Measurement and Analysis of Implicit Cognitions Using a Lexical Decision Task

Research Stage: Analysis

ASC Abstract (2021)

Objective: This project evaluates the utility of a lexical decision task (LDT) for criminological research. LDTs are reaction-time tests where people quickly read sentence stems then indicate whether a subsequent letter string forms a word. LDTs are an alternative to popular yet problematic Implicit Association Tests for measuring implicit cognitions. The current study assesses whether LDT data are useful for: (1) measuring implicit theoretical constructs; (2) indicating survey respondent attention or data quality problems.

Methods: Data were collected from 1325 people through the online platform Amazon Mechanical Turk (MTurk). Participants completed an LDT and a corresponding self-report survey tapping two potentially criminogenic cognitions: Hostile Attributions/Calculative View of Others and Retaliatory Ethic/Toughness Image.

Results: We leverage various sources of information from the LDT data, including response times to construct-congruent and construct-incongruent words and word/nonword errors. We use structural equation models to assess single latent versus dual latent attitude models and measure survey respondent characteristics (speed; error rate). We also assess predictive validity by examining correlations between latent variables and other theoretically relevant self-report items (e.g., deviant behavior; attention checks).

Conclusions: As online-distributed surveys proliferate, LDT methods may serve dual purposes: to improve construct measurement and to screen out low-quality data.

This project is a component of the Coercive Control and Deviance project.

Note: This project will be presented at the 2021 American Society of Criminology conference in Chicago, Illinois.