The study focused on adults who were not involved in an exercise program at the time of recruitment. Participants were selected based on continuous review of screening questionnaires, and individuals with the lowest relative scores on validated measures of physical activity and fitness, including the stage of change in physical activity (19), weekly moderate-to-vigorous physical activity (MVPA) according to the short version of the International Physical Activity Questionnaire (IPAQ) (20), self-rated physical fitness (21), and self-rated fitness percentile, were preferentially invited. Further exclusion criteria were any acute injuries, physical illnesses, mental illnesses, medication use, and any other medical conditions that did not allow jump exercises. The training study was part of a larger study in which the participants carried out further experiments for which additional exclusion criteria were set (e.g., no tattoos due to MRI requirements). Out of the 890 individuals who completed an online screening, 75 met inclusion criteria and were enrolled in the study. Sixty-five of those enrolled completed the Post + 0 measurements and were therefore included in further analysis (see Figure 1 for details). These 65 participants reported a stage of change in physical activity of 2.6 ± 0.7 (Likert scale from 1 = precontemplation to 5 = maintenance), MVPA of 137 ± 197 min/week, a self-rated physical fitness level of 2.6 ± 0.7 (Likert scale from 1 = very poor to 5 = very good), and a self-rated physical fitness percentile of 32 ± 16 during screening.

The study was approved by the Ethics Committee of the University of Konstanz (No. 31/2022) and conducted in accordance with the latest revision of the Declaration of Helsinki. At Pre, the 65 participants (34 females, age 23 ± 3 years, body height 173 ± 8 cm, body mass 71 ± 12 kg, BMI 23.7 ± 3.1, resting heart rate 78 ± 13 min⁻¹, systolic blood pressure 125 ± 13 mmHg, diastolic blood pressure 83 ± 10 mmHg) signed an informed consent form, after we provided detailed verbal and written information about the study design and procedures, exclusion criteria and potential risks.

We randomly assigned participants to a training group (n = 35; 18 females, age 23 ± 3 years, body height 172 ± 8 cm, body mass 70 ± 10 kg, BMI 23.4 ± 3, resting heart rate 78 ± 13 min⁻¹, systolic blood pressure 125 ± 15 mmHg, diastolic blood pressure 84 ± 11 mmHg) or a control group (n = 30; 16 females, age 23 ± 2 years, body height 173 ± 9 cm, body mass 73 ± 14 kg, BMI 24.1 ± 3.4, resting heart rate 78 ± 13 min⁻¹, systolic blood pressure 124 ± 11 mmHg, diastolic blood pressure 82 ± 9 mmHg) using a computer-generated randomization sequence. Participants of the training group had reported a stage of change in physical activity of 2.7 ± 0.7, MVPA of 140 ± 170 min/week, a self-rated physical fitness level of 2.7 ± 0.7, and a self-rated physical fitness percentile of 35 ± 15 during screening. Corresponding values in the control group were 2.6 ± 0.6, 134 ± 228 min/week, 2.6 ± 0.7, and 30 ± 17, respectively. As we assumed a higher drop-out rate in the training group, we assigned 30 per cent more participants to this group. The participants in the training group performed high-intensity jump training 3 times a week for 8 weeks. At Pre, we administered a battery of tests to assess the physical baseline performance of all participants. After 8 weeks (Post + 0; 0 days after the intervention) and 16 weeks (Post + 63; 63 days after the intervention), we reassessed all participants using the same set of tests with the identical order as in the Pre session.

We measured body mass and height at the beginning of each test session using a calibrated mechanical seca 700 scale with an integrated vertical measuring rod and headpiece (seca, Hamburg, Germany). Afterwards, participants performed a warm-up consisting of 30 s of jumping jacks, 30 s of countermovement jumps (CMJs), and 30 s of high knee running with 30 s of rest between tasks. All exercises were initially demonstrated by the experimenter.

After the warm-up, we assessed CMJ performance, hop performance, and static balance of participants using a Leonardo Mechanograph® GRFP force plate (Novotec Medical GmbH, Pforzheim, Germany) operating at 800 Hz. First, participants performed 3 maximal CMJs with their hands on their hips. We instructed them to stand still until they heard a loud beep, after the beep, they should perform the CMJ with the aim of jumping as high as possible while keeping the legs straight during the flight phase. Participants had a 1-min break between each jump. We used custom written MATLAB© scripts (MathWorks, Natick, MA, USA) to analyze all force traces off-line and conducted all analyses according to previous methodology (22); for details see also Supplementary Material S1. For in depth analysis of jump kinetics we chose specifically the following variables: jump height [m], absolute [kW] and relative (to body mass) [W kg⁻¹] peak power, absolute [kN] and relative [N kg⁻¹] force at zero velocity, upward movement range [m], and upward and braking movement duration [ms]. Jump height was calculated using the impulse-momentum method. For comparability with previous literature using field-based methods, jump height was also calculated using flight time (23) (Supplementary Material S1). For statistical comparisons, we calculated the average of the 3 CMJs for each variable.

