Epenthesis

This paper analyses vowel and consonant epentheses in the history of German from the typological perspective of syllable and word languages (Auer 1993, 2001; Szczepaniak 2007a). The prosodic analysis reveals a shift of the epenthetic domain from the phonological syllable in Old High German (OHG) towards
the phonological word in Early New High German (ENHG): The OHG epentheses improved syllables by simplifying consonant clusters, e.g. OHG durh > duruh ‘through’ (CVCC > CV.CVC) or resolving hiatuses as in OHG sāen > sāwen ‘to sow’ (CV.VC > CV.CVC). In contrast, epentheses operating since ENHG have highlighted the phonological word by producing hiatuses and, thus, raising the word-internal sonority, e.g. MHG bûr(e) > (E)NHG Bauer ‘peasant’ (CVC > CVṾVC) or by producing complex word-external syllable codas, e.g. MHG mâne > (E)NHG Mond ‘moon’ (CVCV > (CVC) > CVCC). This domain shift is part of the phonological change of German from a syllable to a word language type, which has brought about a series of word-optimizing processes.

"Fonologia catalana" is a comprehensive classical account of the phonology of Catalan, written in the 1990s within the tennets of the standard generative phonology theory, with some additional information  cast in autosegmental theory. The book includes chapters on vowel phenomena, syllable structure, consonant word-final phenomena, assimilatory and other consonantal contact phenomena,  pronominal clitics, and suprasegmentals. Each chapter includes some exercises.

The behavior of epenthetic consonants has received a great deal of attention in the recent phonological literature because of the significantly different ways in which it is treated by constraint-based and rule-based theories and because of the formal consequences attendant on the empirical comparison of these two treatments. Optimality Theory requires that the choice of (regular) epenthetic consonant in a given language be predictable from the interaction of independently-motivated inventory constraints and well-formedness constraints. Rule-based formalisms on the other hand allow rules of the type “Ø → [g] / V_V”, wherein the choice of epenthetic segment is synchronically arbitrary. In this paper I provide empirical evidence from a range of languages demonstrating that the prediction of rule-based phonology is correct: a language can choose any consonant for insertion by regular rule. I argue that unnatural systems of this type are accounted for most efficiently and insightfully in a rule- driven framework; existing OT implementations can be altered to account for the relevant phenomena, but only at the cost of abandoning the central theoretical tenets that have been claimed to give OT the advantage over derivational theories. The OT perspective moreover loses essential insights of rule-based phonology into the relationship between historical change and language acquisition, and between automatic and morphologically-conditioned phonology.

Este artigo tem como objetivo fazer uma análise teórica da epêntese vocálica no português brasileiro (PB), à luz da teoria da fonologia
lexical. Dessa forma, o trabalho desenvolve-se, sobretudo, a partir da busca de evidências na bibliografia já produzida sobre esse assunto.
Com base nesse ponto de vista, pode-se observar uma tendência entre os estudiosos em classificar a epêntese como uma regra tanto lexical como pós-lexical. Este trabalho sugere que a regra de epêntese em PB é um processo fonológico que opera, eminentemente, no domínio lexical.

This study attempts to present an Optimality Theoretic (Prince and Smolensky,1993) analysis of the svara sandhi changes occurring in Odia (An Indo-Aryan language spoken in the eastern state of Odisha, India). Odia, like other major Indian languages such as Bengali, Telugu, Malayalam, etc., has been influenced by Sanskrit and has hence subsumed the phenomenon of sandhi occurring in Sanskrit.

