Wise S P, Murray E A, Gerfen C R
Laboratory of Neurophysiology, National Institute of Mental Health, NIH Animal Center, Poolesville, MD 20837, USA.
Crit Rev Neurobiol. 1996;10(3-4):317-56. doi: 10.1615/critrevneurobiol.v10.i3-4.30.
The primate basal ganglia receives information from most of the cerebrum, including the frontal cortex, but projects (via the dorsal thalamus) primarily to the frontal lobe, perhaps in its entirety. As such, the frontal cortex and basal ganglia constitute an integrated, distributed neuronal architecture. We review evidence that the frontal lobe and basal ganglia specialize in different, but related, aspects of response learning. Frontal cortex acts when new rules need to be learned and older ones rejected, whereas the basal ganglia potentiate previously learned rules based on environmental context and reinforcement history. Such potentiation increases the probability that the central nervous system will select a particular rule to guide behavior. We outline a possible mechanism for the basal ganglia's proposed role in rule potentiation, one that involves both the direct and indirect striatal output pathways and their dopaminergic input. It has previously been proposed that direct-pathway neurons recognize a pattern of corticostriatal inputs, which promotes activity in recurrent, positive-feedback modules (or loops) of which they are an integral part. We propose that this recurrent activity potentiates a rule associated with those modules. If so, then the dopaminergic system is well situated and organized to modulate rule potentiation in both the short and long term. Dopaminergic neurons of the midbrain increase activity during learning and other periods of relatively unpredictable reinforcement. Dopamine enhances gene expression and other forms of activity in striatal neurons of the direct pathway, while suppressing neurons of the indirect pathway. In the short term, then, dopamine may augment the activity of modules triggered by a recognized context, whereas in the long term it may promote context-dependent activation of the same modules. Together, these modulatory influences could support both rule potentiation and learning the context for potentiating that rule.
灵长类动物的基底神经节接收来自大脑大部分区域的信息，包括额叶皮质，但主要（通过背侧丘脑）投射到额叶，可能是整个额叶。因此，额叶皮质和基底神经节构成了一个整合的、分布式的神经元结构。我们回顾了证据表明额叶和基底神经节在反应学习的不同但相关方面具有专门化功能。当需要学习新规则并摒弃旧规则时，额叶皮质发挥作用，而基底神经节则根据环境背景和强化历史增强先前学习的规则。这种增强增加了中枢神经系统选择特定规则来指导行为的可能性。我们概述了一种可能的机制，用于解释基底神经节在规则增强中所提出的作用，该机制涉及直接和间接纹状体输出通路及其多巴胺能输入。此前曾有人提出，直接通路神经元识别皮质纹状体输入的模式，这促进了它们作为不可或缺部分的循环正反馈模块（或回路）中的活动。我们提出这种循环活动增强了与这些模块相关的规则。如果是这样，那么多巴胺能系统在短期和长期内都处于良好的位置并具备调节规则增强的组织结构。中脑的多巴胺能神经元在学习和其他相对不可预测强化的时期会增加活动。多巴胺增强直接通路纹状体神经元中的基因表达和其他形式的活动，同时抑制间接通路的神经元。那么，在短期内，多巴胺可能增强由识别出的背景触发的模块的活动，而在长期内它可能促进相同模块的背景依赖性激活。这些调节影响共同作用，可以支持规则增强以及学习增强该规则的背景。
