November 10, 2003
False Versus Real Memory Recall Looks Different In MRI Brain Scan

Functional Magnetic Resonance Imaging (fMRI) can detect differences in brain patterns when brains correctly and incorrectly think they have seen an image before.

Neuroimaging techniques can help determine if the neural processes driving this retrieval of inaccurate memories are different from those that drive the retrieval of accurate memories. Several research groups are using functional magnetic resonance imaging (fMRI) to address this question. The hope is that neuroimaging can help determine the various potential sources of false memories.

Daniel Schacter, PhD, and his colleagues at Harvard University have looked at neural activity associated with the creation of false memories. Previous studies had focused on neural activity associated with the retrieval of false memories.

Relying upon earlier work that showed the right fusiform cortex is involved in encoding the exact visual details of objects and the left fusiform cortex is involved in more general processing, Schacter’s group designed an experiment to test the role of the right fusiform area in avoiding the formation of false memories for objects similar to those seen previously but not exactly the same.

In the study, led by graduate student Rachel Garoff, participants underwent brain scans using fMRI as they made judgments about the size of various objects. A surprise memory test was then given when the patients were outside the MRI scanner. During the test, patients saw objects identical to those seen earlier, objects similar to those they had seen earlier, and new objects they had not seen at all.

Although the study is still in progress, results to date indicate that the right fusiform area was more active in these individuals during the encoding of objects participants later labeled the “same” as objects they had seen before. The right fusiform area was less active when patients incorrectly labeled objects the same when they were only similar, or when they labeled objects similar when they were actually the same.

“This preliminary finding supports the idea that the right fusiform area is tied to the encoding of specific visual details,” Schacter said. “It also suggests that false memories of objects can be reduced through additional activity of the right fusiform area during encoding.”

In another study, Schacter’s group showed that visual processing regions of the brain were reactivated during true memory but not during false memory.

Scott Slotnick and Schacter constructed “prototype” shapes by adjoining four curves into various shapes, then they distorted these prototypes to form “exemplar” shapes. The twelve individuals who have taken part in the study thus far were instructed to remember each shape and whether it appeared to the left or the right on a screen. “True memory” was defined as recognizing a shape that was seen previously, and “false memory” referred to mistakenly recognizing a shape that resembled a shape seen earlier but that was not actually seen. In the next step, fMRI was used to determine which areas of the brain were associated with true and false memory.

“We found that participants gave the same response regarding whether an object was “new” or “old” during true and false memory, which leads you to expect that the associated brain activity might be indistinguishable,” Schacter said. “But fMRI revealed there is a different activation of brain regions involved in visual processing during true versus false memory. What we need to do now is understand the meaning of this difference.”

This is not as impressive as it first sounds. The researchers can not say in each instance whether the memory being recalled is true or false. They only see a difference on average over many experimental runs.

However, Schacter points out that their work currently averages brain activity over many trials, so detecting the accuracy of a single memory is not yet possible.

Still, these experiments suggest that it might be possible to train people to be more aware of the strength of their own memory recall. If there are differences in brain activity when recalling accurate and inaccurate matches between viewed images and memory it might be possible to develop a training regime to let people know when they are making fake matches between memories and viewed objects. By getting that immediate feedback people might be able to calibrate their own sense of certainty and develop a better sense of just how strong the feeling of seeing a match has to be in order for it to be likely to be accurate.

Share |      Randall Parker, 2003 November 10 02:49 PM  Brain Memory

Chris Genovese said at November 12, 2003 8:36 PM:

Randall, very interesting, thanks. I'm hesitant to make too much of a news report, but here are a few quick reactions based on the news release.

* In the first study mentioned, it appears that Schacter et al. have found qualitatively the same results as other studies in the literature: that one can predict memory performance as a function of encoding depth and that the fusiform area is implicated in encoding. The former is not too surprising, and it explains their new results simply. False memories do not seem necessary here, except perhaps for news-worthiness. (Other alternative explanations include differences in attention or elaboration at study, or differential item difficulty on a per person basis.)

* Moreover, the news report on the first study conflates "false memories" and retrieval errors, but these are not quite the same. To deal with this distinction in such "false memory" studies, it is common to use what is known as the Remember-Know paradigm, where the participants are asked not just whether each test object is old or new but whether they actually remember it or just "know" it, say, via familiarity. Guessing that a stimulus seems familiar (a "know" response) is not necessarily responding to a false memory.

Now, Schacter and company are pros, so they probably followed this paradigm. But it's not clear from the report, and the question needs to be asked.

* Say you create several prototype dot patterns and tweak each around the prototype to produce patterns that you will show subjects. If you just present the tweaked exemplars (but not the prototypes), people tend to claim falsely that they saw the prototypes more than they claim correctly that they saw the individual exemplars. This is often called the prototype effect. There is an ongoing debate on whether people are falsely remembering the prototype by building a neural representation of it (at encoding) or if the error comes in at retrieval because the prototype is closely related to all the exemplars.

The second study in this report uses this basic design, and if the prototype is being represented, it is reasonable to call this a false memory. But I don't quite see how their results (as reported) distinguish among the two theoretical possibilities under debate, because both theories make the same prediction that the prototype would be relatively advantaged at test while the imaging data are only collected at test. The activation differences that were observed are vaguely described as "in visual processing", which suggests the researchers are not seeing differences in the regions they expected. It's possible that the differences in visual areas do reflect memory-related processing, but that's a hard call to make with these data alone.

* Hippocampal activation in the third study would be expected to be sensitive to novelty. It's unclear if or how they controlled for this. If the participant encoded a "key" feature well in the original phase, it will be well remembered at test and the information processed during the second phase will be less novel. Thus, good original encoding => good memory for original AND higher hippocampal activation in original. If the participant didn't encode an original feature well, it will be poorly remembered at test and the information processed will be more novel. Thus, poor original encoding => poor memory for original AND lower hippocampal activation in original. Such a novelty effect explains the results pretty simply and would surprise no one. I'd like to see the evidence for the more dramatic explanation they offer. (This is basically the same encoding-strength argument I mentioned in the first study, which makes it a rather parsimonious explanation.)

* Finally, do you know if this work has been accepted for publication? Are the papers available? It seems odd to have news releases on studies in progress at an early a stage.

Randall Parker said at November 12, 2003 9:36 PM:

Chris, There's a neuro conference going on and tons of press releases coming out of it. The press releases are not mentioning journal names unfortunately. Usually if a university puts out a press release they include a journal name date. You'd have go to digging for the universities of these folks and see if you can find anything on their sites.

Chris Genovese said at November 14, 2003 8:45 AM:

Ah, sorry. I read it quickly and didn't notice that the release was a meeting summary.

Thanks again.

Post a comment
Name (not anon or anonymous):
Email Address:
Remember info?

Go Read More Posts On FuturePundit
Site Traffic Info
The contents of this site are copyright ©