Atelier de phonologie

Atelier de phonologie

Bienvenue ! Welcome!

L'Atelier de phonologie est un séminaire de recherche informel consacré à la phonologie : représentation, computation, acquisition et modalités d'apprentissage, interaction avec les autres modules linguistiques, etc. Y sont présentés des travaux en cours, des préparations de conférences, des revues de la littérature, ainsi que des présentations plus accomplies. Le séminaire, hébergé par le laboratoire SFL (Structures Formelles du Langage), est ouvert à la communauté de phonologues locale et internationale. Il se réunit environ toutes les deux semaines, le mercredi matin de 10h à midi, et est suivi d'un déjeuner sur place. Les archives des réunions passées (depuis l'automne 2012) est disponible ici.

The Atelier de phonologie is an informal research seminar dedicated to phonology: its representations, its computations, its acquisition and learnability, its interaction with other linguistic modules, etcetera. The seminar features informal presentations of work in progress, practice talks, literature reviews, as well as more polished talks. It is open to the local community as well as to international visitors. It is hosted at the SFL lab and it meets approximately every two weeks, always on a Wednesday morning, followed by lunch. The archive of past meetings (since fall 2012) is available here.

Où et quand / When and where

Où/Where et quand /when:

59/61, rue Pouchet, Paris (cartes et informations sur les transports : here you can find a map and information concerning public transportation).

le mercredi, de 10h à 12h. On Wednesdays, from 10am to noon.

Calendrier / Schedule 2020-2021

1er semestre

09 oct: Francesc Torres-Tamarit

23 sept: Noam Faust & Shanti Ulfsbjorninn

7 oct: Eva Zimmermann

21 oct: Benjamin Storme

4 nov: Noa Handelsman

18 nov: Mathilde Hutin

2 dec: Ollie Sayeed

16 dec: Nicola Lampitelli

2nd semestre

20 jan Françoise Rose

03 fév Claudia pons-Moll

17 fév Daniel Asherov

03 mar Marijke De Belder

17 mar Timothée Prémat

31 mar Fabian Zuk

14 avr Maria del Mar Vanrell

05 mai Marie-Hélène Côté

09 sep

Francesc Torres-Tamarit (SFL, CNRS & Université Paris 8)

Typological aspects of contrastive vowel length in Romance

By way of comparative historical reconstruction and cross-dialect synchronic descriptions, Loporcaro (2015) arrives at establishing that contrastive vowel length (CVL) in Northern Romance in stressed syllables is metrically-governed, and that its distribution is implicational. The presence of CVL in proparoxytones implies CVL in paroxytones, but not the other way around. Likewise, the presence of CVL in paroxytones implies CVL in oxytones, but not

vice versa. The same distribution is found for derived (non-contrastive) vowel length as a result of a process of open syllable lengthening, and also for stress-dependent gemination. Building on Loporcaro (2015), in this talk I will

show that by combining internally layered (ternary) feet with uneven trochees, an OT, foot-based analysis of the distribution of CVL in Northern Romance is not only descriptively adequate, but increases explanatory power because it avoids both over- and under-generation problems, as opposed to

analytical alternatives that exclude layered feet (but use instead final extrametricality as in Jacobs 2019). The theoretical contribution of this investigation is to give additional support for minimal layering of feet, an issue that has recently received attention in the literature on metrical theory.

23 sep

Noam Faust (SFL, CNRS & Université Paris 8) and Shanti Ulfsbjorninn (University of Deusto)

Prosodically-driven harmony in Strict CV: a Celto-Semitic case


In Barra Gaelic (BG; Boseh de Jong 1997), stress (underlined) is generally word-initial and correlates with a high tone [aHran] ‘bread’. A harmonic (or “copy”) epenthetic vowel is inserted in the environment /#(C)VC1_C2(…)/, where C1 is any sonorant and C1 and C2 are hetero-organic: e.g. /t̪ɔrɣ/ => [t̪ɔrɔHɣ] ‘fishing line’. As shown in the transcription, this epenthetic vowel is doubly interesting: i. it carries the high tone despite being peninitial, and ii. it is as stressed as the initial vowel.

