Express Visuomotor Responses Reflect Knowledge of Both Target Locations and Contextual Rules during Reaches of Different Amplitudes

Abstract

When humans reach to visual targets, extremely rapid (∼90 ms) target-directed responses can be observed in task-relevant proximal muscles. Such express visuomotor responses are inflexibly locked in time and space to the target and have been proposed to reflect rapid visuomotor transformations conveyed subcortically via the tecto-reticulo-spinal pathway. Previously, we showed that express visuomotor responses are sensitive to explicit cue-driven information about the target, suggesting that the express pathway can be modulated by cortical signals affording contextual prestimulus expectations. Here, we show that the express visuomotor system incorporates information about the physical hand-to-target distance and contextual rules during visuospatial tasks requiring different movement amplitudes. In one experiment, we recorded the activity from two shoulder muscles as 14 participants (6 females) reached toward targets that appeared at different distances from the reaching hand. Increasing the reaching distance facilitated the generation of frequent and large express visuomotor responses. This suggests that both the direction and amplitude of veridical hand-to-target reaches are encoded along the putative subcortical express pathway. In a second experiment, we modulated the movement amplitude by asking 12 participants (4 females) to deliberately undershoot, overshoot, or stop (control) at the target. The overshoot and undershoot tasks impaired the generation of large and frequent express visuomotor responses, consistent with the inability of the express pathway to generate responses directed toward nonveridical targets as in the anti-reach task. Our findings appear to reflect strategic, cortically driven modulation of the express visuomotor circuit to facilitate rapid and effective response initiation during target-directed actions.

SIGNIFICANCE STATEMENT Express (∼90 ms) arm muscle responses that are consistently tuned toward the location of visual stimuli suggest a subcortical contribution to target-directed visuomotor behavior in humans, potentially via the tecto-reticulo-spinal pathway. Here, we show that express muscle responses are modulated appropriately to reach targets at different distances, but generally suppressed when the task required nonveridical responses to overshoot/undershoot the real target. This suggests that the tecto-reticulo-spinal pathway can be exploited strategically by the cerebral cortex to facilitate rapid initiation of effective responses during a visuospatial task.

Introduction

Target-directed actions require knowledge of both the hand and target positions (Sabes, 2011; Proske and Gandevia, 2012). To catch a falling object, for example, the sensed hand-to-target distance must be transformed into accurate motor commands to generate the muscle force and, in turn, accelerate the joints so that the object can be intercepted before it hits the ground. Greater activation of agonists and inhibition of antagonist muscles are, therefore, required to enhance the limb acceleration.

Historically, target-directed visuomotor behavior was thought to be the exclusive domain of the cerebral cortex. This, however, is challenged by mounting evidence showing that human limb muscles start responding to visual targets for reaching at latencies (70-120 ms) that leave little time for cortical visuomotor transformation (Goonetilleke et al., 2015; Glover and Baker, 2019; Gu et al., 2019; Billen et al., 2023; Selen et al., 2023). Notably, the onset time of these express visuomotor responses is far less variable than the mechanical reaction time (RT) (Contemori et al., 2022), which depends mostly on the long-latency (>120 ms; plausibly cortically driven) muscle response components. Express visuomotor responses are also inflexibly tuned to reach the real target, even when a nonveridical response is required, such as in the anti-reach task (Gu et al., 2016). Given their temporal and spatial stimulus-locked attributes, express visuomotor responses were proposed to be conveyed subcortically via the tecto-reticulo-spinal pathway (Pruszynski et al., 2010).

Delineation of the factors that influence express visuomotor responses should provide clues about their origin and relationships to well-studied (putatively transcortical) visuomotor pathways. Previous work showed that the requirement to avoid rapid target-directed responses impaired the generation of express visuomotor responses (Pruszynski et al., 2010; Wood et al., 2015; Atsma et al., 2018). More recent work showed that express visuomotor responses are modulated by explicit cues about the temporal (Contemori et al., 2021a) and spatial (Contemori et al., 2021b) presentation of visual stimuli, and incorporate advance expectations about the required movement to reach the target (Gu et al., 2018; Contemori et al., 2022). In all, these findings appear to reflect cortically driven modulation of the putative subcortical express circuit. Here we asked whether express visuomotor responses are modulated compatibly with the required movement amplitude to accomplish a visuospatial task. If so, it would suggest that the circuits responsible for express limb activity produce control signals that account for the details of reach metrics, rather than merely the initial reach direction.

We conducted two experiments to explore express visuomotor responses to targets that required different movement amplitudes via modulation of: (1) the physical hand-to-target reaching distance; and (2) explicit instruction to overshoot, undershoot, or stop at the target. Experiment 1 showed that express visuomotor responses were facilitated by increasing the hand-to-target distance, suggesting that the express system encodes both the direction and distance metrics of veridical target-directed reaches. Experiment 2 showed significantly fewer and smaller express visuomotor responses, and longer RTs, for both overshooting and undershooting tasks compared with veridical target-directed reaching actions. This suggests that express visuomotor behavior is generally inhibited in circumstances requiring sensory-to-motor transformation for abstract targets, a task that is probably incompatible with the stimulus-locked output of the putative subcortical express circuit (Gu et al., 2016). The findings support the idea that the cerebral cortex strategically exploits the express pathway when its motor output is functional for rapid initiation of veridical target-directed actions, but suppresses the express network when it is incapable of meeting the current task demands.

Materials and Methods

Participants

Fourteen adults completed Experiment 1 (6 females; mean age: 30.9 ± 9 years), and 12 of them also participated in the Experiment 2 (4 females; mean age: 31.8 ± 9.2 years). The sample size was selected to be comparable with previous studies investigating express visuomotor responses (Pruszynski et al., 2010; Goonetilleke et al., 2015; Wood et al., 2015; Atsma et al., 2018; Gu et al., 2019; Kozak et al., 2020; Contemori et al., 2021a; Billen et al., 2023; Kearsley et al., 2022; Selen et al., 2023). All participants were right-handed, had normal or corrected-to-normal vision, and reported no current neurologic or musculoskeletal disorders. They provided informed consent and were free to withdraw from the experiment at any time. All procedures were approved by the University of Queensland Medical Research Ethics Committee (Brisbane, Australia) and conformed to the Declaration of Helsinki.

Experimental setup and task design

Experimental setup

For both experiments, the participants performed visually guided target-directed reaches using a two-dimensional planar robotic manipulandum (the vBOT, Fig. 1A) (Howard et al., 2009). In the vBOT setup, the visual feedback is provided via an LCD computer monitor (120 Hz refresh rate) mounted above the robot handle and projected to the participant via a mirror, which occludes direct vision of the arm (Fig. 1A). The visual stimuli were created in Microsoft Visual C++ (version 14.0, Microsoft Visual Studio 2005) using the Graphic toolbox. The hand position was virtually represented by a blue cursor (∼1 cm in diameter) whose apparent position coincided with actual hand position in the plane of the limb. During the experiments, the upper arm was supported on a custom-built air sled positioned under the right elbow to minimize sliding friction (Fig. 1A).

Figure 1.

