What happens to the information once it has been recognized? Where is the information being recognized? According to a general information processing model of memory, after the pattern is recognized it is stored in STM.
STM can be considered in terms of its structure (Storage bin) and in terms of its functions (processes).
The duplex notion of memory (that we have two types of memory) makes sense when we introspect about our own memories
Ebbinghaus - was the first researcher to support the dualistic notion of memory. Using nonsense syllables, he found that he could recall a list of 7 or fewer items perfectly after hearing the list once. As the list grew past 7, his performance declined. Ebbinghaus' procedure was called Serial Learning. Following list presentation, the task was to recall the items in order of presentation. His finding suggest a brief immediate memory (recency effect) and a longer memory for the beginning items on the list (primacy effect) (SERIAL POSITION EFFECT).Problems with the multi-store models of memory:LOSS OF RECENCY
Glanzer and Cunitz (1966) found the serial position effect; they then did same list learning task with an interpolated task at the end (counting backward for 10 seconds). Findings: loss of recency information.
Postman and Philips (1965) recency effect obliterated by an arithmetic task. These studies serve as research support for the difference between immediate and long term memory.
STM is our conscious awareness, our immediate or working memory, active memory. William James distinguished between our primary and secondary memories.
Primary memory - relates to information that remains in consciousness after it has been perceived, it forms a part of the psychological present. Primary memory has limited capacity and duration. However, it has more than one capacity.
Secondary memory - contains information about events that have left consciousness, the psychological past
Rehearsal- a major means by which information is stored in LTM from STM is rehearsal (RUNDUS).
The memory stores have been demonstrated to be qualitatively different in terms of storage capacity, temporal duration, forgetting mechanism, and the effects of brain damage ( double dissociation).
1) frequency may not affect transfer from ST to LT memory
2) neurophysiological evidence: intact STM and an impaired LTM and vice versa. If LTS depends on STS should not patients with a defective STS be impaired in learning? Shallice and Warrington studied JB who has an impaired STS - she has problems with verbal memory span tasks but leads an active and successful life as a personal secretary and mother (many processes). Warrington & Shallice (1974) discovered that KF's ( a sufferer of Korsakoff's) STM forgetting of auditory letters and digits was considerably greater than his forgetting of visual stimuli and that his ST forgetting of other auditory stimuli (animal sounds, other sounds) was less impaired.
3) recency effects - assumed to illustrate STS, troublesome finding:
Tseng (1973) immediate and delayed recall task: 1) Ss given a list of unrelated words, 2) in between the presentation of each word Ss counted backward - should interfere with rehearsal
Findings: Ss could remember words - the recency effect survived
General Features of STM:
(ATKINSON AND SHIFFRIN MODEL)
Limited Capacity
Sir William Hamilton (19th century philosopher); if you throw a large number of marbles in front of you you can only "see" approximately 7 items at one time.
George Miller (1956) - The magic number 7 ... regardless of the kind of data
Miller - a model of memory where 7+2 units of information can be simultaneously held. Individual letters represent individual pieces of information and each letter fills in a slot. The letters that compose a word are CHUNKED into one unit and each unit takes up one slot. Thus, increased capacity comes about by encoding the letters into words or other meaningful units Chunking expands the capacity of STM; it also helps us understand how so much information can be processed through STM in a relatively short amount of time.
Capacity - Miller 5-9 items, chunking (Bower), Simon (1974) the size of the chunks matters - the number of chunks in the span is smaller the larger the chunks.
Problem with span measures of STM: LTM plays a part in determining the span; for example - when digit strings are presented for immediate serial recall, and one digit string is repeated surreptitiously performance on that string exceeds that of the other non-repeated strings . This finding suggests that LTM is holding some info about the repeated string (Bower & Winzenz, 1969).
Bower and Springston (1970) familiar vs. unfamiliar groups of letters recalled
Chess experts (DeGroot); Chase & Simon; electronics circuitry for technicians
Baddeley & Hitch (1974) when Ss are required to solve simple problems and are at the same time, required to repeat digits aloud, performance on the problems was impaired and it took longer to solve the problems.
