Abstract
The present studies test whether French grammatical gender affects bilingual children’s classification of objects as boys or girls in English, in children aged 3 to 5 years (Study 1) and aged 8 to 10 years (Study 2), compared to monolingual children to control for possible cultural biases. In both studies, children tended to classify more objects as boys than as girls. In Study 1, the bilingual children showed a reduced boy bias relative to monolinguals. Only the older children showed a by-object effect of French gender. The bilinguals’ and monolinguals’ classifications were highly correlated. In Study 3, English-speaking adults classified object names as boys or girls. The adults’ classifications were highly correlated with the children’s. The authors argue that the classification of objects by gender is affected by cultural biases as well as knowledge of French. The effect of French knowledge is modified by age.
Many researchers have argued that the habitual use of a language can influence adults’ thinking (e.g., Boroditsky, 2003; Lucy, 1997, 1998; Matlock, Ramscar, & Boroditsky, 2005; Roberson, Davies, & Davidoff, 2000; Whorf, 1956). For example, responses about time can be primed with different spatial arrangements, corresponding to language-specific spatial metaphors for Mandarin and English speakers (Boroditsky, 2001). Not all studies report a clear or simple effect of language on thought (e.g., Imai & Gentner, 1997; Li & Gletiman, 2002). This inconsistency across studies has led some researchers to argue that some cognitive domains may be sensitive to the influence of linguistic structure across languages (e.g., time and space; see Levinson, 1996; Lucy, 1997; see also Coulson & Matlock, 2001). One challenge in testing hypotheses about linguistic relativity is that in comparing the cognitive effects of two languages, researchers often compare people from two different cultures. For example, Chinese has a more transparent number-naming system than English. To see if Chinese-speaking children, therefore, have a greater facility with numeracy, researchers compare children in China with children in the United States, often showing that Chinese children do, indeed, have an advantage with number processing (e.g., Miller, Smith, Zhu, & Zhang, 1995). It is unclear whether these effects are due to the transparency of the Chinese language or a cultural importance placed on early math skills. In fact, one study held cultural effects constant by testing Canadian children who spoke both Chinese and English and found no advantage for processing numbers in Chinese (Rasmussen, Ho, Nicoladis, Leung, & Bisanz, 2006). These results suggest that cultural variables might be more important than linguistic relativity, at least in the domain of early numeracy.
In the present study, we examine both linguistic and cultural effects of the gender of objects. In previous studies, researchers have found evidence suggesting that language shapes thought in gender-marked languages. Gender marking refers to the use of linguistic units to identify people or objects as masculine or feminine (and sometimes neuter). English has limited gender markings (i.e., with the pronouns he and she, applying most often to people). In contrast, some languages, such as French and German, have a gender associated with every noun, marked on determiners and adjectives. Grammatical gender is associated to some extent with natural gender in the world (so words labeling females like femme “woman” in French are often feminine). In other cases, grammatical gender can be fairly arbitrary (e.g., in French, figure is feminine while visage is masculine; both mean “face”). In French, there are a few reliable phonological cues (i.e., the last sound and/or syllable) that predict gender (Karmiloff-Smith, 1979; Tucker, Lambert, & Rigault, 1977). The gender marking on nouns is often mastered early in childhood. French-speaking children use grammatical gender correctly from about 2 years on (Clark, 1985) and can use the phonology of novel words to predict gender in early childhood (Karmiloff-Smith, 1979). Grammatical gender could influence people’s conceptualizations of categories of natural gender.
Developmental changes in the effects of language on thought are expected (Vigliocco, Vinson, Paganelli, & Dworzynski, 2005). For example, an influence of color terms has been detected on adults’ categorical perception of color (Roberson et al., 2000). However, 2- to 4-year-old children do not show these effects (e.g., Franklin, Clifford, Williamson, & Davies, 2005). It is possible that years of language use are required in order for language to affect thought.
