This study adopts a systematic review methodology to compile and analyze peer-reviewed journal articles that focus on MGS. A systematic review is a rigorous and structured method of synthesizing existing research to answer a clearly defined research question. This process involves identifying, selecting, and critically appraising relevant studies based on predetermined inclusion and exclusion criteria, and follows a transparent and replicable procedure (Higgins and Green, 2011).

The Web of Science database was selected as the sole data source for this review due to its comprehensive indexing of high-impact scholarly journals, particularly those included in the Social Sciences Citation Index (SSCI) and Emerging Sources Citation Index (ESCI). These indices are widely regarded for their academic credibility and coverage of rigorous, peer-reviewed publications. This approach was intended to enable a focused and reproducible search of peer-reviewed outlets. It should be acknowledged, however, that Scopus, ERIC, and PsycINFO index partially non-overlapping corpora (e.g., education-focused venues, psychology-specific journals, practitioner outlets, and conference proceedings) that may not be fully covered by WoS. The search was last conducted in October 2024. No other databases or sources were used.

All retrieved records were screened for eligibility by the author based on predefined inclusion and exclusion criteria. Screening was performed manually by reviewing titles and abstracts, followed by full-text assessment for potentially eligible studies. As this study is single-authored, no independent dual screening was performed. No automation tools were used in the screening process.

The literature search was conducted using the keywords “MGS” and “sex stereotype math.” The search was finalized in October 2024. A total of 343 articles were retrieved using the first keyword, and 408 articles were identified with the second. After removing 70 duplicates, the remaining records were screened based on predefined criteria.

Figure 1 presents the PRISMA flow diagram, which details the flow of information through the phases of the systematic review.

Only studies that included the terms “math and gender stereotypes,” “mathematics and gender stereotypes,” or “mathematical gender stereotypes” in their titles and/or abstracts were retained for further analysis. Following this initial screening, 274 articles were shortlisted for a more detailed eligibility assessment.

The following inclusion criteria were applied to refine the final list of studies:

Data were extracted independently by one reviewer from each report. The reviewer systematically collected information based on predefined criteria and a coding framework. Although this systematic review was not prospectively registered, the research was guided by the code and category framework developed by Baş and Özturan Sağırlı (2017). No automation tools were used in the data codding process. The code and category list used for analyzing each article is presented in Figure 2.

The framework presented in Figure 2 encompasses eight categories: field, subject matter, methodology, sample, data collection tools, data analysis methods, year, and country. Meta-analysis was not conducted due to the data’s unsuitability for quantitative synthesis.

For each synthesis category (e.g., discipline, methodology, sample, data collection tools, country, year), all included studies were reviewed and coded according to a predefined coding framework. The studies were tabulated and categorized under relevant headings. No studies were excluded from individual syntheses unless they lacked information specific to the category being analyzed.

However, in some studies, MGS were not directly measured using a specific instrument (e.g., questionnaire, scale, or test), but were introduced through experimental manipulation within the research design. In these cases, stereotypes were treated as an independent variable; however, no measurement tool or statistical analysis related to the stereotype variable was reported. Therefore, “ST manipulation” was noted as the data collection tool, while the analysis section was left blank for these studies.

In addition to these structured extractions, we conducted an integrated qualitative synthesis covering stereotype-focused definitions, stereotype-related measurement instruments and their reported reliability, and the substantive findings of each study. All three components were analyzed with the same descriptive-interpretive thematic coding approach. In the first coding cycle, texts were read closely and explicit definitions and conceptual framings of stereotypes, the instruments/scales employed together with their psychometric reports (e.g., internal consistency, structural validity, test-retest), and each study’s principal findings were open-coded. In the second cycle, initial codes were clustered into hierarchical schemes reflecting definitional axes and subthemes, measurement types/instrument families and reported reliability indicators, and direction of findings (female-disadvantaging/female-advantaging/null/mixed) alongside the domain of effect (e.g., performance, attitudes/anxiety, selection/intention, instructional context).

Although no formal risk-of-bias assessment was conducted due to the qualitative content-analysis design, reliability was supported through several procedures: a pilot calibration on a small, randomly selected subset prior to full coding (to clarify inclusion/exclusion criteria, and stabilize thematic categories); independent double coding of approximately 30% of the corpus (n = 45) by a second researcher with expertise in education, using pre-specified codes and categories applied to randomly selected studies from the 152 articles; resolution of discrepancies via discussion and consensus with subsequent revisions to the coding scheme as needed; and computation of inter-coder agreement using the Huberman coefficient (Miles and Huberman, 1994), which yielded 83.5%—a level generally considered acceptable in qualitative content analysis. Throughout, a detailed codebook documenting definitions, rules, and representative excerpts was maintained, all updates were logged. In addition, no formal assessment of risk of bias due to missing results (reporting bias) was conducted, as this study employed qualitative content analysis and included all available data from the selected studies.

The coding process was conducted across 10 categories (Supplementary Appendix 4). The main categories were: field, subject matter, methodology, sample, data collection tools, data analysis, country, year, definitions, and conclusions.

Figure 3 shows the distribution details of the examined articles across the fields in which they were conducted.

Figure 3 shows that the majority of articles on MGS were published in the field of psychology. Psychology is followed by education/educational sciences and women’s studies. Although there are studies on MGS in various disciplines such as communication or science and technology, the number of published studies appears to be limited.

Figure 4 shows the findings of the distribution of the subject matters covered in articles on MGS.

As shown in Figure 4, the most frequently addressed subject in MGS-themed articles is MGS themselves. This is followed by topics such as mathematical ST, counter-stereotypical information regarding mathematical ability, the masculinity of mathematics, and gender equity in mathematics education.

Figure 5 presents the findings on the distribution of research methods/designs employed in the reviewed articles.

As shown in Figure 5, studies on MGS were predominantly designed as quantitative research. Among quantitative studies, the experimental design was the most commonly used, followed by survey studies and scale development. In contrast, qualitative and mixed-method designs were the least frequently employed research methodologies.

Figure 6 illustrates findings of the distribution of the samples studied in the reviewed articles.

Figure 6 illustrates that the majority of data in MGS-themed articles were collected from undergraduate students. Additionally, some studies focused on high school students, middle school students, primary school students, and documents. However, fewer studies were conducted with adults, graduate students, early childhood students, teachers, and parents.

Figure 7 illustrates the distribution of data collection tools used in MGS-themed articles.

As shown in Figure 7, the most commonly used data collection tool in MGS-themed articles was surveys/questionnaires. ST manipulations, categorized under “other,” were also widely utilized. Consistent with the findings on research methods/design, only a few studies employed interviews, observations, and documents.

Table 1 shows the distribution of instrument families and sub-types across the included studies.

Questionnaire-based approaches clearly dominate: non-psychometric survey items and validated psychometric scales together constitute the largest share. Within the experimental family, stereotype-threat manipulations are the modal sub-type, while other experimental tasks are comparatively rare. Qualitative instruments are infrequently used and, when present, typically serve as supplements rather than primary measures. Taken together, Figure 7 and Table 1 indicate a literature anchored in survey and experimental paradigms with limited qualitative triangulation.

Across the 152 studies, we identified 42 distinct psychometric instruments. Reliability evidence was reported for 23/42 (54.8%)—most often Cronbach’s α—while 19/42 (45.2%) reported none; where provided, evidence was largely confined to internal-consistency coefficients, pointing to a shortfall in psychometric reporting transparency. Qualitative techniques appeared in 23/152 (≈15.1%) studies; 18/23 (≈78.3%) reported no study-specific trustworthiness indicators. The remaining 5/23 (≈21.7%) offered primarily procedural assurances aligned with Lincoln & Guba (e.g., audio/video recording and verbatim transcription; protocol standardization; triangulation across interviews, observations, and documents; inductive/thematic coding by multiple trained researchers), with a small subset quantifying inter-coder agreement (Fleiss’ κ = 0.425–0.461; Kendall’s W = 0.489; both p < 0.001), indicative of moderate agreement by common benchmarks.

