Syncope in East Cree: Phonological or Phonetic?
CARRIE DYCK, ALETHEA POWER, AND KEVIN TERRY Memorial University of Newfoundland
INTRODUCTION1
Syncope in Algonquian and other languages is often analyzed as a phono- logical process resulting in syllable (vowel) deletion. However, recent work within the Gestural Phonology framework (Browman and Goldstein 1990) suggests that as a consequence of adjustments in articulatory timing, reduced syllables could also be phonemically unchanged but perceptually vowel- less. Evidence includes acoustic measurements that reveal covert contrasts between apparently identical strings such as the [sp] in English s’pose and sport (Davidson 2006:91). While syncope in some Algonquian languages is phonological (i.e., resulting in a reduction of the syllable count), it may be phonetic (i.e., appar- ent, but not affecting the syllable count or structure) in others. Preliminary acoustic measurements, described here, suggest that East Cree syncope is phonetic. Moreover, since East Cree syncope has previously been analyzed as phonological, the data described here suggest the need to re-examine the status of syncope in at least some Algonquian languages. Prior to presenting the data, background information on East Cree syncope is provided. Differences between phonological and phonetic syn- cope are then outlined, leading to the hypothesis that East Cree syncope is a phonetic process that leaves phonological structure intact. The methodology is then provided, followed by ¿ndings and conclusions.
1. Research for this paper was partially funded by SSHRC grants #410-2004-1836 (2004, Brittain, Dyck, and Rose), #410-2008-0378 (2008, Brittain, Dyck, Rose, and MacKenzie), and #856-2004-1028 (Junker, MacKenzie), as well as by the Memorial Undergraduate Career Experience Program. Unless otherwise noted, East Cree data was provided by Darlene Bearskin and Lucy Salt. Special thanks to Yvan Rose and to two anonymous reviewers for insightful comments.
120 SYNCOPE IN EAST CREE 121
BACKGROUND
East Cree, spoken on the East Coast of James Bay, belongs to the Cree- Naskapi-Montagnais dialect continuum. There are two main dialects of East Cree, Northern and Southern. Southern East Cree has both a long ê and â, but in Northern East Cree, long ê has merged with â. (For clarity, we use ߧêߨ instead of traditional orthographic ߧeߨ to represent long /e:/.) Phonetic details for vowels and consonants will be described as necessary throughout the paper. However, it is important to note at the outset that the East Cree plosives /p, t, k/ display little aspiration before vowels. The East Cree phonemic inventory and syllable structure (1) are simi- lar to other Cree-Montagnais-Naskapi dialects (c.f., Plains Cree; Wolfart 1996). Approximate pronunciations for the vowels are î [i(ʁ)], û [u(ʁ)], â [æ(ʁ), İ(ʁ)], ê [˭(ʁ), e(ʁ)], i [ȣ, ȡ, ԥ], u [Ƀ], a [ȣ, ˭, ȡ, ԥ, Ʌ]. In words without syncope (1), East Cree has a CVC syllable template. Onsets can consist of any consonant in the language. Codas are restricted to the fricatives [s, ȼ, and h]. In addition, an extra consonant, drawn from the set of onset con- sonants, is also allowed word-¿nally. (1) Syllable structure in words without syncope; e.g., a.châh.kw ‘soul’
In the literature, syncope is described as a phonological deletion pro- cess that is more likely to occur between voiceless /p,t,k,s,ȼ ɠ/, and also between homorganic consonants such as /t/ and /n/, or /p/ and /m/ (MacK- enzie 1980). East Cree syncope is illustrated in (1).
(2) Northern East Cree syncope examples Orthographic form Phonetic realizations Gloss ’is.pi.kun [s.ɻpܺ.kݜn], [ັT.ɻpܺ.kݜn] ‘taste uh.pi.nim [x.ɻpܼn.nԥm], [ɃI.ɻpܼn.nԥm] ‘s/he lifts it’ uh.tâ.wî.mâu [x.tæ.ޖwi.mæw], [ኺI.tæ.ޖwi.mæw] ‘his/her father’ ush.chî.shikw [ɧȼ.ɻEɋiޝ.ȼȣLɮ], [Ƀȼ.ޖEɋiʁ.ȼȣkɮ] ‘his / her eye’ 122 CARRIE DYCK, ALETHEA POWER, AND KEVIN TERRY
The transcriptions in (2) suggest that syncope is an optional process; for example, the pronunciation of ushchîshikw (2) displays free variation between word-initial [Ƀ] and [ɧ]. Syncope only affects syllables containing a short vowel (MacKen- zie 1980). Moreover, within sequences of short vowels, only certain short vowels are affected, suggesting that the location of syncope is predictable from foot structure; however foot structure is largely abstract, since there is only one pitch-accented vowel per word and hence only one “audible” foot (Dyck et al. 2006, 2008). The optionality shown in (2) suggests a phonetic process and the fact that only certain short vowels are affected in short-vowel spans sug- gests a phonological environment. In our conclusions, we reconcile these observations.
