Sex and frequency of practical work as determinants of middle-school science students’ learning environment perceptions and attitudes

In this study of 431 Grade 9 and 10 students, we investigated gender and frequency of practical work as determinants of science students’ perceptions of their learning environment and attitudes. We assessed classroom environment with the Science Laboratory Environment Inventory (SLEI) and attitudes with the Students’ Adaptive Learning Engagement in Science (SALES) questionnaire and a scale involving students’ future intentions to study science. The surveys exhibited sound factorial validity and reliability. Interesting differences were found in the learning environment and student attitudes according to student gender and three different frequencies of practical work (namely, at least once a week, once every 2 weeks, or once every 3 weeks or more). More-frequent practical work was more effective than less-frequent practical work in terms of perceived open-endedness, integration and material environment in the laboratory environment and more-positive task value and self-regulation attitudes (with modest effect sizes exceeding one-third of a standard deviation). Although small gender differences existed for some scales, increasing the frequency of practical work was not differentially effective for male and female students.


Introduction
Reviews of contemporary science and STEM education around the world highlight the current crisis of low student achievement and enrolments in science courses and the need for reforms focused on engaging young people in STEM learning (Office of Chief Scientist, 2014;Tytler 2007). The report Realizing Human Potential in the Fourth Industrial Revolution recognizes claims that the world's educational systems are failing to prepare children for the workplaces of the future focused on STEM (World Economic Forum, 2017). Recent changes to global economies have increased the importance of teaching STEM subjects (Kennedy & Odell, 2014) and the need for STEM literacies and skills in employment and society (Naisbitt & Aburdene, 1990). Because declining enrolments in school STEM subjects is a concern internationally, young people's interest in science and other in STEM subjects needs to be stimulated and maintained throughout schooling so that they choose to continue with studies in these fields at the university level, thereby ultimately contributing to redressing the skills shortage in STEM (e.g. Australian National Engineering Taskforce, 2010).
In Australia, where our study was conducted, students' performance in science and mathematics has declined in international or national testing programs such as the Trends in International Mathematics and Science Study (TIMSS) and Programme for International Student Assessment (PISA). For TIMMS in 2015, Australian mathematics students were outperformed by students from 21 countries, including Ireland, England and the USA, Singapore, Hong Kong, Korea, and Japan (ACER, 2016a). For PISA, Australian students recorded a significant decline of 17 points in scientific literacy and 10 points in mathematical literacy between 2012 and 2015 (ACER, 2016b).
Although important, these reviews provide minimal insights into factors that contribute towards the decline in student interest (Velayutham et al., 2011). Although involvement in practical work is reported in these reviews as one of the solutions to increasing student engagement in science (Lyons & Quinn, 2010), little research evidence supports this widespread belief. Some research has identified the differential effectiveness of laboratory classes for males and females (Williams et al., 2003). Because the classroom learning environment affects both the cognitive and affective outcomes of students, the provision of a suitable learning environment is important (Aldridge & Fraser, 2008;Fraser, 2012). Specifically for the laboratory learning environment, numerous researchers have identified links between students' attitudes and achievement and their perceptions of their classroom environment (Fisher et al., 1997;McRobbie & Fraser, 1993;Wong et al., 1995). Obtaining a better understanding of associations between the frequency of practical work and students' perceptions of their laboratory learning environment and their attitudes formed the foundation of this study.
The terms 'practical work', 'laboratory work' and 'experiments' are often used interchangeably, yet "not all practical work is carried out in a laboratory, and not all laboratory work comprises experiments" (Hodson, 1988, p. 53). For the purpose of this study, 'practical work' was defined as a method for learning the processes and skills involved in the development of scientific literacy (Bradley, 2005). Practical work can include: teacher demonstrations, during which students make observations about phenomena and good laboratory practice; traditional laboratory exercises, in which students follow sets of instructions to practise laboratory skills; practical design tasks, in which students design and implement an experimental method to answer a research question; investigations in the field, in which students collect and analyse data from external environments; and open-ended exercises in which students determine their own line of inquiry based on their individual interests, research questions and methods of data collection.
Our main objective was to investigate whether gender and the frequency of practical were determinants of students' (a) perceptions of the science laboratory environment and (b) motivation towards science and aspirations to study science in grades 11 and 12.

