Reach Corrections Toward Moving Objects are Faster Than Reach Corrections Toward Instantaneously Switching Targets

—Visually guided reaching is a common motor behavior that engages subcortical circuits to mediate rapid corrections. Although these neural mechanisms have evolved for interacting with the physical world, they are often studied in the context of reaching toward virtual targets on a screen. These targets often change position by disappearing from one place reappearing in another instantaneously. In this study, we instructed participants to perform rapid reaches to physical objects that changed position in diﬀerent ways. In one condition, the objects moved very rapidly from one place to another. In the other condition, illuminated targets instantaneously switched position by being extinguished in one position and illuminating in another. Participants were consistently faster in correcting their reach trajectories when the object moved continuously. (cid:1) 2023 The Author(s). Published by Elsevier Ltd on behalf of IBRO. This is an open access article under the CC BY license (http://creativecommons.org/ licenses/by/4.0/).


INTRODUCTION
Visually guided reaching is an extremely common activity in everyday life. However, in many cases we reach toward objects that are not stationary -like when we reach to pet a cat, only for it to quickly move to avoid our affection. Many groups have studied how quickly we can react to sudden changes in a reach target, finding that we are able to change an ongoing reaching movement quite quickly, with kinematic changes occurring less than 150 ms following the change of target position (Pe´lisson et al., 1986;Prablanc and Martin, 1992;Brenner and Smeets, 1997;Day and Lyon, 2000;Franklin and Wolpert, 2008;Gritsenko et al., 2009;Pruszynski et al., 2010;Gu et al., 2016). These corrections do not appear to rely on conscious perception of target movement  and probably involve subcortical pathways (Alstermark et al., 1987;Day and Brown, 2001); they are often considered automatic because they guide the hand to a moving target even if this conflicts with verbal instructions to do something else when the target moves, such as moving in the opposite direction (Day and Lyon, 2000;Johnson et al., 2002) or stopping an ongoing movement (Pisella et al., 2000).
The motivation for reach-correction studies often relates to the ability of the motor system to reach for or grasp objects. A number of groups have studied the effect of changes in object location (Paulignan et al., 1991;Gentilucci et al., 1992;Carnahan et al., 1993;Castiello et al., 1998;Camponogara and Volcic, 2019) or other physical properties like size, weight, and/or orientation (Patchay et al., 2006;van de Kamp and Zaal, 2007;Voudouris et al., 2013;Zaal and Bongers, 2014) on grasp kinematics. While studies of grasp tend to necessitate the use of physical objects, fewer studies that investigate reaching corrections have used physical objects as the target of the reach (D'Mello et al., 1985;Blouin et al., 1993;Battaglia-Mayer et al., 2013;Shaw et al., 2022), especially in the context of very rapid reaching corrections.
Instead of reaching toward physical objects, participants in many rapid reach correction studies reach toward images on a computer or projection https://doi.org/10.1016/j.neuroscience.2023.06.021 0306-4522/Ó 2023 The Author(s). Published by Elsevier Ltd on behalf of IBRO. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). screen (''virtual targets"), although some of these paradigms can involve making physical contact with the target by tapping the screen where they are displayed (Oostwoud Wijdenes et al., 2011;Brenner et al., 2023). Physical objects are less convenient to use in laboratory settings than virtual targets, but the motor system is likely specialized to interact with physical objects (Cuijpers et al., 2008;Schenk, 2012). Physical objects elicit different neural activity than virtual objects even when these objects are merely presented to the field of view (Gallivan et al., 2009;Fairchild et al., 2021). Further, placement of physical ''graspable" objects within the workspace (Gomez et al., 2018) appears to additionally modulate brain activity.