Different theorists mean different things by space or capacity - storage capacity how much information the memory system holds; processing capacity how much information can be processed or how many separate operations can be performed at one time (Zechmeister & Nyberg, 1982). Regardless of the definition, we know that working memory capacity is limited. (Does the processing itself take up capacity - see rehearsal)
Time Limitations
STM is also limited by how long it can hold information - its duration
Brown (1958) Peterson & Peterson (1958) task - a rehearsal prevention task. Ss are given 3 consonants and then are given an interpolated task (counting backwards by 3s). The interval between presentation of the tbr items and the recall task is varied. Findings: when rehearsal is prevented, forgetting occurs in several seconds (18 seconds)
Murdock (1961) the size of the memory unit has no effect on how rapidly the information is forgotten (chunking). Repeated the procedure using three letters, one three letter word (dog), 3 three letter words (dog, rat, tag). The first and the last conditions have the same number of items. Recall performance declined at the same rate for 3 letters or 3 words.
The results of these experiments imply that forgetting in STM occurs through decay. This position is quite different from the interference position implied by research on spatial constraints on STM - George Miller's simple slot model. If the information is not rehearsed, it is lost
These results also point out the importance of rehearsal in maintaining information in STM and in transferring it to LTM. Problem: why isn't the interpolated task considered interfering? Why aren't the preceding trials interfering.
LTM plays a part in determining the span of STM.
Bower & Winzenz (1969) when digit strings are presented for immediate serial recall and one string is surreptitiously repeated several times, performance on the repeated string becomes progressively superior - suggesting that some information about the repeated string is stored in LTM.( Perceptual fluency - Jacoby et al.)
Coding in STM
Acoustic Encoding
In what form is information held in STM? Originally thought that STM information is acoustically encoded.
Consider the case of the telephone operator giving you a number 969-3192. This number must be maintained in your immediate memory until you can write it down, what do you do - repeat it to yourself - saying the digits (an acoustic code, an auditory representation of the numbers). Thus, from a common sense standpoint, we hold info in an acoustic form.
Remember Conrad's experiment and the type of errors obtained?
Sperling and Speelman (1970) looking at the errors made in recalling items from STM found that STM items are held in acoustic form and may be lost phoneme by phoneme. At recall try to construct the items from the sounds that remain thus producing words that sound like the ones on the list.
Visual Encoding
suggests that some of the time the STM code is visually encoded
Posner & Keele (1969) showed Ss two letters either simultaneously or the second letter followed the first by a brief interval (.5s,1s,1.5s). The task of the S was to indicate if the letters were the same; the dependent measure was reaction time. AA Aa AB Ab
AA identical in form and name
Aa identical in name
AB different in both
Ab different in both
Under which condition do you think RT was longest?Further research by Posner found that RT for identical codes is much shorter in the early phase of memory processing but as time passes, the differences disappear (at approximately 2s).
550
500 . . ..
450 .
400msec.
---------------------Relationship between high and low verbal SATs
Perceptual deficit? readers.
Semantic codes - Wickens - release from PI
Proactive interference - reduced capability of recalling new information due to interference from old information
Retrieval from STM
Retrieval from STM search must be rapid because of its short time duration. First scan STM contents selecting some information for further examination; second we rehearse further encoding information to prevent loss; third, we transfer infor. to LTM
Sternberg (1966,67,69)
serial scanning and probe task; S is shown a series of numbers each letter is displayed for 1.2 secs. After the S decides that the numbers are stored (8, 3, 11, 60), she presses a button. At that time, a probe digit is presented. The probe is either a member of the set or not. The S must then indicate by pressing a reaction time button whether the probe was or was not a member of the set. Performance will be perfect, so the dependent measure is not correctness but reaction time (how long it takes the S to decide).Findings: the longer the set, the longer the decision; RT changed uniformly according to the number of items in the memory set (each member of the set seemed to require a fixed amount of processing time (38msec); Rts were nearly identical for items that were in the set and items that were not in the set!
