In question-answering, speakers display their metacognitive states using filled pauses and prosody (Smith & Clark, 1993). We examined whether listeners are actually sensitive to this information. Experiment 1 replicated Smith and Clark's study; respondents were tested on general knowledge questions, surveyed about their FOK (feeling-of-knowing) for these questions, and tested for recognition of answers. In Experiment 2, listeners heard spontaneous verbal responses from Experiment 1 and were tested on their feeling-of-another's-knowing (FOAK) to see if metacognitive information was reliably conveyed by the surface form of responses. For answers, rising intonation and Longer latencies led to fewer FOAK ratings by listeners. For nonanswers, longer latencies led to higher FOAK ratings. In Experiment 3, electronically edited responses with 1-s latencies led to higher FOAK ratings for answers and lower FOAK ratings for nonanswers than those with 5-s latencies. Filled pauses led to lower ratings for answers and higher ratings for nonanswers than did unfilled pauses. There was no support for a filler-as-morpheme hypothesis, that ''um'' and ''uh'' contrast in meaning. We conclude that listeners can interpret the metacognitive information that speakers display about their states of knowledge in question-answering. (C) 1995 Academic Press, Inc.