Express Arm Responses Appear Bilaterally on Upper-limb Muscles in an Arm Choice Reaching Task

When required, humans can generate very short latency reaches towards visual targets, like catching a falling cellphone. During such rapid reaches, express arm responses are the first wave of upper limb muscle recruitment, occurring ~80-100 ms after target appearance. There is accumulating evidence that express arm responses arise from signaling along the tecto-reticulo-spinal tract, but the involvement of the reticulo-spinal tract has not been well-studied. Since the reticulospinal tract projects bilaterally, we studied whether express arm responses would be generated bilaterally. Human participants (n = 14; 7 female) performed visually guided reaches in a modified emerging target paradigm where either arm could intercept the target. We recorded electromyographic activity bilaterally from the pectoralis major muscle. Our analysis focused on target locations where participants reached with the right arm on some trials, and the left arm on others. In support of the involvement of the reticulospinal tract, express arm responses persisted bilaterally regardless of which arm reached to the target. The latency and magnitude of the express arm response did not depend on whether the arm was chosen to reach or not. However, on the reaching arm, the magnitude of the express arm response was correlated to the level of anticipatory activity. The bilateral generation of express arm responses supports the involvement of the reticulo-spinal tract. We surmise that the correlation between anticipatory activity and the magnitude of express arm responses on the reaching arm arises from convergence of cortically-derived signals with a parallel subcortical pathway mediating the express arm response. New and Noteworthy Express arm responses have been proposed to arise from the tecto-reticulo-spinal tract originating within the superior colliculus, but the involvement of the reticulo-spinal tract has not been well studied. Here, we show these responses appear bilaterally in a task where either arm can reach to a newly appearing stimulus. Our results suggest that the most rapid visuomotor transformations for reaching are performed by a subcortical pathway.


Introduction
lower (18,19), and muscle responses evoked from ipsilateral motor areas tend to have longer 88 latencies and smaller magnitudes (20)(21)(22). To date, express arm responses have been studied 89 only in unimanual reaching tasks. The goal of this study is to test whether express arm responses 90 would be expressed bilaterally when either arm can be used to reach to a visual target. 91 Previous work has shown an emerging target paradigm, wherein a moving target 92 transiently disappears and then emerges from behind a barrier, elicits robust express arm 93 responses in the reaching arm in almost every participant (5,12,23,24). Here, we modified this 94 paradigm by increasing the number of potential locations of target emergence and allowing the 95 subject to reach toward the emerging target with either arm. These modifications elicited reaches 96 by either the left or right arm for different target locations, and at certain locations elicited left 97 arm reaches on some trials and right arm reaches on other trials. Muscle recruitment for reaches 98 toward these latter locations is critical for our primary aim, which is to determine whether the 99 expression of express arm responses depended on whether the arm was chosen to reach to the 100 target or not. Further, as our task requires participants to choose which arm to move toward the 101 emerging target, a secondary aim was to determine when limb muscle activity indicated whether 102 the associated arm would reach to the target or not. In doing so, we can assess the presence or 103 absence of any relationship between the commitment to move a particular arm and the express 104 arm response. Overall, we found that express arm responses evolved on both the chosen and non-105 chosen arm. We also found that the time at which limb muscle recruitment indicated which arm 106 would reach to the target was highly variable and was unrelated to the timing of express arm  appearance was fixed at 1.5 s for all target locations. The target was only presented in its entirely 151 after it moved beneath the occluder, preventing the presentation of a half-moon stimulus with a 152 lower overall area. Upon target emergence, participants were instructed to reach toward the 153 emerging target as quickly as possible and were told that they could use either arm to do so. Surface electromyographic (EMG) activity was recorded from the clavicular head of the right 160 and left pectoralis major muscle (PEC) with double-differential surface electrodes (Delsys Inc. 161 Bagnoli-8 system, Boston, MA, USA). To ensure consistency, the same individual placed 162 electrodes on the right and left PEC for all participants, using anatomical landmarking and 163 muscle palpation to determine location. EMG signals were amplified by 1000, sampled by the 164 KINARM data system at 1000 Hz, then full wave rectified off-line. Kinematic data was also 165 sampled at 1000 Hz by the KINARM data system. At the time of target emergence, a visual 166 stimulus unseen by the subject was also presented to a photodiode, and all EMG and kinematic 167 data were aligned to this time.

