Figure 1 - uploaded by Jonah Katz
Content may be subject to copyright.
Words used in the construction of each artificial language (top); examples of utterances in the spirantization lenition condition (middle); example of a test pair heard in the spirantization condition, with target first and foil second (bottom).
Source publication
This paper presents evidence that 'spirantization', a cross-linguistically common lenition process, affects English listeners’ ease of segmenting novel “words” in an artificial language. The cross-linguistically common spirantization pattern of initial stops and medial continuants (e.g. [gußa]) results in improved word segmentation compared to the...
Contexts in source publication
Context 1
... list of words in each condition, three examples of utterances played during the expo- sure phase, and a test pair for the spirantization lenition condition are shown in Figure 1. ...
Similar publications
Oral and written productions of language seem to correspond to ontologically separate entities. In this paper, we shall not argue against this basic assumption. However, it will be proposed that a careful examination of the writing systems and of particular written productions can provide phonologists with important information about the nature of...
Citations
... • If the goal of word-ini\al strengthening is to ensure reliable cues to word segmenta\on (Katz and Fricke, 2018;White et al., 2020), ideally speakers should ensure clear acous\c or ar\culatory cues in the ini\al por\on of words which happens to be co-extensive with a lexical stem. ...
Language documentation corpora are rich sources of information. This workshop presents methods for accessing phonetic data from these corpora for the purposes of improving language description, phonetic fieldwork, typological research, and more theoretically-motivated laboratory phonological research. Topics covered include transitioning from a documentation corpus to a phonetic corpus, creating forced alignment for languages, pronunciation dictionary preparation, enriching phonetic annotation with Praat scripts, and engaging with theoretical questions through the use of these tools. Illustrative examples involving glottal stop variation in Triqui and consonant lenition in Mixtec are provided. Supplementary materials include all scripts discussed in the presentation and an example sound file.
... Conjointly with the opposite processes of fortition, i.e., devoicing to a voiceless obstruent and strengthening to a voiced stop or affricate, processes classified as continuity lenition result in a surface pattern of weaker allophones appearing within prosodic domains and stronger allophones appearing at domain edges. The functional basis of continuity lenition is that it creates a segmental distribution that conveys information to the listener about the prosodic structure of the utterance (Keating 2006;Kingston 2008; for supporting evidence from psycholinguistic experimentation, see Katz and Fricke 2018;Katz and Moore 2021). "The motivation behind this pattern is that it aligns auditory disruptions with constituent boundaries, and lack of disruption with lack of boundaries, which plausibly helps a listener detect where the boundaries are" (Katz 2016, p. 47). ...
... Adopting Lozano's (2021) implementation of Katz's (2016) approach to fortition and continuity lenition in OT, we identified a constraint ranking that generates the allophonic distribution of voiced prepalatal obstruents while obviating the need to choose either of the alternating consonants as underlying. Our analysis contributes to a growing body of work in phonological theory seeking to flip the conventional narrative that consonant lenition is primarily a speaker-based phenomenon (Keating 2006;Kingston 2008;Katz 2016;Katz and Fricke 2018;Broś et al. 2021;Katz and Moore 2021;Lozano 2021). The spirantization of voiced stops in Spanish has commonly been analyzed in terms of articulatory factors within the speaker's grammar, such as effort reduction or continuancy assimilation (see, e.g., Kirchner 1998Kirchner , 2001Piñeros 2002;González 2006;Colina 2016Colina , 2020Martínez-Gil 2020). ...
