Predicting implicit abstract stimulus attributes from patterns of event-related potentials

• ¹ University of Melbourne, Melbourne School of Psychological Sciences, Australia
• ² University of Cologne, Department of Psychology, Germany
• ³ University of Melbourne, Department of Finance, Australia

Background
In everyday life, not every piece of perceived information can be processed in a conscious, deliberate fashion. Automatic brain activity below the level of conscious awareness has been linked to preferences for unattended stimuli, and can predict preference-based decisions. However, it is an open question whether automatic, attention-independent encoding is restricted to simple value judgements, or whether it also extends to more abstract stimulus characteristics. Here we investigated whether participants’ post-experiment ratings of task-irrelevant, positive background visual stimuli for the dimensions of a) arousal and b) subjective chronology (time) could be predicted from patterns of event-related potentials (ERPs) during image viewing.

Methods
64-channel electroencephalogram (EEG) was recorded from 21 participants (18-31 years) performing a demanding foreground choice-reaction task while one of 24 images (depicting objects, people and scenes) were presented in the background. After the experiment, participants rated the images with respect to a) how arousing they are, and b) how strongly they are related to the present or future. A multivariate linear support vector regression (SVR) analysis was performed to predict participants’ ratings from EEG data.

Results
The arousal and time ratings were uncorrelated (mean absolute r = .17, n.s.). No significant differences between rating increments were found for single-channel ERPs. The SVR analysis, however, showed that both arousal and time ratings could be predicted from spatiotemporal patterns of ERPs from ~200 ms post-stimulus time-windows. For the arousal rating prediction, high feature weights, indexing importance for prediction, were found for occipito-parietal channels; for the time rating prediction, high feature weights were most prominent for prefrontal channels.

Discussion
Using a novel, sensitive SVR approach, our study shows that fast, automatic brain activity reflects both arousal and subjective chronology judgements. Arousal might be tied to emotions and physical arousal, which, in turn, might explain differences in brain activity. In contrast, rating subjective chronology requires more abstract semantic judgements; however, ratings could still be decoded from ERP-patterns. Thus, we reason that an integration of several automatic evaluations of different semantic aspects of stimuli might be the basis for fast preference formation, thereby constructing a meaningful conscious experience.