PCA of pairwise distances.

To explore which of the pairwise distances contributed to the underlying observed effect of prosthesis categorisation we ran a data-driven analysis (PCA) on the 5 distances of the one-handed group. Values in the table are the weights given to each distance within a component. The first component shows a ‘main effect’ of interindividual differences across participants, in which some individuals have overall larger distances than others across all condition pairs. In our calculated indices, we control for this effect by normalising the individual’s selectivity indices by their Hands ↔ Tools distance (see ‘Methods‘). The second component explains almost half of the remaining variance (after accounting for the interindividual differences in component 1). In the second component, individuals showing greater distances between the active prostheses and the tool condition also show greater similarity between the active prosthesis and the cosmetic prosthesis conditions. In other words, when the active prosthesis condition moves away from the tool category, it also tends to get closer to the cosmetic prostheses (as can be seen by the high weights and opposite signs of these 2 distances in the second component). This data-driven analysis provides further support for the hypothesised categorical shift of prosthesis representation. PCA, principle component analysis
(XLSX)