Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 1 Nick Danis 2 Rutgers University 1 Introduction (1) Doubly-articulated stops: IPA: k p, g b 3 "have two simultaneous articulations of the same degree of stricture" (Ladefoged and Maddieson 1996: 328) Have two unordered phonological places: a complex segment (Chomsky and Halle 1968; Sagey 1986; Clements and Hume 1995, a.o.) (2) Today's talk: a. Simplification of Representation: Segments are represented by a set of place features with equal status, i.e. primary vs. secondary articulation is not represented b. Simplification Processes: Universal markedness in OT (Prince and Smolensky 1993/2004; Lombardi 2001; de Lacy 2006) controls complex stop alternations i. e.g. /kp/ [p], but never /kp/ [k] (Amele, Efik) c. Constraints refer to features, not to general segment type (e.g. no "complex" or "doubly-articulated" in constraint definitions) (3) Example Representations a. Voiceless labial-velar: [kp] = { [labial], [dorsal], [+back], [ voice] } b. Voiced labial-palatal: [ɟb] = { [labial], [dorsal], [ back], [+voice] } (4) Place Markedness Violation Profiles m:kpt 4 m:kp m:k t, d 1 0 0 p, b 1 1 0 k, g 1 1 1 Dorsals are most marked, followed by labials then c, ɟ 1 1 1 coronals. kp, gb 2 2 1 Segments with two place specifications (complex cp, ɟb 2 2 1 segments) receive up to two markedness violations. 1 Thanks to Alan Prince, Akin Akinlabi, Bruce Tesar, Paul de Lacy and others including Mike Cahill, Will Bennett, and several audiences at Rutgers where past versions of this project were presented. 2 nick.danis@rutgers.edu 3 The tie bar will be omitted and assumed throughout the rest of this handout; [kp] always represents a single segment, never a cluster. 4 Abbreviations: K =dorsal place, P = labial place, T = coronal place (in constraint definitions), m: = markedness constraint, f: = faithfulness constraint, [ɟ] = voiced palatal stop, [c] = voiceless palatal stop
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 2 2 Representations/GEN S (5) Assumption: labial-dorsals are the only type of complex stop (Bennett 2014) a. labial-coronals are actually labial-palatals and phonologically dorsal b. the velar/posterior closure in clicks is an implementation of their lingual airstream mechanism (see also Halle 1995) (6) Place features: 5 [labial], [dorsal], [coronal] (7) [±back] marked only for dorsals (Chomsky and Halle 1968; c.f. Hall 1997) a. [dorsal +back] = velar place b. [dorsal -back] = palatal place (8) Relevant featural representation of segments under consideration: Place [±back] [±voice] t coronal d coronal + p labial b labial + k dorsal + g dorsal + + c dorsal ɟ dorsal + kp labial, dorsal + gb labial, dorsal + + cp labial, dorsal ɟb labial, dorsal + 3 Constraints/CON S 3.1 Markedness (9) Markedness constraints are based on a universal hierarchy (Prince and Smolensky 1993/2004; Lombardi 2001; de Lacy 2006): [dorsal] > [labial] > [coronal] where > signifies more marked than (10) Constraints are defined stringently rather than in a fixed universal ranking (de Lacy 2006): all three constraints assigns a violation to dorsals, only two assign a violation to labials, and only one assigns a violation to coronals. 5 Glottal place is ignored here for simplicity's sake. Its inclusion does not change any of the crucial conclusions, as long as glottal place is represented featurally (opposed to lack of all place features), as it is in Lombardi (2001); de Lacy (2006).
