Quantifying center bias of observers in free viewing of dynamic natural scenes.

Human eye-tracking studies have shown that gaze fixations are biased toward the center of natural scene stimuli ("center bias"). This bias contaminates the evaluation of computational models of attention and oculomotor behavior. Here we recorded eye movements from 17 participants watching 40 MTV-style video clips (with abrupt scene changes every 2-4 s), to quantify the relative contributions of five causes of center bias: photographer bias, motor bias, viewing strategy, orbital reserve, and screen center. Photographer bias was evaluated by five naive human raters and correlated with eye movements. The frequently changing scenes in MTV-style videos allowed us to assess how motor bias and viewing strategy affected center bias across time. In an additional experiment with 5 participants, videos were displayed at different locations within a large screen to investigate the influences of orbital reserve and screen center. Our results demonstrate quantitatively for the first time that center bias is correlated strongly with photographer bias and is influenced by viewing strategy at scene onset, while orbital reserve, screen center, and motor bias contribute minimally. We discuss methods to account for these influences to better assess computational models of visual attention and gaze using natural scene stimuli.

The role of memory in guiding attention during natural vision.

What is the time frame in which perceptual memory guides attention? Current estimates range from a few hundred milliseconds to several seconds, minutes, or even days. Here, we answer this question by establishing the time course of attentional selection in realistic viewing conditions. First, we transformed continuous video clips into MTV-style video clips by stringing together continuous clip segments using abrupt transitions (jump cuts). We then asked participants to visually explore either continuous or MTV-style clips and recorded their saccades as objective behavioral indicators of attentional selections. The utilization of perceptual memory was estimated across viewing conditions and over time by quantifying the agreement between human attentional selections and predictions made by a neurally grounded computational model. In the critical condition, jump cuts led to sharp declines in the impact of perceptual memory on attentional selection, followed by monotonic increases in memory utilization across seven consecutive saccades and 2.5 s. These results demonstrate that perceptual memory traces play an important role in guiding attention across several saccades during natural vision. We propose novel hypotheses and experiments using hybrid natural-artificial stimuli to further elucidate neurocomputational mechanisms of attentional selection.

Visual causes versus correlates of attentional selection in dynamic scenes.

What are the visual causes, rather than mere correlates, of attentional selection and how do they compare to each other during natural vision? To address these questions, we first strung together semantically unrelated dynamic scenes into MTV-style video clips, and performed eye tracking experiments with human observers. We then quantified predictions of saccade target selection based on seven bottom-up models, including intensity variance, orientation contrast, intensity contrast, color contrast, flicker contrast, motion contrast, and integrated saliency. On average, all tested models predicted saccade target selection well above chance. Dynamic models were particularly predictive of saccades that were most likely bottom-up driven-initiated shortly after scene onsets, leading to maximal inter-observer similarity. Static models showed mixed results in these circumstances, with intensity variance and orientation contrast featuring particularly weak prediction accuracy (lower than their own average, and approximately 4 times lower than dynamic models). These results indicate that dynamic visual cues play a dominant causal role in attracting attention. In comparison, some static visual cues play a weaker causal role, while other static cues are not causal at all, and may instead reflect top-down causes.