One estimate of the capacity of human memory is that over the course of a lifetime, a person will store more than 500 times as much information as appears in the entire Encyclopedia Britannica. Somehow, with all that information in there, a lot of it is immediately available on a moment’s notice, with no apparent effort involved in the retrieval of it. For instance, if a person is asked, “What’s your mother’s middle name?” if it’s a piece of information that the person knows at all, it will have just popped into their head when asked, with no deliberation necessary.
The three essential processes involved in memory are acquisition, storage, and retrieval. For information to go into memory, it must come from somewhere (acquisition). It must then be stored, sometimes for a long time. It hasn’t truly been remembered, however, until it is retrieved. There are several competing models of how the human memory system works—the most popular one, known variously as the information-processing model or by the alliterative name of modal model, is presented here.
The memory system can be seen as consisting of three different storage systems, which differ from each other in capacity as well as in the duration of the memory. When information ﬁrst enters the system (coming in from the senses when reading a book or listening to a conversation, for example), it is held in the sensory buffer, or sensory memory. The capacity of this store is quite large, as everything detected by the senses at any given moment is present there, but its duration is very brief (less than one second). If information held there is not attended to and perceived, it fades away almost instantly. If it is perceived, however, it enters into the next stage, short-term memory. The term working memory has largely supplanted short-term memory among cognitive psychologists, to reﬂect the growing recognition that it contains everything that is currently being attended to, both sensory information and information retrieved from long-term storage. In fact, some writers are even equating working memory with the conscious self. Everything we are thinking about at any given moment is in working memory.
Short-term memory (STM) has a fairly brief duration and a fairly small capacity, and whatever a person is currently paying attention to is held in it. Once information enters into it, it will remain for twenty seconds or less unless it is processed further. Keeping information in working memory involves a process known as maintenance rehearsal, which is familiar to anyone who has ever been told to call a particular telephone number right away and hasn’t written the number down. The tendency is to repeat the number over and over until dialed, and it is completely forgotten once the phone conversation begins, and the number is no longer rehearsed. This idea of repeating the number to oneself, even with the mouth shut, is actually a very accurate representation of what is actually going on. People store things in STM using an auditory code. When people are interrupted while rehearsing a list of letters, for example, the resulting errors tend to involve sound-alike letters rather than letters that resemble the correct ones physically. Other laboratory evidence shows that hearing a separate list of letters while trying to rehearse another will interfere with memory, while seeing the new list on a screen will not. This is because rehearsal actually involves listening to the items in one’s head.
The capacity of STM is approximately seven chunks, give or take two (ranging, in other words, from ﬁve to nine chunks). A chunk is simply a meaningful unit of information; it can be a digit, a word, a letter, or an entire sentence, depending on the task. When a person is rehearsing a seven-digit telephone number, and somebody interrupts them (“Hey, did you want eight or twelve of these?”), the number is lost immediately, because the capacity of STM was exceeded. People can easily use larger chunks than one digit, however. Consider a social security number. Most people think of it in three discrete chunks (three digits, two digits, and four digits) rather than as nine separate pieces of information.
Maintenance rehearsal is good for keeping information active in STM, but deeper processing is required to move it into long-term memory (LTM). Although some information enters LTM automatically, deliberate placement of information into it requires elaborative rehearsal. That term is a bit of a misnomer, as no rehearsal is involved, at least not in the sense of repeating the information. What is actually involved is making connections between the new information and information that is already in LTM. The moment such a relationship is established, the information is in long-term memory.
There are actually at least three basic kinds of long-term memory. Episodic memory is any memory of an episode in life, any speciﬁc event at which one was present. This kind of memory is also known as autobiographical memory. Semantic memory is general knowledge of the world that can be described in words and doesn’t involve recollection of a particular event. Knowledge of a complex series of actions, or of how to do something, that cannot be adequately described in words, constitutes procedural memory. So, answering the question, “What is a bicycle?” involves semantic memory, recalling the long-distance bicycle race you rode in last week is an episodic memory, and knowing how to ride a bicycle is a procedural memory. Neurologically, procedural memories seem to involve different brain mechanisms than other kinds of long-term memory. Brain damage that destroys a person’s ability to form new semantic and episodic memories may leave the ability to form new procedural memories intact. In other words, it may be possible to learn how to do a new task without any explicit memory of having learned it.
