34 Choosing What to Do
The Basal Ganglia: Selecting, Suppressing, and Scaling Action
34.1 From prediction to selection
The preceding chapter (Chapter 33) asked how a controller anticipates consequences and keeps action from falling behind delayed sensory feedback. This chapter asks how one action gains influence when several are possible. Prediction and selection interact throughout behavior: the expected consequence can favor a candidate action, and selecting that action changes what the nervous system must predict.
A visible cup can be reached for, ignored, moved aside, or passed to another person. The eyes can remain on the cup or turn toward a sound. Walking and reaching can proceed together, but the same hand cannot simultaneously grasp the cup and remain in a pocket. An ongoing action can also lose its claim. A reach that was appropriate a moment ago must be interrupted when someone else takes the object.
The nervous system therefore needs a way to regulate access to the body. Spinal and brainstem circuits organize posture, orienting, and recurring patterns of movement. Parietal and frontal networks specify targets, grasps, and learned actions. A body with limited effectors, energy, and time must recruit compatible actions and restrain incompatible ones.
The basal ganglia make a major contribution to this control problem. Their output continuously inhibits thalamic, collicular, and brainstem targets. Changing that inhibition changes the influence of the controllers downstream.
Basal-ganglia loops help selected actions gain access to control, restrain competing actions, and regulate the vigor with which an action is pursued.
The familiar go/no-go model provides the starting mechanism. One route reduces inhibitory output and facilitates action; another strengthens inhibition and restrains action. Understanding those opposing effects will explain why movement can be released by inhibiting an inhibitory neuron. It will also provide a basis for understanding how the system delays commitment when goals conflict and regulates graded access: an action can be difficult to initiate, weakly expressed, or vigorously pursued.
Parkinson’s and Huntington’s diseases make the stakes visible. In Parkinson’s disease, wanted movement can become slow, small, and difficult to initiate. In Huntington’s disease, involuntary movements intrude while voluntary control also deteriorates. The clinical question is how a circuit helps available movements become coherent action.
34.2 An ancient solution to competition among actions
A swimming vertebrate can turn toward prey, evade a threat, accelerate, or hold position. Its spinal and brainstem networks can generate the required patterns, but those patterns compete for the same body. The basal-ganglia architecture that helps resolve this competition is deeply conserved.
Sten Grillner and colleagues made that conservation experimentally accessible in lamprey. Using tract tracing, molecular markers, and electrophysiology, they identified striatal, pallidal, and subthalamic homologues connected through direct and indirect routes with the same principal excitatory and inhibitory signs found in mammals. Subsequent work identified a second output region homologous to the substantia nigra pars reticulata [@stephensonjones2011basal; @stephensonjones2012dual]. Lampreys are living members of an early-diverging lineage. The shared organization supports an inference about the vertebrate common ancestor; it does not make a living lamprey a preserved ancestor of mammals.
The output physiology is especially revealing. Pallidal and nigral neurons maintain tonic activity and release GABA onto tectal and other motor-related targets. Those downstream systems already possess the circuitry needed for orienting and locomotion. Basal-ganglia activity changes the inhibitory influence over them [@stephensonjones2011basal; @grillnerrobertson2016].
This arrangement supplies the mechanism of disinhibition. If neuron A inhibits neuron B, and neuron B inhibits a motor controller, activating A reduces the inhibition reaching that controller. A decrease in inhibition can therefore increase movement. The basal ganglia gain considerable control through this double inhibitory connection: they can release a downstream system without having to generate its detailed motor pattern.
In lamprey, the distinction can be followed through the body. Brainstem reticulospinal neurons recruit spinal networks whose alternating activity produces swimming. Basal-ganglia output influences whether and how associated motor systems gain access to that machinery. The forebrain does not need to specify each left–right contraction to influence the animal’s course.
Mammals retained these outputs to the superior colliculus and brainstem while greatly expanding recurrent cortex–basal ganglia–thalamus–cortex loops. An enlarged cerebral cortex supplied more differentiated sensory information, learned rules, possible actions, and expected outcomes. Basal-ganglia output could influence behavior through those cortical loops as well as through older subcortical controllers [@grillnerrobertson2016; @grillneretal2020motion].
This is the evolutionary argument developed in Chapter 30 and Chapter 31: new capacities were built through expanded access to existing controllers. The basal ganglia provide a particularly clear example because their conserved inhibitory output can regulate both an orienting response organized in the midbrain and a learned action prepared in cortex.
34.3 The anatomy of a recurrent gate
The basal ganglia are a set of interconnected nuclei rather than one compact organ. Their boundaries are anatomically distinct, but their operation depends on recurrent loops among striatum, pallidum, subthalamic nucleus, substantia nigra, thalamus, cortex, superior colliculus, and brainstem. The principal structures divide the work of receiving input, transforming it, and distributing inhibitory output.
34.3.1 The striatum receives information about state and possible action
The striatum is the largest input structure of the basal ganglia. In the dorsal striatum, the internal capsule separates the caudate nucleus from the putamen, but the two are continuous around the fiber bundles that pass between them. The ventral striatum includes the nucleus accumbens and adjoining territories. These subdivisions have different patterns of input and connectivity. Sensorimotor input is especially prominent in posterior and lateral putamen; associative input is especially prominent in anterior caudate and putamen; limbic and motivational input is especially prominent ventromedially. Between these territories lie gradients and convergence zones through which current goals, learned context, bodily state, and possible movements can interact [@alexanderetal1986parallel; @mcfarlandhaber2000convergent].
The principal output cells of the striatum are spiny projection neurons, or SPNs, traditionally called medium spiny neurons. They are GABAergic. Their axons therefore inhibit the pallidal or nigral neurons they contact. SPNs receive excitatory input from wide regions of cortex and from several thalamic nuclei. They also receive dopamine from the midbrain, inhibition from local neurons and other SPNs, and modulation from cholinergic interneurons. Each SPN integrates cortical and thalamic input with the recent activity and modulatory state of the striatal network.
