jim.shamlin.com

Chapter 5: Memory Models

Memory is a mechanism by which we retain information from the present for use in the future (storage), as well as that by which we access information stored in the past for use in the present (retrieval).

As seen in theprevious chapter, memory is a critical component in perception: we recognize items because we have encountered them before, and remember them. It is also critical to cognition because though is based on previous experience: we do not imagine a solution or plan a course of action based only on what we see before us at a given moment, but on the lifetime of experience we have accumulated.

As seen in theprevious chapter

Measuring Memory

In studying memory, researchers have devised various tools that require participants to remember information - to interpret, store, retain, and retrieve it in various ways. In many ways, the design of the experiments has shaped our understanding of memory, so the author wishes to consider them before wading more deeply into the topic.

Recall versus Recognition

There are two basic applications of memory, recall and recognition, which are differentiated by the strength of an immediate cue. A person who can bring something to mind without a view is said to be able to recall it, whereas a person who remembers seeing something before when they are shown it again is said to recognize it.

(EN: Here there is room for argument in the precise distinction, as all memory is cued by something and it is only the degree of specificity in the cue that distinguishes one from another. Asking "have you ever heard of brand x detergent" is said to be recognition because the brand is mentioned, and "what is a brand of detergent" is said to be recall because the brand is not named in the question - though the word "detergent" serves as a cue to recall the associated brand name. It is a significant difference, but a rather subtle one.)

Various types of recall/recognition tasks are described:

Implicit versus Explicit Memory

All of the tasks above are explicit memory tasks, in which the subject is aware that he needs to remember something and attempts to bring it to mind. There are also tasks of implicit memory, in which things must be remembered in order to perform a task.

In the latter instance, we recognize that implicit memory is necessary to performing basic tasks. The task of carrying on a conversation requires the implicit memory of the meaning of words - we are not attempting to retrieve definitions when we listen or speak, but our ability to do so enables us to understand what is being said and to express ourselves fluidly: conversation fails when a participant cannot remember the definition of a key term.

The difference between explicit and implicit tests is the way in which a subject is "primed" to remember something. IF asked a direct question, subjects are leveraging explicit memory; if asked to perform a task, subjects leverage implicit memory and their success or failure indicates where knowledge is not recalled.

Traditional Model of Memory

There are various models of memory that are in present employment.

The notion of short-term and long-term memory, which distinguishes information that is stored fin a readily-accessible manner for immediate use from information that is stored for a longer period in a medium from which it is more difficult to retrieve goes back to the nineteenth century (William James).

In the mid-sixties, Richard Atkinson proposed a refined model that differentiated the sensory store (a small amount of recently-gathered sense data), the short-term store (somewhat longer but of relatively limited capacity), and long-term store (very large capacity and possibly infinite duration of storage).

It is admitted that these models are hypothetical constructs - theoretical concepts that are not readily measurable or observable and have not been associated to specific regions of the brain - but instead are notions that are used conversationally based on observation. Even so they have been useful in describing the way in which memory seems to function.

Sensory Store

The sensory store is the initial repository of information received from the environment, along with our perceptual interpretation of it. Sensory data is not stored in its raw form, but is organized into "icons" that are mental representations.

Consider the example of a person who "writes" something by tracing letters in the air with a fingertip: the words are never actually seen, but interpreted from the motions that suggest the shape of letters, and what registers in our mind is not the sequence of motion, but the letters themselves.

The notion of an iconic store originated with a dissertation by Sperling (1960), in which it was considered how much information gathered from the immediate experience is applied. Among his findings is that subjects are able to recall information, such as four of letters and numbers, if they are flashed on a screen for as little as 50 milliseconds (0.05 seconds)

The experiment was refined to present three rows of four characters each, asking participants to recall only one line (they were not told which), and discovered that participants were generally accurate if asked immediately after the flash, and could recall nine of the twelve characters, but after a delay of several seconds their recall was decreased to four or five. This research suggests that sensory information fades very rapidly from memory.

There is some criticism of his research because of the small amounts of time involved - in effect, that the subject may have forgotten some of the characters during the time that he was speaking them, as well as the suggestion that they were reading them from an optical after-image rather than recalling them from memory. In spite of these flaws, his research still seems to suggest that short-term sense memory exists, has a very limited capacity, and is highly perishable.

Further experimentation refines his findings by suggesting that there is a limited capacity of storage (when the number of characters is increased to 16, participants' immediate recall dropped to 75% ) and its limited duration (participants can recall the characters they just saw, but not the set just before, indicating the new set erased the old).

