Behavioral Social Neuroscience Seminar

Multivariate pattern recognition has recently emerged as a powerful tool in human cognitive neuroscience. Its power lies in its ability to go beyond traditional brain activation studies to investigate the specific task-related "information" encoded in cortical representations. Here we will present work from our lab that has applied this approach to visual perception and human decision making. In one series of experiments we investigated the information flow through the visual system during varying levels of conscious access. Using pattern recognition techniques we were able to reveal that the brain appears to use different networks for perceptual decision-making depending on the visibility of sensory information: Whereas choices for highly visible objects were best predicted from high-level sensory regions, choices under low visibility were best predicted from supramodal regions of parietal cortex. In another series of experiments we investigated "free" decisions and were able to predict subjects' choices for different actions (button presses) several seconds before the choices entered awareness. This long-term predictive activity was confirmed by further studies using other choices (adding / subtracting). Furthermore, we found that free choices have highly similar cortical representations as choices made during perceptual decision when subjects believe to be "purely guessing". In an extension of this research we also investigated reward encoding and choices where subjects have a clear preference for one or the other option. For this we studied the evaluation of consumer products (cars) and politicians. We were able to predict with high accuracy which car someone would buy, even when the car was presented outside the focus of attention. Taken together these results reveal the power of multivariate pattern recognition for investigating human choice behavior.