You are viewing the SNL 2018 Archive Website. For the latest information, see the Current Website.

Poster E31, Saturday, August 18, 3:00 – 4:45 pm, Room 2000AB

Anterior temporal lobe in identification of specific entities

Nicholas Riccardi1, Rutvik Desai1;1University of South Carolina

Introduction: The anterior temporal lobe (ATL) has been shown to have an important role in the semantic system. Evidence from neuropsychological and neuroimaging investigations have also demonstrated its role in several other domains, including social and emotional processing, naming of unique entities, sentence processing, and memory. Some past studies have used associative learning paradigms wherein facts are learned about novel stimuli. ATL involvement is attributed to the retrieval of this semantic or social information. Here, we examined the role of the ATL in the identification of specific familiar entities, while controlling for other factors such as semantic retrieval. We hypothesized that the ATL would be involved in the identification of familiar compared to novel entities, even when these conditions are matched for semantic or social content, demonstrating ATL involvement in memory and identification of specific entities separate from semantic retrieval. Methods: Twenty-six healthy adults were trained to recognize 144 previously unfamiliar stimuli from six categories (objects, persons, buildings, words, nonwords, numbers), with no additional information attached. Twenty-four hours after training, subjects were presented with blocks of familiar stimuli along with blocks of novel stimuli from the same categories and had to judge them as either familiar or unfamiliar while undergoing fMRI. Searchlight (5 mm radius) multivoxel pattern analysis (MVPA) was used within a bilateral ATL mask to train a linear support vector machine (SVM) model to classify stimuli as either ‘familiar’ or ‘unfamiliar’, using each participant’s time course-dependent voxel activations as input. Classification results were evaluated within-subject using 6-fold, leave-two-out cross-validation wherein the model was trained on the voxel patterns from five familiar and five novel blocks of the same category and then asked to classify the two remaining blocks as either familiar or novel. Prediction accuracies were based only on test data and were independent of the training set. To evaluate significance of cross-subject accuracy, the accuracies of every within-subject analysis were averaged for each category separately and subsequently tested against a simulated binomial cumulative distribution with a sample size of n=26 and 2-class classification. Informative voxel clusters were identified via t-tests against chance accuracy, and voxelwise thresholding at p < 0.001. Results: The model was able to accurately discriminate between familiar and novel stimuli for each category: persons (82.7%, p <0.001), buildings (82.7%, p < 0.001), objects (78.2%, p < 0.005), words (80.4%, p < 0.005), nonwords (83%, p < 0.001), and numbers (79.5%, p < 0.005). Searchlight analysis revealed cross-subject clusters of informative voxels for each category, with clusters in the left ATL being associated with verbal stimuli and right ATL with pictorial stimuli. Conclusions: The ATL contains information that identifies specific/familiar entities from novel entities that are visually similar, even when controlling for associated social and semantic content. This holds for a variety of categories, including those with low semantic or social value such as nonwords and numbers. These findings support a role for the ATL in the identification of specific entities that can be dissociated from social and semantic processes.

Topic Area: Meaning: Lexical Semantics