Short-Term Store

The short-term memory, to which e have introspective access, is reckoned to hold a larger amount of information for a longer amount of time than the sensory store. Definitions of "short term" vary, but are generally around half a minute, during which we have very accurate recall of recent events.

How much information can be held in short-term memory is open to debate. Miller (1956) observed that it seems to be about seven items, plus or minus two. (EN: Miller's "rule of seven" is often cited, but it stands to note the article is an observation, considering various kinds of experiments, in which the design of the experiments tend to use groups of five, seven, and nine in various ways, and he infers that this is significant, but does not make a case for the reason it is so, nor conduct any original research to substantiate his observation.)

Further, there is some clarity on what an "item" is, as an effect of constructs. That is to say that five random characters (U8IK3) is considered to be five objects, but five characters that create a word (MOUSE) is remembered as a single unit. Further, a sequence of words that describes a single concept ("large screen television set") or the qualities of a given object ("a white shirt with a dark blue ink stain on the pocket") are also consolidated into a single unit.

There's also evidence related to linguistics, that items with shorter names (cat, dog, fox, car, tree) are easier to recall than a list of items with longer names (umbrella, telephone, electrocardiograph, watermelon, firefighter)

These experiments, while interesting, are tainted by their design: that it is not merely memory that is being texted, but perception and verbal skills: to recall a list of items, the subject must first understand what the object is (perception) and then be able to speak its name (verbal).

To remove the influence of these factors, additional experiences were conducted (Cowan 2001) that showed participants amorphous shapes and ask them to choose a matching shape from an array of options. This demonstrated that participants were able to hold roughly four patterns in memory. However, even this experiment had impurities in that memory replied upon the ability to recognize shapes and patterns: the more complex the pattern or the more similar the options, the less accurate their responses, which can be attributed to processes other than memory.

In spite of its flaws, this research has made various suggestions about the breadth and depth of short-term memory, and while the results cannot be accepted as entirely precise, they do support the general observation that short-term memory has a relatively limited capacity and duration.

Long-Term Store

The long-term store of memories persists information for very long periods of time, perhaps indefinitely. Information is retrieved from long-term storage on a daily basis to bring to mind information that was gained an hour ago, or even decades ago. Recognizing people, knowing where we put things, knowing where we are to go on any given day of the week all draw upon long-term memory stores.

It is not known for certain how much information is contained in long term memory of for how long it is persisted. There hasn't been an experiment designed that can measure the capacity, nor has neuroscience been sufficient to fully assess the amount of information in storage. But we do know it is quite remarkable in terms of the volume and persistence.

Work with epileptics and Alzheimer's patients demonstrates that people in their eighties can vividly recount incidents from their early childhood that has not been recalled for many years. This suggests that memories are permanent, and it is only out ability to retrieve them that makes them seem less so.

There have been some objections to this studies: that there are a small number of reports among a large number of patients who were able to recall early childhood memories, and some conjecture as to whether the recollections are actual memories or fabricated incidents.

Other evidence is less disputable. Some researchers (Bahrick et. al, 1975) used high school yearbooks to prompt recollection, and found that participants were able to recall the names of classmates they had not seen in decades.

Levels-of-Processing Model

The author refers to the levels-of-processing model as "a radical departure" from the stages-of-memory model, in that it proposes that there are not distinct stages but a continuum in terms of the depth of coding. Specifically, there is no part of the brain devoted to recent memories as opposed to distant ones, but the brain contains all information as a single repository, with some data closer to the surface than other.

Moreover, data does not exist in free form, but is accessible by relationships - one memory will trigger another. Some experimentation was done with word-association to demonstrate that words that are closely associated in memory (dog and cat) are easier to remember than those that are arbitrarily paired for the sake of experiment (dog and umbrella).

A separate experiment showed better recall when words are related to the subject - people recall adjectives that they feel describe themselves better than those that do not. The implication is that each person is receptive to stimuli that are relevant to himself, and access this information more readily than general information.

There is some criticism of the LOP theory, mainly that it is somewhat circular in its logic: a memory that is harder to retrieve is considered to be deeper simply because it is harder to retrieve and vice versa. There's also the observation that word-pairs that rhyme are even easier to retrieve than terms that are closely related (chair-hair is a strong pairing even though there is no connection between the concepts). As such it's questionable whether the "depth" of a memory is as relevant to recall than its relationship to present stimuli.

