Participants
We recruited 25 participants from the research pool of the University of Western Ontario (UWO). All participants reported normal vision and were compensated with course credit. We excluded data from 8 participants due to hardware malfunction (n = 3) or chance-level task performance (n = 5), leaving 17 participants (median age: 18-years, range 18–23). The Research Ethics Board of UWO approved this study.
Equipment
A robot arm repeatedly moved objects through 90° on the horizontal plane ~30-cm in front of the participant’s eyes (45° either side of forward-facing; manufactured by Bonneville Scientific Inc., Salt Lake City, Utah, USA). One full oscillation lasted 10-s. An Eyelink II (SR Research Ltd, Mississauga, Ontario, Canada) tracked the movements of the left-eye (sample rate: 500-Hz; Fig. 1). The experimenter recalibrated the eye-tracker prior to each trial. The object stimulus (a ball) was mounted to the back of the mirror at the end of the robot arm, and the appropriate stimulus turned to face the participant on each trial. This ensured that the size of the moving portion of the arm was consistent across stimulus conditions. Prior to completion of mirror trials, the experimenter ensured that the mirror was placed in such a way that the participant could see the reflection of their face throughout the robot arm’s range of movement.
Cognitive task
To increase cognitive load while participants pursued the stimuli, we developed a two-back category-matching task. Participants heard a series of spoken letters (3-s onset asynchrony) and indicated via button press whether or not each item was from the same category as that two items previously. Five letters that contain horizontal lines when written in their capital form comprised one category (E, F, H, L, T) and five that did not (V, W, X, Q, S) comprised the other. Such an orthographic comparison task was hypothesised to tax visual processing and interfere with pursuit demands. A pseudo-random series of 90 letters (22 matches) was created per trial.
Design and procedure
Participants completed four trials from a two-by-two design — two stimuli (mirror, ball) x two levels of cognitive load (pursuit only, pursuit plus two-back) – in counterbalanced order. Prior to the experiment, participants completed a practice two-back task until they understood the procedure. Participants were instructed to smoothly pursue the moving stimuli with their eyes without moving their head. Each trial lasted 4.5-min followed by a brief rest.
Analyses
To rule out potential task non-compliance, we excluded those participants with chance performance on either trial. Specifically, we shuffled the recorded button presses within trials and calculated the corresponding discrimination value (p[hit] minus p[false alarm] [11]), thus controlling for response biases. We performed this procedure 1000 times, creating a distribution under the null hypothesis that discrimination was not different from chance. Participants with discrimination within the lower 95% of the surrogate distribution (i.e. p > .05) were excluded from analyses (conducted with MATLAB).
Frequentist and equivalent Bayesian comparisons with default priors were conducted with JASP Version 0.7.1.12 [12, 13]. Specifically, to complement the t-tests, the Jeffrey-Zellner-Siow Bayes factor (JZS-BF10) tested the strength of the evidence for each observed effect size [14]. A JZS-BF10 ANOVA approach contrasted the strength of evidence for models reflecting the null, main effects, and interaction [15]. A JZS-BF10 between 1/3 and 3 is considered to be only weak/anecdotal evidence for an effect; 3–10: substantial evidence; 10–100: strong evidence; >100: very strong evidence [16]. The same category descriptions hold for the inverse.