Dilution as a model of forgetting

Dilution as a model of forgetting 2 Abstract We develop a model of long term forgetting based upon three ideas: i) Memory for a stimulus can be described by a population of accessible traces; ii) The probability of retrieval after a delay is predicted by the proportion of traces in this population that will be defined as correct if sampled; and iii) This population is diluted over time by null traces that, if accessed, block retrieval. We model this dilution as a linear function of time and the outcome of accessing memories by their temporal organisation. We apply our model to five published experiments studying forgetting in cued recall, four recognition experiments, and one using savings methods. The model specifies the different form of the retention function in each case well, and provides a principled explanation for some puzzling characteristics of forgetting without recourse to mechanisms such as decay or consolidation. Dilution as a model of forgetting 3 A recent article by Wixted (2004) is one of many on the nature of the forgetting Chater (2007); among many others. The mathematical specification of the forgetting function endures as a central theoretical issue because it has proven difficult to resolve convincingly and because it is seen as an important constraint upon how fundamental processes of memory are modelled. In his article, Wixted specifically argues that the significance of Jost " s Law 1 to this debate has been neglected. This law states that of two memory traces of equal strength at a given time, the younger trace will decay more rapidly than will the older. This is illustrated in Figure 1 in which two stimuli, A and B, are learned at different times and to different degrees, with Stimulus A being the earlier and receiving more learning trials. Typically, the retention function for Stimulus B learned at the later time declines faster and eventually intersects that of A; with Jost " s Law, as stated above, applying clearly at that point of intersection. For the purposes of this article, the significance of Wixted " s argument is that the available evidence constrains the possible theoretical explanations of Jost " s Law to models that assume that the proportional likelihood of forgetting decreases with the age of the trace. Rather than being a footnote in the history of memory research, or trivial in the sense that it follows merely from the curvilinear …