TL;DR
Our perception of a stable visual world is remarkable, given how much our eyes move. This study uses intracranial EEG in humans to show that while early visual areas are highly sensitive to visual shifts caused by eye movements (saccades), higher-level visual areas actively suppress these signals. This suggests a hierarchical mechanism for maintaining visual stability, with higher brain regions acting as gatekeepers against self-generated visual noise.
What Problem Does This Paper Address?
We experience a remarkably stable visual world, even though our eyes are constantly making rapid movements called saccades. These saccades cause the image on our retina to shift dramatically. To maintain this sense of stability, our brains must somehow compensate for these self-generated visual disturbances. Previous work has shown that the brain suppresses signals related to eye blinks in higher visual areas but not in early visual cortex. This study investigates whether a similar suppression mechanism exists for saccades, and crucially, how this suppression unfolds across the visual hierarchy, from early visual areas to more complex, higher-level processing regions. The key question is: where does the brain start to distinguish between visual input caused by external events versus self-generated eye movements?
Methods
The researchers utilized invasive intracranial EEG (iEEG) recordings from 12 neurosurgical patients who were undergoing monitoring for clinical reasons. This technique provides high spatial and temporal resolution of brain activity. During the experiment, patients viewed still images of faces and other objects. The core of the study involved comparing the brain's response to two types of visual events:
- External Stimulus Displacements: The image on the screen was shifted by a small amount (1.3° or 3.7° visual angle).
- Saccades: Patients made small, natural eye movements (1.3° to 3.7° visual angle) while viewing the images.
Crucially, eye movements were precisely tracked using an eye-tracker synchronized with the iEEG recordings. This allowed the researchers to identify saccades that occurred during image viewing. They then analyzed the high-frequency broadband (HFB) activity, a proxy for neuronal firing rate, in various brain regions. A Finite Impulse Response (FIR) General Linear Model (GLM) was used to deconvolve the overlapping neural responses to these events and estimate their unique contributions. Statistical comparisons were made between the HFB responses to external displacements and saccades across different visual regions of interest (ROIs), including early visual areas (V1-V3), V4, VO, and higher-level face-selective and non-face-selective areas.
Datasets
The study used data from 12 neurosurgical patients. The visual stimuli consisted of still images of faces and non-face control stimuli. The study also references and builds upon data from a previous study by Golan et al. (2016) concerning blink-related transients. No new datasets were introduced.
Key Insights and Ideas
- Hierarchical Suppression: The most striking finding is that the suppression of saccade-related visual signals is not uniform across the brain. Early visual areas (V1-V3) show similar responses to both external visual shifts and saccade-induced shifts, indicating they are highly sensitive to retinal input. However, as you move up the visual hierarchy to areas like V4, VO, and higher-level processing regions (both dorsal and ventral streams), the response to saccades is significantly suppressed, while the response to external displacements remains strong.
- Extra-retinal Influence: This hierarchical suppression strongly suggests an "extra-retinal" mechanism. This means the brain is not just passively reacting to the visual input on the retina but is actively using information about the intended eye movement (an "efference copy") to dampen the processing of self-generated visual noise.
- Shared Suppression Mechanism: The study found a significant overlap between the brain regions that suppress saccade-related transients and those that suppress blink-related transients (from a previous study). This suggests that the brain might employ a shared pathway or mechanism for suppressing visual disruptions caused by different types of oculomotor activity.
- Ventral Stream Involvement: Unlike previous studies that focused primarily on dorsal visual areas (like MT/MST), this work demonstrates that saccade suppression also occurs in ventral high-level visual areas, including face-selective regions. This highlights the broad impact of oculomotor feedback on visual processing.
Key Results
- In early visual cortex (V1-V3), both external stimulus displacements and saccades elicited similar positive HFB activity increases.
- In higher-level visual areas (V4, VO, face-selective, and non-face-selective), external displacements triggered significant HFB activity increases, while saccades elicited almost no HFB response modulation (i.e., suppression).
- Statistical analysis revealed that significantly more electrodes showed greater HFB responses to external displacements compared to saccades in regions anterior to V3, with the strongest differences observed in higher-level areas.
- No electrodes showed significantly greater HFB increases for saccades compared to external displacements.
- The latency of peak HFB responses to external displacements and saccades was similar in early visual sites, suggesting a retinal origin for these initial responses.
Limitations and Future Work
- Saccade Magnitude: The study focused on relatively small saccades (1.3°-3.7°). Larger saccades can bring new objects into view, potentially leading to different neural responses. Future work could explore how these suppression mechanisms operate for larger saccades.
- Retinal vs. Extra-retinal Contributions in Early Visual Cortex: While the study provides strong evidence for extra-retinal suppression in higher areas, it acknowledges that weak reductions in HFB activity in some early visual sites might be partly explained by retinal factors (e.g., momentary blurring during saccades) rather than solely extra-retinal signals. Further studies are needed to disentangle these contributions.
- Specific Mechanism of Suppression: The precise mechanism by which saccade-related transients are suppressed in the ventral stream remains an open question. It could be a general gating mechanism or a more specific offsetting process. Future experiments, perhaps involving controlled visual image displacement concurrent with saccades, are suggested to clarify this.
- HFB as a Proxy: The study relies on HFB activity as a proxy for neuronal firing. While a good correlate, it's acknowledged that single-unit activity or slower iEEG frequencies might reveal different patterns.
Summary updated