Quelles expériences pour étudier le sens de la prosodie? J.-M. Marandin LLF (CNRS & Paris Diderot)

Intention Partager mon expérience de monteur d’expérience pour étudier le sens de la prosodie. In progress ! (Diapositives en français et anglais)

2

Introduction 1/5

Introduction Point de départ: Les difficultés liées à l’interprétation des données obtenues grâce à une expérience de production à base de lecture de phrases (script-based production experiments). Exemple: InfoFoc(Beyssade et al. 2010)

3

Introduction 2/5

Introduction

Les réponses pour dépasser ce problème: – Expérience à base d’interaction: Map Task, – Une alternative basée sur une tache de

Répétition/reproduction en contexte (replication in Context) Exemple: Small Talk.

4

Introduction 3/5

Introduction

Dans la seconde partie de l’exposé, je dépasse le détail de la conception d’expérience pour me focaliser sur la tache elle-même: répéter en contexte. Je prendrai deux exemples d’expériences basées sur des taches de jugement: – En syntaxe: Dative alternation (Bresnan et al.) – En prosodie: Topicalizing Reprises (Marandin et al.)

5

Introduction 4/5

Introduction

Deux conclusions: 1. Une conclusion pratique: les expériences sont non seulement possibles, mais donnent des données pertinentes. 2. Une conclusion théorique: ces résultats donnent de l’intérêt aux théories qui définissent la compétence en termes de réproduction (entre autres, Keller 1994). 6

Introduction 5/5

Plan •  Partie 1: La répétition/reproduction comme stratégie expérimentale 1.1. Problèmes avec les expériences à base de script lu. (Ex.: InfoFoc). 1.2. Replication-based experiment (Ex.: Small Talk). •  Partie 2: Répétition/reproduction et compétence 2.1. Duplication et jugement métalinguistique. 2.1.1. Syntaxe (Ex.: Dative alternation) 2.1.2. Prosodie(Ex.: Topicalizing Reprises) 2.2. Compétence et usage du langage.

7

Part I Replication as a strategy in experiment design

1.1. Script-based production experiments

1.1. Script-based .. 1/8

The problem •  Laboratory phonology (Pierrehumbert, Beckman & Ladd 2000). •  Concern with data elicited through Script-based production experiments •  Script-based experiments: Subjects play a role reading lines in a scenario. The lines are given in order to tightly control the segmental stuff utterances are made of (syllables, types of consonants, etc.)

10

1.1. Script-based .. 2/8

InfoFoc Heuristic question: (1) How are narrow vs broad information foci prosodically marked in French? -  Information Focus corresponds to the phrases that contribute the content resolving questions -  General design: comparison of answers to wh/partial questions and answers to broad questions. InfoFoc (Beyssade, Hemforth, Marandin,& Portes 2010). 11

1.1. Script-based .. 3/8

InfoFoc. Design Subjects are presented with the description of a context (2) and a question: either a partial question (3) or a broad question (4). They are asked to read aloud the answer (5) as if they were actually participating in the dialogue. (2) Context Richard is a policeman. He has to treat various documents (films, leaflets, K7s) seized in a terrorist cache.

(3)

Le responsable :

Qu'as-tu visionné la nuit dernière ? What did you screen last night?

(4)

Le responsable :

Où en es-tu dans ton enquête ? What’s up with your investigation

(5)

Richard :

J'ai visionné les vidéos la nuit dernière. I screened the videos last night

– 14 participants, 112 target answers, 107 analyzed (5 rejected)

13

IR in declaratives. Results (1)

Pattern A Nuclear Low at the right edge of the Direct Object NP (DO)

Pattern B Nuclear Low at the right edge of the Object NP + initial accentuation of the DO

Pattern C Initial accentuation of DO + register enhancement of DO (Nuclear Low at the right edge of the sentence)

Pattern D Nuclear Low at the right edge of the sentence

14

IR in declaratives. Illustration (2) Illustration: J’ai élargi le gilet avec du velours noir. I let out the vest with black velvet.