Validity reporting was even sparser: 30/42 (71.4%) psychometric instruments provided no study-context validity evidence. Among the remaining 12/42 (28.6%), evidence centred on structural validity (CFA/EFA/PCA), with fewer instances of convergent/discriminant/criterion and adaptation/procedural evidence. Most reported CFA solutions showed good-excellent fit (typically CFI ≈0.98–0.99, TLI ≈0.96–0.99, SRMR ≈0.02–0.03, RMSEA ≈0.04–0.07), though a minority were marginal (e.g., CFI ≈0.93; RMSEA ≈0.08). EFA/PCA findings commonly supported single-factor structures with high loadings, but KMO/Bartlett statistics and/or follow-up CFAs were frequently omitted. Convergent/criterion evidence included parallels with implicit and explicit indicators, age-appropriate known-group differences, and prediction of mathematics self-concept; discriminant evidence indicated separability from ability-stereotype measures.

Among the 23 studies that employed qualitative measurement tools, eight provided information regarding validity. The reported evidence was primarily grounded in content- and process-oriented indicators, including triangulation across field applications (observations, interviews, and materials), unannounced/randomized visits to capture typical lessons, video/audio recording and verbatim transcription, protocol standardization (shared opening prompt), thematic/iterative coding conducted by multiple researchers, prolonged engagement, and ethical safeguards (e.g., minimizing coercion, use of pseudonyms), as well as phenomenological reduction and context-specific descriptions. Notably, no quantitative validity evidence (e.g., convergent/discriminant/criterion relations, factor analysis, and measurement invariance) was identified in these studies; the reported indicators were confined to content and process assurances. Importantly, the κ/W values reported for expert agreement should be regarded as evidence of reliability, rather than validity.

Figure 8 illustrates the distribution of data analysis methods used in MGS-themed articles.

Various data analysis methods have been used in articles themed on MGS. As seen in Figure 8, the most frequently used analysis method in quantitative studies is inferential/parametric analysis, which was employed in 191 studies. Among the inferential parametric analysis methods, regression analysis (38 studies), ANOVA (36 studies) and t-test (35 studies) stand out as the most commonly used methods. Additionally, correlation analysis (26 studies), and ANCOVA (13 studies) were also frequently preferred. The next most frequently used analysis method is statistical descriptive analysis, which was employed in 40 studies. Among the inferential non-parametric analysis methods, chi-square test (11 studies), Mann Whitney-U test (2 study), Kendall correlation coefficient (1 study), and binomial distribution test (1 study) were employed. For qualitative analysis methods, content analysis (9 studies), document analysis (9 studies), discourse analysis (1 study), and descriptive analysis (2 studies) were used.

Figure 9 displays the findings regarding the distribution of countries in the reviewed articles.

It is observed that a large proportion of articles themed on MGS have been published in Western countries. Countries such as the United States (51 studies) and Germany (18 studies) are prominent, while other Western countries like Spain (6 studies), the United Kingdom (8 studies), France (8 studies), and Italy (6 studies) also show a noticeable concentration. Other countries, on the other hand, host a relatively smaller number of studies. In the context of MGS research, contributions from Ethiopia (1 study), Uganda (1 study), India (1 study), Chile (2 studies), and Mexico (1 study) remain limited. This distribution suggests that the majority of studies are concentrated in the Western world, whereas regions such as the Global South contribute far less to the field. Accordingly, it may be argued that research on gender and mathematics in the Global South is still at an early, developmental stage.

Figure 10 illustrates the annual distribution of publications on MGS.

Figure 10 presents information on the distribution of the examined articles over the years. According to the data in the graph, there are noticeable fluctuations in the number of articles published on MGS. In 1999 the number of articles was limited to 2, while in 2002, it increased to 6. In 2012, however, there was a significant rise, reaching 10 articles. In 2013, 2014, and 2015, the number of articles remained around 6, while in 2016 and 2017, there was an increase, with eight articles published. In 2020, the number dropped further to 2. However, there was a resurgence in 2022, with seeing 15 articles. In 2023, the number reached 11, but in 2024, it dropped back to 4. These findings indicate that research in this area peaked particularly in 2012 and 2022, with fluctuations in other years.

Figure 11 illustrates the distribution of definitional axes and subthemes across the included studies.

Across the 158 definitional assignments, the literature conceptualizes stereotypes primarily along two axes: a belief/domain-ownership axis centered on male superiority, and a process-based axis centered on ST. The MGS (MGS) axis accounts for 89 (56.3%) of all instances; within MGS, Superiority of Males in Math is by far the most prevalent subtheme (n = 59; 66.3% of MGS). Additional MGS subthemes appear less frequently: Math Is for Males (n = 8; 9.0%), Math Is a Male/Gendered Domain (n = 7; 7.9%), Males Are Associated with Math (n = 5; 5.6%), Math Is a Masculine Domain (n = 5; 5.6%), Affective Dimension (n = 3; 3.4%), and Achievement Dimension (n = 2; 2.3%). The second major axis, ST (ST), comprises n = 64 (40.5%) of assignments. Here, Risk of Confirming the Stereotypes is most common (n = 35; 54.7% of ST), followed by ST—Performance Decrement (n = 21; 32.8%) and Public Stereotype Salience (n = 8; 12.5%). Other definitional axes collectively form a long tail (n = 5; 3.2%) including Stereotype Internalization (n = 2), Mathematical Stereotype Endorsement (n = 1), Counter-Stereotypic Role Model/Exemplar (n = 1), and Stereotype Lift (n = 1).

Consistent with the quantitative profile, the qualitative synthesis indicates a concentration of stereotypes along the belief/domain-ownership axis—particularly the emphasis on male superiority. Across many articles, the stereotype is framed as a direct belief in gender-based ability superiority. In our qualitative document analysis, multiple texts articulated this stance; one source states: “One of the most obvious forms of stereotyping relates to explicit beliefs alleging a male or female ability-superiority in domains such as mathematics and language arts” (Supplementary Appendix 1, Study 12, p. 2). In doing so, such texts position mathematical success and ability as more properly male, permeating both individual attitudes and contextual expectations. The most concrete instantiation of this theme reduces to the explicit claim that “boys are better at math” (Supplementary Appendix 1, Study 86, p. 597). This formulation recurred verbatim; for example, one text states: “gender stereotypes often manifest as the belief that boys are better at math than girls” (Supplementary Appendix 1, Study 142, p. 1). Together, these formulations show that superiority is articulated not only through implicit associations but also through explicit declarations, with implications for self-efficacy, sense of belonging, and expectancy structures.

In parallel, some studies do not state male superiority explicitly yet underscore the associative linkage between mathematics and masculinity; for example: “… stereotypical beliefs that associate math and gender (i.e., math-gender stereotypes, where math = male)” (Supplementary Appendix 1, Study 4, p. 638). Such associations function as an implicit filter for the questions “for whom” and “to whom it is suited,” thereby aligning domain belongingness with gender and reinforcing the psychosocial mechanisms noted above. Several texts go further, framing mathematics as male-owned rather than merely male-associated, marking a shift from belongingness to normative exclusion. A representative formulation states: “Mathematics-gender stereotype is the false idea that mathematics is for men, not for women” (Supplementary Appendix 1, Study 18, p. 123). These formulations legitimize institutional expectations that construct mathematics as a naturally male domain and, by implication, position girls as guests.

Additionally, several texts frame mathematical ability in essentialist, masculine terms—for example: “… mathematical ability’ as natural, individual and masculine…” (Supplementary Appendix 1, Study 88, p. 204), which portrays ability not as a developable skill but as an innate attribute aligned with masculinity. The corpus also frequently labels mathematics as a gendered domain; for example: “… stereotypically male domains such as mathematics” (Supplementary Appendix 1, Study 110, p. 233). Such usage constructs the field’s cultural image in a male-centered manner. Finally, some definitions extend beyond domain ownership/superiority to invoke affective (enjoyment/interest) and achievement (expectancies of success) dimensions. For instance: “It is a common stereotype that boys/men are more likely to enjoy and succeed in mathematics while girls/women are more likely to enjoy and succeed at language arts subjects that require more reading and writing skills” (Supplementary Appendix 1, Study 127, p. 173). These formulations reproduce the view that boys/men enjoy and excel in mathematics, whereas girls/women enjoy and excel in language-heavy subjects, thereby reinforcing gendered expectations in valuation, self-efficacy, and performance beliefs.

