Cognitive Science: An Introduction/Human Memory

Memory
Memory is important to all complex systems. Without memory, a system or a complex organism would have to relearn all procedures and facts that have been useful for performing relevant tasks in the past. Without memory, animals would not be able to locate their hidden cache of food. People would not remember important information such as their address, their names etc. Memory is also important for planning for the future. For example, we need memory to remember our schedule and what we need to do this weekend yet at the same time, we do not remember everything we have learned at any single moment in time. What then are the characteristics of memory, and how does cognitive science define it?

What is memory?
Memory is usually thought of as a storage system that retains information for a period of time. From a computational approach, in addition to retaining information, memory also stores There are three cognitive processes that are closely associated to the study of memory:
 * instructions to be executed
 * the results of any computations that are executed
 * 1) Encoding: The process by which information is reformatted into a representation that facilitates transfer into memory.
 * 2) Learning: The process by which information is transferred to and retained in long-term memory
 * 3) Retrieval: The process by which information retained in long-term memory is accessed and made available.

Memory has a variety of uses. Minds can use things they know about the world and how things are perceived to aid with perception. For example, if you see someone's head poking out of a window, your mind can use its knowledge of the human body to conclude that it's probably a whole person, rather than just a head. It affects our emotional responses and preferences to things in our environment, based on the feelings we associate with things. It allows us to report things that have happened to us, and to use the past to help plan for the future and make better decisions.

Why can't we remember things exactly? There are two possible answers. The first is that there is some adaptive advantage to it. The second is that it is impossible given the nature of how our brain evolved. The few people with perfect photographic memories had trouble using them, because they were less meaningful. We will discuss the neuroscience reasons for memory errors later.

What is immediate memory?

 * Immediate memory can be thought of as the memory system that can retain information for a few seconds while people perform other activities.
 * Immediate memory was eventually referred to as short-term memory, when Miller (1956) postulated that there were limits to the amount of information capable of being held in immediate memory.

SENSORY MEMORY
Sensory memory is an extremely short-term storage system that is capable of holding vast amounts of information captured from the senses without any manipulation for short periods of time.

ICONIC MEMORY

As early as the 1950s and 1960s, psychologists studying visual memory were faced with a puzzle described by George Sperling

"When complex stimuli consisting of a number of letters are tachistoscopically presented, observers enigmatically insist that they have  seen more than they can remember afterwards, that is report, afterwards" (Sperling, 1960, p1)

This implies that while people could see more than they reported, they were unable to report everything they had seen. Thus begun an area of research that attempted to identify the properties and limitations of memory that contributed to this observation and how to study the contents of this visual memory system, that seemed able to temporarily store vast amounts of information for a brief period of time.

Iconic memory or visual sensory memory is a short-term memory system that briefly holds visual information captured directly from our eyes - the primary visual organ. Information can be held in iconic memory for up to 500ms.

Iconic memory was first demonstrated using two methods: (1) whole-reports and (2) partial reports of arrays (George Sperling, 1960) Participants are shown a display of letters in a 4 by 3 array for a brief display of time: for example they could have seen this array for 400ms. Q  S   E   R      F   T   Y   P      Z   X   V   N In whole-reports of arrays: participants are shown the array then asked to report the entire letter display: typically most people can only remember 4 or 5 letters. In partial reports of arrays: participants are shown the array then required to report the letters of a single row in the array. The row that participants are required to report is cued with a tone. So for example: If they heard a high tone, there are required to report the letters in the top row. Typically, participants are able to report 3 to 4 letters in the row (75% to 100% accuracy). Sperling inferred that this accuracy rate of (75% to 100%)for the partial report condition could by extension suggest that participants would have been able to report 75 to 100% of all information in their iconic memory.

Based on his findings, Sperling proposed that a large amount of visual information is initially available to observers, but the majority of this visual information decays rapidly until eventually only the amount of information roughly equivalent the span of immediate memory is remembered and can be reported.

ECHOIC MEMORY

Echoic memory or Auditory sensory memory is a short-term memory system that briefly holds auditory information captured directly from our ears - the auditory visual organ.

