7.1: Introduction


The foundations of cognition are built upon the sensory-motor loop -- processing sensory inputs to determine which motor action to perform next. This is the most basic function of any nervous system. The human brain has a huge number of such loops, spanning the evolutionary timescale from the most primitive reflexes in the peripheral nervous system, up to the most abstract and inscrutable plans, such as the decision to apply to, and attend, graduate school, which probably involves the highest levels of processing in the prefrontal cortex (PFC) (or perhaps some basic level of insanity... who knows).

Learning Rules Across the Brain
Learning Signal Dynamics
Area Reward Error Self Org Separator Integrator Attractor
Basal Ganglia +++ --- --- ++ - ---
Cerebellum --- +++ --- +++ --- ---
Hippocampus + + +++ +++ --- +++
Neocortex ++ +++ ++ --- +++ +++

Table $$7.1$$: Comparison of learning mechanisms and activity/representational dynamics across four primary areas of the brain. +++ means that the area definitely has given property, with fewer +'s indicating less confidence in and/or importance of this feature. --- means that the area definitely does not have the given property, again with fewer -'s indicating lower confidence or importance.

In this chapter, we complete the loop that started in the previous chapter on Perception and Attention, by covering a few of the most important motor output and control systems, and the learning mechanisms that govern their behavior. At the subcortical level, the cerebellum and basal ganglia are the two major motor control areas, each of which has specially adapted learning mechanisms that differ from the general-purpose cortical learning mechanisms described in the Learning Mechanisms chapter (see Comparing and Contrasting Major Brain Areas for a high-level summary of these differences -- the key summary table is reproduced here (Table $$7.1$$)). The basal ganglia are specialized for learning from reward/punishment signals, in comparison to expectations for reward/punishment, and this learning then shapes the action selection that the organism will make under different circumstances (selecting the most rewarding actions and avoiding punishing ones; Figure 7.1). This form of learning is called reinforcement learning. The cerebellum is specialized for learning from error, specifically errors between the sensory outcomes associated with motor actions, relative to expectations for these sensory outcomes associated with those motor actions. Thus, the cerebellum can refine the implementation of a given motor plan, to make it more accurate, efficient, and well-coordinated.

There is a nice division of labor here, where the basal ganglia help to select one out of many possible actions to perform, and the cerebellum then makes sure that the selected action is performed well. Consistent with this rather clean division of labor, there are no direct connections between the basal ganglia and cerebellum -- instead, each operates in interaction with various areas in the cortex, where the action plans are formulated and coordinated. Both basal ganglia and cerebellum are densely interconnected with the frontal cortex, including motor control areas in posterior frontal cortex, and the prefrontal cortex anterior to those. Also, as discussed in the prior chapter, the parietal cortex is important for mapping sensory information to motor outputs (i.e., the "how" pathway), by way of computing things like spatial maps, and relative spatial relationships among objects in the environment. Thus, parietal representations drive motor action execution as coordinated by the cerebellum, and cerebellum is also densely interconnected with parietal cortex. In contrast, the basal ganglia are driven to a much greater extent by the ventral pathway "what" information, which indicates the kinds of rewarding objects that might be present in the environment (e.g., a particular type of food). They do also receive some input from parietal, but just not to the great extent that the cerebellum does.

Both the cerebellum and basal ganglia have a complex disinhibitory output dynamic, which produces a gating-like effect on the brain areas they control. For example, the basal ganglia can disinhibit neurons in specific nuclei of the thalamus, which have bidirectional excitatory circuits through frontal and prefrontal cortical areas. The net effect of this disinhibition is to enable an action to proceed, without needing to specify any of the details for how to perform that action. This is what is meant by a gate -- something that broadly modulates the flow of other forms of activation. The cerebellum similarly disinhibits parietal and frontal neurons to effect its form of precise control over the shape of motor actions. It also projects directly to motor outputs in the brain stem, something that is not true of most basal ganglia areas.

We begin the chapter with the basal ganglia system, including the reinforcement learning mechanisms (which involve other brain areas as well). Then we introduce the cerebellar system, and its unique form of error-driven learning. Each section starts with a review of the relevant neurobiology of each system.

This page titled 7.1: Introduction is shared under a CC BY-SA license and was authored, remixed, and/or curated by O'Reilly, Munakata, Hazy & Frank.