This article traces the development of voiced prepalatal obstruents /dʒ/ and /ʒ/ in Judeo-Spanish, the language spoken by the Sephardic Jews since before their expulsion from late-15th century Spain. Using Medieval Spanish as a comparative starting point, we examine diachronic innovations in the phonological status and distribution of affricate /dʒ/ and fricative /ʒ/ in Judeo-Spanish during the diaspora, focusing in particular on the effects of lexical borrowing from Turkish and French in territories of the former Ottoman Empire. In contemporary Sephardic communities that are in contact with non-Sephardic varieties of Mainstream Spanish, some speakers occasionally replace syllable-initial /∫/, /dʒ/, and /ʒ/ in certain Judeo-Spanish words by a voiceless velar /x/ in efforts to accommodate the pronunciation of the corresponding Mainstream Spanish cognate form. We provide a novel analysis of Judeo-Spanish voiced prepalatal obstruents, including their diachronic and synchronic variation under language contact. The analysis combines a constraint-based approach to phonological alternations, as formalized in Optimality Theory, with a usage-based representation of the mental lexicon, as proposed in Exemplar Theory, to account for speaker- and word-specific variability. A hybrid theoretical model provides a more nuanced understanding of the relationship between lexicon and grammar in Judeo-Spanish phonology than is available in previous structuralist descriptions.
... Saffran et al., 1996aSaffran et al., , 1996b: all else being equal, lower-probability transitions are more likely to be inferred to span a group boundary than higher-probability transitions. The relevant aspect of this research for our current purposes is evidence, from artificial and natural languages and from listeners of all ages, that all of the acoustic cues to grouping discussed in section 2.2 can reinforce or conflict with statistical cues and affect segmentation (Bagou et al., 2002;Christophe et al., 2003;Katz and Fricke, 2018;Kim, 2004;Millotte et al., 2011;Nakatani and Dukes, 1977;Saffran et al., 1996b). ...
This paper reviews evidence concerning the nature of grouping in music and language and their interactions with other linguistic and musical systems. I present brief typological surveys of the relationship between constituency and acoustic parameters in language and music, drawing from a wide variety of languages and musical genres. The two domains both involve correspondence between auditory discontinuities and group boundaries, reflecting the Gestalt principles of proximity and similarity, as well as a nested, hierarchical organization of constituents. Typically, computational-level theories of musical grouping take the form of a function from acoustic properties through grouping representations to syntactic or interpretive constituents. Linguistic theories tend to be cast as functions in the opposite direction. This study argues that the difference in orientation is not grounded in principled differences in information flow between the two domains, and that reconceptualizing one or both theories allows for gains in analytical understanding. There are also obvious differences between musical and linguistic grouping. Grappling with those differences requires one to think in detail about modularity, information flow, levels of description, and the functional nature of cognitive domains.
... Musical event hierarchies encode constituency, which is associated with Gestalt grouping principles on one hand and with tonal and motivic interpretation on the other. Prosodic structure encodes constituency, which is associated with Gestalt-relevant sound patterns (Jeon & Nolan, 2013;Katz & Fricke, 2018;Kenter & Féry, 2013) on the one hand and syntactic and information-structural interpretation on the other (e.g. Selkirk, 1984;Steedman, 2000). ...
This paper reviews some basic elements of musical structure, drawing from work in traditional and cognitive musicology, ethnomusicology, psychology, and generative textsetting. Music features two different hierarchical representational components that can both be visualised in grid notation: metrical structure and event hierarchies (also referred to as Time‐Span Reduction). Metrical structure is an abstract pattern of stronger and weaker points in time that form a temporal ‘scaffold’ against which auditory events occur, but is partly independent from those occurring events. Event hierarchies encode the constituency (referred to as grouping) and prominence of actually‐occurring musical events. Basic principles of these components are illustrated with examples from Western children's, folk, and popular song. The need for both metrical hierarchy and event hierarchy is illustrated using mismatches between metrical and rhythmic structure. The formal, conceptual, and empirical features of musical metre are quite different from linguistic stress and prosody, despite the frequent analogies drawn between them. Event hierarchies, on the other hand, are shown to resemble linguistic prosodic structure with regard to all of these features.
