Effects of speech rhythm on spoken syntax A corpus-based study on Brazilian Portuguese and Italian

This paper presents a preliminary study on the syntax/prosody interface in spoken Brazilian Portuguese and Italian, within the Language into Act Theory theoretical framework (L-AcT, Cresti 2000). According to this theory, spoken syntax has to be studied taking into consideration the way the speaker combines information along one or more information units, since each of them is understood as a syntactic/semantic island (Cresti 2014). Then, linearized syntax defines proper dependency relationships, performed within the same information unit, while patterned constructions , syntactic structures performed across more than one information unit, are to be analyzed considering the pragmatic functions of the information units involved (Cresti 2014). According to a non-discrete conception of speech rhythm (Barbosa 2000, 2006), BP can be defined as a more stress-timed language with respect to IT, and its tonal unit can host more phonological syllables than IT’s, for BP’s rhythm leads to strong re-syllabification phenomena. Based on data from the C-ORAL-BRASIL (Raso & Mello 2012) and Italian C-ORAL-ROM (Cresti & Moneglia 2005) spontaneous speech corpora, this paper aims at verifying if such prosodic features partake in the way the two languages display preference for linearized syntax or patterned constructions .


Spontaneous speech and the Language into Act Theory: the role of prosody
The study presented in this paper is based on the Language into Act Theory (henceforth: L-AcT; see Cresti 2000;Moneglia & Raso 2014;Raso 2012), an inductive theory on spoken language built through decades of empirical work on the LABLITA (Linguistic Lab of the Italian Department of Florence University, Italy) spoken corpora (Moneglia 2011).Based on L-AcT, the C-ORAL-ROM spoken corpora for Italian, Spanish, French and European Portuguese (Cresti & Moneglia 2005), and, later, the C-ORAL-BRASIL corpus were built (Raso & Mello 2012).
In the analysis of speech, L-AcT is innovative in taking into significant consideration its prosodic dimension, understood as the fundamental one for several reasons which will be illustrated further on in this section.The emphasis on prosody could sound kind of obvious in the analysis of speech, but it's not: for a long time, traditional studies on spoken language have based their analysis on speech transcriptions, which means on written texts and, hence, on linguistic categories derived from and adequate to written, not spoken, language.
L-AcT identifies the unity of reference for speech in the utterance, understood as the shortest linguistic unit prosodically and pragmatically interpretable as autonomous, i.e. as a speech act.Following Austin's Speech Acts Theory (Austin 1962), L-AcT assumes that the speech act is made up by three simultaneous acts: locutionary act, corresponding to the transmitted linguistic content; -illocutionary act, corresponing to the action (illocution) the speaker performs through the speech act; -perlocutionary act, corrisponding to the affective impulse that leads the speaker to perform the speech act 1 .
1 In Austin's view, on the contrary, the perlocutionary act corresponds to the speaker's intended goal.
2 Starred abbreviations stand for the speakers' names in the C-ORAL corpora; in bfamdl acronym b refers to Brazilian corpus (i is used to refer to the Italian one), fam to familiar-private context (vs.pub for public), dl to dialogue (vs.mn monologue, cv conversation).The acronym is followed by the number of the recording session, and then the number of the utterance.
Besides conveying the illocutionary value of the utterance, according to L-AcT prosody plays an important role in the segmentation of the speech flow, through two different kinds of prosodic breaks: terminal breaks (i.e.perceived as having a terminal intonation) signal the utterance boundaries ("//" in the C-ORAL-ROM and C-ORAL-BRASIL transcriptions, see (1) above), while non-terminal breaks segment tonal units within the utterance (marked with "/").
A prosodic break does not necessarily correspond to a pause: according to L-AcT, it is defined as a "perceptively relevant prosodic variation in the speech continuum such that it causes the parsing of the continuum into discrete prosodic units" (Moneglia & Cresti 2006: 92).
Following the IPO -Institute of Perception Research of the University of Eindhoven model of speech perception, L-AcT assumes that the f 0 (fundamental frequency) exhibits a series of movements within the speech flow, among which some are voluntary (although unconscious) and perceptually relevant, thus carrying informational values4 (t 'Hart, Collier & Cohen 1990;Firenzuoli 2003).
The prosodic units can, therefore, convey Information Units (IUs), defined by specific prosodic profiles5 , functions and distribution within the utterance.According to the L-AcT's definition, then, the utterance needs at least one IU which carry the illocutionary force in order to be pragmatically interpreted as autonomous.Such a IU is the Comment, and in (1) above most of the utterances are formed by this single unit (simple utterances: Urano //), but utterances can also be formed by the Comment plus one or more IUs (compound utterances: see copos / que tem aí //).
IUs are divided into two main groups, according to their macro-functions: textual IUs form the text of the utterance, i.e. convey its semantic and syntactic content, or serve in order to interpret the utterance or part of it (they are: Topic, Comment, Multiple Comment, Bound Comment, Appendix of Topic/Comment, Parenthesis, Locutive Introducer); dialogic IUs serve in order to regulate the interaction (they are: Phatic, Allocutive, Conative, Incipit, Expressive, Discourse Conector) 6 .
Units without information value are also found in speech and described by the L-AcT, as for example Time Taking units (i.e.filled pauses) and Empty units (i.e.retractings, or last units of interrupted utterances).For this paper's purposes, it is particularly worth mentioning, within this group, the Scanning units (see example (5) in section 3.2.2).These units don't bear any prosodic nucleus nor informational value and are used in order to scan the linguistic content of an IU (usually, Topic, Comment or Locutive Introducer).Actually, Scanning units correspond to different tonal parts of the same IU when the locutive content exceeds its capacity, or when it is effectively scanned by the speaker, be it for stylistic or performance reasons.

