Strong Positions and Laryngeal Features in Yukatek Maya∗ Scott AnderBois, UC Santa Cruz [email protected] NELS 39 November 7, 2008

1

Goals • Present evidence from two phonological phenomena in Yukatek Maya which challenge a ‘pure privilege’ view of positional asymmetries by inserting marked forms only in certain ‘weak’ positions. • Demonstrate that rather than being random exceptions, these processes are part of a robustly attested cross-linguistic pattern: a preference for certain laryngeal features at the end of certain prosodic constituents. • Argue that this pattern is the phonologization of a gradient pressure: the difficulty of maintaining voicing throughout relatively large prosodic constituents.

2

Positional Privilege in Yukatek Maya • Yukatek Maya (YM) has two phonological phenomena which occur only in certain ‘weak’ positions. Yukatek Maya has the following consonant inventory1 :

(1)

Plain Stop Ejective Implosive Fricative Nasal Glides/Liquids

Labial p p’ b m

Alveolar t ts t’ ts’

Palatal ch ch’

s n l

x

Velar k k’

Glottal ’

h

y

w

• Yukatek Maya has a five vowel system with the expected vowel qualities: [a], [e], [i], [o], [u]. • Additionally, it has a four way suprasegmental distinction between short vowels written as [a], long low tone vowels [aa], long high tone vowels [´aa], and creaky-voiced vowels [a’a]. ∗

Many thanks to all my Yukatek Maya consultants for their patience and cooperation. Thanks also to Junko Ito for her guidance and support, to Armin Mester, Judith Aissen, audiences at LASC and TREND 2008, Jaye Padgett, and all of my fellow participants in the Winter ’08 UCSC Research Seminar. All errors are of course my own. 1 A few orthographic notes: ‘ch’ is used to represent IPA [Ù], ‘x’ to represent the palatal fricative [S], ‘b’ to represent the bilabial implosive [á], ‘y’ to represent the palatal glide [j], and an apostrophe to represent the glottal stop as well as phonemic creaky voice on vowels.

1

Strong Positions and Laryngeal Features in Yukatek Maya

2.1

ω-final aspiration

According to Blair & Vermont-Salas (1965), allophonic aspiration in YM occurs in all codas and nowhere else as seen in (2). (2)

a. b. c. d. e.

suuth ‘return’ (*suut, *suut^) s´ıinikh ‘ant’ (*s´ıinik, *s´ıinik^) atan ‘wife’ (*ath an) cheel ‘rainbow’ (*chh eel) ts´ıimin ‘horse’ (*tsh´ıimin)

Given the above data, however, there is an obvious alternative hypothesis: aspiration only occurs in word-final codas. • Distinguishing these two hypotheses is non-trivial since there are morphological restrictions on native lexical stems requiring them to be either CVC or CVCVC. • The crucial data cited by Blair & Vermont-Salas (1965) all comes from morphologically complex forms such as N-N compounds (CVC.CVC) which are likely prosodically complex as well. • The only place where we see morpheme-internal codas is in the output of an optional syncope rule seen in (3) where such stops are not aspirated. atan-il → [at.nil] (*[ath .nil]) ‘wifehood’

(3)

• Additional evidence that word-internal codas are not aspirated comes from the pronunciation of monomorphemic English names and nonce words containing word-internal codas (e.g. [putnam], *[puth nam]). • In our pilot study, such coda consonants were uniformly unaspirated, in contrast to word-final ones in the same environments (e.g. [path ] *[pat]). Conclusion: Stops in Yukatek Maya are aspirated in prosodic word-final position and unaspirated elsewhere.

2.2

φ-final [h]-epenthesis

In addition to aspiration, which appears to privilege a cross-linguistically weak position (the final segment of the phonological word), there is a similar process which operates at the level of the phonological phrase (φ). 2.2.1

Independent evidence for φs in YM

• Straight (1976) and Orie & Bricker (2000) describe a debuccalization process in YM where a consonant’s supralaryngeal features are deleted when it is adjacent to one with identical ones as seen in (4)-(6).2

2

See also McCarthy (1988) and Bessell (1992) for further analysis of these facts.

