Results

Descriptive statistics
Means, standard deviations, and scale ranges for each media processing and critical thinking measure in samples A and B are shown in Table 1. The means for the measures were relatively consistent across samples.
Table 1 Descriptive Statistics, Internal Consistency Reliability and Test-retest Reliability for Media Processing and Critical Thinking Variables
Scale range Sample A (N = 216) Sample B (N = 208)
Time 1 Time 1 Time 2
Mean (SD) α CR Mean (SD) α CR Mean (SD) ICC
Media attitudes questionnaire (MAQ)
 Realism scepticism subscale 1 to 5 4.13 (0.94) .80 a .80 3.80 (1.06) .87 .86 4.10 (0.99) .64***
 Similarity scepticism subscale 1 to 5 3.49 (1.12) .83 .85 3.37 (1.00) .76 .79 3.49 (1.04) .81***
 Desirability scepticism subscale 1 to 5 2.90 (1.03) .82 .84 2.94 (0.98) .79 .80 3.16 (1.04) .70***
Critical thinking about media messages 6 to 36 17.56 (6.91) .90 .90 NA NA NA NA NA
Critical thinking about media messages – Appearance focus 6 to 36 15.78 (7.94) .95 .95 NA NA NA NA NA
Fake subscale 1 to 5 NA NA NA 3.04 (0.99) .92 .93 3.06 (1.07) .80***
Note. α Cronbach’s alpha, CR construct reliability, ICC intraclass correlation coefficient, NA data not available
*p < .05; **p < .01; ***p < .001
aSpearman-Brown coefficient calculated for internal consistency of a two-item scale

Factor structure

MAQ
The PCA of the ten items from the MAQ in sample A revealed three factors with eigenvalues greater than one. Inspection of the scree plot revealed a clear break after the first component, and a weak break after the third component. Parallel analysis supported retention of three factors. Following oblique rotation, the majority of the 10 items loaded strongly and separately on the three components (see Table 2), with the exception of items 3 (the models in advertisements are real people) and 7 (models in ads are intelligent) which had low item loadings, and also loaded across each of the three components with none of the loadings >.40. In addition, the communality values for both item 3 and 7 were relatively low (.42 and .38, respectively). Items 3 and 7 were omitted.
Table 2 Items and Factor Loadings for the Media Attitudes Questionnaire
Component loading from PCAa Factor loading from CFAb
Item (original subscale designation [6, 25]) 1. Similarity scepticism 2. Realism scepticism 3. Desirability scepticism 1. Similarity scepticism 2. Realism scepticism 3. Desirability scepticism
1. Normally women (in real life) look like models in ads (Ad, Realism; C, Realism) -.02  .92 .05 .89
2. Normally women (in real life) are as thin as the models in ads (Ad, Realism; C, Realism) -.06  .90 .00 .86
3. The models in advertisements are real people (Ad, Realism; C, Realism) .40 .33 -.07 - - -
4. I could look like the models in ads (Ad, Realism; C, Similarity)  .94 -.05 .06 .84
5. I could be as thin as the models in ads (Ad, Realism; C, Similarity)  .97 -.10 .03 .90
6. Most women could be as thin as the models in ads by exercising and/or dieting (Ad, Realism; C, Similarity)  .57 .16 -.22 .47
7. Models in ads are intelligent (Ad, Desirability; C, Desirability) .23 .33 -.25 - - -
8. Models in ads are beautiful (Ad, Desirability; C, Desirability) .01 -.06  -.88 .73
9. Models in ads have perfect bodies (Ad, Desirability; C, Desirability) -.01 -.06  -.92 .94
10. Models in ads have lots of fun (Ad, Desirability; C, Desirability) -.04 .07  -.80 .53
Average variance extracted .57 .76 .59
Major loadings on components are in bold
Note. PCA principal components analysis, CFA confirmatory factor analysis, Ad adolescent sample [6], C college sample [25]
aSample A only (N = 216); bSample B only (N = 201)
Data from sample B were subjected to CFA using the PCA in sample A to identify the model to be tested. Three items (4, 5, and 6) were specified for factor 1 (Similarity Scepticism), two items (1 and 2) for factor 2 (Realism Scepticism), and three items (8, 9, 10) for factor 3 (Desirability Scepticism). The CFA analysis revealed adequate fit to the data, χ 2(13) = 27.77, p = .010, CFI = .98,TLI = .95, RMSEA = .08. Three items were retained for the Similarity Scepticism subscale, two were retained for the Realism Scepticism subscale, and three were retained for the Desirability Scepticism subscale as per the component structure identified from the PCA in sample A. Factor loadings are shown in Table 2.