168
To allow cross-muscle comparisons, we normalized EMG activity to baseline, dividing 169 EMG activity on each trial by the average EMG activity between -500 to -100ms before target 170 onset across all trials. Normalized muscle activity was only used when comparing the 171 magnitudes of recruitment across different muscles, otherwise, source EMG voltages were 172 analyzed.

173
RT was calculated as the time from target appearance below the occluder, indicated by 174 the photodiode, to the initiation of the reaching movement by the arm that intercepted the target.

175
The reach RT for each trial was determined using a custom MATLAB (version 2014b, The 176 MathWorks, Inc., Natick, Massachusetts, United States) script that found the time when the hand 177 exceeded 5% of its peak velocity of the hand after target onset, and then moved backwards in 178 time to find the point at which hand acceleration following target onset exceeded the 95% 179 confidence interval of acceleration data taken from a period of 100 ms before to 50 ms after 180 target onset. The offset of hand motion was the time at which hand velocity fell below 5% of its 181 peak velocity. The onset and offset of movements were confirmed offline by an analyst in a 182 graphical user interface and adjusted if necessary. We excluded trials with RTs less than 100 ms 183 due to presumed anticipation, and trials with RTs exceeding 500 ms due to presumed 184 inattentiveness. 16% of trials were excluded using these RT constraints, primarily due to 185 anticipatory movements. We also excluded trials consisting of multiple movement segments 186 toward the target, excluding ~2% of trials.

187
Arm-choice was defined simply as the arm that intercepted the target. A psychometric 188 function was generated using the proportion of right arm reaches as a function of target location.

189
For each participant a logistic regression was fit to the data, using the 'logit' MATLAB function: 190 f(p) = log(p/(1-p)), where p is the proportion of right arm reaches. Using the fitted curve, we 191 estimated the theoretical point where a target would be intercepted with either the left or right 192 arm with equal likelihood. The closest target location to this point, referred to as the target of 193 subjective equality, was then used for further analyses, as this target location permitted the best 194 within-muscle comparison of recruitment when that arm was chosen to reach to the target or not.

195
Previous work examining the express arm response has used a time-series receiver-196 operating characteristic analysis, contrasting EMG activity for movements into or away from a 197 muscle's preferred direction (1, 27). Because a given arm only moved in one direction in our 198 study (e.g., all targets lay to the left or right of the right or left arm, respectively), we developed a 199 novel method for detecting and quantifying the express arm response. Our method involves a 200 two-or three-piece linear regression, fitting lines to EMG activity in a baseline, anticipatory 201 (only used for the three-piece linear regression), and post-target interval (see (5, 13) for methods 202 based on a two-piece linear fit). Our rationale for using a three-piece linear regression was based 203 on qualitative observation of mean EMG recruitment, which often started to increase in an 204 anticipatory fashion above baseline before and just after target appearance ( Figure 1B). 205 To determine the presence or absence of an express arm response, we took the following 206 steps. First, we ensured that there were at least 25 reaches from a given arm to a particular target 207 (only one arm was used to reach to most target locations). Whenever there were enough reaches 208 from a given arm, we further analyzed the muscle activity from both the left and right PEC, as 209 this provides us with EMG activity from both the reaching and non-reaching arm. We then fit the where there was no increasing anticipatory activity between baseline and target related activity.

216
To account for situations where anticipatory activity was present, we fit the data with a three-217 piece linear regression, enforcing a minimum of 10 ms between the first and third pieces. Doing 218 so involved finding two inflections points that minimized the loss, delineating the baseline 219 activity (spanning from -200 ms to the first inflection point), anticipatory activity (spanning from 220 the first to second inflection point), and the target-related interval (spanning from the second 221 inflection point to the peak EMG activity; see Figure 1B). As a three-piece linear regression 222 always decreases loss compared to a two-piece linear regression, we determined whether a three-223 piece regression would be warranted by calculating the ratio of the loss between the two-and 224 three-piece linear regressions. If the ratio was below 0.7, we used the three-piece linear 225 regression. If the ratio was above 0.7, we used the two-piece linear regression. We also 226 calculated the loss ratio between the two-piece linear regression and regular linear regression. A 227 two-piece linear regression was used if the loss ratio was below 0.6, otherwise a linear regression 228 was used.