We recently described aspects of rapid reach corrections under tactile guidance using an oriented edge or the relative motion of a textured surface under the fingertip (Pruszynski et al., 2016;Reschechtko and Pruszynski, 2020). These studies involved reaching toward a physical object that was physically linked to the tactile stimulus. This physical object paradigm differs from most previous reaching paradigms for two main reasons. First, as previously mentioned, the target of the reach is a physical object; second, as a physical object, the target moves in a physically possible way, i.e. continuously, rather than disappearing in one location and instantaneously reappearing in another location (''switching").
Sarlegna & Mutha provide a comprehensive review of studies investigating the role of visual target changes on the control of movements directed to those targets (Sarlegna and Mutha, 2015). In the context of reach corrections, studies have often used instantaneous change of target position (Soechting and Lacquaniti, 1983;Prablanc and Martin, 1992;Brenner and Smeets, 2003;Veerman et al., 2008), instantaneous movement of a background to affect the visual impression of target movement (Saijo, 2005;Gomi et al., 2006), or instantaneous change in position of the cursor representing the hand (Franklin et al., 2016), although some previous studies have used continuously moving targets (Brenner and Smeets, 1997;Day and Lyon, 2000;Crowe et al., 2022). Franklin and Wolpert used smooth hand feedback perturbations in a series of experiments (Franklin and Wolpert, 2008), although they did not focus specifically on the effect of smooth target movement, and other groups have also used smooth background drifts (de la Malla et al., 2018;Crowe et al., 2022). Brenner and Smeets investigated reaching to moving virtual targets (Brenner and Smeets, 1997) and reported that response times were slightly higher when targets moved slowly compared to when they moved more rapidly or switched instantaneously. Numasawa and colleagues (Numasawa et al., 2022) also investigated the effect of various target movement speeds (not including instantaneous switching) on reaching corrections and found that, for a range of target speeds, correction latency was unaffected.
Here, we report the results of an experiment that investigates rapid reach corrections using physical objects that move in two categorically different ways. In one condition, participants reached toward a physical object that was illuminated; switching the illumination to another, identical object in a different position as if the physical object instantaneously ''switched" positions similar to common reach correction paradigms using virtual targets; in a second condition, the object moved continuously and very rapidly by rotating about a highspeed motor similar to one we used previously (Pruszynski et al., 2016;Reschechtko and Pruszynski, 2020). In contrast to previous findings using virtual targets, we found that participants updated their reach trajectories more quickly when reaching toward objects that rapidly changed position compared to those that switched instantaneously.

