This presentation focuses on two aspects of spoken language corpora which may be relevant in designing sign language corpora and comparing oral vs. sign languages, namely what the relevant units of analysis of spoken language are and how to design corpora for a correct sampling of spoken language variation. Both questions will be addressed taking as point of departure the experience of LABLITA in storing and describing child and adult spoken language corpora and the theoretical approach to the study of intonation developed in the laboratory by E. Cresti.
The spoken domain shows different structural differences when compared to that of written language. Undoubtedly, one of these is the presence of intonation as a necessary component for the realisation of speech. Intonation has a minimum restitution in the conventional writing code, and even if the latter is a derived diamesic system, it is characterised by textual organisation, sentences based on syntax, which is different from speaking. The previous assertion deserves some commentary. In particular, the basic linguistic notion useful for describing spoken language is quite different from that used for written language. The flow of speech is necessarily divided into utterances and the relation of the notion of utterance to syntax and semantics is not obvious. (See Harris, 1951; Bar-Hillel,1967; Sornicola, 1981; Miller & Weinert, 1998).
In particular, spoken language utterances frequently do not coincide with the structure of a clause, therefore the process of segmentation of speech continuum is frequently arbitrary (Moneglia & Cresti,1997). Examples 1 and 2, make the point clear. 1 In the first example a student addresses the question of whether or not her professor needs a photocopy like other people in the class. The second is a dialogue between two workers fixing a car. Both are extremely common empirical situations dealing with spontaneous speech. The flow of speech turns out to be continuous (no pauses) and the process of its partition is radically underdetermined.
The audio files can be roughly transcribed in the CHAT format as follows. The reader must click on the highlighted letters for the multimedia link to the audio files:
*SUS:
lei gliene serve una anche a lei ? una in più o no no lei ha questa //
you do you need one you too ? one more or not no you have this one //
%add:
to the professor
%sit:
while SUS is going to go for a photocopy
In example 1 a clear question intonation allows the dialogic turn to be divided into two parts. However some other segmentation is needed in order to allow full interpretation of the text, which still remains obscure. For example in the first segment we do not know the linguistic status of the first pronoun "lei”. Could it be considered a nominal utterance or not? In the second segment, given that "you have this one” should be a sentence, we do not know the structure of the items "una in più o no no”. In both cases there is no verb on the surface leading to a possible solution, and the mapping of utterances on clause structure turns out to be arbitrary. Moreover, given that there is no pause in the flow of speech, any definition of utterance like "silent to silent” will lead us to interpret 1) as just one single utterance.
Many people who study spoken language, take the view of eliminating these materials in order to avoid arbitrary decision-making processes (see for example Biber et alii, 1998). The words with no clear clause structure should be considered "fragments” as in the square brackets below:
Example 1')
*SUS:
[lei] gliene serve una anche a lei ? [una in più o no no] lei ha questa //
[you] do you need one you too ? [one more or not no] you have this one //
Taking this view, more than 40% of spoken language locution turns out to be made up of fragments. That is an extremely large quantity of spoken language that will be considered out of scientific analysis.
The second example presents similar problems. It also shows, however, that the process of separating sentences from the collection of fragments happens to be arbitrary:
| *MIC: | che macchina l' è codesta Punto ? |
| which [kind of] car is that Punto ? | |
| *OPR: | Punto milledue mi guardi ? |
| Punto 1200cc can you check it out for me? | |
| *MIC: | i' ché ti guardo guarda come tu se' brutto costì sensore temperatura acqua raffreddamento motore |
| what I schoud check for you look how bad you are there sensor temperature water cooling engine |
For example, the first dialogic turn could be parsed in just one utterance with the word "Punto” within the clause. However it will be parsed in two utterances by considering it as a first clause followed by a nominal utterance. Both are possible clause structures in Italian. The other two dialogic turns present similar problems. In particular the last turns should be comprised of fragments in almost all the final part, which is long and obscure:
2')
| *MIC: | i' ché ti guardo - guarda come tu se' brutto [costì sensore temperatura acqua raffreddamento motore ] |
| what I schoud check for you look how bad you are there sensor temperature water cooling engine |
From my point of view the basic unit of analysis of spoken language cannot be the concept of clause since it is both too weak and too strong in capturing evidence from spoken language corpora.
