Thinking Clearly About Violent Cognitions

Abstract

The purpose of the current study was to explore whether measures such as the Violence Scale of the Revised Measures of Criminal Attitudes and Associates (MCAA-R-V) and the Criminal Attitudes to Violence Scale (CAVS) assess attitudes toward violence (i.e., evaluation of violence) and whether attitudes and the cognitions assessed by the MCAA-R-V and CAVS are independently associated with violent behavior. Participants (568 undergraduate students) completed the MCAA-R-V and the CAVS, as well as measures of evaluation of violence, evaluation of violent people, identification of self as violent, and past violent behavior. Exploratory factor analyses revealed that the MCAA-R-V and CAVS items formed correlated but distinct factors from the items of the evaluation of violence, evaluation of violent people, and identification of self as violent scales. Regression analyses indicated that evaluation of violence and identification of self as violent correlated with violent behavior independently of the MCAA-R-V and CAVS. Our results suggest that attitudes toward violence may be distinct from other cognitions often referred to as “attitudes” in the criminological literature, and both attitudes and these other cognitions may be relevant for understanding violent behavior.

Keywords

attitudes cognition violence factor analysis

Attitudes can be important determinants of behavior. Many explanatory models and theories of violent behavior hypothesize attitudes toward violence influence whether one engages in violent behavior (e.g., Anderson & Bushman, 2002; Andrews & Bonta, 2010; Bandura, 1973). This is also the case for behavior in general (e.g., Ajzen, 1991, 2001; Ajzen & Fishbein, 1980, 2005). Although the influence of attitudes on behavior is moderated by a number of variables, such as the accessibility and stability of the attitude (Glasman & Albarracín, 2006; Kraus, 1995), the relationship between attitudes and behavior can be strong. Meta-analyses have revealed medium (r = .38, k = 88; Kraus, 1995) to large (r = .52, 95% confidence interval [CI] = [.49, .54], N = 4,598; Glasman & Albarracín, 2006) average correlations between attitudes and subsequent behavior. In sum, theory and research suggest that attitudes are important for understanding behavior.

Attitudes toward violence are considered important in theory, research, assessment, and treatment because of their potential influence on violent behavior. Moreover, attitudes are potentially dynamic causal factors, meaning that changing them may provide a means by which to reduce violent behavior (Andrews & Bonta, 2010; Douglas & Skeem, 2005; Kraemer et al., 1997). But what are attitudes? To date, there has been a lack of clarity and precision in the conceptualization of, and terminology for, various cognitive constructs, including attitudes, in theory and research on violence (e.g., Maruna & Mann, 2006; Polaschek, Collie, & Walkey, 2004). Although there is debate in the social-psychological literature about the exact nature and definition of attitudes, there is general agreement that the key feature of attitudes is evaluation (e.g., Conrey & Smith, 2007; Eagly & Chaiken, 2007; Fazio, 2007; Gawronski & Bodenhausen, 2007; Petty, Briñol, & DeMarree, 2007; Schwarz, 2007). More specifically, attitudes are typically defined as evaluations (e.g., positive vs. negative) of psychological objects, such as people, things, or behaviors (e.g., Ajzen, 2001; Eagly & Chaiken, 1993, 2007; Fazio, 2007). In one of the most widely cited definitions, Eagly and Chaiken (1993) defined attitudes as “a psychological tendency that is expressed by evaluating a particular entity with some degree of favor or disfavor” (p. 1). Another widely accepted definition of attitudes is that they are associations in memory between a psychological object and a summary evaluation of that object (Fazio, 1995, 2007; Fazio, Chen, McDonel, & Sherman, 1982). Similarly, Ajzen (1991) has defined attitudes toward behavior as “the degree to which a person has a favorable or unfavorable evaluation or appraisal of the behavior in question” (p. 188). From these definitions, attitudes toward violence are evaluations of violence.

Some researchers have begun to explore attitudes toward violence as defined above (e.g., Archer, 2004; Bluemke & Zumbach, 2012; Eckhardt, Samper, Suhr, & Holtzworth-Munroe, 2012; Polaschek, Bell, Calvert, & Takarangi, 2010; Robertson & Murachver, 2007; Snowden, Gray, Smith, Morris, & MacCulloch, 2004). In these preliminary studies, attitudes toward violence have been assessed in two general ways. One approach has used self-report measures such as semantic differential scales in which participants rate violence on bipolar evaluative scales (e.g., good vs. bad). The use of semantic differential scales is a widespread, long-standing, and validated approach to assessing attitudes in social-psychological research (e.g., Gawronski & Bodenhausen, 2006; Glasman & Albarracín, 2006; Hofmann, Gawronski, Gschwendner, Le, & Schmitt, 2005; Osgood, Suci, & Tannenbaum, 1957; Taylor, 1971). The other approach has used Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998) measures to assess implicit evaluation of violence. In these IAT measures, implicit attitudes are inferred from the relative speed with which one sorts stimuli into differently configured categories representative of the associations between violence and evaluation (e.g., good vs. bad). Self-report attitude measures are thought to primarily reflect more deliberative propositional evaluative processes, whereas implicit attitude measures are thought to reflect more immediate evaluative associations in memory (e.g., Gawronski & Bodenhausen, 2006).

There is some evidence that more positive implicit attitudes toward violence are associated with more violent behavior (Eckhardt et al., 2012; Robertson & Murachver, 2007), higher estimated risk of violent behavior (Polaschek et al., 2010), and higher levels of psychopathy (Snowden et al., 2004). However, self-reported evaluations of violence have not been significantly associated with estimated risk of violent behavior (Polaschek et al., 2010), trait-aggressiveness (Archer, 2004), or psychopathy (Snowden et al., 2004). Although the available evidence does not suggest that self-reported attitudes toward violence are particularly important, these are only three preliminary studies, two of which had no non-violent comparison groups and relatively low statistical power due to small samples. Taken together, the evidence suggests that implicit attitudes toward violence may be associated with violent behavior, but the available research on self-reported attitudes toward violence has been less informative.

Notwithstanding the research noted above, the term attitude is more often used to refer to a wide range of cognitions that appear to extend beyond evaluation of violence in the criminological literature. In theory, research, assessment, and treatment with violent offenders, attitude is often used as a synonym or superordinate label for terms such as excuses, justifications, rationalizations, neutralizations, and moral disengagement. For example, consider the following scale items: “Sometimes a person may have to carry a weapon to protect themselves”; “When your main business is crime, being violent is just part of the job”; “Sometimes you have to threaten people to get them to do the right thing, even if you don’t like doing it”; “I see myself as a violent person”; “People who use violence get respect”; and “Lots of people are out to get you.” These items are drawn from the following self-report scales: Violence Scale of the Measures of Criminal Attitudes and Associates (MCAA-V; Mills, Kroner, & Forth, 2002), Criminal Attitudes to Violence Scale (CAVS; Polaschek et al., 2004), Justifications for Violence (JFV) Scale (Kelty, Hall, & Watt, 2011), Maudsley Violence Questionnaire (Walker, 2005), and Attitudes Towards Violence Scale (ATVS; Funk, Elliott, Urman, Flores, & Mock, 1999). These measures are often interpreted as reflecting attitudes toward violence, but it is not clear that agreement with the items above would indicate or require a positive evaluation of violent behavior. Instead, perhaps some items from measures such as these may reflect violent identity (i.e., extent to which one identifies oneself as a violent person) or evaluation of people who are violent (e.g., Funk et al., 1999; Maruna & Copes, 2004). Although these and other measures include other items that seem more likely to reflect evaluation of violence (e.g., “If somebody insults me or my family I feel better if I beat them up”), they also include items like the examples listed above. As a result, the extent to which these scales measure attitudes toward violence versus other cognitive constructs is unclear.

