Set up a searchlight on multiple object tracking: functional determinants of training intervention

Poster Session F - Tuesday, April 16, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Anna-Maria Felßberg^1,2 (anna.felsberg@med.ovgu.de), Christian Merkel^1,2, Nadine Schönemann^1,2, Mircea Ariel Schoenfeld^1,3, Jens-Max Hopf^1,2; ¹Medical Faculty Otto von Guericke University, Magdeburg, ²Leibniz Institute, Magdeburg, ³Kliniken Schmieder, Heidelberg

Previous research described two distinct behavioral patterns in Multiple Object Tracking, linked to object-based and location-based mechanisms (Merkel et al., 2015). A linear relation between performance and level of probe-target-congruency signifies a location-based representation of the individual targets. Error rates and reaction times decreased significantly when a probe encompassed all targets, indicating an object-based representation. This study focuses on representation-specific learning processes during tracking. 36 subjects underwent two imaging sessions in a 3T scanner, to track four out of eight items covertly. In-between they received three training sessions. Participants had to indicate whether a probe showed full congruence or not. Functional data were analyzed using Multiple Voxel Pattern Analysis with a 5mm searchlight and SVM classifier, trained to binary discriminate between increasing congruencies to the m0 and the m4 condition. For areas encoding the objecthood of the entire target-set, the discriminability between any match-condition and m4 should be equally high, whereas a location-based encoding would exhibit an increase in discriminability between conditions the more they differ in their match. Error rates and reaction times showed congruence-dependent increases, with exception for full matches. A repeated-measures ANOVA revealed a significant main effect of session, indicating overall improvement after training, confirmed by Bonferroni-adjusted paired t-tests for all conditions. The classification maps were subjected to a regression with the linear and the deviant function as covariates. The t-values for every covariate at each time point were examined. Tracking-related patterns for both representations within the visual cortex exhibited shifts towards higher-order cortical areas after training.

Topic Area: ATTENTION: Spatial