Introduction

Human motor control spans a continuum between reflexive and voluntary actions, but a clear demarcation between what constitutes a “reflex” versus a “voluntary” response has been debated for many years (see Prochazka et al. 2000, for a Review). Adding to the complexity, a bidirectional relationship exists between some reflexive and voluntary responses. This interaction between reflexive and volitional contribution is exemplified in the muscular activity elicited when the upper limb is perturbed. When sufficient stretch of muscle occurs, short- (M1: < 50 ms) and long-latency (M2: ~ 50–100 ms) responses are observed in the electromyographic (EMG) recording of lengthened musculature (Hammond 1956). Both responses are sensitive to peripheral factors such as background state of the motoneuron pool and magnitude of the perturbation (Pruszynski et al. 2009), but the M2 response shows the remarkable feature of also modulating based on the voluntary intention of the participant (e.g., Hammond 1956; Lee and Tatton 1975; MacKinnon et al. 2000; Pruszynski et al. 2008). For example, when participants are instructed to “resist” or “compensate” against the perturbation, the M2 response appears large and continuous with a third (voluntary) response. By contrast, the instruction to “let-go” or “not-intervene” with the perturbation results in a reduced M2 and no voluntary response. It has been proposed that the M2 response displays instruction/goal-dependent modulation because the perturbation engages overlapping neural circuitry with the generation of voluntary activity (Scott 2004; Pruszynski and Scott 2012).

The elicitation of stretch responses can also influence the onset latency of the voluntary response. Voluntary RT to auditory or visual signals is typically > 120 ms (e.g., Jaeger et al. 1982; Forgaard et al. 2015); however, the mechanical perturbations used to elicit stretch responses can initiate the voluntary response at a reduced latency (< 100 ms) such that it overlaps the M2 response (Crago et al. 1976; Evarts and Granit 1976; Evarts and Vaughn 1978). The early voluntary response following a perturbation causing muscle stretch has been called a “triggered reaction” (e.g., Crago et al. 1976; Houk 1978; Manning et al. 2012). It remains a matter of contention whether the appearance of goal-dependent M2 modulation arises from excitability changes of the underlying circuitry (Hammond 1956; Lee and Tatton 1975; Pruszynski et al. 2008) or from superimposition of the voluntary response onto the M2 response (Crago et al. 1976; Rothwell et al. 1980; Manning et al. 2012; Ravichandran et al. 2013). Indeed, distinguishing between these alternatives has proven problematic because most studies have used tasks where the voluntary response occurred in the same muscle as M2, and thus resolving when the voluntary response begins to make contributions to the appearance of M2 modulation has been difficult to determine (see Forgaard et al. 2015).

Another example of how reflexive and voluntary actions interact can be seen when an individual is startled. The generalized startle response is a diffuse primarily upper body reaction following presentation of sufficiently intense stimuli. It begins with closure of the eyes followed by contraction of the neck, trunk, and arm muscles (Landis and Hunt 1939; Brown et al. 1991; Carlsen et al. 2011). A number of studies over the past 20 years have investigated the effects of a startling auditory stimulus (SAS) on a preplanned action. When a voluntary response is preprogrammed and the SAS elicits a startle response, the intended action is initiated at a latency usually less than 100 ms. This phenomenon has been termed the StartReact effect (for Reviews see Rothwell 2006; Valls-Solé et al. 2008; Carlsen et al. 2012; Nonnekes et al. 2015). Despite the early elicitation of the response, EMG and kinematic characteristics of the intended response remain relatively unchanged when compared to non-SAS trials (Valls-Solé et al. 1999; Carlsen et al. 2004b). Although the exact mechanism underlying the StartReact effect remains a matter of debate (see “Discussion”), it is generally accepted that neural activation from circuitry involved in the startle response acts to hasten the voluntary response (Valls-Solé et al. 2008; Carlsen et al. 2012; Nonnekes et al. 2015).

Investigations of the StartReact effect have shown that activation in sternocleidomastoid (SCM; within 120 ms of the SAS) is the best indicator that a startle response has occurred (Carlsen et al. 2007; Maslovat et al. 2015). For example, trials with a startle response in SCM show consistently short RTs (~ 80 ms) irrespective of intensity of the SAS, whereas trials with no activity in any startle indicators or activity in only orbicularis oculi (OOc) display a stimulus intensity effect, in which RT decreases as sound intensity increases (Carlsen et al. 2007). OOc is also activated as part of an auditory blink reflex (Brown et al. 1991) which is produced from separate circuitry from the startle response and is not implicated in the StartReact effect (Carlsen et al. 2007). It is important to note that almost all trials with true startle-related SCM activity are also accompanied by a response in OOc (Carlsen et al. 2007; Forgaard et al. 2015, 2016).

While the StartReact effect has been consistently found when using a SAS, it has been suggested that mechanical perturbations can also act as a stimulus that evokes a startle response and hence a possible initiator of the StartReact effect (Ravichandran et al. 2013; Forgaard et al. 2016). However, a majority of these studies that have examined triggered reactions and goal-dependent M2 modulation in the upper limb have used expected perturbations (e.g., Lee and Tatton 1975; Crago et al. 1976; Rothwell et al. 1980; Pruszynski et al. 2008; Manning et al. 2012; Forgaard et al. 2015). We have recently shown that while expected mechanical perturbations do elicit activity in SCM within the startle time criteria (< 120 ms), this was not accompanied by an early response in OOc and, therefore, likely not a true startle response (Forgaard et al. 2016). Instead, we suggested that the SCM activity was a result of postural responses associated with head/neck stabilization during the performance of a ballistic upper limb movement (see also Dean and Baker 2017). This calls into question using SCM as the sole indicator of startle when delivering upper limb mechanical perturbations.

The purpose of the present work was twofold: (1) we sought to determine whether the SCM activity evoked by a mechanical perturbation is produced from the reflexive startle response or postural control: and (2) whether a startle response is a prerequisite for the early (< 100 ms) elicitation of a voluntary response following a perturbation imperative signal. One way in which postural- and startle-related SCM activity can be dissociated is through the use of prepulse inhibition (PPI). PPI is the inhibitory influence of a non-startling sensory stimulus on the reflexive startle response, the effects of which are typically maximal when presented ~ 100 ms prior to a startling stimulus (Valls-Solé et al. 2005; Yeomans et al. 2006; Maslovat et al. 2012). In Experiment 1, we employed a PPI stimulus (80 dB auditory signal) in a perturbation paradigm to examine whether perturbation-evoked SCM activity results from the elicitation of a startle response, or is instead a result of postural stabilization. Participants were instructed to perform a ballistic wrist flexion response against a wrist extension perturbation (Lee and Tatton 1975; MacKinnon et al. 2000; Manning et al. 2012; Forgaard et al. 2015, 2016). On random trials, a PPI stimulus was delivered 100 ms prior to the perturbation. If the perturbation-evoked SCM activity is the result of startle, we hypothesized that PPI would attenuate SCM responses (similar to auditory startle findings). By contrast, the PPI stimulus was not expected to have an inhibitory influence on SCM if it is activated as part of its postural role in stabilizing the head/neck.

