1

Nasalization as a Repair for Voiced Obstruent Codas in Noon

John Merrill

U.C. Berkeley jtmm@berkeley.edu

2

Nasalization as a Repair for Voiced Obstruent Codas in Noon In the Noon language (Cangin: Senegal), voiced stops have nasal allophones in coda position

/ɓed/ [ɓen] ‘accompany’ /ɓed-ee/ [ɓedee] ‘accompanied’ /gog/ [goŋ] ‘snake’ /gog-ii/ [gogii] ‘the snake’

This allophonic pattern can be seen as a repair for voiced obstruent codas, which are cross-linguistically marked Why is this pattern noteworthy? Why is it so rare in the world’s languages?

3

Coda devoicing and the Too Many Solutions problem Voiced obstruent codas are cross-linguistically marked Avoided by devoicing (see Myers 2012 for an overview) /tab/ → [tap]

But why devoicing? Other potential repairs: Nasalization /tab/ → [tam] Gliding /tab/ → [taw] Deletion /tab/ → [ta] Vowel epenthesis /tab/ → [tabǝ] ... In OT, a simple re-ranking of faithfulness constraints would be enough to get these other repairs Yet it is often noted that the only attested repair is devoicing (Steriade 2008, Blumenfeld 2006, Hermans and Ostendorp 2007) The Too Many Solutions problem

4

The P-map Steriade (2008) proposes the P-map to solve the Too Many Solutions problem The P-map is a principled way of ranking the constraints of a language such that repairs change the UR in the most perceptually minimal way “The aim, in any departure from the UR, is to change it minimally to

achieve compliance with the phonotactics.” (154) As the voicing contrast is less perceptible in coda position than a nasality contrast, IDENT[nas]/CODA will outrank IDENT[voi]/CODA For a UR /tab/, [tap] is perceptually closer to [tab] than any other conceivable repair

5

The P-map and coda nasalization Because the perceptual distance between coda [p] and [b] is less than that between coda [m] and [b], devoicing is always preferred Steriade (2008: 153) on nasalization: “...one does not encounter sound systems in which all the final voiced stops,

and only they, turn to nasals...” Describes what this “unattested” system would look like:

before vowel: tib-a tud-a tag-a top-a tat-a tek-a word final: tim tun taŋ top tat tek

6

The P-map and coda nasalization Because the perceptual distance between coda [p] and [b] is less than that between coda [m] and [b], devoicing is always preferred Steriade (2008: 153) on nasalization: “...one does not encounter sound systems in which all the final voiced stops,

and only they, turn to nasals...” Describes what this “unattested” system would look like:

before vowel: tib-a tud-a tag-a top-a tat-a tek-a word final: tim tun taŋ top tat tek

Noon:

imperative: [yab-a] [bíd-a] [lag-a] [tap-a] [hot-a] [ɓak-a] bare verb: [yam] [bín] [laŋ] [tap] [hot] [ɓak]

‘land’ ‘write’ ‘close’ ‘pound’ ‘see’ ‘set aside’

7

Senegal

http://0.tqn.com/d/goafrica/1/S/_/2/senegal.gif

8

The Cangin languages

Soukka 2000: 16

9

The Noon language

Soukka 2000: 17

Dialects: Thiès (Lopis 1981) Padee (Soukka 2000) Saawii (Fieldwork with Christine Diop) Northern Noon

10

Consonant inventory of Northern Noon

All voiced obstruents in Noon are stops Tautomorphemic prenasalized stops are rare, found only in borrowings

Note: Vowels are irrelevant to this discussion. <ë, é, í, ó, ú> represent [+ATR] vowels in the Cangin languages

labial coronal palatal velar glottal vl. stop p t c k ʔ vd. stop b d j [ɟ] g nasal m n ñ [ɲ] ŋ prenas. stop mb nd nj [ɲɟ] ng implosive stop ɓ ɗ ƴ vl. continuant f s h vd. continuant w l (r) y

