Measuring Exposure to Media with Antisocial and Prosocial Content: An Extended Version of the Content-based Media Exposure Scale (C-ME2)

ABSTRACT

The present research developed a measure for exposure to both antisocial and prosocial media content by revising and extending a previous Content-based Media Exposure Scale (C-ME). The validity and reliability of the C-ME2 was tested in two independent samples (N = 678), among young adults (Study 1) and adolescents (Study 2). Results of Confirmatory Factor Analyses showed good fit, in both studies, for both antisocial and prosocial dimensions of media content, and for both males and females. Furthermore, the C-ME2 explains unique variance beyond previous measures of violent and general media exposure. Evidence is presented of reliability, discriminant and predictive validity of the C-ME2, measuring both frequency and exposure to specific content of media. The C-ME2 covers all media platforms, is easy to use in all research designs, and allows for standardization and systematic comparisons across studies.

Antisocial and risk behaviors (e.g., swearing, stealing, fighting, binge drinking) are frequently depicted in the media. This type of media is also popular among youth and they are especially susceptible to its negative influence (e.g., Brown & Witherspoon, 2002; Dahl & Hariri, 2005; Nije Bijvank, Konijn, & Bushman, 2012; Parkes, Wight, Hunt, Henderson, & Sargent, 2013; Strasburger, Jordan, & Donnerstein, 2010). Thus, it is crucial to investigate the impact of exposure to antisocial and risk behavior media content on youth, which requires reliable and valid measurement instruments for all media platforms. The Content-based Media Exposure Scale (C-ME; Den Hamer et al., 2014) was developed for this purpose. This article briefly describes the C-ME and explains the need for the C-ME2, which is an expanded version of the C-ME.

C-ME

In a special issue of Communication Methods and Measures (Fishbein & Hornik, 2008), three recommendations were offered for instruments measuring media exposure. First, the instrument should measure both the general frequency of media exposure (e.g., “how often do you watch TV?”), and the content of that exposure (e.g., Annenberg Media Exposure Research Group, 2008; Bleakley et al., 2008; Lee, Hornik, & Hennessy, 2008; Sargent, Worth, Beach, Gerrard, & Heatherton, 2008). Second, the measurement of content should be weighted by the frequency of exposure (e.g., “how often do you watch violent content?”; Annenberg Media Exposure Research Group, 2008; Bleakley et al., 2008). Third, the instrument should not be focused on one particular medium, but should transcend the various media platforms (Bleakley et al., 2008). Given today’s media landscape, media are watched on mobile devices, tablets, computers, TV screens, etcetera. The C-ME meets all three recommendations. The original C-ME contains eight antisocial items (e.g., fighting, vandalism, stealing, substance abuse) and nine neutral filler items.

C-ME2

Prosocial behaviors are also often portrayed in the media (e.g., helping, sharing, cooperation). Exposure to prosocial media content can increase prosocial behaviors, such as cooperating, helping others, and empathy (e.g., Gentile et al., 2009; Happ, Melzer, & Steffgen, 2015; Harrington & O’Connell, 2016), also demonstrated in meta-analytic reviews (Greitemeyer & Mugge, 2014; Mares & Woodard, 2005). Long-term effects of prosocial media use further increases empathy and helpfulness among youth (Prot et al., 2014). Because the C-ME did not assess consumption of media with prosocial content, we replaced the nine neutral filler items of the original scale with ten items that measure prosocial behaviors in media content (e.g., helping, comforting, standing up for others). These ten items were based on previous research on prosocial behavior (e.g., Crick & Grotpeter, 1996; Greitemeyer & Mugge, 2014).

Another deficiency in the C-ME is that it did not measure exposure to relational aggressive content in the media, defined as intentionally harming another person’s social relationships, feelings of acceptance, or inclusion within a group (e.g., Archer & Coyne, 2005; Crick & Grotpeter, 1995; Underwood, 2003). Some examples include gossiping, spreading rumors, and giving someone the “silent treatment”. Previous research has shown that relational aggression is more common in females than in males (Coyne & Archer, 2004; Gentile, Coyne, & Walsh, 2011; Puckett, Aikins, & Cillessen, 2008; Underwood, 2003). Thus, we added 4 relational aggression items to the C-ME2.