Two broad SPN populations anchor the classical pathway diagram. Direct-pathway SPNs project strongly to the output nuclei and are enriched in dopamine D1 receptors. Indirect-pathway SPNs project strongly to the external globus pallidus and are enriched in dopamine D2 receptors [@gerfenetal1990d1d2]. These two populations provide the entry points to the go/no-go model: their inhibitory axons reach the output nuclei through routes with different net effects.
34.3.2 GPi and SNr provide tonic inhibitory output
The internal segment of the globus pallidus, or GPi, and the substantia nigra pars reticulata, or SNr, are the principal output nuclei. Their neurons are GABAergic and often fire at substantial rates even in the absence of an overt movement. This tonic activity continuously inhibits downstream targets.
GPi projects prominently to motor and intralaminar regions of the thalamus. Those thalamic nuclei influence motor, premotor, supplementary motor, prefrontal, and other cortical regions. The pathway is recurrent: cortex influences basal-ganglia input, basal-ganglia output changes thalamic activity, and thalamus returns that influence to cortex.
SNr projects strongly to the superior colliculus and to several thalamic and brainstem targets. Its collicular projection provides a direct route for regulating gaze shifts. The physiological evidence for this route will give the pathway diagram a concrete behavioral meaning.
GPi and SNr distribute inhibition to different combinations of thalamic, collicular, and brainstem targets. The consequence of an output change depends on the population affected: a pause in neurons projecting to an eye-movement system has a different immediate effect from a pause in neurons influencing a cortical hand-control loop.
34.3.3 GPe and STN form a recurrent internal network
The external segment of the globus pallidus, or GPe, is the first pallidal target of the indirect pathway. Its neurons project to STN, GPi, SNr, striatum, and other structures, while receiving input from striatum, STN, and additional sources. These connections give GPe a central role in shaping activity within the basal ganglia.
Two well-characterized GPe populations, established chiefly through rodent experiments, illustrate this organization. Prototypic neurons project strongly to STN and other basal-ganglia nuclei. Arkypallidal neurons project back to striatum and can suppress striatal activity. The populations differ in molecular markers, intrinsic firing, and behavioral timing [@malletetal2012dichotomous; @abdietal2015prototypic]. The return projection to striatum will become important when we examine stopping.
The subthalamic nucleus, or STN, is the major glutamatergic nucleus within the basal ganglia. Its projections excite GPi, SNr, and GPe. STN receives inhibition from GPe and excitatory input from several cortical regions. The reciprocal GPe–STN connections create a network capable of rapidly changing the timing and distribution of basal-ganglia output. The direct cortical projection to STN forms the hyperdirect pathway described below.
34.3.4 Dopamine changes excitability and learning
Dopamine reaches the dorsal striatum principally from the substantia nigra pars compacta, or SNc. Dopamine changes the responsiveness of striatal neurons and the strength of their active synapses. Through D1- and D2-family receptors, it alters membrane excitability, transmitter release, and corticostriatal plasticity. These effects help determine which activity patterns are amplified, which are weakened, and how past outcomes change future action.
The neighboring ventral tegmental area supplies much of the dopamine reaching ventral striatum and limbic-associated forebrain. SNc and VTA projections form overlapping gradients across striatal and other forebrain territories. This chapter concentrates on their contribution to access, initiation, and vigor. The next chapter will develop how dopamine-dependent learning changes what actions become worth selecting.
Cortical–basal-ganglia circuits are commonly divided into motor, oculomotor, associative, and limbic loops. Different cortical regions preferentially reach different striatal territories, and basal-ganglia output returns through correspondingly organized thalamic routes [@alexanderetal1986parallel].
This topography permits selective control. A change in activity within an oculomotor loop can influence gaze without requiring the same change throughout every motor system. Convergence permits integration: cortical and thalamic inputs overlap within parts of striatum, and recurrent connections link information about action, context, and expected outcome [@mcfarlandhaber2000convergent].
The combination explains how a learned rule can change a reach or an expected reward can change a saccade. Anatomical specialization supplies selective access; communication among the loops allows the action to reflect more than one source of information.
34.4 Three routes through the human basal ganglia
The direct, indirect, and hyperdirect pathways explain how input changes the tonic inhibition supplied by GPi and SNr. In the traces below, + marks excitation and − marks inhibition. The signs follow the axons leaving each named structure. The starting condition is crucial: the output nuclei are already firing and inhibiting their targets.
34.4.1 The direct pathway facilitates action: go
The direct route is:
Cortex or thalamus (+) → direct-pathway SPN (−) → GPi/SNr (−) → target
Excitatory input recruits direct-pathway SPNs. Those neurons inhibit the output nuclei. Output-nucleus firing falls, so less inhibition reaches the thalamic, collicular, or brainstem target. The target is disinhibited.
This is the basis of the go pathway. Its action opens access by removing restraint. A collicular population receiving less nigral inhibition can more readily recruit a gaze shift; a thalamic population receiving less pallidal inhibition can exert greater influence within a cortical motor loop. Excitatory drive and the target’s current state determine how it responds to that release.
The double inhibition is worth following explicitly. The striatal neuron does not excite the downstream controller. It inhibits the neuron that was inhibiting the controller. An inhibitory first step can therefore facilitate the eventual action.
34.4.2 The indirect pathway restrains action: no-go
The classical indirect route is:
Cortex or thalamus (+) → indirect-pathway SPN (−) → GPe (−) → STN (+) → GPi/SNr (−) → target
Indirect-pathway SPNs inhibit GPe. Reduced GPe activity removes inhibition from STN. STN then provides more excitation to GPi and SNr, increasing their inhibitory output. The downstream target encounters greater restraint.