With regard to gender, for children to be able to associate grammatical gender with natural gender, they must know something about biological sex as well as cultural concepts of gender. By the time children are about 2.5 years, they usually know whether they are a boy or a girl and that this identity is consistent (see Martin, Ruble, & Szkrybalo, 2002, for a review). At this age, children can classify people as male or female, often based on salient perceptual features such as hair length (e.g., Leinbach & Fagot, 1986). Preschool children know whether males or females stereotypically use objects (e.g., skirts and girls; Guiora & Sagi, 1978; Tap, 1985). Development of gender concepts continues through the later preschool years, when children are actively learning the sociocultural ramifications of gender (e.g., what typical behaviors are associated with genders; Bussey & Bandura, 1992; Martin et al., 2002).
Many studies using different methodologies have shown that adult speakers of Romance languages associate grammatical gender with natural gender (Ervin, 1962; Sera, Berge, & del Castillo Pintado, 1994; Sera, Elieff, Forbes, Burch, Rodríguez, & Dubois, 2002; Vigliocco et al., 2005). The results with children are much more varied.
Some studies have shown that preschool children associate grammatical gender with natural gender. Guiora, Beit-Hallahmi, Fried, and Yoder (1982) argued that Hebrew marks for gender more pervasively than Finnish or English, with gender marking the speaker being male or female. Hebrew-speaking children were better at identifying their own gender between 25 and 30 months than the Finnish- or English-speaking children. However, by 35 months, the same levels of gender identification were attained. Similarly, Martinez and Shatz (1996) found some evidence that Spanish-speaking preschoolers find grammatical gender to be a meaningful category. They asked Spanish- and English-speaking preschoolers to sort pictures of objects and people that went together. Most of the children sorted according to animate versus inanimate categories. Six of the 18 Spanish-speaking children sorted according to grammatical gender; none of the English-speaking children did. When explicitly asked to sort the stimuli into masculine and feminine categories, almost all the Spanish-speaking children did so.
Other studies have shown that preschool children do not associate grammatical and natural gender (Karmiloff-Smith, 1979; Sera et al., 1994; Sera et al., 2002). These studies have shown that older children (about 7 to 9 years) do show such an association (see also Bassetti, 2007). For example, Sera et al. (1994) found that school-aged Spanish-speaking children but not younger ones chose male or female voices for objects with masculine or feminine grammatical gender, respectively.
In all of the previous studies, participants were tested in the language that marks for grammatical gender (e.g., Spanish speakers were tested in Spanish). As a result, the use of the language itself could lead to participants’ associations between natural and grammatical gender (e.g., Sera et al., 1994). One study compared adult native German and Spanish speakers who also spoke English. When describing various objects in English, grammatical gender in their first language predicted the kinds of adjectives used in descriptions (Boroditsky, Schmidt, & Phillips, 2003; see also Ervin, 1962). For example, the English word key was associated with adjectives like little and golden for Spanish speakers (the word is feminine in Spanish) and with adjectives like hard and useful for German speakers (the word is masculine in German).
It is also possible that there are biases within a culture in classifying objects as boys/girls. For example, English speakers sometimes refer to boats as she and personify computers/printers/fax machines. As noted earlier, children associate gender-typed objects (like doll or truck) from at least the age of 3 years on (Carter & Levy, 1988). Braun (1999) reported that Turkish speakers (Turkish does not use gender-marking) have a strong preference to associate particular people words with a gender, most often masculine. To our knowledge, no one has tested the possibility that objects not directly associated with gender-typical activities might also have associations with one gender or another within a culture (like boats being girls for English speakers). If this were the case, then, one possible reason that previous results with preschool children have been inconsistent is that researchers have not controlled for biases in children’s conceptualizations of objects. It is possible that effects of grammatical gender could be seen in preschool years, once these biases are controlled for. In order to test for possible cultural biases, it is important to test monolinguals as well. If there were similarities between monolinguals’ and bilinguals’ classification of objects, this result would be suggestive of cultural biases, since both groups of children are growing up in the same culture, but differ on languages spoken.
