Developing a model of guilty plea decision-making: Fuzzy-trace theory, gist, and categorical boundaries

Abstract

Objective(s): To date, most research on plea bargaining has used some form of the Shadow of the Trial model (SOT) to frame defendant decisions. In this paper we proposed and tested a new conceptual model of plea decision making, based on Fuzzy-Trace Theory (FTT; Reyna, 2012), for the context in which a non-detained, guilty defendant chooses between a guilty plea or trial, where both the plea and potential trial sentence entail incarceration. Hypotheses: Plea decisions will be affected by (1) meaningful, categorical changes in conviction probability (e.g., low-tomoderate; moderate-to-high), and not by more granular changes within categories, and (2) the presence and magnitude of categorical distinctions between plea offer and potential trial sentence, rather than fine-grained differences between individual offers. Method: We conducted three vignette-based experiments (N1 = 1,701, N2 = 1,098, N3 = 1,232), using MTurk participants. In studies 1 and 2, we manipulated potential trial sentence and conviction probability, and asked participants to indicate either the maximum plea sentence they would accept (Study 1), or whether they would plead guilty to a specific offer (Study 2). In study 3, we manipulated plea discount and potential trial sentence, and measured plea acceptance. Results: Maximum acceptable plea sentences were similar within, and different between, “groupings” of meaningfully similar conviction probabilities (Study 1). Plea rates were similar within, and different between, “groupings” that comprised plea offers of similarly meaningful distance from potential trial sentence (Study 3). Results also provide insight into the plea rates that might be expected under different combinations of the independent variables (Studies 2 & 3). Conclusions: Results support a new conceptual model of plea decision-making that may be better suited to explaining case-level differences in plea outcomes than SOT and suggest that future research extending this model to a wider range of contexts would be fruitful.