Cognitive effects of grammatical gender in L2 acquisition of Spanish: Replicability and reliability of object categorization

1.1 Grammatical gender and its cognitive effects

In this article, we focus on the potential cognitive effects of language experience in the domain of grammatical gender. Grammatical gender is a type of noun class system in which nouns within a language are classified based on how other words, e.g., adjectives, agree with them. For example, Spanish nouns are either feminine or masculine depending on the form of associated adjectives (Beatty-Martínez and Dussias 2019). The Spanish word for ‘beautiful’ differs in form depending on the noun that it is describing: la luna es bonita ‘the moon is beautiful’ vs el sol es bonito ‘the sun is beautiful’. Grammatical gender, thus, provides good conditions to test relativity because it is (1) part of language structure (unlike differences in color terms, for example), (2) unnecessary for communication (some languages do not have grammatical gender), and (3) differs across languages (the noun ‘bed’ is masculine in Italian but feminine in Spanish).

A large set of literature provides evidence that grammatical gender can affect (or ‘interfere’ in the words of Vernich (2017)) speakers’ categorization of inanimate objects (e.g., Almutrafi 2015, Belacchi and Cubelli 2012, Beller et al. 2015, Bender et al. 2011, 2016, Lambelet 2016, Pavlidou and Alvanoudi 2018, Phillips and Boroditsky 2003, Vernich et al. 2017). Effects have been found in speakers of different ages (e.g., children in Belacchi and Cubelli 2012, Flaherty 2001, Saalbach et al. 2012, Sera et al. 2002, Sera et al. 1994, and adults in Belacchi and Cubelli 2012, Beller et al. 2015, Bender et al. 2011, Kurinski et al. 2016, Lambelet 2016) and a variety of tasks, including voice-attribution (e.g., Athanasopoulos and Boutonnet 2016, Kurinski et al. 2016, Vernich 2017), sex assignment (e.g., Belacchi and Cubelli 2012, Pavlidou and Alvanoudi 2018, Sera et al. 1994), properties judgment (e.g., Flaherty 2001, Imai et al. 2014; Saalbach et al. 2012), among others.

Even with this increasing body of evidence for the effect of grammatical gender in categorization, the validity of the link between grammatical gender and effects in cognitive tasks is not unquestioned because differences between groups of speakers may not be solely influenced by grammatical gender. Other differences between the groups in question, like cultural factors, may also play an important role. In order to disentangle the relationship between culture and language, one must single out the specific property of language to be scrutinized and exclude additional and/or alternative cultural factors that could play a role in explaining the cognitive phenomena. Doing this is complicated because most studies involve a cross-linguistic method which compares “two grossly separate groups of speakers” (Beller et al. 2015, 333). The same can be said when comparing speakers of languages with grammatical gender systems to those without – it is difficult to prove that cognitive behavior is not influenced by cultural differences in such a case.

1.2 Validity and cognitive effects of grammatical gender in L2 Spanish acquisition

Recent studies have addressed this issue of validity by studying the cognitive effects of acquiring grammatical gender. This allows for a comparison of participants from the same culture: native monolingual speakers of a language without grammatical gender vs native speakers of the same language who also speak a second language with grammatical gender. In fact, over the past two decades, linguists have conducted experiments comparing monolingual English speakers to native English speakers who also speak a grammatically gendered second language (e.g., Forbes et al. 2008, Kousta et al. 2008, Kurinski and Sera 2011). These studies explore the concept of linguistic relativity, but they also avoid the issue of cultural differences among participants.

