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

A sample of 210 published data sets were assembled that (a) plotted amount remembered versus time, (b) had 5 or more points, and (c) 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, 1989, 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, philo-