Study 1
The primary purpose of this study was to test the hypothesis that preschool children associate grammatical gender in French with natural gender across languages. We compared French-English bilingual with monolingual English-speaking children in English on their categorization of objects (some masculine and some feminine in French) as boys or girls. If there are preexisting biases in how children from a particular culture classify objects, the bilinguals may not necessarily classify objects according to French gender the majority of the time, but simply more often than the monolinguals. A secondary purpose of this study was to describe any cultural biases that might emerge. These biases could emerge either within a group (say, as a general tendency among children to classify objects as either boys or girls) or by particular objects. If preschool children associate particular objects with a gender because of cultural biases, the monolinguals and bilinguals should show similar trends in how they classify objects.
Method
Participants
Twenty-seven French-English bilingual children, aged between 3;5 (years;months) and 5;1, participated in this study: 15 boys and 12 girls. The average age of the bilingual children was 49.9 (SD = 5.0) months. The children were growing up in Alberta, a part of Canada in which English predominates. French is one of Canada’s official languages so the small French-speaking population has access to many French resources, including daycares, schools, and one university campus. The bilingual children were all simultaneous bilinguals, having learned both languages from birth, and heard French in daycare or preschool. The other sources of the two languages varied. Some children’s home language was French and others’ English/French, resulting in variability in the children’s relative knowledge of French and English. We included a vocabulary test because bilinguals’ vocabulary correlates with exposure time to a particular language (Pearson, Fernández, Lewedeg, & Oller, 1997).
As a comparison group, we originally tested 41 monolingual English children between 3 and 5 years. We included the results of 27 children age-matched with the bilinguals. The resulting comparison group included 12 boys and 15 girls, aged 3.4 to 5.1. The average age of the monolinguals was 50.2 (SD = 5.1) months. There was no significant difference between the ages of the monolinguals and bilinguals (F < 1). None of the monolingual children knew more than a few words of another language.
Materials
Bilingual children sometimes score lower on vocabulary tests than monolinguals (Nicoladis, 2008), a difference that could affect the results, as the amount of experience with a language has sometimes been linked causally with the effects of language on thought (Franklin et al., 2005). English vocabulary was measured with the Peabody Picture Vocabulary Test III, Version A (PPVT; Dunn & Dunn, 1997). In French, we used the Échelle de Vocabulaire en Images Peabody (EVIP; Dunn, Thériault, & Dunn, 1993).
Sixteen objects were used to test children’s categorization. As animacy had an effect in some previous studies (Martinez & Shatz, 1996; Vigliocco et al., 2005), half the objects were toy animals. Within these groups, half the objects had masculine and half feminine names in French. The animals with masculine names in French were cat, snake, pig, and zebra; with feminine names: giraffe, cow, turtle, and frog. The four inanimate toys with masculine names in French were pencil, whistle, knife, and yoyo; with feminine names: ball, star, fork, and saw. All the objects looked gender-neutral (e.g., bright orange saw; black cat; no pink or blue objects). With the exception of the saw, the objects were not obviously associated with gender-stereotypical activities for either gender.
Procedure
A native speaker of the relevant language administered the vocabulary tests according to the tester’s manual (Dunn & Dunn, 1997; Dunn et al., 1993). The vocabulary tests were administered before the gender categorization (in English) or naming (in French). For the bilingual children, the French session was on a different day from the English session (usually within a week).
For the English sessions, after the PPVT, the experimenter pulled the toys out of a bag in random order, asking “Is this one a boy or a girl?” In the few cases of hesitation or “I don’t know,” the children were encouraged to guess.
In French, the bilingual children were asked to name the objects in random order.
Coding and analysis
The raw scores rather than standard scores on the vocabulary test were analyzed because standard scores may not be valid for bilingual children (Gathercole, Thomas, & Hughes, 2007).
The children only occasionally answered something other than boy or girl (e.g., don’t know, both boy and girl). Out of the total possible 16, the average number of boy/girl answers for the bilingual children was 15.9 (SD = 0.6) and for the monolingual children 15.6 (SD = 1.0). There was no significant difference between the groups.
Our original analyses compared animals versus inanimate objects. Animacy had no effect on the results and will not be mentioned further.