Common sense would tell us that if the memory set was 8,3,11,60 and 8 was the probe then it should take less time to indicate that 8 was in the list than to indicate that 11 was in the list or that 13 was not in the list (serial search); however, what is found is that regardless of where the item is in the list or even if it is not on the list, the decision time is the same. So, when searching STM, instead of doing a compare/decide, compare/decide operation, we seem to compare, compare, compare/decide (serial exhaustive search).
Parallel processing (Baddeley & Ecob, 1973) the rate at which the comparisons can be performed is a function of how active the items are in memory
There is a limitation to the amount of activation available in STM. When there are more items in the set, the activation must be distributed among these items, consequently, each item in the set will be compared less rapidly with the target item. Time is spent inspecting all items at once rather than in a serial search;
Repeated items in a memory set 4,3,6,4,5 (probe 4) findings: RT faster for 4 than for any single item in the set (e.g. probe 3). Conclusion: the rate at which comparisons are made is a function of how active the items are.
Rehearsal: how do we maintain information in STM? How do we make information permanent?
Sperling referred to rehearsal as implicit speech.
Rundus had Ss rehearse items aloud; words were presented 1 per 5secs and Ss selected the words that they wanted to rehearse (the rehearsal set). He later had Ss recall as many items from the list as they could. Findings: Ss recalled the items that they repeated most often. Rundus also noted that the Ss selection of items to be rehearsed was somewhat determined by their meaning ( an item was more likely to be repeated if it fit in with other items in the set) (implies a relationship with LTM).
A major criticism of the Rundus study is that it is correlational. Do rehearsals determine recall or do Ss rehearse those items that would be most readily recalled?
Conclusions: rehearsal serves to increase the strength of items in STM and we use information in LTM to determine what will be rehearsed.Craik and Watkins (1973) tested whether rehearsal determined recall in a cause and effect design.
These results have led to distinctions between two types of rehearsal: maintenance (used to maintain items in STM presumably by renewing them before they are forgotten); and elaborative (maintains information in STM by thinking about, adding to, connecting, etc. the items. This type of rehearsal enhances LTM)
Forgetting
Passive decay theory strength of item decreases with the passage of time. Evidence: Brown/ Peterson & Peterson
Interference the strength of an item decreases because a new item enters STM
Simple slot or replacement model
STM has 5-9 slots. Each slot holds a chunk of information. each entering item occupies a slot. When there are no more empty slots, when new items enter, old items get pushed out.
Implications: the first few items entering the system shouldn't interfere with each other (as forgetting occurs only after the first 7 slots are full) evidence: serial position/ primacy & recency; because each slot contains an item, the item should be totally there or gone (no partial forgetting)
Strength model
each item in STM has a certain strength. When an item is newly entered or just rehearsed, its at its full strength. Forgetting occurs as strength declines and the item cannot be recovered. The cause of the decline of strength may be the entry of new information; the extent of the forgetting is the similarity of the new items to the old.
LEVELS OF PROCESSING MODEL (Craik & Lockhart, 1972)
stimuli can be analyzed at different depths of processing from shallow to deep. Several degrees of depth that are hierarchical in order: physical , acoustic and semantic. Each of these requires a deeper amount of processing.
Assumption: deeper procssing leads to better memory
Hyde and Jenkins, Craik & Tulving
Assumption: only deeper processing , not continued processing at the same level leads to better memory ( argues against maintenance rehearsal) - G experimentLevels retains the distinction between ST and LT memory but has changed the emphasis from separate stores where different encoding processes occur to "coding" as the most important variable. Trace durability is a function of how the info was encoded. Implies a continuum of depth rather than a series of separate stores.
