Phonology Circle will meet on Mondays this semester from 5–7pm in 32-D831 unless otherwise noted. The Phonology Circle is a weekly forum for the presentation of current research in phonology and phonetics. If you want to receive the email announcements, or have any other comments about Phonology Circle, please email Michael Kenstowicz.
Peter Graff and Steven Keyes
What Makes a Base?
Learning nonconcatenative morphological units via Bayesian inference
I will demonstrate an extension to the Goldwater model of Bayesian morphological segmentation (Goldwater, Griffiths & Johnson 2009) to handle the unsupervised learning of nonconcatenative morphologies for languages such as Arabic and Hebrew. I will then discuss a number of linguistic questions that can receive empirical answers by applying the extended model to actual lexica.
Miwako Hisagi (MIT RLE Speech Communication Group)
Perception of Japanese vowel duration contrasts by L1 and L2 learners of Japanese: An EEG/MEG study
(Joint work with Shigeru Miyagawa, Valerie Shafer, Hadas Kotek, Ayaka Sugawara, Dimitrios Pantazis)
One challenge of second language (L2) acquisition research is to evaluate to what extent experience with an L2 leads to changes in automaticity of L2 speech perception. It is important to address whether L2 perception becomes more automatic with increasing experience. The present study investigated the MMN/MMF component. 12 native speakers of Japanese (JP), 12 naïve American English (AE) listeners (i.e., no knowledge of JP) and 12 L2 learners of JP who have acquired some knowledge of JP (i.e., one semester of Japanese) were tested on a vowel duration contrast (tado-taado) to determine whether experience with JP in a classroom leads to sufficiently robust selective perception routines (SPRs) to indicate automatization of speech perception at least for some learners. We used a visual attention task in which attention was directed away from the auditory stimuli. In result, the native JP listeners showed the largest MMN and the naïve AE listeners showed the smallest MMN, with the AE learners of JP showing an intermediate MMN. This study suggests that experience with the L2 (JP) leads to increasing robustness of discrimination of L2 phonemic contrasts, but that these representations (or SPRs) are still less robust than for L1 listeners and that attention modulates these findings.
A linguistic deficit in Huntington's disease? Preliminary evidence from a dissociation between production and perception in a morpho- phonological task.
There is a lively debate in the recent literature about whether the striatum holds a specific role in linguistic processing, or whether it contributes to linguistic processing indirectly, through its role in executive control, memory and attention (Ulmann 2004; Teichmann et al. 2005, 2008; Chan et al. 2012; Mestres-Missé et al. 2012).
Huntington's Disease (HD) offers a unique model of primarily atrophy of the striatum with simultaneous decline in various cognitive functions. Although language impairment in this disease has been described in the literature (Teichmann et al., 2005, 2006, 2008, 2009, Sambin et al. 2012), evidence is scattered and the specific nature of the deficit has yet to be understood.
In this talk I present data from two experiments conducted with French HD patients. The experiments tested the morphophonological knowledge of the patients, looking specifically at gender alternations within the adjectival paradigm (e.g. [petit] FEM- [peti] MASC 'small' ). Although patients had been previously reported to be impaired in a similar task, I show that the deficit is present in a production task, but absent in perception (grammaticality judgment task).
I discuss a possible grammatical explanation for this data and try to put it in the wider context of the role of the striatum in language and other linguistic deficits observed in this disease.
Data collection was completed only very recently and its analysis is still in its early stages. I am looking forward to comments and discussion.
Keiichi Tajima (Professor, Dept. of Psychology, Hosei University / Visiting Researcher, Speech Communication Group, Research Laboratories of Electronics, MIT)
Perception of prosody in a non-native language: The case of Japanese listeners’ perception of English syllable structure
Learners of a second language (L2) are known to have difficulty in the production and perception of not just phonetic contrasts that are not found in their native language (L1), but also prosodic properties that diverge from their L1, such as syllable structure, rhythm, and intonation. In the present study, I report results from a series of studies that investigated the extent to which native Japanese listeners have difficulty perceiving syllables in spoken English, given the fact that English generally has more complex syllable structures than Japanese. Results show that Japanese listeners indeed have great difficulty accurately counting syllables in spoken English words. Perceptual identification training, however, significantly improves their performance. The non-native listeners’ difficulty was strongly related to phonological factors such as the syllable complexity of the English words, but it was not related to phonetic factors such as the speaking rate of the words, nor to lexical factors such as the presence or absence of loanwords in Japanese that are etymologically / semantically related to yet phonologically distinct (with divergent syllable structures) from the source words, e.g., English source word “stress” and its corresponding Japanese loanword “sutoresu”.
Suyeon Yun (MIT)
The Role of Acoustic Cues in Consonant Cluster Adaptation
In this talk I will argue that cross-linguistic asymmetries in nonnative cluster repairs - namely vowel epenthesis and consonant deletion - result from perceptual similarity between the clusters and their repaired forms. Expanding my survey in Yun (2012), I suggest some new typological generalizations. First, when the edge consonant is a stop, a vowel is most likely to be epenthesized after the stop. If not, the edge stop deletes, while the non-edge stop does not (e.g., gdansk (Polish) → [g?dænsk] (English; Yun 2012), ‘compact’ → [kh?mp?kth?] (Korean; Yun 2012), ‘blanket’ → [l?nketti] (Finnish; Karttunen 1977), ‘compact’ → [kompak] (Indonesian; Yun 2012)). Second, when the edge consonant is a sonorant, a vowel is epenthesized before the sonorant. (e.g., lbovskij (Russian) → [ylbovskij] (Kirghiz; Gouskova 2003), rubl’ (Russian) → [rub?l] (Kirghiz; Yun 2012)). While vowel-adjacent sonorants frequently delete, word-edge sonorants do not delete; even in a language where deletion can be a repair strategy, word-edge sonorants undergo epenthesis rather than deletion (e.g., ‘Swaziland’ → [suasilaan] vs. ‘Seattle’ → [siiaatul] (Inuktitut; Pollard 2008)). In addition, I will report several interesting asymmetries concerning positions, source languages, places of articulation, and voicing in stop adaptations, which cannot be explained by standard phonological constraints. Based on the typology, I will argue that acoustic cues, such as stop release bursts and sonorant-internal cues, play a crucial role in consonant cluster adaptation and formulate them as phonetically-based faithfulness constraints. And it will be shown that interactions between fixed rankings of the phonetically-based constraints (based on the P-map hypothesis (Steriade 2001/2008)) and markedness constraints successfully account for the comprehensive cross-linguistic patterns of the nonnative cluster repairs.