Exp Brain Res

DOI 10.1007/s00221-006-0610-8

RESEARCH ARTICLE

Startle produces early response latencies that are distinctfrom stimulus intensity eVects

Anthony N. Carlsen · Chris J. Dakin · Romeo Chua · Ian M. Franks

Received: 11 May 2006 / Accepted: 14 June 2006© Springer-Verlag 2006

Abstract Recent experiments pairing a startling stim-ulus with a simple reaction time (RT) task have shownthat when participants are startled, a prepared move-ment was initiated earlier in comparison to voluntaryinitiation. It has been argued that the startle acts totrigger the response involuntarily. However, an alter-native explanation is that the decrease in RT may bedue to stimulus intensity eVects, not involuntary trig-gering. Thus the aim of the current investigation was todetermine if RT simply declined in a linear fashionwith increasing stimulus intensity, or if there was apoint at which RT dramatically decreased. In the pres-ent experiment participants completed 50 active wristextension trials to a target in response to an auditorystimulus of varying stimulus intensity (83–123 dB). Thepresented data show that RTs associated with a startleresponse are separate from stimulus intensity facili-tated responses. Furthermore, this startle facilitation ismore highly associated with sternocleidomastoid elec-tromyographic (EMG) activity, rather than the EMGfrom the widely used startle response indicator muscleorbicularis oculi.

Keywords Startle · Electromyography · Reaction time · Orbicularis oculi · Sternocleidomastoid

Introduction

The startle response is characterized by a stereotypicalpattern of muscle activity and can be elicited by a loudacoustic stimulus (Davis 1984). Recently, a startlingstimulus has been used as the “go” signal in reactiontime experiments. An interesting eVect of this methodhas been that when participants were startled, theintended movement was initiated signiWcantly earlierin comparison to voluntary initiation (Carlsen et al.2004a; Valls-Solé et al. 1999). Because of the extremelyshort response latencies reported, it was argued thatthe startle acted as an early trigger for the preparedresponse. Two lines of evidence support the notion of astartle-elicited response. First, the response-relatedelectromyographic (EMG) activation pattern (e.g.Wadman et al. 1979) was similar in both burst durationand timing whether or not participants performed thetask in the presence of the startling stimulus (Carlsenet al. 2004a; Valls-Solé et al. 1999). Second, taskaccuracy was maintained during the startle-elicitedresponse (Carlsen et al. 2004a). These observationsindicated that the intended prepared response hadbeen triggered and that it was not simply a later volun-tary response superimposed upon an early startle reXex(Seigmund et al. 2001). For this to occur, it was sug-gested that the prepared response might be held inreadiness in subcortical structures that mediate thestartle response (Carlsen et al. 2004a; Valls-Solé et al.1999).

However, there have been varied reports on theloudness required for a stimulus to elicit a startleresponse, and there exists a point of contention overwhat actually constitutes a “startle response.” Themost widely used response indicator of startle is EMG

A. N. Carlsen · C. J. Dakin · R. Chua · I. M. FranksSchool of Human Kinetics, University of British Columbia, Vancouver, Canada

A. N. Carlsen (&)210-6081 University Boulevard, Vancouver, BC, Canada V6T 1Z1e-mail: carlsen@interchange.ubc.ca

123

Exp Brain Res

activity in the orbicularis occuli (OOc) muscle (blinkresponse, Blumenthal et al. 2005; Davis 1984). How-ever, the physiologically separate (non-startle) audi-tory blink response can still be elicited following startlehabituation, making OOc an ambiguous indicator ofstartle (Brown et al. 1991). Although a startle blinkresponse has been reported to occur due to acousticstimuli as low as 85 dB, many studies have utilizedstimuli between 90 and 130 dB (Blumenthal 1996;Brown et al. 1991). Yet, most of these studies have notattempted to distinguish between startle and non-star-tle blinks. Thus it was suggested that activity in thesternocleidomastoid (SCM) muscle may be a betterindicator of startle as it is among the last indicators tobecome habituated after repeated exposure to the star-tling stimulus, and it is less equivocal in nature (Brownet al. 1991; Carlsen et al. 2004b).

