Brain imaging studies have shown that episodic encoding into long-term memory preferentially activates the left prefrontal cortex and retrieval activates the right prefrontal cortex. However, it is unclear to what degree verbal analysis contributes to the left prefrontal activation during encoding. The present study was designed to avoid verbal analysis during encoding by using abstract pictures and computer-generated sounds which were difficult to code verbally. Sounds and pictures were grouped into six stimulus-stimulus pairs. When the sound from a pair was presented, the subjects were instructed to recall and visualize the associated picture. After 2.0 s the associated picture and another picture appeared on the screen and the subjects were required to identify the associated picture. Feedback about the choice was then given. Regional cerebral blood flow (rCBF) was measured with [15O]butanol and positron emission tomography (PET) in 10 subjects during initial training on the paired-associates task (encoding scan) and after 35 min of training (retrieval scan). Performance during the encoding scan was 59% correct and during the retrieval scan 98% correct, with a mean reaction time of 709 ms during retrieval. The rCBF was also measured during a control condition without any instruction to encode or retrieve. Compared with retrieval, encoding showed significant activation of the posterior part of the right middle frontal gyrus, the right inferior parietal cortex, the cingulate cortex, the left inferior parietal cortex and the left inferior and middle temporal gyri. The rCBF increase during encoding was strongly correlated with the rate of encoding. Retrieval was compared with both encoding and control. In none of these comparisons was there any prefrontal activation. The lack of prefrontal activation during near-perfect performance of the retrieval task suggests that the prefrontal cortex is not necessarily active when retrieval is fast and accurate, or what might be called automatic. Encoding was not associated with more activation of the left than the right prefrontal cortex. This result presents a limitation to the generality of left prefrontal activation during episodic encoding, which has been found in several previous brain imaging studies. Differences between studies in the relative activation of left and right prefrontal cortex during encoding and retrieval might be due to differences in paradigms, the type of stimulus used, and the demand for working memory and verbal analysis.