Kurinski and Sera (2011) focused on the manner in which the acquisition of Spanish in native English-speaking adults affected their categorization of inanimate objects. A total of 50 students of beginning, university-level Spanish courses completed two tasks involving grammatical gender. The first task measured the participants’ acquisition of grammatical gender. In a classroom setting, participants were presented with images of nouns and asked to provide the Spanish words for them (including the definite articles, which show gender agreement). The second task measured the participants’ categorization of inanimate objects. The participants were asked to assign a male or female voice to the same images of nouns used in the first task. The nouns included 8 control items (nouns with biological sex: four male and four female), and the remaining 40 nouns were divided into 4 groups of 10 by the following categories: 1) artificially occurring feminine nouns, 2) artificially occurring masculine nouns, 3) naturally occurring feminine nouns, and 4) naturally occurring masculine nouns (where naturally occurring refers to nouns produced by the natural world, and artificially occurring refers to human-made nouns). Kurinski and Sera (2011) distinguished between naturally and artificially occurring nouns because of a study by Sera et al. (1994), which found that monolingual English speakers were more likely to classify natural objects as feminine and artificial objects as masculine. The control items ensured that the participants actually matched male voices to males and female voices to females. Finally, the participants completed both tasks four times throughout the academic year. The data showed that change occurred after just 10 weeks of Spanish instruction. As participants’ understanding of grammatical gender increased, they were more likely to assign male voices to masculine nouns and female voices to feminine nouns. Specifically, participants most often matched the male voice to artificially occurring masculine objects and the female voice to naturally occurring feminine objects. Kurinski and Sera’s (2011) study suggested that acquiring Spanish as a second language impacted participants’ categorization of inanimate objects.

A similar study sought to explore the cognitive effects of acquiring grammatical gender by comparing English monolinguals to L2 Spanish speakers whose native language was English. Kaushanskaya and Smith (2016) assembled a list of 22 English nouns, 11 of which had masculine Spanish translations and 11 of which had feminine Spanish translations. The nouns were matched with proper names, taking note to avoid phonological similarities. The proper names were previously proven to be statistically considered either male or female names, and the pairings were then divided into two groups: (1) gender-congruent pairs (the Spanish translation’s gender matching the biological sex of the referent) and (2) gender-incongruent pairs (the Spanish translation’s gender not matching the biological sex of the referent). English monolinguals, low Spanish-exposure bilinguals, and high Spanish-exposure bilinguals were presented with the pairings in random order and subsequently tasked with recalling the names matched with each noun. Additionally, a small control group of native Spanish speakers was tested to ensure that the procedure would actually bring about the predicted grammatical gender effects, and “the data from this group showed a congruence effect, with gender-congruent proper names retrieved more successfully at testing” (p. 35). Kaushanskaya and Smith (2016) then averaged the correctly-recalled gender-congruent proper names vs the correctly-recalled gender-incongruent proper names, using paired-samples t-tests to compare the three groups’ performance in accurately recalling the gender-congruent vs gender-incongruent condition. The tests revealed that the English monolinguals and low Spanish-exposure bilinguals were equally accurate at recalling the gender-congruent proper names as the gender-incongruent proper names. However, the high Spanish-exposure bilinguals experienced a significant gender-congruence effect, retrieving the gender-congruent proper names more accurately than the gender-incongruent names. Such results suggested that higher exposure to second language learning can have significant cognitive effects.

1.3 Reliability and cognitive effects of grammatical gender in L2 Spanish acquisition

While studies on language acquisition and grammatical gender provide strong evidence for a link between grammatical gender and cognitive tasks, the evidence is not consistently replicated across different types of languages and across different types of experimental tasks.

Vernich (2017) investigated how speakers of Lithuanian, a language with two grammatical genders, exhibited different patterns of categorization of inanimate objects after learning languages with different gender systems: English (no gender), Russian (three genders), German (three genders and gender-marking articles), and Italian (two genders and gender-marking articles). They found significant differences between the groups of speakers in a sex-assignment task in which speakers assigned a sex (male or female) to inanimate objects. They found that Lithuanian speakers who had learned a second language with a gender system were less likely to assign a sex that matches the Lithuanian gender of inanimate objects. This finding further substantiates a link between gender associations and (second) language knowledge. Interestingly, however, the differences between groups did not correspond to the gender system of the second language. They suggest that “learning a foreign language weakens the strength of the link between each item and its gender” (p. 418). Their results replicate the effect of second language acquisition on grammatical gender and object categorization, but do not replicate the direction of the effect. In studies like Kurinski and Sera (2011), the effects corresponded with gender patterns in the second language, whereas for speakers of a gendered language like Lithuanian, they did not, making the relationship between gender in a second language and object categorization more complex than previously thought.

