Introduction and background

The research reported in this article was part of a larger study carried out in Abu Dhabi, United Arab Emirates (UAE; Khalil 2015). In 2005, just prior to the commencement of the study, the Abu Dhabi Education Council (ADEC) adopted a system-wide initiative, which involved a shift from traditional, teacher-centred teaching and learning to a student-centred learning environment. The reform was introduced to advance students as creative, independent thinkers and problem-solvers who acquire knowledge through exploration and experimentation (ADEC 2013). An important aspect of the implementation of the new reform was to elevate school quality to international standards and assist learners to contribute positively to society by developing successful careers (Badri and Al Khaili 2014). It was anticipated that the new inquiry-based teaching methods would improve students’ perceptions of their learning environment in science classes and their interest in science as a school subject, thus impacting on students’ decisions to study for careers in the sciences (ADEC 2013). Thus, this study examined the impact of the reform on the relationships between the learning environment and students’ attitudes towards science in Abu Dhabi, UAE.

Field of learning environments

Students spend approximately 20,000 h in the classroom by the time they graduate from university, which is why their reactions to their teaching and learning experiences are of considerable importance (Fraser 2012). Further, research has revealed strong and consistent associations between the learning environment and a range of student outcomes, both cognitive and affective, indicating that learning environments play an important role in effective teaching and learning (Fraser 2001, 2012). Because educational researchers have found strong relationships between the learning environment and academic outcomes, it is now widely recognised that the learning environment plays an important role in improving the effectiveness of learning (Fraser 2001; UNESCO 2012).

A striking feature of the field of learning environments is the availability of a variety of economical, valid and widely-applicable instruments that have been developed and used for assessing students’ perceptions of the classroom environment (Fraser 2012, 2018). These instruments have been used at various educational levels and have been translated into many languages for use in numerous countries. Instruments have been developed to assess students’ perceptions of their classroom learning environment in terms of: whether students actively participate in class or sit and listen to the teacher; whether students cooperate and discuss with each other what they are learning or whether they work alone; whether the class is dominated by the teacher or is student centred; whether the teacher is supportive and approachable; and whether the students have a say in the choice of teaching and assessment methods (Fraser 2012).

Although there is growing recognition of the usefulness of learning environment perceptions in evaluating educational innovations, a review of literature indicated that only a handful of instruments have been developed in the Arabic language to assess the learning environment (Al Zubaidi et al. 2016; Afari et al. 2013; Hasan and Fraser 2015; MacLeod and Fraser 2010). Two of these studies involved modifying and translating the WIHIC for use at the college level (Afari et al. 2013; MacLeod and Fraser 2010). To the best of our knowledge, the study reported in this article was the first to be carried out at the high-school level in the UAE, thereby extending these past studies. Furthermore, this study is unique in investigating relationships between students’ perceptions of their learning environment and attitude outcomes in Abu Dhabi, UAE. This section provides a review of research on associations between student outcomes and learning environment that was pertinent to the aims of this study.

Environment-outcome associations

One of the aims of this study was to examine the relationships between the learning environment and student outcomes. This section provides a review of literature pertinent to the field of learning environments and variables that are closely related to those included in the present study, namely, attitudes and motivation.

Findings of past research strongly suggest that students have more positive attitudes towards the class or subject when they perceive their classroom learning environment to be more positive (Dorman and Fraser 2009). For example, when Dorman and Fraser (2009) investigated student attitudes, classroom environment and antecedent variables (gender, grade level and home computer and internet access) with a sample of 4146 high-school students, they found that improving the classroom environment had the potential to improve student attitudes and that the antecedents did not have any direct effect on the outcomes. Other studies of relationships between student attitudes and the learning environment include Kenar et al. (2013), Mink and Fraser (2005) and Teh and Fraser (1994). Past research has also provided strong and consistent evidence to suggest that the learning environment influences student attitudes in a range of subject areas, including: science (Aldridge and Fraser 2000; Aldridge et al. 1999; Kerr et al. 2006; Koul and Fisher 2005; Lay and Khoo 2012; Martin-Dunlop and Fraser 2008; Peer and Fraser 2015; Shadrek 2012; Telli et al. 2006; Wolf and Fraser 2008; Zaragoza and Fraser 2017), mathematics (Afari et al. 2013; Ogbuehi and Fraser 2007); mathematics and science (Fraser and Raaflaub 2013); and geography and mathematics (Chionh and Fraser 2009).