A

C

B

D

15

Answers to partial questions

Pattern

NPA placement IR (+ register change)

A

11%

B

49%

C

23,6%

D

16,4%

Right edge of DO: 60%

Final (right edge of S): 40%

IR on DO: 72,6%

16

Answers to broad questions

Pattern

NPA placement IR (+ register change)

A

17,3%

B

13,5%

C

19,2%

D

50%

Right edge of DO: 30,8%

Final (right edge of S): 69,2%

IR on DO: 32,7%

17

IR in declaratives. Synthesis

Answers to

Placement of the nuclear pitch accent (mostly L* in the corpus)

Initial rise on DO

Right edge of Right edge of DO S Partial questions

60%

30,8%

72,6%

Broad questions

40%

69,2%

32,7%

18

1.1. Script-based .. 6/8

Script-based experiment. Diagnosis.

Recurrent situation with script-based production experiments: – Several patterns for one phenomenon, – Sharp differences in pattern frequencies. But, we are in no position to account for: – The meaning/usage of the patterns, – Their relative frequency. Why? – No idea of how the patterns are used in actual interactions, – Suspicion about their possible fakeness Hence: – A feeling of inconclusiveness and possible deception. 19

1.1. Script-based .. 7/8

Script-based experiment. Diagnosis.

Diagnosis shared by many Lab Phon linguists. “Investigators must keep questioning to what extent findings with scripted laboratory speech account for phonetic phenomena during natural, unmonitored speech acts. We also need to remain cautious upon making claims about observed prosodic patterns and pragmatics, as the script-reading task may induce particular prosodic patterns that may not be observed frequently in natural conversations.” Ito & Speer 2006 in Methods in Empirical Prosody Research 20

1.1. Script-based .. 8/8

Conclusion intermédiaire (expériences à base de script lu)

Solutions? – Abandonner ce type d’expérience (les cantonner aux seules questions de détail phonétique): --> Expérience à base d’interaction: (entre autres): Map Task. – Comparer le discours en milieu naturel et en milieu expérimental: --> Expérience à base de duplication: Rep Task. 21

1.2. Rep Task A strategy to evaluate the relevance of elicited data: replication-based experiments.

1.2. Rep Task. 1/17

Rep Task •  Leading idea: Compare the behavior of Subjects in the lab and the behavior of Speakers in actual interactions. •  Rep Task: Ask subjects to reenact a dialogue whose video and audio recording is available. Compare the renditions in the most detailed way possible. •  Illustration: Small Talk (Laurens, Marandin, Patin & Yoo, 2010).

23

1.2. Rep Task. 2/17

Small Talk Design and protocol: – 12 minutes’ dialogue taken from a 1 hour dialogue between two young men (CID corpus, Bertrand et al. 2008). – The dialogue was edited by removing disfluencies, truncated expressions or overlappings. – Small talk about being a father (dialogue goes from fathers’ attending child delivery, to the preparation of the room for the new-born, the choice of first and last names, anecdotes about the length of names, …). – The two subjects alone in a room facing each other. – Three renditions of the dialogue. (The first one without prior reading, the second and third ones after having discussed freely about the dialogue.)

24

1.2. Rep Task. 3/17

Small Talk. Outcome

•  Gross observation: the resemblance between the renditions. - The original dialogue: - The first rendition : NB.: Script and translation on the hand-out.