11

Coda nasalization in Northern Noon The voiced stops /b, d, j, g/ surface as [m, n, ñ, ŋ] in coda position bare verb punctual -Ca past -ee kúb [kúm] kúb-ba [kúmba] kúb-ee [kúbee] ‘macerate’ kod [kon] kod-da [konda] kod-ee [kodee] ‘cry’ paj [pañ] paj-ja [pañja] paj-ee [pajee] ‘marry’ awaag [awaaŋ] awaag-ga [awaaŋga] awaag-ee [awaagee] ‘open mouth’ bare noun pl. proximal def. -c-ii proximal definite -ii yíib [yíim] yíib-c-ii [yíimcii] yíib-ii [yíibii] ‘chin’ nad [nan] nad-c-ii [nancii] nad-ii [nadii] ‘spider’ bíig [bíiŋ bíig-c-ii [bíiŋcii] bíig-ii [bíigii] ‘back’ These alternations are purely allophonic: exceptionless, completely

predictable from phonological environment Coda /j/ is rare (only 1 known example), but /b, d, g/ are common Knowing that voiced obstruent codas are marked cross-linguistically, this

allophonic pattern can be seen as a repair for this marked structure

12

Voiced stop-final roots contrast with nasal-final roots, which show no allophony bare verb perfect -in bare verb perfect -in ab [am] ab-in [abin] ‘hold’ ñam [ñam] ñam-in [ñamin] ‘eat’ tab [tam] tab-in [tabin] ‘be forbidden’ tam [tam] tam-in [tamin] ‘be hot’ mad [man] mad-in [madin] ‘resemble’ an [an] an-in [anin] ‘drink’ dag [daŋ] dag-in [dagin] ‘be taut’ daŋ [daŋ] daŋ-in [daŋin] ‘be viscous’ In coda position, the distinction between /b, d, j, g/ and /m, n, ñ, ŋ/ is

completely neutralized Thus, the alternation must be seen as nasalization of underlying stops, not

denasalization of underlying nasals Voiceless stop codas are common There is no general dispreference for stop codas— only voiced stop codas There is no reason why coda devoicing would not be an available repair

bare verb perfect -in ap ap-in ‘kill’ ɓít ɓít-in ‘be heavy’ oc oc-in ‘scratch’ ɗaak ɗaak-in ‘hide’

13

Other evidence for a constraint against voiced obstruent codas Noon allows no voiced obstruent codas, and exhibits several different repairs depending on the obstruent in question Implosive stops /ɓ, ɗ, ƴ/ have allophones [wʔ, ʔ, yʔ] in coda position

/ɗoɓ/ [ɗowʔ] ‘bite’ /ɓúɗ/ [ɓúʔ] ‘collapse’ /aƴ/ [ayʔ] ‘lean’

Alternation triggered by vowel deletion:

/jégíɗ/ [jégíʔ] ‘teach’ /jégɗ-oh/ [jéŋɗoh] ‘teacher’

14

Other evidence for a constraint against voiced obstruent codas Voiced prenasalized stops (ND) are never found in coda position. Earlier ND-final borrowings show up as voiced stops

Wolof, Sereer bind ‘write’ → Noon bíd [bín] More recent borrowings make use of vowel epenthesis

Wolof dund ‘live’ → Noon dúndú Voiceless prenasalized stop codas are rare, but are borrowed unaltered,

despite not appearing in native vocabulary Wolof Ngeent ‘Neighborhood in Thiès’ → Noon Ngeent Coda nasalization is clearly part of a larger conspiracy to avoid voiced obstruent codas in Noon

15

Coda nasalization in other languages Other cases of nasalization of voiced obstruent codas exist (some collected in Flynn 2007) But few truly result in the “unattested” pattern described in Steriade (2008)

Cangin languages Essentially the same pattern found in Northern Noon is found in some of the

other Cangin languages

16

Coda nasalization in other languages Southeastern Tepehuán (Uto-Aztecan: Willett 1991: 17) /b, d, g/ have allophones [ʔm, ʔn, ʔŋ] in coda position

