Dopamine & Learning: The Neurochemistry of Reward, Motivation and Canine Training

2. What Is Dopamine? A Neurochemical Primer

Dopamine is a monoamine neurotransmitter synthesized from the amino acid tyrosine. It is produced primarily in two midbrain nuclei: the ventral tegmental area (VTA) and the substantia nigra pars compacta (SNc) . From there, dopamine neurons project to various forebrain regions, including the striatum (caudate nucleus and putamen), nucleus accumbens, and prefrontal cortex.

Crucially, the dopaminergic reward system appears to share broad anatomical similarities across mammals. Several fMRI studies in dogs have demonstrated activity in the striatum that is consistent with reward prediction error signals for both primary rewards (such as food) and social rewards (such as praise). However, it is important to note that fMRI measures changes in blood oxygenation (BOLD signal), not direct dopamine release. Conclusions about dopamine function in dogs are therefore largely inferred from human and rodent research and from indirect measures.

Dopamine’s effects are mediated through five receptor subtypes (D1–D5), broadly classified into two families:

• D1‑like receptors (D1, D5) – Excitatory, involved in reward processing and motor control.

• D2‑like receptors (D2, D3, D4) – Inhibitory, involved in modulating reward, motivation, and impulse control.

The balance between D1 and D2 receptor activation likely shapes how dopamine influences learning and behavior. Too little dopamine – or too few functional receptors – may impair motivation and learning capacity; too much may contribute to impulsivity and compulsive behaviors.

For a detailed exploration of dopamine’s role in impulse control and frustration, see The Neurobiology of Frustration in Dogs.

3. Reward Prediction Error: The Proposed Engine of Learning

The most influential model of dopamine’s role in learning is reward prediction error (RPE) – the discrepancy between an expected reward and the actual reward received. This signal is thought to be the engine of reinforcement learning, allowing animals (including dogs) to update the value of actions and cues based on outcomes.

3.1 The RPE Signal Explained

• Positive RPE (better than expected) – When a reward is larger or more valuable than expected, dopamine neurons fire in a phasic burst. This is believed to strengthen the neural pathways associated with the preceding behavior, making the dog more likely to repeat it.

• Negative RPE (worse than expected) – When a reward is smaller than expected or entirely absent, dopamine neuron activity is suppressed below baseline. This signals that the previous behavior was less valuable than thought, reducing the likelihood of repetition.

• No RPE (exactly as expected) – When the reward matches expectations perfectly, dopamine neurons show little net change. Learning may stagnate because there is no prediction error to drive updating.

As animals learn the association between a cue and a reward, the timing of dopamine release shifts, becoming associated with the cue itself rather than the reward. This likely explains why a clicker or a verbal marker becomes reinforcing: it predicts reward and thus may trigger dopamine release even before the treat is delivered.

3.2 Evidence from Canine fMRI

Using awake canine fMRI, researchers have observed activation in the ventral caudate of dogs in response to a hand signal that predicted food reward, relative to a signal that predicted no reward (Berns et al., 2012). The mean differential caudate response (0.09%) was similar to comparable human studies, suggesting that reward prediction error signals are preserved across species. However, as with all fMRI research, this measures BOLD signal – a correlate of neural activity – not direct dopamine release.

3.3 A Note on Causality vs. Correlation

It is important to emphasize that dopamine signaling is associated with reward prediction and learning, but the precise causal role in dogs has not been experimentally established. Most of the causal evidence comes from rodent and primate studies using optogenetics or direct dopamine antagonists. In dogs, the evidence is largely correlational.

For a deeper understanding of how emotional states can override learned behaviors – even when dopamine signaling is intact – see Learned Behavior vs. Emotional Response: Why Dog Training Sometimes Fails.

4. Action Prediction Error (APE): Emerging Evidence from Rodent Models

Recent research, primarily in rodents, has identified a second type of dopamine teaching signal: action prediction error (APE) . APE encodes the difference between the action that is taken and the extent to which the action was predicted. Unlike RPE, APE is thought to be a value‑free teaching signal that reinforces repeated associations, supporting the formation of habits and repetitive behaviors independently of reward value.

4.1 Current State of Evidence

It is important to note that the APE concept is emerging evidence and has been primarily derived from rodent models. Whether APE operates in the same way in dogs – or even exists as a separate system – is not yet fully validated in canine neuroscience. The following discussion is therefore speculative and should be interpreted as one possible mechanism among several.

Computational modeling and rodent experiments suggest that APEs alone cannot support reward‑guided learning, but when paired with RPE circuitry, they may serve to consolidate stable stimulus–action associations in a value‑free manner.