The purpose of the current investigation was todetermine the eVect of stimulus intensity on: (1) thepresence or absence of startle response indicators, and(2) voluntary reaction time (RT). Since increases instimulus intensity have also been associated withdecreased RTs for over a century (see Woodworth1938, p. 318), it has been suggested that the observedstartle-associated decrease in RT may be due to stimu-lus intensity eVects, not subcortical triggering (Carlsenet al. 2004a). We designed this experiment to deter-mine if RT simply declined in a linear fashion withincreasing stimulus intensity, or if there was a point atwhich a more dramatic decrease in RT was associatedwith a startle response. In other words, is the RT short-ening observed in the presence of a startle more thanan intensity eVect alone?

Methods

Participants

Ten participants (seven females, three males; mean age26 § 8 years) with no obvious upper body abnormali-ties, or sensory or motor dysfunctions volunteered toparticipate in the study. All participants reported nor-mal hearing. All participants gave written informedconsent, and the study was conducted in accordancewith the ethical guidelines set by the University of Brit-ish Columbia.

Apparatus and task

Participants sat facing a computer monitor with theirright arm resting on a custom made manipulandumthat allowed measurement of wrist angular displace-

ment. The details of the experimental set-up and appa-ratus have been previously documented (see Carlsenet al. 2003). The task was to perform a 20° right wristextension movement to a Wxed target as quickly and asaccurately as possible following an auditory stimulus(1,000 Hz, 40 ms duration, < 1 ms rise time). Partici-pants were oVered a monetary bonus for fast reactionsconsisting of 1 cent per ms faster than either their pre-vious best RT or 120 ms.

Instrumentation and stimuli

The intensity of the auditory “go” signal (imperativestimulus) ranged from 83 to 123 dB in 10 dB incre-ments (resulting in 5 stimulus intensity levels) and waspresented via a loudspeaker located directly behind theparticipant. Each participant performed 30 trials at83 dB (control) and 5 trials at each of the other fourstimulus intensities which were randomized betweentrials. Four catch trials (no imperative stimulus) werealso included to discourage anticipation. Electromyo-graphic (EMG) data were collected from the musclebelly of the right Xexor carpi radialis (FCR) and exten-sor carpi radialis longus (ECR), as well as the indepen-dent startle response indicators: left OOc, and SCM.Wrist displacement and EMG data were collected at1,000 Hz and stored for oVline analysis.

Target and feedback

The target was a Wxed point in space located at 20° ofangular displacement into extension with respect to theright wrist’s starting position. A computer screenplaced directly in front of the participant provided realtime position feedback during trials. After each trial,feedback information including displacement error atthe end of the initial impulse (deg), and reaction time(ms) were displayed on the same computer monitordisplay.

Training

Participants were allowed to practice the task prior totesting to familiarize themselves with the task andequipment. The participants were instructed that theywould Wrst hear a warning tone, followed by a Wxedforeperiod (2,500 ms, duration unknown to the partici-pants), and Wnally a “go” tone (imperative stimulus).Instructions emphasised fast reaction times and fastmovement times, as well as minimizing target error.Participants were also instructed that the loudness ofthe stimulus would be variable. Participants receivedblocks of ten practice trials, and were deemed to have

123

Exp Brain Res

reached an adequate level of task competence to startthe testing trials when they could successfully hit thetarget (§ 5°) four out of the last Wve practice trials in ablock. No participants performed more than two prac-tice blocks.

Data reduction and analysis

Movement onset was deWned as the Wrst point of achange of more than 0.2° of angular displacement fromthe starting position following the stimulus. Peak dis-placement was determined by identifying the point atwhich velocity returned to zero following movementonset. The Wnal position of the movement was deWnedas the Wrst point at which angular velocity remainedbelow 8°/s for at least 150 ms. Movement time was deW-ned as the time (in ms) between movement onset andWnal position.

Surface EMG burst onsets were deWned as the pointat which the EMG Wrst began a sustained rise abovebaseline levels. The location of this point was deter-mined by Wrst displaying the EMG pattern on a com-puter monitor with a superimposed line indicating thepoint at which activity increased to more than 2 stan-dard deviations above baseline (mean of 50 ms ofEMG activity preceding onset). Onset was then veri-Wed by visually locating and manually adjusting theonset mark to the point at which the activity Wrstincreased. This method allowed for correction of errorsdue to the strictness of the algorithm. Premotor RT(PMRT) was deWned as EMG onset in the ECR mus-cle. EMG oVsets were marked in a similar fashion, withthe activity between EMG onset and EMG oVset beingdeWned as a distinct burst. PMRTs greater than1,000 ms and less than 40 ms were discarded ifobserved in the lowest intensity (83 dB) condition.These comprised less than 1% of trials. In all otherconditions, no trials were discarded, however, none felloutside these ranges.