In French, the children’s gender choice in naming agreed with the intended target gender (M = 85%). Most of the exceptions were because the grammatical agreement with the children’s label was correct but with a word different from the target word (e.g., ballon [masculine] when our target word was balle [feminine]). The results presented below are based on the target gender of the word; analyzing according to the child’s gender choice in French did not change the pattern of results.
To analyze the results by subjects, two different dependent measures were used: (1) the number of classifications corresponding to French grammatical gender and (2) the number of “boy” classifications made. Classifications according to grammatical gender were the number of times children said “boy” for words that are masculine in French and “girl” for words that are feminine in French. Classifications by “boy” were made to test for possible cultural biases: since the children almost always answered either “boy” or “girl,” it made no statistical difference whether we tested for a boy bias or a girl bias. The choice of “boy” was made by flip of a coin. Some distributions were strongly skewed so all of the statistical analyses were carried out on log-transformed data. The results are reported in actual numbers and so reference is made to the log transformations when necessary for clarity.
In order to see if there were any particular cultural biases for individual objects, we also present the number of children who answered “boy” for each object. Again, since most of the children’s responses were either “boy” or “girl,” almost all of the other responses were “girl.”
Results
Classification according to French gender
The first analyses were done regarding the children’s classification according to gender in French. The monolingual children averaged 7.4 (SD = 1.2) out of 16 categorizations corresponding to French gender; the log-transformed data were significantly lower than chance on a one-sample t test, t(26) = −2.72, p = .011. The average rate of corresponding categorizations for masculine objects was 5.0 (SD = 1.7) out of 8 and for feminine objects 2.4 (SD = 1.5) out of 8. The total number of corresponding categorization choices was not correlated with age, r(25) = .005, ns, or PPVT, r(25) = −.386, ns.
The bilingual children averaged 7.6 (SD = 1.8) out of 16 corresponding categorizations, not significantly different from chance, t(26) = −1.56, p = .13. The average rate of corresponding categorizations for masculine objects was 4.4 (SD = 2.0) out of 8 and for feminine objects 3.2 (SD = 2.2) out of 8. The total number of corresponding categorization choices was not correlated with age, r(25) = .024, ns, PPVT, r(25) = −.178, ns, or EVIP, r(25) = −.194, ns.
There was no significant difference between bilinguals and monolinguals on their choice of corresponding categorization on a one-way ANOVA (F < 1,
Classification as boys and girls
The monolingual children were strongly likely to categorize objects as boys (M = 10.3 out of 16; SD = 3.0), significantly above chance, t(26) = 3.57, p = .001. The number of “boy” choices was not correlated with age, r(25) = .200, ns, or PPVT, r(25) = −.011, ns.
The bilingual children also showed a tendency to categorize objects as boys (M = 9.2 out of 16, SD = 3.8) but not significantly higher than chance (t < 1). The rate of categorizing objects as boys was not correlated with age, r(25) = .107, ns, PPVT, r(25) = .096, ns, or EVIP, r(25) = −.098, ns.
To compare the monolinguals’ and bilinguals’ “boy” classifications, we used a 2 × 2 ANOVA [Language Group × Sex], with Sex referring to the sex of the participant (see Figure 1). There was a significant main effect of Group, F(1, 50) = 6.66, p = .013,
Average Number of “Boy” Classifications Given by Bilingual and Monolingual Male and Female Participants
By-object analyses
Figure 2 graphs the number of children who categorized each object as “boy.” As a general rule, more than half of the children in both groups categorized objects as boys, except the cat and the star. There are some differences between the two groups but not necessarily in support of the hypothesis. For example, more bilinguals categorized the frog and the fork as girls than monolinguals (predicted: these are feminine in French). More bilinguals categorized the pencil and the whistle as girls than monolinguals (not predicted: these are masculine in French).
Number of Children in Each Group Classifying Each Item as “Boy” (Study 1)
We correlated the number of monolingual and bilingual children who responded “boy” to each object and found a significant positive correlation across the 16 objects, r(14) = .624, p < .01.
