Intonative Structure of Focalization in French and Greek David Le Gac & Hi-Yon Yoo Université Paris 7, UPRES EA 333 « Atelier de Recherches sur la Parole », Laboratoire de phonétique, UFR de Linguistique - 2, pl. Jussieu, 75251 Paris cedex 05 [email protected]

Abstract According to Lambrecht (1994), focus and topic are two main components of information structure, where topic expresses the aboutness of an utterance, and focus, its unrecoverable element. As intonation reflects formally this structure, this paper investigates the phonological markers of focus and topic intonation in French. We claim that these two constituents take place in a prosodic structure and that their intonative patterns formally mark dominance relations within this structure. Moreover, we show that the structure defined for French also applies in Modern Greek but poses the problem of tonal alignment with the segmental string.

1. INTRODUCTION The organisation of information in speech communication has been a largely debated subject in all fields of linguistics, from semantics to syntax, as well as prosody. However, the different opinions about the definition of terms such as focus and topic remains less clear. Without getting in the discussion itself, we assume along with Lambrecht (1994) that focus and topic are the two main components of sentence information structure, where topic expresses the aboutness of an utterance, and focus its unrecoverable element1. More specifically, we consider focus as the element of information of a sentence that “cannot be taken for granted at the time of speech”2 while topic sets the referent on which the proposition is construed. This information structure can be formally expressed with grammatical markers, the ordering of the different constituents but also prosodic organisation. Thus, this paper focuses on how prosody expresses formally, i.e. as phonological markers, these constituents in two languages with different prosodic properties, French and Modern Greek, in order to show that information structure has a straight and language independent impact on prosodic structure. Our interest on French may become clearer if we take a look at the following data of French intonation3 : LH H L LH(↓) L↓ L↓ Vendredi à l'école c'est VALERIE qui l'a grondé Jean-Marie Friday

at school

it is Valérie

who scowled him

Jean-Marie

"Friday, at school, Valérie scowled Jean-marie." Representation 1: French declarative sentence "Vendredi, à l'école, c'est VALERIE qui l'a grondé Jean-Marie".

As it appears in Representation 1 and in figures 2 and 3, the intonative pattern of the Focus constituent (c'est VALERIE) in French declarative sentences is characterised by a low tone (symbolised by "L") occurring on the last syllable (-RIE). This pattern has been noticed by numerous authors, among others Martin (1981, 1982), Rossi (1981, 1999), Di Cristo (1981, 1999), Clech-Darbon & al. (1999), Touati (1987). However, focus is also often characterised by its high tone (noted "H") generally found on the left edge of the focalised item (cf. Di Cristo 1998, 1999, Clech-Darbon & al. 1

The notions of topic and focus "has to do with the speaker’s assessment of the relative predictability vs. unpredictability of the relations between propositions and their elements in given discourse situations". The topic and focus dichotomy is one of the most important categories of information structure (Lambrecht, 1994, p.6). 2 Ibid. p. 207 3 The data we present were gathered through an experimental procedure. For details about how the data were gathered, we report to Le Gac (2001).

1

1999, Rossi 1999, Touati 1987). Notice that the placement of the high tone of focus on the left edge is not exclusive, for it can also appear on other syllables, for example at the end of the second syllable "-LE-" as illustrated in Figure 2 (see also Di Cristo 1999). Rossi (1981, 1999) and Di Cristo (1998, 1999) reported that focus conveyed a strong emphasis or an expressiveness when the high tone is associated with the final syllable.

Figure 1: f0 (Hz) of example 1. The focalised proper noun "VALERIE" is marked by a low tone (L) on its ultimate accented syllable "-RIE" and by a high tone (H) at the end of its first syllable "VA-". The PP "à l'école" is on the same tonal level than the NP "Vendredi" (cf. infra).

Figure 2: f0 (Hz) of another instance of example 1. The high tone on the focus is realised at the end of the penultimate syllable "-LE-". In this example, the PP "à l'école" is lowered (H↓).

