Karolina Broś | University of Warsaw (original) (raw)

Books by Karolina Broś

Survival of the Fittest provides an in-depth analysis of weakening processes attested in Spanish ... more Survival of the Fittest provides an in-depth analysis of weakening processes attested in Spanish and English within the framework of Optimality Theory (OT). The book examines fricative lenition as an instance of sound change in progress, contributing to the study of phonological change and the notion of strength in phonology. It also provides motivation for the introduction of a derivational stage in OT analysis. A critical discussion of various OT sub-theories presented by the author leads to interesting conclusions concerning the way in which lenition and opacity processes should be addressed in OT.

Furthermore, under the assumption that language change should be conceived in evolutionary terms, the book concludes that sounds undergo continuous modification which is not at all accidental. The direction of change tends to be a constant on the temporal axis, and the leniting character of a large portion of phonological processes observed in the world’s languages points to the universal tendency for sounds to gradually fade and give way to other, stronger segments, which may be interpreted as an instantiation of ‘'natural selection’ within language.

In taking a broader perspective on language, the book considers phonological processes to be successors of phonetic innovations, and predecessors of morphological and lexical shifts. Thus, in order to encompass more than just a formal discussion of certain phonological phenomena, this book pursues the more profound question of why and how certain regularities within irregularities are attested across languages. The empirical data from Chilean Spanish and English serve as instantiations of these universal patterns.

Papers by Karolina Broś

Phonology, Feb 1, 2021

This study explores ongoing lenition of postvocalic /p t k b d g/ in the Spanish of Gran Canaria.... more This study explores ongoing lenition of postvocalic /p t k b d g/ in the Spanish of Gran Canaria. Duration, intensity and harmonics-to-noise ratio of 16,454 sounds produced by 44 native speakers were measured, with the latter phonetic parameter used for the first time to investigate lenition. The results show a path of gradual sound shortening and opening from voiceless stops to open approximants, as well as systematic use of six different variants depending on the underlying representation and phonological context: two types of [p t k], two types of [b d g] and two types of [ꞵ ð ɣ]. We interpret this as continuity lenition that leads to the flattening of the intensity contour and harmonicity of the target segment with respect to the flanking sounds. We argue that a phonological analysis of this process that accounts for its non-neutralising character requires the use of a scalar [aperture] feature.

Romanica Cracoviensia, Jun 30, 2022

The aim of the paper is to present a thorough description of the terms weakening and lenition in ... more The aim of the paper is to present a thorough description of the terms weakening and lenition in the context of language change, and to present major theories of lenition proposed in the framework of generative phonology. Among the most recent theories of lenition, we mention the proposal by Katz (2016) based on Kingston (2008) in which a distinction is made between loss and continuity lenition. We then present empirical data from the Canary Islands dialect of Spanish in which both types of lenition can be found, making the dialect a model example of weakening language change.

Journal of Neurolinguistics, Feb 1, 2021

It is unclear whether word stress in a language is stored as part of the word or whether it is ge... more It is unclear whether word stress in a language is stored as part of the word or whether it is generated by a rule. We test the generativist hypothesis of lexical storage stating that only unpredictable stress is stored in long-term memory against the contrasting usage-based approach assuming that all phonetic information regardless of its (un)predictability is stored in the mental lexicon together with the word. In a correctness judgment task involving correctly and incorrectly stressed penults and antepenults, we found that incorrectly stressed penults do not evoke an N400 effect, whereas incorrectly stressed antepenults do: there is increased negativity with a peak latency around 350-600 ms from word onset. Only changes to words with exceptional stress cause lexical inhibition, hence exceptional but not default stress markers are stored in the lexicon. Additionally, differences in processing patterns between the N400 and the late positivity component window point to an integration of two stages of word processing: pre-lexical stress recognition and stress-to-meaning matching. The results of the study support the view that stress should be understood as abstract phonological information.

Poznan Studies in Contemporary Linguistics, Mar 1, 2018

Spanish dialects show substantial variation in coda s weakening. Yet, to provide a comprehensive ... more Spanish dialects show substantial variation in coda s weakening. Yet, to provide a comprehensive treatment of this phenomenon, a bigger prosodic constituent than just the coda position should be analysed. Crucially, two aspirating varieties of Spanish are considered. The Granada dialect weakens s to [h] inside words, at word edges and at prefix edges. The process may be either transparent (esto [éh.to] 'this', des-calzar [deh.kal.sáɾ] 'to unshoe', las cosas [lah.kó.sah] 'the things') or opaque (des-hecho [de.hé.tʃo] 'undone', las aguas [la.há.ɣwah] 'the waters'). Chilean Spanish, on the other hand, presents transparent (esto [éh.to] 'this', des-calzar [deh.kal.sáɾ] 'to unshoe') and opaque (las aguas [la.há.ɣwa] 'the waters') aspiration, as well as deletion (las cosas [la.kó.sa] 'the things'), and no aspiration across a prefix boundary (des-hecho [de.sé.tʃo] 'undone'). The reported variable behaviour calls for an integrated approach to segmental weakening across all prosodic constituents, and for a revision of the present understanding of contiguity. The boundary between the prefix and the stem is protected by the grammar despite the weak coda position of the prefix-final s, therefore the domain of application of the CONTIGUITY constraint should be extended to the supramorphemic level.

