Word-final Devoicing in Heritage Polish Paulina Lyskawa, Ruth Maddeaux, Emilia Melara, & Melanie Michaud University of Toronto Eighth Heritage Language Research Institute Harvard University Cambridge, MA June 1-4 1
Word-final obstruent devoicing: Standard Polish Polish is known for its robust system of obstruentfinal devoicing (Gussmann 2007) TABLE 1. Data motivating a word-final devoicing rule in Polish Singular Plural Gloss Underlying representation klup klubi club klub trut trudi labor trud wuk wuji lye wug (adapted from Kenstowicz & Kissberth 1979, cited in Slowiaczek & Dinnsen 1985) /teraz/ [teraz] now /teraz/ [teras] now 2
Word-final obstruent devoicing: Toronto English English also exhibits word-final devoicing (Smith 1997, Smith 2013) /haʊsəz/ [haʊsəz] houses /haʊsəz/ [haʊsəs] houses 3
Word-final obstruent devoicing Little written on the conditions for word-final obstruent devoicing in English and specifically in Toronto English (Smith 2013) Research on how/whether the Polish neutralization rule is retained in a Polish variety in intense contact with English (e.g. Heritage Polish) has yet to be represented in the literature on language contact and heritage speech 4
Research program We investigate Toronto English contact effects on word-final obstruent devoicing in the speech of Heritage Polish speakers Goal: To determine the degree to which heritage Polish speakers in Toronto devoice word-final obstruents compared to that of Canadian English speakers and Homeland Polish speakers. 5
Research program We ll show that: Homeland Polish and Toronto Anglo English devoice in different ways Heritage Polish speakers appear to employ the devoicing processes of both Homeland Polish and Toronto Anglo English, evident from frequency effects and variable grammar 6
Roadmap Introduction Background Methodology Results Discussion Conclusion 7
Background: Word-final obstruent devoicing Word-final obstruent devoicing found in many languages e.g., Dutch, Catalan, Polish [-SONORANT] > [-VOICE] / # 8
Background: Word-final devoicing in Standard Polish Polish devoicing is said to apply: before a pause (Gussmann 2007: 289; Slowiaczek and Dinnsen 1985: 327) [-SONORANT] > [-VOICE] / ## chleb [xlɛp] bread nóż [nuʃ] knife chleb-a [xlɛb-a] gen. sg. noża [nɔʒ-a] gen. sg. 9
Background: Word-final devoicing in Standard Polish Polish devoicing is said to apply: before a voiceless obstruent (Gussmann 2007: 293; Slowiaczek and Dinnsen 1985: 327) [-SONORANT] > [-VOICE] / [-SONORANT, -VOICE] wróg publiczny [vruk publitʃnɨ] public enemy sąd karny [sɔnt karnɨ] criminal court wrog-a [vrɔg-a] gen. sg. sąd-u [sɔnd-u] gen. sg. 10
Background: Word-final devoicing in Standard Polish Polish devoicing is said to apply: before sonorants (Gussmann 2007: 300; only vowels: Slowiaczek & Dinnsen 1985: 327, fn. 3) [-SONORANT] > [-VOICE] / [+SONORANT] obiad mdły [ɔb j at mdwɨ] bland dinner obiadu [ɔb j ad-u] gen. sg. wróg ojczyzny [vruk ɔjtʃɨznɨ] enemy of a state wrog-a [vrɔg-a] gen. sg. 11
Background: Word-final devoicing in Standard Polish Polish devoicing is said to apply: NOT before voiced obstruents (Gussmann 2007: 292; Slowiaczek and Dinnsen 1985: 327) *[-SONORANT] > [-VOICE] / [-SONORANT, +VOICE] mózg doktora [muzg dɔktɔra] doctor s brain sąd wojenny [sɔnd vɔjɛnnɨ] court martial 12
Background: Toronto English Smith (2013): Toronto English stops tend to be devoiced at the end of a phrase and before a voiceless consonant 13 (Smith 2013:2)
