Children with ASD who were included in the study were screened based on the following criteria:

Inclusion criteria: (i) all participants had a diagnosis of ASD according to Diagnostic and Statistical Manual of Mental Disorders [Fifth Edition; DSM-5] from a tertiary care hospital; (ii) the age range was from 3 to 8 years old; (iii) The parents/caregivers agreed to participate in the study and signed an experimental informed consent form; (iv) All included participants did not have specialized musical training or special musical experience; (v) Involvement of a parent or caregiver able to consistently complete assessments throughout the study.

Exclusion criteria: (i) Poor adherence to the study protocol; (ii) Any co-morbid psychiatric or neurodevelopmental disorder or condition that interferes with the subject’s ability to participate in the study or produces an adverse reaction to it, including deafness, or intellectual disabilities according to Diagnostic and Statistical Manual of Mental Disorders [Fifth Edition; DSM-5]; (iii) No response to the TOMATIS test, including no listening curve, no biased lateralization sampling points, and a fit score of 0; (iv) Lifetime history of psychotic disorder; (v) Current or former treatment with cannabinoids.

A total of 90 children, who met the eligibility criteria were initially screened to assign 33 in the cross-sectional study and 57 in the longitudinal study.

In the cross-sectional study, the children were then paired according to their sex, chronological age, and severity of ASD (measured with CARS and ABC). Finally, within each pair, children were randomly assigned to one of two groups by random draw: participants were randomly assigned to the experimental (17) and control (16) groups, who received either 2 sessions of TOMATIS training or intervention training with placebo music over 34 days. The final analyses excluded 2 participants who did not undergo a completed TOMATIS training assessment after random assignment and 1 participant who did not have baseline CARS and ABC scores.

In the longitudinal study, participants received at least 6 phases of intervention training over 7.5 months (90 weeks). Thirty-two participants were excluded because of loss or missing data, and a total of 25 participants ultimately completed the trial. We also recorded the responses and underlying behaviors of the longitudinal study participants when they first took the listening test (see details in Supplementary Table S7).

Of all the participants included in the analysis of intervention effects, the distribution of age groups was as follows: participants ranged in age from 3 to 8 years. The majority of participants (83% in the cross-sectional and 76% in the longitudinal study) were male. Participants in both studies completed all scheduled assessments. Demographic characteristics, baseline scores, symptom severity, and concomitant substance use were similar (Table 1), and none of the subjects had any co-morbid psychiatric or neurodevelopmental disorders.

In the cross-sectional study, the age of the experimental group was 6.30 ± 2.74 years and the age of the control group was 7.55 ± 3.48 years, there was no significant difference between the two groups, p = 0.284; there was no significant difference between the two groups in terms of gender, p = 0.624; there was no significant difference between the experimental group and the control group in terms of ABC scores prior to the training, p = 0.746, and no significant difference in terms of CARS scores, p = 0.211. In the longitudinal experiment, another group of 25 children with ASD (76% male, 24% female) were finally included, with a mean age of 5.47 ± 2.15 years.

This study included a cross-sectional experiment as well as a longitudinal experiment. The cross-sectional experiment employed a double-blind randomized controlled study design, ensuring that trainers, parents, and children in both the experimental and control groups remained oblivious to the disparities in intervention methods between the two groups; the intervention was administered by four proficient TOMATIS trainers during each training session.

The cross-sectional study encompassed brief training sessions for children in both groups, spanning 34 days. Prior to the start of the intervention, the researchers held a meeting for the parents of the experimental group, where the treatment process and the purpose of the study were explained to the parents, and the Childhood Autism Rating Scale (CARS) and the Abnormal Behavior Checklist (ABC) were distributed to the children in the experimental group and the control group. The TOMATIS method, as advised by the American Academy of Pediatrics (AAP), involves repeating the 10-day cycle multiple times during the intervention, incorporating 10 days of training in period 1, 7 days of rest, followed by a second period of 10 days of training, and concluding with 7 days of rest after the training, resulting in a total of 2 periods. The researchers created two 10-day training sessions and 7 days of rest for the children in the experimental group. For the duration of the training, music had to be heard 1–2 times a day for 1–2 h, with each session lasting 10 days to guarantee that each child had a minimum of 30 h and a maximum of 40 h of practice. The experimental group was TOMATIS trained, listening to 4 different pieces of music per day. The program of tracks listened to by each participant was personalized according to the participant, with settings for the percentage of sounds of different frequencies, the strength of air conduction, bone conduction, and the processing time delay. The control group listened to the same music as the experimental group but without any processing of parameters such as frequency. Training was not interrupted in both the experimental and control groups.