There is also much evidence to suggest that students’ perceptions of the learning environment are related to their self-reports of motivation (Gilbert et al. 2014; Velayutham and Aldridge 2013). As with attitudes, strong, positive environment–motivation associations have been found for a range of subjects, including: science (Barak et al. 2011; Koul et al. 2012; Nolen 2003; Velayutham and Aldridge 2013) and mathematics (Gilbert et al. 2014; Opolot-Okurut 2010). These findings were also found across different levels of education, including: primary school (Saeed and Zyngier 2012), high school (Hanrahan 1998; Velayutham et al. 2011) and college (Al Zubaidi et al. 2016; Afari et al. 2013; Baeten et al. 2013).

Gaps and concerns

Although there is growing recognition of the usefulness of learning environment perceptions in evaluating educational innovations (Fraser 2012), to date, there is limited evidence to suggest that this happens in terms of reform efforts taking place in the UAE. The reasons for this could be twofold. First, there could be a lack of reliable surveys and, second, there is a dearth of research to support the notion that learning environment perceptions impact on student outcomes in Arab countries. Further, to the best of our knowledge, there are no studies that have been carried out in the field of learning environments in the UAE in high-school science classrooms. Therefore, the research reported in this article fills a gap in research and extends the existing literature by using a modified and translated learning environment instrument to measure students’ perceptions of the learning environment in the UAE as part of the reform efforts.

Research methods

Sample

Prior to the main administration, the translated version of the modified WIHIC was pilot tested in one randomly-selected grade 7 class (n = 28). In addition, six students were interviewed to provide support for the face validity of the surveys. The selection of these students was based on their willingness to participate.

The data for the main study were collected from a sample of 784 students (365 male and 419 females) from 34 classes in eight public schools in Abu Dhabi, UAE. These schools were all middle schools, catering for students from grades 6–9 and aged between 12 and 15 years. Given that all of the schools were single sex, as required by the Abu Dhabi Education Council (ADEC), four of the schools were all-female and four were all-male. All eight schools were situated on Abu Dhabi Island and were resourced by ADEC. The schools catered for both Emirati (mostly) and expatriate (Arab) students. This sample of schools provided a range of student abilities and was considered to be representative of public schools within the Abu Dhabi Emirate.

Two instruments were used to collect the quantitative data: the What Is Happening In this Class? (WIHIC), initially developed by Fraser et al. (1996), to assess students’ perceptions of their learning environment; and the Science Attitudes and Engagement Survey (SAES) to assess students’ motivation, attitudes and aspirations towards science. The process for modifying and translating the instruments into Arabic to make them suitable for the UAE context is described below.

What Is Happening In this Class (WIHIC)

Given the robust nature of the WIHIC across a number of settings and in a variety of different languages, it was considered suitable for the present study. The WIHIC has been used in many studies to investigate classroom learning environments in different subject area, at different age levels and in different countries. The WIHIC has been validated in science in a number of countries, including: Australia and Taiwan (Aldridge and Fraser 2000; Aldridge et al. 1999), South Africa (Aldridge et al. 2006b); Myanmar (Khine et al. 2018); and Canada (Zandvliet and Fraser 2005). The WIHIC has been translated into numerous languages, including Mandarin (Aldridge and Fraser 2000; Aldridge et al. 1999), Korean (Kim et al. 2000), Malay (Khine and Fisher 2001) and Indonesian (Margianti et al. 2001).

