable woman (she might in fact be a very friendly woman), but the speaker of that sentence considers her mean and hateable. The problem of analysing epithets in Löbner's framework seems to lie in the difficulty of singling out a particular discourse referent by means of the sortal information alone. As I understand his conception of the universe of discourse, the nodes (i.e. objects in our example) that an anaphoric NP refers to have to fulfil (at least part of) the descriptive content of the anaphoric NP (cf. Löbner 1985: 309) and the descriptive content, in turn, has to be sufficient to single out a particular node in the network. It is not at all obvious that the descriptive content of an epithet is really sufficient to establish an unambiguous anaphorical link to a previously mentioned object, usually additional information (e.g. the attitude of the speaker towards the referent) has to be taken into consideration. In Löbner's proposal, however, does not seem to allow for the accommodation of such additional information. Furthermore, it is not the case that the node that is referred to always shares at least part of the predicative information that is given by the lexical 22

The adjectives that can precede nouns that are used as epithets do not seem to contribute much to the meaning of the whole phrase. Because of this, I only propose to assign an additional lexical meaning to the nouns and not to the adjectives. (They should be regarded as semantically vacuous in most cases.??)

36

meaning of an epithet. In (17) the predicate woman, which is part of the lexical meaning of the epithet, is most probably contained in the predicative information of the corresponding node. But epithets like the dragon can be used to refer both to men and to women and the meaning of the epithet does not provide even such basic predicates as man or woman which could help to single out a particular discourse referent. Maybe the thematic roles of the verb could provide an unambiguous link to a previously mentioned object, but I do not think that this solution would be in Löbner's spirit. "VPs either supply predicative information for existing nodes or introduce new objects or event nodes." (Löbner 1985: 317) Löbner apparently does not consider verbs as being able to establish anaphoric links.

In this section we have tested Löbner's hypothesis about contractions with a large set of examples. Although the theory correctly predicts that endophoric, anaphoric and deictic NPs cannot be preceded by contracted forms, Löbner's claims about contractions in front of semantic definites are less convincing. It is not a matter of choice whether or not contractions are used, phrases like "particularly frequent" (Löbner 1985: 312) or "cliticization [...] is possible" (Löbner 1985: 311) do not characterise the use of contracted forms correctly. Furthermore, we have found examples for semantic definites that cannot be preceded by contractions. Epithets appear to resist an analysis along the lines of Löbner's theory, too. In short, Löbner's hypothesis seems appropriate for pragmatic definites (with the exception of epithets), but with respect to semantic definites this hypothesis seems rather inadequate. In the following section I will propose a different theory of contractions. It is inspired by Donnellan's theory of definite descriptions (1966), the primary reason being that epithets can be dealt with rather easily in Donnellan's framework.

4. A New Theory of Contractions In this section I want to propose my own theory of contractions. I will introduce three different uses of NPs: The contextual, the small-world and the generalised use. The latter two require contracted forms, while the first can only be preceded by un-contracted forms. The contextual and the small-world use can be subsumed under the category of the specific use of definite descriptions, which roughly cor37

responds to Donnellan's referential use, the generalised use being somewhat similar to the attributive use of definite descriptions. We will also see that some attributively used NOMs can occur both with contracted and un-contracted forms. This is not arbitrary, though, and I will suggest that syntactic characteristics can override the constraints imposed by the different uses of definite descriptions.

4.1 Motivation Let me briefly explain why Donnellan's theory of definite descriptions might be a good starting-point for an analysis of contractions. During the discussion of Löbner's theory we saw that epithets seem to resist a thorough analysis and that generic statements are not included in Löbner's considerations. In my opinion, both epithets and generic statements can be dealt with in a rather straightforward fashion in the framework proposed by Donnellan. As I said earlier, epithets are always used anaphorically. They refer to a specific object (that has been introduced already). If a definite description is used to refer to someone particular, it is used referentially in Donnellan's sense. The descriptive content of the NP is just a tool to enable the audience to pick out the right referent (Donnellan 1966: 249), it is not necessary that the referent actually fits that description. As we have seen in Section 3, the information that is provided by an epithet does not have to be consistent with other sortal information that is associated with the referent of the NP. The reference to an object can be successful even if that object does not fit the description provided by the NP (Donnellan 1966: 249). So, the problems that we ran into earlier in Löbner's framework, viz. that the meaning (or descriptive content) of the epithet did not enable us to refer to a particular node in the universe of utterance, do not seem to arise if we apply Donnellan's analysis. Note, that anaphoric and referential are not synonymous. The referential use of a definite description does not depend on previous mention of an object that fits the definite description. Epithets are rather special in the sense that they are referential expressions that can only be used anaphorically.

