The study investigates the causal link between gendered languages and gender disparities in mathematics performance. Many languages grammatically encode gender, often employing masculine generics to refer to mixed-gender groups. Research indicates a correlation between gendered languages and greater gender inequality across various domains. However, establishing causality is challenging due to the cultural embedding of language. Previous studies have shown correlations between gendered languages and gender inequality, and demonstrated the impact of gendered language on attitudes. However, they lacked evidence of causal effects on performance. This study addresses this gap by focusing on the prevalent use of masculine generics in gendered languages to address women. Using a large, representative sample of Hebrew-speaking adults in Israel, and exploiting the inherent feature of Hebrew grammar, the researchers test whether addressing women in masculine versus feminine forms affects their math performance.

Existing research demonstrates a correlation between gendered languages and gender inequality across various sectors like labor, credit, education, and household labor distribution. Experimental studies show that addressing individuals in gendered languages influences their attitudes, perceptions, and motivations. For example, answering surveys about sexist attitudes in French or Spanish led to higher reported sexism compared to answering in English. Addressing women in the masculine in academic motivation questionnaires resulted in lower reported task value and intrinsic goal orientation. Although correlations exist, causality remains challenging to prove due to the cultural context intertwined with language. This study differentiates itself by employing an experimental design to investigate causality.

The study used a large, representative sample of Hebrew-speaking adults in Israel (963 participants, 491 women, 472 men). Participants were randomly assigned to one of two experimental conditions: being addressed in the feminine or masculine forms of Hebrew during a standardized math test (SAT-type). The math test comprised six questions from previous university entrance exams. Participants' performance was measured by the number of correct answers, and time spent on the test. After the math test, participants completed an Implicit Association Test (IAT) and a questionnaire assessing their attitudes and beliefs about the association of women and men with science and the liberal arts. Demographic data, including age, education, income, and immigration history (for those born outside of Israel), were collected. The study included a control for age at immigration to assess the influence of language proficiency on responses. Two supplementary experiments were conducted using different tasks (reading comprehension on empathy, and a word generation task) to explore the roles of task-specific and general sex stereotypes in influencing performance.

The main experiment revealed that addressing women in the masculine resulted in significantly lower math scores compared to addressing them in the feminine. This effect was more pronounced among native Hebrew speakers, highlighting the role of language proficiency. Women addressed in the masculine spent significantly less time on the test and reported a stronger belief that "science is for men." Regression analyses, controlling for demographic factors, confirmed these effects. The impact of addressing women in the masculine reduced their math scores by 6.14 percentage points compared to addressing them in the feminine. The effect of language diminished with increased age at immigration, suggesting a link between language proficiency and sensitivity to gendered addressing. Supplementary experiments supported the influence of task-specific and general stereotypes on performance. In an empathy-related task (positively stereotyped for women), women performed better when addressed in the masculine. In a gender-neutral word generation task, women performed better when addressed in the feminine. In the main experiment, the effects on men were less consistent and robust, possibly due to pre-existing stereotypes about men's superior competence in math.

The findings strongly suggest a causal link between gendered language and women's performance in mathematics. Addressing women in the masculine negatively impacts their performance by activating a sense of alienation and making gender stereotypes more salient, thus creating a stereotype threat. The results contribute to the understanding of how seemingly subtle linguistic features can influence behavior and perpetuate gender inequality. This study provides experimental evidence, unlike previous correlational work, supporting the impact of language on performance. The interplay between language proficiency and stereotype threat is crucial, highlighting the importance of linguistic environment in shaping perceptions and behaviors.

This study provides robust experimental evidence demonstrating the significant impact of gendered language on women's mathematics performance. Addressing women in the masculine negatively affects their scores, effort, and perceived association of science with men. The findings highlight the unconscious influence of linguistic structures on beliefs and behaviors, reinforcing the importance of addressing gender inequality within language itself. Future research could explore the impact of interventions aimed at modifying gendered language practices in educational settings and investigate cross-cultural variations in these effects.

The study's limitations include the online, home-based testing environment, which may differ from classroom settings. The absence of a control for the potential influence of the test administrator's gender and the interaction of gendered addressing with other contextual factors within a classroom setting warrant further investigation. Also, the effect sizes observed in the study, while statistically significant, may not be uniformly large across different contexts and populations. Future studies may benefit from exploring these nuances to increase the generalizability of these findings.