... In Experiment 3, we model these lenition patterns using a series of deep neural networks and find that, even with limited training data, we can achieve reasonably high accuracy in the automatic categorization of lenition patterns. The results of this research both complement recent work on the phonetics of lenition in the world's languages (Katz and Fricke, 2018;White et al., 2020) and provide computational tools for modeling and predicting patterns of extreme lenition. ...
... As a result, listeners are regularly exposed to speech with different acoustic cues than one observes in idealized contexts. At the same time, reduction frequently coincides with weak prosodic positions (unstressed syllables; word-internal, phrase-medial, and intervocalic contexts) whereas obstruent fortition frequently coincides with strong prosodic positions (stressed syllables; word-initial, phrase-initial, and phrase-final contexts) (Bouavichith & Davidson, 2013;Cho & Keating, 2001;Fougeron & Keating, 1997;Katz, 2016;Keating, Cho, Fougeron, & Hsu, 2004;Katz & Fricke, 2018). Though reduction often reduces distinctness of a segment relative to its neighbors, the relationship between reduction and prosodic boundaries nevertheless aids the listener in lexical segmentation. ...
... Though reduction often reduces distinctness of a segment relative to its neighbors, the relationship between reduction and prosodic boundaries nevertheless aids the listener in lexical segmentation. Listeners are able use the degree of reduction to make decisions about word boundaries (Katz & Fricke, 2018) just as they are able to use more general prosodic cues for the purpose of word segmentation (Saffran, Newport, & Aslin, 1996;White, Mattys, Stefansdottir, & Jones, 2015). ...
Word-level prosody plays an important role in processes of consonant lenition. Typically, consonants in word-initial position are strengthened while those in word-medial position are lenited (Keating, Cho, Fougeron, & Hsu, 2003). In this paper we examine the relationship between word-prosodic position and obstruent lenition in a spontaneous speech corpus of Yoloxóchitl Mixtec, an endangered Mixtecan language spoken in Mexico. The language exhibits a surprising amount of lenition in the realization of otherwise voiceless unaspirated stops and voiceless fricatives in careful speech. In Experiment 1, we examine the relationships between word position, consonant duration, and passive voicing and find that word-medial pre-tonic position is the locus of both consonant lengthening and less passive voicing. Non-pre-tonic consonants are produced with more voicing and shorter duration. We also find that the functional status of the morpheme plays a role in voicing lenition. In Experiment 2, we examine manner lenition and find a similar pattern – word-medial pre-tonic stops are more often realized with complete closure relative to non-pre-tonic stops, which are more often realized with incomplete closure. In Experiment 3, we model these lenition patterns using a series of deep neural networks and find that, even with limited training data, we can achieve reasonably high accuracy in the automatic categorization of lenition patterns. The results of this research both complement recent work on the phonetics of lenition in the world’s languages (Katz and Fricke, 2018; White et al., 2020) and provide computational tools for modeling and predicting patterns of extreme lenition.
... This study failed to extend to children the finding that spirantization aids word segmentation for 501 adults (Katz & Fricke, 2018). This is unsurprising, because the study only found a marginal 502 effect of learning in any phonetic condition or experimental session. ...
... For the first session, performance is higher on the spirantization condition than the anti-542 spirantization condition, as reported by Katz & Fricke (2018). Even for the anti-spirantization 543 condition, about two thirds of the distribution is above chance (50%). ...
... replicates the findings of Katz & Fricke (2018). A second major pattern is that performance is 569 lower in the second session. ...
Purpose
The aim of the study was to investigate the effects of specific acoustic patterns on word learning and segmentation in 8- to 11-year-old children and in college students.
Method
Twenty-two children (ages 8;2–11;4 [years;months]) and 36 college students listened to synthesized “utterances” in artificial languages consisting of six iterated “words,” which followed either a phonetically natural lenition–fortition pattern or an unnatural (cross-linguistically unattested) antilenition pattern. A two-alternative forced-choice task tested whether they could discriminate between occurring and nonoccurring sequences. Participants were exposed to both languages, counterbalanced for order across subjects, in sessions spaced at least 1 month apart.