2.
Premises to the study: L-AcT and spoken syntax Before analyzing the syntax-prosody interface, two premises are necessary: the first one concerns how is spoken syntax conceived within L-AcT, and it will be presented in the following subsection.
2.1 L-AcT and spoken syntax: linearized syntax vs. patterned constructions Since a significant amount of the linguistic content in speech is made up by strongly fragmented structures, problems arise when traditional syntactic criteria are applied to the analysis of spoken syntax.Thus, the sentence as a unit of reference for speech has been easily discarded by scholars, among which most choose the clause or clause complexes instead (Chafe 1984;Quirk et al. 1985;Halliday 1985;Voghera 1992;Miller & Weinert 1998).Nonetheless, the clause as well is not a completely satisfactory unit of reference for spontaneous speech, and not only much of spoken language is made up by loose phrases, interjections or fragments, but also prosodic and syntactic boundaries often do not overlap (cf. the idea that intonation "is not married to syntax", Bolinger 1984: 413; see also Mithun 2009).A common approach to the "broken" surface of spoken language has been that of postulating an underlying syntactically well-formed structure, a high-level grammar in which syntactic boundaries are very clear, but then transformed and "chopped" in speech due to extralinguistic, interactional, and even performance factors.
Against this idea, Heath (1985, apud Miller & Weinert 1998) on Ngandi aborigenal language assumes that such a fragmented configuration of syntax can indeed be a high-level feature of languages, like Ngandi, very different from English or most European languages.Nonetheless, Miller & Weinert (1998: 356) observe that such a discrepancy between the grammars of Ngandi and English appears strongly reduced if spoken English is analyzed.In this way, they strongly affirm the importance of, first, avoiding the bias of a long tradition of studies on written language and, then, studying spoken language(s) as having its specific grammar and structures (see also Miller & Fernandez-Vest 2006).
Firmly against the idea of speech as a "corrupted" surface form of languages' grammars, Sornicola (1981) claims that the syntactically fragmented nature of spoken syntax is indeed speech-specific and intentionally produced in such broken-up structures by the speakers, while the extralinguistic context and the interlocutors' world knowledge make up for the strongly reduced syntactic linkage.
From similar assumptions, and from the fact that no syntactic unit could consistently cope with such a fragmentation of speech, L-AcT chooses a nonsyntactic unit of reference for speech, i.e. the utterance, which is prosodically and pragmatially defined, independently from its morphosyntactic content.
L-AcT conception of spoken syntax is, instead, closely related to the way the information is packaged by the speaker: within the utterance, each single IU is conceived as a syntactic and semantic island, hence hosting true dependency relationships, while the relationship between different IUs is not syntactic, but related to their informational functions, hence, pragmatic.
Therefore, L-AcT distinguishes between linearized syntax, i.e. proper instances of syntactic dependency, performed within the same IU, and patterned constructions, i.e. syntactic structures performed across more than one IU, to be analyzed, therefore, through pragmatic/informational criteria (Cresti 2014).Examples are provided in (2) and (3): (2) *FLA: eu acho que no saco do lixo nós já passamos //=COM= 'I think that we have already passed through the trash bag (aisle) // (bfamdl01, 506) (3) *LUZ: porque eu acho que no mesmo concurso /=TOP= cê não pode fazer duas //=COM= (bfamdl03, 181) 'because I think that in the same competition / you can't run for two (positions) //' Examples ( 2) and ( 3) are instances of complement clauses performed in linearized and patterned configuration, respectively.In (2), main predicate and complement are performed within the same IU and, therefore, a true dependency relationship is detectable.But in (3), main clause and complement are performed by the Topic/Comment information pattern: in this case, the matrix clause in Topic is not to be analyzed as the head of regency, because a relationship of pragmatic aboutness exists between the two IUs.Therefore, in (3) the matrix clause in Topic is serving the pragmatic function of defining the domain of application of the illocution conveyed by the Comment, in which, by the way, the complement clause is performed 7 .