A, Experimental setup. Participants' hand positions were virtually represented via a cursor (B–D, blue dot) displayed on the monitor and projected into the (horizontal) plane of hand motion via a mirror. The head position was stabilized by a forehead rest (not shown here). B, Schematic diagram of the timeline of events in the emerging target paradigm. Once the cursor was at the home position, the + sign for fixation was presented underneath the barrier. After 1 s of fixation, the target started dropping from the stem of the track at constant velocity of ∼30 cm/s until it passed behind the barrier (i.e., occlusion epoch) for ∼480 ms, and reappeared underneath it at ∼640 ms from its movement onset time. C, Task conditions in Experiment 1. In the control condition, the right and left potential target locations (unfilled gray circles underneath the barrier) had equal distance from the reaching hand. In the long-reach condition, the target required a longer reach relative to control condition; whereas in the short-reach condition, the target appeared closer to the reaching hand relative to control conditions. For all conditions, the target moved initially toward the fixation spot. In these examples, shifting the visual elements toward the left required long and short reaching distances to address the location of left and right targets, respectively. By contrast, rightward shifts of the visual element generated the opposite direction × distance conditions of reaching. D, Task conditions in Experiment 2. Here, the right and left target potential target locations (unfilled gray circles underneath the barrier) had equal distance from the reaching hand across all conditions. In the control condition, the vertical lines underneath the barrier were colored black and the hand had to stop at the actual target location. By contrast, the hand had to overshoot or undershoot the actual target location when the vertical lines underneath the barrier were green (i.e., overshooting condition) or red (i.e., undershooting condition), respectively.

In both experiments, the target was a filled black circle of 3 cm in diameter presented against a light gray background (target luminance ∼0.5 cd/m², background luminance ∼150 cd/m²; Cambridge Research System ColorCAL MKII). This created a high-contrast (Wood et al., 2015) and low spatial-frequency stimulus (Kozak et al., 2019), both features that have proven effective to facilitate express visuomotor responses and rapid correction of ongoing movements (Veerman et al., 2008; Kozak et al., 2019). The target was presented via an emerging moving target paradigm (Fig. 1B) (Kozak et al., 2020; Contemori et al., 2022; Kearsley et al., 2022). To start the trial, the participants had to align the cursor and gaze at a “home” position (a blue ring of ∼2 cm in diameter) located at the center of the monitor and aligned with the mid-body line. At this point, the ring changed to a + sign that defined the gaze fixation spot. For Experiment 1, the fixation spot position was not always coincident with the starting hand position, but rather changed as a function of the trial condition to ensure equal eccentricity for the left and right targets (for details, see Experiment 1: task design; Fig. 1C). Simultaneously, a constant rightward load of ∼5N was applied to enhance the activity of the shoulder transverse flexor muscles, including the clavicular head of the pectoralis major muscle, which was shown to facilitate the generation of detectable express visuomotor responses (Wood et al., 2015). At the same time, we displayed the target close to the top of the monitor and within a vertical track (Fig. 1B). After ∼1 s of fixation, the target fell at constant velocity (∼30 cm/s) toward the fixation spot, disappeared behind the barrier, and reappeared underneath it by making one single flash of ∼8 ms of duration at the right or left of participants' right hand and fixation spot (Fig. 1B,C). The participants, therefore, were presented with transient and temporally predictable targets, both attributes that facilitate express visuomotor responses (Contemori et al., 2021a).

The participants were instructed to not break fixation until the target emerged from behind the barrier and to start moving the hand toward the target as rapidly as possible. For both experiments, horizontal gaze-on-fixation was checked online with bitemporal, direct current EOG. The EOG signal was sampled at 1 kHz, amplified by 1000, and filtered with a 3-3000 Hz bandwidth filter by a Grass P5 AC Series amplifier (Grass Technologies Product Group, Astro-Med). “Fixation” or “Too fast” errors were shown if the participants did not respect the gaze fixation requirements or moved before the target presentation, respectively, and the trial was reset. The time at which the stimulus was displayed on the monitor was recorded with a photodiode that detected a secondary light appearing at the bottom-left corner of the monitor and simultaneously with the actual target. The photodiode fully occluded the secondary light, thus making it invisible for the participants.

Experiment 1: task design

In Experiment 1, we investigated whether express visuomotor responses are modulated by the physical hand-to-target reaching distance. To this aim, we varied the target distance from participants' reaching hand to create: (1) a control-reach condition, when the hand-to-target distance (∼8 cm) was equivalent for both right and left targets; (2) a long-reach condition, when the hand-to-target distance was longer (∼13 cm) than control; and (3) a short-reach condition, when the hand-to-target distance was shorter (∼3 cm) than control. The hand-to-target distance was modulated by shifting the target, track, and visual barrier ∼5 cm rightward, or leftward, relative to the static home position of the hand. Therefore, distinct long and short reaches were required for left and right targets (e.g., leftward shift → left-long/right-short reaches; Fig. 1C). The shift of the visual elements happened >1 s before the target presentation to ensure unambiguous interpretation of the trial context. It is also important to note that the between-target distance (∼16 cm) was kept constant, and the fixation point was shifted by ∼4 cm such that the target had the same visual eccentricity across conditions.

To control the oculomotor behavior, the EOG was calibrated before the main experiment by asking the participants to look at a target located at the center of the monitor (consistent with the fixation spot location in control conditions; Fig. 1C) for ∼10 s. Then the target jumped laterally right/left at three different distances (i.e., 6 direction × distance conditions), stayed there for ∼2 s before returning back to the initial one, and made another jump only after another ∼5 s. For consistency with the main experiment, the target was a filled black circle 3 cm in diameter presented against a light gray background and jumped ±8, ±13, and ±3 cm relative to the starting central position. The target jumped laterally 5 times for every direction and distance condition (i.e., 30 total trials). Importantly, this procedure allowed us to define the within-subject absolute EOG signal values across different eye positions and, thereby control the gaze fixation online.

For the main experiment, each participant completed 6 blocks of 48 reaches/block (24 for each direction), with each block consisting of 16 control-reach, 16 long-reach, and 16 short-reach trials, randomly intermingled.

Experiment 2: task design

Experiment 1 showed modulations of express visuomotor response as a function of the reaching distance (for details, see Experiment 1 results). This could indicate that the physical hand-to-target distance was encoded along the express sensorimotor circuit. Alternatively, the data might reflect context-based preparation of long, or short, movements regardless of the real target distance from the reaching hand. Although these alternatives are not mutually exclusive, we ran a second experiment asking the participants to execute movements of different amplitudes as a function of the explicit instruction to: (1) stop at the target (control); (2) overshoot the target; and (3) undershoot the target (Fig. 1D). The control condition replicated that of Experiment 1 as the participants had to stop at the actual target location within the two vertical black lines underneath the barrier (Fig. 1D: control condition). For the overshoot condition, we displayed green vertical lines underneath the barrier and instructed the participants to overshoot the actual target location by ending the movement at least beyond the outermost vertical green line (Fig. 1D: overshoot condition). For the undershoot condition, we used red lines beneath the barrier and asked the participants to undershoot the actual target location by ending the movement before the innermost vertical red line (Fig. 1D: undershoot condition). On every trial, the target always appeared at ∼8 cm to the right or left of participants' right hand. The design of Experiment 2 did not require distinct movement amplitudes for different target locations (e.g., right-overshoot vs left-undershoot). The motivation for providing advance and equal task instructions for both the right and left targets was to dissociate the executed reach from the target location without adding complexity for the trajectory-endpoint decision at the time of target presentation. To this aim, and consistent with Experiment 1, the trial condition (i.e., the color of the lines underneath the barrier) was made explicit to the participants for >1 s before the target presentation.