This is the basis of the no-go pathway. Its net effect opposes the direct pathway: it makes an action harder to release or sustain. In the classical model, facilitating a selected response and suppressing alternatives are complementary requirements for an orderly movement [@albinyoungpenney1989; @delong1990primate].
The direct–indirect distinction thus explains a fundamental feature of motor control. Successful action requires both activation and inhibition. A response that can be generated but cannot be restrained is as much a control problem as a response that cannot be released.
34.4.3 The hyperdirect pathway provides rapid restraint
The hyperdirect route is:
Cortex (+) → STN (+) → GPi/SNr (−) → target
Cortical excitation reaches STN without passing through striatum. STN excites the output nuclei, and those nuclei increase inhibition of their targets. This route provides fast access to the inhibitory output side of the basal ganglia [@nambuetal2002hyperdirect].
Its speed is useful when an impending action must be interrupted or delayed. In the classical temporal account, an early hyperdirect influence can impose restraint before the striatal pathways resolve which responses should remain suppressed and which should proceed. This gives the system a way to delay commitment when alternatives compete strongly, as well as to interrupt a response when circumstances change. The conflict and stopping experiments below test these uses of rapid restraint.
| Route | Principal change in the canonical chain | Consequence for the downstream controller |
|---|---|---|
| Direct: go | Inhibits GPi/SNr and reduces their inhibitory output | Facilitates access |
| Indirect: no-go | Recruits increased GPi/SNr output through pallidal–subthalamic circuitry | Strengthens restraint |
| Hyperdirect: rapid restraint | Excites GPi/SNr through cortical input to STN | Rapidly delays or suppresses access |
34.4.4 Dopamine supports the expression of action
Dopamine acts differently on the two principal striatal populations. Direct-pathway SPNs are enriched in D1 receptors; indirect-pathway SPNs are enriched in D2 receptors [@gerfenetal1990d1d2]. In the classical account, D1 signaling increases the responsiveness of active direct-pathway neurons, while D2 signaling reduces the responsiveness of indirect-pathway neurons. Both effects favor the expression of movement: the go route gains influence and the no-go route is moderated.
The same arrangement explains why dopamine loss is so disruptive. Reduced direct-pathway support combines with increased indirect-pathway influence, shifting the circuit toward greater inhibitory output. An action can be adequately prepared elsewhere yet remain difficult to initiate or too weakly expressed. This is the first circuit explanation of parkinsonian bradykinesia.
Dopamine also changes synaptic plasticity, linking the current expression of an action to what experience has taught the system. For the present chapter, its crucial role is in setting the responsiveness of the action system. The next chapter develops how dopamine-dependent learning changes that system over time.
The model now makes clear predictions: activating the direct pathway should facilitate movement, activating the indirect pathway should restrain it, and dopamine loss should impair the release and scaling of movement. Those predictions can be tested.
34.5 Testing the go/no-go model
34.5.1 Releasing a gaze shift from inhibition
The nigral projection to the superior colliculus provides one of the clearest demonstrations of disinhibitory control. Hikosaka and Wurtz recorded SNr neurons in macaques making saccades. Many neurons fired tonically between movements and paused in association with particular saccades. A pause in an inhibitory projection could release collicular neurons that help generate the gaze shift [@hikosakawurtz1983snr].
They also manipulated GABA signaling within the superior colliculus. Muscimol, which activates GABA receptors, delayed saccades toward the affected movement field and reduced their amplitude. Bicuculline, which blocks GABA receptors, produced irrepressible saccades toward that field [@hikosakawurtz1983colliculus].
The recording and intervention fit together. The recording identified a timed withdrawal of inhibitory output; the pharmacology showed that changing inhibition at the target changed the release of movement. This is a concrete example of access control: the colliculus supplies machinery for orienting, while nigral inhibition regulates its opportunity to act.
34.5.2 Activating the two pathways produces opposing effects
Kravitz and colleagues tested the direct–indirect distinction by expressing the light-sensitive channel channelrhodopsin in either D1- or D2-defined striatal neurons in mice. Light could then activate one pathway population while the investigators measured movement. Activating the indirect pathway increased freezing and reduced locomotor initiation. Activating the direct pathway increased locomotion and reduced freezing [@kravitzetal2010parkinsonian].
The investigators then depleted dopamine in a mouse model of Parkinson’s disease. Activating the direct pathway rescued the measured deficits in freezing, movement speed, and locomotor initiation. The surviving circuitry could produce movement when direct-pathway activity supplied sufficient influence.
These results give the go/no-go model causal force. They show that the two pathways can exert opposing effects and that direct-pathway recruitment can overcome major motor consequences of dopamine depletion. The next question is how the pathways are recruited naturally when the animal acts.
34.5.3 Both pathways participate when an action begins
Cui and colleagues measured calcium signals from genetically identified direct- or indirect-pathway SPN populations in freely moving mice. An implanted optical fiber sampled activity while the animals moved between levers and a food magazine. Signals from both pathway populations rose around action initiation and were stronger for movements toward the side opposite the recorded striatum [@cui2013concurrent].
This result challenged a particular version of the model: direct-pathway neurons active during movement, indirect-pathway neurons active during immobility. The measurements were population signals, so they did not identify which individual cells controlled the chosen action or its alternatives. They did establish that participation in movement is shared by both pathways.
Tecuapetla and colleagues then perturbed the pathways at different stages of learned action sequences. Suppressing either pathway before a sequence delayed its initiation; suppressing either pathway after initiation impaired ongoing performance. The indirect pathway was needed for smooth action as well as for restraint [@tecuapetlaetal2016complementary].