Previous versions of the WIHIC, translated for use in the UAE, had either been modified to suit college-level students (Afari et al. 2013) or did not report strong factor analysis results for all scales (MacLeod and Fraser 2010); therefore, it was necessary to modify and translate the WIHIC to ensure suitability for the present study. Five of the original seven scales that were considered relevant to a learning environment involving cooperative learning strategies were included, namely: student cohesiveness (the degree to which students know, help and are supportive of one another); teacher support (the degree to which the teacher assists, befriends, trusts and is interested in students); involvement (the degree to which students have attentive interest, participate in discussions, perform additional work and enjoy the class); cooperation (the degree to which students cooperate rather than compete with one another on learning tasks); and equity (the degree to which students are treated equally by the teacher).

The final 40-item version of the modified WIHIC included five scales, namely: student cohesiveness, teacher support, involvement, cooperation and equity. The items were all positively-worded and responded to using a five-point frequency–response scale consisting of almost always, often, sometimes, seldom and almost never. Responses were scored 5, 4, 3, 2 or 1, respectively, with all omitted or invalid responses given a score of three.

Science Attitudes and Engagement Survey (SAES)

To assess students’ attitudes towards science, the Science Attitudes and Engagement Survey (SAES; see Khalil 2015 for details related to the reliability and validity of the instrument) was administered. The development of the SAES drew on two existing surveys, namely, the Test Of Science-Related Attitudes (TOSRA) developed by Fraser (1981) and the Student Adaptive Learning Engagement Survey (SALES) developed by Velayutham et al. (2011).

The SAES consists of eight scales: social implications (to assess whether science is worthy of time, resources and money); adoption of scientific attitudes (the degree to which the student is curious and open to finding out about new things as well as listening to other people’s views); engagement (the degree to which the student is involved in activities and tasks); self-efficacy (students’ confidence and beliefs in their own ability to perform tasks); self-regulation (the degree to which the student controls and regulates the effort put into performing tasks); learning goal orientation (the degree to which students perceive themselves to be participating in science classes for the purpose of learning, understanding and mastering science concepts as well as improving skills); task value (the degree to which the student perceives the science tasks as having interest, importance and utility); and science career aspirations (the degree to which the student is interested in or aspires to a science career). Each scale of the SAES had eight items, providing a total of 64 items. The items were responded to using a five-point Likert response scale of strongly agree, agree, uncertain, disagree and strongly disagree. The items in the SAES were scored 5, 4, 3, 2 and 1, respectively, with omitted or invalid answers assigned a score of three.

Translation of the instruments

To make the modified WIHIC and SAES instruments usable in the UAE context where students speak English as a second language, they were translated into Arabic. The Arabic translation of the surveys was generated using the process of translation, back-translation, verification and modification that is recommended by Ercikan (1998) and Warwick and Osherson (1973). Researchers agree that back-translation of an instrument is essential for its validation and use in a cross-cultural study (John et al. 2006) and is important for maintaining equivalence between the original and translated versions (Behling and Law 2000). The translation process involved, firstly, an independent expert in both languages translating the English version into Arabic. Secondly, an additional expert in both languages who was not familiar with the original English version back-translated each item into English. The two English versions were then compared to ensure that they were consistent in meaning. During the process, minor modifications to the wording were made to make it usable in the UAE context, although care was taken to ensure the intended meaning of each item was not changed. For example, “I knew how to proceed with hard work” was changed to “I can figure out how to do difficult work” and “I have the desire to know the world that we live in” was changed to “I am curious about the world in which we live”.

Although students spoke Arabic as their first language, students were taught science predominantly in English. Thus, it was considered desirable to provide students with a dual layout that included the Arabic version of each item followed by the English version directly underneath. This layout had been used successfully in similar studies by Aldridge et al. (2006a) in South Africa and Afari et al. (2013) in the UAE.

Data analysis

As a first step, we provided evidence to support the reliability and validity of the instruments when used in the UAE. Our analysis was guided by the framework suggested by Trochim and Donnelly (2006), which suggests that both translation and criterion validity should be fulfilled. Hence, an instrument has high construct validity if it can establish content, face, convergent, concurrent and discriminant validity. The analysis used to fulfil each of these criteria is discussed below.