The phenomena of two sounds combining to form a new sound ( /ɔ/ + /ɔ/ = /a/) e.g. sɔsɔ + ɔŋkɔ = sɔsaŋkɔ ‘moon’ or the insertion of glides in certain other constructions (/ɪ/ + */ɪ/ = /j/) e.g. ɪt̪ɪ + ad̪ɪ= ɪt̪jadɪ ‘etcetera’, (/u/ + */u/ = /w/) e.g. /su/ + /aɡɔt̪ɔ/ = swaɡɔt̪ɔ ‘welcome’ or how two vowels of differing heights combine to result in a sort of Vowel Harmony where instead of one vowel influencing the other, both vowels influence each other and result in a sound which has qualities of both vowels (/ɔ/, /a/ + /ɪ/ = /e/ ) e.g. /ɡɔɳɔ/ + /ɪsɔ/ = /ɡɔɳesɔ/ ‘Lord Ganesha’; these are some of the processes that will be looked into.

This paper uses the Optimality Theoretic framework to explain these processes. Newly developed constraints such as COALESCENE, *Diphthongs, (low, back V + low, back V = /a/) are proposed in this study along with certain other common and well-established faithfulness constraints such as MAX-V and DEP-V. The study analyses the svara sandhi process in Odia which has evolved from Sanskrit and shares the same features in most constructions while simultaneously showing the applicability of Optimality Theory in such a study.

Keywords: sandhi; svara sandhi; Odia; Optimality theory; vowel harmony; epenthesis

I present previously unreported data from Serbo-Croatian suffixation, as well as from the language game šatrovački, that contradicts generalizations made by Zec (2002, 2007) on the distribution of syllabic sonorants in this language. Contrary to Zec (2002, 2007), I argue that two productive strategies for repairing syllable structure are syllabification and deletion. I show that instead of assuming different grammars for native lexicon, loanwords and šatrovački forms, all the data can be shown to follow the same grammar. I explain the apparent counterexamples by recourse to stem allomorphy (McFadden, 2016; Steriade, 2016), and marking at the underlying level.

"This dissertation explores the phonological representation and the phonetic realization of prosodic prominence in Persian. It comprises two related parts: the first part addresses prosodic phrasing in Persian sentences, while the second part deals with phonetic correlates of prosodic prominence by reporting conducted production and perception experiments. The phonological part is carried out within the framework of Prosodic Phonology, and aims at determining the prosodic structure of Persian from foot level, up to utterance level. By adopting Optimality Theory, it tries to explain how morphosyntax-phonology interface constraints together with prosodic markedness constraints form the prosodic structure of the language. It begins with foot level and suggests that in languages like Persian which have one non-iterative weight-insensitive edgemost stress per word, a single foot which is edge-aligned with the minimal Phonological Word best explains the prosodic pattern at word level. This part also focuses on prosodic differences between lexical words and weak function words (clitics) and shows that any attempt to describe the prosodic structure of Persian without addressing this crucial difference, will not be able to provide explanation for a wide range of phenomena. This study suggests that proclitics and enclitics behave asymmetrically in Persian: enclitics prosodize as affixal clitics, while proclitics are free clitics. Next, it addresses the problem of weak function words which are not a part of their preceding or following XPs, and demonstrates how the phonological well-formedness constraints determine the direction of cliticization in these XP-external function words. It also deals with the issue of clitic clusters in Persian which was not explored in the previous works. Another contribution of this study is reclassification of so-called exceptionally initial-stressed words.
This dissertation also deals with the longstanding problem of Ezafe constructions and by reviewing previous proposals on the prosodic structure of these constructions, based on phonological evidence and phonetic observations suggests that each lexical word in an Ezafe construction maps onto a Phonological Phrase, and the Ezafe morpheme phrases with its preceding material to satisfy the phonological well-formedness constraint ONSET. The prosodic structure of XP-external clitics such as the Ezafe morpheme is explained by adopting a syntax-prosody interface constraint namely MAP-XP, that bans two sister XPs inside a single Phonological Phrase. This study proposes a ranking of OT constraints by which the prosodic structure of Ezafe constructions and other syntactic phrases such as DPs and VPs can be predicted and explained uniformly. It also proposes that the interaction between morphosyntax-phonology interface constraints and prosodic markedness constraints determine prosodic constituents of all levels and their heads, and other constraints require the heads of phonological phrases to be associated with audible accents. The rightmost Phonological Phrase in an Intonational Phrase is the head. This head associates with an accent which is perceived more prominently than the other accents. One further issue explored here is the fact that in lexicalized Ezafe constructions and also in the ones containing given/old information, some words may appear without audible accent.
The phonetic difference between final and non-final accents is the subject of the second half of this dissertation. Previous research on Persian has shown that the main acoustic correlate of prosodic prominence is f0. This study reports production and perception experiment results conducted in order to answer the question whether final (nuclear) accents are perceived more prominently than the other ones only because they are not followed by any other accent, or because they are phonetically different from the non-final (pre-nuclear) accents. The results of production experiments reveal that nuclear accented syllables have a lower f0 range, but a longer duration in comparison with pre-nuclear accented ones. Other parameters such as overall intensity, spectral tilt and vowel quality do not differ significantly in the two types of accents. Perception experiments reveal that native listeners can indeed distinguish the two types of accents without having access to the portion of the utterance that follows the final accent. This proves that the two types of accents are phonetically different. Perception tests also show that the difference between the shapes of f0 curves in the two types of accents is the main acoustic parameter that helps the listeners distinguish them from each other. In pre-nuclear accented words, the f0 peak is at the right edge of the metrically strong syllable, and the curve has a rising slope at this point. In these syllables, the peak may even occur on the initial syllable of the following word. However, in the syllables associated with nuclear accents, the f0 peak is located inside the syllable, and the curve has a falling slope at the right edge of the syllable.
If the f0 at the right edge of a nuclear accented syllable is manipulated and raised so that the f0 peak is moved to the right edge, the native listeners will perceive the word containing this syllable as a pre-nuclear accented word. This study also shows that duration alone cannot cue the difference between the two types of accents. However, when accompanied by f0 changes, it can help the listeners distinguish the two accents more easily and more efficiently.
Downloadable at: http://roa.rutgers.edu/content/article/files/1314_hosseini_1.pdf
"