Linearized syntax vs. patterned constructions in spoken Brazilian Portuguese and Italian
The second premise to this paper consists in a previous study on the syntax of explicit complement clauses in spoken Brazilian Portuguese and Italian (Bossaglia 2014), whose results have inspired the research presented here in its first steps.
In Bossaglia (2014)  7 From a pragmatic perspective, then, it is the complement, and not the matrix clause, that performs the most important function (i.e.carrying the illocution), while the matrix clause serves as background information.It goes beyond the aims of this paper to go in-depth into such a pragmatic organization of spoken syntax.For further and more detailed discussion see Cresti (2014) and Bossaglia (2014Bossaglia ( , 2015)). 8http://lablita.dit.unifi.it/app/dbipic/That survey on complement clauses aimed at mapping such syntactic constructions in both languages and analyzing them from a contrastive perspective with respect to the types and frequencies of complementizers, the information patterns involved, and the ratio of linearized syntax and patterned constructions.
BP and IT displayed a very similar realization of complement clauses in the DB-IPIC corpora: the same complementizers were detected (in order of frequency, PB/IT: que/che 'that', se/se 'whether', plus various indirect question complementizers in both languages) and showed comparable frequencies; the same information patterns appeared to be used in PB and IT (mainly Topic/Comment and chains of Multiple Comments or Bound Comments; IT data displayed also a few textually more articulated patterns than PB); linearized configurations were strongly predominant on patterned constructions in both languages.
In view of such a great, and expected, homogeneity between BP and IT in the realization of complement clauses, the only interesting deviation was found in the ratio of linearized syntax and patterned constructions: 83,9% vs. 16,1% in BP, 67,3% vs. 32,7% in IT (Bossaglia 2014: 48).
Our hypothesis correlates such a discrepancy to a prosodic difference between PB and IT, namely their rhythmic typology.In this paper, its preliminary experimental verification is presented.