Translation validity is related to the interpretation of the construct and whether it is detailed, accurate, based on theory and can be understood by the participants. Because the scales used in the present study were drawn from previously established instruments, validation focused only on establishing face validity and criterion validity. To ensure face validity, the instruments were pilot tested in two stages: first, administration of the instruments; and, second, interviews with students.

Evidence to support the criterion validity involved whether the construct affords the assumptions that are anticipated. First, principal axis factor analysis with oblique rotation was used to examine the factor structure of the instruments. The criteria for retaining an item was that it should have a loading of at least 0.40 on its own scale and less than 0.40 on all other scales (Field 2009). Once the factor structure was established, the internal consistency reliability (using Cronbach alpha coefficient) was examined. A cut-off of 0.60 was considered to be satisfactory and alpha coefficients of 0.70 were considered to be ‘good’ (Cohen et al. 2009).

To support the discriminant validity, the component correlation matrix obtained from oblique rotation in exploratory factor analysis was examined. Brown (2006) and Field (2009) explain that oblique rotation in exploratory factor analysis provides a realistic representation of how factors are interrelated. Whilst Field (2009), noted that based on theoretical grounds, there should be a moderately strong relationship between factors, according to Brown (2006), factor correlations above 0.80 indicate an overlap of concepts and point towards poor discriminant validity. Finally, to support the concurrent validity, the ability to differentiate between the groups that it is expected to differentiate between, an analysis of variance (ANOVA) was conducted using class membership as the independent variable. The η2 statistic, based on the ratio of the between-group effect to the total amount of variance in the data (Field 2009), provided information about the amount of variance attributed to class correlations.

To examine whether relationships exist between students’ perceptions of a cooperative learning environment and their attitudes, motivation and science-related aspirations, simple correlation and multiple regression analyses were employed. As a first step, simple correlation analysis was used to examine the bivariate relationships between each attitude and each learning environment scale. In the second step, multiple regression analysis was undertaken using the set of five scales of the WIHIC survey as independent variables and each attitude scale as the dependent variable. This analysis provided more parsimonious information about relationships between correlated independent variables and reduced the risk of a Type I error often linked with simple correlation analysis. To identify which of the individual learning environment scales were significant independent predictors of student attitudes, the standardised regression weights (β) were examined.

Results

This section first reports evidence to support the reliability and validity of the instruments. Next, we report the results of the simple and multiple correlations used to examine relationships between students’ perceptions of the learning environment and the outcomes assessed in the SAES.

Pilot testing of instruments

Both instruments were pilot tested in one all-female class (n = 28) by administering them and then interviewing selected participants. Students were carefully observed during administration to ensure that no technical difficulties were encountered and they were encouraged to ask questions if item clarification was required. Students took approximately 15 min to complete the SAES and approximately 10 min to complete the WIHIC, with no technical difficulties being encountered or reported by students.

The purpose of the interviews was to ensure the readability and comprehensibility of individual items and to ascertain whether they were interpreted in ways that were intended by the researcher (Cohen, Manion and Morrison 2000). The interviews indicated that the participants interpreted the items in ways that were intended and it was established that items were suitable for use in the UAE context. During the interviews, when students were asked to give examples of their responses, there was much discussion amongst the students, who were eager to give their opinions. The interviews indicated that the participants interpreted the items in ways that were intended and it was established that items were suitable for use in the UAE context. For example, when asked to give examples for “I am involved in hands-on activities”, students responded by citing experiments and group work. Additionally, when asked to give examples for “A career in science would be exciting”, responses such as brain surgeon and petroleum engineer were provided. Student interviews confirmed the face validity of the WIHIC and SAES. The Arabic/English versions of the WIHIC and SAES were judged to be valid and reliable when used in the UAE context. In all cases, students were able to explain their response choices for the survey items and were able to provide examples for their choice of response.