Before we discuss the three different uses of NPs that I want to propose in detail, I want to emphasise that the context of utterance is crucial for deciding whether an NP is classified as small world, contextual or generalised. There are in general no 38

such things as "contextual nouns", but "a definite description occurring in one and the same sentence may, on different occasions of its use, function in either [of the three] way[s]" (Donnellan 1966: 247).

(2) Fritz ist beim Arzt. [Fritz is CONTR-with-the doctor.] (Fritz is with the doctor.) (3) Fritz ist bei dem Arzt. (Fritz is with the doctor.)

Let us now look at different situations in which the above sentences could be uttered.

(2a) Anna, Felix and Fritz live in a small village. There is only one doctor in that village. Anna asks Felix about Fritz' whereabouts. (2b) Fritz has been feeling ill for some time already and went to see a doctor. Felix tells Anna that. (3a) Anna is at a party and asks her friend Felix where Fritz is. Both Anna and Felix know that there is one doctor at the party and Felix thinks that Fritz is talking to that doctor.

The noun Arzt (doctor) in (2) is used rather differently in situations (2a) and (2b). In (2a) Felix refers to a particular person, viz. the person who is practicing medicine in the village. The definite description is used referentially in Donnellan's sense. In (2b), on the other hand, the referent of the noun Arzt is not determined. It is totally irrelevant which doctor it is that Fritz is seeing, but it is essential that the referent fits the description, i.e. that he is a GP. This use of the NP is similar to the attributive use of definite descriptions, Fritz is seeing someone or other who fits the description, i.e. who is a doctor. Sentence (3) has a structure very similar to that of (2). But, again, the use of the noun Arzt is rather different from those in (2). The description is used referentially, Felix has a particular person in mind that he makes a statement about. In that sense, (3) and (2) uttered in the situation (2a) are similar. However, in (2a) the description is not only a tool to enable the audience to pick out the right referent, as is the case in (3). It also indicates that the person referred to plays a unique role in a locally restricted domain, here the 39

(unique) role of a person who practises medicine in the village. So, even though (3) and (2) uttered in situation (2a) resemble each other to a large extent, I would like to distinguish the uses of the respective NPs as the small-world (in (2a)) and the contextual use (in (3)) of a definite description, because they behave differently in the presence of a preposition. The contextual use requires the noncontracted forms, whereas the small-world use (and also the generalised use in (2b)) can only be preceded by contracted forms.

The threefold distinction between uses of nouns can also be found in the following examples.

(4) Anna and Felix are talking about a party some time ago. Anna says: An dem Abend habe ich mich richtig besoffen. [CONTR-in-the evening have I got myself really drunk.] (I got really drunk that night.) (5) Felix tells Anna about the plans for his wedding next Sunday. Am Abend gibt es dann noch ein Feuerwerk. [CONTR-in-the evening there will be also a display of fireworks.] (There will also be a display of fireworks in the evening.) (6) Anna talks to Felix about her daily routines and says: ... und am Abend trinke ich ein Glas Rioja und rauche eine Zigarette. [... and CONTR-in-the evening drink I a glass of Rioja and smoke a cigarette.] (... and in the evening I have a glass of Rioja and a cigarette.)

In (4) and (5) the noun Abend (night or evening) is used referentially, in (6) it is used similarly to the attributive use. The use of the definite description in (5) differs from that in (4) in that the evening that is referred to plays a unique role in a small community, viz. the role of being the evening of Felix' wedding. Abend in (6), in contrast, does not refer to a particular evening, but rather to some evening or other. And, again, we see that the preposition-article contraction cannot appear in front of a noun that is used contextually (as in (4)).

