The items discrimination index (corrected item-test correlation) ranges between .92 and .39, i.e. all are within acceptable values. The test is highly reliable, with an alpha coefficient of .99. In turn, the results of the Kolmogorov-Smirnov test do not permit the univariate normality hypothesis to be maintained either for item scores or total scale score.

The results of the exploratory factor analysis using the Parallel Analysis carried out recommend the extraction of a single factor, both Bartlett's test statistics (13104.9 -g.l.=190, p=.000010-) and the Kaiser-Meyer-Olkin test (0.97) indicating that the data meet the correct conditions for undergoing factor analysis as regards item intercorrelations; the total variance percentage explained by the factor (67.5%) and the values of both GFI (.999) and RMSR (.048) show a good fit of data in a one-dimensional model.

The confirmatory factor analysis carried out with each of the remaining random samples enables us to confirm that the values obtained indicate excellent goodness-of-fit of the data to the one-dimensional model (Table 1).

Using the complete sample, the factor weights of each one of the variables were calculated. The results show very high factor weights of the factor items, ranging from .45 to .98 (Table 2).

On the other hand, data analysis by ITR using Samejima's graded response model produces the Test Characteristic Curve (TCC) where the unbroken line is the information function and the dotted line is typical measurement error. The x-axis represents score in the variable studied. In this way and as can be seen, the test is more accurate, with greater information and lower typical error for the central values of the variable (see Figure 1). Likewise, the results shown in Table 3 reveal that the idea most easy to change would be “People should be treated equally, regardless of their sex” (item 2) and the most stubbornly held and hardest to change and intervene on would be “I think it is right that in my circles of friends, my future domestic activity is considered more important than my professional activity” (item 14).

Figure 1.

Item information function.

(0.1MB).

Lastly, interindividual differential analysis using the non-parametric Mann-Whitney U test will show statistically significant differences depending on the educational level of study participants - university vs. Vocational Training - (p<.001), with a value of Z= -30.16, and between males and females (p=.014), with a value of Z= -2.46. The grading can be seen in Table 4.

The purpose of the study has been to construct an improved scale for the assessment of gender role attitudes. Based on the existing scales in the literature on sexist attitudes, we have tried to include in our proposal the principal processes in our socialization - family, social interrelations and employment. To do so, items from different scales have been taken into account, in addition to other elements considered in the literature or in questionnaires previously used with victims of violence experienced within an affective relationship, and that reflected both egalitarian and sexist attitudes.

The result is the GRAS made up of 20 items with five answer options for each item. Items are classified in terms of attempting to identify attitudes held by both males and females that encourage aggressive behavior within dating relationships among adolescents, (Baber & Tucker, 2006; Glick & Fiske, 1996, 1999; López-Cepero et al., 2013). This grouping has been distributed into two categories - transcendent attitudes vs. sexist attitudes - which in turn are subdivided into three areas: family, social interrelations and employment. The final vision of the instrument has been set up as a scale representing a bipolar dimension classified by the research team into: two items of Transcendent Attitudes in Family Function; four showing Transcendent Attitudes in Social Interrelation Function; the same number showing Sexism in Family Function and Social Function; and lastly six for Sexism in Employment Function.

At the same time as confirming the existence of a single factor which fulfills all the necessary fit indices (percentage of explained variance, GFI, RMSR, RMSEA, CFI, and SRMR), the psychometric properties of the scale offer a reliability coefficient of .99. Similarly, the factor weights results for each of the variables show high values, without this meaning that their relevance for determining the factor is lessened.

On the one hand, the results obtained, in addition to confirming their repetition, show the commonality found in the information in our scale for each of the categories set out in the different areas considered. On the other, accepting that a change in behavior first requires a change in attitude, we have attempted to discover what positions of the instrument used offer the highest likelihood of being able to be changed. In this respect, we have seen that the easiest idea to change belongs to the category Transcendent Attitude in Social Function: “People should be treated equally, regardless of their sex”; likewise, two other attitudes liable to be changed belong to this same category: “Boys have the same obligations to help with household chores as girls” and “Household chores must not be allocated by sex”. These statements show that treating both sexes equally could mean lower likelihood of aggressive behavior by either partner in adolescent dating relationships.

Bearing in mind that our study population sample is young, the results obtained invite us to be cautiously optimistic and think that upbringing styles appear to be changing in our society (Ferrer et al., 2006). This opens up the possibility of developing a more egalitarian gender role attitude, which we will need to confirm in subsequent studies where both members of the couple exclude violent conduct from their behavioral repertory.

By contrast, the category which appears to be most resistant to change is that shown by the items of Sexism in Social Function: “I think it is right that in my circles of friends, my future domestic activity is considered more important than my professional activity”, followed by those in the Employment Function grouping: “A father's main responsibility is to help his children financially”, and “Some jobs are not appropriate for women”, which may be a true reflection of the reality observed in our context - it is, furthermore, reasonable to think that it really is very difficult to change this belief and this attitude to role so deeply embedded in the gender roles to be performed by each sex. That is to say, these attitudes involve the most deeply rooted positions where gender is imposed by power - males vs. females - encouraging discrimination, particularly in employment. This is coherent with the studies of Herrera et al. (2012), Lila et al. (2014) and López-Cepero et al. (2013), who all point out a greater predisposition towards violence within affective relationships as a product of such attitudes, which should make victimization and perception of abuse possible in the world in which we live.

In sum, it is important to highlight the use of a sample made up of young people, since very little research has been carried out on this group. Using the knowledge acquired in the area, this is highly relevant for preventing the appearance and upholding of attitudes that have a negative bearing on relationships in couples and are the precursor to acceptance of violent behavior. Another strength of the research lies in the instrument designed to carry it out, which attempts to offer a wide perspective on attitudes by organizing them into two categories (egalitarian and sexist) and looking at three different contexts (family, social and employment). Also worth noting are the differences existing around two important variables (sex and educational level), our GRAS offering better information on attitudes for assessing change, one of its strengths being that it includes males and females (without segregating the information) and, on the other, that it allows the researcher to work without having to take the sexual orientation of the person being assessed into account, which has practical implications in the healthcare area. To sum up, the analysis enables the attitudes more liable to be changed to be identified, proposing the alternatives possible in order to set out the guidelines for designing prevention and intervention programs for dealing with abuse situations.

We are aware that these implications have a limitation that needs to be corrected in future studies in this area of research, insofar as establishing its construct validity and discrimination capacity will be included in more extensive assessments of abuse perceived or unperceived by victims, or its relationship with the presence or absence of violent behavior suffered or received in dating relationships between young couples.

This research was funded by the Ministry of Health, Social Policy and Equality (SUBMINMU012/009).