Semester of Graduation

Spring 2024


Master of Arts (MA)



Document Type



Complex visuospatial memory tasks typically require information to be encoded about individual item identities as well as their locations relative to one another and the bounds of the environment. Little is known about how viewers balance these competing goals when encoding time is limited, and whether the emphasis placed on one type of encoding relative to another influences memory accuracy. In two experiments, we examined the utility of hidden Markov models (HMMs) to address this gap by modeling visuospatial memory formation as a series of transitions in and out of two or more “encoding states” that differentially emphasize individual item details (item memory) versus spatial relations (relational memory) over 16-second study periods. Using fixation frequency and fixation duration as observed variables, we found evidence in both experiments that at least three distinct encoding states were operative during study, one of which was associated with behaviors consistent with item encoding and the other two with forms of relational encoding. Participants in both experiments also demonstrated a pronounced “item-to-relational shift” over the course of individual study trials, with the majority of item encoding occurring in the first half of each trial. Finally, the amount of time spent in the item state relative to the relational states was associated with memory accuracy in both experiments, although the nature of this relationship was inconsistent. Overall, these data demonstrate the potential for multistate modelling of eye movements to inform neurocognitive models of memory formation by providing insight into how learners prioritize and integrate multiple forms of information.



Committee Chair

Lucas, Heather

Available for download on Thursday, April 01, 2027