Rule following as choice: The role of reinforcement rate and rule accuracy on rule‐following behavior
Clients pick rules that pay best, so load payoff on the rule you want followed.
01Research in Context
What this study did
The team asked 24 college students to pick between two written rules. Each rule pointed to a different computer task. One task paid points on a rich schedule. The other paid on a lean schedule. Sometimes the rule lied. The scientists tracked which rule the students followed across rounds.
Sessions ran in a small lab room. Students could switch rules at any time. Points converted to gift cards. The researchers varied both payoff size and rule accuracy. They wanted to see which factor drove rule choice.
What they found
Students gravitated toward rules tied to the richer payoff. When the accurate rule led to fewer points, most students dumped it. Accuracy only mattered if payoffs were equal. The data traced smooth choice curves, just like pigeon data from the 1970s.
Even when a rule lied once, students kept using it if it still paid better. The authors say rule-governed behavior is simply choice governed by reinforcement history.
How this fits with other research
Kelly (1973) first showed that pigeons shift response patterns when reinforcement rates change. Udhnani et al. (2025) now show the same law applies to human rule following. The species differ, but the principle holds.
Anonymous (1995) found that rats matched their foraging responses to reinforcement frequency. The new study mirrors that pattern with rules instead of nose pokes. Together they build a cross-species case: choice follows payoff, whether the response is a lever press or a verbal rule.
Hackenberg (2018) reminds us that token systems need careful backup-reinforcer rates. The current data say the same logic applies to written rules. If the "backup" payoff behind a rule is thin, clients will drop the rule just like they drop tokens.
Why it matters
You can treat rule following like any other choice. Make the desired rule the richer schedule and clients will stick with it. If the natural payoff is weak, supplement it with praise, tokens, or breaks. Check that the rule actually delivers—an accurate but lean rule will still lose. In practice, pair high-probability reinforcers with the rules you want followed, and watch compliance rise without nagging.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Track the payoff rate behind each instruction you give; add immediate tokens or praise if the rate is low.
02At a glance
03Original abstract
Rules can control the listener's behavior, yet few studies have examined variables that quantitatively determine the extent of this control relative to other rules and contingencies. To explore these variables, we employed a novel procedure that required a choice between rules. Participants clicked two buttons on a computer screen to earn points exchangeable for money. During training, participants were exposed to rules from two simulated individuals. Rule compliance was measured using free-operant choice periods. In the test phase, both simulated individuals appeared simultaneously, providing different rules, followed by a free-operant period of extinction to assess participants' preferences. Experiment 1 varied the reinforcement rate associated with each rule provider, showing that participants systematically preferred the rule provider with the highest reinforcement rate. In the control condition without rules, participants' preferences tended toward indifference. Experiment 2 varied rule accuracy. This time, participants' preferences favored the icon correlated with accurate rules. However, preferences were not exclusive to the alternatives instructed by this rule provider and tended to match the reinforcement rate obtained for this rule provider during training. These findings suggest that rule-following behavior is a form of choice governed by the relative distribution of reinforcement available in the listener's environment.
, 2025 · doi:10.1002/jeab.70048