Sex as a possible source of group inequivalence in Lovaas (1987).

Clarke (1998) wrote a short theory paper. It asked one question: did the famous Lovaas (1987) study have too many boys in one group?

The paper did not run new kids. It looked at the published numbers and warned that sex mix can hide real effects or fake ones.

The author showed that the treatment and control groups had different boy-girl ratios. If you compare each group to the general autism sex ratio, the groups are not equal.

That mismatch could make the treatment look stronger or weaker than it really was.

James et al. (2023) later proved the point with data. They found that parent and expert scores line up differently for boys and girls with ASD. The 1998 warning was theory; the 2023 paper is evidence.

Backer van Ommeren et al. (2017) add another layer. Girls with autism show more reciprocal play than boys. If your groups are sex-skewed, social gains might come from the girls, not the treatment.

Rutherford et al. (2016) show girls get referred later. Late entry can make them look “less severe” at intake, again tilting group balance. Together these studies say: watch sex from screening to final graph.

Before you claim an intervention works, open your baseline table. Count boys and girls in each condition. If the splits diverge from the 4:1 male ratio seen in most autism clinics, weight, match, or analyze separately. A five-minute check can save your study from a fatal confound and your clients from the wrong program.