Peak EMG amplitudes in the startle response indi-cators were deWned as the largest EMG amplitude, rec-tiWed and Wltered with a 25 Hz lowpass elliptic Wlter,recorded within an interval of 100 ms following EMGburst onset. To normalize the EMG for comparisonbetween participants, EMG burst amplitudes wereexpressed as a percentage of the mean peak EMGamplitude for each respective muscle in the 124 dBcondition for each participant.

Trials were separated by stimulus dB intensity, aswell as by whether or not an EMG burst was present instartle response indicators (No response indicator,OOc only, or SCM + OOc). Since SCM did not occurwithout accompanying OOc activity, this grouping is

called SCM. Startle response indicators were classiWedas startle if onsets occurred in less than 120 ms follow-ing stimulus onset.

Integrated EMG values for OOc were computed bynumerically integrating the normalized and rectiWedEMG for each participant. First the EMG was recti-Wed, amplitude and baseline normalized, and onsetnormalized (aligned to a common onset). These valueswere numerically integrated (ms time base) over aperiod of 160 ms from the common onset (0 ms). Sub-sets were taken consisting of 0–60 ms (to capture theentirety of the Wrst burst observed in OOc only condi-tions) and 60–160 ms (to quantify any further activity).

Statistical analyses

Dependent measures were analyzed where appropriateusing 5(dB level) £ 3(indicator) repeated measuresanalysis of variance (ANOVA), to determine if diVer-ences existed between trials. DiVerences with a proba-bility of less than 0.05 were considered to be signiWcant.Tukey’s honestly signiWcant diVerence (HSD) post-hoctests as well as simple eVects tests were administered todetermine the locus of the diVerences.

Results

Voluntary response

All participants were accurate in performing therequired 20° movement. One-way ANOVAs showed nosigniWcant diVerences in either Wnal position, F(4,36) =1.53, P = 0.347, or movement time, F(4,36) = 0.861,P = 0.497, across all stimulus intensities.

Premotor RT (PMRT) was compared between stim-ulus dB intensities as well as between startle responseindicators using a two-factor ANOVA. Because forsome participants, no responses were observed in cer-tain conditions (e.g. OOc only at 113 dB), for statisticalanalysis, missing cells were Wlled with mean values forthat condition. Thus because true variability isdecreased increasing the possibility of Type 1 error,main eVects were only considered signiWcant ifP < 0.01. Results are presented in Fig. 1. SigniWcantmain eVects on PMRT were found for both indicator,F(2,18) = 54.82, P < 0.001, and dB level, F(4,36) = 4.81,P = 0.003, ad a signiWcant interaction was found,F(8,72) = 8.51, P < 0.001. Post-hoc tests revealed thatwhen SCM was present, PMRT was signiWcantlyshorter than if no response indicator (NRI) was pres-ent regardless of dB level (P < 0.05). Furthermore,when SCM was present, PMRT was signiWcantly

123

Exp Brain Res

shorter than if only OOc was present at both 93 and103 dB (P < 0.05). Finally, when only OOc was presentRT was signiWcantly shorter than if NRI was observedonly at the two highest dB levels (P < 0.05).

Blink response

We sought to asses whether diVerences existed in theproWle of the OOc EMG based on the presence orabsence of SCM activity. This was due to the strongerapparent association between short RTs and SCMactivity, as compared to OOc activity. An example ofthe ensemble average of rectiWed, baseline and ampli-tude normalized OOc EMG activity from a single par-ticipant is presented in Fig. 2a. When SCM waspresent, OOc activity included a larger, longer second-ary component that was absent when SCM was alsoabsent. This pattern was analyzed across participantsby comparing numerically integrated amplitude-nor-malized OOc EMG from onset for 160 ms (Fig. 2b).The integral values were collapsed across stimulusintensity since no main eVect was observed (P = 0.224).A one-way ANOVA analyzing Integrated EMG(IEMG) between “OOc only” and “SCM present”groups indicated that IEMG was signiWcantly largeracross participants, F(1,7) = 19.231, P = 0.003, whenSCM was present (29.64 § 17.70%*ms) compared towhen SCM was absent (7.68 § 7.66%*ms). Further-more, the proportion of the integrated value contrib-uted by the time period form 60–160 ms following

onset (representing the second component) was muchsmaller when SCM was absent (27.2 %) than whenSCM was present (53.2%) (Fig. 2b).

