Memory: an anthology, Edited by Harriet Harvey Wood & AS Byatt

We think of memory as a single faculty, but in truth it is a hotchpotch of many distinct abilities. We know this because brain damage sometimes knocks out just one while leaving the rest intact. That is what happened to the man who mistook his wife for a hat. Oliver Sacks's patient was fine, except that he'd forgotten how to recognise objects. Other kinds of brain damage can remove nothing but the ability to identify faces, or to perform familiar manual tasks. Then there are the amnesiacs who cannot recall incidents prior to some brain injury, even though they can keep track of recent activities. Other amnesiacs suffer the opposite plight, conveyed so effectively in the film Memento. They can remember their life up to their injury, but can't form new memories.

When all our memory capacities are intact, they enable us to preserve information from the past. Psychologists distinguish three broad categories of remembered information. Procedural memory retains the kind of practical information that is rarely forgotten, like how to ride a bicycle. Semantic memory preserves factual knowledge, like the date of the Battle of Hastings. And episodic memory delivers first-hand recall of events from our own experience, like your first day at school or last year's holiday.

AS Byatt and Harriet Harvey Wood's intriguing anthology has two parts: the first a series of commissioned essays by literary and scientific eminences, the second a selection of brief extracts, grouped under such headings as the "Idea of Memory" and "Memory and Imagination". The editors aim to cover all aspects of the subject, but their main focus is on episodic memory and its literary embodiment. The library rather than the laboratory is the natural habitat for both Byatt and Harvey Wood (for many years head of literature at the British Council) and their first concern is the way experience gets woven into texts. Scientists and philosophers do get a look in, but strongly outnumbered by novelists and critics.

By and large, the literary contributions take it for granted that our episodic memories provide a genuine record of past experiences. Craig Raine airs doubts about Proust's explanation of the pleasures of memory, but doesn't query Proust's presupposition that memories recall real incidents. This literary confidence is not supported by the scientists.

According to myth, everybody over 50 is supposed to remember exactly where they were when they heard that Kennedy died. Psychologist Ulrich Neisser explains that this just isn't so. They may think they remember, but they're likely to be wrong. When the space shuttle Challenger exploded in 1986, Neisser immediately asked a sample of people where they heard the news. Three years later, he asked them again. A quarter of his sample gave an account wrong in every particular, and most were mistaken.

Many people resist the idea that vividly imagined episodes might be fabrications. In the anthology, Hilary Mantel's extract blankly rebuffs queries about her reliance on episodic recall in her memoir Giving up the Ghost ("I believe strongly in the power and persistence of memory"). Against that can be placed a contribution by Oliver Sacks, which confesses how he couldn't possibly have experienced an episode described in his memoir Uncle Tungsten. He says that he would have sworn in court that he remembered a bomb falling next door in the Blitz, until his brother pointed out that he was at school at the time, and knew about it only from a detailed letter.

The authenticity of literary texts isn't the only thing called into question by the unreliability of memory, as acknowledged in a short section on "False Memories". Ian Hacking and Elizabeth Loftus both urge caution about the "recovered memories" on which many accusations of past child abuse are based. Neisser's essay touches on another danger of faith in episodic memory. DNA evidence now regularly leads to the quashing of wrongful convictions. Most turn out to have been based on eyewitness evidence. Of the first 200 prisoners released because of new DNA evidence in the US, over 75 per cent were in jail because of what someone claimed to have seen.

It may seem surprising that episodic memory should be so capricious. But this makes some sense. A number of contributions conjecture that episodic memory is peculiar to humans. This is disputable. If dogs can dream – and there is plenty of evidence they can – perhaps they can also imaginatively replay incidents from their past. Still, what they clearly can't do is arrange these incidents into a coherent narrative, the story of their life.

It is this narrative ability that is peculiar to humans. Our episodic replay is just one way in which we fill in the chapters of our life stories. Given this, we can see why we might often want to augment our first-hand experiences. There are lots of other ways of finding out about our pasts, most obviously from parents and others. This will sometimes encumber us with false information, but that is the price for opening ourselves to all the sources.

If the ability to construct narratives is peculiar to humans, where did it come from? Is it aided by our genetic heritage, or purely a cultural phenomenon? Unfortunately, these questions are absent from this anthology. There is an interesting commissioned essay by Patrick Bateson on "Memory and Evolution", but this is about the way evolution can turn habits into instincts, not our facility at story-telling. Still, this anthology isn't really designed for those who want to know about the scientific origins of story-telling. For those who want to know how literature makes stories out of memories, on the other hand, it will be a very useful companion.

David Papineau is professor of philosophy at King's College, London

Chatto & Windus £25 (412pp) £22.50 (free p&p) from 0870 079 8897