Construct validity of the WIHIC

To investigate the construct validity of the translated (Arabic) version of the WIHIC when used in the UAE, the data collected from 784 students were used to examine the instrument’s factor structure, internal consistency reliability, concurrent validity and discriminant validity. The multivariate normality and sampling adequacy of the data were tested for the WIHIC. Bartlett’s test of sphericity indicated that χ2 = 19,719.950 and that this value was statistically significant (p < 0.001). The Kaiser–Meyer–Olkin measure of adequacy was high (0.97), confirming the appropriateness of the data for further analysis.

Exploratory factor analysis using principal component factor analysis with oblique rotation was then carried out to extract salient factors. The result reported in Table 1 show that the factor loadings for each item were at least 0.40 on their a priori scale and less than 0.40 on each of the other four scales. The percentage of variance accounted for by different scales, reported at the bottom of Table 1, ranged from 3.16 to 40.12%, with the total variances explained being 61.08%. Also, the bottom of Table 1 shows that the eigenvalues for different scales ranged from 1.27 to 16.05, which are was greater than 1, indicating that they are acceptable as recommended by Kaiser (1960).

Table 1 Factor loadings, internal consistency reliability (Cronbach alpha coefficient) and ability to differentiate between classes (ANOVA results) for the WIHIC
Full size table

The internal consistency reliability, using Cronbach’s alpha coefficient, was calculated for each learning environment scale with the individual as the unit of analysis. The bottom of Table 1 shows that alpha coefficients for the five scales ranged from 0.86 to 0.93, which are higher than the minimum of 0.70 for satisfactory reliability as recommended by Cohen et al. (2009).

To examine whether each WIHIC scale could distinguish between students in different classes, a one-way ANOVA was used. The ANOVA results, reported at the bottom of Table 1, were statistically significant (p < 0.01) for all five WIHIC scales, suggesting that the Arabic version of the WIHIC was able to distinguish between the perceptions of students in different classes. The η2 values ranged from 0.13 to 0.26.

Finally, the correlation matrix, generated through the oblique rotation, was used to examine the discriminant validity of the WIHIC scales. The component correlation matrix, obtained from the oblique rotation (available from the corresponding author upon request), showed that the highest correlation was 0.62 and thus all values met the requirements of adequate discriminant validity according to Brown (2006).

Construct validity of the SAES

To provide evidence to support the construct validity of the Science Attitudes and Engagement Survey (SAES), the data gathered from 784 students were used to examine the convergent validity (factor structure and internal consistency reliability), concurrent validity (ANOVA) and discriminant validity. The multivariate normality and sampling adequacy of the data were tested for the SAES. Bartlett’s test of sphericity indicated that χ2 = 25,347.258 which was statistically significant (p < 0.001). The Kaiser–Meyer–Olkin measure of adequacy was high (0.96), confirming the appropriateness of the data for further analysis.

Principal component factor analysis with oblique rotation was carried out to examine the factor structure of the SAES. Three items (Items 11, 32 and 49) were found to be problematic and were removed from all further analyses. The factor loadings for the remaining 61 items in 8 scales are reported in Table 2, showing that all but one of the remaining 61 items (the exception being Item 10) had a factor loading of at least 0.40 on its own scale and less than 0.40 on the other seven scales. The exception was retained, however, as its inclusion strengthened the structure of the scale and increased its internal consistency.

Table 2 Factor loadings, internal consistency reliability (Cronbach alpha coefficient) and ability to differentiate between classes (ANOVA results) for the SAES
Full size table

The bottom of Table 2 shows that the percentage of variance accounted for by different scales ranged from 1.93 to 31.76%, with the total variances explained being 56.54%. Also, the bottom of Table 2 shows that the eigenvalues for different scales ranged from 1.18 to 19.37, which are greater than 1 and are acceptable as recommended by Kaiser (1960). These results support the factorial validity of the SAES when used with the sample of 784 students.

The internal consistency reliability, using Cronbach’s alpha coefficient, was calculated for each of the SAES scales. The results at the bottom of Table 2 indicate that the alpha coefficients for the eight scales ranged from 0.80 to 0.90. Based on the cut off of 0.70 recommended by Cohen et al. (2009), these reliability coefficients were considered to be satisfactory.

