Exploring the pre-instruction gender gap in physics.

There is a substantial body of work in physics education looking at gender disparities in physics. Recent work has linked gender disparities in college physics course performance to disparities in high school physics preparation, but to our knowledge, the origin of the disparity in high school physics preparation is still underexplored. In a select sample, we found that women on average had lower force and motion conceptual evaluation (FMCE) pre-scores (the FMCE is a short conceptual assessment of Newton's laws), and FMCE pre-score entirely mediated the effects of high school preparation and social-psychological factors on exam performance. The gender gap in FMCE pre-scores could not be explained by differences in the number of physics courses taken in high school. Instead, we find that the gender gap in the FMCE is partially explained by female students' higher levels of general test anxiety. We hypothesize that the format of the FMCE, a timed assessment, triggers stereotype threat in female students despite being a low-stakes assessment. Therefore, instructors and researchers should take care in interpreting the results of such concept inventory scores and should re-think the way they assess understanding of physics concepts. Results of this work aligned with previous findings on gender disparity in timed exams call upon investigating gender equitable assessment formats for evaluating physics knowledge to replace timed assessments, either high or low stakes.

Introduction

Many demographic groups are currently underrepresented in STEM fields [1,2]. To address this under-representation, we need to make sure the paths to choose and excel in STEM fields are accessible to underrepresented demographic groups [3]. STEM introductory courses are an influential part of this path [4]. Previous works have shown that student performance in introductory STEM significantly affects STEM retention [5]. Furthermore, low academic performance in these courses can also have negative psychological consequences for marginalized students (Nardo et al., in press). One of the main introductory STEM courses is calculus-based physics 1. This course is offered in all STEM higher education institutions with similar course content and a pre-requisite for many STEM fields, and hence a gatekeeper for them. The interviews with students who have left STEM fields in the last two decades have shown that challenges in this course is one of the main reasons for leaving STEM fields despite their initial aspirations [5].

Firstly, previous studies have shown that the common instructional practices in physics 1, like other introductory courses, are particularly disadvantageous for marginalized students. The strongest predictor of students’ performance in these courses, including physics 1, is STEM preparation from high school experiences, referred to as incoming preparation hereafter. Many marginalized students attend high schools which are under-resourced particularly in STEM fields [6,7], leading to lower STEM preparation. Salehi et al. [3,8] have shown that while marginalized students underperform in physics 1 across three vastly different institutions, these demographic performance gaps can almost entirely be explained by the students’ lower incoming preparation. After controlling for variation in incoming preparation of students, there was no difference in performance of marginalized and over-represented students. In their work, they had two measures of incoming preparation: 1) general incoming preparation as measured by Math SAT/ACT–a standardized college entrance exam in the United States, 2) physics-specific incoming preparation as measured by physics concept inventories, such as the Force and Motion Conceptual Evaluation (FMCE). They show that while general incoming preparation can explain most of the performance gaps for non-white and non-Asian under-represented racial minority (URM) and first-generation students, the gender performance gaps are mainly due to lower physics-specific incoming preparation. For women, lower scores on the math SAT/ACT could explain only about one-third of the gap, with physics-specific incoming preparation explaining the remaining gap.

Secondly, previous studies have shown that high-stakes exams disadvantage women in introductory STEM courses, while in non-exam assessments (e.g., homework, quizzes), if anything, women overperform compared to men [9,10]. These studies have shown that women have higher test anxiety regardless of their incoming preparation and test anxiety negatively affects their exam performance. For men, on the other hand, test anxiety was negatively correlated with their incoming preparation and did not affect their exam performance when controlled for incoming preparation. Similar findings have been replicated across different STEM courses [11]. In this study, we expand on the intersection of these two lines of work. We examine what are the root causes of women’s lower physics-specific incoming preparation, and whether the measurement format impacts the gender gap in both incoming preparation and physics 1 performance.