2

Strong Positions and Laryngeal Features in Yukatek Maya

(4)

Debuccalization deletes most consonants a. k’am makal → k’a makal ‘receive elephant ears’ b. p’is sakal → p’i sakal ‘measure cloth’

(5)

Ejectives surface as [P] a. saak’ kaal → saa’ kaal ‘itchy neck’ b. sop’ bak’ → so’ bak’ ‘sell meat’

(6)

Plain (allophonically aspirated) stops become [h] a. p´aaph p’aak → p´aah p’aak ‘spicy tomato’ b. xoth t’´ u’ul → xoh t’´ u’ul ‘cut rabbit’

Debuccalization Schematized: (7)

Given a string ...C1 ]ω [C2 ..., if C1 and C2 are identical except for laryngeal features, then delink all non-laryngeal features of C1 .

Straight (1976) and Orie & Bricker (2000) both characterize the process as applying whenever the linear environment in (7) is met. • This is true with respect to morphological boundaries and the prosodic word. • However, looking more closely at the examples provided by Orie & Bricker (2000), we find that their examples fall into a limited class of particular syntactic types as in (8): (8)

Debuccalization tokens by Syntactic Configuration:

Syntactic Configuration Compound Adj + Noun Verb + Derivational Suffix Verb + Bare Direct Object

Number of tokens 4 11 3 16

In our fieldwork examining different syntactic configurations in (9)-(11), however, we find that debuccalization does not always occur. • The object, puut, in the contrastive focus position in the left periphery: (9)

[puuth ]φ [t-u maan-ah]φ [Hwaan]φ . papaya PFV-A3 buy-CMP Juan. ‘It was papaya that Juan bought’

• A wh-word, m´ aax, in the left periphery: (10)

[M´aax]φ [xot le t’u’ul-o’]φ who cut DEF rabbit-D2 ‘Who cut the rabbit?’

3

Strong Positions and Laryngeal Features in Yukatek Maya

• Unlike the object, the subject, Carmen, is external to the VP. (11)

[puuth ]φ [k-u maan-ikh ]φ [karmen]φ . papaya IMP-A3 buy-CMP Carmen. ‘It was papaya that Carmen bought.’

• The difference between examples like (9)-(11) (no debuccalization) and debuccalizing forms is that the linear environment for debuccalization, (7), crosses a φ-boundary in the latter. • We propose a right edge-based φ-construction algorithm, as formulated in OT by Ito & Mester (in prep), and seen in (12) (though any Selkirk & Shen (1990)-style edge-based algorithm will work). (12)

LexP-To-φ: Align(LexP, Right, φ, Right) ‘Every lexical maximal projection is right-aligned with a phonological phrase.’

Ranking LexP-To-φ and other constraints enforcing parts of the Strict Layer Hypothesis over *Struc derives phonological phrase boundaries as in (13): (13)

(Focus)φ (Asp V DPobj )φ (DPsubj )φ

• This correctly predicts the inability of debuccalization to apply in (9) - (11) while predicting its occurrence in the syntactic structures where it applies. Conclusion: Phonological phrases, independently supported by debuccalization, are rightaligned with lexical maximal projections. 2.2.2

Epenthesis of [h] φ-finally

Having established phonological phrasing independently on the basis of debuccalization, the following surface-true generalization emerges: (14)

Phonological Phrases in Yukatek Maya always end in a consonant

• The synchronic source of this property is obscured by morpheme structure constraints forcing lexical stems to be CVC or CVCVC. • There are two places where we do see an active phonological alternation: epenthesis of [h]. First, vowel-final borrowings from Spanish epenthesize a coda [h] when φ-final as in (15). This behavior is in contrast with native words containing a final [h] (as in (16)) where the [h] is retained in all positions. (15)

a. b.

estudiante → YM Pestudiante[h] ‘student’ estudiante[P]il maaya t’´aan ‘student of Maya’ (*estudiante[h]il)

(16)

a. b.

huuh → huu[h] ‘iguana’ huu[h]il ha’ ‘water iguana’ (*huu[P]il)

4

Strong Positions and Laryngeal Features in Yukatek Maya

Second, while we do not see this alternation in native lexical words, it does seem to be present in the function word w´aa in (17)-(19). (17)

Juan-w´ aah]φ uk’ le sa’-o’ Juan-w´aa drink DEF atole-D2 ‘Did Juan drink the atole?’

