The sample was made up of 2,364 students in their last year of compulsory education in the Principality of Asturias (Spain). The sampling technique used was a probability proportional to size (PPS). Gender, type of school and geographical origin were also considered in the selection of the sample. In the study, 1,197 women (50.7%) and 1,162 men (49.2%) – in 5 cases they did not indicate their gender –answered the questionnaires. The students were aged 14 to 18 (M = 15.65 and SD = 0.74). Analysis of age, by gender, revealed statistically significant differences (F = 10.8, p <0.01). In girls, the average age was 15.60 (SD = 0.72) and in boys, 15.70 (SD = 0.76). On referring to geographical origin it is worth noting that 77.1% of the sample (1,823 students) attended schools located in the center/north of the region, which included highly populated cities; 10.2% (241 students) went to schools located in the coalfields zone, 3.8% (91 students) lived in the eastern area of the region, and finally, 8.9% (209) lived in the west.

Science and Technology Outcome Expectations Scale. It is a modified version of the Mathematics/Science Outcome Expectations Scale (MSOES, Fouad, Smith, & Enochs, 1997). This scale assesses high school students’ beliefs about the potential consequences of studying science and technology subjects in high school (1 = strongly disagree to 6 = strongly agree). Internal consistency (α) in the present sample was 0.95, while Fouad and Smith’s (1996) study yielded an α of 0.71, and Fouad et al. (1997) obtained an α of 0.70.

Science and Technology Self-Efficacy Scale. This scale is a Spanish translation and adaptation of the Science Grade Self-Efficacy Scale (Britner & Pajares, 2001; 2006). The scale consists of 3 items assessing high school students’ confidence in their abilities to perform science and technology tasks successfully. Participants rated each item on a 6-point scale ranging from 1 (not confident at all) to 6 (completely confident). For the full sample, the scale yielded an α of 0.88. Britner and Pajares (2006) found an α of 0.85, and an earlier study by the same authors (Britner & Pajares, 2001) yielded an α of 0.86.

Science and Technology Interest Scale. This scale included 9 Likert-type items, and was an adaptation of the Engineering Fields Questionnaire (Lent et al., 2003). The original scale was intended for university students, but the research team adapted it to test adolescents’ interest in science and technology activities. Participants rated each item on a 6-point scale ranging from 1 (very low interest) to 6 (very high interest). Reliability of this scale was 0.94. In Lent et al. (2003) this scale got an α of 0.83, while in Lent et al. (2005) the α was 0.80.

Sources of Science and Technology Self-Efficacy Scale. This is a Spanish translation and adaptation of the Sources of Science Self-Efficacy Scale (Britner & Pajares, 2006). The scale consists of 59 items measuring the effects of mastery experience, vicarious experience, social persuasion, and psychological states (1 = strongly disagree to 6 = strongly agree). Britner and Pajares (2006) reported an α of 0.90 for mastery experience, 0.80 for vicarious experience, 0.88 for social persuasion, and 0.91 for emotional state. In our study, instrument yielded three subscales from factor analysis: perceived social supports, perceived social barriers, and emotional state. The mastery experience factor appears to be somewhat unclear; some items are found in the social persuasion factor and others in vicarious learning. The first subscale was the 21 items referring to “perceived social supports” (α = 0.95 in the present study), meaning the supports students have from significant others. The second subscale includes 11 items which measure variables that act as barriers to career development and is called ‘‘perceived social barriers’’ (α = 0.85). The last subscale, “emotional state” with five items (α = 0.86), measures students’ anxiety about science and technology subjects.

Four items from this scale which have been analyzed in detail conform one of the predictive variables: “students’ perception about parental academic expectations”: “My father believes that I have high abilities to get good marks in technology/computing subjects”; “My mother believes that I have high abilities to get good marks in technology/computing subjects”; “My father believes that I have high abilities to get good marks in science subjects”; “My mother believes that I have high abilities to get good marks in science subjects”.

Gender Roles Attitudes Scale. This scale consisted of 7 Likert-type items, also developed by the research team, for assessing the individual’s gender stereotypes in relation to science and technology studies. Participants rated each item on a 6-point scale ranging from 1 = strongly disagree to 6 = strongly agree. Alpha coefficient in the present sample was 0.85. We evaluated the content validity using Lawshe’s (1975) content validity ratio (CVR). Twenty-three specialists, all with expert knowledge about gender differences in a range of scientific areas, took part in this evaluation. CVR values ranged from 0.82 to 1.

Four items from this scale had a direct relation with the objectives of our investigation. They conform the second predictive variable: “students’ perception about parental gender stereotypes”: “My father believes that technology/computing is for boys”, “My mother believes that technology/computing is for boys”, “My father believes that science is for boys”, “My mother believes that science is for boys”.