As for the topic sequences, they are distinguished according to their position in relation to the focus. The pattern found on pre-focal constituents, (Vendredi and à l'école) confirm authors such as Martin 1981, 1982, Rossi & al. 1981, Rossi 1999, Di Cristo 1997, 1998, 1999, Clech-Darbon & al. 1999, Touati 1987. The last syllable of the topic is characterised by a Low High (LH) pattern. Rossi (1999) pointed out that in a succession of such sequences, two contours were possible: Either all LH patterns are on the same melodic level (Figure 1), or the first tonally prominent pattern is followed by a subsequent lowered high tone (LH↓), thus creating a prosodic hierarchy (Figure 2). In the case of post-focal topics (qui l'a grondé) and (Jean-Marie), intonation is globally low. However, Di Cristo & Jankowski (1998), Di Cristo (1999) and Rossi (1999) reported that this final lowering created a steadily declining melodic configuration (L↓…L↓). As far as interrogative utterances are concerned, we observe the inverted patterns of declarative utterances, with a high tone on the last syllable of the focalised constituent (see Representation 2 and Figure 3). Thus, the focus is no more characterized by a final low tone preceded by a high tone, but, by an ultimate high tone preceded by a low one (Rossi 1981, 1999, Martin 1981, Di Cristo 1981). The pre-focal topics are marked by falling patterns (HL) on their final syllable, instead of rising ones (idem). It has to be noticed that the falling pattern on the second topic may be lowered (HL↓) as in the declarative sentences. As for the post-focal topics, the steadily declining low tones are replaced by steadily declining high tones (Figure 3). HL Vendredi

à

Friday

At school

HL(↓) l'école

LH c'est VALERIE

LH↓ qui l'a grondé

LH↓ Jean-Marie ?

It is Valérie

Who scowled him

Jean-Marie

"Friday, at school did Valérie scowl Jean-marie ?" Representation 2: French interrogative sentence "Vendredi, à l'école, c'est VALERIE qui l'a grondé Jean-Marie ?".

2

Figure 3: Intonative patterns in interrogative utterances in French

Following these data, we propose to give a phonological analysis to French intonation. We discuss a recursive and hierarchical prosodic structure inspired by previous theoretical works by Martin, Rossi and Di Cristo. We claim that the focus is the central element of the structure and that all the prosodic elements maintain formal relations with the focus, namely by two rules, Tone Inversion and Tone Copy. We will first give a general outline of previous works on French intonation (section 2) before presenting our own analysis of French intonation based on focalisation (section 3). We then discuss an application of this model to a stressed language, Modern Greek, where a parametric constraint of tone alignment with the segmental string must be applied (section 4). Finally, conclusive remarks are presented at the last section (section 5).

2. PREVIOUS PHONOLOGICAL ANALYSES The data we discussed in the introduction support previous data in different works on French intonation. The idea that the different melodic patterns show their relative prominence in a hierarchical prosodic structure has often been discussed among French scholars. In this section, we present a general survey of different prosodic structures proposed for French through three major authors, Rossi (1999), Di Cristo (1997, 1998, 1999) and Martin (1981, 1982). The data we described in the introductory part will be used to illustrate the different theoretical frameworks.

2.1. Rossi (1999) The heart of Rossi (1999)’s theoretical framework is that the tonal shapes observed in French are interpreted as intonative morphemes marking the hierarchical informational structure. Thus, our Representation 1 can be interpreted as follows (see Representation 3): - Pre-focal topics are realised as the intonative morpheme CT (H) or CD (H↓), - Focus is marked by the intonative morpheme CC. Moreover, the H tone linked to the focus is interpreted by Rossi as a specific morpheme of emphasis symbolised by AF. - According to Rossi (1999), post-focal topics do not have a proper intonation because they are at the lowest level of the informational hierarchy. Therefore, the post-focal topics copy some of the intonative features of CC, for instance the feature L, and thus become cc morphemes. CT AF CC cc cc CD (Vendredi) (à l'école) (c'est VALERIE) (qui l'a grondé) (Jean-Marie) Friday at school It is Valérie Who scowled him Jean-Marie Representation 3: Reanalysis of the utterance in representation 1 according to Rossi’s theoretical framework (1999) CT

As far as the interrogative utterances are concerned, CC is replaced by the intonative morpheme CI. Following the same copy rule as in declarative sentences, the subsequent topics exhibit a reduced version of CI which is represented by ci. The pre-focal topics present an unmarked variant of the CT morpheme, which realises as low in interrogative utterances.

