Continuous recording and interobserver agreement algorithms reported in the Journal of Applied Behavior Analysis (1995-2005).

Bigby et al. (2009) read every article in the Journal of Applied Behavior Analysis from 1995 through 2005. They looked for papers that used continuous recording and wrote down which IOA formula each one used.

They found three main ways teams check agreement: exact agreement, block-by-block, and time-window. The paper simply counts how often each shows up.

Most studies picked one of the three algorithms, but many forgot to say which one they used. The review gives no winner—it just maps the field.

No numbers on best practice are offered; the paper is a road map, not a verdict.

Bigby et al. (2009) extends its own catalog. The same year the team ran an experiment that tested the three algorithms head-to-head. That sister study shows exact agreement can punish high-rate behavior, while block-by-block and time-window may hide errors when responses clump.

Jennings et al. (2022) covers later JABA studies and finds the same gaps—many papers still skip IOA details. The problem seen in 1995-2005 is alive in 2015-2020.

Gilchrist et al. (2018) and Maharaj et al. (2020) move the job to sensors. Their automated counters reach 80-92% agreement with humans, hinting that someday we may skip manual IOA altogether.

If you use continuous recording, pick an algorithm before you start and state it in your report. Exact agreement is strict but can sink your IOA for rapid behavior. Time-window is the safest first choice for high-rate or long-duration responses. And if you pilot an accelerometer or Kinect, you now have evidence that machine counts can stand in when human observers are costly.