Covington, N. 1 , Brown-Schmidt, S. 2 & Duff, M. 1
1 University of Iowa
2 University of Illinois at Urbana-Champaign
Schapiro and colleagues (2014) reported data from a patient with severe amnesia and extensive medial temporal lobe damage, including hippocampus, who displayed significant disruptions in statistical learning of stimulus triplets relative to a healthy comparison group (overall accuracy; patient = 41%, comparison group = 70%). These results, along with evidence from recent neuroimaging studies showing hippocampal activation during statistical learning tasks, point to a role for the hippocampus in statistical learning. The present study investigated the necessity of hippocampus for statistical learning by replicating the Schapiro and colleagues study with a severely amnesic patient with focal bilateral hippocampal damage. Participants completed a canonical statistical learning paradigm in four modalities: syllables, shapes, scenes, and tones. During an exposure phase, participants watched or listened to a continuous stream of stimuli composed of randomly repeating triplets for 4.8 minutes. Following exposure, participants completed forced-choice familiarity judgments composed of studied and novel triplets. Patient performance (accuracy collapsed across modalities = 68%) was similar to healthy comparison participants (N=4) (63%). These preliminary results suggest that hippocampus is not necessary for statistical learning. Hippocampal-implicated impairments may be more likely in situations that tap activities known to involve hippocampus such as flexible generalization or learning second-order bindings.