Passive limb displacement is routinely used to assess muscle tone. If we attempt to quantify muscle stiffness using mechanical devices, it is important to know whether kinematic stimuli are able to trigger startle reactions. Whether kinematic stimuli are able to elicit a startle reflex and to accelerate prepared voluntary movements (StartReact effect) has not been studied extensively to date. Eleven healthy subjects were suspended in an exoskeleton and were exposed to passive left knee flexion (KF) at three intensities, occasionally replaced by fast right KF. Upon perceiving the movement subjects were asked to perform right wrist extension (WE), assessed by extensor carpi radialis (ECR) electromyographic activity. ECR latencies were shortest in fast trials. Startle responses were present in most fast trials, yet being significantly accelerated and larger with right versus left KF, since the former occurred less frequently and thus less expectedly. Startle responses were associated with earlier and larger ECR responses (StartReact effect), with the largest effect again upon right KF. The results provide evidence that kinematic stimuli are able to elicit both startle reflexes and a StartReact effect, which depend on stimulus intensity and anticipation, as well as on the subjects’ preparedness to respond.
Displacement of larger body parts, for example, a limb segment, or the whole body, may elicit generalized reactions, for example, postural adjustments, that frequently include muscle responses from upper limbs and neck [
Startle reflexes habituate less when the subject is prepared to perform a voluntary movement [
Notably, an accelerated release of a motor response may also be directly related to the intensity of the IS [
Startle reflexes and StartReact effect are of clinical interest in both diagnostic and therapeutic facets of neurorehabilitation. (1) Startle reflexes could potentially “contaminate” the measurement and quantification of muscle stiffness with robotic exoskeletons in patients affected by spasticity. (2) Startling stimuli may serve to trigger prepared actions, for example, reaching and grasping, in patients suffering from stroke, whose corticospinal tract is damaged and thus incapable of conveying the necessary neural impulses for mediating the requested movements [
With these premises in mind, we were interested in further exploring whether incremental kinematic stimuli in the lower limbs are able to shorten RTs in the upper limbs and whether they are able to elicit a startle reflex and a StartReact effect. Specifically, we wanted to separate voluntary responses from stretch reflexes, and therefore we used different limbs for stimulation and recording. In order to document the presence of a startle reflex, we recorded from OOc and SCM. Responses in either muscle should help differentiating whether an accelerated voluntary response is due to the strength of the stimulus or due to a superimposed StartReact effect. Recording of OOc and SCM activity when performing the requested upper limb movement in response to expected kinematic stimuli in the lower limbs aimed to confirm that the applied stimulus-response protocol excludes OOc or SCM activity as part of the motor task, or as a part of a postural reaction (i.e., head stabilization), which, if present, could be erroneously mistaken as expression of a startle reflex. Prepulses, which are known to suppress startle responses in OOc and SCM while preserving the StartReact effect, were included in some trials with unexpected high intensity stimuli, in order to compare these responses with those without prepulse, and to confirm the startling value of the delivered stimulus.
Eleven healthy subjects (9 females, 2 males, age 29–51 years) took part in the experiments. All were self-reported right-handers with normal or corrected-to-normal vision and were free from any neurological deficit that could affect the execution of the task. Subjects gave their written informed consent for the experiment in accordance with the Declaration of Helsinki, which was approved by the local Institutional Review Board.
Subjects were placed in a driven electromechanical gait robot, Lokomat (Hocoma, Switzerland), with each lower limb strapped to an exoskeleton, adjusted to individual anthropometric measures. The Lokomat produced passive knee joint movements at various angular velocities. Subjects were suspended in the air during trial periods by a harness around the torso attached to an over-head Body Weight Support System (with deflection pulleys). Handrails on either side of the subject at waist level allowed supporting with the hands if necessary. A band strapped between both bars in front of the subject enabled resting the forearms. A possible startle reaction was monitored by recording surface electromyographic (EMG) activity from the right OOc and SCM. Surface EMG related to wrist extension (WE) was recorded from right extensor carpi radialis muscle (ECR). Single sweeps of 4 s were recorded, including a 900 ms prestimulus delay using routine electrodiagnostic equipment (Viking IV, Nicolet Biomedical, Madison, Wisconsin). Filter settings were 10–10000 Hz. Electrical constant voltage stimuli of 0.1 ms duration and 1.5 times perception threshold were generated with a Digitimer D180A. These stimuli served as prepulse and were only occasionally delivered in few trials through ring electrodes placed on the left index finger, 100 ms preceding KF. Both Lokomat and electrodiagnostic system were synchronized, with the sweep being triggered by the Lokomat as soon as a KF of 80° was initiated from a starting point with completely extended knees, irrespective of induced angular velocity. Subsequent knee extension back to the starting point was always performed at a low angular peak velocity (less than 10°/s).