•  Detailed comparison: – A contour labelled « rising of list » (RL)* (Portes, Bertrand & Espesser 2007a, b),

– The intonation patterns associated with Topic reprise declaratives. * « Rising of list » wrongly sugests that the contour is intrinsically linked to listing or coordination. We keep it as an arbitrary label. 25

1.2. Rep Task. 4/17

Small Talk. Script •  •  •  •  •  •  •  •  •  •  •  •  •  •  • 

Tu avais assisté à l’accouchement, toi, de ? Nan. J’ai pas voulu. Tu as pas voulu ? Nan nan. T’avais eu? Cétait quoi ? T’avais? Bah en plus c’était une césarienne donc euh Ah bon, elle a accouché avec une césarienne, Myriam ? Ah d’accord. Alors ouais alors là, c’est clair. Ouais ouais mais même au départ quoi, moi je C’est un truc qui me dit rien du tout quoi. Moi, j’a i C’est vrai ? Si tout va bien je vais essayé de le faire mais j’en sais rien si je vais pas tourner de l’oeil ou j’en sais rien ouais tu vois ? Ouais. Ou ce que tu vas garder comme image. Ouais voilà. Parce que j’ai vu. On a fait les cours de préparation à l’accouchement. Ouais. Tu y es allé, toi ? Ouais. J’y suis allé.

26

Small Talk Comparison I: Rising of list

1.2. Rep Task. 6/17

Rising of list (RL)

RL is a nuclear contour showing 4 features: - The syllable anchoring the maximum of F0 shows an important lengthening, - Pitch tends to remain constant during the realization of the lengthened syllable, resulting in plateau patterns most of the time, - The pitch is situated in the mid or upper range or the speaker’s register, - Modal voice. See Portes, Bertrand & Espesser 2007a, b for a formal characterization.

Illustration: O

R1

R2

R3

– et puis tu vois pour les parents ça peut être un choc si tu veux. And you see for the parents it might be quite a shock

– Ouais ouais c’est ça. 〈Pourquoi il a pas pris le mien〉. Yeah yeah that’it. Why didn’t he choose mine (my name)

28

1.2. Rep Task. 7/17

Rising of list

Raw distribution: - 13 occurrences in the original - 10 of them replicated at least once - 6 of them replicated more than once - 3 replicated three times - 4 occurrences added by the participants

29

1.2. Rep Task. 8/17

Rising of list

Context feature favoring the replication of RLs: –The discourse relation “Elaboration” holds with the preceding segment, – The segment showing RL is a less abstract description of what is elaborated from the preceding segment. Ex.: RL6-7 – Et là, c’est bonnard quoi quand tu l’as. T’as une meuf qui le prend, qui lui lave le nez. and then it is great when you get him. A women takes him, washes his nose

30

1.2. Rep Task. 9/17

Rising of list

RL6-7 – Et là, c’est bonnard quoi quand tu l’as. T’as une meuf 〈qui le prend〉, 〈qui lui lave le nez〉

RL14 – et puis tu vois pour les parents ça peut être un choc si tu veux. – Ouais ouais c’est ça. 〈Pourquoi il a pas pris le mien〉.

RL15 – Isabelle me disait par exemple Tain t’emmerde pas avec deux noms ça va être. 〈tu vas voir après quand tu vas avoir les papiers〉.

31

1.2. Rep Task. 10/17

Rising of list

Semantics: -  RL signals that E stands for E’ that is more general and/or abstract than E. -  Similar to synecdoches where an expression referring to a concrete object stands for an expression referring to an abstract one. (1 a.) # Il lui pose toujours la même question. Pourquoi t’as pas pris mon nom ? b.) # Il lui a posé deux questions. Pourquoi t’as pas pris mon nom ? Pourquoi t’as changé de prénom ? c) Il a été bouleversé quand il a appris qu’il avait choisi le nom de sa mère. Pourquoi t’as pas pris mon nom ? Qu’est-ce que je t’a fait ?

NB.: « # » = inappropriate on the underlined N.

32

1.2. RL.Conclusion

Rising of list. Conclusion intermédiaire

Du point de vue expérimental – Les sujets reproduisent le contour RL quand il y a suffisamment d’indices pour son emploi dans le contexte. – Intérêt de RepTask: les contextes où RL est reproduit et la fréquence de la reproduction sont des indications sur sa sémantique. En particulier: RL n’est pas spécifiquement lié à la coordination (avec ou sans conjonction).

