Most studies focus on average behavior across a whole block of trials. However, there is recent emerging evidence for a direct influence of past experience on perceptual judgments. When judging the orientation of a grating, responses of observers are systematically biased towards the orientation presented in the previous trials. Such effects that even occur at the trial-by-trial level have been termed serial dependence and occurs at the whole processing hierachy from orientation to facial expression or identity.

In a series of study we are investigating whether such serial dependence effects also occur for actions. We are also looking at the relevant signals that could mediate such an influence of previous trials.

The stupid oculomotor system?

The nature of the relevant signals for these sequential effects is still hotly debated. A recent study published by Cicchini and colleagues elegantly demonstrated that prior history information influencing perceptual decisions is based on high level sensory information with integrated contextual information. In direct contrast to this, our results show that sequential effects for oculomotor behavior are based on early levels of sensory signals. Thus, at a first glance the oculomotor system could seem stupid as it is ignoring visual context and common perceptual constancy features. 

We measured observers’ eye movements as they tracked sequences of cars and blobs moving at different perceived depth levels across linear perspective drawings (see Figure). Making use of the Ponzo illusion allowed us to elegantly dissociate the retinal speed and size from the perceived speed and size of moving objects. Our results reveal that sequential effects depend only on the retinal speed of the previous movements, ignoring the perceived speed even when perceived and retinal speeds differed. Sequential effects even transferred from one target type to the other, indicating the relevant signals for serial dependence on oculomotor behavior ignore context effects and object consistency and seem to be based solely on retinal error signals in the early levels of visual processing. 

These results show that beyond a new perception-action dissociation, the demonstrated serial dependence effect uncovers fundamentally different goals for perception and oculomotor control. The perceptual system integrates and interprets retinal information in order to accomplish size and velocity constancy. In contrast, the oculomotor system ignores these constancy mechanisms and simply tries to bring and keep any object of interest close to the fovea, which in turn then allows for an optimal perception of the moving object. Highlighting the differing goals of the two systems provides an exciting framework to interpret action-perception dissociations, which would be relevant to a broad audience working in vision, behavioral and clinical neuroscience, decision-making and beyond. 

Goettker, A., & Stewart, E. E. (2022). Serial dependence for oculomotor control depends on early sensory signals. Current Biology, 32(13), 2956-2961.

Sequential effects in oculomotor behavior

Here we investigated which differences in the previous trial affect subsequent oculomotor behavior. We observed that previous velocity as well as previous position of the target can lead to changes in subsequent oculomotor behavior, and that this effect is especially strong if the previous stimulus was high contrast. This suggests a reliability-weighted integration of previous information with the current sensory input.

By using trial-by-trial variability in how observers tracked the moving target (either pure pursuit or pursuit and additional corrective saccades –> see research on Saccade-Pursuit interactions), we could show that the mediating signal for subsequent oculomotor behavior seems to be the retinal velocity including the saccadic epochs. Interestingly, the short-term adaptation of subsequent oculomotor behavior seems to be maladaptive: A previous trial with a forward corrective saccade led to a slower pursuit response in the next trial and therefore the increased need for a larger forward corrective saccade. Thus, on the trial-by-trial level this adaption was unstable.

Goettker, A. (2021). Retinal error signals and fluctuations in eye velocity influence oculomotor behavior in subsequent trials. Journal of Vision, 21(5), 28-28.