Temporal evolution of the central fixation bias in scene viewing

Lars Oliver Martin Rothkegel, Hans A. Trukenbrod, Heiko H. Schütt, Felix A. Wichmann, Ralf EngbertJournal of Vision 2017 · DOI

TL;DR

Researchers have found that when people look at images, their eyes tend to fixate on the center first – a phenomenon called the "central fixation bias" (CFB). This study shows that by delaying the first eye movement after an image appears, this bias can be significantly reduced, suggesting it's partly an artifact of sudden image onset rather than purely image content. This has implications for how we design experiments to study visual attention.

What Problem Does This Paper Address?

When we view natural scenes on a screen, our eyes exhibit a strong tendency to initially fixate near the center of the image, a behavior known as the central fixation bias (CFB). This bias can obscure the underlying mechanisms of visual attention, such as how image features or cognitive processes guide our gaze. While previous research has explored explanations for the CFB, this paper aims to understand its temporal evolution and how it can be manipulated. Specifically, the authors investigate whether delaying the initial eye movement after an image appears can reduce the CFB and what role the timing of this first saccade plays.

Methods

The study employed four experiments involving human participants viewing natural scenes. The core manipulation involved varying the "pre-trial fixation time" – the duration participants were instructed to fixate on a cross after the image appeared but before they were allowed to freely explore it. This effectively delayed the first saccade (eye movement) relative to image onset. Eye movements were recorded using an eye-tracker. The researchers analyzed the "distance to center" (DTC) over time as a measure of the CFB, with smaller DTC indicating a stronger bias. They also used computational models (including a modified SceneWalk model) to simulate scanpaths and compare them to empirical data, aiming to understand the mechanisms underlying the CFB.

Datasets

The study used a set of 120 images from Tatler's (2007) original study on the central fixation bias. These images comprised indoor scenes, outdoor scenes with man-made structures, and outdoor scenes without man-made structures. No new datasets were introduced.

Key Insights and Ideas

  • CFB is not solely image-driven: The CFB can be significantly reduced by decoupling image onset from the start of exploration, suggesting it's partly an artifact of sudden stimulus presentation.
  • Timing is crucial: A delay of as little as 125 ms between image onset and the first saccade can lead to a substantial reduction in the early CFB.
  • Saccade latency is a key mediator: The time it takes for the first saccade to occur (saccade latency) strongly influences the CFB, with shorter latencies leading to a stronger bias.
  • Sudden onset as a trigger: The CFB might be an orienting response to the sudden luminance change of an image appearing on screen.
  • Experimental paradigm matters: The standard scene viewing paradigm, where exploration begins immediately after image onset, may artificially inflate the CFB, masking other attentional processes.

Key Results

  • In Experiments 1 and 2, delaying the first saccade by 125 ms or more significantly reduced the early CFB compared to a 0 ms delay (immediate exploration).
  • A delay of 125-250 ms was found to be sufficient for a strong and reliable CFB reduction. Longer delays had diminishing returns and altered the overall viewing pattern.
  • The latency of the first saccade was a significant predictor of the distance to the image center for the second fixation across all experiments. Shorter latencies were associated with a stronger bias towards the center.
  • Model simulations showed that a modified SceneWalk model, incorporating a central activation map at image onset, could qualitatively reproduce the CFB and its temporal evolution, outperforming other models.

Limitations and Future Work

  • Generalizability: The authors acknowledge that the extent to which these findings generalize to other laboratory tasks (e.g., visual search, face perception) that also use sudden stimulus onsets remains an open question.
  • Mechanism of reduction: While the study shows that delaying the first saccade reduces CFB, the precise underlying neural or cognitive mechanisms for this reduction (beyond the initial orienting response) are not fully elucidated.
  • Model limitations: While the SceneWalk StartMap model performed well, it still underestimated the initial CFB in some experiments, suggesting that stronger initial biases might be needed for even better predictions.
  • Future work: The authors suggest using modified scene viewing paradigms that control for the CFB to better study bottom-up and top-down attentional processes. They also propose that initial fixation locations should be more evenly distributed across the image, or even biased towards the periphery, to avoid artificially influencing early fixations. Further investigation into the reasons for the reduction of CFB after 2 seconds of exploration is also warranted.

Summary updated

Abstract

When watching the image of a natural scene on a computer screen, observers initially move their eyes towards the center of the image --- a reliable experimental finding termed central fixation bias. This systematic tendency in eye guidance likely masks attentional selection driven by image properties and top-down cognitive processes. Here we show that the central fixation bias can be reduced by delaying the initial saccade relative to image onset. In four scene-viewing experiments we manipulated observers' initial gaze position and delayed their first saccade by a specific time interval relative to the onset of an image. We analyzed the distance to image center over time and show that the central fixation bias of initial fixations was significantly reduced after delayed saccade onsets. We additionally show that selection of the initial saccade target strongly depended on the first saccade latency. Processes influencing the time course of the central fixation bias were investigated by comparing simulations of several dynamic and statistical models. Model comparisons suggest that the central fixation bias is generated by a default activation as a response to the sudden image onset and that this default activation pattern decreases over time. Our results suggest that it may often be preferable to use a modified version of the scene viewing paradigm that decouples image onset from the start signal for scene exploration and explicitly controls the central fixation bias. In general, the initial fixation location and the latency of the first saccade need to be taken into consideration when investigating eye movements during scene viewing.
BibTeX
@article{rothkegel2017temporal,
  author = {Lars Oliver Martin Rothkegel and Hans A. Trukenbrod and Heiko H. Schütt and Felix A. Wichmann and Ralf Engbert},
  title = {{Temporal evolution of the central fixation bias in scene viewing}},
  journal = {Journal of Vision},
  year = {2017},
  doi = {10.1167/17.13.3},
}

Loading related content...

Figure