Scanning Short Term Memory

You will be shown a sequence of items which you are to remember. After an interval, a   ?   followed by an item will appear. Say   Yes   if the item was in the original sequence of items that you saw. Say   No   if it was not in the original sequence of items that you saw. Make your decision as quickly and accurately as possible. Now click on the image to the right labeled Experiment 1 to begin the experimental trial.

	This single trial illustrates one of the more well-known experiments that uses reaction time to measure the speed of a basic mental process. In this case, the process is presumably the scanning of short term memory for a matching item. The basic experimental paradigm involves: selecting a set of items, in this case the integers 0 through 9; constructing subsets of this set that vary in size where the items within the subset occur in a random, as opposed to numeric order. The subset is then presented one item at a time to the subject. Then, after an interval, the target item is shown and the subject must determine as quickly as possible whether or not that item was in the original subset.      If we assume that there is a component of the brain that is used to temporarily store a set of items, a short term memory (STM), then this experiment provides data on the time that is required to access, search and match a target item to the items that have been stored in STM. In this type of experiment, Sternberg found that the time to determine whether or not the target was in the original set was a linearly increasing function of the number of items in the original set. He considered two types of scanning processes that might yield this result. The first he termed exhaustive search. In this type of process all of the items are examined before an answer is made.
The theoretical predictions associated with this type of scanning process are shown at the top of the figure below. Note that there is no order effect associated with this type of search. Since all of the items are examined, the position of the target item in the original list will have no effect on the reaction time.
The other type of scanning process is referred to as self-terminating search. The predictions associated with this type of process are shown at the bottom of the figure above. In this type of process, the search terminates as soon as a match is found. If the search is serial, then on average the target item will be found after half of the target list has been searched. This is the reason why the curves diverge as a function of the length of list for the self-terminating search but remain parallel for the exhaustive search. In both types of search, if the target item is not on the list, then all of the items must be looked at. But, if it is on the list, then for self-terminating search only half, on average, of the items need to be scanned. Thus, the differing slopes of the Negative and Positive Target curves.      Sternberg's evidence was for parallel curves with a slope of roughly 40 msec. per item. He interpreted this as consistent with an exhaustive search. This result ran counter to many people's intuitions, and this added to the interest in this type of research. (Psychologists tend to be particularly impressed if an experimental result runs contrary to their intuitions. Whether or not this is a desirable trait is another story for another time.)      Now, recall the criticism of the subtractive method. Sternberg wanted to determine that this scanning process was a basic component of the mind/brain. He had shown that one variable, the length of the list to be scanned, affected this process in a very systematic fashion. Did this mean that the scanning process was a basic component of the mind/brain?      Well, the logic to answering this question is this. If, it is argued, one can find other independent variables that could potentially affect this component, and they affect this component independently of each other; then one's belief that you have found a basic component is strengthened. Why, you may ask? Well, I don't really know. There seems to be a bias here that in order for something to be "basic" it should be independent of other components...independence needn't be a prerequisite of what one decides to call "basic"....but its a nice thing, so...why not? Perhaps the bias reflects the preference for linear functions.
	Well, now to testing for independence to increase our confidence that we have a basic component. What we need is another variable. If you click on the image labeled Experiment 2 on the left then you can do a trial from another condition which will illustrate the variable that Sternberg chose to study.

This was the same type of trial that you did above except that we have made the contrast between the items and the background somewhat less pronounced. This type of manipulation Sternberg referred to as stimulus degradation.      The figure to the right illustrates the rationale for this study. Sternberg thought of this task as involving three basic processes. The encoding of the test stimulus (shown in blue); the exhaustive search process (shown in green); and the decision process (shown in red). Now, it was argued that the stimulus degradation manipulation could affect both the encoding and the search process. If it affects the search process, then it should add to the time it takes to make each comparison. If this is the case, then these two variables--stimulus degradation and list length would interact...that is, the amount of time it takes to make a decision will depend not only on the number of items but also on the degree of stimulus degradation.      The presence of an interaction of these variables would then lessen the confidence that these folks would have in their identification of the basic processes involved in this task. Luckily for them, stimulus degradation did not interact with list length and they felt it appropriate to continue to believe that they had not only identified a basic process of the mind/brain; namely, scanning short term memory for a match; but also identified the nature of the process--exhaustive search--and the speed with which an individual comparison could be made.      To appreciate in a more detailed fashion this argument about independence and the use of the interaction test in an analysis of variance as a test, you are encouraged to look over the page on Additive Factors and Analysis of Variance.

Back to Mental Olympics		Experimental Decomposition of Mental Processes
© Charles F. Schmidt