Ulric Neisser coined the term echoic memory and argued for its existence as follows:

"Perhaps the most fundamental fact about hearing is that sound is an intrinsically temporal event. Auditory information is always   spread out in time; no single millisecond contains enough information to be very useful. If information were discarded as soon as it    arrived, hearing would be all but impossible. Therefore, we must assume that some 'buffer", some medium for temporary storage, is    available in the auditory cognitive system. (Neisser, 1967, p 199-200)

However, the auditory system also does not retain raw auditory information for an extensive period of time, as confusion will result if all auditory traces were to be retained permanently. Auditory stimuli such as spoken words and numbers consist of sounds spread out over a period of time (as compared to visual information spread out over space in the case of iconic memory). Hence, a slightly different paradigm is required to study echoic memory.

Darwin, Turvey and Crowder (1972) used a slight modification of Sperling's whole and partial report condition to investigate the properties and limitations of echoic memory. Participants wore headphones and listened to speech recordings of three lists that seemed to come from three different locations: left ear, right ear and a list that was played into both ears but seemed to come from the middle of the head. Each list consisting of three items made up of letters or numbers. So a participant could hear "B 2 F" coming from the left ear, "3 J 4" from the right ear, and "M 5 Q" that sounded like it came from the middle. In the whole-report condition, participants were required to report all letters and numbers that they had heard.In the partial-report condition, participants were cued visually as to which list they were required to report. In the addition, the cue could also appear at an interval of 0s, 1s, 2s or 4s after the end of the list.

In the whole-report condition, slightly less than 5 out of 9 items were correctly identified when they were cued to recall items immediately after presentation of the last item (i.e. after a delay of 0s). This dropped to less than 4.5 out of 9 items after 4s. In the partial report condition, more than 50% of the items were correctly identified in the cued ear after a delay of 0s. This decreased to approximately 50% (or 4.5 out of 9 items) after a delay of 4s.

These findings suggest that the capacity of echoic memory is not as efficient as iconic memory. A smaller amount of auditory information can be remembered upon immediate presentation of a cue. However, like iconic memory, echoic memory also decays over time.

SHORT-TERM MEMORY
Short-term memory is the component of the human memory system that temporarily retains information in conscious awareness for a period of time

There is a limit to the capacity of short-term memory. The exact amount of information that can be held in short-term memory is still currently being debated. However, the capacity limits serve as a bottle-neck limiting the amount of information that can be consciously retained or attended to

People occasionally engage in strategies such as chunking that allow them to increase the amount of information held in short-term memory. The following are some concepts of chunking highlighted by George Miller (1956)
 * Chunking: The process of organizing or grouping information into larger units
 * Chunk: The unit that is encoded in short-term memory
 * Recoding: The process of chunking items, then remembering the newly formed chunks.

To enhance the amount of information that that can be stored in memory, information are chunked into something that is meaningful. For example: If asked to memorize the letters c, h, u, n, k. An example of chunking would be to group these letters into the word "chunk" and memorize the word instead of the individual letters. The importance of chunks in memory retention is best summarized by Miller:

"Since the memory span is a fixed number of chunks, we can increase the number of bits of information that it contains simply by       building larger and larger chunks, each chunk containing more information than before." (Miller, 1956, p 93)

As larger chunks emerge amount of information that can be remembered increases. Remembering information can be thought of as a continual process of forming chunks that go together, until there are few enough chunks so that all items can be remembered. Chunking is a way of stretching the capacity limit of short-term memory.

WORKING MEMORY
Working Memory: “A Brain system that provides temporary storage and manipulation of information necessary for complex tasks such as language comprehension, learning and reasoning” (Baddeley, 1992).

The original conception of working memory by Alan Baddeley had three working memory systems: A fourth memory system was more recently added.
 * The phonological loop
 * The visuospatial sketch pad
 * The central executive
 * the episodic buffer

PHONOLOGICAL LOOP
The phonological loop was the first component system of working memory identified by Alan Baddeley. The phonological loop consists of two subcomponents

The  phonological store is the passive component of the phonological loop.

The  articulatory loop involves rehearsal of information in the phonological store.

Sound or speech related information to be retained in memory are first encoded as acoustic codes. Continual elaborative and maintenance rehearsal of acoustic codes eventually results in the transfer of verbal information into long-term memory. The phonological loop is also known as the articulatory loop as rehearsal usually involves vocal or subvocal articulation of acoustic codes

Several reliable effects provide some insight into the characteristics of the phonological loop.