Results
Children showed little evidence for learning in either the phonetically natural or unnatural condition nor evidence of differences in learning across the two conditions. Adults showed the predicted (and previously attested) interaction between learning and phonetic condition: The phonetically natural language was learned better. The adults also showed a strong effect of session: Subjects performed much worse during the second session than the first.
Conclusions
School-age children not only failed to demonstrate the phonetic asymmetry demonstrated by adults in previous studies but also failed to show strong evidence for any learning at all. The fact that the phonetic asymmetry (and general learning effect) was replicated with adults suggests that the child result is not due to inadequate stimuli or procedures. The strong carryover effect for adults also suggests that they retain knowledge about the sound patterns of an artificial language for over a month, longer than has been reported in laboratory studies of purely phonetic/phonological learning.
Supplemental Material
https://doi.org/10.23641/asha.13641284
... In Experiment 3, we model these lenition patterns using a series of deep neural networks and find that, even with limited training data, we can achieve reasonably high accuracy in the automatic categorization of lenition patterns. The results of this research both complement recent work on the phonetics of lenition in the world's languages (Katz and Fricke, 2018;White et al., 2020) and provide computational tools for ...
... As a result, listeners are regularly exposed to speech with different acoustic cues than one observes in idealized contexts. At the same time, reduction frequently coincides with weak prosodic positions (unstressed syllables; word-internal, phrase-medial, and intervocalic contexts) whereas obstruent fortition frequently coincides with strong prosodic positions (stressed syllables; word-initial, phrase-initial, and phrase-final contexts) (Bouavichith and Davidson, 2013;Cho and Keating, 2001;Fougeron and Keating, 1997;Katz, 2016;Keating et al., 2004;Katz and Fricke, 2018). Though reduction often reduces distinctness of a segment relative to its neighbors, the relationship between reduction and prosodic boundaries nevertheless aids the listener in lexical segmentation. ...
... Though reduction often reduces distinctness of a segment relative to its neighbors, the relationship between reduction and prosodic boundaries nevertheless aids the listener in lexical segmentation. Listeners are able use the degree of reduction to make decisions about word boundaries (Katz and Fricke, 2018) just as they are able to use more general prosodic cues for the purpose of word segmentation (Saffran et al., 1996;White et al., 2015). ...
Word-level prosody plays an important role in processes of consonant lenition. Typically, consonants in word-initial position are strengthened while those in word-medial position are lenited (Keating et al., 2003). In this paper we examine the relationship between word-prosodic position and obstruent lenition in a spontaneous speech corpus of Yoloxóchitl Mix-tec, an endangered Mixtecan language spoken in Mexico. The language exhibits a surprising amount of lenition in the realization of otherwise voiceless unaspirated stops and voiceless fricatives in careful speech. In Experiment 1, we examine the relationships between word position , consonant duration, and passive voicing and find that word-medial pre-tonic position is the locus of both consonant lengthening and less passive voicing. Word-initial consonants are produced with more voicing and shorter duration. We also find that the functional status of the morpheme plays a role in voicing lenition. In Experiment 2, we examine manner lenition and find a similar pattern-word-medial stops are more often realized with complete closure relative to word-initial stops, which are more often realized with incomplete closure. In Experiment 3, we model these lenition patterns using a series of deep neural networks and find that, even with limited training data, we can achieve reasonably high accuracy in the automatic categorization of lenition patterns. The results of this research both run counter to expected patterns of lenition in the world's languages (Katz and Fricke, 2018; White et al., 2020) and provide computational tools for modeling and predicting patterns of extreme lenition.
... The opposing fortition processes (strengthening) increase disruption in the presence of a prosodic boundary by reducing the intensity of the target segment. Katz & Fricke (2018) show that listeners expect reduced forms to occur in non-boundary positions in an artificial language experiment. We found no evidence that stress affects intensity directly, but that evidence does not directly oppose prosody-based accounts. ...
