Abstract
It was previously shown by fMRI studies that unfair offers during an ultimatum bargaining game activate regions in the brain associated with emotions and conflict, leading to decisions inconsistent with standard economic theory. The temporal dynamics of emotional processing and mental attributes were not clear due to the coarse temporal resolution in those studies (∼2 s). Here, the ultimatum game was studied by EEG recorded from the responders in 19 channels. EEG time series were first in-put to independent component analysis. An equivalent current dipole model was used to localize the sources of the independent components in EEGLAB. The Talairach coordinates of the dipoles were matched with references in the Brede neuroinformatics database. Dipole magnitudes, anatomical regions, and mental attributes were used to explain the rejection of the offers by applying multiple regression as a function of time in epochs with a median resolution of 250 ms. The results are consistent with previous studies regarding responder behavior and activated regions (e.g., anterior cingulate cortex, frontal gyrus, insular cortex). There are three main findings observed with the higher temporal resolution: (1) regression results from fine-scale temporal data showed activations not captured when the analysis was done by using time-averaged data; (2) temporal analysis detected the individual significant epochs and fluctuations (positive and negative correlations) in regions and for the associated mental attributes (e.g., reward/harm perception, anger, unfairness); (3) there was a sequential activation of anterior cingulate cortex and insular cortex, respectively, leading to the rejection response. Overall, regression models could explain a large percentage (∼80%) of out-of-sample behavioral responses. The results are promising for the prospect of using EEG and source localization techniques in neuroeconomics to study finer temporal dynamics of neural activation.
Acknowledgments
This study was supported by University of Chicago Department of Economics Research Grant to Dr. List and Boğaziçi University Research Fund 07HX103 to Dr. Güçlü. We thank Andrew Snyder and Benjamin Duval for organizing the dipole data.
Notes
1Logit regression of temporal dynamics yielded several cases of perfect prediction. Since the activity patterns were similar for both logit and linear regression, we report only the results from the linear regression applied to each epoch.
2We should note that although we employed the Benjamini–Hochberg multiple comparison correction, it is possible that we may have detected a number of unplanned contrasts which generated significance in the mental attributes (e.g., obsessive-compulsive). As in “whole-brain” vs. “region-of-interest” based analyses in the fMRI literature, choosing a small number of variables based on a priori backing from previous literature may be a more reliable way to investigate activity in particular regions.