The problem may be approached on empirical grounds considering that spoken language events have a prosody which envelops each utterance and specifies to our perception the illocutionary force of the utterance itself, namely the communicative value of a single linguistic action (Austin, 1963). From this point of view the textual structure of spoken language turns out to be comprised of speech acts whose quality is mainly defined by intonation. Written language crucially lacks both properties.
This in not new. It is quite obvious that intonation says something about the action quality of the utterances. At the same time it is also obvious that intonation can allow the interpretation of even a single word.
The approach we developed for the analysis of spontaneous speech links the previous general remarks to three topics regarding intonation and its linguistic role: 1) the idea that utterances are systematically parsed in groups of tone units (intonation pattern), the type of which is discriminated at a perceptive level (’t Hart et alii, 1990); 2) the notion of utterance has prosodic constraints (Crystal, 1975; Halliday, 1976); 3) the functional value of prosody is predictable (Cresti, 1994).
The melodic pattern which scans a possible utterance can be simple, composed of a single tone unit, or complex, in which case it is made up of two or more tone units linked melodically together. The possibility that non terminal tone units occur within an utterance (Pierrehumbert, 1980) corresponds to the scanning of an utterance by means of a complex pattern. The following, for example, is a possible patterned declarative sentence in Italian:
On the other hand, an important tradition of intonation studies (Karcevsky, 1931; Crystal, 1975) has always highlighted the fact that there is no such thing as an utterance without a profile of terminal intonation. Considering the question on theoretical grounds this classic statement means that we cannot get two utterances in the same prosodic contour. Of course this constraint holds only for the notion of utterance, which must have a prosodic counterpart. However, it does not hold for the notion of clauses and sentences which are independent of prosody.
At the same time it has been noted that, within the possible tone units of an utterance, the tone information which enables one to identify the illocution, or modality, of the utterance lies in a specific tone unit (Martin, 1978). For example in 3) the hat movement in the second unit on the right.
There are interesting consequences of this statement. For example, the following performance of the sentence "Carlo va a Roma”, where the two prosodic units have both received a hat movement on the last tonic syllable, must be considered a sequence of two separate utterances for prosodic reasons. We have strong perceptual evidence for it that the reader can hear by listening to audio file 4)
Given the above, one can deduce that the tone units comprising an utterance must not be considered mere scan units since they have a structural role. Tone structure conveys informative values (Bally, 1950; Halliday, 1976). These values are the object of judgements based on the perception of prosodic cues.
By considering these properties, we have arrived at a meaningful generalisation for analysing spontaneous speech related to the value of prosody. The idea that in each parsed utterance there is always a particular tone unit with the structural role of expressing illocutionary value which can be generalised in the analysis of spoken texts. The utterance limit can be identified each time the prosody makes it possible to perceive the completion of a speech act (Cresti, 1994; 1996) allowing a pragmatic interpretation of the text.
For example, given that we have perceptual evidence that both prosodic units lead to the accomplishment of a speech act in 4), we also have perceptual evidence that only the second unit accomplishes a speech act in 3). The first unit cannot be interpreted as an utterance because of its intonation. The reader can easily obtains the perceptual result listening in isolation to all segments in the audio files.
It is has been noted that in spoken language utterances performed in prosodic patterns have one, and only one, tone part which contains prosodic cues allowing its pragmatic interpretation also in isolation. We call such units comment.2 This unit of information is necessary in order to have a speech act. The other possible elements of the prosodic pattern do not have these cues and therefore cannot be interpreted as independent utterances, but they are necessarily part of the linguistic action performed by the comment unit.3
In conclusion an illocutionary criterion permits one to perceptively recognise the scan unit which contains the prosodic indices of the linguistic action, and, at the same time to distinguish such units from other information units of an utterance, which are optional and can never be pragmatically interpreted.4 Generally speaking we have a method for segmenting the speech continuum in utterances.
The previous relation between elements of a prosodic pattern holds in natural languages like Italian, Franch, or Dutch apart from considerations regarding syntax. These relations depend on the action nature of utterances of spoken language and have a surface linguistic signal in prosodic patterning. The following principles summarise the notions useful for a correct identification of the basic units of spoken language:
Interpretation: illocutionary force specifies how to relate locutive content to the pragmatic context of the utterance.