Experiment 2 investigated whether a startle response is required for a perturbation to elicit the voluntary response at < 100 ms. One confound when participants perform a movement in opposition to a perturbation (i.e., compensate/resist task) is that RT cannot be determined because the voluntary response occurs in the same muscle and over a similar time-course as the M2 response (Hammond 1956; Manning et al. 2012; Forgaard et al. 2015). To overcome this, in Experiment 2 participants performed wrist extension movements following a wrist extension perturbation (i.e., assist task; Jaeger et al. 1982; MacKinnon et al. 2000; Miscio et al. 2001). Similar to Experiment 1, the perturbation elicited stretch responses (M1 and M2) in wrist flexors. Moving the voluntary response to wrist extensors allowed RT to be determined on a trial-by-trial basis because the voluntary response was not preceded by stretch responses in the same muscle. To permit startle incidence and RT comparisons with the StartReact effect, on random trials the perturbation was either replaced with a SAS probe, or was accompanied by a SAS probe. To obtain a high percentage of startle responses, intensity level of the startle probe is usually > 120 dB (e.g., Valls-Solé et al. 1999; Carlsen et al. 2011). However in Experiment 2, we specifically used a less intense, 115 dB SAS, to obtain a distribution of trials both with and without a startle response (Carlsen et al. 2007, 2009). This procedure also allowed us to obtain two RT distributions on SAS trials; one which was consistently short (< 100 ms) and was associated with startle, and another which was longer (> 100 ms) and not accompanied by a startle response (Maslovat et al. 2015). Of primary interest was the distribution of RTs obtained on the control (perturbation only) trials. If a voluntary response can be elicited at latencies < 100 ms only in the presence of a startle response, we hypothesized that these early RTs would only be observed on the trials where participants were startled. By contrast, if a mechanical perturbation can elicit the voluntary response at < 100 ms even in the absence of startle, we reasoned that the preprogrammed voluntary response would consistently be elicited early.

Methods

Participants

A total of twenty (11 female, 9 male, aged 20–27 years) healthy right-handed volunteers participated in these experiments. Fourteen participants were tested in Experiment 1 and thirteen were tested in Experiment 2. Six participants completed both Experiments (testing sessions were separated by a minimum of 2 weeks). All protocols were approved by the University of British Columbia Behavioural Research Ethics Board and conducted in accordance with the Declaration of Helsinki. Informed written consent was obtained from each participant prior to testing.

General setup and procedures

The experiments took place in an acoustically dampened chamber. Participants sat on a height-adjustable seat and were positioned with the right arm in a manipulandum that constrained movement to flexion/extension of the wrist. The right elbow was flexed at 100° and the hand was semi-pronated with the wrist joint aligned to the rotational axis of the manipulandum. Connected to the manipulandum was an Aeroflex torque motor (TQ 82W-1C). A metal handle adjoined to the motor shaft was placed near the right metacarpophalangeal joints and lateral wrist movements were prevented by padded stops. Continuous position feedback of the right wrist was provided on an oscilloscope placed 0.5 m in front of the participant. An LED lightbox was placed on top of the oscilloscope and a loudspeaker positioned 30 cm behind the participants’ ears. Each trial started with an auditory warning signal (80 dB, 50 ms, 500 Hz) and an extension preload increased over 500 ms to 0.25 Newton metres (N m). Participants were instructed to “resist by lightly activating wrist flexors against the load and to hold their wrist at the home position”. An evenly distributed random foreperiod (2500–3500 ms) followed the warning signal and was terminated by an imperative signal (which differed based on the experiment, see below). Following the imperative signal, the preload level of extension torque (0.25 N m) was maintained for 1000 ms.

Experiment 1

For Experiment 1, thermoplastic casting surrounded each participants’ right hand such that wrist movement could occur without having to grasp onto the handle. We have previously used this setup for stretch response studies as it allows the wrist to move with minimal activation of hand musculature (Forgaard et al. 2015, 2016). The home position was 0° of wrist flexion and visually marked on the oscilloscope. Participants were instructed to “flex their wrist as fast as possible following onset of the imperative signal”. On some random trials, a PPI stimulus (80 dB, 50 ms, 1000 Hz) was presented prior to the imperative signal but participants were instructed that “this auditory stimulus was irrelevant to the task”. This experiment consisted of two protocols, one (Experiment 1A) to replicate the effects of PPI on the auditory startle response and StartReact effect (Valls-Solé et al. 2005; Maslovat et al. 2012) and a second (Experiment 1B) to examine the effects of the same PPI stimulus on perturbation-evoked SCM activity. Protocol order was counterbalanced between the 14 participants.

In Experiment 1A, the imperative signal was a green LED light appearing for 150 ms. On 10% of trials, a SAS (120 dB, 50 ms, 1000 Hz) was presented simultaneous with the LED (condition name: SAS). On another 10% of trials, a PPI stimulus was presented 100 ms before the SAS (PPI-SAS). An additional 10% of trials had a PPI stimulus 100 ms before the imperative signal (PPI-Control) and a final 10% of trials had a PPI stimulus but the imperative signal was withheld (PPI-Catch). Prior to testing, participants were given 20 practice trials for Control, PPI-Control, and PPI-Catch. Within the testing block (1 block of 50 trials), trial order was randomized with the stipulation that no trials with a SAS occurred within the first 3 trials, nor were any presented on consecutive trials.

For Experiment 1B, the imperative signal was a wrist extension perturbation (1.5 N m, 150 ms). On 40% of trials, only the imperative signal was presented (Flex). On another 40% of trials, a PPI was presented 100 ms before the imperative signal (PPI-Flex). The remaining 20% of trials consisted of a PPI stimulus but no imperative signal (PPI-Catch). Trial order was fully randomized. Practice involved 20 trials and the testing phase consisted of two blocks of 40 trials. To prevent fatigue, 2 min of rest was provided between blocks.

Experiment 2

For the second experiment, participants were instructed to grasp onto the handle connected to the motor shaft. The home position for each trial was 10º of wrist flexion and defined on the oscilloscope. The imperative signal was a wrist extension perturbation (1.5 N m, 50 ms). Pilot testing determined that it was easier to respond with an extension response to a shorter perturbation (50 vs 150 ms). The longer perturbation used in Experiment 1B resulted in the wrist moving too far in extension to comfortably perform a voluntary extension response. Participants were instructed to “extend their wrist as fast as possible following onset of the perturbation”. On 10% of trials, a SAS probe (115 dB, 50 ms, 1000 Hz) was presented simultaneous with the perturbation (SAS-Extend). On a separate 10% of trials, a SAS probe was presented but no perturbation was delivered (SAS-Catch). On an additional 10% of trials, no perturbation and no SAS probes were presented (Catch).

