What is the capacity of short term memory PDF

Preview

Summary

the capacity of short term memory...

Description

What is the capacity of short term memory? Short-term memory is the memory for a stimulus that lasts for a short while (Carlson, 2001). In practical terms visual short-term memory is often used for a comparative purpose when one cannot look in two places at once but wish to compare two or more possibilities. Tuholski and colleagues refer to short-term memory as being the concomitant processing and storage of information (Tuholski, Engle, & Baylis, 2001). They also highlight the fact that cognitive ability can often be adversely affected by working memory capacity. It is particularly important to be clear on the normal capacity of short term memory as, without a proper understanding of the intact brain’s functioning it is difficult to assess whether an individual has a deficit in ability (Parkin, 1996). This review outlines George Miller’s historical view of short-term memory capacity and how it can be affected, before bringing the research up to date and illustrating a selection of ways of measuring short-term memory capacity. The historical view of short-term memory capacity

Span of absolute judgement The span of absolute judgement is defined as the limit to the accuracy with which one can identify the magnitude of a unidimensional stimulus variable (Miller, 1956), with this limit or span traditionally being around 7 + 2. Miller cites Hayes memory span experiment as evidence for his limiting span. In this participants had to recall information read aloud to them and results clearly showed that there was a normal upper limit of 9 when binary items were used. This was despite the constant information hypothesis, which has suggested that the span should be long if each presented item contained little information (Miller, 1956). The conclusion from Hayes and Pollack’s experiments (see figure 1) was that the amount of information transmitted increases in a linear fashion along with the amount of information per unit input (Miller, 1956). Figure 1. Measurements of memory for information sources of different types and bit quotients, compared to expected results for constant information. Results from Hayes (left) and Pollack (right) cited by (Miller, 1956)

Bits and chunks Miller refers to a ‘bit’ of information as ‘the amount of information needed to make a decision between two equally likely alternatives’. Thus a simple either or decision requires one bit of information; with more required for more complex decisions, along a binary pathway (Miller, 1956). Decimal digits are worth 3.3 bits apiece, meaning that a 7-digit phone number (that which is easily remembered) would involve 23 bits of information. However, an apparent contradiction to this is the fact that, if an English word is worth around 10 bits and only 23 bits could be remembered then only 2-3 words could be remembered at any one time, obviously incorrect. The limiting span can better be understood in terms of the assimilation of bits into chunks. Miller distinguishes between bits and chunks of information, the distinction being that a chunk is made up of multiple bits of information. It is interesting to note that whilst there is a finite capacity to remember chunks of information, the amount of bits in each of those chunks can vary widely (Miller, 1956). However, it is not a simple case of being able to remember large chunks immediately, rather that as each bit becomes more familiar, it can be assimilated into a chunk, which is then remembered itself. Recoding is the process by which individual bits are ‘recoded’ and assigned to chunks. Thus the conclusions that can be drawn from Miller’s original exposition is that,

whilst there is an accepted limit to the number of chunks of information that can be stored in immediate (short-term) memory, the amount of information within each of those chunks is able to be quite high, without adversely affecting the recall of the same number of chunks. The modern view of short-term memory capacity Millers magic number 7+2 has been more recently redefined to the magic number 4+1 (Cowan, 2001). The challenge has come from results such as those from Chen and Cowan, in which the predicted results from an experiment were that immediate serial recall of absolute numbers of singleton words would be the same as the number of chunks of learned pair words. However in fact it was found that the same number of pre-exposed singleton words was recalled as the number of words within learned pairs - eg 8 words (presented as 8 singletons or 4 learned pairs). However 6 learned pairs could be recalled as successfully as 6 pre-exposed singleton words (Chen & Cowan, 2005). This suggested a different mechanism for recall depending on the circumstances. Cowan refers to the maximum number of chunks that can be recalled as the memory storage capacity (Cowan, 2001). It is noted that the number of chunks can be affected by long-term memory information, as indicated by Miller in terms of recoding - with additional information to enable this recoding coming from long-term memory.

Factors affecting apparent short-term memory Rehearsal The propensity to use rehearsal and memory aids is a serious complication in accurately measuring the capacity of short-term memory. Indeed many of the studies ostentatiously measuring short-term memorycapacity have been argued to be actually measuring the ability to rehearse and access long-term memory stores (Cowan, 2001). Given that recoding involves rehearsal and the use of long-term memory formation, anything that prevents or influences these will obviously affect the ability to recode successfully (Cowan, 2001).