3.
Syntax-prosody interface in spoken BP and IT

Speech rhythm typologies: BP and IT
It is widely accepted that rhythm is a language-specific, psychologically relevant category.Many experimental studies focused on perception have observed that adult and newborn individuals are highly sensitive to languages' rhythm, and capable of clearly discriminating between rhythmic typologies of different and unfamiliar languages (see, for example, Bertoncini & Mehler 1981;Mehler et al.1996;Nazzi, Bertoncini & Mehler 1998;Ramus, Nespor & Mehler 1999).
Traditionally, natural languages have been assigned a stress-timed or syllable-timed rhythmic typology according to whether they present isochrony, i.e. periodicity, of temporal duration between, respectively, stressed syllables or syllables (Pike 1945;Abercrombie 1967).Such a dichotomy was early proven to be problematic, since natural languages, in general, don't manifest only one rhythmic typology depending on the context of the communication (Roach 1982;Dauer 1983;Bertinetto 1989).Besides, focusing exclusively on the temporal dimension for the definition of rhythmic typologies means disregarding many of the variables involved in speech rhythm.More recently, different approaches have managed to better account for speech rhythm by analyzing it through a more complex perspective (among others: Saltzman & Munhall 1989;Cummins & Port 1998;Cummins 2002;Barbosa 2000Barbosa , 2004Barbosa , 2006)).The inspiration for such a new perspective in the study of speech rhythm can be found in studies on human and animal motricity (e.g.Kelso, Tuller & Harris 1983;Schöner & Kelso 1988;Kelso 1995) that observed the existence of synergic systems, in which different physical components interact both one with another (functional coupling) and with the environment.According to the Dynamical Model of speech rhythm proposed by Barbosa (2000Barbosa ( , 2004Barbosa ( , 2006Barbosa ( , 2007) ) the temporal dimension is still taken into account, of course, in the production of speech rhythm, but within a self-organized synergy whose components are (Barbosa 2006: 10-11; see also Barbosa 200, 2001Barbosa 200, , 2004Barbosa 200, , 2007)): a. interaction of high-level linguistic information with an accentual oscillator; b. funcional coupling of two abstract oscillators, accentual and syllabic; c. coupling of the syllabic oscillator with gestural level9 .
In this way, a less dichotomic approach to speech rhythm is proposed, since it is postulated that both the accentual and the syllabic oscillators interact in speech rhythm production in any natural language.This means that the same mechanism is responsible for both the stress-timing and syllable-timing tendencies according to a greater force of, respectively, the accentual or the syllabic oscillator in their functional coupling (Barbosa 2001: 33;Barbosa 2000Barbosa , 2004Barbosa , 2006)).
It follows that within this perspective speech rhythm is a non-discrete category: there aren't any tout court stress-timed nor syllable-timed languages (as it was claimed by Abercrombie 1967: 67), and the rhythmic typology of any natural language can be described both as more stress-or syllable-timed depending on the languages to whom we compare it with10 .
The hypothesis presented here springs from the assumption that Brazilian Portuguese displays a stronger stress-timing tendency with respect to Italian, and, conversely, Italian displays a stronger syllable-timing tendency with respect to Brazilian Portuguese (Barbosa 2000: 392) 11 .
The consequence of such a difference in BP and IT speech rhythm is that BP's tonal unit (see section 1) can in principle host more phonetic and phonological material than IT's, since a stronger entrainment of the accentual oscillator over the syllabic one results in a sort of "temporal compression", by which less time is needed in order to pronounce a higher number of syllables (Barbosa 2001: 32-33).
Therefore, the higher ratio of linearized complement clauses in BP mentioned in the previous section could be due to BP stress-timing tendency with respect to IT.In the next sections, the first methodological steps in order to verify such a hypothesis are presented.