Data were collected during a regular school day by the research team. A total of 45 public and private educational institutions participated in the study. After a brief presentation in which the researchers described the purpose of the study, the students were asked to fill out a questionnaire which took about 30-45 minutes to complete. All of them participated voluntarily, and no parents withheld their consent. The validation of the questionnaire, which has been published in Inda-Caro, Rodríguez-Menéndez and Peña-Calvo (2016), was created taking into account several scales.

In order to study the relationship between students’ gender and their perceptions of parental academic expectations and gender stereotypes about their performance in science and technology subjects, we carried out a descriptive analysis. We calculated chi-square (χ²), contingency coefficient (C), and standardized residuals (z _residuals) (H1). Secondly, we made an analysis of variance to determine the effect sizes of students’ gender on fathers’ and mothers’ beliefs about their abilities and performance in science and technology subjects (H1). Finally, we carried out structural equation modeling with IBM_AMOS22 software in order to test the theoretical model proposed in the state of the art. We aimed to assess the influence of students’ perception about parental expectations and gender stereotypes on self-efficacy beliefs, outcome expectations and interest, considering the students’ gender variable (H2, H3).

The percentage of missing cases in each of the variables of the study was analyzed. The percentage of missing data ranged from 1.9% to 4.7%. Besides, differences between missing/non-missing values cases in the predictor and criterion variables were studied. No significant differences were found in any of the variables (p > 0.05).

When boys and girls were asked if their father thought they had the qualities for success in different subjects, and taking into consideration the gender variable, we found that the relationship was statistically significant for all subjects: technology (χ² = 135.442, p < 0.001; C = 0.237, p < 0.001); science (χ² = 17.940, p < 0.01; C = 0.088, p < 0.01). Moreover, when they were asked about their mothers’ opinion in the same question, similar conclusions were obtained: technology (χ² = 137.029, p < 0.001; C = 0.236, p < 0.001); science (χ² = 29.027, p < 0.001; C = 0.111, p < 0.001) (Table 1).To sum up, boys and girls considered that their parents believed they had qualities for technology and science, but this conviction was always higher in boys, and these differences were more evident in technology than in science.

Table 1 Table 1 Distribution of observed frequencies and standardized residuals

G = Girls; B = Boys; DF = Definitely false; MF = Mostly false; SF = Slightly false; AT = Almost True; MT = Mostly true; DT = Definitely true. Z_r = Standardized residuals; fo = Observed Frequency; fe = Expected frequency. Positive values Z_r = fo > fe; Negative values Z_r = fo < fe

Two other items from the questionnaire allow us to enlarge upon this interpretation. Firstly, students were asked if their father/ mother believed that certain types of subjects were boys’ or girls’ “things”. It is interesting that students’ responses indicated a relationship between gender and this perception about fathers’ belief for all subjects: technology (χ² = 145.957, p < 0.001; C = 0.247, p < 0.001); science (χ² = 183.692, p < 0.001; C = 0.275, p < 0.001). Mothers’ answers confirmed the same relation: technology (χ² = 156.372, p < 0.001; C = 0.253, p < 0.001); science (χ² = 194.125, p = 0; C = 0.280, p < 0.001).

Again, boys tended to show a more stereotypical perception of their parents’ opinion towards science: boys scored higher than expected in the items where they were enquired if their father/mother considered technology to be a male thing (result also present for science), whereas girls did not report such a high perception (Table 1).

The normality criterion was confirmed in all variables except “My father believes that science is for boys” and “My mother believes that science is for boys”. These two items showed more than -1 and 1 skewness and kurtosis. On the other hand, the univariate homoscedasticity principle did not hold and Levene’s test showed significance in all criterion variables (p <0.001). However, the sample was large, and there was a balance among two groups; due to this fact, an analysis of variance and effect sizes was calculated with ƞ ² to evaluate differences between girls and boys in the criterion variables. In order to confirm that previous analysis was adequate, Man-Whitney test was run with these two items with a non – normal distribution.

As Table 2 shows, girls considered that their parents believed they had less ability than boys to perform well in technology subjects. Boys scored significantly higher (p < 0.001) in relation to fathers’ and mothers’ beliefs in all items except for parents’ beliefs about their abilities to get good marks in science subjects, where there were no significant differences. We can confirm that in their daughters’ and sons’ opinions, fathers and mothers believed certain subjects seem more appropriate for boys than for girls, and where boys are supposed to have more capability than girls.

Table 2 Table 2 Analyses of parental beliefs about students’ abilities *

p < 0.05

Prior to the path analysis, we examined Pearson correlations among factors, not only in girls but also in boys, as shown in Table III. This analysis was run employing maximum likelihood estimate method. Subsequently, diverse fit indices were evaluated; parsimony measures were calculated: χ² value, CMIN/DF (minimum of discrepancy function/degrees of freedom), were considered adequate when showing values below 2. Concerning comparative indexes: CFI (Comparative fit index) and TLI (Tucker-Lewis index) values were set above 0.95 in order to get a good fit and, finally, RMSEA (Root mean square error of approximation) values were set lower than 0.08 for an acceptable fit.