To examine whether the SAES scales could distinguish between students in different classes, a one-way ANOVA was used. The ANOVA results, reported in Table 2, were statistically significant (p < 0.01) for all eight scales, indicating that students in the same class have similar attitudes, while attitudes vary between classes. The last row in Table 2 indicates that the η2 values ranged from 0.12 to 0.23.

Finally, the correlation matrix, generated through oblique rotation, was used to examine the discriminant validity of the SAES. The results (see Khalil 2015) showed that the highest correlation was 0.47 and that this value satisfies the requirements of acceptable discriminant validity suggested by Brown (2006).

Relationships between the learning environment and attitudes

To ascertain whether students’ perceptions of the learning environment were related to their attitudes towards science and science-career aspirations, simple correlation and multiple regression analyses were employed. The simple correlations reported in Table 3 indicate that all five WIHIC scales were statistically-significantly (p < 0.01) and positively related to each of the eight attitude outcomes (self-efficacy, self-regulation, learning goal orientation, task value, social implications, engagement, adoption of scientific attitudes and science career aspirations). The multiple correlations reported at the bottom of Table 3 were positive and statistically significant (p < 0.01) for all eight attitude outcomes. The multiple correlations (R) ranged between 0.40 and 0.49 for each outcome. To identify which of the learning environment scales contributed to the variance in students’ attitudes, the standardised regression weights (β) reported in Table 3 were examined as discussed below for each learning environment scale.

Table 3 Simple correlation and multiple regression analyses for associations between learning environment and attitude scales
Full size table

The student cohesiveness scale assesses the extent to which students know, help and are supportive of one another. The standardised regression weights (β) reported in Table 3 indicate that student cohesiveness was a statistically-significant (p < 0.01) independent predictor of four of the eight outcomes: learning goal orientation, task value, social implications and science career aspirations.

The teacher support scale assesses the extent to which the teacher assists, befriends, trusts and is interested in students. Table 3 suggests that teacher support was a significant (p < 0.01) independent predictor of seven of the eight attitude outcomes, the exception being engagement.

The involvement scale assesses the extent to which students have attentive interest, participate in discussions, perform additional work and enjoy the class. The standardised regression weights (β) (Table 3) indicate that involvement was a significant (p < 0.01) independent predictor of five of the eight attitude outcomes: self-efficacy, self-regulation, task value, engagement and science career aspirations.

The cooperation scale assesses the extent to which students cooperate rather than compete with one another on learning tasks. Table 3 suggests that cooperation was a significant (p < 0.05) independent predictor of six of the eight attitude outcomes: self-efficacy, learning goal orientation, task value, social implications, engagement and adoption of scientific attitudes.

The equity scale assesses the extent to which students are treated equally by the teacher. Table 3 suggests that equity was a significant (p < 0.01) independent predictor of all eight attitude outcomes.

Discussion

The aim of this study was twofold: to investigate students’ perceptions of their science learning environment using a modified and translated (Arabic/English) version of the What Is Happening In this Class? (WIHIC); and to examine relationships between the learning environment and students’ attitudes, motivation, engagement and science career aspirations in the United Arab Emirates (UAE). The modified and translated (Arabic/English) version of the WIHIC was employed to assess students’ perceptions of their learning environment and the Science Attitudes and Engagement Survey (SAES; Khalil 2015) was used to assess students’ motivation, attitudes and aspirations towards a career in science. The data collected from 784 students in 34 science classes in Abu Dhabi, UAE was analysed to establish the validity and reliability of the WIHIC instrument and to identify relationships between the learning environment and student attitudes.

