• motivation;
  • gender;
  • attitudes;
  • cognitive science;
  • international education


  1. Top of page
  2. Abstract
  3. Theoretical Background
  4. This Study
  5. Method
  6. Results
  7. Discussion
  8. References
  9. Supporting Information

The present study is based on the empathizing-systemizing (E-S) theory of cognitive science. It was hypothesized that the influence of students' gender on their motivation to learn science is often overestimated in the research literature and that cognitive style is more important for motivation than students' gender. By using structural equation modeling, and based on previous research, a precise causal model was formulated to test this hypothesis. Then, using multiple group confirmatory analysis, the model was tested in a cross-cultural context that included four countries—Malaysia, Slovenia, Switzerland, and Turkey—and 1,188 upper secondary students. Data were collected using standard questionnaires on cognitive style and motivation to learn science. The results showed full mediation of systemizing—the second dimension of the E-S theory—between gender and motivation. That is, gender had no direct impact on motivation, but systemizing explained 27% of the variation in students' motivation scores. The indirect impact of gender was significant but very low; it explained 1.5% of the variance, in favor of boys. Empathizing—the first dimension of the E-S theory—had no impact on students' motivation scores. This causal model proved to be similar (invariant) in all four cultures. The results suggest that considering students' cognitive style, instead of or in addition to their gender, could lead to a better understanding of students' motivation to learn science. Science teaching methods that support both cognitive styles—systemizing and empathizing—could enhance students' learning of science. © 2013 Wiley Periodicals, Inc. J Res Sci Teach 50: 1047–1067, 2013

Motivation to learn science is a matter of international concern (Fensham, 1995; Osborne & Dillon, 2008). It is known to be strongly associated with science achievement (e.g., Cavallo, Rozman, Blinkenstaff, & Walker, 2003) and with career choices (cf. Zeldin, Britner, & Pajares, 2008). The well-known “swing away from science” is a challenging problem in many countries (Osborne, Simon, & Collins, 2003).

Empirical evidence reveals many factors that contribute to low motivation in school science, such as the lack of perceived relevance, inadequate narrative quality in science education, lack of pedagogical variety, less engaging teaching than in other school subjects, male-oriented content, and an assessment system that encourages rote and performance learning rather than mastery learning for understanding (Osborne & Dillon, 2008). Other factors linked to declining motivation are changes in the classroom environment, less student-centered instruction, fewer classroom discussions and debates, and more learning, note-copying and adherence to textbooks (Vedder-Weiss & Fortus, 2011).

What is common among these findings is a focus on external factors (Byrnes, 1996) that affect students' motivation to learn science. Only a minority of studies have tried to explore internal student characteristics that may or may not motivate them to learn science. If they do, then their approach is mostly cultural (cf. Aikenhead, 2000), or value driven (Haste, 2004).

Compared to these classical attempts to understand motivation to learn science, the current study emphasizes a cognitive style concept, also called brain type (Baron-Cohen, Knickmeyer, & Belmonte, 2005), that has its foundations in cognitive science. It is grounded within a non-cultural paradigm, which suggests that human cognition is domain-specific. It also assumes a neurobiological background for cognitive dispositions (traits) and their cross-cultural stability (Hirschfeld & Gelman, 1994). Its theoretical framework is the empathizing-systemizing (E-S) theory (Baron-Cohen et al., 2005; Billington, Baron-Cohen, & Wheelwright, 2007).

E-S theory suggests that a systemizing cognitive style is an important and so far underestimated cognitive factor for motivation to learn science. This relationship is assumed to be fundamental in the sense that it underlies other factors related to motivation to learn science, like teaching quality, learning environment, cultural influences, and gender. The present study uses structural equation modeling to formulate a precise structural model based on previous research. Using multiple group confirmatory analysis, this model is tested in a cross-cultural context that includes four countries—Malaysia, Slovenia, Switzerland, and Turkey—and 1,188 upper secondary school students. In doing so, this study extends a series of previous studies that have been conducted with Swiss students (Zeyer, 2010; Zeyer, Bölsterli, Brovelli, & Odermatt, 2012; Zeyer & Wolf, 2010 see below). In this new cross-cultural context, the basic research question was whether the hypothesized causal relationship between gender, systemizing cognitive style, and motivation to learn science, which has already been found to be remarkably constant in different Swiss schools, is invariant in different cultures as well.

The focus of this study is on cognitive style and its relationship to motivation. However, since the distribution of cognitive style is gender-dependent according to E-S theory, gender is a factor that had to be considered. This is why, though the main part of the next section will be dedicated to introducing E-S theory, a subsection will discuss the literature on gender differences in motivation to learn science. Then, the relevant literature for the main hypothesis that empathizing and systemizing are predictors of motivation to learn science will be presented.

Theoretical Background

  1. Top of page
  2. Abstract
  3. Theoretical Background
  4. This Study
  5. Method
  6. Results
  7. Discussion
  8. References
  9. Supporting Information

Empathizing-Systemizing (E-S) Theory

The E-S theory was theoretically and empirically substantiated by Baron-Cohen et al. (2005). According to this theory, there are two fundamental psychological dimensions: systemizing and empathizing. These relate to the “consciousness of a physical world” and the “consciousness of a mental world,” respectively (Baron-Cohen, Wheelwright, Stone, & Rutherford, 1999).

Systemizing describes a person's ability to perceive “physical things” and understand these objects and their function in the context of a system. A systemizer's access to the world is organized in such a way as to identify details and their interconnections. For this reason, the gestalt perception is repressed. The goal of this cognitive dimension is to identify rules that determine a system and understand how to predict the behavior of a system.

Empathizing is the ability to identify and perceive mental states. It is concerned with understanding people and their psychological makeup. Empathizing involves a cognitive and an affective component (Baron-Cohen & Wheelwright, 2004; Davis, 1980). The cognitive component primarily includes consciously understanding other people's thoughts and feelings. The affective component includes emotional reactions to other people's mental conditions. Therefore, the affective component involves a response phenomenon, while the cognitive component brings observations into perspective from a rational distance.

The basic principle of E-S theory is that people use both of these psychological dimensions. However, one of them is generally predominant. A person who is more influenced by systemizing is described as having a systemizing cognitive style; those who are more influenced by empathizing have an empathizing cognitive style. People who are equally equipped with both dimensions are described as balanced. This is known as the E-S model. The E-S model is determined by two personal assessment questionnaires (see below), from which a systemizing quotient (SQ) and an empathizing quotient (EQ) that measure an individual's ability to use each of these cognitive dimensions can be calculated. Both questionnaires have been translated into different languages and have been extensively tested for validity and reliability (Baron-Cohen, Richler, Bisarya, Gurunathan, & Wheelwright, 2003; Baron-Cohen & Wheelwright, 2004).

An important result that has emerged from investigations of the E-S model is that—though most people display a balanced cognitive style irrespective of gender—on average women are empathizers (i.e., their normalized EQ is generally larger than their normalized SQ, E > S), and males are systemizers (S > E). Of course, there are also systemizers among women and empathizers among men (Baron-Cohen et al., 2005).

It was this result why gender had to be considered in our current study, although its focus is on cognitive style and its relationship to motivation. Indeed, research on the relationship of gender and science learning is ambiguous and inconclusive and asks for a change of paradigm that may be inspired by E-S theory.

Gender and Motivation to Learn Science

During early childhood and the elementary school years, boys seem to be more interested and motivated than girls to learn science-related topics (Alexander, Johnson, & Kelley, 2012). A significant gender difference among 11- to 13-year-old pupils has also been reported, with girls favoring languages and humanities, and boys favoring physical education and science (Osborne et al., 2003). However, overall, relatively little work has explored the development of science-related aspirations among children younger than 14 years (DeWitt et al., 2011).