Table 3 Table 3 Pearson correlations, means and standard deviation in girls and boys

Note. Girls’ intercorrelations (n = 1197) are presented above the diagonal, and boys’ intercorrelations (n = 1162) below it. Means and standard deviations are presented in the vertical columns for girls and in the horizontal rows for boys.

Firstly, we wanted to launch the measurement model with the full sample to determine the fit to theoretical model proposed. The model showed in Figure 1 was tested, and the fit indexes were (χ² = 15.71; df = 5, p < 0.01), CMIN/DF = 3.14; CFI = 0.99; TLI = 0.99, RMSEA = 0.03). All indexes showed good values except for CMIN/DF, and the weights of students’ perception about parental expectations and gender stereotypes on outcomes expectations were almost zero or zero. Considering this finding, we also tried a more parsimonious model, without these paths. The resulting model was Model 0. As it can be seen in Table 4 the fit indexes were very optimal. Afterwards we tested the model for girls (Model 1) and for boys (Model 2). The measurement model fitted both for girls and boys (Table 4). A multiple-group analysis was also carried out to explore the model fit (Figure 2). Table 4 presents the fit indices for these alternative models taking into consideration the recommendation made by experts to report several indices of model fit (Hoyle, 1995). Model 3 depicted a good fit for gender variable (Model 3). Furthermore, the three models were considered to test the structural invariance of the model. The first model (constrained weight model) considered the same loads for girls and boys (Model 4) and revealed a very good fit. The comparison of model 3 and model 4, using χ² differences test, indicated that there were no differences (Δχ² = 8.05, p = 0.25). On the other hand, two other models were tested, one of them considering the intercepts equality (Model 5) and another one seeing the residual variances equality. These last models showed worse fit than Model 3. In addition, there were significant differences among them (Table 4). These analyses showed that structural invariance could not be observed, as long as significant differences were found between Model 3 and Model 5 and Model 6. However, the best fit was obtained for Model 3, in which there was no constrainment.

Table 4 Table 4 Fit indices for girls, boys and constrained models

χ² difference = the difference in χ² comparing the two models.

To start with, we must highlight that although the coefficients of the predictive variables (students’ perceptions about parental academic expectations and gender stereotypes) were not significant, it is worth considering the sign of the path. It can be stated that when a student perceived that his or her father did not show gender stereotypes, self-efficacy beliefs tended

to have a positive tendency. If we focus on mothers’ influence, we obtained that when girls and boys perceived gender stereotypes in their mothers, their self-efficacy beliefs also showed the same tendency. Moreover, self-efficacy beliefs tended to rise when boys and girls perceived good academic expectations in their fathers. On the other hand, when boys and girls perceived mothers’ positive academic expectations, self-efficacy beliefs seemed to have a negative tendency.

Figure 2 Figure 2 Path model depicting influence of perceived family expectations and gender stereotypes Note. Multiple-group path analysis by gender. Standardized weights are the paths coefficients. For girls, R ² self-efficacy = 0.4%; R ² Outcome expectations = 8% and R ² interests = 22%. For boys, R ² self-efficacy = 0.6%; R ² Outcome expectations = 9% and R ² interests = 22%. G = Girls; B = Boys (n= 2,359) * p < 0.05 **p < 0.01 *** p <0 .001

As shown in Figure 2, the complementary relation between fathers’ academic expectations and gender stereotypes, and mother’s academic expectations and gender stereotypes need to be reviewed since the correlations were not high. Despite this, it should be born in mind that the sign of the correlation between the perception of mothers’ academic expectations and gender stereotypes in boys was negative: when a boy perceived gender stereotypes in his mother, the perception of good academic expectations (good marks) seemed to be affected and decreased. A relation between the perception of parental gender stereotypes in boys and girls could be evidenced. This may reflect a coherence in the educational perspective of the couple in gender issues. The same situation could be shown in the academic expectations: when a boy or a girl perceived high parental expectations in science and technology, they perceived them simultaneously in their fathers and mothers.

As regards the relationship between boys’ and girls’ perceptions about parental academic expectations and gender stereotypes, the research has found little influence on self-efficacy beliefs (the percentage of explained variance was low). In contrast, self-efficacy beliefs had a determinant influence on students’ interest and outcome expectations in science and technology.

All the data from this research have been obtained from boys’ and girls’ answers. We are aware of the need to put these in context with fathers’ and mothers’ real beliefs in the qualitative phase of the investigation. In the future, it would be important to replicate the study with other samples in more ethnically and diverse geographical contexts, in order to explore if the results maintain the intensity or the interesting tendency shown in perceived academic expectations and gender stereotypes.

Finally, the cross-sectional nature of this research is a methodological limitation. The study can indicate statistical relationships between particular variables but cannot support causal inferences. It would also be important to analyze if there might be previous variables which could be configuring students’ perceptions and which have not been considered in the present study.