Theoretical framework

Physics education researchers have long thought about the gender disparities in physics, particularly when it comes to representation and course performance. Indeed, representation is an issue that starts before students enter college, with fewer women on average taking advanced physics courses [12]. Sax et al. [13] discuss several potential explanations for these observed gender disparities: familial expectations and beliefs, K-12 experiences, psychological factors, and perceptions of physics as a field. Familial expectations set by parents’ influence on their children’s course taking, as well as the careers of the parents, may result in fewer women taking advanced science and math coursework in high school–either due to lack of interest or being discouraged from doing so [14]. Lack of K-12 preparation as well as negative experiences in those environments can also dissuade women from pursuing STEM [15]. These experiences also impact things like self-efficacy and interest at the college level [16].

Most studies of gender in physics have focused primarily on the social psychological aspects of interest and persistence [17] despite the clear evidence that these are deeply intertwined with prior academic experiences and achievement. We adopt a version of Social Cognitive Career Theory (SCCT) in which learning experiences, background characteristics, contexts, and performance are interrelated and dynamic [3,18]. In this model we hypothesize that background characteristics (socioeconomic status, parental education and careers) influence students’ learning experiences. For example, students from lower socioeconomic strata may not have access to the same quality of instruction as students from higher strata. These early learning experiences influence students’ psychology such as their motivation, self-efficacy, interest in a subject, test-anxiety. There is then a feedback loop between psychology and learning experiences that impact students throughout their academic careers. This can be reinforced or mitigated by contextual influences.

This theory, along with our prior empirical work, inform our research questions. We assume that students’ background characteristics (in this case, gender) will influence their prior academic experiences and social psychological profile. For example, women are less likely to take AP physics courses on average nationally. They may also be more likely to suffer from test anxiety and/or stereotype threat due to early academic experiences. These factors will influence the students’ level of preparation for college physics. We then test whether this level of preparation is the sole mediator for future physics achievement, or whether these social-psychological factors and personal characteristics also play some direct role. Our research questions are thus:

Are there gender disparities in physics1 performance as measured by exam performance? (RQ1)

Are there gender disparities in physics 1 incoming preparation as measured by FMCE pre-test? (RQ2)

What learning experience, background characteristics, and contextual factors contribute to these gender disparities (i.e., what are the mediators of gender disparities in exam score and FMCE pre-score?)? (RQ3)

Note that there are a number of studies that have already investigated RQ1 and RQ2. We repeat them for the current sample to provide more context to the results of RQ3.

Methods

Data and measures

We collected data from a private, highly selective research-intensive university in the western United States during the years 2018 and 2020. This project was approved by the Stanford University IRB, protocol number 48006. In both years, the FMCE and a background survey were administered for attendance credit at the first recitation section. The background survey asked about students’ prior physics and math coursework, and in 2020, also probed several social-psychological constructs based on previously validated surveys [3]. Table 1 summarizes these factors and some previous works examining their effects on STEM performance. The same survey was administered for attendance credit in the final recitation section of the term as well. A binary measure of students’ gender was obtained from the office of institutional research for both years. We acknowledged that binary measure of gender has limitations, but that is the only format used in institutional data sets. At the beginning of the course (the time of the FMCE), students gave written consent (checked a box) to allow their data to be used for research purposes. We only included data for students for whom we had consent. All participants were over the age of 18. This study was determined exempt from review by the Stanford University IRB, protocol number 48006.

Table 1

List of social-psychological constructs probed by the survey and relevant references.

Social Psychological Factor	Reference
Test Anxiety	Measures feelings of anxiety and unease during exams [19]
Growth Mindset	Measures tendency to view intelligence as fixed [20]
Sense of Belonging	Measures social integration in the science classroom [9]
Self-Efficacy	Measures how well students think they are able to do science [3]
Science Identity	Measures how important science is to person’s self-esteem [21]
Motivation to Learn	Measures whether motivation to learn is performance-motivated or intrinsic [3]
Interest in Science (Physics)	Measures whether students find subject material interesting [19]
Ethnicity Stereotype Threat	Measures whether they think their peers perceive them differently due to their race [21]