Background: Word-final devoicing in English and Polish The two processes are different (Iverson & Salmons 2011): Polish phonological complete neutralization English phonetic partial devoicing 14
Methodology Data taken from three corpora: Heritage Language Documentation Corpus (Nagy 2009), N=7 Homeland Polish Corpus (Nagy 2009), N=10 Contact in the City Corpus (Hoffman & Walker 2010), N=10 15
Methodology Data: Hour-long sociolinguistic interviews ~120 tokens for each speaker with word-final underlyingly voiced obstruents 16
Methodology Excluded: affricates (too few tokens in each speaker group) inflectional suffixes (confound in Polish, only gen. pl. suffix /-uv/) 17
Methodology 3-level impressionistic coding 2 independent coders and 3 rd one if case of discrepancy Dependent variable: voicing (voiced vs. devoiced) 18
Methodology Linguistic variables: type of obstruent (fricative vs. stop) following segment (pause, voiced obstruent, voiceless obstruent, vowel, nasal, approximant) type of word (open vs. closed) Social variables: sex and age 19
Methodology For Heritage Polish, also: Ethnic Orientation score (0 English-oriented; 2 - Polish-oriented) (Keefe & Padillla 1987) Subset of EO pertaining to language use Code-switching per minute Multivariate analysis conducted in Rbrul 20
Results: Frequency FIGURE 1. Rate of word-final obstruent devoicing across groups Difference between Heritage and Homeland Polish statistically significant (p = 7.07e-05) 21
Results: Multivariate analysis FIGURE 2. Social factors conditioning word-final devoicing SEX HERITAGE HOMELAND ENGLISH TOTAL: 687 TOTAL N: 1050 INPUT: 0.67 TOTAL N: 803 INPUT: 0.25 female 0.542 0.63 78% 200 0.33 0.58 70% 633 0.285 0.57 31% 373 male -0.542 0.37 72% 487-0.33 0.42 58% 417-0.285 0.43 26% 430 AGE Range 36 Range 16 Range 14 youngest -0.957 0.28 71% 558 0.252 0.56 66% 431 [ ] [ ] middle 0.957 0.72 88% 129-0.302 0.43 65% 313 [ ] [ ] oldest / / / / 0.051 0.51 65% 306 [ ] [ ] Range 64 Range 13 22
Results: Multivariate analysis FIGURE 2. Social factors conditioning word-final devoicing SEX HERITAGE HOMELAND ENGLISH TOTAL: 687 TOTAL N: 1050 INPUT: 0.67 TOTAL N: 803 INPUT: 0.25 female 0.542 0.63 78% 200 0.33 0.58 70% 633 0.285 0.57 31% 373 male -0.542 0.37 72% 487-0.33 0.42 58% 417-0.285 0.43 26% 430 AGE Range 36 Range 16 Range 14 youngest -0.957 0.28 71% 558 0.252 0.56 66% 431 [ ] [ ] middle 0.957 0.72 88% 129-0.302 0.43 65% 313 [ ] [ ] oldest / / / / 0.051 0.51 65% 306 [ ] [ ] Range 64 Range 13 23
Results: Multivariate analysis Linguistic factors English 3 significant factor groups following segment, obstruent, word type Homeland Polish 1 significant factor group following segment Heritage Polish 2 significant factor groups following segment, obstruent 24
Results: Multivariate analysis FIGURE 3. Linguistic factors conditioning word-final devoicing HERITAGE HOMELAND ENGLISH TOTAL: 687 TOTAL N: 1050 INPUT: 0.67 TOTAL N: 803 INPU T: 0.25 FOLLOWING SEGMENT factor logodds FW % N logodds FW % N logodds FW % N voiceless obstruent 1.835 0.86 92% 222 1.975 0.88 94% 366 0.282 0.57 35% 122 pause 1.698 0.85 92% 136 1.643 0.84 90% 136 2.105 0.89 71% 120 nasal -0.239 0.44 68% 84-0.614 0.35 54% 157-0.734 0.32 19% 36 vowel -0.377 0.41 71% 66 0.791 0.69 82% 74-1.072 0.26 12% 323 approximant -0.53 0.37 67% 75-1.078 0.25 44% 107-0.298 0.43 29% 96 voiced obstruent -2.387 0.08 24% 104-2.718 0.06 13% 210-0.283 0.43 26% 106 Range 76 Range 82 Range 63 OBSTRUENT fricative 0.569 0.64 76% 494 [ ] [ ] 0.639 0.66 38% 461 stop -0.569 0.36 69% 193 [ ] [ ] -0.639 0.35 16% 342 Range 28 Range 31 WORD TYPE closed [ ] [ ] [ ] [ ] 0.28 0.57 36% 425 open [ ] [ ] [ ] [ ] -0.28 0.43 21% 378 Range 14