The DB-IPIC corpora
The data presented in this study come from the two comparable minicorpora of C-ORAL-BRASIL and Italian C-ORAL-ROM, available online at the DB-IPIC platform 12 .The Data-Base of Information Patterning Interlinguistic Comparison was built as a rich query interface allowing contrastive studies based on the C-ORAL corpora.In this online platform, access to the Italian Informal C-ORAL-ROM corpus, a minicorpus extracted from it, and a minicorpus extracted from the Informal C-ORAL-BRASIL is possible.All DB-IPIC corpora are provided with the transcriptions of the recordings implemented with prosodic break annotation, while the audio file of each utterance is available thanks to the textto-speech alignment.Besides the informational annotation (i.e.annotation of the IUs), Italian corpora are provided with PoS tagging as well (at the moment, PoS tagging for BP is available only in the C-ORAL-BRASIL DVD, where the whole corpus is available).
The two BP and IT DB-IPIC minicorpora are made up by 20 recorded sessions, and were built in order to be representative of their reference corpora and comparable one with another.As regards their representativeness, the proportions between monological and dialogical interactions within them is the same as within the C-ORAL-BRASIL and the Italian C-ORAL-ROM corpora, i.e. one third of monologues vs. two thirds of dialogues and conversations (interactions with more than two active participants).The same holds for the proportions between Family/Private (15 recorded situations over 20 in BP, 16 over 20 in IT) and Public communicative contexts.As in their reference corpora, a high degree of diaphasic variation is detectable in the minicorpora, in order to best represent the richness of spontaneous speech situations (for more details, see Panunzi & Mittmann 2014;Panunzi & Gregori 2012).
The comparability of the architecture and size (in terms of number of words and prosodically Terminated Sequences, i.e.Utterances and Stanzas13 ) of DB-IPIC minicorpora is described in Table 1 below, adapted from Panunzi & Mittmann (2014: 140): Although the two minicorpora differ in the number of words (31,465 vs. 37,355), it is important to notice that in terms of number of units of reference they are perfectly comparable (5,512 vs. 5,663).Such a comparability represents a key factor for the purposes of this study.

Data extraction and analysis
In order to test our hypothesis that BP tonal unit can host more phonological syllables than IT's, and that this would be one of the factors involved in the higher rate of linearized syntax than in IT, we followed three methodological steps.
First, we extracted from the two minicorpora all the simple utterances (formed by the COM unit only) and simple scanned utterances (formed by the COM unit, performed through more than one tone unit) that contain a verbal predication14 , as it is exemplified in ( 4) and ( 5), respectively: (4) *ART: qualcuno dice che ci si mette meno //=COM= (ifamdl04, 262) 'someone says that it takes less time //' (5) *LIA: ma se ti dico /=SCA= che lei /=SCA= tutte le mattine /=SCA= con uno straccetto hhh //=COM= (ifamcv01, 748) 'but if I tell you / that she / every morning / with a rag //' Since they are formed by a single IU, choosing simple and simple scanned utterances for our analysis was meant to use the smallest unit of reference for speech.
In both groups of utterances (simple and simple with SCA), we searched for complex syntactic structures (i.e. with a verbal head, like complex sentences, cleft sentences and so on) in both languages.Our expectation was to find a higher rate of complex syntactic structures per simple utterance (i.e.linearized complex syntactic structures15 ) in BP than in IT, in light of the the highest rate of linearization already found by Bossaglia (2014) for complement clauses.
Finally, the amount of phonological syllables per simple utterance (understood as the smallest information/prosodic unit) was calculated, expecting, once again, to find a higher rate in BP than in IT.Simple scanned utterances were excluded in this last step: although they represent the smallest unit of reference from an informational point of view, they do not from a prosodic perspective, because they are made up by more than one tonal units.