Of course, we are not arguing here that these scales, and the cognitive constructs they measure, are unimportant for theory and research on violence; we are simply questioning the extent to which they actually reflect attitudes toward violence. Whatever constructs are being assessed by measures like the MCAA-V and CAVS, they generally appear to be relevant to understanding and predicting violence as evidenced by their observed association with past and future violent behavior (e.g., Mills, Kroner, & Hemmati, 2004; Polaschek et al., 2004). However, if these scales are measuring constructs in addition to or instead of attitudes toward violence, then it would be beneficial for researchers and clinicians to be aware of this. It is possible that different types of cognitions have different causal associations with violent behavior; for example, some constructs may be more important than others, some may facilitate persistence whereas others may facilitate desistance, and some may be causes of violent behavior whereas others may be consequences of violent behavior (e.g., Gannon, Ward, Beech, & Fisher, 2007; Maruna & Copes, 2004). If such distinctions do exist, addressing these cognitions separately may facilitate more informative assessments and more effective interventions.

The purpose of the current study was to explore the extent to which existing self-report measures designed to assess attitudes toward violence actually do assess evaluation of violence. We also explored the practical value of distinctions between these potentially different cognitive constructs. More specifically, we addressed two main research questions:

Research Question 1: Do measures such as the Violence Scale of the Measures of Criminal Attitudes and Associates–Revised (MCAA-R-V) and CAVS assess attitudes toward violence (i.e., evaluation of violence)?

Research Question 2: Are attitudes and the cognitions assessed by the MCAA-R-V/CAVS independently associated with violent behavior?

To address the first question, we conducted exploratory factor analyses (EFAs) of the MCAA-R-V, CAVS, and semantic differential scales assessing evaluation of violence, evaluation of violent people, and identification of self as violent. If the items from the evaluation of violence scale formed a distinct factor from the MCAA-R-V and CAVS, then that would suggest the MCAA-R-V and CAVS may be measuring cognitions other than attitudes toward violence. Conversely, if MCAA-R-V or CAVS items loaded on the same factor as the evaluation of violence, evaluation of violent people, or identification of self as violent scale items, that would provide some insight into what constructs the MCAA-R-V or CAVS items are actually measuring. We were unable to make specific hypotheses because, to the best of our knowledge, no published research has examined the extent to which violent cognition scales like the ones noted above overlap with or are distinct from attitudes. Nevertheless, we expected that at least some of the MCAA-R-V and CAVS items would be distinct from the evaluation of violence items.

To address our second research question, we conducted regression analyses with all the scales noted above as predictors and self-reported violent behavior as the outcome. If the evaluation of violence scale and the MCAA-R-V or CAVS were independently associated with violent behavior, then that would further suggest distinctions between attitudes toward violence and other cognitive constructs regarding violence may be important. We hypothesized that evaluation of violence would be associated with violent behavior independently of the MCAA-R-V and CAVS.

Method

Participants

Participants were 790 students in first- or second-year undergraduate psychology courses at Carleton University. Of the 790 participants, 222 (28.1%) were missing data on at least one variable of interest (MCAA-R-V, 11.1%, n = 88; CAVS, 15.7%, n = 124; Evaluation of Violence semantic differential scale, 10.3%, n = 81; Evaluation of Violent People semantic differential scale, 9.7%, n = 77; Identification of Self as Violent semantic differential scale, 5.9%, n = 47; Violent Behavior Scale [VBS], 9.2%, n = 73), or reported that they could not understand written English (0.3%, n = 2 reported no; 2.9%, n = 23 had missing data). Little’s MCAR test was non-significant and it was concluded that data were missing completely at random, χ²(91) = 93.63, p = .404. As a result, listwise deletion was used for missing data resulting in a final sample size of 568 participants. Listwise deletion would not be expected to bias our findings or render our statistical power inadequate, given that data were missing completely at random and the remaining number of participants with complete data was large (see Allison, 2001). Of the 568 participants, the majority were female (76.20%, n = 426), 23.8% were male (n = 133), and 1.6% did not report gender (n = 9). Participants’ age was measured using categories ranging from 17 or younger to 55 to 59. The median age category was 18 to 19 (57.1%, n = 324), 3.5% of participants were 17 or younger (n = 20), 23.3% were 20 to 21 (n = 132), 6.5% were 22 to 23 (n = 37), 1.9% were 24 to 25 (n = 11), the remaining 7.58% (n = 43) of participants were in the categories ranging from 26 to 59, and 1 participant did not report age. The majority of participants were single (51.6%, n = 293), 39.1% were in a romantic relationship (n = 222), 6.2% were living with a romantic partner (n = 35), 2.8% were married (n = 16), and less than 1% were separated or divorced (n = 2).

Measures

Demographic Questionnaire

Participants were asked questions about their gender, age, marital status, and ability to understand written English.

Semantic Differential Scales

Three sets of 7-point semantic differential scales were used to assess evaluation of violence (seven items), evaluation of violent people (seven items), and identification of self as violent (two items). Scale anchors are presented in Table 2. Total scores were computed by averaging the item scores for each scale and could range from 1 to 7 with higher scores reflecting, respectively, more positive evaluation of violence, more positive evaluation of violent people, and more identification of self as violent. Scores for all three scales showed adequate internal consistency in the current sample (see Table 1).

Table 1.

Descriptives and Intercorrelations.

	1	2	3	4	5	M	SD	α
1. Evaluation of violence						1.67	0.81	.94
2. Evaluation of violent people	.61**					2.09	0.87	.93
3. ID of self as violent	.47**	.42**				2.54	1.15	.84
4. MCAA-R-V	.49**	.34**	.40**			18.34	5.36	.80
5. CAVS	.56**	.38**	.47**	.76**		40.08	13.35	.93
6. VBS	.37**	.23**	.46**	.37**	.43**	5.13	7.44	.75

Note. N = 568. MCAA-R-V = Violence scale of the Measures of Criminal Attitudes and Associates–Revised; CAVS = Criminal Attitudes to Violence Scale; VBS = Violent Behavior Scale; ID = identification.

p < .001.

MCAA-R-V

The MCAA-R-V (Mills & Kroner, 2007) consists of 10 self-report items (see Table 2) rated on a 4-point scale from 1 (disagree) to 4 (agree). Ratings are summed to compute a total score, which can range from 10 to 40. Although there is not yet any published reliability or validity data available for scores on the MCAA-R, scores on the Violence scale of the original MCAA (Mills et al., 2002) have shown adequate internal consistency in both offender and student samples (e.g., Mills & Kroner, 2001; Mills et al., 2002), are associated with violent criminal history (Mills et al., 2002), and are predictive of violent reoffending (Mills et al., 2004). Internal consistency was also adequate with the current sample (see Table 1).

Table 2.

MCAA-R-V EFA Rotated Factor Loadings and Standardized Scores.

