The Rubber Hand Illusion: Top-down attention modulates embodiment

Abstract

The Rubber Hand Illusion (RHI) creates distortions of body ownership through multimodal integration of somatosensory and visual inputs. This illusion largely rests on bottom-up (automatic multisensory and perceptual integration) mechanisms. However, the relative contribution from top-down factors, such as controlled processes involving attentional regulation, remains unclear. Following previous work that highlights the putative influence of higher-order cognition in the RHI, we aimed to further examine how modulations of working memory load and task instructions-two conditions engaging top-down cognitive processes-influence the experience of the RHI, as indexed by a number of psychometric dimensions. Relying on exploratory factor analysis for assessing this phenomenology within the RHI, our results confirm the influence of higher-order, top-down mental processes. Whereas task instruction strongly modulated embodiment of the rubber hand, cognitive load altered the affective dimension of the RHI. Our findings corroborate that top-down processes shape the phenomenology of the RHI and herald new ways to improve experimental control over the RHI.

Keywords: Rubber hand illusion; body ownership; bottom-up processes; consciousness; selfhood; top-down cognition.

Figure 1.

Rubber hand illusion setup. Left: experimental setup with fake silicon arm between the two real hands and the occluder. Right: participant view during synchronous stroking.

Figure 2.

Exploratory Factor Analysis. Numbers on the left represent item numbers; numbers on the lines represent item loadings on their primary factor or intercorrelation between factors. The red lines represent negative loadings.

Figure 3.

Model comparison: confirmatory versus exploratory factor. Left panel: the large BIC difference suggests a much better fit for Exploratory Factor Analysis (EFA) than Confirmatory Factor Analysis (CFA). Right panel: distributions of bootstrapped (10,000 samples each) total variances explained by the CFA and EFA models, respectively. Note that for the CFA, 250 bootstrapped variances explained (out of 10,000) were greater than one due to the bootstrapping process, so were excluded for this figure. The overlapping distributions suggest both the CFA and EFA models explain comparable total variances, with the EFA explaining marginally more. BIC: Bayes Information Criterion.

Figure 4.

Embodiment factor. Averaged standardised factor scores for the “Embodiment of rubber hand” dimension (y-axis). In the original scale, participants could choose between 0 (“I do not agree at all”) and 7 (“I agree completely”). Regression analyses revealed that Instructions were a statistically reliable predictor of embodiment, (β = 0.5, SE = 0.1, 95% CI = [0.29, 0.7]. Error bars represent 95% bootstrapped confidence intervals.

Figure 5.

Affect factor. Averaged standardised factor scores for the “Affect” dimension (y-axis). In the original scale, participants could choose between 0 (“I do not agree at all”) and 7 (“I agree completely”). Regression analyses revealed that load was a statistically reliable predictor of affect (β = –0.17, SE = 0.09, 95% CI = [–0.35, –0.009]). Error bars represent 95% bootstrapped confidence intervals.

Figure 6.

Loss of one’s own hand factor. Averaged standardised factor scores for the “Loss of own hand” dimension (y-axis). In the original scale, participants could choose between 0 (“ I do not agree at all ”) and 7 (“I agree completely”). There were no significant effects. Error bars represent 95% bootstrapped confidence intervals.

Figure 7.

Feeling of having two right hands factor. Averaged standardised factor scores for the “Feeling of having two right hands” dimension (y-axis). In the original scale, participants could choose between 0 (“I do not agree at all”) and 7 (“I agree completely”). There were no significant effects. Error bars represent 95% bootstrapped confidence intervals.