Following the CMJs, participants performed 2 sets of 10 reactive hops with the instruction to hop as high as possible and to keep ground contact time as short as possible. From each hop trial, hops 4–8 were retained for further analysis. The following variables were calculated, and averaged across the 5 selected hops, and the two trials: contact and flight time [ms], reactive strength index (ratio of flight to contact time), hop height [m], absolute [kN] and relative [N kg⁻¹] peak force. The calculations were performed using previously reported methods and algorithms (11) (Supplementary Material S2).

Thereafter, participants performed a single-leg standing test to assess static balance. We instructed participants to stand as still as possible on their dominant leg (determined as the leg used to kick a ball) and keep their hands on their hips for 15 s. We measured the change in position of the center of pressure (CoP) on the force plate and derived an ellipse with the area covering 95% of the CoP path [cm²].

We used a dynamometer (Isomed 2000, D&R Ferstl GmbH, Nuremberg, Germany) to determine maximum isometric leg press strength. Participants were positioned in a leg press module with 60° hip flexion, 90° knee flexion and 90° ankle plantarflexion. The leg press was equipped with a force plate to measure ground reaction forces during the maximal leg extension trials. Participants were instructed to “push themselves off as hard as possible from the force plate”. Each trial lasted 3 s, and we repeated it 3 times, with 30 s rest between trials. The positioning data for each participant was stored in the dynamometer, but technical issues led to incorrect standardization for some participants. As a result, we needed to exclude leg press data from 14 training group participants at Pre and Post + 0, and 15 at Post + 63. In the control group, we needed to exclude 18 at Pre and Post + 0, and 17 at Post + 63. Thus, we ended up with 31 participants in total who were included in the final analysis of leg press data. We took the maximum force value from each trial and determined the MVC for each participant as the mean of the maximum force values of the 3 trials.

To assess potential transfer effects in activities of daily living, we conducted a 14-step stair climb and a gait test. The 14-step stair climb was performed at habitual speed. Prior to the stair climb, we asked participants to stand still for 2 min. Immediately, after the 2 min we measured their heart rate [min⁻¹] by a sports watch (Polar Vantage 2©, Polar OY, Kempele, Finland) with a chest strap and a heart rate monitor (Polar M10 HR-monitor©, Polar OY, Kempele, Finland). Thereafter, the participants climbed up the stairs, and we assessed the climbing time [s] with a stopwatch. Then participants stood still, and we measured the heart rate peak [min⁻¹] within a 30 s-window after the stair climb test. Following the stair climb test, we conducted 8 gait test trials using the OptoGait© system (Microgate©, Bolzano, Italy) over a 10 m corridor. We instructed participants to walk at an “everyday pace, as if walking calmly to the bus stop”. We extracted gait variables from the proprietary OptoGait© software, which provides the mean plus standard deviation for stride time [s], stride length [m] and mean walking speed [ms⁻¹] of the 8 trials.

Lastly, we performed a cardiopulmonary exercise test (CPET). Participants laid down for 5 min and we determined resting heart rate [min⁻¹] (Polar M10 HR-monitor©, Polar OY, Kempele, Finland) and blood pressure (Boso medicus X©, BOSCH + SOHN GmbH u. Co. KG, Jungingen, Germany). Thereafter, participants performed an incremental ramp test on a cycle-ergometer (Ergoline GmbH, Bitz, Germany) while we measured heart rate and breath-by-breath respiratory gas exchange with an Ergostik© gas analyzer (Geratherm Respiratory GmbH, Bad Kissingen, Germany) and the Blue Cherry© diagnostic platform. The test started with a 5-min warm-up at 35 W, during which we instructed participants to “ride straight ahead for 5 min on a flat road at a comfortable everyday pace as if taking a leisurely Sunday stroll”. After the 5min warm-up, power output increased by 3 W every 12 s (15 W per min) and participants kept their cadence between 70 and 80 min⁻¹. We determined exhaustion when participants could no longer maintain 60 min⁻¹, or until they could go no further. Further, we determined V˙O2peak [ml kg⁻¹min⁻¹] as the peak value of a 30-s moving average of breath-by-breath oxygen uptake. We determined peak power [W] and peak heart rate [min⁻¹] as the maximum values that were achieved during the test. During off-line analysis we identified artifacts in O2 and CO2 breath-by-breath data that affected V˙O2peak detection, likely due to equipment failure or incorrect mask positioning. In total we needed to exclude five measurements, which led to a drop-out of five participants specifically for the CPET (3 control, 2 training).

