The phonological processes known as "lenition" have traditionally been explained as articulatory effort reduction. However, such a motivation for lenition has never been directly demonstrated; in addition, there are reasons to doubt the articulatory explanation.This paper focuses on a particular type of lenition (intervocalic spirantization of voiced stops) and presents two experiments that investigate what role, if any, perceptual considerations might play in lenition. Experiment 1 shows that spirantization of intervocalic voiced stops is a less perceptually salient change than devoicing of intervocalic voiced stops (an unattested process). Using the line of reasoning of Steriade's P-map hypothesis, perceptual facts offer an alternative to the articulatory account: lenition of intervocalic voiced stops yields spirants rather than voiceless stops because the latter change is perceptually highly salient. The results of Experiment 2 show that the perceptual facts differ by place of articulation, such that the difference between stops and spirants is greater for labials than for dorsals. These results do "not" match the attested typology; if anything, languages are more likely to spirantize labials than they are to spirantize dorsals. Thus, perceptual facts have the potential to explain some, but not all, of the typology of lenition. (Contains 10 figures, 2 tables and 9 notes.)