Movement Improves the Quality of Temporal Perception and Decision-Making

Abstract

A critical aspect of behavior is that mobile organisms must be able to precisely determine where and when to move. A better understanding of the mechanisms underlying precise movement timing and action planning is therefore crucial to understanding how we interact with the world around us. Recent evidence suggests that our experience of time is directly and intrinsically computed within the motor system, consistent with the theory of embodied cognition. To investigate the role of the motor system, we tested human subjects (n = 40) on a novel task combining reaching and time estimation. In this task, subjects were required to move a robotic manipulandum to one of two physical locations to categorize a concurrently timed suprasecond. Critically, subjects were divided into two groups: one in which movement during the interval was unrestricted and one in which they were restricted from moving until the stimulus interval had elapsed. Our results revealed a higher degree of precision for subjects in the free-moving group. A further experiment (n = 14) verified that these findings were not due to proximity to the target, counting strategies, bias, or movement length. A final experiment (n = 10) replicated these findings using a within-subjects design, performing a time reproduction task, in which movement during encoding of the interval led to more precise performance. Our findings suggest that time estimation may be instantiated within the motor system as an ongoing readout of timing judgment and confidence.

Keywords: decision-making; movement; perception; time perception.

Copyright © 2019 Wiener et al.

Figures

Figure 1.

Task design for the three experimental types. In each experiment subjects held the handle of a robotic manipulandum and moved a screen cursor to one of two targets to classify a tone as short (left target) or long (right target). A, In Experiment 1 (free-movement), subjects were held in place in the starting position, while a ready cue was presented. Following this, the arm was released for a warm-up period, in which subjects could freely move. Onset of the tone proceeded for one of seven possible durations (log-spaced), and subjects were free to move to one of two target locations, but could only enter a location after tone offset occurred. B, In Experiment 2 (hold) no warm-up period was provided, and subjects were only released once the tone duration was complete. C, In Experiment 3 (different hold positions) subjects performed the hold experiment, but with one of two possible starting positions, in different blocks.

Figure 2.

Behavioral data from free-movement and hold experiments. A, Middle, The average proportion of trials on which subjects classified durations as long as a function of tested duration. Psychometric curves represent fits to the average data. Red symbols represent data from Experiment 2 (hold); blue symbols represent data from Experiment 1 (free-movement). Bottom, Average chronometric data, with reaction time as a function of duration. Reaction time was notably similar for both groups for durations under 2 s, but then became increasingly faster for free-movement subjects. Top, Individual bisection points, derived from fitted curves. Boxplots display the mean (red line), bounded by the 95% confidence interval (red shaded region) and the SD (blue region). B, Top, CV data for both groups, demonstrating significantly lower variability for the free-movement experiment than the hold experiment. Bottom, The average movement length for subjects in the free-movement experiment for 2 s trials classified as long or short; subjects chose long more often when they moved more during the 2 s tone interval. Dashed line represents the identity. Results remained significant with removal of the outlier. Error bars represent SE.

Figure 3.

Movement trajectories for an example subject in Experiment 1 (free-movement). Top row, Layout of the start position (green filled circle), and short (blue filled circle) and long targets (red filled circle) is the same as in Figure 1C. Distance is in meters. Each panel represents the total number of movements for one of seven intervals used in our temporal bisection task. Color-coding of traces reflects hand movement through the progression of the given trial (black, warm-up period; green, during the tone duration; red, after the tone until the decision). As can be observed, for relatively short durations (<2000 ms), the majority of movements smoothly progress from the starting position at the bottom of the panel to the short target location (left target). As durations increase (1580, 2000, and 2520 ms), choices split between the short and long locations (note the directions of the red traces). Notably, for relatively long durations (>2000 ms), subject arm movements initially move toward the short location, but then shift toward the long location (right target) once the interval has elapsed past a certain point. Bottom row, Density estimates for arm location only during the interval portion of each trial (brighter pixels represent a greater proportion of the trial occupied in that location), displaying the shift to the long duration target with longer intervals.

Figure 4.

Movement trajectories are idiosyncratic between subjects, yet consistent within subject. Layout of the start position (green filled circle), and short (blue filled circle) and long targets (red filled circle) is the same as in Figure 1C. Color-coding of traces reflects hand movement through the progression of the given trial (black, warm-up period; green, during the tone duration; red, after the tone until the decision). Top row, Trajectory data for 2000 ms trials for three example subjects from Experiment 1 (free-movement). Classification of a short duration is the left target; a long classification is the right target. Each subject employed a separate strategy, with one subject (S4) moving in a circular pattern between short and long locations, another (S1) rotating in a leftward arc before moving in between both targets, and a third (S2) moving in an up–down fashion before shifting from the middle to a target location. Bottom row, Three example subjects for 2000 ms trials from Experiment 2 (hold). Changes-of-mind are evident, including a shift from long to short (S1), short to long (S3), and both (S2).

Figure 5.

Changes of mind occur in the appropriate direction. Displayed are the mean proportion of changes-of-mind that occurred for each interval in our stimulus set for the hold group for each transition direction. Accordingly, when subjects switched from the long to the short target, it was generally for intervals

Figure 6.