Analysis

We used a combination of multivariable linear regression and structural equation modeling (SEM) to investigate the relationships between gender, incoming preparation, social psychological factors and course performance as measured by the final exam score. We first used multivariable linear regression as a tool physics education researchers are more familiar with compared to SEM. Multivariable linear regression allows one to simultaneously investigate the effects of multiple variables on a particular outcome. We were interested in the effects of gender, social psychological factors, and incoming preparation on final exam performance. An example regression model looks like: Compared to multivariable regression, SEM is a more complex method and has the advantage of not only estimating the effects of multiple variables on an outcome, but also allowing to examine the structure of relationships between the independent variables. SEM tests how well a structure, as specified by a set of linear models and model covariances, can predict the observed covariance matrix of the entire data set using maximum-likelihood estimation. An example of a simple SEM model would be: In the above example models, while the multivariable regression method only allows one to test whether and to what extent gender predicts exam score, SEM methods allows one to also test different paths for the effect of gender on exam score. In the above SEM model, gender affects exam score through two paths: indirectly through incoming preparation and directly. The indirect effect is that gender predicts level of incoming preparation and the level of incoming preparation in turn predicts student exam score. In other words, incoming preparation mediates the effect of gender on exam score. The SEM model represented as two equations above can also be visually represented as shown in Fig 1. The size of mediation effect can be calculated by multiplying the coefficients of the mediation path, in this case γ1 and β2.

Fig 1

Visual representation of the structural model explained above.

In this work, we used SEM to examine the mediators of gender differences in exam performance. For more details on SEM analysis, the advantages and shortcomings as well as different tools to conduct such analysis, we refer the reader to Ballen and Salehi [22].

Results

RQ1. Are there gender disparities in physics 1 performance as measured by exam performance?

Table 2 summarizes the results of the regression analysis. First, we examined whether course performance and/or physics incoming preparation measured by FMCE pre-score varied across gender. The first model in Table 2 predicts students’ performance only based on gender. As found previously in Salehi et al. [8], Model 1 shows that there is a gender gap of 0.22 standard deviations in exam performance indicating that on average women gained 0.22 standard deviation lower exam score compared to men.

Table 2

Multivariable linear regression predicting the effects of gender and FMCE pre-scores on exam grades.

	Model 1	Model 2	Model 3
Gender (F = 1)	-0.22* (0.10)	0.093 (0.093)	0.095 (0.095)
FMCE Pre-score		0.52***(0.046)	0.47***(0.062)
FMCE x Gender			0.11 (0.095)
R-squared	0.0092	0.27	0.25
N	366	366	366

*** p < 0.001

** p < 0.01

* p < 0.05.

Then we tested how physics 1 performance is impacted by students’ gender after controlling for physics incoming preparation, as measured by FMCE pre-score. Based on previous research, we hypothesize that there would be a gender gap in exam grades, but that gap would be explained by differences in FMCE pre-scores. The second model in Table 2 uses both gender and physics prior preparation. The gender gaps captured by Model 1 becomes small and insignificant in Model 2 when controlling for FMCE pre-scores, suggesting that the gap in exam grades can be traced to differences in male and female students’ prior physics preparation as measured by FMCE pre-score. This confirms our hypothesis that there exists a gender gap in exam performance, but it can be explained by variation in physics incoming preparation, which aligns with previous studies. We found no interaction between gender and FMCE pre-score as shown in Model 3, suggesting that the relationship between preparation and exam performance is the same for male and female students.

RQ2. Are there gender disparities in physics 1 incoming preparation as measured by FMCE pre-test?

We then examined to what extent FMCE pre-scores depend on gender (see Table 3). We found a gender gap of 0.48 standard deviations on the FMCE, with female students being disadvantaged. This is consistent with previous investigations at this university [8].

Table 3

Multivariable linear regression predicting the effects of gender and high school physics coursework on exam scores.

	Model 1	Model 2	Model 3
Gender (F = 1)	-0.48*** (0.10)	-0.48*** (0.099)	-0.48*** (0.10)
No. Physics Courses		0.31*** (0.057)
AP Physics 1			0.24* (0.11)
AP Physics C			0.55*** (0.13)
R-squared	0.054	0.12	0.093
N	366	366	366

*** p < 0.001

** p < 0.01

* p < 0.05.

RQ3. What learning experiences, backgrounds characteristics, and contextual factors contribute to these gender disparities?