The control group passively listened to the same repertoire as the experimental group but used a normal Bluetooth headset, which was not affected by the TOMATIS parameter through the Talksup program manager v6.0.2 program, so the trainer did not know whether the parameter being set was for the experimental or control group, and the cell phone could be connected to the TOMATIS programming platform for listening, accompanied by a parent or a trainer to ensure that the music was listened to in its entirety. A parent or trainer will accompany the child during the training to ensure that the child listens to the music in its entirety. The trainer will check the completion of the training and record feedback through online meetings or face-to-face.

The longitudinal study spanned over 7.5 months, with more than 6 sessions, each lasting 10–15 days of training and 2–4 weeks of rest in between. We compared the audiometric core indexes before and after the training and also collected the CARS assessment form at the end of the training. We did multiple regression analyses between the CARS form scores and the audiometric core indexes. During the intervention, parents collaborated to complete the Parent Ten Program Self-Assessment Form and to monitor the initial state of the research, all other training sessions were conducted, resulting in a total of over 6 training sessions being gathered. Parent self-assessment forms were filled out throughout the pre-training initial phase, the intervention phase during training, and the maintenance phase after training.

To ensure consistency and fidelity of the intervention protocol, we trained all trainers in standardized practices prior to the start of the study to ensure consistency in their understanding and implementation of the study protocol. In addition, we implemented regular fidelity checks throughout the intervention, including randomized observation of treatment sessions and assessment of therapist operations to ensure strict adherence to the study protocol.

TOMATIS programming follows the seven steps: initial assessment, parent interview, individualized training program design, training, breaks, wrap-up, and feedback, in which the trainer monitors and supports the entire process and the parents are involved in the entire training process. Individualized programs are designed based on each parent interview, hearing test, and training feedback record. Participants were initially given a comprehensive hearing assessment, including measures of sensitivity and responsiveness to different frequencies of sound.

The TOMATIS training program employs gating as a means to regulate the sound frequency entering the middle and inner ear, complemented by lateralized adjustments in the left and right ears, to engage the trainee’s auditory system. The amount of time spent listening is determined by the participant’s habits and adherence. Every session comprised of a preliminary warm-up and a structured training session, known as the listening session (refer to https://advancedbrain.com/ for further information). The goal of this was to give the listener the opportunity to get used to the music before achieving the best level of sensory stimulation. The listening program was composed of three stages: being open to the music (passive acceptance of the music), taking breaks to integrate it, and expressing oneself (listening to the music while practicing vocally). The initial 10-day program primarily comprises Mozart sonatas and other classical compositions, employing music that is softly progressive, typically abundant in harmonics and high frequencies that stimulate the parasympathetic nerves, and can be adjusted to enhance the distinct frequencies of the sounds using specialized TOMATIS headphones; the subsequent 10-day intervention procedure mirrors the initial one, albeit with the distinction that it incorporates 30 min of sound connection, each accompanied by a microphone, to enhance vocal control through self-listening. The study also excluded children who failed to meet the training criteria. Participants were found to have weak areas of auditory frequency based on scores from psychological tests and clinical interviews, and individual music sessions were organized based on the number of musical pieces or chants filtered and the length of the program. The age, diagnosis, duration, and substance of the listening programs received by the retained participants were relatively alike.

The TOMATIS audiometric study process utilized the Madaule and Belfrage (1984) scoring system to assess the eight primary parameters, yielding individual scores spanning from 0 to 10 and composite scores ranging from 0 to 80. Hearing threshold, curve slope, homogeneity, parallelism, balance, spatiality, selectivity, and dominance were among the 8 main parameters that were indexed.

Furthermore, this research modified the assessment criteria. In this study, only the longest time interval data from the initial treatment was utilized when comparing multiple time intervals. Furthermore, excluding the scale data, the scores were modified to account for directionality to align with the notion that “greater scores signify more significant progress,” and the coding of the data was verified by two researchers.

In contrast to the control group, children with ASD showed notable enhancements in auditory status and behavioral symptoms both before and following TOMATIS training (Table 2). Supplementary Table S6 provides comprehensive information on the pre- and post-intervention data for each participant in the experimental group.