Each participant completed 6 blocks of 48 reaches/block (24 for each direction), with each block consisting of 16 control-reach, 16 overshoot-reach, and 16 undershoot-reach trials, randomly intermingled.

Data recording and analysis

Kinematic data recording and analysis

The kinematic data of the vBOT handle were recorded via two optical encoders at a sampling rate of 1 kHz. To define the mechanical RT, we adopted the “extrapolation” technique (Veerman et al., 2008; Oostwoud Wijdenes et al., 2014; Zhang et al., 2018a,b) as it returns reliable RT measurements, even in circumstances requiring short movements evolving at low velocities (Brenner and Smeets, 2019). Briefly, we defined the first peak of the radial hand velocity after the time point at which it first exceeded the baseline value (i.e., average velocity recorded in the 100 ms preceding the target onset time) by >5 SDs. We then fitted a line to the hand velocity data enclosed between 25% and 75% of the peak velocity and indexed the RT as the time at which this line crossed the baseline velocity value. Trials with RT <160 ms (∼5%) or >500 ms (<1%) were excluded during offline analysis.

To determine the response correctness, we measured the initial reach direction by adopting a procedure previously described by Contemori et al. (2022). Briefly, we compared the initial hand-trajectory direction (i.e., slope of a line connecting the hand position coordinates at the RT and the 75% of the peak velocity) with the actual target location. We then computed the movement endpoint by searching for the point in time at which the total hand velocity fell below 0.5 m/s after having reached its peak value. We reasoned that a trial was correct if the hand initially moved toward the actual target and ended at the location specified by the trial condition.

For correct trials, we computed the movement time (i.e., RT-to-endpoint time), and the time to reach the maximal velocity. We also conducted a trial-by-trial temporal normalization for the whole movement duration to test whether the movement evolved similarly across conditions despite task-dependent differences in movement time. This allowed us to index the point (%) within the movement at which the hand velocity reached its peak. For both experiments, the kinematic data were averaged across the left and right directions to limit potential biases related to the leftward preloading robot force.

EMG data recording

Surface EMG activity was recorded from the clavicular head of the right pectoralis muscle (PMch) and the posterior head of the right deltoid muscle (PD) with double-differential surface electrodes (Delsys Bagnoli-8 system). The quality of the EMG signal was checked offline with an oscilloscope by asking the participants to flex (PMch activation-PD inhibition) and extend (PMch inhibition-PD activation) the shoulder in the transverse plane. The sEMG signals were amplified by 1000, filtered with a 20-450 Hz bandwidth filter by the Delsys Bagnoli-8 Main Amplifier Unit, and sampled at 2 kHz using a 16-bit analog-digital converter (USB-6343-BNC DAQ device, National Instruments).

Trial-by-trial, the EMG signal was saved on a secondary computer via a custom MATLAB script that also generated live plots of the recorded data. This gave us the opportunity to interrupt the experiment in case the EMG signal deteriorated (e.g., loss of electrode-on-skin contact). The sEMG data were then downsampled to 1 kHz and full-wave rectified offline.

Detection of muscle response onset time

To detect the earliest stimulus-driven muscle response, we adopted a single-trial analysis method, named the detrended-integrated signal method, that we recently developed and validated (Contemori et al., 2022). Briefly, we initially computed the integral of the full-wave rectified EMG signal recorded between 100 ms before and 300 ms after the target onset time. We then computed the linear regression function of integrated EMG signal enclosed in the background period (from 100 ms before to 70 ms after the stimulus presentation) and subtracted this function from the entire 400 ms window, thus detrending the integrated EMG trace. We then computed the average and SD values of the detrended-integrated signal in the background epoch. We indexed the “candidate” muscle response onset time as the first time the detrended-integrated signal exceeded the background average value by more (i.e., signature of muscle activation), or less (i.e., signature of muscle inhibition), than 5 SDs.

We previously showed that the occurrence of false-positive express muscle response detection (i.e., candidate onset times earlier than 70 ms after the target presentation) is lower than 5% by using a 5 SD threshold (Contemori et al., 2022). Here, we also tested the occurrence of muscle responses on an earlier time window at 20-60 ms from the target presentation. No muscle response was detected in this “pre-express” time window; neither with 5, 4, or 3 SDs as threshold for the candidate response onset time. We are therefore confident using 5 SDs as the threshold to index the candidate onset time of express visuomotor responses.

Critically, the candidate response onset time does not exactly correspond to the earliest deflection-from-background of the signal. To find this time point, we ran a linear regression analysis around the candidate muscle response onset time and indexed the time at which the linear trendline intercepted the zero value of the detrended-integrated signal (for details, see Contemori et al., 2022, their Fig. 3). A muscle response was classified as “express” if it was initiated within 70-110 ms after the target presentation. By contrast, the muscle responses initiated later than 110 ms were classified as “long-latency.” We used a shorter express time window relative to previous work (i.e., 70-120 ms) (Gu et al., 2016; Contemori et al., 2021a,b, 2022) to prevent contamination of the express epoch by the long-latency (plausibly cortically mediated) EMG activity of faster trials. Further, we found that the delay between the onset time of long-latency muscle response and RT of the corresponding trials was on average 40 ms. Thus, even for the earliest RT trials included in the data analyzed (160 ms RT cutoff; see Kinematic data analysis), the long-latency EMG response should have started >110 ms from the target presentation. Importantly, this allowed us to minimize the risk that rapid muscle responses from the long-latency phase contaminated the signal enclosed in the express epoch.

Figure 3.

Experiment 1 condition-dependent variations of the RT (A), maximal velocity (B), movement time (C), time to the maximal velocity (D), index of the maximal velocity within the movement (E), and variability of the trajectory endpoint (F). On each plot, thin light gray lines indicate the 14 subjects who completed Experiment 1. Thick black dotted line indicates the average across subjects. Horizontal thick dark gray lines on top of the subjects indicate the between-condition statistically significant differences: (A) RT, short-reach versus control and long-reach, p < 0.001; control versus long-reach, p = 0.1; (B) movement time, short-reach versus control, p = 0.016; long-reach versus short-reach and control, p < 0.001; (C) maximal velocity, p < 0.001 for all between-condition contracts; (D) time to maximal velocity, p < 0.001 for all between-condition contracts; (F) endpoint trajectory variability, short-reach versus control, p = 0.96; short-reach versus long-reach p = 0.045, control versus long-reach, p = 0.008.

