Proactive Interference, Retroactive Interference -- What About Self-interference? A New Interpretation of the Recency-Primacy Shift

Eugen Tarnow, PhD


Medscape General Medicine. 2005;7(1):5 

In This Article

Abstract and Introduction

The recency-primacy shift (RPS) indicates that memory for early list items improves and memory for later items becomes worse as the retention interval between study and test increases. In this contribution, this puzzling experimental finding - memory improving with time -- is found to be consistent with a model in which recognition is temporarily interfered with by its own storage process (self-interference). I show that this interpretation can qualitatively better account for the RPS experimental data than can the dimensional distinctiveness model, the only other outstanding explanation of the RPS. Two experimental predictions separate the 2 models: The dimensional distinctiveness model predicts no RPS for 2-item lists, in contrast to self-interference, and as the overall timescale is changed, the dimensional distinctiveness model predicts no difference in the RPS whereas self-interference predicts significant changes.

It is recognized that "the naïve layperson might expect psychological theories of memory to make detailed quantitative claims about the course of forgetting.[1]" Similar frustration is echoed elsewhere: "not one detailed effort to grapple with the theoretical implications of Jost's law [an older memory trace will decay less rapidly in a given period of time than a younger one] can be identified in the 20th-century literature (or, so far, in the 21st-century literature).[2]"

Curve fitting may be the key to progress. In the hard sciences, curve fitting to experimental data is of fundamental importance: The data can elicit testable theoretical models and theoretical models can be tested by new experimental data. What emerges is a powerful synergy that allows for an efficient development of theory and experiment. In the memory field, curve fitting to discriminating data is a relatively new endeavor. It was not until 1996 that an effort was made to fit the various retention data with different types of curves.[3] At that time, it was realized that the experimental data were not good enough to show a preference for an algebraic decay, an exponential decay, or even an asymptotic nondecay. Once the curve fitting effort had begun, the lack of quality experimental data bothered apparently nobody but the investigators, who then took it upon themselves to come up with a much better set of experimental data with small enough error bars to be theoretically useful.

The data[1] probed recall at time intervals of 6 seconds. At smaller time intervals than 6 seconds, an interesting anomaly in experimental memory research, the recency-primacy shift (RPS), occurs under some circumstances.[4,5,6,7] Typically, when we try to remember serially presented items, the most recent item is the one best remembered (referred to as recency). However, in some experimental situations, the first item is the one remembered (referred to as primacy). Although the data on RPS are not as accurate as other data,[1] they seem to show conclusively the existence of the RPS in some experimental situations.

The existing attempt at an explanation in the literature for the RPS is the dimensional distinctiveness model (DDM).[5] This model suggests that the most distinct stimuli in a series are the first and last, and that recall the first and last stimuli is better than recall of the stimuli in the middle. DDM does not, however, produce the RPS for stimuli presented at constant intervals as used in experiments, but only for stimuli presented at successively longer intervals (a "decreasing schedule[7]"). One does not find this mentioned in the literature.[5]

Rather than starting from the DDM, this contribution goes back to the fundamentals of interference theory. Apparently for the first time, a general equation is written down that includes the effects of both proactive and retroactive interference; as a result, an extra term is given to us by mathematics, namely, the self-interference term. This term seems to explain the RPS data reasonably well and is interpreted as the presence of a memory-storage process that interferes with recognition about 5 seconds after stimulus presentation.