Mirroring the quantitative distribution, the qualitative synthesis shows a marked concentration along the process-based axis around the conceptualization of ST. First, across many texts, ST is framed as the risk of confirming a negative in-group stereotype. One source states this explicitly: “ST is the risk of confirming, as self-characteristic, a negative stereotype about one’s group” (Supplementary Appendix 1, Study 1, p. 62). Second—analogous to the superiority discourse—the most concrete outcome-level manifestation of ST is performance decrement; for example: “The threat of being negatively stereotyped in mathematics can impair the performance of women on difficult math tests, a phenomenon referred to as ST” (Supplementary Appendix 1, Study 151, p. 13). Third, some definitions—without invoking male superiority per se—emphasize public/social salience of the stereotype as the trigger of threat: “ST is the sense of threat that can arise when one knows that he or she can possibly be judged or treated negatively on the basis of a negative stereotype about one’s group” (Supplementary Appendix 1, Study 112, p. 437). Such framings suggest that expectations about who is doing the judging and by what criteria operate as an implicit filter, shaping perceptions of evaluative contexts and potentiating the threat experience.

Notwithstanding their small share (3.2%), these long-tail axes introduce distinct focal points that enrich the conceptual landscape and offer fine-grained process insights. In several reviewed studies, the emphasis shifts from situational activation to individuals’ agreement with or endorsement of the stereotype: “Specifically, mathematics-gender stereotype endorsement (MGS endorsement) regards the degree of agreement with or endorsement of this stereotype” (Supplementary Appendix 1, Study 18, p. 123). Other texts conceptualize the stereotype as a multi-stage process of internalization progressing from awareness to self-ascription: “Stereotype internalization is usually defined as the incorporation of negative societal views in the self-concept: People first become aware of societal stereotypes (e.g., their group reputation); then some of them tend to endorse these stereotypes (i.e., they believe the stereotype is true about their group); finally they come to internalize the stereotype believing that the stereotype is true about themselves” (Supplementary Appendix 1, Study 74, p. 858). This definition delineates a pathway awareness → endorsement → self-attribution, indicating durable effects on self-perception and suggesting interaction with—indeed, potential amplification of—stereotype-threat processes.

A further set of texts emphasizes the regulatory effect of information and figures that reverse the stereotypic association: “By definition, a counter-stereotype plays on stereotypes by reversing them. For Pedulla (2014), the central idea is that counter-stereotypical information provides positive associations between a perceiver and the negatively stereotyped individual or group” (Supplementary Appendix 1, Study 36, p. 4). Analytically, this framing posits positive associative bridges between the perceiver and the negatively stereotyped target, thereby weakening endorsement/internalization pathways and, contextually, attenuating the threat experience. Finally, some texts highlight a mechanism that operates asymmetrically within the same ecosystem: “stereotype lift,” which can be defined as “a performance boost that occurs when downward comparisons are made with a denigrated outgroup” (Supplementary Appendix 1, Study 1, p. 62). This indicates that downward social comparisons vis-à-vis a devalued out-group can yield performance gains—a process distinct from, yet mirroring, ST, with implications for how evaluative contexts distribute cognitive and motivational resources across groups.

Taken together, the corpus conceptualizes MGS predominantly through belief/domain-ownership formulations centered on male superiority and process-based formulations centered on ST. Less frequent, yet conceptually informative, are definitional strands concerning endorsement, internalization, counter-stereotypic exemplars, and stereotype lift. This layered map clarifies where the field’s definitional center of gravity lies and highlights under-articulated mechanisms that future research could leverage for theory development and intervention design.

This section synthesizes the reported findings in the literature on MGS into six outcome categories: qualitative findings, descriptive findings, correlational findings, mediation findings, meta-analytic findings, and experimental findings.

Figure 12 displays the thematic map of qualitative results in the reviewed articles.

Qualitative findings converge around five themes: (i) School Ecology and Representation. One line of research documents how teacher discourse and material representation reproduce gendered expectations; for example, one study reports the following observation: “Everyone does the same exercises. The girls have a problem finding solutions by themselves…” (Supplementary Appendix 1, Study 63, p. 6). Another study examining imbalances in visual depictions arrives at this conclusion: “Altogether, of the 423 characters shown carrying out a professional activity, 146 are women, as compared to 277 men.” (Supplementary Appendix 1, Study 35, p. 1,491). (ii) Domain/Identity Stereotypes. A text documenting conventional framings states: “These discourses are oppositional and gendered; they inscribe mathematics as masculine…” (Supplementary Appendix 1, Study 88, p. 217). A study reporting more balanced cases notes: “It shows mostly balanced gender distribution in freetime and shopping categories…” (Supplementary Appendix 1, Study 38, p. 231). Another finding pointing to unconventional early-childhood patterns is conveyed as follows: “5-year-olds of both genders thought that girls liked math more than boys did” (Supplementary Appendix 1, Study 25, p. 1,273). (iii) ST Dynamics. Narratives differentiating by susceptibility record, for low susceptibility: “[In sixth grade] me and about four other people in our class were at a higher level than the rest of the class…” (Supplementary Appendix 1, Study 148, p. 614); and for high susceptibility, the same study offers: “Females who are good in math are smart… But males are probably gonna end up using math in their future…” (Supplementary Appendix 1, Study 148, p. 616). (iv) Measurement and Instrumentation. An example in which reliability is explicitly documented reports: “… Kendall’s W = 0.489, p < 0.001” (Supplementary Appendix 1, Study 17, p. 1). (v) Affect and Self-Perception. Among studies describing mathematically aversive profiles, one states: “These women do not like math…” (Supplementary Appendix 1, Study 105, p. 139); for successfully encouraged profiles, the same corpus reports: “These women appeared to have very positive attitudes toward math…” (Supplementary Appendix 1, Study 105, p. 136); stereotypically discouraged profiles are characterized as follows: “Two things are most noticeable about this group…” (Supplementary Appendix 1, Study 105, p. 139). Taken together, these findings indicate that mathematical gender stereotyping operates in a multi-layered manner across classroom ecology, domain-identity constructions, affective and self-belief processes, threat dynamics, and the quality of measurement.

In the descriptive strand of the literature, Figure 13 displays two higher-order themes: (i) Domain/Identity Stereotypes and (ii) ST. Codes cluster predominantly under the former. “MGS” (n = 19) is the dominant category; the masculinized framing of mathematics emerges early in development. “Egalitarian beliefs” (n = 3) point to more balanced representations. “Math-gender misconceptions” (n = 1) capture educator-specific misunderstandings. “In-group bias” (n = 2) and “unconventional” patterns (n = 1) are rare. Across descriptive studies, ST is reported only infrequently.

Within the correlational strand of the literature, the code co-occurrence network (Table 2) indicates that gender-stereotype endorsement (23 co-occurrences) and MGS beliefs (11) occupy central positions. The most frequent pairings are performance × stereotype endorsement (n = 8), performance × mathematical stereotype beliefs (n = 3), and academic intention × mathematical stereotype beliefs (n = 2). Math attitudes, self-ascribed ability, self-perception, participation, and math anxiety also co-occur primarily with stereotype endorsement. These analyses, however, do not establish causality; the co-occurrences reflect joint reporting within the same textual segments. The subsequent sections on moderation/mediation and experimental evidence elaborate the directionality of—and mechanisms underlying—these associations.

Within the mediation strand of the literature, as summarized in Table 3, evidence from 11 SEM and 19 experimental mediation tests indicates that the dominant pathway links stereotypes to performance and intentions via self-beliefs and affect (e.g., self-concept, self-efficacy, anxiety). In two studies, parental math-gender stereotypes were associated with girls’ non-STEM orientations through intrusive support as a social-transmission mechanism. Consistent with this account, most experimental findings show that stereotype threat elevates anxiety, activates performance-avoidance goals, and elicits dejection, thereby undermining performance. By contrast, conditions incorporating self-affirmation/positive achievement identity, counter-stereotypic role models, or self-monitoring attenuate the threat effect and are associated with improved performance. A subset of studies reported null or mixed mediation effects.

Within the experimental strand of the literature, as summarized in Table 4, the modal pattern is that stereotype threat reduces women’s mathematics-related outcomes, whereas effects for men are typically null. The effect intensifies when the threat is made salient and in mixed-gender settings; conversely, conditions involving self-affirmation/positive achievement identity, counter-stereotypic role models, self-monitoring, and selected mindfulness practices attenuate—indeed, in some cases, reverse—the threat effect.

