All of the experiments in this study received written informed consent for children and their parents and were approved by the Human Research Ethics Committee (Approval number #2738) and Animal Experiment Ethics Committee (#080618-1) of Azabu University in accordance with the World Medical Association Declaration of Helsinki.

In total, 106 children (34 boys and 72 girls, 10–12 years old), consisting of 80 students from an elementary school’s upper grades (fourth, fifth, and sixth graders) and 26 primary school students (fourth, fifth, and sixth graders) in Sagamihara City, were gathered by public recruitment. They had no history of neurological or psychological impairment. They heard an overview of the experiments and then decided to volunteer. They gave informed consent and were comfortable in having limited riding experience. Of those, 54 children (15 boys and 39 girls), who had ≤1 horse-riding experience, were the subjects for the horse-riding experiments. They were separated into three subgroups, A, B, and C horses. There were 5 boy and 13 girl riders in each subgroup. The other students (19 boys and 33 girls) were used in the walking (9 boys and 17 girls) and the resting (10 boys and 16 girls) groups.

A Polar^® RS800CX digital HR monitor and a Polar^® Wearlink strap with a transmitter (Polar^® Electro Öy, Kempele, Finland) were attached to the left wrist and the chest, respectively, of each subject to measure the HR and HRV. The sensors (4.5 cm × 4.5 cm × 1.2 cm; MVP-RF8-GC-2000; Microstone, Nagano, Japan) for monitoring accelerations were attached to the backsides of the children. Half of the children in each group, 13 each in the resting and walking groups, and 9 each in the riding groups, were randomly chosen for the attachment of the HR monitor and transmitter. However, some of the children had difficulty attaching those instruments and thus were replaced with the experiment on monitoring accelerations.

We used three different kinds of horses, with a mean age of 20 years (range 18–22) in this study: a half-breed horse (mare) that was 155-cm high and 190-cm wide (horse A); a Kiso (gelding), which is a Japanese traditional horse, that was 141-cm high and 184-cm wide (horse B); and a pony (gelding) that was 135-cm high and 171-cm wide (horse C). The horses were kept in individual stalls (4 m × 4 m) with natural lighting at Azabu University (Kanagawa, Japan) and trained to be ridden by the disabled and children mainly by two of us, who had over 10 years experience with horse training. The horses were also ridden almost every day by Azabu University students. The horses had no history of medical problems in the preceding 6 months, had not received any pharmacological treatment for 3 months prior to the study, and were considered to be clinically healthy by the two veterinarians.

Horseback riding and walking, as the control, was carried out at the 50 m × 50 m outdoor field at Azabu University.

All of the subjects rested on the seats along the field’s fence for 10 min and performed the Go/No-go pre-task (Pre) during the rest (Figure 1). The riding, or the walking, was performed for 10 min with a 5 min rest to perform the Go/No-go post-task (Post). The tasks were applied to all of the subjects, including the resting group, and performed just after the 10 min of riding, walking, or resting (Experiment 1). After the task, the riding and the walking were resumed, along with a 10 min rest (Figure 1). All of the children completed the arithmetic problems after the second 10 min session of riding, walking, or resting (Experiment 2).

The tasks were presented in three colors, red, blue, and yellow, serially on a 15″ computer screen (Sony, Tokyo, Japan) for 200 ms each. The “Go” reaction required the pressing of the space bar on the keyboard when blue or yellow was displayed, while the ‘No-go’ reaction required that no key was pressed for more than 2,000 ms when red was displayed. Incorrect answers were recorded if the “Go” reaction/color was displayed and the space bar was not pressed for more than 2,000 ms, and if the “No-go” reaction/color was displayed and any key was pressed. The colors were presented, in a random order, 10 times each. In addition, the intervals between the color presentations were chosen at random from 2,000, 2,500, or 3,000 ms. In this experiment, the children were seated on a chair placed in front of the 15″ monitor with the keyboard on a desk after the first 10 min of riding, walking, or resting (Figure 1).

Thirty problems of 1-digit + 1-digit addition were randomly chosen and printed on one side of an A4-size paper (210 mm × 297 mm). The answers were written on the printed side of the paper. Children were seated after the second 10 min session of riding, walking, or resting and then given 30 questions to answer (Figure 1). They pressed the space bar of the computer just before answering and then pushed the bar again when they had finished the problems. The computer calculated the duration, in seconds, that the children spent on the arithmetic problems.

The three-dimensional accelerations of the three horses were measured using an accelerometer. The sensors for monitoring accelerations were attached to the backsides of the children, and the data from the sensors were sent to a tablet computer (Acer Iconia TAB-W500S, New Taipei City, Taiwan) wirelessly through Bluetooth (Parani-UD100, Anaheim, CA, USA), over a distance of ~25 m. Nine of 18 riders per horse participated in the measurement of three-dimensional accelerations.

