Motor control drives visual bodily judgements

Abstract

The 'embodied cognition' framework proposes that our motor repertoire shapes visual perception and cognition. But recent studies showing normal visual body representation in individuals born without hands challenges the contribution of motor control on visual body representation. Here, we studied hand laterality judgements in three groups with fundamentally different visual and motor hand experiences: two-handed controls, one-handers born without a hand (congenital one-handers) and one-handers with an acquired amputation (amputees). Congenital one-handers, lacking both motor and first-person visual information of their missing hand, diverged in their performance from the other groups, exhibiting more errors for their intact hand and slower reaction-times for challenging hand postures. Amputees, who have lingering non-visual motor control of their missing (phantom) hand, performed the task similarly to controls. Amputees' reaction-times for visual laterality judgements correlated positively with their phantom hand's motor control, such that deteriorated motor control associated with slower visual laterality judgements. Finally, we have implemented a computational simulation to describe how a mechanism that utilises a single hand representation in congenital one-handers as opposed to two in controls, could replicate our empirical results. Together, our findings demonstrate that motor control is a driver in making visual bodily judgments.

Keywords: Amputees; Body representation; Embodied cognition; Motor simulation; Phantom limb; Visuomotor.

Fig. 1

Hand laterality judgment stimuli and results. (A) Example stimuli used in the hand laterality judgement task. (B) Group performance (absolute RT, left; Accuracy, right) in the hand laterality judgement task is shown for controls (grey), amputees (blue) and congenital one-handers (orange) for the intact and missing hands (light vs dark shades, respectively). Dots correspond to individual performance. (C) Group performance (absolute RT, left; Accuracy, right) in the hand laterality judgement task is shown for easy and hard postures in controls (grey), amputees (blue) and congenital one-handers (orange). Values indicate means ± standard error. Congenital one-handers exhibit slower RTs in hard postures compared to controls. Congenital one-handers, but not amputees and controls, also show an accuracy difference between the two hands. CT = controls; AM = amputees and CG = congenital one-handers. For RT, absolute RT values are plotted, however all statistical analyses were performed on log-transformed RT values (see supplementary figure S5 for plots with log-transformed RT values). Pink diamonds depict the predicted performance values from the computational model. D = dominant side images and ND = non-dominant side images. I = intact side images and M = missing side images (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

Fig. 2

Amputees phantom hand motor control correlated with hand laterality judgement performance. Mean RT (ranked) for laterality judgements (both intact and missing hand images) is significantly correlated to RTs (ranked) in the phantom hand motor control, r_s(13) = 0.695, p = 0.004. This suggests that existing motor control, rather than visual experience, relates to laterality judgements. Smaller values on the phantom motor task denote faster RTs while executing sequential finger-thumb oppositions with the phantom hand. Age at amputation does not correlate with RT for laterality judgements, r_s(14) = 0.174, p = 0.52. A direct comparison between the correlation revealed there is a significant difference between the two correlations. Thus, for amputees, motor control of the phantom hand is a better predictor of performance on the laterality task than the lack of visual experience of the missing hand.

Fig. 3

A schematic diagram illustrating the laterality decision making process as simulated by our computational approach. The top panel illustrates a decision process example in a two-handed control (depicted as the woman in the grey dress). In response to the hand image stimulus, left- and right-hand posture models are simultaneously activated, each collecting evidence to either accept or reject whether the hand posture in the stimulus could be generated with that hand. In the example above, once enough evidence is collected by the left-hand posture model to accept that the visual hand posture can be generated by the left hand, the two-handed individual (correctly) judges the hand stimuli as a left hand. Because the actual stimulus is left, the drift rate is positive and negative for the left- and right-hand simulators, respectively. Once the left-hand model collected enough evidence to reach the decision threshold, the right-hand model abandons its process. The bottom panel illustrates the process in response to the same visual stimulus in a right-handed congenital (orange dress). In this example, we illustrate how having a single hand posture model can be less efficient and result in slower reaction times. Since this individual never had visuomotor experience of a left hand, we assume this individual can only utilise a single right-hand posture model to judge hand laterality. In the example above, the right-hand model takes longer to reach the decision threshold, to conclude the presented posture cannot be generated with the right hand. Note that if sufficient evidence had not been collected to reach a decision before the quit timer, then a “reject” decision would be forced.