Identifying participants exhibiting stimulus-locked express visuomotor responses

One of the most distinctive attributes of express visuomotor responses is that their onset time is more locked to the target presentation time than the mechanical RT (Pruszysnki et al., 2010; Wood et al., 2015; Kozak et al., 2019, 2020; Contemori et al., 2021a,b, 2022; Kozak and Corneil, 2021). Critically, the broad range of delays for the long-latency motor signal to reach the RT detection threshold is consistent with polysynaptic nature of cortical sensorimotor networks to transform sensory inputs into deliberate decisions for actions. By contrast, the strikingly short-latency and relative temporal consistency of express visuomotor responses imply a small range of delays in motor signal conduction time, consistent with the few synapses of the tecto-reticulo-spinal pathway. To test the extent to which the express visuomotor response onset times were independent from the RT, we adopted a procedure previously described by Contemori et al. (2022). We first selected the trials showing an express muscle response and the gathered the corresponding RTs. We then divided these trials into “express-fast” and “express-slow” subsets based on whether the associated RT laid above or below the median RT of the full class of express trials. We then computed the average express responses initiation time of the express-fast and express-slow trials as well as the average RT of the corresponding fast and slow trial bins. Finally, we fitted a line to the express-fast and express-slow average data to test whether the muscle response onset time did not covary with the RT (i.e., line slope >67.5 deg) (for details, see Contemori et al., 2022; see also Contemori et al., 2021a, their Fig. 3; Contemori et al., 2021b). Participants with express response onset times that did not covary with the RT for both the right and left trials and among all task conditions were classified as an express visuomotor response producer (for details, see Results). For these subjects, we computed the condition-dependent express response initiation time by averaging this metric across the express visuomotor response trials and then across the right and left target locations. We also computed the condition-dependent express response detection rate by averaging the percentage of express visuomotor response trials within the dataset across the two target locations. Further, we quantified the condition-dependent express response magnitude by computing the average EMG activity recorded in the 10 ms after the response initiation time for each rightward and leftward trial exhibiting an express visuomotor response. We then averaged this metric across the express response trials and computed the difference between the left and right targets (Contemori et al., 2022).

Test whether express visuomotor responses reflect contextual visuomotor behavior

We and others previously showed that larger express visuomotor responses are associated with earlier RTs (Pruszysnki et al., 2010; Gu et al., 2016; Contemori et al., 2021a). Here, we found that express visuomotor responses were facilitated in task conditions that also facilitated the reach onset time (for details, see Results). Further, modulating the reaching amplitude correlated with task-dependent variation in movement velocity and, thereby, the long-latency muscle response magnitude (LLRM; for details, see Results) that was defined, trial-by-trial, by taking the average EMG signal from 5 ms before 5 ms after the RT.

Although we minimized the risk of contamination of the express epoch from the long-latency EMG signal (for details, see Detection of express visuomotor response), we also verified whether the task-dependent modulation of the express response reflected the contextual visuomotor behavior. To this aim, we tested express visuomotor responses on data samples with matched RTs across conditions by adopting a trial-matching procedure akin to that used by Dash et al. (2018) and Kozak et al. (2019). Further, we also retested express visuomotor responses on data samples with matched LLRM across conditions. We reasoned that, if the express visuomotor response reflected task-dependent modulations of the express circuit, then similar between-condition contrasts should be observed in both original, RT-matched, and LLRM-matched datasets. These trial subsets were generated for each participant who exhibited express visuomotor responses across all the three task conditions (see Results). We first defined the range of RT and LLRM values by pooling all the correct trials across the three task conditions. We then verified the presence of at least one trial per condition for each RT ±2 ms value of the full data sample and repeated this procedure for each LLRM ±5 µV value. The ±2 ms and ±5 µV tolerances were applied to be conservative on the number of nonmatching RT trials to discard, which would otherwise increase by searching for perfect value-match between conditions. Participants were excluded from this analysis if this procedure discarded >50% of the original trials in one, or more, of the three task conditions. These procedures generated three condition-specific datasets having similar distributions of the variables of interest, but different numbers of trials across conditions. To create compatible datasets, we binned the RT-matched trials every 20 ms from the smallest RT value, and then binned the LLRM-matched trials every 20 µV from the smallest LLRM value. For all task conditions, we then resampled with replacement the binned trials 100 times by using a bootstrapping approach. For each bin of trials, we selected the same number of trials per condition based on the lowest number of trials across conditions for that bin in the original dataset. Finally, we reran the detrended-integrated signal analysis methods on the RT-matched and then on the LLRM-matched datasets.

Statistical analysis

To test the statistical differences across conditions, we ran repeated-measures ANOVA, unless otherwise stated, as the normality of the distributions was verified by the Shapiro–Wilk test. Specifically, for the kinematics variables, we ran the repeated-measures ANOVA on the mechanical RT, movement time, maximal hand velocity, time to maximal hand velocity, percentage of the movement at which the maximal hand velocity was reached, and variability of the movement endpoint. For the EMG, the repeated-measures ANOVA was run on the detection rate, onset time, and magnitude of express muscle responses, as well as on the LLRM. The repeated-measures ANOVAs were conducted in SPSS (IBM SPSS Statistics for Windows, version 25) with Bonferroni correction and task condition (three levels: Experiment 1, control long-reach, short-reach; Experiment 2, control, overshoot, undershoot) as within-participant factors. When the ANOVA revealed a significant main effect, we estimated the effect size by computing the Partial η squared (${\eta }_p^2$) and ran Bonferroni tests for post hoc comparisons. The detectable effect size with our smallest sample size (N = 10; see the EMG results of Experiment 2) and statistical power of 0.8 was estimated to be medium-to-large (effect size f 0.44; G*Power, version 3.1.9.4, Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany). For all tests, the statistical significance was designated at p < 0.05.

Results

Experiment 1

Kinematic results

In Experiment 1, the participants reached to visual targets that could appear at different rightward or leftward distances from their dominant hand (for details, see Experiment 1 task design). They successfully achieved the task goal in >90% of the trials across the three experimental conditions.

Figure 2A shows exemplar correct hand-to-target trajectories of a participant who completed Experiment 1. For this subject, the targets requiring short reaching distances resulted in longer RT relative to control and long-reach conditions (Fig. 2B, dashed vertical lines). After its initiation, the movement evolved at faster and slower velocities than control for the long-reach and short-reach conditions, respectively (Fig. 2B, dotted vertical lines). The task-dependent variation in maximal velocity did not fully compensate that in reaching distance, thus leading to longer movement times to complete longer than shorter reaches (i.e., RT-to-endpoint time; Fig. 2B). The participants, however, were not required to complete the movement within a specific time (see Materials and Methods). Also, the velocity profiles were symmetrically bell-shaped regardless of peak velocity such that the maximum hand velocity was reached at around the movement half across all conditions (Fig. 2C).

Figure 2.

Kinematic data of an exemplar participant for Experiment 1. A, Hand trajectories in the control (black traces), long-reach (green traces), and short-reach (red traces) conditions. B, Condition-dependent hand velocity traces. The time is relative to the target presentation. Vertical dashed and dotted lines indicate the mechanical RTs and maximal velocities across conditions, respectively. C, Time-normalization of the hand velocity traces for the entire movement duration and point of the movement at which the peak velocity was reached. Data are plotted as mean (solid lines) and SD (shaded area around the mean lines).

