Focusing on a cross-linguistically unusual contrast inequity between stems and affixes, this paper argues for an approach to morphological patterning based on psycholinguistically-relevant factors. These factors include neighborhood density and token frequency, which play an important role in lexical retrieval. The paper first shows that the proposed root-affix faithfulness metaconstraint of McCarthy and Prince (1995) is factually incorrect, given the existence of patterns in which affixes control the phonology of stems. The paper then turns in detail to the case of Modern Hebrew, and shows that despite the unusual features normally associated with nonconcatenative morphology the morphophonological patterning observed there is readily explained once factors relating to lexical retrieval are taken into consideration.