Utterance: an utterance is each meaningful expression (of each type) which can be interpreted in a pragmatic context.
Illocutive criterion: prosodic cues specify at the perceptual level the accomplishment of a speech act. Therefore speech flow can be segmented into utterances every time a linguistic action is judged from prosodic patterning.
Comment principle: illocutionary cues are placed on one, and only one, tone unit in a patterned utterance. According to the illocutive criterion we can discriminate whether or not a tone unit bears illocutionary information; i.e. whether it can be interpreted pragmatically or not.
On the basis of illocutionary criterion the transcription of spoken language must take into account, as a basic level of tagging, both intonation and the segmentation of speech flow into speech acts. Annotation of the utterances and their prosodic parsing is a basic part of the transcription of spoken language; it only allows one to see the sound partitions operated by intonation, which help us interpret the spoken text (Moneglia & Cresti, 1997). This annotation is very simple; for instance a simple bar "/" for each not terminal tone unit, double bar "//" to mark the end of an utterance, question mark "?" for end of interrogative utterances, three dots "…" for a suspended utterance. The following signs can easily be implemented in
| Terminal tone units // ? !... | |
| Not terminal tone units / |
The previous criteria have been successfully applied to both corpora of spontaneous adult speech and infant speech (see Cresti, (in press); Cresti & Moneglia (1993)). This has already permitted the signalling of their organisation in utterances and essential prosodic indices in the transcription of the texts. The following is the transcription with prosodic punctuation and segmentation of utterances in 1 and 2, in accordance with the illocutionary criterion.
Example 1'')
| *SUS: | lei / gliene serve una anche a lei ? una in più / o no ? no // lei ha questa // |
| you / do you need one you too ? one more / or not ? no // you have this one // |
Example 2'')
| *MIC: | che macchina l' è / codesta / Punto ? |
| which [kind of] car is / that one / Punto ? | |
| *OPR: | Punto / milledue // mi guardi ? |
| Punto / 1200cc // can you check it out for me? | |
| *MIC: | i' ché ti guardo // guarda come tu se' brutto / costì // sensore temperatura acqua // raffreddamento motore // |
| what I should check for you // look how bad you are / there // sensor temperature water // cooling engine // |
The reader can verify the perceptual relevance of previous segmentation by listening in isolation to each prosodic group in the multimedia files. Accordingly to the illocutionary criterion and the comment principle only one segment of each patterned utterance can be interpreted as a speech act, while all simple utterances can receive possible interpretation in isolation.5
In conclusion we have briefly shown that the partition of spoken texts in utterances relies on a proper analysis of prosodic cues with their functional values rather than the arbitrary process of assigning a syntactic structure to the spoken text. Apart from the fact that the text suddenly becomes interpretable, we have just proposed a criterion for segmenting spoken language which has consequences on the evaluation of spoken texts, for both qualitative and quantitative analysis. For example there are no fragments in the previous text, which is properly patterned by intonation and divided into utterances allowing full interpretation of both complex and simple prosodic structures. The idea of a large proportion of fragments in spoken language is the consequence of assigning a clause structure without considering the action value of utterances in spoken language.
The choice of a basic level intonation tagging might appear to be an oversimplification compared to different transcription systems of intonation and especially compared to ToBi, which has been frequently proposed as a standard (See Gibbon et alii 1997). Some clarifications seem to be in order. The annotation of prosodic parsing it is not a transcription of the intonation like ToBi, or MARSEC, in the sense that it does not load external signs, such as numbers or indices or letters onto the text so as to reproduce in a ciphered manner the characteristics of the intonation profiles. Annotation of prosodic parsing is a definition of the prosodic units which have functional value not an evaluation of prosody according to a pre-theoretical phonological typology (see Pierrehumbert 1980; Ladd 1996). The demarcation of utterances and their prosodic parsing according to the illocutionary criterion, which are annotated in the transcription, will constitute the functional correlation of any later analysis of prosodic cues.
It may be interesting in the context of this conference to underline that the structural link between intonation and utterance as an action notion is strongly based in the process of first language acquisition and therefore, from a theoretical point of view, the previous theory is independently motivated at the language learning level (see Moneglia 1994).