	Rotated Factor Loading (Standardized Score)
Item	Factor 1	Factor 2	Factor 3	Factor 4
Evaluation of Violence semantic differential
Violence is very positive	.78 (20.35)	.06 (1.58)	.07 (1.79)	.04 (1.51)
Violence is extremely fun	.91 (29.87)	−.05 (−1.58)	.02 (0.87)	.05 (1.96)
Violence is very pleasant	.98 (35.49)	.00 (−0.00)	−.02 (−0.74)	−.07 (−2.26)
Violence is very right	.85 (22.71)	.05 (1.32)	.04 (1.14)	−.02 (−0.68)
Violence is very good	—	—	—	—
Violence is very enjoyable	.80 (22.37)	.07 (2.35)	.03 (1.09)	.02 (0.77)
Violence is very moral	.75 (20.11)	.15 (3.90)	−.02 (−0.52)	.03 (0.84)
Evaluation of Violent People semantic differential
Violent people are very positive	−.05 (−1.14)	.81 (21.40)	.08 (2.72)	.07 (1.93)
Violent people are extremely fun	.03 (0.59)	.85 (27.95)	.13 (3.16)	−.10 (−2.88)
Violent people are very pleasant	−.00 (−0.16)	.91 (35.34)	.13 (3.26)	−.09 (−2.58)
Violent people are very right	.08 (2.21)	.78 (23.84)	−.06 (−2.48)	.19 (4.54)
Violent people are very good	−.02 (0.46)	.83 (27.04)	−.12 (−3.52)	.20 (5.18)
Violent people are very enjoyable	.08 (2.24)	.79 (25.41)	.05 (1.64)	−.01 (−0.50)
Violent people are very moral	.06 (1.40)	.72 (19.83)	−.11 (−3.28)	.17 (3.90)
Identification of Self as Violent semantic differential
I am very violent	.06 (1.25)	.02 (0.86)	.08 (2.49)	.81 (16.08)
I am very aggressive	−.01 (−0.58)	.02 (0.93)	.06 (2.46)	.88 (17.78)
MCAA-R Violence scale
It’s none of my business if I saw someone being robbed	−.17 (−1.99)	.13 (1.83)	.36 (5.63)	−.01 (−0.18)
It’s understandable to hit someone who insults you	.10 (1.58)	.00 (0.05)	.69 (16.91)	.02 (0.45)
Sometimes a person may have to carry a weapon to protect themselves	.08 (1.08)	.04 (0.61)	.42 (8.48)	.01 (0.20)
Sometimes you have to fight to keep your self-respect	.08 (1.47)	−.01 (−0.22)	.78 (21.77)	−.03 (−0.77)
It is reasonable to expect a fight from someone you cheated	.09 (1.22)	.00 (0.09)	.61 (14.22)	.01 (0.32)
Ignoring a store being robbed is not wrong	−.18 (−1.94)	.22 (2.98)	.35 (5.64)	−.08 (−1.37)
There is nothing wrong in beating up a child molester	.11 (1.56)	−.05 (−1.00)	.42 (8.90)	.10 (1.92)
It’s not wrong to fight to save face	−.04 (−0.97)	.10 (2.45)	.74 (23.13)	.03 (0.85)
Someone who makes you very angry deserves to be hit	−.04 (−0.83)	−.01 (−0.25)	.82 (22.98)	.06 (1.46)
It’s all right to fight someone if they stole from you	.12 (1.99)	−.03 (−0.76)	.78 (22.83)	.04 (1.06)

Note. For the standardized score, the critical z score was 3.40, α = .0003. Bolded values indicate significant loading on the factor (p < .0003). MCAA-R-V = Violence scale of the Measures of Criminal Attitudes and Associates–Revised; EFA = exploratory factor analysis.

CAVS

The CAVS (Polaschek et al., 2004) consists of 20 self-report items (see Table 3) rated on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree). Ratings are summed to compute a total score, which can range from 20 to 100. Scores on the CAVS have shown excellent internal consistency and are associated with violent criminal history (Polaschek et al., 2004). Internal consistency was also excellent with the current sample (see Table 1).

Table 3.

CAVS EFA Rotated Factor Loadings and Standardized Scores.

	Rotated Factor Loading (Standardized Score)
Item	Factor 1	Factor 2	Factor 3	Factor 4
Evaluation of Violence semantic differential
Violence is very positive	.80 (20.54)	.06 (1.62)	.03 (0.96)	.04 (1.63)
Violence is extremely fun	.89 (28.49)	−.03 (−1.19)	.04 (1.49)	.05 (1.92)
Violence is very pleasant	.95 (31.90)	.00 (0.06)	−.01 (−0.28)	−.05 (−1.56)
Violence is very right	.86 (22.13)	.05 (1.30)	.02 (0.60)	−.01 (−0.43)
Violence is very good	—	—	—	—
Violence is very enjoyable	.78 (21.00)	.08 (2.89)	.04 (1.53)	.02 (0.60)
Violence is very moral	.76 (19.75)	.16 (4.40)	−.03 (−1.11)	.01 (0.42)
Evaluation of Violent People semantic differential
Violent people are very positive	−.05 (−1.07)	.82 (23.31)	.07 (2.20)	.06 (1.65)
Violent people are extremely fun	.05 (0.85)	.84 (27.78)	.08 (1.78)	−.08 (−2.45)
Violent people are very pleasant	.01 (0.29)	.90 (41.05)	.08 (2.26)	−.07 (−2.24)
Violent people are very right	.07 (1.88)	0.82 (21.98)	−.02 (−0.88)	.14 (3.36)
Violent people are very good	−.03 (−0.76)	.87 (23.75)	−.07 (−2.35)	.16 (3.60)
Violent people are very enjoyable	.08 (2.04)	.79 (27.66)	.05 (1.64)	−.02 (−0.77)
Violent people are very moral	.05 (1.23)	.76 (20.11)	−.07 (−2.01)	.12 (2.75)
Identification of Self as Violent semantic differential
I am very violent	.06 (1.46)	.12 (1.12)	.08 (2.12)	.80 (20.62)
I am very aggressive to very gentle	.04 (0.87)	.14 (1.33)	.04 (1.76)	.78 (17.54)
CAVS
If somebody insults me or my family I feel better if I beat them up	.07 (1.30)	−.05 (−1.30)	.64 (15.25)	.20 (5.28)
Lots of people are out to get you so you have to be violent	.06 (0.98)	−.10 (−2.15)	.72 (19.79)	.03 (0.91)
When I get violent, what I want most is to teach the other person a lesson	−.02 (−0.40)	.00 (−0.01)	.66 (16.80)	.17 (4.01)
Men should be allowed to sort their differences out by fighting	.30 (4.72)	−.13 (−2.56)	.55 (12.38)	.09 (2.31)
If somebody puts me down, I feel like I have to fight them to get back my pride	−.05 (−1.01)	−.08 (−1.92)	.82 (22.55)	.10 (2.40)
The best thing about being violent is that it gets my anger out of my system	−.00 (−0.02)	.09 (1.98)	.62 (16.11)	.14 (3.41)
Fighting between men is normal	.11 (1.69)	.03 (0.64)	.54 (11.95)	.03 (0.75)
After a fight I feel happy if I won and depressed if I lost	−.04 (−0.57)	.04 (0.83)	.68 (16.50)	.08 (1.64)
Some people have to be treated roughly because they lack feelings that can be hurt	−.11 (−1.96)	.01 (0.36)	.79 (21.31)	.06 (1.53)
My loyalty to my friends or gang is more important than avoiding violence	.08 (1.09)	−.10 (−1.83)	.60 (13.64)	.01 (0.17)
I am more likely to be violent when another person shows me up in public	.04 (0.57)	.01 (0.21)	.69 (17.78)	.04 (0.90)
The best lesson a man can teach his son is how to fight	−.02 (−0.32)	.02 (0.36)	.67 (14.77)	.04 (1.11)
It is important to fight when your gang’s honor is threatened	.06 (1.21)	−.02 (−0.51)	.79 (21.66)	−.11 (−2.68)
I believe that you have to use violence to get through to some people	.01 (0.13)	.08 (2.16)	.85 (29.80)	−.10 (−2.61)
The best thing about being violent is that it makes the other person get into line	−.06 (−1.43)	.11 (2.95)	.90 (34.96)	−.06 (−1.59)
When your main business is crime, being violent is just part of the job	.06 (0.81)	.07 (1.21)	.54 (12.17)	−.12 (−2.51)
It’s necessary to carry a gun or a knife if you live in a rough neighborhood	.05 (0.81)	.07 (1.34)	.64 (16.36)	−.22 (−4.66)
If a person hits you, you have to hit them back	.10 (1.48)	−.05 (−1.07)	.54 (12.84)	.10 (2.52)
If I assault or rob someone, chances are I’ll get away with it	−.02 (−0.26)	.04 (0.89)	.68 (16.00)	−.16 (−3.16)
Violence is an important part of my culture, even if it is against the law	.13 (1.84)	.02 (0.52)	.64 (11.76)	−.07 (−1.44)

Note. For the standardized score, the critical z score was 3.60, α = .0002. Bolded values indicate significant loading on the factor (p < .0002). CAVS = Criminal Attitudes to Violence Scale; EFA = exploratory factor analysis.

