Proust, it turns out, was not really Proustian at all. Lost time did not come flooding back the moment Marcel tasted those madeleine crumbs sopped in lime-blossom tea. What actually happens, if you read the passage carefully, is that the narrator is ambushed by an intense sensation that he can't account for. It actually takes him 10 goes before he links his bliss to a childhood memory of Aunt Léonie, cakes and Combray. From there it requires 3,000 pages of painstaking prose to corral the past into some sort of meaningful order.

It is this effortful quality, this way that memory is something that you have to "do" that Charles Fernyhough wants us to take away from Pieces of Light, his attempt to explain what he calls "the new science of memory". In fact, this new science turns out to be pretty old science if you happen to be an academic who works in one of memory's cognate disciplines: psychology, neuroscience, even evolutionary biology. But Fernyhough, who is a popular science writer as well as an academic psychologist, is worried that most lay people still think of memory in terms of a vast personal DVD library. When we want to recall that long, hot summer of 1976 or last week's committee meeting we imagine ourselves reaching for the file in which that experience is stored, neat as a new pin. People with bad memories are assumed simply to have lost the knack of finding their way along their own library shelves.

In fact, as Fernyhough persuasively shows, memory is far more mutable than that. Every act of remembering is an act of creation, a confabulation stitched together from an array of different cues. We know this, really, when we get into a muddle over whether we actually recall an incident from childhood or whether we've simply been told about it or seen a photo.

What is harder to accept is that all our memories are equally provisional, created not out of a stable if sometimes cloudy past, but from the urgent needs of the present. We remember what we remember because it helps us negotiate who we are today and what we might become tomorrow. But that's not all. Each act of remembering, and especially each act of retelling, subtly changes the memory itself. What we end up with is a smudgy copy of a copy of a copy, over which the officious present has drawn a sharp new outline and now dares us to disagree.

To make all this clearer Fernyhough serves up the latest findings in neuroscience, quoting academic studies in which the hippocampus, amygdala and prefrontal cortex flash on and off like fairy lights. He mixes these with in-depth case histories of people such as Claire whose memory failed at the age of 43 when her brain was diced by herpes simplex; and poor Colin, a lorry driver who remains haunted by a fatal traffic accident that was not his fault. If Fernyhough had pushed harder he might have produced the kind of book reminiscent of Oliver Sacks at his early to mid-career best. In other words, he might have been able to use the poetic oddness of Claire's and Colin's personal wiring as a way of getting closer to the heart  or the brain  of the human condition. But this he never quite manages to do, and instead Claire and Colin  and Patrick, Peter and Nanna Martha and all his other case studies  remain locked inside their own narrative dilemmas, unable quite to join hands with the rest of the book.

It's not as though Fernyhough is uncomfortable with using personal testimony to put warm flesh on hard science. A sizeable chunk of the book is taken up with him exploring his own personal past. He tells us about a potty-training accident when he was three, birthday thank you letters he wrote when he was six, awkward sailing trips a little later with his newly divorced Dad. He strides around Sydney and Cambridge, cities where he lived long ago, and frets whether his deeply felt personal memories are actually rehashed from holiday brochures and promotional blurbs.

These sections do not add greatly to the book. Fernyhough namechecks WG Sebald a couple of times and it is hard not to feel as we tramp along the Blackwater with him in search of memories of his late father that he is straining too hard after a luminous opacity. Where the book really springs to life is, ironically, in its more journeyman parts. Fernyhough is a crisp and knowledgeable guide to all the data that generally stays buried deep in specialist journals. We hear about the New Zealand twin study, in which fractious siblings claim heroic memories for themselves and smuggle the painful or embarrassing ones on to their other halves. Then there's the discovery that a group of people who are allowed to discuss an event actually remember less about it than the same people tested individually. And that's not forgetting the clever pigeons who quickly learn to distinguish between random squiggles and photographs of natural scenes and can still tell the difference two years later.

Fascinating though these snippets are, they do not quite add up to the "new science" that the book's subtitle hopefully proposes. What we get instead is an episodically enlightening meditation on the complex business of remembering, forgetting, and re-remembering all over again.

 Kathryn Hughes's biography of Mrs Beeton is published by HarperPerennial.

During the initial stages of the Watergate inquiry, the White House counsel to President Nixon, John Dean, was known as "the human tape recorder" because of his ability to recall precise details of events. Of course it soon transpired that there were actual tape recorders at the White House deployed to record those same events. In a famous paper, the late cognitive psychologist Ulric Neisser used the comparisons between Dean's testimony and the recordings to show that Dean frequently recalled the gist of conversations with Nixon, but that his memory was often systematically distorted. For example, Dean recounted that a particular meeting started with the president asking him to sit down (Nixon didn't do this, but it's often how meetings start), and he recalled the president saying he'd "done a good job" and had "appreciated how difficult a task it had been". (All Nixon did was describe Dean's work as "skilful".)

Many of the errors in Dean's memories are normal: we don't remember things that happen in a verbatim way; instead it is the general sense of events that we recall. Like Dean, we incorporate what we "know" to be true about the world into our memories, and these memories can often be selfserving. Neisser contributed to an evolving view in psychology, dating back to the 1930s, that memories are not simply encoded verbatim reports: unlike the memories in Total Recall, which can be erased, inserted or replaced, human memory is something that we do, and remembering is an active process. We reconstruct events rather than replaying a tape recording, and that process involves many factors, such as how we desire the things to be.

In his new book on memory, psychologist and novelist Charles Fernyhough provides an outstanding overview of how this dynamic understanding can be applied to many different facets of memory. He travels to Sydney to find locations he wrote about in his first novel, The Auctioneer, and shows how his subsequent imaginings of these places have irrevocably overwhelmed what he actually saw. He also uses landscapes to bring memories and people back to life, re-enacting walks he took with his late father, and movingly describing the ways that we keep the dead alive, for ourselves and for those around us, by telling our memories, our stories. He shows how we use memories to connect ourselves to the past but also to tell stories to make the past what we want it to be.

Fernyhough's skills as a writer are evident both in the beautiful prose and in the way he uses literature to illustrate his argument. As he points out, since memories are the stories we tell, we can learn much about memory from the words of writers. He draws on both science and art to marvellous effect in his exploration of why early memories are so often filled with sunlight. He also shows that a "Proustian" rush of memory has little to do with what Proust actually described. But a rush of memories is what I experience when, for example, I smell Vidal Sassoon hairspray, as I glimpse tantalising sensations and emotions of being a teenager with complicated hair. (It was the 80s.) Strikingly, if I read my diaries from those years I get nothing like the same sense of familiarity. As Fernyhough might put it, perhaps the stories I wanted to tell about my life when I actually was 17 are not the stories I now want to tell myself about those sunny, hairsprayed days.

Sophie Scott is professor of cognitive neuroscience at University College London