3

2.2. Di Cristo (1997, 1998, 1999) In a more phonological approach to intonation close to the intonational theory developed by Pierrehumbert & Beckman (1988), Di Cristo (1997, 1998, 1999) propose that the high tones on prefocal constituents and the low tone on the focus must be seen as boundary tones associated to the right edge of Intonative Units (or IU). The High tone linked to the focus is interpreted as a floating element, delimiting the focus. Then, tones are placed at the output of a grid, which provides informally the accentual hierarchy of syllables in an abstract way. Moreover, Di Cristo claims that post-focal elements are deaccented, and are analysed as nonautonomous intonative phrases embedded in a Macro Intonative Unit. In this unit, a IU "mother", which carries the focus dominates the IU "daughter", which includes the post-focal constituents. This kind of structure is based on the notion of Compound Prosodic Domain proposed by Ladd (1996).Di Cristo (1999) and Di Cristo & Jankowski (1999) assumes that this subordination triggers a copy of the final tone of the IU "mother" on the IU "daughter". In the following representation (Representation 4), we give as a re-analysis of our own example in Di Cristo’s terms: Friday

At school

Vendredi à l'école [L H]IU [L H]IU

it is Valérie

Who scowled him

Jean-Marie

c'est VALERIE [L L]IU"mother" [ Hfloating

qui l'a grondé [ L

Jean-Marie L]IU"daughter" ]MacroIU

Representation 4: Intonative structure of the declarative utterance along the lines of Di Cristo (1997, 1998, 1999)

In interrogative utterances, the low tone which ends the IU "mother" (i.e. the focus) is replaced by a high tone. This induces slightly declining high tones on the following IU "daughter" (i.e. the post-focal sequence). Di Cristo points out that the pre-focal IUs are generally pronounced "with a global pitch pattern similar to that found in basic declarative utterances" (Di Cristo 1998:212-213), that is to say, the pre-focal IUs manifest a final low tone preceding by a rising pitch movement (or LHL).

2.3. Martin (1981, 1982) Finally, for Martin (1981, 1982), the prosodic structure is strictly right-headed, and is instantiated by the “inversion de pente” between focus and the pre-focal sequences (rheme/theme distinction for Martin) which are on a same level of the prosodic hierarchy. If we re-examine our examples in Martin’s terms, as it appears in Figure 4, the falling pattern of the focus generates twice the same rising pattern on the two pre-focal sequences in declarative sentences. In interrogative sentences the rising pattern on the focus triggers falling patterns on the pre-focal sequences. The patterns are on the same level. Rheme

Vendredi

à l'école

Theme

c'est Valérie

qui l'a grondé

Theme

Jean-Marie

Declarative utterances Interrogative utterances

Figure 4: Intonative structure of the declarative and interrogative sentences according to Martin (1981, 1982)

Unlike the two other authors, Martin does not include post-elements in his prosodic structure. Therefore, they present undetermined and reduced patterns with low patterns in declarative sentences and rather high patterns in interrogative sentences.

4

2.4. Problem The different studies above are interesting because reveal some prosodic structures based on the informative structure. The informational organisation is manifested by formal relations between prosodic elements such as the “inversion de pente” and the phonological copy. However, we consider that they are not sufficient, because this kind of analysis is not homogeneous and all the formal relations among the different constituents do not clearly appear. For Rossi and Di Cristo, the patterns on pre-focal and focal sequences are independent patterns while the patterns on post-focal sequences are determined by the pattern on the focus (by a rule of phonological copy). On the other hand, Martin’s theory reveal opposite phonological relations between the different sequences: what is phonologically determined by the focus are the pre-focal sequences and not the post-focus sequences. The post-focal sequences are considered somehow “hors structure”, that is without clearly determined patterns.

3. INTONATION STRUCTURE OF FOCALISATION The examination of previous analyses of French prosody shows that there is a lack of an unified and homogeneous model. However, the data we described in the introductory part shows clearly the existence of an hierarchical prosodic structure where all the constituents of the utterance are strictly related. Therefore, in this section, we present the prosodic model we developed on the basis of the information structure of an utterance.

3.1. Hierarchical and recursive structure We argue for a recursive tree structure of prosody based on the information structure, and more specifically on the focus constituent of an utterance. This structure is directly inspired by Ladd (1996) and his structure of tunes which updates the proposals of the X-bar theory of constituent structure. In this hierarchical structure, we assume that the minimal intonative unit of an utterance is the INTONATIVE WORD (henceforth IW) in the spirit of Martin (1981, 1982). The IW includes minimally a lexical head (noun, verb etc.) and clitics (determinants, prepositions etc.). It is right bounded by an intonative pattern associated to the accented syllable of the word: Declarative Interrogative

LH H L LH(↓) HL LH HL(↓) (Vendredi)IW (à l'école)IW (c'est VALERIE)IW

L↓ LH↓ (qui l'a grondé)IW

L↓ LH↓ (Jean-Marie)IW

Representation 5 : Declarative and interrogative French utterances organised in intonative words IW (shown in parentheses). The accented syllables are underlined.