The HR monitor and the strap with a transmitter (Polar^® Electro Öy, Kempele, Finland) were used to measure the inter-beat (RR) intervals. The data were filtered using Kubios HRV software, version 2.0 (The Biosignal Analysis and Medical Imaging Group at the Department of Physics, University of Kuopio, Kuopio, Finland), and the power spectrum for the frequency was calculated.

The measurement of the HRV was started just before the rest. The values were calculated from data captured from 5 min of each session, just before the task or the arithmetic problems [pre-values, HRVs (1-1) and (2-1), Figure 1].

Statistical analyses of the effects of horseback riding on the Go/No-go tasks and on the arithmetic problems, whose performance rates are shown as percentages with non-normal distributions, were performed using the appropriate Kruskal–Wallis, Mann–Whitney, or chi-squared test, and Welch’s test after a one-factor ANOVA. Spearman’s rank correlation coefficient between the Go/No-go tasks and the HRs was also calculated. Statistical analyses of the HRV data were carried out using the Wilcoxon signed-ranks test to compare values obtained during riding, walking, or resting with the pre-values. The significant differences between the accelerations were determined using Welch’s t-test. P values < 0.05 or better were considered significant.

Horseback riding produced better Go/No-go results than walking or resting (Figure 2A). In 25 of 54 children (46.3%), the performance of Go/No-go tasks increased after 10 min of riding, while it improved in 7 of 26 children after 10 min of walking (26.9%). There were significant differences among the horses in the effects of riding on the Go/No-go tasks (Figure 2B). Riders of horses A and C performed better than those of horse B (A vs B, χ²(2) = 30.7, P < 0.01; B vs C, χ²(2) = 16.3, P < 0.01).

In the pre-tasks, 17.9% of all subjects had incorrect answers for a single question and 2.8% of them had two incorrect answers, while in the post-tasks, 16.0% of all subjects had single incorrect answers and none of them had two (or more) incorrect answers. There was no significant difference in the solving of arithmetic problems between the pre- and the post-tasks (Figure 3A). However, riding on horses A and C shortened the testing time significantly (Figure 3B), compared with riding on horse B (A vs B, χ²(1) = 4.7, P < 0.05; B vs C, χ²(1) = 4.5, P < 0.05).

The changes in HRs measured by the HR monitor and transmitter were calculated as shown in delta (⊿). In total, 26 children had increased and 19 children had decreased HRs. Spearman’s rank correlation coefficient between the changes in Go/No-go reactions and those in HRs (R²) was 0.0301, which was not significant (Figure 4). However, when the increasing and decreasing HR groups (Figures 1 and 5) were compared, the riding affected the Go/No-go reactions significantly, but not the arithmetic problems.

Heart rate variability was analyzed by fast Fourier transformation using text files from Polar ProTrainer (version 5, Polar Electro Oy, Kempele, Finland) imported into Kubios (version 2.0, 2008, Biosignal Analysis and Medical Imaging Group, University of Kuopio, Finland, MATLAB). We determined the HF component (0.15–0.4 Hz, HF), which reflects parasympathetic nervous activity, and we also determined the ratio of the LF component (0.04–0.15 Hz, LF) to the HF component (LF/HF) as an appropriate indicator of sympathetic nervous activity.

A significant increase in the HF (normalized unit) of HRV [HRV (1-1)] occurred in riders on horse B, while significant decreases in the HFs of riders were observed when riding horses A (EX. 1) and C (EX. 2) (Table 1). There were no changes in this parameter during resting or walking.

The LF/HF ratios increased significantly when riding horse C in EX. 1 and horse A in EX. 2, while riding horse B, walking or resting did not affect the LF/HF ratios significantly (Table 1).

Three-dimensional accelerations, which had a range of ±2 × g (g = 9.8 m/s²), were measured for the horses. The three-dimensional sensors were attached to the backsides of the children. Measurements were initiated after the calibration of the sensors. A sampling rate of 50 ms was used for recording the accelerations. We collected 3,600 data points over 3 min in each riding experiment and then extracted 1,024 consecutive data points (every 51.2 s). The three-dimensional data were the averages of the maximum values of the acceleration in the X- (forward and backward), Y- (left and right), and Z- (up and down) axes. The calculated data for horses A, B, and C with riders during walking (upper row) and the statistical data of the acceleration on the Z-axis (lower row) are shown in Figure 6.

There were statistically significant differences on the Z-axis among horses A, B, and C (Figure 6). The Z-axis represents the up and down “shaking” movement, and thus riding horse B gave the children a greater up and down movement than the other horses. In contrast, the other axes of acceleration were not altered by horse walking.