Abstract
The role of encoding/retrieval conditions compatibility was investigated in a reality-monitoring task. An experiment was conducted which showed a positive effect of reinstating distinctive encoding operations at test. That is, generation of a low-frequency word from the same word fragment at study and test significantly enhanced item recognition memory. However, reinstating of relatively more automatic operations of reading or generating a highfrequency word did not influence recognition performance. Moreover, LF words were better recognized than HF words, but memory for source did not depend on the encoding/retrieval match or on the word-frequency. In comparison with reading, generating an item at study significantly enhanced source memory but generating it at test had no effect. The data were analysed using a multinomial modelling approach which allowed ruling out the influence of a response bias on the measurement of memory ability.