Differences in cognitive performance between boys and girls have been reported in the literature for many years (Maccoby & Jacklin 1974, Bell 2001), but their interpretation remains contentious (Feingold 1992a). Important issues have arisen over such matters as: the methods to be used to assess girls' and boys' performances on cognitive ability and achievement tasks; the appropriate statistical procedures to be used to compare those performances, including meta-analytic procedures; and whether the sex differences are cultural or biological in origin. Although some reported differences are relatively large, there is recent evidence of progressive changes in girl/boy relativities (Nowell & Hedges 1998), suggesting that much of the sex difference in cognitive performance is attributable to non-biological factors. The present research on chemistry problem solving in secondary students is positioned within this literature.

As with some other areas of science (Lee & Burkham 1996), and mathematics (Beller & Gafni, 1996; Hedges & Nowell, 1995), the evidence with which the present research commenced indicated that boys' achievement in chemistry was higher, on average, than girls', and that this difference was not attributable to item difficulty confounds (Walding, Fogliani, Over, & Bain, 1994). Of the various explanations considered for this effect, the most promising appeared to be sex-correlated differences in science-related interests and in orientations to sex roles. For example, many students engage in activities, interests and sports that provide them with experiences relevant to science. Such experiences are likely to result in an incidental knowledge and understanding of these domains that the students might otherwise not possess. Sex differences in these science-related experiences may be associated with sex differences in incidental learning and ultimately with performance on formal chemistry tests. A student's sex role orientation, irrespective of biological sex, may also influence her/his experiences of the world in ways that influence the uptake and use of science-related knowledge. The main research reported in this thesis examined these possibilities, culminating in a structural equation model demonstrating sex influences on chemistry problem solving partly mediated by science–related knowledge and sex role orientation.

An empirical test of science-related incidental knowledge was developed and trialled with 542 secondary school chemistry students. Clear evidence was found for sex differences in this knowledge (some areas favoured girls, others boys), and the differences thus revealed were used to construct a chemistry achievement test with three sub-scales: one biased in favour of girls' incidental knowledge, another biased in favour of boys, and a third consisting of neutral items. Performance on this test was correlated with students': science-related incidental knowledge; their sex role orientation (Antill & Cunningham, 1982); and their enrolment status for a concurrent physics course (because earlier research suggested this also might be a relevant predictor of chemistry achievement —Walding et al., 1994).

Preliminary analyses showed that two of the three chemistry sub-tests performed as expected. The boy-gendered subtest produced a significant difference in scores favouring boys and with the neutral subtest there was no significant difference between boys' and girls' scores. However, the girls' subtest was expected to show a sex difference favouring girls but this did not eventuate.

As mentioned above, two main variables were examined as possible factors giving rise to this sex difference in chemistry achievement. The correlation between performances on the boys' science-related incidental knowledge sub-test and the boy-gendered chemistry sub-test supported the notion that incidental knowledge influences achievement on formal chemistry tests. However, no such effect was found for the girls' science-related incidental knowledge and the girl-gendered chemistry sub-test. Although it was found that the science-related incidental knowledge scale was clearly girl-biased, this was not the case with the girl-gendered chemistry sub-test as mentioned above. The absence of a relationship between the two may reflect the lack of a girl bias in the girl biased chemistry sub-test or an absence of a relationship per se.

The other variable thought likely to give rise to sex differences in formal chemical knowledge was sex role orientation. Sex role orientation – defined as masculine, feminine, androgynous and undifferentiated – was associated with differential performance on the chemistry test. Firstly, highly masculine students tended to perform significantly better than weakly masculine students on the boy gendered chemistry questions whereas highly feminine students performed significantly better than weakly feminine students on the girl gendered chemistry questions. Lastly, as anticipated, there was no sex role differentiation on the unbiased chemistry questions.