In Modern Hebrew (MH), stress is generally final: [mufsak] ‘begin.pass.prtc’. A harmonic process transforms [a] to [e] before a word-final unstressed sequence [eC]: [mufsek-et] ‘begin.pass.prtc-fm’. The unstressed [e] must be analyzed as epenthetic/weak. Like BG peninitial epenthesis, the MH case is typologically strange: the weak, unstressed vowel triggers harmony on the lexical stressed one.

We propose a Strict CV account of both patterns that highlights their similarities. In both languages, a prosodic domain must be edge-aligned (to the left in BG, to the right in MH). When epenthesis obliges the aligned domain to span two V-slots, two effects follow. First, its non-aligned edge must also be marked: this is done by H in BG and by stress in MH. Second, the span of two V-slots must be signaled by harmony.

07 oct

Eva Zimmermann (Leipzig Universität)

Phonological exceptions result from gradient constraint violations: An argument for Gradient Symbolic Representations


The assumption of Gradient Symbolic Representations that phonological elements can have different degrees of activity (Smolensky and Goldrick, 2016; Rosen, 2016; Zimmermann, 2018, 2019) allows a unified explanation for the typology of phonological exceptions.

Exceptional (non)triggers and (non)undergoers of otherwise regular phonological processes are predicted from gradient constraint violations: The activation of a phonological element in an underlying morpheme representation determines 1) how much the element is preserved by faithfulness constraints and 2) how much it is penalized by markedness constraints. I argue that this simple mechanism predicts the attested typology of phonological exceptions.

In this talk, I argue that the predictions made by the GSR account are empirically more adequate than the ones made by alternative approaches to exceptionality based on autosegmental defectivity (Lieber, 1987; Tranel, 1996; Zoll, 1996) or lexically indexed constraints (Pater, 2006; Flack, 2007; Mahanta, 2012).

21 oct

Benjamin Storme (UNiversité de Lausanne) parler de

Not only size matters: limits to the Law of Three Consonants in French phonology

Grammont’s Law of Three Consonants (LTC) states that French schwa is obligatorily pronounced in any CC_C sequence to avoid three-consonant clusters. Although schwa presence has been shown to be sensitive not only to cluster size but also to the nature of consonants in post-lexical phonology, the LTC is still considered as accurate to describe schwa-zero alternations in lexical phonology. The paper uses judgment data from French speakers in France and Switzerland to compare the behavior of schwa in derived words (lexical phonology) and inflected words (post-lexical phonology). The results show that schwa-zero alternations are conditioned not only by cluster size but also by cluster type in lexical phonology. Moreover, the same phonotactic asymmetries among consonant clusters are found in lexical and post-lexical phonologies. The data therefore support a weaker version of the lexical-phonology hypothesis than what is usually assumed for French. Lexical and post-lexical phonologies do not require different phonotactic constraints but only different weights for the same constraints.

4 nov

Noa Handelsman (Tel Aviv Universtiy)

Category-specific phonology in the acquisition of Hebrew

In Hebrew, noun-stems and verb-stems are prosodically restricted – they are usually disyllabic with final codas. These prosodic restrictions are identical for nouns and verbs such that they may be indistinguishable; for example, kaˈtav means both ‘a reporter’ and ‘to write’. The contrast between nouns and verbs emerges with the morphological paradigm, as nouns and verbs employ different suffixes and different morpho-phonology (Bat-El 2008). However, these morpho-phonological means to distinguish between nouns and verbs are not available during early speech, when children produce stem-like forms without overt morphological structure (Levy 1980, Armon-Lotem & Berman 2003, Adam & Bat-El 2009).

The talk will address the contrast between nouns and verbs in children’s productions during this morphology-free period, when nouns and verbs are produced without the morphology that allows distinguishing between them. The question addressed is whether children make an overt distinction between nouns and verbs during this period, and if so, how.