Learning environments: Science Laboratory Environment Inventory (SLEI)
There has been growing recognition worldwide that the learning environment is centrally important in education OECD, 2013OECD, , 2017Zandvliet & Fraser 2019), not only because of convincing evidence that it influences a broad range of students' outcomes, but also because a positive learning environment is both a worthwhile end itself and a means to a valuable end (Fraser, 2012(Fraser, , 2019. The growth of the field of learning environments was accelerated by the establishment in 1984 of the American Educational Research Association (AERA) Special Interest Group (SIG) on Learning Environments, which has its own program each year at the AERA's annual meeting. The next landmark in the evolution of the field was the establishment in 1998 of Learning Environments Research: An International Journal published by Springer Nature (previously Kluwer), which was in its 25th volume in 2022 and is rated in Scimago Quadrant 1 in several fields including Education.
A wide range of research methods can be used for assessing and investigating learning environments, including direct observations, surveys of students' and teachers' perceptions, ethnographic and phenomenological approaches, natural inquiry, case studies, narratives, and behaviour settings (Fraser, 1994(Fraser, , 2012. Tobin and Fraser (1998) advocate combining qualitative and quantitative approaches in mixed-methods approaches research on learning environments because using multiple methods not only allows the unique strengths of each different method to emerge, but also reduces or overcomes each method's weaknesses.
Research specifically on classroom learning environments took off about 50 years ago when the work of Walberg & Anderson (1968) and Moos (1974) spawned diverse research programs around the world (Aldridge & Fraser, 2008;Fraser, 2007Fraser, , 2014. During the ensuing decades, numerous questionnaires were developed to assess students' and teachers' perceptions of specific learning environments -such as science laboratories, constructivist classrooms, teacher-student interpersonal interactions, or computer-assisted and online settings -so that currently few fields have such a rich diversity of valid and economical instruments (Fraser, 1998;Fraser et al., 2010;Teh, & Fraser, 1995;Trinidad et al., 2005). These questionnaires have been used extensively in many countries for a wide variety of research applications, especially for evaluating educational programs (Lightburn & Fraser, 2009), investigating associations between student outcomes and the nature of the learning environment (Aldridge et al., 2018;Zandvliet & Frases, 2009) and guiding practitioners' attempts to improve learning environments (Fraser & Aldridge, 2017).
The Science Laboratory Environment Inventory (SLEI), which was used in our study, was created in response to the need to assess the learning environment specific to science laboratory classes, particularly in upper-secondary and higher-education contexts . It measures the five dimensions of Student cohesiveness, Open-endedness, Integration, Rule clarity and Material environment. The initial field testing and validation of the SLEI was unusual in that it simultaneously involved 5477 students in 269 classes in six different countries (the USA, Canada, England, Israel, Australia and Nigeria) .
Since its initial development and cross-national validation, the SLEI has been validated and used in numerous countries and languages. In Florida, Lightburn and Fraser (2009) used the SLEI with 761 high-school biology students in 25 classes in evaluating the effectiveness of using anthropometry activities. This study not only crossvalidated the SLEI, but it also supported the effectiveness of using anthropometric activities in terms of both positive learning environments and student attitudes.
In Korea, Fraser & Lee (2009) developed a Korean-language version of the SLEI to investigate differences between the learning environments of three academic streams (science-independent, science-oriented, humanities). With a sample of 439 high-school students, the Korean version was crossvalidated and it was found that the science-independent stream perceived their laboratory classroom environments more favourably than students in the other two streams.
In two studies in Singapore, the SLEI was modified to form the Chemistry Laboratory Environment Inventory (CLEI) for use specifically in high-school chemistry laboratory classes. Wong & Fraser (1996) administered the CLEI to 1592 grade 10 chemistry students and reported that it was valid for that sample and that CLEI scales were related to a range of student attitudinal outcomes (Wong et al., 1995). The second study in Singapore with a sample of 497 gifted and non-gifted students studying chemistry (Quek et al., 2005) confirmed its validity with gifted students, identified differences between the genders and between gifted and non-gifted students in perceptions of laboratory classroom environments, and reported modest associations between the learning environment and student attitudes to chemistry.
In a study involving 513 senior high-school students in Taiwan, a Chinese-language version of all SLEI scales was crossvalidated via confirmatory factor analysis using structural equation modeling (Lee et al., 2020). Also, the SLEI scales of Open-endedness and Rule clarity were found to be significantly related to student self-efficacy.