ASSUMPTIONS
We assume two levels in the grammar, Phonology and Phonetic Implementa- tion (PI; Cohn 1993). At the phonological level, we assume abstract metrical units, including feet, syllables, the sub-syllabic constituents of the onset and rhyme, and timing units. (The latter account for the phonemic vowel length distinction in East Cree.) At the Phonetic Implementation level, we assume that abstract phonological units translate into more concrete representations relevant for the timing of physical gestures. We adopt elements of the model of Gestural Phonology (Browman and Goldstein 1990) to demonstrate ges- tures and timing. However, we depart from Gestural Phonology in assuming the presence of abstract phonological units such as the syllable. Two types of vowel deletion—real (or phonological) and apparent (or phonetic)—are predicted within a two-level model. Real vowel deletion occurs at the Phonological level (typically in metrically weak environments); a syllable nucleus and its timing unit are deleted, resulting in the deletion of the audible portion of the vowel via Stray Erasure (Steriade 1982). In contrast, apparent vowel deletion occurs at the level of PI. The nucleus and syllable are unaffected; meanwhile, the melodic content of the nucleus is either less audible or inaudible, due to phonetic reduction processes. (Both phonological and phonetic analyses assume that syncope only happens in metrically weak environments. See, for example Algeo 1975; Hooper 1978; Zwicky 1972 for phonological accounts of English syncope, and Beckman 1996 for a summary of phonetic accounts.) SYNCOPE IN EAST CREE 123
If syncope occurs at the Phonological level in East Cree, then it is dif¿cult to characterize East Cree syllable structure. The problem can be illustrated with the word amiskw ‘beaver’ which is typically pronounced as [Ʌmskɮ] in Northern East Cree: neither a sequence of three consonants nor a coda /m/ can occur in words without syncope, illustrating that words with syncope can depart radically from the syllable template in (1). In contrast, if East Cree syncope occurs at the level of PI, then syllable nuclei would not be deleted. Instead, they would be pronounced differently, with some phonetic characteristic(s) or trace(s) occurring in place of an audible vowel (Beckman 1996; Browman and Goldstein 1990). We would also expect the rule to be gradient, a hallmark of PI rules. (To illustrate gradiency, Cohn 1993 contrasts English and French. English vowel nasaliza- tion is gradient, whereas French vowel nasalization is categorical. In Eng- lish, the degree of vowel nasalization/orality varies with the distance from the nasalizing consonant; in contrast, in French, vowels are either greatly nasalized or completely oral. Cohn attributes the difference to the status of nasalization in English versus French, arguing that vowel nasalization is phonetic in English and phonological in French.) Research within the Gestural Phonology (GP) framework shows that at least in some cases, syncope is a process of apparent but not actual vowel deletion. Apparent syncope is an acoustic by-product of articulatory restructuring, motivated by concerns such as “motor economy” or “ease of articulation,” in metrically weak (i.e., reduced) syllables. Due to articulatory restructuring, consonants in the syllable margin can overlap and sometimes completely eclipse the vocalic gestures, to the point where the vowel is inau- dible (Beckman 1996; Coleman 1992, 1994, 2001; Davidson 2006; Dirksen and Coleman 1997; Goad and Brannen 2003). Articulatory correlates of consonant gestural overlap are discussed in the Results section. To illustrate the concept of gestural overlap, we use a modi¿ed GP diagram, shown in (3) and (4). A gesture is represented as a series of timed events, beginning with the onset of the constellation of consonant or vowel gestures (left-edge diagonal line), the target gesture, a period of time during which the target is maintained (horizontal line), and the release and offset (right-edge diagonal line). Example (4) illustrates the hypothetical timing of gestures in non- elided syllables where there is relatively little gestural overlap. The timing of the offset of [k] and the onset of [ȼ] overlap with the vowel, but there is a period of time when only the vocalic gestures (dotted line) occur. The 124 CARRIE DYCK, ALETHEA POWER, AND KEVIN TERRY vowel is audible because there is time for enough glottal pulses to support audible vowel formants.