To sum up, the occasion of an utterance is crucial for determining the way in which a definite description is used and a distinction into three different kinds of 40

uses is motivated by linguistic evidence. Let us now turn to a more detailed description of the three uses of NPs.

4.2 Three Different Uses of Definite Descriptions I will begin with the generalised uses of a definite description and then we will proceed to the specific use, which can again be subdivided into the small-world and the contextual use of an NP.

4.2.1 The Generalised Use As stated above, the generalised use of a definite description roughly corresponds to the attributive use of Donnellan, in the sense that the speaker does not refer to a particular object, but rather to a set of objects that fit the description or to an institution or a common (i.e. not necessarily unique) role. Additionally to (2) uttered in situation (2b) and to (3), all of the following are examples for the generalised use of a noun.

(7) Mein Freund arbeitet im Krankenhaus. [My friend works CONTR-in-the hospital.] (My friend works in a hospital.) (8) Cäsar starb im März. [Cesar died CONTR-in-the March.] (Cesar died in March.) (9) Anders als beim Menschen, bei dem die Nase aus dem Gesicht ragt, geht beim Leguan der Kopf einfach in die Schnauze über.23 [Different CONTR-from-the human, whose nose protrudes from the face, CONTR-at-the iguana the head simply merges with the mouth.] (In contrast to humans, with their noses protruding from the face, the iguana's head simply merges with the mouth.)

The identity of the referent of hospital in (7) is irrelevant, what is important is that the friend works in some place or other that fits the description, i.e. that he works

23

from anatomie.htm

http://www.green-iguana.net/gruener-leguan/leguan-

41

in a hospital.24 The speaker of (7) might not even know which hospital it is that his friend works in, still the sentence is perfectly acceptable. Also in (8), the speaker might not even be aware of the fact it was March 44 BC in which Cesar died, so he is not referring to a specific object, but only to some object that fits the description. Sentence (9) could be regarded as a statement about a general property of iguanas; it is a characterising sentence. The phrase beim Leguan (roughly: the iguana's) refers to the kind Iguana-Iguana, it is a kind-referring nominal. (cf. Krifka (1995)). So here we have an example for a generic sentence that contains a generic NOM and I propose to subsume both of these notions of genericity under the category of generalisingly used definite descriptions. The idea to regard kindreferring nominals as being used generalisingly seems to be rather obvious. But what about object-referring, i.e., non-generic definite descriptions that occur in characterising sentences? Consider the following example:

(10) Im Winter ist es kalt. [CONTR-in-the winter it is cold.] (It is cold in winter.)

(10) is a generic sentence and the noun Winter is not kind-referring. The sentence does not express "specific episodes or isolated facts, but instead [reports] a kind of general property, that is, [reports] a regularity which summarises groups of particular episodes or facts" (Krifka (1995): 2). The nominal does not refer to a particular object and should therefore be regarded as an instance of the generalised use.

It is often possible to refer to a particular instance of the kind described by the NOM, for example with the help of pronouns. Consider the following example.

(11) Anna war gestern beim Zahnarzt. Er hat ihr einen Zahn gezogen. (Anna went CONTR-to-the dentist yesterday. He pulled one of her teeth.)

24

The indefinite article in the English translation already indicates the non-determinedness of the referent of the nominal. In German, however, the definite article is used, if we assume that the contracted forms are built with the help of a preposition and the definite article.

42

I suggest that the anaphoric reference to dentist is accomplished by accommodation. The phrase beim Zahnarzt (roughly: at the dentist) does not introduce a discourse referent. Otherwise (12) would not be possible.

(12) Fritz war heute nicht beim Friseur. Es gibt nämlich keinen in dem kleinen Dorf, in dem er wohnt. (Fritz did not go CONTR-to-the hairdresser's today, because there is none in the small village he lives in.)

Admittedly, the sentences are a little awkward, but acceptable. The example shows that it seems inappropriate to suggest the introduction of a discourse referent via the phrase beim Friseur (roughly: at the hairdresser's). We can also find other examples that seem to support the idea that a generalising definite description does not introduce a discourse referent. (13) Ruth geht morgen zum Zahnarzt. *Der andere Zahnarzt ist im Urlaub.25 (Ruth is going CONTR-to-the dentist tomorrow. *The other dentist is on vacation.)

