Introduction
The concept of body image has changed constantly throughout history, because its description depends on factors such as time, culture, biological processes and glo balization (Grogan, 2008; Gonçalves & Bedin, 2016). As a result of the evolution of the social context, it is currently necessary to have a flawless and extremely thin body, as this amounts to having a "perfect body" (Amaya, Alvarez, & Mancilla, 2010).
Women who want to achieve the ideal of beauty, generally modify their cognitive schemes, reinforced by social conception. Over time, these ideas are adopted and become behaviors that put their physical health at risk (Neziroglu, Khemlani-Patel, & Veale, 2008; Rodríguez & Cruz, 2008; Jones & Morgan, 2010; Ramírez et al., 2015), which increases the probability of suffering from an eating disorder (Gerbasi et al., 2014; Jauregui & Bolaños, 2011).
Current social influences determine the importance of achieving a socially acceptable body. Failure to do so can cause concern for weight. If this concern is constant it can cause body discomfort, and if such discomfort is not resolved in time, it is likely that it will lead to body dissatisfaction, which in turn could cause low self-esteem and low perception of social support (Vázquez et al., 2011; Vázquez, Alvarez, & Mancilla, 2000).
Social feedback creates a unilateral perception, since it postulates that a woman with a good body image is percei ved as a pleasant, successful, beautiful and self-confident person. For this reason, having the ideal body becomes so important and women decide to make long days of exer cise, extreme diets, take laxatives, among other harmful behaviors, despite knowing the negative implications that these practices have on their health (Herraiz-Serrano et al., 2015; Trejo-Ortíz, Mollinedo, Araujo, Valdez-Esparza, & Sánchez, 2016).
The relationship between body image and social in fluences is evidenced every time the physical appearance evolves because of various factors such as the media, social media, verbal messages and social relationships, among others (Cash, 1990). However, it has not been determined which variables have the greatest impact, how they are correlated or how they influence body image (Balantekin, Birch, & Savage, 2018).
Social influence is more relevant in economically deve loped environments, where the mass media and the social context play an important role in promoting the dissemination of idealized and extremely thin bodies (Bearman, Presnell, & Martínez, 2006; Frederick, Forbes, Grigorian, & Jarcho 2007; Vaquero-Cristóbal, Alacid, Muyor, & López-Miñarro, 2013; Ferguson, Muñoz, Garza, & Galindo, 2014; Moreno-Murcia, Marcos-Pardo, & Huéscar, 2016).
It is expected that in these social contexts the modeling of behavior and the verbal messages from peers or relatives influence the development of bodily discomfort (Balantekin et al., 2018). The opinion of the immediate social context reinforces the behaviors that were developed to achieve the desired body and in turn encourage comparison among peers, so that women develop a sense of achievement when their body is most praised by their friends and family (Wasylkiw & Williamson, 2013; Francisco, Narciso, & Alarcão, 2013; Saffon & Saldarriaga, 2014; Gerbasi et al., 2014; Cruwys, Leverington, & Sheldon, 2016). It is important to bear in mind that social recognition for having a perfect body is more valuable when the comparison is with a real and not an idealized model, such as those presented in the different media.
Advertising is also transcendental, since it shows incre asingly unattainable bodies that can decisively influence the dissatisfaction with body image, in such a way that ideas and behavior that trigger symptoms or lead to a diagnosis of an eating disorder are developed or promoted (Mas-Manchón et al., 2015; Losada, Leonardelli, & Magliola, 2015).
It is known that it is important to analyze the influence of peers, advertising, and the media on body dissatisfaction. That is why previous investigations have found that there is greater influence of social media and peers. However, the results on the influence of advertising are inconclusive (Fardouly, Pinkus, & Vartanian, 2017; Ferguson et al., 2014; Rodgers, Paxton, & McLean, 2014). Other studies focus on explaining how the interaction of body dissatisfaction with social factors causes poor body perception, contingent self-esteem and high comparison rates with others (Bailey & Ricciardelli, 2010; Ferguson, Winegard & Winegard, 2011; Salazar-Mora, 2008).
On the other hand, the problem of body dissatisfaction is increasing, so that, in developing countries like Mexico, it is more common to find teenage girls with a tendency to develop excessive concern for body image. Traits that pre cede the development of an eating disorder are increasingly common in Mexican adolescents (Vaquero-Cristóbal et al., 2013). Particularly, women are being affected by stereotypes of extreme thinness, which can lead to low self-esteem and a greater desire to be slim (Botella, Ribas, & Ruiz, 2009; Vaquero-Cristóbal et al., 2013).