Hofstein, Levy Nahum and Shore (2001) used the SLEI in research into inquiry-type laboratories that were implemented in grade 11 chemistry in Israel with 130 students in an inquiry group and 185 students in a comparison group. Relative to the comparison group, the inquiry group perceived the actual laboratory environment more positively and had smaller differences between their actual and preferred laboratory learning environments.
When a Malay version of the SLEI was used with 800 science teachers in 100 secondary schools (half urban and half rural) in Selangar, Malaysia, satisfactory reliability for SLEI scales was reported by Che Nidsam et al., (2013). As well, positive associations were reported between teachers' satisfaction with teaching and their perceptions of their science laboratory learning environments as assessed with the SLEI.
A shortened four-scale version of the SLEI in the Spanish language (the SASLEI) was administered to 174 first-year undergraduate biology students in a college of health sciences in Spain (De Juan et al., 2016). This Spanish SASLEI was validated by conducting factor analysis and reliability estimation. Aladejana & Aderibigbe (2007) administered the SLEI in Nigeria to 328 senior-secondary chemistry students randomly selected from four purposively-selected schools. The authors concluded that students could assess the SLEI's five components, which were found to be related to students' academic performance.
In a study of 371 students from 11 high schools in Israel, the SLEI was used to reveal numerous interesting differences (Hofstein et al., 1996). Significant differences emerged between actual and preferred environment for all scales; gender differences were relatively small for actual environment, but larger for preferred environment; and the laboratory environments of chemistry and biology classes were perceived significantly differently in terms of Integration and Open-endedness. Significant differences between perceptions of actual and preferred learning environment also were reported for a sample of 109 university entomology students in Florida (Burleson & Myers, 2013).
The SLEI also has been crossvalidated in Australia. With 489 senior high-school biology students in 28 senior biology classes in 8 schools in Tasmania (Fisher et al., 1997;Henderson et al., 2000), student-perceived learning environment based on the SLEI was more strongly associated with attitudinal outcomes than to either achievement or practical skills. With 1594 senior high-school chemistry students in 92 classes in Queensland, the SLEI exhibited sound factorial validity and significant associations with students' cognitive and affective outcomes McRobbie & Fraser, 1993).
Our study contributes to the field of learning environments research by using the SLEI in middle-school classrooms, where it has had limited use (with exceptions being the studies of Che Nidsam et al., 2013;Fraser & Lee, 2009;Lightburn & Fraser, 2007), and by applying learning environment criteria for the first time in evaluating the effectiveness of increasing the frequency of practical work in science education.

Attitudes
Attitudes have been central in psychology for a long time (Allport, 1935;Schwarz & Bohner, 2001) and, in science education specifically, the measurement and investigation of students' attitudes have a long history and continue to be of contemporary interest internationally (Fraser, 1977;Khine, 2015;Saleh & Khine, 2011;Tytler & Osborne, 2012;Kind et al., 2007) identified numerous important and long-standing problems plaguing many existing attitude questionnaires (e.g. lack of clear descriptions of the constructs to be measured; combining conceptually-different constructs to form one unidimensional scale; and low reliability), which were overcome in the Test of Science-Related Attitudes (TOSRA; . The main instrument used to assess students' attitudes to science in our study was the Students' Adaptive Learning Engagement in Science (SALES). Velayutham et al., (2011) developed the SALES questionnaire to assess students' motivation and self-regulation in science learning with four scales: Learning goal orientation, Task value, Self-efficacy and Self-regulation. The SALES questionnaire was validated by Velayutham et al., (2011) with a sample of 1360 students in 78 classes across grades 8, 9 and 10. Because the SALES questionnaire is a new instrument, it was further validated in our study.
Modified versions of the three SALES scales of Task value, Self-efficacy and Self-regulation were used among 327 college students in the USA, the United Arab Emirates and Turkey. UAE students were provided with both Arabic and English versions. Both explor-atory and confirmatory factor analyses supported the SALES' factorial validity (Pasha-Zaidi et al., 2019).
In Afghanistan, Sayed & Fraser (2019) administered a Dari translation of one SALES scale (namely, Self-regulation) to 1610 grade 10-12 science students in a range of school sizes and of various ethnic and social groups. All schools were single-gender because there is no co-education in Afghanistan. The Self-regulation scale was found to have the high reliability value of 0.88 and to be related to learning environment scales.
