Fifty-three healthy children (30 females and 23 males; age: 11.3 ± 0.6 years; body height: 161.1 ± 7.7 cm; body mass: 57.6 ± 14 kg; body mass index: 22.1 ± 4.7 kg/m²) voluntarily participated in this study. All participants attended the sixth grade of a public secondary school in the Ruhr area of North Rhine-Westphalia, Germany. Subjects were unaware of the training protocol and had no previous experience with the implemented set of tests. All participants and their legal guardians provided informed consent before entering the study. The study protocol was approved by the Human Ethics Committee of the University of Duisburg-Essen, Faculty of Educational Sciences (EA-PSY20/23/04102023).

Using G^*Power (version 3.1.9.7; Faul et al., 2007) the power analysis (f = 0.25, α = 0.05, 1-β = 0.80, number of groups: n = 2, number of measurements: n = 2, correlation between testing: r = 0.30, drop-out rate per group: 10% due to injury reasons not attributable to treatments) revealed that a total sample size of N = 50 participants would be sufficient to detect statistically significant treatment effects.

The study was conducted from October to December 2023. As the investigation was performed within a school setting, group allocation on a class basis took place by lottery. One class of 27 pupils (12 females and 15 males) was therefore assigned to be the intervention group (INT) and performed a 45-min spatial ability intervention twice per week. The other class of 26 pupils (11 females and 15 males) was assigned to be the control group (CON) and attended regular PE class using the same volume. Prior to the intervention, a pretest was performed with all participants. The identical assessment was performed as posttest upon completion of the intervention period (Figure 1).

The pretest and posttest each took place on 3 testing days and covered a variety of spatial ability aspects. Standardized verbal instructions were provided prior to each assessment. The 1st testing day was executed in the students' classroom, where participants individually performed three paper-and-pencil assessments from a test-booklet. A visual demonstration of each test was performed during instructions. The Paper Folding Test (PFT; Ekstrom et al., 1976) was chosen as a measure of spatial visualization. The task consisted of the depiction of 20 sheets of paper that were folded in different ways, a hole was then punched through them, and the paper was unfolded again. For each item, participants had to select the correct answer out of five options. The PFT was divided into two parts with ten items each. Each part had a time-limit of 3 min separated by a 3-min break. The PFT included one practice item that had to be completed prior to the first part. A maximum of 20 points could be reached. To evaluate mental rotation, the Mental Rotation Test (MRT) Version A by Peters et al. (1995; based on Vandenberg and Kuse, 1978) was selected. A three-dimensional block figure was depicted as a target object. Out of four possible answers, participants had to decide which two showed rotated and tilted versions of the target object. Similar to the PFT, the MRT comprised of two parts with 12 items each. Each part had to be solved within a time-limit of 3 min with 3 min break in between. Four practice items with a time-limit of 5 min were completed in advance. One point was allocated if both answers were correct resulting in a maximum of 24 points. Lastly, the Water Level Task (WLT; Piaget et al., 1956) in the version by Yingying Yang (University of Alabama; Merrill et al., 2016) was employed to test spatial perception. Participants had to imagine how the water level would look like when 12 empty jars tilted to various degrees were half filled with water. They were then asked to draw the respective waterlines. The time-limit was 3 min. For lines within the tolerance range of ±10° from horizontal, participants received one point, amounting to a maximum of 12 points. Reliability of the PFT (ICC = 0.78), MRT (ICC = 0.81), and WLT (ICC = 0.88) has been shown in one of our previous studies (Morawietz et al., 2024).

Spatial learning, visuospatial short-term- and working memory were assessed on the 2nd testing day using a computerized version of the Corsi Block Test (CBT; Corsi, 1972). The self-programmed test was based on the online-demo of Millisecond Software, Seattle, USA and in line with the normative data provided by Kessels et al. (2000). While instructions took place in small groups of five students, the test was performed by each participant individually in a separate room. Nine blue squares were presented on a black laptop screen. In sequences of increasing length some of the squares lit up in yellow. Starting with two squares (block sequence one and two), one square was added to every subsequent sequence. Using a computer mouse, participants were asked to immediately repeat the sequences. At least one out of two trials per block sequence length had to be repeated correctly. Otherwise, the test was instantly terminated. The CBT was performed without time-limit and the block span (max. 9 points), and total score (max. 144 points) were calculated. The CBT (span: ICC = 0.95; composite score: ICC = 0.93) proved to be reliable in one of our previous studies (Morawietz et al., 2024).