VBS

The VBS consists of eight self-report items regarding various violent behaviors and involvement with the criminal justice system for violent behavior since age 16 (e.g., “From when you were 16 years old to today, how many times have you started a physical fight with someone?” “From when you were 16 years old to today, how many times have you been arrested for a violent offence [e.g., assault]?”). Participants rate each item on a 10-point scale ranging from 0 (never) to 9 (9 times or more) and the total score is computed by summing the items. Scores can range from 0 to 72, with higher scores indicating more violent behavior. The VBS was created for the current study by combining selected items from other antisocial behavior scales (e.g., S. L. Brown, 1992; Elliott, Huizinga, & Ageton, 1985; Smith & Thornberry, 1995), modifying some wording and the response scale, and adding items about arrest and conviction. Similar self-report measures of antisocial behavior are generally associated with official criminal justice indicators of antisocial behavior (e.g., Thornberry & Krohn, 2000). Scores on the VBS showed acceptable internal consistency with the current sample (see Table 1).

Statistical Analyses

EFA was used to examine the underlying latent constructs of the violence attitude measures, the violence identification measures, and the violence cognition scale items. The primary goal of EFA is to reveal any latent factors that cause the measured variables (or items) to covary (Costello & Osborne, 2005; Fabrigar, Wegener, MacCallum, & Strahan, 1999; Tabachnick & Fidell, 2007). MPlus version 6.0 (Muthén & Muthén, 2010) was used to conduct the EFAs (for a more detailed discussion of the EFA methodology used, see Hermann, Babchishin, Nunes, Leth-Steensen, & Cortoni, 2012). Factors were extracted from polychoric correlation matrices using the Weighted Least Square (WLSMV) method. Polychoric correlations estimate the relationship between two unobserved continuous latent variables from two observed ordinal variables (e.g., Flora & Curran, 2004; Holgado-Tello, Chacón-Moscoso, Barbero-García, & Vila-Abad, 2010) and are more consistent and robust correlation estimates for ordinal data (e.g., Holgado-Tello et al., 2010). Factors were rotated using an oblique rotation method (Geomin), thus allowing factors to correlate. The number of factors retained was based on four-factor retention methods: (a) Kaiser’s Criterion (i.e., keeping the number of factors with eigenvalues greater than 1), (b) Scree plot, (c) Parallel Analysis, and (d) Velicer’s Minimum Average Partial test (MAP test). Each factor retention method has advantages and disadvantages and using multiple factor retention methods is preferable (Henson & Roberts, 2006). Nevertheless, of the four retention methods, parallel analysis and the MAP test are the least subjective methods because they are based on statistics and, as a result, are given the most weight when making decisions about the number of factors to retain (O’Connor, 2000; Schmitt, 2011).

Once the number of factors to retain has been determined, the overall factor structure fit is assessed using fit indices. The root mean square error of approximation (RMSEA) assesses the lack of fit in a factor structure relative to a perfect factor structure (Tabachnick & Fidell, 2007), and generally should not exceed .06 (Hu & Bentler, 1999; Schmitt, 2011). The comparative fit index (CFI) assesses the factor structure fit relative to a baseline model where there are no relationships between items (T. A. Brown, 2006), generally a CFI of .95 or greater indicates good fit. The standardized root mean square residual (SRMR) is the average difference between the original correlations in the input matrix and correlations predicted by the factor structure (T. A. Brown, 2006), generally values less than .08 indicate acceptable fit (Hu & Bentler, 1999; Schmitt, 2011). Schmitt (2011) noted that currently in the literature there is a debate about the validity of fit indices, particularly the cutoff values, and as a result they should be only used as guidelines.

Items have conventionally been sorted into their respective factors using the magnitude of the factor loadings. Typically, items with factor loadings of .30 to .40 or greater are said to load onto a particular factor (Cudeck & O’Dell, 1994; Schmitt & Sass, 2011). In addition to using this rule-of-thumb, standardized factor loadings can be used to assess whether a factor loading significantly loads onto a particular factor (Cudeck & O’Dell, 1994; Schmitt, 2011; Schmitt & Sass, 2011). A significant factor loading is determined by testing whether the factor loading significantly differs from 0. To determine significance and protect against Type 1 error, a correction procedure for correlated factors is used to compute the appropriate α level (see Cudeck & O’Dell, 1994, and Schmitt, 2011, for an overview). The z score associated with the determined α level is then used as the critical point for determining significance.

Results

Intercorrelations (Pearson product–moment), means, standard deviations, and internal consistencies (Cronbach’s α) for all measures are presented in Table 1. All measures were significantly intercorrelated, with the coefficients ranging from medium to large. Of the three semantic differential scales, the Evaluation of Violence scale correlated most strongly with the MCAA-R-V and the CAVS. The largest correlation was between the MCAA-R-V and the CAVS. The Evaluation of Violence scale, MCAA-R-V, and the CAVS showed similar correlations with self-reported violent behavior on the VBS. Scores for all measures were adequately reliable as indicated by Cronbach’s alpha values. Although the frequency of violent behavior was generally low on the VBS, the majority of participants reported at least some violence; 68.3% of the total sample, 82.7% of the men, and 63.6% of the women had a VBS score equal to or greater than 1.

Do the Scales Form Distinct or Overlapping Factors?

Two EFAs were conducted for the current study. The first EFA examined the underlying structure of the MCAA-R-V and the three semantic differential measures (i.e., Evaluation of Violence scale, Evaluation of Violent People scale, and Identification of Self as Violent scale). The second EFA examined the underlying structure of the CAVS and the three semantic differential scales. Two polychoric correlation matrices were computed, one for each EFA. Many of the polychoric correlations were greater than .30, suggesting that each of the matrices were factorable. The polychoric correlation matrices were also examined for correlations exceeding .90, as this would suggest issues with multicollinearity. Items 4 and 5 of the Evaluation of Violence scale had a correlation of .91, suggesting that multicollinearity was an issue. To address this problem, the analyses were computed three ways: (a) with all of the items present, (b) with Item 4 removed, and (c) with Item 5 removed. For both the MCAA-R-V and the CAVS EFA, the pattern of results did not differ between the three sets of analyses. For both the MCAA-R-V and CAVS EFAs, the results reported are from the dataset excluding Item 5 of the Evaluation of Violence scale.

As previously mentioned, to determine the number of factors to retain, the Scree plot, Kaiser’s criterion, parallel analysis, and the MAP test were used. For the MCAA-R-V EFA, the Scree plot suggested retaining five factors, Kaiser’s criterion suggested retaining five factors, parallel analysis suggested retaining seven factors, and the MAP test suggested retaining three factors. Model fit statistics were then used to determine the number of factors to retain. The final model was based on acceptable fit statistics and parsimony and consisted of four factors. The rotated factor loadings and their associated standardized scores are presented in Table 2. The eigenvalues, proportion of variance explained by each factor (before rotation), and the cumulative proportion of variance explained are presented in Table 4. Table 5 presents the correlations between each factor. Generally, the model demonstrated an acceptable fit with a RMSEA of .08 (90% CI = [.07, .08]), CFI of .97, and SRMR of .04.

