A method for combining occurrence and nonoccurrence interobserver agreement scores.

Harris et al. (1978) wrote a short math note. They showed a new way to blend two IOA numbers into one.

The new formula gives extra weight to the rarer event. It keeps chance agreement from looking too good.

The paper gives the formula, not new data. No kids, no graphs, no stats.

The authors say, "Use this when behavior is very common or very rare."

Cox et al. (2025) now say the 1978 fix is too narrow. They give eight newer scores like F1 that do the same job and more.

Rolider et al. (2012) ran fake data sets and proved the old worry: high-rate behavior can fool simple IOA. Their tests back up why C et al. bothered.

Jones et al. (1977) had already told writers to show both occurrence and non-occurrence IOA. The 1978 paper just merged those two lines into one number.

If you still use the weighted IOA from 1978, know that better tools exist. Cox et al. (2025) give ready-made Excel code for precision, recall, and F1. Switch when you need clean numbers for grant reviews or court reports. Keep the old formula only if your team has years of files in that format.