En este artículo se describe la epéntesis vocálica en el zapoteco de San Bartolo Yautepec y se argumenta que es un proceso fonológico que opera tanto en el plano léxico de la palabra fonológica como en el postléxico dentro de la frase entonativa, aunque con distintas características en cada dominio prosódico. Asimismo, se sostiene que la epéntesis vocálica cumple la función de reparar secuencias de conso-nantes poco perceptibles. Así pues, los enclíticos y la palabra fonológi-ca manifiestan el mismo patrón de epéntesis, lo cual permite incluir a los primeros en el dominio de la palabra fonológica por un proceso de adjunción que no afecta la posición del acento en raíces.

Through comparison of regular sound correspondences in three closely related Tibeto-Burman (TB) languages, Ersu, Lizu, and Duoxu (collectively "ELD"), informed by external comparison with other TB languages and recent phonetic analyses of the production of voiceless nasals, we reconstruct *fricative-nasal sequences in their common ancestor, Proto-ELD. In the development of these historic clusters, two pathways of change can be recognized. Their difference lies in the divergent relative phasing of velic and oral gestures in the original fricative-nasal sequences: (i) fricative weakening (from a tight cluster): *FN > N ̥ > h̃ > x (ii) fricative strengthening (from a loose cluster): *F-n > *F-t > t > k or *F-n > s The different reflexes observed in Ersu, Lizu, and Duoxu represent different points along these two developmental pathways. These reconstructions and pathways of development have implications for our understanding of both universal (phonetic) and language-specific aspects of change in fricative-nasal sequences. The first pathway makes it possible to explore the process of nasal devoicing beyond voiceless nasals so as to enrich our understanding of nasal devoicing in natural languages. The coexistence of two opposite pathways of change, on the other hand, provides insights into the morphological and syllabic structure of words with contiguous fricative-nasal sequences in ELD languages at different points in time-insights that may be valuable in examining the history of other languages and language families beyond the ELD cluster.

