• No se han encontrado resultados

Capítulo 4. Tinta vs. Pixel

4.1. Análisis de formato impreso

4.1.2. Revista Vogue

Task order (i.e., the order of implicit, explicit and weight/height measurements) did not have a significant influence on any of the other measures. Native and non-native speakers also did not differ in their scores on any of the measures. Participants scored an average of 21.15 (SD=6.13) on the RSES, which falls within a normative range (Schmitt & Allik, 2005). RSES reliability in this sample was satisfactory at α=.896. Participants scored 1.86 (SD=1.30) on the EDE-Q. Mean scores for the restraint, weight, eating and shape subscales were 1.59 (SD=1.47), 2.29 (SD=1.63), 1.07 (SD=1.17) and 2.50 (SD=1.50), respectively. These values are somewhat higher than the normative sample data obtained by Mond, Hay, Rodgers and Owen (2004), in which female participants scored, on average, 1.29 on the restraint, 1.64 on the weight, 0.59 on the eating, and 2.16 on the shape concern subscale. Global EDE-Q score (1.42) was significantly lower than in the present sample, t(262)=2.821, p=.005, which indicates elevated levels of ED pathology in the current study sample. However, participants in the Mond et al. (2004) study were also older (mean age 35.3) and, on average, slightly overweight (mean BMI 25.2). Reliability for the EDE-Q in the current sample was found to be satisfactory, at α=.931.

91

Participants scored 46.23 (SD=4.00) on the MPS; reliability was found to be relatively low, at α=.671. Finally, mean IAT-D score was .50 (SD=.41) across the full sample (N=69) and was significantly different from zero, t(68)=10.13, p=.000. This indicates that participants performed significantly quicker in the congruent, compared to the incongruent, IAT trial block – which suggests stronger positive than negative, automatic associations with the self. A

correlation matrix of these variables is found in Table 2. EDE-Q scores were significantly, negatively correlated with both the RSES, and the IAT-D. This suggests that higher ED pathology was associated with both lower explicit and implicit self-esteem. Surprisingly, EDE-Q was not significantly correlated with MPS (p=.611).

Table 2 Study 1 correlations 1 2 3 4 5 1. RSES 1 2. MPS .061 1 3. EDE-Q total -.647** .062 1 4. IAT-D .280* -.279* -.321** 1 5. BMI -.116 .076 .250 -.202 1

Correlations marked * are significant at the .05 level Correlations marked ** are significant at the .01 level

92

6.2. HYPOTHESIS 1

The first aim was to replicate previous findings which suggest that higher levels of ED pathology are associated with low explicit (RSES), but not lower implicit (IAT) self-esteem. In an analysis procedure consistent with Cockerham et al (2009) and Hoffmeister et al. (2010), participants were categorised into high- and low-pathology groups using a median split procedure. Participants who scored below the EDE-Q median (1.76) were categorised as “Low EDE-Q” (N=34), and those who scored above were categorised as “High EDE-Q” (N=34). An independent samples t-test found significant group differences on the RSES score t(66)=-4.144,

p=.000, 95% CIs [-8.237, -2.880]. A Shapiro-Wilk test with RSES score as the dependent variable

and pathology group as factor was not significant, p=.281, indicating normal distribution for this measure. Participants in the High EDE-Q group scored significantly lower on the RSES (M=18.32) compared to the Low EDE-Q group (M=23.88). Contrary to hypothesis, a similar pattern was found on the implicit measure using an independent-samples t-test: participants in the High EDE-Q group scored significantly lower on the IAT-D (M=.377) compared to those in the Low EDE-Q group (M=.614), t(66)=-2.462, p=.016, 95% CIs [-.428, -.044]. This suggests that participants with elevated EDE-Q scores reported both lower explicit and lower implicit self- esteem.

Mean IAT-D values were also tested for difference from zero in each group using a one- sample t-test; the calculation of the IAT-D suggests values above zero to be indicative of faster performance in the congruent block (positive self-associations) and values below zero to be indicative of faster performance in the incongruent block (negative self-associations).

Participants in the low EDE-Q group scored significantly above zero on the IAT-D, t(33)=9.066,

p=.000, d=3.156. Likewise, IAT-D in the High EDE-Q group was also greater than zero

93

in the congruent trial of the SE-IAT, which suggests that both groups held stronger positive automatic associations with the self than negative.