Isogloss, Mar 7, 2019

This paper takes a series of lenition phenomena from Gran Canarian Spanish as a point of departur... more This paper takes a series of lenition phenomena from Gran Canarian Spanish as a point of departure to discuss the influence of phonology on the phonetics component. Based on phonetic and phonological data, it can be concluded that a blocking effect ensues between the process of coda deletion and post-vocalic voicing, giving rise to phonetic opacity. Against the assumption that the latter process is phonological in nature, acoustic data suggest that it is highly gradient, coarticulatory and variable, in which case it is inexplicable why it is blocked by phonological segment deletion. The proposed solution set forth in this paper is that the phonetic component has access to deep structure beyond featural specifications of sounds. What is more, evidence from prosody indicates that structural information concerning prosodic boundaries is also transposed into phonetics and influences production. Thus, the type and amount of information computed at the phonetics-phonology interface needs to be revised and supplemented by turbid structures in order to account for surface variability and both inter-and intra-speaker differences.

Onomázein Revista de lingüística filología y traducción, 2013

This article examines the so-called s-aspiration as well as s-deletion in Chilean Spanish. It ana... more This article examines the so-called s-aspiration as well as s-deletion in Chilean Spanish. It analyses these phonological processes as segment weakening in interaction with resyllabification across word boundaries, and then moves on to the presentation of Chilean opacity in various theoretical frameworks prevailing throughout the 1990s and the beginning of the 21st century. This article demonstrates that a combination of concepts introduced by lexical phonologists and markedness governing the optimality framework is the correct path to follow in the analysis of Chilean opacity effects, with derivation playing a crucial role in their reanalysis within the framework of Derivational Optimality Theory.

Linguistics vanguard, Apr 6, 2023

The COVID-19 pandemic has shown that alternative methods of data collection are necessary to cont... more The COVID-19 pandemic has shown that alternative methods of data collection are necessary to continue working in certain fields of linguistics. This is a challenge for (socio)phoneticians and phonologists who have to rely on good quality sound but cannot do fieldwork or gather recordings in a traditional manner. In this paper, I show that audio recordings made via social media can help alleviate this problem. To this end, I compared samples from five speakers of dialectal Spanish recorded in a laboratory setting and via a social media application (WhatsApp). The analysis of temporal and spectral characteristics of consonants in postvocalic position shows that recordings made via social media can be successfully used for at least some types of sociophonetic analysis. They also provide some additional advantages for researchers: ease of data collection, potentially large speech corpora, and access to authentic, naturalistic speech which is uninhibited by laboratory conditions or the presence of a researcher and a professional recording device.

Linguistica Copernicana, Mar 18, 2018

Phonetica, May 1, 2019

Background/Aims:This paper examines the process of postvocalic voicing in the Spanish of Gran Can... more Background/Aims:This paper examines the process of postvocalic voicing in the Spanish of Gran Canaria from the point of view of language change. A perception-production study was designed to measure the extent of variation in speaker productions, explore the degree to which production is affected by perception and identify variables that can be considered markers of sound change in progress.Methods:20 native speakers of the dialect were asked to repeat auditory input data containing voiceless non-continuants with and without voicing.Results:Input voicing has no effect on output pronunciations, but voicing is highly variable, with both phonetic and social factors involved. Most importantly, a clear lenition pattern was identified based on such indicators as consonant duration, intensity ratio, absence of burst and presence of formants, with the velar /k/ as the most affected segment. Furthermore, strong social implications were identified: voicing degrees and rates depend both on the level of education and on the gender of the speaker.Conclusion:The results of the study suggest that the interplay of external and internal factors must be investigated more thoroughly to better address the question of phonetic variation and phonologisation of contrasts in the context of language change.

Lingwistyka Stosowana / Applied Linguistics / Angewandte Linguistik, 2014

Phonica, Dec 23, 2022

This paper discusses the corpus of Gran Canarian Spanish gathered in 2016 in order to provide an ... more This paper discusses the corpus of Gran Canarian Spanish gathered in 2016 in order to provide an in-depth sociolinguistic account of the lenition processes identified in the dialect. After a detailed description of the methodology and database preparation, two case studies showcasing the utility of such corpora are presented. First, we show the phonetic and social factors governing the distribution of different surface variants of the underlying coda /s/, pointing to generalised variation and hence incompleteness of any of the weakening options. Second, we provide a comparison of the spontaneous speech produced by the 6 informants of the corpus with their productions from a laboratory study, which leads to the conclusion that variation is subject to yet another important factor: social setting and that the options chosen on each occasion are reflections of competing stages on the same lenition trajectory that is systematically applied by language users. All in all, the paper shows the advantages of using fieldwork data vis à vis lab speech elicitations, both as an independent sociophonetic database and as a starting point for comparative studies on sound weakening.

Glossa: a journal of general linguistics

In this paper, we present novel data from Spanish spoken on Gran Canaria which show an interactio... more In this paper, we present novel data from Spanish spoken on Gran Canaria which show an interaction of two lenition processes: final consonant deletion and vowel apocope. We show that in certain positions in an utterance, these processes optionally combine in a fed counterfeeding interaction. Furthermore, the variation present in the dialect due to process optionality uncovers a latent opacity pattern, i.e. an additional opaque interaction that is not motivated by any individual input-output mapping, but only by the quantitative aspect of variation. This takes the form of mutual counterfeeding, a rarely reported phenomenon, thus creating a novel test case for theories of opacity. The second part of the paper provides a formal analysis of the opacity-ridden data, taking into account process optionality and variation. The theoretical analysis and learning simulations using the Expectation-Driven Learner demonstrate that a probabilistic variant of Serial Markedness Reduction can capture...

