Consider an experiment in which subjects are asked to make a recognition judgment—that is, to decide for each in a list of stimuli whether they believe to have seen that item in a particular earlier study episode. One subject might not care much about the advancement of science, want to get out and get to lunch, and thus zip his way through our task as quickly as humanly possible, making each decision after only the least amount of deliberation. Another subject might feel as though the experimenter will treat her score as a measure of intelligence, character, or trustworthiness and thus pore over each test stimulus to extract every available mote of information from memory before making a recognition decision. Such individual differences are commonplace in decision tasks like this one. Even if we use some between-subject manipulation of learning, for example, we have faith that random assignment will wash away such strategic differences over our sample.
But what if our entire sample was like the first hypothetical subject described earlier? This scenario is not entirely unlikely at many major American universities. Our laboratory might be aesthetically unappealing, or our experimenters might have bad breath; such factors can also influence strategy selection in our subjects.
Hypothetical group means are shown in the top panel of Figure 24.1 and indicate no effect of our learning manipulation. It would be useful to know if there is a restriction placed on our data by an inadequate range of decision speeds. In this case, all subjects performed the task quickly, but we have no way of assessing that fact. Even if we measured decision response time (RT), we would be ill equipped to make any such judgments without a sense of what the "full" parameter range of response speeds should be. The solution to this problem is to create a within-subjects variable along which we manipulate the decision placement along the speed-accuracy trade-off spectrum. We might, for example, use payoffs for different combinations of correct or speedy decisions. We might simply instruct the subjects to make decisions quickly or to take their time. Perhaps most effectively, we can force subjects to withhold their response until a delimited amount of time has elapsed and then force them to make their response within a given time window (Reed, 1973). If we use such a strategy, we ensure the collection of performance data across a reasonable range of decision speeds. We
ICatuMon 1 Condition 2
Was this article helpful?