Basal Ganglia: Action Selection, Learning, and Movement
Author: Ukrainian Psychological Hub · Published: October 12, 2026 · Editorial Policy
The basal ganglia are interconnected nuclei deep in the brain that help regulate what actions are selected, when they begin, how intensely they are expressed, and how behavior changes after experience. They contribute to voluntary movement, reinforcement learning, habits, action sequences, motivated behavior, and aspects of cognitive control. They operate through extensive connections with the cerebral cortex, thalamus, brainstem, and midbrain. Their functions emerge from coordinated activity across distributed neural circuits. Costa (2019); Park et al. (2020).
Consider reaching for a glass while talking. The nervous system evaluates competing possible actions, initiates an appropriate reach, scales its speed and force, and updates future behavior if the glass is unexpectedly heavy. Cortical, cerebellar, brainstem, spinal, sensory, and basal ganglia circuits participate in different aspects of this process. The basal ganglia are especially important for regulating the selection, initiation, and specification of behavior and for adapting action policies with experience. This example illustrates network cooperation; it does not mean that any one nucleus performs the whole task.
What are the basal ganglia?
The basal ganglia are groups of densely interconnected subcortical nuclei. Principal components include the striatum, the globus pallidus, the subthalamic nucleus, and the substantia nigra. Anatomical conventions differ slightly about which structures are included. “Basal nuclei” is an anatomically defensible alternative name because a ganglion usually denotes neuronal cell bodies outside the central nervous system. “Basal ganglia” remains the widely used clinical and scientific term. Fazl and Fleisher (2018).
Basal ganglia circuits receive extensive inputs from cortical and thalamic regions. The striatum integrates many of these inputs with neuromodulatory signals, especially dopamine; interconnected basal ganglia nuclei influence activity in the thalamus and brainstem; thalamic projections return to the cortex. The system is therefore well positioned to modify action control across movement, associative cognition, and motivational processing. Historically influential models described approximately parallel cortico-basal ganglia loops, while subsequent research demonstrated considerable interaction among them. Alexander, DeLong, and Strick (1986).
Main structures and functional anatomy
Striatum: the principal input territory
The dorsal striatum comprises the caudate nucleus and putamen in common human anatomical descriptions. The ventral striatum includes the nucleus accumbens and adjacent regions. These territories differ in connectivity and often in their contributions to movement, associative learning, incentives, and choice. They nonetheless participate in overlapping networks, so the dorsal striatum cannot be described as exclusively motor and the ventral striatum as exclusively emotional or pleasurable.
The striatum contains large populations of inhibitory, GABA-releasing spiny projection neurons. Direct-pathway-enriched neurons commonly express D1-family dopamine receptors; indirect-pathway-enriched neurons often express D2-family receptors. This classification is useful for understanding core circuitry, but receptor expression, local interactions, neuron populations, and projection targets are more complex than a strict two-cell-type diagram. Cholinergic and GABAergic interneurons shape processing, while dopamine modulates excitability and plasticity in ways that depend on receptor, location, state, and experience. Calabresi et al. (2014); Lee and Sabatini (2025).
Globus pallidus: external and internal segments
The globus pallidus is divided into external (GPe) and internal (GPi) segments. In classical pathway models, GPe has a prominent position in indirect-pathway interactions, and GPi serves as an important output nucleus. Pallidal populations also make additional recurrent and long-range connections. Many GPi output cells exert ongoing GABAergic inhibition of downstream targets; changing that activity can release specific target circuits from inhibition. A mechanism called disinhibition helps explain selective facilitation without implying that pallidal neurons act as isolated switches. Fazl and Fleisher (2018).
Subthalamic nucleus: a hub for rapid adjustment
The subthalamic nucleus (STN) contains predominantly excitatory glutamatergic projection neurons. It participates in indirect-pathway interactions and receives direct cortical inputs through the hyperdirect pathway. These relationships have motivated research into rapid action adjustment, stopping, conflict, and response thresholds. Which interpretation is appropriate depends on timing, task, neural population, and broader network state. The STN is also an established stimulation target in selected people with Parkinson’s disease. NINDS, Deep Brain Stimulation.