The  word length effect is the finding that "memory span is inversely related to word length across a wide range of materials; when number of syllables and number of phonemes are held constant, words of short temporal duration are better recalled than words of long duration." (Baddeley, Thompson & Buchanan, 1975, p575).

The word length or span of a word is strongly correlated to the number of syllables in a word. SO an example of a word length effect would be that people can remember less words from lists that have more multisyllable or multi-phoneme words such as "ember" and "human" then lists with mono-syllable or single-phonemic words such as "queen" and "prime"

The articulatory suppression effect is that finding that when participants "saw lists of letters for immediate recall, but had to say a neutral sound as they perceived each letter, this substantially reduced recall of letters" (Murray, 1967).

The articulatory suppression effect has also been extended to poorer memory for lists of words when people are asked to articulate something else while they were trying to remember the words. For example. when people are asked to say "bah-bah-bah" while remembering lists of words, they will remember less words in the list, than if they were not given this instruction.

The phonological similarity effect is the finding that when people are asked to remember a list of word where words are phonologically similar for example "bore, board, book, bode" people find it harder to remember words from this list than a comparison list with words that are not phonologically similar like " newt, keep, soul, find" (Baddeley, 1966)

VISUOSPATIAL SKETCH PAD
When people are asked to visualize or imagine visual information, visual codes are represented with spatial reference to one another.

The visuospatial sketch pad is a system in Alan Baddeley's model of working memory that is involved in the processing and manipulation of spatial and visual information.

CENTRAL EXECUTIVE
The central executive is a component of working memory that is involved in the planning of behaviour and how retrieved and retained information is used. The primary function of the central executive is to initiate and control ongoing cognitive processes. It is also involved in allocating cognitive resources to required tasks, hence, it is responsible for determining the order in which a sequence of tasks is performed. The central executive is also known as the executive control system.

Without the central executive system, we would be able to look at an arithmetic problem, such as 9 - 5 x 6 and retain the visual code, but we would not know what to do with the code (i.e., 9 - 5 x 6 = ?). The executive function involves integrating these visual codes with information and procedures retrieved from long-term memory and determines the sequence of procedures required to perform the problem

EPISODIC BUFFER
The original three subsystems described a clear distinction between visual and verbal codes. However, the distinction posed a problem for the original working memory model - model was unable to resolve the binding problem described by Baddeley as follows:

"the question of how information from a range of separate independent sensory channels is bound together to allow the world to be perceived as comprising a coherent array of objects." (Baddeley, 2000, p420)

There was a need to find some way for distinct visual and verbal codes to be integrated into a single representation - For example there is a need to explain the impact of the visual codes "rude" and "rood" on verbal codes. As such a fourth component of Working memory: the episodic buffer was conceptualized. The buffer was described by Alan Baddeley as follows

"The episodic buffer is assumed to be a limited-capacity temporary storage system that is capable of integrating information from a variety of sources. It is assumed to be controlled by the central executive, which is capable of retrieving information  from the store in the form of conscious awareness, of reflecting on that information and, where necessary, manipulating  and modifying it. The buffer is episodic in the sense that it holds episodes whereby information is integrated  across space and potentially extended across time." (Baddeley, 2000, p 420)

The buffer was seen as a means of integrating information form a variety of memory systems - the visuospatial sketch pad, the phonological loop and long-term memory. The buffer temporarily retains this information and can manipulate and modify this information. The buffer differs from other working memory components, in that it can hold both visual and verbal information as a single episode, that can extend over a period of time.

WHAT'S THE DIFFERENCE BETWEEN SHORT-TERM MEMORY AND WORKING MEMORY?

The usage of the terms "short-term memory" and "working memory" differ based on the function of the storage system. The term "Short-term memory" is used to describe a storage system that retains information for a short period of time before transferring the information into a permanent store. When cognitive scientists study "Short-term memory", the primary focus is the limitations of the temporary storage system (i.e What are the capacity limits? How long does information need to be retained in short-term memory before it can be transferred into long-term memory?) "Working memory" is typically used when describing a group of memory systems that allow information to be manipulated. Capacity limits for individual components in working memory might occasionally be the focus of research. However, the main focus of study tends to be on how information is coded in the different working memory systems, and how these codes might be manipulated while people perform various tasks.