However, for older students, a gender difference has not been confirmed. At the college level, Glynn, Taasoobshirazi, and Brickman (2007) found no relationship between gender and motivation to learn science. In PISA (2006), gender differences varied unsystematically from country to country (Bybee, 2012). Likewise, in a study with 14- to 16-year old students, Bryan, Glynn, and Kittleson (2011) found that boys and girls scored similarly on motivation to learn science, with one exception: Among students who did not aspire to advanced college placement, boys scored higher than girls in intrinsic motivation. However, the size of this effect was small (Bryan et al., 2011). Similarly inconclusive results have been obtained for motivational sub-dimensions like self-efficacy (Britner, 2008; Glynn, Taasoobshirazi, & Brickman, 2009), confidence (Britner & Pajares, 2006; Kahle, Parker, Rennie, & Riley, 1993), self-determination (Glynn et al., 2009; Glynn, Brickman, Armstrong, & Taasoobshirazi, 2011), and examination anxiety (Britner, 2008).

Thus, remarkably, in a review on the “girls and science” issue, Brotman and Moore (2008) concluded that a “paradigm shift in research on gender and science” is needed, that recognizes the “existence of elements of both genders in all of us” (p. 876). Referring to Gilbert and Calvert (2003), they argued “that masculinity and femininity are cultural constructions that have come to be seen as mutually exclusive—associated exclusively with either males or females—when in fact traits culturally associated with masculinity can be part of females' identities and vice versa” (ibid.). The basic idea of our study is that that cognitive style is such a trait.

Cognitive Style and Motivation to Learn Science

This idea has its origins in the field dependence/field independence cognitive style theory (Zhang & Sternberg, 2006). People who tend to first view the world in terms of relationships and only later in terms of categories are said to be field-dependent. In contrast, people who tend to use an articulated style, who break up the world into abstract categories which can then be organized into larger chunks, are said to be field-independent (Lavenda & Schultz, 2007). Males are generally more field-independent than females (Witkin, Dyk, Faterson, Goodenough, & Karp, 1962). It has been shown that field-independent students achieve higher levels in mathematics and physical sciences, and field-dependent students have higher achievements in social sciences, literature, and human services (Morgan, 2001; Varma & Thakur, 1992).

Baron-Cohen et al. (2005) refer to field-independence as one element of a systemizing cognitive style. However, E-S-theory is more encompassing and, in particular, conceived as culturally invariant (Wakabayashi, Baron-Cohen, Uchiyama, Yoshida, & Wheelwright, 2007). This is in stark contrast to the field dependence/field independence approach, which is situated within a cultural paradigm (Faiola and Matei, 2005; Zhang & Sternberg, 2006).

Billington et al. (2007) showed that a systemizing cognitive style was a predictor for entering sciences rather than humanities. Indeed, a previous study conducted by Wheelwright et al. tested the cognitive styles of science and humanities students. They found that physical sciences students scored significantly lower on empathizing and significantly higher on systemizing compared to humanities students, independent of gender (Wheelwright et al., 2006). Billington et al. confirmed these results using both questionnaires and performance tests. They concluded that “irrespective of their gender, if an individual's systemizing is at a higher level than their empathizing, it is this profile that leads them into disciplines that require an analytical style to deal with rule-based phenomena” (Billington et al., 2007, p. 266).

This Study

  1. Top of page
  2. Abstract
  3. Theoretical Background
  4. This Study
  5. Method
  6. Results
  7. Discussion
  8. References
  9. Supporting Information

A limitation of both Wheelwright et al. (2006) and Billington et al. (2007) is that they used the choice of discipline as a rough measure of motivation to learn science. However, it is well known that career choices, especially for girls, are affected not only by motivational disposition, but also by families, teachers, peers (Zeldin & Pajares, 2000), and societal role models (Jacobs & Bleeker, 2004; Simpson & Oliver, 1990). Additionally, by using a binary variable (physical sciences vs. humanities), the statistical method used in these studies was limited to inferential statistics and logarithmic prediction, which do not allow for causal inferences. To overcome these limitations, we introduced Glynn and Koballa's Science Motivation Questionnaire (SMQ; Glynn & Koballa, 2006), and used structural equation modeling (SEM).

Science Motivation Questionnaire (SMQ)

SMQ scale is based on social cognitive theory, the goal of which is to “explain how people acquire competencies, attitudes, values, styles of behavior, and how they motivate and regulate their level of functioning” (Bandura, 2001, p. 54). Thus, motivation is defined as “the internal state that arouses, directs, and sustains students' behavior towards achieving certain goals. In studying the motivation to learn science, researchers attempt to explain why students strive for particular goals, how intensively they strive, how long they strive, and what feelings and emotions characterize them in this process” (Glynn et al., 2009, p. 128). The motivation to learn, as conceptualized in social cognitive theory, is multi-dimensional (Glynn et al., 2011). It consists of six important motivational constructs, reflected by the six SMQ subscales. These include intrinsic and extrinsic motivation, relevance to personal goals, self-determination, self-efficacy, and assessment anxiety (Glynn & Koballa, 2006).

We used the SMQ in our studies for three reasons. Most motivation questionnaires assess single, specific components of motivation, such as self-efficacy (cf. Britner, 2008) or interest (cf. Linn & Hyde, 1989). The SMQ is ideally suited for our purposes because it combines a number of key components in a single scale. Other existing questionnaires also do this; however, few studies establish their validity (cf. Tuan, Chin, & Shieh, 2005). Finally, our previous studies (Zeyer, 2010; Zeyer et al., 2012; Zeyer & Wolf, 2010) used the SMQ so that the results could be compared with those obtained by Glynn et al.

Previous Studies and Research Concept

Using the SMQ, and applying the concept of mediation of structural equation modeling (Iacobucci, 2010), the basic idea of our research could easily be translated into the stringent hypothesis that cognitive style is a mediator between gender and motivation to learn science. After two pilot studies (Zeyer, 2010; Zeyer & Wolf, 2010), the hypothesized structural model was tested on a sample of 500 Swiss upper secondary school students (Zeyer et al., 2012). SEM confirmed that systemizing fully mediated the relationship between gender and motivation to learn science. In other words, gender had no direct effect on motivation, but had an indirect impact via systemizing. The impact of a systemizing cognitive style on motivation was fairly high, explaining about 25% of the variance on the SMQ. However, there was no relationship between empathizing, the other cognitive dimension of E-S theory, and motivation to learn science.

The current study uses these results and the structural model. It was tested in a cross-cultural context that included four countries—Malaysia, Slovenia, Switzerland, and Turkey—and 1,188 upper secondary school students. In this new cross-cultural context, the basic research question was whether the causal relationship between gender, systemizing cognitive style, and motivation to learn science, which was found to be remarkably constant in different Swiss schools, is also invariant across different cultures.

Multiple group comparison in SEM provides a powerful statistical instrument to deal with cross-cultural issues (Steinmetz, 2011). Testing our basic hypothesis was equivalent to investigating three hierarchically ordered research questions that directly translate into three escalating SEM tests of cross-cultural invariance. The three research questions (RQ) and the corresponding SEM tests are as follows:

  • RQ 1: Can the model be confirmed with a sample that includes other cultural contexts in addition to the Swiss context? In the SEM framework, this means testing the structural invariance of the model.
  • RQ 2: Can the causal relationships between the variables in the model be compared across cultures, and, if so, do they have a similar pattern in each culture? In the SEM framework, this means testing the metric invariance of the model across cultures and equality of the impact factors between variables.
  • RQ 3: Can the values of the endogenous variables be compared across cultures, and, if so, how do they vary across cultures? In the SEM framework, this means testing the scalar invariance of the model across cultures and inequality of the endogenous variables.