Experiment 2 began with 30 trials of practice in which only Control (Extend; 27 trials) and Catch (3 trials) trials were presented. The testing phase included five blocks of 40 trials (200 total trials). Each block contained 28 Extend, 4 SAS-Extend, 4 SAS-Catch, and 4 Catch trials. Trial order was pseudorandomized such that no trials with a SAS probe occurred within the first three trials of a block and no two trials in a row contained a SAS probe. Two minutes of rest was provided after each block.

Analysis and classification of startle

Surface EMG recordings were taken from the muscle bellies of right extensor carpi radialis (ECR), right flexor carpi radialis (FCR), left and right orbicularis oculi (OOc), and left and right sternocleidomastoid (SCM) using pre-amplified surface electrodes connected to an external amplifier (Model DS-80, Delsys Inc., Natick, MA). EMG signals were amplified at 3K and bandpass filtered from 20 to 450 Hz. Signals were digitized at 2 kHz using a 1401plus data acquisition system and Spike2 (CED, Cambridge, UK) computer software and a custom-written LabVIEW (National Instruments, Austin, TX) program was used for offline data analysis.

At the beginning of analysis, EMG data were DC offset corrected (mean baseline value was determined and subtracted from each data point) and full-wave rectified. Burst onsets [in left/right OOc and SCM (both experiments), right FCR (Experiment 1A), right ECR (Experiment 2)] were marked as the first point at which rectified EMG traces began a sustained rise above baseline levels (mean of activity for 200 ms preceding the imperative signal on a trial-by-trial and an individual muscle basis). EMG traces were displayed on a computer monitor with a superimposed marker indicating the initial point where activity increased to more than 3SD above baseline and remained above this level for at least 10 ms. Onset markers were verified visually and adjusted if necessary (due to the strictness of the algorithm) to the onset of the appropriate muscle burst. For Experiment 1B, integrated wrist flexor EMG activity was calculated (on a trial-by-trial basis) from the pre-defined stretch response epochs (M1: 25–50 ms; M2a: 50–75 ms; M2b: 75–100 ms). Due to the elicitation of these stretch responses, RT could not be determined (Manning et al. 2012; Forgaard et al. 2015), but differences between conditions could be inferred from the M2b time period (Ravichandran et al. 2013; Forgaard et al. 2015).

In StartReact studies, short-latency activation (< 120 ms) in SCM is used as the gold-standard measure of startle detection (Carlsen et al. 2011). Trials with only an auditory blink reflex (observed in OOc), or no responses in any startle indicator muscle do not show consistently early RTs (Carlsen et al. 2007). However, trials with true startle-related SCM (following either a perturbation or a SAS) have consistently early RTs and are accompanied by a response in OOc (Carlsen et al. 2007; Forgaard et al. 2016). By contrast, trials with just postural SCM activity do not typically have early OOc activity (Carlsen et al. 2004b; Forgaard et al. 2016). Therefore, to aid in separating startle and postural SCM activity, we defined a positive startle response as bilateral activity in OOc and SCM within 120 ms of a given stimulus (Forgaard et al. 2015, 2016).

To quantify the effects of the PPI stimulus on SCM activity (Experiment 1A/B), rectified SCM data were collapsed across the left and right sides and integrated from 40 to 90 ms. Previous work has shown that an auditory PPI stimulus presented 100 ms before a SAS, significantly attenuates SCM activity during this interval (Maslovat et al. 2012). The PPI-SAS data were expressed relative to integrated SAS data (Experiment 1A) and the PPI-Flex data were expressed relative to the integrated Flex data. A normalized value of less than 1.0 reflects attenuation and a value greater than 1.0 indicates facilitation.

Data reduction

Analysis for the Experiment 1A focused on the eight participants who consistently (> 50%) displayed a startle response on (non-PPI) SAS trials. 6.0% of trials were omitted due to false starts [either responding before the IS (18 trials), or responding on a PPI-Catch trial (6 trials)]. Experiment 1B focused on data from 10 participants who displayed SCM activity on more than 20% of Flex trials. This included the same eight participants from Experiment 1A and an additional two participants who did not startle on SAS trials. Experiment 1B had more trials (32 trials per condition; see above) compared to the SAS trials in Experiment 1A (5 trials); thus we felt confident including participants with a lower incidence of SCM compared to startle incidence in Experiment 1A. For Experiment 1B, 2.1% of trials from the 10 participants were omitted due to false starts [responding before imperative signal (3 trials) or responding on a PPI-Catch trial (14 trials)] and 0.9% were excluded because no voluntary response appeared following the imperative signal (7 trials).

For Experiment 2, data were omitted from one participant due to an abnormal response profile on SAS-Catch trials. Although this participant consistently was startled on these trials, the response that was elicited in the forearm was one of flexion followed by extension (the task required extension). Because this participant appeared to have been preprogramming a different motor response than what was specified, we excluded their data. No other participants displayed this behaviour on SAS-Catch trials. From the remaining 12 participants, only 1.9% of trials were removed due to any errors. Reasons for omission included trials where participants began their response prior to the imperative signal (i.e., false starts; 22 trials) or did not respond (4 trials). In addition, we excluded any trials that fell more than three standard deviations from a participant’s mean RT within a given condition (19 trials).

Dependent measures and statistics

All incidence data were converted to a proportion of trials from a given condition and arcsine transformed prior to statistical analysis. Non-transformed mean values are reported alongside statistical results conducted on the transformed data. Onset Latency values were only taken on trials from a given muscle if a response was detected in that muscle.

For Experiment 1A, Startle Incidence and RT data were analyzed using a SAS (SAS, No SAS) × PPI (PPI, No PPI) repeated measures analysis of variance (ANOVA). OOc/SCM Onset Latency values were compared on SAS trials with a Condition (SAS, PPI-SAS) × Side (Left, Right) ANOVA. For Experiment 1B, Startle Incidence and wrist flexor stretch response epoch (M1, M2a, M2b) data were evaluated using paired samples t-tests comparing Flex and PPI-Flex. Startle indicator measures (OOc/SCM Incidence, Onset Latency) were examined with a Condition (Flex, PPI-Flex) × Side (Left, Right) repeated measures analysis of variance (ANOVA). For both Experiment 1A and 1B, the normalized integrated SCM data were analyzed with a one-sample t-test against a value of 1.0.

For Experiment 2, repeated measures ANOVA tests were used to compare the effect of Condition (Extend, SAS-Extend, SAS-Catch) on Startle Incidence and RT. We also report incidence and onset latency of responses in startle indicator muscles (Left/Right OOc and Left/Right SCM) on Extend trials.