This research tests the internal and external reliability and validity of the revised and extended C-ME2, using a sample of young adults (Study 1) and a sample of adolescents (Study 2). Because adolescents are heavier consumers of media than young adults (Brown & Witherspoon, 2002; Rideout, Foehr, & Roberts, 2010; Strasburger et al., 2010), we expected stronger relations for predictive validity in Study 2 than in Study 1.

Study 1

Study 1 had two primary purposes: (1) to test the simple structure of the antisocial and prosocial media content factors using Confirmatory Factor Analysis (CFA), and (2) to test the convergent and discriminant validity of the C-ME2 with personality traits (i.e., trait aggressiveness, empathy) and other media use measures.

Method

Participants and procedure

Participants were 216 undergraduate university students (age range 18–24 years; M_{age =} 21.62, SD_{age =} 1.82, 34.3% male). An additional 68 participants were excluded because they did not finish the survey. Participation was voluntary (without credit). The anonymous survey was completed online. All items were scored using a 5-point scale (1 = never to 5 = very often). A debriefing followed.

Measures

Antisocial media content was measured using 12 items (the original 8 items (den Hamer et al., 2017) plus 4 new items measuring relational aggression), and prosocial media content was measured using 10 items (Table 1; Appendix A).

Table 1. Item loadings and descriptive statistics from items in Study 1.

	Antisocial factor		Prosocial factor
Item	Men	Women	Men	Women
Each of the items start with: “How often do you watch, on the Internet/TV/games/mobile phone/DVD, people who…”
… openly talk about sex?	.71	.72
… drink (a lot of) alcohol?	.69	.76	-	-
… fight?	.66	.58	-	-
… use drugs?	.66	.64	-	-
… make a fool of someone else?	.65	.68	-	-
… are having sex?	.62	.66	-	-
… laugh at another persons’ expense?	.60	.74	-	-
… destroy someone else’s belongings?	.59	.67	-	-
… say negative things about another person behind their back?	.58	.72	-	-
… steal?	.57	.75	-	-
… make someone else trip and fall for fun?	.47	.66	-	-
… shoot at another person?	.45	.52	-	-
… help someone?	-	-	.75	.78
… are willing to do something for someone else?	-	-	.74	.88
… comfort others?	-	-	.72	.87
… are nice to another person?	-	-	.70	.71
… give something to another person to make them happy?	-	-	.68	.80
… stand up for someone?	-	-	.65	.70
… are in love?	-	-	.64	.68
… cooperate with someone else to do good?	-	-	.58	.75
… understand how another person is feeling?	-	-	.56	.74
… put someone in the spotlight in a positive way?	-	-	.48	.64

Trait aggressiveness was measured using the physical aggression subscale of the Aggression Questionnaire (adapted version from Buss & Perry, 1992; in Konijn, Nije Bijvank, & Bushman, 2007; Cronbach α = .83; M = 1.74, SD = 0.60), which consists of 9-items (e.g., “I have threatened people I know”; “I get into fights a little more than the average person”).

Empathy was measured using the 20-item Basic Empathy Scale (BES; Jolliffe & Farrington, 2006; e.g., “When someone is feeling ‘down’ I can usually understand how they feel”; “I get caught up in other people’s feelings easily”). However, a CFA revealed that the reverse-scored items had low factor loadings (<.20). Therefore, the 8 reverse-scored items were removed, leaving 12 items (Cronbach α = .86; M = 3.79, SD = 0.49).

General media use was measured using an often used scale that asked for frequencies of using several media platforms (Rideout et al., 2010). Participants indicated how often they used television, social network sites, video games, YouTube and other websites during a normal weekday, and during a normal weekend separately (1 = less than 1 hour, 2 = between 1–2 hours, 3 = between 2–4 hours, 4 = more than 4 hours). Participants also indicated how many movies they generally watched per week (1 = 0 movies, 2 = less than 2 movies, 3 = 2–5 movies, 4 = more than 5 movies). Weekday and weekend scores were combined to obtain an overall measure of media use. Scores ranged from 1–16. A mean score index of general media use was created by averaging across media platforms (Cronbach α = .63; M = 4.76, SD = 2.08).

Violent media use was measured by asking participants to list their three favorite television programs, films, and video games, to rate how often they consumed it (1 = Almost never to 5 = Almost every day), and to rate how violent it was (1 = Not at all violent to 4 = Very violent). Frequency scores were multiplied by violence ratings. Thus, scores ranged from 1–20 (Cronbach α = .67; M = 5.26, SD = 1.86). This measure has been used in several previous studies (e.g., Anderson & Dill, 2000; Gentile et al., 2011; Ostrov, Gentile, & Crick, 2006; Swing & Anderson, 2014).