4.2 Potential Implications for Training (with Caution)

If APE-like mechanisms exist in dogs, they might help explain:

• Why behaviors become automatic (habit formation) with repetition, even without continuous reinforcement.

• Why both desirable and undesirable habits are difficult to break once established.

• Why some dogs develop compulsive, repetitive behaviors that are resistant to extinction.

However, these implications remain speculative. Trainers should base their methods on the well‑established RPE framework rather than on the more speculative APE model.

For more on how habits and automatic behaviors develop, see The Neurology of Dog Behavior – How the Brain Affects Dog Training.

5. Dopamine in the Canine Brain: What fMRI Reveals

Advances in awake canine fMRI have allowed researchers to observe brain activity related to reward processing in dogs. However, it is crucial to understand the limitations of this method (see Section 10).

5.1 The Caudate Response to Reward-Predicting Cues

In a landmark study, dogs were trained to respond to hand signals indicating the imminent availability of a food reward. Consistent with RPE theory, researchers observed significant activation in the ventral caudate of dogs in response to the hand signal that indicated “reward” relative to the signal that indicated “no reward” (Berns et al., 2012). This suggests that the canine caudate encodes reward expectancy in a manner similar to humans.

5.2 Social vs. Food Reward: Individual Differences

A follow‑up study (Cook, Prichard, Spivak, & Berns, 2016) used fMRI to probe the neural basis for dogs’ preferences for social interaction vs. food reward. Using the ventral caudate as a measure of intrinsic reward value, the researchers found that the caudate was significantly more active to reward‑predicting stimuli and showed roughly equal or greater activation to praise vs. food in 13 of 15 dogs.

Importantly, the relative caudate activation to food‑ and praise‑predicting stimuli was a strong predictor of each dog’s sequence of choices in a subsequent Y‑maze task. This suggests a neural mechanism for preference in domestic dogs that is stable within, but variable between, individuals. For some dogs, social praise may be more rewarding than food; for others, the opposite may be true.

5.3 Implications for Training

These findings have direct practical applications:

• Reinforcers are not universal – What works for one dog may not work for another. Individual dopamine‑related sensitivity to different reward types should be assessed.

• Social interaction may act as a primary reward – For many dogs, owner attention and praise appear to trigger reward‑related brain activity comparable to or greater than food.

• Reward value can be inferred from behavior – Simple choice tests (e.g., Y‑maze preferences) can help identify what a dog truly finds rewarding.

For a deeper look at how the prefrontal cortex and reward systems interact in self‑control, see The Role of the Prefrontal Cortex in Canine Self‑Control.

6. Dopamine and Training: Practical Applications

Understanding the dopamine‑based framework of reward prediction error translates directly into training strategies. The following principles are derived from well‑established RPE research, not from speculative models.

6.1 Variable Reinforcement Maintains Dopamine Responsivity

When rewards are entirely predictable, dopamine responses flatten. The dog knows exactly what will happen, there is little positive prediction error, and motivation may decrease. This is why continuous reinforcement (rewarding every correct response) is excellent for initial learning but may be suboptimal for maintaining long‑term motivation.

Variable reinforcement schedules (e.g., random ratio or random interval) are thought to maintain higher dopamine responsivity because each reward has the potential to generate a positive prediction error. The dog never knows exactly when the next reward will come, keeping the dopamine system engaged.

6.2 Surprise Rewards Likely Generate Strong Dopamine Bursts

Unexpected rewards – “jackpots” or surprise treats – generate the largest positive prediction errors, likely producing the strongest dopamine bursts and the most powerful reinforcement of the preceding behavior. This is why occasionally surprising a dog with a high‑value reward can dramatically strengthen a behavior.

6.3 Anticipation May Be More Powerful Than Consumption

Some researchers suggest that it is the anticipation of reinforcement, not the reinforcement itself, that is most enjoyable. Learning is driven by motivation, and motivation is fueled by anticipation. This is why creating excitement and anticipation in training – using cues that predict rewards, varying the timing of reinforcement, and building a “game” around the training – may be neurobiologically more effective than simply delivering treats robotically.

6.4 A Caution: Partial Rewarding in Naïve Dogs

A 2021 study on partial rewarding during clicker training in naïve dogs (Cimarelli et al., 2021) found an important caution: while variable reinforcement is powerful in experienced learners, partial rewarding did not improve learning speed in naïve dogs and was associated with a more pessimistic affective state. Dogs that were partially rewarded (60% of the time) showed a more pessimistic bias in a cognitive bias test than dogs that were continuously rewarded, suggesting that partial rewarding might negatively affect welfare in inexperienced dogs.

