Considering an effect size of d = 1.55 to discriminate between RT differences in motor tasks (Carlsen, 2015) (α = 0.05, power = 0.95), 8 participants were needed as calculated using G*Power software (version 3.1.9.6, Germany) (Faul et al., 2007). Fifteen healthy adults [7 women; age: 25.9 (5.2) years old; weight: 68.3 (11.4) kg; height: 172.8 (9.9) cm] were recruited using a convenience sample for a single experimental session to ensure that smaller effects can be detected. To be included, participants needed to be aged between 18 and 40 years old. Exclusion criteria were: presence of low back pain limiting everyday activities, pathology of the auditory system, idiopathic scoliosis (Hatzilazaridis et al., 2019) and any major pathologies that could interfere with the tasks tested in this study. The study was approved by the Institutional Research Ethics Committee of the Centre Intégré Universitaire de Santé et de Services Sociaux de la Capitale-Nationale—Réadaptation (Project No. #2019-1778), all experiments were performed in accordance with the Declaration of Helsinki, and all participants provided their written informed consent before participation.

The susceptibility to StartReact effect was tested using startling and non-startling acoustic stimuli (SAS/NSAS—see StartReact protocol). The acoustic stimuli were delivered within a simple precued RT task in which participants performed a postural or a voluntary task (see Study design).

Shapiro–Wilk test was used to assess the normality of distribution. If normality was not met, a log-transformation was used to normalize the data sets.

The significance level was set at p < 0.05. Statistical analyses were performed using SPSS (IBM SPSS statistics version 30)¹ and figures were created using Prism software (Graphpad Prism for Windows version 10.4.2).² Data are presented as [mean (Standard deviation)] throughout the manuscript unless otherwise specified. To compare the StartReact effect of LES during voluntary and postural tasks, a linear mixed model was computed on fixed factors Task (postural, voluntary) and Condition (80, 116 dB) for RT with participant’s intercept as a fixed factor and a scaled identity covariance matrix. RT alone was used as a covariate considering it was significantly different between tasks. A sensitivity analysis using the same LMM model but with the CDF variables (80 dB, 116 dB_slow, 116 dB_fast) as fixed factors was also computed. Bonferroni corrections were applied to pairwise comparisons.

Figure 3 illustrates the EMG response of the LES, AD and SCM during SAS and NSAS without motor preparation for one participant. Table 1 presents the mean latency and occurrence rate of the startle reflex for the different responses and conditions. SCM data for 3 participants were excluded because of technical issues.

SAS alone elicited activation of LES, AD and SCM in 70.0 (26.7)% (14/15 subjects; 105/150 stimulations), 19.3 (24.6)% (8/15 subjects; 29/150 stimulations) and 46.7 (28.1)% (11/12 subjects; 56/120 stimulations) of trials at a mean latency of 104.9 (29.5) ms, 93.4 (30.4) ms and 71.6 (31.6) ms, respectively. As illustrated in Figure 3, a short-latency inhibition period occurred in LES with SAS in 34.0 (29.7)% (12/15 subjects; 51/150 stimulations) of trials at a mean latency of 60.4 (22.2) ms and was negligible with NSAS (2/15 subjects; 2/150 stimulations). For NSAS, responses in LES and SCM were elicited in only 8.7 (9.9)% (7/15 subjects; 13/150 stimulations) and 2.5 (6.2)% (2/12 subjects; 3/120 stimulations) of trials at a mean latency of 91.3 (27.1) ms and 89.2 (11.9) ms, respectively. No response was elicited in AD. An additional participant with an obvious short-latency inhibition period is depicted in Supplementary Figure 1.

The data was not normally distributed, so it was log-transformed. Nonetheless, non-transformed data are reported to facilitate interpretation unless otherwise specified.

Figures 4, 5 present average and individual EMG traces of a participant for RT alone, NSAS (80 dB) and SAS (116 dB) in LES, AD and SCM for the postural and voluntary tasks, respectively. Note the progressive reduction in muscle onsets from the “RT task alone” (later) to SAS (earlier) conditions for all muscles tested. Also note the similar effects of auditory stimuli on the whole motor pattern which is particularly obvious on the second burst of activation for both AD and LES in the postural task (Figure 4).