Results and discussion

See Table 2 for descriptive statistics for all variables measured in Study 1.

Table 2. Descriptive statistics for the measures included in Study 1 for all participants (N = 216), and for males (n = 74) and females (n = 142) separately. Independent t-tests and standardized mean differences (i.e., Cohen’s d) are given for gender differences.

Variable	Mean (SD)	t	d
Antisocial content	Overall: 2.72 (0.63); Males: 2.96 (0.57); Females: 2.60 (0.62)	4.14**	0.60
Prosocial content	Overall: 3.30 (0.52); Males: 3.11 (0.48); Females: 3.39 (0.52)	−3.83**	−0.56
Trait aggressiveness	Overall: 1.74 (0.60); Males: 1.95 (0.63); Females: 1.63 (0.56)	3.76**	0.55
Empathy	Overall: 3.40 (0.67); Males: 3.09 (0.60); Females: 3.56 (0.65)	−5.10**	−0.75
General media use	Overall: 4.76 (2.08); Males: 5.32 (2.44); Females: 4.46 (1.81)	2.95*	0.42
Violent media use	Overall: 5.26 (1.86); Males: 6.18 (1.97); Females: 4.78 (1.61)	5.80**	0.81

Note. Df = 390, *p < .005, **p < .001

CFA of antisocial and prosocial content

To assess the factor structure of the C-ME2, a Confirmatory Factor Analysis (CFA) was conducted using Structural Equation Modeling (Mplus-V.6.12). Maximum-likelihood extraction (WLSMV) with polychoric or categorical variables was used. The following fit indices are reported: the Comparative Fit Index (CFI), the Tucker Lewis Index (TLI), the Root Mean Squared Error of Approximation (RMSEA), with its 90% confidence interval, and the Chi-Square with its corresponding degrees of freedom. A good fit is assessed with a CFI and TLI of .95 or higher, an RMSEA of .06 or lower (RMSEA<.10 indicates an adequate fit), and a non-significant Chi-Square (Chen, 2007; Cheung & Rensvold, 2002; Hu & Bentler, 1999; Kelloway, 1998). However, Chi-Square tests are often significant in large samples even when the fit is good (Hu & Bentler, 1999).

Because we expected that males and females would differ in their use of antisocial and prosocial media content, a CFA with multi-group analysis on participant sex was conducted separately for antisocial items and prosocial items. Because males and females had different variances on various items, the thresholds of some items were unconstrained and the error variances of several items were allowed to correlate. The model for antisocial media content showed a good fit; CFI = .98, TLI = .97, RMSEA = .08 (90% CI: .06-.10), χ²(111, N = 216) = 186.08 (χ²_males = 80.16, χ²_females = 105.92), p < .001. The model for prosocial media content also showed a good fit; CFI = .99, TLI = .99, RMSEA = .07 (90% CI: .04-.09), χ²(73, N = 216) = 133.25 (χ²_males = 81.08, χ²_females = 52.17), p < .01.

Both the antisocial media content factor and the prosocial content factor were internally consistent (antisocial: Cronbach α = .89; M = 2.72, SD = 0.63; prosocial: Cronbach α = .88; M = 3.30, SD = 0.52; total: Cronbach α = .90; M = 3.00, SD = 0.47). As expected, males reported more exposure to antisocial media content (M = 2.96, SD = .57) than did females (M = 2.60, SD = .62; t(214) = 4.14, p < .01, d = 0.60), whereas females reported more exposure to prosocial media content (M = 3.39, SD = .52) than did males (M = 3.11, SD = .48; t(214) = −3.83, p < .01, d = 0.56). Over all, both males and females reported more exposure to prosocial media content than to antisocial media content (males: t(73) = −2.37, p < .05, d = −0.30; females: t(141) = −15.05, p < .001, d = −1.38). Exposure to antisocial media content was significantly related to exposure to prosocial media content for both males (r = .42, p < .001) and females (r = .40, p < .001).