A contrast is not permissible, because there is no discourse referent that could be contrasted by the phrase der andere Zahnarzt (the other dentist). I said above that it is often possible to refer to a particular instance indicated by the generalised use of a noun. In (14), for example, the pronoun seems inappropriate (at least according to my intuitions).

(14) Ruth geht morgen zum Zahnarzt. ?Er wohnt in der Lerchenstrasse. (Ruth is going CONTR-to-the dentist tomorrow. ?He lives in Lerchenstreet.)

Maybe accommodation is only possible if the sentence containing the pronoun somehow expresses a property that is associated with the concept of the descriptive content of the nominal, the concept of a dentist (whatever that means) in our

25

This example is taken from Bosch (1995).

43

example (cf. example (11)). But this is just a guess and needs further investigation.

Nominalised verbs seem to represent a class of nouns that are almost always used generalisingly. Consider (15), repeated here from Section 3.

(15) Vom Nachdenken bekommt Paula immer Kopfschmerzen. [CONTR-from-the thinking hard gets Paula always a headache.] (Thinking hard always gives Paula a headache.)

This might be due to the meaning of such nouns, but this suggestion would need further examination, too.

For the generalised use of a definite description it is essential that the description fits. It is not used as a tool for identifying a particular referent, it rather classifies objects with the help of the descriptive content of the respective nominal. In the presence of an appropriate preposition the contracted forms have to be used.

4.2.2 The Specific Use The specific use of a definite description resembles the referential use of NPs in Donnellan's sense. In the specific use the speaker always has a particular object in mind that he makes a statement about. It is not always necessary that the description actually fits the object referred to. The specific use of definite descriptions does not in general exhibit a uniform distribution of contracted or un-contracted forms. A subdivision of specifically used definite descriptions seems necessary.

4.2.2.1 The Small-world Use The small-world use of a definite description always requires the contracted forms. In Section 4.1 we already saw a few examples for the small-world use. Generally speaking, definite descriptions belong to the small-world category if they refer to a particular object that has a unique function or role in a regionally restricted domain or in a small community. There are, however, two different ways in which small-world nominals can be used: They can be used in bridging anaphors and as local names. The reference of a nominal that is used in bridging

44

anaphors is established via accommodation. Local names, on the other hand, are used to refer to individuals that are unique in a particular community. Let us begin with the latter kind of nominals that belong to the small-world category.

In certain situations all of the following examples are instances of nominals being used as local names.

(16) Anna war gestern beim Priester. [Anna was yesterday CONTR-at-the priest.] (Anna went to see the priest yesterday.) (17) Felix hat seinen Rucksack im Rathaus vergessen. (Felix forgot his rucksack CONTR-in-the town hall.) (18) Der Sohn vom Bürgermeister wurde am Sonntag verhaftet. (The son CONTR-of-the mayor was arrested on Sunday.)

Suppose that Anna lives in a village that only has one priest. If (16) is uttered in such a situation, then the nominal Priester (priest) refers to the village's priest and we could say that the nominal is used as a name here. Of course, this interpretation is only possible in a restricted domain, that is why I propose the term local name for this kind of small-world nominals. The nominal Rathaus (town hall) in (17) can be understood as referring to the local town hall. In this case the NOM would, again, be used as a local name. An analogous interpretation can be given for (18): A city usually has only one mayor and members of a particular community can use the nominal Bürgermeister (mayor) as a local name that refers unambiguously to a certain person. In the above examples, the descriptive content of the nominal fitted the referent. It is not always necessary, though, that the description fits the object that is referred to.

(19) Das Buch liegt da vorne beim Tisch. (The book lies over there, CONTR-near-the table.) (20) Felix wurde gestern vom General zum Essen eingeladen. (Yesterday, Felix was invited to dinner CONTR-by-the general.)