Linguistics Vanguard

Unlike some of its related languages (e.g. Catalan or Portuguese) Spanish seems to have a stable,... more Unlike some of its related languages (e.g. Catalan or Portuguese) Spanish seems to have a stable, not particularly crowded, vowel inventory and despite significant consonant weakening, it rarely exhibits vowel reduction. It is thus worth examining whether there is a correlation between stress and reduction processes. The primary assumption here is that a language's stress pattern and the nature of its vowel inventory are strictly connected with the freedom of reduction. A disruption of the stress pattern in a syllable-timed language, such as Spanish, may inhibit comprehension and speech perceptibility, vowels being the principal stress and melody carriers. Two perception tests have been conducted among Spanish speakers to examine the perceptibility of vowel contrasts and speakers' sensitivity to stress shift and unstressed vowel quality/duration changes. One of the principal aims of the experiments was to see whether stress shifts affect comprehension and word retrieval from...

Phonology, Feb 1, 2021

Romanica Cracoviensia, Jun 30, 2022

Journal of Neurolinguistics, Feb 1, 2021

Poznan Studies in Contemporary Linguistics, Mar 1, 2018

Isogloss, Mar 7, 2019

Onomázein Revista de lingüística filología y traducción, 2013

Linguistics vanguard, Apr 6, 2023

Linguistica Copernicana, Mar 18, 2018

Phonetica, May 1, 2019

Lingwistyka Stosowana / Applied Linguistics / Angewandte Linguistik, 2014

Phonica, Dec 23, 2022

Glossa: a journal of general linguistics

Linguistics Vanguard

Romanica Cracoviensia

Gran Canarian Spanish: A Reflection of Weakening Language Change in the Spanish-Speaking World Th... more Gran Canarian Spanish: A Reflection of Weakening Language Change in the Spanish-Speaking World The aim of the paper is to present a thorough description of the terms weakening and lenition in the context of language change, and to present major theories of lenition proposed in the framework of generative phonology. Among the most recent theories of lenition, we mention the proposal by Katz (2016) based on Kingston (2008) in which a distinction is made between loss and continuity lenition. We then present empirical data from the Canary Islands dialect of Spanish in which both types of lenition can be found, making the dialect a model example of weakening language change.

Introduction: post-vocalic voicing in Gran Canaria Broś (2016) : recordings of several speakers f... more Introduction: post-vocalic voicing in Gran Canaria Broś (2016) : recordings of several speakers from 2014/15 Oftedal (1985) : previous reports, phonetic data, systematic Post-vocalic voicing: word level /p/ de[b]artamento 'apartment' /t/ fone[d]ica 'phonetics' /k/ má[g]ina 'machine'

Kiparsky (2000, 2013) and Bermúdez-Otero (2003, 2011) propose that opacity can be dealt with in O... more Kiparsky (2000, 2013) and Bermúdez-Otero (2003, 2011) propose that opacity can be dealt with in
Optimality Theory by distinguishing between three levels of derivation. This paper contributes to the
discussion on the feasibility of such a multilevel approach by showing that, contrary to the expectations, two
independent phrase-level processes can interact opaquely. I present previously unreported data from Gran
Canarian Spanish, collected through my 2014/2015 fieldwork, which illustrate two cases of morphologically
unbound underapplication: a synchronic chain shift and a case of counterfeeding deletion. As such, they pose
a challenge to OT accounts that do not assume opaque mappings stratum-internally. I argue that the very
existence of processes involving non-transparent level-internal mappings calls for a revision, and possibly
redesign, of stratal frameworks that do not assume explicit process precedence relations, and for their
refocussing on within-stratum interactions.
Theoretical background. Synchronic chain shifts are a type of opaque mappings that challenge OT's
property of strict domination (Prince & Smolensky 1993). Strategies employed within parallel OT that seem
to be able to deal with some of the resultant faithfulness cumulativity effects (Farris-Trimble 2008) include
constraint conjunction (Kirchner 1996), ternary scales (Gnanadesikan 1997) and contrast preservation
(Łubowicz 2003). Meanwhile, Stratal OT (Bermúdez-Otero 2003, Kiparsky 2000) is usually believed to
specialise in dealing with opaque counterfeeding interactions. Yet, as noted by McCarthy (2007), one of the
pillars of Lexical Phonology (which served as a basis for level definition) was not transposed to Stratal OT to
the detriment of the framework: extrinsic rule ordering inside each stratum. Thus, all the processes observed
within a given level must be transparent. A problem arises, however, when this transparency cannot be
granted. Contrary to Kiparsky's (2013) claim that there are maximally two degrees of opacity (precisely at
stratum junctures), the data below show that stratum-internal opacity is attested. As will be demonstrated,
Stratal OT, though particularly apt at dealing with morphosyntactically-conditioned opacity, fails to resolve
counterfeeding cases that do not involve passing from one stratum to another. The cyclicity model assumed
by Stratal OT is therefore either inadequate or too restrictive (see also Ettlinger 2007; Newell 2015).
Postvocalic voicing. In Gáldar, in the north-west of Gran Canaria, Spain, an extended process of
noncontinuant voicing (1; cf. intervocalic voicing reported by Oftedal 1985) can be observed alongside
generalised Spanish spirantisation, (2), that turns voiced stops into approximants of differing strength
(aperture). The contexts of occurrence of the two processes partially overlap, creating a chain effect.

