Reconsidering overlap-based measures for quantitative synthesis of single-subject data: what they tell us and what they don't.

Carter (2013) wrote a think-piece, not an experiment.

He looked at overlap tools like PND and PEM.

These count how many treatment points sit outside the baseline range.

Critics say the tools miss big effects that still overlap.

Mark asked: what are we really measuring?

The paper says overlap indices are not broken.

They simply tell you if the baseline ‘box’ was cracked.

That is experimental control, not how large the change is.

To know size, you need a different ruler.

Gaily et al. (1998) and Reid et al. (1999) already said, “go ahead, average those graphs.”

Mark keeps their green light but adds a speed-limit sign: use overlap only for control.

Cohn et al. (2007) ran numbers and crowned IRD the best overlap index.

Mark nods at their winner yet warns it still can’t speak for magnitude.

Dodd (1984) showed the C statistic can lie when data line up straight.

Mark widens the warning to every overlap tool, giving the old critique a broader coat.

Next time you pool single-case graphs for a parent or an IEP team, report two lines.

Line one: “Control was strong; 90 % of points escaped baseline.”

Line two: “Size is still unknown; look at the actual level change.”

This split keeps your summary honest and avoids the ‘overlap means small’ trap.