LONG-TERM MEMORY
Long-term memory is best described as a permanent storage system of information that we have learned. This information may not be readily available and is not immediately conscious. However, information in long-term memory can be retrieved and brought to conscious awareness usually by cues: for example, the cue word "your phone number" would bring to your conscious awareness a string of numbers that would not have been immediately available.

In 1950 Karl Lashley trained rats on memory tasks and then destroyed parts of their brains to find out where the memories were stored. But it didn't work. The more brain he destroyed, the worse performance got, suggesting that memories have some kind of distributed representation, rather than particular memories being stored at particular locations.

Each perception, and any memory encoding of it, has some specific neural correlate. Any complex memory will involve many things, such as concepts, perceptions, and emotions. At the time of memory recall, if we could re-activate the memory perfectly, we would re-experience the memory exactly the same way. But this never happens, because our internal instructions for accessing it are imperfect, the neurons or the connections between them might have degraded or otherwise changed over time. This leads to inaccurate, or dim memories. Because different memories share similar features, when we try to retrieve one we often retrieve pieces of another, resulting in interference. One of the reasons unusual things are remembered better is because they have less competition in the brain,, and one of the suggested explanations for why memory gets worse with old age is that we know so much that there's more interference.

TYPES OF LONG-TERM MEMORY
Larry Squire provided a taxonomy between different types of long-term memory (Squire, 1986)


 * Declarative memory: memory for knowledge about facts or events. For example: When a person is asked to recall the largest city in Canada or when Ottawa became the capital of Canada, the person is using declarative memory to retrieve information about that fact or event.
 * Semantic memory: a type of declarative memory that contains knowledge for facts and events.
 * Episodic memory: a type of declarative memory that contains knowledge about personally experienced events. For example, if you remember what you had for lunch last Friday, or describe the last day of your summer holidays, you would be using episodic memory, as you are trying to remember knowledge about an event or a sequence of events that happened to you personally.
 * Procedural memory: memory for performing a sequence of actions. For example: Learning how to ride a bicycle, drive a car or tie shoelace are examples of using procedural memory.

One classification of memory involves a distinction between explicit memory and implicit memory
 * explicit memory is long-term memory knowledge that can be retrieved and consciously reflected upon
 * implicit memory is knowledge that can be drawn upon without any involvement that the knowledge has been consciously retrieved.

ENCODING
Before Learning can occur, encoding usually takes place. Encoding is the process by which information in the environment is taken and transformed into a code that can be used by the cognitive system. Information needs to be encoded before it can be stored in memory.

Three types of codes have been identified There is still ongoing debate about the existence and nature of other types of codes.
 * 1) Visual code: A code based on the visual information of a stimulus
 * 2) Acoustic code: a code based on speech or sound related information for a stimulus
 * 3) Semantic code: a code that represents the meaning of a stimulus

LEARNING

 * A Memory trace or an engram is a physical change in a particular area of the brain that indicates that learning has occurred.


 * Consolidation is the process by which information is transferred from short-term (or working) memory into working memory.

In humans, information is transferred into the storage system through rehearsal. Rehearsal is the process that consolidates memory traces allowing learning to occur. Rehearsal results in a clearer, more precise code for information to be stored in memory. Two types of rehearsal have been identified: For example: associating the word "sound" with other memories such as "starts with s", "five letter word" or "has auditory properties" is elaborative rehearsal. "Sound" is transferred to memory, but is also linked with other memories
 * 1) Maintenance Rehearsal: This involves repeating information over and over. The representation stored in memory is therefore an exact replication of the information that has been repeated. For example: continuously repeating the word "sound" vocally or subvocally is called maintenance rehearsal. The acoustic code for "sound" is stored in memory
 * 2) Elaborative Rehearsal: This involves first encoding the information to be retained, then linking the code to other memories by trying to find meaningful associations.

Maintenance rehearsal results in retention of a single code without any associations to other memories Elaborative rehearsal results in associating of the code to other memories that have previously been retained.

