In production, word-initial voicing contrast of plosives in Tokyo Japanese is not robustly based on VOT, since young speakers tend to devoice previously voiced plosives. Meanwhile, speakers rely heavily on f0. The present study aims to examine the role of VOT and f0 cues in perception. We conducted an identification test using resynthesized stimuli along a VOT continuum (-60 to +40 ms) orthogonal to an f0 continuum. Results suggested a categorical perception of VOT, while f0 was especially useful when VOT was around 0 ms. Higher f0 contours affected the response rate more than lower f0 contours, suggesting that the perceptual role of f0 raising was more important than f0 lowering. Hence, plosive devoicing in production is not preceded by listeners' misparsing of low f0 and prevoicing. Their misattribution of high f0 to voicelessness may play a more important role in the shift of VOT-f0 cue weighting in perception and production.