In the event of any violations to the assumption of sphericity for an ANOVA with more than two factors (determined by Mauchly’s test), Greenhouse-Geisser corrected degrees of freedom were reported, alongside the corrected p values. Partial eta-squared (\(\eta _{p}^{2}\)) was used as a measure of effect size. Dunn–Bonferroni corrected t-tests were employed to interpret significant main effects and significant interaction effects were qualified using simple main effects analysis. Standard error of the mean is reported as an indicator of variability and the level of statistical significance for each test was set at p = 0.05.

Results

Experiment 1A: replicating the effects of PPI on the startle response and StartReact effect

Analyzing startle incidence data revealed a significant main effect for SAS, F(1,7) = 382.42, p < 0.001, \(\eta _{p}^{2}\) = 0.98, that was superseded by a SAS × PPI interaction F(1,7) = 8.41, p = 0.023, \(\eta _{p}^{2}\) = 0.55. Simple main effects analysis indicated that PPI-SAS had a lower startle incidence [p = 0.032; 61.3% (± 9.8%)] than SAS [91.9% (± 3.7%)]. On non-SAS trials by contrast, presentation of a PPI stimulus did not influence startle incidence [p = 0.305; PPI-Control: 5.0% (± 3.1%) vs. Control: 0.4% (± 0.4%)].

Overall, on SAS trials, the PPI stimulus appeared to attenuate activity in startle indicator muscles (see Fig. 1). An analysis of OOc Onset Latency uncovered main effects for PPI, F(1,7) = 7.06, p = 0.033, \(\eta _{p}^{2}\) = 0.50, indicating that OOc onset was delayed [52.9 ms (± 4.4) vs. 40.9 ms (± 1.3)] following the PPI stimulus. There were no significant main effects for Side (p = 0.699) or PPI × Side interactions (p = 0.488). For SCM, there was also a PPI main effect for Onset Latency, F(1,7) = 5.90, p = 0.046, \(\eta _{p}^{2}\) = 0.46. SCM onset was also significantly later on PPI-SAS trials [100.0 ms (± 8.3) vs. SAS: 77.6 ms (± 6.0)]. The Side main effect (p = 0.139) and PPI × Side interaction (p = 0.825) were non-significant.

Fig. 1
figure 1

Group ensemble startle indicator data for Experiment 1A. SAS condition solid black lines; PPI-SAS solid grey lines

The analyses above relied on an EMG burst marking algorithm to determine muscle onsets. It is possible that on some PPI-SAS trials, responses in startle indicators were too small to be detected. Thus, we quantified the integrated EMG activity regardless of the presence of a muscle burst. For PPI-SAS trials, the integrated SCM activity from 40 to 90 ms was significantly reduced, t(7) = − 5.13, p = 0.004. The mean normalized value was 0.47 (0.12) which corresponds to a 53% reduction in integrated SCM activity, compared to the SAS condition (see Fig. 1).

An examination of RT revealed a main effect for SAS, F(1,7) = 103.27, p < 0.001, \(\eta _{p}^{2}\) = 0.94, but the main effect for PPI (p = 0.281) and the SAS × PPI interaction (p = 0.127) were non-significant. This indicates that the presentation of a SAS significantly reduced RT [SAS: 106.0 ms (± 7.2) and PPI-SAS: 108.9 ms (± 5.8)] compared to Control conditions [Control: 211.7 ms (± 10.7) and PPI-Control: 179.1 ms (± 15.4)]. In summary, the findings of Experiment 1A confirm previous reports (Castellote et al. 2017; Maslovat et al. 2012; Valls-Solé et al. 2005) that a PPI stimulus given 100 ms prior to SAS can attenuate the startle response without a significant impact on the StartReact effect.

Experiment 1B: a PPI stimulus does not attenuate perturbation-evoked SCM activity

The purpose of Experiment 1B was to examine whether a PPI stimulus could influence SCM activity following a perturbation. Rather than attenuating responses in startle indicators, as would be expected for the startle response, the PPI stimulus significantly advanced the SCM responses and had no influence on OOc activity (see Fig. 2). A low proportion of trials in this Experiment were classified as displaying startle (Flex: 4.8%; PPI-Flex: 8.8%; p = 0.501). On average, SCM activity was observed on 92.1% (± 10.7) of trials but no incidence differences were observed between left and right SCM (p = 0.567) or between trials with or without a PPI stimulus (p = 0.734). By contrast, analysis of SCM onset data showed main effects for PPI, F(1,9) = 45.43, p < 0.001, \(\eta _{p}^{2}\) = 0.84, and Side, F(1,9) = 5.99, p = 0.037, \(\eta _{p}^{2}\) = 0.40, without a significant interaction (p = 0.529). SCM activity occurred 17.4 ms earlier on PPI-Flex trials and 7.7 ms earlier in the right SCM compared to left [L-PPI-Flex: 123.0 ms (± 6.0); L-Flex: 139.8 ms (± 6.8); R-PPI-Flex: 114.7 ms (± 6.8); R-Flex 132.7 ms (± 6.7)]. OOc responses by contrast, appeared on 57.2% of trials at a mean onset of 164.9 ms. No OOc measures were significantly different between conditions [Incidence (Side: p = 0.827; PPI: p = 0.203; Side × PPI: p = 0.871); Onset Latency (Side: p = 0.258; PPI: p = 0.946; Side × PPI: p = 0.945)].

Fig. 2
figure 2

Group ensemble displacement and EMG data for Experiment 1B. Flex condition solid black lines; PPI-Flex solid grey lines. The perturbation moved the wrist into extension and participants executed a wrist flexion movement. Stretch responses and the voluntary response appeared in FCR. Panel A: All EMG data on same Y-Scale. Note the postural activity in SCM. Also note the increased M2b activity and earlier voluntary response on PPI-Flex trials. In panel B, SCM data have been magnified. Note how the responses were advanced on PPI-Flex trials

A one-sample t-test comparing the normalized integrated SCM activity from 40 to 90 ms on PPI-Flex to a value of 1.0, did not reveal any significant attenuating effects of the PPI stimulus, t(9) = 1.71, p = 0.122. The mean normalized value was 1.42 (± 0.23) indicating that activity was in the direction of facilitation on PPI-Flex trials. Given that the startle response to a perturbation or a somatosensory stimulus is ~ 20 ms later than the auditory startle response (Álvarez-Blanco et al. 2009; Forgaard et al. 2016; Ravichandran et al. 2013), we moved the integration window 20 ms later (60–110 ms) and we re-ran the same analysis. However, rather than finding attenuation of SCM, we now observed significant facilitation on PPI-Flex trials, t(9) = 2.50, p = 0.034. Integrated SCM activity was 1.61 (± 0.23) times larger on PPI-Flex trials compared to Flex trials (see Fig. 2).