The participants of the training group performed jump training sessions of 15 min on 3 days a week for 8 weeks. We instructed the participants of the control group to maintain their normal daily routine for 8 weeks and they did not receive any training or any other intervention during this time. For the training group, we gradually increased the intensity and volume of the training, with a minimum of 24 h rest between 2 sessions. We instructed the training group to perform 3 sessions per week, with 1 session taking place under controlled laboratory conditions at the Human Performance Research Centre (HPRC), University of Konstanz. We fully instructed the participants on how to perform the remaining 2 weekly sessions, which were performed at home. Each training session consisted of 4 parts: a standardized warm-up (30 s jumping jacks, 30 s submaximal CMJs, and 30 s high knee running), hops, CMJs and high-intensity CMJ training. We varied CMJs and hops in number of repetitions and length of rest to progressively increase the training effort throughout the training program (see Supplementary Material S3 for information about the training programming per session). We thoroughly explained all parts of the training during the first training session, which was held at the HPRC. We further provided a detailed training plan to the participants (Supplementary Material S4) containing video links that allowed participants to review the warm-up and exercises at home.

We provided all participants with a sports watch (Polar Vantage 2©, Polar OY, Kempele, Finland) to store and track their training units. The watch guided the participants through each training session, automatically indicating the start and end of each exercise, as well as the duration, number of repetitions and length of breaks. Jump training adherence was defined as the number of recorded jump training sessions lasting longer than 10 min, expressed relative to the prescribed total of 24 sessions. Peak training intensity was calculated as the peak heart rate during each session, expressed as a percentage of the individual's peak heart rate determined during the CPET at Pre. While this measure reflects cardiovascular training intensity, neuromuscular load was maximized by instructing participants to perform all jumps with maximum effort.

We conducted all statistical analyses in R (version R4.2.2) and created figures using ggplot2 or MATLAB© (TheMathworks, Natick, USA). To test for baseline differences between groups, we ran independent t-tests for all variables at Pre using the R stats package. To examine the effects of training over time, we computed linear mixed effects models using the lmer function (lme4 package). We treated performance measures as dependent variables, with fixed effects for time (Pre, Post + 0, Post + 63), group (training, control), and their interaction (time × group). We added a random intercept for participants to account for within-subject variability and evaluated model significance by type III ANOVA using the Anova function (car package). In case of significant group × time interaction effects, we executed post-hoc pairwise comparisons with estimated marginal means (emmeans) using the Bonferroni correction. We calculated effect sizes as Cohen's d for pairwise comparisons and partial eta squared (η²) for ANOVA main and interaction effects.

To gain more detailed insights into the observed changes in CMJ technique, we applied functional principal component analysis (24) to the normalized relative force-time traces, which is an unsupervised learning technique used to identify the functions explaining the greatest variance in relative force from pre-training to Post + 0. Details on this procedure are provided in Supplementary Material S5.

To explore relationships between baseline scores and changes in CMJ and aerobic performance, we calculated pre-post delta scores for CMJ height and V˙O2peak, e.g., V˙O2peak changes as ((V˙O2peak post − V˙O2peak pre)/(V˙O2peak pre × 100)). We used Pearson's correlation coefficients for analysis and visualized the relationships with scatter plots. Furthermore, we fitted linear regression models to the delta plots and plotted confidence intervals using geom_smooth in ggplot2.

Performance improvements in CMJs can primarily be attributed to an optimized movement strategy. A greater countermovement depth increased the time available for impulse generation, thereby enhancing takeoff velocity and jump height. Notably, these benefits came at the expense of relative force during much of the movement. Relative force-velocity traces demonstrated that, after training, force production was reduced during both braking and upward phases, except near takeoff where higher relative forces at greater velocities facilitated elevated peak power output. This observation aligns with recent findings showing that deeper countermovements compromise force generation at low velocities but enhance it at higher velocities, ultimately leading to improved performance (25). As no external feedback was provided during training, these technique-related adaptations likely emerged spontaneously.

Functional principal component analysis confirmed a decrease in relative force across most of the movement, with higher relative force emerging only near takeoff (first fPC and associated scores). No similar changes were seen in the control group, supporting the conclusion that CMJ performance gains were primarily technique-driven. Importantly, CMJ improvements were independent of baseline performance levels. This can be explained by neural plasticity that usually evolves early during training and leads to task specific increases in neuromuscular performance, independent of the initial performance in novices to the trained task (26).

