Individual examples of the time course of trans-saccadic recalibration and saccade adaptation from
Experiment 2 are depicted in
Figure 6. To compute an aggregate measure of trans-saccadic learning, we averaged the gain values across the four segments of each session (trials 1 to 30 without trans-saccadic size change or target step, trials 31 to 100, and trials 101 to 170 with opposite trans-saccadic manipulations, and trials 170 to 210 again without trans-saccadic manipulation). In
Figure 7 we plotted the averaged data for training trials, test trials at the same location, and test trials at the opposite location, respectively, and separately for the different training sequences (decrease-increase vs. increase-decrease for trans-saccadic recalibration and inward-outward vs. outward-inward for saccadic adaptation).
A quick inspection of the plots in
Figure 7 reveals a clear crossover interaction in the gain values in the middle of the session in all plots but the one relative to saccade adaptation at the opposite location. This crossover interaction is the diagnostic sign of learning in our paradigm, as it mirrors the reverse order of the training manipulation. The fact that it is absent specifically in the case of saccadic adaptation at the opposite locations seems to suggest that saccadic adaptation acquired at one location, contrary to trans-saccadic recalibration of perceived size, does not transfer to a mirror location in the other hemifield.
Notice also that there is a general tendency to overshoot in the first trial block of the saccade adaptation session. This might have to do with the fact that observers were trained to produce one clear saccadic movement toward the target in the previous session, to ensure that the trans-saccadic manipulation could be applied as the saccade was in flight. Furthermore, notice that the relative amplitude of saccades at the end of the session is generally the same in trials 101-170 as it was in trials 1-30, or slightly lower. This is compatible with the idea that saccade amplitudes are determined by the adaptation effect and by a general tendency to saccade amplitude decrease (
Cassanello et al., 2019). This decrease is probably less evident in our paradigm owing to the fact that observers were already producing saccades in the previous trans-saccadic recalibration session.
To summarize the main finding of the experiment, and to allow for a direct statistical analysis of the main experimental question, we further aggregated the data by computing, for each panel in
Figure 7, the difference in the average gain in the segments in which the training trans-saccadic change was increase/outward, and the average gain when the training was decrease/inward. The values are shown in
Figure 8. Evidently, all values are positive (as expected if the training affects the adjustments/saccadic amplitudes), and mostly larger in the case of saccadic adaptation, except for the opposite position, in which there is no sign of saccadic adaptation but there is still size recalibration. For comparison, if learning would be complete, the total gain difference would be 0.3 (twice the 0.15 trans-saccadic manipulation), so when present, saccadic adaptation reaches 50% of the maximum expected effect and size recalibration approximately 25%. The general impression was confirmed by an analysis of variance with task (size recalibration vs. saccade adaptation) and trial type (training, test at the same location, and test at the opposite location) as factors, which revealed a significant two-way interaction,
F(2,22) = 11.867,
p < 0.001,
ηp² = 0.519. This was further explored by performing one-sample
t-tests against a value of 0 in each cell of the experimental design (
Table 1), which evidenced significant effects of learning in all but the case of saccadic adaptation at the opposite location, again confirming that spatial transfer of trans-saccadic learning at mirror locations takes place in the case of size recalibration but not in the case of saccadic adaptation.