Calculating contingencies in natural environments: issues in the application of sequential analysis.

The authors wrote a how-to guide, not an experiment. They looked at how behavior analysts use lag-sequential math on everyday video or ABC logs.

They asked: are we mixing up three ideas? Contiguity (events touch in time), contingency (one event truly changes the odds of the next), and dependency (a causal rule is in place).

They showed that most software only tests contiguity. It can scream “reinforcement!” when the real world is only serving up coincidences.

The paper gives a checklist. Split your data into baseline and consequence windows. Run two separate chi-squares. First asks: does the consequence follow the response more than we expect by chance? Second asks: does the response come back more often after that consequence?

Only when both answers are yes do you have evidence of contingency. If only the first is yes, you have contiguity, not control.

Dugan et al. (1995) used the old one-step lag method and said, “Descriptive analysis matched brief FA in 12 of the kids.” Moss et al. (2009) reply, “Match rate could be luck, not function.” The two papers don’t fight; the newer one just tightens the rules.

Ghaemmaghami et al. (2016) showed that kids behave better when reinforcement is truly response-based, not time-based. J et al. give the math tool to spot that difference in natural data.

Layton et al. (2022) proved that context changes contingency control. J et al. add: if you ignore context in your sequential math, you may label a context effect as a reinforcement effect.

Before you write “escape maintained” in the report, run the two-step test. It takes five extra minutes in Excel. You will catch false positives early and spare the kid an unnecessary punishment procedure. Share the checklist with your RBTs so they stop counting every response-consequence pair as proof of function.