The decision process is represented within the movement trajectories and force patterns. A, Position and force data for both experiments as a function of tone duration. Red symbols represent data from Experiment 2 (hold); blue symbols represent data from Experiment 1 (free-movement). The left axis displays the ratio of the Euclidean distance from arm position to the long and short targets for the free-movement group; values >1 (horizontal blue dashed line) indicate the subject was closer to the short location, and vice versa. The right axis displays the force direction, in degrees, that subjects were exerting on the arm at tone offset. Values above 90° (horizontal red dashed line) indicate the subject was pushing toward the short target (105°), whereas below 90° indicate pushing toward the long target (75°). B, Intertrial variability (SD) of the arm position was measured for the final 100 ms of arm movements across trials for each duration. Free-movement subjects (blue trace) demonstrated linearly increasing SD with tone duration, whereas hold subjects (red trace) displayed linearly decreasing SD. Note the axis break and scale difference in the y-axis, demonstrating that hold subjects exhibited greater variability overall compared with free-movement subjects. Error bars represent SE.

Figure 7.

Differences in movement dynamics across tasks. Movement dynamics for the free-movement group (A) and hold group (B) during the presentation of each interval. Data are shown for the last 1000 ms of each presented interval. Dashed lines represent the 50% inflection point between the long and short duration targets, with values above this line representing a closer relative location (top) or a closer force direction (bottom) to the short target location. All displayed data are sigmoid transformed to reduce the influence of extreme values on logistic regression (see Materials and Methods). For both sets, movement dynamics display a transition from proximity to the short duration target to the long duration target as the interval elapses; additionally, differences in position or force could be distinguished well before the interval offset. C, Intertrial variability of movement dynamics displayed in free-movement and the hold group (D). Both panels display the average within-subject SD of movement data from their corresponding left panels. For both groups, intertrial variability increases as the presented interval elapses, but with different profiles. Further, variability is dramatically lower for the free-movement group compared with the hold group (note the difference in the vertical scale). Shaded regions display SE.

Figure 8.

Movement dynamics predict eventual choice. All three panels display data for the middle interval of the stimulus set (2000 ms). A, Normalized movement ratios for the free-movement group. Eventual choice could be determined ∼500 ms before interval offset. B, Normalized force direction for the hold group. In contrast to the free-movement group, no distinguishable difference in direction was observed between short and long response choices. C, Normalized force direction for the hold group during the interval before stimulus onset. Here, a distinguishable difference was detected in the final 100 ms before target onset; note that the direction subjects are exerting force is pointing in the opposite direction of the choice they will eventually make on that trial. Shaded error represents SE. Gray shaded regions indicate significance at p < 0.05 for logistic regression.

Figure 9.

Comparison of behavior for different starting positions. Behavioral results from Experiment 3 (different hold locations) are displayed for the (A) bisection point, (B) CV, and (C) average reaction time across all durations. Red symbols represent individual subjects at the close hold location; blue symbols represent the long hold location. No significant difference between starting location was found for either the bisection point or coefficient of variation, but a significant difference was found for reaction time, with subjects responding sooner when the starting location was closer to the target locations.

Figure 10.

Movement improves temporal reproduction. A, Task schematic for the temporal reproduction task. Subjects (n = 10) performed alternating blocks of trials in which they were presented with an auditory stimulus for one of seven possible intervals (1–4 s; encoding), and then required to reproduce that interval (reproduction). In the hold condition, subjects were held in one of six possible starting locations during encoding and the subsequent ISI. In the free-movement condition, subjects were allowed to freely move the arm during the encoding phase, but were forced back to one of the six possible locations when the interval ended. Following the ISI, in either condition, subjects were required to reproduce the interval by moving the robotic arm for the same amount of time as the tone interval they had just heard. No feedback regarding cursor location was presented. Lines within demonstrate seven random trials from a representative subject, with darker lines indicating trials on which shorter intervals were presented. Green circles indicate the starting locations of the respective phases in these trials. B, Temporal reproduction performance. Gray lines indicate individual subject performance; bold lines indicate mean reproduced intervals, with a linear regression fit to these values. Subjects reproduced all intervals well, with a notable central tendency effect between across the stimulus set. A significant difference in the slope of reproduced to presented intervals was found, with free-movement trials exhibiting more central tendency (inset, individual slope values between conditions). C, CV values between conditions. Individual points represent CV values for each of the seven intervals across all 10 subjects. Significantly higher CV values were found across all seven tested durations in the hold condition compared with free-movement, indicating better precision in these trials.

Figure 11.

Movement strategies for subjects in the temporal reproduction task. Similar to the temporal bisection task, subjects used a number of different strategies, yet were consistent with the strategy chosen. Three representative subjects are shown with distinct movement strategies across the three moving conditions. Layout is the same as in Figure 10; green circles indicate the starting position. Traces displayed above are from seven trials chosen at random for each subject, with one for each of the seven intervals tested; lighter green colors indicate longer interval trials. Left column, Reproduction movements from trials on which the subject was held in place during the estimation phase. Middle and right columns, Movements made during the estimation and reproduction phases in the free-movement condition. Each subject again adopted a different strategy, such as a random pattern (S2), an arc movement (S5), or a patterned movement consisting of sharp turns (S8).