The findings answer different questions. Kravitz established the opposing effects available through the two routes. Cui established their concurrent recruitment during natural behavior. Tecuapetla established that disrupting either route can impair the organized action. Together they require a model in which opposing circuit influences cooperate over time.
34.5.4 From a global switch to selective control
A useful way to understand that cooperation is to consider reaching while standing. The selected reach must gain influence, another reach must remain unexpressed, and postural support must continue. These requirements provide a functional interpretation of coactive pathways: facilitation and restraint can act on different populations and at different moments within the same behavior.
This is the action-channel account. Related cortical inputs recruit striatal and output populations that influence particular downstream controllers. Reducing inhibition in one set of outputs can favor a response while maintained or increased inhibition elsewhere limits interference. Natural actions recruit combinations of these populations, just as reaching earlier in this unit combined transport, hand shaping, gaze, and posture.
Focused facilitation accompanied by broader restraint is often described as center–surround selection. The useful idea is that increasing the influence of one candidate and limiting its competitors work together. Here, center and surround describe a functional relationship among actions; the participating populations can be distributed and overlapping, and the selected action can itself require several cooperating channels.
The specific assignment of a population to a chosen response or a competing component is an experimental question. Coactivation alone does not supply that assignment. The explanatory advance is that the go/no-go distinction can operate within an organized action instead of dividing the entire animal into moving and motionless states.
Circuit anatomy provides further means of coordination. GPe projects directly to the output nuclei as well as to STN; STN excites GPe; and arkypallidal neurons return inhibition to striatum. Direct-pathway axons can also send collaterals into GPe. These connections allow the timing and distribution of restraint to change as behavior unfolds. The canonical signs remain the foundation for tracing those effects.
34.6 Conflicting goals and the threshold for commitment
34.6.1 Several goals can support incompatible actions
A hungry animal approaching food may also detect a possible predator. Hunger favors approach; threat favors withdrawal. Both goals remain biologically important, but their immediate actions compete for the same body. Conflict can also arise between two ways of satisfying one goal, such as approaching either of two food sources. The control problem is to turn these competing claims into a coherent course of behavior.
The affordances introduced in Chapter 32 therefore acquire their influence in relation to current goals. The same object can support different actions as bodily state and context change. A sandwich affords eating, carrying, offering, or throwing; hunger, a social goal, or a threat changes the relative support for those possibilities. Sensory evidence specifies opportunities, learned experience supplies expectations, and current need changes their significance.
A response surface provides a useful way to visualize this competition. In such a representation, the height associated with a candidate action can express its current support. Changing hunger, threat, expected benefit, or effort reshapes the surface. Several candidates can be strongly supported at once, and a small change in state can change which candidate has the strongest claim. The surface represents the outcome of interacting influences across the control system. Basal-ganglia loops contribute by changing how effectively those supported actions gain access to downstream controllers.
Preparation can develop before commitment. An animal can orient toward a faint sound while continuing to forage; a person can begin shaping the hand for a likely grasp while visual evidence about the object accumulates. These provisional preparations reduce the time needed to act once the situation becomes clearer. Selection concerns when a prepared possibility gains enough influence to become the next action.
34.6.2 Restraint can buy time for a decision
Having the strongest current support and having sufficient support to act are distinct conditions. A decision threshold specifies how much accumulated evidence is required before a response is released. Raising that threshold allows more information to be considered; lowering it permits earlier commitment. This is a property of the decision process, expressed through populations and their interactions, rather than the voltage threshold for an individual neuron’s spike.
Conflict gives temporary restraint a purpose. When incompatible responses are strongly activated, cortical input through STN can increase inhibitory output and delay commitment. During that interval, additional evidence can change the balance between candidates. Striatal and recurrent pallidal influences can then favor the selected response while maintaining restraint over alternatives. In the response-surface picture, changing the relative support for the candidates and changing the conditions for release are two ways to alter behavior.
Delay has a cost as well as a benefit. Waiting can prevent a premature reach toward the wrong object, but excessive delay can lose an opportunity to grasp it. An urgent escape may require commitment before uncertainty has been resolved. Effective control adjusts the demand for evidence to the consequences of acting and waiting.
34.6.3 Testing the link between conflict and restraint
Frank and colleagues tested people with Parkinson’s disease with subthalamic deep-brain stimulation switched on or off. With stimulation off, patients slowed their decisions when alternatives created high conflict. With stimulation on, that adjustment was disrupted: patients instead sped up under high conflict [@franketal2007hold]. The intervention changed the use of restraint when it was needed for deliberation.
Cavanagh and colleagues connected this behavior to frontal activity and a formal model of decision making. Participants chose between symbols with learned reward probabilities. Choices between similarly valued options created greater conflict than choices between a good and a poor option. The investigators recorded scalp EEG and fitted a drift-diffusion model, in which noisy evidence accumulates until it reaches a decision boundary.
In healthy controls and patients with stimulation off, greater medial-frontal theta activity was associated with a higher estimated decision threshold particularly under high conflict. Subthalamic stimulation reversed that relationship: greater theta activity accompanied a lower threshold [@cavanaghetal2011threshold]. The threshold was estimated from choices and response times; EEG measured the accompanying physiological activity. Their relationship supports an account in which frontal–subthalamic interactions help delay commitment when competing responses demand further evaluation.
These experiments give the gate model an additional function. Restraint can preserve time for a decision while alternatives remain in preparation. Once an action has gained influence, the same need for flexible control appears in another form: can it be interrupted when new evidence makes it inappropriate?
34.7 Stopping an action already in preparation
A stop-signal task exposes the competition between releasing and withholding a response. The participant prepares an action following a go cue. On some trials, a later signal instructs cancellation. Increasing the delay before that signal makes stopping harder because preparation has advanced farther.