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 3 (11) m:kpt "Don't be dorsal, labial, or coronal" Assign one violation: for every segment S where [dor] S and for every segment S where [lab] S and for every segment S where [cor] S (12) m:kp "Don't be dorsal or labial" Assign one violation: for every segment S where [dor] S and for every segment S where [lab] S (13) m:k "Don't be dorsal" Assign one violation: for every segment S where [dor] S (14) Additionally, a markedness constraint assigns violations to palatals: (15) m:[-back] "Don't be [-back]" Assign one violation for every segment S where S is [-back] (16) Place Markedness Violation Profiles m:kpt m:kp m:k m:[-back] t, d 1 0 0 0 p, b 1 1 0 0 k, g 1 1 1 0 c, ɟ 1 1 1 1 kp, gb 2 2 1 0 cp, ɟb 2 2 1 1 (17) Voiced stops are marked based on the following constraint. (18) m:kpt/[voice] "Don't be a voiced dorsal, labial, or coronal" Assign one violation for every [+voice] segment S where [dor] S for every [+voice] segment S where [lab] S for every [+voice] segment S where [cor] S
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 4 (19) This assigns violations for each [+voice]/place feature pair in a segment. (20) Violation Profiles: m:kpt/[voi] and *[+voice] m:kpt/[voi] *[+voice] p, t, k 0 0 kp, cp 0 0 b, d, g 1 1 gb, ɟb 2 1 (21) This is crucially different from a simple *[+voice] constraint, as the bordered cells indicate. (22) This difference is relevant in the typology of languages without voiced complex stops but voicing distinctions elsewhere. 3.2 Faithfulness (23) Place is subject to a single place faithfulness constraint. (24) f:place "Don't add or remove place features" If S1 is a segment in the input, and S2 is a segment in the output, and S1 and S2 are in correspondence, and Assign one violation for every x such that: x S1 & x S2 and Assign one violation for every x such that: x S2 & x S1 where x {[lab], [cor], [dor]} (25) Violation Profile for f:place I O f:place Comment kp t 3 [dor] and [lab] not in output, [cor] not in input kp p 1 [dor] not in output kp k 1 [lab] not in output kp c 1 [lab] not in output kp kp 0 No change in major place kp cp 0 No change in major place
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 5 (26) Ident (McCarthy and Prince 1995) constraints for features [back] and [voice]: (27) f:[back] "Don't change the value for [back]" If S1 is a segment in the input, and S2 is a segment in the output, and S1 and S2 are in correspondence Assign one violation if S1 is [αback] and S2 is [ αback] (28) f:[voice] "Don't change the value for [voice]" If S1 is a segment in the input, and S2 is a segment in the output, and S1 and S2 are in correspondence Assign one violation if S1 is [αvoice] and S2 is [ αvoice] (29) Violation Profile for f:[back] I O f:[back] Comment C C 0 Faithful C K 1 /C/ is [-back], [k] is [+back] C T 0 [T] is undefined for [±back] 4 Typology and Rankings (30) Subranking for Languages with Complex Segments: f:place m:kpt m:kp m:k Faithfulness dominates all place markedness, so all segments are faithful w.r.t. place, including complex segments (31) Subranking for Languages without Complex Segments: m:kp m:kpt f:place m:k m:kpt only differentiates between complex stops and simple stops (2 violations vs. 1), meaning complex stops are banned but all simple stops are faithful under this ranking m:kpt t, d 1 p, b 1 k, g 1 c, ɟ 1 kp, gb 2 cp, ɟb 2
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 6 4.1 Predicted Surface Inventories (32) Predicted Surface Inventories of Languages with Complex Stops Surface Inventory t p k c kp cp d b g ɟ gb ɟb Attested Languages kp only, voiceless t p k kp Bora, Lama, Lokpa (Cahill 2009) kp and cp, voiceless t p k c kp cp kp only, voiced simple stops t p k kp d b g Late, Sekpele, Krachi (WAPS) kp and cp, voiced simple stops t p k c kp cp d b g ɟ kp and gb t p k kp d b g gb Soso, Mende, Dagbani, Bantonu, Ga, Ewe, Avatime (Siya), Siwu (Lolobi), Ge, Bini, Ora, Ijaw (Upper), Kambari, Tiv, Krio (WAPS) fully faithful t p k c kp cp d b g ɟ gb ɟb Dagbani (Olawsky 1999) (33) Compared against languages surveyed in Ladefoged (1968); Cahill (2009) 4.2 Reduction of Complex Stops (34) In the following languages, [KP] surfaces as [p] in certain contexts: a. Amele: /gb/ [p] in codas, paradigmatic alternation (Roberts 1987) b. Efik: [kp] in onsets, [p] in codas, phonotactic alternation (Welmers 1973) (35) In fact, the distribution of labial-dorsals is almost exclusively limited to certain initial positions cross-linguistically as discussed in detail in Cahill (2000) (36) In present typology, not only is [p] less marked than [k], but /kp/ [k] is harmonically bounded: (37) Violation profiles of /kp/ [p] vs. /kp/ [k] m:kpt m:kp m:k m:]-back] f:place f:[back] m:kpt/[voi] f:[voi] Input Output kp t 1 0 0 0 3 0 0 0 p 1 1 0 0 1 0 0 0 k 1 1 1 0 1 0 0 0 kp 2 2 1 0 0 0 0 0 (38) As dorsals are more marked than labials, kp p will always receive a subset of the violations of kp k, meaning kp k is never a possible optimum in any ranking. (39) The cases of Amele and Efik can be modeled without primary vs. secondary place marked in the representation (see also Cahill 2000: 78).