The variables included in structural equations models are: (a)) body dissatisfaction, which occurs if the ideal body is internalized and when the person compares with it concludes that her body does not correspond to that ideal (Acosta & Gómez, 2003); (b) weight worry which refers to the individual's weight in relation to food intake and the concern developed by the unsightly aspects of obesity (Vázquez et al., 2011); (c) body discomfort, caused by anxiety when facing situations that question the beauty of the body (Vázquez et al., 2000).
In the structural equation models, these three variables will be interacting with other factors such as : (a) adver tising influence, which refers to the interest aroused by the publicity of slimming products; (b) verbal messages influence, which refers to the interest in articles, reports, books and conversations about weight loss; (c)) social models influence, which is the interest in the silhouette of actresses, advertising models and passers-by; and (d) social situations influence, which refers to the subjective social pressure experienced in food intake situations and the social acceptance attributed to thinness (Vázquez et al., 2000).
Figure 1 illustrates the theoretical models proposed for the present investigation.
Method
Design
The present research used a non-experimental, corre lational-explanatory and cross-sectional design, where information was gathered through a survey including two measurement instruments (Thompson, Diamond, McWilliam, Snyder & Snyder, 2005).
Participants
An incidental sampling was used. The sample consisted of 206 women with an average age of 22.12 years (SD = 4.21), all residents of the city of Monterrey, Mexico. Of the sample, 125 (60.7%) reported being students, 72 (35%) were employed, and nine (4.4%) were housewives. Regarding marital status, 113 (54.9%) reported being single without a partner, 69 (33.5%) were single with a partner, 23 (11.2%) were married and one (.5%) was divorced. As inclusion criteria it was proposed that the participants were Mexican women between the ages of14 to 30 years; and as exclusion criteria, not meeting the age range required or be male.
Instruments
Body Shape Questionnaire (BSQ;Cooper, Taylor, Cooper, & Fairbum, 1987). The Spanish version by Raich et al. (1996) and validated in Colombia by Castrillón, Luna, Avendaño and Pérez-Acosta (2007) was used. This questionnaire has 34 items, of which 20 assess body dissatisfaction and 14 are related to weight. The questionnaire is answered by means of a a six points Likert-type scale (1 = never to 6 = always). In its original version, the BSQ has an internal consistency of a = .96, whereas in the present study, these measurements were α = .96 and ω = .95.
Questionnaire on Influences of the Body Aesthetic Model (CIMEC-26,Toro, Salamero, & Martínez, 1994). An abbreviated version of 26 items was used, where eight of them assess bodily discomfort; eight, the influence of advertising; three, the influence of verbal messages; four, the influence of social situations, and three items, the in fluence of social models. The questionnaire is answered based on a three points Likert-type scale (0 = no, never to 2 = yes, always). The internal consistency of CIMEC-26 original version in patients with anorexia is α = .93 and in population with absence of eating disorder is α = .91. In the present study an internal consistency of α = .93 and ω = .91 was obtained.
Procedure
An open call was made through social networks to participate voluntarily in the study. The contact with the participants was face-to-face and the corresponding informed consent was obtained. Subsequently, they were told to open a league on the internet that gave them access to respond to the questionnaires previously described from a computer equipment provided by the research team. In the case of minor participants (29 in total), the informed consent of the parent or guardian was requested at the time of applica tion of the instruments. All participants were guaranteed absolute confidentiality and were reassured that the results obtained would only be used for academic purposes. In ad dition, the option of psychological care was given in case any participant requested it. Finally, it is worth mentioning that no monetary or material remuneration was given in exchange for participation in the present investigation. This study was approved in its technical and ethical aspects by the Doctoral Committee of the Faculty of Psychology of the Universidad Autónoma de Nuevo León [Autonomous University of Nuevo León].
Data Analysis
The IBM SPSS 24 software was used for the descriptive statistical analysis, (Pearson's r) correlations (Pearson's r) and Cronbach's alpha (α) for reliability. The AMOS 24 was used to calculate the multivariate normality coefficient (Mardia) and to perform the structural equations modeling.
In order to establish the omega (ω) reliability value, the mathematical expression found in McDonald (1999) was used. There were no lost cases in the sample. In each test, the typical scores were calculated to determine, by Z score, the existence of atypical cases that had a value of ± 3 and in the same way the Mahalanobis distance procedure (D 2) was performed to search for multivariate atypical cases that exceeded the limit of significance, p <.001 (Tabachnick & Fidell, 2013). After reviewing both statistics, the non-existence of outliers in the sample was established.
