Development and Validation of an Expanded Version of the Three-Factor Levenson Self-Report Psychopathy Scale

abstract

The Levenson Self-Report Psychopathy Scale (LSRP; Levenson, Kiehl, & Fitzpatrick, 1995) is a brief self-report questionnaire frequently used in psychopathy research. Although the scale has many desirable properties such as brevity and being available in the public domain, there are also several psychometric concerns associated with it, including low internal consistency, problematic construct validity, and incomplete conceptual coverage of several components of psychopathy. In 2 studies, we provide evidence that additional items can augment the LSRP to address the aforementioned concerns. In the first study, using a large sample of students and members of the general Australian community (n = 729), we found that an expanded 36-item, 3-factor version of the LSRP was associated with improvements in internal consistency and construct coverage with little degradation in model fit. In the second study, using another Australian community sample (n = 300), we replicated the results of Study 1 and demonstrated improvements in construct validity for the expanded 36-item, 3-factor scale compared to the 19-item, 3-factor scale. Our results indicate that, although slightly longer, the expanded version of the 3-factor LSRP ameliorates many of the concerns associated with its original counterpart.

Notes

¹ To test if the factor loadings between the two samples differed, a factor analysis of the LSRP items in both samples was conducted in which the items were allowed to freely load on their respective factors. Then a factor analysis in which the same factor loadings found in the university sample was applied to the community sample. A χ² test was used to test the significance of the degradation in model fit when constraining the factor loading between samples. No significant differences between the groups on factor loadings was found, χ²(33) = 38.92, p = .22.

² To avoid copyright infringements, the items specific to the scale are not shown in this section. Please see Study 2 and Table 3 where items were rewritten, but capture the same phenomena.

³ We also tested for gender invariance with respect to factor loadings. This analysis revealed a significant difference in factor loadings, χ²(33) = 53.87, p = .01, which was no longer significant when the loadings for one item was freely estimated across men and women, χ²(32) = 43.50, p = .08. The item “For me, what's right is whatever I can get away with” loaded more strongly on the Egocentricity factor for women (λ = .82) than men (λ = .73). The magnitude of the difference was considered to be small enough that it was theoretically insignificant and analyses proceeded with the men's and women's samples combined. Previous research has also found the LSRP three-factor model to be gender invariant (Sellbom, 2011). Invariance testing for ethnicity, however, was not possible due to an insufficient sample size.

⁴ Loadings with specific items could not be shown to avoid violations of copyright. Standardized loadings are reported.

⁵ (a) “When I get frustrated, I often ‘let off steam’ by blowing my top” with “I have been in a lot of shouting matches with other people”; (b) “For me, what's right is whatever I can get away with” with “In today's world I feel justified in doing anything I can get away with to succeed”; (c) “I don't plan very far in advance” with “I like planning things out”; (d) “My main purpose in life is getting as many goodies as I can” with “Making money is my most important goal”; (e) “Before I do anything, I carefully consider the possible consequences” with “I like planning things out”; and (f) “I often do things before thinking them through” with “Before I do anything, I carefully consider the possible consequences.”

⁶ The same criteria as in Study 1 were used to identify noncooperative responses.

⁷ Exclusion rate consistent with Study 1 and previous research with online samples (Downs et al., 2010; Shapiro et al., 2013).

⁸ The sample size was considered too small to attempt to test for measurement invariance between genders or ethnic groups.

⁹ It was noted that the mean scores for the 6-point and 4-point Likert scales were both just below the midpoint of the scales range. Based on the results of Study 2 (see later), it was concluded that there is little difference in use between the 6-point and 4-point Likert scales. This is also consistent with extant literature (e.g., Finn, Ben-Porath, & Tellegen, 2015). However, given the additional scale points associated with the 6-point scale, researchers using this scale in the future might opt for the 6-point scale, as it allows more differentiation of scores across the scale.

¹⁰ We examined whether gender moderated any of these regression-based associations. Regression paths were freely estimated for each gender and then we constrained the female group to the same regression estimates as the male groups and tested for a significant degradation in model fit using a χ² diff-test in Mplus. The results of these analyses revealed only one significant gender difference, χ²(3) = 8.76, p = 033. In the-36 item LSRP scale, the Egocentricity subscale was significantly associated with TriPM Boldness for women (β =.29, p = .002), but not men (β = .07, p = .53). In general, given the large number of analyses run with null findings, we concluded that gender was not a meaningful moderator in this context, although future research should follow up on this one difference to determine the possibility of Type I error versus meaningful difference.