We also examined to what extent FMCE pre-scores depend on gender when controlling for the number and type of physics courses a student took in high school. We hypothesized that the gender gap in FMCE pre-scores is due to gender differences in either the number or the quality of physics courses a student took in high school (Table 3). The gender gap in FMCE pre-score remains unchanged when controlling for high school physics course. This indicates that gender differences in FMCE pre-scores are uncorrelated with the number of physics courses taken in high school for students at this university. This contrasts with reports that female students are less likely to take physics in high school [23]. The size of the gender gap also remains the same when controlling for whether a student had taken an AP mechanics course (AP Physics 1 is an algebra-based college-level physics course, AP Physics C is the calculus-based version of the same course). Overall, these findings, in contrary to some previous speculations, suggest that in this sample the gender variation in incoming preparation as measured by FMCE pre-score is not due to gender differences in the type or the number of high school physics courses. We expect that this is specific to this sample because of the highly-selective nature of the institution.

We next examined gender differences in various social-psychological factors at the beginning of the course (see Fig 2). Female students reported much higher test anxiety than male students as also shown in previous work [9,11] (d = 0.51 ± 0.10, p < 0.001) and a lower sense of belonging (d = 0.31, p < 0.01). However, female students reported higher interest in physics (d = 0.22 ± 0.11, p < 0.05) and a greater sense of physics identity (d = 0.23 ± 0.11, p < 0.05).

Fig 2

Gender gaps on social psychological constructs.

Positive values indicate the female students score higher on that construct than male students. Error bars are the standard error of the coefficient. * p < 0.05, ** p < 0.01, *** p < 0.001.

Using the SEM model, we tested how these social psychological factors as well as high school physics courses can mediate the effect of gender on FMCE pre-scores (see Fig 3 for the model structure). By all fit indices, this mediation model is a good fit for the data covariance matrix (acceptable ranges: CFI > 0.95, TLI > 0.90, RMSEA < 0.08, SRMR < 0.05). The model results indicate that there is a large direct effect of gender (0.425 standard deviations) on FMCE pre-score, but there is also an indirect effect of gender mediated by Test Anxiety and Sense of Belonging: female students have higher test anxiety (and lower belonging) and test anxiety in turn negatively affects FMCE pre-score (effect size = 0.057 standard deviations) while belonging positively affects FMCE pre-scores (effect size = 0.030). Furthermore, number of high school courses does not mediate the effect of Gender (effect size = 0.0070): while number of high school courses is correlated with FMCE scores, there is no gender difference in the number of high school courses. Contrarily, the positive benefits of female students’ higher sense of science identity and interest in physics are small (indirect effects 0.015 standard deviations and 0.0083 standard deviations, respectively).

Fig 3

SEM model of the effects of gender on FMCE pre-scores as mediated by number of high school physics courses, test anxiety, science identity, and interest in physics.

*** p <0.001, ** p < 0.01, * p <0.05, † p <0.10. Acceptable ranges: CFI > 0.95, TLI > 0.90, RMSEA < 0.08, SRMR < 0.05 and p > 0.05 for whole model. Dashed links are not significant at the p = 0.10 level. Red links indicate negative effects, while green links indicate positive effects.

We triangulated the findings of the SEM model with multivariable regression (Table 4). The results are the same. There is a gender gap of 0.48 standard deviations which remains unchanged when you control for number of high school physics courses taken. Both sense of belonging and test anxiety are strongly correlated with FMCE scores, though the correlation with sense of belonging disappears when you account for Gender and HS coursework.

Table 4

Multivariable linear regression predicting the effects of gender, high school physics coursework, and social psychological factors on FMCE scores.

	Model 1	Model 2	Model 3	Model 4
Gender (F = 1)	-0.48*** (0.10)	-0.49*** (0.099)		-0.43*** (0.10)
High School Courses		0.28*** (0.057)		0.24*** (0.057)
Test Anxiety			-0.16** (0.052)	-0.11* (0.05)
Science Identity			0.058 (0.059)	0.076 (0.058)
Interest			0.022 (0.052)	0.037 (0.050)
Sense of Belonging			0.15** (0.059)	0.094 (0.057)
R-squared	0.057	0.11	0.065	0.14
N	366	366	366	366

*** p < 0.001

** p < 0.01

* p < 0.05.