For the first time, the SALES was used among gifted students by Koren & Fraser (2019). The sample consisted of 495 middle-school students in 31 classes in 10 schools in Florida. All four SALES scales were used but they were modified somewhat (e.g. specific reference to the school subject of 'science' was removed and explanatory text was added in parentheses to some items to clarify meaning). Factor analysis supported the structure of this version of SALES, with all items having a factor loading of 0.40 or greater on its own scale and less than 0.40 on all other scales. The total proportion of variance explained was approximately 60%. Each scale's alpha reliability was high and ranged from 0.89 to 0.91.
In Singapore, Pramathevan & Fraser (2020a) used three of the SALES' four scales (namely, Task value, Self-efficacy and Self-regulation) in evaluating technology-based science teaching among a sample of 722 grade 10 students. Approximately half of the sample were in technology-based science classrooms and the other half were in regular science classes. This version of the SALES exhibited strong factorial validity and reliability for this Singaporean sample (Pramathevan & Fraser, 2020b).
Robinson & Aldridge (2022) used a translated Arabic version of the SALES among 291 grades 6-9 female mathematics students in four schools in the United Arab Emirates. Only the two scales of Self-efficacy and Task value were chosen. The independence of these two sales was supported by factor analysis and their reliability was supported by Cronbach alpha coefficients.

Gender differences
Among STEM and science education researchers, there has been much historical and contemporary interest in gender differences in students' achievements, attitudes and learning environment perceptions (Scantlebury, 2012;Parker et al., 1996;Koch et al., 2014;Young & Fraser, 1994).
Prior studies document more-positive attitudes among boys than girls towards integrated STEM or areas within STEM such as science or mathematics (e.g. Weinburgh 1995). Analysis of TIMSS data from OECD and non-OECD countries revealed that, relative to girls, boys had more-positive attitudes towards science and mathematics and stronger self-efficacy beliefs (Reilly et al., 2019). In response to gender imbalance in attitudes and enrolment rates in STEM subjects, researchers have tried to explain females' lower level of participation (e.g. Cheryan et al., 2017;Koul et al., 2021;McLure, Koul & Fraser, in press;Penner & Willer 2019).
Numerous studies of gender differences in perceptions of classroom environment in science, mathematics and technology over the past decades have revealed gender differences in favour of females (e.g. Hofstein et al., 1996;McRobbie et al., 1990;Welch et al., 2014), but these differences tend to be relatively small and statistically-nonsignificant for most learning environment scales.
Research has shown gender to be a strong predictor of science subject choice, with physics classes being dominated by boys, biology classes by girls, and chemistry classes being fairly evenly balanced in terms of student gender (Ainley et al., 2008;Fullarton et al., 2003). One of the most promising explanations for the imbalance in physics enrolments focuses on students' self-efficacy in science and the interaction of this with their conceptions of the relative difficulty of physics (Lyons & Quinn, 2010).
The Programme for International Student Achievement (PISA) (2006) revealed that, despite achieving similar scores for scientific literacy, Australian girls reported lower levels of self-efficacy, self-concept, enjoyment of science, instrumental motivation, and future orientation to study or work in science than boys. Females expressed more-favourable attitudes to the importance of doing well in science (Thomson & DeBortoli, 2008). The 2015 PISA survey showed that Australian boys had higher interest and self-efficacy in science than Australian girls, despite achieving similar scores for scientific literacy (Thomson et al., 2017).
When Williams et al., (2003) researched gender differences in attitudes to practical work, physics and biology students considered practical work interesting, but that males valued practical exercises more than females. However, females valued the relevance of physics more than males.
Hofstein & Mamalok-Naaman (2011), in their review and analysis of the literature about attitudes towards and interest in learning chemistry, concluded that more attention should be drawn to differences between learners (e.g. gender, motivational patterns, and learning styles). In terms of gender differences in attitudes in prior research, more-favourable attitudes were exhibited by females in some studies but by males in other studies.

Science practical work and its frequency
Although laboratory teaching is widely recognized as a hallmark of science education, concerns remain about whether the considerable staffing and equipment costs are really justified (Boud et al., 1986;Hegarty-Hazel, 1990;Tobin, 1990;Woolnough, 1991). Research on science laboratory instruction has been inconclusive in establishing its claimed advantages (Hofstein & Kind, 2012;Hofstein & Lunetta, 1982, with some reasons proposed as being teachers' lack of the skills needed to teach it (Layton, 1989) and the closed-ended nature of activities that involve students in investigations for which they already know the desired answer (Hodson, 1988).
