How to be proactive about interference: lessons from animal memory

Abstract

Processes of proactive interference were explored using the pigeon as a model system of memory. This study shows that proactive interference extends back in time at least 16 trials (and as many minutes), revealing a continuum of interference and providing a framework for studying memory. Pigeons were tested in a delayed same/different task containing trial-unique pictures. On interference trials, sample pictures from previous trials reappeared as test pictures on different trials. Proactive-interference functions showed greatest interference from the most recent trial and with the longer of two delays (10 s vs. 1 s). These interference functions are accounted for by a time-estimation model based on signal detection theory. The model predicts that accuracy at test is determined solely by the ratio of the elapsed time since the offset of the current-trial sample to the elapsed time since the offset of the interfering sample. Implications for comparing memory of different species and different types of memory (e.g., familiarity vs. recollection) are discussed.

Fig. 1

Example trial sequence for proactive-interference testing. In this example, an interference stimulus is presented as the sample on trial n − 1. This same stimulus appears again on the next trial (n) as the test stimulus. The correct response on trial n is a “different” response (a peck to the white rectangle). However, having seen this stimulus on the previous trial increases the chances of an error (making a “same” response by pecking the test picture) due to confusing the previous sample (on trial n − 1) with the current sample (on trial n). On each trial, pigeons were required to peck the sample picture 20 times (a fixed-ratio 20, or FR 20, schedule) before the delay began. The relative sizes of the pictures, white rectangle, and monitors in this illustration are not to scale.

Fig. 2

Mean results of 4 pigeons on the same/different task at 1-s and 10-s delays between the offset of the sample stimulus and onset of the test stimulus. Percentage correct is shown as a function of the number of trials separating the interfering stimulus and the test. On the left of the x-axis, “n − 1” refers to the condition in which the interfering stimulus occurred on the immediately preceding trial. On the right of the x-axis, “no-PI” refers to the no-interference, or different-trial baseline, condition. Error bars represent standard errors of the mean.

Fig. 3

Model and data fit. In our signal detection model of proactive interference (a), the signal is given by the difference between the log of the retention-delay time (log T_C) and the log of the time between the offset of the interfering sample presented in a previous trial and the onset of the test-trial picture (log T_I). The observer’s internal representation of elapsed time is noisy, as indicated by the width of the normal distributions. The graph (b) shows the mean model fit to the data from Figure 2. The model was fit simultaneously to the proactive-interference functions for both delays for each individual subject. On the right of the x-axis, “no-PI” refers to the no-interference, or different-trial baseline, condition. The shaded areas show 1 SEM of the individual fits (see Model Fitting in the Supplemental Material).