The head-free recording, together with the large fixation window (see Methods), in principle allowed for large deviations between eye position and the fixation dot. Contrary to this expectation, we found that the animals’ fixation accuracy (mean distance from the fixation target) was excellent, regardless of injection type (saline: mean = 0.67 dva, CI, 0.67–0.68; ketamine: mean = 0.73 dva, CI, 0.73–0.74). The difference between the ketamine and saline trials was small but statistically significant (mean difference [M] = 0.06 dva; CI, 0.05–0.07; p ≤ 0.001, permutation test). In other words, ketamine reduced fixation accuracy. Fixation instability (the total distance traveled during stimulus presentation; see Methods) was low as well (saline: mean = 1.46 dva; CI, 1.43–1.49; ketamine: mean = 1.17 dva; CI, 1.15–1.2), and lower during ketamine (M = –0.28 dva; CI, –0.32 to –0.24; p ≤ 0.001, permutation test).
Next, we investigated whether eye movements affected the animal's perceptual choices in the illusory CvCS condition. A logistic regression (see Methods) showed that fixation accuracy did affect the animals’ choices, β = 0.65, SE = 0.18, t(1879) = 3.68, p < 0.001; however, fixation instability did not, β = –0.03, SE = 0.03, t(1879) = –0.81, p = 0.42. In other words, the animals chose the test grating with surround significantly more often in trials with lower fixation accuracy; the illusion was weaker in trials with lower fixation accuracy. Importantly, even after accounting for this influence of fixation accuracy, ketamine was still a strong predictor for the animals’ choices, β = 0.27, SE = 0.1, t(1879) = 2.5, p = 0.011. This shows that eye movements may have modulated the illusion strength, but ketamine also reduced the illusion in an eye-movement–independent manner.