To look at the effect of all these variables and their interplay on exam performance, we extended the above mediation model by adding exam performance to the model. This model examines the potential effects of high school preparation, FMCE performance, Gender, and social-psychological factors on exam performance (Fig 4). All fit indices indicate that this mediation model is a good fit of the data covariance matrix.

Fig 4

Extension of the model in Fig 1.

We look at the effects of gender on final exam scores, as mediated by high school preparation, social-psychological factors, and performance on the FMCE. *** p < 0.001, ** p < 0.01, * p < 0.05, † p <0.10. Acceptable ranges: CFI > 0.95, TLI > 0.90, RMSEA < 0.08, SRMR < 0.05 and p > 0.05 for whole model. Dashed links are not significant at the p = 0.10 level. Red links indicate negative effects, while green links indicate positive effects.

Note that there is minimal change to all the coefficients present in Fig 2 that are also in Fig 3. We find that the FMCE pre-score is the sole mediator of the effects of Gender, social psychological factors, and high school preparation on exam performance (direct effect = 0.492 standard deviations, indirect effect = -0.009). In other words, test anxiety and belonging mediate the effect of gender on FMCE pre-scores, and FMCE pre-scores in turn predict exam performance. Therefore, a female and a male student with the same FMCE pre-score would obtain a similar exam score regardless of their physics courses and their social psychological orientation. Said more plainly, any effects of Gender, social-psychological factors, or high school physics courses on students’ final exam performance is captured by the FMCE. We again triangulated these results with multivariable regression (Table 5). These results confirm that essentially all variables in exam performance are mediated by the FMCE pre-score.

Table 5

Multivariable linear regression predicting the effects of gender, social psychological variables, and FMCE pre-scores on exam grades.

	Model 1	Model 2	Model 3	Model 5
Gender (F = 1)	-0.22* (0.10)	0.093 (0.093)	0.073 (0.096)	0.085 (0.10)
FMCE Pre-score		0.52***(0.046)	0.49*** (0.049)	0.49*** (0.050)
HS Courses			0.095. (0.055)	0.094. (0.056)
Test Anxiety				0.010 (0.049)
Science Identity				-0.034 (0.055)
Interest				0.015 (0.048)
Sense of Belonging				0.041 (0.055)
R-squared	0.0092	0.27	0.26	0.25
N	366	366	366	366

*** p < 0.001

** p < 0.01

* p < 0.05.

Previous work shows that test-anxiety mediates the gender differences in final exams. Here we see the similar effect of test anxiety on exam of FCME. Therefore, maybe it is plausible that lower performance of female students may be bound by the timed format of an assessment regardless of its weight. To test that, we look at the relationship between gender, FMCE pre-test, and performance on the first homework assignment, which is not a timed assessment like the final exam and FMCE (from 2018) using a regression analysis (Table 6). We find no gender gap on the first homework, while FMCE scores are correlated with homework 1 scores (indicating some dependence on high school preparation, β = 0.14). The interaction between gender and FMCE pre-score is not significant and does not improve the fit of the model (ANOVA, p = 0.23), therefore, Model 2 is the best-fitting model in Table 4. Taken together with Fig 3, it seems likely the gender difference in final exam and pre-FMCE is due to their timed format partly as performance on timed assessments triggers test anxiety. Note that we did not perform a corresponding SEM for Table 6 because social psychological data was not available in 2018.

Table 6

Regression model of homework 1 scores on FMCE pre-scores and gender.

	Model 1	Model 2	Model 3
Gender (F = 1)	0.044 (0.098)	0.11 (0.10)	0.12 (0.10)
FMCE pre-score		0.14** (0.052)	0.091 (0.065)
Gender x FMCE			0.12 (0.10)
R-squared	0.00	0.014	0.015
N	518	518	518

All coefficients are standardized to be in units of standard deviations.

