The Environmental Basis of Memory

Memory should make more available things that are more likely to be needed. Across multiple environmental domains it has been shown that a system that behaved in this way would match memory effects evolving repetition, delay, and spacing (Schooler & Anderson, 2016). We examined the patterns with which words appear in two large-scale data sets: tweets from popular sources and comments on popular subreddits. These sources are sufficiently large that they enable accurate measurement of how detailed patterns of past appearance predict probability of occurring again. None of a candidate set of mathematical models of memory does very well at predicting the observed patterns. Anderson & Milson (1989) had proposed a model of the environment which assumes that there is a distribution of desirability of items, that their desirability decays according at different rates, and that items occasionally have revivals to their original desirability. While Monte Carlo simulation of the original model, which assumed exponential decay, does not fit these two sources particularly well, simulations of a model which assumes power-law decay does. A mathematical model of human memory based on the revised Anderson and Milson environmental model did better than other models at predicting the environmental data and a wide range of behavioral studies focused on how spacing of items affects probability of recall. This model also predicts latency effects associated with presentation patterns at frequencies that are too high to study with probability of recall.