RETRIEVAL
Psychologists in the 1950s and 1960s studied retrieval by asking people to remember lists of items and then to recall the items. During free recall, "the order in which they recall the items depends upon the probability of recalling the item" (Deese & Kaufman, 1957; Bousfield, Cohen, & Silva, 1956). However, Deese & Kaufman (1957) further stressed
 * In free recall people are asked to recall the list items in any order
 * In serial recall people recall the list items in the order in which they were presented.

"In general, we recall the first words first and the last words last. Thus, our recall of ordinary prose approximates the order of recall forced by the method of serial anticipation." (Deese & Kaufman, 1957, p 180)

THE SERIAL POSITION CURVE

The free recall task produces a finding that is best highlighted by plotting a serial position curve. The Serial position curve is obtained by presenting people with lists of items and asking them to recall the items. The serial position of each item refers to the position of that item on the list. For example if a list contains the words in the following order: student - teacher - college - school

"college" in this list would have a serial position of 3. Participants in recall tasks are generally given several lists to recall and the proportion of correctly recalling an item based on its serial position in the original list is plotted on a graph to obtain the serial position curve.

The primacy and recency effect

The serial position curve usually produces a "U-shaped curve" With higher accuracy for items at the beginning and end of the list. The primacy effect refers to better recall for items at the beginning of the list The  recency effect refers to better recall for items at the end of the list. The ability to recall items in the middle of the lists are lower than items at the beginning and end leading to the "U-shaped curve".

PRIMING
 * Priming occurs when the processing of a stimulus is affected by presentation of a related stimulus.
 * Typically Priming (or positive priming) is observed when cognitive processing of a current stimulus is facilitated by processing of a previous stimulus resulting in a reduced response time. For example: after seeing the word "raven", people respond faster to the word "black" as compared to the word "white"
 * Negative priming is observed when one or more cognitive process involved in the processing of a previous stimulus need to be inhibited as it otherwise interferes with processing of a current stimulus. For example: people see pairs of colored shapes and are always asked to respond to the triangle". In the first pair of shapes people see a white triangle and a black square, in the second pair of shapes, the see a black triangle and a red circle. People would have responded slower to the black triangle then if the triangle had been another color (lets say blue).

Changing Memories by Retrieving them When a memory is retrieved it becomes vulnerable to change. The context in which a memory is retrieved can change the memory, leading to false memories.

FORGETTING
Forgetting is the generic term used to describe the failure to retrieve information that has previously been remembered.

The Forgetting Curve

Herman von Ebbinghaus a German psychologist in the late 20th century, using himself as the only subject created lists of nonsense syllables in the form of consonant-vowel-consonant trigrams (for example: QAL, TER, YUP) and learned the lists according to the beat of a metronome. He then set aside the lists for a period of time, then relearned the lists. Ebbinghaus used nonsense syllables to minimize the influence of previous knowledge on learned material.

In the context of the Ebbinghaus studies:
 * Retention interval is the period of time in days that the list was set aside and not learned.
 * Savings score is the reduction in the number of trials required to relearn a list relative to original learning of the list.

For example if it took 10 trials to learn a list but 4 trials to relearn a list The savings score is 6 trials, and there is a savings of [(10-4)/10 x 100] = 60%.

The rationale behind the concept of savings is that the more information is forgotten, the greater the number of trials that are required to relearn the information to the criterion of accuracy obtained at original learning of the list.

The forgetting curve is a graph of savings (as percentage) as a function of retention interval. The forgetting curve shows a dramatic reduction in savings at retention intervals of less than two days. After 20 mins, savings decreased dramatically to 58%, and decreased to 44% after 1 hour. Savings dropped to approximately 22% after 2 days. Savings then remained relatively consistent at about 18% after 31 days.

To put this in perspective, if Ebbinghaus had required 100 trials to learn a list of nonsense syllables such that he could accurately recall all the items in that list; after 20 minutes, he would have needed to learn that same list 42 times to recall all the syllables in that list. After 1 hour, he would have needed to learn that list 56 times. After 2 days, he would have needed 78 trials; and he would have needed 81 trials after 31 days

The Ebbinghaus studies were the first to demonstrate experimentally the rate at which material was lost from memory.

Decay and Interference

Around the 1940s to the 1950s: two theories have been suggested as to why people are unable to remember information.


 * Decay is the loss of information over time from memory.
 * Interference is a competition of related or recent information that results in loss or failure to remember target information.