Predictive and discriminant validity

Predictive analyses were conducted between the antisocial media and prosocial content factors of the C-ME2 and trait aggressiveness, violent media use, general media use, and empathy. Furthermore, discriminant analyses were used to assess whether the antisocial factor discriminated between violent media use and general media use. In order to test discriminant validity, 1- and 2-factor models were conducted (following Hayes, Glynn, & Shanahan, 2005). First, a 1-factor model was tested for the antisocial factor and violent media use measure; all items were constrained to load on one factor. Second, a 2-factor model was tested in which the items were constrained to load on their expected latent variable (i.e., items of the antisocial factor of the C-ME2 were allowed to load on one factor, whereas items of the violent media use measure were allowed to load on the other factor). The 1- and 2-factor models were then compared, using χ²-difference tests, as well as fit indices.

As expected, exposure to antisocial media content related significantly to violent media use (r = .55, p < .001; r_males = .50, p < .01; r_females = .42, p < .001). Also, the 2-factor model fit the data significantly better than the 1-factor model (Table 3), which indicates that the two factors discriminated from each other. Thus, measuring exposure to antisocial media content via the C-ME2 explains unique variance beyond measuring violent media content with a traditional measure.

Table 3. Predictive and discriminant validity of the content-based media exposure-2 scale in Study 1.

Variable		Single factor model	Two factor model	Improvement in fit from single to two factor model
		χ^2(df)	χ^2(df)	Δχ^2 (df)
		CFI TLI	CFI TLI	ΔCFI ΔTLI
		RMSEA	RMSEA	ΔRMSEA
Violent media use and antisocial media content	Men	232.52 (399)	228.10 (395)	4.42 (4)
	Women	339.61 (399)	282.89 (395)	56.72 (4)
		.94	.96	.02
		.94	.96	.02
		.06	.05	−.01
General media use and antisocial media content	Men	211.38 (285)	187.18 (281)	−24.20 (4)
	Women	230.93 (285)	214.02 (281)	16.91 (4)
		.95	.96	.01
		.95	.96	.01
		.07	.06	−.01
General media use and prosocial media content	Men	251.65 (240)	194.62 (236)	57.03 (4)
	Women	167.25 (240)	161.31 (236)	5.94 (4)
		.93	.95	.02
		.93	.95	.02
		.08	.07	−.01

Note: Cronbach α is the internal consistency reliability estimate for measurement of the construct in that row. CFI = Confirmatory fit index. TLI = Tucker Lewis index. RMSEA = Root mean squared error of approximation. All changes in χ² are statistically significant at p < .05.

As expected, the general media exposure measure correlated positively to the antisocial factor of the C-ME2 (r = .44, p < .001; r_males=.59, p < .001; r_females=.37, p < .001). Furthermore, the 2-factor model fit significantly better than the 1-factor model (Table 3). Thus, measuring exposure to antisocial media content via the C-ME2 also explains unique variance compared to measuring general media exposure in a commonly used way.

Similarly, when considering prosocial and general media use, the 2-factor model was significantly better than the 1-factor model (Table 3). Thus, measuring exposure to prosocial media content via the C-ME2 explains unique variance above measuring general media exposure. In addition, the general exposure measure was positively correlated to exposure to prosocial media content (r = .20, p < .01; r_males = .52, p < .001), but not for females (r_females = .21, p = .08).

Predictive validity

Exposure to prosocial media content correlated positively to empathy (r = .30, p < .001; r_males = .19, p = .11; r_females = .23, p < .007), although the correlation was not significant for males. Exposure to antisocial media use was not significantly correlated to empathy (r = -.08, p = .25; r_males = .08, p = .52; r_females = .13, p = .12). In addition, empathy was not significantly correlated to violent media exposure (r = -.11, p = .11; r_males=.10, p = .41; r_females = -.04, p = .61), as expected.

Exposure to antisocial media content was significantly related to trait aggressiveness (r = .30, p < .001; r_females = .28, p < .001; and only marginally significant for young adult males (r_males = .20, p = .09). Exposure to prosocial media content was not significantly related to trait aggressiveness (r = .003, p = .97; r_males = .06, p = .60; r_females = .08, p = .38).

Study 2

Study 2 was designed to replicate and extend the findings of Study 1 and the original C-ME using a sample of adolescents (aged 11–17).

Method

Participants and procedure

Participants were 392 secondary school pupils from two rural schools in The Netherlands (M_age=13.01, SD_age=0.89, 56.6% males; 92.3% Caucasian). The procedure was the same as in Study 1. In one school, the survey was completed online. In the other school, the survey was administered by paper-pencil due to limited access to computers. Results were similar for the two schools, so the data were merged. Missing data were handled using hotdeck imputation (Myers, 2011), with the decks being gender and age. Participant consent rate was 100%; parent consent rate was 99.7%.

