Preprint -

Laboratorio linguistico del Dipartimento di italianistica

Emanuela Cresti

Speech act units and informational units

1995 - Università di Firenze

Il laboratorio linguistico rende disponibili alla lettura esclusivamente i risultati delle ricerche e i corpora in esso elaborati sotto forma di preprint, reprint e pubblicazioni in proprio. E' vietata la vendita e la riproduzione non autorizzata, anche parziale, per qualsiasi uso. [2] Pubblicato in E. Fava (ed.), Speech Acts and Linguistic Research, proceedings of the Workshop, July 15-17, 1994, Center for Cognitive Science , State University of New York at Buffalo, Nemo, Padova 1995

For correspondence Emanuela Cresti Università degli studi di Firenze Dipartimento di italianistica Piazza Brunelleschi, 4 50121 FIRENZE E-mail ELICRESTI@ cesit1.unifi.it

Emanuela Cresti

SPEECH ACT UNITS AND INFORMATIONAL UNITS

1. The subject of my paper regards speech act units, about which we can raise some preliminary questions, such as : Why do we speak of speech act units? What do we mean by "speech act unit"? What does it correspond to? Is a speech act composed of many units, and, if so, what is their nature? Since the answer to each question is not obvious, I shall try to answer by describing how my research on speech act units started as grammar research on spoken language, and specifically on spoken Italian. Observations of recorded spontaneous conversations, but also radio and movie dialogues, and adult-child (0-36 months) conversations, led me to take into main consideration the role of intonation. Spoken language is performed by phonetic sounds, but also by their suprasegmental features (fundamental frequency (F0), intensity, lenght). After we were able to visualize and measure F0 variations, which is generally believed to be the main physical parameter of intonation, through the device of pitch meters 1, the main question was to distinguish the function of intonation and its regularity with regard to any linguistic level (syntax, semantics, pragmatics, phonology and information). Many answers presented themselves, but the problem, far from being fully solved, still remains open. However, I have my own answer, which can be summarized as follows: intonation functions as a signalizer of speech acts and of its components in the flow of speech. In my research I have shown that there is a very general correspondance between a speech act and an intonation pattern. But what is a speech act and what is an intonation pattern?

2. Speech acts and patterns of intonation My operative definition of a speech act derives from Austin, i.e., a complex act simultaneously performed as a locutionary act, an illocutionary act and a perlocutionary act. The complex act includes both primitive and not-primitive speech acts, among which, from a very general point of view, no difference can be made, even if a great linguistic difference can be found among them. For instance, a primitive speech act can correspond only to a noun phrase or even to a noun (Help!), but this is not the case for non-primitive speech act (Let us go to the cinema). My definition of intonation pattern is less immediate, conceived as it is in a framework of formal and perceptual studies of intonation, which have a point of reference in the work of the Eindhoven team on Dutch and English intonation 2, that of Philippe Martin on those of French and English 3, and in my own research on that of Italian 4. Roughly speaking, our hypothesis can be expressed as follows: an intonation pattern is a model with many possible variants, and its type is determined by one of its possible elements. A pattern can be composed of one or more tonal units, formally and typologically differentiated, but only one (comment unit) is necessary and sufficient to form the pattern itself. So a pattern can be established with only the necessary tonal unit, comment unit, or with that in addition to many other optional tonal units, typologically differentiated: Pattern A = (a), (b),...A,...(c),(d); Pattern B = (a), (b),...B,...(c), (d)... The correspondance between a speech act and an intonation pattern can be explained in terms of performance; a speech act must be performed not only as a phonetic sequence but necessarely as a phonetic sequence patterned by intonation. Every utterance is melodically organized. A speech act cannot be established without intonation and, in case of monotone performance, it is not pragmatically interpretable 5. Because of this correspondance we contemplate the possibility that a speech act can also be composed, at least from an intonational point of view, of many tonal units belonging to one pattern.

3. Modality and intonation 3.1. But we may ask why a speech act must be performed by an intonation pattern. Since antiquity, grammarians have observed that there is a correspondance between the melodic performance of an utterance and what has been called in many different ways, and what we shall call for the moment, "modality" 6. Types of sentences, employment of verbal moods, kinds of rhetorical periods, expression of emotions, which are generally included under the heading of "modality", often display a specific melodic "packing". This assumption is very easy to verify in a language like Italian, where the very same lexical and morphological expression can express different "modalities" only through different intonational performances. Let us see some examples with their respective variations of F0: 1) Piove. F= assesment (It's raining)

2) Mangia. F=assesment (S/he is eating)

2a) Mangia? F=question (Does s/he eat?)

2b) Mangia! F=order (Eat)

1a) Piove? F= question (Does it rain?)