Results: Multivariate analysis FIGURE 3. Linguistic factors conditioning word-final devoicing HERITAGE HOMELAND ENGLISH TOTAL: 687 TOTAL N: 1050 INPUT: 0.67 TOTAL N: 803 INPUT: 0.25 FOLLOWING SEGMENT factor logodds FW % N logodds FW % N logodds FW % N voiceless obstruent 1.835 0.86 92% 222 1.975 0.88 94% 366 0.282 0.57 35% 122 pause 1.698 0.85 92% 136 1.643 0.84 90% 136 2.105 0.89 71% 120 nasal -0.239 0.44 68% 84-0.614 0.35 54% 157-0.734 0.32 19% 36 vowel -0.377 0.41 71% 66 0.791 0.69 82% 74-1.072 0.26 12% 323 approximant -0.53 0.37 67% 75-1.078 0.25 44% 107-0.298 0.43 29% 96 voiced obstruent -2.387 0.08 24% 104-2.718 0.06 13% 210-0.283 0.43 26% 106 Range 76 Range 82 Range 63 OBSTRUENT fricative 0.569 0.64 76% 494 [ ] [ ] 0.639 0.66 38% 461 stop -0.569 0.36 69% 193 [ ] [ ] -0.639 0.35 16% 342 Range 28 Range 31 WORD TYPE closed [ ] [ ] [ ] [ ] 0.28 0.57 36% 425 open [ ] [ ] [ ] [ ] -0.28 0.43 21% 378 Range 14
Results: Multivariate analysis FIGURE 3. Linguistic factors conditioning word-final devoicing HERITAGE HOMELAND ENGLISH TOTAL: 687 TOTAL N: 1050 INPUT: 0.67 TOTAL N: 803 INPUT: 0.25 FOLLOWING SEGMENT factor logodds FW % N logodds FW % N logodds FW % N voiceless obstruent 1.835 0.86 92% 222 1.975 0.88 94% 366 0.282 0.57 35% 122 pause 1.698 0.85 92% 136 1.643 0.84 90% 136 2.105 0.89 71% 120 nasal -0.239 0.44 68% 84-0.614 0.35 54% 157-0.734 0.32 19% 36 vowel -0.377 0.41 71% 66 0.791 0.69 82% 74-1.072 0.26 12% 323 approximant -0.53 0.37 67% 75-1.078 0.25 44% 107-0.298 0.43 29% 96 voiced obstruent -2.387 0.08 24% 104-2.718 0.06 13% 210-0.283 0.43 26% 106 Range 76 Range 82 Range 63 OBSTRUENT fricative 0.569 0.64 76% 494 [ ] [ ] 0.639 0.66 38% 461 stop -0.569 0.36 69% 193 [ ] [ ] -0.639 0.35 16% 342 Range 28 Range 31 WORD TYPE closed [ ] [ ] [ ] [ ] 0.28 0.57 36% 425 open [ ] [ ] [ ] [ ] -0.28 0.43 21% 378 Range 14
Results: Multivariate analysis FIGURE 3. Linguistic factors conditioning word-final devoicing HERITAGE HOMELAND ENGLISH TOTAL: 687 TOTAL N: 1050 INPUT: 0.67 TOTAL N: 803 INPUT: 0.25 FOLLOWING SEGMENT factor logodds FW % N logodds FW % N logodds FW % N voiceless obstruent 1.835 0.86 92% 222 1.975 0.88 94% 366 0.282 0.57 35% 122 pause 1.698 0.85 92% 136 1.643 0.84 90% 136 2.105 0.89 71% 120 nasal -0.239 0.44 68% 84-0.614 0.35 54% 157-0.734 0.32 19% 36 vowel -0.377 0.41 71% 66 0.791 0.69 82% 74-1.072 0.26 12% 323 approximant -0.53 0.37 67% 75-1.078 0.25 44% 107-0.298 0.43 29% 96 voiced obstruent -2.387 0.08 24% 104-2.718 0.06 13% 210-0.283 0.43 26% 106 Range 76 Range 82 Range 63 OBSTRUENT fricative 0.569 0.64 76% 494 [ ] [ ] 0.639 0.66 38% 461 stop -0.569 0.36 69% 193 [ ] [ ] -0.639 0.35 16% 342 Range 28 Range 31 WORD TYPE closed [ ] [ ] [ ] [ ] 0.28 0.57 36% 425 open [ ] [ ] [ ] [ ] -0.28 0.43 21% 378 Range 14
Results: Multivariate analysis FIGURE 4. Continuous linguistic and social factors conditioning word-final devoicing in Heritage Polish HERITAGE TOTAL: 687 Logodds EOQ continuous [ ] LANG EOQ continuous 1.154 CODE-SWITCHING RATE continuous 0.267 strong correlation between code-switching and frequency of devoicing (r=0.74) 29
