Edinburgh Research Explorer

Sandhi, mutation and contrast: laryngeal phonology in

Plougrescant Breton

Citation for published version: Iosad, P 2009, 'Sandhi, mutation and contrast: laryngeal phonology in Plougrescant Breton', Paper presented at Toronto-Tromsø Phonology Workshop, Toronto, Canada, 8/10/09 - 10/10/09.

Link: Link to publication record in Edinburgh Research Explorer

Document Version: Peer reviewed version

General rights Copyright for the publications made accessible via the Edinburgh Research Explorer is retained by the author(s) and / or other copyright owners and it is a condition of accessing these publications that users recognise and abide by the legal requirements associated with these rights.

Take down policy The University of Edinburgh has made every reasonable effort to ensure that Edinburgh Research Explorer content complies with UK legislation. If you believe that the public display of this file breaches copyright please contact [email protected] providing details, and we will remove access to the work immediately and investigate your claim.

Download date: 05. Oct. 2021 Talk outline

Laryngeal phonology in Plougrescant Breton: sandhi, mutation, and contrast >. Laryngeal phonology in a Breton dialect Pavel Iosad ?. Final devoicing is loss of contrast, not loss of feature Universitetet i Tromsø/CASTL @. Sandhi voicing is phonetic implementation (mostly) [email protected] A. Devoicing sandhi do not need [−voice] B. Privative laryngeal features will do C. Implications Torontø–Tromsø Phonoløgy Workshop October F–>> ?==F University of Toronto

Background Previous work

Krämer (?===)

I Île de Groix Breton (Ternes, >FD=) Argued to exhibit a ternary contrast between [+voice], [ voice], I Breton: a Celtic language, closely related to Cornish and Welsh I − and [=voice] segments I Mostly described by Celtologists, dialectologists, and historical linguists I Evidence for binary features Final devoicing is loss of features I Breton phonology remains seriously understudied (as opposed to I syntax)

I Few proper phonetic studies, mostly aural transcriptions Hall (?==E)

I What can we do? I Same dialect, same source

I Privative features with feature geometry

I Feature disalignment

I Final devoicing is loss of features and loss of contrast The present approach Breton dialects

I Traditionally divided into four groups Cornouaillais, Trégorrois, Léonais (KLT): relatively homogeneous, I Work in progress, (almost) nothing is Vnal I basis for standard language I Features are privative with feature geometry I Vannetais (south-east): very divergent, sometime even served by I “Final devoicing” is loss of contrast own literary tradition (Guillevic & Le GoU, >F=?) Devoicing sandhi is I I Île de Groix is a Vannetais dialect I Either lexical phonology Source rather messy (“phonemic” approach, not very systematic) Or failed mutation due to geminate inalterability I I Here: attempt to look at a less messy data point Argument for substance-free phonology I I Plougrescant is a Trégorrois dialect; description by Jackson (>FC=) Tested on Plougrescant Breton (Jackson, >FC=) I I more systematic

I Further outlook: extend approach to Île de Groix if possible

Consonant inventory Vowel inventory

Place i:• y: • u Manner Labial Alveolar Postalveolar Palatal Velar Uvular Laryngeal I• Y e:• Stop p b t d c é k g e• • o(:) Fricative f v s z S Z X h • œ(:) Nasal m n ñ Lateral l L E(:)• • O(:) Rhotic r Glide w j a• a:• ¯ I Length contrast for all consonants except voiced obstruents

I Length is only licensed by (main) stress Restrictions on laryngeal features Restrictions on laryngeal features

I Voiced and voiceless obstruents contrast word-initially; short I Following long stressed vowels, consonants can only be short; allophones voiceless obstruents do not occur:

(>) a. ["pesk] ‘Vsh’ (@) a. ["o:ber] ‘to do; to make; to work’ b. ["bœ:rE] ‘morning’ b. ["li:z@r] ‘letter’ c. ["lo:gOt] ‘mice’ c. ["me:l@n] ‘yellow’

I Voiced and voiceless obstruents contrast immediately following I Following short stressed vowels, consonants are long; voiced unstressed vowels; short allophones: obstruents cannot be long, so they are excluded:

(?) a. [bO"to:] ‘shoes’ (A) a. ["tap:ut] ‘to take’ b. [Sa"dEn:@t] ‘chained (participle)’ b. ["jaX:OX] ‘more healthy’ c. [kY"ry:no] ‘peals of thunder’ c. [skY"dEl:o] ‘basins’