Schmidt and colleagues recorded across basal-ganglia structures in rats performing such a task. STN neurons responded rapidly to stop cues on both successful and failed stop trials. SNr responses distinguished the outcomes: the stop-related increase in output was expressed when stopping succeeded [@schmidtetal2013canceling].
Recordings and modeling identified a race at SNr. Excitatory STN input promoted continued inhibition of downstream targets. Movement-related striatal input inhibited SNr and promoted release. For cancellation to succeed, the stop-related influence had to arrive before the movement-related influence made the action difficult to arrest. The striatal competitor in this experiment was therefore the go-related input, not a second, slower stop command.
The behavioral implication is direct. Detecting the stop cue is insufficient; the resulting neural influence must reach the appropriate output population in time. The fast route through STN gives the system an opportunity to interrupt an impending response.
Human recordings connect this account to frontal control. Chen and colleagues combined inferior-frontal and STN recordings with stimulation in patients undergoing DBS surgery. Short-latency cortical responses to STN stimulation supported a direct connection. During stopping, cortical activity preceded STN activity, and stronger coupling was associated with faster cancellation [@chenetal2020hyperdirect]. Measurements extending from frontal activity through motor cortex to muscle reveal a corresponding cascade toward the motor output [@janaetal2020stopping].
A rapid interruption can buy time for more lasting cancellation. Mallet and colleagues examined identified GPe populations during a rat stop task. Arkypallidal neurons responded to stop cues earlier and more strongly than prototypic neurons. Their response timing matched the suppression of developing go-related activity in striatum [@mallet2016stopsignal].
The anatomy gives that timing a mechanistic interpretation. Arkypallidal neurons send GABAergic projections back to striatum. The authors proposed a two-step process: an STN–SNr influence first pauses release, while pallidostriatal inhibition then suppresses the striatal activity supporting the prepared action. This pause-then-cancel account distinguishes a rapid interruption from withdrawal of the developing motor drive.
Stopping an overt movement also imposes mechanical demands. An action already underway cannot be erased: the controller may shorten it, reduce its force, prevent its next component, or launch a compensating movement. A person carrying a full cup must maintain grip and balance while halting forward progress. The sensory and motor systems must preserve those useful components and change the part of the action that has become inappropriate. Switching adds a further requirement: a replacement response must gain influence as the previous response loses it. The conceptual sequence in Figure 34.4 places the experimentally established stopping mechanisms within that broader control problem.
34.8 From go/no-go to movement vigor
The pathway model initially explains whether an action is facilitated or restrained, including how commitment can be delayed when alternatives compete. Its next extension is to how strongly the action is expressed. A reach that begins too late or moves too slowly can fail even when the correct object has been selected. Time spent deciding and speed once moving are separable: a person can deliberate over which cup to take and then reach for it briskly.
34.8.1 Access is graded
The output nuclei change firing rate, timing, and coordination across populations. These changes produce degrees of influence over their targets. In a graded version of the gate model, reducing inhibition makes a downstream response easier to recruit, while increasing inhibition makes it harder. Changes in that balance can affect onset, amplitude, speed, and persistence.
This supplies a bridge from go/no-go to vigor, the speed and intensity with which an action is pursued. Choosing to approach an object and deciding how rapidly to approach it draw on related information: expected benefit, effort, urgency, and the current state of the body. A small change in benefit may alter the vigor of the same action before it changes which action is selected.
The physiology that realizes this influence extends beyond average inhibitory firing. A brief pause and a sustained reduction in output need not have the same effect. Synchronized bursts can constrain when a target responds. The relative importance of these patterns becomes especially clear in Parkinson’s disease.
34.8.2 Can a person move faster than the speed they usually produce?
Mazzoni, Hristova, and Krakauer tested seven people with Parkinson’s disease and seven controls in an arm-reaching task. Target distance and the required speed range varied across conditions. Participants continued until they had made twenty movements within the prescribed speed range [@mazzonietal2007vigor].
Patients could produce the required fast movements with accuracy comparable to controls. They nevertheless needed more attempts to accumulate the twenty qualifying movements, particularly as movement demands increased. Their slowing could not be explained simply as a strategy for preserving accuracy at higher speeds.
The authors interpreted this pattern as a change in the implicit balance between energetic cost and the benefit of moving quickly. The experiment measured the movements and attempts; the cost–benefit account explains why sufficient motor capacity could coexist with a reduced tendency to use it. The task separates what a patient can achieve from how readily that performance is produced.
This is a consequential distinction for the chapter. An available action can be difficult to mobilize without the force-generating apparatus being lost. Vigor concerns the recruitment of that capacity.
34.8.3 Dopamine helps mobilize the action
Experiments in mice connect this behavioral distinction to nigrostriatal dopamine. Panigrahi and colleagues found that dopamine depletion reduced movement and disrupted striatal representations of vigor; restoring dopamine improved both [@panigrahi2015vigor]. Da Silva and colleagues manipulated dopamine-neuron activity before self-initiated movement and changed the probability and vigor of subsequent movements [@dasilva2018dopamine].
These results establish a contribution to the expression of action on the current occasion. Dopamine-dependent learning, developed in the next chapter, alters how the system responds on later occasions. Both processes can change the vigor with which a familiar movement is pursued.
The physical result still depends on the controllers described earlier in this unit. Motor cortex and descending pathways recruit spinal circuits; motor pools generate muscle force; cerebellar circuits calibrate timing and coordination. Basal-ganglia and dopamine systems help determine how readily and vigorously that machinery is engaged.
Yttri and Dudman trained mice to move a joystick for reward and delivered pathway-specific stimulation contingent on movement speed. Depending on which pathway was stimulated and which portion of the speed distribution triggered stimulation, subsequent movements became faster or slower. Either pathway could support a learned change in either direction [@yttridudman2016].