Invariance testing
Multi-group analysis of the items of the MAQ was conducted with sample A and sample B to test for invariance of the identified factor structure between samples. The unconstrained χ 2(13) = 50.76, p < .001, and constrained χ 2(34) =76.30, p < .001 models were not significantly different χ 2(21) = 25.4, p = .230, indicating that the identified factor structure was invariant across sample A and sample B.

CTMM
The CFA of the 6 items from the CTMM revealed good fit to the data, χ 2(7) = 9.06, p = .327, CFI = 1.00, TLI = 1.00, RMSEA = .03. The model fit in this sample supported the one factor structure of the original scale.

CTMM-AF
The CFA of the 6 items from the CTMM-AF also revealed good fit to the data, χ 2(7) = 11.24, p = .129, CFI = 1.0, TLI = .99, RMSEA = .05. The one factor structure for the CTMM-AF was also supported in this sample.

Fake
The CFA of the 5 items of the Fake subscale showed good fit to the data, χ 2(2) = 1.08, p = .406, CFI = 1.00, TLI = 1.00, RMSEA = .00. The one factor structure of the Fake subscale was supported in this sample. Items and factor loadings from CFA outcomes for the CTMM, CTMM-AF, and Fake subscale are shown in Table 3.
Table 3 Items and Factor Loadings for the CTMM, CTMM-AF, and Fake Scales
Measure and item Factor loading
Critical Thinking about Media Messagesa
 3. I think about what the people who made the media message want me to believe .90
 4. I think about the things the advertisers do to get my attention .82
 2. I think about who created the message I see on the ad .73
 1. I think about the purpose behind a message I see on television .71
 5. I think about whether the things that advertisers want me to do are good for me .71
 6. I try and think about how true or false an advertisement is .68
Average variance extracted .60
Critical Thinking about Media Messages – Appearance Focusa
 4. When I see ads with very attractive female models I think about the things the advertisers do to get my attention .92
 3. When I look at ads with thin female models I think about what the people who made the media message want me to believe .91
 2. When I see very attractive female models used in ads I think about who created the message I see on the ad .90
 1. When looking at thin female models on television I think about the purpose behind the message I see .87
 5. When looking at ads with thin female models I think about whether the things that advertisers want me to do are good for me .83
 6. When I see ads about ways to be more attractive I try and think about how true or false an advertisement is .75
Average variance extracted .76
Fakeb
 3. They probably used computer re-touching to make her look like that .92
 2. Nobody looks like that without computer tricks .90
 5. It takes a lot of camera tricks to make someone look that good .84
 4. That kind of perfection isn’t real .81
 1. She’s airbrushed .68
Average variance extracted .71
aSample A only (N = 216); bSample B only (N = 204)

Internal consistency reliability
Cronbach’s alpha and Spearman-Brown coefficient values, as well as construct reliability values for internal consistency of measures are shown in Table 1. All measures, in both samples, showed adequate to excellent internal consistency of items.

Test-retest reliability
Test-retest reliabilities of the MAQ and Fake subscales were examined in sample B and for the CTMM and CTMM-AF in the test-retest only sample. Intra-class correlation coefficients, shown in Table 1, indicated adequate agreement (≥.70) between Time 1 and Time 2 scores for the MAQ and Fake subscales, with the exception of the Realism Scepticism subscale where agreement was slightly below the adequate threshold. In the test-retest only sample, scores on the CTMM (ICC = .78, p < .001, M T1 = 14.50, SD T1 = 5.73, M T2 = 15.41, SD T2 = 7.15) and CTMM-AF (ICC = .70, p = .001, M T1 = 13.77, SD T1 = 5.98, M T2 = 13.47, SD T2 = 5.97) demonstrated adequate test-retest reliability.