We note that HW1 covers much of the same content as the FMCE–basic 1D kinematics–thus it is notable that there is no gender gap on the assignment. Indeed, the assignment is still strongly predictive of final exam score (r = 0.26, p < 0.0001). This is not quite as strong as the correlation with FMCE score, likely because the FMCE covers a wider range of content and is thus more predictive of overall course performance. In prior work [3] we have seen similar patterns of the lack of gender gaps in measures that include homework scores.

Discussion

There are several notable findings in this work. First, the size of the gender gap on the FMCE pre-test remains unchanged when controlling for prior physics coursework in high school. This is because, in this sample, there is no correlation between gender and a number or type of physics courses a student took in high school. Second, there is no gender gap on the first homework assignment, which should be similar to FMCE in its dependence on prior physics preparation, but different in format. Finally, we see that the gender gap in FMCE is partially explained by differences in female and male students’ test anxiety and sense of belonging, even though the correlation between these factors and FMCE scores is small overall.

This leaves us with two potential explanations about the origin of the gender gap on the FMCE pre-test. First, it is possible that, though female students take physics in high school at similar rates as male students in this sample, they had negative experiences that impact their learning. This seems plausible given the reported influence of masculinity in physics [24] and the chilly or sometimes hostile climate that female students report in physics classrooms [25]. It is curious, however, that the female students who do persist to take physics in college have higher interest in physics and a higher sense of science identity. A plausible explanation is that these social-psychological motivators may have been necessary for the female students to overcome their negative experiences in high school physics courses and continue pursuing physics courses in college.

The second explanation is that the format of the FMCE as a “timed” closed book evaluation of physics competency induces stereotype threat in female students. As Spencer et al. [26] have shown, if an evaluation is presented to measure a competency of different groups in the topic, regardless of the stakes of the evaluation, the stereotyped group would underperform due to stereotype threat. They have also shown that anxiety is a mediator of the stereotype threat, and that stereotype threat is associated with lower sense of belonging. We observed the same pattern here. Even though the FMCE is zero-stakes, it is still framed as evaluative–assessing how much physics the students already know. This is likely to trigger stereotype threat. General test anxiety and sense of belonging partially mediate the gender gap in FMCE and explain 25% of the gap. Indeed, on the untimed first homework assignment, we see no gender gap even though homework grades also depend on prior physics preparation. Women suffer from the negative effect of stereotype threat despite higher sense of science identity and interest in Physics. In fact, Steele and colleagues have shown that stereotype threat is more consequential for the members of stereotyped groups who are more invested in the stereotype fields.

We see that the gender gaps in exam performance are entirely mediated by FMCE pre-scores, suggesting that any effects of test anxiety on final exam performance are captured by the timed nature of the FMCE. This would suggest that the stakes of the assessment do not seem to matter as much as the fact that the assessment is timed. This factor appears to explain a portion of the gender gap, but a substantial amount of the gender gap is left unexplained by variables we collected. Because we find no gap in Homework 1 scores, we hypothesize that some other social-psychological measure might explain the remainder of this gap. There are many aspects of student thinking for which we do not have validated psychometric scales. For example, math anxiety is considered to be a separate construct from test anxiety and is reported to be higher in female students than male students. In the future, we will collect math anxiety measures to see if this explains part of the gap, as math is an important part of physics preparation and the FMCE contains many graphical interpretation questions.

There have been previous studies examining potential gender bias in physics conceptual assessments, including the FMCE, the instrument used here. In Henderson et al. [27], they conducted differential item functioning and found only one item which seemed to favor men over women. However, there are important methodological limitations to this finding. First, the DIF calculation assumes that the post-test concept inventory score is an accurate measure of students’ ability. As our previous research has shown, this is less a measure of ability and more a measure of educational privilege, so the fundamental assumptions underlying this comparison are suspect. Indeed, the difference between men and women at post-test could be because of gender bias in the classroom itself, rather than actual understanding of physics knowledge. Even if this assumption held true, it would imply that intrinsic gender bias in this instrument is small compared to other factors like academic performance and social psychological effects. To the extent that our study overlaps with this investigation, it agrees with the finding that there are gender performance gaps.