The capacity of LTM is widely believed by psychologists to be unlimited. There is no real way to test this, but available evidence shows that people do store truly vast amounts of information for a very long time. One classic study using yearbooks showed that people were amazingly good at recognizing the faces of their high school classmates after not having seen them for over twenty-ﬁve years. This study found that a fair amount of information was lost within the ﬁrst ﬁve years, but that anything that was remembered for those ﬁve years was still remembered after another twenty, forming part of what has been called the permastore. Another study found that people do surprisingly well on tests of a foreign language or high school algebra some ﬁfty years after having formally studied the subject.
Long-term memory is not always as accurate as we expect it to be, however. People often have extremely vivid memories of highly emotional events (known as ﬂashbulb memories), and the vividness leads us to believe more in the accuracy of these memories than in that of less-sharp memories. This may be a mistake, as a landmark study by Neisser and Harsch demonstrated. The day after the space shuttle Challenger exploded on liftoff, shocking the nation, they surveyed a large sample of college freshmen, asking them where they were when they found out about it, who was with them, and how they found out about it. They then re-interviewed the students three years later, and their results were fascinating: compared to what they recalled one day later, a large majority remembered things incorrectly three years later, and their memories often included surprisingly large errors such as failing to remember the correct location, or remembering a friend’s presence even though they hadn’t known that person at the time. This is especially notable given that almost everyone in the sample was sure, when asked, that they remembered it very accurately.
A similar study of ﬂashbulb memories involved the reading of the verdict in the O. J. Simpson murder trial. A group of college students was asked to describe what they were doing at the moment when they heard the verdict, and they reported their recollections on three separate occasions: just three days later, after ﬁfteen months, and after thirty-two months. At the last interview, almost all of them were sure that they still remembered it quite accurately, but more than 70 percent of their memories were distorted and inaccurate. As in the Challenger study, the most important thing to learn from this study is that the students were completely unaware that their memories had become distorted.
Many factors can inﬂuence the extent to which we are able to retrieve accurate memories, and some of them can be used by students to improve their likelihood of success on tests. A common piece of advice given to students who are struggling is to ﬁnd an opportunity to study in the room in which the test will be given. Based on memory research, this turns out to be good advice, as memory retrieval is sometimes context-dependent. When a memory is ﬁrst encoded and stored, other information present at the same time may also be encoded along with it, so that almost any feature of the environment present at the time of learning can also serve as a retrieval cue. One of the most intriguing studies of this effect was carried out by cognitive psychologists Baddeley and Hitch, along with a group of scuba-trained student volunteers. The students were divided into two groups to learn some information. Half of them learned it on a boat, while the other half were underwater, in full scuba gear, when they learned it. The groups were then divided again to be tested on the material, with half of each group being tested on the boat and the other half getting evaluated on the ocean ﬂoor (special boards that could be written on underwater were used). The results showed that people performed best when the context of retrieval matched the context in which the material was learned, even when that context was the bottom of the sea.
The external environment isn’t the only one that matters, however; sometimes people also encode information about how they were feeling at the time the information was learned. This is known as state-dependent memory, as emotional state can also act as a retrieval cue. If a person is happy and peaceful at the time of learning, for example, he will perform better on a test if he is also happy and peaceful then. Conversely, if a person is extremely angry when learning something, she should enlist a friend to enrage her right before the test.
In a rather extreme test of this effect, people were taught new information after smoking marijuana, and they tended to do somewhat better on a test of the material if they were also under the inﬂuence of marijuana at the time of testing. Similar effects have been documented for caffeine, alcohol, and various other psychoactive drugs. Please note that this is not an endorsement of alcohol or drug use. In all cases, the best performance was shown by those who did not indulge in the substances on either occasion.
Even with context properly matched, of course, people sometimes forget things. Much forgetting is simply a failure to retrieve information correctly, though some is certainly due to simple decay: over time, some information is simply lost. A lot of forgetting, however, is due to interference, meaning that some information in memory interferes with the ability to retrieve other information. There are two kinds of interference, proactive and retroactive. In proactive interference, older information interferes with the ability to recall what was learned more recently. If a student who already knows Spanish takes a French class, for example, she or he may occasionally make errors in class by using a Spanish verb known for years rather than the French verb learned last week. Retroactive interference, on the other hand, is what happens when new learning interferes with retrieval of older memories. If that same student, in speaking with a grandparent who speaks only Spanish confuses her or him by inadvertently using a French expression, retroactive interference is to blame.
