Abstract
Models of visual attention postulate the existence of a saliency map whose function is to guide attention and gaze to the most conspicuous regions in a visual scene. Although cortical representations of saliency have been reported, there is mounting evidence for a subcortical saliency mechanism, which pre-dates the evolution of neocortex. Here, we conduct a strong test of the saliency hypothesis by comparing the output of a well-established computational saliency model with the activation of neurons in the primate superior colliculus (SC), a midbrain structure associated with attention and gaze, while monkeys watched video of natural scenes. We find that the activity of SC superficial visual-layer neurons (SCs), specifically, is well-predicted by the model. This saliency representation is unlikely to be inherited from fronto-parietal cortices, which do not project to SCs, but may be computed in SCs and relayed to other areas via tectothalamic pathways.
Introduction
In a world overflowing with visual information, the brain must prioritize what deserves attention. Traditionally, neuroscientists have focused on the neocortex to find "saliency maps"—internal representations of what is visually conspicuous. However, an alternative hypothesis suggests that the superior colliculus (SC), an evolutionarily ancient midbrain structure, may be the original seat of this mechanism. This study aims to distinguish between bottom-up saliency (feature-driven) and top-down priority (goal-driven) by recording SC activity during the free viewing of complex, dynamic natural videos.
Methodology and Data
Researchers recorded extracellular activity from 60 single neurons (34 in the superficial layers, SCs; 26 in the intermediate layers, SCi) in three rhesus monkeys. The monkeys viewed a library of 516 high-definition video clips, including footage from the BBC’s Planet Earth.
To analyze the data, the team employed the Itti-Koch saliency model, which decomposes video into feature maps such as color, motion, and edges. These maps were transformed into a log-polar coordinate system to match the physical topography of the SC. The analysis focused on "naturally occurring fixations"—brief periods of stillness between eye movements—to correlate model predictions with neural firing rates and gaze behavior.
Key Insights: Saliency vs. Priority
The study revealed a clear functional split between the layers of the SC. The superficial layers (SCs) appear to encode a "pure" saliency map. These neurons are feature-agnostic, responding to the integrated conspicuousness of a scene regardless of the animal's behavioral goals. In contrast, the intermediate layers (SCi) act as a "priority map," where saliency coding is only present if the monkey is planning an eye movement toward that specific location.
Because the SCs does not receive direct projections from fronto-parietal "attention" areas, this saliency map is likely computed locally or inherited from early visual structures. This suggests that complex attention-guiding mechanisms exist independently of the neocortex.
Results and Behavioral Impact
The findings were highly robust, with 79% of SCs neurons showing a significant positive correlation with the saliency model. Furthermore, the model effectively predicted behavior: locations with high predicted saliency were more likely to be targeted by the monkeys' eyes. High saliency also correlated with shorter fixation durations (299 ms vs. 314 ms), indicating that conspicuous stimuli trigger faster neural processing and gaze shifts.
Conclusion and Future Directions
This research demonstrates that the primate SC encodes a comprehensive visual saliency map during natural viewing. While the study shows a strong correlation, future work is needed to definitively prove causality and trace the tectothalamic pathways (such as the pulvinar) that relay this midbrain map to the rest of the brain. These results underscore the importance of subcortical structures in fundamental cognitive processes and suggest that our mechanisms for attention are rooted deep in evolutionary history.
Summary updated