Measures

Trait aggressiveness (Cronbach α = .88, M = 1.95, SD = 0.79) and empathy (Cronbach α = .73, M = 3.37, SD = 0.63) were measured in the same way as in Study 1.

General media use was measured differently because the general media use measure had low reliability in Study 1. In Study 2, we measured participants’ general use of TV, music, the Internet, and video games. Furthermore, we refined the rating scale to measure participants’ media use frequency more precisely than in Study 1 with scores ranging from 1 = 0 hours to 9 = 7 hours or more. However, these adjustments did not improve the reliability of the scale (Cronbach α = .54; M = 7.09, SD = 2.95).

Violent media use was measured in the same way as in Study 1 (Cronbach α = .79; M = 16.06, SD = 9.80).

Results and discussion

See Table 4 for descriptive statistics for all variables measured in Study 2.

Table 4. Descriptive statistics for the measures included in Study 2 for all participants (N = 392), and for males (n = 222) and females (n = 170) separately. Independent t-tests and standardized mean differences (i.e., Cohen’s d) are given for gender differences.

Variable	Mean (SD)	t	d
Antisocial content	Overall: 2.17 (0.79); Males: 2.33 (0.81); Females: 1.96 (0.72)	4.69**	0.48
Prosocial content	Overall: 2.83 (0.77); Males: 2.66 (0.73); Females: 3.05 (0.76)	−5.10**	−0.53
Aggressiveness	Overall: 1.95 (0.79); Males: 2.13 (0.83); Females: 1.73 (0.66)	5.29**	0.53
Empathy	Overall: 3.37 (0.63); Males: 3.13 (0.55); Females: 3.68 (0.58)	−9.76**	−0.98
General media use	Overall: 16.06 (9.79); Males: 17.41 (10.20); Females: 14.29 (8.97)	3.22**	0.32
Violent media use	Overall: 7.09 (2.95); Males: 8.15 (3.01); Females: 5.69 (2.19)	9.38**	0.92

Note. df = 390, ** p < .001

CFA of antisocial and prosocial media content

As in Study 1, the model for antisocial media content showed a good fit; CFI = .99, TLI = .99, RMSEA = .06 (90% CI: .04-.07), χ²(122, N = 392) = 187.40 (χ²_males = 73.71, χ²_females = 113.69), p < .001. The model for prosocial media content also showed a good fit; CFI = .97, TLI = .98, RMSEA = .07 (90% CI: .06-.09), χ²(90, N = 100) = 190.63 (χ²_males = 96.00, χ²_females = 94.63), p < .001.

Both the antisocial (Cronbach α = .89; M = 2.17, SD = 0.79) and prosocial (Cronbach α = .88; M = 2.83, SD = 0.77) factors were statistically reliable. As expected, the adolescent males reported significantly more exposure to antisocial media content (M = 2.33, SD = 0.81) than did females (M = 1.99, SD = 0.72), t(390) = 4.69, p < .001, d = 0.47. In contrast, females reported significantly more exposure to prosocial media content (M = 3.05, SD = 0.76) than did males (M = 2.66, SD = 0.73), t(390) = 5.10, p < .001, d = 0.52; Table 5). As in Study 1, both males and females reported on average significantly more exposure to prosocial media content than to antisocial media content (males: t(221) = −4.66, p < .001, d = −0.44; females: t(169) = −15.24, p < .001, d = −1.47). Exposure to antisocial and prosocial media content did not correlate for the overall sample or for males (r = .04, p < .42; r_males = .03, p = .67), although there was a significant positive correlation for females (r_females = .21, p < .01).

Table 5. Item loadings and descriptive statistics for items of the C-ME2 used in Study 2.