The changes accumulated across trials, persisted after stimulation ended, and were prevented by dopamine antagonists. This separates an immediate motor effect from a learned adjustment of movement. Opposing pathway influences can help modify the parameters of an action as well as its probability of occurring.
34.8.4 Current need changes the competition
The opportunities described in Chapter 32 acquire different significance with the state of the organism. Thirst increases the benefit of drinking. Fatigue increases the cost of climbing. Threat increases the urgency of orienting or escape. An instruction can require leaving an attractive object untouched.
These influences reach the basal ganglia through distributed inputs. Prefrontal, orbitofrontal, cingulate, insular, motor, and parietal cortices project to different but overlapping striatal territories. Thalamic input supplies additional information about events and ongoing state. Hypothalamic and brainstem systems influence arousal, autonomic support, and neuromodulation. The pattern reaching striatum therefore combines what can be done with its significance under present conditions.
Learned synaptic weights help convert that input into differential influence over output. In a thirsty person, a visible cup can recruit an approach-and-drink response. In a task requiring restraint, rule-related input can favor withholding the same response. The useful controlled variable includes both the probability of acting and the resources committed to the act.
34.8.5 Learned sequences acquire boundaries
An action can consist of several movements performed as a unit. Beginning a practiced sequence and ending it place different demands on control from repeating one of its internal movements. Nigrostriatal recordings show this distinction developing with learning.
Jin and Costa trained mice to earn a reward through eight lever presses. As performance became organized into bouts, neurons in striatum and substantia nigra increasingly showed activity associated with the first or last press of a sequence. Some responses distinguished sequences performed on different levers [@jincosta2010startstop].
The informative comparison is between similar presses at different positions in the learned bout. Activity concentrated at its boundaries indicates sensitivity to the organization of the sequence, beyond the occurrence of an individual press. Such boundary signals provide a candidate mechanism for recruiting a practiced action as a unit and terminating it at the appropriate point.
This connects sequence control to selection and vigor. A useful action must start, continue with sufficient intensity, and yield when it is complete or no longer appropriate. The learning that makes a sequence efficient, and the conditions under which it becomes habitual, are developed in the next chapter.
34.9 What Parkinson’s and Huntington’s diseases reveal
The go/no-go model gives Parkinson’s and Huntington’s diseases an intelligible relationship. Parkinsonian dopamine loss shifts the classical circuit toward greater restraint; early loss of indirect-pathway influence in Huntington’s disease weakens restraint. These predictions explain important features of the motor syndromes. Following each disease beyond those features reveals what rate changes alone leave unexplained.
34.9.1 Parkinson’s disease: action is available but difficult to mobilize
The defining pathology of Parkinson’s disease includes degeneration of dopamine neurons in the substantia nigra pars compacta and severe loss of dopamine in dorsal striatum. The cardinal motor sign is bradykinesia, slowness of movement accompanied by declining amplitude or speed during repeated action. Difficulty initiating movement and reduced spontaneous movement are also prominent. Facial expression, arm swing, handwriting, speech amplitude, and automatic adjustments can all be diminished. Rigidity and resting tremor are common, and postural instability becomes prominent in many patients.
The classical pathway model gives a useful first account. Dopamine loss reduces support for direct-pathway activity and removes D2-mediated modulation from indirect-pathway SPNs. The resulting circuit tends to increase inhibitory GPi/SNr influence over thalamic and brainstem targets. Controllers that can still generate movement have greater difficulty gaining sufficient access. The direct-pathway rescue in dopamine-depleted mice described above supplies an experimental counterpart: increasing that pathway’s activity restored the measured locomotor behaviors [@kravitzetal2010parkinsonian].
Dopamine replacement improves movement, as this account predicts. Recordings then extend the explanation from the amount of output to its temporal organization. Parkinsonian basal ganglia show abnormal timing, synchrony, and oscillation as well as changes in average firing. Activity in the beta range—roughly 13–30 Hz—can occur in prolonged bursts across cortex and basal ganglia. Longer beta bursts are associated with greater motor impairment, and dopaminergic medication or effective deep-brain stimulation reduces their prevalence [@tinkhauseretal2017bursts; @tinkhauseretal2017adaptive]. The duration and coordination of these bursts provide information that the average firing rate alone cannot supply.
Deep-brain stimulation, or DBS, makes the same point. High-frequency stimulation of STN or GPi can improve bradykinesia, rigidity, and tremor even though stimulation does not simply silence every neuron near the electrode. It changes local firing, axonal transmission, network synchrony, and the temporal pattern reaching connected structures. Modern adaptive systems can deliver stimulation in relation to pathological activity rather than continuously, using circuit dynamics as part of the treatment target.
Parkinsonian bradykinesia also exposes the relation between selection and vigor. The intended action may be known, and the relevant muscles may be strong enough, yet the spontaneous movement is delayed or underscaled. External cues can sometimes improve gait or movement amplitude by recruiting alternative sensory and cortical routes. Changing the sensory context changes how the movement is prepared and recruited.
Resting tremor is the feature least well explained by a simple increase in basal-ganglia output. Tremor involves interactions among basal ganglia, thalamus, cerebellum, cortex, and peripheral feedback. Its presence is a useful reminder that a disease can reveal a circuit’s function without every symptom following from one pathway sign.
Parkinson’s disease thus connects two levels of explanation. The pathway model explains why dopamine loss shifts the system toward poverty of movement. Network physiology explains why the timing and coordination of activity must also be restored for fluent control.
34.9.2 Huntington’s disease: unwanted movement escapes while voluntary control deteriorates
Huntington’s disease is caused by an expanded CAG repeat in the HTT gene and is inherited in an autosomal dominant pattern [@hdcollaborative1993gene]. The mutation produces a progressive disorder of striatum, cortex, and other systems. Motor, cognitive, and psychiatric changes can all be prominent.
