In three experiments that used levels of processing and study time manipulations as independent variables in a word fragment completion task, the validity of the assumption of independence between recollection and automatic influences of memory was assessed. This assumption underlies the use of the process dissociation procedure (Jacoby, 1991), a tool suggested for distinguishing the different contributions of recollection and automatic influences of memory. Overall, it appeared that semantic processing, as compared with physical processing at study, positively affected recollection but negatively affected automatic influences of memory in word fragment completion. This negative effect on the automatic influences was reduced when the available study time decreased. The incompatibility of these results with the assumption of independence between recollection and automatic influences of memory and their impact on the applicability of the process dissociation procedure are discussed.