(18)

Juan hant-ik le hanal-w´ aa’-o’]φ Juan eat-INC DEF food-w´aa-D2 ‘Did Juan eat the food ?’

(19)

tak in hant-ik w´ aa ba’ax]φ wish A1 eat-INC w´aa what ‘I want to eat something’

Conclusion: Like ω-final aspiration, φ-final [h]-epenthesis also preferences a particular marked form, [spread glottis], in a weak position.

3

Cross-Linguistic Context

ω-final aspiration and φ-final [h]-epenthesis are examples of a robust cross-linguistic pattern which we term ‘Final Laryngeal Strengthening’ (FLS).

3.1

Final Laryngeal Strengthening

Aspiration • Sierra Popoluca has essentially the same aspiration pattern as Yukatek Maya (Elson (1947)). • Hupa has three stop series (e.g. /t th t’/) with the plain unaspirated and aspirated stop neutralizing to the plain [t] in word-medial codas and [th ] in word-final codas. (Vaux & Samuels (2005) cite many other similar examples) • German word-final devoicing appears to be another instance of this pattern if, following Jessen & Ringen (2002), we take [s.g.] rather than [voicing] to be the relevant feature in this process. Glottalization • Tigre neutralizes ejectives and plain stops in word-medial coda positions, but maintains the contrast in all other positions including word-final codas. (Fallon (2002) p. 92) Epenthesis of [h]/[P] • Afar has a rule which epenthesizes [h] in stressed open syllables utterance-finally and sometimes φ-finally (Barnes (2006) p. 120-1). • Japanese has a process in certain quotational structures which epenthesizes a coda glottal stop at the end of the quoted passage (which also presumably is a prosodic boundary of some sort) if unembedded and geminates the following onset if a complementizer follows. (Junko Ito, p.c.) • Dagbaani epenthesizes [P] in certain word-final open syllables (Barnes (2006) p. 125). 5

Strong Positions and Laryngeal Features in Yukatek Maya

3.2

The limits of FLS

Given this broad pattern of Final Laryngeal Strengthening, it is instructive to see what kinds of formally similar processes do not seem to be attested. In particular, there seem to be two interesting restrictions on FLS processes. FLS only refers to [spread glottis] and [constricted glottis] • We do not see clear cases of φ-final voicing (though see the debate between Blevins (2006) and Kiparsky (2006) for discussion of potential ω-final cases). • Also, we do not see word or phrase final epenthesis of other segments such as [n], [r], [t], etc. FLS only targets prosodic units at least as large as ω • We have already argued in §2.1 that aspiration in Yukatek Maya is properly analyzed as being word-final rather than syllable-final. • The only other putative counterexample that I am aware of is Sierra Popoluca aspiration as described by Elson (1947). Here, too, however, many of the data points cited are clearly compounds (e.g. peth .kuy ‘broom’ from pet- ‘sweep’ and kuy ‘wood’) and other morphologically complex forms which could plausibly be parsed into multiple prosodic words. • This restriction stands in stark contrast to initial aspiration in English (foot-initial), Korean (syllable-initial), and other languages.

4

An OT Account of Final Laryngeal Strengthening

The natural way to capture this pattern is using Generalized Alignment (McCarthy & Prince (1993)) constraints aligning particular features ([s.g.] or [c.g.]) with the right edge of prosodic constituents (see Davis & Cho (2003) for such an approach to the initial aspiration in English and Korean). • To highlight the asymmetries between initial alignment and FLS, we introduce a family of constraints as in (20). (20)

Final Laryngeal Strengthening(α-laryngeal feature, β-prosodic unit): ‘Every β-final segment should have specification [α]’ where α ∈ {[s.g.], [c.g.]} and β ∈ {ω, φ, IP, u}.

