PRIMARY WORKING MEMORY STORAGE CAPACITY
Working memory refers to the temporary storage of information in connection with performing other, more complex, tasks (Baddeley, 1986). In the multi-store modal model of memory, it is the short-term memory (or short-term store) that acts as a key working memory system to allow information to transfer into the long-term memory (or long-term store) and thereby allow learning to take place.
As mentioned above, a number of findings, such as the lack of a positive (or negative) relationship between the amount of rehearsal of presented items and how well they are recalled from short-term memory (Craik & Watkins, 1973), cast doubt on the validity of the multi-store
modal model of memory, and in particular on the assumption implicit in this model that holding information in short-term memory (or the short-term store) necessarily leads to information transfer into long-term memory (or the long-term store). Furthermore, since in this modal model the short-term memory (store) acts as the working memory that is essential for learning, it would be expected that patients with short-term memory (store) impairment should manifest impaired long-term learning. However, Shallice and Warrington (1970) described the case of a patient with a severely affected short-term memory (store) who nonetheless had a normal long-term learning capacity; this person had a memory span of just two digits and almost no recency effect in the free-recall task (in which the subject is asked to free-recall as many of a previously presented list of unrelated words as he/she can, in any order). Moreover, when a short-term memory (store) deficit is experimentally induced in normal subjects by giving them digits to rehearse concurrently with a grammatical reasoning task, even with eight digits the reasoning time increases by only around 50%
and the error rate remains around 5% (the same as with fewer digits to rehearse) (Baddeley & Hitch, 1974), and this in spite of the fact that a digit load of eight should have totally filled the short-term memory (store) according to the multi-store modal model. Tasks using a similar digit load concurrent with comprehension and free-recall learning also show that the long-term memory (store) can be impaired but that the recency effect still occurs, again contrary to the modal model prediction.
In response to these difficulties, Baddeley and Hitch formulated the working memory model shown diagrammatically in Figure 1.5. The central executive is an attentional controller which is supported by two active slave systems, the articulatory or phonological loop, responsible for the maintenance of speech-based information, and the visuospatial scratch-pad or sketch-pad, which can hold and manipulate information in the visuospatial domain. This model was compatible with the findings mentioned in the previous paragraph. For example, concurrent verbal (articulatory) activity and visual or spatial activity appear to interfere with two different systems. Subjects using a mnemonic based on spatial location to remember word lists have better recall of the lists than those who use a simple rote rehearsal procedure, but this advantage disappears if the former subjects are required to carry out a visuospatial task concurrently (Baddeley & Lieberman, 1980). Again, a patient with gross impairment of digit span would be hypothesized to have a defect of their phonological loop functioning; if there were no co-existent impairment of the functioning of the central executive or visuospatial sketch-pad, then normal learning should still be possible.
Figure 1.5 The working memory model of Baddeley and Hitch. (After Baddeley & Hitch, 1974).
Working memory capacity can be measured using a task in which, after reading a series of sentences, the subject is required to recall the last word of each of these sentences (Daneman &
Carpenter, 1980). This task therefore requires both comprehension and recall (as opposed to just recall in the simple word-span task).
Central executive
Visuospatial scratch-pad or sketch-pad
Articulatory or phonological
loop
Factors affecting memory 21 There is a strong positive correlation between working memory tests and intelligence quotient (IQ) tests. Furthermore, standard intelligence quotient tests appear to be more susceptible to the subject’s previous knowledge, whereas working memory tests appear to have a greater relationship to the speed of processing (Kyllonen & Christal, 1990).
In terms of the model of working memory of Baddeley and Hitch (1974), the central executive is a limited-capacity system which provides the link between the two slave systems (see Figure 1.5) and long-term memory; it is responsible for planning and selecting strategies (Baddeley, 1986). The visuospatial scratch-pad or sketch-pad appears to have a visual component, which is concerned with such factors as shape and colour, and a spatial component, which is concerned with location (Baddeley & Lieberman, 1980; Baddeley, 1986). The articulatory or phonological loop also appears to be made up of two components: a memory store which can hold phonological information for one to two seconds; and an articulatory control process (or processor) (Baddeley, 1986; Paulesu et al., 1993). Memory traces in the phonological loop can be refreshed by means of subvocal articulation. (Subvocal or vocal articulation may also be used to provide an input into this slave system upon the visual presentation of objects by the subject articulating the names of those objects.) The phonological loop is held to provide the basis for digit span. In particular, the number of items retained in the digit span is believed to be a function of both the rate of fading of the memory trace in the phonological loop and the rate of refreshing of memory traces by means of subvocal articulation. The size of the storage capacity can be reduced in the following ways:
∑ The phonological similarity effect. This involves trying to remember items with similar sounding names.
∑ Presenting irrelevant spoken material. This gains access to the store and corrupts the memory trace.
∑ The word length effect. As the length of the words increases, the memory span decreases, presumably because of the longer time required for longer words to be rehearsed, leading to a greater probability of memory trace decay.
∑ Articulatory suppression. This requires a subject repeatedly to articulate an irrelevant speech sound which interferes with subvocal rehearsal.
THE PRINCIPLE OF CHUNKING
Chunking increases the amount of information stored in ‘short-term memory registers’ (in the multi-modal model) by allowing one entry to cover several items. While the number of chunks is restricted, their content is not. With the help of long-term memory, new material can be recoded thereby increasing the content of chunks.
For example, British trainee and qualified psychiatrists are unlikely to require eight ‘registers’ in order to remember the letters MRCPSYCH. Similarly, the string DSMIVTRICDEEG can readily be split into the chunks DSM-IV-TR, ICD, and EEG.
It has been suggested that the acquisition of skills through practice may involve the grouping of sets of mental entities (be they motor or perceptual) as chunks (Newell & Rosenbloom, 1981).
SEMANTIC MEMORY
Semantic memory refers to the subject’s knowledge of facts, language, concepts, and the like, and is an aspect of long-term/secondary memory that is consistent with the finding that verbal information is stored in terms of meaning rather than exact words (Tulving, 1972). It is easier to remember words paired with meanings (Bower, 1972), and to recall words synonymous to those in a given list (Sachs, 1967). Therefore semantic encoding is a more efficient way than simple rehearsal of transferring information from the ‘short-term memory’ to the long-term/secondary one.
EPISODIC MEMORY
Episodic memory is an aspect of long-term/secondary memory that refers to the memory for events. It provides a continually changing and updated record of autobiographical material (Tulving, 1972).
SKILLS MEMORY
Skills memory, or procedural memory, is an aspect of long-term/secondary memory that supports skilled performance.
OTHER ASPECTS OF LONG-TERM/SECONDARY MEMORY
Ryle (1949) distinguished between procedural knowledge and declarative knowledge; whereas the former referred to knowledge that supported the performance of tasks, the latter referred to factual knowledge. Tulving (1985) distinguished between autonoetic awareness (or remembering) and noetic awareness (or knowing).