Likewise, the Mardia coefficient was calculated to confirm the multivariate normal distribution, which proved to be adequate since it was less than 70 for each model (M1 = 45.70, M2 = 22.17, M3 = 19.33; Rodríguez & Ruiz, 2008). On the other hand, with the purpose of confirming the theoretical models proposed, the structural equation modeling technique was used. Statistical analyzes were performed by the maximum likelihood criterion and no items were eliminated from the scales used in the study.
Seven indices were considered to prove the goodness-of-fit of each model: the ratio between the chi-square and the degrees of freedom (χ2/gl), the goodness-of-fit index (GFI), the adjusted goodness-of-fit index (AGFI), the comparative fit index (CFI), the mean square error of approximation (RMSEA), the residual normalized square root mean (SRMR) and the parsimony ratio (PR). For the aforementioned coeffi cients the expected values were: χ2/df ≤ 2, GFI and CFI ≥.95, AGFI ≥.90, RMSEA and SRMR ≤.08 and PR >.75, which are indicators of a good adjustment to data (Byrne, 2016).
The models proposed in Figure 1, have as constant feature the same four independent variables: advertising influence, verbal messages influence, social models influence, and social situations influence. The only change from one model to another with respect to the theoretical scheme was the dependent variable, that is, body dissatisfaction, weight worry, and bodily discomfort. The study consisted solely of self-reported measures, no anthropometric data such as weight or height were obtained.
Results
In this section, correlations, measures of central tenden cy and reliability values of the instruments used are first shown. Subsequently, the goodness of global adjustment of the models proposed is shown. Finally, the direct and indirect effects, as well as the values of the variance ex plained are indicated.
Descriptive statistics, correlations between study variables and reliability
Previously to the structural equations modeling, a Pearson's correlation analysis was performed between all variables studied, in addition to estimating the means (M) and standard deviations (SD). On the other hand, reliability values were reported for each sub-scale (See Table 1). It is important to note that all the variables correlated positively and significantly at a level of p <.01. The highest correla tions were between body dissatisfaction and weight worry (r = .871, p <.01), between body dissatisfaction and bodily discomfort (r = .848, p <.01), in addition to concern about weight and bodily discomfort (r = .826,p <.01), whereas the lowest correlation was between verbal messages influence and social situations influence (r = .320, p <.01).
Note. *p <.01, V1 = body dissatisfaction, V2 = weight worry, V3 = bodily discomfort, V4 = advertising influence, V5 = verbal messages influence, V6 = social models influence, V7 = social situations influence, M = mean, SD = standard deviation, α = Cronbach´s alpha, ω = McDonald´s omega
The correlations between the dependent variables are high (r ≥ .80), ergo, there is multicollinearity and with this there could be a direct affectation on the goodness of fit and explanatory level indicators of the models proposed (Kline, 2015). Therefore, the high statistical relationship between the constructs body dissatisfaction, weight worry and bodily discomfort, must be differentiated in a theoretical way, because by not doing so, errors of categorization and statistical inconsistency would occur, which could cause a misinterpretation of the results (Mancilla, Vázquez, Mancilla, Amaya, & Álvarez, 2012). Due to the above, it is justified to use separate models for each of the depen dent variables and thus be able to know the influence that advertising, verbal messages, social models and social situations have on them.
Goodness of fit of the models proposed
The resulting parameters indicate that the three models proposed have good goodness of fit in the seven indicators reported (See Table 2). It is important to mention that no covariance error was correlated and no items were elimi nated to improve the models.
Note. M 1 = body dissatisfaction model, M 2 = weight worry model, M 3 = bodily discomfort model. Values that indicate a good fit to the data: χ 2 /df ≤ 3, GFI and CFI ≥ .95, AGFI ≥ .90, RMSEA and SRMR ≤ .08, PR > .75.
As regards to the results, it is important to note that the value of χ2/df for the M1 and M3 models was less than one, which could be an overestimation indicator (Escobedo, Hernández, Estebané, & Martínez, 2016). However, this hypothesis is rejected, since the three models studied ob tained a PR greater than .75 (Byrne, 2016; James, Mulaik & Brett, 1982; Mulaik et al., 1989). Therefore, because the acceptance of any model is given by multiple indicators (Byrne, 2016), it is possible to affirm that the models pro posed are adjusted in an adequate way to the empirical data.
Direct and indirect effects between the variables studied
Figures 2, 3 and 4 show the direct effects and Table 3 presents the indirect effects of the three models proposed.