A second issue of reliability within studies on the effects of grammatical gender is that results are not always replicated, especially across different types of experimental tasks. As mentioned above, Samuel et al. (2019) show that differences in experimental design, stimulus selection, language structures, and participant demographics can affect the likelihood of getting positive, mixed, or no support for linguistic relativity. This leads to questions of how reliable the effects of linguistic gender are in two ways. First, some studies with the same experimental task may find contradictory results (see Boroditsky et al. 2003 vs Mickan et al. 2014 for an instance with gender, and Boroditsky 2001 vs Chen 2007 and January and Kako 2007 for an instance in the domain of conceptualization of time). Even if such differences can be attributed to other differences in the studies, e.g., languages involved or speaker attributes, the lack of full replicability indicates that the effects of grammatical gender depend on other factors. Second, studies that employ different experimental tasks differ widely in terms of whether they find support, and in instances where they do find support, how strongly their results support grammatical gender as affecting cognition. Samuel et al. (2019) find that tasks in which participants explicitly assign a voice/sex to inanimate objects are more likely to find results that support relativity (e.g., Almutrafi 2015, Athanasopoulos and Boutonnet 2016, Belacchi and Cubelli 2012, Beller et al. 2015, Bender et al. 2016, Kurinski et al. 2016, Pavlidou and Alvanoudi 2018, Ramos and Roberson 2011, Sera et al. 1994, 2002, Vernich 2017, Vernich et al. 2017) than tasks in which sex is not explicitly mentioned, e.g., when participants are asked to judge the properties of the inanimate objects that are gendered in the language they speak (e.g., Flaherty 2001, Imai et al. 2014, Landor 2014, Mickan et al. 2014, Montefinese et al. 2019, Semenuks et al. 2017). The fact that similar tasks, i.e., classifying inanimate objects in some way, provide mixed or even conflicting results, brings up questions about the type and strength of connection between grammatical gender and cognitive processes. Such a lack in replication across studies and task types suggest that the relationship between gender and cognition is complex and interacts with a variety of factors. In this article, we contribute to the ongoing discussion of how this interaction plays out in different tasks and language (learning) situations to better our understanding of what contexts are most likely to reflect an interaction between language structure and cognition.

3.1 Participants

Students were recruited from two courses taught at Brigham Young University: SPAN 321 (Third Year Reading, Grammar, and Culture) and WRTG 150 (Writing and Rhetoric) with the assistance of the course professors. These courses were chosen to provide a monolingual group (WRTG 150) and two groups of L2 Spanish speakers (both from SPAN 321) from the same population of BYU students. The university recommends SPAN 321 as the first Spanish class for students who learned Spanish on their LDS missions. LDS missionaries leave their homes for 18–24 months, during which time they attend an intensive L2 training center for up to 2 months before departing for their country of assignment, which is not chosen by the missionary but rather the leadership of the LDS Church. Once they arrive in the area in which their target language is spoken, missionaries are assigned a fellow missionary companion with whom they work constantly. As a part of the rigorous schedule, missionaries dedicate at least 1 h to official language study each day. The rest of the time is spent meeting, teaching, and serving the people in the area to which the missionaries are assigned. There is relative uniformity in age, mother tongue, and L2 motivation among LDS missionaries (Hansen 2012). The Spanish department at BYU suggests that students who register for SPAN 321 can speak, write, read, and listen at an Advanced Low level, according to the ACTFL Oral Proficiency Interview scale (ACTFL). In order to further measure Spanish language proficiency, participants self-evaluated their language abilities as a part of the survey questions using the Oral Proficiency Interview scale. Self-evaluation of language proficiency is not as objective as external measures of proficiency; however, the validity of self-assessment has been shown to be quite high (see discussion in Oscarson 1989) and self-assessment has been found to significantly correlate with language performance in meta-analyses over dozens of studies including over 68,000 students (Li and Zhang 2021). In addition to evaluating their language proficiency, participants reported whether or not they served an LDS mission and if they served abroad or in the United States. All participants were entered into a drawing from which six were randomly selected to receive $50 Visa gift cards for their participation.

