Molecular analyses of the principal components of response strength.

The authors built math models to explain three core facts: how often a response happens, how fast it starts, and how likely it is. They call the trio 'response strength' and treat each measure as a random draw from the same hidden process.

The paper is pure theory. No kids, no rats, no sessions. Just equations that show how a single stochastic engine can spit out rate, latency, and probability at the same time.

The model fits the curves. When the engine runs hot, responses come fast, start quickly, and almost always occur. When it runs cold, all three drop together.

The math gives you a ready-made ruler. You can plug in your session data and read off one number that tells you how 'strong' the behavior is right now.

Fox et al. (2001) ran the experiment first. They used real pigeons and PCA to prove that one hidden factor—response strength—links rate, latency, and persistence. The 2002 paper keeps the single-factor idea but swaps the stats: random processes instead of component scores.

Storm (2000) showed that Herrnstein’s k and R0 drift when schedule order changes. The new stochastic frame explains why: the underlying Poisson engine never sits still; its parameters wander within sessions.

Marr (1989) begged for Newton-style equations. This paper answers the call with concrete functions you can paste into Excel or R.

Next time your client’s rate jumps around, don’t average it away. Plot the session minute-by-minute and feed it into the refractory-Poisson sheet. If the latent strength estimate climbs, you know reinforcement is working even when raw counts look messy. One number, three windows, clearer decisions.