Discussion
The primary purpose of this study was to test the hypothesis that grammatical gender in French influences preschool bilingual children’s categorization of objects as boys or girls, even when asked in English. The results showed that the bilingual children did not classify objects as boys or girls according to grammatical gender in French, neither the majority of the time nor more frequently than monolingual children.
There were, however, differences in the children’s classifications, depending on whether they spoke French or not. As a group, both bilingual and monolingual children showed a higher tendency to categorize the toys as boys than girls. This boy bias could be due to the default use of he in English (Hyde, 1984). However, the bilinguals categorized a significantly lower number of toys as boys than the monolingual English-speaking children. This result suggests that there is an effect of French on the children’s English categorizations, but not on a by-object basis (see Sera et al., 2002, for similar global effects in older participants). Instead, the effect was a more global one, a reduced boy bias. Feminine grammatical gender in French may suggest to children that objects can be girls. Previous studies have generally not shown by-object effects in monolingual children until around 9 years of age (e.g., Karmiloff-Smith, 1979). There was no correlation between French vocabulary scores and children’s categorizations, suggesting that amount of exposure time to French had little effect within this age range.
While the effect of speaking French was significant, the similarities between the monolingual and bilingual groups were quite striking. First, both groups showed different tendencies depending on whether the participants themselves were male or female. The male participants categorized the majority of the objects as boys; the girls in both groups were less likely to do so. One possible interpretation is that, in the absence of stereotypical or known gender cues, preschool children reason through their own gender (see Martin, Eisenbud, & Rose, 1995). A second similarity between the two groups was that there was a strong correlation between the two groups in the children’s tendency to classify objects as boys. This correlation is suggestive of cultural biases in children’s classifications. Will these cultural biases still be observed in older children, who might also show a cross-linguistic effect of language (Karmiloff-Smith, 1979)?
Study 2
The purpose of this study was to test whether French gender classification affected older children’s classification of objects as boys or girls, even when asked in English. One previous study found that Italian, monolingual 9-year-olds chose voices corresponding to grammatical gender in Italian while Italian-German bilinguals did not (Bassetti, 2007). This result was interpreted as an effect of grammatical gender (in German) on the bilinguals’ conceptualization, which then led to a lower link to the Italian gender of objects. This interpretation could be strengthened if bilingual children classify objects according to grammatical gender in a genderless language like English. In this study, we expected to show by-object effects of French grammatical gender on classifications of objects done in English.
We also tested a group of monolingual children of the same age range, again allowing for the possibility of cultural biases in classification.
Method
Eighteen French-English bilingual children aged 8;9 to 10;6 participated in this study: 10 boys and 8 girls. The average age was 113.7 (SD = 7.6) months. All of the children were growing up in Edmonton, Alberta, Canada. All of the bilingual children had been exposed to French from birth, had at least one French-speaking parent at home, and went to French schools (i.e., schools targeting native French-speaking children, not French immersion). The children’s source of English was varied, from regular exposure from birth in the home to minimal exposure from the general community and media. We asked the parents to report the children’s relative proficiency in French and English on a 9-point scale, with 1 representing very minimal English knowledge, 5 representing approximately balanced knowledge of the two languages, and 9 representing very minimal French knowledge. Sixteen parents gave a report of their children’s proficiency, averaging 3.8 (SD = 2.0), with a range of 1 to 7. As a group, the children were strongly French dominant.
As a control group, we tested 30 monolingual children within the same age range and included the 18 who were closest in age. These 18 children ranged in age from 8;5 to 10;7. There were 9 boys and 9 girls. The average age of the children was 113.1 (SD = 6.2) months. There was no significant difference between the ages of the bilinguals and the monolinguals (F < 1). Five of the monolingual children had had 1 year or less of French immersion but did not differ noticeably from the rest of the monolinguals in their gender classifications.
Materials and procedure
We used the same materials and procedure as in Study 1, with one exception: We did not administer the vocabulary tests to the older children. Recall that the vocabulary scores had no relation to the results of Study 1. As a result, all children were tested only in English.