For the entire group, the repeated-measures ANOVA showed statistically significant task condition (control vs long-reach vs short-reach) main effects for RT (F_(2,12) = 20, p < 0.001, ${\eta }_p^2$ = 0.6), movement time (F_(2,12) = 23.1, p < 0.001, ${\eta }_p^2$ = 0.54), maximal hand velocity (F_(2,12) = 366, p < 0.001, ${\eta }_p^2$ = 0.97), time to maximal hand velocity (F_(2,12) = 42.5, p < 0.001, ${\eta }_p^2$ = 0.77), and endpoint movement variability (F_(2,12) = 4.5, p = 0.02, ${\eta }_p^2$ = 0.26). The short-reach target condition led to significantly longer RT (Fig. 3A), significantly shorter movement time (Fig. 3B), and involved significantly lower maximal hand velocities (Fig. 3C) that were reached significantly earlier (Fig. 3D) than the control conditions. By contrast, the long-reach target condition led to the opposite results, except for the RT that was not statistically different from control. When the peak-velocity event was indexed relative to the whole movement duration, however, we did not find statistically significant differences between conditions (F_(2,12) = 2.9, p = 0.07; Fig. 3E). The endpoint of the movement trajectory was significantly more variable for the long-reach than the other conditions (Fig. 3F), plausibly reflecting a trade-off between speed and accuracy to accomplish the task.

Overall, these results indicate that the participants were biased by the hand-to-target distances such that they took more time to start moving toward targets appearing close to their hand. Once the movement started, the hand velocity was modulated according to the hand-to-target distance but the greater hand speeds for longer reaches were insufficient to complete the task within the same time across conditions. Nevertheless, the hand was always accelerated for approximately half the movement distance before being decelerated to stop at the target, resulting in similar movement profiles for all hand-to-target distances.

EMG results

Figure 4 shows EMG data recorded from the PMch of an exemplar participant who met the conditions for positive express visuomotor response detection (see Materials and Methods) across all conditions of Experiment 1. In the first two columns of raster plots of Figure 4, express visuomotor responses appear as a vertical band of either muscle activations (left targets) or inhibitions (right targets) at 70-110 ms after the target presentation time. For this subject, the number of trials with an express visuomotor response initiation increased, and that of long-latency responses decreased, by increasing the hand-to-target reaching distance (Fig. 4A–I, red and magenta scatters and bars). Specifically, the detection rate of express visuomotor response was 55%, 77%, and 81% for the short-reach, control, and long-reach conditions respectively. In addition, the average EMG signal enclosed in the express time window (Fig. 4J, gray patch) was smaller for the short-reach condition than the other conditions. The express visuomotor responses onset time, however, was ∼90 ms after the target presentation across all conditions (Fig. 4J).

Figure 4.

Surface EMG activity of the PMch muscle during the leftward and rightward movements executed toward the target requiring short (∼3 cm; A, B), control (∼8 cm; D, E), and long (∼13 cm; G, H) reaching distances of an exemplar participant who completed Experiment 1 and exhibited an express response in each of the three different hand-to-target distance conditions (see Materials and Methods). Each raster plot represents the rectified EMG activity from individual trials sorted by RT. Brighter white colors represent greater EMG activity. White vertical line at 0 ms indicates the target presentation time. Blue scatters represent the RT. Red and magenta scatters represent the express and long-latency muscle response initiation times, respectively. C, F, I, The distribution of express (red histograms) and long-latency (magenta histograms) muscle response onset times as is that of RT (blue histograms) across the left and right target conditions. J, The average EMG activity across all trials. Thick lines indicate left target reaches. Thin lines indicate right target reaches. Gray patch at 70-110 ms represents the average muscle activity enclosed in the express time window. In this panel, the vertical dashed lines indicate the onset time (averaged across right and left target-directed express trials; see Materials and Methods) of express visuomotor response in the three task conditions.

Ten participants (i.e., 71% of the sample) exhibited express visuomotor responses on the PMch in all three conditions of Experiment 1. For these subjects, the repeated-measures ANOVA showed a statistically significant task condition (control vs long-reach vs short-reach) main effect for the detection rate of express visuomotor response (F_(2,8) = 39.6, p < 0.001, ${\eta }_p^2$ = 0.81). The post hoc analysis revealed that the prevalence of express visuomotor responses was significantly lower for the short reach than the other task conditions (Fig. 5A). Although the express response onset time tended to decrease with the hand-to-target-distance (Fig. 5B), we did not find statistically significant contrast between the three conditions (F_(2,8) = 1.26, p = 0.3). The express visuomotor response magnitude was significantly modulated by the hand-to-target distance (F_(2,8) = 11, p < 0.001, ${\eta }_p^2$ = 0.55) as it was significantly smaller for the short reach than the other conditions (Fig. 5C). It is worth noting that these results are unlikely to reflect fixation-dependent differences in target perception, as the left and right targets had equal visual eccentricity among the three task conditions (see Materials and Methods).

Figure 5.

Condition-dependent variations of the express visuomotor response detection rate (A), onset time (B), and magnitude (C). On each plot, each light gray line indicates 1 of the 10 subjects who exhibited the express response behavior across all the three conditions of Experiment 1. Black dots represent the average across subjects. Horizontal dark gray lines on top of the subjects indicate the between-condition statistically significant differences: (A) express response detection rate, short-reach versus control and long-reach, p < 0.001; control versus long-reach, p = 0.13; (C) express response magnitude, short-reach versus control, p < 0.001; short-reach versus long-reach, p = 0.01; control versus long-reach, p = 0.78.

The short-reach condition led to significantly fewer (Fig. 5A) and smaller (Fig. 5C) express visuomotor responses, but also significantly longer RTs (Fig. 3A) and smaller LLRM responses than the other conditions (short-reach 39 ± 15 µV; control 48 ± 19 µV; long-reach 54 ± 21 µV; F_(2,8) = 7.7, p = 0.004, ${\eta }_p^2$ = 0.46; short-reach vs other conditions p < 0.01). We sought to differentiate the reaching-distance effects on express visuomotor response from task-dependent underlying variability of responsiveness. To this aim, we tested the express responses on RT-matched and LLRM-matched data samples (for details, see Materials and Methods). Notably, the between-condition contrasts in express visuomotor response metrics for the RT-matched and LLRM-matched datasets (Table 1) were consistent with those of the original data samples (Fig. 5). This indicates that the express visuomotor response was influenced by the physical distance to reach the target, but not by the time at which the movement was initiated or the LLRM. It is also worth noting that these results are unlikely to reflect fixation-dependent differences in target perception, as the retinal location of the target was kept equal across the hemi visual fields for all task conditions (see Materials and Methods).

Table 1.

Experiment 1: task-dependent variation of express visuomotor response metrics for data samples with matched RT and matched LLRM between conditions^a

Matched variable	Express visuomotor response metric	Task condition			Statistics
					Repeated-measures ANOVA	Post hoc comparisons
		Short reach (SR)	Control (C)	Long reach (LR)		SR vs C	SR vs LR	C vs LR
RT	Detection rate (%)	53 ± 23	58 ± 24	63 ± 16	F(2,6) = 8.4 p = 0.004* ${\eta }_p^2$ = 0.54	p = 0.016*	p = 0.009*	p = 0.11
	Onset time (ms)	95 ± 2	96 ± 3	96 ± 1	F(2,6) = 0.75 p = 0.49	—	—	—
	Magnitude (µV)	25 ± 15	33 ± 16	32 ± 15	F(2,6) = 12 p < 0.001* ${\eta }_p^2$ = 0.63	p = 0.005*	p = 0.009*	p = 0.17
Long-latency muscle response	Detection rate (%)	53 ± 22	68 ± 18	70 ± 21	F(2,8) = 9.15 p = 0.002* ${\eta }_p^2$ = 0.5	p = 0.001*	p = 0.007*	p = 0.67
	Onset time (ms)	96 ± 2	94 ± 3	95 ± 3	F(2,8) = 0.73 p = 0.49	—	—	—
	Magnitude (µV)	24 ± 12	33 ± 13	32 ± 12	F(2,8) = 12.8 p < 0.001* ${\eta }_p^2$ = 0.59	p < 0.001*	p = 0.005*	p = 0.53

^aData are mean ± SD.