To address this question, the spontaneous productions of 3 Hebrew-acquiring children were examined with reference to the development of the lexicon, the phonology (codas and number of syllables) and the morphology (suffixes). The findings suggest that during the period when the productions are morphology-free, children use their own phonological strategy to distinguish between nouns and verbs. This is a case of Category Specific Phonology, often found in adults’ systems, whereby nouns and verbs adhere to different phonological patterns and thus different phonological grammars (McCarthy & Prince 1995, Smith 1997, Antilla 2002, Bat-El 2008).

18 novembre

Mathilde Hutin (Université Paris-Saclay)

Lenition and fortition of stops at word-edges in Romance languages:

A study of voicing alternations in French, Romanian, Spanish, Portuguese and Italian


Mathilde Hutin1, Yaru Wu1,2, Adèle Jatteau3, Ioana Vasilescu1, Lori Lamel1, Martine Adda-Decker1,2

1 Université Paris-Saclay, CNRS, LIMSI, Bât. 507, rue du Belvédère, 91405 Orsay, France

2 Université Paris 3 Sorbonne Nouvelle, CNRS, UMR 7018, LPP, 19 rue des Bernardins, 75005 Paris, France

3 Université de Lille, CNRS, UMR 8163, STL, Lille, France

{mathilde.hutin, yaru.wu, ioana.vasilescu, lori.lamel, martine.adda},

Lenition is a well-known phenomenon defined as a process whereby a consonant is “weakened”: “a segment X is said to be weaker than a segment Y if Y goes through an X stage on its way to zero” (Venneman in Hyman 2008). A refined definition (Szigetvári 2008) distinguishes between “consonantic” lenition, where consonants become more consonant-like when they lose place or laryngeal specifications (ex. s → h → 0), and “vocalic” lenition, where consonants become more vowel-like when they move up the sonority scale (ex. t → ɾ). On the other hand, fortition appears to be the reverse of vocalic lenition exclusively, i.e. a movement down the sonority scale (ex. j → dʒ) while an equivalent of “consonantic fortition” (at least regarding place) does not appear to be on record. The present study will thus focus on vocalic lenition and fortition. More precisely, studies on Romance languages (Carvalho 2008) show that changes concerning the laryngeal feature are instances of such phenomena (ex. Lat. vita → m. fr. vidə → Fr. vie). Following Ohala (1989), the present study is based on the idea that historical processes such as voicing and devoicing are pre-conditioned by synchronic variation. We will thus present an in-depth exploration of voicing alternations in word-initial and word-final position in 5 Romance languages (mostly French and Romanian but also Spanish, Portuguese and Italian) using large-scale corpora (ca. 1000 hours of speech) and automatic alignment. The aims are to investigate (i) whether the languages under survey here display voicing alternations that can be considered lenition and fortition, i.e. that are conditioned by positional factors instead of representative of adjacency effects (Ségéral & Scheer 2008), (ii) on a subsidiary note, whether in connected speech the consonants at word-edges follow the generalizations drawn from word-internal regularities (Ségéral & Scheer 2008), and finally (iii) whether these voicing alternations can be considered as ongoing changes in these languages, i.e. are starting a phonologization process (Hyman 2008).

2 décembre

Ollie Sayeed (University of Pennsylvanya)

f vs θ

The categories [f] and [θ] show an asymmetrical relationship both synchronically and diachronically: [θ] is rarer than [f] typologically (UPSID; Maddieson 1981), [θ] has more variable acoustic cues than [f] (McGuire and Babel 2012), [θ] is more confusable for [f] than [f] is for [θ] (Miller and Nicely 1955), [θ] > [f] is common diachronically while [f] > [θ] is rare or unattested (Honeybone 2016), and [θ] is synchronically more volatile across dialects than [f] (Kjellmer 1995). But why? In this talk, I give preliminary results from a series of studies on [f] and [θ] aimed at teasing apart two explanations: one based on phonological markedness, and one based on asymmetric acoustic distribution.