Coding and analysis
As in Study 1, we analyzed the number of children’s responses that corresponded to French gender as well as the number of “boy” responses the children gave. Unlike in Study 1, the saw showed virtually no variation in the responses: All the monolingual children and 17 out of 18 of the bilingual children classified the saw as a boy. We eliminated this item from the analyses. Out of the remaining 15 items, the bilingual children gave either boy/girl answers for an average of 13.9 (SD = 1.4) items and the monolingual children an average of 13.6 (SD = 3.0). One monolingual child gave no boy/girl responses for inanimate objects and was not included in the analysis for possible effects of animacy. The distribution of the data was sometimes skewed so all the statistical analyses were carried out on log-transformed data.
Our first set of analyses included the animacy of the objects (i.e., animals vs. inanimate objects) and the biological sex of the children themselves. These variables showed no significant results and are not considered further.
Results
Classification according to French gender
The first set of analyses was with regard to gender in French. The monolingual children classified the objects according to gender in French for an average of 6.7 (SD = 2.4) out of 15 of their responses. This rate was significantly lower than chance, t(17) = −2.46, p = .025. The average rate of choosing the corresponding categorization to the French gender for masculine objects was 3.9 (SD = 1.4) out of 8 and for feminine objects 2.8 (SD = 1.4) out of 7. The monolinguals’ number of corresponding categorizations was not correlated with their age, r(16) = −.206, ns.
The bilingual children categorized the objects according to the gender in French an average of 9.0 (SD = 1.8) out of 16. This rate showed a tendency to be greater than chance, t(17) = 1.87, p = .079. For masculine objects, the average rate of corresponding-to-French-gender choices was 4.9 (SD = 1.3) out of 8 and for feminine objects 4.1 (SD = 1.2) out of 7. The total number of choosing the corresponding gender classifications was not correlated with the bilinguals’ age, r(16) = −.210, ns, or parental reports of proficiency, r(16) = −.284, ns.
Comparing the bilingual and monolingual children on a one-way ANOVA revealed a significant difference between the two groups, F(1, 34) = 9.18, p = .005,
Classification as boys and girls
In Study 1, both monolinguals and bilinguals showed a tendency to categorize objects as boys. In this study, the monolingual children categorized the objects as boys for an average of 7.4 (SD = 1.9) out of 15 and the bilingual children an average of 7.2 (SD = 1.8) out of 15. The log-transformed data revealed no significant difference from chance for the monolinguals, t(17) = −1.72, p = .103, and significantly lower from chance for the bilinguals, t(17) = −2.21, p = .041.
By-object analyses
Figure 3 summarizes the results of children’s “boy” categorizations by group and by object. As can be seen, the group differences were carried by most of the objects with the notable exception of the frog and the yoyo (both groups tended to categorize these as boys at equivalent rates). It can also be noted that both groups had a tendency to categorize the whistle, yoyo, snake, pig, ball, and frog as boys. Both groups had a tendency to categorize the cat, zebra, star, giraffe, and cow as girls.
Number of Children in Each Group Classifying Each Item as “Boy” (Study 2)
The correlation on the number of children who responded “boy” by object between the two groups was positive and significant both without the saw, r(13) = .575, p < .05, and with the saw, r(14) = .681, p < .01.
Discussion
This study showed that bilingual children between 8 and 10 years showed a small but reliable effect of French grammatical gender on their classifications in English. This result is consistent with previous research with monolinguals showing effects of grammatical gender on conceptualization in later childhood (e.g., Karmilff-Smith, 1979; Sera et al., 2002). However, our research demonstrates this effect cross-linguistically.
There were also stronger effects of cultural biased associations between objects and natural gender. For example, the inanimate objects whistle, yoyo, and ball may have been more strongly associated as being for boys while stars are for girls (see Leinbach, Hort, & Fagot, 1997). Snakes and pigs may have associations with danger and dirtiness, stereotypically associated with males, while cats, cows, zebras, and giraffes may have associations with gentleness, stereotypically associated with females (see Boroditsky et al., 2003). Frogs often become princes in fairy tales, perhaps leading to an association between frogs and boys.