Two comprehensive meta-analyses indicate that gender-ST is associated with small yet reliable performance decrements for girls/women. A child-adolescent meta-analysis (47 effects) estimated an average effect of d ≈-0.22, with no significant moderators and signals of publication bias. A broader synthesis (86 studies; 224 effects) found a small-to-moderate decrement only under threats targeting women (d ≈0.29), evident for mathematics but not for spatial tasks; heterogeneity was partly explained by task type, experimenter gender, and control condition. No consistent mean effects emerged for stereotype lift or for threats targeting men. Overall, the effects are context-sensitive and small in magnitude (| d| ≈0.20–0.30).

The research findings indicate that the majority of studies are situated within the field of psychology. Additionally, a significant number of articles focused on MGS have been conducted in the fields of education and women’s studies. Given the concept of MGS, the emergence of this finding is quite natural.

Mathematical knowledge is often perceived as entirely rational (Tang et al., 2010); however, McDonald (1989) suggests that individuals have emotional responses to mathematics and that every thought has an emotional component. From this perspective, it can be argued that there is a relationship between emotions and information processing. It is well-established that individuals, particularly when faced with challenging learning experiences, experience emotions such as anxiety, which in turn shape their mathematical learning (Ashkenazi and Danan, 2017; Skagerlund et al., 2019). Additionally, attitudes and beliefs are significant factors that lead individuals to respond to mathematics in various ways. In this context, the impact of gender-based beliefs on mathematics is an important and noteworthy issue. Therefore, it is expected that studies related to MGS are frequently explored within the field of psychology.

Mandler (Adams and McLeod, 1989) posits that experiences lacking value or meaning do not elicit emotional responses. Accordingly, emotional reactions to mathematics may be understood as reflecting the cultural values to which individuals are exposed. For example, White students have frequently been observed to outperform their peers in mathematics (Brown-Jeffy, 2009), and in many countries boys score higher than girls (Ayalon and Livneh, 2013; Gutfleisch and Kogan, 2024). Indeed, the gender gap in mathematics achievement appears even more pronounced among Turkish-origin girls in the fourth grade, who occupy a “double-minority” position as the other within the other (Guiso et al., 2008). The relatively higher achievement of individuals from particular gendered and ethnic subcultures can be linked to the elevated value ascribed to science—particularly mathematics—since the Industrial Revolution, coupled with the pervasive belief that mathematics may not be suitable for everyone (McDonald, 1989). Within this frame, the causes and consequences of MGS naturally fall within the purview of researchers focused on gender equality.

Research indicates that male students outperform female students in cognitive domains that bear on academic achievement, such as problem solving and mathematical reasoning (Altunçekiç et al., 2005; Gallagher et al., 2000; Geary et al., 2000). In affective domains—mathematics anxiety, beliefs, self-confidence, self-efficacy, and attitudes toward mathematics—male students likewise tend to show more favorable outcomes than their female counterparts (Çakiroglu and Isiksal, 2009; Frenzel et al., 2007; Kargar et al., 2010; Keller, 2001; Köğce et al., 2009). Several studies directly examining academic performance in mathematics also report significant differences favoring male students (Tate, 1997; Van de Gaer et al., 2008). Nonetheless, perspectives positing a decisive biological basis for gender differences in mathematical ability (Auyeung et al., 2006) are challenged by evidence showing that studies of biological effects yield contradictory and insufficient results (Ceci et al., 2009). Caplan and Caplan (2005) argue that gender differences in mathematical ability have never been conclusively demonstrated and, when observed, are more plausibly attributable to factors linked to individual experiences. If biological differences do not necessarily exclude women from mathematics and mathematics-adjacent fields, then it is reasonable that researchers have shifted attention to classroom contexts to ask which experiences lead young women to disengage from mathematics (Keller, 2007). Accordingly, factors that may impede learning—such as MGS—have become a central focus for educators within a broad ecological framework spanning classroom practices, teacher attitudes, peer relations, and instructional materials. It follows that scholarship on mathematical gender roles has naturally moved from reductionist accounts emphasizing biological explanations of gender gaps in achievement toward research that centers educational practices.

As a result, it is important to explore MGS not only within the fields of psychology, education, and women’s studies but also across other related disciplines such as philosophy, sociology, communication, and science and technology. Moreover, such interdisciplinary research is expected to provide deeper insights from the perspective of gender equality.

The analysis reveals that “MGS” is the most frequently addressed topic in the reviewed literature, followed closely by “mathematical ST.” Other topics, such as “counter-stereotypical information about mathematical ability,” “gender equity in mathematics education,” and the “masculinity of mathematics,” are explored significantly less.

The prominence of MGS as the most frequently addressed topic in the reviewed literature is an expected outcome, given the scope of this systematic review. Since this study examines research on gender stereotypes in mathematics, the centrality of MGS aligns naturally with the thematic boundaries of the selected literature. Furthermore, this prevalence can be attributed to the foundational role the concept plays. It serves as a starting point for understanding and investigating related phenomena such as ST, counter-stereotype interventions, and gender inequalities in mathematics education.

Research examining how MGS shape individuals’ perceptions of mathematics (Martinot and Désert, 2007; Passolunghi et al., 2014; Tiedemann, 2002), career choice behaviors (Chaffee and Plante, 2022; Liu, 2018), and mathematical achievement (Cvencek et al., 2015; Smetackova, 2015; Song et al., 2016) has long been an important focus. Due to the widespread and profound impact of these stereotypes, it can be argued that they have become a central element in understanding gender inequalities in mathematics and fields requiring advanced mathematical skills. As a result, MGS have become a fundamental topic for both theoretical research and practical interventions.

In the reviewed studies it has been observed that the most frequently investigated topic after MGS is ST. ST can be defined as a psychological situation in which individuals are at risk of confirming a negative stereotype expectation based on their gender (Hyde et al., 2008). Comprehensive research on this threat not only aims to understand the existence of MGS but also explores the negative effects of these threats on the mathematical performance of stigmatized social groups (Hyde et al., 2008) and the psychological mechanisms involved in this process (Bertrams et al., 2022; Casad et al., 2017; Pérez-Garín et al., 2017). Specifically, this threat has been shown to have a significant negative impact on the academic performance of women, who are a group often questioned about their mathematical competence (Bedyńska et al., 2018; Doyle and Voyer, 2016). In this context, ST can be considered a significant reason for the challenges faced by girls and women in science, technology, engineering, and mathematics (STEM) fields. Therefore, research on ST clearly demonstrates that this issue is a critical topic and has been frequently addressed in the literature.

However, the intense emphasis on MGS and ST carries the risk of overshadowing other critical dimensions of the relationship between mathematics and gender. For instance, counter-stereotypical interventions that challenge traditional gender norms can provide valuable insights into mitigating the negative effects of these stereotypes. It is well-established that the presentation of female role models associated with mathematics and science reduces the harmful impacts of MGS (Drury et al., 2011). In a study conducted by Good et al. (2010), the effects of stereotypical (e.g., male scientists) and counter-stereotypical (e.g., female scientists) textbook visuals on high school students’ understanding of a science lesson were examined. The study concluded that female students demonstrated better comprehension when exposed to counter-stereotypical visuals.

Similarly, exploring the phenomenon of masculinity associated with mathematics offers a profound perspective on how male identities are constructed and how this influences students’ engagement in mathematics lessons. Studies reveal strong evidence that mathematics teachers pay more attention to male, students than female students and assign greater responsibilities to male students during the learning process. Furthermore, high-level cognitive questions are systematically directed at male students significantly more often than at female students (Mittelberg et al., 2011; Nurlu-Üstün and Aksoy, 2022). Male teachers are also noted to provide more support and attention to male students during problem-solving activities, and male students predominantly participate in mathematical discussions (Lafrance, 1991). Other research suggests that the increased interaction between teachers and male students places women at a clear disadvantage in mathematics lessons compared to their male counterparts (Black and Radovic, 2018).