Table 4.

Eigenvalues and Proportion of Variance Explained for Retained Factors for MCAA-R-V and CAVS EFAs.

	Eigenvalue	Proportion of Variance (%)	Cumulative Proportion of Variance (%)
MCAA-R-V EFA
Factor 1	11.01	44.05	44.05
Factor 2	3.22	12.87	56.92
Factor 3	1.59	6.35	63.27
Factor 4	1.30	5.18	68.45
CAVS EFA
Factor 1	15.61	44.59	44.59
Factor 2	4.53	12.95	57.54
Factor 3	1.48	4.24	61.78
Factor 4	1.32	3.77	65.55

Note. Oblique rotation was used; as a result, the proportion of variance explained by each factor only applies to the unrotated factor structure solution. MCAA-R-V = Violence scale of the Measures of Criminal Attitudes and Associates–Revised; CAVS = Criminal Attitudes to Violence Scale; EFAs = exploratory factor analyses.

Table 5.

Correlations Between Factors for MCAA-R-V and CAVS EFAs.

	1	2	3
MCAA-R-V EFA
Factor 1	—	—	—
Factor 2	.63*	—	—
Factor 3	.50*	.27*	—
Factor 4	.50*	.39*	.34*
CAVS EFA
Factor 1	—	—	—
Factor 2	.63*	—	—
Factor 3	.60*	.34*	—
Factor 4	.41*	.28*	.38*

Note. MCAA-R-V = Violence scale of the Measures of Criminal Attitudes and Associates–Revised; CAVS = Criminal Attitudes to Violence Scale. EFAs = exploratory factor analyses.

p < .05.

For the CAVS EFA, the Scree plot suggested retaining four factors, Kaiser’s criterion suggested retaining four factors, parallel analysis suggested retaining eight factors, and the MAP test suggested retaining three to four factors. Model fit statistics were then used to determine the number of factors to retain. The final model was based on acceptable fit statistics and parsimony and consisted of four factors. The rotated factor loadings and their associated standardized scores are presented in Table 3. The eigenvalues, proportion of variance explained by each factor (before rotation), and the cumulative proportion of variance explained are presented in Table 4. Table 5 presents the correlations between each factor. Generally, the model demonstrated a good fit with a RMSEA of .06 (90% CI = [.05, .06]), CFI of .97, and SRMR of .03.

Are Evaluation of Violence and the MCAA-R-V/CAVS Independently Associated With Violent Behavior?

To address this question, we conducted two hierarchical multiple regression analyses—one involving the MCAA-R-V and the other involving the CAVS. We entered the MCAA-R-V or the CAVS in the first step, the Evaluation of Violence scale in the second step, the Evaluation of Violent People scale in the third step, and the Identification of Self as Violent scale in the fourth step. As shown in Table 6, the MCAA-R-V, Evaluation of Violence scale, and the Identification of Self as Violent scale (but not the Evaluation of Violent People scale) were all independently associated with self-reported violent behavior. Moreover, the MCAA-R-V and the Evaluation of Violence scale together showed a significantly stronger association with self-reported violent behavior than the MCAA-R-V alone, as indicated by the significant increase in R² for Step 2 over Step 1. The overall association with violent behavior increased significantly again with the addition of the Identification of Self as Violent scale in Step 4. Exactly the same pattern of results was found for the regression analysis involving the CAVS (Table 7). Although most variables in the multiple regression analyses were positively skewed, the same results as above were found when these variables were logarithmically transformed. The same results were also obtained when we entered gender as a covariate in the first step of the regression analyses.

Table 6.

Multiple Regression Predicting Violent Behavior (VBS) From the MCAA-R-V and Semantic Differential Scales.

Predictor	R	R ²	Adjusted R²	ΔR²	B	B SE	β
Step 1	.37	.14	.14	.14*
MCAA-R-V					0.52*	0.05	.37
Step 2	.43	.19	.18	.04*
MCAA-R-V					0.35*	0.06	.25
Evaluation of violence					2.25*	0.40	.24
Step 3	.43	.19	.18	.00
MCAA-R-V					0.35*	0.06	.25
Evaluation of violence					2.29*	0.48	.25
Evaluation of violent people					−0.06	0.41	−.01
Step 4	.52	.27	.26	.08*
MCAA-R-V					0.25*	0.06	.18
Evaluation of violence					1.57*	0.46	.17
Evaluation of violent people					−0.66	0.40	−.07
ID of self as violent					2.19*	0.28	.34

Note. MCAA-R-V = Violence scale of the Measures of Criminal Attitudes and Associates–Revised; VBS = Violent Behavior Scale; ID = identification.

p < .05.

Table 7.

Multiple Regression Predicting Violent Behavior (VBS) From the CAVS and Semantic Differential Scales.

Predictor	R	R ²	Adjusted R²	ΔR²	B	B SE	β
Step 1	.43	.18	.18	.18*
CAVS					0.24*	0.02	.43
Step 2	.46	.21	.20	.02*
CAVS					0.18*	0.02	.32
Evaluation of violence					1.73*	0.42	.19
Step 3	.46	.21	.20	.00
CAVS					0.18*	0.02	.32
Evaluation of violence					1.78*	0.49	.19
Evaluation of violent people					−0.07	0.40	−.01
Step 4	.52	.28	.27	.07*
CAVS					0.13*	0.02	.23
Evaluation of violence					1.28*	0.47	.14
Evaluation of violent people					−0.62	0.39	−.07
ID of self as violent					2.04*	0.28	.31

Note. CAVS = Criminal Attitudes to Violence Scale; VBS = Violent Behavior Scale; ID = identification.

p < .05.

Discussion

The purpose of the current study was to explore whether measures such as the MCAA-R-V and CAVS assess attitudes toward violence (i.e., evaluation of violence) and whether attitudes and cognitions assessed by the MCAA-R-V and CAVS are independently associated with violent behavior. Two EFAs indicated four factors with the items of the MCAA-R-V (or CAVS), Evaluation of Violence scale, Evaluation of Violent People scale, and the Identification of Self as Violent scale forming correlated but distinct factors consistent with their original scales. These findings suggest that the MCAA-R-V and CAVS may be assessing something other than attitudes toward violence. Two multiple regression analyses indicated that the MCAA-R-V (or the CAVS) and the Evaluation of Violence scale were independently associated with self-reported violent behavior. Furthermore, these measures provided complementary information such that together they were more strongly associated with violent behavior than either one alone. These findings suggest that both attitudes toward violence and the cognitive constructs assessed by the MCAA-R-V and CAVS may be associated with violent behavior.

If measures such as the MCAA-R-V and the CAVS are not actually assessing attitudes toward violence, what are they assessing? They may be assessing any of the constructs referred to in the criminological literature, such as excuses, justifications, rationalizations, neutralizations, or moral disengagement, which may serve to reduce cognitive inconsistency pre-offense, post-offense, or both (Maruna & Copes, 2004; Sykes & Matza, 1957). Another interesting possibility is that some items on measures such as the MCAA-R-V and CAVS may be assessing normative beliefs as conceptualized in Ajzen’s (1991) theory of planned behavior. “Normative beliefs are concerned with the likelihood that important referent individuals or groups approve or disapprove of performing a given behavior” (Ajzen, 1991, p. 195). In Ajzen’s theory, both norms and attitudes are distinct factors that influence each other and behavioral intentions, which in turn influence behavior. Whatever constructs are assessed by the MCAA-R-V and CAVS, our results suggest that they may be distinct from attitudes toward violence.