In a level above, IWs form higher constituents called INTONATIONAL PHRASES (henceforth IP). An IP can contain one or several IWs. This assumption is suggested by the melodic patterns presented for the pre-focal topics in section 1. The fact that the second pre-focal topic is lower than the preceding one (cf. sequences LH…LH↓ and HL…HL↓ in representations 1 and 2) puts forward the idea of the existence of a hierarchy between each pair of topics. In the latter case (cf. Figure 5 below), we claim that IWs combine into a unique IP where the prosodic HEAD of the IP is borne by the IW on the left (i.e. "Vendredi"). The same reasoning is applicable for the post-focal sequences (cf. tonal sequences L↓…L↓ and H↓…H↓ in representations 1 and 2) where the IW "qui l'a grondé" is higher than the following IW "Jean-Marie". In the contrary case (cf. Figure 6), i.e. when no melodic lowering is noticed on the second pre-focal IP, we argue that the topic sequence do not maintain a hierarchical relation and that each IW of the sequence forms an independent IP. Further arguments for this analysis will be given in section 3.3 along with the phonological mechanisms that apply in the model.

5

IP

(Vendredi)IW

Decl: H Inter: L

IP''

IP''

IP'

IP'

IP*

(à l'école)IW

(c'est Valérie)IW

H↓ ↓ L↓ ↓

IP

(qui l'a grondé)IW

L H

(Jean-Marie)IW

L↓ H↓

L↓ H↓

IP

IP

(Vendredi)IW

(à l'école)IW

Dec: H Inter: L

Figure 5: sequence with hierarchically organised pre-focal topics

IP*

(c'est Valérie)IW

H L

L H

IP

(qui l'a grondé)IW

L↓ H↓

(Jean-Marie)IW

L↓ H↓

Figure 6: sequence with non-hierarchically organised pre-focal topics

Finally, in an upper level, the post-focal IPs and the IP-focus form together an IP’ as can be seen in figures 5 and 6. The IP’ combines into an IP”, the highest level of the structure. Equally, each prefocal IP combines directly and separately into the IP” level (Figure 6). The last and most important statement we want to put forward is that the IP bearing the focus is the HEAD of the whole structure (represented as a IP* in our figures). We argue that the IP-focus must be considered as the head for the following reasons. First, it is possible to present just the IP bearing the focus as an answer to a Wh-question. Second, in an informational level, the focus is the most important piece of information. Finally, one main reason to regard the IP-focus as the head is found in the role it plays on the phonological level, as it will be shown in the following sections.

3.2. The notion of tonal primitive As it was discussed above, the IP bearing the focus plays a central role in the structure. In the data we presented (section 1), we saw that the IP-focus bears itself a tone on its ultimate syllable (respectively a low and a high tone in declarative and interrogative utterances). Being the head of the structure, we claim that the IP-focus bears the TONAL PRIMITIVE of the utterance. This tone is called the tonal primitive of the utterance for two main reasons. First, the tonal primitive is considered as given contrarily to the other tones: the tonal primitive is fixed and is determined by the modality of the utterance. Second, the tonal primitive allows to derive ALL the other tones of the structure. Thus, as it appears in Figure 7, in the case of declarative utterances, the tonal primitive is a low tone, that we call the Low Tone of Focalisation (henceforth Lf). Lf will derive the other tones following the dominance relations that exist between the IP-focus and the other IPs. In the case of interrogative sentences, the tonal primitive is a high tone associated with the focalised item and noted Hf. The underlying structure remains unchanged and therefore, Hf generates different tones on the remaining IPs by applying the same phonological rules as in the declarative utterances. IP''

IP' Intonative structure

IP

IP*

IP

(Topics)IW

(Focus)IW

(Topics)IW

Interrogative utterance

…L(↓)…

Hf

H↓

Declarative utterance

…H(↓)…

Lf

L↓ Phonological level

Phonological rules

Figure 7: general intonative structure with the tonal primitives of declarative and interrogative utterances.