EXPERIMENTAL PROCEDURES
Participants 20 individuals (ages 19-39; 10 men and 10 women) participated in these experiments. Three participants were left-handed and the rest were right-handed based on self-report; participants used their dominant hand for all experimental procedures. All participants reported normal or corrected to normal vision and no injuries or neurological conditions that would prevent them from making rapid reaching movements. All participants provided informed consent in accordance with procedures approved by the Health Sciences Research Ethics Board at The University of Western Ontario.

Procedures
Participants sat at a table facing the experimental apparatus and all experiments were carried out in a brightly lit room. All potential targets -whether illuminated or not -were always visible. During each experimental trial, the participant used their dominant hand to reach from a stationary starting position to a physical target. The target was a table tennis ball (4 cm diameter), which was illuminated from within with a white LED. Participants began each trial with the index finger of their dominant hand pressing a mechanical switch connected to an Arduino compatible microcontroller; when the participant lifted their finger to reach, they released the switch; following the switch release, the microcontroller initiated a target change following a 30 ms fixed delay. Before initiating the reaching movement, participants received an auditory cue indicating that they could begin reaching toward the initial target at any time. Participants were instructed to reach in the direction the target as quickly as possible and received a second auditory cue 300 ms after they initiated their reach for pacing. We emphasized the rapidity of correction toward the target movement, but we did not require participants to touch or grasp the target and did not emphasize endpoint accuracy; rather, we asked them to make a shooting movement in the direction of the target.
For continuous object movements, the object (i.e. the illuminated table tennis ball) was mounted on the end of a carbon fiber rod, 30 cm from the starting position, which was rotated at high speed from its initial position to final position using a stepper motor with closed-loop control (34HE46-6004D-E1000, OMC Corporation Ltd, China), similar to a previous apparatus (Pruszynski et al., 2016;Reschechtko and Pruszynski, 2020). The object always started in a central position, from which it could move left (counterclockwise rotation, CCW), right (clockwise rotation; CW), or not change (NC) during catch trials. Movement onset was triggered by the microcontroller 30 ms after the participants lifted their finger from the starting position; in this condition, the motor rotated 15 degrees (corresponding to lateral displacement of approximately 7.7 cm) in 50 ms. The position of each participant's head was not controlled specifically, but this procedure resulted in the target moving across a visual angle of $12°, and therefore an average visual velocity of around 240°/sec.
For the switching target paradigm, LED illumination switched from one stationary object (table tennis ball) to another object target, similar to methods used in previous studies to switch physical reaching targets (Paulignan et al., 1991;Gentilucci et al., 1992;Castiello et al., 1998). These objects were positioned along an arc with radius of 35 cm (lateral distance between targets % 9 cm), directly behind potential target positions for the continuous movement condition. We used this arrangement so that the objects for each stimulus condition provided a very similar visual impression. The arrangement of objects on the apparatus is shown in Fig. 1A. When participants triggered target movement for this condition, illumination for the initial (central) object would turn off and illumination of one of the flanking objects (left; CCW, or right; CW) would turn on with the same 30 ms latency as in the moving target condition. In the equivalent of no change trials, the illumination did not switch.
Participants performed 180 reaches to the continuously moving object and an additional 180 reaches to the instantaneously switching target. The target conditions interleaved into 60-trial blocks; within each block, participants experienced 20 CW, 20 CCW, and 20 NC trials in a completely randomized order. Half of the participants began with the continuous movement condition (i.e. 60 continuous, 60 switching, 60 continuous, 60 switching, 60 continuous, 60 switching) and the other half began with the instantaneously switching condition.

Kinematics
We recorded upper limb kinematics at 120 Hz using optical motion capture (OptiTrack V120:Trio; NatrualPoint Inc, Corvallis, OR). We attached retroreflective markers to each participant's dominant arm on the following landmarks: fingernail of the index finger (fingertip); metacarpal-phalangeal joint (MCP) of the index finger; ulnar styloid (wrist); lateral epicondyle (elbow), and acromion process (shoulder); we identified these landmarks using visual inspection and palpation. Primary analyses were carried out using the position of the marker on the index finger (in agreement with previous studies), but some exploratory analyses involve measures related to the position of other arm segments.