We have tried to show in many works that the basic milestone of combinatorial language is the capacity of the child to integrate more than one locutive expression in a single language action, and that prosodic cues mark this process in child language (See Moneglia & Cresti 1993 and references in the LABLITA web site). In particular, in the early transition from the one word period to complex utterances, each word in a child’s dialogic turn allows its interpretation as a single speech act. Such interpretability is a function of prosody that permits each single word to correspond to an action. For example the following dialogic turn is certainly a sequence of distinct acts because of the prosody of each word. Following the illocutionary criterion each word can be interpreted in isolation as a separate utterance because of its intonation. The reader can again verify by listening in isolation to each segment in the multimedia file..
| GIU (1;.9.23) | lilla // lilla // totta // |
| clock // clock // broken // | |
| %sit: | playing with the clock |
Despite the fact that possible cognitive relations and even a predicative structure could be assigned to the sequence of words "clock – broken”, there cannot be one utterance as far as the two words perform more than one illocutionary act. The phase of illocutionary independent words in language acquisition appears as a universal character of language learning.
In order to reach the stage of combinatorial language, the child must learn some prosodic tools to be able to put together two locutive elements in the same act. The child GIU demonstrates this ability in his longitudinal protocols after one month. The following are examples of the two basic strategies a child can follow in this task 6 : two words in the same prosodic envelope (linear strategy) and two words in a prosodic pattern (informational patterning strategy):
| GIU (2;0.20) | quetto / chiae // |
| this / key // |
| GIU (2;0.20) | mimma // etti mimma // |
| child // this child |
Again the reader can easily verify our interpretation by listening in isolation to each word of the multimedia files. The constraint of giving an interpretation to each element as a single act has disappeared. However, on the contrary, by following the comment principle one, and only one, of the two tone elements can be interpreted in isolation.7
Summarising, in the framework we have just designed, intonation is not the prosodic interpretation of the syntactic structure of an utterance. By means of intonation language events, namely utterances, are attributed the value of an action. The detection of prosodic cues, which are the object of perception, allows explicit criteria for the analysis of oral texts to be defined. By comparing sign language, which is an action-language, with verbal language, it would be nice to be able to offer a proper comparison with spontaneous speech, which we have tried to show is an action language too.
In connection with the study of the specific structure of spoken language it is also important to set up corpora, suitable for comparison from the point of view of corpus design. In fact spoken language has peculiar linguistic properties with respect to its variation at many levels: structure of dialogic situations, sociolinguistic features, goals and regulations of the linguistic process. According to this variation many linguistic properties of spoken language such as MLU, omission and variation of morphemes, word frequency, may vary a lot from text to text.8 Therefore all the values which are necessary in order to compare spoken language codes are a function of the pre-theoretical selection of contexts in corpus design. What texts would better testify essential properties of spoken language and what range of variation should those properties have?
I would like to finish this brief talk by presenting the structure of the LABLITA corpus and some consequences that the variation of text may have on inducing the general character of spontaneous speech from corpora.
Since the beginning of the 70s, LABLITA has been collecting a series of corpora of spontaneous spoken language in order to create some databases for the study and identification of linguistic properties of spontaneous spoken language and in particular its intonation. One of the main issues was to distinguish informal everyday spontaneous speech, like in the above examples, from formal speech. There are four LABLITA corpora, all transcribed and electronically recorded in CHAT format (MacWhinney, 1994): 9
In the LABLITA corpus variation is the result of parameters that are considered significant in a large literature (See Bilger, 1997; Labov, 1966; Biber, 1994; 1988; Berruto, 1987; Gadet; 1996). The first two parameters constitute the structural variation of the corpus: (1) dialogical structure (monologues, dialogues, conversations), (2) social domain of use (family, private life, public life, media productions). The other parameters, namely speaker’s age (3), education (4) and job (5) vary in a transversal way in relation to structural parameters showing the diaphasical and diastratical variations of language use.