(1) Postvocalic voicing of /p t k/

/p/ de[b]artamento ‘apartment’ /p/ yo [b]ienso que ‘I think that’
/t/ foné[d]ica ‘phonetics’ /t/ juntos y [d]al ‘together and so on’
/k/ má[g]ina ‘machine’ /k/ de [g]olombia ‘of Colombia’

(2) Post-continuant spirantisation of /b d g/

/b/ lle[β]o ‘I take’ /b/ la [β]oca ‘the mouth’
/d/ po[ð]er ‘be able to’ /d/ el [ð]oble ‘two times’
/g/ ma[ɣ]o ‘magician’ /g/ mi [ɣ]rupo ‘my group’

Both voicing and spirantisation apply inside words and across word boundaries. There is one distributional
difference, however. Whereas spirantisation applies everywhere except after a pause or a nasal/homorganic
lateral, i.e. after all continuant sounds, voicing is strictly post-vocalic and is blocked after any consonant.
While el triple 'three times' surfaces with no voicing, el doble 'two times' undergoes spirantisation. The end
result for the grammar is phonemic overlap: voiced stops can be allophones of both voiceless and voiced
underlying stops (I assume underlying stops but see e.g. Baković 1995).
A phonological account of the two phenomena must incorporate a way of blocking an A → C mapping
in the observed p t k → b d g → β ð ɣ chain shift. This cannot be achieved with a stratal account as: (i) both
processes are phrase level – they apply across word boundaries, (ii) there is no stratum juncture and no
constraint reranking is possible within a stratum. Since both voicing and spirantisation are allowed in the
language, markedness constraints triggering them (*V [-cont, -voice], *[+cont] [-cont, -nasal]) must be
ranked higher than the corresponding faithfulness constraints (IDENT(voice), IDENT(cont)) and a ranking
paradox will ensue whenever the contexts for the application of both processes are met. One solution that is
particularly viable in addressing the problem is constraint conjunction. (3) shows a possible analysis, in
which the high-ranked conjoined constraint prevents spirantisation of underlying voiceless stops by banning
multiple faithfulness violations (in grey). Nevertheless, while incorporating local conjunction works, the
solution is not internal to Stratal OT. Adding an additional constraint type might make the framework too
powerful in terms of cross-linguistic predictions and typology (cf. Jesney 2005 on learnability). A similar
conclusion can be drawn if other external mechanisms, such as contrast preserving constraints, are adopted.

Another way of dealing with the chain shift in Stratal OT would be to enforce faithfulness to the UR rather
than to the output of the previous stratum, a solution that apparently deals satisfactorily with some data
(Hauser, Hughto & Somerday 2014, McCarthy 2007). Nevertheless, given the fact that the Canarian data
involve two changes within one stratum and changes do not apply one at a time in this framework, there is no
intermediate stage in the derivation in which the UR and the current input would differ in terms of voicing or
continuancy. The only solution is to stipulate a constraint requiring that no change in continuancy may occur
in underlying voiceless stops (e.g. Ident UR-O(cont)/[-voice]). This ensures that /p t k/ do not go all the way
to voiced continuants, solving the chain shift problem. Note, however, that Spanish does have continuants
arising from stops in the coda (actuar 'to act' [ax.tu.ar] / [aɣ.tu.ar]) so a high-ranked constraint of the above
type may not be the correct path to follow for the whole of the dialect.
Deletion as a blocker. Regardless of the solution adopted in the case of the chain shift, another
counterfeeding interaction obscures the analysis of the Canarian data under a stratal approach. Interestingly,
Canarian Spanish shows a blocking effect of segment deletion. The widely extended process of coda deletion
typically creates a context for voicing in phrase level phonology, yet no voicing applies.

(3) Deleted segment blocks noncontinuant voicing

la(s) caracterí(s)tica(s) ‘the features’
por pensa(r) tontería(s) ‘for thinking about silly things’
se puede acepta(r) que ‘it can be accepted that’

In (3), segments in parentheses are deleted. The following stops do not change their unvoiced status despite
the fact that they have a vowel to the left (due to deletion). This underapplication again involves no stratum
juncture (both deletion and voicing are sensitive to resyllabification, a phrase-level process). Even if we
assumed that deletion is a word-level process, this would only make matters worse as for a phonological
account to work deletion should apply after voicing. What is more, no local conjunction can be invoked in
this case as the relation between the interacting processes is non-local and such an extension of the
conjunction framework would make it too powerful (cf. Moreton & Smolensky 2002). Furthermore, I can
imagine no harmonic mapping that would make a serial OT analysis of this data feasible, i.e. predict waiting
with deletion until the end of phonology. I suggest that the only possibility is to allow for covert structure in
the grammar. Apparently, the deleted segment is not interpreted phonetically, but still present in the
phonology, which is correctly predicted by containment (Prince & Smolensky 1993). This conclusion is well
matched by articulatory phonology predictions concerning gestural masking whereby cross-tier gestural
overlap may lead to apparent deletion, i.e. the articulatory gestures may be present, but the resultant sound
may fail to be perceived (Browman & Goldstein 1990).