Problems:
1) no objective measure of depth. Does it refer to Processing time? No - can produce tasks that are shallow that take a lot of time (Craik & Tulving, 1975 - CVC study). No evidence that slower processing leads to better memory
2) maintenance rehearsal can lead to Lt memory (GLenberg, Smith, and Green, 1971) Nelson (1977) Horowitz & Pyrtulak (1969) repetition may cause a substantial increment in the ability to recall the item as a whole (Redintegration ) . Repeated presentation may prime the representation of the word which may increase its accessibility.(Repetition priming)
3) shallow processing and durable learning - superficial encoding gives rise to a rapidly fading memory. Evidence against shows that physical aspects of a stimulus can be maintained over long periods of time: Kolers (1976) inverted text - 13 months later subjects remembered. Morris, Bransford & Franks (1977) mode of testing (TAP)
4) discrete domains vs. a processing continuum - each encoding task would process material to a specific depth which would determine its memorability. Clear differences are observed between processing domains: e.g. for words - visual processing leads to poorer retention than phonological coding which is poorer than semantic, but DIFFERENCES WITHIN DOMAINS HAVE NOT BEEN OBSERVED implying that the notion of depth as a processing continuum is less appropriate than a more discrete concept such as processing domains.
5) the linear processing stages assumption; that info is processed through an ordered hierarchy of levels beginning with the physical features to the phonemic to the semantic. Perception involves the concurrent or parallel processing of many dimensions of a stimulus simultaneously rather than linearly. This includes the transfer of info in both a bottom- up direction and a top-down direction (e.g. a typed word - a graphemic form transformed into a phonological rep. which leads to the meaning of the word)
However, Marcel (1983) showed that Ss were sensitive to meaning when they were unaware of the physical stimulus or the phonological characteristics:
presented words tachistoscopically followed by a pattern mask (yellow). Ss sometimes guessed a word that was semantically similar to the masked word (red) but was orthographically different. So a second study was done:
after presenting a word (hill) Ss were asked:1) did a word precede the mask?
2) which did it resemble (hilt or cigarette)
3) which did it semantically resemble (hilt or mountain)
Findings: the 3rd.condition was best under short intervals (just the opposite of the levels prediction)dyslexic patients - process words semantically that they cannot pronounce (acoustic) say water for lake
6) compatibility effect - items given a positive response during the initial encoding e.g. does it rhyme with log? dog or bun
were better remembered. Positive instances are more related to the question hence creating a more integrated and retrievablee unit.7) simple depth of processing is less important than degree of encoding elaboration - under slow presentation conds. when Ss were aware of a subsequent recall test, a phonemic encoding led to poorer retention. The conds. of presentation were so slow that the S had to be aware of the words' meanings. A semantic encoding leads to a richer, more elaborate semantic code than simply registering the word's meaning (C&T).
What we have referred to as sensing, pattern recognition, attention, rote repetition ( all shallow); understanding and interpreting the stimulus, conceptual analysis (deep
WORKING MEMORY: Domains of Processing (BADDELEY)
The levels of processing theory and the multistore models of memory fell into disfavor because they were over simplifiedThe problems with levels:
1) maintaining info at a given level (whether phonological or semantic) will not increase its memorability
2) levels of processing theory implies a continuum of depth rather than a series of separate stores
Working memory holds the most recently activated portion of LTM moving this information in and out of temporary store.Domains of Processing - the memory system does not exist as a continuum
Three components:
1) a modality-free central executive resembling attention;
2) an articulatory loop which holds information in a phonological form (speech-like);
3) a visuo-spatial sketch pad specialized for spatial or visual coding (Baddeley & Hitch, 1974)
1) the central executive - has limited capacity and is used in most cognitively demanding tasks. Responsible for control processes (selective attention) (frontal lobe damage) .The coordination of resources is the prime function of working memory with memory storage being only one of the demands made on the system. The role of the central executive is coordinating information from the slave systems (art. loop, vis-spat).
2) the articulatory loop consists of a passive phonological store which is directly concerned with speech perception and an articulatory process that is linked to speech production. This system is a slave system and is organized in a temporal and serial fashion. Temporal duration determines capacity(Baddeley, Thompson, Buchanan, 1975 - word span study, Ss could provide immediate serial recall for as many words as they could read aloud in two seconds).