Results and data analysis
The utterances gathered according to the mentioned criteria were 1,369 for BP (1,141 simple + 228 simple with SCA) and 1,413 for IT (998 simple + 415 simple with SCA), as it is shown in Table 2 below:  At the outset, it seems that in IT there's much more use of simple scanned (29.4% on the total amount of utterances) utterances than in BP (16.7%), as it is shown in Figure 1 below: Such a difference turns out to be more interesting when observed within each interactional typology.In facts, the mentioned greater amount of simple utterances over simple scanned utterances is detected within conversations, dialogues and monologues in BP, but only within conversations and dialogues in IT.Actually, in monological interactions IT displays a much greater amount of simple scanned utterances (64.2%), as it can be observed in Figure 2: It is worth noticing that in monological interactions BP makes a higher use of simple scanned utterances too.This fact has to do with the interactional typology itself, that is characterized by a lesser degree of actionality (there's oly one active participant in monologues) with respect to dialogues and conversations.The counterpart of such a reduced actionality is a higher textual complexity in monologues, which usually are narratives or argumentative texts.Therefore, the fact that in IT data such a textual complexity is mainly performed through simple scanned utterances could already be, at the outset, an indicator of the fact that IT tonal unit presents a lesser capacity than BP one.
4.1 Complex syntactic structures per simple utterance in BP and IT The search for complex syntactic structures within the two data samples was meant to verify if textually structured locutive content could easily be performed within a single tonal unit in both languages.The expected higher rate of complex syntactic structures per simple and simple scanned utterance in BP was verified in our data, since 16.7% (229) of BP utterances, over the total amount of them, host complex syntactic structures, while only 10.7% (151) do in IT: BP data display a higher rate of complex syntactic structures per utterance not only globally, but also for each interactional typology (with the exception of monologues), as it is shown in Figure 3: As Figure 3 shows, in monological interactions IT displays a strong preference for complex syntactic structures to be performed through simple scanned utterances, while in conversations and dialogues the percentages are quite similar (23 simple vs. 27 simple scanned utterances in cv, 15 simple vs. 10 simple scanned utterances in dl: the discrepancy in these interactional typology is not really significant).
If one looks at the data considering the relative frequency of utterances hosting complex syntactic structures with respect to the total amount of simple and simple scanned utterances separately, other observations are possible: As it is shown in the figure above, both languages display a higher relative percentage of utterances hosting complex syntactic structure within the group of scanned utterances (percentages are calculated on the total amount of simple and simple scanned utterances, respectively).
Scanned utterances' capacity is increased by the addition of one or more tonal units, hence they are at the outset capable of hosting a bigger amount of linguistic content than non-scanned ones.Therefore, a higher relative number of utterances with complex syntactic structures was expectable within the scanned utterances group in both languages.
However, it is clear that within BP simple utterances the amount of nonscanned utterances hosting complex syntactic structures is pretty higher than in IT: 13.2% vs. 5.2%, 16.5% vs. 3.5%, 11.9% vs. 7.1% in conversations, dialogues and monologues, respectively.
Within the sample of utterances hosting complex syntactic structures, the two languages display opposite trends in what concerns the privileged means of conveying such structures, i.e. non-scanned utterances for BP (168), and scanned utterances for IT (103): However, one must have in mind that in monological interactions IT data display a much broader use of scanned utterances in conveying complex syntactic structures than in the other interactional typologies.Once again, monologues turn out to be the privileged locus for textual complexity.
Finally, our data showed that BP simple utterances display a greater amount of complex syntactic structures not only at a global level, but also within nonscanned and scanned utterances, as it is shown in Figure 6: It is noteworthy that the only context in which IT displays a presence of complex syntactic structures comparable to that of BP is the group of simple scanned utterances, i.e. simple utterances whose capacity has been increased by the addition of more tonal units.All the facts illustrated in this section, therefore, seem to be in line with our hypothesis.