Earlier RT were observed in the 116 dB condition [84.7 (14.4) ms] compared to 80 dB [100.4 (21.2) ms—main effect: Condition | F_{1, 40.145} = 28.433, p < 0.001] regardless of the task. Considering that the interaction was not significant (F_{1, 40.145} = 0.17; p = 0.68), the StartReact effect was not different between tasks (Table 2 and Figure 6). Sensitivity analysis using CMF supports the latter results. Indeed, earlier RT was present in the 116 dB_fast condition [72.5 (12.1) ms] compared to 116 dB_slow [99.0 (18.3) ms; p < 0.001] and 80 dB (p < 0.001—main effect: Condition | F_{2, 68.317} = 76.06, p < 0.001), without difference between 116 dB_slow and 80 dB (p = 1.00—Supplementary Figure 2 and Supplementary Table 1). However, RT during the postural task was earlier compared to the voluntary task (main effect: Task | F_{1, 78.148} = 4.29, p = 0.04). Still, the interaction was not significant (F_{2, 68.317} = 0.32; p = 0.73).

At authors’ knowledge, this study is the first to evaluate the startle reflex in LES. Several findings are of interest. First, a rostro-caudal order of activation was present in the muscles tested (SCM: 71.6 ms; AD 93.4 ms; LES 104.9 ms) as already reported for other muscle groups (Brown et al., 1991; Valls-Solé et al., 1995). Second, we found a surprisingly high occurrence of startle responses with the SAS condition in LES (70.0%), compared to AD (19.3%) and even compared to SCM (46.7%). Third, a short latency LES inhibition, prior to excitation, was observed with the SAS condition. Lastly, with participants seated still, the 116 dB condition elicited startle responses in LES, AD and SCM muscles at a far greater occurrence than the 80 dB condition. Altogether these results suggest a brainstem origin of the startle reflex and support the applicability of our parameters in the assessment of LES control using the StartReact paradigm.

Our study tested motor tasks in which back muscles acted as (i) the prime mover of the spine (voluntary task) and (ii) as a postural controller through APA. Overall, StartReact effects were observed for both voluntary and postural tasks suggesting the involvement of the reticulospinal system in the control of low back muscles. Nonetheless, our results do not support a larger reticulospinal drive for the voluntary task using both analytical methods (SAS/NSAS; CDF). Using the CDF analysis, the shortening during the voluntary task was greater although not significant. This larger shortening may be explained by (i) the longer LES RT in the voluntary task (more “space” for RT shortening) and (ii) a potential floor effect for the postural task, as elegantly demonstrated in a recent study (Tapia et al., 2022). Currently, different hypotheses on underlying mechanisms explaining the StartReact effect have been proposed (Carlsen and Maslovat, 2019; Nonnekes et al., 2015). Indeed, SAS may increase the excitability of the reticular formation which accelerate the rise of the motoneuron threshold and shorten the RT (Tapia et al., 2022). Different evidence seems to support the brainstem hypothesis. First, patients with hereditary spastic paraplegia (consisting of retrograde axonal degeneration of the corticospinal tract, but not of the reticulospinal tract) have an intact StartReact effect although their RT is delayed during a voluntary task (Nonnekes et al., 2014), thus strongly supporting a role of the reticulospinal tract in the StartReact effect. Second, a study in non-human primates using in vivo electrophysiology recording and computational model strongly supports the contribution of the reticulospinal tract to shorten the RT rather than the corticospinal tract (Tapia et al., 2022). Third, by using EEG and high density-EMG in a reaction time task, Neumann et al. (2025) found that movement-related cortical potentials emerged only 65 ms after muscle activation during the StartReact paradigm, suggesting that the motor cortex could not be critically involved in accelerating the initiation of movement induced by SAS. Nevertheless, there is also some evidence of cortical influence on the StartReact paradigm, including delayed RT by TMS-induced silent period (Alibiglou and MacKinnon, 2012) and modulation of motor cortex interneurons excitability by SAS (Marinovic et al., 2014). Altogether, it seems unlikely that an exclusive circuit contributes to the StartReact effect (Marinovic and Tresilian, 2016). Still, recent evidence strongly support a substantial contribution of the reticulospinal tract (Tapia et al., 2022) even though other brain areas—including the motor cortex—may certainly have an influence (Carlsen and Maslovat, 2019; Valls-Sole, 2012; Nonnekes et al., 2015).