Holds up to two seconds of spoken material. Use of the loop can be prevented if the material presented is visual and subjects are required to make irrelevant articulations during presentation (causes suppression of the art. loop).A) The acoustic similarity effect. Immediate recall of ordered items is poorer if they are similar rather than dissimilar in sound. Conrad (1964) demonstrated the importance of phonological encoding. Baddeley (1966) phonological similarity messed up recall in serial recall tasks ( mad, man, mat, cad, cap). Conrad (1970) deaf kids showed phonological confusions in remembering consonant sequences (good speakers - articulation).
B) the irrelevant speech effect. Refers to the reduction in recall of lists of visually presented items brought about by the presence of irrelevant spoken material. The semantic characteristics of the material is not important with a language that is unfamiliar being just as disruptive as a familiar language. However, it is not simply distraction as loud bursts of noise do not disrupt.
C) the word length effect. Memory span for words is inversely related to the spoken duration of words. Ss can remember about as many words as they can say (vocally or subvocally) in 2 secs.
D) articulatory suppression. It is possible to disrupt the use of subvocal rehearsal by requiring Ss to utter sound irrelevant sound such as the word "the". Prevents rehearsal and removes the effect of word length. Prevents registering material in the phonological store. Articulatory suppression studies - Murray (1965) the greater the art., the greater the memory. However, art. suppression has very little effect on reading - it seems to affect the ordering of words.(holds word order in memory).
Neurophysiological evidence - patients suffer damage to the slave system subprocesses rather than to the central processor. Patients with aphasia typically have lesions in the areas associated with speech (temporal lobe - Wernickes' and frontal lobe - Broca's area) and aphasia is frequently associated with impaired memory span ( aphasics have impaired articulation and impaired memory).
3) the visual spatial sketch pad is a slave system to the central executive. Imagery is disrupted by performing a visual spatial task simultaneously.
Alzheimer patients and normal elderly Ss were required to perform two tasks simultaneously , one visual and one verbal. Task difficulty of each task was adjusted so that Alz. Ss were making the same proportion of errors as the other Ss, then the Ss had to perform both tasks at the same time. Alz. Ss showed a marked impairment in performing both tasks simultaneously while normal elderly and young Ss could coordinate and do both tasks effectively ( memory and tracking) (Baddeley, Logie, Bressi, Della Salla, Spinnler - Q.J.EXP.PSY. 38A,603,1986)*. As the disease progressed performance on the individual tasks held up well but the coordination of tasks deteriorated markedly. A central executive deficit.
Classic amnesic syndrome patients appear to have gross disruption of the capacity to form new lasting memories but preserved performance on tasks that were assumed to test STM. A second type of patient shows normal LTM but a STM span limited to 1 or 2 items. (a deficit in ST storage). If STstore acts as a working memory being necessary for learning, for the retrieval of old materials, and for the performance of many cognitive tasks, one would expect that patients with a grossly defective STM to show other problems including impaired LTM. However, these patients appear to have intact LTMs and show few other cognitive impairments (dissociation).
To simulate the STM deficit condition Ss were asked to engage in various types of learning tasks while trying to remember digits. As the digit load increased, task performance decreased but the degree of disruption fell short of what would have predicted by a dualist model. Ss whose digit memory was at full capacity could reason and learn quite effectively. The articulatory loop is assumed to hold speech based information (including digit spans) while other tasks can be performed.
Evidence for the importance of the phonological loop: children with a specific language disorder were unable to repeat back nonwords - 8 year olds functioned like 4 yr olds; a study on 118 kids (4-5yrs) nonword repetition measured - highly correlated with vocabulary and a powerful predictor of vocab a yr later.; in another study, kids were taught new words for monsters. Two groups: matched for nonverbal intelligence but differing in nonword repetition abilities. The low nonword rep. Kids had greater difficulty learning the new names. Remember nonrepetition capacity is assumed to depend on a ST phonological store.
PARALLEL DISTRIBUTED PROCESSING
McClelland (1981) McClelland and Rumelhart (1986b)
a connectionist network/ parallel = more than one process occurring at one time; distributed processing - processing occurring in a number of different locations
&n