Restrictions on laryngeal features Summary

I Leaving Vnal devoicing aside for a moment, laryngeal features I Word-Vnally following a stressed vowel, voiced obstruents are are mostly predictable: not permitted. Consonants are short following long stressed Laryngeal contrasts are allowed in the onset of the Vrst syllable vowels and long following short stressed vowels. I and of the stressed syllable

(B) a. ["to:k] ‘hat’ I Otherwise they are predictable: b. ["me:l] ‘honey’ I Voiced following unstressed (always short) vowels Voiced when single and following long stressed vowels (C) a. ["grwEk:] ‘woman, wife’ I I Voiceless (and long) when single and following short stressed b. [mEl:] ‘ball’ vowels

I What is contrastive? What is marked? Final devoicing Final devoicing in monosyllables

I The really interesting part is when a stressed vowel precedes

I Stress is normally penultimate in KLT (but not in Vannetais!), so this is mostly monosyllables and a few words with Vnal stress I At Vrst blush Vnal devoicing looks normal I If it is vowel length that is distinctive, we expect V:C# (D) a. [bYga"lEéjo] ‘children’ b. [bY"ga:lIc] ‘child’ (E) a. ["to:go] ‘hats’ ¯ b. ["to:k] ‘hat’ I But what about vowel length? I And cf. minimal pairs like I This is a good question (F) a. ["kas:] ‘send!’ ([s] never voiced, French borrowing) b. ["ka:s] ‘cat’ (cf. orthographic kaz) ¯

Final devoicing in monosyllables Final devoicing in monosyllables

I This isn’t really devoicing in view of what we know about Does real Vnal devoicing happen? Well, yes quantity and voicing I I There is variation described by Jackson (>FC=) as “free”, and I This is incomplete neutralization especially with coronals I Confer real devoicing: I Context probably unknowable; the ambition here is at best to Vnd (>=) a. [lO"go:d@n] ‘mouse’ which representations are involved b. ‘to hunt mice’ [lO"gOt:a] (>?) [ty:t]∼[tYt:] ‘people’ (orthographic tud)

I Side note: it isn’t always about voicing per se: I More examples to come immediately below, as they involve (>>) a. ["rO:his] ‘people of ar Roc’h’ sandhi to which we now turn b. ["rO:X] ‘ar Roc’h (placename)’ I What about lexically voiceless Vnals? These are relatively few, French borrowings of various antiquity, and behave as expected, I Not really surprising if you know (some) [h] is historically *G, but cf. (F-a) must be accounted for Sandhi Sandhi

I The traditional view (Stephens, >FF@; Favereau, ?==>) is essentially that all consonants are voiced in sandhi before I Plus there is the devoicing sandhi that is the focus of Krämer [+voice] segments (?===) and Hall (?==E) I For Île de Groix Ternes (>FD=) describes it as a lexical distribution: (>@) a. ["pwe:l@z "˜a.˜O] ‘if you saw me’ some words, and only these words, devoice initial obstruents b. [­mab "ne:we] ‘new son’ ¯ following an obstruent c. [­pOb "bi.@n] ‘little youth’ I For Plougrescant, Jackson (>FC=) is less concerned: “sometimes” And voiceless before voiceless consonants I (>B) a. ["la:tt˜ı] ‘said to me’, cf. [d˜ı] ‘to me’ ¯ (>A) a. [­map "hi:r] ‘tall son’ b. ["kankuS] ‘>== times’, cf. ["tErguS] ‘thrice’ ¯ b. [@n ­dyt "kap:ap] ‘the able people’

Sandhi Outline of analysis

I In the narrative texts given by Jackson (>FC=), the sandhi rules are often violated

I Especially with regard to sandhi voicing (>C) a. [­map "dy:] ‘black son’ Outline feature analysis ¯ I b. [­mErX "va:t] ‘good girl’ ¯ I Argue that Vnal devoicing without length permutations is a c. ["dwa:n tœs "di:w˜ı]‘the fear that you have of me’ phonetic process Argue that sandhi voicing is the Wip side of Vnal devoicing I Jackson (>FC=) explains the the texts were dictated at a slow pace I I However, some (in fact most) of the examples, such as (>C-a) and I Unify some devoicing sandhi with “failure of mutation” (>C-b), are transcribed with a secondary–main stress rhythm; I Tentatively propose that other devoicing sandhi are an artifact of these are possibly genuine connected phrases univerbation I Thus failure of sandhi is not necessarily an artefact of dictation I Note that vowels outside main-stressed syllables are shortened, so the preservation of length contrasts under devoicing does not work in the same way when stress is secondary Feature analysis Feature analysis