16 décembre

Noam Faust (Université Paris 8, CNRS SFL) and Nicola Lampitelli (Université de Tours, CNRS LLL) parler de

The transparent conspiracy of gutturals and low vowels

Graphic problems prevent the expoition of the abstract.


20 janvier

Françoise Rose (Dynamique Du Langage, CNRS/Université de Lyon)

A contribution to the typology of syncope: rhythmic and differential syncope in Mojeño Trinitario


Rhythmic syncope, a process deleting the vowels in weak metrical positions is a perfect illustration of the role of the metrical parse in stress placement and phonology in general. Differential syncope, a process by which low-sonority vowels may be banned from syllable nuclei or strong foot branches, and high-sonority vowels banned from weak foot branches, illustrates the role of sonority in prosody. In Mojeño Trinitario, an Arawak language of Bolivia, a metrical analysis provides crucial insight into a rhythmic syncope process that is amenable to neither a strictly phonological, lexical or morphological analysis. However, rhythmic syncope underapplies with sonorous vowels, some of which are lexically-determined as transparent to syncope. Mojeño Trinitario therefore shows a rare case of a hybrid type of syncope, where the application of rhythmic syncope is constrained by the sonority scale.

10 février

Claudia Pons-Moll (Universitat de Barcelona)

Prosodically-driven morpheme non-realization in the Minorcan Catalan DP as evidence for left-edge Strong Start



17 février

Daniel Asherov (MIT)

A problem for frequency-based approaches to segmenthood from Hebrew affricates

(joint work with Christopher Minwoo Yang, MIT)


The ‘segment’ is typically taken to be a fundamental notion in linguistic analysis, yet there is often disagreement about the definition of a segment, and the correct analysis of individual sound sequences as segments or non-segments (Trubetzkoy 1939). For example, a sequence of sounds like [ts] may constitute a single affricate consonant [t͡s] or two adjacent consonants [t] and [s]. Similarly, a sequence of sounds like [mb] may be understood by speakers as a single prenasalized stop [mb] or as two adjacent consonants [m] and [b].

There have been recent attempts to predict which sound sequences would be mapped into a single segment in a given language based on their relative frequency in that language (Gouskova & Stanton 2020). This approach has been successful in predicting segment-like behavior in affricates and prenasalized stops in a variety of languages. For example, it predicts that speakers of Hebrew would learn [ts] as a segment-sized unit, but not speakers of English. This aligns with the traditional analyses of the segmental inventory of the two languages.

In this talk we offer an empirical test of the frequency-based approach based on the phonological patterning of Hebrew (potential) affricates. Hebrew has one very frequent potential affricate (ts), and two infrequent potential affricates which entered the language through loanwords (tʃ,dʒ). The frequency-based model, which is sensitive to this asymmetry, predicts that speakers of Hebrew would learn [ts] as a segment, but would perceive [tʃ] and [dʒ] as sequences of segments. We will assess this prediction using some diagnostics for segmenthood made available by independent Hebrew phenomena. We will find that all three potential affricates do behave as single segments in Hebrew, contra the expectations of a frequency-based approach to segmenthood.

03 mars

Category-specific phonology in Dutch: Evidence against optimal paradigms

Marijke De Belder (University Oldenburg)


In this talk I discuss that Dutch nouns and Dutch verbs have different syllable structures. I present the results of an experiment that shows that this knowledge is part of the native speaker's phonology. I show that the overall pattern in the language could be understood as a side-effect of verbal inflectional requirements: syllables in a verbal syntactic context need to be able to accommodate an extra consonantal position that realises the inflection. Nouns do not have this need in Dutch, as nominal inflection can always be realised as a syllable on its own.

