Using advanced, non-invasive imaging techniques, Yoko Akado and Craig Stark compared the areas of the brain that were active when a subject was encoding a complex event and afterwards, during exposure to misleading information. For example, subjects were asked to watch a vignette comprised of 50 photographic slides showing a man stealing a woman's wallet, then hiding behind a door. A little later, the subjects were shown what they thought was the same sequence of slides but unbeknownst to them the second set of slides contained a misleading item and differed in small ways from the original--the man hid behind a tree, for example, not a door.
Two days later, the subjects took a memory test, which asked them to recall details such as where the man hid, and which presentation--the first, second, or both--contained that information. Memory for a misinformation item was scored as a false memory only if the subject attributed the item to either the original presentation or to both the original and second slide presentations.
Stark and Akado found clear evidence that the subjects' brain activity predicted if their memories of the theft would be accurate or false. Consistent with findings from numerous previous studies that have reported that areas such as the hippocampus are highly active during memory formation, Okado and Stark found activity in the left hippocampus tail as well as perirhinal cortex was correlated with successful encoding of an item in memory, even when the memory that was formed was for a false item. But in subjects who had formed false memories, it was noticeable that activity in other brain areas such as the prefrontal cortex was weak during exposure to the second sequence of slides compared to during the original viewing.
Okada and Stark suggest that activity in the prefrontal cortex is correlated to encoding the source, or context, of the memory. Thus, weak prefrontal cortex activity during the misinformation phase indicaates that the details of the second experience were poorly placed in a learning context, and as a result more easily embedded in the context of the first event, creating false memories.