This study evaluated 79 Japanese students: 30 high-functioning children with ASD and 49 children with TD. Participants from two discontinuous age groups were included: preschool children age 3–6 years with no previous classroom experience and adolescents age 11–15 years who had attended elementary school or junior high school. The study compared the eye gaze behavior of 4 groups: 25 preschool children with TD (7 boys), 12 preschool children with ASD (9 boys), 24 adolescents with TD (11 boys), and 18 adolescents with ASD (11 boys). Exclusion criteria were any past or present psychiatric illness; difficulties in eye movement or visual function; and inability to accomplish the 10-minute experiment described in the Methods section. Written informed consent for all student–participants was obtained from their parents.
Criteria for ASD and TD
High-functioning ASD was diagnosed by specialists in the field of pediatric neurology and/or developmental pediatrics according to the following criteria:
- Autistic disorder or pervasive developmental disorder based on the Diagnostic and Statistical Manual of Mental Disorders, fourth edition, text revision 
- A full scale intelligence quotient (FSIQ) score of ≥70 in the Wechsler Intelligence Scale for Children, fourth edition , for children age >5 years and a developmental quotient (DQ) of ≥70 in the Kyoto Scale of Psychological Development  for children age <5 years
- A score of ≥25.5 in the Childhood Autism Rating Scale (CARS)  or a score above the cutoff value for the relevant age group in the Parent-Interview ASD Rating Scale, text revision (PARS-TR) .
The CARS score indicates the severity of autism and originally considered that scores <30 indicated no autism and scores >30 indicated mild-to-moderate or severe autism . However, Tachimori et al.  recently reported that children diagnosed with Asperger syndrome could be distinguished from those without ASD using cutoff values of 25.5 versus 26.0, respectively. The present study also used this criterion to differentiate children with ASD from children with TD (CARS score 25.5 vs. 26.0, respectively). The Japanese version of the PARS-TR  was administered in a semi-structured interview with a parent or family member of the child. This scale evaluates both the current symptoms and the most pronounced symptoms during infancy, defined by the PARS-TR peak symptoms scale. A significant correlation exists between PARS-TR scores and Autism Diagnostic Interview-Revised (ADI-R) scores, particularly between qualitative abnormalities in the reciprocal social interaction component in the ADI-R score and the social communication component in the PARS-TR score .
The research was approved by the Ethics Committee of Kansai Medical University (No. 1100).
Experiments were conducted in a quiet, well-lit room at Kansai Medical University Medical Center. Participants were seated in front of a 48 × 30 cm monitor for the presentation of visual stimuli, and their chins were placed on a chin rest to minimize head movement. The distance between the monitor and the chin rest was 60 cm. Partitions were placed to ensure that only the monitor was within the participant’s field of vision. On the monitor, two social images—a smiling human face and a classroom scene in a high school setting (Figure 1)—were presented sequentially, once for each image, with no sound. Each stimulus was shown for nine seconds followed by an intertrial interval of one second. The duration of the whole experiment was approximately 10 minutes (Figure 1, row 1). The participants were instructed to freely watch the static visual images on the monitor. The eye gaze position was measured at 250 Hz using an infrared camera attached to the bottom of the monitor (iView X RED, SensoMotoric Instruments, Teltow, Germany). Eye tracking data were analyzed using a customized software program written in MATLAB (MathWorks, Natick, MA, USA).
Of the 79 participants, 5 could not accomplish a 10-minute experiment (4 for a human face; 1 participant for a classroom scene) and thus were excluded from the following analyses.
To compare the eye gaze behavior between children with TD and children with ASD and between preschool children and adolescents, the testing first identified the visual areas that were regions of interest (ROIs) (Figure 1, row 2). The ROIs were set as eyes and mouth for the human face image, and the human face, finger, pointed-at object, and wall for the classroom image. The eye gaze time for each ROI was compared using the Welch t-test between preschool children with ASD and those with TD and between adolescents with ASD and those with TD. Within the 9 seconds of each stimulus presentation, the duration of the gaze in each ROI was measured. Statistical analyses were performed using the statistical software SPSS version 22.0 (IBM Corp., Armonk, NY, USA).