	Antisocial factor		Prosocial factor
Item	Males	Females	Males	Females
Each item starts with: “How often do you watch, on the Internet/TV/games/mobile phone/DVD, people who…”
… openly talk about sex?	.66	.73
… drink (a lot of) alcohol?	.80	.79	-	-
… fight?	.68	.75	-	-
… use drugs?	.79	.77	-	-
… make a fool of someone else?	.65	.74	-	-
… are having sex?	.67	.68	-	-
… laugh at another persons’ expense?	.49	.65	-	-
… destroy someone else’s belongings?	.77	.81	-	-
… say negative things about another person behind their back?	.49	.80	-	-
… steal?	.81	.75	-	-
… make someone else trip and fall for fun?	.46	.49	-	-
… shoot at another person?	.60	.77	-	-
… help someone?	-	-	.68	.75
… are willing to do something for someone else?	-	-	.78	.78
… comfort others?	-	-	.69	.82
… are nice to another person?	-	-	.70	.84
… give something to another person to make them happy?	-	-	.61	.70
… stand up for someone?	-	-	.67	.78
… are in love?	-	-	.58	.62
… cooperate with someone else to do good?	-	-	.61	.71
… understand how another person is feeling?	-	-	.65	.74
… put someone in the spotlight in a positive way?	-	-	.63	.61

Discriminant validity

As expected, exposure to antisocial media content was positively related to violent media use (r = .63, p < .001; r_males = .68, p < .001; r_females = .46, p < .001). The 2-factor model fitted the data significantly better than the 1-factor model (Table 6). Exposure to prosocial media content and violent media use correlated negatively for the adolescents over all (r = -.14; p < .01; nonsignificant: r_males = -.09, p = .17; r_females = .04, p = .58).

Table 6. Predictive and discriminant validity of the content-based media exposure-2 scale in Study 2.

		Single factor model	Two factor model	Improvement in fit from single to two factor model
Variable		χ^2(df)	χ^2(df)	Δχ^2 (df)
		CFI	CFI	ΔCFI
		TLI	TLI	ΔTLI
		RMSEA	RMSEA	ΔRMSEA
Violent media use and antisocial media content	Males	360.78 (410)	209.35 (406)	−151.43 (4)
	Females	452.63 (410)	344.15 (406)	−108.48 (4)
		.91	.97	.06
		.90	.96	.06
		.08	.05	−.03
General media use and antisocial media content	Males	308.52 (230)	150.98 (226)	−157.54 (4)
	Females	276.49 (230)	206.30 (226)	−70.19 (4)
		.93	.97	.04
		.92	.97	.05
		.10	.06	−.04
General media use and prosocial media content	Males	335.71 (192)	132.49 (188)	−203.22 (4)
	Females	299.18 (192)	179.55 (188)	−119.63 (4)
		.88	.97	.09
		.89	.97	.08
		.12	.06	−.06

Note: Cronbach α is the internal consistency reliability estimate for measurement of the construct in that row. CFI = Confirmatory fit index. TLI = Tucker Lewis index. RMSEA = Root mean squared error of approximation. All changes in χ² are statistically significant at p < .05.

As expected, the general media exposure measure correlated positively with the antisocial factor of the C-ME2 (r = .40, p < .001; r_males = .31, p < .001; r_females = .38, p < .001). Again, the 2-factor model showed a better fit than the 1-factor model (Table 6). The correlation between exposure to prosocial media content and general media exposure was nonsignificant (r = .00, p = .96; r_males = -.01, p = .95; r_females = .13, p = .051). Also, the 2-factor model of prosocial media use and general media use fit better than the 1-factor model (Table 6).

Predictive validity

As expected, antisocial media use was negatively correlated to empathy (r = -.14, p < .01; r_males = -.12, p = .07; r_females = .07, p = .39). Exposure to prosocial media content correlated positively to empathy (r = .37, p < .001; r_males = .33, p < .001; r_females = .26, p < .001).

Also expected, exposure to antisocial media content was significantly related to trait aggressiveness (r = .57, p < .001; r_males = .53, p < .001; r_females = .53, p < .001). The correlation between trait aggressiveness and prosocial media content was nonsignificant (r = -.09, p = .07; r_males = -.08, p = .27; r_females = .04, p = .61).

Regression analysis showed the antisocial factor of the C-ME2 was the best predictor of aggressiveness for males and females (β_males = .38, p < .001; β_females = .39, p < .001), followed by general media use (β_males = .19, p < .01; β_females = .25, p < .001), followed by violent media use (β_males = .18, p < .01; β_females = .17, p < .05; Table 7).

Table 7. Regression analyses predicting aggressiveness in Study 2.

