An "aha" effect in memory was first reported by Auble, Franks, and Soraci (1979). They demonstrated that recall was greater for sentences that were initially incomprehensible but which were eventually comprehended, as compared with sentences that were understood from the outset. The present studies extend this "aha" effect to memory for pictorial stimuli. In Experiment 1, a recall advantage for pictures encoded by connecting the dots as compared with those encoded by tracing or visual scanning occurred only in the absence of foreknowledge of the picture (i.e., an "aha" effect). In Experiment 2, we replicated this finding and obtained evidence that conceptually based, verbal foreknowledge does not function in a similar manner as does pictorial foreknowledge in suppressing the "aha" recall advantage. These results place important constraints on previous research on generation effects for visual stimuli and attest to the cross-modal generalizability of the "aha" effect.