Evaluating spatial orientation, an adapted version of the Medicine Ball Number Run (Jung, 1983 as cited in Hirtz et al., 2010), the Numbered Cones Run (NCR), was performed on the 3rd day. Instruction and assessment took place in small groups of five students. Five cones were randomly numbered and set up in a semicircle with 1.5-m distance from each other. The starting point was marked at 3-m distance. Standing with their back to the numbered cones, a number was called out to the participants, and they were asked to run to the respective cone as fast as possible, touch it and return to the starting point. Upon return, the procedure was repeated immediately for a second and third time. The running order was predetermined by a random number generator and differed between participants. Reliability of the NCR (ICC = 0.91) has been shown in one of our previous studies (Morawietz et al., 2024).

The training took place during regular PE class over a period of 6 weeks (twice per week). Each PE class lasted 60 min and included a 5–10-min warm-up and 30–35 min of exercise. Fifteen minutes were allocated to changing and setting up/taking down of equipment. While the CON followed the regularly scheduled curriculum with wrestling exercises, the INT performed a specifically designed training program focusing on motor coordination with spatial ability elements (Table 1). Training of the INT was performed by a graduated sport scientist. The training program consisted of 12 self-contained sessions and was conceptualized on the basis of various intervention studies in the field of spatial abilities (e.g., Jansen et al., 2011, 2013; Blüchel et al., 2013; Dirksen et al., 2015; Pietsch et al., 2017; Boraczyński, 2019; Latino et al., 2021). Training sessions included Le Parkour inspired obstacle courses; obstacle courses that needed to be completed as a team; stations with coordination exercises; ball coordination exercises including throwing and catching; juggling; football and volleyball coordination; dancing and coordination games. Possibilities for progression were offered and encouraged for all tasks applicable.

Prior to the conduction of parametric analyses, normal distribution (Shapiro–Wilk Test) and variance homogeneity (Mauchly Test) were checked and confirmed. Data were presented as group mean value ± standard deviation (SD). Afterwards, a series of 2 (test: pretest, posttest) × 2 (group: intervention, control) repeated measures ANOVA were performed. If a significant test by group interaction occurred, Bonferroni-adjusted post-hoc tests (i.e., paired t-test) were applied. The significance level was a priori set at α <0.05. For the ANOVA, effect size was calculated as partial eta-squared (ηp2) and reported as small (0.02 ≤ ηp2 ≤ 0.12), medium (0.13 ≤ ηp2 ≤ 0.25), or large (ηp2 ≥ 0.26; Cohen, 1988). For the paired t-test, effect size was calculated as Cohen's d and stated as small (0 ≤ d < 0.50), medium (0.50 ≤ d < 0.80), or large (d ≥ 0.80; Cohen, 1988). All analyses were performed using SPSS version 28.0 (IBM Inc., Chicago, IL).

Our findings are supported by previous research evaluating the effect of motoric interventions on spatial abilities in children and adolescents. Mental rotation performance of the INT increased from 7.2 to 10.4 points (group mean values) from pre- to posttest but remained stable in the CON (pretest: 7.0 points, posttest 6.9 points) indicating that the intervention positively affected mental rotation performance of the INT. Blüchel et al. (2013) for example, report that mental rotation performance in children (mean age ± SD: 9.06 ± 0.45 years) improved significantly after 2 weeks of specific coordinative motor training including throwing, catching and bouncing tasks, skateboarding or motor memory games compared to children attending regular classroom-based lessons. The MRT and the test conduction was identical to the one used in the present study. Similar results are disclosed by Jansen et al. (2013) who recorded improvements in mental rotation performance of children (mean age ± SD: 7.68 ± 0.503 years) after 5 weeks of creative dance training as well as Pietsch et al. (2017) who found positive effects after performing a Life Kinetik-motion program with cognitive, coordinative, and visual task complexes respectively with children (mean age ± SD: 8.65 ± 0.482 years). Control groups in both studies attended regular PE-class and child-adapted MRTs were applied with animals or letters as stimuli.

