Without a good model of L1 speech perception, one cannot determine whether effects observed in L2 perception are due to L1 transfer, or to an L2 learning strategy that applies irrespective of L1. Bohn's Desensitization Hypothesis predicts that Spanish learners of English will use duration to distinguish English /I/ and /i/ because exposure to Spanish does not sensitize them to the relevant spectral difference. I propose that Spanish listeners are in fact sensitized to the spectral difference, but may avoid using it to distinguish the vowels because it may cue some other contrast in Spanish. In earlier studies, I found that some Spanish listeners used spectral differences in the vowel to distinguish an English postvocalic stop voicing contrast, /t/-/d/. The present study examined native Spanish listeners' perception of Spanish word-final postvocalic /t/ and /d/ produced by native Spanish speakers. Listeners could distinguish whether speakers had produced /t/ or /d/ at levels significantly above chance. A discriminant-analysis implementation of Nearey's NAPP model, constructed using the acoustic properties of the speakers' productions, correlated well with the listeners' perception pattern. F2 towards the end of the vowel was the best predictor of consonant identity: low F2 was correlated with /d/ perception.