Body dissatisfaction model (M 1 ). In the case of the body dissatisfaction model (see Figure 2), the proposed parameters mostly obtained a level of significance of p <.01. There are direct effects of the advertising influence (β = .39, p <.01) and social situations influence (β = .53, p <.01) on body dissatisfaction. However, the regression coefficient for the social model’s influence was not significant (β = .11, p> .05). Other direct effects in this model were the verbal messages influence on advertising influence (β = .62, p <.01) and the verbal messages influence on social situations (β = .60, p < .01) and on social models (β = .66, p <.01).
Regarding indirect effects, the verbal messages influence had an indirect effect on body dissatisfaction, mediated by the advertising influence (β = .31, p <.01). In addition, if the route of the social situations influence is added, a double mediating effect of verbal messages on body dissatisfaction is obtained (β = .64, p <.01), but by using the path of social models influence the result is a non-significant effect on body dissatisfaction (β = .12, p >.05). On the other hand, the advertising influence had an indirect effect on body dissatisfaction mediated by the social situations influence (β = .39, p <.01). However, in the case of social models influence it was not significant (β = .077, p >.05). Other indirect effects found were the verbal messages influence on the social situations influence (β = .29, p <.01) and the social models influence (β = .36, p <.01) using the adver tising influence as a mediator variable.
Lastly, it was found that the influence of verbal messages explained 38% of the common variance of the advertising influence, while verbal messages influence and the adverti sing influence explained 44% of the social models influence variable and 36% of the social situations influence.
In short, the verbal messages influence in mediation with the advertising influence (β = .31, p <.01) and the verbal messages in mediation with the advertising influence and the social situations influence (β = .64, p <.01) explained 79% of the common variance of body dissatisfaction.
Weight worry model (M 2 ). With respect to the weight worry model (see Figure 3), the paths proposed mostly obtained a significance level of p <.01, with the exception of the path social situations → weight worry, in addition to publicity → social models → weight worry and verbal messages → advertising → social situations → weight worry that were significant at the level p <.05.
On the other hand, direct effects of the advertising in fluence (β = .42, p <.01), social situations (β = .21, p <.05) and social models (β = .29, p <.01) on weight worry were found. Other direct effects on this model were the verbal messages influence regarding the advertising influence (β = .63, p <.01), the verbal messages influence on social situations (β = .60, p < .01) and on social models (β = .66, p <.01). These last effects had little or no variability with respect to the previous model (M1).
As regards to indirect effects, the verbal messages in fluence had an indirect effect on weight worry, mediated by the advertising influence (β = .52, p <.01). In addition, if the path of the social situations influence is added, a double mediating effect of verbal messages on weight worry is obtained (β = .37, p <.05), and when using the path of the social models influence also results in a double mediating effect on weight worry (β = .46, p <.01). On the other hand, the advertising influence had an indirect effect on weight worry, mediated by the social situations influence (β = .39, p <.01) and, in turn, by the social models influence (β = .31, p <.05). Other indirect effects found were the verbal messages influence on the social situations influence (β = .37, p <.01) and the social models influence (β = .38, p <.01) using the advertising influence as a mediator variable.
Finally, it was found that the verbal messages influence explained 40% of the common variance of the advertising influence. In addition, the verbal messages influence ex plained 44% of the social models influence variable and 36% of the social situations influence. Together, the verbal messages influence in mediation with the advertising in fluence (β = .52, p <.01), verbal messages in mediation with the advertising influence and the social situations influence (β = .37, p <.01) and verbal messages in mediation with the advertising influence and the social models influence (β = .37, p <.01) explained 62% of the common variance of weight worry.
Bodily discomfort model (M 3 ). With respect to the bodily discomfort model (see Figure 3), the proposed paths indicated a level of significance of p <.01, with the exception of the path social models → bodily discomfort, p <.05. Regarding direct effects, it was found that the advertising influence (β = .38, p <.01), social situations (β = .36, p <.01) and social models (β =. 26, p <.05) have a significant effect on bodily discomfort.
Other direct effects in this model were the verbal mes sages influence regarding the advertising influence (β = .63, p <.01), social situations (β = .60, p <.01) and social mo dels (β = .66, p <.01). These last values have little or no variability with respect to the previous models (M1 and M2).
Concerning indirect effects, the verbal messages influence had an indirect effect on bodily discomfort, mediated by the advertising influence (β = .26, p <.01). Likewise, if the path of the social situations influence is added, a double mediating effect of verbal messages on bodily discomfort is obtained (β = .39, p <.01). However, when using the path of social models influence, a significant effect on the bodily discomfort was not obtained (β = .23, p >.05).