References:
Baković, E. 1995. Strong Onsets and Spanish Fortition.  Bermúdez-Otero, R. 2003. The acquisition of
phonological opacity.  Bermúdez-Otero, R. 2011. Cyclicity.  Browman, C. & L. Goldstein. 1990.
Gestural specification using dynamically-defined articulatory structures.  Ettlinger, M. Variation as a
window into opacity.  Farris-Trimble, A. 2008. Cumulative faithfulness effects in phonology. 
Gnanadesikan, A.E. 1997. Phonology with ternary scales.  Hauser, I, C. Hughto & M. Somerday. 2014.
Faith-UO: Analyzing opacity in Harmonic Serialism.  Jesney, K. 2005. Chain Shift in Phonological
Acquisition.  Kiparsky, P. 2000. Opacity and Cyclicity.  Kiparsky, P. 2013. The residue of opacity. 
Kirchner, R. 1996. Synchronic chain shifts in Optimality Theory.  Łubowicz, A. 2003. Contrast
Preservation in Phonological Mappings.  McCarthy, J. 2007. Derivations and levels of representation 
Moreton, E. & P. Smolensky. 2002. Typological consequences of local constraint conjunction.  Newell, H.
2015. Implications of “multi-stratal” affixes.  Oftedal, M. 1985. Lenition in Celtic and in Insular Spanish.

Spanish is notorious for its vast array of leniting sound changes – its consonants are not partic... more Spanish is notorious for its vast array of leniting sound changes – its consonants are not particularly stable and undergo a series of processes analysed jointly under the umbrella term 'weakening'. By the well-known process of aspiration, s is debuccalised and even elided in syllablefinal position. Spirantisation forces b d g to become weaker spirants or approximants [β ð ɣ] with a variable degree of aperture (Harris 1969). Syllable-final fricatives tend to undergo voicing, while word-final consonants are devoiced, spirantised or lost. All of these changes have led to an uneven distribution of sounds: most Spanish dialects lack a contrast in fricatives, voiceless f s x being the only phonemic units. Voiced variants of these emerge only as context-dependent allophones. At the same time, spirantisation weakens the contrast between voiced and voiceless stops, which is maintained only phrase-initially (dos 'two' vs. tos 'cough') and word-medially after a homorganic
sonorant (manda 'commands' vs. manta 'blanket'). All other instances of underlying b d g turn into [β ð ɣ]. Interestingly, the resultant distributional gap is 'filled' in at least one Spanish dialect.
In this paper, I present novel data from a dialect spoken in Galdar on Gran Canaria, which show postvocalic voicing of p t k. Most importantly, the data cannot be analysed as intervocalic or intersonorant voicing due to the asymmetry between the left-hand and the right-hand environments. It appears that a consonantal sonorant on the left (including glides) does not trigger voicing while the same context on the right does not inhibit the process as long as there is a vowel to the left.
a.
a[b]asionado 'enthusiastic'
fone[d]ica 'phonetics'
la fre[g]uencia 'the frequency'
tengo una [b]rima 'I have a cousin'
un [p]ueblecito 'a small village'
juntos y [d]al 'together and so on'
otra [g]lase de 'other type of'

b.
im[p]ortante 'important'
en[t]onces 'so / then'
en un ban[k]o 'in a bank'
el [t]riple 'three times'
super [k]ómodo 'very convenient'

The process applies both inside words and across word boundaries, in the same token as spirantisation (except that the latter extends to post-[r l] contexts). Voicing is blocked after vowels which become adjacent to the stop as a result of elision (both r and s can be deleted word-finally).
Thus: e(s)tas son la(s) caracteri(s)tica(s) 'these are the features' does not present voicing of the stop (in bold) after deleting coda s. Neither does the phrase die(z) primo(s) 'ten cousins' or por pensa(r) tontería(s) 'for thinking about silly things' after eliding s and r, respectively. Moreover, voicing is blocked if a voiceless segment stands to the right (cara[k]terísticas 'features'). Coda obstruents undergo other types of weakening in this position, e.g. spirantisation, lack of plosion or elision.
Interestingly, the process described here very much resembles historical changes. In French, lenition started with the spirantisation of voiced stops, followed by the voicing of obstruents, and the resultant sounds were then spirantised and lost completely (except [β] > [v]). The context for voicing was exactly the same as in modern Canarian: aprilem>avril 'April' fratre>frère 'brother' but rumpere>rompre 'to break' (Bichakjian 1972). In Spanish, voiced stops were spirantised and then lost (credo > creo 'I think'), while voiceless stops were voiced and then spirantised (lupum>modern Spanish lo[β]o, Lloyd 1987), and tend to be elided (habla(d)o 'spoken'). It seems, therefore, that we have come full circle with the lenition process, at least in some part of the Spanish-speaking world.
The treatment of lenition in generative phonology ranges from autosegmental feature spreading and underspecification, through positional markedness in OT, to articulatory-, effort- and perception-based analyses (e.g. Harris 1969; Mascaró 1987; Lubowicz 2002; Alber 2014; Gurevich 2014; Kirchner 2001; Piñeros 2002; Kaplan 2010). The main problem is how to incorporate various types of phonetic and functional grounding into a formal phonological framework. Looking at the whole of the apparent chain shifts observable as lenition patterns from the contrast-preserving perspective, I will discuss the systemic consequences of sound change in Canarian Spanish: phonemic overlap (Bloch 1941) and the narrowing of perceptual distinctness between sounds. My OT analysis of these facts presents articulatory, perceptual and functional factors consequentially, without incorporating them into the tableaux in the form of (non-categorical) constraints.