Contrary to the common impression people have of their memories as fairly accurate representations of how things really were, many factors actually can cause our memories to be inaccurate or even completely false. The human mind seeks regularity and predictability, and it works to provide it even where it doesn’t exist. Inevitably, any event or situation has elements to which a person does not pay attention, and so they do not encode them in memory. Where such gaps occur, memory is constructive; so rather than consciously realize that the gaps are there, they are ﬁlled in from generalized knowledge of the world, as well as expectations and beliefs. People have schemas in their heads for various common situations, based on experiences, and they ﬁll in the details they missed according to these schemas.
Consider your kitchen schema, for example: What are the things one would expect to see in a kitchen? For most Americans, and for the college student research participants in the next example, a sink was in the list somewhere. A group of students was made to wait in a small kitchen before being led into a lab for the experiment they had signed up for. Upon entering the lab, they found that the experiment had already begun, as they were asked a series of questions about what they had observed in the kitchen while they waited. Most students endorsed the statement that a sink had been present, and were then quite surprised to hear that the room didn’t contain one. In a similar study, students waited for several minutes in a room that they had been told was a graduate student’s ofﬁce and were later asked to recall everything that was in the ofﬁce. Most of them speciﬁcally remembered seeing books, despite the fact that the room did not contain a single book. Their expectations of what belongs in an academic ofﬁce overruled the actual evidence of their senses. Again, note that the process of ﬁlling in these gaps is unconscious, and the only reason these people would ever know that their memories were inaccurate was because the researchers told them so.
The constructive nature of memory is an especially important (and dangerous) issue where eyewitness testimony in a courtroom is concerned. The use of eyewitnesses is predicated on two notions: that they will testify truthfully and honestly, and that their memories of what they witnessed will be accurate. Leaving aside the issue of whether witnesses tell the truth as they know it, witnesses can sometimes make serious mistakes without being aware of it, and these errors can have serious consequences. As of late September 2004, 151 death-row inmates in the United States have been released based on DNA evidence that showed they were not guilty of the crime of which they were convicted. In the vast majority of cases, the conviction was based primarily on eyewitness testimony. Like everyone else, an eyewitness can only remember what was perceived and can only perceive what was attended to. There are limits, imposed by the existence of memory gaps, to how accurate their testimony can possibly be, and a large body of evidence shows that those gaps may be ﬁlled in by new information about the crime, and a witness’s memory can even be altered by the wording of a lawyer’s question.
In a classic experiment by cognitive psychologist Elizabeth Loftus that has been widely replicated, subjects saw a ﬁlm of a minor trafﬁc accident and were then asked a question about how fast the cars were traveling when they made contact. The only difference between the groups of subjects was the verb used in the question. People who were asked how fast the cars were traveling when they smashed into each other estimated a signiﬁcantly higher speed than those who were asked how fast the cars were traveling when they hit each other. When interviewed a week later, people who had heard the word smashed also provided a more violent description of the accident than the others. Loftus calls this the misinformation effect. In one variation on this experiment, the ﬁlmed accident was a very gentle collision, with no visible physical damage to either car, but at the delayed interview, people in the smashed group frequently recalled seeing broken glass on the pavement, although none was visible on the actual ﬁlm. In another variation, a stop sign was visible in the ﬁlm, but some of the subjects were asked questions that mentioned a yield sign instead at the delayed interview. These subjects frequently included the nonexistent yield sign in their recollections.