I Before we even discuss Vnal devoicing, we should solve the I In substance-free phonology with emergent privative features, [voice]/[spread glottis] problem this point is rather moot

I Phonetics rather poorly understood I We are interested in the patterning, whether the “voiceless” I Voiceless stops are described as aspirated (at least initially) at Le obstruents are labelled [spread glottis] or [voiceless] (cf. Blaho, Bourg Blanc (Falc’hun, >FB>) and Saint-Pol-de-Léon (Sommerfelt, ?==E) is irrelevant >FDE), but these are both Léonais I Or voiced stops are [voice] or [stiU], of course No mention of aspiration is made for Plougrescant by Jackson I I propose that in Plougrescant Breton “voiceless stops” are (>FC=, >FCD) I [voiceless] and “voiced stops” do not bear a laryngeal feature, but In all cases the voiced stops are described or assumed to be voiced I do have a laryngeal node One possible point: at Plougrescant fricatives underwent a I I return below to why nodes are better than features context-free voicing (“new lenition”), cf. Southern English I Fricative Voicing, which Honeybone (?==Ba) takes as evidence for I Main reason is restricted distribution: only initial and stressed [spread glottis]:∅ syllables, both reasonable contexts for positional faithfulness (Beckman, >FFF; Smith, ?==?) I But Honeybone (?==Ba) himself admits the analysis of fricatives should not be spread to stops uncritically I We need to make reference to this feature to derive the restrictions (but not to describe Vnal devoicing as I argue below)

I In that sense it is “marked” (Trubetzkoyan markedness)

Final devoicing Assumptions of analysis

I I propose that Vnal devoicing is in fact loss of the laryngeal node, i. e. it is the exclusion of the very possibility of contrasting for I Vowel length distinctive in main-stressed syllables: laryngeal features faithfulness  markedness in this context *[voiceless] above M[vcl] I Devoiced stops are a third phonological category: they behave I diUerently from true voiceless stops in that they do not obey I Except for positional faithfulness: M[vcl]/Initial and length-related restrictions M[vcl]/σ´ above *[vcl]

I True voiceless stops cannot follow long vowels; devoiced stops I Bimoraic template for main-stress syllable (M W): can McGarrity (?==@); Bye & de Lacy (?==E)

I In particular, what is the diUerence between Vnal devoicing as in I Final devoicing driven by a constraint *Lar/_]Wd militating against [ty:t] and Vnal devoicing with gemination as in [tYt:]? any segments with a laryngeal node at the end of a (morphological?) Word I No tableaux in analysis (but hopefully it is pretty theory-independent) Medial obstruents: /Vt/ Medial obstruents: /Vd/

I Obstruents are long and voiceless following short stressed vowels I Assuming richness of the base, what happens with voiced σ´ σ obstruents after short vowels? σ´ σ

t a p u t t a b u t Lar ˚= Lar [vcl] I Assume a constraint *Lar/µ: geminates without laryngeal I The voiceless obstruent piggybacks on M W to be speciVcations exist in the language (geminate sonorants) parsed into the stressed syllable and thus keep [vcl] I This is of course outranked by positional faithfulness to [vcl] to I This is assuming (as I do) that faithfulness to vowel length is derive the previous case undominated

Medial obstruents: /Vd/ Medial obstruents: /V:t/

I This is a simple case σ´ σ

I The obstruent loses its laryngeal speciVcation in order to become moraic for the beneVt of M W µ µ I Laryngeally unspeciVed obstruent geminates are realized as voiceless for obvious phonetic reasons l o kg O t I Maybe these are excluded by Lexicon Optimization since the learner never really has to posit /b:/? Lar ˚ = [vcl]

I No superheavy syllables, so [vcl] cannot be saved Final devoicing: voiced stops Final devoicing: voiced stops

