Statistical inference in behavior analysis: Having my cake and eating it?

Bell (1999) wrote a think-piece, not an experiment.

He asked one question: can we keep behavior analysis pure while still using simple stats?

His answer was yes—swap eyeball rules for easy nonparametric tests like randomization checks.

The paper finds no new data.

It argues that quick, assumption-free stats give clearer yes/no answers than staring at graphs.

That clarity, he says, helps both researchers and journal reviewers.

Branch (1999), printed the same year, says the opposite: drop p-values, trust replication instead.

The clash looks huge, but the two papers talk past each other. M wants better tools; N wants a new rulebook.

Taylor et al. (2022) later tested the idea. Objective methods matched visual calls about 84% of the time, showing stats can back up, not replace, your eyes.

Lancioni et al. (2008) gave the reason: different people looking at the same FA graph often disagree, so a numeric backup is wise.

Next time you run a single-case study, add one small nonparametric test to your visual check. Free online calculators make it a five-minute step that shields your decision from “it looks flat to me” critiques and meets today’s publication bar.