Early neuropathology often affects striatal projection neurons of the indirect pathway more severely than direct-pathway neurons. Classic postmortem studies showed disproportionate loss of striatal neurons projecting to GPe and relative preservation of some neurons projecting to GPi and SNr during early stages [@reineretal1988differential; @dengetal2004differential]. Modern single-nucleus analyses refine that result: vulnerability depends not only on direct versus indirect identity but also on striosome–matrix organization and transcriptional state. In the early-grade human postmortem samples studied by Matsushima and colleagues, striosomal indirect-pathway SPNs were the most depleted population [@matsushimaetal2023vulnerabilities].
Loss of indirect-pathway influence can reduce restraint over downstream targets and contribute to chorea, the irregular, flowing intrusion of involuntary movement. Chorea demonstrates why suppression is an active achievement. A healthy nervous system does not merely generate the intended action. It continually prevents fragments, postures, and transitions that would interfere with it.
The person with Huntington’s disease is not simply moving too much. Chorea can coexist with impaired voluntary control, slowed saccades, poor sequencing, motor impersistence, and difficulty using sensory information to guide action. Cognitive and psychiatric changes may precede or accompany the obvious motor syndrome because associative and limbic corticostriatal loops are also affected. As degeneration spreads, later disease can become hypokinetic and rigid. The early release of unwanted movement therefore gives way to a broader failure of the recurrent system.
Huntington’s disease also extends beyond striatal cell death. Corticostriatal communication, interneurons, dopamine signaling, and cortical populations change over the course of the illness. The mutation is present throughout the nervous system, and the decades-long disease process alters development, compensation, and network function. The early indirect-pathway bias explains part of the initial syndrome; the spread of pathology explains why that initial pattern changes.
34.9.3 The comparison reveals a controlled balance
The two diseases expose the importance of both release and restraint. Parkinsonian dopamine loss changes the responsiveness and dynamics of a circuit whose downstream motor capacities remain substantially available. Huntington’s disease progressively damages striatal and cortical populations, often with an early bias that weakens indirect-pathway restraint.
Their clinical courses also distinguish the diseases from simple opposite settings on one control. Parkinson’s disease includes tremor and temporal disturbances that a static rate diagram does not explain. Huntington’s disease includes impaired voluntary movement even during chorea and can later become hypokinetic and rigid. The go/no-go model explains the direction of important early motor effects; cell vulnerability and network dynamics explain why the full syndromes extend beyond that first account.
34.10 How the experiments constrain an explanation
The evidence in this chapter links anatomy, circuit activity, and behavior. Each method contributes a different part of that link.
Comparative anatomy and physiology identify conservation. Molecular markers distinguish cell classes, tract tracing establishes their connections, and electrophysiology tests their operation. Finding the same combination in lamprey and mammals supports an ancient circuit scaffold with continuing access to brainstem controllers. The comparison concerns transmitters, receptor expression, connections, and physiological effects. It supplies a basis for asking how expanded cortical systems entered an existing control architecture [@stephensonjones2011basal; @stephensonjones2012dual].
Anatomical tracing establishes routes of influence. Anterograde tracers reveal where axons from an injected region terminate; retrograde tracers identify the neurons projecting to it. Comparing injections across territories reveals both specialization and convergence, including the overlapping cortical and thalamic inputs to striatum [@mcfarlandhaber2000convergent]. Tracing supplies the wiring needed to interpret activity. Recording and intervention determine how that wiring is used during a particular behavior.
Recording during behavior identifies when a circuit participates. Nigral pauses accompany saccades; striatal population activity rises around movement onset; sequence-related neurons distinguish the beginning and end of a bout. The experimental task determines which interpretations can be separated. Comparing a first press with an internal press asks a different question from comparing movement with rest.
The measurement also determines what can be inferred. Single-unit recordings resolve the spikes of individual neurons. Local field potentials measure extracellular voltage fluctuations generated by population currents and reveal coordinated rhythms. Genetically targeted calcium indicators identify activity in selected cell classes; a bulk signal pools activity across a population, whereas cellular imaging can resolve individual members. These distinctions matter when interpreting concurrent direct- and indirect-pathway activity: two population signals can rise together while their constituent neurons have different response preferences [@cui2013concurrent].
Multisite recordings follow the interaction among structures. Recording from STN and SNr during the same stopping task tests whether the stop cue reaches the circuit and whether the resulting influence reaches output in time [@schmidtetal2013canceling]. Combining cortical, basal-ganglia, and muscle measurements places neural events in relation to the developing action. Timing helps distinguish preparation, output, and sensory consequences within a recurrent system.
Perturbation tests causal influence. Collicular pharmacology changes the inhibition encountered by gaze-control neurons. Genetically targeted optogenetic tools permit brief activation or suppression timed to preparation, initiation, or a sequence transition; chemogenetic tools alter selected populations over longer intervals. Activating a large pathway population tests the effect it can exert, whereas suppressing activity at a particular moment tests its contribution to the behavior then underway. Combining recording with perturbation connects natural recruitment to causal function [@kravitzetal2010parkinsonian; @tecuapetlaetal2016complementary].
Primate experiments connect these approaches to differentiated gaze and limb control. The macaque studies of nigral firing and collicular pharmacology joined single-neuron activity to precisely measured saccades [@hikosakawurtz1983snr; @hikosakawurtz1983colliculus]. Tracing in the larger primate forebrain reveals how extensive cortical territories and dopaminergic projections are organized across striatum [@haberetal2000spiral]. Comparative anatomy, precise behavior, and focal intervention together support translation from a circuit mechanism to human action.