Integrated Model of Working Memory

The working model is the most widely used and accepted in the present day. This theory maintains that certain information is held in a space similar to the processing memory of a computer, while other memory is stored in a separate space similar to the hard drive of a computer, where it is stored for retrieval but not currently in use. The data in working memory comes from two sources: it combines stimuli from the external environment with data retrieved from storage for present use.

This seems similar to long- and short-term memory, but the distinction between them is not the time at which the memory was created, but the time at which the memory was last used. As such, data dredged up from deep storage is as fresh in the mind as something that was seen only seconds ago.

One model of working memory (Baddeley 1990) defines specific elements of the working memory system:

While there are certainly other theories of working memory (such as Craik or Lockheart), this dovetails with many of them in the way in which processes occur and mental tasks are orchestrated.

Whereas the three-stage model emphasizes the receptacles in which data is stored, the working memory model emphasizes the processes in which it is used - the reception, coding, processing, retrieval, and storage of data in the form of memory.

A few different analogies are explored. Consider the loading dock, workspace, and storage racks of a warehouse; or consider the camera, editing deck, and tape library of a film studio. In each metaphor, there is one system for capturing new data, another system for processing it, and a third for storing it - similar to perception, cognition, and memory.

The approach to working memory also has the support of neuroscience in that certain sections of the brain show increased electrochemical activity when memory is accessed - primarily along the borders of the major divisions of the brain: the supplementary motor area between the front and center sections of the brain, the superior parietal area between the midbrain and the hind brain, the posterior parietal area between the hind brain and the temporal lobe, and Broca's area between the frontal and temporal lobes.

The pattern of electrical activity differs depending on the nature of information involved, with greater activity in one of the areas depending on whether the information in question is visual, audio, or motor/somatic. Some further detail is provided about the localization of activity based on the nature of the mental task.

Multiple Memory Systems

The working memory model is consistent with the theory that multiple systems are involved in the storage and retrieval of information, and also suggests that there is communication among these systems, as various sensory details of a significant event can be remembered easily, but facts out of context are not as easily retained: a person may remember the color of the shirt he was wearing on the day his first child was born but not the color of the shirt he wore two Tuesdays ago.

Based on those findings, it is proposed that there are two different kinds of memory: semantic memory that stores general knowledge and episodic memory that stores knowledge related to specific events. It's noted that most laboratory experiments are designed to test semantic memory rather than episodic.

Subjects with lesions on the frontal lobe have a peculiar effect on associating memory to episodes. They recall that a particular stimulus was experienced, but cannot associate it to a specific event. (EN: This also seems to be common in normal individuals, who can recall details out of context). This observation does not substantiate the notion that the two are entirely disconnected systems, but it does suggest that not all semantic information is correlated to an episode.

Neuroscience experiments have also observed asymmetry in hemispheric encoding and retrieval. There is greater activity in the left prefrontal lobe for tasks requiring retrieval of semantic information, and more activity in the right than the left when retrieving episodic details. This observation supports the notion that the two forms of retrieval occur in different areas of the brain - though there are exceptions, as the specific task of remembering a list of terms, which seems primarily semantic, cases more activity in the right side of the prefrontal lobe (EN: Though arguably, the items are being recalled in the context of a list, just as episodic details are recalled in the context of an event.)

A third distinct memory system handles procedural memory - which is the knowledge of tasks performed in a given sequence. The central region of the cerebellum seems to be involved in this type of memory.

Yet another approach considers the difference between declarative memory, in which a person is attempting to recall a specific detail, and implicit memory, in which the person happens to recall a specific detail in the process of bringing something else to mind.

Connectionist Perspective

The connectionist model, also called the "parallel distributed processing" model, maintains that knowledge is based on connections among various nodes, not in each individual node. That is, each memory contains various bits of information, and when a memory is triggered, other memories related to each of its bits is triggered - e.g., the sight of a blue square touches on memories that include the color blue and memories that include a square, since the color blue is associated to water it also triggers every memory associated to water, etc. While connections are seemingly infinite, the processing stops at the limits of working memory, which functionally results in memory being filled with those things most closely related to the original qualities of the stimulus.

The triggering stimulus therefore has a "priming effect" on cognition. Such that the sight or mention of a color has an overall effect on mood by virtue of all of the emotions associated to memories of that color.

Models based on connections seem top have some appeal in that they address aspects of thought that other theories do not. Chiefly, the model suggests the ability of the mind to perform multiple simultaneous operations rather than a single thread, and that the notion of "working memory" does not apply to a separate memory space, but merely the activated portions of long-term memory.

