Modeling source-memory overdistribution

• We demonstrate that source overdistribution can result from guessing processes.
• A reanalysis shows that the one-high-threshold model accounts for overdistribution.
• We conduct a hierarchical-Bayesian modeling of two new experiments.
• The two-high-threshold model provides the best account for the two experiments.
• Overall results dismiss the need of a new model to account for overdistribution.

In a process-dissociation task of source memory, individuals have to judge whether items belong to one of different, mutually exclusive contexts (e.g., Source A, Source B). The acceptance rates to different test probes (e.g., “Source A?”) can be used to estimate the probability that the item is assigned simultaneously to the different contexts (“Source A and Source B”), designated as source overdistribution. Brainerd et al. (2012) have argued that source overdistribution can be used to refute traditional models of source memory such as the One or Two High-Threshold Source-Memory models (1HTSM and 2HTSM; Batchelder and Riefer, 1990; Bayen et al., 1996). We reanalyze previously-published datasets, including Brainerd et al.’s data, and show that there is no support for the rejection of the 1HTSM/2HTSM. Moreover, through a hierarchical-Bayesian model comparison using data from two new experiments, we show that the 2HTSM is not only able to account for source overdistribution, but also provides the best account of the data among different candidate models. These new results suggest that source overdistribution is an outcome of different guessing processes.