On the not so recent invention of interobserver reliability: A commentary on two articles by Birkimer and Brown.

Hartmann et al. (1979) wrote a short commentary. They read two fresh JABA articles about interobserver reliability. The authors wanted to show these "new" stats are actually old tools from other fields.

They dug into history books. They found Fisher’s Exact Test and percent-agreement formulas used decades before JABA began. The paper is a myth-buster, not an experiment.

The team showed that Fisher’s Exact Test and simple agreement formulas existed long before behavior analysts claimed them. The tools came from agriculture and psychology in the early 1900s.

Their main point: give credit where it is due, and stop treating basic stats as ABA inventions.

Jessel et al. (2020) backs the warning. Their big map of functional-analysis papers shows wide swings in how reliability is scored. The lack of standard support proves P et al.’s call for historical honesty still matters.

Greer et al. (2020) and Fisher et al. (2016) seem to clash. One says synthesized FA works fine; the other says it misses functions. Both studies still lean on the same old agreement formulas P et al. traced, showing the field keeps using the grandparent stats without question.

Wolfe et al. (2026) extends the worry into visual analysis. They compare masked versus traditional graph judging and find only modest agreement. The outcome echoes P et al.: reliability methods need routine check-ups, not hero worship.

Next time you write "IOA calculated with 90% agreement," remember the formula is older than JABA. Cite the method, but do not brag like we invented it. Use the saved space to report how you trained observers, how often you checked, and what disagreements looked like. History says the math is solid; our job is to apply it carefully and transparently.