Similar findings were observed for the WLT. While the INT increased their group mean values from 4.3 to 5.9 points between pre- and posttest, slight decreases were observed in the CON (pretest: 5.9 points, posttest: 4.9 points). Even though the test has been extensively researched over the past decades, to our knowledge no other motoric intervention studies made use of the WLT as measure of spatial abilities impeding the discussion of our findings based on previous literature. Various developmental stages of problem-solving have been proposed by Piaget et al. (1956) which are associated with the age of the participants. Their approach has been refuted over the years because even though all strategies appear to be used by children when solving the task, they appear to be used independent of age and switches between strategies occur (Thomas et al., 1993). Thomas et al. (1993) discuss three different response strategies (i.e., a bottom parallel approach, a random-like approach, and a correct approach) which are commonly seen in children and adolescents. While younger children tend to use a bottom parallel approach and the correct understanding tends to develop with increasing age, shifts between strategies and influential factors like the degree of inclination of the vessel or field effects also appear to affect responses (Thomas et al., 1993; Lohaus et al., 1996; Vasta and Liben, 1996). This does not only account for children but also for adults. Moreover, knowledge and understanding of the principle of horizontality of liquids appears to be inevitable to solve the WLT (Hecht and Proffitt, 1995). Hecht and Proffitt (1995) further established that adoption of an allocentric frame of reference is required to master the task successfully. This finding is supported by Barhorst-Cates et al. (2020), who report that children (mean age ± SD: 9.28 ± 0.76 years) who are able to adopt environmental views when asked to draw a map of their home environment tend to perform better on the WLT than children who adopt an egocentric perspective. Looking at our results, only few participants managed to solve the majority of items correctly and deviations of more than 50° occurred at pre- and posttest. Still spatial perception [i.e., spatial relations in regard of the own body's position as defined by Linn and Petersen (1985)] or extrinsic-static spatial skills [i.e., the spatial relation of an object in context of other objects or the environment according to the categorization of spatial abilities by Newcombe and Shipley (2015)] as measured by the WLT appear to be addressed and promoted by our intervention.

The INT also improved their performance on the CBT. The span as well as the CS increased from 4.7 points (pretest) to 5.7 points (posttest) and from 30.8 points (pretest) to 45.4 points (posttest), respectively. In contrast, values of the CON remained stable or even decreased slightly from pretest (span: 5.0 points; CS: 35.6 points) to posttest (span: 5.0 points; CS: 34.6 points). Comparable findings were obtained by Latino et al. (2021) who had adolescents (mean age ± SD: 14.4 ± 0.5 years) perform a 12-week coordinative ability training designed to improve cognitive skills. Performance on the CBT increased significantly from pre- to posttest in the INT but not in the CON, who participated in general psycho-physical wellness program. Similarly, Notarnicola et al. (2012) found significant differences between groups on the CBT forward and backward in favor of the INT after 26 weeks of orienteering lessons and orienteering exercises compared to the jogging CON. Moreover, the INT improved significantly from pre- to posttest on the backward CBT. Participant's mean age in this study was 9 years. Supported by these previous studies, our findings point toward an enhancement of visual-spatial working memory function based on the intervention.

Finally, results on the NCR revealed significantly faster completion times for the INT when looking at pre- and posttest (11.0 and 10.3 s, respectively), indicating positive effects of the motor-coordination intervention on spatial orientation. The CON on the other hand was not able to maintain or improve their performance (pretest: 10.5 s; posttest: 10.7 s). To our knowledge, the NCR is not widely used in the current research community. A study by Dirksen et al. (2015) found a significant main effect of time but not of group and no interaction effects when evaluating the results of 20 weeks of movement-coordination exercises during PE class compared to participation in regular PE class in children (mean age ± SD: 12 ± 0.46 years). Even though training content and participant age are similar the results of the study by Dirksen et al. (2015) and the results of the present study differ notably. This could be due to the fact that the training modalities [15 min/session, 2x/week for 20 weeks in Dirksen et al. (2015) vs. 45 min/session, 2x/week for 6 weeks in the present study] as well as training intensity and levels of load differed. Moreover, in contrast to the present study, post-hoc power analysis revealed that the sample was too small (n = 91) to reveal significant interaction effects due to high drop-out rates (Dirksen et al., 2015).