Probability of observing startle indicators

A signiWcant interaction was found using a two wayANOVA between stimulus intensity and the percentageof trials exhibiting startle indicators, F(8,72) = 19.857,P < 0.001 (Fig. 3). SpeciWcally, at 83 dB no activity wasobserved in either OOc or SCM (i.e. NRI). However,SCM activity (which was always accompanied by OOcactivity) was detected in 2% of trials at 93 dB andincreased with increasing dB level, whereas trials inwhich neither OOc nor SCM were detected decreasedwith increasing stimulus intensity. Post hoc analysis ofthe interaction showed that the only signiWcant diVer-ence in the percentage of trials in which only OOc activ-ity was detected was between 83 and 103 dB (P < 0.05),indicating that at 103 dB the probability of observingOOc without SCM was greatest.

Discussion

Previous investigations have shown that PMRT is con-siderably decreased when a startling acoustic stimulusis used as the “go” signal during a RT task (Carlsenet al. 2003, 2004a, b; Valls-Solé et al. 1999), whereasthe characteristics of the movement and the EMG pro-Wles are unaVected (Carlsen et al. 2004a). These resultshave been explained using a subcortical triggeringhypothesis. It has been suggested, however, that theRT decrease is simply an extreme case of the well doc-umented “stimulus-intensity eVect” (Kohfeld 1969;Luce 1986; Woodworth 1938). Results were unequivo-cal: when a SCM (startle) response was detected, RTwas signiWcantly shorter than when no startle responsewas elicited, regardless of the intensity of the impera-tive stimulus.

Reaction time results are displayed in Fig. 1. Whenno startle activity (neither OOc nor SCM) was present,premotor RT decreased steadily with stimulus inten-sity, and reached asymptote at 113 dB. This WndingconWrms a “stimulus intensity eVect” observed by earlyresearchers (see Woodworth 1938) and is thought tooccur because sensory and perceptual processing aresubstantially faster for more intense physical stimuli(Levick 1973). However, when startle SCM activitywas observed, RT was signiWcantly shortened for all dBlevels above control indicating that the startle facili-tated responses are diVerent than stimulus intensityfacilitated responses.

Fig. 1 Mean (§ SD) premotor reaction time (PMRT) from stim-ulus to EMG onset as a function of stimulus intensity in decibels(dB) and the presence of startle response indicators. Black circlesare means of PMRTs for trials in which neither orbicularis oculi(OOc) nor sternocleidomastoid (SCM) were observed (no re-sponse indicator, nri). Blue triangles are PMRTs for trials inwhich OOc was observed but SCM was not observed. Red squaresare trials in which SCM was observed. *SigniWcant diVerencefrom nri. **SigniWcant diVerence from nri and OOc only

123

Exp Brain Res

Although increases in stimulus intensity (in theabsence of a startle response) resulted in reduction inRT (Fig. 1, black circles), there is no evidence to sug-gest that the information processing pathways for thestimulus were changed. However, due to the extremelyshort RT latencies observed under startle conditions

(Fig. 1, red squares), it has been suggested that thestartle may act upon a diVerent pathway to triggerthe prepared response directly (Carlsen et al. 2004a;Valls-Solé et al. 1999). SpeciWcally, it was argued that ifthe required response could be speciWed in advance, itmay be possible to store the response subcortically forlater voluntary triggering. This hypothesis was testedpreviously using a simple/choice RT experiment, withresults showing that if a response could be prepared inadvance (Simple RT) the response was triggered earlyby startle. However, no RT diVerences were seen dueto startle in a Choice RT condition, when the responsewas speciWed following the stimulus (Carlsen et al.2004b). Thus it appears that a response was triggeredby startle only if it was known (and presumably pre-pared) in advance.

If a response could be speciWed in advance it may bepossible to alleviate cortical demand by oVloadingsuYcient detail of a motor program to a holding area.Thus the only information processing requirementwould be detection of the imperative stimulus andsubsequent triggering of the response. One possiblecandidate for such motor program storage is the mid-brain reticular formation. This is because it plays animportant role in both the involuntary startle response(Yeomans and Frankland 1996) as well as mediating