33

Small Talk Comparison II Topic reprise declaratives

1.2. Rep Task. 11/17

Topic reprise declaratives (TRD) Reprise declarative: declarative whose content does not add to the Common ground, either because it is identical to that of the previous turn (1) or entailed by it (2). (1) – Nan. J’ai pas voulu – T’as pas voulu (2) – L’appréhension s’est transformé en Tu vois, j’arrêtais pas de bailler. – Ouais, ça t’a assommé, quoi – Observation: - High number (9) of reprise declaratives in the original dialogue. - They are associated with only two intonation patterns.

35

1.2. Rep Task. 12/17

TRD. Prosody in the original

•  P1: Recto tono until the last register.

200 Pitch (Hz)

syllable which is slightly rising or falling in an overall compressed

300

100 0 C’ETAIT PEINT

SUR LA TAPISSERIE

0

1.2019 Time (s)

•  P2 : Low plateau until the last normal span of the speaker

200

Pitch (Hz)

syllable which is rising (more or less sharply), the span of the register is the

150

100 70 T’AS PAS VOU 0

LU 0.43776

Time (s)

36

1.2. Rep Task. 13/17

TRD. Prosody in replications

Observation: TRDs in replications show the same intonation behavior in 78 % of the cases. They are associated with: – P1: 7 times (NB.: The register is less compressed than in the original). – P2: 11 times – Subjects use a third pattern P3 in 6 cases.

37

1.2. Rep Task. 14/17

TRD. Prosody in replications

P3: high plateau until the last syllable which is falling. 160

Pitch (Hz)

140 120 100 80 T’AS

PAS

VOU

0

LU 0.46882

Time (s)

38

1.2. Rep Task. 15/17

TRD. Conversation analysis – Literature: - In syntax-semantics, RDs are analyzed as signalling a problem in Grounding or Updating the CG - In CA analysis, it has been reported that RDs may convey a marked acknowledgment. Speaker presents herself as co-asserting. – Here: RDs are used as a device to manage the topic flow. (Button & Casey 1985) They present an aspect of the content of the preceding turn as topicalizable by the addressee for further talk. Hyothesis: The restricted choice of intonation patterns is related to the dialogue role of RDs. – TRDs are incompatible with patterns presenting the speaker as committed to the content of her turn. E. g. TRDs are incompatible with the falling pattern or the questioning H+L pattern. (Beyssade & Marandin 2007) – It the hypothesis is right, Subjects when they abide by the restriction of the choice of contours, show they are sensitive to the dialogue role of the utterances. Thus, their competence includes a knowledge of the usage of contours in dialogue. 39

1.2. Rep Task. 16/17

Interim conclusion (Small Talk) •  Subjects in the lab replicated intonation choices made by Speakers in another situation. •  They replicated other Speakers’s choices when the context gives enough cues inducing the usage of the given prosody. •  Replicating RL or (putatively) the restriction on the intonation patterns of topicalizing reprises requires that the replicating Subjects understand the content and the role of the turns they are reenacting. 40

1.2. Rep Task. 17/17

Interim conclusion (Small Talk)

Implication for the design of experiments: Subjects in the lab behave as Speakers in actual intractions, provided that the protocol of the experiment activates the proper context of use for the phenomenon under study. Accordingly: – Elicited speech may provide one with valid data for semantic or pragmatic analysis. – The inconclusiveness of most script-based experiments is more a matter of lack of proper context for the phenomenon under study than of artificiality of the setting.

