There are data to suggest that the phonological representations of young children who stutter are less specified than their typically fluent peers. The purpose of the present study is to determine if this apparent difference in phonological encoding persists in adults who stutter. Utilizing a false memory paradigm, nine adults who stutter (AWS) were asked to listen to and then recall/produce 12 lists of 12 words each. Each word list was comprised of either semantic, phonological or an equal number of semantic and phonological associates of a single, unpresented, critical ‘lure’ word. Three parameters of recall performance were measured across these three conditions: 1) number of accurately recalled productions, 2) number of lure intrusions and 3) number of other intrusions. AWS produced significantly more accurate recalls in the semantic condition than either the hybrid or phonological conditions, and significantly more lure intrusions in the phonological and hybrid conditions than the semantic condition, but there was no significant difference on measures of other intrusions. These results extend the findings with young children who stutter, and indicate that the phonological representations are less robust than the semantic representations in the lexicon of AWS.