This dissertation examines the realization of vowel sequences across word boundaries in Arabic-Spanish bilinguals and Spanish monolinguals in Puerto Rico, as in the phrase hombre inteligente ‘intelligent man’. Spanish and Arabic have different syllable structures, which cause differences in their strategies of resyllabification. The onset is obligatory in Arabic (Abu-Salim, 1982; Broselow, 2017; Watson, 2007), while this is not the case in Spanish (Hualde, 2014). In Arabic then, resyllabification occurs to accommodate syllable weight and comply with the permitted types of syllables in the language. Conversely, in Spanish vowel sequences may be realized as hiatus, diphthongized, or coalesced. In addition, one vowel may be elided, and infrequently, epenthesis can occur. Given these linguistic differences between Arabic and Spanish, the aim of the present study was to examine how Arabic-Spanish bilinguals are affected by the linguistic constraints of Arabic when resyllabifying vowel sequences across word boundaries in Spanish. Linguistic factors including stress, vowel height, and word length, among others, were examined. Additionally, this study aimed to investigate how the social factors language dominance, age of acquisition, gender and age affect these realizations.
In order to examine vowel sequences across word boundaries, a reading task and sociolinguistic interviews were used. Bilingual participants also completed the Bilingual Language Profile questionnaire in Spanish (Birdsong, Gertken & Amengual, 2012) to elicit information on language dominance, proficiency, age of acquisition, and language use. The participants in the reading task were divided among 10 monolinguals and 14 bilinguals, with a gender distribution of nine women (six monolingual, three bilingual) and 15 men (four monolingual, 11 bilingual); age range 19-65 years. The participants for the sociolinguistic
interviews consisted of four monolingual men and 11 bilingual men; age range 19-55 years. All ix
participants had at least an Associate’s degree. A total of 1921 tokens were produced in the reading task and 427 tokens were produced in the sociolinguistic interviews.
The findings of the present study showed that Arabic-Spanish bilinguals and Spanish monolinguals differed in their realization of vowel sequences. Bilinguals employed a strategy of glottal epenthesis more frequently than their monolingual counterparts in both tasks, although this was not significant. However, the results also showed that bilinguals diphthongized sequences significantly more than monolinguals. Within the bilingual group, those who were dominant in Arabic glottalized significantly more than Spanish-dominant bilinguals in both tasks. However, an examination of the sociolinguistic interviews showed that Spanish-dominant bilinguals favored more reduced forms (diphthongization, elision, coalescence). In both tasks, monolinguals’ and bilinguals’ realizations were significantly impacted by the linguistic factors stress and vowel height. Sequences where one of the vowels was a low or mid vowel were conducive to realizations of vowel retention (i.e., glottalization, hiatus, or two homogeneous vowels), while sequences containing a high or low vowel were conducive to reduction (diphthongization, elision, coalescence). When one of the two vowels in contact was stressed, particularly the second vowel of the sequence, reduction was less likely.
These findings show that language dominance strongly impacts how vowel sequences are realized in Arabic-Spanish bilinguals. Language dominance appeared to be a stronger indication of phonetic outcomes than age of acquisition in these bilinguals. Lastly, although infrequently, the results also showed that monolinguals of Puerto Rican Spanish utilize glottal epenthesis between vowels as a strategy for hiatus resolution. The findings suggest that there are universal linguistic conditions that favor glottalization, in addition to factors associated with bilingualism, such as language dominance and the onset of bilingualism.