6

One may ask why tonal primitive of focus in declarative utterances is a low tone while it is a high tone in interrogative ones. A possible explanation of this choice can be found in Vaissière (1995), presenting the cross-linguistic prosodic similarities: "Falling and low values are associated with the notion of end". Following the same reasoning, we can consider that the tonal primitive in declarative utterances is a low tone because it somehow ends the utterance at the informational level as the main information of a message. Conversely rising patterns are found in the focus of interrogative utterances because these patterns correspond to the well known notion of continuity. Vaissière presents this argument for the rising patterns found in pre-focal topics in declarative sentences. However, while Vaissière’s extralinguist reasoning is built on physiological and psychological explanations, we claim that the tones on topics are determined by strictly phonological reasons, namely by the dominance relations we present in the following subsection.

3.3. The formal dominance relations Once the tonal primitive is fixed, dominance relations within the prosodic structure are manifested between the tonal primitive and each Topic-IP. Two rules apply: the TONE INVERSION RULE (from right to left) and the TONE COPY RULE (from left to right). These rules apply to all structures whatever the modality is. Thus, as we discussed in section 3.1, two cases surface concerning the internal relationship between pre-focal topics. Since the main difference lays on the presence or not of hierarchy between pre-focal topics without changing the rules, we will present the two rules by illustrating them principally in a structure with hierarchically organised topics. We will then turn to non-hierarchically organised topics. 3.3.1. Phonological rules in a structure with hierarchically organised topics In a declarative sentence, when the topics are hierarchically organised, the pre-focal sequence has a tonal contour with an high tone on the first topic and a lowered H↓ on the second topic while a L tone is found on the rest of the sentence. In interrogative utterances, we observe the inverted patterns of declarative utterances. As in declarative sentences, a lowered HL↓ is also found on the second prefocal topic. The following figure (Figure 8) illustrates the application of the two phonological rules within the intonative structure formed by the combination of hierarchically organised IPs: IP''

IP'

IP

(Vendredi)IW

IP*

(à l'école)IW

(c'est Valérie)IW

IP

(qui l'a grondé)IW

(Jean-Marie)IW

Interrogative utterance

L

L↓

Hf

H↓

H↓

Declarative utterance

H

H↓

Lf

L↓

L↓

Copy+Lowering

Copy+Lowering

Tone Inversion

Figure 8: Intonative structure and relations of dominance when the second pre-focal topic "à l'école" is lowered

In such a sentence, the two phonological rules apply as follow : ♦ Tone Inversion: Tone Inversion takes care of the relation of dominance from right to left. This rule which is directly inspired by Martin’s work, makes it possible to derive the tones on the left of the focus constituent. In declarative sentences, Lf generates a High tone on the Head of the prefocal IP, that is on the IW (Vendredi). Following the same principle, the tonal primitive Hf generates a L tone on the Head of the IP in interrogative sentences.

7

Moreover, we claim that the tone inversion rule occurs not only between IPs, but also locally within an IW. Indeed, French is characterised by tonal patterns such as LH and HL (example 1). We assume that the tone on the right edge of the IW inverts the preceding one. For instance, in declarative utterances, H tones of pre-focal topics give locally a L tone, and Lf of the IP-focus a H tone. Thus, from the tonal primitive of the structure, we obtain LH patterns on pre-focal topics for declarative sentences and HL patterns for interrogative ones while we obtain respectively HL and LH patterns for focus on declarative and interrogative sentences (cf. Figure 9). As far as the post-focus sequence is concerned, tone inversion rules also applies in interrogative utterances where the high tones give locally low tones. However, this inversion rule does not clearly appear in declarative sentences. We believe that this Tone inversion rule still takes place but in a weakened manner. (Vendredi)IW