229
Following these steps, we then determined the presence of an express arm response in the 230 following manner. First, the EMG data had to be fit by either a two-or three-piece linear 231 regression; EMG data fit by a linear regression signified the absence of an express arm response.

232
Second, the target related inflection point had to occur within 70-105 ms, and the slope of the 233 first and second piece for a two-piece linear regression, or the second and third piece for a three-234 piece linear regression had to be significantly different at P < 0.05, as determined by a 235 bootstrapping procedure. If these criteria were met, the latency of the express arm response was 236 defined as the time of the inflection point for the two-piece linear regression, or the second 237 inflection point for the three-piece linear regression. The express arm response magnitude was 238 defined as the difference of the peak EMG activity over the next 15ms to the EMG activity at the 239 onset of the response. We also quantified muscle activity immediately preceding the express arm 240 response (in the results, we term this the "anticipatory activity" for simplicity, although we 241 recognize that anticipatory and baseline activity are equivalent for a two-piece linear regression).

242
Anticipatory activity was quantified as the difference between the EMG signal immediately 243 before the express arm response, and the baseline activity.

244
In a separate analysis to determine at what point muscle activity reflected arm choice, we 245 used a time-series receiver-operating characteristic (ROC) analysis from EMG activity recorded 246 when participants reach to the target of subjective equality. This target location provided a large 247 sample of EMG activity from a given muscle on trials where the associated arm or the opposite 248 arm reached to the target. We separated EMG activity based on which arm reached to the target, 249 then analyzed at every time sample (1 ms) from 500ms before target onset to the end of the trial.

250
For each time-point we calculated the area under the ROC curve, which is the probability that an 251 ideal observer could discriminate whether the associated arm would reach to the target or not, 252 based solely on the EMG activity. Values of 1 or 0 indicate perfectly correct or incorrect 253 discrimination respectively, whereas a value of 0.5 indicates chance discrimination. We set the 254 threshold discrimination at 0.6 because this criterion exceeded the 95% confidence intervals   (28).

272
The reticular formation is a likely interface in a tectal pathway mediating express arm responses.

273
Given the bilateral projections from the reticular formation, we wondered whether express arm 274 responses would be expressed bilaterally in a task where participants could choose which arm to 275 use to intercept an emerging target. We recorded muscle activity from the right and left PEC 276 muscles as participants completed a modified emerging target paradigm ( Figure 1A). Targets 277 could emerge at one of seven locations below the barrier, and participants reached to catch the 278 target as fast as possible with either arm. We analysed muscle activity from both the reaching 279 and non-reaching arm to determine the presence of the express arm response. We also examined 280 the time at which muscle activity indicated that the associated arm would reach toward the target 281 or not, relative to the time of any express arm response. representative subject). This location was associated with a high number of reaches from either 292 arm in all participants. Across our sample, the target of subjective equality was at center (n = 10), 293 3 cm left (n = 2) or 7 cm left (n = 2) of center ( Figure 2B). The target of subjective equality 294 permits a within-muscle comparison of recruitment when the associated arm was chosen to reach 295 or not. In general, locations other than the target of subjective equality did not generate enough 296 reaches from both arms for within muscle comparisons.

298
Do express arm responses appear bilaterally? 299 The main question we wanted to address was whether express arm responses evolve bilaterally 300 when either arm could be used to intercept an emerging target. Figure 3A shows the average 301 muscle activity from an exemplar participant (same participant as Figure 2A), across all 302 positions where at least 25 reaches were made by the associated arm. These data show how 303 participants tended to reach with the arm closest to the target (e.g., note how the right or left arm 304 tended to reach for targets in the right or left hemifield, respectively). Using either a two-piece or 305 three-piece linear regression to determine whether there was an express arm response ( Figure   306 1B, see Methods), we observed express arm responses in both the reaching and non-reaching arm 307 (inflection points are denoted by the black dot; express arm responses in Figure 3A are denoted 308 by the first or second dots when a two-or three-piece linear regression was used, respectively) .

309
When detected, express arm responses occurred ~90ms after target appearance in both the 310 reaching and non-reaching arms.

