Abstract:
Subjects in two experiments were timed as they decided whether singly presented probe words belonged to one or the other of two memorized lists, or to neither list. In contrast to previous studies, they were required to respond by pressing different buttons for positive probes, depending on which list they came from, and by pressing a third response button for negative probes. Each list varied in length from one to four words. Reaction times increased linearly with the combined number of words in the two lists: they were not best explained by length of the list the probe was in or by length of the longest list. This suggests that subjects searched serially through both lists in succession, regardless of which list the item was actually in.
When there was no a priori basis for distinguishing the lists, the slope of the function for positive test probes was 33-35 ms per word higher than that for negative probes. The slope for negative probes was 58 ms per word in one experiment and 46 ms per word in another. These results suggest two possibilities: 1) Subjects first scan the combined lists exhaustively to determine whether the probe is there: if it is not, they make a negative response, and if it is they scan again to determine which list it is in, or 2) Subjects might simultaneously initiate two separate scanning processes over the combined lists. One process would be a fast exhaustive scan over the combined lists to determine whether the probe was a physical match of any item in either memory set. The other would be a slower self-terminating scan for whether the probe was in list A or list B. Entry to the combined lists would be random and the search would terminate when a match was found, rather than at a sub-list boundary; this would account for the lack of relationship of RT to the length of individual A or B lists in the various list-length conditions. Negative responses could be initiated at the end of the faster scan.
Both of these possibilities would explain the steeper slope for positive than for negative responses obtained in two list groups where the lists were not conceptually distinct. The simultaneous scan alternative is, however, somewhat more satisfactory. Positive responses were actually faster than negatives at the smallest set sizes, and this cannot be explained easily if negative responses require only one scan and positives require a second, sequential scan. Also, the simultaneous scan alternative ties in logically with the previous work of Burrows and Okada.
When the words in the two lists were taught as conceptually distinct (one list representing animate and the other inanimate objects), the difference in slope between positive and negative responses was reduced to only 6 ms per word (33 ms for positive responses: 27 ms for negative). The data suggest that subjects still scan over a concatenated list on most trials but that, after a single scan for positive/negative, they were able to make decisions about which positive response button to press by making a category check, requiring a constant time regardless of set size, on most trials. On some relatively small proportion of trials categorization may have allowed subjects to restrict scanning to the category that the probe was in.