Although it remains unclear what is being measured by the MCAA-R-V and CAVS, the current and past results suggest that these scales are relevant to understanding and predicting violent behavior as evidenced by their observed association with violent behavior (Mills et al., 2004; Polaschek et al., 2004). An important area for future research would be to explore how attitudes and the construct assessed by scales such as the MCAA-R-V/CAVS may influence each other and violent behavior (e.g., moderation, mediation) and how they may be affected by other factors such as interventions. It would also be worthwhile to explore what cognitive construct(s) scales such as the MCAA-R-V and the CAVS are measuring. This issue could be addressed with the approach we used in the current study. Specifically, responses to items that more unambiguously reflect the relevant construct (e.g., normative beliefs, excuses, etc.) could be entered into a factor analysis with the MCAA-R-V/CAVS items. If these additional items form a distinct factor from the MCAA-R-V/CAVS, that would suggest the MCAA-R-V/CAVS are not measuring that construct. Conversely, if these additional items load onto the same factor as the MCAA-R-V or CAVS items, that would suggest they are measuring that construct.

There are a number of noteworthy limitations in the current study concerning our sample, measures, and design. With regard to our sample, we do not know if our findings with university students would generalize to non-student populations, such as offenders in the community or correctional/forensic settings. The MCAA and CAVS were developed for and typically used with criminal rather than student populations and university students would be expected to differ in many ways from such non-student populations. However, it is not a foregone conclusion that similar results would not be obtained with these populations (e.g., Anderson & Bushman, 1997; Anderson, Lindsay, & Bushman, 1999). For example, the correlation between the Violence scale of the original MCAA and violent behavior is at least as strong among students (r = .27 with self-reported past violent behavior) as among offenders (r = .12 with prior violent convictions; Mills & Kroner, 2001). Furthermore, although the frequency of violent behavior was generally low in the current sample, the majority of our participants reported at least some past violent behavior, and this was the case for women as well as men. More generally, we would argue that research with more accessible populations, such as university students, can make a valuable contribution to the understanding of serious violent behavior (e.g., violence that is more likely to come to the attention of police, courts, and correctional/forensic institutions). The economy and efficiency of conducting research with such populations makes it ideal as a first step in exploring fundamental questions and hypotheses before moving on to the more time-consuming and costly process of attempting to replicate these findings with, for example, a similarly large sample of incarcerated offenders. The fact that we can find distinctions between evaluation of violence and the MCAA-R-V/CAVS with any population is important in its own right and it is the first step in a prudent process of replication and extension. Future research should attempt to replicate our findings with another sample of students, then extend to non-student samples in the community, and ultimately examine samples of violent criminals for whom measures such as the MCAA-R-V/CAVS were developed.

A related concern is that it was not possible to run the analyses separately for the male participants because there were too few in our sample. Thus, it is unknown whether the same distinction between evaluation, identification, and MCAA-R-V/CAVS would be found with an all-male sample. However, it is not unreasonable to expect similar findings. Although women were less violent than men in our sample, the majority of women reported at least some past violent behavior (more than 60%). Furthermore, evaluation of violence, identification of self as violent, and the MCAA-R-V/CAVS were still independently associated with violent behavior even when we entered gender as a covariate in the regression analyses. More generally, the attitudes section of the original MCAA has shown comparable correlations for male and female justice-involved youth with measures of rule breaking (r = .52 and .57, respectively), aggressive behavior (r = .37 and .53, respectively), externalizing behavior (r = .41 and .52, respectively), and risk for recidivism (r = .39 and .53, respectively; Myers, Brown, Greiner, & Skilling, 2012).

The validity of scores on our measures is also an important consideration. Do the semantic differential scales we used in the current study actually reflect attitudes and identification? This is an important question because if they do not, then our study did not actually address our research questions. Although the specific semantic differential scales we used in the current study have not been validated per se, the use of similar semantic differential scales is a widespread, long-standing, and validated approach to assessing key constructs in social-psychological research (e.g., Gawronski & Bodenhausen, 2006; Glasman & Albarracín, 2006; Hofmann et al., 2005; Osgood et al., 1957; Taylor, 1971). For example, Osgood et al. (1957) found that several evaluative semantic differential scales including good–bad, pleasant–unpleasant, and positive–negative loaded together on an evaluative factor. In the current study, these and the other evaluative semantic differential scales we used to assess attitudes toward violence all loaded together on the same factor. Furthermore, we found that the semantic differential scales we used for evaluation of violent people and identification of self as violent all loaded on their respective factors. Taken together, the available evidence allows some confidence regarding the construct validity of scores on our semantic differential scales. Nevertheless, our semantic differential scales are undoubtedly an incomplete assessment of the constructs of interest and future research should examine other self-report measures as well as implicit measures (Bluemke & Zumbach, 2012; Eckhardt et al., 2012; Nunes, Hermann, & Ratcliffe, 2013; Polaschek et al., 2010; Robertson & Murachver, 2007; Snowden et al., 2004).

Another potential concern regarding measurement is our reliance on self-reported violent behavior. Even though participants were informed that their responses were anonymous, they may not have been willing or able to accurately report such information. However, other research clearly indicates that antisocial behavior (e.g., criminal history) can be accurately assessed with self-report measures (e.g., Huizinga & Elliott, 1986; Jones & Miller, 2012; Kroner, Mills, & Morgan, 2007; Thornberry & Krohn, 2000; Woods, Hermann, Nunes, McPhail, & Sewell, 2011).

Finally, because of the cross-sectional and correlational design in the current study, it was not possible to test whether the different cognitive factors we found had different causal associations with violent behavior. Perhaps some of these constructs could be causes of violent behavior, others could be consequences of violent behavior, and still others could be both cause and consequence (e.g., Gannon et al., 2007; Maruna & Copes, 2004). Based on the general social-psychological literature, we would expect attitudes to affect behavior and behavior to affect attitudes (Albarracín, Johnson, & Zanna, 2005). If our findings are replicated, future research should explore questions about the direction of influence using longitudinal designs.

The current study indicates that attitudes toward violence may be distinct from other cognitions often referred to as “attitudes” in the criminological literature. If these results are replicated with other samples and measures, then we may actually know very little about the role of attitudes in violent behavior despite the ubiquity of the term in the criminological literature. It is important to make these distinctions, refine definitions and measurement as needed, and reexamine what we think we know about the role of attitudes in violent behavior. Greater precision and clarity in conceptualization and measurement of violent cognitions will facilitate further advances in understanding of the cognitive underpinnings of violence, which in turn will facilitate further improvements in assessment and treatment aimed at managing and reducing violence.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was facilitated by a grant from the Social Sciences and Humanities Research Council of Canada.

Author Biographies

Kevin L. Nunes is an associate professor in the Department of Psychology and director of the Aggressive Cognitions and Behaviour Research Lab at Carleton University. The main focus of his research is on the role of various cognitions in the initiation and persistence of sexual and non-sexual violence. Other research interests include risk assessment and correctional programming.

Chantal A. Hermann is currently completing her PhD in psychology at Carleton University. She received her BSc (Hons) in forensic science with a minor in psychology from Laurentian University and her master’s in psychology from Carleton University. The main focus of her research is understanding the role of cognition (primarily attitudes) in sexual aggression. Other research interests include online/internet sex offenders and psychologically meaningful risk factors for sex offenders (e.g., emotional congruence with children).