(à l'école)IW

(c'est Valérie)IW

(qui l'a grondé)IW

(Jean-Marie)IW

Interrogative utterance

H L

H L

L

Hf

L H↓

L H↓

Declarative utterance

L H

L H

H

Lf

(H) L↓

(H) L↓

Local Tone Inversion

Local Tone Inversion

Figure 9: local tone inversions in interrogative and declarative utterances

♦ Tone Copy: The rule of Tone Copy takes care of the relation of dominance from left to right, and generates a tone lowering. As it appears in Figure 8, in declarative sentences, this process intervenes between the IP of focus (c'est Valérie) and the Head of the post-focal IP (qui l'a grondé), as well as between the latter and the final IW (Jean-Marie), thus generating steadily declining low tones. In interrogative sentences, Hf triggers iteratively lowered H tones. Moreover, we state that this Tone Copy rule occurs between the Head of the pre-focal IP (Vendredi) and the following IW (à l'école). Thus, in declarative utterances, we will have LH…LH↓ sequences while this rule will trigger HL…HL↓ sequences in interrogative utterances. It has to be noted that the process of Tone Copy we propose is somewhat different of that one described by Rossi (1999), Di Cristo & Jankowski (1999), Di Cristo (1999) and Clech-Darbon & al. (1999). First, our Tone Copy rule implies a lowering of the subsequent tone. Second, this rule does not occur only between the focus and the post-focus as claimed in the previous works, but also locally between pre-focal IWs: we claim that it is a general phonological process, taking care of all relations of dominance from left to right. 3.3.2. Phonological rules in a structure with non-hierarchically organised topics Now, when the sequence of topics is not hierarchically organised (presenting for instance a tonal sequence H…H), the two pre-focal IWs derive directly from the IP'' and each IW form an independent IP. This structure is shown in Figure 10. Thus, Tone Inversion rule apply between the IP-focus and each preceding topic-IP. In our case, The tonal primitive of the IP-focus generates an inverted tone in the right edge of each IP. IP''

IP'

IP

IP

IP

IP

(Vendredi)IW

(à l'école)IW

(c'est Valérie)IW

(qui l'a grondé)IW

(Jean-Marie)IW

Interrogative utterance

L

L

Hf

H↓

H↓

Declarative utterance

H

H

Lf

L↓

L↓

Copy+Lowering Tone Inversion

Figure 10 : Intonative structure and relations of dominance when the two pre-focal topics (vendredi) and (à l'école) are on the same tonal level

8

One may ask on what grounds it is possible to choose between the two structures we argue for the pre-focal topics. We assume that this choice is a result of the informational hierarchy the speaker makes in order to organise the pre-focal topics into informational foreground and background. Thus, in figure 7, the pre-focal topics are placed on the same informational level whereas in figure 6, the second pre-focal topic, which is prosodically dominated, is somehow sent to the informational background. Arguments can be found in the examination of the decreasing configuration. Rossi (1999) also reported that a second pre-focal topic can be lowered. He argued that this tonal lowering was not a linguistically controlled phenomenon but was due to physiological reasons. However, Pierrehumbert & Beckman (1988) put forward the idea that a physiological lowering consists in a progressive decrease of 10Hz/sec. In our data (see for example figure 2), it appears that the tonal interval between the two high tones of the pre-focal topics is about 30Hz (±200 ~ ±170Hz, i.e. ≅5 quarter tones). In other words, the tonal interval in question is too important to be due to a physiologically lowering and suggests that the second high tone is lowered for linguistic reasons. Moreover, the tonal interval of 5 quarter tones that appear in our data is exactly the tonal interval which is found between prosodic boundaries due to syntax, i.e. between the nominal phrase and the verbal phrase of a neutral simple sentence (for more detailed discussion see Touati 1987 and Le Gac 2001), showing that a tonal interval of 5 quarter tones manifests a linguistic hierarchy. We thus believe that this lowering configuration can not be explained by a simple physiological reason but it is a linguistically controlled phenomenon, indicating the informational hierarchy.

3.4. Unmarked tones To end the analysis of the prosodic structure of French utterances, we turn to the question of tones which are not handled by the intonative structure. This completes the account of the whole melodic configuration of an utterance. In fact, these tonal objects do not present any contrast between them, remain unchanged, and therefore do not convey any information. In the case of French, we always find a low tone at the beginning of each IW whatever the modality of the utterance is. We consider that this kind of objects is unmarked tones, that can be found on edges and that are low by default.

4. MODERN GREEK In this section, we aim to extend our analysis of the intonative structure based on focalisation, by examining a “free” stress language, Modern Greek. A general examination of Greek utterances reveals similar patterns than French.

4.1. Intonative structure in Greek declarative utterances The examination of the different utterances produced by our speaker shows that the intonative patterns of Greek are quite similar to those of French, as it appears in Representation 6 and in Figure 11 and Figure 12: pre-focal topics have a LH pattern, focus exhibits a HL pattern, and post-focal topics are steadily lowered, thus confirming the results found by Botinis (1989, 1998) and Baltazani & Jun (1999). L H

(tin deftera)

L H(↓) (sto spiti)

H L

(i Marina)

L↓ (xtipise)

L↓ (ton Manoli)

Art-Acc Monday Loc home Art-Nom Marina Verb-past Hit Art-Acc Manolis4 Representation 6: Greek declarative utterance "On Monday, at home, Marina hit Manolis"

