All data met the normality assumption (p>.057 in all cases), therefore, parametric tests were used throughout (see Data S1). Analysis for the CCT followed Zopf et al.24 (link) Trials in which participants responded too soon (≤150 ms) or too late (≥1500 ms) were discarded from the analyses and only the reaction times (RT) for correct responses were used.24 (link) The CCT effect was calculated as participants’ performance on non-homologous minus homologous trials.24 (link) As accuracy and RT in the CCT are interdependent,24 (link),57 results from each participant were integrated into an Inverse Effectiveness Score (CCT-IES; reported in Figure 1C in the main text).24 (link),57 ,58 Supplemental analyses of participants’ reaction times (CCT-RT) and percentage errors (CCT-PE) showed substantially similar results than the combined CCT-IES measure (see Figure S2). A 2 (effector-type: hand, eye) x 2 (spatial-congruency: congruent, incongruent) repeated measures ANOVA was used to analyse CCT data.
The Onset Time of the RHI was analysed with a similar 2 × 2 rmANOVA. Yet, given that not all participants reported an illusion in all four experimental conditions, statistical analyses were only run on the data from participants who responded in all the blocks (n = 12; Figure 1D). The rationale for this choice is that missing values may potentially be due to a range of reasons, from not experiencing any RHI to forgetting the instructions. For example, during the debriefing at the end of the experiment a few participants even reported they had forgotten to step on the pedal because they were experiencing a strong RHI (see Figure S3 for details on formal analyses supporting this rationale).
Finally, given that Subjective Reports were given on a continuous visual analog scale (from −100 = strongly disagree to +100 = strongly agree) instead of a 7-points Likert scale,22 (link) participants’ responses were also analysed through parametric tests. Participants’ agreement with each statement of the questionnaire were first averaged across question type (i.e., RHI/control items), and then analysed in a 2 (spatial-congruency: congruent, incongruent) x 2 (effector-type: hand, eye) x 2 (item-type: illusion, control) rmANOVA.
For each of the three dependent variables, we expected a main effect of spatial-congruency, in line with the general principle that crossmodal stimulation must be spatially congruent in order to elicit the RHI.37 (link) Crucially, in line with our main hypothesis of similar RHI for hand and eye movements, we also predicted no significant interaction between main effects. Given that the experiment was designed to allow asserting the null hypothesis (i.e., eye movements and hand movements induce the same amount of RHI), non-significant interactions between spatial-congruency and effector-type were further tested through Bayesian t-tests on the difference between spatially congruent/incongruent conditions in each type of movement. This allowed us to determine whether the datasupported the null hypothesis or if, alternatively, the null result could reflect insufficient statistical power.26 (link) All the ANOVAs were conducted using IBM SPSS Statistics for Windows, version 23 (IBM Corp.,USA). Bayesian analyses were run on JASP v. 0.12.1 (JASP Team 2016, University of Amsterdam).
Free full text: Click here