The assessment of autistic children: a selective review of available instruments.

The author read every autism test manual from 1960-1983. He kept the five most used rating scales. He checked each one for reliability, validity, and ease of use.

No kids were tested. The paper is a narrative review of published manuals and studies.

All five scales give steady scores when two people rate the same child. None can clearly separate autism from other delays. None cover every autism symptom.

In plain words: the tests are repeatable but not accurate.

Haring et al. (1988) later showed the CARS can separate autistic teens from non-autistic ones. This extends Parks (1983) by giving the first solid validity numbers for one scale.

Matson et al. (1994) found a CARS cut-off (Social Impairment ≥ 26) that hits 78 % accuracy. This refines the same scale the 1983 paper doubted.

Tassé et al. (2013) gave the Japanese ADI-R strong validity data. This successor study counters the 1983 claim that no scale has good discriminant validity.

Zander et al. (2015) showed pairing ADI-R plus ADOS boosts specificity to 92 %. This offers a practical fix for the 1983 validity gap.

Do not trust a single autism score. Pick tools that later studies have validated, like CARS with the SI ≥ 26 rule or the ADI-R plus ADOS pair. Always pair a parent interview with a direct observation. This guards against the weak discriminant validity first flagged in 1983.