Application of the process-dissociation procedure has shown that conceptual
encoding episodes do not lead to automatic influences of memory on purport
edly data-driven indirect tests of memory. Using 2 variants of the process-
dissociation procedure with the word-stem completion task, the procedure is
shown to underestimate automatic influences of memory when prior encoding
includes a conceptual component. The underestimation is attributed to an aw
areness of past occurrence that is particularly likely with conceptually en
coded items. This effect occurs even in the absence of the signature of a g
enerate-recognize strategy and suggests that prior conceptual encoding may
contribute to automatic influences of memory in stem completion. A multinom
ial generate-recognize model is presented that fits these results and previ
ous results typically taken as support for the assumption that controlled a
nd automatic influences of memory are independent.