Analysis
We imposed mild kinematic criteria on reaches to ensure we did not analyze reaches in which participants did not reach to the correct target. We only analyzed trials where participants moved at least 2.5 cm in the correct direction over 300 ms after the target movement. During switching target changes, the central ''switch" target was illuminated from within; on some trials, after participants initiated a reach, the central target turned off and a different target illuminated simultaneously (rightward or ''CW" target jump illustrated). The moving object (on the rod) was moved out of the workspace during jumping target trials and did not move during these trials. (C) Schematic of continuous object movement trials. The object on the rod (''continuous target") began in a central initial position; during some trials, the rod rotated to move the target to a new position directly in front of one of the other jumping target positions (''CW" rotation illustrated). During continuous movement trials, the moving object was illuminated from within and the switch targets were not illuminated. (D) Schematic of target kinematics. In the continuous movement target paradigm, the object moved to ±15°eccentricity over 50 ms (red traces); in the target switch paradigm, the target moved immediately (blue traces). In both paradigms, target change began 30 ms after participants began their reach, which is represented by the black vertical dashed line (t = 0). (E) Average kinematic traces in the anterior-posterior/medial-lateral plane for an exemplar participant, reaching to moving targets (red targets and traces) and switching targets (blue targets and traces). Numeric callouts refer to time in ms relative to target change (0 ms). Participants were not instructed to touch the targets and continued to reach beyond the time plotted here.
Because we did not emphasize endpoint accuracy or where specifically to end the reach in our instructions to participants, we did not impose any further acceptance criteria.
To determine the time when participants began to correct their reaching movements (time of divergence), we constructed a time series of receiver operating characteristics (ROC) for fingertip lateral velocity in each participant and condition. We then determined the onset of corrective movements by finding when the ROC reached certain threshold values. We define the time of divergence as the last time the ROC crossed 60% discriminability (i.e. correctly differentiating between 60% of left-and rightward movements) before reaching a threshold of correctly differentiating between 75% of movements to left-and rightward targets. We considered a variety of ROC criteria for determining divergence time and they yielded similar results.
To investigate potential differences in the gain of movement corrections, we analyzed fingertip position for each participant from 50 to 100 ms following the time their movements were determined to diverge via ROC. We compared the difference in fingertip position on the medial-lateral axis between reaches following leftward target movements to those following rightward target movements, expressed as a proportion of lateral target displacement (target displacement from the starting target position was approximately ±7.7 cm in movement trials and ±9 cm in switch trials), to see if this distance was affected by the type of target movement. Because the distance between the targets and start positions were different between the target presentation paradigms, we also analyzed difference in medial-lateral fingertip position after matching the anterior-posterior distance from the fingertip to the target for both target paradigms.
We also performed several exploratory analyses to investigate whether reaching strategies might differ between the target paradigms even before reach corrections began. To do this, we computed the maximum speed of reaches in the anterior-posterior direction, the position where participants achieved that maximum speed, and their anterior-posterior position when the target movement began, on a block-by-block basis. We used anterior-posterior speed for this analysis rather than 3D velocity so that we could include NC trials, in which peak velocity could not be influenced by corrective movements and could have resulted in apparently later peak velocity onsets during trials in which target position changed.
We carried out paired t-tests to compare divergence times and lateral displacements between target paradigms. We performed t-tests using the ttest_rel function in Scipy library (Virtanen et al., 2020) in Python (Python Software Foundation, Beaverton, OR). We used mixed model ANOVAs with participant as a random factor to compare maximum speed, position of maximum speed, and speed at the time of target displacement between target paradigms (2 levels: moving and switching) and blocks (3 levels: blocks 1-3). We carried out the ANOVAs using the mixed model workflow in JASP (JASP Team, Version 0.16.4). These ANOVAs modeled participants with random intercepts but not slopes and used Kenward-Roger approximation for degrees of freedom. When investigating a significant effect with more than two levels, we used Bonferroni corrections for post-hoc analysis. Unless otherwise noted, descriptive statistics are presented as mean ± SEM.

RESULTS
Participants successfully completed trials without notable exception. For NC trials, we required that participants stay within 2.5 cm of either side of the target. Of the 60 trials (across three blocks) for each target movement direction, the median number of accepted trials for the movement paradigm was 60, 60, and 58.5 for CW, CCW, and NC trials respectively, and the median number of accepted trials for the switching paradigm was 60, 58, and 58 for CW, CCW, and NC trials. Across all trials from all participants (3600 per paradigm), 102 were discarded in the movement paradigm and 178 were discarded in the target switching paradigm. The entire experiment (360 reaches) usually took around 45 minutes. Although the reaches were unconstrained, most fingertip movement was limited to the task-relevant (transverse) plane.

Participants corrected reaches earlier when the target moved continuously
Our main question was whether reach corrections toward continuously moving objects differed from corrections toward targets that switch positions instantaneously. The latency when the object moved (median = 125 ms) was lower (faster) compared to instantaneous target switching (median = 146 ms). These times are generally similar to those observed and reported in previous literature (Prablanc and Martin, 1992;Smeets, 1997, 2003;Day and Lyon, 2000;Franklin and Wolpert, 2008) and are also similar to those we reported previously using a physical object (Reschechtko and Pruszynski, 2020). These results are summarized in terms of the lateral velocity in Fig. 2A; traces diverge earlier for moving targets (solid lines) than switching targets (dashed lines). A paired t-test on these divergence times confirmed that reaches toward moving objects consistently diverged earlier (t 19 = 3.16; p = 0.005). Overall divergence times for the two conditions are shown in Fig. 2B.