(3) [ȼ, n] in coda; no syncope; pronounced as [kɃȼ, kȣn]
k ݜ ݕ k ܼ n
Example (3) illustrates the hypothetical timing of gestures in “elided” syllables where there is a relatively greater degree of gestural overlap. The onset of [ȼ/n] eclipses the vowel almost completely. In cases where overlap is not as extreme and the vowel is longer, a partly devoiced vowel could result; conversely, the vowel could be rendered completely inaudible if there were not enough glottal pulses to support audible vowel formants; (see Beckman 1986:102–103 for examples of partial-to-complete vowel devoic- ing / shortening in Korean, Japanese, and possibly, French). As well, the [ȼ/n] could be elongated if the syllable length were unchanged, or it could simply be produced more immediately after the onset, if the syllable length were shortened.
(4) Syllables with [n, ȼ] in coda; gestural overlap (syncope); pronounced as [kȼȼ/kȼ, kᆱ/kn]
k ݜ ݕ o k ܼ o n
Finally, the location of elided syllables can depend on phonological environments, particularly metrical ones. (See the Conclusion for further discussion.) SYNCOPE IN EAST CREE 125
HYPOTHESIS
We hypothesize that East Cree syncope does not affect abstract syllable structure, but is instead a matter of Phonetic Implementation. (The default hypothesis, that syncope is phonological, was explored in the Assumptions section.) We predict that the gestural timing of segments in metrically weak syllables will differ from the gestural timing of segments in metrically strong syllables. In particular, we might expect increased consonant length in syn- cope environments as a consequence of gestural overlap. An open question is whether the inherent duration of an apparently eclipsed vowel would be affected in syncope environments; it would depend on whether or not syncope at the level of Phonetic Implementation involved the independent process of syllable shortening.
METHODOLOGY
To determine the nature of East Cree syncope, we performed four pilot experiments, using data from one Southern East Cree speaker and one Northern East Cree speaker. (While it is not ideal to mix sub-dialects, these are pilot studies.) The data for acoustic analysis consisted of word lists produced at a relatively slow speech rate. (We mention speech rate because in English, the degree of pre-stress syncope in words like potato and sup- pose is partly dependent on fast speech rate, at least for some speakers; Davidson (2006:95).) To determine whether syncope would involve shortening, Dyck mea- sured the length of elided and unelided syllables with a CV and CVh struc- ture in Southern East Cree; Logan measured the length of unelided vowels that were pitch-accented and non-pitch-accented in Southern East Cree. To examine the issue of consonant length, Terry measured the length of non-word-¿nal /s, ȼ, m, n/ in onset and coda positions in elided and non- elided syllables in Southern East Cree. (Only [s, ȼ] occur in both onset and coda position. After determining that there was no interaction between onset and coda length, we conÀated the positions.) Examples include Southern East Cree NIkus [OȣɻkɃsʁ] ‘my son’ versus NIshikî [ᆱɻȼȣkiʁ] ‘my skin.’ Finally, Power measured the length of plosive release of /p, t, k/ in onsets in elided and non-elided syllables in Northern East Cree. Examples include Northern East Cree TIpâchimuwin [tԥpatȼȣɻmuwԥn] ‘story’ and aTIHkw [ɻɅUɧLɮ> ‘caribou.’ 126 CARRIE DYCK, ALETHEA POWER, AND KEVIN TERRY
Measurements were made using Praat (Boersma and Weenink 2009). Units were identi¿ed through a combination of impressionistic transcriptions and examination of the waveform and spectrogram. Examples of measure- ments are provided in (5)–(6). (5) Plosive aspiration, no syncope ânisKUtâpân ‘great grand-parent’ [… kɃ …] SYNCOPE IN EAST CREE 127
(6) Plosive aspiration, syncope aKUHp ‘coat’ [… kހ… ]
FINDINGS
Syllable length
We found no signi¿cant difference in length between non-elided CV syllables (M = 0.14262, SD = 0.00132) and elided CV(h) syllables (M = 0.13659; SD = 0.00079); (p > 0.05). Logan (2010) also found no 128 CARRIE DYCK, ALETHEA POWER, AND KEVIN TERRY
signi¿cant difference in length between pitch-accented vowels (M = 0.0716, SD = 0.0032) and non-pitch-accented vowels in unelided CV syllables (M = 0.0647, SD = 0.0018); (p > 0.05).
