A series of experiments was conducted to examine whether listeners could utilize subtle acoustical cues in reconstructing words that were missing from the speech signal. Preliminary tests revealed that listeners can make use of semantic context in such reconstruction, analogous to earlier results in reading. In Experiment 1, listeners were able to use intonational cues at the beginning of a string to detect whether or not a sentence-initial word had been missing. In Experiments 2 and 3, listeners utilized coarticulatory cues for palatalization in sentence-internal locations to guide their lexical selection of reconstructed words. In Experiment 4, listeners were influenced in lexical selection by the duration of a silent interval that had replaced the missing word. Taken together, the results indicate that listeners can utilize the acoustical fine structure of speech to both detect the occurrence of a missing word and guide its lexical reconstruction.