The use of young children’s eyewitness testimony is especially problematic, given what is known about their memory and how susceptible it is to distortion. In the 1980s, several high-proﬁle trials showed just how serious these problems can be. The McMartin family, who ran a preschool/day care center in California, became the defendants in the longest and costliest criminal proceeding in U.S. history, when well over 100 counts of child physical and sexual abuse were levied against them. The accusations included a number of fantastical allegations, including a secret room children were taken to through hidden tunnels to be abused, but no secret room or tunnels were ever found. Other children testiﬁed that they had arrived at the secret room by being ﬂushed through the toilets, but were cleaned up upon their return. In a similar case involving the Little Rascals day care center in Edenton, North Carolina, accusations included a secret basement aquarium ﬁlled with trained sharks, used for disciplinary purposes. At one point in the trial, a child alleged that the entire class had been taken on a boat out into the ocean, where a child may or may not have been thrown to these trained sharks, which had been released for this purpose. There were also allegations about the disciplining of a child by hanging him upside down from a tree and setting him on ﬁre. This child later testiﬁed, absent any evidence of burns.
In a New Jersey trial, a young teacher named Kelly Michaels was accused of, among other things, playing the piano naked, having children eat feces and drink urine, having children lick peanut butter off her naked body, and of amputating several boys’ penises (presumably this last claim would have been fairly easy to dispel with physical evidence), as well as occasionally killing babies and having the children drink their blood (again, this would presumably have left some physical evidence, not to mention reports of missing babies).
What the above cases all have in common (other than the patent absurdity of many of the charges) is that all resulted in convictions, which were only overturned on appeal years later. The defendant in a similar case, Gerald Amirault, who was accused of raping children with kitchen knives (again, no physical evidence of the “crimes” was ever introduced), was ﬁnally released on parole by the state of Massachusetts in April 2004, a full eighteen years after ﬁrst going to jail. One reason that he wasn’t released sooner was his refusal to undergo sex-offender counseling while in prison, since he wasn’t actually a sex offender. The state Board of Pardons had actually recognized the ﬂaws in his prosecution and recommended his immediate release years earlier, but the governor at the time refused to endorse the pardon.
How could sensible judges, prosecutors, and juries have let things go this far? In each case, a major part of the prosecutor’s argument was a question about the children: Why would they lie about this? The unspoken assumption was that if the children were not lying, then everything they were saying must really have happened, and the absence of physical evidence simply meant that child rapists are unusually clever at covering up their misdeeds. The truth is of course subtler: the children did not lie, but what they alleged in court was not true either. To the person doing the remembering, false memories do not differ from real ones, except that the false ones are often a bit “fuzzier.” The children were telling the truth as they understood it, rather than either lying or telling the objective truth. Research on young children’s ability to remember a traumatic event is especially relevant here. Obviously, it would be highly unethical to conduct an experiment in which children were randomly assigned to be abused, but there is another experience that preschoolers normally go through that involves an unfamiliar adult making them take off their clothes, touching them in uncomfortable ways, and sometimes causing pain: a trip to the doctor’s ofﬁce.
In a research design that has now been replicated numerous times, a pediatrician and a nurse ﬁll out a checklist for each child who comes for an examination, listing everything that was actually done during the examination. Meanwhile, the children are interviewed immediately after the examination, and again after two to six weeks. An interesting pattern of results has emerged in these studies: younger children are far more prone to memory errors than older children, and the type of question asked matters a great deal. When they are asked open-ended questions, like “What did the doctor do?” preschool aged children provide very little information spontaneously, but what they say tends to be accurate. When speciﬁc, yes/no questions are asked about their experiences, however, their accuracy drops to chance levels (a 50/50 shot at answering correctly, in other words) both for questions about what really happened and for questions about things that did not happen, even when the interview occurs immediately after the examination.
Another factor in the questioning of children that is known to inﬂuence children’s testimony, present in all of the cases described above, is repeated questioning. When a very young child is asked the same question repeatedly, even when the answer is a ﬁrm “No,” that child will eventually change his answer to satisfy the adult, because the adult is conveying quite clearly that the answer is wrong. The misinformation effect was also a factor in all of the cases. Most of the bizarre details came originally from the adults, and the children eventually incorporated them into their own testimony after hearing them repeatedly, with their own imaginations taking over to ﬁll in gaps.
Another factor in the trials was stereotype induction, which is what happens when children are repeatedly given prejudicial information about a person (“He’s a bad man”) and later incorporate it into their memories of that person. In a classic study by Stephen Ceci and his colleagues, a guest visited a preschool classroom several times and told stories about his clumsy, accident-prone friend named Sam Stone. After several weeks of this, another guest visited the classroom. Introducing himself as Sam Stone, he simply greeted the class, stayed a while, and then left. The next day, the teacher drew the children’s attention to a broken toy and a damaged book and asked them who had done it. When the children were later interviewed and asked what happened when Sam came to visit, many of them described his breaking of the toys, often accompanied by comments like, “He’s so clumsy!”