Although comparisons between groups were only partially consistent with the

hypothesis (i.e., the predicted difference in explicit self-esteem was found, but the differences in implicit self-esteem were unexpected), a repeated-measures ANOVA was carried out to assess a possible interaction between ED pathology and self-esteem. EDE-Q group (Low vs. High) was used as the independent variable, while IAT-D and RSES were used as outcome variables for implicit and explicit self-esteem, respectively. Consistently with previous results, a main effect of self-esteem was found, F(1,66)=962.220, p<.000, η2

p=.936. The interaction was

also found to be significant, F(1,66)=16.045, p<.000, η2

p=.196. These findings suggest that

despite the fact that participants both high and low in ED pathology held overall positive automatic associations with the self, the discrepancy between implicit and explicit self-esteem was indeed greater in participants with elevated pathology. Self-esteem discrepancy and ED pathology was then assessed systematically for Hypothesis 2.

6.3. HYPOTHESIS 2

The aim was to determine whether discrepant self-esteem (i.e., low explicit, high

implicit) was a better predictor of ED symptomatology than low explicit self-esteem alone. First, a hierarchical regression model was run to determine whether both variables (RSE and IAT-D) explained more variance in Global EDE-Q scores than the RSES alone. The first model was significant, indicating that RSES significantly predicts EDE-Q, F(1,68)=48.228, p=.000, R2=.419, β=

-.647. The second model, incorporating both factors as independent variables, was also significant: F(2,68)=25.891, p=.000, R2=.440. However, R2 change for the second model ( .021)

failed to reach significance, p=.120. This suggests that implicit self-esteem may not contribute significant regression value to the model. However, the hypothesis was also specifically made with regards to self-esteem discrepancy, and a multiple regression model does not account for

94

an interaction between the variables entered into the model. For this reason, a second regression analysis was carried out.

First, a “self-esteem discrepancy (SED)” value was calculated. The RSES and IAT-D scores were standardised and their values subtracted: i.e., (z-IAT-D) – (z-RSES). As a result, a larger SED indicates a greater discrepancy between a participant’s implicit and explicit self-esteem. A regression model was then conducted using SED as the predictor and EDE-Q as the outcome variable. A hierarchical stepwise regression was used with Global EDE-Q as the dependent variable, and SED, IAT-D and RSES as independent variables. The steps were as follows: 1) predicting Global EDE-Q from RSES, 2) predicting Global EDE-Q from RSES and IAT-D, and 3) predicting Global EDE-Q from SED. The overall regression model was significant, F(1,67)=48.228, p<.000, R2=.419. However, only the RSES variable reached significance, β=-.137, p<.000 and was

therefore included in the model. Neither the SED not the IAT-D significantly predicted the EDE- Q score, β=-.182 and -.151, respectively, p>.1. These results suggest that self-esteem

discrepancy did not predict ED pathology better than RSES alone.

6.4. HYPOTHESIS 3

The third hypothesis was concerned with fitting the key variables (self-esteem discrepancy, ED pathology and perfectionism) to a structural equation model. However, this hypothesis was predicated on the assumption that perfectionism was correlated with both ED pathology and self-esteem discrepancy. In fact, EDE-Q was not correlated with the MPS, r=.062,

p=.611, nor with any of its subscales. Participants in the High EDE-Q group did not score higher

on the MPS compared to the Low EDE-Q group, F(1,66)=.415, p=.522. Similarly, no differences were found on the Self, Other or Social sub-scales. As a consequence, it was not possible to construct a model. However, it is worth noting that MPS was significantly correlated with the IAT-D, r=-.279, p=.020, which suggests that higher perfectionism was associated with lower implicit self-esteem in this sample.

95

As noted previously, the MPS is comprised of three subscales, designed to assess self- oriented, other-oriented, and socially prescribed perfectionism. Although it was not possible to address Hypothesis 3 in the intended way, some post-hoc analyses were carried out in order to assess any relationship between implicit self-esteem and the different facets of the MPS. No significant differences were found between high- and low-pathology groups on the Self,

F(1,68)=.057, p=.812, Other, F(1,68)=.742, p=.392, or Social, F(1,68)=.143, p=.707, subscales of

the MPS. Correlations between the MPS subscales and self-esteem measures were also carried out, but found no significant relationships between either a) the MPS subscales and the RSES (all ps >.1), or b) the MPS subscales and the IAT-D (all ps >.1). These results suggest that there was no significant relationship between either measure of self-esteem and the facets of the MPS.

7. Discussion