For its part, the advertising influence had an indirect effect on bodily discomfort, mediated by the social situations influence (β = .20, p <.01), but not significant through the social models influence (β = .16, p >.05). Other indirect effects found were the verbal messages influence on the social situations influence (β = .29, p <.01) and the social models influence (β = .38, p <.01) using the advertising influence as a mediator variable.
Finally, it was found that as in the M2 model, the verbal messages influence explained 40% of the common variance of advertising influence, 44% of the variable influence of social models and 36% of the influence of social situations. Collectively, the influence of verbal messages in mediation with the advertising influence (β = .26, p <.01) and verbal messages in mediation with the advertising influence and the social situations influence (β =. 39, p <.01), explained 72% of the common variance of bodily discomfort.
Discussion
Generally, each of the models proposed are an evidence that sociocultural variables have a significant influence on the creation of a negative body image, and in particular, when the effect of the verbal messages influence over the advertising influence was evaluated, the parameters got meaningful outcomes in the three models proposed. From this specific outcome is inferred that it is important for women to get feedback from their social context, since from this depends the importance and the approach given to the messages received from the different media (Fardouly et al., 2017; Ferguson et al., 2014; Rodgers et al., 2014).
In addition to this, in the models of body dissatisfaction, bodily discomfort and weight worry, a significant influence of advertising was found, mainly due to its direct effect on social models, social situations and verbal messages, which coincides with previous findings where the direct influence of communication media and advertising on the body image of women is described (Fardouly et al., 2017; Ferguson et al., 2014; Rodgers et al., 2014).
On the other hand, the social situations showed that they are directly related to body dissatisfaction, weight worry, and bodily discomfort. This result supports the theory in dicating that in contexts where beauty canons are unlikely for women to attain these have an influence in developing a wrong perception about their body image, whatever patho logical or normative this might be (Ferguson et al., 2014).
In the particular case of the social models variable, this showed a significant relationship with the perception of weight, which supports the theory that women only take into consideration the advertising aesthetic models when they want to compare their weight, but when they want to make a general corporal comparison they prefer their peers instead of unreal models (Brudzynski & Ebben, 2010).
Taking into account that for the three models the effects were studied considering mediating variables - this in order to know how they interact with each other with respect to the formation of body image -, the results obtained support the theoretical proposal.
In the case of the first model, regarding body dissatis faction, a significant relationship between verbal messages, advertising, social models, and body dissatisfaction was obtained. This shows that the messages that are given within the social context (parents, peers, mates, friends, etc.) influence the kind of publicity that is chosen to see, and this also impacts the reproduction of social situations which favor body dissatisfaction (Fardouly et al., 2017).
Regarding the weight worry model, a relationship was found between verbal messages, advertising, social situ ations, and weight worry. This shows that the interaction between the comments about weight, the advertising fos tering weight loss, and the situations that promote a thin body must result in weight worry (Caccavale, Farhat, & Iannotti, 2012). Weight worry is also related to verbal mes sages, publicity, and social models. This is consistent with other previous investigations (Bailey & Ricciardelli, 2010; Fardouly et al., 2017; Ferguson et al., 2011; Ferguson et al., 2014; Rodgers et al., 2014; Salazar-Mora, 2008). Thus, listening to comments about weight, seeing advertisements that promote thinness, and looking at people who represent extremely thin social models results in restlessness because of the amount of weight to lose, since it is necessary to be in accordance with the socially accepted beauty canon ( Rodgers et al., 2014).
Finally, in the third model, referring to bodily discomfort there is a relationship between verbal messages, advertising, and social situations. This explains that messages coming from the social environment influence the kind of publicity that is seen, since this promotes paying more attention or participating in social situations related to an ideal body image. The combination of these factors results in a corporal uneasiness which through time can lead to bodily dissatis faction (Saffon & Saldarriage, 2014). It is worth mentioning that this does not occur when bodily discomfort is related to verbal messages, advertising and social models, and also when trying to relate body dissatisfaction with verbal messages, advertising and social models.
This may happen because women do not believe that their body image can be comparable to a model that appears in some type of advertising, since they are aware of the various modifications that a model has to undergo in order to look good (Bailey & Ricciardelli, 2010).
Based on the above, it is concluded that social situations have a great impact in women's perceptions about their own body image. This means that it could be evidenced that social influences are a cause of weight worry as well as of body discomfort and body dissatisfaction. Therefore, it is necessary to carry out preventive interventions in which the objective is to increase resilience, change cognitive schemes and develop healthy behaviors that allow redu cing the impact of socially accepted standards of beauty (Espósito, 2015).
Finally, the limitations of the study include the sam ple size and the scarce participation of adolescent girls. Additionally, it is suggested for future research to consider anthropometric measures.