Discussion: Heritage Polish speakers employ multiple rules That Heritage Polish speakers devoice significantly more than Homeland Polish speakers suggests they are employing the devoicing rules of both the Homeland Polish variety and English Code-switching results support this: code-switching provides the context in which to employ both rules 30
Discussion: Other possible accounts A. Heritage Polish speakers overgeneralize the Homeland rule a. We don t find Heritage Polish speakers favouring devoicing before a vowel 31
Discussion: Other possible accounts B. Final obstruents are underlyingly voiceless for Heritage Polish speakers a. Should we expect a distinction between fricatives and stops, then? b. This would mean they are behaving non- Homeland- and non-english-like c. Not possible because then we would see forms that remain voiceless despite the following vowel suffix (see Table 1) 32
Discussion: Other possible accounts C. Homeland speakers devoice final obstruents less than Heritage speakers because of the orthography a. How do we test this? 33
Discussion: Further research If we re right with this multiple rule application hypothesis: We should find instances of it in other areas of Heritage Polish grammar We should find this in the grammar of other Heritage languages 34
Conclusion Heritage Polish speakers tend to devoice word-final obstruents: Before voiceless obstruents (like Homeland Polish and English) Before a pause (like Homeland Polish and English) When the obstruent is a fricative (like English; unlike Homeland Polish) 35
Conclusion Heritage Polish speakers devoice word-final obstruents significantly more than Homeland Polish speakers and Toronto English speakers 36
Conclusion Heritage Polish speakers appear to employ both Homeland Polish and English devoicing rules, resulting in their higher devoicing frequency 37
References Gussmann, Edmund. (2007). The phonology of Polish. OUP Oxford. Iverson Gregory K. and Joseph C. Salmons (2011). Final Devoicing and Final Laryngeal Neutralization. Eds. Marc van Oostendorp, Colin J. Ewen, Elizabeth Hume and Keren Rice. The Blackwell Companion to Phonology. Hoffman, M. & Walker, J.A. (2010). Ethnolects and the city: Ethnic orientation and linguistic variation in Toronto English. Language Variation and Change 22(1):37-67. Keefe, S. & Padilla, A. (1987). Chicano Ethnicity. Albuquerque, NM: University of New Mexico Press. Nagy, N. (2014). A sociolinguistic view of null subjects and VOT in Toronto heritage languages. Lingua http://dx.doi.org/10.1016/j.lingua.2014.04.012. Slowiaczek, L. & Dinnsen, D. (1985). On the neutralizing status of Polish word-final devoicing. Journal of Phonetics 13:325-41. Smith, C. (1997). The devoicing of /z/ in American English: effects of local and prosodic context. Journal of Phonetics 25(4):471-500. Smith, J. (2013). Sociophonetic variation of word-final stop voicing in Toronto English. [Handout] Presented at Change and Variation in Canada 7. University of Toronto. May 4-5. 38
Acknowledgements Many thanks go out to Naomi Nagy, Aaron Dinkin, Keren Rice, and Darcie Blainey for their invaluable advice on this project. Additional thanks go out to the participants of the sociolinguistic interviews from which the data for this study were taken. Contact: Paulina Lyskawa: paulina.lyskawa@mail.utoronto.ca Emilia Melara: emilia.melara@mail.utoronto.ca 39