I No Lar node word-Vnally I Final consonant is extrametrical (so maybe no Lar node not licensed by prosodic structure?) I Laryngeally unspeciVed obstruents in pausa are realized as voiceless, phonetic reasons are well-known I Stress: ultimate if V: in Vnal syllable, else penultimate. Moraic trochee, but then Vnal VC˘ must be L I What if our [vcl] is really [spread glottis] in this dialect? It is apparently unproblematic to have aspiration as the Wd I phonetically natural realization of phonological underspeciVcation (Vaux & Samuels, ?==B) σ´ I What about cases such as [ty:t]∼[tYt:]? µ µ I I propose this is real Vnal devoicing, i. e. the imposition of the [vcl] feature at word (phrase?) edges (Iverson & Salmons, ?==D)

t o g I First let’s look at underlying voiceless obstruents ˚= Lar

Final voiceless stops True Vnal devoicing

I The [vcl] obstruent becomes moraic to satisfy M W, I In this scenario, forms such as [tYt:] for /tyd/ imply that the so the restrictions on vocalic quantity hold constraint driving Vnal devoicing is ranked over faithfulness for vowel length. Wd Wd

σ´ σ´ = µ µ µ µ

k a s = t y dt Lar Lar [vcl] [vcl] Final devoicing: summary Voicing sandhi

I I have argued that what looks like normal Vnal devoicing is in fact the deletion of a Lar node, or absence of contrast I In this system, voicing sandhi arise from two sources Before sonorants: laryngeally unmarked stops are voiced in the I Further evidence: Vnal /v/ does not always neutralize with /f/ I phonetically: Jackson (>FC=) writes [v] phonetics ˚ Sonorants do not contrast for laryngeal features, so they do not I We know [v] is aerodynamically complicated (Padgett, to appear) I have a [Lar] to spread I So this would be consistent with a phonologically underspeciVed /v/? I Explains variability (pause-sensitivity?) ˚ No need to have (contrastive) laryngeal features for sonorants I Final devoicing as Vnal fortition (Iverson & Salmons, ?==D) is I distinct from this process and also attested (Krämer, ?===; Blaho, ?==E; Hall, ?==E) Grazing other dialects: Vnal devoicing is optional at I [­mab "ne:we] = /mab ne:we/ I ¯ ˚ Saint-Pol-de-Léon (Sommerfelt, >FDE) (?)

Voicing assimilation sandhi Devoicing sandhi

I Some examples of devoicing sandhi

Before obstruents, we are faced with two options (>D) a. ["la:tt˜ı] ‘said to me’ I b. ¯ ‘I Vnd, I consider’ (lit. ‘I get to Same as above [me "gaft˜ı] I me’) Explains possible devoicing even before voiced obstruents I c. [­dO "wen:@k "tit] ‘your two sous’ (lit. ‘two sous to I Possibly predicts that under certain phonetic circumstances Vnal consonants may be voiced before voiceless consonants? you’) Spread of Lar, with [vcl] if need be I I Prepositions are overrepresented Variation must have a phonological explanation (stochastic I Actually, this is also true of Île de Groix! ranking?) I Devoicing sandhi crucial piece of evidence in favour I (>E) [tra n@"vaNktemp] ‘we don’t miss anything’ (lit. ‘nothing is missing to us’)

I What’s with the prepositions? Detour >: mutation Detour ?: prepositions

I Breton is (widely?) known for its initial consonant mutation I Historically, prepositions in Brythonic have tended to undergo the eUects of soft mutation/lenition in a context-free way I Here we are only interested in lenition Old Welsh and Old Breton gurth ‘through’, Modern Welsh wrth, Underlying I p t k b d g m Modern Breton ouzh Mutated b d g v z h v I Old Welsh di ‘to’, Modern Welsh i (via *[Di]) The interesting bit is the voicing of voiceless stops I I Modern Welsh variation: trwy ∼ drwy ‘through’

Detour ?: prepositions Back to Breton: devoicing sandhi

I Why is this important? I I propose that (some) Breton devoicing sandhi reWect the same I At least in Welsh, there is evidence that the new initial consonant incomplete lexicalization of the voiced stops is not fully lexicalized I Consider lavare[t t]iñ In particular, *gan ‘with’ is historically *kant I t [L] t I The conjunction a ‘and’ causes a mutation whereby voiceless ˚ stops are spirantized to [f T X] but voiced ones are unaUected Lar I We expect *a gan for ‘and with’, but it is actually a chan (Morgan, >FB?; Ball & Müller, >FF?) [vcl] The same is true of dros and drwy though there the variants with I Normally, [L] docks to the following , e. g. due to MF  the voiceless stop survive in the modern language I /t/ (Wolf, ?==D) I So maybe gan is really [L]can underlyingly I But not when the Lar node spreads to a preceding root node I Where [L] is the autosegment (Wolf, ?==D) Devoicing sandhi Devoicing sandhi