  1. Top of page
  2. Abstract
  3. Theoretical Background
  4. This Study
  5. Method
  6. Results
  7. Discussion
  8. References
  9. Supporting Information


Our intention was to find researchers from countries with different cultures who were willing and able to take part in this study. The student samples had to be upper secondary school level, and all selected students had to be enrolled in general science education that included the three science subjects, biology, chemistry, and physics. Ultimately, four countries became involved: Malaysia, Slovenia, Switzerland, and Turkey. The assumption that these four countries are culturally different in ways that are relevant to this study is supported by Olsen and Lie (2011) and Kjaernsli and Lie (2011). These researchers used data from PISA (2006) and identified clusters of countries that were culturally related in their profiles of interest and motivation in science. They found that Central/South Asian countries, German countries, and Slavic countries formed different cultural groups with high internal consistency. Turkey displayed a moderate correlation with the group of Central/South Asian countries (Kjaernsli and Lie, 2011; Olsen & Lie, 2011).

It is beyond the scope of this article to provide an exhaustive description of cultural and educational differences and similarities. However, Supporting Information accompanying the online article is available. It provides a short general description of each country, followed by a comparison using aspects as identified in the reports of the United Nations Development Program (2011), and a description of each country's school system, and of its science education.

In Malaysia, upper-level students from three secondary schools in the state of Penang participated. These schools were selected after permission was granted from relevant authorities. Located in urban, suburban, and rural areas, they were assumed to be representative of Malaysian secondary schools. Students in these schools were from various racial, religious, and social backgrounds, reflecting typical Malaysian secondary school students.

In Slovenia, secondary school students from four schools in central Slovenia participated. All participants were selected from general upper-level secondary schools, and are therefore considered to be representative of typical Slovenian secondary school students. Parental permission was obtained for all participating students.

The Swiss subsample consisted of students from a high school in the central part of the country. Consent was obtained from the school's board. As every student, in principle, has access to every school in Switzerland, this school was assumed to be representative of a common Swiss social background.

The Turkish sample consisted of students enrolled in a general public high school in Ankara. A general public high school was selected in order to be representative of Turkish secondary school students. The school was convenient and had a large student body, and the school administration was willing to help with data collection.


The students were visited in their classes where they were required to fill in the questionnaires. They were informed that a study was being conducted to further understand students' motivation to learn science. The general research conditions were presented, and the questionnaires were distributed. The students had a break between each questionnaire, and the completed questionnaires were then collected. The same process was used for each class.

In each country, adapted versions of the EQ, SQ, and SMQ questionnaires were used. Question wording was tested in a pilot study in each country. Tests of cross-cultural validity and reliability of the measurement instruments were part of the structural-modeling process (see below).

Questionnaire Part A

Part A of the questionnaire consisted of Baron-Cohen's SQ and EQ questionnaires (Baron-Cohen, 2003; Baron-Cohen & Wheelwright, 2004). These consist of 40 cognitive and 20 control (i.e., items that are not included in the final score) forced-choice items. The SQ includes questions such as “If I had a collection (e.g., coins, CDs, stamps), it would be highly organized,” and “When I learn a language, I become intrigued by grammatical rules.” The EQ includes items such as “I am good at predicting what someone will do” to measure cognitive empathy, and “I usually stay emotionally detached while watching a film” to measure the affective component of empathy. In both the EQ and SQ questionnaires, participants were asked to select “definitely agree,” “slightly agree,” “slightly disagree,” or “definitely disagree” for each item. Approximately half the items were reverse-scored to avoid response bias.

Questionnaire Part B

Part B of the questionnaire was the 30-item Science Motivation Questionnaire SMQ (Glynn & Koballa, 2006). Typical items for this questionnaire were “I enjoy learning science,” “Earning a good science grade is important to me,” or “I am confident I will do well in the science labs and projects.” Students responded to each of the 30 randomly ordered items on a five-point Likert-type scale, ranging from 1 (never) to 5 (always). Items on anxiety about science assessment were reverse-scored such that a higher score on this component corresponds to less anxiety.

The Structural Modeling Procedure

Due to the complexity of the empirical test, a two-step process was used to confirm the first-order model (Jöreskog, 1993). All the estimates were produced using AMOS 16.0 (Airbuckle, 1997) and the maximum-likelihood estimation method. As a first step, the EQ, SQ, and SMQ measurement models were tested through confirmatory factor analysis. The full structural model was tested directly in the second step. The structural model reflected the core hypothesis of this study, that the impact of gender on motivation to learn science would be fully mediated by SQ. In other words, gender has only an indirect influence on motivation to learn science via SQ. However, at the start of the testing process, the basic structure of the full model also included EQ scores. The role of EQ was assumed to be symmetrical to the role of SQ, even though the Swiss model showed that EQ had no impact on SMQ. Indeed, this was an important aspect of the first research question in this study: Can the lack of impact of EQ on SMQ be confirmed across cultures, or is it idiosyncratic to the Swiss culture?

Item Reduction

To include SQ, EQ, and SMQ as single latent variables in the model, each underwent substantial item reduction. Given the sample size and the number of SMQ items, a final number of 12–15 items was deemed adequate for the structural model (Kim, 2005).

Parceling is a widely used and theoretically and experimentally well-sustained approach for item reduction (Bandalos & Finney, 2001). Parceling can be particularly advantageous over using the original items when estimating large numbers of items is likely to result in spurious correlations. Parceling may also be advantageous when items from a large pool will likely share specific sources of variance that may not be of primary interest. It is particularly adequate if the aim of the study is to examine the nature of a set of constructs, rather than to understand the structure of a set of items (Little, Cunningham, Shahar, & Widaman, 2002). All of these considerations apply to our research project.

An important issue in parceling is item dimensionality. EQ and SQ have been conceptualized as strictly unidimensional constructs, and this unidimensional structure has been confirmed in extensive psychometric testing (Baron-Cohen et al., 2003; Baron-Cohen & Wheelwright, 2004). In terms of the SMQ, our own research has revealed that this scale is a global measure that does not show unidimensionality of subdimensions (Zeyer et al., 2012).

Therefore, two different strategies were used for EQ and SQ on the one hand, and SMQ on the other. Given the unidimensional item sets in the EQ and SQ questionnaires, a random assignment method was used (Little et al., 2002). Each EQ and SQ item was randomly assigned to three parcels without replacement. This method is appropriate when items are drawn from a common pool, as is the case for items on the EQ and SQ questionnaires.

For the multidimensional SMQ item set, a domain-representative approach (Kishton & Widaman, 1994) was used. In this method, parcels are constructed by combining items from different dimensions into item sets. This approach accounts for multidimensionality by creating parcels that encompass not only common variance, but also unique facets of the multiple dimensions. Every parcel includes one item from each dimension, such that the construct's overall complexity is mirrored in every parcel. Given that each of the six sub-domains of the SMQ consists of five items, the domain-representative approach produced five parcels of six items, with each item in a given parcel drawn from a different sub-dimension.


  1. Top of page
  2. Abstract
  3. Theoretical Background
  4. This Study
  5. Method
  6. Results
  7. Discussion
  8. References
  9. Supporting Information

Descriptive Measures

A total of 1,431 students from four countries (Malaysia, Slovenia, Switzerland, and Turkey) were studied. Data were excluded if a student had not answered every question of all scales, or if answers could not definitively be identified. After this raw data cleaning, the sample included 1,188 students (243 omitted cases, 16.98%). The mean age was mage = 16.59 years (SD = 0.905). Table 1 shows that the sample includes more girls (712, 59.9%) than boys (476, 40.1%; see limitations section below).

Table 1. Distribution per country and gender
Country MaleFemaleTotal
 % within country44.255.8100.0
 % of total10.613.424.0
 % within country32.967.1100.0
 % of total918.427.4
 % within country45.654.4100.0
 % of total10.012.422.0
 % within country39.160.9100.0
 % of total10.416.226.7
 % within country40.159.9100.0

Additional descriptive measures, including SQ, EQ, and SMQ, and the SMQ sub-scale scores by gender and by country, are available as Supporting Information accompanying the online article.