41

Part II Replication and linguistic knowledge

2.1. Replication in judgment-based experiments

2.1.1. Syntax Dative alternation (Bresnan)

2.1. Judg. task 1/9

Replication in a judgment task. Syntax Construction choice is a well known problem (i. a. Rosenbach 2003, Bresnan 2007). Example: dative alternation in English. (1)

a. He gave the pony to my children b. He gave my children the pony

The choice of (1a) over (1b) depends on many different factors such as the semantics of the verb, length of the XPs, pronominality, activation statu of the referent, etc. It has been studied through: – Probalistic modeling of the choice in a corpus, – Judgment-based experiments. I focus on one of these experiments I call Dative alternation (Bresnan 2007). 45

2.1. Judg. task 2/9

Dative alternation

Design and protocol •  A corpus of utterances (with context) showing one of the dative construction is gathered, •  For each utterance, an alternative paraphrase is constructed showing the other construction, •  Each pair of utterances is presented to the Subjects •  Subjects have to rate which alternative is the most natural in context.

46

2.1. Judg. task 3/9

Dative Alternation. Example

Context: (1) About twenty-five, twenty-six year ago, my brother-in-law showed up in my front yard pulling a trailer. And in this trailer he had a pony, which I didn’t know he was bringing. And so over the weekend, I had to go and find some wood and put up some kind of a structure to house that pony,

Which construction (2) or (3) yields the most natural continuation: (2) because he brought the pony to my children (3) because he brought my children the pony 47

2.1. Judg. task 4/9

Dative alternation. Results

Subjects tend to pick out the same choices made by the original speakers. Moreover, their choices correspond to the predictions of the probabilistic model accounting for the distribution of the constructions in the corpus. See Bresnan 2007 for the quantitative details. 48

2.1. Judg. task 5/9

Dative Alternation. Conclusion.

– Subjects’ choices (in a metalinguistic task) replicate Speakers’ choices in actual world situations. Note that Subjects have access to the relevant contextual knowledge. – The linguistic knowledge put to use in the judgment task includes knowledge of the usage of the construction in point in contexts.

49

2.1.2. Intonation Topicalizing Reprises (under construction)

2.1. Judg. task 6/9

Experiment Topicalizing Reprises Design and protocol: – Replication of Dative alternation for the choice of intonation patterns, – Fragments of dialogue are chosen from the first replication of Small Talk – Subjects have to choose the most natural intonation renditions of a topicalizing reprise in the given context.

Hypothesis. Subjects pick out the intonation rendition chosen by the Subjects in Small Talk. NB.: Contrary to alternatives constructions in Syntax, alternative intonation pattern are not equivalent pragmatically. 51

2.1. Judg. task 7/9

Topicalizing Reprises. Choice of renditions The two intonation renditions (resynthetized) correspond to two intonation patterns : - The former belongs to the restricted set observed in Small Talk -  The latter is either the falling (assertoric) pattern or the H+L questioning pattern (rise on the penultimate and fall on the final syllable).

Prosodic patterns observed in Small Talk vs.

Other

P1. Recto tono P2. Final rising

P3. Falling P4. (Questioning) H+L

52

2.1. Judg. task 8/9

Topicalizing Reprises. Example

Given the context: – Bon, ben. C’est marrant parce l’appréhension s’est transformée en .. Tu vois, j’arrêtais pas de bailler. –  Ouais, ça t’a assommé, quoi – Tain. J’ai failli sortir à un moment donné. Which rendition is the most natural ? P1: Recto tono

vs.

500

P3: Falling

30 24

400

300

12

200 0 100

–12

0 ouais

ça tʼa asommé

0

quoi

ouais 1.19714

Time (s)

ça tʼa

(as)sommé

0

quoi 1.09392

Time (s)

53

2.1. Judg. task 9/9

Topicalizing Reprises. Discussion

- Two pilots (to-date: 25 / 10 participants) -  Statistical treatment of the results in progress -  Role of the context. -  Alternatives with P2: vs. with P3:

•  First pilot: – Tu avais assisté à l’accouchement, toi, de… – Nan. J’ai pas voulu. – T’as pas voulu – Nan nan.

Results: P2 (final rising): 40% P3 (falling): 60% Falling typical of acknowledgement in closing sequence ?