45

Sentence (19) is acceptable if the object referred to by Tisch (table) has a unique function in a locally restricted domain or in a small community. The descriptive content of the nominal is a tool that enables the audience to pick out the right referent, in our example it also corresponds to the kind of function that the referent has. It appears to be a general characteristic of the small-world use of a definite description that the descriptive content indicates the function or role that a particular object has. This does not mean, however, that the object necessarily fits the description. In (19), for instance, the noun table could well be used to refer to an orange box which happens to function as a table in the situation at hand. So, the description does not fit the referent, but it indicates the function of the referent and is used as a local name. Also in (20) the referent of the NOM does not have to be a general. He might be a man with a pedantic and austere character who never served in the military. Nevertheless, if he is regarded as performing a unique role (i.e. the role of being a general) in a small community, the nominal general is used as a local name. Of course, the same sentence could not be used in the same way if uttered to someone who is not part of that community, so the use of local names is restricted to certain social circles or to certain regions. Nouns that refer to regions or named buildings should also be considered as local names (even though the community in which such names are used is rather large). Nouns of this kind which are usually not accompanied by the definite article belong to this kind of small-world use if they are modified in certain ways (cf. Section 3). In (21) I have listed a few NOMs that cannot occur with uncontracted forms.

(21) im Nahen Osten, am Brandenburger Tor, im Deutschland der Nachkriegszeit, am Mount Everest, im südlichen Italien, beim Olympiastadion, am Funkturm, im Berlin der 20er Jahre, im Jüdischen Museum, zum Kongresszentrum, im Sportpalast (in the Middle East, at the Brandenburg Gate, in post-war Germany, at Mount Everest, in Southern Italy, near the Olympic Stadium, at Funkturm, in Berlin in the twenties, in the Jewish Museum, to the Kongresszentrum, in the Sportpalast)

46

It should be noted that in their discussions of contractions Haberland (1985: 86ff.) and Hartmann (1980: 17f) make use of a notion very similar to local names in our sense. "Einwohner eines Dorfes können auf Grund ihres allgemeinen Wissens über den Gemeinderat, die Kirche, den Pastor, die Mittelpunktsschule usw. reden, Mitglieder einer Familie über den Vater, die Mutter, die Großmutter, die Oma usw." 26(Hartmann (1980): 18). Both Hartmann and Haberland point out that there are some German dialects (most notably Fering and the Mönchengladbachdialect) that have two different definite articles, one being used with local names, the other being used with all other nominals. Apparently, contractions can usually be used in contexts that would allow for the definite article that combines with local names (Haberland (1985): 89). Unfortunately, I am not very familiar with the German dialects mentioned, so it is not possible to verify Haberland's and Hartmann's claims in this paper. Nevertheless, I think their notion similar to local names is very helpful for analysing contractions.

As stated above, what is referred to as "FCs with implicit anaphoric arguments" in Löbner's proposal or sometimes as "bridging anaphors" is also an instance of the small-world use in our sense.

(22) Peter hat neulich sein altes Radio repariert. Am Verstärker war etwas kaputt gegangen. [Peter has the other day his old radio repaired. CONTR-at-the amplifier had something broken.] (The other day, Peter has repaired his old radio. The amplifier was broken.)

The referent of the phrase amplifier is established via accommodation. I regard this use of the definite description as small-world, because the referent has a certain unique function in the situation of the utterance and the identification of the referent is only possible in a restricted domain. In our example, the domain is restricted in the sense that outside of the situation of utterance the referent of the definite description does not perform a unique function. 26

"Due to their general knowledge, inhabitants of a village can talk about the district council, the church, the pastor, the secondary school, etc., member of a family can talk about the father, the mother, the grandmother, the granny, etc." (my translation)

47

In Section 3 I gave an example similar to (22), repeated here for convenience, in which the un-contracted form was preferred.

(20) Fritz hat gestern eine Rezension über ein interessantes Buch gelesen, das er sich heute kaufen wollte. Von dem / Vom Titel hatte er sich allerdings nur den ersten Buchstaben gemerkt. (Yesterday, Fritz read a review about an interesting book that he wanted to buy today. Unfortunately, he could only remember the first letter of the / CONTR-of-the title.)