311
Previous reports have emphasized that the trial-by-trial timing of express arm responses 312 is more aligned to stimulus rather than movement onset (1, 4). As shown in Figure 3B, we 313 indeed found that the timing of express arm responses was more tied to stimulus rather than 314 movement onset, regardless of whether the associated arm reached or not. This characteristic 315 feature of express arm responses appears as the vertical banding of EMG activity in Figure 3B 316 when muscle activity is aligned to stimulus onset, showing a burst of muscle recruitment ~90 ms 317 after target emergence regardless of the ensuing reach RT. Following this bilateral generation of 318 the express arm response, a more prolonged period of increased recruitment was observed only 319 on the muscle associated with the reaching arm.

320
The prevalence of express arm responses is known to vary across paradigms and 321 participants (1,5,9,12). We wanted to know whether all participants had express arm responses 322 in general, and further whether the responses were equally prevalent in the reaching and non-323 reaching arms. As shown in Figure 4, the modified emerging target paradigm elicited express 324 arm responses from at least one participant at each location. Further, almost all participants (n = 325 13) generated express arm responses following target presentation to at least one location.

326
Compared to the null-hypothesis that the response only occurs in the reaching arm, we found that 327 the response also occurred in the non-reaching arm (Chi-squared test: p < 0.001, c2= 52.3858, 328 df=1). We also compared the prevalence of express arm responses in the reaching and non- Next, we were interested in the latency and magnitude of express arm responses recorded 339 bilaterally, and whether these measures differed depending on whether the associated arm was 340 selected to move or not. Previous work has shown that express arm response latency (9) and/or 341 magnitude (4) may differ depending on stimulus properties and task context. We therefore 342 examined these properties at the target of subjective equality, as a function of whether the 343 associated arm was chosen to reach or not (note that this is a within-muscle comparison). Using 344 only paired observations (i.e., when express arm responses were detected in a given muscle 345 regardless of whether the arm was chosen to move or not) we found no difference in express arm 346 response latency with arm choice (paired observations shown as the connected points in Figure   347 5A; p = 0.5299, t = -0.6565, df = 8). Further, using a single factor ANOVA we found no

353
We also investigated whether the express arm response on the reaching arm was different 354 as a function of whether the non-reaching arm also showed an express arm response. To test this, 355 we compared the latency and magnitude of the express arm response in the reaching arm when 356 the non-reaching arm also exhibited an express arm response versus when the express arm 357 response was only observed in the reaching arm. Using a student's t-test, we found no difference 358 in the express arm response latency or magnitude on the reaching arm as a function of the 359 presence or absence of an express arm responses on the non-reaching arm (latency: p > 0.05, t = 360 0.4947, df = 6.4457, magnitude: p > 0.05, t =1.5089, df = 6.9517; data not shown).

361
If mediated by a common source like the reticular formation, we would expect the 362 magnitude of express arm responses on the reaching and non-reaching arm to be correlated 363 across participants and targets (e.g., a larger express arm response on the reaching arm should be 364 associated with a larger express arm response on the non-reaching arm). To analyze this, we 365 identified target locations where an express arm response was observed on both the reaching and 366 non-reaching arm, using one observation for each participant (see Methods), and found that 367 express arm response magnitudes were positively correlated between the muscles (Figure 5C, 368 Pearson correlation, p = 0.0204, r = 0.7141; every point represents a unique observation for a 369 participant; note magnitudes are normalized here since this is an across-muscle comparison).

370
Thus, larger express arm response magnitudes on the reaching arm tended to be associated with 371 larger express arm response magnitudes on the non-reaching arm. Interestingly, on average, the 372 magnitude of the express arm responses was about 1.5 times as large on the reaching versus non-373 reaching arm (p = 0.0631, t = 2.1190, df = 9).

374
In our paradigm, participants knew in advance that targets would appear medial relative 375 to the starting position of both the left and right arm, leading us to wonder if participants 376 anticipated which arm to use prior to target emergence. We analyzed the potential influence of 377 such anticipation and found greater anticipatory activity when the associated arm was chosen to 378 reach to the target of subjective equality (Figure 5D; paired t-test, p = 0.0080, t = 3.5082, df = 379 8). This relationship between anticipatory activity and arm choice can be seen in Figure 3A on 380 the right PEC at the 0 cm target; note how anticipatory activity preceding the express arm 381 response was greater when the right rather than left arm reached to the target. This level of 382 anticipatory activity related to the magnitude of the ensuing express arm response in the reaching 383 arm (n.b., the latter measure quantifies the EMG magnitude above anticipation; Figure 5E  The express arm response is a brief period of muscle recruitment that increases muscle force.

