A novel hierarchical later process model: Evaluating latent sources of variation in reaction times of adult daily smokers

Reaction time data from cognitive tasks continue to be a key way to assess decision-making in various contexts to better understand addiction. The goal of this paper is twofold: to introduce a nuanced modeling approach for reaction time data and to demonstrate the novel insights it can provide into the decision processes of nicotine-dependent individuals in different contexts. We focus on the Linear Approach to Threshold with Ergodic Rate (LATER) model, which is a cognitive process model that describes reaction time data in terms of two distinct aspects of cognitive functioning: speed of information accumulation (“accretion”) and threshold amount of information needed prior to execution (“caution”). We introduce a novel hierarchical extension to the LATER model to simultaneously account for differences across persons and experimental conditions, both in the accretion and caution parameters. This approach allows for the inclusion of person-specific predictor variables to explain between-person variation in terms of accretion and caution together with condition-specific predictors to model experimental condition manipulations. To highlight the usefulness of this model, we analyze reaction time data from a study on adult daily cigarette smokers. Participants performed a monetary incentivized Go/No-Go task during two testing sessions, once while following their typical smoking patterns and again following 12 h of verified smoking abstinence. Our main results suggest that regardless of trial type, smokers in a period of abstinence have faster accretion rates, and lower caution thresholds relative to smoking as usual.