Vom Titel (CONTR-of-the title) in (22) can be regarded as an instance of the bridging-anaphoric use of small-world nominals. The referent of the nominal is established via accommodation and the contracted form is required. The uncontracted form, on the other hand, indicates that the nominal is used anaphorically, because anaphoric NOMs do not allow contracted forms. I want to suggest that the apparent indifference with respect to the contracted or the un-contracted form in sentences like (22) arises from an ambiguity of the interpretation of the respective nominal: The nominal being used in a bridging anaphor on the one hand and the indication of some other kind of anaphoric use (i.e., contextual use in our sense, cf. Section 4.2.2.2) on the other.

There is a last set of examples that should also be considered as instances of the small-world use; I am not really sure, though, whether these examples can be subsumed under either of the two different kinds of small-world nominals that I introduced earlier. Maybe we should regard them as yet another kind of small-world NOMs, viz. as nominals that imply a set of alternatives. Above I stressed the importance of the occasion of an utterance for the determination of the use of a particular nominal. But certain syntactic characteristics also have to be taken into account. Ordinals, superlatives, and other prenominal adjectives that single out a particular object from a set of possible alternatives (like next, previous, or last) can be used to indicate that a definite description is used specifically in our sense.

48

(23) im ersten Moment, beim nächsten Mal27 [CONTR-in-the first moment, CONTR-at-the next time] (in the first moment, next time) (24) im letzten Sommer, am dritten Tag, zur nächsten Woche, beim letzten Treffen, am höchsten Berg [CONTR-in-the last summer, CONTR-on-the third day, CONTR-to-the next week, CONTR-at-the last meeting, CONTR-at-the highest mountain] (last summer, on the third day, until next week, at the last meeting, at the highest mountain) (25) Im letzten Monat hat Paula eine Telefonrechnung von 500 Euro gehabt. (CONTR-in-the previous month, Paula's phone bill was 500 Euros)

I would consider the nominals in (23) and (24) as belonging to the small-world category because they refer to objects that have a unique function in the situations at hand in the sense that there can always be only one first moment or one last meeting. In sentences like (25) the definite description is obviously used specifically and I want to claim that the object referred to also has a unique function in the context of the utterance. Furthermore, the nominal seems to refer to a particular object in a set of possible alternatives. It was not this month or two months ago that Paula's phone bill was so high, but it was last month. The same observations can be made in the following sentences.

(27) Anna asks Fritz whether he enjoyed Felix' wedding. Fritz replies: Am Abend war mir langweilig. (I was bored CONTR-in-the evening.) (28) Intrusion Detection am Beispiel von Snort.28 [Intrusion Detection CONTR-at-the example of Snort.] (Intrusion Detection exemplified by Snort.) 27

It appears to be the case that the nouns in (23) can only occur with adjectives of the kind mentioned. An apparent exception can be found in the following sentence. (26) Ich weiß im Moment auch nicht weiter. [I know CONTR-in-the moment not further.] (I am at a loss right now.)

I suggest that im Moment is an idiom, since the phrase can be used interchangeably with jetzt gerade (right now). 28 from http://www.pro-linux.de/work/snort/print/index.html

49

(29) Am Kopf des Toten haben wir eine Schädelfraktur festgestellt. (We found a skull fracture CONTR-on-the head of the body.)

Fritz' answer in (27) seems to imply that he was bored in the evening, but that he enjoyed the rest of the wedding day. In (28) am Beispiel (roughly: exemplified by) indicates that intrusion detection could just as well have been explained with the help of Network Flight Recorder, but that Snort was chosen instead. The person who utters (29) implies that the rest of the body was also examined, but that he is focusing on the head at the moment. To be honest, I have no idea why these implications can arise if a small-world definite description is used, but I think it is an interesting observation nonetheless.

The small-world use of a definite description requires the contracted forms. In this use the definite descriptions function referentially, in the sense that a particular object is referred to. We can distinguish between different kinds of small-world nominals: Local names and bridging anaphors. The third class consists of nominals that imply that the referent of the NOM is a particular object in a set of possible alternatives.

4.2.2.2 The Contextual Use If a nominal is used contextually, it cannot occur with contractions. Endophoric, anaphoric and deictic uses of definite descriptions belong to the contextual use. The identification of the right referent of these nominals is successful if the audience is able to pick out the right referent of the respective NOMs. The description of the NP used does not have to fit the referent, neither does it indicate a certain unique function or role of the referent (as in the small-world use of definite descriptions). Information additional to the purely descriptive content of a nominal sometimes have to be taken into consideration in order to make the identification possible. This has become obvious during the discussion of epithets above. Epithets are clearly instances of the contextual use of definite descriptions, they refer to a specific object and they are used anaphorically.

