Predicting Perceptual Performance From Neural Activity

Abstract

A key aim of neuroergonomics is to gain an understanding of human neural function in relation to cognitive and behavioral performance in real world tasks. Here we investigated the relationship between neural activity and human performance in a rapid perceptual categorization task. We compared two different modalities to indirectly measure neural activity, functional magnetic resonance imaging (fMRI) and electroencephalography (EEG). We applied a multivariate pattern classifier to predict the individual variability in perceptual performance during a difficult visual categorization task using single trial EEG and fMRI separately. Twenty observers perceptually categorized images of cars and faces embedded in filtered noise while EEG activity from 64 electrodes was concurrently recorded. Another twenty observers performed the same task while their fMRIBOLD responses were recorded. Our results showed significant correlations between the neural measures and perceptual performance (p <0.05; r = 0.69 for EEG; r =0.66 for fMRI). We were able to reliably identify from their neural activity the best performing individual from two randomly sampled observers (84% for EEG; 75% for fMRI; chance = 50 %). Finally, we demonstrated that EEG activity predicting the performance across individuals was distributed through time starting at 120 ms and sustained for more than 400 ms post-stimulus presentation, indicating that both early and late components contain information correlated with observers’ behavioral performance. Together our results highlight the potential to relate individual’s neural activity to performance in difficult perceptual tasks and show a convergence in predictive ability across two different methods to measure neural activity.