I Further evidence for this approach comes from so-called “failure of mutation” (Jackson, >FCD, §AE>)

I Lenition of voiceless stops is said to “fail” when an adjective (given the necessary morphosyntactic conditions) follows an This can be for any number of reasons I obstruent-Vnal noun I Some version of geminate inalterability I Structure sharing inhibits weakening processes (Honeybone, I But with sonorant-Vnal nouns or voiced stops mutation happens ?==Bb) I Cf. kaer ‘beautiful’ I Under certain assumptions, the structure shown is not convex (Scobbie, >FFD) (>F) a. un dimezell gaer a maiden beautiful b. ur vaouez kaer a woman beautiful

I Morphosyntax actually irrelevant, since other triggers of this mutation are sonorant-Vnal

Devoicing sandhi What, autosegments?

I In previous work I have doubted that the autosegmental approach I This is the same phenomenon: an autosegment normally leading is suited to Brythonic Celtic mutations (cf. also Green, ?==C) to voicing is inhibited by spreading of the Lar node I I think these data are actually pretty solid evidence for I Following sonorants the Lar node can’t spread since sonorants autosegments or at least for a phonological analysis with a Lar node are never well-formed I Breton is less problematic than Welsh morphosyntactically But this time we have much better evidence for the autosegment I I Breton mutation seems to be genuinely sensitive to prosody being there (Pyatt, ?==@) The same data are described by Ternes (>FD=) in an extremely I I There is still the problem of doing mutation phonologically: Wolf convoluted way. . . (?==D) covers only a small subset The generalization: if an obstruent is voiced by an autosegment, I I In particular, the autosegment should cause deletion of [vcl] in it can resist voicing by spreading Lar to a preceding obstruent the current approach

I Problem! But see Bye & Svenonius (?==F) for an approach. . . More devoicing sandhi More devoicing sandhi

I Jackson (>FCD, §AED): “provection in common phrases” I Are these actually phrases or words?

I Saint Gonéry is the patron saint of the local chapel I Other types of devoicing sandhi do not seem to fall under this rubric

(?=) a. [sankO"ne:ri] ‘Saint Gonery’ b. ["kankuS] ‘thrice’, cf. ["tErguS] ‘thrice’

I I propose that here devoicing is due to univerbation, i. e. the relevant words are now compounds

I Word-internally voiceless obstruent clusters are (nearly) universal (also noted by Hall, ?==E for Île de Groix) I ‘Thrice’ might well be a single word, cf. Welsh dwywaith ‘twice’, and in fact *[guS] is the reduced form, cf. stressed gwej ‘time, occasion’

I Etc. Photo credit: SteUen Heilfort. Source.

Summary and outlook: sandhi Loss of feature or loss of contrast

I Voicing sandhi are mostly due to phonetic implementation of laryngeally unspeciVed obstruents in a phrasal context I Here I have argued that Breton presents examples two types of I Some devoicing sandhi are due to inhibition of autosegmentally Vnal devoicing

induced voicing I Final devoicing as loss of contrast: cf. the arguments of Harris (?==F) for FD as weakening I Others might possibly be not phrasal sandhi at all I Final devoicing as edge alignment: Vnal fortition (Iverson & I Both of these phenomena seem to be cross-dialectal, so the Salmons, ?==D) account possibly extends to Île de Groix: Take-home message here: there is no process of “Vnal devoicing”, I Prepositions I “Vnal weakening” or “Vnal fortition” that we can speak of in I More examples: the “devoicing” word [b@"nak] ‘any’ is Middle Breton pennac (Lewis & Piette, >FC?, §AB) universal terms I The “provection in common phrases” (univerbation) is described Argument for substance-free phonology as pan-Breton. Examples of devoicing sandhi in Île de Groix I include ‘grey peas’ and ‘little Vnger’—intuitively good candidates for univerbation Final devoicing as phonetics Ternary contrasts