•  Second pilot: – Tu avais assisté à l’accouchement, toi, de… – Nan. J’ai pas voulu. – T’as pas voulu – Nan nan. – T’avais eu .. C’était quoi? – Ben en plus c’était une césarienne donc euh – Ah bon elle a accouché avec une césarienne Myriam? Ah d’accord, là c’est clair.

Results: P2 (final rising): 70% P3 (falling): 30% Topicalizing role recovered ? 54

2.2 Replication and linguistic knowledge

2.2. Linguistic .. 1/5

Synthèse – Dative alternation: les jugements lexico-syntaxique opérés par les sujets en condition expérimentale correspondent aux choix effectués par les locuteurs en “milieu naturel”. – Small Talk: (indique fortement que) les choix prosodiques des sujets en condition expérimentale correspondent aux choix des locuteurs “en milieu naturel”. – Topicalizing Reprises: (suggère qu’il est bien possible que) less choix prosodiques soient comme les choix lexico-syntaxiques. – Tous ces résultats mettent l’accent sur la nécessité de fournir aux sujets en condition expérimentale un contexte approprié. 56

2.2. Linguistic .. 2/5

Linguistic knowledge

S’il est vrai que -  les jugements des sujets en condition expérimentale refètent les choix des locuteurs en milieu naturel, -  Les choix des sujets en condition expérimentale reflètent les choix des locuteurs en milieu naturel, -  Alors -  Ce que l’on teste, c’est la capacité des sujets à reproduire les choix des locuteurs dans leur interactions en milieu naturel. 57

2.2. Linguistic .. 3/5

Linguistic knowledge

La capacité à reproduire les choix des autres locuteurs – Capacité massive: : “Every time, we open our mouth to speak, we replicate grammatical structures –sounds, words and constructions– that have occurred in prior utterances to which we have been exposed.” (Croft 2006). Mais surtout et crucialement: – Capacité distribuée: c’est chaque construction constitutive d’une langue qui est associée à ses contraintes de reproduction. 58

2.2. Linguistic .. 4/5

Reproduction et conception d’expériences

•  Les expériences qui mettent en jeu/scène une tache de reproduction exploite directement la compétence des sujets-locuteurs. •  Elles exploitent la dimension conventionnelle du savoir linguistique tel que la définit Rudi Keller “Talk in a way which you believe the other would talk if he or she were in your place.” (Keller 1994: 99)

59

2.2. Linguistic .. 5/5

Reproduction et changement linguistique • 

Plusieurs théoriciens de la variation linguistique (synchronique et diachronique) font de la capacité des sujets à reproduire les choix des autres membres de la communauté linguistique le moteur de la stabilité ou, au contraire, du changement linguistique (Croft 2006, Keller 1994, Thomsen 2006).

• 

“The replication process is conversation, that is, the production of utterances in language use. [..] Language change occurs in a population through replication. [..] A theory of language change must also explain why languages do not change much of the time as well. In language change, the chief causal mechanism [..] is convention. [..] Normal replication is conformity to convention.” (Croft, 2006).

• 

Perspective (surprenante) qui permet de rapprocher l’étude du langage conçu comme une capacité des locuteurs et l’éude du langage conçu comme un objet socio-historique évoluant au gré des performances de ceux qui l’utilisent (i. a. Rosenbach 2003, Bresnan & Ford (to app.)).

60

Acknowledgement

Acknowledgement

Lab linguistics requires the collaboration of many people with different skills. This talk is based on collaborative involving:

– InfoFoc : Claire Beyssade, Barbara Hemforth, Cristel Portes. – Small Talk : Fréderic Laurens, Cédric Patin, Hiyon Yoo. – Topicalizing Reprises : Clément Planck, Hiyon Yoo. Special thanks to F. Laurens, C. Planck & Hiyon Yoo for the preparation of this talk. All errors are mine. Research supported by the ANR: Projet PRO-GRAM.