Visual gain was similar between target movement paradigms
After determining that people began to change their reach trajectories faster for moving targets, we investigated whether the magnitude of correction after the time of divergence was different depending on target paradigm. For each participant, we computed the average position of their fingertip in leftward versus rightward reaches for the 100 ms following the time at which their reaches diverged. The average gain, computed from fingertip position from 50 to 100 ms following time of divergence was marginally larger for the moving target than the switching target (4.75 ± 0.49 cm compared to 3.81 ± 0. 30 cm; paired t-test: t 19 = 2.12; p = 0.047). Fig. 2C shows the temporal evolution of difference in fingertip position between movements to left-and rightward target positions aligned by time of divergence, while Fig. 2D shows participants' average differences over 100 ms following the time of divergence. The intertarget distance was larger for reaches to switching targets than moving targets, so we performed the same analyses on gain normalized by inter-target difference, which increased the difference between paradigms (proportion of inter-target distance: 0.31 ± 0.03 vs 0.21 ± 0.02; t 19 = 3.39; p = 0.003). The distance to the switch targets was also longer than the distance to the moving target because of the way the targets were positioned (Fig. 1A), so we also analyzed the distance between leftward-and rightward trajectories when the anterior-posterior distance was matched across target paradigms; in this case, gain was similar between target paradigms (proportion of inter-target distance: moving: 0.33 ± 0.15; switching: 0.34 ± 0.13; t 19 = 0.25; p = 0.80).

Relative onset of divergence is uniform across upper limb segments
Due to the kinematic redundancy in this unconstrained task, it was possible for people to make corrective actions without all segments of the upper limb moving in a correction-specific direction. We thought that people might change their coordination strategies in the different movement paradigms because we previously observed some participants who appeared to make a final correction with the hand only via wrist ab/ adduction, while the wrist appeared to remain relatively stationary. Therefore, we tracked the positions of various parts of the arm during this study and examined whether the corrective action is implemented by a proximal arm segment (e.g. movement about the shoulder), or a distal one (e.g. movement about the elbow or wrist). One potential reason people might select one strategy versus another is that the distal segments have lower inertia; on the other hand, descending commands reach proximal musculature more quickly than distal musculature.
We carried out an exploratory analysis by computing divergence times (as we did for the fingertip) for markers attached to the MCP, wrist, and elbow joints. Using ROC curves, it was possible to distinguish between movements toward leftward and rightward target movements from all segments and for both movement paradigms. In general, divergence times were later for segments that were more proximal. For reaches toward switching targets, median divergence times were 142, 150, 158, and 229 ms following target change onset for the fingertip, MCP, wrist, and elbow markers, respectively (Fig. 3A). For reaches toward moving objects, median divergence times were 125, 133, 142, and 208 ms following target movement (Fig. 3B). Overall, there was no clear effect of the target paradigm on arm segment use: we observed the same relationship (more rapid divergence toward objects that moved continuously than targets that switched) regardless of the segment we analyzed.
Finally, we investigated whether, within individual participants, earlier divergence at one segment was related to earlier divergence at another segment by correlating divergence times across each pair of arm segments for each target condition. Divergence times at the fingertip and MCP joint were positively correlated for reaches to switching targets (r = 0.87; p = 4.93e À7 ) and moving objects (r = 0.59; p = 0.006), and divergence times for the MCP and wrist were positively correlated during reaches to switching targets (r = 0.5; p = 0.024), but none of the other correlations were significant. Across target movement paradigms, elbow divergence was usually the latest of all segments (all participants for the moving object and 18 for the jumping target). While the order of divergence recapitulated the median times of divergence (most participants diverged at fingertip, then MCP, and then wrist), there was notable variability. When reaching toward the moving target, four participants showed earliest time of divergence at the wrist for the moving object paradigm; however, when moving toward the switching target, only one of those participants showed earliest divergence at wrist -even though four participants overall showed earliest time of divergence in the wrist for this movement paradigm as well. These results likely point to the kinematic redundancy that people naturally exploit during this unconstrained task, despite its rapid timescale.

DISCUSSION
Our results indicate that, at least under certain circumstances, people correct ongoing reaching movements more quickly if the target they are aiming for moves very quickly than if the target they are aiming for instantaneously switches position. This may occur because different target movement paradigms activate different circuits that control movement production or because the information available to participants to track target movements between conditions is different.

