A runs-test algorithm: contingent reinforcement and response run structures.

Hachiga et al. (2010) worked with rats in a lab.

The rats pressed two levers.

A computer watched the order of presses.

If the rat showed a streak—like left-left-left—the runs-test spotted it.

Food only came when the streakiness dropped.

The team ran two experiments to see if the rule could wipe out the pattern.

The rats stopped making long runs.

Their choices became almost random.

Sequential dependencies vanished while the food rule stayed in place.

When the rule ended, mild patterns crept back.

Mechner (1958) first saw these streaks.

That paper only described them; Hachiga et al. (2010) removed them.

The new study extends the 1958 baseline into a working fix.

Macdonall (1998) showed stay-versus-switch rules shape run length.

Yosuke’s algorithm goes further—it targets the dependency itself, not just the length.

Katz et al. (2003) proved reinforcement rate alters bout size.

The 2010 paper adds a smart gate: it lets food drop only when the bout structure is flat.

You can break hidden streaks in any repeated choice.

Program a quick runs-check into your data sheet.

Reinforce the learner only when streakiness falls below a cutoff.

Try it during coin-reward games, token boards, or response-allocation tasks.

You may see fairer, less biased responding by Monday afternoon.