Memory impairments, known in technical parlance as “amnesias,” have always occupied a central place in neuropsychology. Not surprisingly, a process as complex as memory can disintegrate in a vast variety of ways. Memory impairment is almost never global. It is almost always partial, producing a number of different forms of amnesia.One of the main distinctions made in neuropsychology is between “anterograde amnesia” and “retrograde amnesia.” Anterograde amnesia is a loss of the ability to learn new information after brain damage had set in. Retrograde amnesia is an inability to recall information acquired before the damage took place. Someone who suffered brain damage in a car accident last year and is now unable to recall what he read in the newspaper yesterday may suffer from anterograde amnesia. And if this person does not remember the name of the company that employed him for five years before the car accident, he is likely to suffer from retrograde amnesia. It is not uncommon to develop the two forms of amnesia together as a result of brain damage, and our poor fellow may be both unable to recall what he learned recently and to access the information acquired before the accident.The distinction between anterograde and retrograde amnesia depends on our knowledge of the exact time when brain damage occurs, which is not always easy to figure out. If a previously healthy individual suffered traumatic brain injury in a car accident, the exact time of the event is usually easy to establish. But in a case of dementia this is not possible, since in dementia the decline is gradual, unfolding over years. By the time a patient is diagnosed with dementia, he or she will have already been ill for a long period of time, often measured in years, not months.Despite these diagnostic quagmires, the distinction between anterograde and retrograde amnesia has been very useful to neuropsychologists and neurologists for years. These two forms of amnesia often appear together; but for reasons idiosyncratic rather than logical, anterograde amnesia has always been given more attention, and it was assumed that it is more common and more severe than retrograde amnesia.My own clinical experience contradicted this widely held assumption. I thought that we had witnessed the effects of a common logical error, the absence of evidence mistaken for the evidence of absence. (Because researchers did not pay nearly as much attention to retrograde amnesia as they did to anterograde amnesia, they did not find it.) In my own work I have always been particularly intrigued by retrograde amnesia, since I felt that it offered a unique window into the way knowledge is organized and stored in the brain.Among other things, retrograde amnesia tells us about the time course of long-term memory formation. When memory of the past suffers following brain damage, not all memories suffer to an equal extent. Virtually without exception, relatively recent memories will be more affected than the memories for a very distant past. This phenomenon is known as the temporal gradient of retrograde amnesia.Someone who suffered head injury in a terrible car accident may have lost his memories for the events of a month or two months, even a year or two years, before the accident, but he is more likely to remember the events of a decade or two decades earlier. The same is true for a patient suffering from dementia. This is why the common sense argument that someone’s memory cannot be all bad, since he remembers the names of his grade school teachers, really proves nothing. A patient suffering from Alzheimer’s disease will have such memories of very distant past preserved well into the advanced disease stages, while the memories for more recent events will be gone relatively early in the course of the disease.The temporal gradient is counterintuitive. Many years ago I conducted a casual survey of a few friends from various walks of life, asking them to venture a guess as to what memories are more likely to suffer in brain disease: relatively recent or very distant. Guided by their common sense but unencumbered by the technical knowledge of brain science or neuropsychology, they all guessed wrongly without exception that the more distant memories would suffer first. This common sense—defying property of retrograde amnesia may serve as a great clinical tool for telling apart the memory loss caused by brain damage from the memory loss caused by psychological factors such as hysteria, or from plain malingering.But the temporal gradient teaches us more than how to confound the unsuspecting public. It tells us volumes about how long-term memories are actually formed. Indeed, if memories remain vulnerable as long as they depend on the reverberating patterns of activation, then the extent of the temporal gradient allows us to estimate the amount of time necessary for long-term memory formation to run its course. And it turns out that retrograde amnesia may affect memories going back years and even decades.It is known, for instance, that hippocampal ablation may result in retrograde amnesia going back as far as fifteen years. This means that it may take that long for a permanent, structural, relatively invulnerable long-term memory to be formed in the brain.The process is gradual and incremental, rather than an abrupt, precipitous emergence of the long-term trace where there had been nothing before a second ago. The gradual nature of the long-term trace formation is illuminated by another peculiar feature of the temporal gradient—its “shrinkage.” As we already know, it is not uncommon for a patient who had just suffered brain damage in an accident to have memory loss extending years and even decades back. But with the passage of time, some of the memories will return, and the recovery of memories will follow an orderly temporal course.The span of the memory loss will “shrink.” (This casual and somewhat inelegant word has in fact been adopted as a technical term in memory research, and scientists write about “shrinking retrograde amnesia,” or “shrinking temporal gradient.”) Like so many other features of retrograde amnesia, this mysterious process defies common sense. The shrinkage unfolds backward, memory for the more distant events returning before memory for the more recent ones. But the shrinkage is usually incomplete and the memories for the most recent events never recover. Just how extensive the permanent memory loss is varies from patient to patient and depends on the severity of brain injury. This permanent memory loss is genuine and intractable. No amount of hypnosis or “truth serum” will help recover the lost memories, and any attempt to do so will simply reflect a lack of neuropsychological sophistication.The orderly and gradual process by which memories recover in “shrinking retrograde amnesia” tells us about the gradual nature of long-term memory formation. The farther along the process is, the more rapidly the memory will recover. But the memories whose formation has been assaulted by brain damage at very early stages are too fragile to rebound. They will be lost forever.So the most critical obstacle on a memory’s path into long-term storage is time itself. It takes years, or possibly even decades, for a long-term memory to form in the brain. Since there is no perpetuum mobile in the physical world, the reverberating loops stand a good chance of being extinguished on their own, and most of them are. Most reverberating loops become extinguished before the structural engram has had a chance to be formed. Nature appears to be very protective of the permanent memory store in the brain and the plank for being admitted into it is very high.*24\302\2*