Consonant Length
We found consonants to be signi¿cantly shorter in syllables with full vowels, and signi¿cantly longer in syllables with apparently elided vowels. For plo- sives, the aspirated release (as opposed to the consonant itself) was longer. As shown in (7), the fricatives /s, ȼ/ in Southern East Cree are shorter when followed by a full vowel (M = 0.1241, SD = 0.0007), than when followed by an apparently deleted vowel (M = 0.1639, SD = 0.0016); (t(88) = 6.1061, p < 0.01). (7) Duration of non-word-¿nal fricatives
Average duration of fricatives
0.18 0.16 0.14 0.12 0.1 Mean ms 0.08 0.06 0.04 0.02 0 No syncope Syncope condition
As shown in (8), the nasals /m, n/ in Southern East Cree are shorter when followed by a full vowel (M = 0.0234, SD = 0.0000), than when followed by an apparently deleted vowel (M = 0.0628, SD = 0.0005); (t(103) = 4.1578, p < 0.01). SYNCOPE IN EAST CREE 129
(8) Duration of non-word-¿nal nasals
Average duration of nasals
0.09 0.08 0.07 0.06 0.05 Mean ms 0.04 0.03 0.02 0.01 0 No syncope Syncope condition
As shown in (9), the plosives /p, t, k/ in Southern East Cree have little aspiration when followed by a full vowel (M = 0.0234, SD = 0.0000), but are heavily aspirated when followed by an apparently deleted vowel (M = 0.0855, SD = 0.0013); (t(52) = 11.6619, p < 0.01). (9) Length of release for non-word-¿nal plosives
Plosive aspiration
0.09 0.08 0.07 0.06 0.05 Mean ms 0.04 0.03 0.02 0.01 0 No syncope Syncope condition 130 CARRIE DYCK, ALETHEA POWER, AND KEVIN TERRY
Moreover, our ¿ndings are replicated cross-linguistically. First, there is a relationship between fricative length and elision in English, where, for [ԥ]-ጸ alternations such as s’pose/suppose, “… the duration of /s/ is signi¿cantly longer in tokens with elision than in tokens which exhibit the vowel . . .” (Davidson 2006:91). Second, plosive aspiration [ɧ] is in complementary distribution with schwa [ԥ] in a number of Salishan languages; as in East Cree, conditioning factors include unstressed position, and location of [ԥ] between voiceless segments (Urbanczyk 1977:77–80). There is also a relationship between aspiration and schwa in English; as in East Cree, conditioning factors deter- mining whether [ԥ] or [ɧ] are perceived or produced include unstressed posi- tion, location of [ԥ] between voiceless segments, and, for some speakers, speech rate (Beckman 1996; Davidson 2006). Finally, alternations between aspiration [ɧ] and vowels other than schwa are attested cross-linguistically: for example, voiceless vowels are realized as aspiration with vowel-like formant structure in Cayuga (Doherty 1993:276–94) and in Comanche (Jakobson et al. 1967:52). These observations suggest that aspiration is an alternative way of realizing vowel nuclei.
CONCLUSIONS East Cree vowels and syllables appear to be the same length, regardless of whether they are pitch-accented, non-pitch-accented, or elided. This obser- vation suggests that syncope in East Cree does not involve shortening and that it does not affect phonological timing units. Moreover, consonants are lengthened in elided syllables; this observation is compatible with our hypothesis that syncope is a matter of gestural overlap: consonants are longer and vowels are shorter (or impressionistically deleted) in syncope environments. These observations suggest that syncope is fundamentally a matter of phonetic implementation. However, in previous work cited above, we found the location of apparently deleted syllables to be phonologically determined: apparently deleted syllables always occur in metrically weak environments. Both sets of observations can be reconciled by assuming that syncope is funda- mentally a phonological process, with effects that play out at the level of Phonetic Implementation. The environment for syncope (metrically-weak SYNCOPE IN EAST CREE 131 versus strong syllables) is determined in the Phonology. Then, at the level of Phonetic Implementation, apparent deletion in weak syllables occurs as an acoustic by-product of articulatory restructuring, for reasons cited earlier. Assuming this analysis, East Cree syllable structure is unchanged and East Cree syllables conform to the syllable template in (1); elided syllables retain their nucleus and timing unit, but the vowels are imperceptible due to overlap by surrounding consonants. The above analysis raises the question of whether vowel alterna- tions described as “deletion” or “devoicing,” should be re-examined in other Algonquian languages. Some descriptions suggest that a second look might be warranted. For example, in Passamaquoddy, the phonetic length of “unstressable” [ԥ]’s which survive the syncope process, varies with the phonetic length of the preceding syllable:
In a word like SVpKSΩQΩN ‘onions,’ for example, in which e is phonetically a long vowel [æޞ] because of the following hC, the Ω of the next syllable is generally quite short and frequently omitted altogether. After the relatively short a of iNΩPΩN ‘snowshoes,’ however, unstressable Ω is comparatively long and would not ordinarily be deleted except at a high rate of speech. (LeSourd 1993:119)
As well, in Cheyenne, devoiced [อ] becomes [s] or [š] when next to LTTT> ‘short distance’ (Frantzڔs/ or /š/ respectively, as in /kah.ke.se/ [ka/ 1972:10). Such processes may be more insightfully viewed as matters of Phonetic Implementation than as categorical phonological rules.