*Statistically significant.

Overall, Experiment 1 results show that the express visuomotor response was modulated by the metrics of the visuospatial reaching task, such that targets reachable via small (or large) hand displacements inhibited (or facilitated) the generation of frequent and robust muscle responses within similar express time limits.

Experiment 2

Kinematic results

Experiment 2 tested whether the context-dependent results of Experiment 1 reflected encoding of the veridical hand-to-target reaching metrics, or preparation of movements of different amplitudes regardless of the real hand-to-target distance. Specifically, the participants were required to overshoot, undershoot, or stop at the target as a function of explicit trial-based instructions (for details, see Experiment 2 task design). They successfully achieved the task goal in >90% of the trials across the three experimental conditions.

For an exemplar subject, the requirement to stop at the actual target location (i.e., control condition) resulted in earlier RTs relative to the other tasks (Fig. 6B, dashed vertical lines). Higher velocity and longer movement time were observed for the overshoot than control conditions, whereas the undershoot task led to the opposite results (Fig. 6B). Also, the hand was accelerated for longer in conditions with higher peak velocities (Fig. 6B). The movement, however, evolved similarly across conditions with a single acceleration phase terminating within the first half of the movement (Fig. 6C).

Figure 6.

Task-dependent hand trajectories (A), velocity traces (B), and time-normalized hand velocity traces showing the point at which the hand reached the peak velocity within the entire movement (C) of an exemplar participant for Experiment 2 (same format as in Fig. 2).

For the entire group, the repeated-measures ANOVA showed statistically significant task condition (control vs target-overshoot vs target-undershoot) effects for RT (F_(2,10) = 31.9, p < 0.001, ${\eta }_p^2$ = 0.74), movement time (F_(2,10) = 50.9, p < 0.001, ${\eta }_p^2$ = 0.82), maximal hand velocity (F_(2,10) = 338, p < 0.001, ${\eta }_p^2$ = 0.97), time to maximal hand velocity (F_(2,10) = 81.7, p < 0.001, ${\eta }_p^2$ = 0.88), point of the maximal hand velocity within the movement (F_(2,10) = 6.7, p = 0.005, 0.38), and endpoint movement variability (F_(2,10) = 19.8, p < 0.001, ${\eta }_p^2$ = 0.64). The movement started significantly earlier in the control than other conditions, and significantly earlier in the undershoot than overshoot condition (Fig. 7A). The movement endpoint was reached significantly earlier and later than control for the undershoot and overshoot conditions, respectively (Fig. 7B). The hand moved significantly slower than control for the undershoot and significantly faster than control for the overshoot conditions (Fig. 7C). Peak velocity occurred significantly earlier than control for the undershoot and significantly later than control for the overshoot conditions (Fig. 7D). The peak velocity occurred at around half of the movement distance for all conditions, but significantly earlier for the overshoot condition (Fig. 7E). The trajectory endpoint variability was not significantly different from control for the undershoot condition (Fig. 7F). Notably, this indicates that the undershooting movements were oriented toward an abstract target location that was reached trial-by-trial with a precision error akin to that of the reaches terminating at the veridical target in the control condition. The movement endpoint was significantly more variable for the overshoot than the other conditions (Fig. 7F). This result, however, was consistent with the contrast between the long-reach condition and the other task conditions of Experiment 1 (Fig. 3F) and could reflect a speed-accuracy trade-off.

Figure 7.

Experiment 2 task-dependent variations of the RT (A), maximal velocity (B), movement time (C), time to the maximal velocity (D), index of the maximal velocity within the movement (E), and variability of the trajectory endpoint (F; same format as in Fig. 3). Statistically significant differences: (A) RT, undershoot versus control and overshoot, p < 0.001; control versus overshoot, p = 0.002; (B) movement time, p < 0.001 for all between-condition contracts; (C) maximal velocity, p < 0.001 for all between-condition contracts; (D) time to maximal velocity, p < 0.001 for all between-condition contracts; (E) index of the maximal velocity within the movement, undershoot versus control, p = 0.89; undershoot versus overshoot, p = 0.02; control versus overshoot, p < 0.001 (F) endpoint trajectory variability, overshoot versus undershoot and control, p < 0.001.

These results show that the requirement to not reach the actual target location increased the time to initiate the response, whereas the online movement evolved similarly across conditions in addition to expectable task-dependent variations in movement velocity and time. These results are consistent with a delay in the long-latency muscle response arising from the need to compute an abstract movement endpoint, rather than using the hand-to-target metrics directly.

EMG results

Figure 8 shows exemplar EMG data recorded from the PMch of a participant who exhibited express visuomotor responses across all the three conditions of Experiment 2 (same subject as in Fig. 4). For this subject, the express visuomotor response detection rate was 79% for the control condition, 49% for the overshoot condition, and 52% undershoot condition (Fig. 8A–I, red scatters and bars). Consistently, fewer long-latency muscle responses were detected for the control than the other conditions (Fig. 8A–I, magenta scatters and bars). The EMG signal enclosed in the express time window started diverging from baseline ∼85-90 ms after the target presentation and exhibited a similar incremental rate across the three task conditions up to 100 ms (Fig. 8J). For both the overshoot and undershoot conditions, however, the express EMG signal returned close to the background level before initiation of the long-latency response (Fig. 8J, red and green arrows). By contrast, for the control condition, the long-latency EMG signal followed the express EMG signal with little or no pause between these two phases (Fig. 8J, black arrow in the inset plot); this was observed also across all task conditions of Experiment 1 when participants always executed veridical target-directed reaches (Fig. 4J).

Figure 8.

Surface EMG activity of the PMch muscle and distributions of express response onset time, long-latency response onset time, and RT during the undershooting (A–C), control (i.e., stop at the target; D–F) and overshooting (G–I) visuospatial tasks of an exemplar participant who completed Experiment 2. A-I, Same format as in Figure 4. J, The task-dependent average EMG activity computed across all trials. Arrows highlight the transition of the EMG signal across the express (gray patch at 70-110 ms from the target presentation time) and long-latency epochs (>110 ms from the target presentation; for details, see Materials and Methods).

For Experiment 2, 10 participants (i.e., 83% of the sample) met the conditions for positive express visuomotor responses detection on the PMch across the three conditions. For these subjects, the repeated-measures ANOVA showed a statistically significant effect of task condition on the detection rate of express visuomotor response (F_(2,8) = 9.3, p = 0.002, ${\eta }_p^2$ = 0.51), which was significantly larger in the control than the other task conditions (Fig. 9A). The express visuomotor response tended to initiate earlier for the control than the other conditions (Fig. 9B), but this between-condition contrast was not statistically significant (F_(2,8) = 1, p = 0.37). The express visuomotor response magnitude was significantly modulated by the task conditions (F_(2,8) = 5, p = 0.02, ${\eta }_p^2$ = 0.36) as it was significantly smaller than control for both the overshoot and undershoot task conditions (Fig. 9C).

Figure 9.

