By Emma Young
Does the language that you speak influence what you think? And do languages that assign a gender to most nouns — such as French and Spanish — lead speakers to feel differently about women versus men, compared with languages that don’t — such as Chinese? Both questions have been hotly debated. But now a major new study, involving an analysis of millions of pages of text in 45 different languages from all over the world, concludes that gendered languages shape prejudice against women.
Gendered languages, such as French and Spanish, Russian and Hindi, dictate that most nouns are male or female. For example, “the ball” is la pelota (female) in Spanish and le ballon (male) in French. In these languages, adjectives and verbs also change slightly depending on the gender of the noun. Then there’s a group of “natural gender” languages, in which nouns aren’t gendered, but pronouns do reflect gender (such as he and she in English, which belongs to this group.) Finally, there are genderless languages, in which a pronoun can refer to either gender, such as Chinese and Finnish.
Earlier research has found that in countries where gendered languages are spoken, women earn lower wages and are less likely to succeed in politics. But researchers have questioned whether the nature of the language itself has anything to do with this. Perhaps, some have suggested, languages simply reflect differences in broader cultural attitudes to men and women.
In the new study, David DeFranza and colleagues at the University of Utah used a technique called Natural Language Processing (NLP) to analyse vast amounts of text taken from Wikipedia and the Common Crawl project, which holds snapshots of text taken from the internet since 2013, comprising more than 630 billion words. Text in 45 different languages, which include the first or second languages of more than half of the world’s population, were included.
The NLP software scanned these texts, looking for close co-occurrences of gender words (such as woman, she, hers, her and man, he, his, him in English) and positive or negative terms. The list of positive terms included love, pleasure and lucky, for example, while the negative terms included abuse, sickness and ugly.
The team found that the gendered languages were much more likely to show skewed associations, with more positive terms being associated with men than with women. For example, in the Commons Crawl component of the study, 19 of the 28 gendered languages used in this analysis — 67% — showed this pattern, whereas none of the 17 genderless languages did.
Various geographic, cultural and demographic factors have been shown to influence gender equality. But the team did control for various of these factors in their analysis (for example, they used data from the Global Gender Gap index developed by the World Economic Forum and the Human Development Index, developed by the UNDP), and the effect still held. “Using languages that span continents and cultures, we find evidence that gender prejudice exists more in gendered languages,” they conclude.
A further analysis used a similar method to explore gender stereotypes within these languages. The team found that in gendered, but not genderless, languages, male words, more than female words, were associated with being “competent”, “skilful” and “confident” (attributes traditionally associated more with men than women) but also “warm” (which traditionally has been associated more with women). In gendered languages, men are seen in a more positive light, overall, it seems.
What might explain this pro-male bias?
The researchers speculate that as gendered languages personify even inanimate objects with a specific gender, the regular use of such terms could frequently, unconsciously reinforce implicit attitudes about men and women. So, if men are traditionally viewed in a more positive light, using masculine nouns could bring to mind the positive associations regarding masculinity and reinforce a pro-male bias. If this is how the effect works, then the broader attitudes of a culture are clearly important, but language type has its own role to play.
There are a few limitations to the study. Most notably, this is an analysis of written text on the internet. Perhaps it doesn’t accurately reflect the attitudes of societies more broadly. However, this text was voluntarily produced by millions of people, and it’s debatable whether studies that involve asking people what they think are better at revealing actual beliefs. Indeed, the researchers argue that their work shows the value of analysing the extraordinary amounts of text held in Wikipedia and Common Crawl for research in social psychology.
Future studies could explore whether other stereotypes, not considered in this study — such as weakness vs power, home vs career, or perceptions of emotional versus rational decision-making — might show different patterns of gender association in gendered vs genderless languages, the team suggests.