A number of studies have addressed the dimensionality and the measurement invariance of the Levenson Self-Report Psychopathy scale (LSRP; Levenson, Kiehl, & Fitzpatrick, 1995). However, previous investigations exclusively relied on North American samples and mainly on statistical methods that assume data as normally and continuously distributed. The aim of this study was to further investigate dimensionality and differential item functioning for gender in the 26-item LSRP scale in a Brazilian population-based sample using analytical methods adequate for ordered categorical variables. Findings revealed strong evidence for a 2-factor model, similar to the original factor structure of the instrument. Only 3 items presented negligible differential functioning for gender. We address the implications of these results, and stress the need for expanding research on the measurement and structural invariance of the LSRP scale by means of cross-cultural investigations.
Psychopathy includes interpersonal exploitation, profound emotional deficits, and a tendency toward impulsive, reckless behavior (Cleckley, 1976;Cooke & Michie, 2001). The Levenson Self-Report Psychopathy scale (LSRP; Levenson, Kiehl, & Fitzpatrick, 1995) represents a widely used instrument for assessing psychopathic traits in nonforensic (and occasionally forensic) populations (Lilienfeld & Fowler, 2006). The LSRP consists of two positively correlated scales: Primary Psychopathy (LSRP-PP; 16 items, α = .82), and Secondary Psychopathy (LSRP-SP; 10 items, α = .63). Levenson et al. (1995) reported that the LSRP-PP scale was associated with harm avoidance (r = -.17), disinhibition (r = .34), and boredom susceptibility (r = .39), whereas the LSRP-SP scale was associated with academic performance (r = -.23), stress reaction (r = .41), disinhibition (r = .16), and boredom susceptibility (r = .27). Hence, results provided discriminant validity for both scales, adding evidence in favor of the two-factor structure of the instrument.
PSYCHOMETRIC STUDIES OF THE LSRP SCALE
Lynam, Whiteside, and Jones (1999) sought to replicate the original two-factor structure of the LSRP in a large sample of 1,958 undergraduates using confirmatory factor analysis. Although authors reported that "[t]he 2-factor structure of the inventory was strongly replicated" (Lynam et al., 1999, p. 110), actually, results showed modest fit to the data, even after correlating 17 errors, and allowing for Item 26 ("Love is overrated") to cross-load, χ 2 = 1,097, p < .001, Comparative Fit Index (CFI) = .916, root mean square error of approximation (RM-SEA) = .040. Subsequently, Brinkley, Schmitt, Smith, and Newman (2001) found poor fit to the data for the two-factor model with the 17 correlated errors in two samples (total N = 549) of