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 7 5 Labial-Coronals (40) Far fewer languages are reported to have labial-coronals, and the cases are more contentious. (41) While Ladefoged (1968) cites four languages as having labial-coronals, counterarguments have been made since then (as summarized in Bennett 2014): (42) Margi, Bura: a. Maddieson (1983, 1987) argues labial-coronals are sequences of labial and coronal stops, not a single labial-coronal segment, based on: i. non-simultaneous closures ii. longer overall duration than simple segments iii. shortening of a preceding vowel due to resyllabification b. Sagey (1986) still considers these segments phonologically complex (43) Dagbani, Nzema: a. Labial-velars become labial-palatals before front vowels (Ladefoged 1968; Wilson and Bendor-Samuel 1969; Cahill 1999; Olawsky 1999; Cahill 2008) b. Because this contact is palatal, its [coronal]-ness is up to debate; it is assumed here to be [dorsal] but could very well be [coronal] (44) However, cases of labial-coronals in Yeli (Henderson 1995) are most definitely not dorsal 6 : a. [t a:] 'bushknife' b. [ka:] 'fan palm' c. [k pa:] 'stealing' d. [t pa:] 'turn over' (Henderson 1995: 8) (45) The place of articulation in a. and d. is dental. There is also a post-alveolar labialcoronal. (46) New Assumption: Labial-dorsals and labial-coronals are both licit phonological segments: they are admitted into GEN S (47) New Prediction: /tp/ reduces to [t] and not [p], as [t] is less marked than [p] (48) Margi: labial-coronals reduce to coronals for some speakers in fast speech (Sagey 1986: 266, citing Hoffmann): a. /bdəәli/ [dəәli] 'Dille' b. /ptəәl/ [təәl] 'chief' (49) Sagey (1986) cites the above alternation as evidence that [coronal] is the major articulator/primary place in Margi, but here this falls out from the typology 6 Thanks to Mike Cahill for stressing the importance of these.
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 8 6 Other Typological Results 6.1 Voicing of Complex Segments (50) The voicing markedness constraint m:kpt/[voi] as defined assigns more violations to voiced complex stops than it does to voiced simple stops. (51) In languages that show a voicing distinction in simple stops but lack [gb], voicing faithfulness dominates voicing markedness, which dominates place faithfulness. f:[voi] m:kpt/[voi] f:kpt (52) Voiced outputs with one place feature incur fewer violations than complex segments, so in these languages, /gb/ [b] (53) Languages whose surface inventories match those produced in the typology: Late, Sekpele, Krachi (Ladefoged 1968) 6.2 Implications of Complex Stops (54) The typology makes many implicational predictions, two of which are below: (55) Attested prediction: the presence of CP implies the presence of KP Languages: Dagbani, Nzema (Ladefoged 1968; Olawsky 1999) (56) Not captured: Plenty of empirical languages have [gb] as their only complex stop (Cahill 2009); here, markedness predicts that the presence of [gb] implies [kp] 7 Conclusion (57) The reduction of complex stops to simple stops can be explained with independentlymotivated universal markedness hierarchies and no longer requires primary versus secondary articulation in segment representations, language-specific ranking of place faithfulness constraints, or other similar devices. (58) Towards a complete phonology of complex segments, more work is needed: a. Fully including labial-coronals in the typology and capturing any phonological processes b. Investigating further the case of gb-only languages and the implications between complex stops and gapped simple stop inventories (Cahill 2009)
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 9 8 Appendix 8.1 Full Factorial Typology See full discussion of typology in Danis (2014a, 2014b) (59) KP0 Factorial Typology Inputs- > t p k c kp cp d b g j gb jb Lg. 1 t.cor t t t t t t t t t t t t Lg. 2 d.cor t t t t t t d d d d d d Lg. 3 t.lab t p t t p p t p t t p p Lg. 4 d.lab t p t t p p d b d d b b Lg. 5 t.dor.s.maj t p k t p p t p k t p p Lg. 6 d.dor.s.maj t p k t p p d b g d b b Lg. 7 t.dor.c.maj t p k t kp p t p k t kp p Lg. 8 d.dor.c.maj.b t p k t kp p d b g d b b Lg. 9 d.dor.c.maj.gb t p k t kp p d b g d gb b Lg. 10 t.dor.s.f t p k c p p t p k c p p Lg. 11 d.dor.s.f t p k c