Substantia nigra: dopamine and output circuits
The substantia nigra pars compacta (SNc) contains dopamine-producing neurons whose projections strongly influence dorsal striatal processing. Loss of these neurons is central to the movement-related pathology of Parkinson’s disease. The substantia nigra pars reticulata (SNr) is a primarily inhibitory output structure that influences thalamic and brainstem targets, including circuits important for orienting and eye movements. These two divisions have different cell populations and functions; speaking of “the substantia nigra” as a homogeneous dopamine center obscures that distinction. Fazl and Fleisher (2018); Hikosaka et al. (2019).
The ventral tegmental area (VTA), another midbrain dopamine region, connects prominently with ventral striatal and cortical territories. It is related to, but distinct from, SNc. Dopamine and reinforcement therefore involve wider networks than one anatomical nucleus. For a transmitter-focused explanation, see Dopamine: Reward Prediction, Learning, Motivation, and Movement.
How basal ganglia circuits regulate action selection
Action selection is the process by which possible behaviors are favored or suppressed based on goals, sensory circumstances, learned associations, anticipated results, and current bodily state. Cortex and thalamus provide evolving input to basal ganglia circuitry. Changes in striatal, pallidal, subthalamic, and nigral activity influence the participation of downstream thalamic and brainstem networks. This contributes to choosing a response, organizing its onset, changing its vigor, maintaining a sequence, switching actions, or withholding inappropriate responses. Costa (2019).
A classical concept is disinhibition: certain basal ganglia output populations tonically inhibit target circuits; reducing inhibition makes activity in those target circuits easier. This is a biologically grounded explanation of selective facilitation, not a literal claim that one region opens a single gate for every movement. Whether an action occurs depends on simultaneous activity in cortex, basal ganglia, cerebellum, brainstem, spinal circuits, and sensorimotor feedback.
Direct pathway: selective facilitation
The classical direct pathway begins with cortical excitation of striatal projection neurons, often enriched in D1-family dopamine receptors. These neurons inhibit GPi and SNr output neurons, thereby reducing inhibition of particular downstream targets. The model thus predicts facilitation of selected actions through disinhibition. The common label “Go pathway” captures this basic mechanism. It does not establish that each direct-pathway neuron always promotes every movement or that activity in the pathway alone is sufficient for complex voluntary action. Calabresi et al. (2014).
Indirect pathway: suppressing alternatives and learning from failure
In the classic indirect pathway, D2-enriched striatal neurons project to the GPe and influence networks involving the STN and GPi/SNr. Through these interactions, inhibitory output toward selected targets may increase, potentially suppressing competing actions. The shortcut “NoGo pathway” remains widespread but conveys only part of its proposed function. A 2025 perspective proposes that synaptic plasticity in indirect-pathway circuits supports avoiding actions with unfavorable outcomes while allowing exploration of alternative options. This mechanism is an active theoretical proposal rather than a completed description of every human decision. Lee and Sabatini (2025).
Hyperdirect pathway: fast influence through the STN
The hyperdirect pathway connects particular cortical areas to STN without passing first through striatum. Its comparatively rapid influence on basal ganglia output has been investigated in response stopping, conflict, and action-threshold theories. “Emergency brake” is a convenient metaphor but does not describe a universally acting inhibitory button. Motor and associative territories, context, and connectivity matter, and an experimental signal in STN cannot automatically be interpreted as a person deciding not to act.
Why the simple Go/NoGo model is incomplete
The direct-versus-indirect diagram is valuable for introducing anatomical projections and for understanding aspects of movement disorders. Yet experimental evidence challenges the assumption that these pathways operate as a pair of exclusively alternating switches. In a mouse experiment, simultaneous changes in activity of direct- and indirect-pathway striatal neurons preceded action initiation. The result demonstrated coordinated recruitment in those animals and tasks; it did not prove that all human actions recruit the pathways in identical ways. Cui et al. (2013).
Subsequent accounts emphasize cell-population specificity, timing, local striatal interactions, projection heterogeneity, and plasticity. A pathway’s functional consequences can differ according to the action represented, the point in the action sequence, and the behavioral conditions. The classical model offers a foundation; modern research provides the richer, testable network account. Calabresi et al. (2014); Park et al. (2020).
