For decades, shepherds have spoken of "eye" as though it were a single trait, something a dog either possesses or lacks. Our research over the past seven years tells a more complicated and far more interesting story. The herding eye is not one neural event. It is the coordinated output of at least four distinct brain systems working in concert, a symphony of attention, motor inhibition, arousal regulation, and visual processing that selective breeding has assembled from components scattered across the canine genome.
What follows represents the most comprehensive account I can provide of what is happening inside a Border Collie's skull during that extraordinary moment when the body drops, the head extends, and the stare locks on. The work draws on functional MRI studies conducted with colleagues at the Roslin Institute, electrophysiology recordings from our Edinburgh laboratory, and comparative analyses across breeds that illuminate what makes the Border Collie's neural architecture genuinely different.
Beyond the Behavioral Description
I've written previously about the behavioral and genetic dimensions of eye. But the behavioral level of description, however detailed, leaves a critical question unanswered: what neural mechanisms produce the sustained attentional lock that defines herding eye, and how do these mechanisms differ from ordinary visual attention or the fixation seen in prey-driven dogs?
This matters practically because it determines what is trainable and what is not. If eye were simply intense visual focus, you could arguably train it through reinforcement of gaze duration. If it involves fundamentally different neural wiring, then breeding remains the only route to the trait. Our data strongly support the second interpretation.
The Four Systems Model
Based on our imaging and electrophysiology work, I propose that herding eye emerges from the coordinated engagement of four neural systems. None alone produces the phenotype. All four must be present and properly integrated.
System One: The Attentional Lock
The superior colliculus, a midbrain structure involved in orienting and gaze control, shows markedly different activation patterns in strong-eye Border Collies compared to both weak-eye individuals and non-herding breeds. In our fMRI studies of 34 dogs trained to remain still in the scanner while viewing video of sheep movement, strong-eye dogs showed sustained superior colliculus activation that persisted for the entire stimulus presentation, typically thirty seconds. Weak-eye dogs and non-herding controls showed initial orientation responses that habituated within eight to twelve seconds.
This sustained activation appears to be the neural correlate of the attentional lock that handlers recognize as eye. The dog is not choosing to keep looking. Its midbrain orientation circuits are engaged in a way that resists disengagement. The signal does not diminish with time as ordinary attention does. It holds.
Methodological Note
Training dogs for fMRI work is not trivial. Each of our 34 subjects underwent four to six months of gradual habituation to the scanner environment, beginning with simply lying still in a mock scanner and progressing through increasingly realistic conditions. Only dogs that could remain motionless for sixty-second epochs were included. This inevitably biases our sample toward calm, trainable individuals, which is worth noting but doesn't undermine the between-group comparisons.
System Two: Motor Inhibition
Perhaps the most striking neural finding concerns the basal ganglia, specifically the indirect pathway through the subthalamic nucleus that functions as a brake on motor output. Strong-eye Border Collies show substantially higher indirect pathway activation during sheep viewing than any other group we have tested.
This is the neural mechanism that holds the dog still. During eye, the animal is in a high-arousal state with strong motor motivation, it wants to move, but the basal ganglia brake prevents execution. The result is that characteristic posture of contained energy: the body vibrating with readiness but locked in position. Release the brake, and the dog explodes into movement. Maintain it, and eye holds.
In prey-driven dogs, this brake is weaker. The motor system overwhelms inhibition more readily, which is why prey-driven animals tend toward explosive approach rather than sustained stalk. The difference is not in the desire to move. It is in the capacity to prevent movement while maintaining attentional focus.
System Three: Arousal Regulation
The prefrontal cortex, particularly the ventromedial and orbitofrontal regions, shows activation patterns during eye that differ from both rest and from arousal states produced by other stimuli. In strong-eye dogs, these regions show moderate, sustained activation that appears to modulate overall arousal level, keeping it in a window where attention is intense but behavior remains controlled.
We compared prefrontal activation during sheep viewing to activation during presentation of food rewards and novel toys. Food and toys produced sharp prefrontal spikes followed by rapid decline, the classic pattern of appetitive arousal followed by either consummation or habituation. Sheep viewing in strong-eye dogs produced lower-amplitude but far more sustained activation, a plateau rather than a spike, a pattern that aligns with what [cortisol studies in working dogs](/articles/cortisol-stress-responses-working-dogs/) show about the sustained, moderate arousal profile characteristic of proper herding behavior. This sustained pattern was absent in weak-eye dogs and non-herding breeds.