Experiment 2 task-dependent variations of express visuomotor response detection rate (A), onset time (B), and magnitude (C; same format as in Fig. 5). A, Express response detection rate, undershoot versus control, p = 0.03; undershoot versus overshoot, p = 0.15; control versus overshoot, p = 0.003. C, Express response magnitude, undershoot versus control, p = 0.003; undershoot versus overshoot, p = 0.67; control versus overshoot, p = 0.04.

As was noted in Experiment 1, the task condition that facilitated express visuomotor responses corresponded to that leading to earlier RTs (i.e., control condition; Fig. 7A). The LLRM response magnitude, however, correlated with the movement amplitude (F_(2,8) = 17, p < 0.001, ${\eta }_p^2$ = 0.65), as it was significantly higher for the overshoot (76 ± 30 µV) than both the control (62 ± 31 µV; p = 0.02) and undershoot conditions (41 ± 21 µV; p < 0.001). Therefore, we tested whether the between-condition contrasts in express visuomotor response metrics held true for RT-matched and LLRM-matched datasets (for details, see Materials and Methods). Again, the results of the RT-matched and LLRM-matched subsets of trials (Table 2) were consistent with those obtained from the original data samples (Fig. 9). This indicates that the express visuomotor responses were modulated by the contextual task instructions, regardless of the RT or the LLRM.

Table 2.

Experiment 2: task-dependent variation of express visuomotor response metrics for data samples with matched RT and matched LLRM between conditions^a

Matched variable	Express visuomotor response metric	Task condition			Statistics
					Repeated-measures ANOVA	Post hoc comparisons
		Undershoot (US)	Control (C)	Overshoot (OS)		US vs C	US vs OS	C vs OS
RT	Detection rate (%)	52 ± 23	61 ± 16	56 ± 20	F(2,6) = 4.3 p = 0.03* ${\eta }_p^2$ = 0.35	p = 0.032*	p = 0.1	p = 0.17
	Onset time (ms)	96 ± 3	94 ± 3	95 ± 3	F(2,6) = 1.79 p = 0.2	—	—	—
	Magnitude (µV)	32 ± 11	37 ± 11	30 ± 9	F(2,6) = 4.3 p = 0.03* ${\eta }_p^2$ = 0.36	p = 0.035*	p = 0.52	p = 0.015*
Long-latency muscle response	Detection rate (%)	53 ± 22	65 ± 15	51 ± 15	F(2,8) = 8.7 p = 0.002* ${\eta }_p^2$ = 0.49	p = 0.02*	p = 0.4	p = 0.03*
	Onset time (ms)	95 ± 4	94 ± 3	94 ± 3	F(2,8) = 0.37 p = 0.69	—	—	—
	Magnitude (µV)	29 ± 12	39 ± 13	30 ± 10	F(2,8) = 7.2 p = 0.005* ${\eta }_p^2$ = 0.46	p = 0.016*	p = 0.4	p = 0.015*

^aData are mean ± SD.

*Statistically significant.

Our results suggest that the reaching amplitude modulates the express visuomotor response differently for veridical target-directed reaches (Experiment 1: Fig. 5; Table 1) versus reaches that undershoot or overshoot the location of physical targets (Experiment 2: Fig. 9; Table 2). To test whether the qualitive differences between the two experiments were the statistically significant, we ran a two-way repeated-measures ANOVA with experiment type (2 levels: Experiment 1; Experiment 2) and task condition (three levels: short/undershoot, control, long/overshoot) as within-subject factors. This analysis was conducted only for those subjects (n = 8) who completed both experiments and exhibited the express behavior across all conditions (see Materials and Methods). The two-way repeated-measures ANOVA showed that the task condition significantly modulated the detection rate and magnitude of express visuomotor responses (detection rate: F_(2,6) = 10.12, p = 0.002, ${\eta }_p^2$ = 0.59; magnitude: F_(2,6) = 9.33, p = 0.003, ${\eta }_p^2$ = 0.57). Notably, however, the effect of task condition on express response detection rate and magnitude differed for the two experiments as shown by a statistically significant interaction between experiment type and task condition (detection rate: F_(2,6) = 13.86, p < 0.001, ${\eta }_p^2$ = 0.66, Fig. 10A; magnitude: F_(2,6) = 4.69, p = 0.003, ${\eta }_p^2$ = 0.4, Fig. 10C). No significant main effect or interaction was found for the express response onset time (Fig. 10B).

Figure 10.

Between-experiments contrasts in express visuomotor response detection rate (A), onset time (B), and magnitude (C). Each thin line indicates 1 of the 8 subjects that completed either the first (black lines) or second (dark-red lines) experiment and exhibited express visuomotor responses across all conditions (Experiment 1: short, control, and long reaches; Experiment 2: undershoot, control, and overshoot tasks; see Materials and Methods). Thick lines indicate the average across subjects. For each subject, the data are normalized to the control condition.

Overall, the results of Experiment 2 indicate that matching the real target with the task-goal endpoint facilitates express transformation of visual inputs into appropriate motor outputs compared with reaches toward nonveridical target locations. When express responses occurred, however, they reflected express sensory-to-motor transformations of the target location within similar times for both veridical and nonveridical target-directed reaches.

Discussion

This study showed that express visuomotor responses reflect both the physical hand-to-target reaching distance and explicit instructions about the required movement amplitude during a visuospatial task. This suggests that the express visuomotor outputs can be strategically exploited by the cerebral cortex to facilitate rapid and appropriate responses to visual targets. A schematic representation of a possible circuit organization is outlined in Figure 11.

Figure 11.

Proposed cortico-subcortical coordination for visuomotor response generation in which cerebral cortex areas inhibit or facilitate the express tecto-reticulo-spinal system according to whether the task involves veridical or nonveridical targets to reach. The reach-goal location resulting from integration of physical target location and cued instructions is converted into the extent of the reach based on multisensory signaling of the current arm position (kinematics), the required joint torques (kinetics), and muscle activation via spinal interneurons and motoneurons. Dashed black lines with ? indicate uncertainty about whether the kinematic-kinetic metrics for the later (voluntary) reach component are computed by a transcortical or subcortical network.

Mechanisms for express visuomotor responses during veridical target-directed reaches

Experiment 1 showed that the hand-to-target reaching distance modulated express visuomotor transformations putatively performed along a tecto-reticulo-spinal circuit. The involvement of the superior colliculus in the generation of express visuomotor responses (Corneil et al., 2004, 2008; Pruszynski et al., 2010) is consistent with its capability to encode the location of either target or distractor stimuli within ∼40-70 ms (Boehnke and Munoz, 2008). Further, the surface layer of this midbrain structure is organized in a retinotopic map, whereas its deeper layers are organized in somatotopic maps (for review, see Boehnke and Munoz, 2008; Basso and May, 2017). Therefore, the superior colliculus could integrate visual and somatosensory information to compute the direction and distance of a target-directed action. Downstream from the superior colliculus, the reticular formation also receives inputs from somatosensory afferents (Leiras et al., 2010). Thus, a tecto-reticulo-spinal pathway has the required sophistication to compute and generate appropriate express visuomotor responses during veridical target-directed actions of different amplitudes (Fig. 11).