Information overload Short-term memory capacity may be limited when information overload precludes recoding (Cowan, 2001). For instance, if attention is directed away from the target stimulus during presentation there is too much information being processed to attend properly to the target stimulus. Therefore fewer items would be remembered as they would have been replaced by information from this alternate direction. Similarly, but actually distinguished quite definitively by Cowan, are techniques such as the requirement to repeat a separate word during the target stimulus presentation, which acts to prevent rehearsal.

Altering stimulus frequency and format It has been found that, if a word list contains words of long and short length words, recall is better for the length that occurs least frequently, thus is more individually distinct (Chen & Cowan, 2005). Similarly the word length effect indicates that memory span is higher for words with a shorter spoken duration; syllable length varying as long as the spoken duration remains relatively constant (Parkin, 1996). This is similar to Miller’s chunking of information, if one were to assume that the spoken duration was a chunk of information and the syllable length was the bit of information.

Associations between components of information

Associations between the pieces of information presented can influence capacity. Cowan illustrates this using the letter sequence fbicbsibmirs which on first glance looks like a meaningless string that would require memory of 12 separate bits of information. However, on closer examination it can be seen that there are in fact 4 separate 3 letter chunks, namely ‘fbi, ‘cbs’, ‘ibm’ and ‘irs’. Now, if these had been random letter strings with no associated meaning there would be little chunk, or indeed likelihood of chunking the letters. However it is suggested that the well known acronyms of governmental and industry organisations considerably aids recoding, thus memory. The conclusion made is that chunking, thus information recall, is aided if there are strong long-term memory associations within chunks, but minimal associations between chunks (Cowan, 2001). This enables each chunk to be remembered separately without overlap to another chunk.

Time limitation Short-term memory has traditionally be assumed to be time limited, in that information is only able to stay in the memory store for a specific time. However this assertion has been challenged and instead a form of information replacement has been suggested, whereby a finite capacity to short-term memory ensures that the entry of a new piece of information displaces an older one (Cowan, 2001).

Methods of measuring the capacity of short-term memory There are a variety of methods used to measure the capacity of short-term memory. These include enumeration, whole report and alphanumeric span tasks (taken to have a 4 chunk upper limit), and recall of visual stimuli, multi-object tracking and repetition priming (all argued to show an upper limit of less than 4) (Avons, Ward, & Russo, 2001). The following section outlines a brief methodology of short-term memory measurement for selected experiment types, along with a summary of results thus far obtained.

Enumeration Enumeration tasks involve presenting a participant with n objects to count, and measuring the reaction time for each number. It is argued that the smaller the working memory capacity, the steeper the reaction time slopes would be (Tuholski et al., 2001). As can be seen from figure 2 below; using lines as the object to count; the reaction time is relatively constant until more than 4 lines are presented, at which point reaction time increases sharply. This indicates that 4 lines is the easy upper limit in terms of this particular version of short-term memory. The authors conclude that it is the controlled processing element of counting that limits the working memory span. This has been described as subitizing, in which a few items can be readily and rapidly attended, but more items require a steep increase in both reaction time and overall time required to attend to the items (Cowan, 2001). Figure 2. An example of results obtained from an enumeration task (adapted from (Tuholski et al., 2001) This ‘elbow’ in the enumeration curve has been proposed to be caused by an increase in memory load, specifically a less automatic method of processing, which allows more time in which engrams within the short-term memory can be overwritten, thus reducing accuracy (Green & Bavelier, 2005). It could be argued, however, that enumeration isn’t measuring short-term memory as much as counting ability. Further it has been indicated that enumeration is invariably only related to individuated items (Cowan, 2001), eg bits rather than chunks, so it is not clear what results would occur if it were not.