Little research evidence supports the belief that frequent, authentic practical work enhances the engagement of students. Bradley (2005) studied 537 years 8-10 students in Tasmania and Western Australia who perceived that science practicals are essential for science learning and wanted an increase in the amount of practical work.
Chen & Howard (2010) completed a study of the effect of live simulation on students' science learning and attitudes with 311 middle-school students. The simulation allowed them to access and interpret satellite data and images and to design investigations. The findings included positive changes in students' attitudes and perceptions towards scientists. Male students adopted more positive scientific attitudes than female students. Rennie et al., (2001) reported that open investigations had not penetrated Australian schools. Of their sample of 2802 students, 33% of students stated that they never planned and completed their own experiments, and a further 25% of students stated that they com-pleted their own experiments only once a term or less. 70% of the students surveyed reported that they completed an experiment by following instructions about once a week or more. Shepardson and Pizzini's (1994) study of 268 year 7 and 8 students established an association between inquiry and perceptions of science classes. Students were more positive about learning science when teachers used an inquiry-based approach rather than workbooks or traditionally-structured laboratory classes.
An inquiry-based laboratory approach was the focus of a study by Hofstein, Shore and Kipnis (2001) who used the SLEI with Israeli high-school chemistry students to compare the perceptions of students who experienced an inquiry-type laboratory learning environment with the perceptions of students enrolled in a more-conventional laboratory course. Students in the inquiry group found the laboratory learning environment more open-ended and more integrated with the conceptual framework that they were developing than did students enrolled in the conventional course. These findings suggest that inquiry-type practical experiences can promote a positive and healthy learning environment.
Sheldrake et al., (2017) published a study about the importance of conveying the applications and relevance of science. In England, the Programme for International Student Assessment (PISA), which involved 4935 students across 171 schools in 2006 and 5194 students across 206 schools in 2015, asked students about the kind of job that they expected to have when they were around 30 years old. The analysis involved coding these into 'science-related career aspirations'. Various science-specific attitudes and beliefs were measured across both surveys, including students' generalized interest in science and perceived utility of science. Teaching approaches were measured across both surveys, and specifically covered students reporting the frequency of receiving or encountering 'interaction/debate', 'practical/handson activities', 'student-led investigations' and teaching that focused on the 'applications of science'. Explaining the wider context or applications of science improved students' interest and perceived utility of science, but inquiry-based teaching approaches (including handson/practical activities and student-led investigations) did not lead to higher student interest.

Instruments
Data collection involved the administration of two questionnaires: 1. All five scales from a modified version of the SLEI assessed students' perceptions of their learning environment (Cohesiveness, Open-endedness, Integration, Rule clarity and Material environment). 2. Attitudes were assessed with all four scales from the SALES questionnaire to measure student self-reports of engagement in science classes and the value that students place on learning science (Learning goal orientation, Task value, Self-efficacy and Selfregulation). Additionally, to assess students' aspirations, the Career Interest in Science scale from the Test of Science-Related Attitudes (TOSRA, Fraser 1981) was modified to form a Future Intentions to Study Science scale. The wording of TOSRA items was changed to include a broader range of careers that require an assumed knowledge of senior-secondary science subjects.
The frequency of practical work undertaken by each student was assessed by asking students whether, in their science lessons, practical work is completed: (1) on average, once every three or more weeks; (2) on average, once every two weeks; (3) on average, at least once a week."

Sample
The sample consisted of 431 year 9 and 10 science students from two typical Catholic and two typical independent schools in Adelaide, South Australia. The schools were coeducational and students were of mixed abilities.

Data analysis and interpretation
The structure of the SLEI and attitude questionnaire (SALES and Future Intentions scale) was examined using principal axis factor analysis with varimax rotation and Kaiser normalisation. The internal consistency reliability of each scale was assessed using Cronbach's alpha coefficient.