We recently documented circumstances of express visuomotor response modulation secondary to explicit cues (Contemori et al., 2021a,b, 2022). Prestimulus information might facilitate processing of expected stimuli at the superficial superior colliculus (for review, see Corneil and Munoz, 2014) and/or initiation of expected responses along the express pathway (Basso and Wurtz, 1998; Cisek and Kalaska, 2005; Contemori et al., 2022). Here, the eccentricity (hence saliency) of opposite targets was constant, but local thresholds for responding could be modulated asymmetrically by corticotectal projections (Boehnke and Munoz, 2008). Alternatively, or additionally, the advance information about the required distance for each possible target could bias the tecto-reticular circuit to generate larger express signals for targets requiring larger movements. These signals would be more likely to cross the spike threshold for neurons along the express pathway, consistent with the increased number of large express visuomotor responses associated with increased reaching distance. Such a mechanism would also be expected to result in slightly shorter latencies for the larger responses, but the differences would be difficult to detect from these EMG signals.

Previous work suggests that the express responses contribute to the volitional visuomotor behavior because larger express outputs were associated with earlier RTs (Pruszysnki et al., 2010; Gu et al., 2016; Contemori et al., 2021a). Mechanically, RT detection depends on muscle force accelerating the arm to the threshold velocity. Notably, the muscle force will rise more rapidly if the same motor units receive close temporospatial summation of express and long-latency motor signals, which would enhance intramuscular calcium release and diffusion (i.e., the catch property of muscle) (for review, see Tsianos and Loeb, 2017). In Experiment 1, more frequent and larger express visuomotor responses were associated with the amplitude of the long-latency EMG signal (Fig. 4G) that reflected the required reaching length (Fig. 2). As expected, summation of larger express and long-latency muscle recruitment generated earlier RTs and higher peak velocities. It is also possible that the weaker express visuomotor responses associated with shorter reaches reflect, at least partially (Wong et al., 2017), the higher complexity (i.e., RT delay) inherent in planning short movements that offer less time for online correction.

Mechanisms for express visuomotor responses during reaches to nonveridical targets

In Experiment 2, the requirement to overshoot or undershoot the real target led to fewer and smaller express visuomotor responses relative to control. Why should the express response be inhibited in circumstances requiring nonveridical responses?

The fact that express visuomotor responses rigidly encode the visual stimulus location (Wood et al., 2015; Gu et al., 2016; Atsma et al., 2018) is consistent with their proposed subcortical origin (Pruszysnki et al., 2010). Nevertheless, this also suggests that the express motor output might be counterproductive when the task requires nonveridical responses. For instance, Gu et al. (2016) showed that large pro-target express responses delayed the initiation of correct anti-target reaches, plausibly because of larger time costs to override the express pro-target muscle forces (Gu et al., 2016). The express system, however, appears to be flexible to contextual task rules when these are predictable. For instance, Wood et al. (2015) recorded express target-directed muscle responses in delayed-reach task trials that were randomly intermingled with no-delay trials (i.e., task condition unpredictability). By contrast, express visuomotor responses were obliterated when only delayed-reach trials were presented within a block (Pruszynski et al., 2010). Notably, these results are consistent with more recent evidence of cortically driven modulation of express visuomotor responses (Contemori et al., 2021a,b, 2022). Considering that express visuomotor responses aid rapid movement initiation (Pruszysnki et al., 2010; Gu et al., 2016; Contemori et al., 2021a), their inhibition could reflect strategic cortico-subcortical inhibition in contexts requiring longer RTs, such as for reaching toward nonveridical targets (Fig. 11).

The overshoot/undershoot tasks resulted in longer RTs than control, which could reflect increased complexity to plan a nonveridical response trajectory (Wong et al., 2016). Cortical planning of appropriate responses for achieving complex task goals can modulate the networks downstream from the cerebral cortices (Selen et al., 2012; for review, see Kurtzer, 2015). Critically, these include the reticulo-spinal circuits that are proposed to process the superior colliculus signals to generate express visuomotor responses (Corneil et al., 2004, 2008; Pruszynski et al., 2010) (Fig. 11).

The mammalian reticular formation is involved in the control of static posture (Sherrington, 1898; Rhines and Magoun, 1946; Magoun and Rhines, 1946). More recent neurophysiological and behavioral work also suggests that this brainstem structure contributes to volitional upper-limb movements (Alstermark and Isa, 2012; Contemori et al., 2021c) and reflexive responses to mechanical perturbations of static upper-limb postures (Kurtzer, 2015). Notably, the reticular formation receives descending signals from both the superior colliculus (Boehnke and Munoz, 2008) and cortical brain areas (Keizer and Kuypers, 1984, 1989; Fregosi et al., 2017; Darling et al., 2018; Fisher et al., 2021). The reticular formation, therefore, appears to be well placed to integrate descending collicular signals encoding the physical stimulus location (Everling et al., 1999; McPeek and Keller, 2002) with cortical premotor signals affording task-related rules (e.g., to overshoot/undershoot the target). In this circumstance, express stimulus-driven motor signals may be inhibited (or even obliterated) to delay the RT when there is uncertainty about the reach goal, such as for our nonveridical reaching tasks.

A common subcortical network for rapid initiation and online control of reaching?

Both the superior colliculus (Werner, 1993) and reticular formation (Buford and Davidson, 2004; Schepens and Drew, 2004) are active before and during upper limb reaching movements. The tecto-reticulo circuit, therefore, might contribute to both the reach initiation (Pruszysnki et al., 2010; Gu et al., 2016; Contemori et al., 2021a) and rapid kinematic (Day and Lyon, 2000; Day and Brown, 2001; Veerman et al., 2008; Smeets et al., 2016; Brenner et al., 2023) and postural (Fautrelle et al., 2010; Zhang et al., 2018a,b) adjustments during ongoing reaching actions. Notably, this is consistent with recent evidence of express visuomotor responses to correct the online movement trajectory in a jump-target task (Kozak et al., 2019). Furthermore, the size of muscle responses starting ∼90-120 ms after visual perturbation of virtual hand position showed a nonlinear scaling for perturbation amplitudes >2 cm (Cross et al., 2019). Although we did not characterize a function that defines express response modulation according to movement distance, our observation that the express response to the farthest target was not greater than that to the middle target is consistent with observations of response saturation in the dynamic tasks of Cross et al. (2019).

Initiation and online control of real-world visuospatial actions rely on visuomotor circuits that must integrate multisensory information about the body and target positions, which is inherently variable and noisy. The current data are consistent with previous evidence suggesting that the putative subcortical express circuits can be primed to generate flexible context-dependent motor outputs that support the accomplishment of visuospatial tasks from both static (Kurtzer, 2015; Weiler et al., 2019; Contemori et al., 2022) and dynamic postures (Cross et al., 2019; Kozak et al., 2019; Weiler et al., 2021).

In conclusion, this study shows that express visuomotor responses can be flexibly modulated to achieve visuospatial task-goals. The data are consistent with the idea of a subcortical visuomotor pathway whose motor output is strategically exploited by the cerebral cortex to facilitate rapid initiation of veridical target-directed reaches. It remains to be determined whether the longer latency, presumably cortically driven, motor responses rely on some or all of the same subcortical circuits to convert reaching targets in extrapersonal space coordinates into patterns of muscle activity that will achieve the desired limb movements. Overall, our findings emphasize the need for consideration of subcortical sensorimotor circuits in theories of human motor control and behavior.