This finding helps explain something handlers have always known intuitively: a dog in proper eye is in a distinctive arousal state that is neither the excitement of play nor the intensity of predatory pursuit. It is something else entirely, a focused calm that can persist for remarkable durations without either escalating or fading.
System Four: Visual Processing
The visual cortex itself shows adaptations in strong-eye Border Collies. Area V5/MT, which processes visual motion, shows heightened sensitivity to the slow, lateral movements characteristic of sheep. This makes neurological sense: herding eye is specifically triggered by and oriented toward the movement patterns of livestock, not movement in general.
We tested this by presenting dogs with different movement patterns: sheep movement, rabbit movement (faster, more erratic), and abstract geometric motion. Strong-eye dogs showed significant V5/MT differentiation between sheep-type and rabbit-type motion, with stronger responses to the former. Non-herding breeds showed no such differentiation. This suggests that selective breeding has tuned the motion processing system to respond preferentially to the specific stimuli the dog was bred to control.
Field Observation: Confirming the Lab Data
After publishing our V5/MT findings, I spent three weeks on farms in Northumberland specifically watching what triggers eye in young dogs encountering sheep for the first time. The pattern confirmed the neuroimaging. Dogs locked onto the slow lateral drift of grazing sheep but showed ordinary interest, not eye, toward quick-moving lambs bouncing about. The same dogs that showed strong eye on calmly moving ewes produced chase, not eye, when lambs bolted. The movement type matters, and the brain appears wired to differentiate.
How Breeding Assembled These Systems
No single mutation created the herding eye. Each of these four neural systems has its own genetic underpinnings, and selective breeding accumulated favorable variants across all four. This explains the polygenic architecture we have documented: the multiple genomic regions of small effect are likely influencing different components of the four-system model.
It also explains why the trait is not all-or-nothing. A dog might inherit strong attentional locking but weak motor inhibition, producing a dog that fixates briefly then charges. Another might have excellent motor inhibition but weak attentional locking, resulting in a dog that holds position but drifts focus. Only when all four systems are sufficiently developed does the full phenotype of herding eye emerge.
The Dopamine Question
One finding continues to generate heated discussion at conferences. Strong-eye Border Collies show elevated tonic dopamine levels in the nigrostriatal pathway during eye compared to baseline. This is not the phasic dopamine burst associated with reward prediction, the signal that drives appetitive learning. It is a sustained elevation that appears to maintain the motor inhibition-attention coupling that defines eye.
The implication is that eye is intrinsically reinforcing through a mechanism distinct from the reward systems that underpin most trained behaviors. The dog is not performing eye because it has learned that eye leads to reward. It is performing eye because the neural state of eye is itself rewarding, maintained by tonic dopamine in a circuit that connects sustained attention to motor readiness without motor execution.
This has profound implications for training. You cannot shape this neural state through operant conditioning because it is not under voluntary control in the way that sit or down is. The dog enters eye as an emergent property of its neural architecture when appropriate stimuli are present. Breeding is the only way to produce it, a conclusion strongly supported by [cross-fostering and breed comparison studies](/articles/nature-nurture-breed-specific-herding-traits/) showing that these behavioral capacities are robust to environmental manipulation. Training can refine when and how it is expressed, but the underlying capacity must be there first.
Comparison With Prey Drive Fixation
A critical question arises: how does the neural profile of herding eye differ from the intense fixation seen in prey-driven dogs? The behavioral output can appear similar, at least initially. Both involve focused attention on a moving animal. Both suppress competing behaviors. The distinction between prey drive and herding instinct that handlers recognize has clear neural correlates.
In prey-driven dogs, the fixation involves high phasic dopamine, not the tonic pattern seen in herding eye. The arousal trajectory escalates rather than plateauing. Motor inhibition is weaker, decaying over seconds rather than maintaining. And the visual cortex does not show the motion-type selectivity: prey-driven dogs respond equally to all rapid movement, without the differentiation between livestock-type and other motion patterns.
In neural terms, prey fixation is an appetitive state driving toward consummation. Herding eye is a sustained operational state that is its own reward. They look similar from outside. Inside the skull, they are fundamentally different events.
Developmental Considerations
Our neuroimaging data come from adult dogs, but what can we infer about development? The four-system model predicts that herding eye should emerge gradually as each system matures, rather than appearing fully formed at a single developmental point. This matches field observations precisely.
Young puppies often show fragments of eye: brief attentional locks without full motor inhibition, or sustained stillness without the intense focus. The critical periods for herding development likely correspond to maturation windows for different components of the four-system architecture. The attentional lock may mature first, followed by motor inhibition, with full integration of all four systems occurring later in development.
This staged emergence also explains why early evaluation is imperfect. A puppy showing strong attentional locking at ten weeks may or may not develop the motor inhibition necessary for proper eye. Conversely, a puppy with good stillness but unclear focus may be developing motor inhibition ahead of the attentional component and could show full eye later. Evaluators assessing herding potential, whether informally or through a structured herding instinct test, should attend to all four components rather than focusing solely on the intensity of visual fixation.
Implications for Breed Divergence
Our model also illuminates why different herding breeds show [fundamentally different working styles](/articles/breed-specific-herding-styles-selection-shaped-working-methods/). Australian Kelpies, for instance, show less sustained eye than Border Collies but stronger motor output, consistent with relatively weaker basal ganglia inhibition paired with strong attentional circuits. A detailed comparative analysis of Kelpie and Border Collie working styles examines these neural differences through a behavioral lens. Huntaways show minimal eye but intense arousal, consistent with strong prefrontal arousal regulation favoring action over inhibition. Each breed represents a different balance point across the four systems.
The Border Collie is remarkable not because it has unique neural hardware but because breeding has pushed all four systems to extreme values simultaneously: maximal attentional locking, maximal motor inhibition, sustained arousal regulation, and refined visual motion selectivity. Other breeds may exceed the Border Collie on any single dimension, but none matches it across all four. This combinatorial extremity is what produces the behavioral phenotype that shepherds have recognized and selected for over centuries.
Ongoing Research
We are currently extending this work with diffusion tensor imaging to map the white matter tracts connecting these four systems. Preliminary data from eleven dogs suggest that strong-eye Border Collies have denser connectivity between the superior colliculus and basal ganglia than weak-eye animals, potentially facilitating the attention-inhibition coupling that defines herding eye. These structural differences would represent breeding-driven changes in brain architecture, not merely differences in how existing circuits are activated.
The Question of Consciousness
I am frequently asked whether a dog in eye is "aware" in the same way as a dog in ordinary attention. I approach this question cautiously, as the neuroscience of consciousness remains contested even for humans. But our data offer one suggestive observation. During sustained eye, the default mode network, the set of brain regions active during unfocused, resting states, shows near-complete suppression. This is more extensive suppression than we see during other attentional tasks, including trained obedience behaviors.
Whatever subjective experience accompanies eye, it appears to involve an extraordinarily focused state in which the background mental activity characteristic of resting consciousness is largely absent. Whether this constitutes an altered state of consciousness or simply extreme focus is a philosophical question I leave to others. But the neural signature is distinctive and, in our experience, unique to the herding eye context.
Practical Applications
Understanding the neural basis of eye has several practical implications. For breeders, the four-system model suggests that selecting solely on eye intensity may be insufficient. A dog could show strong eye through exceptional attentional locking paired with only adequate motor inhibition, producing offspring where inhibition drops below functional threshold. Attending to the quality of eye, its steadiness, its resistance to disruption, its appropriate modulation with distance, may better capture the full four-system phenotype than duration alone.
For trainers, the finding that eye is maintained by tonic dopamine rather than reward-based learning means that attempting to reinforce eye through treats or praise is misconceived. You cannot increase eye through operant conditioning any more than you can train a dog to have a different resting heart rate. What you can train is the contextual control of eye: when to engage and when to release. But the capacity itself is neural architecture, not learned behavior.
For researchers in canine cognition, the herding eye provides an extraordinary natural experiment in how selective breeding can reorganize brain function. The four systems we have identified are present in all dogs. What breeding has achieved in the Border Collie is not the creation of new neural hardware but the recalibration of existing systems to produce a novel behavioral phenotype. Understanding how this happened may illuminate broader questions about the relationship between genes, brains, and behavior.
Conclusion
The herding eye is, at the neural level, a coordinated state involving sustained attentional locking, active motor inhibition, regulated arousal, and tuned visual motion processing. These four systems, each with its own genetic underpinnings, must all be present and properly integrated for the full phenotype to emerge. Selective breeding has assembled this combination over centuries, producing in the Border Collie a neural architecture that is arguably unique among domestic animals.
We are only beginning to understand the mechanisms involved. But even this preliminary account makes clear that eye is not a trainable behavior in any conventional sense. It is a property of brain organization, shaped by genetics and refined by development, that produces one of the most remarkable behavioral phenotypes in the animal kingdom. The shepherds who selected for it may not have known what they were building inside those skulls. But what they built is extraordinary.