Practical implication: Use continuous reinforcement during initial learning phases to establish clear associations. Once the behavior is well‑established, introduce variability to maintain motivation.

6.5 Novelty May Drive Dopamine Release

Rodent research suggests that novelty itself drives dopamine release. Transient dopamine levels in the basal ganglia encode novelty, contributing to an uncertainty representation that efficiently drives exploration. This may explain why dogs are often more engaged when training includes novel elements – new locations, new props, variations on familiar cues.

For more on the interplay between dopamine and stress hormones in training, see Aversive Training Methods – Neurological Effects in Dogs.

7. Dopamine and Problem Behaviors: Associations and Correlations

Dysregulation of the dopamine system has been associated with several canine behavioral problems. However, it is crucial to distinguish correlation from causation.

7.1 Dopamine and ADHD‑like Behavior in Dogs

Dogs can exhibit a constellation of behaviors resembling Attention‑Deficit/Hyperactivity Disorder in humans, including impulsivity, attention issues, hyperactivity, and sometimes aggression. Researchers have used the term “ADHD‑like” to describe this phenotype. It is important to note that this term is a behavioral construct and not a formally established diagnostic category in veterinary medicine.

A study evaluating serum dopamine and serotonin levels in dogs with ADHD‑like symptomatology (González‑Martínez et al., 2023) found that dogs clinically classified as ADHD‑like showed lower dopamine concentrations compared to control dogs. Dopamine levels were also associated with aggression, hyperactivity, and impulsivity.

The expression of ADHD‑like behavior in dogs appears to depend on a complex gene–environment interaction, as is the case with many neurological disorders in humans. This means that genetic predisposition combined with environmental factors (stress, inconsistent training, lack of structure) may determine whether these behaviors emerge.

7.2 Dopamine and Compulsive Behaviors

The dopaminergic system has also been implicated in canine compulsive disorders (stereotypic behaviors) such as tail chasing, flank sucking, pacing, and shadow chasing. Dopamine receptor antagonists have been investigated as potential treatments. A randomized placebo‑controlled crossover study at Tufts University examined the effects of a dopamine receptor antagonist on dogs exhibiting ritualized, stereotypic pacing or digging.

One possible mechanism is that compulsive behaviors may reflect sensitized dopamine circuits, where the behavior itself becomes a source of dopamine release, creating a self‑reinforcing loop. This could explain why compulsive behaviors are notoriously difficult to treat with behavioral modification alone – they may be driven, at least in part, by neurochemical dysregulation.

7.3 Dopamine and Impulsivity

The relationship between dopamine and impulsivity is complex. Some studies suggest that high baseline dopamine release in the nucleus accumbens is associated with increased impulsive choice, while others find that low dopamine function is also associated with impulsivity. This may reflect a U‑shaped relationship where both too much and too little dopamine can impair impulse control.

For more on impulse control and its neurobiological basis, see The Neurobiology of Frustration in Dogs and Reactivity in Dogs – A Neurological Perspective.

8. Dopamine and the Reward Cascade: A Simplified Model

Dopamine does not act in isolation. It is part of a complex, bidirectional network involving multiple neurotransmitters. One popular didactic model is the reward cascade.

8.1 The Simplified Cascade Model

According to this simplified representation:

1. Serotonin is released from hypothalamic neurons.

2. Serotonin triggers the release of met‑enkephalin (an opioid peptide) in the VTA.

3. Met‑enkephalin inhibits GABA‑ergic neurons that normally suppress dopamine neurons.

4. This disinhibition allows dopamine release in the nucleus accumbens and prefrontal cortex, contributing to reinforcement and a sense of well‑being.

8.2 Important Caveats

This model is a simplified representation of a far more complex and bidirectional system. Modern neurobiological models emphasize:

• Extensive feedback loops (dopamine also influences serotonin release).

• Multiple other neurotransmitters involved (glutamate, endocannabinoids, etc.).

• Region‑specific and receptor‑subtype‑specific effects.

• Significant species differences.

Nevertheless, the cascade model usefully illustrates that a disruption anywhere in the system – for example, low serotonin due to chronic stress – may reduce dopamine release, potentially impairing reward processing, motivation, and learning. This may explain why SSRIs (which increase serotonin availability) are sometimes effective for impulsivity and aggression even when the primary problem appears to be dopaminergic.

9. Clinical Implications: Using Dopamine Understanding in Training and Behavior Modification

Understanding the dopamine‑based framework of learning translates directly into practical strategies.