The following matrix shows the variation of dialogical structure on the vertical axis and the variation of social domain of use on the horizontal. A more accurate classification of each horizontal field distinguishes if the texts, all spontaneous, are to some level regulated or not.
|
FAMILY
|
PRIVATE
|
PUBLIC
|
||||
| Free |
Regulate
|
Free | Regulate | Free | Regulate | |
| Dialogue | ||||||
| Conversation | ||||||
| Monologue | ||||||
| 20% | 35% | 15% | 10% | 20% |
Click here for the list of texts of the LABLITA
The LABLITA corpus assures a data base for the distinction between formal and informal speech through scanning the various domains of spoken language use. For example texts which are "public” and "regulated” frequently will be more formal than texts which are family conversations. At the same time, as the percentage clearly shows, the corpus cover a huge proportion of spontaneous/informal speech with respect to formal speech. In so doing the LABLITA corpus tries to highlight the central area of its use, assuring a proper base for induction. This choice is essential for a better understanding of the specific properties of spoken language and to identify its quantitative limits. It is also important considering that in general we think of language as an "ideal written language”, potentially with no limits.
In particular, from a quantitative point of view, the difference between the basic properties of formal and informal speech is not well known. The following is an example of a formal spoken text (taken during a seminar on child language at the University of Florence). Its transcription was obtained again by applying the illocutionary criterion. The difference between 1) and 2) is impressive, but what is it about exactly?
| *CEC: | brevemente / quello che era stato detto / è che c' era sta-to / un incremento elevatissimo di protoforme / e |
| [ briefly / what we told / is that it had occurred / a huge increase of proto-morphemes / and | |
| infatti lo si vede dal grafico / a ven-ti mesi / c' è proprio un picco / e / s' era parlato / appunto / sulla / | |
| actually we see it from the graphic / at twenty months / there is right a rise / and / we talked / in facts / about some / | |
| con-tem-po-ra-nei-tà / anche / dell' esplosione lessicale / ma su questo / i dati devono essere riguardati / an-co-ra devo fare ... | |
| contemporaneity / also / with lexical spurt / but about this fact / data must be revised / I still must do...] |
This distinction can be demonstrated quite easily once we get reliable measures of spoken texts not only at the word level but also at the level of structure. From this point of view the approach we have briefly presented leads to immediate results. Demarcation of the utterances and their prosodic patterning can receive automatic verification giving rise to a series of basic speech measurements. Measurements are based on concepts of dialogic turn, informative vs. non-informative words, tone units and utterances. The following is a simplified version which takes into account four parameters.
On the basis of the previous indices we arrive at a quantitative description of the basic structure of the single texts. By comparing these numerical data it is possible to define a first classification of spontaneous texts where the distinction between formal and informal speech may be better appreciated. We have verified (Tizzanini 1999) the descriptive capacity of the measurement on a sub-corpus of six texts, four of which (a family conversation, a country conversation, a conversation between work colleagues and another between university students) brought together in Group A, are marked by a strong degree of spontaneity. The other two texts (a university lecture and a radio interview) are formal texts (Group B). The following table summarises some of the results obtained:
| MLT
(utter./turn) |
MLU
(words / utter.) |
MTU
(tone unit/utter.) |
MLTU
(words/tone unit) |
|
| Group A | 2.24 | 6.03 | 1.98 | 2.84 |
| Group B | 4.04 | 11.53 | 4.22 | 2.58 |
The quantitative difference between the two groups is immediately obvious. The most spontaneous texts (A) are generally characterised by quick exchange turns (the sign of a very restricted dialogue exchange) and short utterances (sign of a simple informative structure). On the contrary, the Group B texts present longer exchange turns and more complex utterances both at the word level (MLU) and prosodic level (see Cresti & Scarano 1998) (these factors are highlighted by the values which are almost twice those of the Group A texts).10 Hence, example 7 is an instance of this tendency: just one utterance, comprised of 53 words divided into 17 tone groups.
I would like to end the talk by stressing that such an easy result strictly depends on the choice of the basic unit of analysis of spoken language we adopted: the action notion of utterance. This result cannot be obtained by describing the structure of the previous spoken texts through the concepts of sentence or clause: 7) is made up of six/seven sentences and their length is irrelevant as a measure of spoken text.
Austin, L. J. (1962) How to do things with words, Oxford University Press, Oxford.
Bacelar Nascimento M.F. (1998) Quelques resources linguistiques de base pour l'étude des variétés du portugais contemporain, Revue Française de Linguistique Appliquée, Vol. III – 1.
Bally C. (1950), Linguistique générale et linguistique française, Francke Verlag, Berne.
Berrendonner A. et alii (1983), Principes de grammaire polyectale, Presses Universitaires, Lyon.