Movement strategies toward switching and moving targets
Due to the experimental setup and use of physical targets, the switching targets were farther from the participants than the moving targets. The center of moving target moved about a circle with 30 cm radius (the starting position was in the center of the circle), whereas the switching targets were positioned along a circle with 35 cm radius. Participants adjusted their reach kinematics to these different target positions: although fingertip speed was not significantly different at the time of target change, participants reached higher peak velocities when reaching toward the farther (switching) targets and achieved those velocities later in the reach. Brenner and colleagues showed that movement corrections are more vigorous when they occur closer to the time of target contact .
If participants anticipated arriving at the targets earlier when reaching toward a moving target, we might have identified an earlier time of divergence because they corrected more vigorously. Our investigation of movement vigor (visual gain) is not conclusive in this regard: while visual gain -computed as the difference in lateral position of the finger between 50 and 100 ms following divergence onset -was higher in units of absolute displacement, visual gain was similar when participants were the same distance from the target (even though they achieved different peak anterior-posterior speeds as they reached toward the target). It is also unclear how our instructions, in which participants did not actually touch the targets but were encouraged to make a reach correction as quickly as possible, would affect the constraint of ''time to contact" (participants made physical contact with a screen in Brenner et al. (2022) so contact was presumably a very salient constraint).
The group of Brenner and Smeets have conducted many studies that investigated the effect of several aspects of target movement on reaching behavior. Among the factors investigated was movement speed, including contrasting jumping and continuously moving targets. They found that people made somewhat faster corrections when targets moved faster -and corrections were more rapid when the targets switched instantaneously compared to when they moved continuously (for example, see Fig. 5 in Smeets 1997 andFig. 2 in Brenner &). Numasawa and colleagues (Numasawa et al., 2022) also investigated the effect of different target moving speeds on reaching corrections. Although they did not compare the effect to a jumping target, they did not observe differences in correction latency depending on target velocity. In contrast, they report that gain was higher for higher target velocities.
One potential difference between our experimental paradigm and those previously mentioned is our use of physical targets. Participants in Numasawa and colleagues' study used a pointer to track virtual target on a screen. Brenner and Smeets have used a variety of experimental paradigms in these studies, including many requiring ''physical contact" with virtual targets by touching the screen where the target is displayed (Oostwoud Wijdenes et al., 2014;Zhang et al., 2018;Brenner et al., 2023). It might seem surprising that merely using physical targets could make a difference -especially because our instructions did not require participants to grasp or touch the targets. However, judgements regarding ''graspability" (as determined by effector and/ or target) could be important for modulating the neural circuits involved in the action (Freud et al., 2018;Gomez et al., 2018), and observing physical objects elicits different neural responses than observing pictures of those objects (Snow and Culham, 2021). Because circuits for rapid reach corrections are presumably specialized for interactions with physical objects, these circuits may also induce more rapid corrective behavior, although we did not directly test this hypothesis here.

Eye movements and arm movements
When the eyes fixate on an object and that object begins to move, the movement of the object on the retina (''retinal slip") drives smooth pursuit eye movements to bring the object back toward the fovea smoothly (Krauzlis, 2004;Ono, 2015). In contrast, when a target jumps to a new position instantaneously, the target shift drives saccadic eye movement in which the fovea is shifted ballistically to the new position (Leigh and Zee, 1999). Onset of smooth pursuit movements is reportedly 25-75 ms earlier than that of saccades (Krauzlis and Miles, 1996;Adler et al., 2002).
We did not record eye movements in the current study, but rapid reaching corrections likely involve subcortical circuits (Alstermark et al., 1987;Day and Brown, 2001), and those circuits could be related to some of those involved in the production of smooth pursuit. If some of the same circuits are involved in reach corrections and eye movement, the increased speed of reach corrections to continuously moving targets could be a product of the same processes that yield faster responses during pursuit eye movement. In our moving target paradigm, target change was very rapid ($50 ms); this may not have evoked eye movement before the onset of reach correction, but in different paradigms researchers have reported both eye movements preceding hand movements (Georgopoulos et al., 1981;Neggers and Bekkering, 2000;Land and Hayhoe, 2001) and hand movements preceding eye movements (Abekawa et al., 2014). Regardless of whether eye or arm movements come first, onset of eye and hand movements are highly correlated, which is consistent with the idea that retinal slip may drive rapid reach corrections even if it does not first induce eye movement.