Subjects were informed that they were going to be suspended in the Lokomat and that there would be a series of trials and that unless otherwise advised they should perform a fast WE (simple RT paradigm) as soon as they perceived the IS, that is, perception of passive KF induced by the robot. They were told that if they felt uncomfortable due to the maintained position they should tell the experimenters immediately, so that they would be lowered to standing on the floor, or, if desired, would be detached from the system. The study included also established stops and lowering of the subjects from the electromechanical device at different times. The study contained a predetermined workflow of trials. A trial included a verbal warning signal for the subject to be prepared “ready!”, the IS delivered at a variable latency of 1 to 3 s following the warning signal, and recording of the subject’s responses. The experimental session was composed of three consecutive blocks of trials. Within each block, trials were separated by a minimum of 45 seconds, time required by the Lokomat system to again reach the starting position, to provide sufficient time for subjects to achieve a comparable resting condition, and to avoid influence of one trial upon the subsequent one.
In a pilot test, one subject was randomly exposed to different passive left KFs reaching peak velocities between 6°/s and 240°/s, which were all clearly perceived. The fastest velocity was previously used for measurements of muscle stiffness in patients with upper motoneuron lesions [
In block 1, at the beginning of each experimental session, low intensity tone bursts (60 dB nHL, 500 Hz, 10 ms duration) were used as IS. Subjects were instructed to perform only a fast WE upon hearing the tone (condition “WE-only”). Five WE-only trials were repeated in order to accustom the subject to the suspension in the Lokomat system and to depict the subject’s movement pattern employed when briskly raising their hand, specifically to check whether there was any EMG activity in SCM associated with the task, for example, anticipatory postural adjustments, which could interfere with analysis of responses in the subsequent blocks of trials.
In block 2, subjects were instructed to respond with a fast WE upon perceiving the IS, from that point on passive KF. Different angular velocities of left KF interspersed with occasional fast right KF were applied in pseudorandom order, without informing subjects beforehand about the type of upcoming stimuli. Accordingly, based on velocity and presence/absence of startle signs, resulting recordings of each trial were post hoc grouped and categorized for analysis into different “conditions.” The type of trials and resulting conditions were as follows: Trials at 240°/s passive left KF, resulting in conditions “240React” and “240StartReact” Trials at 60°/s passive left KF, resulting in conditions “60React” and “60StartReact” Trials at 6°/s passive left KF, resulting in condition “6React”; none of these trials showed any startle signs; hence there was no condition “6StartReact” Trials at 240°/s passive left KF, in which subjects received the prepulse in addition to passive KF, resulting in conditions “240PrepReact” and “240PrepStartReact” Trials at 240°/s passive KF, in which the contralateral (right) leg was moved as IS, resulting in conditions “240ContraReact” and “240ContraStartReact”; in order to render these stimuli less expected, these trials were only introduced during the last fourth of the experimental block; as subjects were previously continuously exposed to left leg movements, they were supposedly less prepared to anticipate their right leg to be moved
Each of the 240°/s trials was interspersed with at least 5 trials at 6°/s and occasional trials at 60°/s, in order to avoid rapid habituation. For each condition, OOc and SCM activity was visually checked online in order to obtain a sufficient number of recordings containing startle signs, that is, at least three 240StartReact, two 240React, and two 240ContraStartReact recordings per subject. Thus, at least 40 trials at lower angular velocities were delivered per subject. If a subject generated too few overt startle responses in any given condition, additional trials were added at 240°/s, each interspersed with at least 5 trials at lower angular velocities. Thus, the total number of trials varied across subjects as it was not known beforehand how many trials at 240°/s eventually had to be repeated in order to achieve the required number of startle responses per condition.
At the end of this block, subjects were exposed to a subset of left KF trials at 240°/s, in which they were asked to remain relaxed and specifically not to perform WE. These trials were interspersed with some trials at 6°/s and 60°/s in order to keep subjects alert but uncertain about the velocity of the upcoming IS. This subset of trials at 240°/s was analyzed specifically to rule out EMG activity in SCM as part of postural adjustment to the leg displacement and served as controls for other trials at the same velocity. The resulting conditions were “240Control” and “240StartControl.”
Finally, block 3 was comprised of five trials in which subjects were explicitly informed about both side and velocity of the forthcoming leg movement (left leg, 240°/s) and in which they were asked to respond with fast WE. Resulting recordings were categorized accordingly into conditions “240Known” and “240StartKnown.”
