© 2004 Gunnar Ólafur Hansson. WCCFL 23 Proceedings, ed. B. Schmeiser, V. Chand, A. Kelleher and A. Rodriguez, pp. 318-331. Somerville, MA: Cascadilla Press. Tone and Voicing Agreement in Yabem* Gunnar Ólafur Hansson University of British Columbia 1. Introduction Yabem, an Oceanic language of Papua New Guinea (Dempwolff 1939, Bradshaw 1979, Poser 1981, Ross 1988, 1993, 1995; see also Bradshaw 1999, Hansson 2001), displays a sound pattern which superficially looks like voicing agreement (voicing harmony), i.e. a cooccurrence restriction on voiced and voiceless obstruents. In the last two syllables of a (non- compound) word, a domain best interpreted as a right-aligned iambic foot (Ross 1993), stops are either all voiced or all voiceless: (1) Apparent voicing agreement within roots: a. -kato ‘make a heap’ b. -a ba ‘untie’ teka ‘his bone’ a de ‘his cousin’ típ ‘all at once’ dìb ‘thud’ (2) Prefix alternations apparently due to voicing agreement with root: a. k -p ‘shove (2sg realis)’ b. o -de ‘move to (2sg realis)’ ta-p ‘shove (1pl.incl.)’ da-de ‘id. (1pl.incl. realis)’ At first glance, this agreement even appears to be sensitive to the relative similarity of potential trigger-target pairs, a highly salient property of consonantal agreement patterns (Walker 2000, Hansson 2001, Rose & Walker to appear), in that stops only interact with other stops, not with the (sole) fricative /s/: (3) No voicing agreement between stops and /s/: a. “Disagreeing” roots b. Prefix-root “disagreement” saì ‘house partition’ da-su ‘push (1pl.incl. realis)’ saba ‘potsherd; spleen’ s -b ‘incise (3pl realis)’ * I am grateful to WCCFL 23 audience members for helpful comments on the work presented here, in particular Jill Beckman, Maria Gouskova, José Hualde and Rachel Walker. All errors and infelicities are my own. Except where explicitly noted otherwise, the Yabem data cited in this paper are all taken from Dempwolff (1939) and Ross (1993, 1995).