フランス語を学ぶ多くの人は音節に対して非常に困難を感じているように見受けられます。フランス語を母語とする者は音節を自然に区切ることができます。しかし、日本語を母語とする話者は、日本語の音節構造に基づいてフランス語を発音しようとするので、しばしば区切り方を誤ってしまいます。よくある間違いは「音挿入」です。これは言葉の音、とくに子音連結を単純化するために子音の間に母音を挿入してしまう現象です。この発表では日本語への借用語と第２外国語としてのフランス語の発音を比較してご紹介します。

Added syllable complexity, whereby a non-targeted consonant is added
next to a targeted consonant in the syllable, has received relatively little
attention in studies of children with speech sound disorders (SSD) and
typically developing children. Despite the scarcity and subtlety of the
pattern, evidence in child and adult data indicates universality. The
present article examines the pattern in a bilingual child’s longitudinal
speech in English from age 2;7 to 4;0, focusing on word-initial consonant
addition. The purpose is to identify phonological and psycholinguistic
processes associated with the pattern. It is suggested that the complexity pattern with both legal and illegal outputs is a systemic developmental behaviour linked to the child’s level of phonological acquisition, facilitating acquisition of the complex CCV rule as well as of non-acquired singleton consonants. Implications of the results for children’s SSD intervention techniques and for adult degenerative speech are discussed.

A number of runic inscriptions from the entire Germanic area from between A.D. 200 and 800 exhibit non-etymological, epenthetic vowels, such as worahto for *worhto ‘did’. An analysis of all (likely) instances of epen­thesis in early Ger­manic languages shows that epenthesis developed only in clusters involv­ing /r/, /l/ or /n/.

Epenthesis was an optional feature of nearly every early Germanic dialect, being most abundantly attested in southern Sweden. There is no statis­ti­cally sig­nif­icant evidence of an increase or decrease in the amount of epenthesis dur­ing the period. A detailed analysis reveals two different phonological en­vi­ron­ments for epenthesis. Scandinavian attestations of epenthesis oc­cur most­ly in heterorganic consonant clusters, irrespective of their sonority se­quence, where epenthesis is a result of a transition in articulatory gestures. The epenthetic vowels appear as a (or ᴀ) in Scandinavia. In inscriptions from south­ern Germany, however, epenthetic vowels are concentrated in clusters with a marked sonority sequence, irrespective of their place of artic­u­la­tion. While the epen­thetic vowels in the inscriptions from Germany are either a or u, the few po­tential instances of epenthesis in marked sonority se­quences in Scan­di­navia are rendered by vowels other than a. The epenthetic vowels in Anglo-Frisian in­scrip­tions resemble the Scandinavian type, but only partially.

This study has as main aims to discuss the motivation for the non-occurrence of the epenthetic vowel at the beginning of the word in the context of word junction in the first hundred Cantigas de Santa Maria (CSM – Afonso X, 1221-1284). Through this work and from a non-explored corpus (especially as for linguistic themes – SNOW, 1987, p. 478-480; MASSINI-CAGLIARI, 2005, p. 24-26), it is intended to contribute to reaching a deeper knowledge of Portuguese prosody history in the period of its recognition as an " independent " language in relation to Latin.

This paper focuses on languages that exhibit processes of copy epenthesis, specifically those where the similarity between a copy vowel and its host extends to prosodic or suprasegmental resemblance. We argue that copy vowels and their hosts strive for identity in all prosodic properties, and show that this drive for prosodic identity can cause misapplication in the assignment of properties such as stress, pitch, and length. To explain these effects, we argue that any successful analysis of copy epenthesis must involve a correspondence relation (following Kitto and de Lacy 1999). Our proposal successfully predicts the extant typology of prosodic identity effects in copy epenthesis; alternative analyses of copy epenthesis relying solely on featural spreading (e.g. Kawahara 2007) or gestural realignment (e.g. Hall 2003, 2006) do not naturally capture the effects discussed here.

