Embodied decisions during walking

Abstract

Research on embodied decision-making only recently started to examine whether and how concurrent actions influence value-based decisions. For instance, during walking humans preferably make decisions that align with a turn toward the side of their current swing leg, sometimes resulting in unfavorable choices (e.g., less reward). It is suggested that concurrent movements influence decision-making by coincidental changes in motor costs. If this is true, systematic manipulations of motor costs should bias decisions. To test this, participants had to accumulate rewards (i.e., points) by walking and turning toward left and right targets displaying rewards across three experiments. In experiments 1a and 1b, we manipulated the turning cost based on the current swing leg by applying different symmetric turning magnitudes (i.e., same angles for left and right targets). In experiment 2, we manipulated the turning cost by administering asymmetric turning magnitudes (i.e., different angles for left and right targets). Finally, in experiment 3, we increased the cost of walking by adding ankle weights. Altogether, the experiments support the claim that differences in motor costs influenced participants' decisions: experiments 1a and 1b revealed that the swing leg effect and stepping behavior were moderated by turning magnitude. In experiment 2, participants showed a preference for less costly, smaller turning magnitudes. Experiment 3 replicated the swing leg effect when motor costs were increased by means of ankle weights. In conclusion, these findings provide further evidence that value-based decisions during ongoing actions seem to be influenced by dynamically changing motor costs, thereby supporting the concept of "embodied decision-making."NEW & NOTEWORTHY Motor processes of concurrent movements have been shown to influence embodied decision-making. It is hypothesized that this is driven by coincidental changes in motor costs. We tested this claim by systematically manipulating motor costs of choice options during walking. In three experiments we show how variations in motor cost (e.g., turning angle or stepping constraints) bias decision-making, thereby supporting the concept of "embodied decision-making."

Keywords: decision-making; embodied choices; gait; motor cost; reward.

Graphical abstract

Figure 1.

Experimental setup of experiment 1a. A: participants started by positioning their feet in the required starting position. The projection at top shows the time course of cues. The starting position was displayed with a projector on the opposite side of the room (see top). The font color represented available lateral targets to finish a trial (here red, representing 90° targets). The German word “Bereit” was used in the experiment instead of “Ready” as displayed here. After participants took the required starting position, a “+” appeared as the Go signal and they walked toward the central zone. After the third step, rewards appeared on the left and right sides of the screen. Participants had to step into the central zone and walk toward a target area to finish a trial and receive rewards. B: example of a lateral step. Given that the right foot stepped into the zone and participants chose to walk to the left side, a lateral step can be taken. C: example of a crossover stepping strategy. Given that the right foot stepped into the zone, we assumed that participants make a crossover step toward the right side.

Figure 2.

Experimental setup and hypothesis plots for all experiments. A: experiment 1a displays the basic design used for the rest of the experiments. Bottom: the hypothesis of experiment 1a: we expected the swing leg effect (SLE) to increase with turning magnitude. B: based on the results of experiment 1a, in experiment 1b a stepping constraint was placed with carpet on the central zone. Additionally, only the 15° and 90° target angles were included and the cue for targets changed to rectangles on the sides (here blue for 15° turning magnitude). For experiment 1b we had the same hypothesis as in experiment 1a, pictured at bottom. C: in experiment 2 the target angle could be asymmetric (here left 15° in blue and right 90° in red). The target angles were displayed either before participants started walking (top left, control condition) or with rewards in the third step (top right). Bottom: the hypothesis of experiment 2: we expected for both the early and more importantly the late target timing a preference to walk toward the side with a smaller turning magnitude (angle effect, AE), in addition to the SLE. D: In experiment 3 participants walked with and without ankle weights. Additionally, only 52.5° turning magnitude and the same obstacle as in a prior study (14) were used instead of the cones. Bottom: the hypothesis of experiment 3: we expected the weights to increase the SLE.

Figure 3.