To investigate the statistical significance of differences based on the independent variables of gender and frequency of practical work, MANOVA was conducted with the set of 10 dependent variables (namely, five SLEI scales, the Future Intentions scale and four SALES scales). If the multivariate test yielded statistically-significant results in terms of Wilks' lambda criterion, then the univariate two-way ANOVA would be interpreted for each individual scale. In addition, effect sizes were used to provide information about the magnitude of differences between genders or different frequencies of practical work. Cohen's d effect size is calculated by dividing the difference between two means by the pooled standard deviation; it expresses a difference in standard deviation units.

Validation of the SLEI and attitude questionnaire
To check the appropriateness of our data for factor analysis, the Kaiser-Meyer-Olkin (KMO) test for sampling adequacy and Bartlett's test of sphericity were conducted. As a measure of sampling adequacy, the KMO test specifies the proportion of variance among variables that is common, with a minimal cut-off value of 0.5 recommended by Cerny & Kaiser (1977). Bartlett's test of sphericity can be used to check that the correlation matrix generated from a dataset is an identify matrix, suggesting that variables are unrelated (Tobias & Carlson, 2010). In our study, the KMO measure of sampling adequacy for the learning environment and attitude scales was high and above 0.9, confirming the appropriateness of our data for factor analysis. Although Bartlett's test of sphericity indicated that χ 2 (861) = 7839, p < 0.01, suggesting that the data were suitable for reduction, we retained all items to permit comparison of my results with past research.
The criteria used for retaining any SLEI item were that it must have a factor loading of at least 0.40 for its a priori scale and less than 0.40 with each other scale. The application of these criteria led to the retention of all items except SC7 and OE10 (Table 1 in Appendix A).
After the removal of SC7 and OE10, all items supported factorial validity. The percentage variance accounted for by different SLEI scales varied from 5.13 to 24.94% with a total of 52.78% of the variance accounted for. Eigenvalues varied from 1.79 to 8.72 for different scales, which exceed the minimum value of 1.00 recommended by Kaiser (1960).
Likewise, the criteria used for retaining any SALES or Future Intentions item were that it must have a factor loading of at least 0.40 for its a priori scale and less than 0.40 with each other scale. All attitude scales exhibited satisfactory factorial validity ( Table 2 in Appendix B), with the percentage variance ranging from 4.96 to 42.18% for different attitude scales and with a total variance of 66.36%. The eigenvalue varied from 1.88 to 16.03 for different attitude scales.
With the removal of SC7 and OE10, all learning environment and attitude scales exhibited sound internal consistency reliability (see the bottom of Tables 1 and 2 in Appendices A and B), replicating past research with the SLEI Fraser & Lee, 2009) and the SALES questionnaire (Velayutham et al., 2011).

Student gender and frequency of practical work as determinants of learning environment and attitudes
Prior to conducting MANOVA, the suitability of data for analysis was checked using Levene's Test of Equality of Error Variances to confirm that the error variance for each of the 10 dependent variables (classroom environment and attitudes) was equal across groups (genders and frequencies of practical work). The F value from Levene's test for each of the 10 dependent variables was statistically nonsignificant (p < 0.05), indicating that error variances were equal across the comparison groups for all scales and therefore that the data were suitable for analysis via MANOVA. Additionally, Box's Test of Equality of Covariance Matrices was conducted separately for the set of 10 dependent variables to check whether the observed covariance matrices of the dependent variables were equal across the comparison groups. Statistically-nonsignificant results (p < 0.05) for the set of 10 classroom environment and attitude scales suggest that our data's covariance matrices were adequate for conducting MANOVA (Huberty & Petoskey, 2000).
To investigate the statistical significance of differences between student genders and between different frequencies of practical work, a two-way MANOVA was performed for the set of five learning environment and five attitude scales as the dependent variables. MANOVA was initially conducted to reduce the Type I error rate that would be associated with performing a separate univariate test of differences between groups for each individual dependent variable. When MANOVA was conducted using SPSS for our sample of 431 students, the multivariate test using Wilks' lambda criterion yielded a statistically-significant overall difference, F(10, 420) = 10.18, p < 0.01. This justified interpreting the separate oneway ANOVA for each of the 10 individual learning environment and attitude scales.
Although MANOVA revealed a significant multivariate effect for both student gender and frequency of practical work, the interaction between gender and frequency was nonsignificant. This suggests that having more-frequent practical work was not differentially effective for male and female students. Consequently, univariate ANOVA results were not considered further for the gender x frequency interaction.