9.1 Assess Individual Reward Value

Not all rewards are created equal. Use simple choice tests (e.g., Y‑maze preferences, treat vs. praise tests) to determine what your dog truly finds rewarding. Individual variation in caudate response to different reward types suggests that preference is biologically grounded.

9.2 Build Anticipation

Create anticipation through variable timing, cue‑reward pairings, and predictable structures that the dog can learn to expect. Anticipation, not just consumption, appears to drive dopamine release and motivation.

9.3 Use Variable Reinforcement Strategically

• Initial learning phase – Use continuous reinforcement (reward every correct response) to establish clear associations.

• Maintenance phase – Introduce variable reinforcement (random ratio or random interval) to maintain dopamine responsivity.

• Caution – Avoid partial rewarding in naïve, anxious, or inexperienced dogs, as it may induce frustration and pessimistic affective states (Cimarelli et al., 2021).

9.4 Incorporate Surprise and Novelty

Use “jackpot” rewards (unexpected high‑value treats or extra rewards) to generate strong positive prediction errors. Introduce novel elements to training sessions to potentially drive novelty‑related dopamine release.

9.5 Recognize When Dopamine Dysfunction May Be Contributing

If a dog consistently shows:

• Poor motivation despite apparently high‑value rewards

• Extreme impulsivity or hyperactivity

• Compulsive, repetitive behaviors

• Difficulty learning from reinforcement

… then dopamine dysregulation may be a contributing factor. A veterinary behaviorist can assess whether dopaminergic medications (e.g., selegiline, fluoxetine) might be appropriate to restore neurochemical balance and enable learning.

9.6 Address Chronic Stress First

Chronic stress elevates cortisol, which may reduce dopamine synthesis and release. Before expecting dopamine‑driven learning to be effective, the dog’s baseline stress should be managed. A dog whose cortisol is chronically elevated may not experience normal reward processing.

For a comprehensive look at chronic stress and its effects, see Neurobiology of Chronic Stress in Dogs – Cortisol Impact.

10. Limitations of Current Dopamine Research in Dogs

Any scientifically responsible discussion of dopamine in dogs must acknowledge the significant limitations of the current evidence base.

10.1 Limited Direct Neurochemical Measurement

Most canine dopamine research relies on:

• fMRI – measures BOLD signal (blood oxygenation), a correlate of neural activity, not direct dopamine release.

• Serum/plasma dopamine – peripheral levels do not necessarily reflect brain synaptic dopamine.

• Behavioral inference – observing that a dog works for a reward does not prove a specific dopamine mechanism.

10.2 Reliance on fMRI Proxies

fMRI is an excellent tool for localizing brain activity, but it has well‑known limitations: it is indirect, has low temporal resolution relative to dopamine firing, and BOLD signals can be influenced by non‑dopaminergic processes. Conclusions about dopamine based on fMRI are therefore inferences, not direct measurements.

10.3 Cross‑Species Inference

Much of what we “know” about dopamine and learning comes from rodent and primate studies using invasive techniques (optogenetics, direct microdialysis, electrophysiology). While the mammalian dopamine system is broadly conserved, there are important species differences. Extrapolating from rodents to dogs – a species with vastly different social ecology and cognitive demands – requires caution.

10.4 Small Sample Sizes

Canine fMRI studies typically involve small numbers of highly trained, cooperative dogs (often 10–20 subjects). These samples may not represent the broader population, particularly behaviorally challenged dogs.

10.5 Lack of Causal Evidence in Dogs

No study has directly manipulated dopamine in dogs (e.g., via receptor antagonists or optogenetics) and measured the causal effects on learning. All evidence is correlational or inferred from human/rodent research.

Takeaway: The dopamine model presented in this article is the best available framework for understanding reward‑based learning, but it should be viewed as a working model with significant gaps.

11. Summary Table: Dopamine in Canine Learning (CMS‑ready bullet structure)

Reward Prediction Error (RPE)

• Signal: Difference between expected and actual reward

• Proposed effect: Positive RPE → dopamine increase → reinforcement; Negative RPE → dopamine decrease → extinction

• Training implication: Use variable reinforcement and surprise rewards

• Evidence strength: Strong (rodent, primate, human; canine fMRI consistent)

Action Prediction Error (APE)

• Signal: Difference between predicted and actual action

• Proposed effect: Value‑free teaching signal; may reinforce repetition and habit formation

• Training implication: Currently speculative; not yet validated in dogs

• Evidence strength: Emerging (primarily rodent models)

Dopamine Pathways

• VTA → ventral striatum: Value‑based learning (RPE)

• SNc → dorsal striatum: Motor learning and habit formation

• Training implication: Different training contexts may engage different pathways

