Our previous recognition studies have shown a profound effect of negative probe recency on false alarm rate. Further work revealed this effect to be a function of the amount of intervening time, not of the number of intervening items. When false alarms are plotted against the amount of time since the item last occurred either as a probe or as a memory set item, an exponential decay function provides a better fit to the data than does a power function. In addition, the serial position data based upon positive probes demonstrates a recency effect but not a primacy effect. The fitted exponential function above provides an excellent prediction of this serial position function; the power function does not. We hypothesized that the phenomenon is based upon recognition in the presence of an episodic recollection of the occurrence of a probed item, and that the episodic recollection disappears with time. In the present study, participants indicated the presence or absence of such an episodic recollection. Results confirm our hypotheses and produce a replication of earlier results.
In the study of recognition memory, both Monsell (1978) and McElree and Dosher (1989) have demonstrated that a negative probe that has recently appeared on a previous trial is less likely to be rejected by subjects than one which was presented on a more distant trial. More recently, Boneau and Daily (1992, 1994) have extended these findings and demonstrated a persistent effect of stimulus-item presentation. This effect was manifested in a procedure that involved a Sternberg-like task utilizing memory sets of 18 items followed by 5 recognition probes. The result of interest was a prominent tendency for false alarms to negative probed items that had previously occurred, as high as 60-70% for negative probes on the trial immediately following the presentation of the item. This tendency to false alarm decays, but persists for approximately 6 or 7 trials (about 6 or 7 minutes). In another experiment we demonstrated that the effect and its decay are fundamentally a function of the time since the probed item was last seen and only fortuitously a function of the number of intervening items.
We were led to hypothesize that the probability of "yes" ("just seen") responses is a function of the time since the item was last seen. For positive probes, the time since occurrence would be short and the proportion of "yes" (in this case, correct) responses would be high. On subsequent trials with negative probes, the proportion of "yes" (in this case, false alarms) responses would decrease as the time since the probed item was last seen increases. In Figure 1 we plot the data from Experiment 1 as a function of the time since the probed item was last seen (response to positive probes, .82, is included as the first point). Here we have fitted the data with an exponential decay function and a power function. Further, if the decay is a function only of time, then we should see a tendency for earlier items in the memory set to show a smaller probability of "yes" responses on positive probes than for later items, because of the longer decay time; a serial position function should emerge. Figure 2 presents these serial position data along with predicted serial position functions constructed by using parameters from the exponential and power fits from Figure 1. These results prompted us to favor the exponential over the power function as a fit for our results.
Based upon our own experience in the experimental setup, we hypothesized that most "just seen" judgments occurred when the probed item produced an episodic recollection of the occurrence of the item in a/some list and that these episodic states declined in probability with the passage of time since the item had last occurred. The present experiment was conducted to test this hypothesis directly by asking the subject to respond "yes" only when episodic recollections occurred in the presence of the probed item.
The experimental procedure was identical to that of Experiment 1 except that the basis for the subject's judgment of the probed item was changed from "Probed item was present in most recent list" with feedback on accuracy to "Probed item produced an episodic recollection of its occurrence" with no feedback.
The subjects were 29 George Mason University undergraduate psychology majors enrolled in a class in Memory and Cognition. All subjects participated to fulfill a class requirement.
IBM AT compatible personal computers equipped with VGA monitors were used to present the stimuli, to control events, and to record the subject's responses and response times.
Subjects received a total of 48 trials. The memory set on each trial consisted of 18 words 4 to 6 letters in length selected from a pool of 230 words drawn from all categories of Thorndike and Lorge (1944). During the memory set presentation, subjects engaged in a semantic decision task to control rehearsal. Subjects judged whether or not the presented word had a necessary human involvement. Words were presented sequentially and were separated only by a .4-sec. feedback that occurred when the subject responded.
Following a brief pause at the termination of the memory set sequence in which subjects were+ informed of the change in tasks, subjects were presented with a sequence of 5 recognition probes to which they were to respond "Yes" only if they had an distinct recollection of having just seen the probed item, and "No" otherwise. No feedback was given for responses to probes.