However, it's been noted that connectionist models have thus far failed to provide clear predications or explanations of recall and recognition memory that follows a single episode or exposure to a semantic datum in human experimentation - though it naturally lends itself very readily to computer simulation.

Memory in the Real World

Some researchers are looking into studying memory in the natural environment, or at least making the laboratory more natural, based on the notion that memory is highly ecological, and that the contrived nature of experimentation provides little valid and practical information.

While there is naturally criticism that such field experiments are largely anecdotal (hence unscientific) due to the lack of a controlled environment, the laboratory approach fails to study real phenomena that are in any way relevant to the everyday use of memory, and are therefore irrelevant and misleading.

Further, even laboratory experiments lack true control, in that memory derives from the experiences of the subject throughout their lifetime, hence researchers can control the laboratory environment in which memory occurs, but cannot control the randomness and idiosyncrasy of the memories of the subject they bring into the lab.

The author predicts a tend in favor of field research as researchers become increasingly interested in the function of memory. This may take the form of field experiments and observation, or it may attempt to replicate behavior from the real world in the laboratory setting.

Exceptional Memory and Neuropsychology

Reviewing case studies with people who have abnormal memory (enhanced or deficient) provides "some interesting insights" into the nature of memory in general.

Outstanding Memory

The term "mnemonist" has been coined to describe individuals who have outstanding memory - or more aptly, an outstanding ability to recall a large number of granular details over a long period of time.

One interesting case (Lorin) is an anonymous individual (an entertainer who used his memory to amuse audiences) who sought out a psychologist to test his memory: this patient had the ability to accurately recall long lists of words over an extended period of time, not merely within a laboratory session, but over a number of years. Lorin recorded sessions with the subject over a thirty-year period, and found him able to recall lists from sessions 15 or more years in the past. While the patient did admit to having certain tricks to memorization (such as associating words to visual images, some of them quite abstract, such as using the number 3 to recall a gloomy person). Much of his recall was quite effortless and unintentional.

Another case (Hunt) involved a patient who could remember long strings of numbers, and who indicated he was able to do so by converting them into dates and recalling what he was doing on that day. Still another (Ericsson) indicated the ability to remember strings of numbers by relating them to running times for different races, either actual or plausible. The latter subject showed improvement over a two-year period and was eventually able to recall a string of over 80 digits, though he was foiled by any string that could not be converted into sequences that were plausible running times.

While we may envy these abilities, it's also noted that patients with extremely enhanced memory also find that they have difficulty in other areas: they can remember information but cannot process it, and they have considerable difficulty in social situations. In essence mnemonists are likened to parrots, remarkable for their ability to memorize and repeat, but rather stupid and obnoxious for their lack of ability to do anything else.

The author mentions "hypermnesia," as the ability to recall many facts and details, and the theory that nothing we experience is ever lost or forgotten, but is resident somewhere in memory if only it can be recalled. The relevance of this theory to mnemonists is that the normal brain regards a majority of sensory information as unimportant and insignificant - it may be perceived and even stored, but it is irrelevant to anything and therefore lacks the ability to be retrieved.

For example, you may recall something that was said, but not who said it or what color of shirt they were wearing at the time - this does not mean you did not perceive these details, just that they were not relevant to the conversation, and while they are stored in memory they are regarded as irrelevant to the information of primary interest. Though they may persist as cues that cause us to remember the conversation.

For the hypermnesiac subject, these filters are dysfunctional and they are unable to focus solely on the relevant information - and often lose track of the relevant details for their attention to a broad range of unrelated stimuli. For example, Lorin's subject reported difficulty in following conversations, as the sound of speech became a blur as he noted physical details of the environment.

Deficient Memory

Amnesia is a condition in which memories are either lost or irretrievable. The most familiar form is retrograde amnesia, in which subjects have a loss of memory as a result of a traumatic event, which causes memories formed during or immediately prior to the event, Even a relatively mild concussion can cause some degree of retrograde amnesia.

One case study (Russell 1946) involved a 22-year-old groundskeeper thrown from a motorcycle in 1933, after which he seemed to have normal mental functions except that he recalled the present year as 1922 and considered himself to be a schoolboy, such that eleven years of memory seemed to be erased. His memory gradually returned over the following several weeks, until the point he regained his place in time and most of recent memory up to a few minutes prior to the accident. In some instances, the events before a trauma are completely lost.

A more common condition is infantile amnesia, the ability to recall events that occurred when we were very young. It is perfectly normal to have very little memory of anything that occurred before the age of five years, and extremely rare to have any memory prior to age three.