Studies of variable lenition patterns converged on two phonetic properties as characteristic of lenition: reduced duration and increased intensity. However, the causal precedence of the two factors remains unclear. We focus on the causal structure of variable lenition. Study 1 examines the relationship between three correlates of lenition: speech rate, stress, and low information content, and their effect on reduced duration and increased intensity. We find that though increased intensity is more prototypically viewed as the core aspect of lenition, the effect of the three correlates on intensity is mediated by duration. Study 2 shows that all frequent lenition processes in the Buckeye corpus involve durational reduction. The contribution of this paper is a proposal with a fairly simple principle, with few auxiliary assumptions: Reduced duration precedes increased intensity in variable lenition. *
... Contrastive glottalization and a very complex tonal inventory; root-final stress realized via lengthening (DiCanio et al., 2018(DiCanio et al., , 2019 Stops may be produced with variable degrees of lenition in spontaneous speech. (Bouavichith and Davidson, 2013;Davidson, 2011;Hualde et al., 2011;Katz, 2016;Katz and Fricke, 2018;Katz and Pitzanti, 2019;Lewis, 2001;Torreira and Ernestus, 2011;Warner and Tucker, 2011) Across languages, manner/voicing lenition is more common in word-medial position and in the onsets of unstressed syllables than in word-initial position or in the onsets of stressed syllables. What about here? ...
... voicing by position in words of different sizeNovel finding! Previous work argues that lenition is resisted at word boundaries(Katz and Fricke, 2018;Katz and Pitzanti, 2019). ...
I discuss two areas where the marriage of technical, computational skills and language documentation produces advances beyond each discipline. The first area is the development of computational models predicting surface phonetic allophony. Deep neural networks were able to predict observed surface allophony (voiced, devoiced, spirantized) with 90% accuracy when trained on an annotated, spontaneous speech corpus of Yoloxóchitl Mixtec (Mexico) (DiCanio et al, submitted). This result is not only relevant for endangered language phonetics, but also for the diagnosis of speech apraxia, a disorder typified by incomplete stop closure (Davis et al 1998). The second area focuses on the development of an automatic alignment system for Itunyoso Triqui (Mexico) and the challenges of examining corpus tonal variation in complex prosodic systems, a current topic in speech recognition (c.f. Lin et al 2018). These examples demonstrate how a close collaboration of computational and documentary approaches in linguistic research broadly advance science.
... And if fortis forms occur initially in prosodic domains, they disrupt the speech stream more. By aligning auditory discontinuities with prosodic boundaries, even probabilistically, this should help listeners chunk the speech stream into constituents; Katz and Fricke (2018) provide some preliminary evidence from a word-segmentation experiment that this prediction is correct. Another version of the listener-oriented approach holds that it is information that is aligned with prosodic boundaries (Harris, 2003;Cohen Priva, 2017). ...
This paper gives a detailed description of the consonant system of Campidanese Sardinian and makes methodological and theoretical contributions to the study of lenition. The data are drawn from a corpus of field recordings, including roughly 400 utterances produced by 15 speakers from the Trexenta and Western Campidanese areas. Campidanese has a complex lenition system that interacts with length, voicing, and manner contrasts. We show that the semi-automated lenition analysis presented in this journal by Ennever, Meakins, and Round can be fruitfully extended to our corpus, despite its much more heterogeneous set of materials in a genetically distant language. Intensity measurements from this method do not differ qualitatively from more traditional ones in their ability to detect lenition-fortition patterns, but do differ in interactions with stress. Lenition-fortition patterns reveal at least three levels of prosodic constituent in Campidanese, each of which is associated with medial lenition and initial fortition. Lenition affects all consonants and V-V transitions. It reduces duration, increases intensity, and probabilistically affects qualitative manner and voicing features in obstruents. Mediation analysis using regression modeling suggests that some intensity and most qualitative reflexes of lenition are explained by changes in duration, but not 'vice versa'.