	Aggressiveness
	Males		Females
Predictor	R^2	β	R^2	β
Model	.35		.38
Antisocial Media Use		.38***		.39***
Violent Media Use		.18**		.17*
General Media Use		.19**		.25***
N	222		170

Note. *p < .05, **p < .01, ***p < .001

General discussion

The results from Study 2 generally replicated the findings from Study 1. However, adolescents were exposed to more violent media content (M = 16.06, SD = 9.80) than were young adults (M = 5.26, SD = 1.86). As expected, predictive validity was also greater for adolescents than for young adults. Likewise, among young adults, exposure to prosocial media content correlated positively to empathy for females but not for males, whereas among adolescents exposure to prosocial media content was positively correlated to empathy for males and females. The C-ME2 seems to be a better measure for adolescents than for young adults.

The psychometric properties of the prosocial factor were comparable to the psychometric properties of the antisocial factor. Adding the prosocial factor improves the balance of measuring content-based media exposure and is useful in investigating the effects of both antisocial and prosocial media content. Although a number of studies have investigated the negative impact of violent media content on youth, far fewer studies have investigated the positive impact of prosocial media content on youth. Thus, the C-ME2 can help fill this gap in the literature.

Consistent with previous research, males were heavier users of media with antisocial content than females (e.g., Dal Cin, Stoolmiller, & Sargent, 2012; Denniston, Swahn, Hertz, & Romero, 2011; Linder & Gentile, 2009; Möller, Krahé, Busching, & Krause, 2011; Parkes et al., 2013; Den Hamer et al., 2014), whereas females were heavier users of media with prosocial content than males. Likewise, predictive validity was comparable, and in the adolescent sample somewhat stronger, to the results for the original C-ME (den Hamer et al., 2017). Importantly, the factor structure of the C-ME2 (antisocial vs. prosocial media content) provided a good fit for both males and females, consistent with the original C-ME.

One limitation with the C-ME2, and most other media exposure measures, is that it relies on self-report data. Previous research has shown that self-reported data can be problematic (e.g., Nisbett & Wilson, 1977). Future research should compare C-ME2 responses to actual exposure to antisocial and prosocial media content. Another limitation in both studies is the low reliability of the measure of general media exposure (cf. Lee et al., 2008).

Although the C-ME2 includes items that measure exposure to different types of antisocial content in the media, these individual items loaded on the same factor. However, researchers interested in specific types of antisocial behavior could use those specific items.

In conclusion, the C-ME2 measures both the frequency and differential exposure to a wide variety of media content, in particular antisocial and prosocial content, covering all media channels, today often converging through computer screens. The C-ME2 is a valid, reliable, and easy to use instrument in all research designs, and allows for standardization and comparisons across studies.

Appendix A.

All items of the C-ME2 scale

Please report for every question how often you watch this on TV/Internet/DVD. This could be clips on You Tube, music videos, quiz shows, television shows, video games, cinema, etc. So, it does not matter where you watch it, but how often you watch it.*

For every statement, circle one number.

Table

		Never	Hardly ever	Sometimes	Often	Very often
	How often do you watch (on the Internet/TV/games/mobile phone/DVD) …
1	… people who fight?	1	2	3	4	5
2	… people who openly talk about sex?	1	2	3	4	5
3	… people who use drugs?	1	2	3	4	5
4	… people who destroy someone else’s belongings?	1	2	3	4	5
5	… people who shoot at another person?	1	2	3	4	5
6	… people who make a fool of someone else?	1	2	3	4	5
7	… people who drink a lot of alcohol?	1	2	3	4	5
8	… people who make fun of another person?	1	2	3	4	5
9	… people who are having sex?	1	2	3	4	5
10	… people who say negative things about another person behind their back?	1	2	3	4	5
11	… people who make someone trip and fall for fun?	1	2	3	4	5
12	… people who steal?	1	2	3	4	5
13	… people who are nice to another person?	1	2	3	4	5
14	… people who help someone?	1	2	3	4	5
15	… people who are in love?	1	2	3	4	5
16	… people who understand how another person is feeling?	1	2	3	4	5
17	… people who put someone in the spotlight in a what positive way?	1	2	3	4	5
18	… people who are willing to do something for someone else?	1	2	3	4	5
19	… people who comfort others?	1	2	3	4	5
20	… people who give something to another person to make them happy?	1	2	3	4	5
21	… people who stand up for someone?	1	2	3	4	5
22	… people who cooperate with someone else to do good?	1	2	3	4	5

*In the reported studies, we measured media exposure as a momentary state whereas media use also develops over time. The C-ME2 is open to include time-indicators in the instructions (e.g., asking for media exposure in specific weeks or over the last month) and can therefore likewise be used for long-term and longitudinal purposes. Items should be presented in randomized order.