Movement initiation, movement vigor, and action sequences
Initiating and scaling a movement
The basal ganglia influence not only which action is favored, but also when an action starts and how dynamically it unfolds. Reaching, walking, orienting the eyes, and many self-initiated movements require context-sensitive control of onset, force, direction, and speed. Research on action specification suggests that the basal ganglia help regulate graded aspects of behavior as well as categorical choice. They operate together with motor cortex and descending premotor systems. Park et al. (2020).
Movement vigor refers broadly to how energetically or rapidly an action is carried out under given conditions. Dopamine-related and corticostriatal processes can influence vigor, effort, and action initiation. Vigor should not be treated as a direct biological synonym for enthusiasm or a universal measure of motivation. Its interpretation depends on the task, costs, expected benefits, and the organism’s state. Frank (2025).
Learning and performing action sequences
Many actions consist of ordered sequences: typing, playing an instrument, preparing food, or taking several steps before turning. Basal ganglia circuits contribute to acquiring and performing sequences, including aspects of initiation, transition, and termination. Sequence learning unfolds across striatal, cortical, and other networks. Models based on reinforcement learning explain important features, while goal-directed hierarchical sequences challenge oversimplified models. Garr (2019).
A sequence is an ordered pattern of actions. A habit is classified by the way behavior is controlled, particularly by reduced sensitivity to outcome devaluation and action-outcome contingency under appropriate tests. A practiced sequence can remain goal-directed or become more habitual. These distinctions are central to interpreting experimental studies of striatal dopamine and skill learning. Song and Lee (2024).
Gaze and orienting
Basal ganglia outputs influence eye-movement and orienting pathways. For example, projections involving caudate and SNr affect the superior colliculus, a midbrain structure central to gaze orientation. Work in primates has explored how learned object value and basal ganglia output alter selection of eye movements. This broadens the concept of action selection beyond limbs to the organization of attention-related overt behavior. Hikosaka et al. (2019).
Reinforcement learning and dopamine
Reinforcement learning describes changes in behavior driven by experienced consequences. Corticostriatal synapses and dopaminergic modulation participate in learning about the value of choices under different circumstances. A prominent computational hypothesis proposes that some brief dopamine responses convey a reward prediction error: a difference between the outcome expected and the outcome received. This account has deep experimental support, yet contemporary studies show additional dopamine signals related to movement, salience, effort, and behavioral state. Gershman et al. (2024); Frank (2025).
When an action produces an unexpectedly beneficial result, a prediction-error-like signal may help alter synaptic learning in circuits representing relevant circumstances and actions. When an expected benefit fails to appear, learning signals may change in another direction. The direction and outcome of plasticity depend on receptor pathways, local circuits, timing, and previous experience. “Dopamine equals pleasure” and “more dopamine always means more learning” are scientifically unreliable shortcuts.
Direct human evidence and its limits
A 2023 study measured subsecond fluctuations of extracellular dopamine in the human striatum during a probabilistic reward-and-punishment learning task in patients undergoing deep brain stimulation surgery. Results supported temporally differentiated signals associated with reward and punishment prediction errors. Its unusual direct chemical measurement is a major strength; the selected clinical sample and surgical context limit generalization to every person and everyday situation. Sands et al. (2023).
Animal recordings and pathway-specific manipulations continue to supply unusually precise causal insights, while human neuroimaging, clinical electrophysiology, PET, and medication studies offer complementary evidence. A correct scientific account distinguishes a computational fit to behavior from a directly measured neurotransmitter signal and from a causal perturbation of a circuit.
Goal-directed control and habits
A goal-directed action depends on the anticipated outcome and the causal relationship between action and result. A habit is less sensitive to outcome devaluation or changed contingencies under a valid experimental test. Work in animals implicates dorsomedial associative and dorsolateral sensorimotor striatal territories differently in these forms of control, although these territories participate in larger systems and their functions are not strictly exclusive. Cross-species interpretation requires care because human striatal topography and behavioral experiments differ. Yin and Knowlton (2006).
Habits can be useful adaptations that support efficient behavior. Repetitive activity becomes a clinical concern depending on distress, impairment, inflexibility, and disorder-specific features, not from a hypothetical brain scan of “habit dominance.” To explore the wider brain-behavior relationship, see How the Brain Produces Behavior: Neurons, Circuits, and Networks.
