TL;DR
This study investigates how neurons in the MSTd brain region compensate for the visual distortions caused by eye movements during self-motion. They found that MSTd neurons adjust their responses more at faster pursuit speeds, suggesting a sophisticated mechanism for maintaining accurate heading perception regardless of how quickly we move our eyes. This compensation is driven by both visual cues and extra-retinal signals, highlighting MSTd's crucial role in egomotion processing.
What Problem Does This Paper Address?
When we move through our environment, our eyes are rarely still. Smooth pursuit eye movements, common during walking or driving, cause the visual scene on our retinas to shift. Specifically, the "focus of expansion" (FOE) – a key visual cue for determining our direction of self-motion (heading) – gets displaced from its true location. This displacement becomes larger with faster eye movements. Despite these visual distortions, humans can still accurately perceive their heading. This raises a fundamental question: how does the brain achieve this accurate perception? Previous research suggested that the MSTd area of the brain might be involved in this "pursuit compensation." However, it was unclear if this compensation mechanism scaled with the speed of pursuit, meaning if it adjusted more for faster eye movements that cause larger visual shifts. This paper aims to answer that question by examining how MSTd neurons' responses change with varying pursuit speeds.
Methods
The researchers recorded from 40 neurons in the MSTd area of two macaque monkeys. They employed a multi-stage experimental design:
- Preferred Optic-Flow and Pursuit Direction: First, they characterized each neuron's preferred response to different types of visual motion (optic flow patterns like expansion, contraction, rotation) and preferred direction of smooth pursuit eye movements at various speeds.
- Pursuit Compensation Experiment: This was the core of the study. Monkeys performed tasks under three conditions:
- Fixed Gaze: The monkey fixated a stationary target, and optic flow stimuli were presented.
- Real Pursuit: The monkey pursued a moving target at different speeds (2.58, 5.05, and 9.22 deg/s), while optic flow stimuli were presented. This condition involved both visual and extra-retinal (eye movement-related) signals.
- Simulated Pursuit: The monkey fixated a stationary target, but the entire optic flow display was rotated counter to the pursuit direction at the same speeds as real pursuit. This condition isolated the effect of visual cues by mimicking the retinal image of real pursuit without the extra-retinal signals.
For each condition and each neuron, they measured neural responses to optic flow stimuli with varying FOE positions. They then used cross-correlation analysis to quantify how much each neuron's tuning curve shifted to compensate for the FOE displacement caused by pursuit. This allowed them to calculate a "compensation index" (CI) representing the degree of compensation relative to perfect compensation.
Datasets
Not discussed in the paper. The study used simulated visual stimuli generated by the researchers.
Key Insights and Ideas
- Speed-Dependent Compensation: The most striking finding is that MSTd neurons don't just compensate for pursuit-induced visual shifts; they do so more at faster pursuit speeds. This is crucial because faster eye movements create larger visual distortions.
- Retinal Cues Alone Contribute: Even when only visual information was available (simulated pursuit), MSTd neurons showed a significant level of compensation, indicating that the visual scene itself provides cues for this adjustment.
- Extra-retinal Cues Enhance Compensation: The compensation was significantly greater during real pursuit (when both visual and extra-retinal signals were present) compared to simulated pursuit. This highlights the integration of different sensory inputs for robust egomotion perception.
- Antialignment for Integration: A systematic "antialignment" was observed between a neuron's preferred pursuit direction and its preferred visual motion direction. This suggests a neural architecture designed to combine these two types of signals for effective compensation.
Key Results
- Population-Level Compensation: Across the population of MSTd neurons, the median compensation rate increased with pursuit speed. The median compensation index (CI) for real pursuit was 50.7%, meaning neurons compensated for about half of the required shift. For simulated pursuit, the median CI was 29.6%.
- Speed Scaling: The increase in compensation with pursuit speed was significant. The median rate of compensation increase with pursuit speed was 51% of that required for perfect compensation in real pursuit.
- Retinal vs. Extra-retinal Contribution: Retinal cues alone contributed about 30% of the compensation, while the combination of retinal and extra-retinal cues led to a significantly higher compensation (difference P < 0.05).
- Preferred Direction Alignment: A significant peak in the distribution of angular differences between preferred pursuit and preferred laminar motion directions was observed around 180 degrees, indicating a tendency for these preferences to be opposite.
Limitations and Future Work
- Display Size: The researchers note that their relatively small display size (20x20°) might have contributed to the observed compensation, as larger displays have shown less effective neural compensation in previous studies. They speculate that the smaller display might make the stimulus boundary motion more salient.
- Stimulus Complexity: While the study focused on expansion patterns, it acknowledges that other optic flow patterns (contraction, rotation) were also tested and showed similar compensation trends. However, the detailed analysis was primarily on expansion.
- Future Work: The authors suggest further investigation into how neurons integrate retinal and extra-retinal cues, and how increased neural activity with pursuit speed might drive these compensatory shifts. They also propose exploring the role of stimulus size and speed regimes in shaping these responses, drawing parallels to previous work on the antialignment of visual and pursuit preferences.
Summary updated