Analysis of the wrist flexor stretch response data showed no significant differences during the M1 (p = 0.672) or M2a epochs (p = 0.453). M2b activity, however, was significantly increased, t(9) = 3.12, p = 0.012, on PPI-Flex trials [4.2 mV*ms (± 0.7) compared to Flex 3.4 mV*ms (± 0.6)], suggesting that the voluntary response may have occurred earlier on PPI-Flex trials which resulted in increased superimposition onto the end of the M2 response (Fig. 2).

Experiment 2: a mechanical perturbation elicits an early voluntary response in the absence of startle

An analysis of Experiment 2 startle incidence data revealed a significant main effect of Condition, F(2,22) = 47.47, p < 0.001, \(\eta _{p}^{2}\)= 0.81. The highest incidence of startle occurred for SAS-Extend [p values < 0.001; 68.2% (± 8.8%)] compared to SAS-Catch [18.1% (± 7.6%)] and Extend [2.8% (± 1.6%)]. The latter two conditions also differed significantly (p = 0.029). With respect to our initial objective, the 115 dB SAS produced a distribution of trials with and without a startle responseFootnote 1 (see below) and the near zero incidence of startle for the Extend condition was similar to Experiment 1B as well as previous findings that a mechanical perturbation imperative signal does not consistently elicit a startle response (Forgaard et al. 2015, 2016). An analysis of RT showed a main effect of Condition, F(1.02,11.24) = 6.51, p = 0.026, \(\eta _{p}^{2}\) = 0.37. RT for the SAS-Extend condition was significantly shorter [80.8 (± 2.2) ms] than SAS-Catch [p = 0.035; 118.6 (± 12.9) ms] and Extend [p < 0.001; 96.8 (± 1.9) ms]; however the values for SAS-Catch and Extend conditions did not differ significantly (p = 0.379).

Based on our startle response criteria (see “Methods”), every trial was categorized as either displaying a startle response (Startle+) or not displaying a startle response (Startle−).Given that the objective of Experiment 2 was to determine whether a mechanical perturbation could elicit StartReact-like RTs in the absence of a startle response, we re-analyzed RT values from the trials for Extend Startle−with the trials for SAS-Extend Startle+ and SAS-Catch Startle+. Despite reducing the number of participants to nine (three participants had no SAS-Catch Startle+ trials), we observed a significant main effect of Condition, F(2,16) = 29.68, p < 0.001, and a stronger effect size (\(\eta _{p}^{2}\) = 0.79), compared to the preliminary test of RT. The post-hoc test also showed a different pattern of results. RT for the SAS-Extend Startle+ [76.4 (± 2.2) ms] and SAS-Catch Startle+ [77.7 (± 3.0) ms] conditions were not significantly different (p > 0.999), but both SAS conditions were earlier than Extend Startle− [95.6 (± 1.8) ms; SAS-Extend Startle+ p < 0.001; SAS-Catch Startle+ p = 0.002]. This finding provides evidence that a mechanical perturbation does not consistently elicit RTs as short as the StartReact effect. See Fig. 3 for ensemble average displacement and kinematic data from the SAS-Extend Startle+, SAS-Catch Startle+, and Extend Startle− conditions. A distribution of RTs (in 10 ms bins) from these conditions is also presented in Fig. 4a. Note how the SAS-Extend Startle+ and SAS-Catch Startle+ RT distributions were consistently early and centred around 70–80 ms (Fig. 4a) and the Extend Startle− RT distribution was centred around 90–100 ms. Although this Extend Startle− distribution was significantly later than SAS-Extend Startle+ and SAS-Catch Startle+, 67.9% of trials still had RT values of 100 ms or less.

Fig. 3
figure 3

Group ensemble displacement and EMG data for Experiment 2. Panel A: Extend Startle−. Panel B: SAS-Extend Startle−. Panel C: SAS-Catch Startle+. The perturbation moved the wrist into extension and participants executed a wrist extension movement. Stretch responses appeared in FCR (c.f. panel C) and voluntary response appeared in ECR. No stretch responses were observed in Panel C because the perturbation was not delivered. In panel A, note how the voluntary response began at < 100 ms in the absence of a startle response. Postural activity was observed in SCM muscles. In panels B and C, note how a startle response was observed (early activation of OOc and SCM) and the voluntary response began at ~ 70 ms

Fig. 4
figure 4

Experiment 2 RT data distributed into 10 ms bins. Panel A: Extend Startle− (solid grey), SAS-Extend Startle+ (solid black), and SAS-Catch Startle+ (dashed grey) conditions. Y-axis normalized by total number of trials classified as displaying the startle response (SAS-Extend Startle+ and SAS-Catch Startle+) or not displaying startle (Extend Startle−). Panel B: SAS-Catch Startle+ (dashed grey) and SAS-Catch Startle− conditions normalized in the same manner as A. The number of trials in SAS-Catch Startle− appears smaller because a response in ECR was only elicited on 37.0% of these trials. By contrast a response in ECR was elicited on 97.8% of SAS-Catch Startle+ trials

Further analysis of the Catch and SAS-Catch conditions provides interesting insight into the StartReact effect. Recall that the imperative signal was not presented on these trials; therefore, correct performance required that participants withhold the preprogrammed extension response. On the Catch trials with no SAS probe, a response was only initiated on 1.7% of trials suggesting that participants successfully withheld the voluntary response when no perturbation or SAS was presented. By contrast, on SAS-Catch trials, the preprogrammed response was elicited on 46.4% (± 8.9%) of trials but the probability and latency of response triggering differed considerably in the presence versus the absence of the startle response. On SAS-Catch Startle+ trials, the preprogrammed response was elicited on 97.8% (± 2.1%) of trials with a mean RT of 77.7 ms (± 3.0 ms). For the SAS-Catch Startle− trials, the preprogrammed response only appeared on 37.0% (± 8.8%) of trials and the mean onset latency was 131.4 (± 13.4) ms. See Fig. 4B for a distribution of RT values for SAS-Catch Startle+ and SAS-Catch Startle−. The SAS-Catch Startle+ RT values were consistently early and at a typical StartReact latency suggesting that the voluntary response was triggered by the SAS. By contrast, the SAS-Catch Startle− trials fell into two distributions; one which was early (i.e., StartReact-like) and another which was later and more variable.

In a previous study (Forgaard et al. 2016), and Experiment 1B of the current project, we found that when participants compensated against a wrist extension perturbation, a considerable number of trials (> 50%) showed postural SCM activity. We observed similar SCM responses in Experiment 2 where participants performed Extend responses to the perturbation. SCM activity was observed on 68.3% (± 10.9%) of trials at a mean onset latency of 130.2 (± 7.2) ms, which followed onset of the ECR response (i.e., RT) by 37.3 (± 7.8) ms. In Experiment 1B, we found R-SCM to occur ~ 7 ms earlier than L-SCM; however in Experiment 2, L-SCM (127.3 ms) now preceded R-SCM (133.1 ms) by ~ 6 ms, but this was not significantly different (p = 0.283). OOc activity was observed on 50.4% (± 10.3) of trials at a mean onset of 145.5 (± 13.2) ms.