The Full Structural Model

The first step of the SEM analysis was to establish and test the full structural model that would be used for testing cross-cultural invariance. The confirmatory analysis of the three parceled measurement models of the SQ, EQ, and SMQ each showed a good fit for the overall sample. For reasons of space, these results are omitted. Figure 1 presents the basic, full structural model that includes all involved factors for gender, EQ, SQ, and SMQ.


Figure 1. Full structural model including empathizing quotient. Standardized estimates. ***p < 0.001.

Download figure to PowerPoint

In this model, all factor loadings were highly significant (p < 0.001) except for the direct impact of gender on SMQ and the direct impact of EQ on SMQ, which were both very low and not significant. Therefore, as in the Swiss model, it made sense to omit the latent variable of EQ and remove the factor loading of gender on SMQ. Both steps improved the goodness of fit of the model. The final structural model is shown in Figure 2.


Figure 2. Definitive full structural model. Standardized estimates.

Download figure to PowerPoint

Given the good fit (see technical details in the following paragraph), the high amount of multiple square correlations and the high, equally distributed and highly significant factor loadings, this model was considered appropriate as a full model for the presented theory. It reflected the theoretical background of the E-S theory. It confirmed the hypothesis that motivation to learn science was highly influenced by the systemizing cognition, while the EQ, representing the empathizing cognition, had no significant influence on motivation. It also confirmed (through the SQ) the hypothesis that gender indirectly affects the motivation to learn science.

The factor loadings of the measurement instruments were statistically highly significant (p < 0.001), and the corresponding signs concurred with the hypotheses. The standardized estimates, from 0.74 to 0.89, confirmed the formal validity of the individual items (c.f. Bollen, 2002). The explained variance of the items lay between 0.54 and 0.79, which is a highly acceptable range of magnitudes. Descriptively, the model worked very well, which was first indicated by a highly acceptable goodness-of-fit index (GFI) of 0.982. Secondly, the baseline comparison (CFI) was 0.987. From an inferential point of view, the model was compatible with the data (CMIN/DF = 3.876). Finally, RMSEA = 0.049 and PCLOSE = 0.530 indicated a very good fit (for fit measures, see Airbuckle, 1997, 551ff).

The standardized regression weight of the SQ on the SMQ (.52) was considerably high. Thus the explanatory power of the model was high. The impact of the SQ explained 27% of the variation of the SMQ.

There was also a highly significant factor loading of gender on the SQ. The standardized regression weight was negative (−0.29) because this variable represented “female” with the value 1 and “male” with the value 0. Therefore, in this model, males had a higher SQ (which aligned with the results in the descriptive part of this study). The impact of gender explained 9% of the variation of the SQ.

The indirect effect of gender was calculated by bootstrapping. The standardized indirect effect of gender on the SMQ (−0.152) was highly significant but very small. Since there was no significant direct loading of gender on the SMQ, the effect of gender on the motivation to learn science was strictly indirect and minimal.

In order to test configural invariance, the method of simultaneous multiple group comparison was used, which allowed for examination of the causal relations in both the measurement and the structural model (see Byrne, 2010). A configural model (with no equality constraints imposed) was created by introducing the four data sets of the four different cultures. As expected, it reproduced the goodness-of-fit results of the basic model (CFI = 0.993, RMSA = 0.019, PCLOSE = 1). It was concluded that the full model showed configural invariance, that is, the four cultural groups showed a comparable model structure.

In order to test metric invariance, metric measurement and (structural) models were created and tested for invariance, that is, it was assumed that the factor loadings in all measurement models were the same across cultures. The fit of the measurement model was consistent with that of the configural model (CFI = 0.985; RMSEA = 0.025). The difference in the comparative fit ΔCFI = 0.008 was smaller than 0.01, which accepted the invariance hypothesis (Cheung & Rensvold, 2002). Metric invariance allows for the comparison of impact factors across cultures. Testing for inequality hypotheses showed that differences between the impacts of gender on the SQ across cultures were not significant. The same held true for the impact of the SQ on the SMQ; it was the same across all cultures. It was concluded that the model showed metric invariance with equal impact factors across all four cultures.

In order to test scalar invariance, scalar measurement models were created and tested for invariance, that is, it was assumed that in addition to the factor loadings, the intercepts in all measurement models were the same across cultures.

The fit of the scalar model was not consistent with the fit of the configural model (CFI = 0.911; RMSEA = 0.056). Therefore, the intercepts were released in both SQ and SMQ, for one resp. three indicators. The released model presented a much better fit (CFI = 0.979, RMSEA = 0.028). The CFI difference test yielded ΔCFI = 0.006, which accepted the hypothesis of partial scalar invariance. This allowed for comparison of factor loadings and latent means between the four groups (Byrne, Shavelson, & Muthen, 1989; Steenkamp & Baumgartner, 2000). Table 2 shows the squared multiple correlations, the implied means of the SMQ, and the means of the SQ for all four cultures.

Table 2. Squared multiple correlations (SMC) and the implied latent means of the SQ and the SMQ
 SMC SQSMC SMQLatent Mean SQLatent Mean SMQ

There was no significance for the difference of the mean SMQ across cultures. The only significant difference was found between the mean SQs. The Swiss mean SQ was the lowest, followed by the mean SQ in Slovenia (p < 0.01). The mean SQs of Turkey and Malaysia were higher, but equal. It was concluded that the full model displayed partial scalar invariance and that the means of the latent variables were the same across cultures except for the means of the SQ in the different countries.


Three statistical limitations must be discussed in the context of this study. First, there is ongoing discussion about whether multicollinearity is a problem in SEM. Multicollinearity is a well-known statistical phenomenon in multiple regression models where two or more predictor variables are highly correlated. In that case, the coefficient estimates may change erratically in response to small changes in the model or data. It is still unclear whether multicollinearity is a problem in SEM. Many researchers assume that structural equation models are robust against this type of phenomenon or that they can even remedy multicollinearity problems (cf. Malhotra, Peterson, & Bardi Kleiser, 1999). In any case, Monte Carlo simulations have shown that effects of multicollinearity are not expected if correlation values are less than .5, internal reliability of the variables involved is high (i.e., Cronbach's alpha > 0.8), and sample size is large (i.e., >300; Grewal, Cote, & Baumgartner, 2004). Our study meets these conditions.

Second, as previously noted, the sample included more girls than boys (Table 1). The gender ratio was nearly identical before and after the cleaning process (874 boys, 38.9%, vs. 874 girls, 61.1%), which indicates that the cleaning process did not result in a biased sample. Rather, the sample reflects a trend observed in all OECD countries, where more young women tend to graduate from general upper secondary school programs than young men. In 2009, the average OECD graduation rate from general programs was 55% for young women and 43% for young men. In Eastern European countries, such as Poland, the Czech Republic, and Slovenia, young women outnumber young men as graduates by at least three to two (OECD, 2011). We decided that the gender ratio of students attending the investigated schools should be represented in the sample because the focus of this study is not gender specifically, but rather the influence of cognitive style on motivation to learn science. Given the results of the pilot studies, we did not expect that the structural model or its factor loadings would be biased by the gender ratio. This was confirmed by the structural model and its robustness across cultures, which showed that the impact of gender was low and the impact of systemizing cognitive style on motivation was stable and high, independent of the gender ratio in each country. Moreover, reanalysis of the data in a classical partial correlation framework controlling for gender (gender affected the latent variables for which invariance was tested) confirmed the stability of the statistical results across gender.

Third, it must be emphasized that descriptive measures of SQ, EQ, and SMQ (available as Supporting Information accompanying the online article) are provided for an overview in terms of conventional scores, and should only be used in combination with SEM results. Since the structural equation model in this study does not include the SMQ sub-scales, inferences about cultural differences on SMQ sub-dimensions are speculative. Traditional inferential statistics in a cross-cultural context are seriously flawed when Likert-type scales and aggregated items are used to compute composite scores. Nevertheless, many researchers still rely on these traditional methods (e.g., analysis of variance [ANOVA] and t-test) to compare mean composite scores across cultural samples (Steinmetz, 2011). However, this approach ignores differences between observed and latent variables. Group differences for an observed composite score can be solely attributed to a latent mean difference when intercepts and item loadings are invariant (Steinmetz, 2011). SEM allows for a combination of estimation of latent means (and their comparison across groups) and analysis of factor structure (Ployhardt & Oswald, 2004). Therefore, multiple-group confirmatory factor analysis (Jöreskog, 1993) was the method of choice for the present study.