Phonological syllables per simple utterance in BP and IT
As we mentioned, only non-scanned utterances were used in order to calculate the phonological capacity of BP and IT simple utterances.As a first confirmation of our hypothesis, BP displayed a higher maximal number of phonological syllables per utterance than IT: 30 vs. 23, as examples ( 6) and ( 7) shown below 16 .
(6) *BAL: ela não conseguiu carregar a pilha por causa que isso aqui tava em duzentos-e-vinte //=COM= (bfamdl02, 37) 'she couldn't charge the battery because this thing was set at 220 volts' 16 Phonological syllables were counted manually.As a methodological choice, aphetisms (e.g.BP tá for está "(he/she/it) is") were counted as whole forms (i.e.including the lost syllable in the calculation as well), with the exception of the second-person singular personal pronoun cê "you" (whole form você), which is completely grammaticalized in spoken BP (see Ferrari 2015 for details).
(7) *PAO: < è il ventitreesimo anniversario >17 dell' assassinio di Aldo Moro //=COM= (ipubmn01, 117) 'It's the 23rd anniversary of Aldo Moro's murder' Besides, in each interaction typology BP displayed not only a higher maximal number of phonological syllables per utterance, but also higher mean, median and mode numbers, as is shown in Table 4: This first empirical test showed that BP tonal unit seems to be capable of hosting a higher number of phonological syllables than IT.

Conclusions
In this preliminary study on the prosody-syntax interface in spoken BP and IT we tried to verify the hypothesis that the difference in speech rhythm between the two languages could have consequences in the way their spoken syntax is structured.We assumed that the stronger preference for linearization BP displayed in the codification of complement clauses with respect to IT (Bossaglia 2014) could be related to the more stress-timed nature of BP rhythm: this is because a more stress-timed rhythm allows the tonal unit to host a higher number of phonological material than a more syllable-timed one.
The search for the rate of complex syntactic structures per simple (and simple scanned) utterance within the two minicorpora led to satisfactory results, since we could verify that BP actually presents a higher rate than IT.This means that within the domain of linearized syntax (i.e. a single IU, the COM) BP seems to be much more capable of performing complex syntactic structures than IT.One more verification of this fact was that IT displays a comparable 'ability' only in scanned utterances (see Figure 6 above), i.e. utterances whose capacity is increased by the addition of one or more tonal units.So, we can admit that BP is characterized by a higher predisposition for linearized syntax than IT, probably due to its rhythmic characteristics.
Such correlation between prosodic features and syntax was confirmed also by our calculation of the number of phonological syllables the two languages present in their non-scanned simple utterances.As it was expected, BP appears to have higher mean, median and mode numbers in each interactional typology (with the only exceptions of lower mode number in conversations, and equal median and mode numbers in monologues), beside higher maximal number of phonological syllables per utterance.
Being these first results so promising, we intend to advance this study by searching for the ratio between phonological and phonetic syllables (viz.V-V units, see Barbosa 2000Barbosa , 2006) ) in BP and IT simple utterances, which could represent a more reliable measure for their tonal units' capacity.
the realization of explicit complement clauses in spoken Brazilian Portuguese and Italian is analyzed, based on two comparable subcorpora of the C-ORAL-BRASIL and Italian C-ORAL-ROM (available online in the DB-IPIC -DataBase for Information Patterning Interlinguistic Comparison platform 8 ; see Panunzi & Mittmann 2014 and section 3.2.1 for details).

Figure 1 .
Figure 1.Simple and simple scanned utterances in BP and IT IPIC minicorpora

Figure 2 .
Figure 2. BP and IT simple and simple scanned utterances per interactional typology

Figure 3 .
Figure 3. Simple and simple scanned utterances hosting complex syntactic structures in BP and IT

Figure 4 .
Figure 4. Simple and simple scanned utterances hosting complex sentences in BP and IT

Figure 5 .
Figure 5. Typology of utterances hosting complex syntactic structures in BP and IT

Figure 6 .
Figure 6.Complex syntactic structures in BP and IT simple (scanned) utterances

Table 1 .
Size of BP and IT DB-IPIC minicorpora

Table 2 .
Simple and simple scanned utterances in DB-IPIC BP and IT minicorpora

Table 3 .
Complex syntactic structures per simple and simple scanned utterance in BP and IT

Table 4 .
Phonological syllables per utterance in BP and IT