414
Previous work with unimanual anti-reach, delay, or stop-signal tasks has shown that express arm 415 responses can produce small, task inappropriate, movements toward a target (4, 29, 30). The 416 non-reaching arm provides a further opportunity to study the kinematic consequences of express 417 arm responses in isolation from ensuing reach-related activity. First, we looked at the velocity of 418 both the reaching and non-reaching arm at every location and consistently saw a small movement 419 towards the target in the non-reaching arm. This can be seen in Figure 7A where we have plotted 420 horizontal velocity from the exemplar participant for both the reaching and non-reaching arms at 421 every location. As expected, the velocity is much higher in the reaching arm than in the non-422 reaching arm, but there is clearly a small deviation of the non-reaching arm toward the target 423 (represented at an increased scale in the insets in Figure 7A). To quantify the non-reaching 424 arm's peak velocity and allow cross-participant comparisons, we normalized it by the peak 425 velocity of the reaching arm. We found on average the non-reaching arm had a peak velocity that 426 was 8.11 ± 2.27% of the reaching arm. Compared to a null hypothesis that no movement occurs 427 in the non-reaching arm, the non-reaching arm did indeed move towards the stimulus (Student's 428 t-test, p < 0.001, t = -13.3950, df = 13). Next, we compared the peak velocity in the non-reaching 429 arm based on whether an express arm response was observed but did not find any difference in 430 peak velocity based on whether an express arm response was observed (peak velocity: 8.94 ± 431 2.22%) or not (peak velocity: 7.80 ± 2.68%) (Figure 7B; student's t-test, p > 0.05). Thus, 432 although the non-reaching arm did move toward the target, the peak velocity of this movement 433 was unrelated to the detection of an express arm response. This is a somewhat surprising result; 434 however, this could be due to a failure of the surface EMGs to reliably detect all express arm 435 responses, especially in situations with a low signal to noise ratio.

436
Another feature that is apparent in the velocity traces of the non-reaching arm is that the 437 small movement toward the target is followed by a brief reversal in velocity. This reversal 438 reflects a small returning movement of the non-reaching arm back toward the starting position.

439
Interestingly, the EMG correlates of this returning movement on the non-reaching arm are 440 apparent in Figure 3A, where recruitment levels after the express arm response drop below the 441 levels of anticipatory recruitment attained just before the express arm response.

442
Given the presence of anticipatory EMG activity, we examined whether the reaching arm 443 drifted slowly inwards, given that all targets appeared medial relative to starting hand positions.

444
To do this, we compared the position of the hand at baseline versus immediately before the 445 express arm response and observed no relationship between the level of anticipatory activity and 446 any change in hand position (p > 0.05). This suggests the anticipatory activity did not move the 447 hand, perhaps because any arising forces were insufficient to overcome the inertia of the hand, or 448 because of co-contraction of unrecorded antagonist muscles.

