Modeling Preference Reversals in Context Effects over Time

Context effects constitute benchmark behavioral phenomena for theories of choice. Previous research using signal-to-respond paradigms suggests that these phenomena develop differentially over time. Importantly, however, such time pressure manipulations correspond to an externally controlled stopping rule, which may produce different behavior compared to an internally controlled stopping rule. The first goal of this work was to examine context effects as a function of internally controlled response time. Across eight data sets, we find a consistent crossover pattern such that the attraction and compromise effects are associated with slower response times and the similarity effect is associated with faster response times. The second goal is to utilize state-of-the-art methodology (Evans et al. in Psychonomic Bulletin & Review, 26, 1–33, 2019; Turner and Sederberg in Psychonomic Bulletin & Review, 21(2), 227–250, 2014) to determine whether four information-accumulation models (MDFT, the LCA, the AAM, and the MLBA) can correctly capture the direction in which preference evolves over time in each context when fit to joint choice and response time data with an internal stopping rule. Whereas the MLBA best fit the means, and MDFT and the AAM partially captured the qualitative patterns, none of the models captured the full spectrum of results. Despite a strong historical trend for empirical work in consumer choice, recent response time modeling has focused on fundamentally different non-consumer tasks. The third goal is to generalize that modeling to the consumer choice domain, providing a bridge to a wide range of prior work. The present work provides new insights into the relationship between choice and response times and sets important constraints for models that seek to account for such behavior. © 2020, Society for Mathematical Psychology.