Despite hops being an integral training component, hop test performance did not improve. Previous studies have shown that hop height can improve in both older adults and athletes, yet such gains typically stem from changes in the mechanical properties of the muscle–tendon complex rather than altered motor strategies (27, 28). This may explain why CMJ performance improved while hops did not. In this study, neuroplastic adaptations seemed insufficient to affect hop performance, and changes in muscle–tendon function were likely too small to drive measurable improvements (9, 27, 28).

Although CMJ peak power improved, no training effects were observed in maximal isometric leg extension strength or functional assessments, including one-leg stance, 14-step stair climb, and 10 m gait. These findings are consistent with previous research showing that jump-specific adaptations do not necessarily generalize to unrelated motor tasks (29). It is also plausible that baseline function in these tests was already high, leaving limited scope for further improvement (30–32).

In gait analysis, both groups exhibited reduced stride length and gait velocity after training. As no group-by-time interaction was detected, these changes likely reflect test-related familiarization effects rather than training-induced adaptations (33). This interpretation is supported by the reduced stair climb times observed in both groups at follow-up assessments.

Training also produced baseline-dependent effects on aerobic performance. Within the training group, participants with low baseline aerobic fitness showed clear improvements, whereas those with moderate or high fitness levels did not. This negative correlation likely reflects the overload principle: adaptations occur only if the physiological system is challenged beyond its habitual load (34, 35). Although aerobic training was not a central focus of the intervention, some sessions included continuous jump bouts of ≥3 min, resembling high-intensity interval protocols (36). Despite the limited volume, this load may have been sufficient to induce cardiovascular adaptations in participants with low initial capacity. High-intensity training is generally more effective than moderate continuous exercise for untrained adults (37, 38), but a dose–response relationship has not yet been established for jump training (39).

Evidence from our previous bed rest studies supports the potential of jump training to preserve aerobic capacity. During 60 days of bed rest, a jump training intervention with higher weekly frequency (5–6 sessions) and similar session length almost fully preserved aerobic performance compared to controls (11, 13, 17). In the present study, the lower frequency (3 sessions/week) and absence of confinement likely reduced the overall cardiovascular load, limiting effects to individuals with lower baseline fitness. Exploratory threshold estimation suggested a baseline aerobic capacity below ∼37 mL kg⁻¹min⁻¹ may be required for measurable improvements, though this finding should be interpreted cautiously as it was not hypothesis-driven. In addition, because baseline V˙O2peak is included in both the predictor and the relative change score, the observed association may in part be driven by regression-to-the-mean effects.

The current training program was designed to be scalable, requiring only 15 min per session and three sessions per week. This minimal dose was sufficient to elicit meaningful adaptations in unfit but otherwise healthy young adults compared to baseline levels in literature (40–44). Since it has already been shown that jump training can also elicit more generic improvements in untrained participants (45), future research should evaluate strategies to optimize its effectiveness. Extending session duration (20–30 min), increasing frequency (4–5 sessions per week), or a combination of both may amplify adaptations. Additionally, tailoring training to baseline performance (e.g., focusing more on aerobic vs. anaerobic capacity) may further expand outcomes across diverse populations.

Participants were recruited as non-exercising young adults. Although the recruitment strategy aimed to identify individuals with relatively low levels of self-reported physical activity, baseline screening data indicated that average MVPA was below but close to the World Health Organization recommendation of at least 150 min/week (46). Thus, as objective activity thresholds were not used as inclusion criteria, the sample can be characterized as non-exercising but relatively active in daily life rather than strictly inactive. This may have limited the potential for training-induced improvements in generic performance outcomes due to ceiling effects. Furthermore, the sample consisted of young, healthy adults, and the present findings may not generalize to older adults, clinical populations, or individuals with mobility limitations. Future studies should therefore incorporate objective activity-based screening procedures to more precisely target low-activity populations and examine the effects of low-volume, high-intensity jump training across older and clinical cohorts. Additionally, continuous activity tracking during the intervention could be useful to better control for potential influences from changes in habitual physical activity.

The 8-week high-intensity jump training proved to be a feasible intervention for a previously non-exercising population, resulting in meaningful improvements in CMJ height and peak power. However, these adaptations did not translate to gains in maximal isometric strength or functional performance measures such as balance, walking, or stair climbing. Aerobic benefits were observed in individuals with low baseline fitness, while overall effects on aerobic performance remained limited. Collectively, these results emphasize the importance of adjusting the load and progression of jump training to the individual capacities of the participants to achieve general enhancements in physical performance.