A broader research agenda that includes these less-explored dimensions can help us grasp the complexities of gender dynamics in mathematics in a more nuanced manner. Future studies that go beyond focusing solely on MGS and ST could more effectively address the multiple and intersecting factors contributing to gender imbalances in mathematics and other related disciplines. Such an approach would foster a more comprehensive and holistic understanding of how gender operates in educational contexts and the broader professional world.

Studies on MGS predominantly employ quantitative methods, with mixed and qualitative approaches being less commonly utilized. The prominence of quantitative research can be attributed to its inherent strengths. For instance, quantitative findings are often generalizable to entire populations or subpopulations due to their reliance on large, randomly selected samples (Carr, 1994). Additionally, the processes of data collection and analysis in quantitative research are typically time- and cost-efficient, often leveraging online surveys, forms, or statistical software.

Beyond these advantages, the dominance of quantitative research paradigms in this field can also be explained by the historical evolution and characteristics of the disciplines in which these studies are conducted. Most research on MGS has been carried out in psychology and educational sciences—fields that have historically favored quantitative approaches. This is consistent with findings in psychology (Gezici-Yalçın and Coskan, 2021) and educational sciences (Göktaş et al., 2012; Gül and Sözbilir, 2015; Nurlu-Üstün, 2023; Sağırlı-Özturan and Baş, 2020), where quantitative methodologies are predominant.

In psychology, this tendency is rooted in the discipline’s emergence as an independent science distinct from philosophy and medicine. From its inception, psychology adopted the deductive research methods of the natural sciences (Mayring, 2002). Early on, a dominant belief held that an objective reality existed independently of human perception or interpretation (Tebes, 2005). Consequently, experimental methods were advocated as fundamental to psychological research (Walsh et al., 2014). However, addressing complex societal issues such as MGS requires psychology to transcend these methodological limitations. There is a growing need for sophisticated approaches that integrate qualitative and mixed-methods research to provide a more nuanced understanding.

Similarly, educational research is also dominated by the quantitative paradigm. This prevalence is attributed to the scientific backgrounds of many education scholars, who often assume that research must produce statistical results, offer generalizable conclusions, and follow traditional methodologies (Ekiz, 2004). Yıldırım and Şimşek (2013) argue that some scholars trained in the positivist tradition reject qualitative methods that deviate from this framework as unscientific. Ekiz (2004) adds that these scholars have, either directly or indirectly, hindered the development and acceptance of qualitative methodologies in educational research. Nonetheless, studies exploring the impacts of MGS on educational settings, classroom practices, teacher behaviors, and student outcomes would benefit significantly from employing diverse qualitative research designs. Such approaches can offer a more comprehensive and holistic perspective, ultimately addressing critical gaps in the literature and advancing the field.

In conclusion, the predominance of quantitative methods in research on MGS stems from both the advantages of this methodology and the historical tendencies of disciplines such as psychology and educational sciences. However, relying solely on quantitative approaches may be insufficient to fully capture the complex social dynamics of this field. Therefore, future research should enhance methodological diversity to broaden the scope of the field and provide a more comprehensive perspective.

The research findings indicate that studies on MGS are most commonly conducted with undergraduate students, followed by middle and high school students. It has been observed that undergraduate students are the most common group sampled in studies on MGS. Similarly, Henrich et al. (2010) note that, particularly in the fields of psychology and cognitive science, samples are predominantly drawn from Western, Educated, Industrialized, Rich, and Democratic (WEIRD) populations, and more specifically from American undergraduate students. As Gül and Sözbilir (2015) have pointed out, conducting research with undergraduate students is often easier and more cost-effective.

The second most commonly studied groups are middle school students and high school students. Because the middle and high school years represent a critical period during which individuals’ career expectations begin to take shape and their vocational preferences are influenced (Correll, 2001). It is suggested that the initial steps of career planning are taken during this period (Göncü-Akbaş and Okutan, 2020). These years are regarded as a pivotal transition phase into either the workforce or higher education (Rowland, 2004). Consequently, the decisions made during this period play a significant role in shaping individuals’ vocational preferences in adulthood (Çakır, 2004). In this context, the increasing focus on middle and high school samples in studies examining the impact of MGS on career choices is noteworthy. The rise in research targeting these age groups can be considered a crucial step toward identifying stereotypes during this critical developmental period and developing intervention strategies. Such studies contribute significantly to the academic literature by facilitating the early detection and mitigation of negative stereotypes that could influence individuals’ career expectations.

However, it has been found that articles themed around MGS focus less frequently on elementary school students, preschool children, parents, teachers, and documents such as media elements and educational materials. One of the findings of this study is the limited number of studies conducted with samples from the preschool and elementary school years. The scarcity of studies may partly stem from the methodological difficulties of conducting survey-based research with children, including issues of comprehension and response validity. MGS weaken women’s and girls’ connection to mathematics, their sense of belonging (Good et al., 2010), and their willingness to engage in activities aimed at improving their mathematical abilities (Appel et al., 2011). This, in turn, negatively affects their mathematical achievement (Cvencek et al., 2015; Kiefer and Sekaquaptewa, 2007) and prevents them from pursuing careers that require advanced mathematical knowledge (Correll, 2001). Along with these negative effects of MGS, it is suggested that early ages are critical for the formation and reinforcement of such stereotypes (Cvencek et al., 2011; Del Río and Strasser, 2013; Herbert and Stipek, 2005). However, the limited number of studies conducted at the early childhood and elementary school levels indicates that gender stereotypes in these critical periods have not been sufficiently explored.

In addition, the limited focus on groups such as adults, teachers, and parents in studies on MGS may lead to a lack of comprehensive understanding of how these groups influence or are influenced by such stereotypes. It is well-established that both teachers and parents play a pivotal role in socializing children’s academic values and attitudes. A substantial body of research documents how parents’ and teachers’ expectations, gender stereotypes, and attributions affect children’s attitudes and performance in mathematics (Eccles et al., 1990; Ing, 2013; Tiedemann, 2000; Yee and Eccles, 1988). The limited research on groups such as adults, teachers, and parents regarding MGS results in a lack of understanding about how these groups influence or are influenced by such stereotypes. From this perspective, understanding how parents and teachers shape children’s perceptions of mathematics is crucial for developing interventions aimed at breaking these stereotypes.

In the literature on MGS, it has been observed that research based on documents is relatively scarce. However, studies that analyze documents such as textbooks and various media materials provide an important methodology for examining the historical and social context of MGS. For instance, textbooks, which are considered the primary source of knowledge (Çalışkan and Uymaz, 2022) and one of the most commonly used materials in classrooms (Kılıç and Seven, 2002), are distributed for free to students by the government in countries like Turkey. Therefore, these materials, which are accessible to every student, are shaped according to social and cultural norms and values (Wu et al., 2016). Zhang and Zhou (2008) emphasize that mathematics textbooks have a long-term and deep impact on students’ MGS, influencing their future mathematical learning processes. Studies have shown that while there is an attempt to maintain gender balance in mathematics textbooks, these materials still play a significant role in reproducing traditional gender stereotypes through elements such as occupational and family roles, as well as the gender of characters involved in mathematical activities (Guichot-Reina and De la Torre-Sierra, 2023; Moser and Hannover, 2014; Nurlu, 2021).

Similarly, media, ranging from movies to comic books and video games, has the power to convey stereotypical gender representations and mathematical content to new generations (Binark and Bek, 2009). Popular media tools aimed at children, including children’s books, television programs, films, websites, and video games, carry traditional gender stereotypes and contain gender-biased messages related to mathematics (Fellus et al., 2022; Hall and Suurtamm, 2020; Ladd, 2011; Wille et al., 2018). Similarly, posts on TikTok and X illustrate how the phrase “girl math” functions as a discursive practice. Within consumer and shopping contexts, the jargon frames women’s mathematical reasoning as “illogical” or “wrong.” Humorous examples circulate under this label, such as “Anything under five dollars feels like it’s pretty much free.” While presented as lighthearted or ironic, such usage reproduces the longstanding stereotype that girls and women lack mathematical competence. In this way, girl math normalizes gendered assumptions about cognitive ability and embeds them into everyday consumption practices and self-perceptions (Salma and Leiliyanti, 2024). Exposure to such media content has been shown to lead both male and female students to adopt these gender stereotypes (Hall and Suurtamm, 2020). In this context, studies on documents such as media materials and textbooks can provide valuable insights into how mathematical gender representations are shaped by societal norms and cultural values. However, the limited number of studies in this area suggests that the historical and cultural contexts have not been adequately addressed, highlighting the need for a broader perspective on MGS.

