Toward a methodology of withdrawal designs for the assessment of response maintenance.

Rusch et al. (1981) wrote a how-to paper. They sketched three new withdrawal designs. The goal was to test if a skill sticks around after you pull out the intervention.

The designs are called sequential, partial, and partial-sequential withdrawal. Each one removes parts of the program in a different order.

This was a concept paper. No kids, no data, no graphs. The team simply gave step-by-step blueprints that other researchers could run.

Tanious (2022) also adds new tools, but for a different design. Tanious gives randomization tricks for changing-criterion studies, while R et al. give withdrawal tricks for maintenance checks.

Killeen (1978) sets rules for when baseline is stable. Rusch et al. (1981) build on that spirit by setting rules for when you can say 'behavior maintains.' Both papers hand the analyst clear decision points.

Moeyaert et al. (2023) sweep 60 single-case meta-analyses. Many of those syntheses could include studies that used the very withdrawal designs R et al. proposed, so the 1981 paper still feeds modern stats work.

Next time you want to prove a behavior will last, pick one of these three withdrawal plans instead of yanking everything at once. You will get cleaner evidence and fewer relapses. The paper is short, free, and gives ready-made phase labels you can drop straight into your next reversal study.