Sacha Maimone is currently completing her master’s in psychology at Carleton University. She received her BA Honors in psychology from Carleton University. The main focus of her research is understanding the role of various cognitions (primarily attitudes) in the initiation and persistence of violent behavior.

Mandie Woods is currently completing her master’s in psychology at Carleton University. She received her BA Honors in psychology from Carleton University. The main focus of her research is understanding the relationship between violent self-concept and violent behavior. Other research interests include examining implicit measurement techniques for use with various cognitions, such as self-concept.

References

Ajzen

(1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-211. doi:10.1016/0749-5978(91)90020-T

Ajzen

(2001). Nature and operation of attitudes. Annual Review of Psychology, 52, 27-58. doi:10.1146/annurev.psych.52.1.27

Ajzen

Fishbein

(1980). Understanding attitudes and predictive social behavior. Englewood Cliffs, NJ: Prentice Hall.

Ajzen

Fishbein

(2005). The influence of attitudes on behavior. In Albarracin

Johnson

B. T.

Zanna

M. P.

(Eds.), The handbook of attitudes (pp. 173-221). Mahwah, NJ: Lawrence Erlbaum.

Albarracín

D. E.

Johnson

B. T.

Zanna

M. P.

(2005). The handbook of attitudes. Mahwah, NJ: Lawrence Erlbaum.

Allison

P. D.

(2001). Missing data (SAGE University Papers Series on Quantitative Applications in the Social Sciences, No. 7-136). Thousand Oaks, CA: Sage.

Anderson

C. A.

Bushman

B. J.

(1997). External validity of “trivial” experiments: The case of laboratory aggression. Review of General Psychology, 1, 19-41.

Anderson

C. A.

Bushman

B. J.

(2002). Human aggression. Annual Review of Psychology, 53, 27-51. doi:10.1146/annurev.psych.53.100901.135231

Anderson

C. A.

Lindsay

J. J.

Bushman

B. J.

(1999). Research in the psychological laboratory: Truth or triviality? Current Directions in Psychological Science, 8, 3-9. doi:10.1111/1467-8721.00002

10.

Andrews

D. A.

Bonta

(2010). The psychology of criminal conduct (5th ed.). Cincinnati, OH: Anderson.

11.

Archer

(2004). Which attitudinal measures predict trait aggression? Personality and Individual Differences, 36, 47-60. doi:10.1016/S0191-8869(03)00050-3

12.

Bandura

(1973). Aggression: A social learning analysis. Englewood Cliffs, NJ: Prentice Hall.

13.

Bluemke

Zumbach

(2012). Assessing aggressiveness via reaction times online. Cyberpsychology: Journal of Psychosocial Research on Cyberspace, 6(1), article 1. doi:10.5817/CP2012-1-5

14.

Brown

S. L.

(1992). Validation of the Psychopathy Checklist: Screening version in a noncriminal population (Unpublished Honour’s thesis). Carleton University, Ottawa, Ontario, Canada.

15.

Brown

T. A.

(2006). Confirmatory factor analysis for applied research. New York, NY: Guilford Press.

16.

Conrey

F. R.

Smith

E. R.

(2007). Attitude representation: Attitudes as patterns in a distributed, connectionist representational system. Social Cognition, 25, 718-735. doi:10.1521/soco.2007.25.5.718

17.

Costello

A. B.

Osborne

J. W.

(2005). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical Assessment, Research & Evaluation, 10(7), 1-9.

18.

Cudeck

O’Dell

L. L.

(1994). Applications of standard error estimates in unrestricted factor analysis: Significance tests for factor loadings and correlations. Psychological Bulletin, 115, 475-487. doi:10.1037/0033-2909.115.3.475

19.

Douglas

K. S.

Skeem

J. L.

(2005). Violence risk assessment: Getting specific about being dynamic. Psychology, Public Policy, and Law, 11, 347-383. doi:10.1037/1076-8971.11.3.347

20.

Eagly

A. H.

Chaiken

(1993). The psychology of attitudes. Fort Worth, TX: Harcourt Brace Jovanovich.

21.

Eagly

A. H.

Chaiken

(2007). The advantages of an inclusive definition of attitude. Social Cognition, 25, 582-602. doi:10.1521/soco.2007.25.5.582

22.

Eckhardt

C. I.

Samper

Suhr

Holtzworth-Munroe

(2012). Implicit attitudes toward violence among male perpetrators of intimate partner violence: A preliminary investigation. Journal of Interpersonal Violence, 27, 471-491. doi:10.1177/0886260511421677

23.

Elliott

D. S.

Huizinga

Ageton

S. S.

(1985). Explaining delinquency and drug use. Beverly Hills, CA: Sage.

24.

Fabrigar

L. R.

Wegener

D. T.

MacCallum

R. C.

Strahan

E. J.

(1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods, 4, 272-299. doi:10.1037/1082-989X.4.3.272

25.

Fazio

R. H.

(1995). Attitudes as object-evaluation associations: Determinants, consequences, and correlates of attitude accessibility. In Petty

R. E.

Krosnick

J. A.

(Eds.), Attitude strength: Antecedents and consequences (pp. 247-282). Hillsdale, NJ: Lawrence Erlbaum.

26.

Fazio

R. H.

(2007). Attitudes as object-evaluation associations of varying strength. Social Cognition, 25, 664-703. doi:10.1521/soco.2007.25.5.603

27.

Fazio

R. H.

Chen

McDonel

E. C.

Sherman

S. J.

(1982). Attitude accessibility, attitude-behavior consistency and the strength of the object-evaluation association. Journal of Experimental Social Psychology, 18, 339-357. doi:10.1016/0022-1031(82)90058-0

28.

Flora

D. B.

Curran

P. J.

(2004). An empirical evaluation of alternative methods of estimation for confirmatory factor analysis with ordinal data. Psychological Methods, 9, 466-491. doi:10.1037/1082-989X.9.4.466

29.

Funk

J. B.

Elliott

Urman

M. L.

Flores

G. T.

Mock

R. M.

(1999). The Attitudes Towards Violence Scale: A measure for adolescents. Journal of Interpersonal Violence, 14, 1123-1136. doi:10.1177/088626099014011001

30.

Gannon

T. A.

Ward

Beech

A. R.

Fisher

(2007). Aggressive offenders’ cognition: Theory, research, and practice. Chichester, UK: John Wiley.

31.

Gawronski

Bodenhausen

G. V.

(2006). Associative and propositional processes in evaluation: An integrative review of implicit and explicit attitude change. Psychological Bulletin, 132(5), 692-731. doi:10.1037/0033-2909.132.5.692.

32.

Gawronski

Bodenhausen

G. V.

(2007). Unraveling the processes underlying evaluation: Attitudes from the perspective of the APE model. Social Cognition, 25, 687-717. doi:10.1521/soco.2007.25.5.687

33.

Glasman

Albarracín

(2006). Forming attitudes that predict future behavior: A meta-analysis of the attitude-behavior relation. Psychological Bulletin, 132, 778-822. doi:10.1037/0033-2909.132.5.778

34.

Greenwald

A. G.

McGhee

D. E.

Schwartz

J. L. K.

(1998). Measuring individual differences in implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology, 74, 1464-1480. doi:10.1037/0022-3514.74.6.1464

35.

Henson

R. K.

Roberts

J. K.

(2006). Use of exploratory factor analysis in published research: Common errors and some comment on improved practice. Educational and Psychological Measurement, 66, 393-416. doi:10.1177/0013164405282485

36.

Hermann

C. A.

Babchishin

K. M.

Nunes

L. K.

Leth-Steensen

Cortoni

(2012). Factor structure of the Bumby RAPE Scale: A two-factor model. Criminal Justice and Behavior, 39, 869-886. doi:10.1177/0093854812436802

37.

