Working Memory Model
_____________________________
| |
| Central Executive |
|___________________________|
/
/
_______/________ _______\________
| | | |
| Phonological | | Visuospatial |
| Loop | | Sketchpad |
|______________| |______________|
(Willingham, 2001)
According to Baddeley (1992), working memory is the "mental workbench" that is the site of applied conscious mental effort. Information in working memory is encoded into meaningful chunks. It is an exciting place where work happens, problems get solved and things come into the mind. The model has an executive control system, the central executive, and two "slave" systems, the phonological loop and the visuospatial sketchpad. The central executive plans future actions, initiates retrieval of long-term memories (LTM), uses decision-making processes and integrates new information. The phonological loop is the speech and sound related component of working memory and holds verbal and auditory information. The visuospatial sketchpad holds visual and spatial information.
This model helps explain how we solve algebra problems. Say we have the problem (3+4)*2/3+(12/3). First of all, the central executive retrieves rules and facts about solving mathematical problems. Then it chunks the problem into manageable units, and uses the phonological loop to temporarily store information. Thus, solving the problem would go something like this: add 3+4 using facts and rules retrieved from LTM, store resulting 7 in phonological loop. Multiply 7x2 and store resulting 14 in phonological loop. Divide 12 by 3 and add resulting 4 to 3 yielding 7. Divide 14 by 7 and get 2.
According to the model, there are three major limitations of the slave systems. The first limitation is that the slave systems are only responsible for low-level processing of simple information. The second limitation is that the slave systems are highly domain specific (i.e., they are specialized for particular codes or types of information). The third limitation is that they have limited attentional resources (i.e., limited capacity). If the slave systems become overwhelmed, they may get more resources from the central executive.
The Dual Process component of the model overcomes some of the limitations of earlier models because it allows for a division of tasks into primary and secondary tasks. For any two tasks performed together, there are three possible outcomes: the tasks are independent of each other, the tasks are dependent on each other, or the tasks are partially dependent on each other. If the tasks are independent of each other, then the tasks use separate mental mechanisms. If one task disrupts another task, then the two tasks use the same mental mechanisms. If the two tasks interfere with each other in some conditions, then partial sharing is occurring. In order to test whether two tasks are dependent or independent of another, we can vary the difficulty of each task separately and observe when performance begins to suffer.
Baddeley and Hitch (1974) performed such an experiment. They showed subjects a simple letter combination such as "AB" then asked increasingly complex and difficult questions about it. They also had subjects do an increasingly difficult "articulatory suppression task" while they answered the questions. The articulatory suppression task was either to repeat "the", repeat the numbers from 1 to 6 or to repeat random numbers. When there was no suppression task, reaction time increased as the questions became more difficult. There was little difference in reaction times between the "the" group and the "1-6" group. The group that had to repeat random numbers had much worse performance. This was interpreted to mean that the central executive was drained by the task and therefore less able to reason and understand the increasingly complex questions.
According to Willingham (2001) the phonological loop has two parts: the phonological store and the articulatory control process. The phonological loop can repeat about 2s of auditory information and the articulatory control process writes information into the phonological store. This is the supposed mechanism of "self-talk." Baddeley, Lewis and Vallar (1984) found evidence for the articulatory nature of the phonological loop. Participants were asked to repeat "blahblahblah" while looking at a list of words to be remembered. When these subjects were compared with subjects who only tapped their fingers, it was found that the "blah" subjects did not confuse words like "mad, man, mat" while the tapping subjects did. This result indicates that the "blah" subjects were using the resources of the phonological loop so the similar sounding words were not translated into sounds and therefore not confused.
Other research on the phonological loop has found that its capacity is highly correlated with vocabulary in children. Baddeley, Gathercole and Pagano (1998) reviewed a number of studies investigating this correlation. The researchers found correlations between digit span and vocabulary size and between ability to repeat long nonwords and vocabulary size.
The visuospatial sketchpad appears to have separate processes for determine what something looks like (visual) and where it is located (spatial). Baddeley and Lieberman (1980) designed an experiment to test this hypothesis. Subjects had to track a swinging pendulum with a flashlight. At the same time they had to do another spatial task. Other participants had to perform a task that was visual, but not spatial. It was found that performance suffered more when both tasks were spatial, indicating that visual and spatial information is processed separately.
The Working Model of memory also accounts for chunking better than the Modal Model. Evidence for chunking was found by Chase & Simon (1973). They gathered subjects who were either experts or non-experts at chess. The subjects were given 5 s to view pieces on a chessboard of a game in progress. Later they were asked to rearrange pieces on a board as they had seen them. Novices were able to get 8-10 pieces correctly, while experts usually got all of the pieces correct. When the pieces on the board were arranged randomly rather than as a game in progress, both the novices and experts had similar recall ability. This indicates that experts were able to chunk the information into larger meaningful units because of past experience playing chess and not because of superior memory capacity.
Given this experimental evidence, it seems that the Working Memory model is superior to the short-term model in that it includes everything in the short-term model and takes into account some things the short-term model has difficulty explaining (e.g. solving algebra problems).
References
Willingham, D.B. (2001). Cognition: The thinking animal. Upper Saddle River, NJ: Prentice Hall.