Unexpectedly, the PFT only showed a main effect of test but no test × group interaction effects. There are several possible reasons for this. For one, a recent paper by Uttal et al. (2024) aptly summarizes the difficulties in relation to spatial testing. Disagreement about spatial constructs in general and about the question which tests evaluate which spatial constructs in particular are widespread in the research community (Uttal et al., 2013, 2024). Therefore, the possibility exists that the PFT does not measure spatial visualization as expected by us. Burte et al. (2019) for example, claim that frequently different strategies (i.e., not only spatial visualization) are used to solve the tasks of the PFT and that participants might switch between strategies depending on the test item. Moreover, Uttal et al. (2024) point out that many tests have been designed and are available for certain age groups only with tests lacking that can be used across age groups. According to the test's authors, the PFT is suited for students from grade nine to 16 (Ekstrom et al., 1976). We tried to overcome this issue by evaluating test-retest reliability and minimal detectable change (MDC_95%; i.e., the clinically relevant effect between repeated measurements of one subject) in younger children (mean age ± SD: 11.4 ± 0.5 years) and adolescents (mean age ± SD: 12.5 ± 0.7 years) prior to administering the test in the present study (Morawietz et al., 2024). Results indicated that the PFT would be suitable for this age group (ICC = 0.78), however, the test might not be sensitive enough to capture changes in the present age group or our sample might have been too small and not representative. Other spatial ability tests might have been more suitable to detect changes by our intervention. Another possible reason for the lack of improvement on the PFT is that spatial visualization might not have (sufficiently) been trained and affected by our intervention. Considering the categorization by Linn and Petersen (1985), spatial visualization involves complex processing of spatial information which might be executed in multiple steps or on multiple levels. According to the more recent categorization by Newcombe and Shipley (2015) the PFT can be assigned to the intrinsic-dynamic spatial skills, which involves the mental transformation and modification of the spatial information of objects (e.g., folding, rotation, or deformation). Even though contents like juggling and different ball coordination exercises should have addressed these skills, their frequency throughout the intervention period might have been too low. The small improvement from pre- to posttest in both groups can be attributed to learning and memory effects, due to repeated exposure to the same test stimuli. These learning effects have frequently been discussed in the literature before (e.g., Scharfen et al., 2018; Fehringer, 2023). Similar findings to ours have also been reported by Bakker (2008) who investigated the effect of the Tridio^® learning material (i.e., exercises with cubes and mosaics to enhance spatial abilities) in 5th-graders and did not find significant improvements in the INT compared to the CON.

Our findings show that already a targeted intervention on a rather small scale (i.e., 6 weeks of training, twice per week for 45 min/session) can significantly facilitate spatial abilities in healthy children. Other research found that even shorter intervention periods (e.g., 2 weeks or 5 weeks of training) resulted in improvements of specific spatial skills (Blüchel et al., 2013; Pietsch et al., 2017). While schools have a strict curriculum for PE-classes like for any other subject, it would be worthwhile to integrate such small-scale interventions within the everyday school life or PE. The present intervention does not only affect spatial skills which in turn might impact on overall academic achievement but also addresses fundamental movement skills like the different aspects of motor coordination. As discussed earlier, these and other motor skills are increasingly declining in today's children and adolescents (Bös and Ulmer, 2003; Tomkinson et al., 2019; Masanovic et al., 2020). The diverse structure of the intervention program, the focus on participation rather than performance and the provision of individual variations in the level of difficulty for all tasks help to motivate students and allow the experience of success also for students that are less keen on sports. This is also backed by our results with attendance rates of 95.7 and 94.6% in the INT and CON, respectively. Moreover, such positive experiences might help in providing an access to a sports and movement culture in their leisure time for students that did not have this possibility beforehand. Therefore, integrating a coordinative motor training within the school context can contribute to tackle several current academic and health problems at once.

While the overall outcomes of this study are very promising, several limitations need to be taken into account. Even though our sample size was large enough to generate meaningful outcomes, the overall sample was rather small and there is a need for larger studies to provide a stronger base of evidence. Moreover, only healthy children of a specific age were investigated in the present study. Our results are therefore neither transferable to other age groups nor to participants with (cognitive) impairments. Future research should replicate the present study with other age groups to confirm our findings, to work out discrepancies and to facilitate knowledge about developmental steps with regard to spatial abilities. The same accounts for the spatial ability measures used. Findings do neither account for any other measure of spatial ability nor for any adaptation of the tests used not even if they might be assigned to the same category of spatial tests. As discussed earlier, there is disagreement about the constructs of spatial abilities in the research community with two commonly used and in parts overlapping categorizations being those of Linn and Petersen (1985; i.e., spatial perception, mental rotation, and spatial visualization) and Newcombe and Shipley (2015; i.e., intrinsic-static, intrinsic-dynamic, extrinsic-static, and extrinsic-dynamic; Uttal et al., 2013, 2024). Generally speaking, reliable and valid spatial ability tests appear to be lacking, many measures and/or answer keys are difficult to access or expensive and there are inconsistencies regarding the question which test is suitable to measure which spatial construct (Uttal et al., 2024). While some tests have names that sound similar, they actually measure very different constructs. On the other hand, there are tests that allegedly measure different constructs when in fact they evaluate the same spatial skills (Hegarty and Waller, 2005; Uttal et al., 2024). Even though extensively researched, the possibility exists, that the tests used in the present research do in fact measure different spatial constructs than we intended to, particularly as we used them in a different age group than most tests were developed for. Also, other spatial ability measures might have been more suitable and informative to evaluate our intervention. This might particularly be true for the NCR. Even if improvements from pre- to posttest were observed in the INT, the NCR might not be specific enough to give an impression of real-world orientation. For one, it is no real measure of large-scale spatial abilities, as the whole test set-up can be viewed from a single vantage point and does not require movement (Heil, 2020 based on Weatherford, 1982). Therefore, it does not provide sufficient spatial orientation and no spatial navigation information, which are essential for a complete picture of spatial abilities. Further, running speed and reaction time play essential roles for mastery of this task (Dirksen et al., 2015). Participant's motivation and state of mind on the testing day might additionally affect outcomes. Future research might make use of different, appropriate measures of spatial ability that can be used across age groups to evaluate whether theses capture intervention effects even more precisely. To date, virtual reality (VR) might be the most suitable approach to evaluate large-scale spatial abilities in a comparable, standardizable, and reproducible way. This however requires, that researchers have open access to testing material and measures and that they can access VR equipment (Uttal et al., 2024).