Hofmann

Gawronski

Gschwendner

Schmitt

(2005). A meta-analysis on the correlation between the Implicit Association Test and explicit self-report measures. Personality and Social Psychology Bulletin, 31, 1369-1385. doi:10.1177/0146167205275613

38.

Holgado-Tello

F. P.

Chacón-Moscoso

Barbero-García

Vila-Abad

(2010). Polychoric versus Pearson correlations in exploratory and confirmatory factor analysis of ordinal variables. Quality & Quantity, 44, 153-166. doi:10.1007/s11135-008-9190-y

39.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. doi:10.1080/10705519909540118

40.

Huizinga

Elliott

D. S.

(1986). Reassessing the reliability and validity of self-report delinquent measures. Journal of Quantitative Criminology, 2, 293-327. doi:10.1007/BF01064258

41.

Jones

Miller

J. D.

(2012). Psychopathic traits and externalizing behaviors: A comparison of self- and informant reports in the statistical prediction of externalizing behaviors. Psychological Assessment, 24, 255-260. doi:10.1037/a0025264

42.

Kelty

S. F.

Hall

Watt

B. D.

(2011). You have to hit some people! Measurement and criminogenic nature of violent sentiments in Australia. Psychiatry, Psychology and Law, 18, 15-32. doi:10.1080/13218710903566961

43.

Kraemer

H. C.

Kazdin

A. E.

Offord

D. R.

Kessler

R. C.

Jensen

P. S.

Kupler

D. J.

(1997). Coming to terms with the terms of risk. Archives of General Psychiatry, 54, 337-343. doi:10.1001/archpsyc.1997.01830160065009

44.

Kraus

S. J.

(1995). Attitudes and the prediction of behavior: A meta-analysis of the empirical literature. Personality and Social Psychology Bulletin, 21, 58-75. doi:10.1177/0146167295211007

45.

Kroner

D. G.

Mills

J. F.

Morgan

R. D.

(2007). Underreporting of crime-related content and the prediction of criminal recidivism among violent offenders. Psychological Services, 4, 85-95. doi:10.1037/1541-1559.4.2.85

46.

Maruna

Copes

(2004). What have we learned from five decades of neutralization research? Crime and Justice, 32, 221-320.

47.

Maruna

Mann

R. E.

(2006). A fundamental attribution error? Rethinking cognitive distortions. Legal and Criminological Psychology, 11, 155-177. doi:10.1348/135532506X114608

48.

Mills

J. F.

Kroner

D. G.

(2001). Measures of criminal attitudes and associates: User guide. Unpublished manuscript.

49.

Mills

J. F.

Kroner

D. G.

(2007). Measures of criminal attitudes and associates–Revised. Unpublished measure.

50.

Mills

J. F.

Kroner

D. G.

Forth

A. E.

(2002). Measures of Criminal Attitudes and Associates (MCAA): Development, factor structure, reliability, and validity. Assessment, 9, 240-253. doi:10.1177/1073191102009003003

51.

Mills

J. F.

Kroner

D. G.

Hemmati

(2004). The Measures of Criminal Attitudes and Associates (MCAA): The prediction of general and violent recidivism. Criminal Justice and Behavior, 31, 717-733. doi:10.1177/0093854804268755

52.

Muthén

L. K.

Muthén

B. O.

(2010). MPlus user’s guide. Los Angeles, CA: Muthén & Muthén.

53.

Myers

Brown

S. L.

Greiner

Skilling

(2012, June). Examining the reliability and validity of the Measure of Criminal Attitudes and Associates (MCAA) in a sample of male and female justice-involved youth. Poster presented at the 73rd Annual Conference of the Canadian Psychological Association (CPA), Halifax, NS.

54.

Nunes

K. L.

Hermann

C. A.

Ratcliffe

(2013). Implicit and explicit attitudes towards rape are associated with sexual aggression. Journal of Interpersonal Violence, 28, 2657-2675. doi:10.1177/0886260513487995

55.

O’Connor

(2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicer’s MAP test. Behavior Research Methods, Instruments, and Computers, 32, 396-402. doi:10.3758/BF03200807

56.

Osgood

C. E.

Suci

G. J.

Tannenbaum

P. H.

(1957). The measurement of meaning. Urbana: University of Illinois.

57.

Petty

R. E.

Briñol

DeMarree

K. G.

(2007). The Meta-Cognitive Model (MCM) of attitudes: Implications for attitude measurement, change, and strength. Social Cognition, 25, 657-686. doi:10.1521/soco.2007.25.5.657

58.

Polaschek

D. L. L.

Bell

R. K.

Calvert

S. W.

Takarangi

M. K. T.

(2010). Cognitive-behavioural rehabilitation of high-risk violent offenders: Investigating treatment change with explicit and implicit measures of cognition. Applied Cognitive Psychology, 24, 437-449. doi:10.1002/acp.1688

59.

Polaschek

D. L. L.

Collie

R. M.

Walkey

F. H.

(2004). Criminal attitudes to violence: Development and preliminary validation of a scale for male prisoners. Aggressive Behavior, 30, 484-503. doi:10.1002/ab.20081

60.

Robertson

Murachver

(2007). Correlates of partner violence for incarcerated women and men. Journal of Interpersonal Violence, 22(5), 639-655. doi:10.1177/0886260506298835

61.

Schmitt

T. A.

(2011). Current methodological considerations in exploratory and confirmatory factor analysis. Journal of Psychoeducational Assessment, 29, 304-321. doi:10.1177/0734282911406653

62.

Schmitt

T. A.

Sass

D. A.

(2011). Rotation criteria and hypothesis testing for exploratory factor analysis: Implications for factor pattern loadings and interfactor correlations. Educational and Psychological Measurement, 71, 95-113. doi:10.1177/0013164410387348

63.

Schwarz

(2007). Attitude construction: Evaluation in context. Social Cognition, 25, 638-656. doi:10.1521/soco.2007.25.5.638

64.

Smith

Thornberry

T. P.

(1995). The relationship between childhood maltreatment and adolescent involvement in delinquency. Criminology, 33, 451-481. doi:10.1111/j.1745-9125.1995.tb01186.x

65.

Snowden

R. J.

Gray

N. S.

Smith

Morris

MacCulloch

M. J.

(2004). Implicit affective associations to violence in psychopathic murderers. The Journal of Forensic Psychiatry & Psychology, 15, 620-641.

66.

Sykes

G. M.

Matza

(1957). Techniques of neutralization: A theory of delinquency. American Sociological Review, 22, 664-670.

67.

Tabachnick

B. G.

Fidell

L. S.

(2007). Using multivariate statistics (5th ed.). Boston, MA: Pearson Education.

68.

Taylor

H. W.

(1971). Semantic differential factor scores as measures of attitude and perceived attitude. The Journal of Social Psychology, 83, 229-234.

69.

Thornberry

T. P.

Krohn

M. D.

(2000). The self-report method for measuring delinquency and crime. In Duffee

Crutchfield

R. D.

Mastrofski

Mazerolle

McDowall

Ostrom

(Eds.), Criminal justice 2000: Innovations in measurement and analysis (pp. 33-83). Washington, DC: National Institute of Justice.

70.

Walker

J. S.

(2005). The Maudsley Violence Questionnaire: Initial validation and reliability. Personality and Individual Differences, 38, 187-201. doi:10.1016/j.paid.2004.04.001

71.

Woods

Hermann

C. A.

Nunes

K. L.

McPhail

I. V.

Sewell

(2011, June). The validity of self-reported risk among sex offenders. Poster presented at the 2nd North American Correctional and Criminal Justice Psychology Conference, Toronto, Ontario, Canada.