I Growing body of work on Vnal devoicing (and generally laryngeal assimilation) as a “low-level phonetic process” I Krämer (?===) argues that the presence of both voicing and I The Paradestück here is of course Dutch (Ernestus & Baayen, devoicing necessitates binary features, i. e. a ternary contrast ?==C, ?==D; Jansen, ?==D) I Related issue: UUmann (?==F) asks how to distinguish between I Possibly others (e. g. the disputed claim for Polish) categorically voiceless and laryngeally unspeciVed stops in a

I Breton seems to show quite good evidence for incomplete privative system neutralization I The answer is of course feature geometry Laryngeally unspeciVed segments interpreted by the phonetics as I I Objection of UUmann (?==F): but this is an overgenerating devoiced or aspirated rather than [−voice] or [spread glottis] notational variant of binary features speciVed

I Needs careful cross-linguistic study

Ternary contrasts Tiers or features?

I One answer: who says we never need bigger feature geometry trees? It is correct that arboreal representations can have many I Here I use class nodes (as in e. g. Avery, >FFC) levels, but maybe this is empirically better? I Blaho (?==E): no need for nodes if features can do the job, e. g. I Related answer: binary features are no more God-given/less substitute Lar with [obst] since only obstruents are laryngeally stipulative: [=voice], [>voice] and [?voice] are also a notational speciVed variant, but these are as overgenerating as trees I Gives strange results for Breton, since Vnal devoicing is driven by I Reason: three independent values of [F] cannot capture *[obst]: works formally but how insightful is it? Are the devoiced implication relations in the same way that feature geometry can obstruents sonorants? (Well, why not) Here I argue that the feature geometry/underspeciVcation I I Here: nodes are necessary approach is empirically more adequate than one based on [±voice] spreading Tiers or features? Summary

I If features can only attach to nodes, the presence of a node (even with no features) is the formal correspondent of contrastive I New interpretation of Breton data speciVcation I Possible cross-dialectal extension I Sort of answers the concern of UUmann (?==F) on the diUerence Privative features can do the job between two types of feature absence I I Feature/node geometry is preferable to binary features and I Without nodes, how do we deVne tiers and all the autosegmental (possibly) to node-less geometry. phenomena that come with them?

I Null hypothesis: all and only features dependent on a speciVc Trugarez! node are on the same autosegmental tier