Cognition, motivation, and emotion
Motor, associative, and motivational cortico-basal ganglia loops provide a framework for studying working-memory updating, behavioral flexibility, cognitive control, reward valuation, and effort-based action. These functions involve distributed systems; no single striatal or pallidal region produces thought, emotion, or personality on its own. Circuit interactions matter more than a one-region-one-mental-function chart. Alexander et al. (1986); Frank (2025).
Ventral striatal activity is commonly investigated in incentive learning and motivation. Increased activation on a brain scan cannot by itself prove that someone is happy, desires an object, has formed an attachment, or is experiencing pleasure. Reverse inference is especially problematic because the same region participates in multiple computations across tasks. The larger reward-circuit explanation is addressed by Brain Reward System: Circuits, Reinforcement, and Motivation.
Recent cost-benefit models propose that D1- and D2-related circuitry may coordinate local decisions about individual actions and broader choices about engaging a course of behavior. Models are useful when they produce clear predictions that can be tested against behavior and neural recordings. Their mathematical variables should not be mistaken for direct measurements of universal human desires or motives. Frank (2025).
How scientists study the basal ganglia
Structural MRI and diffusion imaging
Structural magnetic resonance imaging reveals major nuclei, their size, shape, and certain tissue characteristics. Diffusion MRI measures patterns of water diffusion and may support models of white-matter organization. Tractography estimates probable fiber trajectories; it does not photograph individual axons or establish the existence and direction of every reconstructed connection. Group-level differences in striatal volume do not automatically predict an individual’s decision-making ability or psychiatric diagnosis.
Functional MRI
Functional MRI usually measures blood oxygen level-dependent (BOLD) changes. These are hemodynamic signals associated with underlying neural and metabolic activity, not direct recordings of action potentials, dopamine molecules, or subjective states. An fMRI activation in the putamen or nucleus accumbens may support an experimentally specified contrast but does not establish which of several possible psychological processes is present. Interpretation depends on design, controls, preprocessing, statistics, and convergent methods.
PET and neurotransmitter imaging
Positron emission tomography uses radiotracers; according to the tracer and protocol, it can estimate receptor or transporter binding, metabolism, or changes relevant to neurotransmitter release. Such measurements depend on kinetic modeling and spatial or temporal constraints. PET does not offer a general instantaneous map of pleasure or intent. Combining PET with fMRI can add complementary information, although correlations between signals are not automatically causal explanations.
Recordings, stimulation, and animal studies
Human intracranial electrodes can record local field potentials or, in certain contexts, individual neuronal activity at high temporal resolution. Local field potentials reflect aggregated electrical dynamics and cannot simply be translated into one thought. Direct dopamine sampling has been achieved in special surgical contexts, adding evidence beyond imaging proxies. Sands et al. (2023).
Animal studies can combine behavioral tests with pathway-selective optical recordings, optogenetics, or pharmacological manipulations. They allow strong causal investigation of specified cells and circuits under defined conditions, as illustrated by the mouse study of concurrent direct- and indirect-pathway activation. Translating those effects to human cognition requires convergence across species, tasks, and methods. Cui et al. (2013).
Human lesion studies and clinically indicated deep brain stimulation can provide important intervention evidence. Yet lesions and stimulation influence networks of interconnected regions and fibers. Treatment effects do not prove that a stimulated nucleus exclusively generates a mental function. NINDS, Deep Brain Stimulation.
Basal ganglia dysfunction and neurological conditions
Parkinson’s disease
Parkinson’s disease involves degeneration of dopamine-producing neurons in the SNc and abnormalities in connected basal ganglia networks. Common motor manifestations include bradykinesia, rigidity, rest tremor, and gait or postural problems. The disorder can also affect sleep, autonomic function, cognition, and mood. Dopamine loss is central to its pathophysiology, while the exact expression of symptoms involves complex network activity and cannot all be explained by a simple direct/indirect balance. Clinical diagnosis is based on neurological evaluation, with selected tests when indicated. NINDS, Parkinson’s Disease.
Medications including levodopa can improve motor symptoms in many patients. Deep brain stimulation of selected targets, including STN or GPi, can help certain individuals whose symptoms require additional management. DBS adjusts circuit activity and requires specialized selection, programming, and long-term care. It is not a cure for neurodegeneration and cannot be assumed to improve every motor or nonmotor symptom. NINDS, Deep Brain Stimulation.