Even though no gender-specific differences were found in the present study, this factor should still be considered in future research. While sex differences in spatial abilities in favor of males have been discussed for many measures in adult populations (Linn and Petersen, 1985; Voyer et al., 1995), research in children is less conclusive. While some studies find sex differences in specific spatial abilities in pre-pubertal children and adolescents, others do not (e.g., Pavlovic, 2009; Neuburger et al., 2011; Blüchel et al., 2013; Jansen et al., 2013; Nazareth et al., 2019; Barhorst-Cates et al., 2020). Participant's age, task characteristics or solving strategies might play a critical role in the emergence of sex differences. It is further known that sex differences can be reduced to some degree by goal-oriented interventions (Feng et al., 2007; Tzuriel and Egozi, 2010). Even though some reasons are suspected (e.g., hormonal differences, exposure to spatial toys and games during childhood, cognitive developmental stages, usage of different strategies, or gender beliefs), it would be of great interest for future research to investigate the reasons for the development of sex difference in more detail (Linn and Petersen, 1985; Newhouse et al., 2007; Tzuriel and Egozi, 2010; Merrill et al., 2016; van der Heyden et al., 2016). This is of special interest with regard to academic achievements and occupational success, particularly in STEM subjects. Moreover, it would be of value to evaluate whether sex difference occur in the present study and whether the coordinative motor training addresses participants of all genders equally. Based on these findings, training programs could be tailored more specifically to facilitate spatial skills in all sexes and contribute to the reduction of sex differences.

Additionally, it can be discussed whether the implemented coordinative motor-training could be designed even more specifically to yield larger effects on spatial skills. The present intervention was built upon previous successful motor intervention studies (e.g., Blüchel et al., 2013; Jansen et al., 2013; Pietsch et al., 2017; Latino et al., 2021) and structured in such a way that the different sessions did not built upon each other and targeted different aspects of motor coordination. It was thus ensured that students could easily participate again at any time, even if they missed a session, which is essential within the school context. Moreover, the individual training sessions were designed with a focus on variety to keep participants motivated throughout the intervention period. Further, training content could be adapted to the participant's individual performance level in terms of complexity and degree of difficulty to forestall too high or too low loads. However, to date, little is known about the ideal training modalities (i.e., frequency, intensity, and duration) and content to maximize impact on spatial abilities. Previous research varies considerably and ranges from 20 min/session, 5x/week for 2 weeks to 40 min/session, 2x/week for 12 weeks (Blüchel et al., 2013; Latino et al., 2021). Future research should therefore consider to determine the most effective components of the present training to design more specific interventions. Beyond that, evaluating quantitative exercise parameters (e.g., heart rate, rating of perceived exertion, number of steps) might help to report exercise intensity more comprehensively in the future. Moreover, different training modalities should be evaluated and compared to find out how much training is required to yield significant effects and which training modalities yield the largest effects.

Building on this, it is further essential to establish how long-lasting intervention effects are. In the present study and most previous studies only the immediate post-intervention effects were evaluated. More research is needed to evaluate whether spatial skills continue to improve with ongoing training, whether a plateau is reached after some time as well as whether and how long intervention effects are maintained after the training is ceased.