Available online at www.sciencedirect.com The dynamics of dopamine in control of motor behavior Mati Joshua 1,2 , Avital Adler 1,2 and Hagai Bergman 1,2,3 The basal ganglia are known to control behavior using reward information; however, recent experiments have revealed that the basal ganglia contribute to the processing of salient non- rewarding events as well. Here, we suggest that the temporal dynamics of the response of dopaminergic neurons (DANs) enable the basal ganglia to have a dual role. The fast DAN response to salient events is mediated thorough the brainstem- basal ganglia loop. Forebrain loops enable the second phase of the dopaminergic responses that require highly processed information. The convergent encoding of fast/salient and slow/ detailed information suggests that the basal ganglia control the tradeoff between fast and immediate responses to environmental events and slow responses that are only executed after substantial environmental information has been gathered. Addresses 1 Department of Medical Neurobiology, The Hebrew University- Hadassah Medical School, Jerusalem 91120, Israel 2 The Interdisciplinary Center for Neural Computation, The Hebrew University, Jerusalem 91904, Israel 3 Eric Roland Center for Neurodegenerative Diseases, The Hebrew University, Jerusalem 91904, Israel Corresponding author: Bergman, Hagai (hagaibe@ekmd.huji.ac.il) Current Opinion in Neurobiology 2009, 19:615–620 This review comes from a themed issue on Motor systems Edited by Abdel el Manira and Krishna Shenoy Available online 10th November 2009 0959-4388/$ – see front matter # 2009 Elsevier Ltd. All rights reserved. DOI 10.1016/j.conb.2009.10.001 Introduction The neural network of the basal ganglia is embedded in the cortical and subcortical loops [1]. The cortex is con- nected to the main input structures of the basal ganglia— the striatum and the subthalamic nucleus. Recent studies have shown that subcortical structures directly activate the input structures of the basal ganglia [2  ,3]. The striatum and the subthalamic nucleus directly and indirectly project to the output structures of the basal ganglia that project back to the frontal cortex through the thalamus [4]. However, the output of the basal ganglia is not restricted to the thalamo-cortical pathway but is also directed to subcortical motor nuclei [5,6]. Basal ganglia activity is modulated by the dense dopaminergic inner- vation of the striatum. As for the striatal neurons, the dopaminergic neurons (DANs) receive convergent inputs from cortical and subcortical structures. Input to dopa- minergic neurons not only comes partly from the brain- stem [7,8] but also from more dorsal sources such as the frontal cortex [9], the striatum [10], globus pallidus and the substantia nigra pars reticulata (SNpr) [11,12] and the pathways from the lateral habenula [13  ,14]. Reinforcement learning models have emphasized the involvement of the basal ganglia in controlling behavior. These models are based mainly on the response of the DANs to rewarding events [15]. However, these models do not take into account the different inputs to DANs that may lead to multi-phasic temporal dynamics in their responses to behavioral events. Here we hypothesize that the temporal dynamics of DAN responses reflect two different modes of the basal ganglia network: a fast phase that enables fast behavioral responses and a slow phase that enables motor control via richer processing of information. We suggest that these phases are driven by the brain stem and the forebrain networks, respectively. Dopamine can encode both saliency and reinforcement learning error Studies of midbrain dopaminergic neuron activity have shown that these neurons respond to mismatch in reward prediction [16–20]. These results prompted reinforce- ment learning models of the basal ganglia in which the dopaminergic signal encodes the temporal difference error. These models predict that DANs decrease their discharge rate for aversive events or for cues predicting future aversive events. In fact, several studies have shown that some DANs decrease their rate during noxious stimuli [21–24]. Similarly, recordings in monkeys avoid- ing a punishment have shown that DANs do not increase their discharge rate for aversive events as is the case for appetitive events [25]. The increase in discharge rate for aversive events was attributed to generalization of the reward predicting cue. Other studies have found that some of the DANs increase their discharge rate for aversive events [26,27  ] in line with early neurochemical studies that reported an increase of striatal dopamine during aversive conditions [26]. Recently we [28  ] and others [29,30] have shown that during a classical con- ditioning task DANs increase their firing rate to an aversive airpuff and to cues that predict it. These results support the notion that DANs also encode the saliency of external events [26,31,32]. Two recent studies have reported a correlation between the location of DANs and their response to aversive events [27  ,29]. Thus, the disparity in previous results may be due to the spatial separation of DANs into distinct www.sciencedirect.com Current Opinion in Neurobiology 2009, 19:615–620