If we analyze the data, many facts of an intonational nature may be observed: a) there are systematic variations of F0, whose form appears perceptually relevant and which are repeated (1=2; 1a=2a). As we have stated, they are the main characteristic which the different melody depends on and they seem to be conventionally codified; b) systematic variations of F0 are formally codified as the same melody, even if they perform different words ( 1=2, but piove vs mangia), and even if they present strong differences of micromelody, showing to be phonetically independent; c) every codified intonation is opposed to all the others (1### 1a,2a,2b; 1a ### 2,2b), mantaining its "modal meaning". 3. 2. But it may also be noted that the traditional term of modality, which, as we have said, includes among others relevant aspects of language such as the expression of emotions, as concerns types of sentences and employment of verbal moods, partially clashes with an illocutionary classification 7. Let us examine our examples: 1) is an assesment as 2), 1a) is a question as 2a), 2b) is an order. So, in very simple linguistic examples there is a one-to-one correspondance between melody or intonation (conventionally codified F0 variations) of a simple locutive expression, only one word, and illocution (assesment, question, order). It is also evident that the relation between intonation and force is independent from locutive contents, that is, different lexical entries (piove, mangia) with the same kind of intonational variation express the same illocutionary force, and the same lexical entry piove with two different conventionally codified kinds of intonation expresses two different illocutions (piove vs piove?). Now we must generalize this point by observing that every linguistic expression which is significant can be the locutive content of a speech act if it is performed through the device of a suitable intonation. Let us see: 0) Una rosa // F=assesment (A rose)

0a) Una rosa? // F=question (A rose)

0b) Una rosa // F=order (A rose)

Our examples are respectively 0) a primitive speech act with the force of an assesment or an answer, 0a) with the force of a question, and 0b) with the force of an order, as in the case of 2), 2a), and 2b), even if no predicate or sentence is realized. In conclusion, conventional intonational forms have illocutionary meanings, or rather, illocutionary functions, because different intonations succeed in transforming the same locutive content into different speech acts 8. 4. Utterances longer than one word The systematic relation between intonation and force holds also if we consider a longer utterance, such as a locution composed not only of one word but of several words. Let us see: 3) Mangia un panino. F=asses. (S/he is eating a sandwich)

3a) Mangia un panino? F=question (Is s/he eating a sandwich?)

3b) Mangia un panino! F=order (Eat a sandwich!)

From a viewpoint of perceptual relevance, the form of intonation 9 of 3) is the same as 1) and 2), that of 3a) is the same of as 2a), and that of 3b) is the same as 2b). Then 3) expresses assesment force as 1) and 2), 3a) expresses question force as 1a) and 2a), 3b) expresses order force as 2b). In all our examples the intonation is characterized by only one relevant perceptual variation of F0, both over one word and over two words, so that only one intonational unit is realized. This can happen also with many words which can be "linearised" in the same tonal unit 10, expressing all together a single illocutionary force, as for instance: 4) E' capo assistente di Medicina. F=asses. (S/he is chief assistent of Medical departement) 5) Apri meno la finestra! F=order (Open less the window!) 6) Ha preso il treno delle sette? F=question (Did s/he take the seven o'clock train?)

Generally speaking, only one intonational unit can be performed as well over several words. From our examples we can verify that a single intonational unit is typologically the same, whether this was performed over one word or over more words ( 1=3=4; 2=2a=3a=6;2b=3b=5). In our examples the unit is the same and it has the same function, that of expressing a commom illocution for all the words, which are linearised in that tonal unit. We call this particular kind of tonal unit comment. In conclusion, a locutive content of one or more words can be performed as a speech act by the device of conventionally codified intonation. The same locutive content can

function as many different speech acts having different illocutionary forces, according to different intonational performances.

5. Another intonational unit 5. 1. But we must ask what happens when the utterance is longer, as very often it is 11. Literature has always shown that a peculiar aspect of spoken language is that of so-called segmentation. Spoken "style" very often is characterized by topicalization, left-and-right dislocations, birematic nominal sentences, cleft and pseudo-cleft sentences, repetition, kinds of focusing and emphasis, insistence, which reach a bipartition between neutral and stressed parts of an utterance, and so on. Let us look at some Italian examples: 7) Il giardino / pensa a tutto la nonna //COM F=answer (The garden, 0 takes care of everything grandmother) 8) Il caffè / lo voglio bello forte //COM F=asses. (The coffee, (I) it want good and strong) 9) Lo voglio bello forte /COM il caffé // F=asses. ( (I) it want good and strong, the coffee) 10) Da domani / dieta //COM F=promise (From tomorrow, (on) diet) 11) E' Carlo /COM che è venuto da noi // F=contrast ( (It) is Charles, that has come to (visit) us) 12) Il calcio / il calcio non mi piace //COM F=judgement (Soccer, soccer does not (to ) me appeal)

Bally proposed that the basic structure of spoken language is a theme-rheme pattern, see 14), which according to him is opposed to the bound structure of written language (canonic sentence with syntactic relations), see 13): 13) Sono contento perchè Carlo ha vinto. (I'm glad because Charles won) canonic sentence, syntactically bound. 14) Carlo ha vinto/ sono contento// ( Charles won, I'm glad) spoken structure theme-rheme.