  1. Top of page
  2. Abstract
  3. Theoretical Background
  4. This Study
  5. Method
  6. Results
  7. Discussion
  8. References
  9. Supporting Information

The results of the testing process are consistent with the theoretical background of E-S theory, and confirm the results of the Swiss studies (Zeyer, 2010; Zeyer et al., 2012; Zeyer & Wolf, 2010) in a cross-cultural context. The goodness of fit for the structural model is excellent. Moreover, the model displays configural, metric, and partial scalar invariance across all four cultures (cf. Steinmetz, 2011). These findings provide evidence for answering the three research questions.

Based on the configural invariance of our model, we can conclude that the model indeed reflects a cross-cultural pattern; in other words, the model represents a causal structure that is invariant across all four cultures. All the salient aspects of the Swiss studies were consistently reproduced: the direct impact of SQ on motivation, the absence of any impact of EQ on motivation, and full mediation of the relationship between gender and SMQ by SQ.

Because the full model showed metric invariance, the impact factors involved can be numerically compared. SEM showed that they are equal across all four cultures. The impact of SQ on motivation to learn science is fairly large. On average, it explains 27% of the motivational variance. The more a person systemizes, the greater their motivation to learn science. EQ, however, does not influence motivation to learn science.

Gender has a negative impact on SQ, and explains 9% of the SQ variance. The effect is negative because the female gender was coded as +1. On average, female students score lower on SQ than male students—a finding in agreement with E-S theory (Baron-Cohen, Wheelwright, Burtenshaw, & Hobson, 2007). A classical t-test (p < 0.01, t-test for equality of means, two-tailed) revealed that the size of the effect of gender on SQ is d = 0.56, which is consistent with the effect sizes found in literature (0.44, cf. Nettle, 2007)

Gender has only an indirect (mimic) effect on motivation to learn science. It is negative and remarkably small (though significant). It explains only 1–2% of the motivational variance in all four cultures. If a student is a strong systemizer, they will show high motivation to learn science, almost independent of their gender. Since men tend to be stronger systemizers than women, men tend to be more motivated to learn science. However, if a man is a low systemizer, he will be less engaged in science despite his gender, and vice versa.

Because the structural model displays also scalar invariance, the comparison of latent means across cultures is possible. Mean SQ for the Slovenian and Swiss students were equal, and lower than mean SQ was observed for the Turkish and Malaysian students (which, again, were equal). This is consistent with the fact that the Turkish and Malaysian samples included a higher proportion of males than the Slovenian and Swiss samples.

It is important to discuss these findings in a broader context. The lack of a direct impact of gender on motivation supports the findings of Bybee (2012), Glynn et al. (2007), and Bryan et al. (2011). All these studies took a wide range of motivational aspects into account, and none of them found significant motivational differences between genders.

The positive impact of systemizing on motivation to learn science reflects the intrinsic systematic approach of science to phenomena in the material world (Searle, 2004). It is crucial, however, to note that the disposition of systemizing has broader applications beyond the fields of science and technology, because systemizing can be applied to systems that can be found in many areas of daily life, such as, music, language, collections of objects, etc. (Baron-Cohen et al., 2005). Conversely, a truly scientific approach to nature includes more than systemizing, such as aspects of explanation or finding consensus in the scientific community (cf. Cobern & Loving, 2001). Therefore, it is reasonable to assume that a systemizing cognitive disposition fosters motivation to learn science (as the structural model suggests), rather than vice versa.

In contrast, empathizing had no effect on motivation in the structural model. This does not contradict previous research by Wheelwright et al. (2006) and Billington et al. (2007) that showed that, on average, humanities college students were empathizers. It would be inappropriate to conclude from this that empathizing should have a negative impact on motivation to learn science. For example, students who choose an STEM subject as one of their favorite subjects at the secondary educational level may nevertheless chose not to study STEM at the tertiary level (Holmegaard, Madsen, & Ulriksen, 2012). However, our results indicate that a converse effect, namely, a positive influence of empathizing on motivation, does not exist either.

It seems remarkable that the numerical pattern of causal factors is the same across all four countries, since the style of science teaching is very different in each of these countries. In Slovenian upper secondary school science education, frontal teaching dominates (EURYDICE, 2011). Swiss science education emphasizes hands-on science and experiments (Stern, Metzger, & Zeyer, 2009). Turkish science education leans toward a transmissive teaching style and infrequent use of science laboratories (Özden, 2007). Finally, in Malaysia one-way instruction techniques, rote learning methods, and teaching to test are common (Syed Zin & Lewin, 1993; see also Supporting Information). Nevertheless, we found no significant cross-cultural differences between these countries.

Even the SMQ means did not differ significantly between the four cultures. This result is rather unexpected. Indeed, Kjaernsli and Lie (2011) were able to show significant motivational differences in PISA (2006) for different clusters of countries, some of them resembling the four countries involved in this study in many aspects. However, since their scale measured “future-oriented motivation to learn science,” their results may not be strictly comparable to ours. More research is needed to better understand cross-cultural differences in science motivation. Our finding that the SMQ demonstrated scalar invariance across the four countries suggests that the SMQ is a reliable and valid instrument for further cross-cultural motivational research.

Cross-cultural Invariance of SQ

One of the most salient results of our study is the cross-cultural invariance of SQ and its impact on motivation. This has methodological and conceptual implications.

On a methodological level, our results are the statistically strictest confirmation of cross-cultural invariance so far. To our knowledge, this is the first cross-cultural test of the E-S theory using multiple group comparisons. An earlier study (Wakabayashi et al., 2007) used traditional inferential statistics (see limitations above). From a methodological point of view, metric invariance in multiple group comparisons is a fairly strong assertion, suggesting that the SQ questionnaire items operationalize a construct that is stable in all of the cultures tested. Indeed, a system in the E-S context is defined as an entity that takes inputs which can then be operated on in variable ways to deliver different outputs in a rule-governed manner (Baron-Cohen et al., 2003). Defined in this general way, systems can be found in various contexts (cf. Baron-Cohen, 2009). We interpret the metric invariance of the SQ construct as confirmation of the culturally independent epistemic core of this definition.

On a conceptual level, as previously noted, Baron-Cohen et al. rely on biomedical reasoning to explain the cross-cultural invariance of SQ and its dependence on gender. They summarize a large body of empirical research findings not only in psychology, but also in various bio-medical disciplines, such as neurology, anatomy, and endocrinology (Baron-Cohen et al., 2005).

However, (neuro-)biology cannot explain our central finding, that the impact of SQ on motivation is stable across cultures. We interpret this result in terms of an often neglected aspect of E-S theory. Indeed, researchers often overlook that, in addition to the epistemological dimension (“systemizing” and empathizing”), the theory includes an ontological dimension that differentiates between “physical things” as the objects of systemizing, and “mental states” as the goal of empathizing. Referring to philosophy of mind (cf. Searle, 2004), Baron-Cohen et al. (1999) define “physical things” as phenomena with an objective (“third-person”) ontology, while mental things are intentional phenomena with a subjective (“first-person”) ontology. The differentiation between third- and first-person ontology is universal (cf. Searle, 2004). We argue that it is this ontological differentiation that contributes to the stability of our overall model in a cross-cultural context.

More empirical and theoretical research is needed to better understand the conceptual essence of this new approach. Nevertheless, in the following section, we would like to infer some implications of our structural model for science teaching, with due caution.

