The saccadic flow baseline: Accounting for image-independent biases in fixation behavior

Abstract

Much effort has been made to explain eye guidance during natural scene viewing. However, a substantial component of fixation placement appears to be a set of consistent biases in eye movement behavior. We introduce the concept of saccadic flow, a generalization of the central bias that describes the image-independent conditional probability of making a saccade to (xi+1, yi+1), given a fixation at (xi, yi). We suggest that saccadic flow can be a useful prior when carrying out analyses of fixation locations, and can be used as a submodule in models of eye movements during scene viewing. We demonstrate the utility of this idea by presenting bias-weighted gaze landscapes, and show that there is a link between the likelihood of a saccade under the flow model, and the salience of the following fixation. We also present a minor improvement to our central bias model (based on using a multivariate truncated Gaussian), and investigate the leftwards and coarse-to-fine biases in scene viewing.

Decision making in slow and rapid reaching: Sacrificing success to minimize effort

“…It is important to note that biases and heuristics can boost performance above a completely random baseline, but without the computations required for computing an optimal strategy. This idea is formalised in a model of eye movements during visual search (Clarke, Green, Chantler, & Hunt, 2016), in which the sequence of eye movements was selected at random, but from a population of eye movements participants made from that region of the search area (see also Clarke, Stainer, Tatler, & Hunt, 2017). This random walk incorporates natural tendencies of the saccade system to, for example, make eye movements of a particular size and angle, and to saccade toward the centre more than other regions.…”

Variable and sub-optimal responses to a choice problem are a persistent default mode

“…Although this strategy is simple to understand and implement, the results of demonstrate that in this and two other tasks (visual detection and memorizing strings of digits), naive participants not only fail to maximize their potential accuracy on the task, most of them do not modify their behaviour in a way that varies systematically with manipulations of task difficulty at all (see also Morvan and Maloney, 2012). In a follow-up study, James, Clarke and Hunt (2017) ruled out an explanation based on missing or inaccurate information about throwing ability. They asked half of the participants, just before they executed each throw on the 45 trials in the choice phase of the experiment, to report their expected odds of success to hit each of the two targets from their chosen standing position.…”

“…In a complex and unpredictable environment, a more efficient approach to these kinds of "small" problems may be to solve them with variability , allowing the constraints of the immediate environment to shape the set of viable choices and randomly varying within that set to allow for flexibility and learning to occur. Consistent with this notion, a stochastic model of fixation selection during visual search, which selects fixations at random from a population of common saccade vectors, describes human search behaviour reasonably well (Clarke, Green, Chantler & Hunt, 2016;Clarke, Stainer, Tatler & Hunt, 2017). A similar process of random selecting from a population of possible responses could guide many decisions, preventing stereotyped behaviour while avoiding overthinking of minor choices.…”

Practice-related changes in eye movement strategy in healthy adults with simulated hemianopia

“…In this model, each eye movement during search is randomly selected from the population of eye movement vectors that a participant tends to make from that region of the search array. Thus while the saccade selection process is random, the population of saccade vectors is constrained, making some locations more likely to be fixated than others (see also Clarke et al, 2017). These constraints on the population of vectors, we argue, come from a combination of motor, perceptual, and attentional biases that have evolved or develop gradually to make "random" search more efficient.…”

“…But this model's premise is inconsistent with research revealing profound failures to direct eye movement to locations that could maximise information gain Morvan & Maloney, 2012;Verghese, 2012). The results of these latter studies are more consistent with a stochastic model of eye movements during search, where each eye movement is randomly selected from a population of the eye movements participants can make from that region of the screen (Clarke, Greene, Chantler & Hunt, 2016;Clarke, Stainer, Tatler & Hunt, 2017).…”