The 2 (experimental, control) × 2 (pre-training, post-training) repeated measures ANOVA on CARS scores revealed a significant difference between pre- and post-training CARS scores, F(1, 28) = 37.95, p < 0.001, ηp2 = 0.575. The between-groups differences were not significant, F(1, 28) = 0.22, p = 0.641. The interaction was significant, F(1, 28) = 43.33, p < 0.001, ηp2 = 0.607. Simple effects analyses showed that there was no significant difference in CARS scores between the groups before training, p = 0.211, and that there was a significant difference in CARS scores between the groups after training, p = 0.047, Cohen’s d = 0.76, which means that the experimental group’s symptoms of autism were significantly more likely to be improved than the control group after training. Improvement. A significant reduction in CARS scores was observed in the experimental group from pre- to post-training, indicating a substantial effect of the intervention (p < 0.001, Cohen’s d = 1.42).

The 2 (experimental, control) × 2 (pre-training, post-training) repeated measures ANOVA of ABC scale scores revealed a significant difference in ABC scores between pre- and post-training, F(1, 28) = 89.54, p < 0.001, ηp2 = 0.762. The between-group effect was not significant, F(1, 28) = 2.48, p = 0.126. the interaction was significant, F(1, 28) = 100.22, p < 0.001, ηp2 = 0.782. Simple effects analysis showed that there was no significant difference in ABC scores between the groups before training, p = 0.746; the difference in ABC scores between the groups after training was significant, p = 0.009, Cohen’s d = 0.99, indicating that there was a significant improvement in autism symptoms in the experimental group after training compared to the control group. There was a significant difference in scores before and after training in the experimental group, p < 0.001, Cohen’s d = 1.69, with a reduction in autism symptoms; there was no significant difference in the control group before and after training, p = 0.701 (Table 2).

Analysis of the TOMATIS listening ability curves revealed that, compared to the control group, participants in the experimental group had better quality of completion at the end than in the first training period after 2 training periods, participants’ cooperation increased, audiometric curve index scores improved, left- and right-ear curve balance improved and became more symmetrical, the slopes of the curves in the frequency bands flattened, spatial localization errors decreased, and left- and right-ear laterality was reduced. The repeated-measures ANOVA revealed significant improvements in scores following mid- and high-frequency training within the experimental group, demonstrating the intervention’s effectiveness, t = −6.403, p < 0.001, Cohen’s d = 1.66; a significant difference in elevation before and after high-frequency training, t = −4.669, p < 0.001, Cohen’s d = 1.21; and a significant difference in elevation before and after mid- and high-frequency training, t = −4.669, p < 0.001, Cohen’s d = 1.21. A significant difference in pre- and post-lift, t = −6.158, p < 0.001, Cohen’s d = 1.59 (Figures 2, 3).

Repeated-measures ANOVA was performed for the percentage improvement of audiometric indices at different frequencies in the experimental group posttest (Figure 4 and Table 2). Pairwise comparisons of different frequencies in the experimental group revealed that improvements in middle and high frequencies, laterality, and hearing thresholds were significantly better than those in low frequencies, p < 0.001; there were no significant differences in the percentages of improvement in the other frequency bands.

There is an inverse relationship between high frequencies and imitation (words and movements), visual response, and somatic usability, indicating enhanced imitation and movement-related behaviors with elevated audiometric indexes, as well as bias versus. There is an inverse relationship between threshold and emotional response, verbal communication, and high activity levels, indicating enhanced emotional and verbal-related behavioral deficits with a greater inclination towards bias. The point at which the audiometric curve reaches its threshold (Table 3).

Paired-sample t-tests showed that comparing the differences in the different frequency domains (Table 3 and Figure 5). There was a significant difference in the pre-and post-training boost in low frequencies, t = −2.463, p < 0.05, Cohen’s d = 0.47; Before and following mid- and high-frequency training, a notable disparity in elevation was observed, with t = −2.572, p < 0.05, Cohen’s d = 0.51; prior to and following high-frequency training, t = −3.006, p < 0.05, Cohen’s d = 0.64; and before and following lateralization and auditory threshold training, t = −2.648, p < 0.05, Cohen’s d = 0.36. The value of Cohen’s d is 0.36, with a p-value of less than 0.05 and a value of 2.648. Over the course of the longitudinal period, there was a consistent upward trend in the total scores of all participants, with children diagnosed with autism displaying notable progress in the areas of coordination, language, socialization, and attention. Additionally, the extraction of parent 10-item self-scores from the odd-numbered longitudinal period was conducted for all participants (Supplementary Table S6 and Table 4).

The negative correlation between HF and interpersonal relationships, total impression, intellectual functioning, verbal communication, imitation (words and actions), adaptation to environmental changes, relationship with inanimate objects, nonverbal communication, visual response, emotional response, and total CARS scores was significant, with improvement in the above behaviors due to elevation of HF metrics; and the negative correlation between LF and proximal sensory response scores was significant, with improvement in proximal sensory response due to elevation of LF metrics (Table 5).