En la mayoría de los dialectos del español moderno abundan procesos de debilitamiento consonántic... more En la mayoría de los dialectos del español moderno abundan procesos de debilitamiento consonántico tales como la aspiración y hasta la pérdida de la [s] implosiva, el cambio de las oclusivas sonoras en aproximantes débiles en posición intervocálica, la confusión de la lateral con la vibrante, la neutralización de las nasales en posición final/implosiva etc. Mientras tanto, las vocales españolas parecen mantenerse estables. No presentan reducciones a diferencia de algunas lenguas vecinas (portugués, catalán, francés).

En los escasos casos de reducción vocálica (sobre todo en México y en algunas zonas andinas), lo único que observamos es ensordecimiento de las vocales acompañado de la falta de debilitamiento consonántico, lo que indica una correlación interesante entre los dos tipos de debilitamiento fonético.

El objetivo del presente artículo es investigar esta correlación planteando, como punto de partida, una hipótesis sobre la percepción del habla que supone una mayor disminución de la comprensión de la señal vocal tras alterar el patrón melódico del idioma. Se supone aquí que una modificación de la señal mediante el desplazamiento de la vocal tónica o reducción de una vocal átona (acortamiento, centralización, elevación/cierre) puede dificultar seriamente e incluso imposibilitar la comprensión de una determinada unidad léxica. Dos experimentos de percepción han sido diseñados para averiguar esta hipótesis. Dichos cambios fueron introducidos en palabras españolas con y sin contexto. También se añadió una parte en la cual aparecen palabras inventadas (hápax), formadas de acuerdo con las reglas de silabeo del español para evitar sesgo como consecuencia de la recuperación de palabras del léxico
(inventario) nativo.

Los resultados del estudio preparado en forma de una encuesta audio online indican una interrupción en el proceso de recuperación y comprensión de palabras españolas tanto en el caso de cambio de acento, como en el caso de reducción de un tipo determinado: centralización (schwa). Es más – en la mayoría de los casos la vocal centralizada no se oye (62%) y cuando sí se entiende, va interpretada como la vocal anterior media [e] en el 70% de los casos, lo que (junto con un par de otros factores) indica que [e] puede tener estatus de vocal por defecto en español (lo que se ve confirmado por procesos de epéntesis y otros fenómenos fonológicos).

Spanish dialects show substantial variation in the treatment of coda s, most of them showing some... more Spanish dialects show substantial variation in the treatment of coda s, most of them showing some degree of segment weakening: debuccalisation to h inside words and at word edges, analysed as spontaneous or effort-driven lenition. This gives rise to systemic changes whereby an otherwise
non-existent sound (h) is added to the inventory and another sound (s) slowly vanishes.
The general process of s weakening encompasses a series of strategies employed to deal with a marked structure: s standing in a prosodically weak position. It extends from preconsonantal contexts inside words and at word boundaries to prepausal environments and prevocalic position in the most advanced Spanish varieties where it is obscured by resyllabification (Lipski 1996). Given the fact that onset s seems immune to such weakening, it may be confirmed that the coda is a weak prosodic position. Nevertheless, a bigger prosodic constituent should be considered in terms of relative strength in some contexts. Although internal coda s typically weakens to h, prefix-final s
not necessarily follows suit, not to mention prefix-final codas resyllabified into onsets (esto [eh.to] 'this' vs. despertar [des.peɾ.taɾ] 'to wake', deshecho [de.se.tʃo] 'undone').
In the most radical dialects a further mismatch can be observed: in Chilean coda s is dispreferred altogether and deleted unless it would disrupt morpheme contiguity (casas verdes [ka.sa.βeɾ.ðe] 'green houses' but conozco [ko.noh.ko] 'I know' – 'weaker' lenition). Interestingly, prefix-final s resists deletion although such a process would have no effect on the integrity of the stem (e.g. descalzar [deh.kal.saɾ] 'to unshoe'). It seems that the morphophonological relationship is of crucial importance: morpheme contiguity needs to be extended to encompass the whole prefixed word. Although the affix is traditionally considered a weak position (Casali 1996), the whole prosodic word containing the prefix should be considered a strong position as it inhibits radical changes to the underlying segment in the same way they are banned morpheme-internally. The boundary between the prefix and the stem is somehow protected by the grammar despite the weak coda position.
The aim of this paper is to look for an integrated theory of positional weakening based on evidence from Spanish and confront the presented data with evidence from other dialects where strong positions are affected (e.g. onsets, by extension in accordance with the laws of language change), as well as discuss the notion of contiguity and crucially extend it to the supramorphemic level.

