Response covariation on self-stimulatory behaviors during sensory extinction procedures.

The team used sensory extinction on children with autism. They picked one self-stim behavior that gave its own sensory payoff, like hand-flapping that feels like wind.

They ran an ABAB design. In the B phases they blocked the sensory payoff. They watched to see if other self-stim behaviors also dropped.

Only behaviors that looked and felt the same to the child went down. If the treated behavior was visual, other visual stims dropped. Auditory stims stayed the same.

No broad spill-over happened. Collateral change stayed inside the same sensory lane.

Davis et al. (1974) saw the opposite with overcorrection. When they punished one self-stim, even topographically different misbehaviors fell. Overcorrection travels wide; sensory extinction stays narrow.

Hangen et al. (2020) extend the picture. They showed that DRO, which includes extinction, can make other behaviors rise instead of fall. Together the papers warn: extinction side effects can go either way, so watch every response class.

Kettering et al. (2018) tested simple extinction unpairing and got mixed results. Their trouble matches why W et al. needed the precise sensory block: removing the payoff only works if you hit the exact sensory channel.

Before you block a self-stim, map its sensory payoff. If you target hand-flapping for visual feedback, do not expect finger-snapping for sound to fade. Collect separate data for each sensory class. If you need broad reduction, pair sensory extinction with a wider plan, or choose overcorrection instead.