One can imagine several ways in which category-specific phonological rules may become part of the speaker’s phonology (and may subsequently be reflected in the lexicon over time). For example, the category-specific phonological rules may be acquired via analogy and statistics: speakers may avoid marked inflected forms, the child notices the gap and learns that polysyllabic morphologically simplex verbs are statistically rare and formulates a verb-specific markedness rule against them. Now, the Optimal Paradigms Approach (McCarthy 2001, 2005, Cable 2004) formulates a very specific strategy: the native speaker evaluates candidates for insertion not as single forms, but rather as entire paradigms. Given that the Dutch verbal paradigm has a higher potential of imposing markedness than the nominal one, verbal bases could end up being more restricted.

I provide experimental evidence against the Optimal Paradigms approach, with its claim that the phonology evaluates entire paradigms rather than single forms. I further discuss the theoretical consequences for our understanding of the PF-syntax interface and for morphosyntax.

17 mars

16th century French final schwa: from variation to diachrony, an OT account.

Timothée Premat (Université Paris 8 / SFL)

Final schwa in considered to be an epenthetic vowel in Standard Contemporary French. This was not always the case: in Old French (8th – 14th centuries), final schwa was a vowel present in both the underlying representation and the surface form. Manuals and grammars of French phonetic/phonological diachrony converge to say that final schwa in the UR begins to be deleted during the Middle French period (14th – 16th centuries), starting by hiatus contexts : ə → Ø / V__# (= contraction, like in amie /ami.ə/ realized [ãmi]) (GGHF, p. 477, §384). The 16th century turns out to be a period of great changes. It saw the trend that started with contraction spread to other contexts and develop into the general apocope of final schwa. Accordingly, the distribution of final schwas in surface representations became independent of its possible existence underlyingly.

This talk begins with data from the testimony of numerous 16th century grammarians and from metrical habits (versification). Generalizations are then provided about final schwa contraction/apocope and elision. Finally, an Optimality Theoretic analysis is put forth that covers both synchronic variation (diastratic/stylistic and diatopic) and diachronic evolution.

We model final schwa behavior in the 16th century using only three constraints (Max, *Hiatus and *Schwa) and three grammatical mechanisms. These are: i. Local Constraints Conjunction (LCC, Green 1993 ; Smolensky 1993); ii. Locus Specification (applied to LCC: Łubowicz 2005); and iii. Partial Ranking (Antilla 1997). We also show that partial ranking accounts not only for synchronic variation, but also for how diachronic evolution can emerge from synchronic variation. The account therefore sheds light on how final schwa in French became an epenthetic vowel. We close with an extrapolation regarding the emergence of the fully epenthetic final schwa of contemporary French.

31 mars

L’apophonie et les dynamiques de changement dans le système verbal de l’arabe égyptien (le parler du Caire)

Radwa Fathi (LLING/ Université de Nantes) parler de

Dans cette présentation, je vais faire une proposition concernant la vocalisation des verbes à la Forme I de l’arabe du Caire. Je vais soutenir que le système est le même que celui de l’arabe classique : les alternances perfectif/imperfectif sont gérées par un système apophonique. Le développement de mon argumentation est présenté comme si les corpus des deux langues étaient d’une stabilité comparable. Mais, ma proposition doit confronter une réalité. Cette réalité est liée à une différence fondamentale : pour des raisons évidentes le corpus de l’arabe classique ne « bouge » plus ; le corpus de l’arabe du Caire, par contre, met en évidence des dynamiques de changement continu. La question que je poserai donc est la suivante : est-ce que ces changements expriment un éloignement ou même une décadence du système apophonique ? Ou au contraire ‘est-ce que ces changements sont pilotés par l’apophonie ?’

L’une des manifestations de l’évolution constante de l’arabe du Caire est la présence de formes alternatives pour la vocalisation des verbes : dans un nombre important de cas, un même verbe peut être réalisé avec différentes vocalisations, FaʕaL ou FuʕuL, FiʕiL ou FaʕaL, FiʕiL ou FaʕaL ou FuʕuL. La même chose est vraie des imperfectifs : yi-FʕaL ou yu-FʕuL, yi-FʕiL ou yi-FʕaL, etc. Il est vrai qu’on trouve le même phénomène en arabe classique mais dans une beaucoup moins grande mesure. Dans le cas de l’arabe du Caire, on peut parler de prolifération.