Bally was also the first to notice that there is a correspondance between this segmented structure of theme-rheme and a specific melodic performance 12, as we can see from the following F0 contour corresponding to 14):

5. 2. Actually, if we consider only some kinds of segmentations, those which imply a so-called left segmentation, such as topics 7), left dislocations 8), nominal birhematic sentences 10), in some way all kinds of theme-rheme structures, we may notice that their segmented structure is systematically interpreted by intonation and systematically in the same way. 7')

8')

10')

The left-segmented locutive part - theme in Bally's terms - is performed by a tonal unit with a specific form, which, generally speaking, we can denote as "not concluded" 13. On

the contrary, the following locutive part, rheme in Bally's terms, is performed by one of the possible tonal units with an illocutionary function. As we saw before, it can be for instance one comment tonal unit of assesment, or question, or order, and so on. Because of the general character of syntactic structures involved in the thematic part, whose main aspect concerns topicalization, we call the correspondent tonal unit topic. Both the segmented locutive parts of the utterance are recognized and interpreted by intonation, which does not operate only a parsing but assigns distinct, conventionally codified intonations to each part, which corresponds to a specific tonal unit: the first one a topic unit (on the left) and the second one a comment unit (on the right). The topic-comment pattern can be considered one of the most common intonational patterns of spoken language; it exhibits the first and most relevant partition of a speech act, which can correspond to two different locutive parts, Ballyan theme and rheme, with their specific intonational performances. 6. The topic is optional 6.1. If the second tonal unit, a comment tonal unit, has in a systematic way an illocutionary function, we may note that the topic too has a "systematic meaning", which for the moment we shall indicate as a thematic one. But we would like to know more about the function of this first intonational unit of a segmented spoken utterance. Before answering, we must put forward some observations regarding the fact that the theme-rheme structure can generate a variant and be realized only through its rheme. From a textual or linguistic point of view, indeed, a theme may always be lacking and not be realized. So we are in the situation in which a theme-rheme structure can be elliptical concerning its theme, while its rheme can never be lacking and must on the contrary always be realized. Let us observe our examples with topicalization and left dislocation without their thematic part: 7a) ... / pensa a tutto la nonna //COM F=answer ( .... 0 takes care of everything grandmother) 8a) ....../ lo voglio bello forte //COM F=asses. ( .....(I) it want good and strong) 10a) ..../ dieta //COM F=promise (... (on) diet)

The versions a) of our examples result as still pragmatically interpretable utterances, and what is missing in their overall meaning, i.e., their theme, appears to be integrable through reference to a pragmatical context. On the contrary, if we consider the same examples, where it is the rhematic part that is elliptical, versions b), we may observe that they look like text fragments, or achieve an effect of suspension; at any rate, they are no longer pragmatically interpretable. 7b) Il giardino /TOP .....// F=?? (The garden /.... ) 8b) Il caffè /TOP.... // F=?? (The coffee/.... ) 10b) Da domani /TOP... // F=?? (From tomorrow/... )

At this point we can conclude on the one hand that the rhematic part, independently of its locutive content and performed as a comment unit, is sufficient to function as a speech act; while on the other hand we must consider that any "thematic" part, performed as a topic, is optional and at the same time cannot be pragmatically interpreted and cannot function by itself as a speech act. 6.2. Many different attempts have been made to explain the optionality of theme (wellknown in literature), the most common of them being that lexical and/or syntactic, or informational reasons are responsible for this phenomenon, but this is not the case, as we illustrate in the following paragraphs. On the contrary it has never been considered that a theme clashes with an intonational performance as topic and actually this performance seems highly relevant. Our question, indeed, is whether in reality what is elliptical is not the locutive function of the theme but, on the contrary, its being the topic. This would imply that topic and comment not only depend on intonational performance but are specific functions of spoken language. A rheme performed as a comment, indeed, is the locutive part which is used to express the illocution of the act: no speech act exist without a force and no locutive content can be pragmatically interpreted if a force has not been established for it. So the rheme, because of its behaving as comment and only because of that, must on the one hand be realized, and on the other it can be interpreted, in practical terms, even in isolation. We know, indeed, that every significant expression can operate as a speech act if properly performed as a comment, even a simple noun; see for instance 0), 0a), 0b), and is also susceptible of pragmatical interpretation.

This means that if an utterance can be segmented into a thematic and a rematic part, whose performance corresponds to a topic-comment intonation pattern, the rheme, clashing with the comment unit expressing an illocution, must be realized anyway, otherwise the speech act is not established. The comment unit indeed is not only sufficient, as version a) of our examples show, but also necessary for the establishment of a speech act, as we can see from version b) of our examples. So a very common spoken utterance with a segmented structure, as theme-rheme, if it is performed as a topic-comment pattern, is composed of two parts: one, the topic, optional and not interpretable in isolation, and whose corresponding locutive part is the theme; and the other, the comment, necessary and sufficient for constituting a speech act, interpretable in isolation, and whose corresponding locutive part is the rheme. But what about the optionality of theme? 7. The optionality of theme does not depend on syntax The most common explication of the optionality of theme is that the rhematic part clashes with a syntactic structure, such as a predicate or a sentence, and the thematic part clashes with a nominal or adverbial phrase, and because of this syntactic characteristic it would be impossible to omit the rheme and possibly also the theme. But even if this syntactic distribution is quite common, even from our few examples it is evident that things happen in another way. Especially concerning so-called nominal sentences, we may observe that they cannot depend on syntactic constraints 14, because both the theme and the rheme are nominal or adverbial phrases and no predicate is performed; but they can equally be realized in an elliptical version, as only one nominal rheme. For instance, let us examine 10a) and 10b), where the principle of the optionality of the theme is valid, but evidently cannot be explained in syntactic terms, because the rheme too is a nominal or an adverbial phrase. Moreover, examples as 14a) and 14b), where both the theme and the rheme are sentences, show that this principle is still operative and cannot depend on syntax: 14a) ....../ sono contento //COM F= asses.(.... (I) am glad) 14b) Carlo ha vinto / ..... //COM F=?? ( Charles won, ......//)

In 14b), although the theme is a sentence, naturally if the topic intonation is maintained for this part, we get an impression of suspension, which prevents the listener from giving a "practical interpretation" to the utterance, lacking an illocution. There are also cases in which the theme is a sentence and a rheme a nominal phrase. Let us examine 15): 15) Leggi solo fumetti /TOP pigrone! //COM F=evaluation ((You) only read comic books, lazybones!)

In this case as well, if the intonation of 15b) is maintened as that of a topic, we get the same effect of suspension, and even if it is a sentence, it is not interpretable; while 15a), mantaining its comment intonation, remains pragmatically interpretable, even if it is an adjective. 15a) ......./ pigrone! //COM F=evaluation (..... lazybones!) 15b) Leggi solo fumetti /TOP ..... // F=?? ((You) only read comic books ...)

So the necessity of rheme and the optionality of theme cannot be explained by syntactic constraint, as the syntactic variety of our examples shows; it appears instead to depend on this effect of suspension. 8. The optionality of theme does not depend on lexical features It might be hypothecized that the optionality of theme depends on lexical features or the semantic meaning of the expression involved, as for instance individual topic features and, possibly, comment features, but we have done some experiments, by which we can show that the semantic meaning, generally speaking, is not relevant in order that an expression function as a topic or as a comment. For instance, if we substitute the occurrence of a rhematic part with that of a thematic part, naturally performing the rhematic part with a topic intonation and the thematic part

with a comment intonation, what we obtain is still an interpretable utterance. But what is more interesting is that the optionality of the topic still works. If we consider an example as 16), a nominal birhematic sentence, whose expression in topic is the same which was the comment in 10) (Da domani/TOP dieta//COM) and that in comment was the respective topic, we can constate that the sufficiency of the comment and the optionality of the topic are still valid, despite the inversion of informational functions for every lexical expressions. 16) Dieta /TOP da domani //COM F=ironic promise ((On) diet, from tomorrow)

16a) ..../ da domani //COM F=generic promise (... from tomorrow) 16b) Dieta/TOP.....//F=?? ( (On) diet ....)

If utterances coinceived to express a specific content are reversed in their theme-rheme structure, they obtain obviously a different semantic result, and from an informational point of view we may find them very stressed. But this specific informational result does not prevent the possibility of changing the roles of both the thematic and the rematic part. As we saw in paragraph 3., the expression of illocution is independent of locutive content, but we claim that the locutive content of an expression does not influence its possibility to behave as a topic too. So the informational functions of comment and topic seem to be on one side constant and on the other independent of syntactic and lexical features, i. e., from their locutive content. 9. The optionality of topic cannot be explained by its given informational feature 9. 1. We have not mentioned till now, among the classic literature on spoken language, the work of Halliday, whose research has indeed been the most important in this field. Halliday systematically studied the relation between intonation and information, espe-

cially asserting the existence of specific spoken units, the information units, as different from sentences or from any kind of written or ideal linguistic entity. His idea of information, which is the most accepted, is based on an principle which implies that information normally progresses from given to new "data". Halliday, though he offers his own general and original framework, agrees with Bally's theme-rheme structure, explaining that structure in conformity with his theory and stating that from an in formational point of view the rhematic part should be new, as opposed to the thematic part which should be given. Likewise, the optionality of theme is conceived as a result of the informational principle self; it would be necessary, indeed, that every message express some new information and on the contrary that it can omit given information. But even if this very general assumption can be shared, problems arise anyway in the search for its linguistic correspondance; there have, indeed, been many studies 15 on this subject which have shown that the informational status of the thematic and rhematic parts, from a point of view of given vs new, changes from one time to the next. Very briefly we may say: a) the thematic part can correspond to a given bit of information, but in most cases it corresponds to the introduction of a new topic in the conversation; b) very often the rhematic part corresponds to new information, but sometimes it is a repetition of something already said, or anyway it is something which we cannot define as being not previously known. The whole matter seems to be strictly motivated by the speaker's emotion, which gives subjective values to objective data, in such a way that something which has not yet been stated in the situation, but which has long resided in the speaker's thoughts, is stated as a theme and "functions" as topic, while something absolutely obvious, already stated, but which has a special meaning for the speaker, as for instance a confirmation or a kind of contrast with the listener, is stated as a rheme and "functions" as comment, conveying a particular illocution. So it is clear that an informational principle based on an objective progression from given to new data, if generally speaking it is acceptable, does not suffice to explain what is going on in spoken utterance, where the criteria for defining what is given and what is new cannot be objective. Halliday too explains that we have not to interpret the principle in a mere objective way, but he does not give us a different key to decide. Because we cannot enter into a speaker's mind, we must look for a different informational principle, existing just the same as a general priciple.

9.2. Very briefly if we analyze the theme-rheme structure independently of any consideration, what persists is that the simple tonal performance of any locutive expression as a topic or as a comment, makes it behave as a theme or as a rheme, and that the expression performed as a comment, behaving as a rheme, must be stated, while that performed as a topic, behaving as a theme, may be omitted. The informational principle, based on the the necessity of the "new" and on the normal progression from the given to the new, can be easily explained considering that only what expresses the illocution is necessary but also new 16. The linguistic expression which is performed and functions as the comment, indeed, is necessarely new, independently from its locutive (lexical and syntactical) content, because the accomplishment of an illocutory act changes the world and makes the expression too become new. Only the tranformation of the world is in effect new, and this is just what a comment realizes accomplishing an act. So the fact that an expression is performed as a comment does effect for it to be the rheme and to be new at the same time. Therefore, the tonal performance has a one-to-one correspondance with an informational level which is independent of syntax and lexicon and cannot be identified with the traditional given-new principle either. For the moment we assume that the performance of a topic-comment intonation pattern corresponds to the expression of illocution by the comment and the "non-expression" of illocution by the topic. The path of our informational principle could be defined pragmatical, as proceeding from a non-illocutionary part, topic, to an illocutionary part, comment. This path is basically different from the traditional given-new one, because it is founded on the differentiation between what linguistically changes the pragmatical context, and in that sense is new, and what does not 17 . 10. Topic functions as the field of application of a comment 10.1. But we must ask what the positive function of the topic is. Let us examine a new time some examples as 7) (Il giardino /TOP pensa a tutto la nonna //COM), 8a) (Il caffè TOP / lo voglio bello forte //COM ), and 10) (Da domani /TOP dieta //COM ), where the locutive parts in topic have no real syntactic link with the respective locutive part in comment. Topics in 7) and 10) can be defined as anacolutha, and in 8) an anaphoric element in the comment, which is the object of the verb, repeats the topic, but the nominal phrase in topic is only a topic!

Nor can our examples be assimilated for semantic reasons; all that the topics share with one another is an informational relation with their respective comments, which we can try to translate as "concerning": concerning the garden or concerning the coffee, something else is assessed, concerning tomorrow a certain promise will be available, and so on. If we observe only the correspective comment unit, we know that the utterance remains interpretable, showing its illocutionary force, and its interpretation can be filled with pragmatical information: a pragmatical prominence in the physical or mental world. 7c) ...../ pensa a tutto la nonna //COM F=answer (.....0 takes care of everything grandmother) concerning something in the situation 8c)...../ lo voglio bello forte //COM F=asses. ((I) it want good and strong) concerning something in the situation) 10c)...../ dieta //COM F=promise ((on) diet) valid here and now

For instance, one of the most common cases the interpretation of which is completed with the pragmatical situation, is an assertive act of evaluation, which very often is performed as a primitive act by only one adjective; in that case its reference to the pragmatical context is evident. 18) Wonderful! //COM F=evaluation (referring to the sunset)

But some different cases are also possible. An answer, for instance, is a locutive content performed by a specific comment intonational unit, in such a way that a listener can interpret it as an answer. But s/he can do so also because in the pragmatical context there is a question, to which it is possible to make reference. 19) A: Who has left?//COM F= question B: Charles//COM F= answer

In 19) Charles, a comment of answer, refers to Who has left?, which functions as a linguistic object, a linguistic prominence in the situation which the comment refers to. So all the examples take their reference in a pragmatical prominence, and in 10c) it is the tense of the speech act, which offers the space-time localization for the utterance, "starting now/ on a diet".

10.2. Performing a speech act implies establishing an illocutionary act simultaneously with a locutionary and a perlocutionary one; moreover, performing an illocutionary act implies referring to something, applying the force to something and in some way. Very often it can be realized through a kind of deixis in the pragmatical context; but deixis, as is known, is underdeterminated, and very often it does not appear clearly what a force is addressed to. But often it happens, and this is more relevant, that locutive contents concern things or events which are not present, or times and places different from here and now, as expression of human thought, and in that case the expression of reference can become highly relevant. So a speaker may need or simply choose to state "something before" its comment, in order to address its illocutionary force in a clear and unambiguous way, stating to the listener the pragmatical or mental prominence of her/his illocution, before and off the force self 18. Therefore we can claim that a topic functions from an informational point of view as a linguistic field of application of the force established by the comment; so it can be referential either in a pragmatical or in a mental way, it can be either given or new from an objective point of view, but what is relevant is that it is the prominence chosen by the speaker to complete the information of the comment, making fully interpretable the utterance only through a linguistic device. Thus the informational function of the topic can be defined as the linguistic field of application of a comment or its linguistic reference. Defining the function of topic in terms of "field of application of the force" is independent of the speaker's interpretation of pieces of information as new or given, even if sometimes its locutive content can be given from a cognitive point of view. The function of the topic depends only on the accomplishment of an illocution, which must be realized in order for a speech act to exist, and which necessarely requires a reference; if the speaker "chooses" to produce a linguistic reference for his comment, then s/he "makes" a topic. So, following our pragmatical point, we claim that speech acts can be realized as spoken utterance very often in the form of an informational pattern of a topic-comment informational units, only the comment of which expresses the illocution, while the topic expresses its linguistic reference, allowing a full and clear interpretation of the utterance. 11. The presence of a topic is not motivated by quantity constraints but by informational reasons.

11.1. We started our reflection about topic by saying that very often our utterance can be longer than eleven syllables, which in Italian for instance is a kind of ideal measure for tonal unit, though some may still feel that topic depends merely on quantity limits. There are, indeed, some theories, which tend to explain the function and rules of intonation in terms of simple parsing of the flow of speech. Intonation should group only a few words at a time, giving a stress structure to each group, without any specific, relevant linguistic function, but just in order to facilitate the performance of utterances and their comprehension by the listener. Just as we showed that optionality of the topic does not depend on syntactic or lexical constraints, now we must verify that it does not happen because of mere quantity reasons. Even in some of our examples it may be verified that topic-comment pattern is realized under the limit of eleven syllables, as 10) or 15). In spoken corpora short speech acts are very common; but also speech acts patterned as topic-comment, shorter than eleven syllables, can be easily found in lively, spontaneous dialogues in Italian. But there is a more cogent reason than the simple verification that quantity limits at least are not necessary to allowing the performance of a topic-comment pattern; that happens when the topic-comment pattern is nedeed to assure the "grammaticality" of the utterance. This is the case with nominal sentences, which we have already introduced and which are so common in spoken corpora. Topic is optional, but, as we have said, sometimes it allows a full linguistic interpretation for a comment; but its relevance shifts considerably, depending on the locutive characteristics of the comment self. If by chance the comment is "filled" by a sentence, it is already fully interpretable by linguistic devices, and the topic can function only as a reduction and to add greater accuracy to its reference. Let us see 19) and 19a): 19) Carlo ha vinto la coppa //COM F=asses.(Charles won the trophy) 19a) Ieri /TOP Carlo ha vinto la coppa //COM F=asses. (Yesterday, Charles won the trophy)

If the comment is a predicate and implies any kind of verb, topic can be useful, but it remains not strictly necessary, because any verb allows the circumscribing of a pragmatical or mental reference through the device of temporal and personal morphemes, and very often through the presence of arguments derived from the semantic structure of the verb itself. Let us see 20):

20) Ha vinto la coppa //COM F=asses. ((S/he) won the trophy)

In Italian utterances such as 20), we may not know who the winner is, but we know for instance that s/he is neither myself nor you nor a generic plural group; and that the event is not occurring right now but has occurred some time in the past, and that the event is complex and comprehends a specific object (the trophy). On the contrary, if the comment is a nominal phrase, no hypothesis can be made about either any kind of argument or its time-space location, and if no topic helps us to acquire any information about other elements of reference, nothing can be known. For this reason nominal sentences need a topic to become pragmatically interpretable; otherwise they receive a generic interpretation which is considered valid for the speaker and the listener in a here-and-now situation, without any specification about the purpose of their telling. See 18) and the part of the answer of 19). Let us examine 21), 22), 23) as well: 21) Attenzione! //COM F=alert (attention!)

22) Profumo //COM F=evaluation (scent)

These are an alert and an evaluation valid here and now for the listener and expressed by the speaker, and their "topic" must be looked for in the situation. They are typical primitive speech acts; they can become much more informative if they are accompanied by a topic, even a simple noun or adverbial phrase. Let us examine 21a) and 22a): 21a) Un politico /TOP attenzione //COM F=alert (A political (man), attention)

22a) Gli aranci /TOP profumo //COM F=evaluation (the orange trees, scent)

If these nominal sentences become birhematic and performed as a topic-comment pattern, they also become interpretable without the aid of the pragmatical situation. 11. 2. But what is most interesting is that in spoken Italian it is impossible to distinguish a birematic nominal sentence in a continous flow of speech, performed as only one tonal unit of comment, we mean without realizing the typical topic-comment pattern, as it happens in 21b) and 22b): 21b) * Un politico attenzione! //COM F= alert (a political (man) attention!)

22b) * Gli aranci profumo//COM F=evaluation

(theorangetree,sscent)

We can verify that examples such as these have no quantity limits but must be performed through a pattern of topic-comment for locutive constraints, otherwise they reach a kind of agrammatical result. 11.3. We can only sketch for what concerns this question that relations among informational units, as for instance topic and comment, and relations among locutive expressions within any informational unit, as for instance in 22b) the orange trees and scent, are totally different. In 22b) two nominal phrases cannot be informationally related within the comment unit, but they had to be related trough a structure where one nominal phrase must be the semantic head of the other, as for instance it should be 23)19 : 23) A: Qual è il tuo più bel ricordo? //COM F= question (What is your best souvenir?) B: Il profumo degli aranci //COM F=answer (The scent of orange trees)

In 23) the scent is the semantic head of a nominal domain of identification, whose semantic specification is played by the prepositional phrase of orange trees and the entire domain is the locutive expression accomplished as a comment of evaluation. So we do not say that utterances such as 21b) and 22b) do not exist at all in spoken Italian, but first of all that they cannot be pragmatically interpreted in the same manner as 21a) and 22a). In conclusion, the topic-comment pattern is not only an intonational parsing of a long utterance, but an informational program chosen by the speaker in order to establish a speech act in such a way as to make it fully interpretable. So topic and comment are not only tonal units, but linguistic marks of specific informational functions: that of force and that of its field of application. We feel that our theory finally clarifies the reason why a theme is optional and a rheme is necessary, shifting optionality and necessity onto the topic and the comment, through

which theme and rheme very often are performed. In this framework, a comment functions as the establishment of the force of the specch act, and thus it is necessary and sufficient to constitute a speech act; on the contrary, a topic functions as the field of application of the force, and thus it can be more or less useful depending on the linguistic features of comment, but it always remains optional because it can be replaced by the pragmatical reference, and it is never interpretable by itself, even if it is a sentence, because a field of application of an action cannot be interpreted without this action's being established.