Visual information regarding movement and position
Another potential reason for different latencies of correction toward target change in our study is related to availability of visual information. In our study's movement condition, information related to both position and motion were available, whereas only position information was available in the switching target condition -although the target movement used in this study was very rapid and stopped before reach corrections began. Previous studies have investigated whether and how people use position and motion information to guide reach toward moving targets (Smeets and Brenner, 1995;Brenner and Smeets, 1997;Brouwer et al., 2003;Oostwoud Wijdenes et al., 2014;Brenner et al., 2022). These results indicate that people may take both position and motion into account, for example by reaching farther in the direction of target movement for faster target speeds (Brouwer et al., 2002). Oostwoud Wijdenes and colleagues found that reach corrections were earlier but less vigorous when motion information about a moving target was withheld by blocking participants' vision of that target during part of the movement (Oostwoud Wijdenes et al., 2014). In a study where participants tapped targets on a screen that rapidly changed positions by a small amount (1.66 mm update at 120 Hz), Brenner and colleagues concluded that people rapidly integrate target position information into reach corrections but use movement cues over a longer time to anticipate when they will arrive at the target (Brenner et al., 2023). If this is the case, target movement information seems an unlikely candidate to drive the behavioral differences in our study because the target movement itself occurred so quickly and the targets in both paradigms were stationary for most of the reach. However, predictions of target movement that are induced by observing motion -or position updates congruent with motion -can alter position perception for hundreds of milliseconds via a process sometimes called ''Representational Momentum" (Freyd and Finke, 1984). Such predictions could lead participants to anticipate a moving target will move farther than a switching target, which could lead to higher visual gain toward moving tar-gets, although whether this occurred in the present study is not clear; a follow-up study emphasizing reach accuracy could serve to elucidate this point further by investigating whether reaches toward moving targets overshoot the lateral position.
While duration of target movement in our paradigm was very short, it is nonetheless possible that participants used different tracking strategies for the different target paradigms. Participants may detect target change at different times depending on the target paradigm: in the case of the continuously moving target, participants may have started to track the target as soon as it began to move, whereas they needed to detect a position change in the target switching condition. Some experiments suggest that the process of detecting a change in target position can increase latency of corrections (Gritsenko et al., 2009;Brenner and Smeets, 2011;Smeets et al., 2016); if an instantaneous target switch prevents participants from continuously updating position information, the target switch paradigm could lead to increased correction latencies because participants had to detect the position change before initiating the correction.

Limitations
A number of considerations make it difficult to directly compare different target kinematics across physical objects and virtual targets. One primary difficulty is matching enough parameters that the physical object and virtual target paradigms are reasonably similar. The virtual target paradigms we have cited here involve different reach distances, instructions, stimuli, and effectors. These differences make it difficult to compare our results to previous results, and to know whether our results generalize to a wider range of movement velocities, target sizes, and so forth. Due to physical constraints of our target paradigm, the presentation of targets in different positions with respect to the starting position; therefore, we cannot completely dissociate effects of different target presentation paradigms and differences which may arise from differing distances to the target. Our primary focus in this study was to investigate the onset of movement corrections; as such, we did not emphasize accuracy of the reach so we cannot evaluate whether different target presentation paradigms led to different reach strategies in the context of endpoint accuracy. We also did not record eye movement, so connections between arm and eye movement are based only on the differences in movement paradigm and eye movements similar paradigms have evoked. Finally, the physical movement of the object during continuous movement produced some audible noise when the motor moved the target; while this noise is not direction-specific, we cannot rule out the possibility that it affected response latency.

DECLARATION OF COMPETING INTEREST
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.