/kaib/ [kaiʔm] ‘it has ripened’ /kaib-aʔ/ [kaibaʔ] ‘it will ripen’ /duud/ [duuʔn] ‘it has rained’ /duud-uʔ/ [duuduʔ] ‘it will rain’ /gaag-gaʔ/ [gaaʔŋgaʔ] ‘he will look around for it’ /gaag-aʔ/ [gaagaʔ] ‘he will look for it’

17

Coda nasalization in other languages Alabama (Muskogean: Montler & Hardy 1988)

The only voiced obstruent, /b/, surfaces as [m] in coda position Pluractional verbs are formed by vowel subtraction:

ɬobafka ‘to have a hole’ bala:ka ‘to lie down (sg.)’ ɬomka ‘to have holes’ balka ‘to lie down (pl.)’

Gemination of /b/ results in [mb]

aba:li ‘high’ hajo:ki ‘deep’ ámba:li ‘(getting) higher’ hájjo:ki ‘(getting) deeper’

But as words are all vowel-final, this alternation occurs only word-medially

18

Coda nasalization in other languages Stoney (Siouan: Shaw 1980) /b, d/ surface as [m, n] before a word or compound boundary (%)

ʧhába owĩʧhané%jaʧ ‘he went hunting beaver’ ʧhám%òné%jáʧ ‘lie down (sg.)’ júda ‘to eat’ juhjún%maʧ ‘I’m going along eating’

But this does not apply to /ʤ, g/, or voiced fricatives (devoicing is

employed) And does not apply to all coda positions (never word-medially; here

devoicing is employed)

19

Coda nasalization in other languages Japanese (Kuroda 2006, Inkelas and Cho 1993) In native vocabulary, voiced obstruents become nasals in coda position

asob-ɯ → asobɯ ‘to play’ asob-te → asonde ‘playing’

But also affects sequences of a voiceless + voiced obstruent

tuk-dasɯ → tundasɯ ‘thrust out’

This can be seen as a repair for voiced geminates, rather than a response to

voiced obstruent codas

20

Coda nasalization in other languages Vimeu dialect of Picard (José and Auger 2004) Voiced stops surface as nasals in coda position

/rɛpɔd/ [rɛpɔ n] ‘to answer’ /rɛpɔd-y/ [rɛpɔ dy] ‘answered’ /gab/ [ga m] ‘leg’ /gab-e/ [ga be] ‘action of kicking one’s leg over the head of a young/short/little person’

But only after a nasalized vowel In fact consistent with the P-map

A similar pattern is found in Canadian French (Flynn 2007: 7)

21

Coda nasalization in other languages Blust (2005) cites two independent cases of historical coda nasalization in Austronesian Northern Batak (Karo and Dairi-Pakpak) Proto-Batak *b, *d, *g > m, n, ŋ / in coda position

PB Karo *abab > abam ‘fine burning ashes’ *alud > alun ‘to massage’ *deleg > deleŋ ‘mountain’

But voiced obstruent codas were rare in Proto-Batak Perhaps for this reason, there are apparently no synchronic D~N

alternations

22

Coda nasalization in other languages Berawan languages (Long Terawan, Batu Belah, Long Jegan) Proto-North Sarawak *b, *d > m, n / in coda position (no information on

coda *g)

PNS Long Terawan *ulu eleb > ulo lem ‘knee’ *kuyad > kuyan ‘grey macaque’ *sulud > sulon ‘comb’

These languages are not well documented, and no resulting alternations are

recorded But *b, *d > k, r / word-medially, so if alternations do exist, they would not

be of the relevant type

23

Coda nasalization in other languages Nasalization as a synchronic repair for voiced obstruent codas is indeed rare Only SE Tepehuán and Cangin languages (Noon, Lehar, Ndut) show it

robustly Why should it be so rare? The answer (for Cangin at least) lies in the historical origin of these

alternations This pattern did not arise from the historical nasalization of final stops, but

from the development of earlier prenasalized stops e.g. the [b ~ m] alternation was once *mb in all positions