12. Others informational units 12. 1. The topic-comment pattern is the most relevant informational parsing of a speech act, but it is not the only one; studies of spoken corpora show that speech acts are much more complex than a pattern topic-comment and that there are many informational units with different functions. Very briefly, we may outline two different kinds of informational units beyond topic and comment, which are ruled by general illocutionary functions. One of these can be very quickly explained as a locutive integrator; by this we mean that the speaker can express a topic or a comment and after having expressed it, "feels" disssatisfied with her/his performance: because s/he made some mistakes, or changed her/his mind, or what has been said is partially wrong or inadequate, and so on. Then the speaker goes on to integrate only locutively her/his performance, adding linguistic "stuff" to the informational unit of topic or comment s/he had already made. This locutive integration is added through a specific tonal unit, already known in literature as a comma tonal unit, which can be defined as lower and less perceptually relevant than the tonal unit of topic or comment it follows. Many locutive integrations may follow a topic or a comment. We call this kind of informational unit appendix, which is ruled only by a function of locutive integration of an informational unit with an already realized illocutionary function. 24) Arcibaldo /TOP figlio di Carlo /APT va a Roma //COM F=asses. (Archibald, son of Charles, goes to Rome)

25) Carlo /TOP va a Roma /COM da Arcibaldo //APC F= asses. (Charles, goes to Rome, by Archibald)

12. 2. The other possibility is that concerning direct interventions of the speaker in her/his speech act in order to make it work better. We call these other informational units dialogical units. There are many different kinds of dialogical units with different functions, either due strictly to the progression of a dialogue (taking turns, ending, interrupting, confirming, contrasting, overturning, and so on), or with general communicative goals, such as request for attention, alert, support, explication (linguistic, metalinguistic, meta-illocutionary), allocution. These last are in reality kinds of comunicative actions, directed to the listener, in order to sustain a specific speech act; we can consider them as secondary illocutions subordinate to the general illocution of the speech act they are "helping". Every dialogic unit is performed through a tonal unit. These can acquire different melodic shapes, but what is relevant is that, on the one hand, every tonal unit has an accent and a relevant perceptual movement of F0, and, on the other hand, this kind of tonal unit cannot be interpreted in isolation and implies linkage to something which enables them to be interpreted. These informational units, generally speaking, have free occurrence, and can be repeated and inserted according to the momentary will of the speaker; in certain cases they can even interrupt a word. Obviously, a speech act can be very simple ( no!), or very complex and composed of many different informational units, all behaving as part of the same speech act, though the only informational unit which must be realized is one of comment, establishing the illocution of the act.

13. Conclusions We began our paper with some questions regarding, above all, the meaning of speech act units, asking if it were possible to imagine their existence. Our answer is positive and implies the consideration of a new linguistic level, that of information, and the definition of it in a new way. If we consider a speech act in its concrete, spoken utterance we cannot ignore its intonational performance, which allows us to individuate a tonal pattern, very often long and

complex, composed of many tonal units. Our point is that every tonal unit corresponds to an informational unit, whose function can be various. Intonation, in our opinion, is the systematic mark of information, whose fundamental principle is that of the establishment of illocution. The most simple variant of a speech act is that composed of only one comment unit, expressing the illocution of the act, which is necessary and sufficient to establish the act itself. A comment can be preceded by the linguistic field of application of its force, which is performed by an informational unit of topic. Topic is optional. Every topic and/or comment can be linguistically integrated by one or more informational units of appendix. This first informational patterning can be accompanied by one or more dialogical informational units, whose functions are various and whose occurrence is for the most part free. Our conclusion is that speech act units do exist and that their nature is informational, concerning a pragmatical definition of information. References A. Arnauld & C. Lancelot (1660), Grammaire générale et raisonnée, Port-Royal. L.J. Austin (1962), How to do things with words, Oxford University Press, Oxford. C. Bally (1963), Linguistica generale e linguistica francese, Il Saggiatore, Milano. G. Berruto (1985), Per una caratterizzazione del parlato:l'italiano parlato ha un'"altra" grammatica?, in Holtus & Radtke (ed.), Gesprochenes Italienisch in geschichte und gegenwart, Narr, Tubingen. D. L. Bolinger (1989), Intonation and its uses: Melody in grammar and discourse, Stanford University Press, Stanford E. Cresti (1987), L'articolazione dell'informazione nel parlato, in Accademia della Crusca, Gli italiani parlati, Firenze. E. Cresti (1991), Criteri informativi per la definizione dell'enunciato, in Accademia delle Scienze, Romance linguistics: Semantics and translation, Mosca. E. Cresti (1992a), La scansione del parlato e l'interpunzione, in E. Cresti, N. Maraschio, L. Toschi (a cura di), Storia e teoria dell'interpunzione. Atti del convegno internazionale sull' interpunzione, Bulzoni, Roma. E. Cresti (1993), L'ontogenesi della predicazione: dalle linearizzazioni alla formazione del predicato, in E. Cresti & M. Moneglia (a cura di) Ricerche sull'acquisizione dell'italiano, Bulzoni, Roma. E. Cresti (1994) Information and intonational patterning in Italian, in in Ferguson B., Gezundhajt H., Martin P. (ed.) Accent, intonation et modèles phonologiques, Editions Mélodie, Toronto. M. Fagioli (1972), Istinto di morte e conoscenza, Nuove Edizioni Romane, Roma. M. Fagioli (1974), La marionetta e il burattino, Nuove Edizioni Romane, Roma. M. Fagioli (1975), Psicoanalisi della nascita e castrazione umana, Nuove Edizione Romane, Roma. E. Fava (1984), Atti di domanda e strutture grammaticali in italiano, Libreria Universitaria Editrice, Verona.