Whole Report Whole report procedures involve recalling all possible stimuli from an array presented. This contrasts to partial report procedures, in which only specific stimuli need to be recalled, usually in response to a specific cue. Cowan reports results indicating that short-term memory capacity is 4 for whole report procedures and links this to sensory memory (Cowan, 2001). Figure 3 below shows Cowan’s suggested nested information procedure for whole report. In this any and all information is elevated from the activated long-term memory store into the focus of attention until this latter is full (Cowan, 2001). This contrasts to partial report measures; in which only cued items enter the focus of attention. Figure 3. Processing in whole report procedures (Cowan, 2001) An obvious criticism of whole report measures is that they are assessing the ability to access long-term memory, not necessarily short-term memory capacity.

Multi-object tracking Multi-object tracking is carried out using flashing dots on a computer screen. Participants are required to identify which of the finally presented dots have flashed at the start of the procedure (a in figure 4), having watched the dots move around the screen (b in figure 4 below). Figure 4. An example of the dots used in a multi object tracking procedure (Cavanagh & Alvarez, 2005) Memory capacity appears to be around 4 for this task, as 3 dots can easily be tracked, whereas the majority of participants experience difficulty with 5 (Cowan, 2001). A recent study also found that the limit for tracking independent targets was 4 (Cavanagh & Alvarez, 2005) but Avons and colleagues (Avons et al., 2001) disagree with this (but do not provide a viable alternative) . However, Cavanagh and Alvarez do highlight the need for further research to separate the effects of visual tracking from memory capacity, when measuring performance in multi-object tracking experiments. Further research concludes that visual short-term memory capacity is actually limited by a whole chain of capacity bound operations (Delvenne, 2005).

Repetition priming Repetition priming involves the presentation of a series of words and nonwords, which includes repetition of words with a variable number of other items intervening. The repeated word is said to be primed and the specific measure is the reaction time to this repeated word. It has been found that up to 4 items can be reliably recognised in this way (Cowan, 2001) (see figure 5 below). McKone argues that repetition priming is an accurate measure of short-term memory capacity as the long lists of words prevent rehearsal, as does the inclusion of nonwords (McKone, 2000). Indeed she goes on to explain that capacity, as measured by primed repetition is related to the limited nature of the focus of attention. Figure 5. The reaction time and number of words recognised from primed (old) words in a repetition priming experiment (McKone, 2000)

Conclusion There is still much debate about the capacity of short-term memory and the accuracy of measuring it. It is difficult to separate genuine short-term memory capacity from the more working memory capacity that is affected by rehearsal. Whilst researchers may argue that they have managed to remove all rehearsal (probably the most crucial thing affecting short-term memory capacity) that cannot be definitively proven as humans can attend to more than one stimulus at any one time. Nevertheless whilst Miller’s original work is still seminal in the area of short-

term memory capacity it is true to say that his conclusions of 7 + 2 has now been superseded to 4 + 1.

References    

     



Avons, S. E., Ward, G., & Russo, R. (2001). The dangers of taking capacity limits too literally. Behavioural and Brain Sciences, 24(1), 114-115. Carlson, N. (2001). Learning and memory: Basic mechanisms. Physiology of behaviour (7th ed.) (pp. 423-465). Boston: Allyn and Bacon. Cavanagh, P., & Alvarez, G. A. (2005). Tracking multiple targets with multifocal attention. Trends in Cognitive Sciences, 9(7), 349-354. Chen, Z., & Cowan, N. (2005). Chunk limits and length limits in immediate recall: A reconciliation. Journal of Experimental Psychology. Learning, Memory, and Cognition, 31(6), 1235-1249. Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioural and Brain Sciences, 24(1), 87-114; discussion 114-85. Delvenne, J. F. (2005). The capacity of visual short-term memory within and between hemifields. Cognition, 96(3), B79-88. Green, C. S., & Bavelier, D. (2005). Enumeration versus multiple object tracking: The case of action video game players. [Electronic version]. Cognition, in press, 1-29. McKone, E. (2000). Capacity limits in continuous old-new recognition and in short-term implicit memory. Behavioural and Brain Sciences, 24(1), 130-131. Miller, G. A. (1956). The magical number seven plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 8197. Parkin, A. J. (1996). Spoken language impairments. In A. J. Parkin (Ed.), Explorations in cognitive neuropsychology (1st ed.) (pp. 129-153). Hove, East Sussex: Psychology Press. Tuholski, S. W., Engle, R. W., & Baylis, G. C. (2001). Individual differences in working memory capacity and enumeration. Memory & Cognition, 29(3), 484492....