Implications for Science Teaching

Firstly, our structural model suggests that emphasizing gender in motivational contexts of science teaching may be misleading, because its impact is only indirect and small. For example, various studies have tried to characterize girls who are motivated to learn science. Mujtaba and Reiss (2012) investigated the motivation of girls who intended to study physics post-16 against the gender trend in their classes. Archer et al. (2012) studied girls who strongly identified with science and science learning. Yet, all these girls may be high systemizers, and thus motivated to learn science independent of their gender. Indeed, Mujtaba and Reiss (2012) found few motivational differences towards science subjects between highly motivated girls and boys.

Likewise, many other factors related to science motivation may appear differently in the context of cognitive style. For example, autonomy, relatedness and belonging (Andersen & Nielsen, 2011), or the use of living animals in science lessons (Wilde, Hussmann, Lorenzen, Meyer, & Randler, 2012) may be good motivators for empathizers (who are interested in mental states), but less suitable for systemizers. Different utilities (Bøe, 2012) and school types (Gill & Bell, 2011) may indeed influence motivation to learn science, but through cognitive style, not gender.

According to our structural model, high systemizers are highly motivated to learn science. They are the non-cultural pendant to the potential scientists identified within a cultural paradigm by Aikenhead and colleagues—students that enjoy a smooth transition into the culture of science, without much help from teachers or school culture (Aikenhead, 2000). They may be those students who choose science for identity reasons, such as interests, self-realization, and fit with personal beliefs (Bøe, 2012). In the Swiss studies, we found that only 5% of our sample were high systemizers (Zeyer, 2012a); that is, high systemizers represent a relatively small proportion of an average upper secondary school student population.

Low systemizers, regardless of their gender, seem to be the important challenge to science teaching, because our model suggests, that, by their very cognitive disposition, they are less motivated to learn about physical things and are weak in systemizing. Low systemizers may be those students who perceive STEM as stable, rigid and fixed, and too narrow a platform for developing and constructing desirable identities, and consequently avoid these subjects (Holmegaard et al., 2012). Our structural model suggests that several strategies identified in the literature may be particularly suitable for improving low systemizers' motivation to learn science.

First, ontologically motivated strategies could be based on the assumption that low systemizers are alienated from physical things because they have had few opportunities to come in contact with them. Lack of opportunity is mostly seen in the context of gender (Alexander et al., 2012), but our findings suggest that it might a problem for empathizing boys as well. Notice that, when providing such opportunities, the fact that low systemizers do not have an intrinsic affinity for physical things must be taken into account. A range of educational strategies identified in the literature may help these students to overcome their reluctance, like special teaching activities, labs, field trips, collaborative projects (Bryan et al., 2011), fostering autonomy, relatedness and belonging (Andersen & Nielsen, 2011), preventing feelings of disgust and rejection (Randler, Hummel, & Wüst-Ackermann, 2012), knowledgeable, inspiring, enthusiastic, and caring teachers (Bryan et al., 2011), and family and friends that encourage them to study science (Mujtaba & Reiss, 2012). It can be assumed that these strategies, all of them targeting positive intentional dispositions towards science learning, are particularly helpful for low systemizers (not just girls), and are less important for high systemizers (see below).

Second, on an epistemological level, our structural model suggests that improving systemizing itself in low systemizers may also help. Low systemizers may be those students (not only girls) who associate science with “cleverness” and masculinity, and find physics a particularly difficult subject (Gill & Bell, 2011). Simple and effective teaching practices (Logan & Skamp, 2012) that focus on rule-guided, systematic thinking may foster systemizing ability. Hands-on science (Bryan et al., 2011) may turn out to be another particularly beneficial approach for empathizers, as this type of science teaching embodies systems in experiments and makes them accessible through activity (Swarat, Ortony, & Revelle, 2012).

Finally, a third approach to improving students' motivation to learn science questions our finding that empathizing cognitive style had no effect on motivation to learn science. Is it that science as a subject does not affect empathizing? Or is it the way science is taught in schools? How must science teaching be structured to meet the needs of empathizing students? From an ontological point of view, the solution seems to be simple: science teaching for empathizers must involve “mental states.” But what does this mean?

One strategy identified in the literature is to involve socio-scientific issues in science teaching, because, as Sadler (2004) and others point out, socio-scientific issues involve economic, social, political, and/or ethical considerations. Thus, socio-scientific issues could introduce first-person perspectives into science education. The same may hold true for context-based approaches. Bennett, Lubben, and Hogarth (2007) found evidence that context-based approaches motivate pupils in science lessons, and improve attitudes toward science in both girls and boys. Obviously, context-based teaching frequently involves understanding “mental states,” which is an area in which empathizers can excel, regardless of whether they are male or female.

In particular, health and environment, two often neglected contexts in science education (Zeyer & Dillon, 2012), offer rich first-person connotations (Zeyer, 2012b) that may help empathizers cross the borders into the culture of science. In science education research, health and medicine are frequently considered girls' contexts (Schreiner & Sjøberg, 2004; Bybee, 2012). It is an intriguing idea that, instead, they might be empathizers' contexts, attractive not only to empathizing girls, but also to (often-neglected) empathizing boys.

From the point of view of the structural model, most of these recommendations for science teaching are aimed at low systemizers, independent of their gender, while it is assumed that high systemizers “help themselves” in science education. Given that systemizers are a minority (5%) in an average science classroom, it is not surprising that, in science education research, their needs usually do not come to the surface. Strategies for “good” science teaching identified by traditional science education research may not be suitable for them.

It may be important to give them more freedom to choose what they learn. This strategy has been shown to generally enhance students' motivation to learn science (Vedder-Weiss & Fortus, 2011; Vedder-Weiss & Fortus, 2012). High systemizers may particularly appreciate it.


