Introduction: The Neural Basis of Steady Gaze

To maintain stable vision, the eyes must remain fixed on an object even as the head moves. This ability to hold gaze after a saccade depends on a neural circuit known as the neural integrator. When this system fails, the eyes drift slowly back towards the center, followed by a corrective quick phase. This rhythmic oscillation is called gaze-evoked nystagmus (GEN). Clinical models and computational frameworks have been essential in understanding the pathophysiology behind GEN, particularly its relationship with cerebellar and brainstem networks.

Classical Models of the Neural Integrator

The most widely accepted model for gaze holding is the "lateral integrator" hypothesis. In this model, the brainstem neural integrator — specifically the nucleus prepositus hypoglossi (NPH) and the medial vestibular nucleus (MVN) in the horizontal plane — performs a mathematical integration of the velocity command from the saccadic system. A leaky integrator model explains GEN, where the time constant of the integrator is insufficient, causing the eye to drift centrifugally.

The "Velocity-to-Position" Transformation

This model posits that the eye position command (the step) is derived from the velocity command (the pulse). In healthy individuals, the pulse-step match is perfect. In GEN, the step is mismatched, causing a centripetal drift. Computational models using linear systems theory have shown that a reduction in the integrator's time constant from approximately 25 seconds to just 2-3 seconds produces the classic slow-phase drift seen in GEN.

Rebound Nystagmus: A Model of Adaptive Plasticity

Rebound nystagmus is a transient nystagmus that appears when the patient returns to center gaze after sustaining an eccentric position. It is a hallmark of cerebellar involvement, particularly the flocculus and paraflocculus. The model for rebound nystagmus involves the concept of "velocity storage" and adaptive changes in the neural integrator.

During sustained eccentric gaze, the cerebellar flocculus learns to counter the drift by adding a bias (a prolonged velocity command) to the integrator. When the gaze returns to center, this bias persists momentarily, creating a slow phase in the opposite direction of the prior eccentric gaze — thus the nystagmus "rebounds." This model demonstrates how the cerebellum continuously calibrates the integrator. In multiple system atrophy or spinocerebellar ataxia, this rebound is often exaggerated or absent, depending on the lesion site.

Mathematical and Computational Models

Detailed mathematical models of gaze-evoked and rebound nystagmus have been developed using dynamical systems analysis. These models incorporate:

• Leaky integrators with variable time constants.
• Cerebellar learning rules based on Bekesy's model of adaptive filter banks.
• Positive feedback loops in the brainstem (common in models of the oculomotor integrator).

A key finding from these models is that the direction and amplitude of rebound nystagmus are predictable from the duration and eccentricity of prior gaze holding. This has been validated in clinical studies using video-oculography, confirming that rebound is a manifestation of short-term adaptive changes in the gaze-holding network.

Clinical Models: Localization and Pathophysiology

From a clinical perspective, the presence of GEN without rebound suggests a brainstem lesion (e.g., NPH infarction), while GEN with strong rebound suggests cerebellar disease (e.g., floccular atrophy). A third model involves the "gaze-paretic" nystagmus seen in myasthenia gravis, where the problem is not central but peripheral — a failure of the muscle to sustain contraction, mimicking a leaky integrator at the neuromuscular junction. This highlights that the final common pathway for gaze holding includes both neural and muscular components.

Conclusion: Integrating the Models

Understanding the models of gaze-evoked nystagmus—from classical integrator theory to modern computational approaches—provides clinicians and researchers with a powerful framework for localizing neurological lesions. The interplay between gaze, nystagmus, and rebound phenomenon reveals the elegant plasticity of the oculomotor system. As neuroimaging and modelling techniques advance, these models will continue to refine our understanding of vertigo, oscillopsia, and brainstem-cerebellar connectivity.

Future directions include patient-specific computational models that could predict response to therapy, and the use of machine learning to classify nystagmus types in real time.