Fig. 2 Individual (a) and group (b) orbicularis oculi (OOc) EMG data expressed by whether sternocleidomas-toid (SCM) activity was either detected (red) or not (blue). a Examples are baseline and onset normalized ensemble average of rectiWed OOc activity from a single partici-pant expressed as a percent-age of the participant’s mean peak EMG. b Mean inte-grated EMG (IEMG) values from OOc by time over which the integration was per-formed. Group IEMG values (§ SD) for the time period of 0–160 ms, 0–60 ms, and 60–160 ms following OOc onset

a

b

Fig. 3 Percentage (§ SD) of trials in which the startle indicatorswere observed. Black bars represent percentage of trials in whichneither sternocleidomastoid (SCM) nor orbicularis oculi (OOc)activity was observed (nri). Blue bars represent the percentage oftrials in which OOc activity was observed but not SCM. Red barsrepresent the percentage of trials in which SCM activity wasobserved

123

Exp Brain Res

many types of motor output (Rothwell et al. 2002). Ifthis were the case, it may be possible to trigger themotor program in the absence of the normal corticaltrigger if some event caused adequate activation ofthese subcortical storage structures. We suggest thatwhen SCM activity was observed, there was adequateactivation to trigger a response in the centers thatmediate the acoustic startle response, but additionally,it appears to also be adequate to trigger a stored motorprogram. The data we have provided merely associatesthe presence of SCM and decreased RTs, and althoughseparate pathways for the triggering of the action andtriggering the startle response cannot be ruled out, itappears likely that they are related.

It should be mentioned that the level of motor prep-aration may play a role in the triggering of a responseby startle. For example, lower levels of preparationappear to result in an inability of the stimulus to triggerthe response directly, as well as startle response habitu-ation (see Carlsen et al. 2003). Thus a high level ofpreparation is a requirement of this paradigm. How-ever, we feel that the level of preparation was ade-quately controlled by having all participants maximallyprepared. This was accomplished by having partici-pants in a stable environment, with speciWc instructionsto prepare the movement in advance, and by oVering amonetary bonus.

Detecting a startle response is thus of paramountimportance when investigating the eVect of stimulus inten-sity on RT. Many previous studies have employed theblink reXex (evidenced by EMG activity in OOc) as anindication that a startle response occurred (Blumenthalet al. 2005; Davis 1984). However, some evidencesuggests that OOc may not be a clear indicator of star-tle. For example, Brown et al. (1991) noted that follow-ing startle habituation, an early blink response wasnonetheless elicited by a loud stimulus. It was arguedthat the early OOc EMG component was not part ofthe generalized startle reXex, and reXected a physiolog-ically separate (non-startle) auditory blink reXex(Brown et al. 1991). Our data show a similar single/two-component OOc EMG pattern dissociation(Fig. 2a) when trials were separated based on whetheror not EMG activity was also detected in SCM. Theseresults indicate that OOc activity alone was qualita-tively diVerent than OOc activity when SCM was alsoevident. Integrated EMG analysis indicates that therewas substantially more OOc EMG activity when SCMwas also detected (Fig. 2b). Together, these data sug-gest that OOc was both qualitatively and quantitativelydiVerent depending on the presence of SCM, and thatOOc activity detected in the absence of SCM wasdiVerent and not necessarily indicative of a startle

response. Thus, OOc may not be the best indicator ofstartle as has been previously suggested (Blumenthalet al. 2005).

In addition to having a diVerent EMG proWle, OOcactivity alone (SCM absent) was not consistently asso-ciated with shorter RTs across the range of intensitiesas was SCM activity (Fig. 1, yellow triangles). At lowerintensity stimulus levels (93, 103 dB), RT was no diVer-ent whether or not OOc was detected. However, evenat these low levels, if SCM was detected, RT was short-ened (Fig. 1, red squares). At the two highest stimulusintensities RT was signiWcantly shorter when OOcalone was detected compared to no startle activity,however, the proportion of trials in which OOc alonewas detected was very low at the two highest stimulusintensities (Fig. 3).

The largest proportion of trials in which SCM activ-ity was observed was, unsurprisingly, at the higheststimulus intensity. This result agrees well with previousstartle studies that have shown larger startle responseamplitudes in response to higher intensity stimuli(Blumenthal 1996; Yeomans and Frankland 1996).Additionally, the proportion of trials in which SCMactivity was observed decreased along with dB level(Fig. 3). Conversely, the proportion of trials in whichno startle response was observed decreased steadily upto the highest intensity. The proportion of trials inwhich blink (OOc) activity was observed alone, how-ever, exhibited a somewhat diVerent pattern. Itappears that the middle intensities were suYcient toelicit a greater proportion of non-startle blinks, whileat the higher intensities a greater proportion of trialsresulted in a SCM response. Thus, in order to have thehighest probability of eliciting a SCM (startle)response in experiments in which the eVect of a startleis being examined, the highest stimulus intensityshould be used. However, great care should be taken toselect an intensity that elicits a startle without risk ofdamage to the auditory system (see NIOSH 1998).

