Fig 1.
(A) Example stimuli from the main 4 experimental conditions (columns, left to right): hands (upper limbs), active prostheses, cosmetic prostheses, hand-held tools. One-handers also observed images from multiple viewpoints of their own prosthesis. One image was shown per trial in an event-related design. (B) Probability maps of the body selective ROI. For each participant and hemisphere, the top 250 most activated voxels within the OTC were chosen based on a headless bodies > objects contrast, providing an independent ROI. ROIs from all participants (n = 56) were superimposed, yielding ROI probability maps. Warmer colours represent voxels that were included in greater numbers of individual ROIs. See S1 Fig for the probability maps of each group separately. Data used to create this figure can be found at https://osf.io/4mw2t/. ROI, region of interest.
Table 1.
Prosthesis users’ demographic details and daily prosthesis usage habits.
Fig 2.
Representational structure in body-selective visual cortex.
Multidimensional distances between activity patterns of each of the main condition (hands, tools, cosmetic prostheses, active prostheses) were calculated using representational similarity analysis. (A) Representational dissimilarity matrices for each group showing pairwise distances between the 2 prostheses conditions (active and cosmetic), hands, and tools. (B) To visualise the underlying representational structure a linkage tree (dendrogram) was calculated in each group of participants, combining information from all pairwise distances (two-handed controls, left; one-handed prosthesis users, right). Pairs of stimuli that are closer together in the multidimensional space are clustered together under the same branch. Longer connections across the vertical axis indicate greater relative distances. In controls, cosmetic prostheses are clustered with hands and active prostheses with tools, reflecting their visual similarities. In prosthesis users, however, the 2 prostheses types (cosmetic and active) are clustered together, with tools and hands represented dissimilarly from both prostheses. Data used to create this figure can be found at https://osf.io/4mw2t/.
Fig 3.
Assessing the embodiment and categorisation hypotheses.
(A–B) A hand similarity index was calculated for each participant to quantify the degree to which both prostheses conditions (cosmetic and active) are more similar to hands than tools. A higher index in value indicates greater embodiment (hand similarity of prostheses). (A) A group comparison of the hand similarity index between controls and prosthesis users showed no significant differences (t(54) = 0.47, p = 0.64). Horizontal lines indicate group means and dots indicate individual participants. (B) Correlation between the hand similarity index and prosthesis usage was not significant across users (Pearson’s r(30) = −0.03, p = 0.86). Dark/light circles indicate congenital/acquired one-handers, respectively, and grey shading indicates a bootstrapped 95% confidence interval. (C) Hand (left) and tool (right) distances from users’ ‘own’ prosthesis. Individual distances were normalised by the controls’ group mean distance, depending on the visual features of the ‘own’ prosthesis (hand likeness). A value of 1 indicates similar hand/tool distance to controls’. Users showed significantly greater distances between their own prosthesis and hands (t(25) = 4.33, p < 0.001) contrary to the embodiment hypothesis. Together, these findings demonstrate that hand similarity under the embodiment hypothesis does not adequately explain differences in prosthesis representation in users’ OTC. (D–F) A prosthesis similarity index was calculated for each participant to quantify the degree to which the prostheses representation moved away from their natural categories (hands for cosmetic prostheses and tools for active prostheses) and towards one another. (D) A visual illustration of the prosthesis similarity index formula. Arrows pointing outward indicate distances that should grow in users compared to controls (e.g., hands and cosmetic prostheses) and are therefore positively weighted. Arrows pointing inward indicate distances that should shrink in users compared to controls (i.e., cosmetic and active prostheses) and are therefore negatively weighted. (E) Group comparison of the prosthesis similarity index between controls, and prosthesis users showed greater prosthesis similarity in users (t(54) = −2.55, p = 0.01). (F) The prosthesis similarity index significantly correlated with prosthesis usage; higher prosthesis index associated with greater prosthesis usage (based on wear frequency and skill; Pearson’s r(30) = 0.43, p = 0.01). Together, these findings demonstrate the categorisation hypothesis explains differences in user’s prosthesis representation in the OTC, both with respect to controls and interindividual prosthesis usage. Data used to create this figure can be found at https://osf.io/4mw2t/. n.s., no significance; OTC, occipitotemporal cortex.