• Ranking FLS([s.g], ω) over *[s.g.] derives the aspiration data in YM in (21).

(21)

+

/p´ aap/ p´aap p´aaph ph ´aap ph ´aaph

Dep[Root]

FLS([s.g], ω) *!

*[s.g.] * * **!

*!

6

Strong Positions and Laryngeal Features in Yukatek Maya

• FLS([s.g], φ) helps explain φ-final [h] epenthesis in YM as in the following tableau:

(22) +

/aalbahaaka/ ‘basil’ aalbahaaka]φ aalbahaakah]φ

FLS([s.g], φ) *!

Dep[Root] *

• This derives a surface pattern of consonant-finality but uses independently motivated constraints rather than McCarthy (1993)’s FinalC (see Ito & Mester (in prep) for critical discussion of FinalC). • Since the environment for debuccalization only occurs at ω-boundaries, FLS also helps understand why debuccalization preserves laryngeal features, as in (23), but does not epenthesize them in (24)3 .

(23) +

(24)

+

/p´ aaph p’aak/ [[p´aaph ]ω [p’aak]ω ]φ [[p´aap]ω [p’aak]ω ]φ [[p´aa]ω [p’aak]ω ]φ [[p´aah]ω [p’aak]ω ]φ

Dep[Root]

/k’am makal/ [[k’am]ω [makal]ω ]φ [[k’a]ω [makal]ω ]φ [[k’a’]ω [makal]ω ]φ [[k’ah]ω [makal]ω ]φ

Dep[Root]

FLS([s.g], ω)

OCP-Supralaryngeal *!

*[s.g.] *

*! *! * FLS([s.g], ω) * *

*! *!

OCP-Supralaryngeal *!

*[s.g.]

*

The asymmetry between the prosodic word level (where underlying [s.g] is retained) and the phonological phrase level (where [s.g.] is epenthesized) is derived by the subranking in (25): (25)

FLS([s.g], φ) >> Dep[Root] >> FLS([s.g], ω)

• FLS accords with Flack (2007)’s principle stating that constraints targeting the edges of prosodic domains are universally ranked according to the prosodic hierarchy (i.e. Conu >> ConIP >> Conφ >> Conω )

5

Phonetic Grounding for FLS

Having established that Final Laryngeal Strengthening is amenable to a relatively simple OT analysis, we are left with the deeper question of how to constrain the inventory of OT constraints in a principled way that predicts only the types of FLS processes we in fact see. • In particular, FLS does not target prosodic units smaller than the word (unlike initial positional asymmetries) and does not involve [+Voice]. 3

There is a derivational opacity problem here that is not relevant for present purposes. It can be accounted for under a stratal OT account with aspiration occurring at the lexical level and [h]-epenthesis and debuccalization occurring at the post-lexical level. Alternatively, an O → O correspondence approach would require correspondence in [s.g.] between the word p´ aaph in isolation and in a larger phrase (p´ aah p’aak ). Previous works on YM debuccalization have simply taken the aspirated forms to be underlying, a position that appears untenable given Richness of the Base.

7

Strong Positions and Laryngeal Features in Yukatek Maya

5.1

The role of articulatory knowledge in phonology

These restrictions on FLS can be captured in a uniform fashion given a phonology with access to some amount of articulatory phonetic information. • We propose that Final Laryngeal Strengthening is the phonologization of the articulatory effort required to maintain voicing throughout larger prosodic units. • This pressure is reflected in gradient processes of devoicing or laryngealization at the end of larger prosodic constituents (ω, φ, IP, u). • Barnes (2006) describes the case of phrase-final glottal stop epenthesis in Dagbaani as being a phonologization of this pressure. • The present proposal, then, can be seen as an extension of this logic to a wider array of phenomena. Since the articulatorily ‘costly’ forms to be avoided are consistently voiced large prosodic constituents, final voicing of any prosodic unit and aspiration/laryngealization of smaller prosodic units like the syllable or foot also do not help to reduce articulatory effort.