Control experiment: investigating the presence of a shortening reaction in ECR

A protocol in which the voluntary response occurs in shortened muscle has been used by a number of groups to determine perturbation RT (Evarts and Granit 1976; Evarts and Vaughn 1978; MacKinnon et al. 2000; Miscio et al. 2001; Ravichandran et al. 2013). When using this method, it is important to consider whether any muscular activity in the shortened muscle occurs as a natural consequence of the mechanical perturbation stimulus. It has been reported that passively shortened muscles can display a shortening reaction (Katz and Rondot 1978) which begins > 100 ms after perturbation onset (Miscio et al. 2001). Although the origin of this response remains unknown, it was proposed to be a deliberate response to counteract the effects of the stretch response (Miscio et al. 2001). Due to the close proximity of forearm flexors and extensors, some participants may also display “volume conduction” or “EMG cross talk” in the shortened muscle around the onset of the M2 response in the stretched antagonist (Miscio et al. 2001). The presence of either of these responses in the ECR data of our Experiment 2 protocol could confound our trial-by-trial determination of RT. Therefore, to determine the presence of volume conduction or a shortening reaction in ECR, we re-tested three participants. Testing involved 28 Extend trials and 28 passive trials where they were instructed to “not intervene following the perturbation” (DNI). Surface EMG was recorded from ECR and FCR, and responses were marked in the same manner as Experiment 2. The same marking algorithm was used on the DNI trials to determine the presence of either a shortening reaction or volume conduction in the EMG recording of ECR.

Ensemble displacement and EMG data for each participant is presented in Fig. 5. Participant 1 had a mean RT on Extend trials of 87.2 ms. On 21.4% of DNI trials, the EMG marking algorithm revealed a small response at a mean of 135.4 ms. The Q30 (integrated rectified EMG over first 30 ms of the burst) was 21.0 times smaller than the Q30 of the voluntary response on Extend trials. The mean RT for participant 2 was 94.2 ms; however on 64.3% of DNI trials, the EMG marking algorithm found activity at a mean of 53 ms. This was likely the volume conduction observed by Miscio et al. (2001) as it occurred near the onset of M2 response in FCR. When it appeared, the Q30 was 7.1 times smaller than the Q30 of the voluntary response on Extend trials. For participant 3, no volume conduction was observed, but 14.3% of DNI trials had a response at a mean of 114 ms. This response was 18.3 times smaller than the Q30 of the voluntary response on Extend trials. In summary, the data from Experiment 2 were unlikely to be influenced by a shortening reaction in ECR. Volume conduction around 50 ms could have been observed in some participants but we have two lines of evidence to suggest it did not confound the marking of RT. (1) No Extend trials had RTs of ~ 50 ms and (2) the SAS-Catch Startle+ (which would not have a volume conduction issue in ECR around 50 ms because no perturbation was delivered) had a similar RT distribution to SAS-Extend Startle+ (see Fig. 4A).

Fig. 5
figure 5

Ensemble kinematic and EMG data from the three participants in the control experiment to determine the level of quiescent EMG activity in ECR while participants were exposed to an extension perturbation. Extend data: solid black lines. DNI (passive) data: solid grey lines

Discussion

The long-latency stretch response, perturbation-triggered reactions, and startle-triggered responses are all examples of rapid motor responses that obscure the boundary between “reflexive” and “voluntary” motor control. While auditory startle-evoked responses have traditionally been studied separately from mechanical perturbation responses, recent work has posited that mechanical perturbations may also elicit a startle response and, therefore, the StartReact effect could underlie perturbation-triggered reactions and goal-dependent M2 modulation (Lewis et al. 2006; Shemmell et al. 2009; Ravichandran et al. 2013). The present study examined whether activity in startle indicator muscles following a perturbation imperative signal results from the startle response or postural control (Experiment 1B) and whether an overt startle response is a prerequisite for perturbation RTs of < 100 ms (Experiment 2). Our findings provide clear evidence that a mechanical perturbation does not elicit startle when presented as the imperative signal in a simple RT paradigm. Despite the absence of startle, a majority of trials had RTs of < 100 ms. It would, therefore, appear that a startle/StartReact mechanism is not necessary for the initiation of a perturbation-triggered reaction.

A prepulse inhibition stimulus reveals postural SCM activation

In Experiment 1A, we replicated previous findings showing that activation of SCM can be strongly attenuated when a PPI stimulus is presented immediately before a startling stimulus (Fig. 1; Valls-Solé et al. 2005; Maslovat et al. 2012; Castellote et al. 2017). In stark contrast to what was expected if a mechanical perturbation elicited startle, the SCM responses observed in Experiment 1B were significantly advanced (17.4 ms) on trials with a PPI stimulus (see Fig. 2). Therefore, when participants execute ballistic upper limb movements following a perturbation, rather than the SCM activity being a result of a startle response, we believe it was part of a postural response associated with head and neck stabilization (Forgaard et al. 2016; Dean and Baker 2017). This interpretation is further supported by the asymmetric activation of SCM muscles. A hallmark feature of the startle response is bilateral activation of homologous startle indicator muscles (Carlsen et al. 2011); however on Flex trials, we found that activation in the right SCM preceded the left by 7.1 ms. Despite the advancement on PPI-Flex trials, the asymmetric activation (R-SCM 8.3 ms earlier than L-SCM) was preserved, suggesting that the postural activation in homologous neck muscles was likely preprogrammed and executed together.

In the previous studies examining the effects of PPI on the startle response and StartReact effect, and Experiment 1A of the current work, RT on SAS trials was insensitive to the presence of a PPI stimulus (i.e., the StartReact effect was not influenced by PPI; Valls-Solé et al. 2005; Maslovat et al. 2012; Castellote et al. 2017). Control (non-SAS) trials by contrast were shown to have reduced RTs on PPI trials, a result of inter-sensory facilitation and/or direct cueing of imperative signal delivery (Valls-Solé et al. 2005; Maslovat et al. 2012). Although our Experiment 1B protocol did not allow for a direct determination of RT, differences between conditions could be inferred from the stretch response epoch data. The M1 and M2a epochs were not affected, however activity during the M2b period increased on PPI-Flex trials (Fig. 2). The latter portion of the M2 response (M2b) can be influenced by superimposition of a voluntary response (Rothwell et al. 1980; Ravichandran et al. 2013; Forgaard et al. 2015). We, therefore, propose that the increased M2b activity on PPI-Flex trials resulted from a reduced RT and increased voluntary response superimposition. This finding is in accordance with non-SAS PPI RT data (Valls-Solé et al. 2005; Maslovat et al. 2012) and taken together with the advanced asymmetric SCM responses on PPI trials, provides strong evidence that a mechanical perturbation imperative signal does not elicit startle.

