Feeling the force: returning haptic signals influence effort inference during motor coordination

Abstract

Our brain is known to automatically optimize effort expenditure during motor coordination, such that for example, during bimanual braking of a bicycle, a well-oiled brake will automatically be used more than a corroded, heavy brake. But how does our brain infer the effort expenditure? All previous motor coordination models have believed that the effort in a task is known precisely to our brain, solely from the motor commands it generates. Here we show that this belief is incorrect. Through experiments and simulation we exhibit that in addition to the motor commands, the returning haptic signals play a crucial role in the inference of the effort during a force sharing task. Our results thus elucidate a previously unknown sensory-motor association that has major ramifications for our understanding of motor coordination and provides new insights into how sensory modifications due to ergonomics, stroke and disease can affect motor coordination in humans.

Figure 1. Bimanual two-finger force sharing task.

(A) The task required subjects to press on two force sensors isometrically with their index fingers. In one session, they wore an elastic cloth glove (right hand in figure) on one of their hands. The subjects were given a visual feedback of the total force they applied (blue bar) which they aimed to match the given target level (yellow bar). (B) The force production by the left (cyan trace) and right (orange trace) fingers by a representative subject in two sessions without and with a (left) glove. The target levels were calibrated to the maximum voluntary contraction (mvc) of individual subjects. (C) Gloved finger takes more load: The force distribution in the task by 12 subjects across three target force levels (0.1 mvc, 0.2 mvc and 0.3 mvc) was quantified as the ratio of the right finger force to the total force (R/(R+L)) and averaged for the no glove sessions across subjects for the left gloved subjects(black-blue trace) and right gloved subjects (black-red trace), and for the sessions with left (blue trace) and right (red trace) gloves. A force distribution of 0.5 indicates that both fingers apply equal force, a higher value shows that the right finger applies more force than the left while a force distribution value less than 0.5 indicates that the left finger applies more force than the right.

Figure 2. Results from single-finger session.

The standard deviation of the force at each target force level shows no significant difference between the gloved (blue/red) and no glove (black) sessions across subjects with the left and right hands (two panels). The individual averages are shown as dots while the across subject average and standard error are represented by solid traces.

Figure 3. Effect of tactile sensation on force distribution.

(A) The average change in force distribution (ΔD from Experiment-1) in every individual was plotted against the individual's average change in force sensation (ΔS from Experiment-2). A positive ΔD indicates an increase in the right finger force while a negative ΔD indicates an increase in the left finger force. Similarly a positive ΔS indicates a decrease in tactile sensitivity in the right finger while a negative ΔS indicates a decrease in tactile sensation in the left finger. A significant linear correlation (p < 0.005) was observed between the decrease in force sensation (ΔS) and the change in force distribution (ΔD) across subjects.

Figure 4. Model predictions and data.

(A) A general representation of the optimal feedback control framework with the three glove effect models. The glove effects the tactile sensation (y). The SEeffect (green) assumes the effect of the glove to be restricted to the sensory signal (y). The FMeffect model (grey) assumes the glove affects the forward model as well by inducing an error in the estimated task input matrix (). The EOeffect model (violet) assumes the glove affects, not just the sensory signal but also the effort optimization during the task. (B) The models made distinct predictions on how a change in tactile sensation (ΔS) as measure in Experiment-2 would change the force distribution (ΔD) in Experiment-1. (C) The difference between the change in finger force distribution in the data (ΔD_data) and the finger force distribution predicted by the model (ΔD_model) was significantly different for the SEeffect and FMeffect models during both left (blue bars) and right (red bars) glove sessions. On the other hand, the data conformed well to the EOeffect model for both the left glove (p = 0.79) and right glove (p = 0.47) subjects.