For frequency of practical work, the univariate ANOVA results for each of the 10 learning environment and attitude scales in Table 1 reveals a statistically-significant difference (p < 0.05) between different frequencies of practical work for three learning environment scales (Student cohesion, Open-endedness, Material environment) and the two attitude scales (Task value, Self-regulation). For each learning environment and attitude scale, Table  1 also shows its average item mean and average item standard deviation for the three different frequencies of practical work (once every three or more weeks; once every two weeks; and at least once a week).
In order to explore further the differences between each pair of different laboratory work frequencies, effect sizes were calculated as recommended by Thompson (2001) to portray the magnitude of differences in standard deviation units. Effect sizes provide a measure of the magnitude or importance of differences in scale scores between different practical work frequencies. Effect sizes were calculated by dividing the difference in average item mean for each pair of practical work frequencies by the pooled standard deviation for those practical work frequencies. Also Tukey's HSD multiple comparison procedure was used to determine the statistical significance of differences between each pair of frequencies of practical work. Table 2 shows the magnitudes and statistical significance of differences between pairs of laboratory work frequencies. For simplicity, Table 2 has been restricted to the scales for which there were statistically-significant overall differences based on frequency of practical work (see Table 1). The results in Table 2, for comparisons of different frequencies of practical work, are summarized below.
For the comparison of the lowest frequency of once every 3 weeks or more (≥ 3 wks) with a higher frequency of approximately once every 2 weeks (2 wks), differences were statistically significant for all five scales (Open-endedness, Integration, Material environment, Task value and Self-regulation), with effect sizes ranging from 0.31 to 0.43 standard deviations.
For the comparison of the frequency of approximately once every 2 weeks (2 wks) with an even higher frequency of at least once a week (≤ 1 wk), differences were statistically significant but small for the three scales of Open-endedness (effect size of 0.15 standard .13* *p < 0.05, **p < 0.01, N = 121 (≥ 3 wks), 204 (2 wks), 106 (≤ 1 wk) ≥ 3 wks: practical work frequency of on average once every three or more weeks 2 wks: practical work frequency of on average once every two weeks ≤ 1 wk: practical work frequency of on average at least once a week deviations), Integration (0.10 standard deviations) and Self-regulation (-0.18 standard deviations). It is noteworthy that the negative effect size of -0.18 for Self-regulation suggests that the higher frequency of at least once a week was slightly less advantageous than the lower frequency of approximately once every 2 weeks.
For the comparison of the lowest frequency of once every 3 weeks or more (≥ 3 wks) with the highest frequency of at least once a week (≤ 1 wk), differences were statistically significant for all five scales (Open-endedness, Integration, Material environment, Task value and Self-regulation), with effect sizes ranging from 0.14 standard deviations for Self-Regulation to 0.54 standard deviations for Open-Endedness.
The main overall pattern suggest in Table 2 is that increasing the frequency of practical work from once every 3 weeks or more (≥ 3 wks) to about once in 2 weeks was significantly effective, with modest effect sizes of at least one-third of a standard deviation, in terms of more-positive scores for Open-endedness, Integration, Material environment, Task value and Self-efficacy. However, a further increase in the frequency of practical work from once in 2 weeks (2wks) to at least once a week (≤ 1 wk) seems to have been associated with 'diminishing returns' in that effect sizes for significant differences were small (less than 0.2 standard deviations) and in a negative direction for one scale. Figure 1 depicts in graphical form the mean values for student perceptions of the laboratory learning environment and attitudes to science for each frequency of practical work. This graph demonstrates that the frequency of practical work was a determinant of morefavourable learning environment and attitude scores on numerous scales. Table 3 provides, for gender differences, results for the same types of analyses as reported in Table 1 for differences between frequencies of practical work. Statistically-significant but relatively of small gender differences in Table 3 exist for the learning environment scales of Integration (effect size of 0.19 SDs) and Rule clarity (0.18 SDs), with females being more positive about their laboratory learning environment than males, and for the attitude scales of Future intentions (effect size of 0.23 SDs) and Self-efficacy (0.28 SDs), with males expressing more positive attitudes. This tentative pattern of small gender differences in favour of females for learning environment scales, but in favour of males for attitudes to science, is consistent with some past learning environment research (Quek et al., 2005;Thomson & DeBortoli, 2008 Task value 0.43*** -0.04 0.37*** Self-regulation 0.31** -0.18* 0.14* *p < 0.05, **p < 0.01, *** p < 0.001 N = 121 (≥ 3 wks), 204 (2 wks), 106 (1 wk) Table 2 Effect size and Tukey's HSD multiple comparison for statistical significance of difference between each pair of frequencies of practical work for learning environment and attitude scales for which significant overall differences were found Figure 2 depicts in graphical form the mean value of each laboratory learning environment and attitude scale for males and females. This graph demonstrates that females generally had somewhat more-positive perceptions of their learning environment than males, but males had more-positive attitudes than females. However, it is important to emphasise that the nonsignificant interaction found between gender and frequency of practical via MANOVA suggests that increasing the frequency of practical work was not differentially effective for males and females (Figure 2).