4

Art: article, Acc:accusative, Loc:locative, Nom:nominative,

9

Figure 11 : F0 (Hz) of example 3 where the intonative word (tin deftéra) is on the same tonal level than the intonative word (sto spíti)

Figure 12 : F0 (Hz) of the Greek utterance where the intonative word (sto spiti) is lowered (H↓)

Therefore, we claim that the structures we have proposed for French so far can be applied to a stress language such as Modern Greek : we assume that the tonal primitive is, as in French, the low tone of the HL pattern of focus and that the formal processes of dominance apply in the same way. Thus, the Tone Inversion rule applies as in French and gives LH patterns on pre-focal topics while the Tone Copy rule triggers steadily declining low tones on post-focal topics and a lowered high tone on the second pre-focal topic in sentences where there is an informational hierarchy between pre-focal topics (cf. Figure 12).

4.2. The problem of tonal alignment The main difference between French and Greek is in the alignment of tones with syllables within the intonative word. In Greek, the tonal alignment depends on its accentual mobility: the place of the intonative patterns depends on stressed syllables (stressed syllables are underlined in Representation 6 and in figures 11 and 12). We assume that tonal alignment is a parametric constraint. In both languages, the last tone of a pattern is linked to the last syllable of the intonative word. As for the first tone of the intonative pattern, in Greek, it is associated with the lexical stressed syllable from right to left. It differs from French, where this first tone can associate with any syllable because French words do not bear any lexical stress. Thus, in Greek, Lf appears on the last syllable of the IP-Focus (i Marina), and the H tone is aligned with the stressed syllable -ri-, as illustrated in Figure 13. In French, in the IP-focus (c'est Valérie), Lf also associates with the last syllable, but the high tone can appear either on c'est, Va-, or -lé- (cf. for example Di Cristo 1999), as it is represented with the dotted lines in Figure 145. H Lf H Lf    i Ma ri na Cest Va lé rie Figure 13: Tonal alignment in Greek

Figure 14: Tonal alignment in French

4.3. Interrogative utterances in Greek 4.3.1. Intonative Patterns In the case of interrogative utterances in Greek, pre-focal topics present inverted patterns as in French : HL patterns are found on these sequences, with the High tone on the stressed syllable,

5

It has to be noticed however that this high tone is generally linked to the first syllable.

10

followed by the low tone. This inversion is illustrated in Figure 15 where pre-focal sequences can be put into a hierarchy, in the same way as for the declarative sentences. In the sequence grouping the focus and the post-focal topics, we observe that the focus (i Marina) and the first post-focal topic (xtipise) are characterised by a relatively flat and low configuration whereas the last word of the sequence carries out an LH pattern on its stressed syllable, followed by a final low tone (Baltazani & Jun 1999, Botinis 1989, Ladd 1996).

Figure 15: F0 (Hz) patterns in interrogative utterance in Greek.

4.3.2 Interpretation It seems that the data we encountered for interrogative utterances do not correspond entirely to the structure we proposed for French. However, we claim that in interrogative utterances of Greek, the tonal primitive is still the High tone associated to the last word of the utterance. In fact, the High tone triggers HL patterns on pre-focal topics as in French. However, we assume that in Greek, the focus and the post-focal topics are reorganised into a unique IP which is right-headed. In this IP, the tonal primitive H triggers low tones locally from right to left, that is on the focus and the first post-focal topic. It is generally admitted that the items inside the sequence preceding the last word are marked by a low tone on their stressed syllable (Baltazani & Jun 1999, Botinis 1989, Ladd 1996). Nevertheless, our data, of which Figure 15 is an example, show that stressed syllables exhibit a rising of 10/15 Hz above the preceding syllable. In sum, all the words in the utterance present a HL pattern contrasting with the LH pattern on the last word. As for the low tone at the end of the interrogative utterance, we consider that it is an unmarked boundary tone. This low tone appears at the end of all utterances, whatever the type of modality is. In other words, we assume that this low tone is also present at the end of declarative utterances.

