A Validation Study of the Need for Affect Questionnaire–Short Form in Legal Contexts

ABSTRACT

We evaluated the psychometric properties of scores on the Need for Affect–Short Form (NAQ–S) in 3 samples: undergraduate students (Sample I), jury-eligible community members (Sample II), and forensic clinicians (Sample III). Concerning factor structure, the NAQ–S 2-factor structure displayed good fit to the data in Sample I, with mostly acceptable levels of internal consistency for both approach and avoidance scores. Construct validity patterns were observed such that approach scores were most strongly correlated with female gender and trait agreeableness scores, whereas avoidance scores were most strongly correlated to trait agreeableness scores. Criterion validity associations emerged in that approach scores displayed main effects on mock juror judgments in hate crimes, and forensic clinician judgments of violence risk estimation. Finally, avoidance scores displayed moderating effects on recommended sentencing length by hate crime victim type. Implications are discussed for emotion in legal decision making and future research.

Notes

¹ Study 1 participants were drawn from archival data from a large, multisite investigation of perceptions of expert witness credibility. The previous analyses focused on Five-Factor Model personality or NFA total scores only (see Cramer, Parrott, et al., 2014; Gardner, Titcomb, Cramer, Stroud, & Bate, 2013; Stroud et al., 2014). No identical or conceptually overlapping research questions or analyses are reported in these other studies.

² MLE was selected because it allows for better hypothesis and model testing than other approaches (e.g., least-squares estimation; see Myung, 2003, for further details). Because the NAQ–S items feature original categorical responses, we also tested the CFA using unweighted least squares estimation (ULSE). Rhemtulla, Brosseau-Liard, and Savalei (2012) examined differences in approaches to analysis of categorical data via MLE and ULSE under an array of varying conditions. They reported empirical data demonstrating that both approaches are appropriate for CFAs of data with six or seven ordinal categories (which is the case for the NAQ–S). Model fit and factor loading patterns for the NAQ–S in our study were very similar for the ULSE analysis; therefore, we report only the MLE results. Inspection of NFA subscale score assumption properties detected acceptable residual plots kurtosis and skewness values across samples. Participants with missing NFA scores were few (n = 2 in undergraduate and clinician samples), and were therefore deleted listwise. No missing data were detected in the community sample.

³ Wolf, Harrington, Clark, and Miller (2013) noted that low sample size and number of indicators per factor represent potential threats to accurate parameter estimation in CFA. We discuss this limitation for Samples II and III in the discussion section.

⁴ Regarding nonsignificant effects, statistics are provided only for NFA subscales as they were of central interest. Full model results are available on request.