In total, responses from 149 participants were analyzed: 52 monolingual English speakers, 52 L2 Spanish speakers who served their full LDS mission abroad (36 male, 17 female), and 45 L2 Spanish speakers who did not serve their full mission abroad (15 male, 30 female). Of the 45 L2 Spanish speakers that did not serve their full mission abroad, 27 served their entire mission in the United States, whereas 18 spent some of their service abroad and some in the United States. Due to an error in the survey design, gender and age were not collected for the 52 monolingual English speakers. However, based on the responses from the two subgroups of L2 Spanish speakers, most (if not all) participants likely fell within the age range of 18–24 years old because all but one of the 98 L2 Spanish speakers reported this age range. All 149 participants were members of the Church of Jesus Christ of LDS, making them a relatively homogenous group in terms of religio-cultural background.

3.2 Survey design

Using a similar approach to Kurinski and Sera (2011), a Qualtrics survey was created in which participants were tasked with pairing a list of adjectives typically associated with males or females (Table 1) with a list of ten nouns referring to males and females, ten naturally occurring nouns with feminine Spanish translations, ten naturally occurring nouns with masculine Spanish translations, ten artificially occurring nouns with feminine Spanish translations nouns, and ten artificially occurring nouns with masculine Spanish translations (Table 2). Adjectives were selected from a list developed by Williams and Bennett (1975). Williams and Bennett’s (1975) study presented participants (50 male and 50 female Euro-American college students) with a list of 300 adjectives commonly used to describe people. The participants were asked to select whether each adjective was more regularly associated with men or women. Williams and Bennett (1975) provide a list of adjectives that were selected by at least ¾ of the male and female participants as typically associated with either males or females. In addition, Williams and Bennett classify the adjectives as highly favorable, highly unfavorable, or neutral (based on responses from a different study). For our survey, we use the neutral adjectives that were strongly associated with males and females. Table 1 shows a complete list of adjectives used. We recognize that associations with these adjectives may shift over time; however, we still chose this list because it was empirically based and attempts to avoid the potentially more problematic confound of adjectives that have strong negative or positive connotations.

The survey was conducted entirely in English for all groups. In the survey, adjectives were presented in a random order in pairs of four per question: two male-associated adjectives and two female-associated adjectives. Instead of asking participants to match the nouns with a gendered voice, the survey asked them to select an adjective that they felt most adequately described the nouns listed in Table 2, as mentioned previously (Figure 1). More specifically, participants were told the following.

Figure 1

Survey explanation.

The list of nouns was inspired by, but not identical to, Kurinski and Sera’s (2011) list (see Appendix 1 for original) and was split into two equal lists. One half of the respondents received the first list of nouns, and the other half received the second list of nouns; both lists contained an equal number of nouns from each category and were presented in a random order for each participant. Nouns were separated into two lists in order to reduce the number of overall questions that each participant answered, therefore reducing the survey response time.

The survey was designed and carried out in accordance with the tenets of the Helsinki Declaration and was approved by the authors’ institutional review board. Informed consent was obtained from all individuals included in this study. As a part of the survey, participants consented to participate, answered a series of basic demographic questions, and the L2 Spanish speakers reported or self-rated their language abilities (as previously mentioned). The survey lasted approximately 8–12 min, and it contained 25 experimental questions modeled after the sample question in Figure 2.

Figure 2

Survey sample question.