Measurement of nontargeted problem behavior during investigations of resurgence

Three kids with autism joined a three-phase lab study. First, they got candy for pressing a button or for problem behavior. Next, the candy stopped. Finally, the candy came back, but only for a new response. The team watched both the target behavior and any other problem behavior that had never been treated.

Each child wore a wrist counter. The researchers scored every instance of the taught response and any extra hitting, screaming, or flopping that popped up.

When the new response also lost its candy, the old target behavior returned. At the same moment, untreated problem behavior surged. One boy who had never been trained to hit started hitting again.

All three kids showed the same pattern. Resurgence was not tidy; it spilled into other topographies in the same response class.

Fontes et al. (2018) saw the same spill-over in rats. Punishing a new response brought back the old one. Sullivan et al. now show the effect in children with autism.

Diaz‐Salvat et al. (2020) found that giving kids three ways to ask lowers resurgence. Their fix works because it adds response variety, not because it changes the basic resurgence rule Sullivan confirmed.

Ghaemmaghami et al. (2018) avoided resurgence while shaping complex FCRs. They used tiny criterion shifts. Sullivan’s data say that if you skip that care, untreated forms will still pop up later.

You may think you are only putting one behavior on extinction. Sullivan shows the whole response class can come back. Track every topography during relapse probes, not just the one you trained. If you see a sudden rise in any untreated behavior, treat it early or build more mand options before the probe ends.