However, some inadequacies were apparent with the preliminary analyses and four additional variables were included. Gendered knowledge appeared to be a factor contributing to sex differences in chemistry achievement but the influence of total incidental knowledge was uncertain. Secondly, the effect of masculine and feminine sex orientations was apparent but the influence of non-traditional sex roles was not clear. Both these variables needed to be accounted for in future analyses. The other aspects of interest were the influence of the concurrent curriculum (in this case physics) on chemistry performance and the relationship of the chemistry subscales to a more traditional test of chemistry achievement.

The main analysis was motivated by two purposes: one was to take account of all variables simultaneously. Although a variety of commonly used statistical methods were applied to the data to test and control factors for one another‘s joint influence, they did not take account of measurement error, relationships among predictor variables and unequal interval scaling of independent and dependent variables. For better estimates of effects structural equation modelling was adopted. It allows the determination of construct validity of measures and the fit of the model, none of which could be handled by classical regression analysis. The second aim was to incorporate non-gendered aspects of incidental knowledge and sex role orientation in addition to the gendered aspects.

To achieve these aims, a complex, multistage, mediated-effects latent variable structural model was hypothesised. The measurement design consisted of latent variables for gendered (bipolar) and non-gendered science related incidental knowledge, traditional (bipolar) and non-traditional sex roles, and a latent variable for general chemistry achievement all of which used the variables described earlier as indicators. The structural design consisted of the direct and indirect influences of sex on each of the chemistry sub-tests, with the indirect effects being defined via gendered and non-gendered science related incidental knowledge latent variables, traditional and non-traditional sex roles latent variables, and a concurrent physics enrolment variable.

Several important conclusions were drawn from the structural model. One of the most obvious was that some of the chemistry test performance was sex biased and some not. Some of the influence of sex on these biased parts of the test may derive from prior sex-biased interests in science-related knowledge – an effect resulting from a knowledge difference. However, other influences of sex on the biased parts of the chemistry test may derive from students' sex role orientations – a gender difference. Irrespective of these two 'incidental curriculum' influences – science-related incidental knowledge and sex role orientation – some of the influence may be 'direct', in the sense that it is not attributable to these two major mediating factors under investigation. One interpretation is that there are unmeasured correlates of sex that influenced chemistry achievement; these could include spatial and visual perception (Linn & Petersen 1985), verbal ability (Hyde & Linn 1988), or mathematical ability (Hyde, Fennema & Lamon 1990). However, it is acknowledged that the problem with including more and more mediating variables that correlate with sex and target achievement (in this case the chemistry subtests), is that more and more imperfect substitutes for sex are being added. Hence, it may be possible to eliminate all of the direct sex effect but by splitting it into various surrogates for sex, all imperfect but collectively sufficient to replace sex. It is also possible, however, that neither of the gendered latent variables was optimally defined: a better measurement model might have resulted in a better structural model.

As well as the gendered effects of science-related incidental knowledge (which showed that gendered chemistry subtests show sex differences in response to the variety of students‘ knowledge), there were non-gendered effects suggesting that amount of prior relevant knowledge is important for some aspects of chemistry achievement. In cognate areas such as physics, formal courses might act in the same way as the 'incidental curriculum', both being correlated with sex, and perhaps being responsible for differences in chemistry performance, but the mechanism for the influence of incidental and formal knowledge remains an open question.

In defining the contributions to the advancement of theory and implications for theoretical advance, the mechanism that underlies the influence of science related incidental knowledge on the learning and assessment of formal chemistry is postulated: one concerns a knowledge uptake effect in which the incidental knowledge of events and phenomena is crucial; the other is a testing effect in which the vagaries of assessment and the notion of fair testing are paramount. In this way the implications for future research are elaborated. But more than build an understanding of the nature and influences of sex differences in cognition – the question is posed: what can we do about it; how can practice be changed? Various possibilities are canvassed, culminating in a description about how the new syllabuses in chemistry and physics in Queensland have been written with these questions in mind.