Huntington’s disease
Huntington’s disease is an inherited neurodegenerative disorder strongly associated with injury to striatal and related circuits. It can produce involuntary movement, impaired voluntary motor control, cognitive changes, and behavioral symptoms. These diverse effects illustrate network involvement extending beyond basic muscle movement. Its diagnosis, genetics, clinical course, and treatment are subjects of condition-specific neurological practice, not conclusions from an isolated brain-region description. Fazl and Fleisher (2018).
Dystonia, tics, and other circuit-related conditions
Dystonia and tic disorders have prompted extensive investigations of basal ganglia and interacting cortical and cerebellar networks. Cortico-striato-thalamo-cortical abnormalities are also studied in obsessive-compulsive disorder and other psychiatric conditions. Similar anatomical terminology does not make these conditions interchangeable. Research associations, clinical signs, diagnostic criteria, and intervention evidence must be evaluated separately. There is no universally validated basal ganglia brain-scan signature that diagnoses all such conditions in an individual. Fazl and Fleisher (2018).
Established findings and active debates
There is convergent evidence that basal ganglia circuits participate in movement regulation and learning from action consequences; that striatum receives cortical, thalamic, and dopamine inputs; and that disruptions to specific nuclei and projections can alter motor behavior. Anatomy, physiology, human neurological observations, and animal causal experiments support these conclusions. Costa (2019).
Research continues to refine the division of labor between direct and indirect pathways; the roles of dopamine in prediction, vigor, and state transitions; the human relationship between striatal subterritories, skills, and habits; and the extent to which different learning models describe natural behavior. Reviews published in 2024 and 2025 present competing and integrative models that are scientifically valuable because they generate experimentally testable predictions. They are not proof of one universal computational law. Gershman et al. (2024); Lee and Sabatini (2025); Frank (2025).
Animal tasks, clinical samples, medication status, recording methods, and timescales materially affect the meaning of experimental findings. Correlation does not establish causation. Successful treatment at an anatomical target does not mean that the target uniquely owns the affected psychological process. An imaging pattern observed across a research group is not equivalent to a validated individual diagnostic instrument.
Common misconceptions
“The basal ganglia only control movement”
Movement is a major function, but associative, reinforcement-related, and motivational circuits also involve basal ganglia nuclei. The underlying processing is distributed across multiple networks. Alexander et al. (1986).
“The direct pathway turns everything on and the indirect pathway turns it off”
This is a teaching shorthand for a classical circuit diagram. Both pathways can be active during the same action, and effects depend on cell population, location, timing, and behavioral circumstances. Cui et al. (2013).
“Dopamine is the pleasure chemical”
Dopamine has diverse roles in learning, anticipation, effort, movement, and motivational regulation. Neurochemical signals depend on projection, receptor, timing, and state; a global pleasure-meter model is inaccurate. Gershman et al. (2024).
“Every repeated action is a habit”
Repetition alone is insufficient. Experimental habits are identified by the relationship between behavior, expected outcome, and changed contingencies. Practiced action sequences can remain goal-directed. Yin and Knowlton (2006).
“A scan of the basal ganglia reveals a person’s personality or motivation”
Group-level neuroimaging findings cannot serve as universal individual tests of mental states. Valid clinical imaging addresses particular medical questions under appropriate diagnostic standards.
Frequently asked questions
What is the main function of the basal ganglia?
They help interconnected brain circuits select, initiate, regulate, and learn actions. Their functions include movement control, learned behavior, and certain aspects of cognitive and motivational regulation.
Where are the basal ganglia located?
Their component nuclei are found deep within the forebrain and neighboring subthalamic or midbrain regions. The striatum and globus pallidus are deep forebrain structures, the STN lies in the subthalamic region, and the substantia nigra is in the midbrain.
Are the basal ganglia and striatum the same?
The striatum is a major component of the basal ganglia, not a synonym for the entire system. Other nuclei include the globus pallidus, subthalamic nucleus, and substantia nigra.
What do the direct and indirect pathways do?
Classical models associate the direct pathway with selective facilitation and the indirect pathway with modulation or suppression of competing options. Modern findings show that both can participate during action and that their functional effects are more context-dependent.
What is the difference between the basal ganglia and cerebellum?