Stereotype induction and the misinformation effect are examples of a larger problem that is often a factor in adult false memories as well: source misattribution. When a piece of information is stored in memory, information about the source of the information is also often encoded, but that piece of the memory is often not encoded well, so it fades much faster. The result is that a piece of information may sound familiar, but we no longer remember where we heard it. This is a familiar experience to anyone who has ever known an answer in a trivia game or on a test without any idea of why they know it. As with other memory gaps, people tend to ﬁll these unconsciously with a likely source, rather than realizing that they don’t know where the information came from.
A famous example of this comes from a speech in which President Ronald Reagan told a “true” story he had heard from a World War II veteran, in which the tail gunner in a damaged bomber was hit and couldn’t eject with the rest of the men, so an ofﬁcer remained on the plane so the gunner wouldn’t die alone. The story seemed vaguely familiar to many listeners, which may be because it is a scene from the war movie On a Wing and a Prayer. Although cynics have accused Reagan of simply lying, it actually seems likely that he knew the story but no longer remembered where it came from, and he simply followed the usual human tendency to ﬁll in the gap. When a person no longer recalls the source of a piece of information, they often attribute it to a source anyway. In the misinformation studies described above, subjects associated the vague impression that something was “smashed” with the ﬁlm they had seen, but didn’t remember that it came from the question rather than the ﬁlm.
Adults are just as susceptible to these effects as children, with the added disadvantage of great conﬁdence in the accuracy of their own memories. In the 1980s, during the same period as the assorted multiple-offender/multiple-victim child abuse cases described above, came a near-epidemic of cases in which adults accused others of child abuse that had allegedly occurred long before, sometimes decades earlier, and which had never previously been reported. In these cases, the memories had allegedly been repressed and had only been recovered after many sessions with a therapist. The therapists frequently used hypnosis and guided imagery, along with other suggestive techniques, and after weeks or months of treatment, the patients began to believe in the reality of things that they had never previously remembered or reported. Based on such testimony, and in the absence of any corroborating evidence at all, a number of cases made it to trial, and some resulted in convictions. The problem became so widespread that a group of psychologists founded an organization, called the False Memory Syndrome Foundation, to bring together resources for families who had been falsely accused. On their Web site, they give the following deﬁnition of false memory syndrome: “[A] condition in which a person’s identity and interpersonal relationships are centered around a memory of traumatic experience which is objectively false but in which the person strongly believes” (http://www.fmsfonline.org/?faq=faq).
The foundation has been instrumental in changing public attitudes (and the behavior of judges) where such cases are concerned, and has inevitably suffered a backlash from those clinicians who believe strongly in the Freudian construct of repressed memory. Cognitive psychologists in general, and Loftus in particular, have been accused of lacking sympathy for the victims of abuse, as well as of ignoring the evidence for repression of memories. Childhood sexual abuse is a serious problem, and it is widely recognized that the problem has long remained underreported. People who have been sexually abused deserve our sympathy and any help that we can give them, and nobody associated with the False Memory Syndrome Foundation disagrees with that.
It is important to note that this has become a political controversy rather than a scientiﬁc one, as the scientiﬁc consensus on repression is fairly clear. A review of more than sixty years of research on repression concluded that there is no scientiﬁc evidence for the phenomenon to date, and, as the APA (1995) states, “the reality is that most people who are victims of childhood sexual abuse remember all or part of what happened to them.”
- American Psychological Association. Questions and Answers about Memories of Childhood Abuse. Washington, DC: American Psychological Association, 1995;
- Bruck, M., Ceci, S. J., et al. “External and Internal Sources of Variation in the Creation of False Reports in Children.” Learning and Individual Differences, 9(4) (1997): 289–317;
- Ceci, S. J., and Bruck, M. Jeopardy in the Courtroom. Washington, DC: American Psychological Association, 1995;
- Schacter, D. L. The Seven Sins of Memory: How the Mind Forgets and Remembers. New York: Houghton Mifﬂin Mariner Books, 2002.