Unlike some of its related languages (e.g. Catalan or Portuguese, cf. Mateus & Andrade 2000, Masc... more Unlike some of its related languages (e.g. Catalan or Portuguese, cf. Mateus & Andrade 2000, Mascaró 1978), Spanish seems to have a stable, not particularly crowded, vowel inventory. While a vast majority of Spanish dialects present significant consonant weakening, vowel reduction is particularly rare. The only instances of vowel weakening reported by linguists involve devoicing in certain parts of Mexico and the Andes, interestingly with no accompanying consonant lenition (Lope Blanch 1972, Sessarego 2012, Delforge 2008).
Given this asymmetry, it is worth examining whether there is a correlation between stress and reduction processes.
The primary assumption contemplated here is that a language's stress pattern and the nature of its vowel inventory are strictly connected with the freedom of reduction. This, in turn, is related to the well-known distinction between stress-timed and syllable-timed
languages. In the latter case, it is assumed, a disruption of the stress pattern might inhibit comprehension and speech perceptibility, vowels being the principal stress and melody
carriers. Limited or inexistent vowel reduction levels are less costly in the process of communication, hence vowel weakening remains largely unattested in such languages.
To account for the limited perceptibility of (non-native) vowel contrasts and Spanish speakers' sensitivity to stress shift and unstressed vowel quality and duration changes, a series of perception tests have been conducted on Spanish speakers. The interpretation of the reduced vowel signal by Spanish native speakers was of special interest here. Particularly, it was speculated to what degree the changes in quality and duration of the unstressed vowel would affect its perceptibility and how the reduced vowel would be interpreted with respect to the native inventory. Given the fact that Spanish lacks centralised vowels, its inventory
being limited to corner + mid vowels /i, e, a, o, u/, it is interesting to investigate whether centralised vowels are perceived by native speakers and if so, how they are identified with respect to native vocalic segments. Another important question is whether Spanish words modified in terms of stress and vowel reduction are identifiable i.e. retrievable from the lexicon.
The preliminary study of these questions involved the use of both Spanish native words with vocalic modifications and nonce words imitating Spanish syllabification and stress pattern. The results of both parts of the experiment suggest a possible emergence-ofthe-unmarked effect. Although the perception of schwa follows the patterns reported by
researchers studying ESL acquisition (e.g. Gómez Lacabex & García Lecumberri 2005, Diettes 2010) in that many instances of the centralised vowel are simply inaudible for the average native speaker (with a mean 62% success rate in a group of 32 individuals), an intriguing tendency toward identifying schwa as the mid front vowel /e/ was revealed (70% of the cases). Several variables seem to indicate the existence of a default vowel across all contexts. The gathered data suggest that schwa is not simply perceived as the mid front vowel per se, given certain inconsistencies in pretonic syllables as opposed to word-final position.
What is more, apart from the highly predictable context of pre-/s/ final position (which has been reported as the default word-final, and especially plural value by numerous researchers investigating Spanish varieties), /e/ was identified in some unpredictable environments, which cannot be justified by retrieval from the lexicon or other native speaker bias. This is confirmed by the results from the nonce word test that outright excludes lexical identification.
Thus, while perception tests confirm that changes in stress and vowel quality inhibit comprehension and word identification in Spanish speakers, they also suggest that unknown phonetic categories are interpreted as default segments. The status of /e/ as a default vowel in Spanish is further confirmed by morphology, as well as a series of historical and synchronic phonetic and phonological phenomena, especially vowel epenthesis. The latter takes the form of prothesis (Harris 1969) to repair marked cluster structures in words such as estadio 'stadium' or escándalo 'scandal' (SSG violations), as well as esmaltar 'to enamel' (minimal sonority distance) or eslavo 'slave'. Most of these changes are historical, but new words and loanwords undergo the same process, which is not always reflected in spelling (e.g. snob, esnob 'snob', status 'status', Alfaro 1964).
The same applies to second language acquisition: Spanish speakers consistently insert /e/ before sC clusters (although exceptions of dialectal nature can be found in Latin America, but see Carlisle 1998). The mid front vowel is also the epenthetic plural marker in words ending in consonant (e.g. Colina 2006a). Certain dialectal processes also point to its default status, Dominican 'double plural' being an especially prominent example: mujeres 'women' is realised as [muherese], palomas 'pigeons' as [palomase] in this dialect (Jimenez Sabater 1975, Nuñez-Cedeño 1980, Colina 2006b).
The aim of this paper is to present the results of the preliminary perception experiment together with the results of a follow-up experiment focused specifically on the
perception of schwa and word identification in correlation with centralised reduction (changes in both duration and quality). The experiment has been designed with the possible counterindications in mind, namely consonantal contexts, syllable position (initial, pretonic, post-tonic), morphological and lexical predictability, word frequency effects as well as auditory and acoustic similarity between schwa and the mid front vowel. The test will also be
controlled for possible multilingual or L2 effects (e.g. command of Catalan and similar languages). As with the first experiment, the stimuli will be collected in a silent room setting and manipulated in PRAAT (Boersma & Weenink 2010). The results will show whether the hypothesised existence of a default vowel in reduction contexts is confirmed by native speaker perception.