  1. Top of page
  2. Abstract
  3. Theoretical Background
  4. This Study
  5. Method
  6. Results
  7. Discussion
  8. References
  9. Supporting Information
  • Aikenhead, G. S. (2000). Renegotiating the culture of school science. The contribution of research. In R. Millar, J. Leach, & J. Osborne (Eds.), Improving science education (pp. 245264). Philadelphia: Open University Press.
  • Airbuckle, J. L. (1997). AMOS user's guide. Version 3.6. Chicago: Smallwaters.
  • Alexander, J. M., Johnson, K. E., & Kelley, K. (2012). Longitudinal analysis of the relations between opportunities to learn about science and the development of interests related to science. Science Education, 96(5), 763786.
  • Andersen, H. M., & Nielsen, B. L. (2011). Video-based analyses of motivation and interaction in science classrooms. International Journal of Science Education. 35(6), 906928.
  • Archer, L., DeWitt, J., Osborne, J., Dillon, J., Willis, B., & Wong, B. (2012). “ Balancing acts”: Elementary school girls' negotiations of femininity, achievement, and science. Science Education, 96(6), 967989.
  • Bandalos, D. L., & Finney, S. J. (2001). Item parceling issues in structural equation modeling. In G. A. Marcoulides, & R. E. Schumacker (Eds.), New developments and techniques in structural equation modeling (pp. 269296). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
  • Bandura, A. (2001). Social cognitive theory: An agentive perspective. Annual Review of Psychology, 52, 126.
  • Baron-Cohen, S. (2003). The essential difference: Men, women and the extreme male brain. London: Penguin.
  • Baron-Cohen, S. (2009). Autism: The empathizing-systemizing (E-S) theory. Year in Cognitive Neuroscience 2009, 1156, 6880.
  • Baron-Cohen, S., Knickmeyer, R., & Belmonte, M. K. (2005). Sex differences in the brain: Implications for explaining autism. Science, 310, 819823.
  • Baron-Cohen, S., Richler, J., Bisarya, D., Gurunathan, N., & Wheelwright, S. (2003). The systemizing quotient: An investigation of adults with Asperger syndrome or high-functioning autism, and normal sex differences. Philosophical Transactions: Biological Sciences, 358(1430), 361374.
  • Baron-Cohen, S., & Wheelwright, S. (2004). The empathy quotient: An investigation of adults with asperger syndrome or high functioning autism, and normal sex differences. Journal of Autism and Developmental Disorders, 34(2), 163175.
  • Baron-Cohen, S., Wheelwright, S., Burtenshaw, A., & Hobson, E. (2007). Mathematical talent is linked to autism. Human Nature, 18, 125131.
  • Baron-Cohen, S., Wheelwright, S., Stone, V., & Rutherford, M. (1999). A mathematician, a physicist and a computer scientist with asperger syndrome: Performance on folk psychology and folk physics tests. Neurocase 5, 475483.
  • Bennett, J., Lubben, F., & Hogarth, S. (2007). Bringing science to life: A synthesis of the research evidence on the effects of context-based and STS approaches to science teaching. Science Education, 91(3), 347370.
  • Billington, J., Baron-Cohen, S., & Wheelwright, S. (2007). Cognitive style predicts entry into physical sciences and humanities: Questionnaire and performance tests of empathy and systemizing. Learning and Individual Differences, 17(3), 260268.
  • Bøe, M. V. (2012). Science choices in Norwegian upper secondary school: What matters? Science Education, 96(1), 120.
  • Bollen, K. A. (2002). Latent variables in psychology and the social sciences. Annual Review of Psychology, 53, 605634.
  • Britner, S. L. (2008). Motivation in high school science students: A comparison of gender differences in life, physical, and earth science classes. Journal of Research in Science Teaching, 45(8), 955970.
  • Britner, S. L., & Pajares, F. (2006). Sources of science self-efficacy beliefs of middle school students. Journal of Research in Science Teaching, 43(5), 485499.
  • Brotman, J. S., & Moore, F. M. (2008). Girls and science: A review of four themes in the science education literature. Journal of Research in Science Teaching, 45(9), 9711002.
  • Bryan, R. R., Glynn, S. M., & Kittleson, J. M. (2011). Motivation, achievement, and advanced placement intent of high school students learning science. Science Education, 95(6), 10491065.
  • Bybee, R. (2012). Scientific literacy in environmental and health education. In A. Zeyer & R. Kyburz-Graber (Eds.), Science environment health. Towards a renewed pedagogy for science education (pp. 4967). Dordrecht: Springer.
  • Byrne, B. M. (2010). Structural equation modeling with AMOS. Ottawa: University of Ottawa.
  • Byrne, B. M., Shavelson, R. J., & Muthen, B. (1989). Testing for the equivalence of factor covariance and mean structures: The issue of partial measurement invariance. Psychological Bulletin, 105(3), 456466.
  • Byrnes, J. P. (1996) Cognitive development and learning in instructional contexts. Needham Heights, MA: Allyn & Bacon.
  • Cavallo, A. M. L., Rozman, M., Blinkenstaff, J., & Walker, N. (2003). Students' learning approaches, reasoning abilities, motivational goals, and epistemological beliefs in differing college science courses. Journal of College Science Teaching, 33, 1823.
  • Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233255.
  • Cobern, W. W., & Loving, C. C. (2001). Defining “science” in a multicultural world: Implications for science education. Science Education, 85(1), 5067.
  • Davis, M. H. (1980). A multidimensional approach to individual differences in empathy. JSAS Catalogue of Selected Documents in Psychology, 10(4), 85.
  • DeWitt, J., Osborne, J., Archer, L., Dillon, J., Willis, B., & Wong, B. (2011). Young children's aspiration in science: The unequivocal, the uncertain and the unthinkable. International Journal of Science Education, 35(6), 10371063.
  • EURYDICE (2011). Science education in Europe: National policies, practices and research. Brussels: Education, Audiovisual and Culture Executive Agency.
  • Faiola, A., & Matei, S. A. (2005). Cultural cognitive style and web design: Beyond a behavioral inquiry into computer-mediated communication. Journal of Computer-Mediated Communication, 11, 375394.
  • Fensham, P. (1995). Policy and science education. International Journal of Science Education, 17(4), 411412.
  • Gilbert, J., & Calvert, S. (2003). Challenging accepted wisdom: Looking at the gender and science education question through a different lens. International Journal of Science Education, 25(7), 861878.
  • Gill, T., & Bell, J. F. (2011). What factors determine the uptake of A-level Physics? International Journal of Science Education, 35(5), 753772.
  • Glynn, S. M., & Koballa, T. R. (2006). Motivation to learn in college science. In J. J. Mintzes & W. H. Leonard (Eds.), Handbook of college science teaching (pp. 2532). Arlington, VA: National Science Teachers Association Press.
  • Glynn, S. M., Brickman, P., Armstrong, N., & Taasoobshirazi, G. (2011). Science Motivation Questionnaire II: Validation with science majors and nonscience majors. Journal of Research in Science Teaching, 48(10), 11591176.
  • Glynn, S. M., Taasoobshirazi, G., & Brickman, P. (2007). Nonscience majors learning science: A theoretical model of motivation. Journal of Research in Science Teaching, 44(8), 10881107.
  • Glynn, S. M., Taasoobshirazi, G., & Brickman, P. (2009). Science Motivation Questionnaire: Construct validation with nonscience majors. Journal of Research in Science Teaching, 46(2), 127146.
  • Grewal, R., Cote, J. A., & Baumgartner, H. (2004). Multicollinearity and measurement error in structural equation models: Implications for theory testing. Marketing Science, 23(4), 519529.
  • Haste, H. (2004). Science in my future: A study of values and beliefs in relation to science and technology amongst 11–21 year olds. London: Nestlé Social Research Programme.
  • Hirschfeld, L. A., & Gelman, S. A. (1994). Mapping the mind: Domain specificity in cognition and culture. New York: Cambridge University Press.
  • Holmegaard, H. T., Madsen, L. M., & Ulriksen, L. (2012). To choose or not to choose science: Constructions of desirable identities among young people considering a STEM higher education programme. International Journal of Science Education, 130. doi: 10.1080/09500693.2012.749362
  • Iacobucci, D. (2010). Structural equations modeling: Fit indices, sample size, and advanced topics. Journal of Consumer Psychology, 20(1), 9098.
  • Jacobs, J. E., & Bleeker, M. M. (2004). Girls' and boys' developing interests in math and science: Do parents matter? New Directions for Child and Adolescent Development, 2004(106), 521.
  • Jöreskog, K. G. (1993). Testing structural equation models. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 294316). Newbury Park, CA: Sage Publications.
  • Kahle, J. B., Parker, L. H., Rennie, L. J., & Riley, D. (1993). Gender differences in science Education: Building a model. Educational Psychologist, 28(4), 379404.
  • Kim, K. H. (2005). The relation among fit indexes, power, and sample size in structural equation modeling. Structural Equation Modeling, 12(3), 368390.
