Neuroscience: Recalling the future

Journal name:
Date published:
Published online

Our memories may be tools for working out what's to come and what to do about it, finds Linnaea Ostroff.

Predictions in the Brain: Using Our Past to Generate a Future

Edited by Moshe Bar Oxford University Press: 2011. 400 pp. $99.95 ISBN: 9780195395518

Buy this book: US UK Japan

Memory is the most intimate of abstractions. The matter of how memories are made and stored is a research goal that needs no selling, perhaps driven by a fear of memory loss and our dread of personal obliteration. Yet one question has been neglected: why does memory exist?

Possible answers are explored in Predictions in the Brain, a collection of 25 rigorous, data-laden cognitive-science reviews edited by neuroscientist Moshe Bar. He and his co-authors propose that prediction is a unifying principle of brain function, and that predictions are created from memories. As contributor Yadin Dudai writes: “Memories are made mostly for the sake of present and future.” Memory systems do not store past experiences, but recycle their components into the imagined future.

In throwing evolutionary light on a fundamental process, this idea has legs. As survival advantages go, our ability to envision and plan a nuanced future is a masterpiece — arguably, the root of our success as a species. If memories are used to generate predictions that drive our actions, then the mechanisms of prediction are as worthy of study as those of memory. The growing body of experimental data reviewed in this volume supports a shift in memory research from storage to prediction.


Hitting a baseball requires the brain to predict what might happen next, based on past experience.

From anticipating fine motor movements to imagining our old age, what we think will happen in the future depends on what we know has been. We hit a baseball on the basis of how we have done it before; our experience of traffic lights allows us to anticipate and halt in front of the red one. With only past events to rely on, however, accurate prediction is difficult. A memory system must store information reliably, yet be flexible enough to keep knowledge up to date and to know which items are relevant to a particular purpose. The brain manages these challenges well. It is easy to think of memories as faithful, discrete records of our past, pressed and dried like items in a scrapbook. They are not — and for good reason.

Dudai explores how recalling memories makes them plastic, susceptible to updating or the addition of new associations. This mechanism serves prediction by providing us with the most up-to-date information. For example, if the train to work has been on time 4 out of 5 days, taking it again would seem to be a good decision. But if by next month this has become 6 out of 20 days, choosing the train looks unwise. Each memory of the train must be associated with the previous and subsequent ones as they occur, to estimate the chance of it being late tomorrow.

Beyond the constant generation of everyday predictions (I must push the door to open it; the milk will run out soon), our brains also create detailed pictures of future events. This capacity is thought to be uniquely human, and occupies a significant portion of our mental activity. It is entwined with episodic memory — our personal collection of autobiographical clips. Predictions in the Brain reviews experimental evidence for substantial overlap between memory recall and future simulation. Generally, patients with deficits in episodic memory are unable to envision a complex future; and healthy subjects show similar brain activation when recalling the past and imagining the future.

Several reviews describe the construction system of the brain, which assembles imagined events from our past knowledge. We create altered versions of remembered situations by recombining primitive images with digested information from memory. Our experience of time may be a consequence of a single, versatile storytelling mechanism in the brain that draws its cast, plot and scenery from the memory banks. As authors Demis Hassabis and Eleanor Maguire point out, time is simply one more component of this drama.

As for the neural mechanisms responsible for imagining the future, several brain systems are thought to cooperate in generating mental constructs. Various cortical areas are implicated, and one suggestion is that the hippocampus combines and integrates information stored in other systems. Through this lens, dreams could be seen as shuffles through the mental primitives.

We have always been uncomfortable with explicit prediction: the mythical Cassandra was hated for her supernatural foresight, the major religions forbid fortune-telling, and we are anxious about testing ourselves for genetic diseases that have a delayed onset, such as Huntington's. But Predictions in the Brain shows that this rejection of a certain future is legitimate if much of our mental life is devoted to looking ahead. Memories are not the flotsam of the past: they are the raw materials of cognition.

Additional data