Early ideas of decay and interference

In the 1950s, trace-decay theory has been described as follows:

"a memory trace is established which decays rapidly during the initial phase of its career. Some decay of the trace is assumed to be    compatible with reliable recall. But recall will cease to be reliable if decay of the trace proceeds beyond a critical level" (Brown,   1958, p12)

In the initial stages of entry of information about an item into memory, there is rapid loss of large amounts of information about the item. However, the amount of information that is lost gradually decreases over time. While some information about the item can still be remembered over time; memory for this item is no longer reliable if too much information is lost. According to decay theory, there are two main factors which contribute to the ability to recall an item:
 * 1) the amount of information that has been lost
 * 2) the time interval between the presentation of the stimulus and the recall of information about the item.

"When a sequence of items is presented, the interval between the perception of each item and the attempt to recall that item will    depend on the length of the sequence. If the sequence exceeds a certain length, decay of the memory traces of some of the items will   proceed too far for accurate recall of the sequence to be possible." (Brown, 1958, p13)

Brown (1958, p13) suggests rehearsal as a strategy of countering initial loss of information through decay.

"Everyday experience-of trying to remember telephone numbers, for example-suggests that the effect of such rehearsal may be to  counteract decay of the trace rather than to strengthen it much, since continuing rehearsal tends to be necessary to prevent   forgetting....One way to test the hypothesis of decay of the trace, therefore, is to see whether if recall is delayed for several   seconds forgetting occurs even when the amount of material is well within the memory span. ." (Brown, 1958, p13)

However, when people are asked to remember additional information that prevented rehearsal of an item, they had difficulty in recalling the item. This is best demonstrated by the following:

In a study performed by Peterson & Peterson (1958), participants were asked to attend to a three letter trigram (e.g CHJ). They were then shown a three digit number (521) and asked to count aloud backwards in threes, from this number to the beat of a metronome clicking at the rate of twice every second. They were then asked to report the three letter trigram after a variable time interval had elapsed. In an example of a trial: A participant could be shown the three letter trigram (CHJ), followed by the number (521). The participant would then count aloud (521 - 518 - 515 - 512 - 509 - 506) for a time interval of 3s and then recall the three letter trigram. Even after 3s of counting, memory of the three letter item was slightly better than 50%. Memory of the three letter item had been forgotten, even though the number of items to be recalled was within the memory span. This also suggests that decay is not the only that results in a failure to retain information.

Brown-Peterson Task

These types of tasks are called Brown-Peterson tasks. They require participants to Two Types of Interference
 * 1) remember a list of items ,
 * 2) introduce a task that prevents rehearsal of the items
 * 3) recall the items to be remembered
 * Proactive interference occurs when prior material interferes with recollection or recall of current material. For example: you are trying to remember material definitions from the recent chapter you read, but you recall material from the previous chapter that you studied. Brown-Peterson tasks are also examples of proactive interference.
 * Retroactive interference occurs when more recent material interferes with memory of earlier material. For example, you are trying to describe what you did ten days ago, but you recalled what you did yesterday.

The Function of Forgetting
Another theory is that forgetting is not a side effect of other processes, but something the mind actively does. An rat studies, blocking AMPA receptors prevented rats from forgetting things, suggesting that forgetting is an active process.

Why would forgetting be an important part of memory? Some people have highly superior autobiographical memory (HSAM). Although they can tell you what they were wearing on any particular day, they have trouble generalizing, tend to be obsessive, and are not particularly successful in life. On the other end of the spectrum, people with severely deficient autobiographical memory (SDAM) can remember very little of the details of their lives, and also have trouble imagining the future. They tend to excel at jobs requiring abstract thinking, suggesting that forgetting the nitty-gritty might aid with understanding useful generalizations. This is related to a machine learning concept called overfitting, where too much attention is paid to irrelevant details.

Evolutionary Psychology of Memory
Evolutionary psychology predicts that our memory would be better for those things relevant to survival and reproduction in our ancestral environment. This has shown to be the case in several studies. Words that represent things relevant to survival are better remembered than neutral things, and when reading text, objects that are likely to be relevant to future events are better remembered. For example, a tack described to be on the floor is remembered better than a tack safely in a box