p p d b g j b b Lg. 12 t.dor.c.f t p k c kp cp t p k c kp cp Lg. 13 d.dor.c.f.b t p k c kp cp d b g j b b Lg. 14 d.dor.c.f.gb t p k c kp cp d b g j gb jb Lg. 15 t.dor.s.min t p k k p p t p k k p p Lg. 16 d.dor.s.min t p k k p p d b g g b b Lg. 17 t.dor.c.min t p k k kp kp t p k k kp kp Lg. 18 d.dor.c.min.b t p k k kp kp d b g g b b Lg. 19 d.dor.c.min.gb t p k k kp kp d b g g gb gb (60) Property Notation for Language Names Voicing Voiceless t Voiced d Most Marked Major Place Coronal cor Labial lab Dorsal dor Complexity Simple s Complex c Palatal Realization Faithful f Unfaithful Major Place maj Unfaithful Minor Place min Voiced Complex Segments Reduced Place b Preserved Place gb
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 10 (61) Property Analysis System KP0 Voicing Most Marked Place Voiced f:[voi] >> m:*jgbd /d/ [d] Voiceless m:*jgbd >> f:[voi] /d/ [t] LAB or DOR f:kpt >> m:*ckp COR m:*ckp >> f:kpt /p/ [t] DOR f:kpt >> m:*ck /k/ [k] LAB m:*ck >> f:kpt /p/ [p] Complexity Complex f:kpt >> m:*ckpt /kp/ [kp] Simple m:*ckpt >> f:kpt /kp/ [p] Complex Voicing Palatal Realization Preserved Place f:kpt >> m:*jgbd /gb/ [gb] Reduced Place m:*jgbd >> f:kpt /gb/ [b] Faithful f:[back] & f:kpt >> m:*c /c/ [c] Minor Place Unfaithful f:kpt & m:*c >> f:[back] /c/ [k] Major Place Unfaithful f:[back] & m:*c >> f:kpt /c/ [t] References Bennett, Wm. G. 2014. Some differences between clicks and labio-velars. South African Journal of African Languages, 34(2), 115-126. Cahill, Michael. 1999. Aspects of the phonology of labial-velar stops. Studies in African Linguistics, 28(2), 155-184. Cahill, Michael. 2000. Positional Contrast and Labial-Velars. OSU Working Papers in Linguistics. Cahill, Michael. 2008. The phonetics and phonology of labial-velars in Dagbani. Paper presented at the ACAL 2008, University of Georgia, Athens. Cahill, Michael. 2009. Why labial-velar stops merge to /gb/*. Phonology, 25, 379-398. Chomsky, Noam, & Morris Halle. 1968. The Sound Pattern of English. Cambridge, Massachusetts: The MIT Press. Clements, G. N., & Elizabeth V. Hume. 1995. The Internal Organization of Speech Sounds. In John Goldsmith, Jason Riggle & Alan C. L. Yu (Eds.), The Handbook of Phonological Theory: Blackwell Publishing.
Complex Segments Made Simple: Markedness and Doubly-Articulated Stops 11 Danis, Nick. 2014a. Deriving Interactions of Complex Stops. Ms., Rutgers University. ROA- 1220. Danis, Nick. 2014b. Deriving Interactions of Complex Stops: OTWorkplace Supplement. ROA 1221. de Lacy, Paul. 2006. Markedness: Reduction and Preservation in Phonology. Cambridge, UK: Cambridge University Press. Hall, T. Alan. 1997. The Phonology of Coronals. Amsterdam/Philadelphia: John Benjamins. Halle, Morris. 1995. Feature Geometry and Feature Spreading. Linguistic Inquiry, 26(1), 1-46. Henderson, James. 1995. Phonology and Grammar of Yele, Papua New Guinea. The Australian National University: Department of Linguistics, Research School of Pacific and Asian Studies. Ladefoged, Peter. 1968. A Phonetic Study of West African Languages: Cambridge University Press. Ladefoged, Peter, & Ian Maddieson. 1996. The Sounds of the World's Languages: Blackwell Publishing. Lombardi, Linda. 2001. Why Place and Voice are different: Constraint-specific alternations in Optimality Theory. Segmental phonology in Optimality Theory: Constraints and Representations. Maddieson, Ian. 1983. The Analysis of Complex Phonetic Elements in Bura and the Syllable. Studies in African Linguistics, 14(3), 285-310. Maddieson, Ian. 1987. The Margi Vowel System and Labiocoronals. Studies in African Linguistics, 18(3), 327-355. McCarthy, John, & Alan Prince. 1995. Faithfulness and Reduplicative Identity. Ms. Olawsky, Knut J. 1999. Aspects of Dagbani Grammar: Lincom Europa. Prince, Alan, & Paul Smolensky. 1993/2004. Optimality Theory: Constraint Interaction in Generative Grammar. Malden, MA, & Oxford: Blackwell Publishing. Roberts, John R. 1987. Amele. Sydney: Croom Helm. Sagey, Elizabeth. 1986. The Representation of Features and Relations in Non-Linear Phonology. (PhD), Massachusetts Institute of Technology. Welmers, Wm. E. 1973. African Language Structures: University of California Press. Wilson, W. A. A., & J. T. Bendor-Samuel. 1969. The Phonology of the Nominal in Dagbani. Linguistics, 52, 56-82.