Construct validity
As shown in Table 4, moderate positive correlations between the media processing variables Realism Scepticism, Similarity Scepticism, and Desirability Scepticism were revealed in both samples, indicating convergent validity. The Fake subscale, also considered to be a media processing variable was only correlated with Realism Scepticism and the correlation was of small magnitude. The two critical thinking variables, CTMM and CTMM-AF were very strongly correlated with one another in sample A, demonstrating convergent validity. The related but different types of media literacy variables, media processing and critical thinking, did not correlate strongly, the highest correlation being r = .14 between CTMM and both Realism Scepticism and Similarity Scepticism. The difference between the magnitude of the correlations was confirmed with Steiger’s tests [44] using Lee and Preacher’s software [45]. Correlations between media processing variables, Realism Scepticism, Similarity Scepticism, and Desirability Scepticism, and correlations between critical thinking variables, CTMM and CTMM-AF, were significantly stronger than correlations between any combination of media processing and critical thinking variable (all zs >2.2, all ps <.02).
Table 4 Bivariate Correlations between Media Processing and Critical Thinking Variables
1 2 3 4 5 6
Sample A
 1. Realism scepticism (MAQ) - .36*** .34*** .14* .10 NA
 2. Similarity scepticism (MAQ) - .47*** .14* .06 NA
 3. Desirability scepticism (MAQ) - .07 .04 NA
 4. Critical thinking about media messages - .80*** NA
 5. Critical thinking about media messages - Appearance focus - NA
Sample B
 1. Realism scepticism (MAQ) - .23** .38*** NA NA .15*
 2. Similarity scepticism (MAQ) - .28*** NA NA -.01
 3. Desirability scepticism (MAQ) - NA NA .05
 6. Fake NA NA -
Note. MAQ media attitudes questionnaire, NA data not available
*p < .05; **p < .01; ***p < .001
Item-level convergent validity was further supported by AVE values (see Tables 2 and 3) which for all measures were above the threshold of .5 for adequate convergence [42]. Discriminant validity between media processing and critical thinking variables was examined by comparing the AVE values for each construct with the square of the correlations between constructs. For all pairs of constructs, with the exception of the CTMM compared with CTMM-AF, the AVE value was greater than the squared correlation, supporting discriminant validity. The AVE value for CTMM was smaller than the squared correlation between CTMM and CTMM-AF, indicating that the CTMM did not have discriminant validity from the CTMM-AF. Discriminant validity for measures in sample B was also examined by comparing AVE values with the square of the correlations. For all pairs of constructs, the AVE value was greater than the squared correlation, supporting discriminant validity between MAQ subscales and between the MAQ subscales and the Fake subscale.
Correlations were calculated between the media processing and critical thinking variables and the eating disorder risk factor variables to examine further evidence for construct validity (see Table 5). A number of significant associations were revealed. The majority of observed relationships were in the expected direction such that higher levels of media scepticism for MAQ subscales, and higher levels of critical thinking for CTMM and CTMM-AF, were associated with lower levels of eating disorder risk factor variables. The Fake subscale was not significantly correlated with any of the eating disorder risk factor variables. Unexpectedly, in sample B, a positive correlation was revealed between Similarity Scepticism and body dissatisfaction, indicating that high similarity scepticism scores were related to high body dissatisfaction. The direction of this relationship was inconsistent with other observed relationships.
Table 5 Bivariate Correlations between Media Literacy Variables and Eating Disorder Risk Factor Variables
Age Media exposure Internalisation of the thin-ideal Appearance comparison Body dissatisfaction Dietary Restraint
Sample A
 Realism scepticism (MAQ) -.01 -.06 -.27*** -.13* -.14* -.28***
 Similarity scepticism (MAQ) -.03 -.09 -.29*** -.14* -.04 -.36***
 Desirability scepticism (MAQ) -.05 -.13**** -.53*** -.32*** -.27*** -.37***
 Critical thinking about media messages -.07 -.21** -.13* .00 -.13**** -.02
 Critical thinking about media messages - Appearance focus -.04 -.18** -.09 .03 -.12 -.01
Sample B
 Realism scepticism (MAQ) .23** .08 -.40*** -.15* -.07 -.13****
 Similarity scepticism (MAQ) .04 -.02 -.19** -.09 .25*** .09
 Desirability scepticism (MAQ) .00 -.10 -.45*** -.38*** -.18* -.19**
 Fake -.01 -.13**** -.01 .10 -.02 .07
Note. MAQ media attitudes questionnaire, NA data not available
*p < .05; **p < .01; ***p < .001; ****p < .10
Results from Steiger’s tests comparing the magnitude of relationships between media literacy variables and eating disorder risk factor variables that were expected to be proximally related (internalisation of the thin-ideal and appearance comparison) with the magnitude of relationships between media literacy variables and eating disorder risk factor variables that were expected to be distally related (body dissatisfaction and dietary restraint) were mixed. For example, in Sample A, Realism Scepticism had a stronger relationship with internalisation of the thin-ideal than with body dissatisfaction (z = −2.20, p = .014) but the former relationship was not stronger than the relationship between Realism Scepticism and dietary restraint (z = 0.16, p = .438). Also in Sample A, the relationship between CTMM and internalisation of the thin-ideal was stronger than the relationship between CTMM and dietary restraint (z = −1.65, p = .049), whereas the relationship between CTMM and appearance comparison was not stronger than the relationship between CTMM and dietary restraint (z = 0.29, p = .385). Results from all comparisons are available from the authors upon request.
Age was not related to the media literacy variables with the exception of a positive relationship with Realism Scepticism in sample B showing that older age was associated with higher scepticism. Variables assessing the media processing constructs were not significantly related to media exposure. The critical thinking variables both had small inverse correlations with media exposure such that lower levels of exposure to screen-based media were associated with higher levels of critical thinking.