In conclusion, imbalances in research samples indicate the need for future studies to focus on a wider range of age groups and social roles. For instance, qualitative research on early childhood students and influential figures in their environment (such as parents and teachers), along with studies on documents such as educational materials and media elements, could provide a strong foundation for preventing and transforming MGS.

These findings reveal that data collection tools in research on MGS are largely based on surveys and questionnaires. This can be seen as a reflection of the commonly preferred quantitative research methods in the literature and the generalizability advantages these methods provide. Quantitative data collection methods are frequently preferred because they are believed to yield high-quality data. These methods encourage more honest and sincere responses by ensuring anonymity and typically achieve higher response rates compared to methods like interviews. Additionally, surveys and questionnaires offer the ability to collect a large amount of data in a short period of time and at a low cost (Marshall, 2005).

However, the limited use of qualitative data collection methods such as interviews, observations, and document analysis in the studies examined creates a gap in the more in-depth and holistic exploration of the topic. This is because children and adults may be reluctant or unable to directly express their views on sensitive topics, such as gender stereotypes (Muzzatti and Agnoli, 2007). In this context, the interview method, which allows access to understanding another person’s perspective and gaining insight into their thoughts and stories (Patton, 2014), may enable a more comprehensive understanding of how MGS are shaped and propagated. Similarly, another important qualitative data collection method, observation, allows the researcher to experience the phenomenon first hand through direct observation rather than relying on assumptions. The observation method enables the researcher to engage directly with the environment, establish personal contacts, and gain a holistic understanding of the context in which individuals interact (Patton, 2014). In this regard, observation can be considered a valuable method for studying MGS. Moreover, document analysis, unlike data collected at the individual level, reflects our collective behaviors and reveals the dynamics at the societal level (Lune and Berg, 2017). In this context, document analysis can provide valuable insights into how social and cultural norms reinforce MGS. However, the widespread adoption of the quantitative paradigm in current research may limit the in-depth and qualitative understanding of societal phenomena such as MGS, thereby creating a significant narrowing in this area. Therefore, it should be considered that qualitative and mixed-method approaches could provide a more comprehensive perspective in understanding such complex social phenomena.

The widespread use of data collection tools categorized under the “other” category, such as ST manipulations, indicates a preference for behavioral and experimental approaches in research. These types of manipulations provide the opportunity to directly observe the effects of gender stereotypes, offering in-depth insights into how individuals respond to these stereotypes. However, it should be noted that these tools rely on a cause-and-effect relationship within a limited context, often conducted in laboratory settings, which raises questions about their applicability to persons, environments, treatments, and outcomes not included in the experiment (Shadish et al., 2002). In other words, findings from laboratory settings may not always be applicable to real-world contexts.

In conclusion, the imbalance in data collection tools points to the need for a more holistic approach to addressing MGS. Relying solely on quantitative data in research may be insufficient to understand the impact of stereotypes on individuals’ lives. Therefore, the use of qualitative data collection methods will contribute to a deeper understanding of these stereotypes and the development of more effective intervention strategies. Future studies should aim to balance both qualitative and quantitative research methods, enabling the generation of more comprehensive and accurate findings.

Our findings indicate that methodological transparency remains limited in both quantitative psychometric reporting and qualitative studies. On the quantitative side, only 23 of the 42 instruments identified across 152 studies (54.8%) reported reliability evidence, and most of these were confined to Cronbach’s α; 19 of 42 instruments (45.2%) provided no reliability information whatsoever. Yet the Standards for Educational and Psychological Testing (American Educational Research Association et al., 2014) frame the documentation of validity, reliability/measurement error (precision), and fairness for each intended use as a professional obligation. Likewise, APA JARS-Quant expects researchers to report study-specific reliability coefficients (e.g., internal consistency, test-retest, interrater agreement) alongside relevant validity evidence and implementation details aimed at improving measurement quality (Appelbaum et al., 2018). Methodological work further notes that the assumptions underlying α (tau-equivalence, independence of errors) are often violated, such that α may under- or over-estimate reliability; reporting McDonald’s ω in addition to α is therefore recommended (McNeish, 2017). On the validity side, 30 of the 42 instruments (71.4%) offered no context-specific evidence; the remaining 12/42 (28.6%) relied predominantly on structural validity (EFA/CFA/PCA). Although most CFA solutions reported good-excellent fit (CFI ≈0.98–0.99; TLI ≈0.96–0.99; SRMR ≈0.02–0.03; RMSEA ≈0.04–0.07), critical steps such as KMO/Bartlett diagnostics, parallel analysis, and independent confirmation were frequently omitted; convergent/discriminant/criterion and adaptation/procedural evidence was sparse. This pattern aligns with broader reviews showing that validity is either unreported or disproportionately reliant on structural indicators (Flake et al., 2017; Ntumi and Twum Antwi-Agyakwa, 2022). In such circumstances, the validity of research findings may be largely ungrounded and uninterpretable.

A similar picture emerges on the qualitative side. Only 23 of 152 studies (≈15.1%) employed qualitative techniques; of these, 18 (≈78.3%) did not report study-specific trustworthiness indicators. The remaining five (≈21.7%) provided process-based assurances aligned with Guba & Lincoln (recording-verbatim transcription, protocol standardization, triangulation, thematic coding by multiple researchers), while a small subset quantified inter-coder agreement at moderate levels (Fleiss’ κ = 0.425–0.461; Kendall’s W = 0.489; both p < 0.001). This pattern is consistent with work showing that qualitative reporting is typically at a moderate-low level and that “how trustworthiness was established” is often insufficiently specified (Walsh et al., 2020; Watts and Finkenstaedt-Quinn, 2021). Under the naturalistic paradigm, qualitative quality should be evaluated via credibility, transferability, dependability, and confirmability (Guba and Lincoln, 1985); however, these criteria should not be merely named. Following SRQR/COREQ/JARS-Qual, researchers should report in detail the definition and evidence of saturation, the nature of iteration, researcher positioning (reflexivity), and the scope of triangulation (O’Brien et al., 2014). Moreover, inter-rater reliability (IRR) is regarded as paramount in content analysis (Neuendorf, 2010); indeed, in science education journals, only 19 of 103 studies in 2019 reported IRR (Cheung and Tai, 2021). Process-based assurances are therefore necessary but not sufficient; where appropriate, they should be complemented by quantitative indices such as κ/α/AC1–AC2/W/ICC and reported transparently.

In sum, our results reveal that reporting on measurement quality remains limited in scope and depth across both quantitative and qualitative research. For policy and practice, we recommend: (i) reporting ω alongside α, as well as test-retest and ICC; (ii) supplementing structural validity with convergent/discriminant/criterion and adaptation/procedural evidence; and (iii) in qualitative studies, providing systematic and transparent accounts of reflexivity, saturation, iterative decisions, triangulation, and IRR. Such practices will strengthen the credibility, transferability, and dependability of findings.

This finding provides significant insight into the data analysis methods employed in research on MGS. Firstly, it is observed that the most frequently used methods in quantitative analyses are inferential analyses. This suggests that, whether employing experimental or survey designs, researchers aim to make inferences about variables and demonstrate how sample results can be generalized to a broader population (Creswell and Creswell, 2018). Additionally, it is noted that most of these inferential analyses are parametric, including ANOVA, t-tests, correlation, and regression analyses. The limited use of non-parametric analysis methods indicates a preference for parametric methods that typically validate their assumptions in research (Field, 2013). Chin and Lee (2008) state that parametric analyses offer more robust and reliable results compared to non-parametric ones. Therefore, the widespread use of parametric tests in articles focusing on MGS can be considered advantageous. Descriptive analyses play a significant role in studies focusing on MGS. The frequent use of descriptive analyses indicates that researchers often employ this method to define the group under study and summarize its characteristics using tables, graphs, and statistical measures such as central tendency and variability. This approach provides information about sample and population values (Çakıcı-Eser, 2022). Consequently, the application of descriptive statistics in articles addressing MGS has more clearly revealed the prevalence and acceptance rates of the phenomenon of mathematical gender stereotyping.