E. Fava (1990), Interrogative or relative clauses?, in "Rivista linguistica", 2. E. Fava (forthcom.), Tipi di atti e tipi di frasi, in L. Renzi e G. Salvi (ed.), Grande grammatica italiana di consultazione, Il Mulino, Bologna. M.A.K. Halliday (1987), Sistema e funzione nel linguaggio, Il Mulino, Bologna. J. 't Hart, R. Collier, & A. Cohen (1990), A perceptual study of intonation, Cambridge university Press, Cambridge. Ph. Martin (1978), L'intonation de la phrase en italien, in "Studi di grammatica italiana, VIII, Accademia della Crusca, Firenze Ph. Philippe (1987), Prosodic and rythmic structures in French, in Linguistics, M. Moneglia (1994), The ontogenetic foundation of informational patterning, in Ferguson B., et alii, cit. M. Moneglia & E. Cresti (1993), Formazione dell'atto linguistico complesso e intonazione, in E. Cresti & M. Moneglia, cit. J. Pierrehumbert (1980), The phonology and phonetics of English intonation, M.I.T. Press, Cambridge, Mass. J. Searle (1969), Gli atti linguistici, Einaudi, Torino. G. Tamburini (1994), Tone units and local F phenomena, in Ferguson B., et alii, cit. 0 G. Tamburini (1995), L'ordine dei costituenti e l'articolazione dell'informazione, in Studi di grammatica italiana, Accademia della Crusca, vol XVI.

Footnotes 1

Pitch meters gained widespread use only in the mid-seventies. Begun in the seventies in the Netherlands by 'T Hart, Collier, Cohen and Noteboom. See t'Hart et alii (1992). 3 See Ph. Martin (1978, 87). 4 In the seventies I worked at the Laboratorio Fonetico of the Scuola Normale Superiore di Pisa and since the mid-eighties in the Laboratorio linguistico of the Italian Department of Florence University. 5 We shall see below what "interpretability" means. Any locutive expression cannot be pragmatically interpreted, if it is not performed in a way that an illocution can be assigned to it and that is primary realized trough the intonation. 6 See for instance Arnauld & Lancelot (1660). 7 This correspondance is only partial because some specific illocutions are determined by lexicon; i.e. a promise and a permission can be sentences performable with the same basically assertive intonation, though they express two different illocutions. Intonation is necessary in the computation of real illocution value, and it is present in any case, though it is not always exhaustive of this purpose. 8 See Fava (1984, 1990, forthcom.) for general subjects concerning the relation between types of speech act and types of sentences, types of speech act and their morphological equivalents in Italian and for a wide analysis of italian questions. 9 Clearly the F variations over an expression of two syllables or of eleven syllables, with completely 0 different phonetic stuff, cannot be the same. But there are some positions, as the final part of a tonal unit, which are relevant and which are the seat of tonicity and of perceptual recognizing, while the remaining part can be neutralized from a perceptual point of view. 2

10

See Cresti (1993, 1994). Sometimes what we have to say can be quite drawn out, and it does seem that there are quantity limits for intonational performance; for instance, in Italian this seems to depend on syllable number. The most natural speech flow in Italian can reach up to seven syllables, and it can be stretched as far as eleven, but it starts to become speedy and unnatural near fifteen. 12 See Bally (1963). 13 In standard Italian it has a specific form, see Tamburini (1994). 14 We insist that nominal sentences cannot be considered secondary, because it has been verified that corpora of spoken language - Italian but also French, English and German - show a marked presence of the structure, which seems to amount to about 10% of all the utterances produced. 15 Among others see Berruto (1985), and Tamburini (1995). 16 I thank Massimo Moneglia for discussing all my research, but in particular for helping me in seeing clearly this question. 17 Concerning the sense of this pragmatical principle, we had to open a very long parenthesis which should imply a new definition of perlocution and indicate the general motivation for speaking. In our opinion, these cannot be defined in terms of a conscious will, but we can better comprehend them as unconscious impulses on the part of the listener-partner. But this is not the proper place for this matter. See Cresti (1992). This hypothesis is made in the theoric framework of Fagioli (1971, 1974, 1975). 18 Such behaviour can be understood if we consider that there is no accomplishment of illocution without reference. An illocution must apply its force to something, and it does so in a definite way (directive vs assertive force). See Searle (1969). 19 The nature of relations among informational units and that among expressions within any informational unit is completely different. In particular relations among expressions within an informational unit are semantic and ruled as domain of identification. 11