Discrimination of vowel duration was explored with regard to JNDs, error bias, and effects of modality and consonant context. 90 normal-hearing participants discriminated either auditorily, visually, or audiovisually between pairs of stimuli differing with regard to duration of the vowel /a/. Duration differences varied in 24 steps: 12 with the first token longer and 12 with the second token longer (33-400 ms). Results: accuracy was lower for V than A and AV; step difference affected performance in all modalities; error bias was affected by modality and consonant context; and JNDs (> 50% correct) were not possible to establish.