I Field of empirical inquiry

Ernestus, Mirjam & R. Harald Baayen. ?==D. Intraparadigmatic eUects Avery, Peter J. >FFC. The representation of voicing contrasts. Ph.D. on the perception of voice. In van de Weijer & van der Torre (?==D), thesis, University of Toronto. >B@–>D@. Ball, Martin J. & Nicole Müller. >FF?. Mutation in Welsh. Falc’hun, François. >FB>. Le système consonantique du breton. Rennes: London—New York: Routledge. Pilhon. Beckman, Jill N. >FFF. Positional faithfulness. Ph.D. thesis, University Favereau, Francis. ?==>. Grammaire du breton contemporain. of Massachusetts, Amherst. Morlaix: Skol Vreizh. Blaho, Sylvia. ?==E. The syntax of phonology: a radically substance-free approach. Ph.D. thesis, University of Tromsø. Green, Anthony Dubach. ?==C. The independence of phonology and morphology: The Celtic mutations. Lingua >>C(>>). >FAC—>FEB. Bye, Patrik & Paul de Lacy. ?==E. Metrical inWuences on fortition and lenition. In Joaquim Brandaõ de Carvalho, Tobias Scheer & Philippe Guillevic, Augustin & Pierre Le GoU. >F=?. Grammaire bretonne du Ségéral (eds.), Lenition and fortition, vol. FF, Studies in generative dialecte de Vannes. Vannes: Lafolye Frères. grammar, >D@–?==C. Berlin: Mouton de Gruyter. Hall, Daniel Currie. ?==E. On the voicing system of île de Groix Bye, Patrik & Peter Svenonius. ?==F. Extended exponence and Breton. Presentation at the Workshop on phonological voicing non-concatenative morphology. MS., University of Tromsø. variation, Leiden. Ernestus, Mirjam & R. Harald Baayen. ?==C. The functionality of Harris, John. ?==F. Why Vnal obstruent devoicing is weakening. In incomplete neutralization in Dutch: the case of past-tense formation. Kuniya Nasukawa & Phillip Backley (eds.), Strength relations in In Louis M. Goldstein, D. H. Whalen & Catherine T. Best (eds.), phonology, vol. >=@, Studies in generative grammar, F–AC. Berlin: Laboratory phonology E, ?D–AF. Berlin: Mouton de Gruyter. Mouton de Gruyter. Honeybone, Patrick. ?==Ba. Diachronic evidence in segmental Jansen, Wouter. ?==D. Dutch regressive voicing assimilation as a “low phonology: the case of obstruent laryngeal speciVcation. In Marc level phonetic process”: acoustic evidence. In van de Weijer & van Oostendorp & Jeroen van de Weijer (eds.), The internal van der Torre (?==D), >?@–>B>. organization of phonological segments, vol. DD, Studies in Krämer, Martin. ?===. Voicing alternations and underlying Generative Grammar, @>F–@BA. Mouton de Gruyter. representations: the case of Breton. Lingua >>=. C@F–CC@. Honeybone, Patrick. ?==Bb. Sharing makes us stronger: process Lewis, Henry & J. R. F. Piette. >FC?. Llawlyfr Llydaweg Canol. inhibition and segmental structure. In Philip Carr, Jacques Durand Caerdydd: Gwasg Prifysgol Cymru. & Colin J. Ewen (eds.), Headhood, elements, speciVcation, and McGarrity, Laura W. ?==@. Constraints on patterns of primary and contrastivity: Phonological papers in honour of John Anderson, secondary stress. Ph.D. thesis, Indiana University. >CD–>F?. Amsterdam: John Benjamins. Morgan, Thomas John. >FB?. Y treigladau a’u cystrawen. Caerdydd: Iverson, Gregory K. & Joseph C. Salmons. ?==D. Domains and Gwasg Prifysgol Cymru. directionality in the evolution of German Vnal fortition. Phonology Padgett, Jaye. to appear. Russian voicing assimilation, Vnal devoicing ?A(>). >?>–>AB. doi:>=.>=>D/S=FB?CDBD=D==>>@@. and the problem of [v]. Natural Language and Linguistic Theory . Jackson, Kenneth Hurlstone. >FC=. The phonology of the Breton dialect Pyatt, Elizabeth J. ?==@. Relativized mutation domains in the Celtic of Plougrescant. Études celtiques F. @?D–A=A. languages. In Elsi Kaiser & Sudha Arunachalam (eds.), Proceedings Jackson, Kenneth Hurlstone. >FCD. A historical phonology of Breton. from the Penn Linguistics Colloquium ?C. Philadelphia: University Dublin: DIAS. of Pennsylvania.

Scobbie, James M. >FFD. Autosegmental representation in a declarative constraint-based framework. New York: Garland. Smith, Jennifer L. ?==?. Phonological augmentation in prominent positions. Ph.D. thesis, University of Massachusetts, Amherst. Sommerfelt, Alf. >FDE. Le breton parlé à Saint-Pol-de-Léon. Oslo, van de Weijer, Jeroen & Jan Erik van der Torre (eds.). ?==D. Voicing in Bergen, Tromsø: Universitetsforlaget. Édité par Frañsez Falc’hun et Dutch, vol. ?EC, Current issues in linguistic theory. Amsterdam: Magne Oftedal. John Benjamins. Stephens, Janig. >FF@. Breton. In Martin J. Ball & James Fife (eds.), The Wolf, Matthew. ?==D. For an autosegmental theory of mutation. In Celtic languages, @AF–A=F. London and New York: Routledge. Michael O’Keefe, Ehren Reilly & Adam Werle (eds.), University of Ternes, Elmar. >FD=. Grammaire structurale du breton de l’île de Groix Massachusetts Occasional Papers in Linguistics @?: Papers in (dialecte occidental). Heidelberg: Carl Winter Universitätsverlag. Optimality Theory III, @>B–A=A. Amherst: GLSA. UUmann, Christian. ?==F. To (bi) or not to (bi). Presentation at the Privative Project workshop, Old World Conference in Phonology C, Edinburgh. Vaux, Bert & Bridget Samuels. ?==B. Laryngeal markedness and aspiration. Phonology ??(@). @FB–A@C. doi:>=.>=>D/S=FB?CDBD=B===CCD.