  • Kishton, J. M., & Widaman, K. F. (1994). Unidimensional versus domain representative parceling of questionnaire items: An empirical example. Educational and Psychological Measurement, 54(3), 757765.
  • Kjaernsli, M., & Lie, S. (2011). Students' preference for science careers: International comparisons based on PISA 2006. International Journal of Science Education, 33(1), 121144.
  • Lavenda, R. H., & Schultz, E. A. (2007). Anthropology: What does it mean to be human. New York: Oxford University Press.
  • Linn, M. C., & Hyde, J. S. (1989). Gender, mathematics, and science. Educational Researcher, 18(8), 1719.
  • Little, T. D., Cunningham, W. A., Shahar, G., & Widaman, K. F. (2002). To parcel or not to parcel: Exploring the question, weighing the merits. Structural Equation Modeling, 9(2), 151173.
  • Logan, M. R., & Skamp, K. R. (2012). The impact of teachers and their science teaching on students' ‘science interest’: A four-year study. International Journal of Science Education, 126.
  • Malhotra, N. K., Peterson, M., & Bardi Kleiser, S. (1999). Marketing research: A state-of-the-art review and directions for the twenty-first century. Journal of the Academy of Marketing Science, 27(2), 160183.
  • Morgan, H. (2001). Cognitive style and classroom learning. Westport, CT: Praeger.
  • Mujtaba, T., & Reiss, M. J. (2012). What sort of girl wants to study physics after the age of 16? Findings from a large-scale UK survey. International Journal of Science Education, 120. doi: 10.1080/09500693.2012.681076
  • Nettle, D. (2007). Empathizing and systemizing: What are they, and what do they contribute to our understanding of psychological sex differences? British Journal of Psychology, 98, 237255.
  • OECD (2011). Education at a glance 2011: OECD indicators. OECD Publishing. Retrieved November 15, 2011, from 10.1787/eag-2011-en
  • Olsen, R. V., & Lie, S. (2011). Profiles of students' interest in science issues around the world: Analysis of data from PISA 2006. International Journal of Science Education, 33(1), 97120.
  • Osborne, J., & Dillon, J. (2008). Science education in Europe: Critical reflections. A report to the Nuffield Foundation. London: King's College.
  • Osborne, J., Simon, S., & Collins, S. (2003). Attitudes towards science: A review of the literature and its implications. International Jounal of Science Education, 25(9), 10491079.
  • Özden, M. (2007). Problems with science and technology education in Turkey. Eurasia Journal of Mathematics, Science & Technology Education, 3(2), 157161.
  • PISA (2006). Assessing scientific, reading and mathematical literacy: A framework for PISA 2006. Paris: OECD.
  • Ployhardt, R. E., & Oswald, F. L. (2004). Applications of mean and covariance structure analysis: Integrating correlational and experimental approaches. Organizational Research Methods, 7, 2765.
  • Randler, C., Hummel, E., & Wüst-Ackermann, P. (2012). The influence of perceived disgust on students' motivation and achievement. International Journal of Science Education, 118.
  • Sadler, T. D. (2004). Informal reasoning regarding socioscientific issues: A critical review of research. Journal of Research in Science Teaching, 41(5), 513536.
  • Schreiner, C., & Sjøberg, S. (2004). ROSE (The Relevance of Science Education). A comparative study of students' views of science and science education. Acta Didactica, 4, 1120.
  • Searle, J. R. (2004). Mind: A brief introduction. New York: Oxford University Press.
  • Simpson, R., & Oliver, J. S. (1990). A summary of major influences on attitude toward achievement in science among adolescent students. Science Education, 74(1), 114.
  • Steenkamp, J.-B. E. M., & Baumgartner, H. (2000). On the use of structural equation models for marketing modeling. International Journal of Research in Marketing, 17(2–3), 195202.
  • Steinmetz, H. (2011). Estimation and comparison of latent means across cultures. In E. Davidov, P. Schmidt, & J. Billiet (Eds.), Cross-cultural analysis methods and applications (pp. 85116). New York: Routledge.
  • Stern, E., Metzger, S., & Zeyer, A. (2009). Expertise zu Naturwissenschaft und Technik in der Allgemeinbildung im Kanton Zürich. Zürich: ZHSF.
  • Swarat, S., Ortony, A., & Revelle, W. (2012). Activity matters: Understanding student interest in school science. Journal of Research in Science Teaching, 49(4), 515537.
  • Syed Zin, S. M., & Lewin, K. M. (1993). Insight into science education: Planning and policy priorities in Malaysia. Laporan Kajian Bersama Kementerian Pendidikan Malaysia dan International Institute for Educational Planning, UNESCO. Paris: IIPP's Printshop.
  • Tuan, H. L., Chin, C. C., & Shieh, S. H. (2005). The development of a questionnaire for assessing students' motivation toward science learning. International Journal of Science Education, 27, 639654.
  • United Nations Development Programme (2011). Human development report 2011: Sustainability and equity: A better future for all. New York: Palgrave Macmillan.
  • Varma, O. P., & Thakur, M. (1992). Cognitive style and scholastic achievements. Psycho-Lingua, 22(1), 8192.
  • Vedder-Weiss, D., & Fortus, D. (2011). Adolescents' declining motivation to learn science: Inevitable or not? Journal of Research in Science Teaching, 48(2), 199216.
  • Vedder-Weiss, D., & Fortus, D. (2012). Adolescents' declining motivation to learn science: A follow-up study. Journal of Research in Science Teaching, 49(9), 10571095.
  • Wakabayashi, A., Baron-Cohen, S., Uchiyama, T., Yoshida, Y., Kuroda, M., & Wheelwright, S. (2007). Empathizing and systemizing in adults with and without autism spectrum conditions: Cross-cultural stability. Journal of Autism and Developmental Disorders, 37(10), 18231832.
  • Wheelwright, S., Baron-Cohen, S., Goldenfeld, N., Delaney, J., Fine, D., Smith, R., & Wakabayashi, A. (2006). Predicting autism spectrum quotient (AQ) from the systemizing quotient-revised (SQ-R) and empathy quotient (EQ). Brain Research, 1079, 4756.
  • Wilde, M., Hussmann, J. S., Lorenzen, S., Meyer, A., & Randler, C. (2012). Lessons with Living Harvest Mice: An empirical study of their effects on intrinsic motivation and knowledge acquisition. International Journal of Science Education, 34(18), 27972810.
  • Witkin, H. A., Dyk, R., Faterson, H. F., Goodenough, D. G., & Karp, S. A. (1962). Personality through perception. New York: Harper & Row.
  • Zeldin, A. L., Britner, S. L., & Pajares, F. (2008). A comparative study of the self-efficacy beliefs of successful men and women in mathematics, science, and technology careers. Journal of Research in Science Teaching, 45(9), 10361058.
  • Zeldin, A. L., & Pajares, F. (2000). Against the odds: Self-efficacy beliefs of women in mathematical, scientific, and technological careers. American Educational Research Journal, 37, 215246.
  • Zeyer, A. (2010). Motivation to learn science and cognitive style. Eurasia Journal of Mathematics, Science and Technology Education, 6(2), 123130.
  • Zeyer, A. (2012a). What is a potential scientist? Motivation to learn science—A question of culture or of cognition? In S. Markic, I. Eilks, D. d Fuccia, & B. Ralle (Eds.), Issues of heterogenitiy and cultural diversity in science education and science education research. Aachen: Shaker Verlag.
  • Zeyer, A. (2012b). A win–win situation for health and science education: Seeing through the lens of a new framework model of health literacy. In A. Zeyer & R. Kyburz-Graber (Eds.), Science environment health. Towards a renewed pedagogy for science education (pp. 147173). Dordrecht: Springer.
  • Zeyer, A., Bölsterli, K., Brovelli, D., & Odermatt, F. (2012). Brain type or sex differences? A structural equation model of the relation between brain type, sex, and motivation to learn science. International Journal of Science Education, 34(5), 779802.
  • Zeyer, A., & Dillon, J. (2012). Science environment health—Towards a reconceptualization of three critical and inter-linked areas of education. International Journal of Science Education, 34(2), 327328.
  • Zeyer, A., & Wolf, S. (2010). Is there a relationship between brain type, sex and motivation to learn science? International Journal of Science Education, 32(16), 22172233.
  • Zhang, L.-F., & Sternberg, R. J. (2006). The nature of intellectual styles. Mahwah. New Jersey: Lawrence Erlbaum Associates, Publishers.

Supporting Information

  1. Top of page
  2. Abstract
  3. Theoretical Background
  4. This Study
  5. Method
  6. Results
  7. Discussion
  8. References
  9. Supporting Information

Additional supporting information may be found in the online version of this article at the publisher's web-site.

tea21101-sm-0001-SuppTab-S1.pdf26KTable S1. Descriptive statistics of SQ, EQ, SMQ, and the SMQ sub-scale scores by gender and by country

Table S2. International human development indicators

Table S3. PISA: Comparing countries' performance

Table S4. World Bank: School enrolment

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.