This paper examines the relationship between tone and prosodic positions. I show that prosodic heads prefer higher tone over lower tone, while non-heads exhibit the opposite preference. These generalisations are expressed within Optimality Theory as a family of constraints in a fixed ranking. One set regulates the relation of tone to heads : *H\L *H\M *H\H ; the other deals with tone on nonheads : *N-H\H *N-H\M *N-H\L. These constraints are used to account for the stress system of Ayutla Mixtec : in this language, stress is attracted to a syllable based on its tonal content, but is also influenced by the posttonic syllable's tone. The implications of the theory for other tone-stress interactions-metrically influenced tone placement and neutralisation-are also examined. * I wish to thank John McCarthy, Lisa Selkirk, three anonymous reviewers and the associate editor, who offered many constructive suggestions for improvement. I am also grateful to John Kingston, Alan Prince, Keren Rice and Moira Yip for their comments on various manifestations of this work. For guidance related to the languages discussed, I thank Lee Bickmore, Barbara Hollenbach, Inga McKendry and Ken Pike. 1 In (1) and elsewhere the numbers of parentheses give the page and column in Pankratz & Pike (1967) from which the form is taken. 2 It is quite possible that the Tonal Prominence scale is a total order of all possible heights, which may number as many as six (Odden 1995 : 453ff). The examples I have collected only provide evidence for three tone-height distinctions in relation to stress (see § §3 and 6), so a conservative form of the hierarchy is used here. I am preceded in proposing a tonal hierarchy by Jiang-King (1996 : 99), who offers the hierarchy [jupper] [kraised]. I will show that more than a two-step tonal hierarchy is needed (see §3). 3 This proposal is analogous to Clements' (1997) for sonority-syllable constraints. Constraints against the least marked tone-head combinations would also have an undesirable typological effect. They would allow systems in which foot heads cannot bear tone while non-heads can : *H\H F(Tone) *N-H\H, where F(Tone) is some faithfulness constraint that preserves input tone. I know of no such system, while the opposite type of system-one where tone is only specified on heads-does exist (e.g. Yip 2001). 'phõ:mí phõ:'mí (33) Tone-driven stress in Tibetan a. b.