Making Memory May Mean Modeling and Remodeling


by Moheb Costandi

January 20, 2012

Long-term memory allows us to perform mental time travel, enabling us not only to recollect events from the distant past but also to imagine those that have not yet taken place. During a symposium held at the meeting of the Experimental Psychology Society in London earlier this month, cogntive neuroscientists presented new findings on the neural systems underlying these processes, and discussed current thinking in the field.

Daniel Schacter of Harvard University discussed the relationship between the recollection of episodic memories—memories of life events—and our ability to imagine the future. Recollecting episodic memories is a reconstructive process; the brain stitches together fragments of memories to generate a reliable, but not entirely accurate, representation of an event.

Recent research shows that amnesic and epileptic patients who have damage to their temporal lobes have trouble imagining themselves in future events and novel situations. The ability to imagine the future also declines with age, and is impaired in Alzheimer’s disease and depression, which are also associated with volume reductions in the temporal lobes.

Thus, imagining the future involves remembering the past, and each activates distinct subsystems within a common core network of brain structures. This core network includes the hippocampus, a medial temporal lobe structure known to be involved in episodic memory. [see "One Man’s Continuing Contribution to the Science of Memory"] Indeed, imagining the future produces higher levels of activity in the hippocampus than remembering the past; this may be because encoding a memory takes more energy than simply calling one up.

Schacter and his colleagues propose that the hippocampus plays multiple roles in the processes of episodic simulation; they think of episodic memory as a tool that evolved to help us prepare for future events. The researchers are now investigating the role of emotions in simulating the future. Some visions of the future might worry or excite us, leading them to wonder if the fate of highly emotional simulations differs from that of others.

“We found that people tended to forget simulations of negative experiences more rapidly over time than simulations of positive and neutral experiences,” says Schacter. “We also found that repeatedly simulating negative and positive future experiences, but not neutral ones, made those experiences seem more plausible.”

Do surprises make stronger memories?

Rick Henson of the MRC Cognition and Brain Sciences Unit at the University of Cambridge described the Predictive Interactive Multiple Memory Systems (PIMMS) model of long-term memory. This uses Endel Tulving’s Serial Parallel Independent processing model of memory as a starting point. According to Tulving’s model, events are encoded first in semantic memory (or memory for meaning), and later in episodic memory. They remain stored in both systems, and can be retrieved from either one independently.

The PIMMS model takes the idea further, positing that memory encoding depends upon the accuracy of predictions about the experience. Memory systems interact to generate a prediction about what is happening, and this prediction is compared to actual incoming sensory information.

Successful encoding of an episodic memory is closely related to prediction errors—especially to the novelty of the stimuli being encoded. According to the PIMMS model, most learning occurs when there is a large mismatch between the brain’s predictions and the actual outcome. It is these failures of prediction that drive the encoding of episodic memories.

For example, we are more likely to remember encountering a familiar person in an unusual or unexpected context: if you see your local butcher unexpectedly at your workplace, or on the bus, you are much more likely to remember encountering him than if you see him in the familiar context. Different aspects of this process are likely to be encoded by different memory system components: spatial contexts are thought to be encoded by the hippocampus, and object familiarity, or “the feeling of a memory” by an adjacent structure called the perirhinal cortex.

Call for cross-modal collaboration

In a lecture delivered immediately after the symposium, Jon Simons of the University of Cambridge discussed the approaches used by experimental psychologists to study memory processes. He noted that the different approaches used to investigate memory—such as brain scanning, behavioral tests and examination of brain-damaged patients—often yield conflicting results, and that each one has limitations.

Simons’ work includes investigations of reality monitoring, the process by which we determine whether or not our perceptions and memories are real or imagined. He has shown that failures in attentional systems in the prefrontal cortex can explain certain types of hallucinations. More recently, he and his colleagues published evidence that individual differences in reality monitoring abilities are associated with structural variations in a brain structure called the paracingulate sulcus. At the meeting, he was awarded the 19th Experimental Psychology Society prize, which aims to recognize distinguished research by experimental psychologists at the early stage of their career.

“We’re coming to the realization that any single method isn’t going to give enough information about the relationship between brain and behavior,” Simons said. “Using a single method always has limitations because there are always problems with interpretation. The way to address that is to look for converging evidence across multiple methods, and I think the big breakthroughs in coming years will come from researchers who are doing that.”