Berruto G. (1987), Sociolinguistica dell'italiano contemporaneo, NIS, Firenze
Bettencourt Gonçalves, J. (to be published) Português Falado: variedades geográficas e sociais em Actas do Colóquio Internacional "A investigação do Português em África, Ásia, América e Europa", Instituto Ibero-Americano, Berlim.
Biber D. (1988),Variation across speech and writing, Cambridge University Press, Cambridge.
Biber D. (1994), Rapresentativeness in corpus design, in Linguistica computazionale IX-X.
Biber D., et alii (eds) (1998), Corpus linguistics: investigating language structure and use, Cambridge CUP.
Biber D., et alii (eds) (1999), The Longman grammar of spoken and written english, London, Longman.
Bilger M. (1997), Corpus de portugais et d'espagnol, in Revue de l'Association Français de linguistique appliquée, 2.
Blanche- Benveniste C. (ed) (to be published), Corpus du Français parlé. Echantillonages. Champion, Paris
Blanche-Benveniste C. (1997), Approches de la langue parlée Français, Ophrys, Paris.
Blanche-Benveniste C. Bilger M., Rouget c. Einde van den K. (1990) Le Francais parlé, Editions du CNRS, Paris.
Bortolini U. & Pizzuto E. (eds.) (1997), Il progetto Childes Italia, Del cerro, Pisa.
Cresti E. (1994), Information and intonational patterning in Italian, in B. Ferguson, H. Gezundhejt, Ph. Martin (eds), Accent, intonation, et modéles phonologiques, Editions Mélodie, Toronto pp. 99-140
Cresti E. (1996), Speech acts units and informational units, in E. Fava (ed) Speech acts and linguistic research, Nemo Padova, pp. 89-107.
Cresti E. (1998), Critère illocutoire et articulation informative, in M. Bilger (ed), Acts du colloque Questions de méthode dans la linguistique sur corpus, Perpignan, Champion, Paris.
Cresti E. . (to be published), Corpus di italiano parlato. Campioni, Accademia della Crusca, Firenze
Cresti E. & Moneglia M. (eds)(1993), Ricerche sull'acquisizione dell'italiano, Bulzoni, Roma
Cresti E. Scarano A (1998) Sur la notion de Le parlé spontané, in in M. Bilger (ed), Acts du colloque Questions de méthode dans la linguistique sur corpus, Perpignan, Champion, Paris
Crystal D. (1975), The English tone of voice, Edward Arnold, London.
Crystal D. (1985), Prosodic Development, in Fletcher P. e Garman M., Language Acquisition, Second edition, Cambridge, Cambridge University Press.
De Mauro T. Mancini F. Vedovelli M. Voghera M. (1993) Lessico di frequenza dell'italiano parlato, Etass Milano.
Dictionnaire de Linguistique (1973), Larousse, Paris.
Fasold, R., (1990) The sociolinguistics of Language, Oxford.
Firenzuoli V. (1999), Correlazioni e criteri di variaizone tra articolazione dell'informazione e morfosintassi in un corpus di italiano parlato, submitted to Romanische Forschungen.
G. Berruto (1987), Sociolinguistica dell'italiano contemporaneo, NIS, Firenze
Gadet F. (1996) Variabilité, variation, varieté: le Français d'Europe, in French Language Studies, 6.
Gadet F. (1989), Le Français ordinaire, Armand Colin, Paris.
Gibbon, More, Winski (eds.) (1997) The handbook of Standards and Resources for Spoken language Systems, Mouton de Gruyter, Berlin. Halliday M.A.K. (1976) Sistem and function in Language, (Kress. G. Ed) Oxford
Harris, Z (1951) Methods in structural linguistics, Chicago
Hokett,Ch (1963) the problem of universals in Language in Greeberg (ed) Universals in Language, Cambridge Mass 1-29
Hymes D. (1962), The Ethnography of Speaking, in T. Gladwin, W. C. Sturtevant (a cura di), Anthropology and human behaviour, Anthropological Society of Washington, Washington D.C.
Labov W., (1966) The social stratification of English in New York City, Washington D.C., 1966.
Ladd R.D. (1996), Intonational Phonology, Cambridge University Press, Cambridge.
MacWhinney B. (1994) The CHILDES project: tools for analyzing talk, Lawrence Erlbaum Associates, Hillsdale, New Jersey.