A notable limitation to our study is that it was conducted at a highly selective university with substantial resources. We believe this largely explains why we see no difference in the number of previous physics courses men and women take–the university is highly selective and so all students are likely to take advanced science courses to stand out in the admissions process. However, we do not believe this invalidates our findings. Typically, at this university, the average score on most measures is higher than at other institutions, and the standard deviation is smaller. Thus, the amount of variance in performance outcomes that can be linked to academic preparation or test anxiety is likely lower in our sample than at other institutions. Regardless, the gender disparities in physics 1 exam and FMCE score is still present even in our selective sample. We have good reason to expect that test anxiety may play a larger role at other institutions where students have not been able to be as successful on admissions metrics like the SAT/ACT.

Conclusion

This paper adds greater detail to the understanding of gender gaps in physics performance. While previous work had shown that gender gaps in exam performance and students’ high school physics preparation, the source of those gender gaps are yet to be fully understood. Though nationally female students take physics at different rates, this did not explain the gender gap in physics preparation at this university. Despite the lack of gender difference in physics course taking behavior in high school, the work presented here shows that women in our sample are still at a disadvantage (obtain lower scores) in both FMCE and course final exam. This gender disparity is not due to lower physics preparation in high school as measured by number and type of high school physics courses taken by students, but more the result of social psychological phenomenon such as stereotype threat. There is no difference between men and women in number and type of high school physics courses. However, societal biases women suffer from negative stereotype in physics leads to their lower performance on FMCE and exam as timed assessment of physics competency. This stereotype threat is partially mediated by general test-anxiety and to a small extent, sense of belonging.

We encourage other researchers to perform similar analyses with different student populations. We expect that some of the results (the lack of gender gap in physics course taking in high school, for example) are due to the highly selective nature of the population. Some of the other findings such as the gender gaps in sense of belonging and test anxiety are consistent with other reports, and so we believe are likely more general. It is unclear the extent to which test anxiety plays a role in assessment demographic gaps. We suspect this is somewhat sensitive to the nature of the and population as shown in previous works [28], and thus may not be replicated for other assessment environments.

We conclude with some thoughts on how to avoid triggering this stereotype threat in college physics classrooms. The negative effect of stereotype threat can be diminished by changing the format of assessment. As we have shown here that there is no gender disparity in the first homework. The other works have also shown that women only under-perform on exams and, if anything, they overperform in other types of assessment [9,11]. This suggests that instructors should critically examine how they are assessing students’ understanding of physics concepts. Timed, evaluative physics assessments which are framed to evaluate physics competency, even if low-stakes, are more likely to trigger stereotype threat in female students—in spite of (or because of) these students having higher science identity and interest in the topic. The effect of stereotype threat hampers their performance in these assessments, and even consequently can discourage them from pursuing physics-related STEM fields and have potentially long-lasting psychological effects on women as a marginalized group in STEM. This work adds to the ever-increasing body of education research works that call to transform our common assessment methods in STEM into more equitable formats that do not disadvantage marginalized and/or stereotyped groups.

16 Mar 2022

Submitted filename: referee_report.pdf

Click here for additional data file.

12 May 2022

See attached document.

Submitted filename: RefereeResponse_v2.docx

Click here for additional data file.

27 Jun 2022

Exploring the pre-instruction gender gap in physics

PONE-D-22-04337R1

Dear Dr. Burkholder,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Abhiroop Mukhopadhyay

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

You have addressed all queries of the reviewers. I have accepted the article but some very small typos remain-as pointed out by Reviewer 1. These should be addressed in the final version that is published.

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: No

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I thank the authors for responding to the comments and addressing most concerns. I believe this paper provides interesting insights on an important dimension of the gender gap in STEM. Minor comments and suggestions follow in the attached document.

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

**********

Submitted filename: referee_report2.pdf

Click here for additional data file.

Submitted filename: referee_report2.pdf

Click here for additional data file.

7 Jul 2022

PONE-D-22-04337R1

Exploring the pre-instruction gender gap in physics

Dear Dr. Burkholder:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Abhiroop Mukhopadhyay

Academic Editor

PLOS ONE