Neural Networks 19 (2006) 1120–1136 www.elsevier.com/locate/neunet 2006 Special Issue Hold your horses: A dynamic computational role for the subthalamic nucleus in decision making Michael J. Frank ∗,1 Department of Psychology, Program in Neuroscience, University of Arizona, 1503 E University Blvd, Tucson, AZ 85721, United States Received 31 October 2005; accepted 30 March 2006 Abstract The basal ganglia (BG) coordinate decision making processes by facilitating adaptive frontal motor commands while suppressing others. In previous work, neural network simulations accounted for response selection deficits associated with BG dopamine depletion in Parkinson’s disease. Novel predictions from this model have been subsequently confirmed in Parkinson patients and in healthy participants under pharmacological challenge. Nevertheless, one clear limitation of that model is in its omission of the subthalamic nucleus (STN), a key BG structure that participates in both motor and cognitive processes. The present model incorporates the STN and shows that by modulating when a response is executed, the STN reduces premature responding and therefore has substantial effects on which response is ultimately selected, particularly when there are multiple competing responses. Increased cortical response conflict leads to dynamic adjustments in response thresholds via cortico-subthalamic- pallidal pathways. The model accurately captures the dynamics of activity in various BG areas during response selection. Simulated dopamine depletion results in emergent oscillatory activity in BG structures, which has been linked with Parkinson’s tremor. Finally, the model accounts for the beneficial effects of STN lesions on these oscillations, but suggests that this benefit may come at the expense of impaired decision making. c  2006 Elsevier Ltd. All rights reserved. Keywords: Basal ganglia; Decision making; Subthalamic nucleus; Neural network model; Parkinson’s disease; Reinforcement learning 1. Introduction Deciphering the mechanisms by which the brain supports response selection, a central process in decision making, is an important challenge for both the artificial intelligence and cognitive neuroscience communities. Based on a wealth of data, the basal ganglia (BG) are thought to play a principal role in these processes. In the context of motor control, various authors have suggested that the role of the BG is to selectively facilitate the execution of a single adaptive motor command, while suppressing all others (Basso & Wurtz, 2002; Brown, Bullock, & Grossberg, 2004; Frank, 2005a; Gurney, Prescott, & Redgrave, 2001; Hikosaka, 1994; Jiang, Stein, & McHaffie, 2003; Mink, 1996; Redgrave, Prescott, & Gurney, 1999). Interestingly, circuits linking the BG with more cognitive areas ∗ Tel.: +1 520 626 4787; fax: +1 520 621 9306. E-mail address: mfrank@u.arizona.edu. URL: http://www.u.arizona.edu/ ∼ mfrank/. 1 Portions of this paper were previously presented in conference format at the International Workshop on Models of Natural Action Selection (Frank, 2005b). of frontal cortex (e.g., prefrontal) are strikingly similar to those observed in the motor domain (Alexander, DeLong, & Strick, 1986), raising the possibility that the BG participate in cognitive decision making in an analogous fashion to their role in motor control (Beiser & Houk, 1998; Frank, 2005a; Frank & Claus, 2006; Frank, Loughry, & O’Reilly, 2001; Middleton & Strick, 2000, 2002). Studies with Parkinson’s patients, who have severely depleted levels of dopamine (DA) in the BG (Kish, Shannak, & Hornykiewicz, 1988), have provided insights into the functional roles of the BG/DA system in both motor and higher level cognitive processes (Cools, 2005; Frank, 2005a; Shohamy, Myers, Grossman, Sage, & Gluck, 2005). Of particular recent interest is the finding that deep brain stimulation in the subthalamic nucleus (STN) dramatically improves Parkinson motor symptoms, with both reported enhancements and impairments in cognition (Karachi et al., 2004; Witt et al., 2004). Because the BG consists of a complex network of dynamically interacting brain areas, a mechanistic understanding of exactly how the STN participates in response selection and decision making is difficult to develop with traditional box and arrow models. Computational models 0893-6080/$ - see front matter c  2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.neunet.2006.03.006