Martin Ph. (1978) Questions de phonosyntaxe et de phonosémantique en Française, in Linguisticae Investigationes, 2, pp 93-126.
Miller, J e Weinert R., (1998) Spontaneous spoken Language: Syntax and Discourse, Claredon Press, Oxford.
Moneglia M. (1994b), Intonation, the property of distantiation and the mastering of complex utterance in Italian, Quaderni del Dipartimento di linguistica, Università di Firenze, V.
Moneglia M.(1994) The ontogenetic foundation of informational patterning. In B. G.Ferguson, H. Gezundhajt, P. Martin (eds.), Stress, intonation and phonological models pp.155-186.
Moneglia M., Cresti E. (1993) "Formazione dell'atto linguistico complesso e intonazione: l'ontogenesi delle relazioni informative in italiano", Cresti E., Moneglia M. (eds.), Ricerche sull'acquisizione dell'italiano, Roma, Bulzoni.
Moneglia M. Cresti E. (1997), Intonazione e criteri di trascrizione del parlarto, in U. Bortolini E. Pizzuto (eds), Il progetto CHILDES Italia, Del Cerro Pisa, pp. 57-90
Nespor M. Vogel I. (1986), Prosodic Phonology, Foris, Dordrecth.
Pierrehumbert J.(1980) The Phonology and Phonetics of English intonation, Cambridge, MIT Press.
Sornicola R. (1981), Sul Parlato, Il Mulino Bologna.
't Hart J., Collier R., Cohen A. (1990), A perceptual study of intonation, CUP, Cambridge.
Tamburini G. (1994), Tone units and local F0 phenomena, in B. Fergusson, H. Gezundhajt, P. Martin (a cura di), Stress, intonation and phonological models, Editions Melodie, Toronto.
Tizzanini, G.(1999) Articolazione dell'informazione in corpus di italiano parlato. Unpublished master thesis LABLITA University of Florence.
Voghera M. (1992), Sintassi e intonazione dell'italiano parlato, Il Mulino, Bologna
1 Examples are taken from two texts in the LABLITA corpus of adult spoken Italian, see below for details (back to text)
2The term, introduced by Hockett, is defined here prosodically (See Hockett, 1963) (back to text)
3We are not going to discuss here the role of all units of information that are not a comment unit. The informational relation topic comment (Hockett, 1963) is also conveyed by intonation. For this notion see the theory of informational patterning (Cresti 1996; Tamburini 1994). (back to text)
4The results obtained on the basis of the application of illocutionary criterion are substantially confirmed in the macro-syntactic theory (Blanche-Benveniste 1990; 1998; Berendonner, 1983), for which in spoken language the syntactic "noyau” coincides with the tone unit with illocutionary value.(back to text)
5 The first "lei” in 1) and the word "punto” in the second dialogic turn of 2) should be considered topic units of information in the Theory of informational patterning and, by necessity, cannot receive a pragmatic interpretation as a speech act. (back to text)
6We verified the strategies in a large series of longitudinal studies. See child language corpora of LABLITA listed in http://lablita.dit.unifi.it. (back to text)
7 Notice that the comment principle also applies to linear utterances in early language acquisition, while it cannot be the case in adult language. That is the main transitional character in the acquisition of prosody (See Moneglia, 1994). (back to text)
8The problem of designing a reference corpus for spoken language is discussed with respect to present corpora of spoken Italian, in particular the LIP corpus, and the LABLITA corpus (see De Mauro et al 1993; Cresti (in press). (back to text)
9These corpora probably constitute, altogether, the biggest database presently available on spoken Italian. For all the corpora the audio is available on DAT cassettes, audiocassettes, magnetic tape. The sound source of spontaneous adult spoken Italian corpus is stored on CD in wav format. The acoustic quality is that which is possible in environmental recordings which are often concealed. Corpora are accessible for scientific purposes on explicit agreement conditions. (see Moneglia, 1998 and the web site http://lablita.dit.unifi.it for a detailed description) (back to text)
10 It is worth noting that the average length of the tone unit remains more or less constant in all the texts. Therefore, it is easy to deduce that the tone unit length is due not so much to informative needs, as to natural execution needs.(back to text)
Posted: 15.03.2000