The nanoeconomics of concurrent choice behavior
Second-by-second ARCH models can flag when a client's choices have stabilized, sharpening your baseline data.
01Research in Context
What this study did
Taylor et al. (2019) tracked every peck made by pigeons on two keys. Each key paid off on its own timer. The team fed the bird's moment-to-moment choices into ARCH models. These models come from economics, not behavior analysis.
The goal was to see if the math could spot when a bird was still deciding versus when it had settled into a steady preference.
What they found
The ARCH tool fit the data well. It cleanly split each session into a short transition phase and a long steady-state phase. During transition, choices bounced around. During steady-state, they settled.
The model gave one number for volatility in each phase. That number let the researchers compare how stable choices were second-by-second.
How this fits with other research
Calamari et al. (1987) first drew the bird's local response patterns by hand. Taylor's group now captures the same micro-dynamics with a ready-made model.
Martens et al. (2016) showed preschoolers also move through a transition phase before matching teacher attention. The pigeon data and the kid data tell the same story: choice stabilizes after a short shake-up.
McDowell et al. (2018) used an evolutionary computer model to mimic choice. Taylor uses an econometric model for the same aim. Both prove that you can replicate live choice with code, just using different toolboxes.
Why it matters
If you run concurrent-operant preference assessments, you can borrow the ARCH idea. Export response-by-response data from your session, let the model flag the switch from transition to steady-state, then base your conclusions only on the stable slice. This gives cleaner baselines for treatment decisions and needs no extra lab gear—just free R scripts written for finance.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Split your last five preference-assessment sessions into transition and steady-state halves; compare which half gives better predictions for the child's later choices.
02At a glance
03Original abstract
We introduce to behavior analysis a way of analyzing choice behavior that exploits recent developments in nanoeconomics, financial economics, and econometrics. A response return, modeled on an economic return, is the log differenced count of responses allocated to each of two alternatives during a short time window, compared with that in the immediately preceding window. The response return is a new dependent variable which offers a novel and useful way of looking at operant behavior, especially at the molecular level of analysis. The response-return series is a near-instantaneous measure of an organism's dynamic preferences for each of two alternatives. Analyzing such a series requires time-analytic techniques, including Auto-Regressive Conditional Heteroskedastic (ARCH) models. We illustrate these techniques by analyzing choices between combinations of arithmetic and exponential variable-interval schedules with pigeon subjects. All response-return series were well-fitted by one of three ARCH-family models. The fitted models were differentially sensitive to transition versus steady-state data samples. The particular insights that the ARCH analyses offer are improved understanding of the effects of the instantaneous effects of reinforcers and their absence, of how the distribution of reinforcers in time affects choice, and of the differences between choice in transition and at steady state.
Journal of the Experimental Analysis of Behavior, 2019 · doi:10.1002/jeab.508