449
A key behavioural correlation seen in previous research using unimanual tasks is that 450 larger express arm responses tend to precede shorter-latency RTs (1, 4). Given that this study is 451 the first to study express arm responses in a bimanual task, we examined our data for the 452 presence of any relationships between express arm responses and RTs. We first confirmed that 453 the express arm response magnitude in the reaching arm is negatively correlated to reach RT (left 454 panel of Figure 8A shows trial-by-trial data for the right PEC from the exemplar participant; 455 right panel of Figure 8A shows that the r-values across all participants with an express response 456 at the target of equal selection lay significantly below zero; average r = -0.3710, p < 0.001, t = 457 10.7281, df = 12). Next, we examined whether the magnitude of the express arm response on the 458 non-reaching arm related to the RT of the reaching arm, as a common drive mechanism predicts 459 that a larger express arm muscle response on the non-reaching arm should precede shorter 460 latency RTs on the reach arm. However, we found no relationship between the magnitude of the 461 express arm response on the non-reaching arm and the RT of the reaching arm either in the 462 exemplar participant (left panel of Figure 8B) or across the sample (the distribution of r-values 463 in right panel in Figure 8B does not differ from zero, average r = -0.0016, p >> 0.05, t = 0.036, 464 df = 6). Instead, as we were able to occasionally extract a RT from the movement of the non-465 reaching arm, we found a weaker negative correlation that approaches significance between non-466 reaching express arm response magnitude and non-reaching movement RT (left panel of Figure   467 8C for exemplar participant; right panel of Figure 8C for the sample; average r = -0.1879, p =   The time-series ROC analysis shown in Figure 7 shows greater anticipatory activity in 499 some participants on the arm that reached to the target, showing a degree of commitment before 500 target presentation. Such anticipatory recruitment is only a bias, as participants still reached with 501 hand closest to the most peripheral targets (Figure 2). This bias may result from trial history or 502 fatigue (e.g., favor one arm if the other arm was used on the previous trial). A bias favoring one 503 arm may explain the lack of a relationship between the magnitude of the express arm response on 504 the non-reaching arm and the RT of the reaching arm (Figure 8B), as a common drive to both 505 muscles would predict a negative relationship between the express arm response magnitude of 506 either arm and the reach RT. Instead, since the magnitude of the express arm response is also 507 influenced by anticipatory activity, a bias in anticipatory activity against the non-reaching arm 508 may have muted the magnitude of the ensuing express arm response.

538
This perspective on express arm responses blurs the distinctions between concepts such 539 as target selection, movement planning or preparation, execution, and the commitment to reach 540 with a given arm. In the emerging target task, which engages a high degree of preparation and 541 anticipation, we speculate that EMG activity arises from converging inputs from multiple 542 descending pathways, each of which has distinct characteristics and dynamics during different 543 phases of a trial. Anticipatory EMG activity is characterized by recruitment that can be biased in 544 favour of one arm or the other, starts before target presentation, and is not directed to a particular 545 target. The express arm response is driven by and directed toward a particular target (and hence 546 is arguably synonymous with both target selection and execution) and can be distributed 547 bilaterally. Finally, the commitment to reach with a particular arm is completed after the express 548 arm response and consists of unilateral recruitment of one arm and simultaneous relaxation of the 549 other. Fundamentally, the networks governing muscle recruitment in the context of this task are 550 likely nested (42-44), and as the output of the pathway that most rapidly links vision to action, 551 the express arm response offers a unique window where target-related muscle activity is solely 552 driven by subcortical descending pathways.

Comparison to past studies and methodological considerations 593
Our study is the first to investigate express arm responses when either arm could reach toward a 594 target. Further, we increased the number of potential targets from two to seven. Despite these 595 changes, all but one participant exhibited an express arm response to at least one target. We 596 attribute this to our paradigm maintaining implied motion behind the barrier and a high degree of 597 certainty about the time of target emergence, which have been suggested to be the main factors 598 increasing express arm response prevalence and magnitude (5, 12, 23).

599
Participants chose which arm reached to the target, doing so as quickly as possible.

600
Previous work has shown that arm choice tends to reflect the hemifield of target appearance,  Our task found, for each subject, a target location eliciting reaches with the right arm on 606 some trials, and the left arm on others. Doing so enabled comparison of muscle activity as a 607 function of whether the associated arm was selected to reach or not for movements to the same 608 visual target. For most participants (n = 10), this target of equal selection was the center target.

609
Assuming participants followed task instruction, this center target would be ~1 degree below the when recordings tended to be noisier. When express arm responses were detected, EMG activity 638 displayed the characteristic trial-by-trial changes more aligned to target rather than movement 639 onset (e.g., Figure 3B). 640 Our positioning of targets medial to both hands, with loading forces in the opposite 641 direction, meant that pectoralis major was the only muscle on which the bilateral distribution of  Finally, our participants were either right-handed (n = 12) or ambidextrous (n =2).

651
Previous studies of express arm responses studied few left-handed participants (1, 5, 12), but 652 there has been no suggestion of differences between left-and right-handed participants. We 653 speculate that the express arm response would remain bilateral in left-hand dominant 654 participants, but this remains to be determined.