Given that the WIHIC is already a well-established learning environment instrument, its validation for the purpose of this study involved only the face validity and criterion-related aspects of the construct validity framework suggested by Trochim and Donnelly (2006). All items had factor loadings of at least 0.40 on their own scale and less than 0.40 on the other four scales. The alpha coefficients for the five WIHIC scales ranged from 0.86 to 0.93 with the individual as the unit of analysis, which are all higher than the minimum of 0.70 for satisfactory reliability recommended by Cohen et al. (2009). The ANOVA results were statistically significant (p < 0.01) for all five WIHIC scales, with η2 values ranging from 0.13 to 0.26, thus suggesting that the Arabic/English version of the WIHIC was able to distinguish between the perceptions of students in different classes. The component correlation matrix, obtained from the oblique rotation, showed that the highest correlation was 0.62 and thus all values met the requirements of adequate discriminant validity (Brown 2006).

The findings reported here are similar to those of other studies that used modified English versions of the WIHIC instrument in various countries and found it to be a valid and reliable tool in: Australia (Dorman 2008); Australia and Canada (Zandvliet and Fraser 2005); Singapore (Chionh and Fraser 2009; Khoo and Fraser 2008); India (Koul and Fisher 2005, 2006); Uganda, Africa (Opolut-Okurut 2010); New Zealand (Saunders and Fisher 2006); South Africa (Aldridge et al. 2009) and the United States of America (den Brok et al. 2006; Helding and Fraser 2013; Martin-Dunlop and Fraser 2008; Ogbuehi and Fraser 2007; Pickett and Fraser 2009; Wolf and Fraser 2008). Additionally, our results corroborate those of other studies that have found the WIHIC to be valid and reliable when used in languages other than English, including: Mandarin (Aldridge and Fraser 2000); Korean (Kim et al. 2000; Khine and Fisher 2003); Indonesian (Fraser et al. 2010; Margianti et al. 2001; Wahyudi and Treagust 2004); IsiZulu (Aldridge et al. 2009); Myanmar (Khine et al. 2018); Sepedi (Aldridge et al. 2006b); Arabic (Afari et al. 2013; MacLeod and Fraser 2010); Spanish (Allen and Fraser 2007; Robinson and Fraser 2013); and Greek (Giallousi et al. 2010).

The findings of this study provide evidence to support the reliability and validity of the Arabic/English version of the WIHIC instrument when used in the UAE context. The findings indicate that the Arabic/English WIHIC instrument employed in this study can be used with confidence in future studies.

Simple correlations indicated that all five WIHIC scales were statistically significantly (p < 0.01) and positively related to each of the eight attitude outcomes. Multiple correlations were positive and statistically significant (p < 0.01) for all eight attitude outcomes, and ranged from 0.40 to 0.49 for the eight attitude outcomes. The educational implications of these findings are provided below.

The Student Cohesiveness scale was found to be a statistically-significant (p < 0.01) independent predictor of four of the eight attitude outcomes: learning goal orientation, task value, social implications and science career aspirations. This suggests that, if students are given opportunities to get acquainted and to help and support one another, then they are more likely to have improved attitudes for these scales. Thus, in light of the UAE reform recommendations, it is recommended that teachers use cooperative learning or similar teaching and learning methods to encourage student cohesiveness and thereby improve student attitudes in science.

Teacher support was a significant (p < 0.01) independent predictor of seven of the eight attitude outcomes, with the exception being engagement. These results indicate that the relationships that teachers have with students are critical in the learning environment. The results suggest that teachers wishing to improve student attitudes should consider their interpersonal relationships with students, showing interest in students’ problems, befriending them and assisting them. Given that teachers play a pivotal role in the reform, through the implementation of student-centred teaching and learning, then careful consideration of the levels of teacher support are likely to have a profound impact on attitudes.

Involvement was found to be a significant (p < 0.01) independent predictor of five of the eight attitude outcomes: self-efficacy, self-regulation, task value, engagement and science career aspirations. This finding implies that increasing student involvement in lessons is likely to positively influence the five attitude scales. Implementing inquiry-based, student-centred, hands-on activities in science classes (as per the reform requirements) impact positively on students’ attitudes.

The cooperation scale was a significant (p < 0.05) independent predictor of six of the eight attitude outcomes: self-efficacy, learning goal orientation, task value, social implications, engagement and adoption of scientific attitudes. This finding suggests that increased student collaboration during cooperative learning based tasks and activities in science lessons positively influences student attitudes.