24

The Cangin languages Cangin Western Eastern Noon-Lehar Ndut Palor Safen Noon Lehar Thiès Noon Northern Noon Sources: Ndut: Morgan (1996) Palor: D’Alton (1983) Safen: Mbodj (1983) Lehar: Dieye (2010)

25

Ndut, Palor, and Lehar Ndut and Palor show the same alternations as N. Noon

Ndut, Palor = N. Noon bare verb negative -ay som sob-ay ‘pound’ /soob/ man mad-ay ‘resemble’ /mad/ paŋ pag-ay ‘cook’ /pag/ ‘do’

However, Palor has developed non-alternating plain voiced stops from some earlier implosives, so coda nasalization is not allophonic

The facts in Lehar seem to be exactly the same as in Northern Noon

26

Noon Dialects Northern Noon [b, d, j g] = Thiès Noon [mb, nd, nj ng]

Northern Noon Thiès Noon dooɗ ndooɗ ‘stick’ baal mbaal ‘sheep’ dagal ndangal ‘scorpion’

Coda allophones are the same (nasals) in all dialects. So Thiès Noon shows a [mb ~ m] alternation for Northern Noon [b ~ m]

Northern Noon Thiès Noon [am ~ ab-a] [am ~ amb-a] ‘hold’ [man ~ mad-a] [man ~ mand-a] ‘resemble’ [paŋ ~ pag-a] [paŋ ~ pang-a] ‘do’

27

Safen In Safen, Northern Noon /b, d, j, g/ correspond to [mb, nd, nj, ng] in all positions. Safen N. Noon Onset: njol jol ‘cricket’ paangi péegí ‘grass’ pambi pabi ‘chicken’ ngul gúl ‘pierce’ Coda: kung kúg [kúŋ] ‘fold/bend’ amb ab [am] ‘hold’ pang pag [paŋ] ‘work/do’ rang lag [laŋ] ‘close’

28

Historical origin of coda nasalization These correspondences must be reconstructed as prenasalized stops External evidence: Proto-Cangin *ɗíng = Sereer ɗing ‘fence in,’ PC *mbaal

= Sereer mbaal ‘sheep’ Proto-Cangin Ndut, Palor Safen Lehar N. Noon Thiès Noon *hang haŋ ~ hag- ang aŋ ~ ag- aŋ ~ ag- aŋ ~ ang- ‘be wide’ *hamb ham ~ hab- amb am ~ ab- am ~ amb- ‘hold’ *pang paŋ ~ pag- pang paŋ ~ pag- paŋ ~ pang- ‘do/cook’ *yoond yoon ~ yood- yoon ~ yood- yoon ~ yood- yoon ~ yoond- ‘learn’ *ndangal dagal ndangal dagal dagal ndangal ‘scorpion’ *cangín cigin cangin cëgín cégín ‘worm’ *nd- d- (Ndut) nd- d- d- nd- ‘locative pref.’ *mboos boos (Ndut) mbuus boos mboos ‘well bucket’ Safen retains the original prenasalized stops in all positions ND > N in coda position in all other languages ND > D in onset in all but Safen and Thiès Noon

Crucially, Proto-Cangin did not have plain voiced stops

29

Historical origin of coda devoicing vs. coda nasalization Historical precursor to each alternation: Devoicing: voiced stops in coda position Nasalization: prenasalized stops in all positions, no plain voiced stops Sound changes required to yield each alternation: Devoicing: devoicing of coda obstruents Nasalization: ND > N in coda, ND > D in onset The precursor for coda devoicing is much more common Coda nasalization requires two sound changes rather than one Unconditioned ND > D is found elsewhere, e.g. Kayan (Austronesian)

dialects (Blust 2005: 259), but is not common The historical facts explain the rarity of the nasalization alternation There may be other pathways to coda nasalization, but it cannot arise from a