Figure 3 shows the mean proportion of "yes" responses for probed items (both positive and negative) as a function of the time in seconds since the item was last seen. Also plotted is a baselevel for "Yes" responses to probed items that have no prior occurrences in the experiment. We take this to be the best estimate of the rate of responding that occurs in the absence of the episodic recollection, and in Figure 4 we have plotted the data corrected for this baselevel responding. Also plotted in Figures 3 and 4 are the negative exponential and power functions that best fit the data as presented. These fitted functions with their corresponding R2 values are presented in Table 1 along with the functions from Experiment 1 for comparison purposes.
These functions were used to predict serial position data based upon the average time from the offset of a memory set item in a specific serial position until the occurrence of a probe to that item on positive trials. Because the results for the two sets of functions are nearly identical, only data corrected for base rate "Yes" responding are shown (corresponding to Figure 4) along with the serial position predictions from the functions above. Also depicted is the straight line that best fits these data.
| Curve | Function | R2 |
|---|---|---|
| Exponential (Figure 1) | y=.1570+.843e^-0.0110t | .979 |
| Exponential (Figure 3) | y=.0582+.942e^-0.0086t | .987 |
| Exponential (Figure 4) | y=.0150+.985e^-0.0086t | .989 |
| Power (Figure 1) | y=2.94t^-0.442 | .945 |
| Power (Figure 3) | y=6.89t^-0.637 | .944 |
| Power (Figure 4) | y=8.42t^-0.700 | .931 |
Our findings from Experiment 1 were replicated with this very different task of reporting the presence of an episodic recollection rather than a judgment that the item did or did not occur on the most recent list. It seems reasonable to conclude that our hypothesis is confirmed 1) that judgments of occurrence in the last list are based principally upon the existence of episodic recollections of that occurrence, 2) that these episodic recollections are elicited by the occurrence of the item as a probe, and 3) that the capability of a probe to elicit the episodic recollection decays exponentially. These conclusions are bolstered by the ability of the exponential function to provide an excellent fit to the serial position data. Based upon these results, it seems reasonable to conclude that the presentation of an item either activates some kind of a trace or process which then undergoes exponential decay.
This latter result suggests a possible explanation for the recency effect almost routinely obtained in serial position functions. We have demonstrated earlier that probability of recall of an item is also a function of the time since occurrence of the item (Daily & Boneau, 1994) which suggests that the same process underlies both recall and recognition and that this process may be the one that underlies our phenomenon. A recent paper by Baddeley & Hitch (1993) suggests that the recency effect is attributable to explicit recall of implicit material. Our future work will attempt to relate our phenomenon to the implicit/explicit distinction.
We also are not sure of the basis for the exponential function we have obtained. Some theorists (Anderson & Schooler, 1991; Wixted & Ebbesen, 1991) have suggested that such forgetting curves (which ours could be construed to be) are best fit, not by exponential functions, but by power functions. The issue remains unresolved at this time.
Anderson, J. R., & Schooler, L. J. (1991). Reflections of the environment in memory. Psychological Science, 2, 396 - 408.
Baddeley, A. D., & Hitch, G. (1993). The recency effect: Implicit learning with explicit retrieval. Memory and Cognition, 21,146-161.
Boneau, C. A., & Daily, L. Z. (1992). Short-term recognition memory and LTM activation. (Paper presented at the meeting of the Psychonomics Society, St. Louis, November, 1992).
Boneau, C. A., & Daily, L. Z. (1994). Recognition accuracy for words and nonwords. (Poster presented at the meeting of the Eastern Psychological Association, Providence, April, 1994).
McElree, B., & Dosher, B. A. (1989). Serial position and set size in short-term memory: the time course of recognition. Journal of Experimental Psychology: General, 118, 346 - 373.
Monsell, S. (1978). Recency, immediate recognition memory and reaction time. Cognitive Psychology, 10, 465-501.
Thorndike, E. L., & Lorge, I. (1944). The teacher's word book of 30,000 words. New York: Teacher's College Press.
Wixted, J. T., & Ebbesen, E. B. (1991). On the form of forgetting. Psychological Science, 2, 409 - 415.
Return to Larry Daily's Home Page
Mail comments to:
Larry Z. Daily
Copyright © 1995 Alan Boneau & Larry Z. Daily
