This study develops the hypothesis that the child-specific phenomenon of positional velar fronting can be modeled as the product of phonologically encoded articulatory limitations unique to immature speakers. Children have difficulty executing discrete tongue movements, preferring to move the tongue and jaw as a single unit. This predisposes the child to produce undifferentiated linguopalatal contact, neutralizing the coronal-velar contrast. Adopting a phonetically sensitive model of phonology, I propose that children's difficulty with discrete tongue movement can be encoded in a violable constraint, Move.as-Unit. The positional nature of fronting reflects the fact that discrete lingual movement is penalized more heavily in the motorically challenging context of a larger gesture. This analysis is supported with a longitudinal study of one child (3 ; 9 to 4 ; 4) whose fronting was conditioned by both segmental and prosodic factors. Adopting Move.as-Unit in a Harmonic Grammar framework makes it possible to reframe disparate-seeming conditioning factors as a unified grammatical system.