The observation that qualitative analysis methods—such as content analysis, document analysis, and discourse analysis—are less frequently employed compared to quantitative analyses is noteworthy. This trend may be attributed to the complexity and time-consuming nature of qualitative data analysis. Patton (2014) highlights these challenges, emphasizing the difficulty in reducing the volume of raw data, distinguishing the trivial from the significant, identifying key patterns, and constructing a framework that effectively conveys the essence of the data. However, in areas influenced by emotional and social factors, such as MGS, qualitative analyses can provide in-depth insights. They are instrumental in understanding how such stereotypes are shaped within social and cultural contexts. Therefore, these findings suggest that future research should aim for a more balanced application of both quantitative and qualitative data analysis methods.

The majority of academic research on MGS is concentrated in Western countries. Notably, the United States, Germany, France, and the United Kingdom are at the forefront of publications in this field, whereas academic studies on this topic are significantly less frequent in non-Western countries such as Israel, Ethiopia, Uganda, Turkey, and India. This disparity can be attributed not only to the overall dominance of Western countries in academic publishing but also to their higher levels of gender equality, democracy, and human rights.

The academic pre-eminence of Western nations is reinforced by historical processes, economic investments, and scientific publishing systems. From a historical standpoint, the Scientific Revolution in Central and Western Europe during the 16th and 17th centuries laid the foundation for the systematic production and dissemination of academic knowledge. For instance, in the 18th century, Encyclopédie, ouDictionnaireraisonné des sciences, des arts et des métiers, compiled by Diderot and d’Alembert, played a pivotal role in the structured development of knowledge by integrating scientific information within an interdisciplinary framework. Porter (1990) further asserts that the Scientific Revolution and the Enlightenment were instrumental in the emergence of modern social sciences such as sociology, economics, psychology, and anthropology. Thus, the quantitative dominance of academic output in Western countries has deep historical roots.

Moreover, the substantial financial resources allocated by Western nations for academic research and R and D (Research and Development) further solidify their leadership in scientific production. For example, annual R and D expenditures in the United States exceed $789 billion in 2021 (National Center for Science and Engineering Statistics, 2024). Such significant economic investments render the Western world an attractive hub for researchers, thereby facilitating brain drain from developing nations to more developed regions. Indeed, two-thirds of highly skilled immigrants have settled in North America (Lucas, 2008).

The Western-centric structure of scientific publishing systems further perpetuates this academic dominance. Academic reputation and influence are largely determined by journal rankings, impact factors, and H-indices. Established in 1963 with financial backing from the United States, the Science Citation Index (SCI) encompasses citations from the most prestigious scientific journals. However, the vast majority of these journals are based in the United States and the United Kingdom, while nearly all others originate from Europe. Over time, citation indices such as journal rankings have become key indicators of “reputable” academic knowledge, reinforcing a Euro-American-centered academic publishing landscape. Digitalization and financial investments have further amplified the impact of these citation indices, elevating Western academic networks to an even more dominant position. Consequently, regional academic journals and those publishing in languages other than English face increasing marginalization if they are not incorporated into these citation indices, thereby entrenching existing academic hierarchies. As a result, long-standing regional knowledge ecosystems are weakened, and the legitimacy of journals excluded from these indices is continuously scrutinized (Mills, 2024).

Another critical factor contributing to the concentration of academic studies on MGS in the West is the high level of development in gender equality, democracy, and human rights within these nations. According to the 2024 Global Gender Gap Report by the World Economic Forum, Europe has closed 75% of the gender gap, establishing itself as a global leader in this domain, while North America follows closely with a closure rate of 74.8%. In contrast, non-Western countries such as Israel, Ethiopia, India, and Turkey rank lower due to their comparatively weaker gender equality scores (World Economic Forum, 2024). Similarly, the 2024 Democracy Index classifies Europe and North America as “full democracies” (Economist Intelligence, 2025), while the Freedom in the World 2025 report by Freedom House designates countries in these regions as “free” (Freedom House, 2025). The presence of strong democratic institutions, extensive civil liberties, and robust human rights protections in these nations fosters a conducive environment for research on gender equality, thereby reinforcing the predominance of scholarly literature on this subject within Western academia.

While the academic dominance of Western countries is rooted in historical, economic, and structural factors, expanding research on MGS beyond these regions would enhance the diversity and inclusivity of knowledge production. To achieve this, fostering academic research in non-Western countries, integrating regional journals into international citation indices, and supporting scholars publishing in languages other than English are essential steps. Furthermore, advancements in gender equality, democracy, and human rights within these countries could create a more conducive environment for such studies. A more balanced global distribution of scientific knowledge would not only enrich academic discourse but also contribute to broader societal awareness and policy development.

An examination of the annual distribution of research on MGS reveals fluctuations in the number of publications over time, with notable increases and decreases observed in specific periods. Since 1999, the number of articles has shown a fluctuating trend, with significant peaks in 2012 and 2022. In 2012, the number of publications reached its highest point with 10 articles, followed by a period of relative stability in 2017, when 8 articles were published. In 2022, the number of articles again reached 15, marking another peak. These fluctuations indicate that academic interest in the topic has concentrated in certain periods, influenced by various factors that may have shaped these trends.

The Fourth World Conference on Women, held in Beijing in 1995, significantly increased global awareness of gender equality and initiated a transformative process at the international level (United Nations, 2025). The Beijing Platform for Action, adopted at the end of the conference, facilitated the promotion of research on gender equality and pioneered the creation of international funding opportunities (UN Women, 2000). In this context, the emergence of academic studies on MGS from 1999 onwards can be seen as a consequence of the scientific and political environment shaped by the Beijing Platform for Action.

The increase in academic production in 2012 and 2022 can be attributed to global initiatives and policy changes aimed at advancing gender equality during these periods. Established in 2010, UN Women launched projects in 2011 to support the economic and academic empowerment of women, with various initiatives in the United States and the United Kingdom further complementing these efforts (UN Women, 2025). During the same period, the White House Council on Women and Girls in the U.S. allocated new funding to support women in STEM fields (White House Council on Women and Girls, 2025), while the National Science Foundation (NSF) expanded its ADVANCE Program to develop new policies aimed at empowering female academics (National Science Foundation, 2011). Similarly, in the UK, the Athena SWAN program was expanded to offer awards promoting gender equality in universities and research centers, providing support to female researchers (Athena SWAN, 2025).

The year 2022 stands out as a period during which academic funding agencies implemented stricter criteria for supporting gender equality. Under the European Union’s Horizon Europe Program, the Gender Equality Plan (GEP) became mandatory for research projects, and the inclusion of gender considerations in funding applications was established as an evaluation criterion (European Commission, 2022). Additionally, the “Women TechEU” program provided special funding for female leaders and entrepreneurs (European Innovation Council, 2025). UNESCO’s “STEM and Gender Advancement (SAGA)” initiative offered support to increase the participation of female academics in research, while UN Women and UNDP promoted gender equality awareness within academic institutions through the “Gender Equality Seal for Research Institutions” program (UNDP, 2025; UNESCO, 2025).

The fluctuations in academic interest in MGS at certain periods are directly linked to global policies, available funding, and academic trends. These dynamics help explain the periods of accelerated development in the field. However, the notable decline in the number of relevant publications in 2020, dropping to just two publications, necessitates an investigation into the underlying factors contributing to the fluctuations in scientific production.

Considering the impact of global crises on scientific output, the decline in these years becomes more understandable. For instance, the COVID-19 pandemic, which had a global impact in 2020, caused significant disruptions in academic research. The closure of universities, the transition to remote education, and the suspension of fieldwork greatly hindered scientific production. During this period, many scholars focused on the inequalities created by the pandemic in education (Czerniewicz et al., 2020; Frohn, 2021; Özer et al., 2020), and academic interest shifted toward educational technologies and remote teaching (Başaran et al., 2020). As a result, more specific areas, such as MGS, inevitably became a lower priority in academic agendas.

Moreover, the pandemic led to changes in the peer review and publication policies of many academic journals. As a result, a significant portion of the research published in 2020 focused on the COVID-19 pandemic (Raynaud et al., 2021; Riccaboni and Verginer, 2022). This shift contributed to the lower number of publications on MGS. Considering all these factors, the notable decline in publications in 2020 can be understood in the context of shifts in academic focus and the impact of global crises on scientific production.

