During a naming task, time pressure and a manipulation of the proportion of related prime–target pairs were used to induce subjects to generate an expectation to the prime. On some trials, the presented target was orthographically and generally phonologically similar to the expected target. The expectancy manipulation was barely detectable in the priming data but was clearly evident on a final recognition test. In addition, the recognition data showed that the nearly simultaneous activation of an expectation and sensory information derived from the orthographically and phonologically similar target produced a false memory. It is argued that this represents a blend memory.