Public Release: 

Imaging Studies Reveal Process Of Verbal Memory Formation

Massachusetts General Hospital

Enhanced Imaging Technique Can Pinpoint Specific Structures Key To Making Memories

BOSTON, Mass.--The birth of a memory, the split second when the human brain encodes an event for future reference, has been captured through sophisticated neuroimaging and used to predict accurately whether a specific experience will later be remembered or forgotten, according to a study published in the Aug. 21 issue of Science.

The article reports collaborative research by scientists at the Massachusetts General Hospital (MGH)-NMR Center in Boston, Washington University in St. Louis and Harvard University showing that the level of activity in certain brain structures involved in processing verbal information can predict whether that information will be retained in memory.

"This is the first time we've been able to peer inside someone's brain and predict on average whether or not they will remember what they are now experiencing," says Randy L. Buckner, PhD, senior author of the journal article.

A member of the Psychology Department at Washington University and the MGH-NMR Center, Buckner adds, "Now we can actually see areas of the brain as they go about the process of memorization." Anthony Wagner, PhD, a postdoctoral research fellow at the MGH-NMR Center and also affiliated with the Harvard University Department of Psychology, spearheaded the research and is the article's first author.

"This study provides a firmer biological underpinning for the concept that how we encode information is key to whether or not it is remembered," says Daniel Schacter, PhD, chairman of Psychology at Harvard University and a coauthor of the article who helped develop the study's conceptual framework.

"It is the first work to tie the creation of a specific verbal memory to specific levels of activity in certain areas of the brain."

The Science paper reports on two series of studies carried out at the MGH that used functional magnetic resonance imaging (fMRI) to gauge brain activation during verbal tasks by measuring small but significant differences in blood flow to key regions of the brain.

In the first series, scans were taken of volunteers' brains as they engaged in two different activities: evaluating the meaning of words by determining whether they represented abstract or concrete concepts and evaluating aspects related to appearance, whether they were printed in upper- or lowercase letters. Psychological studies have shown that semantic processing of words, related to the words meaning, like the abstract/ concrete task, tends to result in better memory formation than nonsemantic processing, like the upper-/lowercase task.

The results of this imaging series clearly showed greater activity in certain brain areas, particularly structures within the left frontal and temporal lobes, during semantic processing tasks than during nonsemantic tasks, suggesting that memory formation could be associated with the activity of those structures.

However, because this first series consisted of what are called blocked studies, in which participants repeat the same kind of task for a period of several minutes, it could not pinpoint exactly how differences in brain activity relate to memory formation.

"The blocked study can't tell us whether observed processes are involved with memory formation or with other aspects of carrying out the semantic task, such as enhanced attention or more elaborate performance strategies," says Wagner.

"In addition, the blocked study cannot identify brain activity that predicts whether or not a specific experience will be remembered; such activity would clearly be associated with memory formation."

To search for brain activity clearly related to memory formation, the researchers ran a second series of studies that took advantage of a refinement in fMRI technique. Developed by Buckner and coauthor Anders Dale, PhD, of the MGH-NMR Center, the method known as "event-related" neuroimaging capitalizes on the speed of fMRI to capture images of whole brain activity in intervals as short as 30 milliseconds.

With this precision, the researchers were able to to distinguish brain activity in response to events as brief as two seconds long. In contrast, blocked fMRI trials require 40 to 60 seconds worth of information. Studies using other imaging techniques like positron emission tomography (PET) may require several minutes.

While in the MR scanner, participants watched a rapidly changing series of images, either a nonverbal image (a plus sign) or words for which they had to make an abstract/concrete determination. After the scans were collected, partici-pants reviewed a series of words, some they had seen in the scanner, some which they had not seen, and indicated whether they remembered seeing it during the imaging studies.

When the researchers compared the level of brain activity during processing words that were remembered to the activity for words later forgotten, they found that increased activity in specific structures in the left frontal and temporal lobes could predict whether participants would accurately and confidently remember a word seen in the scanner. These areas were among those that had shown increased activity in the blocked study, but several other areas identified in the first trials did not show increased activity in the second series, ruling out their importance in memory formation.

One of the structures showing increased activity in verbal memory formation is the left parahippocampal gyrus, a main input pathway to the hippocampus, a part of the brain that has long been recognized as crucial for storing and retrieving memories. The authors note that, while this area previously had been implicated in processing unfamiliar experiences, this study suggests the parahippocampal gyrus plays a broader role in memory formation. Even when experiences are similarly novel, differences in parahippocampal activity were seen that predict future memory.

Wagner notes that the availability of event-related imaging was key to the ability to carry out the second series of studies. "This new technique is what allows us to specify how brain activity during learning differs between experiences later remembered and those later forgotten."

Schacter adds: "The split-second thoughts that people have about incoming information play a huge role in whether that experience will be remembered or forgotten. Something that happens in a second can have conseqences for the rest of our life."

The same issue of Science also contains a study from Stanford University that similarly identifies structures in the right frontal and right and left temporal lobes that predicted visual rather than verbal memory. Taken together these studies indicate that there are multiple networks that support different kinds of memory formation.

Other coauthors of the paper are Michael Rotte, MD, and Bruce Rosen, MD, of the MGH-NMR Center, and Wilma Koutstaal, PhD, and Anat Maril of the Harvard Psychology Deparatment. The study was supported by grants from the National Institute on Aging, the National Institute on Deafness and other Communication Disorders, the Human Frontiers Science Program and the Deutsche Forschungsgemeinschaft [German Science Foundation].

Note to reporters/editors:
A graphic illustrating this research is available from the MGH Public Affairs Office. Please e-mail Susan McGreevey at to receive a copy.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.