When developing the ATEC, I first gathered as many autism-related questions as possible from many different sources. Dr. Rimland had written several autism surveys during the previous 30 years, such as the E-2 and the E-3 (109 and 216 questions, respectively); I also collected questions from numerous autism checklists and questionnaires in the public domain. Overall, I compiled more than 1,000 questions.
Dr. Rimland and I then spent several weeks discussing the questions. We removed redundant ones as well as those that would not typically be used to evaluate changes in behavior (e.g., pregnancy and birth complications, child’s eye and hair color). We also reworded almost all of them. We then mailed a rough draft to over a dozen researchers and clinicians and asked them to indicate whether each question was clinically relevant and whether the wording was as clear as possible. Based on this input, we felt confident that the questionnaire had face validity, i.e., the questions assessed what they appeared to be evaluating.
Throughout the development of the ATEC, Dr. Rimland would often stress that the final version should fit on a single page; otherwise, he felt that people would not take the time to complete it. After much effort, I was able to format all 77 questions on one sheet.
We then sent out a preliminary draft of the ATEC to parents and professionals who had contacted ARI in the past, and asked them to complete and return the checklist. In the cover letter, we stressed that we were pilot testing the checklist, and we appreciated their taking the time and effort to respond as accurately as possible. We also asked for feedback regarding the wording of the questions and any other thoughts they wanted to share with us.
A total of 1,358 ATECs were returned to ARI over a six-month period. We then analyzed the data to examine whether the questions within each of the four subscales were assessing the same issue. This was accomplished by calculating a Pearson split-half (internal consistency) coefficient in which the responses to the odd-numbered questions were compared to the even-numbered questions within each subscale and between subscales. For this analysis, one would expect relatively high association between questions within each subscale, and relatively low association between the subscales. For example, questions within the Sociability subscale should be highly correlated with one another, and these questions should have a lower correlation with those in the Sensory/Cognitive Awareness subscale. The results were quite impressive, with correlation coefficients ranging from 0.815 to 0.920.
We then publicized the availability of the ATEC through ARI’s e-newsletters, conferences, and websites. Initially, the checklist was available in hardcopy only, and eventually, it was uploaded to ARI’s website.