The present paper is an experimental study that addresses the insertion of the palatal glide [j] and glottal consonants [h] and [?] to prevent vowel hiatus in Persian. Data included Persian words involving the sound sequence V1-V2 where V1 was the last sound of the root and V2 was the beginning vowel of a suffix (either derivational, inflectional or enclitic). In the previous literature it has been assumed that the insertion of a consonant in VI-V2 sequences in Persian is sensitive to the phonetic specification of the first vowel. Thus, words were selected such that V1 was any of the six vowels of the Persian sound system. Acoustic analyses of the data included measurement of variation of the first and second formant frequencies as wells as variation of energy in the FFT spectra (overall intensity and intensity in the higher frequency bands). Results suggested that while the insertion of [j] changes the frequency of the first two formants F1 and F2, the insertion of the glottal consonants fails to trigger frequency changes in the spectra. The results further suggested that the intensity of higher formant frequencies is more highly to reduce due to the insertion of mediating consonants than overall intensity or the intensity of lower formant frequencies. The results were interpreted to suggest that variation in phonation type is the most noticeable acoustic change that occurs when the glide [j] or the glottal consonants [h] and [?] is inserted in V-V. This change is the least phonetic change that is required to reform vowel hiatus in Persian by turning the ungrammatical sequence V-V to the grammatical VCV one. From the phonological point of view, the findings presented in this research are in line with Chomsky and Halle's (1968) model of phonetic features in which glides and glottal consonants are represented by the same features [-vocalic] and [-cons].  ‫‬

This paper addresses the different repair strategies applied in the Surani Kurdish language for the correction of vowel hiatus, cross-linguistically known as a phonologically marked structure. Data related to vowel hiatus were collected, classified and treated by Optimality Theory. Examination of the data revealed that Surani Kurdish uses three different phonological strategies to repair vowel hiatus which arises during word formation processes. These strategies are epenthesis, deletion and glide formation. The most common phonological pattern applied to prevent V+V sequences from being realized in surface form is glide ([w] and [j]) epenthesis. However, glide epenthesis is not applied in two conditions: (1) when the vowels are both [+low], in which case the second vowel (V2), i.e., the vowel belonging to the suffix, is deleted from the input form; (2) when the vowel in the suffix is /i/, in which case the vowel is turned into the glide [w] through the glide formation process. In an OT account of the phonological data concerned, we introduced a new set of Faithfullness constraints which are sensitive to the morphological information of the input form. Using these constraints, we are able to explain how each strategy turns to be the most optimal pattern in the morpho-phonemic condition in which it applies.

This paper investigates a class of misapplication effects arising in reduplication and copy-epenthesis, specifically those relating to the assignment of phonological properties related to prominence. We argue that extensions of Correspondence Theory (McCarthy & Prince 1995) can explain these effects, given a grammar that contains: (1) a correspondence relation between surface segments, arising when multiple output segments correspond to a single input segment (also Struijke 2000, McCarthy 2002); and (2) faithfulness constraints requiring identity among these correspondents for suprasegmental properties. We show that a grammar with these properties provides a unified account of a range of prosodic misapplication effects.

In this study, we examine the nature and content of some phonetic/ phonological designations to refer to central vowels. Bearing in mind that epenthetic vowels and unmarked vowels are not always central vowels, we will argue in favour of a thorough distinction between these designations and propose the use of: 1. 'Central vowels' to label vowels articulated along the central axis the of oral cavity, 2. 'epenthetic vowels' to refer to postlexically inserted vowels in order to regularize exceptional consonant strings, 3. 'unmarked vowels' to indicate the most frequent epenthetic vowels of a language. Additionally, the term 'schwa' is judged not strictly necessary to phonetic and phonological description, since it does not clearly separate the different meanings associated with more transparent designations such as 'central' or 'epenthetic vowels'.

We discussed the post-lexical character of the vowel epenthesis in Brazilian Portuguese (Collischonn, 1996) based on data from 144 speakers from six cities in southern Brazil from VARSUL database. We brought evidence that this phenomenon has post-lexical behaviour, but may also have lexical characteristics. We propose that epenthesis is in an intermediate stage, in other words, that the phenomenon occurs primarily in the post-lexicon and later on in a few words, is “raising” to the lexical component.

