Reading focus
Syllabus: Why Does Learned Behavior Persist—and When Does It Change?
Compare the commitments of the choice model with the context interpretation and the earlier momentum account. Keep its general relative-value proposal separate from its particular temporal weighting and response-output assumptions.
A question to consider. Propose a change in alternative reinforcement and specify an outcome that would help distinguish competing explanations, rather than merely demonstrate resurgence again.
Why this reading
The same return of behavior can invite different explanations. This theoretical paper asks whether changes in the relative value of past and recent alternatives can account for resurgence, and makes that argument explicit in a quantitative model.
Before you read
Choice here means allocation of behavior across options, not necessarily conscious deliberation. Value is a modeled consequence of reinforcement history. A temporal weighting rule gives different weights to experiences according to their recency. Relative response probability and absolute response rate are different quantities. This substantial paper may take several sittings; use the model summary to connect each equation to its proposed behavioral role.
Read for
- Separate the general proposal about relative value from the particular assumptions used to build the example model. Which assumptions could be revised without abandoning the framework?
- Track why reinforcement history is needed when current reinforcement is zero. Distinguish modeled values, simulations, and empirical observations.
- Examine the criticisms of momentum and context accounts alongside the model's own adjustable parameters. Identify a prospective comparison that could favor one account over another.
Consider afterwards
- Both available responses stop producing rewards. How could one response gain a larger share of behavior while the total amount of responding declines? What would measuring only percentages conceal?
Editorial context
The authors present a proposed account and illustrative simulations, not a definitive contest establishing one theory for all persistence. Their model distinguishes allocation from overall output and includes assumptions about temporal weighting and arousal. Greer and Shahan (2019) develop the framework's implications for durable behavior change. Such applications remain empirical questions; a model's fit to existing results does not establish the effectiveness of every intervention it suggests.