The fluctuations in academic publications on MGS reflect broader global trends, shaped by political, social, and economic factors. While significant peaks in 2012 and 2022 highlight the importance of international initiatives and funding opportunities, the decline in 2020 suggest the vulnerability of research areas to shifts in academic focus and external crises, such as the COVID-19 pandemic. By recognizing and addressing the factors influencing these fluctuations, the academic community can better prioritize and continue to advance research on gender equality in mathematics.

Our findings indicate that the literature organizes the theme of stereotypes along two robust axes: (i) belief- and belonging-based formulations centered on male superiority, and (ii) process-oriented formulations structured around ST. The former casts mathematics as a form of identity-linked “ownership,” shaping choice and persistence through belonging and expectancy structures; the latter attenuates momentary performance via cognitive and affective mechanisms activated within evaluative contexts. Together, these axes align with evidence on achievement disparities (Cvencek et al., 2015; Kiefer and Sekaquaptewa, 2007) and on problems of retention and persistence in the field (Correll, 2001).

Definitions that explicitly assert male superiority or implicitly construe mathematics as a “male domain” (Supplementary Appendix 1, Study 89, p. 237) emerge as the most persistent trope in the literature. This discourse structures not only individual expectancy beliefs but also classroom interaction patterns (Nurlu-Üstün and Aksoy, 2022), representational practices in instructional materials (Guichot-Reina and De la Torre-Sierra, 2023; Nurlu, 2021), and the visibility of role models (Ladd, 2011; Nurlu-Üstün and Uzuner-Yurt, 2023). Normative frames such as “Mathematics is for men, not for women” (Supplementary Appendix 1, Study 18, p. 123) masculinize the field as “naturally” male, thereby weakening girls’ sense of belonging and relegating them to a guest status (Good et al., 2012), and—over the longer term—suppressing course selection and career intentions (Correll, 2001).

Defining ST as the “risk of confirming, as self-characteristic, a negative stereotype about one’s group” (Supplementary Appendix 1, Study 1, p. 62) affirms the centrality of processes that are sensitive to evaluative context. As public visibility and the expectation of being judged increase, threat intensifies, producing performance decrements that are especially pronounced on cognitively demanding tasks (Steele and Aronson, 1995). This pattern helps explain why threat-reducing statements, low-threat task designs, and context-sensitive instructions can be effective.

Although the “long-tail” axes may appear quantitatively small, they enrich the stereotyping ecosystem. Endorsement and internalization delineate a pathway from mere awareness to self-ascription (McKown and Weinstein, 2003), whereas counter-stereotypic role models forge associative bridges that disrupt this chain (Dasgupta, 2011). Meanwhile, “stereotype lift” reminds us that derogating an out-group in comparative contexts can artificially inflate performance, underscoring the importance of fair assessment designs (Walton and Cohen, 2003).

The findings indicate that unidimensional measures may inadequately capture the plural nature of stereotypes. Future work should proceed with multidimensional scales that disentangle belonging/superiority, threat, endorsement, and internalization, and with designs that combine implicit and explicit indicators. At the same time, the conceptual map suggests that interventions must operate on two fronts: (i) representational and role-model strategies that weaken belonging/superiority discourse; and (ii) assessment designs and instructional guidelines that reduce the activation of threat.

Mathematics is widely framed as masculine; this framing is sustained by the school ecology (teacher discourse, material representation) and early-life experiences, and is transmitted to achievement and intentions via cognitive/affective processes. In the correlational network, stereotype endorsement and mathematical gender-stereotype beliefs occupy central positions; SEM and experimental evidence link them to performance through self-beliefs (self-concept/self-efficacy) and affective pathways (anxiety, dejection). This pattern accords with expectancy-value accounts (e.g., self-concept/self-efficacy and domain value (Eccles and Wigfield, 2020) and with stereotype-threat theory (Steele, 1997): gendered contexts erode self-resources, activate avoidance goals, and depress performance.

Meta-analytic evidence indicates small but reliable effects (approximately | d| ≈0.20–0.30) that are sensitive to context. Effects tend to intensify when threat is salient and in mixed-gender settings. However, they can be attenuated—and in some cases even reversed—through self-affirmation/positive achievement identity, counter-stereotypic role models, self-monitoring, and selected mindfulness practices. Experimental mediation findings largely align with a “threat → anxiety/avoidance → performance ↓” pathway (Schmader et al., 2004). Nevertheless, some studies report null or complex mediation, suggesting that task type, measurement timing, and differences in operationalization are decisive factors (Pennington et al., 2016).

Linking parental stereotypes—enacted through behaviors such as intrusive support—to girls’ non-STEM trajectories indicates that stereotypes function as social practices transmitted within the family-school ecology, rather than merely as individual attitudes (Carlana, 2019; Crowley et al., 2001). Likewise, teacher discourse and the uneven representation in instructional materials generate symbolic cues that, as qualitative excerpts illustrate, entrench classroom norms (Blumberg, 2008; Tiedemann, 2002).

Taken together, these findings underscore both conceptual and methodological challenges, while simultaneously pointing toward actionable implications for practice and future research. Methodologically, divergent operationalization of “mathematics-gender stereotype” (identity labels, competence clichés, implicit associations) hinder cross-study comparability. Although some reliability is reported, stronger evidence is needed for measurement invariance across age and cultural groups and for adherence to reporting standards. Empirically, the findings support: (i) teacher professional development (language/comparative feedback), (ii) curriculum-materials review (balanced representation), (iii) pre-exam self-affirmation/identity-supportive micro-interventions, (iv) visibility of counter-stereotypic role models, and (v) adjustments to class composition and task framing—each requiring context-sensitive design and external-validity testing. Priorities for future work include adequately powered, multi-source experiments; time-segmented mediation and multilevel SEM; rigorous measurement invariance by age/sex/culture; standardized operationalization of threat salience, control conditions, and task types; and open-science practices. Longitudinal studies of early-childhood masculinized framings could identify ecological windows for intervention.

The evidence demonstrates that MGS operate in multilayered ways, with effects that are small yet consistent and context-sensitive, and that these effects can be mitigated through appropriate psychosocial interventions and ecological adjustments. Advancing definitional and measurement standardization, alongside strengthening causal research designs, appears critical for the next advancement of the field.

This systematic review has several limitations that should be acknowledged. First, the literature search was restricted to the Web of Science database. Although Web of Science indexes high-quality, peer-reviewed publications, the exclusion of other major databases (e.g., Scopus, ERIC, PsycINFO) may have led to the omission of relevant studies—particularly education-sector proceedings, regional journals, and psychology-specific outlets. This coverage decision could bias the corpus toward internationally indexed journals and may have affected the observed distribution of samples (e.g., early childhood/elementary), methods, and contexts. Future updates should implement multi-database searches (WoS, Scopus, ERIC, PsycINFO), and backward/forward citation chasing to strengthen comprehensiveness and reproducibility.

Second, the review included only studies published in English, potentially excluding valuable research conducted in other languages—particularly in non-Western contexts. This language restriction may have limited the cultural and geographical diversity of the findings.

Third, gray literature such as dissertations, conference proceedings, and institutional reports was not included in the review. As a result, emerging or unpublished research related to MGS may have been overlooked.

No protocol pre-registration (e.g., PROSPERO/OSF) was undertaken for this systematic review. The absence of pre-registration constitutes a limitation, as it may increase the flexibility of decision-making during the study and thereby heighten the risk of selection or reporting bias. To mitigate this risk, however, the inclusion/exclusion criteria and the analysis plan were specified in writing prior to the search, and adherence to the PRISMA 2020 guidelines was maintained (Supplementary Appendix 2).

Finally, although the coding procedure followed a systematic approach and inter-coder reliability was established, some degree of subjectivity in interpreting and categorizing studies is unavoidable. In addition, the review was conducted by a single researcher; the absence of independent dual screening and cross-validation constitutes an additional limitation that may increase the risk of selection and reporting bias. Taken together, these factors suggest that the synthesis may inadvertently reflect some bias. A list of PRISMA items not implemented and the justification for their omission is available in Supplementary Appendix 3.