Further evidence for cultural biases comes from the strong correlations between the by-object classifications in the two groups. Both groups of children are growing up in Canada, suggesting that something about Canadian culture is affecting children’s classifications. If this is true, then these cultural biases should be observable in adults as well.
Study 3
In Studies 1 and 2, the children showed strong correlations by-object in their classifications by gender. These results suggest a cultural bias in the classification of objects. To test that possibility, Study 3 concerned gender classification of objects in Canadian adults. We predicted that the adults’ classifications would be correlated with those of the children from Studies 1 and 2.
Method
Fourteen English-speaking Canadian adults between the ages of 18 and 35 years (M = 22.6, SD = 5.4) who reported that they had spent most to all of their childhood in Canada were included in this study. One other adult participated but reported that he had not spent his childhood in Canada so his data were excluded. Four of the included participants were men and 10 were women.
Materials
A list of 174 concrete nouns, including the 16 keywords referring to the test items from Studies 1 and 2, was generated. By including 174 words, we hoped that the adults would pay no particular attention to the 16 words of interest. The other words were generated as being within the same semantic categories as the 16 key words (e.g., snake was one of the keywords and alligator, dinosaur, and lizard were included). In this study, we analyze only the 16 keywords, as a comparison to the children in Studies 1 and 2.
Procedure
The participants were asked to tick off either “boy” or “girl” (in that order) corresponding to all 174 words with the following instructions: Please give your first impression of whether the following objects are better classified as boys or girls. The participants were told that we were collecting data to compare to the responses of children in other studies.
The participants were given course credit for their participation in this study and completed the list very quickly, suggesting that they did not take the time to consider their responses.
Analysis
Here we present the results of the 16 items corresponding to the test words from Studies 1 and 2. The ratio of “boy” responses for each item was calculated out of the total. Note that the adults in this study responded to all items. There were no significant differences between the men and women’s responses in this study. There was a significant effect for animacy, with the average number of “boy” responses for the eight animals 5.5 (SD = 1.0) and for the inanimate objects 2.1 (SD = 1.0), t(13) = 8.00, p < .001.
In order for a particular item to be considered above chance (according to a chi-square), 11 out of 14 participants had to classify an item as the same gender. This was true for 10 out of 16 items (so the classifications of cow, fork, giraffe, pencil, whistle, and zebra could have been chance).
Results
The adults averaged 11.4 out of 16 (SD = 1.6) of their classifications of the items as boys when the saw was included and 10.4 out of 15 (SD = 1.6) without the saw. Both of these differed significantly from chance, t(13) = 8.57, p < .001, and t(13) = 7.09, p < .001, respectively.
Figure 4 summarizes the by-object analyses for the adults. Table 1 summarizes the correlation coefficients between the number of adults who classified the objects as boys and the number of children from Studies 1 and 2 who classified those same objects as boys. As can be seen in this table, all the correlations are positive and significant.
Number of Adults Classifying Each Item as “Boy” (Study 3)
Correlation Coefficients Between the Number of Adults Classifying Items as Boys and the Numbers of Children From Studies 1 and 2 Classifying Items as Boys
p < .05. **p < .01.
Discussion
Even though there is no systematic grammatical gender in English, the English-speaking Canadian adults in this study showed strong preferences for classifying many objects according to boys/girls, with an overall bias toward boys. The adults’ degree of classifying objects as boys/girls was highly correlated with that of children (both English monolingual and French-English bilingual) growing up in the same culture from the time the children are 3 years old.
The correlation was the lowest for bilingual children between 8 and 10 years, most likely because, as Study 2 showed, these children also show cross-linguistic effects from French gender in their classification of objects.
General Discussion
These studies have shown that both culture and language contribute to children’s classification of objects as boys or girls in English, with the effect of language mediated by age.