Potential mechanisms for a perturbation-triggered reaction

The StartReact effect is the hastening of a preplanned voluntary response by activation startle circuitry. The original explanation was put forth by Valls-Solé and colleagues in the 1990s to account for RTs on SAS trials that were considered too short for the minimal amount of time needed for cortical processing of sound and conduction time from cortex to muscle. These authors proposed that a motor program, or preprogrammed voluntary response, may be stored subcortically (e.g., in reticular formation). Presentation of a SAS elicits the startle response and activation of startle circuitry also triggers the preprogrammed response early via the reticulo-spinal tract (Valls-Solé et al. 1999). Evidence supporting this hypothesis has come from studies showing that the StartReact effect is most reliably observed in muscles with strong reticulo-spinal connections (e.g., Carlsen et al. 2009) and patients with hereditary spastic paraplegia (axonal degeneration of the cortico-spinal tract) have an intact StartReact effect, despite delayed RTs on control trials (Nonnekes et al. 2014). An alternative to the subcortical storage/triggering hypothesis proposes that motor commands for a voluntary response are preprogrammed and stored cortically prior to initiation (Carlsen et al. 2012). On SAS trials, activation of startle circuitry results in ascending neural activation which travels a reticulo-thalamo-cortical route and initiates the preprogrammed response early via the cortico-spinal tract (Alibiglou and MacKinnon 2012; Carlsen et al. 2012; Stevenson et al. 2014). This hypothesis is based on findings that the StartReact effect and startle response can be dissociated. For example (and as replicated in Experiment 1A), presentation of a PPI stimulus ~ 100 ms before a SAS reduces the magnitude of activity in startle response indicators, without impacting the triggering effects of the startling stimulus on the preprogrammed voluntary response (Valls-Solé et al. 2005; Maslovat et al. 2012; Castellote et al. 2017). Along a similar line of evidence, a transcranial magnetic stimulus induced silent period of primary motor cortex delays the StartReact effect, without influencing the startle response (Alibiglou and MacKinnon 2012; Stevenson et al. 2014). Further evidence of a dissociation between the StartReact effect and the startle response are “StartReact-like” RTs on a minority of trials where no startle response was detected (Delval et al. 2012; Maslovat et al. 2015).

Recent work has drawn comparisons between the StartReact effect and perturbation-triggered reactions (Koshland and Hasan 2000; Lewis et al. 2006; Shemmell et al. 2009, 2010; Ravichandran et al. 2013; Shemmell 2015). It is well documented that activation of startle circuitry during situations of motor preparation results in a hastening of the preprogrammed voluntary response and this effect has been most commonly demonstrated using intense auditory stimuli (Valls-Solé et al. 2008; Carlsen et al. 2012; Nonnekes et al. 2015). However, the neurons responsible for the startle response (in the caudal pontine reticular nucleus: NRPc) are not modality-specific; they can also be activated by intense proprioceptive and vestibular inputs (Yeomans et al. 2002). For instance, NRPc neurons respond at short-latency (along with neurons in other pontine reticular formation nuclei) following unexpected platform perturbations in cats (Stapley and Drew 2009). In humans, lower limb perturbations have been shown to be startling, from platform perturbations that elicit strong vestibular and proprioceptive signals (Oude-Nijhuis et al. 2010; Campbell et al. 2013) as well as single-joint knee perturbations (Castellote et al. 2017). In the context of triggered reactions in the upper limb, it has been proposed that the mechanical perturbations used to elicit stretch responses also evokes a startle response and, therefore, the StartReact effect may underlie early initiation of the preprogrammed voluntary response (Lewis et al. 2006; Shemmell et al. 2009, 2010; Ravichandran et al. 2013; Shemmell 2015).

The StartReact effect as a mechanism underlying triggered reactions is appealing and was likely responsible for the early RTs observed by Ravichandran et al. (2013), and potentially other groups that have also used unexpected perturbations (e.g., Koshland and Hasan 2000). However, a majority of studies examining triggered reactions and goal-dependent M2 modulation have presented the mechanical perturbation in an expected manner, usually as the imperative signal (e.g., Crago et al. 1976; Evarts and Vaughn 1978; Rothwell et al. 1980; MacKinnon et al. 2000; Pruszynski et al. 2008; Manning et al. 2012). The present findings complement previous work showing that expected perturbations do not elicit startle (Forgaard et al. 2016). Here, we provide clear evidence that even in the absence of startle, a mechanical perturbation imperative signal can consistently elicit RTs of < 100 ms in shortened muscle. Assuming the same RT distribution when the voluntary response occurs in stretched muscle, a majority of trials will have some superimposition of the voluntary response into the M2b time period.

Despite a lack of an overt startle response on Flex trials (Experiment 1B) and Extend trials (Experiment 2), we should consider the possibility that the perturbation may have still acted as a “sub-threshold” startling stimulus. In Experiment 2, we observed the highest incidence of startle on SAS-Extend trials (~ 68%). Because this was considerably greater than the startle incidence of the SAS-Catch condition (~ 18%), where only the 115 dB auditory stimulus was delivered, it is possible that neural activation from the SAS probe and the mechanical perturbation summated to activate the NRPc startle neurons above threshold. This is in line with animal work showing that cross-modal stimuli (e.g., tactile and auditory) are more effective at eliciting startle than intra-modal stimuli (see Yeomans et al. 2002). It also raises the possibility that on Extend trials, the mechanical perturbation may have produced sub-threshold activation (below the level needed for a startle response) of startle circuitry and is similar to our previous report showing that this sort of wrist perturbation is startling on up to 30% of trials when delivered unexpectedly (Forgaard et al. 2016). Because a somatosensory or proprioceptive startle response is ~ 20 ms later than the auditory startle response (Álvarez-Blanco et al. 2009; Oude-Nijhuis et al. 2010; Ravichandran et al. 2013; Forgaard et al. 2016), the time-course is in accordance with our observation of ~ 20 ms longer RTs on control Extend trials (compared to SAS-Extend Startle+ and SAS-Catch Startle+trials). While these results cannot distinguish between the subcortical storage and triggering hypothesis versus cortical involvement in the StartReact effect, they are in accordance with other work suggesting that the startle response and early elicitation of a preprogrammed voluntary response are sometimes dissociable (Alibiglou and MacKinnon 2012; Maslovat et al. 2012; Dean and Baker 2017). Although this is one potential explanation of sub-100 ms RTs following a perturbation, as no overt startle response was observed on Extend trials, it cannot be considered a result of the StartReact effect. This mechanism of response triggering also operates on the assumption that StartReact circuitry has a lower threshold of activation than startle response circuitry, something that has not been confirmed in the literature.