Human neurosurgery provides direct access to deep nuclei. During DBS implantation, investigators can record single neurons and local field potentials from STN or GPi. Temporary cortical electrodes can be combined with subthalamic recording and stimulation to test the timing and connectivity of frontal–basal-ganglia interactions [@chenetal2020hyperdirect]. Implanted electrodes also provide signals for adaptive stimulation, linking measurement to an intervention that changes with ongoing activity [@tinkhauseretal2017adaptive]. These recordings connect physiological events to the behavior and symptoms of the person being studied.
Disease, medication, and stimulation reveal failures and recoveries over different timescales. Parkinsonian slowness can coexist with preserved capacity for faster movement; medication or stimulation can change the expression of that capacity. Progressive disorders also alter connected regions and recruit compensation. Interpreting human recordings therefore requires knowing the disease stage, medication state, and stimulation condition, then comparing the result with animal experiments and healthy behavior.
Computational models state the proposed causal relationship explicitly. The canonical pathway model predicts output changes from synaptic signs. Decision models distinguish the rate at which evidence accumulates from the threshold required for commitment. The stopping race predicts that the same stop signal can succeed or fail depending on the state of preparation. Vigor accounts predict how effort and benefit alter the use of available motor capacity. A productive model specifies a measurable comparison that could distinguish it from another explanation.
This progression is the standard for the chapter’s organizing claim. Tonic inhibition and disinhibition provide mechanisms for access control. Pathway manipulations establish their behavioral effects. Recordings reveal how those mechanisms are recruited across time. The task is to explain how the observed population patterns select, scale, and interrupt particular actions.
34.11 From selection to learning what is worth selecting
The weights entering action selection change with experience. Success, pain, reward, and repeated practice alter the influence of particular state–action relations. Current need changes the value of an outcome, while learning changes what the organism expects an action to accomplish.
Dopamine participates in both current motor expression and learning. Changes in dopamine can alter how readily a striatal population is recruited; outcome-related dopamine signals can also change the strength of recently active synapses. The next chapter will distinguish those timescales and explain how prediction errors contribute to learning.
The anatomy offers several routes through which these influences can meet. Ventral striatal territories receive especially strong motivational and affective input. Associative territories integrate rules and action–outcome relations. Sensorimotor territories participate strongly in practiced actions. Primate tracing revealed an ascending striato–nigro–striatal spiral through which ventromedial territories can influence dopaminergic projections reaching progressively more dorsal and lateral striatum [@haberetal2000spiral]. This supplies an anatomical route from motivational information toward motor organization; the development of a habit remains a behavioral and learning question.
The present chapter establishes why changing those weights matters. Learned inputs alter the recruitment of basal-ganglia pathways, which alter inhibitory output and the expression of action. The next chapter asks how outcomes produce those changes, how behavior becomes habitual, and why this learning system is vulnerable in addiction.
34.12 Coda: controlling access to control
The basal ganglia influence a motor hierarchy that already contains capable controllers. Spinal networks organize local responses and recurring patterns; brainstem systems coordinate posture and orienting; cortical networks relate the body to targets, tools, and learned rules; cerebellar loops calibrate prediction and coordination. Basal-ganglia output changes the opportunity those controllers have to influence behavior.
The go/no-go model explains the basic mechanism. Direct-pathway inhibition of inhibitory output facilitates action. Indirect-pathway influence strengthens restraint. Hyperdirect input provides a rapid route to increased output inhibition. These opposing effects become selective and graded when distributed across populations and time. They can delay commitment while goals conflict, then regulate the release of an action, its vigor, its continuation, and its interruption.
The comparison with the cerebellum clarifies the division of work. Both systems receive information about action and return influence to cortical and brainstem controllers. Cerebellar circuits predict consequences, estimate evolving state, and recalibrate control from error. Basal-ganglia circuits regulate which controllers gain influence, how readily and vigorously they are recruited, and when their claim should end. In reaching for a cup, prediction helps anticipate the load and timing required to move and grasp; selection determines whether that reach should proceed, yield to another action, or be interrupted. Prediction can inform selection before commitment and continue to guide the movement afterward. These contributions interact throughout the action.
The experiments show how each extension earns its place. Manipulating inhibition changes the release of gaze. Pathway stimulation produces opposing motor effects, while recording and timed suppression show that both pathways participate in organized movement. Conflict tasks test how restraint changes the conditions for commitment. Stop-signal tasks expose the importance of relative timing. Parkinsonian reaching separates motor capacity from the readiness to use it.
Basal-ganglia loops control access to control. This perspective connects an ancient inhibitory architecture with the selection and scaling of embodied action. Its mechanistic content lies in the pathways, their targets, and the changing patterns of activity that make one response more effective and another less so.
Experience changes those patterns. The next chapter asks how the consequences of acting change what the organism will be prepared to do next time.
Established findings. The basal-ganglia scaffold is deeply conserved. GPi and SNr provide tonic inhibitory output; direct, indirect, and hyperdirect routes change that output through distinct synaptic chains. Pathway-specific manipulations produce strong effects on movement, and both principal striatal pathways participate during action. Nigrostriatal dopamine is essential for normal motor expression. Parkinson’s and Huntington’s diseases reveal major failures of movement scaling, coordination, and restraint.
The organizing interpretation. Go/no-go effects explain how the circuit releases and restrains action. Extending those effects across populations and time provides an account of selective access, conflict-dependent commitment, vigor, stopping, and switching. The experiments make this an explanatory framework with testable mechanisms.
Open questions. How are particular action components distributed across the populations? Which output patterns determine initiation, vigor, and sequence transitions? How do rate, synchrony, and timing interact at different targets? These are the questions needed to connect the clear pathway mechanisms to the full organization of natural behavior.