During the voluntary task, the LES onset was reduced by an average of 57 ms compared to trials without sound. This is less than the seminal work from Valls-Sole et al. (1999) that observed a reduction of ≈100 ms while raising the arm, but similar and even larger than studies testing upper limb muscles (Carlsen et al., 2009). For example, Carlsen et al. (2009) observed a reduction in RT of the first dorsal interosseus muscle of ≈35 ms during a finger task and ≈55 ms for the biceps and triceps muscle during arm movement. Greater shortening was also observed during a shoulder task compared to a finger task (Maslovat et al., 2023), and a bimanual compared to a unimanual task of the finger (Maslovat et al., 2020). Authors propose that a greater shortening in RT suggests a larger relative degree of reticulospinal contribution. This latter hypothesis could also apply to our results; strong connectivity from the brainstem to LES was observed in animal studies (Lawrence and Kuypers, 1968b; Galea et al., 2010). Although no difference was present between tasks, a StartReact effect was present during the voluntary task, which suggests a reticulospinal drive in the control of LES for this task. These results are in accordance with recent literature from Eilfort et al. (2025) and Hayman et al. (2025) showing that the StartReact effect is lower in more distal muscles and those that are primarily controlled by corticospinal input (e.g., tibialis anterior, finger muscles), whereas it is stronger for extensors (anti-gravity muscles, lower limbs) and proximal muscles, characteristics that fully apply to the LES. Considering the limited contribution of the corticospinal tract observed in this same task (Desmons et al., 2024), it suggests a greater subcortical control of the lumbar lordosis position in sitting (Lawrence and Kuypers, 1968b).

For the postural task, SAS shortened the RT of low back muscles by ≈44 ms compared to trials without sound. Some studies tested APA elicited by rapid limb movement (Chiou et al., 2024), gait initiation (MacKinnon et al., 2007; Delval et al., 2012) and sit-to-stand maneuver (Queralt et al., 2008). Results differ substantially depending on the task and muscles tested. Using a similar APA task involving rapid arm flexion, Chiou et al. (2024) observed a ≈ 70 ms shortening of the low back muscles RT compared to a trial without sound, Queralt et al. (2008) observed a ≈ 146 ms shortening of the low back muscles RT during a sit-to-stand task. Similarly, MacKinnon et al. (2007) observed a ≈ 123 ms reduction in RT of tibialis anterior muscle during APA elicited by step initiation. Nonnekes et al. (2013) rather observed a reduction in tibialis anterior RT during a backward postural perturbation of ≈20 ms. Nevertheless, APA of the lower limb can be elicited at very short latencies without changing the motor patterns (Valls-Sole et al., 1999; MacKinnon et al., 2007). The shortening observed during our postural task seem to lie in-between those already reported in the literature and can be explained by different tasks and muscles tested. As already discussed, considering that a StartReact effect was observed using our two different analyses, our results support a certain degree of reticulospinal contribution during the postural task. Interestingly, our previous studies using TMS support a modulation of both corticospinal and cortical excitability during this same postural task (Desmons et al., 2024; Masse-Alarie et al., 2018; Chiou et al., 2018). It highlights the possibility of the contribution of a widespread networks of neural areas—including both cortical and subcortical—in the control of back muscles during a APA task, as suggested in animal studies (Drew et al., 2004).

The StartReact paradigm remains one of the few techniques available in humans to evaluate the reticulospinal drive in different tasks and/or muscles. Recent studies have highlighted the usefulness of this approach for understanding muscle recruitment, strength and motor recovery. Findings suggest a higher reticulospinal drive in trained compared to untrained individuals (Akalu et al., 2024) as well as a greater rate of torque development for reaction time tasks with SAS compared to NSAS in the vastus lateralis and medialis (Skarabot et al., 2022), the tibialis anterior (Hayman et al., 2025) and the biceps brachii (Walker et al., 2024) muscles. These findings point toward a significant role of the reticulospinal system in the rapid recruitment of motor units and in the initial development of force, which translates to an accelerated transition from muscle activation to movement (Eilfort and Filli, 2025). Given the role of back muscles in rapid postural adjustments and trunk movement, our results, showing a reticulospinal contribution to both voluntary and postural control of the LES, appear in line with the most recent findings in the field.

Our design did not allow to directly test subcortical contribution considering that it would necessitate invasive stimulations. Some authors propose to use the presence/absence of a SCM startle reflex to consider the StartReact effect as originating from subcortical networks (Carlsen, 2015; Carlsen et al., 2009). Although it was not possible to use this technique in our study, results from the SAS/NSAS analysis were replicated using the CMF analysis (McInnes et al., 2021) increasing the confidence in our findings. Also, we did not measure the movement pattern which is one feature of the StartReact effect (Valls-Sole et al., 1999). Finally, data from only 12 participants were used for the SMC due to EMG technical issues for 3 participants.