Repetition Avoidance in Human Language

Repetition is avoided in countless human languages and at a variety of grammatical levels. In this dissertation I ask what it is that makes repetition so bad. I propose that at least three distinct biases against repetition exist. First, repetition of articulatory gestures is relatively difficult. This difficulty results in phonetic variation that may lead to categorical phonological avoidance. I call this set of claims the Biomechanical Repetition Avoidance Hypothesis (BRAH), and support it with evidence from cross-linguistic patterns in repetition avoidance phenomena, articulatory data from music performance, and a series of phonetic experiments that document the proposed types of phonetic variation. Based on these data, I give an evolutionary account for antigemination in particular. The second anti-repetition bias is a perceptual deficit causing speakers not to perceive one of a sequence of repeated items, of any conceptual category. This bias is already welldocumented, as are the grammatical effects (primarily haplology). I provide here the evidence of gradient variation in production bridging the two, from avoidance of homophone sequences in English corpora. The third factor is a principle disallowing the repetition of syntactic features in certain configurations within a phase domain. I document categorical effects of it in Semitic syntax of possession and relativization. These elicit repair strategies superficially similar to those of phonology (specifically, deletion and epenthesis/insertion). Repetition effects, then, are traceable to a variety of independent, functional biases. This argues against a unitary, innate constraint against repetition. Rather, multiple anti-repetition biases result in particular avoidance patterns, with their intersection producing additional asymmetries. Possible categorical repairs are further constrained by the nature of the formal grammatical system. Thesis Supervisor: Donca Steriade Title: Professor of Linguistics