Neural mechanisms of reward bias in flexible decision-making
Zusammenfassung der Projektergebnisse
Humans have the amazing ability to change the way they respond in seemingly identical situations based on sources of information that are not immediately represented in our sensory environment e.g., the utility of anticipated outcomes (adaptive decision-making). This flexibility enables us to make near-optimal choices while pursuing constantly changing goals in an ever-changing environment. Despite the importance of adaptive decision-making for survival, we are only beginning to study the underlying neural mechanisms. This work funded by the DFG was aimed at studying the neural mechanisms of adaptive decision-making in macaque monkeys that currently provide the best animal model system of human decision-making. The main body of the work was aimed at understanding the role of the frontal eye-field for adaptive decision-making (Aim 1). A parallel line of research provided a detailed description of behavioral variables (choice- and reaction time bias) as well as the development of computational models to account for them (Aim 2). In addition to the detailed study of reward-based biases, a third line of research was established to compare reward-based biases with other types of bias, in this case a Bayesian bias (Aim 3). The work conducted provided a number of novel and exciting insights into the neural mechanisms of adaptive decision-making. First, the work showed that macaque monkeys are capable of adaptive decision-making without receiving any particular training and that adaptive decision-making persists even in highly over-trained animals thus making them an ideal model system to study the neural mechanisms. Second, while animals readily adapted their choices to manipulations of reward, their choices were not affected by manipulations of base rates (Bayesian bias). This suggests that these two types of bias that are identical from a computational point of view, are likely to involve different neural mechanisms. Third, contrary to the hypothesis, the frontal eye field did not seem to play an important role in mediating the observed reward bias. In principle, neurons in the frontal eye field are ideally suited to bias choices towards a more highly rewarded response alternative because they conveniently code visual space and motor actions in a common coordinate frame. However, only a small fraction of the recorded neurons exhibited responses that could bias the animal towards the more highly rewarded response target. Fourth, neurons in the frontal eye field play a different, but no less important role in adaptive decisionmaking: they code information that is related to analysis of the past behavior which may be used to optimize future choices. In particular, the recordings show the first single-cell evidence of an internally generated error signal. Interestingly, the error signal was present in the same neurons that code task difficulty. This finding distinguishes between the two leading accounts of error signals that have been found in human EEG and fMRI studies. Taken together, these findings open up a number of intriguing questions and research possibilities. First, if a motor bias in frontal eye-field does not bias choices towards the more highly rewarded target, what are the neural mechanisms that cause the observed choice bias and where are they implemented? Second, the finding of an internally generated error signal in single neurons of the frontal eye-field provides a unique opportunity to study the neural mechanisms that mediate the online-monitoring of actions to correct potential mistakes and predict outcomes.
Projektbezogene Publikationen (Auswahl)
- (2010). Effects of heartbeat and respiration on macaque fMRI: implications for functional connectivity. Neuropsychologia, 48(7)
Teichert, T., Grinband, J., Hirsch, J., & Ferrera, V. P.
- (2010). Suboptimal integration of reward magnitude and prior reward likelihood in categorical decisions by monkeys. Front Neurosci, 4, 186
Teichert, T., & Ferrera, V. P.
- (2011). Noninvasive, transient and selective blood-brain barrier opening in non-human primates in vivo. PloS one, 6(7)
Marquet, F., Tung, Y.-S., Teichert, T., Ferrera, V. P., & Konofagou, E. E.
- (2011). The dorsal medial frontal cortex is sensitive to time on task, not response conflict or error likelihood. Neuroimage, 57(2)
Grinband, J., Savitskaya, J., Wager, T. D., Teichert, T., Ferrera, V. P., & Hirsch, J.