Anetrograde amnesia is the loss of the ability to create memories, evidenced in patients who have suffered a trauma and are unable to recall events that occurred afterward. One particular case-sturdy is a patient who had brain surgery to remove lesions that were causing epilepsy, and who had excellent recall of events prior to the surgery, but none afterward. This form of amnesia also presents itself in Alzheimer's patients, who can recall their childhood vividly by cannot recall events of the previous day.

One of the general insights of research into amnesia is the distinction between explicit and implicit memory. Explicit memory, or the ability to recall specific events, is often severely impaired whereas implicit memory, the knowledge to perform a task or associate details to an object, is generally not. A bit more detail is provided on the experimentation that supports this theory.

Similarly, amnesiacs show little impairment in procedural knowledge but significant degradation of declarative knowledge - i.e., they are able to perform a task, but cannot describe how the task should be performed if they are not actually doing it.

Neuroscience has also been informed by studying amnesia patients. The location of lesions in the brain and deviations between the electrochemical activities of amnesiacs and normal patients suggests that amnesia has significant physiological features.

Through observational studies, it is found that amnesia has numerous causes that can be characterized by physical damage to the organ or a disruption of its electrochemical processes. Lesions in specific parts of the brain result in specific types of amnesia, though the relationship is still general based on the present level of evidence.

It is also noted, significantly, that specific regions of the brain are not strictly associated to the functions of memory, but that various parts of the brain interact in the formation and recollection of memory. However, it is often observed by contrast that certain parts of the brain show a high correlation to specific types of memory: a person with a lesion in the visual cortex is likely to have greater difficulty recalling visual stimuli than that of other senses, but there does not seem to be clear parity between the effects of long-term and short-term memory as a result of damage or dysfunction to a particular area.

The author then turns to Alzheimer's disease, which often presents dysfunctions related to memory along with a host of other impairments. This presents some difficulty to researchers in diagnosing a connection between the various causes of a given symptom - where a memory cannot be recalled, it may be because of amnesia or dementia.

The definitive diagnosis of Alzheimer's is possible only after the death of a patient, at which time a dissection of the brain will reveal "plaques and tangles" that are not evident in normal patients, but by which time the damage is extensive and the analysis of the progression of the disease is highly theoretical.

Alzheimer's is strongly correlated with age. Between ages 70-75 it is evident in 1% of the population, which grows to 6% by ages 80-85. By another study, 30% of individuals show some symptoms of the disease by age 70, and more than 50% by age 80.

It's also noted that there is a correlation to genetics: people who have the genetic mutation always develop the disease, and symptoms can become evident even by age 20. While the feature sis common to early-onset Alzheimer's, there still remains no reliable evidence of the causes of late-onset Alzheimer's.

Role of the Hippocampus and Other Structures

Early attempts at localizing the function of memory within the brain were unfruitful. Psychologists have discovered many cerebral structures that seem to be involved in memory, but at present have yet to identify a central structure or even define localizations of specific thoughts, events, or stimuli within the organ.

It's presently reckoned that sense memory is stored in the various locations that process sensory stimuli: visual, speech, and olfactory sensations seem to be stored in the areas of the cortex responsible for perception and processing of these senses.

In addition, the hippocampus and some related nearby structures appear to be important for the integration and consolidation of memory in separate parts of the brain, and is believed to coordinate the formation of long-term memory, as well as in complex learning and the coordination of multisensory stimuli.

There is also support for the theory that certain neurotransmitters disrupt or enhance memory storage. Bot serotonin and acetylcholine seem to enhance the level of activity associated with memory. High concentrations of these chemicals are found in the hippocampus of normal people, and low concentrations in Alzheimer's patients.

The consumption of alcohol has also been shown to disrupt the activity of serotonin, impairing the formation of memory - hence the phenomenon of the blackout drunk. Extensive use of alcohol over a long period of time is a contributing factor to Korsakoff's syndrome, which is "a devastating form of retrograde amnesia" that is evidenced in extensive damage to the thalamus and hypothalamus, as well as some damage to the frontal and temporal lobes.

It's also reckoned that adrenaline and other hormones that stimulate the release of glucose in the brain tend to sharpen memory, which is supported by the enhanced memory that occurs during emotionally-charged events, whether the emotions are positive or negative. Such events also are associated to increased activity in the amygdala. It's also noted that there is a gender difference in emotional memories, with women recalling emotional events more readily than do men.