This paper aims at investigating some phonological aspects of syllable structure in Rumthawi Arabic, a Levantine variety spoken in the northern region of Jordan. It basically sheds light on the OT constraint interaction that determines the surfacing onsets and codas of syllables in this dialect. The scope of this paper is more specifically confined to examining the optimal candidates that surface when the definite article morpheme is prefixed. It thus proves that OT constraints in RA interact in an interestingly distinctive way that triggers divergence and sometimes convergence with other dialects due to the parametrical ranking of these constraints in this dialect unlike some other dialects. It is hoped that this humble endeavor will give insight to many interested researchers to deeply investigate various phonological aspects of this dialect.

Kaqchikel (ISO 639-3: cak), a Mayan language of the K’ichee’an branch, has a moderate number (22) of consonant phonemes (Patal Majzul, 2000). Its most notable contrasts are among its stops, which feature two laryngeal series at four places of articulation: a [constricted glottis] set {/ɓ̥  t’  k’  ʛ̥/} and a non-[constricted glottis] set {/p  t  k  q/} (1a, all examples are of Kaqchikel and from own field recordings unless noted differently). Interestingly, at the right edge of the word, members of the latter set become aspirated {[ph  th  kh  qh]}, with a lengthened period of noise following the stop’s release.
In addition to these positionally restricted stop allophones, Kaqchikel’s four non-nasal sonorants {/r  l  w  j/} surface as lengthened voiceless fricatives {[ʂ ɬ f ç]} in the same word-final positions as the aspirated stops. This paper documents these allophones and details the relationship between these phenomena as one of shared phonological [spread glottis] feature insertion (Vaux, 1998; Avery & Idsardi, 2001).
Remarkably, both of these sound changes act to the exclusion of other subsets of Kaqchikel’s consonants. The glottalized series of stops do not have position-sensitive allophones, occurring as voiceless implosives or ejectives in all positions, with short periods of (less noisy) aspiration in word-final position. Likewise, the nasal sonorants {/m n/} of Kaqchikel do not undergo spirantization or devoicing as non-nasal sonorants do in word-final position.
These processes of feature insertion are exceptional. On the one hand, they create non-contrastive allophones, yet are obligatory: all occurrences of word-final plain stops or non-nasal sonorants surface as their [spread glottis] allophones. Contrast this categorical pattern with the case of word-final stop glottalization/ejection in some varieties of English which creates non-contrastive allophones, but only optionally and phonetically (Keyser & Stevens, 2006). On the other hand, the process in Kaqchikel is more selective than in closely related languages: Tz’utujil (tzj) features partially devoiced nasal consonants in word-final position; other sonorant consonants are more fully devoiced. By contrast, Kaqchikel limits edge marking to phonological subclasses of segments. Segments outside of those subclasses surface without [spread glottis] inserted.
The presented findings address several questions: first, are the Kaqchikel sonorants simply devoicing or do they change manner of articulation to voiceless fricatives? Nearly the entire duration of the final segment is made up of aperiodic noise. The only portion of the segment that is periodic and sonorant-like is the transitionary period immediately after the offset of the preceding sonorant vowel.  This segment has become almost entirely a voiceless fricative.
Second, is the domain of insertion the right edge of the phrase, the word, or the syllable? Iverson & Salmons (2007) propose a logical progression of directionality in domain marking from the highest levels of the phrase generalized down to the word, and then the syllable. The data analyzed here show that edge-marking by [spread glottis] insertion occurs for most speakers at the word level, though some speakers have generalized to the later stage of marking every closed syllable, comparable to the devoicing patterns in Tz’utujil, with approximants and liquids devoicing syllable-finally, but nasals partially devoicing only word-finally (Wax Caravallo, 2020).
The evidence presented in this paper show that the insertion of [spread glottis] in Kaqchikel is a phonological, rather than phonetic, process. The locus of insertion is phonologically determined, its effects are phonologically specific, and its implementation is phonologically distinct from other cases of edge marking, even with [spread glottis] (see also Windsor, 2016). All this comes in spite of the fact that [spread glottis] is not a contrastive feature in the language.