• Evidence strength: Inferred from rodent/primate; canine fMRI consistent

Dopamine Dysfunction (Associations)

• Low dopamine (ADHD‑like): Associated with impulsivity, inattention, hyperactivity

• High/sensitized dopamine (compulsive disorders): Associated with repetitive behaviors

• Training implication: May require veterinary assessment and medication

• Evidence strength: Correlational; causality not established

Reward Types

• Food: Primary reinforcer; activates caudate (fMRI)

• Praise: Social reinforcer; activates caudate equally or more in some dogs

• Training implication: Individual assessment of reward value is essential

• Evidence strength: Canine fMRI (Cook et al., 2016)

12. Key Insights (Takeaways)

• Dopamine is primarily about motivation and anticipation, not pleasure – It modulates the “wanting” system, not simply the “liking” system.

• Reward prediction error is the leading model of dopamine‑driven learning – Positive RPEs (better‑than‑expected rewards) are associated with reinforcement.

• Variable reinforcement likely maintains dopamine responsivity – Predictable rewards flatten the dopamine response and may reduce motivation.

• There is emerging evidence for a second dopamine system (APE) – However, this is primarily derived from rodent models and not yet validated in canine neuroscience.

• Individual differences in dopamine‑related reward sensitivity matter – Some dogs find praise more rewarding than food; reward value should be assessed individually.

• Dopamine dysfunction is associated with several problem behaviors – ADHD‑like behavior and compulsive disorders show correlations with dopamine levels, but causality is not established.

• Dopamine does not act alone – It is part of a complex, bidirectional network (the reward cascade, greatly simplified here).

• Training should work with dopamine, not against it – Reward‑based methods align with natural dopamine function; aversive methods likely impair it.

• Current research has significant limitations – Most evidence is correlational, fMRI‑based, or inferred from other species.

13. Conclusion

Dopamine is a central player in the neurochemistry of learning. It modulates motivation, encodes reward prediction errors, and reinforces the behaviors that lead to desirable outcomes. Understanding the dopamine framework transforms training from a mechanical process of “command and response” into a neurobiologically informed partnership between dog and handler.

However, it is essential to interpret this framework with scientific humility. Much of what we know about dopamine comes from rodent and primate studies; direct evidence in dogs is limited, largely correlational, and reliant on indirect measures like fMRI. Concepts such as APE and the reward cascade are useful didactic models but should not be mistaken for established canine neuroscience.

When a dog enthusiastically offers behaviors, seeks out training sessions, and learns quickly, dopamine signaling is likely involved. When a dog loses motivation, struggles to learn, or develops compulsive behaviors, dopamine dysregulation may be a contributing factor – but it is rarely the sole cause. By using variable reinforcement, surprise rewards, individually tailored reinforcers – and by addressing chronic stress and potential neurochemical imbalances – we can train with the brain, not against it.

The most effective training does not suppress behavior. It works with the dopamine systems that make learning intrinsically rewarding.

References

Berns, G. S., Brooks, A. M., & Spivak, M. (2012). Functional MRI in awake unrestrained dogs. PLoS ONE, 7(5), e38027.

Berns, G. S., Brooks, A. M., & Spivak, M. (2015). Scent of the familiar: An fMRI study of canine brain responses to familiar and unfamiliar human and dog odors. Behavioural Processes, 110, 37–46.

Cimarelli, G., et al. (2021). Partial rewarding during clicker training does not improve naïve dogs‘ learning speed and induces a pessimistic‑like affective state. Animal Cognition, 24(1), 107–119.

Cook, P. F., Prichard, A., Spivak, M., & Berns, G. S. (2016). Awake canine fMRI predicts dogs’ preference for praise vs food. Social Cognitive and Affective Neuroscience, 11(12), 1853–1862.

González‑Martínez, Á., Muñiz de Miguel, S., Graña, N., Costas, X., & Diéguez, F. J. (2023). Serotonin and dopamine blood levels in ADHD‑like dogs. Animals, 13(6), 1037.

González‑Martínez, Á., Muñiz de Miguel, S., & Diéguez, F. J. (2024). New advances in attention‑deficit/hyperactivity disorder‑like dogs. Animals, 14(14), 2067.

Head, E., et al. (1996). The effects of 1‑deprenyl on spatial, short term memory in young and aged dogs. Progress in Neuro‑Psychopharmacology and Biological Psychiatry, 20(3), 515–530.

The role of neurotransmitters and the reward cascade in companion animal behaviour. (2023). Veterinary Practice. [Online resource, no page numbers/DOI available]