The resulting number of trials per condition and per subject, which were analyzed, is depicted in Table
Experimental protocol applied to each subject. Except for the first block of trials (WE-only), the sequence of all other trials was pseudorandomized. After performing each single trial, resulting recordings were categorized according to the presence or absence of startle reflex signs into distinct conditions. The column
Velocity (°/s) | Conditions | Number of trials | Prepulse | Imperative signal | Wrist extension |
---|---|---|---|---|---|
— | WE-only | 5 trials | − | Sound | + |
6 | 6React | ≥30 trials | − | KF | + |
6StartReact | |||||
60 | 60React | ≥10 trials | − | KF | + |
60StartReact | |||||
240 | 240React | ≥5 trials | − | KF | + |
240StartReact | |||||
240 | 240PrepReact | ≥2 trials | + | KF | + |
240PrepStartReact | |||||
240 | 240Control | ≥2 trials | − | KF | − |
240StartControl | |||||
240 | 240ContraReact | ≥4 trials | − | Contralateral KF | + |
240ContraStartReact | |||||
240 | 240Known | 5 trials | − | KF | + |
240StartKnown |
EMG activity was full-wave rectified before analysis. For each subject and condition, EMG characteristics were determined separately for the selected muscles, and median values were used for statistical inferences. SPSS 22.0 was used for statistical analysis. Unless otherwise specified, data are shown as median and 95% confidence intervals (in brackets). OOc and SCM activity was accepted to represent a startle reflex, if it appeared at an appropriate latency, lasted more than 50 ms, and exceeded two standard deviations of baseline activity established during a 200 ms time window preceding IS (i.e., movement onset induced by the Lokomat).
Startle reflex onset latency was determined by visual inspection of the respective traces from 40 to 100 ms for OOc and from 50 to 120 ms for SCM following IS. These time windows were used in accordance with previous reports [
For WE, EMG onset latency in ECR was determined by visual inspection in a time window with an upper limit at 500 ms following IS for trials at 60°/s and 240°/s and 2000 ms for trials at 6°/s, as estimated from the pilot subject. Responses were accepted when EMG activity lasted more than 100 ms and exceeded two standard deviations of a 200 ms prestimulus baseline. ECR response magnitude was calculated as iEMG during a 100 ms window following response onset, as respective muscle bursts were usually longer-lasting than those obtained in OOc and SCM. For statistical inference purposes, iEMG values were normalized for each subject relative to individual maximum voluntary activity. Median values and 95% confidence intervals of EMG onset latencies and iEMG were calculated for all subjects and conditions.
Percentages of trials with startle signs were calculated separately for each subject and type of trial and were then compared among trial types using Friedman
Friedman
All subjects performed the study without difficulty. Two subjects asked for intermittent release from suspension in order to stand on their feet for a short period of time during the experiment. In order to acquire a sufficient number of trials per condition at 240°/s containing startle signs, trials of different velocities occasionally had to be repeated according to protocol, thereby amounting to 65 to 85 trials per subject.
In response to passive KF at different angular velocities, used as IS, all subjects performed the required WE. Stimulus intensity had a significant effect on ECR latencies, when comparing conditions 6React, 60React, and 240React (Friedman
In WE-only condition, there was no evidence of SCM activity before WE, documenting no activity associated with head displacements due to brisk wrist movement and/or anticipatory postural adjustments. Condition 240Control (in which subjects were asked not to perform WE) also revealed no associated SCM activity. Thereby we excluded the possibility that SCM activity may appear as part of a postural reaction to passive KF.
There were no startle signs in OOc or SCM at all in trials at 6°/s and only few in trials at 60°/s (0-1/subject). In contrast, at 240°/s startle indicators were present in 45% of trials. Figure
Representative trials from one subject depicting responses from extensor carpi radialis (ECR), orbicularis oculi (OOc), and sternocleidomastoid (SCM) muscles in conditions 240React, 240StartReact, and 240PrepStartReact. The traces show a leftward displacement (anticipation) and larger integrated EMG activity in ECR in the 240StartReact and 240PrepStartReact trials as compared to the 240React trial. Startle-related responses in OOc and SCM (marked with asterisks:
Amount of trials per trial type at 240°/s containing startle signs. For each of the eleven subjects the percentage of trials per trial type at 240°/s containing startle signs in orbicularis oculi or sternocleidomastoid muscles was calculated relative to the total number of respective trials (e.g., number of 240StartReact trials divided by [number of 240StartReact trials plus number of 240React trials] times 100). For descriptive purposes means + standard deviation of these percentages are shown. Note the highest “yield” of startle responses in conditions 240ContraStartReact and 240StartReact, with lower mean percentages in conditions 240PrepStartReact, 240StartControl, and 240StartKnown.
