Molecular (moment‐to‐moment) and molar (aggregate) analyses of behavior

Shimp (2020) built a computer model that treats moment-to-moment responses and whole-session totals as two views of the same data. The model shows how 10-second event counts and end-of-session sums can live on the same graph.

The paper is conceptual. No kids, no rats, just code. It asks: what if we stop fighting about 'molecular versus molar' and simply toggle the zoom lens?

The simulation proved that both grains tell the truth. A scallop seen in averaged data also leaves footprints in 10-second bins. You just have to look at both scales before you pick your intervention target.

No numbers are reported; the finding is the method itself. Plot the two views side-by-side and let the data speak.

Baum (2002) and Marr (2001) argued that molar data give a cleaner story. Shimp agrees they are useful, but adds the molecular view back in, so the papers do not clash—he simply widens the lens.

Hineline (2001) already said 'use both scales.' Shimp turns that call into a concrete recipe: graph 10-second counts next to session totals.

Critchfield et al. (1999) showed, with one child, that time-allocation patterns match reinforcement ratios at both grains. Shimp's model generalizes that single case to any behavior stream.

Next time you run a functional analysis, plot two lines: one counts every 10 seconds, the other shows the whole session sum. If the microscopic line is flat but the macroscopic line trends up, you know rate is increasing because bouts get longer, not faster. That tells you to target bout length, not inter-response time. You get the insight in under five minutes, before you write the treatment plan.