Significance and conclusion
This study was unique in its focus on the relationship between the frequency of practical work and students' perceptions of their laboratory environment and their attitudes towards science. Although practical work is identified in literature reviews as one of the solutions for increasing student engagement in science (Lyons & Quinn, 2010), surprisingly little research has been conducted in this area. Also, relatively few studies in the field of learning environments have been conducted using the SLEI in middle-school science laboratory classes (with exceptions being Che Nidsam et al., 2013;Fraser & Lee, 2009;Lightburn & Fraser, 2007), and the Science Laboratory Environment Inventory (SLEI) has not previously been validated with middle-school students specifically in Adelaide, South Australia; our study addressed these gaps.
Analyses of data from our sample of 431 grade 9 and 10 science students further supported both the factorial validity and internal consistency reliability our five-scale SLEI (assessing Student cohesiveness, Open-endedness, Integration, Rule clarity, and Material environment) and our five-scale attitude instrument (assessing Future intentions to study sci-  ence and the four SALES scales of Learning goal orientation, Task value, Self-efficacy, and Self-regulation). Now that the SLEI and SALES have been validated for use with middleschool students in South Australia, they could be beneficially used by other middle-school science teachers who want to assess their students' perceptions of their laboratory learning environments and attitudes in order to provide the most-positive learning environment and attitude outcomes for students. These validated questionnaires could also be used by teachers for evaluating the effectiveness of instructional strategies and materials designed to improve students' laboratory learning environments and science attitudes.
Our study supported the value of science practical work in that more-frequent practical work (i.e. about once every 2 weeks) was more effective than less-frequent practical work (i.e. once every 3 weeks or more) for the three learning environment scales of Open-endedness, Integration and Material environment and the two attitude scales of Task value and Self-regulation. For these five scales, differences between different frequencies of practical work were statistically significant (p < 0.01) and of modest magnitude (at least one-third of a standard deviation). However, further increasing the frequency of practical work beyond twice a week to at least once per week was associated with diminishing returns in terms of further improvements in learning environment and attitude scales. Interestingly, our analyses also indicated that increasing the frequency of practical work was not differentially effective for male and female students.
Teachers could use information about differences identified in our study to focus their teaching practices on ways that can lead to the laboratory learning environment being perceived as positively as possible by their students and to improve student attitudes and aspirations towards studying science. Our study suggests that one way in which teachers can improve the laboratory learning environment and students' attitudes is to include more frequent practical work as part of their science teaching. Students who undertook morefrequent practical work perceived aspects of their laboratory learning environments more positively than students who undertook less-frequent practical work. Students also placed a higher value on learning science and were more inclined to stay on-task and complete set activities than students who experienced less-frequent practical work.
This study leads to suggestions for future research. Our study revealed some interesting findings for the three frequencies of practical work investigated. Further research could be conducted to identify the optimum frequency of practical work for the most-positive results for a variety of student outcomes. In particular, the effect of the frequency of practical work on student achievement could be studied in future research. The quality of the practical work undertaken by the students was not assessed in our study whose focus was only its frequency. Future research could investigate differences between different frequencies of practical work for different types of practical work (e.g. teacher demonstrations, closed investigations, open investigations and free-choice investigations) in terms of students' perceptions of their laboratory environments, attitudes and aspirations. The quantitative nature of our study restricted our opportunity to provide explanations for differences between the sexes and frequencies of practical work. Further research involving qualitative methods could enhance our findings, provide some insight into reasons for the existence of these differences, and allow triangulation of findings from using qualitative and quantitative methods.   Funding Open Access funding enabled and organized by CAUL and its Member Institutions

Conflict of interest Authors declare no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.