We believe it is more likely that the sub-100 ms RTs following the perturbation resulted from a different mechanism than what underlies the StartReact effect. A comparison of behavioural evidence shows definitive differences between the StartReact effect and elicitation of a voluntary response by a perturbation. For example, one hallmark feature of the StartReact effect is that it only occurs in situations where a voluntary response has been preprogrammed (Valls-Solé et al. 1999; Carlsen et al. 2004a). In choice RT conditions, where no advanced programming can occur, no hastening of the appropriate voluntary response is observed on SAS trials (Carlsen et al. 2004a). By contrast, studies examining perturbation RTs have shown that early voluntary responses occur even in choice conditions where the required motor response is unknown in advance of perturbation delivery (Crago et al. 1976; Evarts and Vaughn 1978; Glencross and Koreman 1979; Manning et al. 2012). Moreover, the onset latency increase over a simple RT condition was shown to be minimal (10–50 ms), likely a result of the high compatibility between the stimulus and the required response (the limb with the required response was directly stimulated by the perturbation). Larger increases (~ 70 ms) were seen in a non-compatible choice condition where the required response involved only responding with the opposite, unperturbed limb (Glencross and Koreman 1979). It is plausible that the sub-100 ms RTs following a perturbation are the result of the stimulus directly activating circuitry involved in producing the required voluntary response.

Upper limb perturbations produce short-latency neural responses in many circuits throughout the central nervous system including areas involved in voluntary movement control such as the cerebellum, red nucleus, posterior parietal cortex, primary sensory cortex, primary motor cortex and dorsal premotor cortex (Evarts and Tanji 1976; Strick 1983; MacKinnon et al. 2000; Pruszynski et al. 2014; Herter et al. 2015; Omrani et al. 2016). These supra-spinal areas also make contributions to the long-latency stretch response and it has been proposed that this stretch response displays characteristics of voluntary movement control because of this shared neural circuitry (Scott 2004; Pruszynski et al. 2008). Extending this, we believe the voluntary response can be initiated at a remarkable (< 100 ms) latency because the perturbation directly facilitates activation of circuitry involved in producing the voluntary response. This explanation could account for early voluntary responses in both simple and compatible choice tasks, whereas the StartReact effect only accounts for voluntary response triggering in a simple RT paradigm.

The triggered reaction

Reports of perturbation-triggered reactions have appeared in the literature since the 1970s; however, there have been disagreements on the precise nature of these responses. Part of the contention appears to stem from different tasks, effectors studied, and various different types of stimuli delivered. The original reports described the triggered reaction in upper limb musculature as a voluntary response that appears at an earlier latency to a proprioceptive stimulus than the voluntary response following stimuli of other modalities (Crago et al. 1976; Evarts and Vaughn 1978; Houk 1978). Similar to this, the StartReact effect is the mechanism underlying startle-triggered reactions, which are also believed to be the same voluntary response as non-startle trials, just initiated earlier and via a separate pathway (Valls-Solé et al. 1999; Carlsen et al. 2004b). Other authors have described triggered reactions as rapid postural adjustments following a mechanical perturbation that compromises whole body stability during standing balance (Nashner and Cordo 1981). However, these rapid postural adjustments are highly stereotyped and shown to be separate from a voluntary response (Nashner and Cordo 1981). By contrast, perturbations during a precision finger grip task (Johansson and Westling 1988) and lip perturbations during speech (Abbs et al. 1984) revealed flexible fast-feedback responses which are coordinated across multiple effectors to achieve a given goal. Interestingly, the M2 response in the upper limb has also been demonstrated to possess similar sophistication, beyond what can be predicted from the superimposition of a triggered reaction (Pruszynski et al. 2011; Weiler et al. 2016). While a simple, open-loop triggered reaction cannot account for all sophistication of rapid motor responses to a perturbation stimulus, in certain paradigms, i.e., simple RT task following upper limb perturbations used to elicit stretch responses, there is evidence that superimposition of a voluntary response does contribute to goal-dependent modulation of the M2 response (Rothwell et al. 1980; Lewis et al. 2006; Manning et al. 2012; Ravichandran et al. 2013; Forgaard et al. 2015). Therefore, to distinguish triggered reactions in the upper limb from other uses of the “triggered response” terminology (e.g., during speech production, whole body postural control, finger manipulation, and the StartReact effect), we appreciate the description provided by Manning et al. (2012, p. 162) “in the context of the long-latency stretch reflex, triggered reactions are RT responses”.

Importance of correctly classifying the startle response

The SAS-Catch condition in Experiment 2 allowed for an examination of the StartReact effect which further demonstrated the importance of correctly identifying the startle response in a StartReact experiment. On SAS-Catch Startle+ trials, the preprogrammed response was elicited on 97.8% of trials, and occurred consistently at a StartReact-like latency (see Fig. 4B). By contrast, on SAS-Catch Startle− trials, the preprogrammed response was only observed on 37.0% of trials and two distinct distributions of RTs emerged. One distribution, which consisted of 36.8% of these trials, was early and at StartReact-like latency suggesting that the preprogrammed response may have been triggered by the SAS. The second distribution (consisting of 63.2% of the trials) was later and more variable (~ 100–200 ms). On these late RT trials, the preprogrammed response was likely voluntarily initiated to the loud sound, similar to participants in a previous study voluntarily responding to the imperative signal (Maslovat et al. 2015). These findings clearly show the importance of separating Startle+ and Startle− trials when investigating the StartReact effect. Researchers can be very confident that when a voluntary response is preprogrammed and a startle response is elicited, the RT will be very early. However, on SAS trials where a startle response is not detected, the response can appear early on a small proportion of trials (e.g., the 13.6% of SAS-Catch Startle− trials), but on the majority of trials it will either not be initiated (63.0% of trials) or will occur late (> 100 ms; 23.4% of trials). This variety of responses to a SAS when no reflexive startle response is observed makes it difficult to draw conclusions about response preparation (as in the current study, we expected a similar preparation level for the CatchSAS trials but observed inconsistent responses on Startle− trials—whereas the responses were very consistent on Startle+ trials). Therefore, it is only when a startle response has been detected that researchers can be confident that sufficient activation of startle circuitry has occurred and can confidently make conclusions about the level or preparation of the intended response and the presence of the StartReact effect.

Conclusion

A mechanical perturbation imperative stimulus can elicit a preprogrammed voluntary response in shortened musculature at a latency of less than 100 ms. The present study demonstrates that these perturbation “triggered reactions” can occur consistently in the absence of an overt startle response. Through the use of a prepulse inhibition stimulus, we show that activity in SCM, a muscle commonly used to indicate startle, was a result of a postural response rather than the startle response. While our study provides evidence that a mechanical perturbation may act as a sub-threshold startling stimulus and therefore the perturbation could still engage circuitry involved in the StartReact effect, important differences also exist between a perturbation-triggered reaction and the StartReact effect. As an alternative explanation, the perturbation imperative signal may activate supra-spinal areas involved in generation of the voluntary response resulting in consistently early RTs, even in the absence of startle.