One hundred years of forgetting: A quantitative description of retention.

Abstract

A sample of 210 published data sets were assembled that (a) plotted amount remembered versus time, (b) had 5 or more points, and (cj were smooth enough to fit at least 1 of the functions tested with a correlation coefficient of .90 or greater. Each was fit to 105 different 2-parameter functions. The best fits were to the logarithmic function, the power function, the exponential in the square root of time, and the hyperbola in the square root of time. It is difficult to distinguish among these 4 functions with the available data, but the same set of 4 functions fit most data sets, with autobiographical memory being the exception. Theoretical motivations for the best fitting functions are offered. The methodological problems of evaluating functions and the advantages of searching existing data for regularities before formulating theories are considered.At the simplest level, this article is a search for regularities. We ask whether there is one retention function that can describe all of memory, or perhaps a different function for each of a small number of different kinds of memory. At a more abstract level, it is about the role of theory and data in psychological research. Can we most rapidly advance psychology as a science by developing theories at the level that commonly fills psychological journals such as this one, or should we first try to describe phenomena that could constrain theories by establishing robust, preferably quantitative, regularities (Rubin, 1985(Rubin, , 1989(Rubin, , 1995? A balance between these alternatives is needed, and here we argue that to obtain such a balance more description is needed.Retention offers the ideal topic to make this abstract, philoDavid C. Rubin, Department of Experimental Psychology, Duke University; Amy E. Wenzel, Department of Psychology, University of Iowa.We thank John Anderson, Alan Boneau, John Cerella, Herb Crovitz, Sean Hinton, Armando Machado, Bennet Murdock, Matt Serra, Harold Schiffman, John Staddon, Wayne Wickelgren, and John Wixted for their suggestions; Richard Palmer for information about the exponential-power function and its history and use in physics; Michael Reed for mathematical advice and help with derivations; Robert Terry for statistical advice; and Jenny Zervakis for help

Assessing the distinguishability of models and the informativeness of data

“…A total of 77 data sets from 16 studies were used in the landscaping analyses. With the exception of five data sets from Rubin et al (1999), they are a subset of those used by Rubin and Wenzel's (1996) review (Some of the methodological details from the studies are presented in Appendix Table 1. Fits of five retention models using ln(ML) to five data sets of Rubin et al (1999;top), the average fit to the remaining 72 data sets (bottom).…”

“…We begin by summarizing recent work comparing models of retention. Rubin and Wenzel (1996) attempted to provide some clarity on the form of the retention function by performing a data-fitting meta-analysis. Some 105 two-parameter functions were fit to 210 data sets that were collected throughout the 20 th century.…”

“…For example, too few variables may be measured, or a poor task might be inadvertently chosen. When trying to distinguish between retention functions, it is virtually pointless to have only three or four retention intervals, since any reasonable function will provide an excellent account of the data (Rubin & Wenzel 1996).…”

The psychology of intelligence analysis: Drivers of prediction accuracy in world politics.

“…There is a long tradition of work on forgetting functions in the adult literature (e.g., Ebbinghaus, 1885;Rubin & Wenzel, 1996;Wixted & Ebbesen, 1991 and in the developmental literature there have been examinations of the variance in long-term recall that can be explained by forgetting shortly after experience of an event (e.g., Bauer, 2005;Bauer, Van Abbema, & de Haan, 1999;Howe & O'Sullivan, 1997; discussed in more detail below). However, until the beginning of the 21st century, there were virtually no data that directly addressed the suggestion that the characteristic distribution of autobiographical memories across the life span-including the phenomenon of childhood amnesia-could be explained in part by the relative vulnerability of memory traces formed early in life.…”

Time scales of memory, learning, and plasticity

“…To answer this question, one should first consider why and how the nervous system forgets. Huge data sets exist on human forgetting [for a review see, for example, (Rubin and Wenzel 1996)] that show, as expected from everyday life, that we forget with higher probability as time passes. The exact shape of this forgetting curve is still under debate [exponential or power law; (Wixted and Ebbesen 1991;Anderson and Tweney 1997)].…”