Given the abundance of vowel reduction phenomena spreading across the different language families... more Given the abundance of vowel reduction phenomena spreading across the different language families, phonetic
reduction is usually analysed with respect to vocalic segments. It is, however, also the domain of consonants. Certain languages, such as Spanish, tend to preserve its vowels (both in quantitative and qualitative terms) regardless of the speech ratio and reduce the articulatory features of the consonants. Most Spanish dialects are very much advanced in weakening phenomena, such as debuccalization and assimilation of preconsonantal codas, spirantization of intervocalic stops (with inter- and intradialectal variation between the resultant spirants and weak approximants) and word-final or intervocalic deletions, pointing very clearly to the universal lenition pattern (Lipski 1996, Colina 2002). These processes can be gradual and very often depend on social factors, producing variation in the speech patterns of different, or even the same speakers, which makes it difficult to analyse them on purely phonological grounds, or establish the phonemic status of such changes.
Closely related languages (e.g. Portuguese or Catalan) seem to present different patterns, with significant reductions in
vowels (Mateus&Andrade 2002, Mascaró 1978). Why is it that Spanish seems to resist one type of reduction while
engaging in another type with such mastery? This paper aims at addressing this issue, suggesting that a language's
intonational and stress patterns are strictly connected with the freedom of reduction. A disruption of these patterns might inhibit comprehension and speech perceptibility, vowels being the principal stress and melody carriers (syllable timing) – a situation which does not occur in languages with rhythmic and phrasal stress patterns (such as French or stresstimed English). Limited or inexistent vowel reduction levels are less costly in the process of communication. Nevertheless, the supposed inexistence of vowel reduction in Spanish may be countered by certain dialectal properties, such as Mexican or Andean vowel reduction (entonces 'then' surfacing as [entons's]).
Interestingly, these varieties have a tendency to preserve their consonants rather than weaken them, which may point to an interesting correlation between the two types of reduction processes. Furthermore, even languages traditionally perceived as non-reducing (e.g. Polish)
have been reported to show schwa-like sounds in certain positions in rapid speech (e.g. Nowak 2006 provides a
thorough analysis of Polish vowel duration reductions).
The author of the paper will present the results of a simulation-based speech perception analysis focused on the Spanish stress pattern and its implications for the (non-)reduction of vowels in the vast majority of its dialects.

An abundant set of cross-linguistic data confirm the observation that a distinction should be mad... more An abundant set of cross-linguistic data confirm the observation that a distinction should be made between word and phrase phonology on the one hand, and between the different advancement stages/domains in which phonological processes apply. In a holistic perspective phonology points to an ascending trend from the lowest (broadest) to the highest (narrowest) domains – as rightly captured by the concept of the life cycle of phonological processes (Bermúdez-Otero 2007). Thus, processes
involving opacity can be explained in terms of domain narrowing, e.g. the differences between Spanish dialects, from phrase level s debuccalisation only, through word-level preconsonantal and opaque prevocalic, to the deletion / debuccalisation conspiracy with complete s loss phrase-finally and debuccalisation word-finally in Chilean (Lipski 1996, Broś 2012). Similar cyclic patterns can be observed in the distribution of English linking r and dark l (Bermúdez-Otero 2011, Turton 2012), as well as Catalan and Quito Spanish (e.g. Colina 2009, Strycharczuk 2012).
In Slavic languages, Polish distinguishes between two dialectal behaviours, one of which involves presonorant assimilation across word boundaries. The Poznań/Kraków dialect presents not only voice assimilation (VA) and final devoicing (FD), but also admits voicing before all sonorants, except word-medially. This is complicated by the specific restrictions on syllabification, with onset maximisation and SSG suspension that result in syllabifying CC and CCN clusters in the onset (ja.sny 'bright', gwie.zdny 'stellar', ża.bki 'frogs', pro.sty 'straight'). This makes traditional accounts of VA implausible in terms of determining directionality (saving onset voice at the expense of the coda), as noted by Rubach (2008), cf. Lombardi (1999). A string-based faithfulness account solves the question of agreement in clusters, but fails to account for the unexpected behaviour of word-final obstruents in Kraków. Traditional pre-OT accounts (Gussman 1992, Rubach 1996) rely on autosegmental delinking cum spreading which requires that word-final obstruents be distinguished from word-medial by the prior
application of FD (underspecification). This is incompatible with the results of the latest studies in voicing oppositions and the life cycle. As noted by Strycharczuk (2012), Kraków/Poznań voicing data suggest that FD is a phrase-final process: full neutralisation in voicing can only be observed prepausally. In all other cases final obstruents share the voicing specification with the following sound. Thus we have bra[t] 'brother', bra[da]dama 'Adam's brother', bra[dm]agdy 'Magda's brother', bra[tk]asi 'Kasia's brother' and bra[dg]osi 'Gosia's brother'. Given the lack of resyllabification in Polish, the last segment in brat invariably stands in the coda, unlike word-medial presonorant obstruents: ja.sny 'bright', za.zna 'will experience'.
I will argue that Kraków Polish has no FD in the traditional sense. Laryngeal contrast can be observed word-medially (purportedly due to syllabification restrictions), albeit with obligatory cluster homogeneity: pstry 'colourful', bzdura 'nonsense', gwiazda 'star', miasto 'town'. General markedness of laryngeal features in obstruents is the driver of both neutralisation across a word boundary (with full
laryngeal agreement before obstruents and sonorants alike) and pre-pausal devoicing interpreted as delaryngealisation. This is presented in a Stratal OT framework where *LAR and AGREE constraints conspire at the phrase level, avoiding unmotivated Duke-of-York effects and stipulative constraint
formulations. The resultant underspecification would then be interpreted as the default value (voicelessness) by the phonetics component of the grammar, giving rise to 'emergence of the unmarked' in the post-phonological component. This perfectly captures the phonetics-phonology interface and phonetic feeding into the phonological component (rule stabilisation and domain narrowing as per the life cycle), especially given the latest insight into the passive/active voicing mechanisms in sound
production and perception (Jansen 2004, Blevins 2004) whereby FD can be attributed to the lack of a voicing target on the right.
The fact that underlying voiced obstruents are more prone to voicing than their voiceless counterparts (Strycharczuk) provides further support for the assumption that there is no FD at the word level in Kraków, although interspeaker variation and optionality might suggest a move in the direction of domain narrowing. Thus, the contrast between Kraków and Warsaw Polish can be interpreted as a difference in the domain application of FD which ascends to the word-level in Warsaw, superseded only by obstruent cluster agreement.