Both contribute to movement and learning. Basal ganglia research often emphasizes action selection, initiation, vigor, and reinforcement-related control, while cerebellar research often emphasizes prediction, timing, and calibration of sensory-motor performance. Their networks interact extensively.
Do the basal ganglia produce dopamine?
Dopamine-producing cells in the substantia nigra pars compacta are part of basal ganglia-related anatomy and project prominently to striatum. Other dopamine cells, including those of the VTA, are located in distinct but connected systems.
Can basal ganglia problems affect decision-making and memory?
Dysfunction in connected networks may affect cognitive flexibility, learning, and motivated behavior as well as motor performance. The precise pattern depends on the neurological condition, location, and broader networks involved.
Can exercise or supplements “reset” the basal ganglia?
Learning and practice can change neural function, and physical rehabilitation is important in many neurological conditions. Claims that one product or routine specifically resets striatal dopamine or basal ganglia pathways need direct clinical evidence. Individual medical decisions should be based on evaluated symptoms and validated interventions.
Conclusion
The basal ganglia influence how the brain organizes action and adapts behavior from experience. Striatum, globus pallidus, STN, substantia nigra, cortex, thalamus, and brainstem participate in interconnected circuits through which movement, learning, motivation, and some cognitive processes are regulated. The basic anatomy and involvement in movement are well established. Contemporary research explains their operation through coordinated, temporally precise, and experience-dependent circuit activity, replacing excessively rigid interpretations of a simple two-pathway switch.
Related Articles
References
Alexander, G. E., DeLong, M. R., & Strick, P. L. (1986). Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annual Review of Neuroscience, 9, 357–381. DOI
Calabresi, P., Picconi, B., Tozzi, A., Ghiglieri, V., & Di Filippo, M. (2014). Direct and indirect pathways of basal ganglia: A critical reappraisal. Nature Neuroscience, 17, 1022–1030. DOI
Costa, R. M. (2019). What, if, and when to move: Basal ganglia circuits and self-paced action initiation. Annual Review of Neuroscience, 42, 459–483. DOI
Cui, G., Jun, S. B., Jin, X., et al. (2013). Concurrent activation of striatal direct and indirect pathways during action initiation. Nature, 494, 238–242. DOI
Fazl, A., & Fleisher, J. (2018). Anatomy, physiology, and clinical syndromes of the basal ganglia: A brief review. Seminars in Pediatric Neurology, 25, 2–9. DOI
Frank, M. J. (2025). Adaptive cost-benefit control fueled by striatal dopamine. Annual Review of Neuroscience, 48, 1–22. DOI
Garr, E. (2019). Contributions of the basal ganglia to action sequence learning and performance. Neuroscience & Biobehavioral Reviews, 107, 279–295. DOI
Gershman, S. J., et al. (2024). Explaining dopamine through prediction errors and beyond. Nature Neuroscience, 27, 1645–1655. Journal
Hikosaka, O., Kim, H. F., Amita, H., Yasuda, M., Isoda, M., Tachibana, Y., & Yoshida, A. (2019). Direct and indirect pathways for choosing objects and actions. European Journal of Neuroscience, 49, 637–645. DOI
Lee, J., & Sabatini, B. L. (2025). From avoidance to new action: The multifaceted role of the striatal indirect pathway. Nature Reviews Neuroscience, 26, 438–449. Journal
National Institute of Neurological Disorders and Stroke. (n.d.). Deep brain stimulation. Official source
National Institute of Neurological Disorders and Stroke. (n.d.). Parkinson’s disease: Challenges, progress, and promise. Official source
Park, J., Coddington, L. T., & Dudman, J. T. (2020). Basal ganglia circuits for action specification. Annual Review of Neuroscience, 43, 485–507. DOI
Sands, L. P., et al. (2023). Subsecond fluctuations in extracellular dopamine encode reward and punishment prediction errors in humans. Science Advances, 9, eadi4927. DOI
Song, M. R., & Lee, S. W. (2024). Rethinking dopamine-guided action sequence learning. European Journal of Neuroscience, 60(1), 3447–3465. DOI
Yin, H. H., & Knowlton, B. J. (2006). The role of the basal ganglia in habit formation. Nature Reviews Neuroscience, 7, 464–476. DOI