5.2

Other potential approaches

While a full discussion of the problems in accounting for FLS under other theories of positional privilege is beyond the scope of the present work, a few thoughts on potential alternative analyses are in order. Pure Prominence • Pure Prominence approaches (i.e. Positional Faithfulness (Beckman (1998)) and Positional Markedness (Zoll (1998))) are too restrictive to account for FLS. • On these approaches, final codas belong to the set of ‘weak positions’ (alternatively, they are not strong positions) and therefore cannot have a marked form that does not occur in strong positions. Evolutionary Phonology • Conversely, the Evolutionary Phonology approach to FLS (Blevins (2006)) appears to be too permissive. It wrongly predicts that every possible combination of laryngeal features and prosodic constituents ought to exist (even though some of them may be quite rare). • In particular, we ought to see syllable-final aspiration of the kind argued against for Yukatek Maya and Sierra Popoluca as well as phonological phrase-final voicing.

6

Conclusion • Focusing on aspiration and φ-final epenthesis in Yukatek Maya, we have examined a systematic exception to strong positions: Final Laryngeal Strengthening. 8

Strong Positions and Laryngeal Features in Yukatek Maya

• We have seen that this phenomenon is both wider and narrower in certain respects than previously assumed. It is wider in that it involves [c.g.] in addition to [s.g.] (and possibly also the retention of laryngeal features in debuccalization processes). It is narrower than previously assumed since it only involves only prosodic units at least as large as the prosodic word. • To capture this entire array of processes in a principled way, we have argued that phonology must have access to articulatory phonetic information (the effort required to maintain voicing in large prosodic units).

References Barnes, Jonathan (2006) Strength and Weakness at the Interface. Mouton de Gruyter. Beckman, Jill (1998) Positional Faithfulness. Ph.D. thesis, University of Massachusetts. Bessell, Nicola (1992) The typological status of /?,h/. In Proceedings of the Chicago Linguistic Society, 56–71. Blair, Robert W. & Refugio Vermont-Salas (1965) Spoken (Yucatec) Maya, ms. Chicago: Department of Anthropology, University of Chicago. Blevins, Juliette (2006) A theoretical synopsis of Evolutionary Phonology. Theoretical Linguistics : 117–166. Davis, Stuart & Mi-Hui Cho (2003) The distribution of aspirated stops and /h/ in American English and Korean: an alignment approach with typological implications. Linguistics 41: 607–652. Elson, Ben (1947) Sierra Popoluca Syllable Structure. International Journal of American Linguistics 13: 13–17. Fallon, Paul (2002) The Synchronic and Diachronic Phonology of Ejectives. Routledge. Flack, Kathryn (2007) The Sources of Phonological Markedness. Ph.D. thesis, University of Massachusetts. Ito, Junko & Armin Mester (in prep) The Extended Prosodic Word, unpublished manuscript, University of California Santa Cruz. Jessen, Michael & Catherine Ringen (2002) Laryngeal features in German. Phonology 19: 189–218. Kiparsky, Paul (2006) The Amphichronic Program vs. Evolutionary Phonology. Theoretical Linguistics : 217–236. McCarthy, John (1988) Feature Geometry and Dependency: A Review. Phonetica 43: 84–108. McCarthy, John J. (1993) A case of surface constraint violation. Canadian Journal of Linguistics . McCarthy, John J. & Alan Prince (1993) Generalized Alignment. Yearbook of Morphology : 79–154, ROA 7-0000. Orie, Olanike Ola & Victoria Bricker (2000) Placeless and Historical Laryngeals in Yucatec Maya. International Journal of American Linguistics 66: 283–317. 9

Strong Positions and Laryngeal Features in Yukatek Maya

Selkirk, Elisabeth & Tong Shen (1990) Prosodic domains in Shanghai Chinese. In The PhonologySyntax Connection, CSLI, 313–337. Straight, H. Stephen (1976) The acquisition of Maya phonology: Variation in Yucatec child language. Garland. Vaux, Bert & Bridget Samuels (2005) Laryngeal markedness and aspiration. Phonology 22: 395– 436. Zoll, Cheryl (1998) Positional asymmetries and licensing, ms. MIT [Available on Rutgers Optimality Archive, roa.rutgers.edu no.282].

10