In terms of culture, it is clear that monolingual English speakers have biases for classifying objects as boys and girls, even though there is no systematic grammatical gender in English. For the 16 items considered here, English speakers classified the majority as boys. While it is possible that this boy bias comes from the default in English to the masculine (Hyde, 1984), it is unlikely that this is the entire explanation. Some items were classified strongly as girls by participants at all ages (e.g., cat, star, giraffe). Some items were clearly classified along stereotypical lines (e.g., for the older children and adults, almost all the participants said a saw was a boy). Previous research has shown that even preschoolers can classify gender-typical objects according to gender stereotypes (Carter & Levy, 1988). The present studies suggest that from the age of 3 years on, people also have gender schemas for objects that go beyond gender-specific activities.
This conclusion has ramifications for testing whether language structure has an effect on thought. Notably, it is important to test for the possibility of cultural biases before concluding that it is something about the structure of the language that affects thought. We cannot rule out the possibility that language is somehow affecting these cultural biases, even if it is not the existence of grammatical gender. It is possible that there is something about language use that leads to the cultural biases. For example, previous studies have shown that words commonly occurring together in phrases or sentences can affect people’s semantic interpretations (e.g., Vigliocco, Vinson, Damian, & Levelt, 2002). It is possible that references to gender around words affects semantic biases of words (Hyde, 1984). For example, if people often talked about turtles in the context of boys playing with them, a bias toward turtles themselves being masculine might emerge. Other cultural representations of objects and animals (such as in fairy tales or in TV shows) might also affect these biases (e.g., Calvert & Huston, 1987).
In addition to the cultural effects, the results from these studies support the hypothesis that language structure can also have a small effect on children’s classification of objects. The 8- to 10-year-old French-English bilingual children in Study 2 were significantly more likely to classify objects according to French gender than English monolinguals. We also argued that the French-English bilingual preschoolers in Study 1 may have shown effects of French gender, in that they classified more objects as girls than English monolinguals. For the bilingual children in both studies, however, the cultural effects outweigh the language effects. The high correlations between the classifications of monolinguals and bilinguals suggest that they are largely using the same criteria to decide on the gender of the objects.
In both Studies 1 and 2, neither age nor measures of the children’s language abilities correlated with the children’s classification of objects. Nonetheless, there were some suggestions that there are changes with age in terms of children’s classifications. Notably, the preschool children (Study 1) showed a strong tendency to classify objects according to their own gender while this tendency was not present in older children (Study 2) nor in adults (Study 3). These effects may therefore be related to preschool children’s ongoing construction of their understanding of natural gender, sociocultural stereotypes of gender, and gender roles (see Martin et al., 2002). There were also some particular items that seemed to change with age, such as zebra, which was classified more frequently as a girl for the children but quite strongly as a boy for the adults. Finally, the adults showed the strongest boy bias over all the items compared to the other two age groups (cf., Braun, 1999). This result suggests that it is possible that an early tendency for a boy bias becomes intensified with age.
All of the interpretations we have made of the present studies are limited by the one kind of task that we asked participants to perform. By explicitly asking participants to classify objects according to boy or girl, we may have been tapping on some aspect of their knowledge that will not generalize to more implicit measures. Before reaching reliable interpretations about the role of language and culture in thought, these results should be replicated using other methodologies, including reaction-timed tasks that might tap implicit knowledge. Additionally, asking participants to explain their classifications might illuminate the source of the gender biases, as in qualities associated with those objects or experience with the objects in particular cultural settings (such as TV shows).
In conclusion, although English does not have systematic grammatical gender, English speakers clearly have gender biases for objects. While there are some changes with age, by and large, these cultural biases appear fairly stable between the preschool years and adulthood. These results also point to the possibility that language can affect these biases: The French-speaking children between 8 and 10 years showed small and reliable effects of French gender on their classification of objects cross-linguistically. However, the effects of language were small compared to the cultural effects. Future research can check for the generalizability of these results with other methodologies and uncover the origins of the cultural effects.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Acknowledgements
We would like to thank the Natural Sciences and Engineering Research Council of Canada for funding this research through a grant to the first author. Allisha Brogan, Mélody César, Monique Charbonnier, Uriah Elago, Kimberly Goddard, Caitlin Hummel, Jessica Lebovic, Carrie Leonard, and Anamaria Popescu helped with the testing.