Effect of the swing leg in decision-making for different turning magnitudes and reward combinations. A: results for experiment 1a. The swing leg effect (SLE) was reliable for all reward combinations. Additionally, the SLE increased when comparing 52.5° to 90° for the equal reward combination (center, marked by red arrow) and in the 60 points right condition (right, marked by red arrow) but not reliable for the other conditions. B: results for experiment 1b. There was a generally stronger SLE compared to experiment 1a. However, there is reliable evidence that the SLE increased with turning magnitude. The y-axis displays the probability of walking toward the rightward side. 0% would mean that participants always went toward the left side; 100% means that participants always went toward the right side. Shown are the model estimates of the mean and 95% credible interval (CrI) for each condition. Note that the y-axis scale differs between equal (center) and unequal (left and right) reward conditions.

Figure 4.

Characteristics of the step after reaching the zone and stepping strategies with varying turning magnitude. Only trials with decisions to walk toward the opposite side of the swing leg were analyzed (and presumably a crossover step was needed). A: position of the step after reaching the zone from the step in the zone in experiment 1a. Participants did not always cross the stance leg (circles, crossover steps) but stepped with the foot in the zone next to the stance leg (triangles, transition steps). The black rectangle represents the central zone. The gray semicircle represents the constraint area for the step after reaching the zone in experiment 1b. For brevity reasons, the figure shows only trials in which participants walked toward the left side with a right swing leg (hence blue). When participants walked toward the right side, the foot positioning was similar but mirrored. B: probability of a crossover step vs. a transition step for different turning magnitudes. Displayed are the mean estimate and 95% credible interval (CrI) on the response scale. Trials for both walking directions were analyzed. C: probability of finishing a trial in time for both stepping strategies. Trials with a transition step and larger turning magnitudes were more often too late compared to trials with a crossover step and smaller turning magnitudes. D: same as A, but for experiment 1b. Again, participants did not always perform crossover steps (circles) but still used transition steps (triangles). However, compared to experiment 1a, the transition step was now mostly outside the constraining carpet. E: same as B, but for experiment 1b. Compared to experiment 1a, participants more often made crossover steps, but the frequency of crossover steps still decreased with the turning magnitude. F: same as C, but for experiment 1b. Again, trials with transition steps and larger turning magnitudes were slower.

Figure 5.

Effect of the turning magnitude and swing leg on decisions for different presentation timings of the targets and reward combinations in experiment 2. A: results when turning magnitudes were displayed before a trial started. B: results when turning magnitudes were displayed with reward while participants were walking. Only asymmetric angle combinations are included, meaning that 15° left indicates that the right turning magnitude was 90°. 0% means that participants always went toward the left side; 100% means that participants always went toward the right side. Lateral steps are realizable in the direction of the swing leg; hence, a swing leg effect would mean that participants went more often leftward given a left swing leg vs. a right swing leg (gray below black). A preference to walk toward the side with a smaller turning magnitude (15°) would be displayed by a positive slope between points. There was a reliable effect of turning magnitude: participants preferred to walk toward the side with a smaller turning magnitude. However, there was no reliable evidence for the swing leg effect (SLE) anymore. Displayed are the model estimates (probability scale) of the mean and 95% credible interval (CrI) for each condition. Note that the y-axis scale differs between equal (center) and unequal (left and right) reward conditions.

Figure 6.

Effect of the swing leg on decision-making with and without ankle weights and for different reward combinations in experiment 3. The x-axis displays the probability of walking toward the rightward side. 0% would mean that participants always went toward the left side; 100% means that participants always went toward the right side. Lateral steps are realizable in the direction of the swing leg; hence, a swing leg effect would mean that participants more often went leftward given a left swing leg vs. a right swing leg (gray below black). The slope of the line between points indicates the influence of weights on the swing leg effect. A stronger swing leg effect with weights would be indicated by a divergence between both lines. There was no reliable evidence that the swing leg effect increased when participants wore ankle weights. Shown are the model mean estimates and 95% credible interval (CrI) for each condition. Note that the y-axis differs between (center) and unequal (left and right) reward conditions.