REFERENCES
Algeo, John. 1975. Syncope and the phonotactics of English. General Linguistics 15(2):71–78. Beckman, Mary. 1996. When is a syllable not a syllable? Phonological structure and language processing: Cross-linguistic studies, ed. by Cutler Otake, pp. 95–123. New York: Mouton de Gruyter. Boersma, Paul, and David Weenink. 2009. Praat: Doing Phonetics by Computer (Version 5.1;
Cohn, Abigail. 1993. Nasalisation in English: phonology or phonetics. Phonology 10(1):43–82. Coleman, John. 1992. The phonetic interpretation of headed phonological structures containing overlapping constituents. Phonology 9:1–44. Coleman, John. 1994. Polysyllabic words in the YorkTalk synthesis system. Phonological Structure and Phonetic Form: Papers in Laboratory Phonology III, ed. by Patricia A. Keating, pp. 293–324. Cambridge: Cambridge University Press. Coleman, John. 2001. The phonetics and phonology of Tashlhiyt Berber syllabic con- sonants. Transactions of the Philological Society. 99(1):29–64. Davidson, Lisa. 2006. Schwa elision in fast speech: segmental deletion or gestural over- lap? Phonetica 63:79–112. Dirksen, Arthur, and John Coleman. 1997. All-prosodic synthesis architecture. Progress in Speech Synthesis, ed. by Jan P. H. van Santen, Richard W. Sproat, Joseph P. Olive and Julia Hirschberg, pp. 91–108. New York: Springer-Verlag. Doherty, Brian. 1993. The acoustic-phonetic correlates of Cayuga word-accent. PhD thesis, Harvard University. Dyck, Carrie, Julie Brittain, and Marguerite MacKenzie. 2006. Northern East Cree accent. Proceedings of the 2006 Annual Conference of the Canadian Linguistics Association. 14 pages. Dyck, Carrie, Julie Brittain, Marguerite MacKenzie, and Yvan Rose. 2008. Syncope in Northern East Cree. Poster presented at the Annual Meeting of the Canadian Linguistic Association/Congress for the Social Sciences and Humanities, University of British Columbia, May 31–June 2, 2008. Frantz, Donald. 1972. Cheyenne distinctive features and phonological rules. International Journal of American Linguistics 38:6–13. Goad, Heather, and Kathleen Brannen. 2003. Phonetic evidence for phonological struc- ture in syllabi¿cation. The Phonological Spectrum, Vol. 2, ed. by Jeroen van de Weijer, Vincent van Heuven, and Harry van der Hulst, pp. 3–30. Amsterdam: John Benjamins. Hooper, Joan B. 1978. Constraints on schwa-deletion in English. Recent developments in historical phonology, ed. by Jacek Fisiak, pp. 183–208. The Hague: Mouton. Jakobson, Roman, C., Gunnar M. Fant, and Morris Halle. 1967. Preliminaries to speech analysis: The distinctive features and their correlates. Cambridge, Massachusetts: MIT Press. Lesourd, Philips S. 1989. Accent and syllable structure in Passamaquoddy. New York: Garland. Mackenzie, Marguerite. 1980. Towards a dialectology of Cree-Montagnais-Naskapi. PhD thesis, University of Toronto. Steriade, Donca. 1982. Greek prosodies and the nature of syllabi¿cation. PhD thesis, Massachusetts Institute of Technology. Urbanczyk, Suzanne. 1997. Patterns of reduplication in Lushootseed. PhD thesis, Uni- versity of Massachusetts, Amherst. Valentine, J. Randolph. 1994. Ojibwe dialect relationships. PhD thesis, University of Texas at Austin. SYNCOPE IN EAST CREE 133
Wolfart, H. C. 1996. Sketch of Cree, an Algonquian language. Handbook of North Ameri- can Indians. Volume 17. Languages, ed. by Ives Goddard, pp. 390–439. Washington: Smithsonian Institution. Zwicky, Arnold. 1972. Note on a phonological hierarchy in English. Linguistic change and generative theory, ed. by Stockwell, R. P. and R. K. S. Macaulay, pp. 275–301. Bloomington: Indiana University Press.