Mester and ItÔ (1989) claim that patterning in the distribution of an autosegmental palatalization morpheme in Japanese mimetic forms constitutes strong evidence for the theoretical claim that among phonological features only those which are redundant are unspecified in underlying representations (Restricted Underspecification). We provide experimental evidence to show that the curious distribution of palatalization in Japanese mimetics, including nonoccurrence with r, on which the argument for underspecification crucially depends, is predictable from facts about the articulatory difficulty of repeated CVCV structures. We show that the distributional regularities involved, when correctly stated, apply not only to mimetics but to the entire vocabulary of Japanese, as well as to at least one language, Spanish, in which palatalization is not an autosegmental morpheme. Underspecification of features, restricted or otherwise, is not required to describe the distribution of palatalization in Japanese mimetics, and the M&I analysis based on underspecif ication is both inadequate and unmotivated.