In summary, several conclusions can be drawn fromthe preceding study: First, the facilitation of voluntaryreactions by startle is diVerent than that brought aboutby increasing stimulus intensity. This is because whenSCM startle activity was observed, PMRT was signiW-cantly shortened to approximately 80 ms irrespectiveof stimulus intensity. Secondly, the probability ofobserving SCM activity increased with stimulus inten-sity, leading to an increased number of observations ofconsistently short PMRTs with increasing dB level.Third, when OOc was observed without the presenceof SCM, PMRTs observed were not diVerent from sim-ple intensity eVects except at the highest intensities(when OOc alone was rarely observed, Fig. 3). Fourth,

123

Exp Brain Res

the EMG of OOc when observed alone was both quali-tatively and quantitatively diVerent compared to whenit was paired with SCM indicating the presence of a dis-tinct response. Thus it appears that the stronger thestartle response that is evoked (i.e. SCM vs. OOc only),the greater will be the shortening of RT. As such, OOcEMG, currently the most employed startle indicator(Blumenthal et al. 2005; Brown et al. 1991), may notalways be the best indicator of a strong startleresponse. EMG from SCM may be a more appropriateindicator of startle than OOc, due to its associationwith a separate RT distribution, and its less ambiguousnature.

Acknowledgment This work was supported by a grant from theNatural Sciences and Engineering Research Council of Canadaawarded to I.M.F. and a Michael Smith Foundation for HealthResearch scholarship awarded to A.N.C.

References

Blumenthal TD (1996) Inhibition of the human startle responseis aVected by both prepulse intensity and eliciting stimulusintensity. Biol Psychol 44:85–104

Blumenthal TD, Cuthbert BN, Filion DL, Hackley S, Lipp OV,Van Boxtel A (2005) Committee report: guidelines for hu-man startle eyeblink electromyographic studies. Psycho-physiol 42:1–15

Brown P, Rothwell JC, Thompson PD, Britton TC, Day BL,Marsden CD (1991) New observations on the normal audi-tory startle reXex in man. Brain 114:1891–1902

Carlsen AN, Chua R, Inglis JT, Sanderson DJ, Franks IM (2003)Startle response is dishabituated during a reaction time task.Exp Brain Res 152:510–518

Carlsen AN, Chua R, Inglis JT, Sanderson DJ, Franks IM (2004a)Prepared movements are elicited early by startle. J Mot Be-hav 36:253–264

Carlsen AN, Chua R, Inglis JT, Sanderson DJ, Franks IM (2004b)Can prepared responses be stored subcortically? Exp BrainRes 159:301–309

Davis M (1984) The mammalian startle response. In: Eaton RC(ed) Neural mechanisms of startle behavior. Plenum, NewYork, pp 287–351

Kohfeld DL (1969) EVects of the intensity of auditory and visualready signals on simple reaction time. J Exp Psychol 82:88–95

Levick WR (1973) Variation in the response latency of cat retinalganglion cells. Vis Res 13:837–853

Luce RD (1986) Response times: their role in inferring elemen-tary mental organization. Oxford University Press, NewYork

National Institute for Occupational Safety and Health (NIOSH)(1998) Publication No. 98–126: criteria for a recommendedstandard. CDC, Cincinnati

Rothwell JC, MacKinnon CD, Valls-Solé J (2002) Role of brain-stem-spinal projections in voluntary movement. Mov Disord17:S27–S29

Siegmund GP, Inglis JT, SandersonDJ (2001) Startle response ofhuman neck muscles sculpted by readiness to perform ballis-tic head movements. J Physiol 535:289–300

Valls-Solé J, Rothwell JC, Goulart F, Cossu G, Muñoz E (1999)Patterned ballistic movements triggered by a startle inhealthy humans. J Physiol 516.3:931–938

Wadman WJ, Denier van der Gon JJ, Geuze RH, Mol CR (1979)Control of fast goal-directed arm movements. J Hum MovStud 5:3–17

Woodworth RS (1938) Experimental psychology. Henry Holt,New York

Yeomans JS, Frankland PW (1996) The acoustic startle reXex:neurons and connections. Brain Res Rev 21:301–314

123