Drawing on insights from recent work on phonetic adaptation, we examined how listeners interpret prosodic cues to two opposing pragmatic meanings of the phrase “It looks like an X” (e.g., “It looks like a zebra (and it is one)” and “It LOOKS like a zebra (but its actually not)”. After establishing that different prosodic contours map onto these meanings (Experiment 1), we demonstrated that prosodic interpretation is shifted by inclusion of another alternative (Experiment 2); the reliability a speaker’s use of prosody to signal pragmatic alternatives (Experiment 3); and most importantly by the distribution of cue values along a continua (Experiment 4). We conclude that listeners derive linguistically meaningful categories from highly variable prosodic cues through rational inference about assumptions that are shared in the conversational context and adaptation to distributional characteristics of prosodic cues.