The correlation-based law of effect.

The author asked a simple question: what if reinforcement is not about one quick event?

Instead, he said we should count how often a response and a good outcome travel together over many minutes or hours.

The paper is pure theory. It sketches new math units to replace the old "one response, one treat" rule.

The idea: think of reinforcement as a correlation, not a single click.

If good stuff usually follows a response across time, that response grows.

This shift lets us measure whole streams of behavior, not just single pushes or pecks.

Jones et al. (2010) picked up the same torch. They built two-part equations that turn the correlation idea into working numbers for concurrent schedules.

Marr (1989) went further and said "treat response rate like physics treats speed." He gave the field Newton-style formulas that fit the 1973 call for new measures.

Staddon et al. (2002) added gears and springs. Their stochastic model shows how probability, latency, and rate link together inside the molar unit the 1973 paper wanted.

You no longer need to chase every single reinforcer moment. Track how often the behavior stream and reinforcer stream overlap in a session. Use that correlation to judge if your intervention is working. It saves data space and matches what later math models already do.