Some variables affecting the superstitious chaining of responses.

Neuringer (1969) tested how to stop superstitious response chains in lab subjects. The team first reinforced long chains of arbitrary responses. Then they tried three ways to prune the chains: a stimulus that meant 'no food,' a 10-second delay after errors, or extra practice trials.

Each subject worked alone in a chamber. The researchers recorded which method broke the chain fastest.

The 'no-food' signal won by a mile. Chains dropped to zero almost as soon as the stimulus appeared.

Delay and extra practice barely helped. Ten seconds of dead time let the old links stay strong.

HOLZ et al. (1963) looked like they disagreed. They used a changeover delay to kill superstitious button pushes and it worked. But the two studies test different things. C et al. blocked accidental reinforcement between two choices. Neuringer (1969) shows how to undo a long chain once it already exists.

Podlesnik et al. (2017) extend the idea. They mixed alternative stimuli with extinction and got even faster drops in responding. Their data say you can boost J's cue effect if you keep the room, therapist, and materials the same while you extinguish.

Cullinan et al. (2001) add a warning. They gave rats free milk during lever training and the lever press became stubborn during later extinction. The lesson: any extra reinforcer, even an unrelated one, can strengthen links you did not want.

When you shape a new skill and the client strings in weird extra moves, slap on an immediate 'not this' signal. A red card, a head shake, or a brief tone beats waiting or repeating trials. Keep the room and people constant while you do it, just like Podlesnik suggests, and the chain will break faster.