As startle signs were absent in trials at 6°/s and scarce in 60°/s and in some 240°/s conditions (i.e., 240PrepStartReact, 240StartControl, and 240StartKnown), subsequent statistical comparisons were only performed between conditions 240StartReact and 240ContraStartReact. The distribution of startle signs in OOc and SCM was similar in both conditions (for OOc:
Presence of startle signs according to stimulus location, obtained in eleven subjects and calculated as described in Figure
In OOc, startle latency was significantly shorter in condition 240ContraStartReact [72 (50, 74) ms] than in 240StartReact [79 (77, 112) ms] (
Latency and area-under-the-curve of startle reflexes according to stimulus location. The plots represent onset latencies and iEMG in orbicularis oculi (OOc) and sternocleidomastoid (SMC) muscles from all subjects obtained in conditions 240StartReact and 240ContraStartReact. Each box represents 50% of all values intersected by the median; the whiskers indicate the smallest and largest values. When the imperative signal was applied in the less expected (right) leg (condition 240ContraStartReact), orbicularis oculi (OOc) latencies were significantly shorter than when imperative signal was applied in the most frequently moved (left) leg (condition 240StartReact). In condition 240ContraStartReact both OOc and SCM showed significantly larger responses. Asterisks above the boxes define the level of significance of between-group comparisons (
In 240°/s trials with WE, latencies of ECR responses were significantly shorter in conditions with startle signs (240StartReact, 240PrepStartReact, 240ContraStartReact, and 240StartKnown, all combined) [210 (192, 254) ms] than without startle signs (240React, 240PrepReact, 240ContraReact, and 240Known, all combined) [233 (207, 249) ms] (
StartReact effects according to condition at 240°/s. Trials containing startle signs (conditions depicted in black) showed shorter latencies and larger responses in extensor carpi radialis (ECR) electromyographic recordings than those without startle signs (corresponding conditions depicted in grey), except for those with prepulse stimulation (data are medians and confidence intervals). Due to study protocol requirements, however, only few trials included a prepulse and, as expected, almost none contained startle signs (condition 240PrepStartReact), while most of them did not contain startle signs (condition 240PrepReact). Both showed ECR latencies and amplitudes similar to other conditions with startle signs, indicating that trials with a prepulse showed a StartReact effect, irrespective of the presence or absence of overt startle signs in orbicularis oculi or sternocleidomastoid muscles.
In order to document a StartReact effect, that is, acceleration of a prepared movement (here: WE) by a startling stimulus and to show the influence of preparedness (here: left KF, “expected side” versus right KF “unexpected side”) we compared latencies and iEMG in ECR responses applying Friedman
The effect of a prepulse on WE was assessed in order to confirm the presence of a StartReact effect. ECR latency was 207 (171, 250) ms in 240PrepReact, which was close to values obtained in conditions containing startle signs. There was a significant difference in ECR latency among 240React, 240StartReact, and 240PrepReact (Friedman
In the present study different intensities of a kinematic stimulus modified RT responses. The highest intensity resulted in shortest RT and largest response magnitude. Kinematic stimuli exceeding a certain intensity level, and particularly when unexpected, may themselves evoke startle reflexes. They also generate a StartReact effect, as explored by a voluntary response task in a muscle distant from stimulus location. Additionally, the magnitude of the StartReact effect is inversely related to the degree of preparedness of a subject to receive the stimulus.
Shorter RTs and larger response sizes associated with stronger stimuli concur with previous reports in other domains [
The movement applied with the Lokomat was able to elicit a startle reflex on some occasions. This was mainly the case in trials with the highest intensity, although not in all conditions. As with other kinds of stimuli, a certain level of intensity and unexpectedness were prerequisite to elicit a startle reflex [
We designed a protocol that included few trials at the highest intensity among a majority of trials at lower intensities in one (left) leg. Thereby we were able to obtain startle reflexes in OOc and SCM, possibly related to some degree of a subject’s surprise (240StartReact). Most startle responses were present in SCM, possibly due to an additional influence of posture (keeping the head upright in a suspended situation) [
Subjects were not explicitly informed that trials would occasionally be interspersed in the other (right) leg during the last fourth of the session. Thereby we were able to obtain largest startle reflexes with shortest latencies in OOc and SCM (240ContraStartReact), possibly related to the highest degree of surprise as the stimulus appeared rather unexpectedly in the leg contralateral to the one which was displaced in the majority of trials throughout the experiment. Consequently, kinematically induced startle reflexes may show EMG responses of different magnitude depending on the level of attention, as previously described in other domains [
The present study design is set out to overcome some limitations of previous studies, which have explored kinematic stimuli without specifically recording from startle reflex indicator muscles [
To rule out that the obtained SCM activity was due to reactive head stabilization as part of postural adjustment related to brisk WE movements [
In order to further confirm that the observed responses in OOc and SCM were indeed startle reflexes, we included some trials containing prepulse stimuli. Prepulses are known to inhibit startle reflexes but at the same time to maintain the StartReact effect [
Considering all factors described above, we conclude that the kinematic stimuli applied at the highest intensity were indeed able to elicit startle reflexes.