Occasionally, nominalised verbs can apparently be preceded by un-contracted forms.

50

(30) Anna's neighbours are renovating. Bei dem (*Beim) Geklopfe kann Anna nicht schlafen. [At the (CONTR-at-the) rapping can Anna not sleep.] (Anna cannot sleep with all the noise.)

Here, the phrase bei dem Geklopfe (roughly: with all the noise) refers to a particular object that is part of the utterance situation. In that sense we could regard (30) as a deictic use in Löbner's sense, and we could claim that deictic uses of definite descriptions require un-contracted forms, that is why the contracted form beim is not permissible in (30). A different interpretation seems more adequate, though. The word dem (the) in (30) is a demonstrative, not a definite determiner. The apparently un-contracted form is in fact a preposition-demonstrative combination and hence does not really belong to the subject matter of this paper. I gave this example nonetheless, because it might help to clarify the theory I am proposing.

Some definite descriptions that are used attributively in Donnellan's sense seem to evade a treatment in the account of contractions presented so far.

(31) In a news report about an unsolved murder. Am Tatort wurden Fingerabdrücke vom / von dem Mörder gefunden. (They found fingerprints CONTR-of-the / of the murderer at the crime scene.) (32) The host of a teetotalers' party is told that one of the guests is drinking a martini. Vom / Von dem Mann mit dem Martini werde ich eine öffentliche Entschuldigung fordern. (I will demand a public excuse CONTR-from-the / from the man with the martini.)

Attributively used definite descriptions do not belong to the category of the specific use in our sense, they should be classified as generalising uses. This classification explains why the contracted forms can be used in the above sentences. With respect to the un-contracted form I want to suggest an analysis along the lines of (30). If the sentences (31) and (32) are used anaphorically (i.e. if they re-

51

fer to a discourse referent that has been introduced earlier via accommodation)29 they are used contextually and hence do not allow the contracted forms.

(33) Der Mörder war verschwunden, aber am Tatort wurden Fingerabdrücke von dem (*vom) Mörder gefunden. (The murderer had disappeared, but they found fingerprints of the (*CONTR-of-the) murderer at the crime scene.)

So, in (31) and (32), the apparent indifference with respect to the contracted or the un-contracted form can be regarded as an interference of the generalised and the contextual use of the respective definite description. As soon as an expression is used anaphorically, endophorically, or deictically, it requires the un-contracted forms and belong to the category of the contextual use.

The un-contracted forms are obligatory if a definite description is used contextually, i.e. if it is used as an anaphoric, endophoric or deictic nominal. Even nominals that are used generalisingly require the un-contracted forms as soon as they are used anaphorically.

A few open questions remain if we follow the theory that I am proposing: The use of pronouns to refer to nominals that are used generalisingly (cf. Section 4.2.1), the reason why nominalised verbs seem to never be used specifically in our sense (cf. Section 4.2.1), and the mechanisms that underlie the implication of alternatives when certain small-world nominals are used (cf. Section 4.2.2.1). Nevertheless, the account presented in Section 4 seems to make correct predictions about the vast majority of the examples that we considered so far, especially about those cases that could not be handled satisfactorily in Löbner's framework.

29

The discourse referents can only be established via accommodation. In section 4.2.1 we saw that if a definite description is used generalisingly (as is the case with attributively used NPs), the respective phrase does not establish a discourse referent. We can use, e.g., pronouns to refer to instances of the kind indicated by the descriptive content, which can then be regarded as discourse referents. So, the discourse referents are introduced via accommodation.

52

5. Conclusion In this paper we have been concerned with preposition-article contractions in German. We saw that Lyon's (1999), Heim's (1991) and Löbner' (1985) accounts of contractions do not make the correct predictions. We discussed Löbner's proposal in detail and found out that his notion of semantic definites is not suitable for an adequate analysis of contractions. It is supposed to subsume too many different phenomena to be coherent. That is why I have been proposing a new account of contractions. Table 2 and 3 summarise the most important aspects of Löbner's and of my proposal, respectively.

