When we scan a room for our keys, we often assume our eyes move in a neutral, systematic way. However, research into human gaze dynamics reveals that our visual system is far from impartial. Instead, it is governed by a set of "navigation rules" and inherent biases that make certain features—like size, color, and orientation—fundamentally unequal in how they guide our attention.
The Asymmetry of Visual Features
A core discovery in this field is that visual search is often asymmetric: it is easier to find a "tilted bar among vertical bars" than the reverse. Zhaoping Li demonstrates in A V1 Model of Pop Out and Asymmetry in Visual Search that this happens because our primary visual cortex (V1) is biased toward "cardinal" orientations (horizontal and vertical).
When distractors have these heavily weighted cardinal features, they create a stronger inhibitory signal in the brain, effectively "silencing" the background and letting a non-cardinal target pop out. Dietmar Heinke and Andreas Backhaus furthered this intuition with the VS-SAIM model, showing that these asymmetries emerge from the interaction between top-down templates and bottom-up feature weighting.
Attractors and Repellers
Not all "elementary" features are created equal when it comes to moving our eyes. Stella X. Yu found that size acts as a powerful "search attractor," where the gaze preferentially transitions between large objects while ignoring smaller ones (Feature Transitions with Saccadic Search). Conversely, color can sometimes act as a "repeller," driving the gaze away from the current attribute and hindering search efficiency.
This hierarchy of features is also reflected in how we learn. Farhan Baluch and Laurent Itti showed that while we can train our top-down attention to better utilize color or spatial frequency, orientation remains stubbornly difficult to optimize (Training Top-Down Attention Improves Performance). This suggests that the neural mechanisms for certain features are more "plastic" than others.
Strategic Adjustments to Scene Statistics
Our search behavior isn't just a fixed biological reflex; it's a strategy that adapts to the target's characteristics. Lars Oliver Martin Rothkegel and colleagues observed that searchers adjust their "eye rhythm"—using larger jumps (saccades) and longer pauses (fixations) when looking for coarse, low-detail targets (Searchers adjust their eye movement dynamics). This allows the brain to process a larger attentional window in the periphery when high-resolution detail isn't required.
Furthermore, these biases may be learned from our environment. Shashi Kant Gupta demonstrated that deep learning models (eccNET) develop human-like search asymmetries when trained on natural images, suggesting that our "quirky" visual biases are actually efficient adaptations to the statistical properties of the world around us.
Space-Filling and Memory
Finally, humans employ local orienting strategies to ensure they don't waste time. Alex Szorkovszky identified a bias called "saccade crossing avoidance," where our eyes systematically avoid crossing the paths of previous movements (Saccade crossing avoidance as a visual search strategy). This acts as a space-filling strategy, ensuring we maximize visual coverage and minimize redundant exploration of the same area.