The equity scale was a significant (p < 0.01) independent predictor of all eight attitude outcomes. All of the significant correlations were positive, suggesting that teachers wishing to improve student attitudes should treat students in a way that is perceived to be fair.

We found that all five WIHIC scales were statistically-significantly and positively related to different attitude outcomes. These findings were consistent with many past studies that reported that students taught using a cooperative learning approach (compared with a conventional learning approach) were likely to have improved attitudes towards science lessons (Akinbobola 2009; Arslan et al. 2006; Campisi and Finn 2011; Demir 2008; Demirici 2010; Hong 2010; Kincal et al. 2007; Korkut-Owen et al. 2009; Kose et al. 2010; Looi et al. 2010; Martin-Dunlop and Fraser 2008; Raviv et al. 2017; Sung and Hwang 2013).

Furthermore, our findings corroborate past studies for which cooperative learning environments improved student motivation towards science (Freeman et al. 2008; Moebius-Clune et al. 2011; Ural et al. 2017). For example, Freeman et al. (2008) found that there was a positive influence of learning communities (collaborative learning) on students’ attitudes, learning experiences and intrinsic motivation in Science, Technology, Engineering and Mathematics (STEM). Further, Moebius-Clune et al. (2011) found that, during inquiry-based teaching, students were motivated and substantively engaged. Kirik and Boz (2012) and Foster (1985) reported that, compared with traditional instruction, cooperative learning improved students’ motivation to study chemistry.

With respect to science career aspirations, the findings of this study corroborate those of past research by Duran et al. (2014) and House (2009). Furthermore, House (2009) found that students involved in activities associated with cooperative learning, such as using active learning strategies, showed positive interest in a science career. Conversely, students who reported that they more frequently listened to the teacher give a lecture-style presentation during science lessons tended to have less interest in a science career.

Our results suggest that students’ perceptions of the learning environment influence their attitudes, particularly self-efficacy and science career aspirations. This study has contributed to these areas of the literature, specifically related to the reform efforts underway in the UAE context. Given these findings, it is recommended that education reform efforts, similar to those examined in this study, consider the importance of providing teachers with knowledge of how to improve the learning environment in science classes in an attempt to improve student attitudes towards science learning. Further, it is recommended that teachers wishing to improve student attitudes look for ways to include more student cohesiveness, teacher support and involvement in their classes. This is an important finding for informing future decisions in the UAE context.

Limitations and conclusion

Although rigorous validity analyses, using Trochim and Donnelly’s (2006) construct validity framework, were carried out in this study, we used exploratory factor analysis to examine the factor structure. It is recommended that, to provide additional evidence for the validity of the WIHIC, future research studies include confirmatory factor analysis. Because this study involved only students from Cycle 2 or middle schools, generalisations to other grade levels should be made with caution. It is recommended, therefore, that future studies include students in other grades. Because data for the present study were gathered from eight schools within Abu Dhabi city, generalising findings to schools in outer regional Abu Dhabi and to other Emirates should be undertaken with caution. It is recommended, therefore, that future studies involve a larger and wider sample of students, both nationally and internationally. Finally, we did not examine the predictive validity of the SAES, which is a fertile area for future studies (to establish that the data predicts something it should theoretically predict).

This research is significant to the field of learning environments and to science education because it was the first study of its kind to examine the impact of cooperative learning on student perceptions of the learning environment and their outcomes in science classes in Abu Dhabi, UAE. The findings of this study suggest that the implementation of cooperative learning strategies in science lessons in Abu Dhabi has improved student attitudes. Although the focus of this study was science learning, the findings could help educators improve their learning environments and student attitudes in other subject areas. These findings have implications for researchers, policy makers, organisations, school administration and teachers wishing to improve the learning environment and student attitude outcomes. It is anticipated that our findings can help to transform future science classrooms for the benefit of all stakeholders. Ultimately, our findings can be used in the future to help to transform future science classrooms in ways that motivate and engage learners who aspire to be the next generation of scientists.