Type of Definite Description

Characteristics

Semantic Definites

- refer unambiguously independent of the utterance situation - proper names, FC1s, FC2s (also with implicit arguments), configurations - contracted forms are possible

Pragmatic Definites

- depend on the utterance situation for unambiguous reference - includes anaphoric, endophoric and the "visible situation use" of deictic definites - require un-contracted forms

Table 2: Characteristics of semantic and pragmatic definites in Löbner's framework

53

Type of Use of Characteristics Nominal Generalised Use

- does not refer to a particular object - the identity of the referent is irrelevant, but the descriptive content of the nominal has to fit the referent - used in generic sentences or as generic nominals - requires contracted forms

Specific Use Small-world Use

- refers to a particular object - used as local names, in bridging anaphors, and sometimes implies a set of alternatives - can only be used in a locally restricted domain or in a particular community - requires the contracted forms

Contextual Use

- includes anaphoric, endophoric, and deictic nominals - can also include generalising nominals if they are used anaphorically - requires the un-contracted forms

Table 3: Different types of uses of nominals in the proposal presented in Section 4.

In contrast to Löbner, the distinction between the two main categories, i.e., the generalised and the specific use of nominals, is based on the kind of reference that is expressed by the nominals, and not on the (in-) dependence on the context of an utterance for the unambiguous reference of a definite description. NOMs that are used specifically refer to a particular object, those that are used generalisingly do not. Such a distinction seems to be suited much better for an analysis of contractions than Löbner's distinction between semantic and pragmatic definites. The subdivision of the specific use into the small-world use and the contextual use of a nominal is useful for making the correct predictions. With the help of these two kinds of specifically used nominals and the notion of the generalised use we are able to give an adequate account of contractions, which is not possible in Löbner's framework.

It still remains unclear why nominalised verbs can only be used generalisingly, why reference to a generalising nominal with the help of pronouns is not always

54

possible, and why certain nominals that belong to the small-world category seem to imply a set of alternatives. Nonetheless, the proposal presented in Section 4 of this paper seems to make the correct predictions for the use of contracted and un-contracted forms in the vast majority of examples.

55

6. References

Bosch, Peter, 1995: Definiteness & Dynamic Knowledge Representation. Lecture slides from: Workshop Anaphora and Reference, Nancy. Carlson, Gregory N. & Pelletier, Francis J. (eds.), 1995: The Generic Book. University of Chicago Press, Chicago. Donnellan, Keith, 1966: Reference and Definite Descriptions. Reprinted in: Martinich, Aloysius P. (ed.), 2001: The Philosophy of Language. Oxford University Press, New York, 2001, pp. 247-259. Haberland, Hartmut, 1985: Zum Problem der Verschmelzung von Präposition und bestimmtem Artikel im Deutschen. In: Osnabrücker Beiträge zur Sprachtheorie 30: 82-106. Hartmann, Dietrich, 1980: Über Verschmelzungen von Präposition und bestimmtem Artikel. In: Zeitschrift für Dialektologie und Linguistik 47: 160183. Heim, Irene, 1991: Artikel und Definitheit (Articles and Definiteness). In: A. v. Stechow, 1991, pp. 487-535. Heim, Irene & Kratzer, Angelika, 1997: Semantics in Generative Grammar. An Introduction. Blackwell Publishers, Oxford. Krifka, Manfred, F.J. Pelletier, G.N. Carlson, A. ter Meulen, G. Chierchia, & G. Link, 1995: Genericity: An Introduction. In: Gregory N. Carlson, 1995, pp. 1-124. Löbner, Sebastian, 1985: Definites. Journal of Semantics 4: 279-326. Lyons, Christopher, 1999: Definiteness. Cambridge University Press, Cambridge, pp. 326-330. Stechow, Arnim von & Wunderlich, Dieter (eds.), 1991: Semantik/Semantics. Ein internationales Handbuch der zeitgenössischen Forschung. An international Handbook of Contemporary Research. Walter de Gruyter, Berlin, New York.

56