Many languages use haplology, suppletion, and the blocking of derivations to achieve avoidance of 'accidental' repetition of surface morphs. At the same time, many languages permit accidental repetition and even encourage 'deliberate' repetition through reduplication. Strong universal constraints against morph repetition therefore fail. This furthermore implies the inadequacy of accounts of morphological processes in terms of matching templates or schemas. We present a psycholinguistic processing model built on evidence from language acquisition, and drawing on activation theory, which affords a unification of the linguistic data while allowing for their variety.