5. CONCLUSION The present paper accounts for an intonative structure where the relations of dominance between constituents are expressed by two formal processes, Tone Inversion occurring from left to right, and Tone Copy implying tone lowering. The pivot of the structure is the tonal primitive, which is Lf in declarative utterances, and Hf in interrogative ones. We have shown that this structure is the same in French and in Modern Greek, with mere parametric differences. This is not astonishing, since other studies bring together the prosodic configuration of Modern Greek with Roman languages such as Romanian (see Ladd 1996). We believe it is interesting to verify the structure we proposed for French and Modern Greek for other languages with other prosodic properties. Italian could be a good example, where it seems that same intonative patterns appear in association with the accentual melodic shape (Rossi, 1998). Thus, the melodic rise under the stress syllable is followed by the melodic rise or the high plateau CT on prefocal topics, while focus and post-focus manifest respectively fall and low values. Beyond the Indo-european phyllum, languages such as Korean and Somali seem to follow the same intonative configurations : focus is generally realised as sharply fall, contrasting with a high "plateau" on the pre-focal topics and a low tone on the post-focal topics (Park 2000, Le Gac 2001). The similarities we can observe in all these different languages, and therefore the application of the two rules, suggest that the intonative structure of focalisation we discussed is a widespread linguistic phenomenon. 11

Bibliography Baltazani M.& Jun S-A (1999), "Focus and topic intonation in Greek" in Proceedings of the XIVth International Congress of Phonetic Sciences, San Francisco, pp.1305-1308 Botinis A. (1989), Stress and Prosodic Structure in Greek : a phonological, Acoustic, Physiological and Perceptual Study, Travaux de l'institut de linguistique de Lund 22, Lund (PhD dissertation) Botinis A. (1998), "Intonation in Greek" in Intonation Systems, Hirst D. & Di Cristo A. eds., Cambridge University Press Clech-Darbon A., Rebuschi G. & Rialland A. (1999) ‘Are There Cleft Sentences in French?’ in G. Rebuschi and L. Tuller eds., The Grammar of Focus, John Benjamins, Amsterdam, 83-118. Di Cristo A. (1997), "Contribution à l'élaboration du cadre accentuel du français", in Proceedings of the 16th International Congress of Linguists. Pergamon, Oxford, Paper n° 0159 Di Cristo A. (1999), "Le cadre accentuel du français contemporain : essai de modélisation : deuxième partie" in Langues vol.2 n° 4 pp.258-267 Di Cristo A. (1998), "Intonation in French" in Intonation Systems, Hirst D. & Di Cristo A. eds., Intonation System, Cambridge University Press, Cambridge Di Cristo A. & L. Jankowski (1999), "Prosodic Organisation and Phrasing after Focus in French", in Proceedings of the XIVth International Congress of Phonetic Sciences, San Francisco, pp.15651568. Hirst D. & A. Di Cristo, eds., (1998), Intonation System, Cambridge University Press, Cambridge Ladd D.R. (1996), Intonational Phonology, Cambridge University Press, Cambridge Lambrecht K. (1994), Information structure and sentence form : Topic, focus and the mental representations of discourse referents, Cambridge Studies in Linguistics 71, Cambridge University Press, Cambridge Le Gac D. (2001), Structure prosodique de la focalisation : le cas du somali et du français, PhD dissertation, Université Paris 7, Paris Martin Ph. (1981), "Pour une théorie de l'intonation" in Rossi M., A. Di Cristo, D. Hirst, Ph. Martin & Y. Nishinuma eds. L'intonation, de l'acoustique à la sémantique, Paris, Klincksieck Martin Ph. (1982), "Phonetic realisations of prosodic contours in French". in Speech Communication 1 (3,4):283-294. Park M-K. (2000), Contribution à l'étude des structures rythmiques en parole lue et en parole spontanée: autour de l'accent focal en coréen et en français. PhD dissertation, Université Paris 7, Paris Pierrehumbert J. & M.E. Beckman (1988), Japanese Tone Structure, MIT Press, Cambridge Rebuschi G. & L. Tuller, eds., (1999), The Grammar of Focus, John Benjamins, Amsterdam Rossi M., A. Di Cristo, D. Hirst, Ph. Martin & Y. Nishinuma (1981), L'intonation, de l'acoustique à la sémantique, Klincksieck, Paris Rossi M. (1998), "Intonation in Italian", in Intonation Systems, Hirst D. & Di Cristo A. eds., Intonation System, Cambridge University Press, Cambridge Rossi M. (1999), L’intonation, le système du français : description et modélisation, Ophrys, Paris Touati P. (1987), Structures prosodiques du suédois et du français, Travaux de l'institut de linguistique de Lund 21, Lund (PhD dissertation) Vaissière J. (1995), "Phonetic Explanations for Cross-Linguistic Prosodic Similarities" in Phonetica n°52

12