single common sound change

30

Conclusions If we accept that sound change is phonetically natural ‘Unnatural’ alternations cannot arise from a single sound change Broadly, the more perceptual distance in an alternation, the more sound

changes must have occurred The P-map is generally successful because it accounts for those alternations

which result from a single sound change But series of sound changes do sometimes result in unnatural alternations which have no synchronic phonetic motivation (Stausland Johnsen 2012) The test for the P-map: If such an unnatural alternation effectively repairs a marked structure, will a language tolerate it despite the availability of a more ‘natural’ repair? Noon suggests that it will

Any mechanism which limits the possible constraint rankings of a language must then allow for some degree of arbitrariness, and not rely entirely on perceived similarity effects

31

Thanks to Larry Hyman, Nico Baier, and Christine Diop!

32

References Blumenfeld, Lev. (2006) Constraints on phonological interactions. Stanford University: Doctoral thesis. Blust, Robert. (2005) “Must sound change be linguistically motivated?” Diachronica 22:2, (pp. 219-269). D’Alton, Paula. (1983) Esquisse phonologique et morphologique du Palor, langue Cangin (Sénégal). Université de la Sorbonne Nouvelle:

Doctoral thesis. Dieye, El Hadji. (2010) Description d'une langue Cangin du Sénégal: Le Laalaa. Dakar: Université Cheikh Anta Diop de Dakar. Flynn, Darin. (2007) “Too Many Solutions: Says Who?” Talk presented at Alberta Conference on Linguistics 2007. Hermans, Ben and Marc van Oostendorp. (2007) “Solving the Too-many-solutions-problem.” Talk presented at the Meertens Institute,

June 4, 2007. Inkelas, Sharon and Young-Mee Yu Cho. (1993) “Inalterability as Prespecification.” Language 69, No. 3, (pp. 529-574). José, Brian and Julie Auger. (2004) “(Final) nasalization as an alternative to (final) devoicing: the case of Vimeu Picard.” Indiana

Working Papers in Linguistics 4(3), (pp. 1–44). Kuroda, Sige-Yuki. (2006) “Feature Geometry and Phonetic Features: A Case Study in Voicing and Coda Nasalization in Japanese.”

Gengo Kenkyu 129, (pp. 91-134). Lopis, Jeanne. (1981) Phonologie et morphologie nominale du noon, parler de Nguente. Université de la Sorbonne Nouvelle: Doctoral

thesis. Mbodj, Cherif. (1983) Recherche sur la phonologie et la morphologie de la langue saafi: le parler de Boukhou, Sénégal. Université de

Nice: Doctoral thesis. Montler, Timothy and Heather Hardy. (1988) “Imperfective gemination in Alabama.” International Journal of American Linguistics

54, (pp. 399-415). Morgan, Daniel Ray. (1996) Overview of Grammatical Structures of Ndut. UT Arlington: Master’s thesis. Myers, Scott. (2012) “Final devoicing: Production and perception studies.” In T. Borowsky, S. Kawahara, T. Shinya, and M. Sugahara

(eds.) Prosody Matters: Essays in Honor of Elisabeth Selkirk, (pp. 151-179). London: Equinox Press, London. Shaw, Patricia A. (1980) Theoretical issues in Dakota phonology and morphology. Outstanding dissertations in linguistics. New York:

Garland Pub. Soukka, Maria. (2000) A Descriptive Grammar of Noon. Lincom Europa. Stausland Johnsen, Sverre. (2012) “A diachronic account of phonological unnaturalness.” Phonology 29, (pp. 505-531). Steriade, Donca. (2008) “The phonology of perceptibility effects.” In Hanson, K. and Inkelas, S. (eds.) The nature of the word, (pp.

151–179). Cambridge: MIT Press. Willett, Elizabeth. (1982) “Reduplication and Accent in Southeastern Tepehuan.” International Journal of American Linguistics 48,

(pp. 168-184).