All subjects performed WE in all trials when required, indicating that also stimuli of lowest intensity were able to trigger these responses. Latency variability was highest for the lowest intensity, consistent with the subjects’ difficulty in discriminating these stimuli [
Exploring startle responses, Ravichandran et al. [
Trials at 240°/s without startle signs had shorter latencies for WE than trials at 6°/s or 60°/s but had longer latencies than those trials at 240°/s containing startle signs. These results suggest, in addition to a progressive influence of stimulus intensity on voluntary movement responses, an additional superimposed startle effect, both being independent of each other [
Furthermore, when subjects are informed about the upcoming stimulus, the probability of ensuing startles is reduced. Accordingly, in situations when subjects knew in advance both side and velocity of leg movement (240Known), no startles occurred in spite of the subjects’ high level of preparedness to react with WE, as evidenced by response latencies which were equally fast in conditions 240Known and 240React. In conditions 240React and 240StartReact, we assumed that subjects expected that the leg which was usually displaced was also to be moved next. However, subjects were uncertain of the velocity, which was usually low (6°/s and 60°/s), and indeed we found only few trials with StartReact effects at low velocities. Most pronounced StartReact effects with shortest latencies and largest EMG bursts in ECR were observed when the movement occurred in the contralateral leg, that is, least expected. This phenomenon of response latency shortening related to a subject’s expectancy of the stimulus has previously been described [
Time was the most critical limiting factor in the present study, and a compromise had to be achieved between number and type of trials to be executed according to study protocol and the subjects’ ability to cope with being suspended in the Lokomat, continuous attention, and increasing fatigue. The main goal was to unequivocally document the presence of startle signs and a StartReact effect in response to kinematic stimuli, while excluding vestibular influences, separating reflexive from voluntary reactions, and differentiating startle signs from muscle activity inherent to postural or voluntary motor patterns. As shortening in RT may be due to a pure intensity effect and as startle signs were present only in few trials, additional trials had to be appended on-line in order to achieve the minimum required number of trials with startle signs. However, each of the 240°/s trials had to be interspersed with at least 5 trials at 6°/s and occasional trials at 60°/s, in order to avoid rapid habituation and to maintain unexpectedness of high intensity stimuli. Adding trials with startling acoustic stimuli as IS for further comparison with previous studies was thus not feasible. Furthermore, additional trials with prepulses were not possible for the same reason, but those included served as additional means to corroborate the StartReact effect even in the absence of an overt startle reflex.
In summary, our results provide evidence that a kinematic stimulus is able to elicit a startle reflex and a StartReact effect. A vestibular route does not seem to be required nor intersensory facilitation. The responses seem to depend on preparedness and expectancy of the subject, as well as the intensity of the stimulus over time and its detection by the subject. These findings have important implications in neurorehabilitation. Interest in quantification of spastic muscle tone by means of robotic devices has recently increased. When measuring muscle stiffness in the Lokomat [
Extensor carpi radialis muscle
Electromyography
Integrated EMG area-under-the-curve
Imperative signal
Knee flexion
Long loop reflex
Orbicularis oculi muscle
Reaction time
Sternocleidomastoid muscle
Wrist extension.
All authors declare that they do not have any conflict of interests.
Juan M. Castellote and Markus Kofler designed the study; all authors collected research data; Juan M. Castellote and Markus Kofler analyzed and interpreted data; all authors drafted the work; Juan M. Castellote and Markus Kofler wrote the manuscript; all authors revised and approved the final version of the manuscript.
The authors would like to thank all subjects who volunteered their time to participate in this study, as well as the staff of Hochzirl Hospital, especially Mrs. Maria Hoch and the colleagues of the Robotics Department, for their assistance during the study. Furthermore, they are grateful to Mrs. Ellen Quirbach for her help with editing the manuscript. This study was supported by Instituto de Salud Carlos III Grant ESPY 1281-15 and Government of Spain Grant PRX12/00349 to Juan M. Castellote.