In a typical false-memory experiment, participants study lists of words containing many items from the same category (e.g., countries or body parts). Subsequently, they often falsely recognize unstudied items from the same category during a memory test. Most explanations suggest these false recognition responses represent genuine false memories, arising from a false sense of familiarity generated by unstudied category words. However, others argue that when stimuli have a clear categorical structure, false recognition responses are "strategic" reasoning that test items from the category are more likely to have been studied.
We present the results of three experiments. In the first, we use a computational model to demonstrate that people's confidence in their recognition responses is weaker when based on categorical information. We argue this is because they are relying on reasoning rather than memory. In a follow-up experiment, we replicate the study while asking participants to think aloud as they perform the task. The results show that participants often explicitly reason about categorical information, and their verbalizations differ from those they produce when relying on memory.
In the third experiment, we manipulate awareness of categorical information using ad hoc categories (e.g., things a dog chases). We show that false recognition responses were vastly inflated when participants were aware of the categorical information. However, we also observe a slight increase in false recognition rates for highly prototypical category items, even when participants are unaware of the category. This suggests that genuine false memories are possible, albeit rare, in such tasks.