61

Reference (1)

References •  •  •  • 

•  •  •  •  • 

Bertrand Roxane et al., 2008. Le CID - Corpus of Interactional Data - Annotation et Exploitation Multimodale de Parole Conversationnelle. TAL 49-3, pp. 1-30. Beyssade Claire & Jean-Marie Marandin, 2007. French Intonation and Attitude Attribution, Proceedings of the 2004 Texas Linguistics Society Conference: Issues at the Semantics-Pragmatics Interface, ed. by Pascal Denis, Eric McCready, Alexis Palmer, and Brian Reese, pp. 1 -12. Beyssade Claire, Barbara Hemforth, Jean-Marie Marandin & Cristel Portes. 2010. Prosodic Markings of Information Focus in French. Proceedings of IDP 09. Bresnan Joan, 2007. Is syntactic knowledge probabilistic? Experiments with the English dative alternation. [Sam Featherston and Wolfgang Sternefeld, eds] Roots: Linguistics in search of its evidential base, Series: Studies in Generative Grammar. Berlin and New York: Mouton de Gruyter, pp. 75–96. Bresnan Joan & Marilyn Ford. To app. Predicting Syntax: Processing Dative Constructions in American and Australian Varieties of English. Button Graham & Neil Casey, 1985, Topic nomination and topic pursuit, Human Studies 8:3 pp. 3-55. Carden Guy. 1976. Syntactic and semantic data: replication results. Language in society 5-1: 99-104. Croft William. 2006. The relevance of evolutionary model to historical linguistics. [Ole Nedergaard Thomsen, ed.] Competing models of linguistic change: evolution and beyond. John Benjamins: 91132. Féry Caroline. 2001. Focus and phrasing in French. [Féry C., C. & W. Sternefeld (eds.), Audiatur Vox Sapientiae. A Festschrift for Arnim von Stechow. Berlin Akademie-Verlag, pp. 153-181.

62

References (2)

•  •  •  •  •  • 

Ito Kiwako & Shari R. Speer, 2006, Using interactive tasks to elicit natural dialogue. [Sudhoff Stefan et al., eds] Methods in empirical prosdy research. Berlin: W. de Gruyter, pp.229-258. Keller Rudi. 1994. On language change: the invisible hand. London: Routledge. Ladd Robert. 1996. Intonational Phonology. Cambridge: Cambridge UP. Laurens Frédéric, Jean-Marie Marandin, Cédric Patin and Hiyon Yoo. 2010. The used and the possible. The use of elicited conversations in the study of Prosody. Proceedings of IDP 09. Marandin Jean-Marie. 2005. Contours as constructions. [D. Schoenefeld, ed.] Constructions all over ; case studies and theoretical implications. http://www.constructions-online.de/articles/specvol1/ Pierrehumbert Janet, Mary E. Beckman & D.R. Ladd. 2000. Conceptual foundations of Phonology as a laboratory science. [ Burton-Roberts Noel, Philip Carr & Gerard Docherty, eds] Phonological knowledge: its nature and status. Cambridge university Press.

•  •  • 

Portes Cristel, Roxane Bertrand & Robert Espesser, 2007a. Contribution to a grammar of intonation in French. Form and function of three rising patterns. Hand-out IDP 07. Available at http://program.linguist.univ-paris-diderot.fr/recherche.php?nav=structure. Portes Cristel, Roxane Bertrand & Robert Espesser, 2007b. Contribution to a grammar of intonation in French. Form and function of three rising patterns in French, Nouveaux cahiers de linguistique française 28, pp. 155-162. Rosenbach Anette. 2003. Aspects of inconicity in the choice between the s-genitive and the ofgenitive in English. [Rohdenburg G. & B. Mondorf, eds] Determinants of gramatical variation in English. De Gruyter : 379-411. •  Szmrecsanyi Benedikt. 2005. Language users as creatures of habit: a corpus-based analysis of persistence in spoken English. Corpus linguistics and linguistics theory 1-1: 113-149. 63