Dans cette présentation, je vais développer trois idées : 1) Que les doublets (ou triplets) en question ne sont pas distribués de façon anarchique. 2) Que les membres des doublets (ou triplets) ne sont pas sur un pieds d’égalité. Par exemple, lorsqu’un verbe d’une même racine peut être réalisé comme FiʕiL ou comme FaʕaL, l’une des deux formes est sur le point de l’emporter sur l’autre. Autrement dit, l’existence de tels doublets ne représente qu’un stade transitoire dans une évolution dont la direction peut être détectée. En particulier, je mettrai en lumière les raisons pour lesquelles FiʕiL est le grand gagnant là où il est en compétition avec d’autres formes. 3) Que le système apophonique est précisément le moteur de cette évolution.

14 avril

Focus marking by Spanish monolingual and heritage speakers

Maria del Mar Vanrell Bosch (Universitat de les Illes Balears)

(joint work with Ingo Feldhausen, Université de Lorraine & ATILF-CNRS)

In this paper, we shed new light on the question of how narrow information and contrastive focus is prosodically and syntactically realized by monolingual (MS) and heritage (HS) speakers of Castilian Spanish (CS). Research on the syntax-prosody interface in the expression of focus by HS is relatively new (see van Rijswijk et al. 2017ab, Kim 2018, Leal et al. 2018). We, thus, add to this body of research by presenting the insights of a production test conducted with 12 Spanish MS and HS (with German as their dominant language). The theoretically oriented literature suggests that information focus must be realized rightmost, whereas stress shift (and other strategies such as clefting or fronting) are possible only for contrastive focus (Zubizarreta 1998). By contrast, experimentally oriented studies suggest that stress shift can be used for both focus types (Muntendam 2009, Gabriel 2010, Leal et al. 2018, among others). However, this apparent discrepancy can often be reduced to diatopic differences (as proposed in Feldhausen & Vanrell 2015). German-Spanish bilinguals are an interesting test case since whereas stress shift is an undisputed strategy to mark focus in German (Uhmann 1991, Féry 1993), it has often been described as dispreferred by Castilian Spanish speakers (Zubizarreta 1998, Vanrell & Fernández-Soriano 2018). Based on data from a production test designed to elicit different focus readings (narrow informational and contrastive focus on the subject and (in)direct objects), the study reveals that whereas HS realize both types of focus almost always by stress shift (L+H*), MS, in turn, distinguish information from contrastive focus through different strategies (e.g., clefting, p-movement, fronting). In the talk we will discuss whether this and other differences between MS and HS can be attributed to the influence of German or whether they rather correspond to default strategies of HS.

6 mai

The phonological status of French schwa (again)

Marie-Hélène Côté (Université de Lausanne)

The phonological status of the French schwa has been a matter of debate for decades. In this talk I take up this issue again, considering its definition, its surface realization, its representation, and its relationship to /œ/ and /ø/. I adopt a comparative perspective between different varieties of French, combining experimental and corpus data. I argue in particular that schwa is absent from the phonemic vowel inventory of northern varieties of French. While the epenthetic status of morpheme-final schwa is now largely accepted in these varieties, I also argue that morpheme-internal schwa should not be treated as distinct from /œ/ and/or /ø/ (depending on the variety), the phonetic differences observed between ‘schwa’ and [œ] and/or [ø] being predictable from the prosodic context, morphology and analogy. In addition, I consider the specific behavior of final schwa in Quebec French, which is largely restricted to a small class of words, unlike in other northern varieties of French.


Les archives des réunions passées (depuis l'automne 2012) sont disponibles ici.

The archive of past meetings of the Atelier is available here.


Pour venir présenter ou être ajouté à la liste de diffusion, merci de nous contacter à l'adresse :

If you would like to come present to the Atelier or to be added to the mailing list, please contact us at: