Finish reading this sentence, then look away from your screen.

In that single moment — looking up and away — three things happened that you didn't consciously orchestrate. Your eyes made a saccade upward. They diverged, widening their angle as your gaze moved from near screen to far wall. Your accommodation relaxed, releasing the focusing tension that close work demands. Three distinct neural systems, three distinct computations, executed in roughly half a second without a single conscious instruction beyond I'm done reading for a moment.

This is what the visual system actually does. Not "eye tracking." Not a single skill. A continuous, layered orchestration of specialized systems that spend every waking moment working in concert — and which, when we understand them individually, tell us something precise and clinically meaningful about what is going wrong when they don't.

What Reading Actually Requires

When a parent brings a child in because an occupational therapist recommended a "tracking evaluation," what they are describing without knowing is one of the most neurologically complex sustained tasks a human performs.

Reading begins before the eyes touch the first word. To bring a page into focus, the visual system must do two things simultaneously and hold them: the eyes must converge, rotating inward to align on the near surface of the page, and they must accommodate, adjusting the lens of each eye to bring that distance into sharp focus. Both systems must then maintain that state continuously — not just achieve it momentarily, but hold convergence and accommodation stable across the entire reading session. This is active, ongoing work.

Then the reading itself begins. The eyes do not glide smoothly across a line of text. They make saccades, rapid, ballistic jumps from one landing point to the next, typically spanning several characters at a time. Between each saccade is a fixation: the eye holds still, and only during that stillness does the brain actually gather and process the visual information on the page. Reading speed is not determined by how fast the eyes move. It is determined largely by how efficiently the fixations are placed and how quickly the saccades carry the eyes to the next useful landing point.

When a word doesn't process, because it's unfamiliar, or fixation landed awkwardly, or attention wavered, the eyes make a regression, a backward saccade to revisit it. Then they return forward and continue. At the end of each line, a long saccade sweeps left and slightly down to find the beginning of the next one. This entire sequence, converging, accommodating, saccading, fixating, occasionally regressing, returning, repeats hundreds of times per page, with every system dependent on every other one performing reliably.

When someone says a child "has trouble with tracking," this is what they are gesturing at. The question is which part.

What Watching Baseball Actually Requires

Sport makes the same point more vividly because the visual demands are less linear and more obviously dynamic.

Standing at the plate or in the outfield, a player is running all four systems simultaneously. Smooth pursuit tracks the pitched or batted ball in flight — the eyes matching the ball's velocity continuously to keep it stable on the fovea. But the binocular system is also triangulating that ball's position in three-dimensional space in real time, using vergence and retinal disparity to judge depth and trajectory well enough to swing at the right moment or position under it.

Meanwhile, the player is also saccading around the field — checking base positions, reading fielder positions, locating the cutoff man. Each of those saccades lands on a target at a different distance, which means each one is accompanied by a vergence shift: converging slightly to see a closer fielder, diverging to read the position of someone deeper in the outfield. And between each of those saccades, fixation must be stable enough to actually extract useful information before the eyes move again.

None of this is experienced as effort. It happens automatically, fluidly, within the normal visual behavior of a reasonably skilled player. But the neural infrastructure supporting it is vast — and every component of it is doing distinct work.

Four Systems, Four Neural Architectures

Understanding why these systems can fail independently requires understanding that they are not just functionally distinct. They are neurologically distinct. Different circuitry, different cortical drivers, different developmental timelines, and, critically, a different relationship between the cortex and the brainstem in each case.

Here is what almost always gets left out of discussions of eye movements: with one significant exception, the eye movement systems are volitionally driven; the exception is the pursuit system. The brainstem executes the movement. But the eyes only move because the cortex decided to move them — to something, for some reason, in service of some goal. The brainstem is a remarkably precise executor. It is not the decision-maker.

Saccades

The saccadic system moves the eyes rapidly to a new target — correcting what is called position error, the difference between where the eyes are pointing and where the brain wants them to point. The motor execution of this movement is handled by the brainstem: the paramedian pontine reticular formation (PPRF) for horizontal movements, the rostral interstitial nucleus of the medial longitudinal fasciculus (riMLF) for vertical ones. Burst neurons in these structures fire at high frequency to drive the movement; omnipause neurons hold them in check during fixation and release them when a saccade is triggered.

But the saccade only happens because the cortex sent the command. The frontal eye fields in prefrontal cortex and the posterior parietal cortex determine where attention should go, translate that into a motor vector, and drive the superior colliculus — a midbrain structure containing a topographic map of visual space — which then activates the brainstem premotor circuitry. This cortical-to-collicular-to-brainstem pathway is the architecture of every intentional gaze shift. The brainstem moves the eyes. The cortex decided where.

This distinction matters clinically because saccadic motor output — the velocity and accuracy of the movement itself — reaches adult-like performance by 4 to 6 months of age. Brainstem circuitry matures early. But the voluntary control of saccades, the ability to direct them efficiently, suppress unwanted ones, and generate them in service of a reading or scanning task, is governed by prefrontal cortex that continues maturing until approximately age 14. A child can have perfectly fast saccades and still be a poor, effortful reader because the cortical direction of those saccades hasn't matured.

Fixation

Fixation is the active decision to stay. After a saccade lands the eyes on a target, the cortex must now suppress the next movement — holding gaze steady long enough to gather information despite whatever else is happening in the visual field. This requires the frontal eye fields to maintain active inhibitory control over the saccade-generating system, tonic activity from fixation neurons in the rostral superior colliculus to suppress orienting responses, and neural integrators in the brainstem to hold eye position against the elastic forces that would otherwise pull the eyes back toward center.

The cortical demand here is essentially an attention demand. Fixation stability and attentional stability are neurologically intertwined — both depend on prefrontal systems that regulate distraction and maintain focus on a chosen target. I evaluated a child that experienced a recent concussion and she demonstrated profoundly unstable fixation during testing. Her mother's description of her at home — scattered, seemingly unable to hold attention on any one thing, easily pulled away by peripheral activity — was not a behavioral observation separate from the visual findings. It was a description of the same underlying difficulty expressed in two different contexts.

Stillness, in other words, is not the absence of a command. It is an active cortical instruction: not yet, stay here, this is where we are.

Vergence

Vergence aligns the two eyes in depth, converging them for near targets and diverging them for far ones. Unlike saccades and pursuit, vergence is disjunctive — the eyes rotate in opposite directions — and it operates on a fundamentally different error signal: retinal disparity, the slight difference between the image each eye receives from the same object at a given depth. The premotor circuitry sits in the midbrain near the oculomotor nucleus, distinct from the conjugate eye movement generators.

The retinal disparity signal is what allows the visual system to land precisely on a chosen target — to park within the lines, so to speak. But the decision of where to park is entirely cortical. Attention directed to a particular object at a particular depth is what triggers the vergence response. The disparity signal then fine-tunes the landing. You choose the parking spot out of intention; retinal disparity is the guidance system that positions you accurately once you've committed.

Vergence dysfunction — insufficiency, excess, or infacility — disrupts this process at the landing stage. The cortical intention is intact; the execution is unreliable. The result is visual fatigue, headache, intermittent blur, and avoidance of close work that is frequently mistaken for attention difficulties, reading disinterest, or learning disability.

Smooth Pursuit

Smooth pursuit is the exception to everything stated above. It is the only eye movement system that cannot be generated voluntarily. The cortex cannot will smooth pursuit into existence — it can only sustain and guide a pursuit response once a moving target provides the initiating stimulus. Without something moving to follow, the system has nothing to work with. Attempts to move the eyes smoothly without a target produce saccades instead.

The pursuit pathway runs from motion-processing cortex — specifically areas MT and V5, which encode target velocity and retinal slip — through the dorsolateral pontine nuclei to the cerebellar flocculus and vermis, which calibrate the response and adapt it based on ongoing error. Cerebellar outputs then drive the brainstem premotor circuitry that produces the smooth conjugate movement.

Because pursuit is calibrated through the cerebellum rather than through prefrontal systems, it is exquisitely sensitive to cerebellar dysfunction, certain medications, and developmental immaturity. It is also the system that matures most slowly — smooth pursuit continues refining through late adolescence, with the ability to track higher-velocity targets and maintain consistent pursuit gain developing well into the teenage years.

When pursuit is impaired, the system compensates with catch-up saccades — brief corrective jumps that reduce the position error that inadequate pursuit has allowed to accumulate. Observed without this framework, those catch-up saccades can appear to be a saccadic problem. The root cause is in the pursuit system entirely.

How They Work Together — And Why That Makes Assessment Hard

The reason "eye tracking" persists as a catchall term is that in real visual behavior, these systems operate in such seamless coordination that their individual contributions are invisible. Reading feels like one continuous act. Watching a ball in flight feels like one continuous act. Looking up from a screen feels like one continuous act.

But the seamlessness is the product of four systems that have learned to coordinate precisely — and when one of them underperforms, the others compensate. The child with poor pursuit gain generates catch-up saccades. The child with vergence insufficiency increases accommodation effort. The child with unstable fixation shortens dwell time and increases saccade frequency to avoid the discomfort of holding still. The system is remarkably good at hiding its own failures, which is exactly why assessment requires more than watching someone's eyes move and noting that something looks off.

A developmental optometric evaluation is designed to disaggregate what seamless visual behavior has assembled — to test each system individually, under conditions specific enough to reveal its true performance, and then to understand how deficits in one system are loading the others. The question is never simply whether a child has a tracking problem. It is which system is underperforming, at what level of the neural architecture, with what developmental expectations for this age, and with what consequences for the tasks that matter to this child's life.

The Cortex Decides. The Brainstem Delivers.

The organizing principle underlying all of this — the one that is always missing from discussions of eye movements — is that the brainstem moves the eyes, but the eyes only move because the cortex wanted them to move somewhere, for some reason, in service of some goal. The brainstem is extraordinarily precise. It is not in charge.

Smooth pursuit is the exception, and it is worth understanding why. Pursuit is the only system without a volitional origin because its job is fundamentally reactive — to match something that is already happening in the world. Every other system begins with a cortical intention: I want to look there, I want to stay here, I want to align on that object at that depth. The brainstem executes. The cortex decides.

This reframes what poor eye movement performance actually means in a child. It is rarely a brainstem problem. Saccadic motor output is adult-like before a child's first birthday. What develops slowly — extending through early adolescence and beyond — is the cortical architecture that directs, controls, and coordinates that output. Prefrontal maturation, attentional development, the refinement of inhibitory control. These are the systems that turn functional brainstem machinery into skilled, effortful, reliable visual behavior.

Which is why "do some tracking exercises" is sometimes the right answer, and sometimes barely scratches the surface.