Modeling modeling.

Parmenter (1999) wrote a short, sharp essay. It tells us to stop asking, “Does my model have the newest bells and whistles?”

Instead, ask, “Does the model explain a lot while staying simple?” Count the benefit, then count the cost.

The paper uses no data. It is a call to think before you add another variable to your spreadsheet.

The author found no new numbers. He found a rule: pick the cheapest model that still tells the story.

A plain ABC log that predicts 80 % of outbursts beats a 50-column monster that predicts 82 %.

Singh et al. (1993) said the same thing six years earlier. They told us to drop long personality tests and watch real behavior. Parmenter (1999) widens the lens to any model you might use.

Christopher et al. (1991) warned that social-validity surveys became empty rituals. Parmenter (1999) agrees: if the survey adds pages but no insight, cut it.

Anonymous (2018) built a tidy FBA ontology years later. The ontology keeps only the fields that matter. That is the benefit-cost idea in code.

Dougherty et al. (1994) ran separate FAs for each behavior shape. Extra sessions cost time, but clearer data save weeks of wrong treatment. Again, benefit beats cost.

Next time you open a data sheet, count the columns. If one extra column only adds a sliver of insight, delete it. Your future self—and every RBT who has to fill the sheet—will thank you. Simple models make fast decisions, and fast decisions help clients sooner.