When people recall items from a specific category (e.g., animals), retrievals occur sporadically, in bursts, decreasing over time. Efforts to gain insight into the nature of retrieval from this fundamental task have focused primarily upon the exponential growth rate of cumulative recall. The dissertation research focused on the inter-retrieval time intervals (IRIs). The guiding hypothesis was that the Lévy flight dynamics of animal foraging behavior might be found in human recall.

In Experiment 1 eight participants recalled exemplars of the category "animals" for 20 min. The exponent of the normalized log-log plot of IRI frequency against IRI for each participant fell between 1 and 3, as expected of a Lévy distribution: recall and foraging seem to be dynamics of the same kind. Given the limitations of regression methods for determining distribution functions and exponents, attention was directed to (a) information-based statistical methods for distinguishing among candidate distributions and (b) numerical methods for deriving the dynamics.

In Experiments 2A and 2B, three participants retrieved from a category and subcategory (higher and lower semantic density, respectively) for one hour each. Analysis through the described methods revealed that the distributions were lognormal (not normal or Pareto) with the implied multiplicative processes confirmed by auto-regressive conditional heteroskedasticity analysis. Analysis of epochs within individual trials and comparison over density conditions both suggest that retrieving from less dense domains (sparser environments) is characterized by higher a (lower μ).

Experiment 3 evaluated the density factor and the fluency (semantic versus verbal) factor with 12 participants retrieving from the categories animal, mammal, and words beginning with t, in three separate trials of 30 min each. Results of manipulations of density were consonant with the results of Experiment 2. Manipulations of task type over different classes of fluency suggest subtly different dynamics, requiring more rigorous dynamical modeling.

Discussion focused on the following: (a) equivalence of foraging and recall phenomena, (b) consequences of an ontology grounded in animal-environment systems, (c) significance of multiplicative models, (d) relevance of the results to human memory in general, (e) combining top-down (distribution identification) and bottom-up (dynamical modeling) methods, and (f) future research directions.