Gender-preferential language use in L1 and L2 argumentative essays? Evidence against lists of ‘gendered’ language features

This study identifies and compares the gender-preferential language features present in the argumentative writing of L1 Indonesian and Indonesian L2 English learners. The data is comprised of 80 English argumentative essays sampled from the International Corpus Network of Asian Learners of English (ICNALE, Ishikawa, 2011) and a comparative corpus of 80 L1 Indonesian argumentative essays collected online from Indonesian university students, both equally divided by gender. Comparison of the data was performed through quantitative analysis of three supposed ‘male-preferential’ features and seventeen ‘female-preferential’ features between the maleand female-produced corpora in L1 and L2 writing. This study investigated (1) the extent of variation in the use of ‘gendered language features’ between male and femaleproduced L1 and L2 texts; (2) whether the use of male/female ‘gendered-language features’ across male/female produced L1/L2 texts match their suggested gender preference, and (3) to what extent L1’s preference for ‘gender language features’ affects male and female learners’ use of such language in L2. The results suggest the majority of supposed gender-preferential features were not significantly different across male/female produced texts, indicating that argumentative essays may be gender-neutral to a certain extent. This study also revealed that L1 preference of gendered language forms does not determine their preferences in the L2. In conclusion, male and female students adopt similar linguistic features to express their arguments. We may claim that gender language forms are not fixed and absolute in academic discourse because instructive texts tend to have a set model to fulfil the pedagogical criteria.


INTRODUCTION
We, as individuals, live in a society where our behaviour is often understood and interpreted based on gender. This interpretation is mediated through a combination of social, cultural, political, and and genetic differences between males and females (Eckert & McConnell-Ginet, 1995;Jule, 2017;Lakoff, 1972;Oakley, 1972;Sunderland, 2000;Talbot, 1998).
The study of language and gender was not prevalent until the breakthrough work of Lakoff (1972). Since then, several studies have emerged in this area, discussing gendered language in a variety of contexts (e.g., Gilligan, 1982;Maltz & Borker, 1982;Swacker, 1975;Tannen, 1990). The debate regarding gendered language use mainly centred around Lakoff's (1972) claims that (1) males and females use different forms of language and (2) the differences in this use are the result of male dominance. Approaches to the investigation of gender in language studies are therefore generally separated into the dominance and the difference approaches, where the former focuses on issues of equality and the latter explores the diversity of language use among men and women as well as the tolerances for such diversity (Eckert & McConnell-Ginet, 2013). Under the difference approach, previous research has sought to identify 'gendered' language features by analysing language variation in texts and speech produced across the genders. Gender-linked language differences are characterised through variation in the use of lexical, syntactic, structural, and content-specific features (Argamon et al., 2003;Koppel et al., 2002;Zheng et al., 2006). Researchers have used various terms in addressing this variation across genders including genderpreferential features (Samar & Shirazizadeh, 2010), gender-based distinctions (Baron, 2004), genderrelated language style (Colley & Todd, 2002), gender-typical style (Rubin & Greene, 1992), gender-based writing styles (Argamon et al., 2003), and gender-specific language characteristics (Sarawgi et al., 2011). Despite variation in their titles, each refers to the notion that certain language features are used extensively and/or exclusively by either men or women when communicating with others. In this research, the term gender-preferential features (Samar & Shirazizadeh, 2010) will hereafter be used to refer to this distinction.
Linguists have categorised the central contrast of language divergence between males and females into two main classifications. Overall, female language is typically associated with an affiliative approach to language use, while male language is heavily characterised by an assertive approach. An affiliative approach suggests women tend to engage and interact positively with their audience (Flynn, 1988;Leaper & Ayres, 2007;Rubin & Green, 1992). Opposite to males, females also tend to focus on managing their relationship with their audience by positively acknowledging the position of others. For example, Leaper and Ayres (2007) listed a range of affiliative attributes, including offering support, a preference for agreement, and a willingness to recognise others' contributions more than seen in male discourse.
On the other hand, the key functions of an assertive approach are giving directive statements, delivering information, as well as disapproving and criticising others' views (Leaper & Ayres, 2007). This approach is claimed to expand males' power to intervene directly and objectively. Furumo and Pearson (2007) also suggested that males tend to use more task-oriented and denotative commands about what others should do. Other studies (e.g. Farrell, 1979;Roen & Johnson, 1992;Taylor, 1978) supported this view, concluding that male language tends to be competitive, antagonistic, and aggressive, while female language is more cooperative and submissive.
In classifying these elements, Biber (1995) outlined an involvement-informational dimension that refers to the differences across male and female language use. Females are claimed to adopt the use of language features indicative of participatory involvement with their audience, including egocentric sequences (e.g., in my opinion, I believe) and modal adjuncts (e.g., maybe, hopefully). Female language has also been claimed to feature heavy use of pronouns (Argamon et al., 2003;Colley & Todd, 2002;Koppel et al., 2002) and tag questions (Baron, 2004;Sterkel, 1988), suggesting females tend to get personally involved in the situation they are discussing, or wish to directly interact with their audience. This is also claimed to be marked by the extensive use of intensifiers, e.g., strongly, really, and very (Mulac & Lundell, 1994;Rubin & Greene, 1992;Sterkel, 1988), affective markers, e.g., excited and anxious (Baron, 2004;Colley and Todd, 2002;Mulac & Lundell, 1994), and diminutives, e.g., a kitty for a cat and veggie for vegetables (Baron, 2004). Various studies (Baron, 2004;Koppel et al., 2002;Lakoff, 1973;Mulac & Lundell, 1994;Rubin & Greene, 1992) have also suggested that female language is strongly characterised by the extensive use of hedges (e.g., somewhat, probably), perceptual verbs (e.g., seems, looks), adversative connectives (e.g., but, otherwise), auxiliaries of possibility (e.g., could, may), qualifiers (e.g., nearly, kind of), and conjunctions (e.g., and, but). These are claimed to reveal an ambience of uncertainty and uneasiness in female language use as a result of perceived maledominated academic fields. In a study, Rubin and Greene (1992) suggested a female genderpreferential language coding scheme based on previous studies (e.g., Flynn, 1988;Hiatt, 1977;Hunter et al., 1988;Rubin & Nelson, 1983;Scates, 1981). Table 1 shows the classification of female linguistic features as suggested by Rubin and Greene (1992), which are to be used in the investigation into gender-preferential features in this study.
On the other hand, males are claimed to exclude such features in their production in favour of 'informational' characteristics indicative of the presentation of facts or information. These include quantifiers, e.g., one, some, and more (Koppel et al., 2002;Mulac & Lundell, 1994;Sterkel, 1988), determiners, e.g., the, a, and an (Argamon et al., 2003;Koppel et al., 2002) and locatives, e.g., above, inside, and left (Mulac et al., 1986;Mulac & Lundell, 1994) to directly present information or facts in their writing. Scates (1981) defined this objective approach as denotative, where the linguistic features used are intended to demonstrate explicit and precise meanings. Although male writing is more likely to exclude expressive or emotional expression, judgmental adjectives, e.g., distracting and badtempered and profanity, e.g., damn and hell (Baron, 2004) are frequently used as a substitution for other, more female-oriented forms. Table 2 summarises the main distinguishing features of male language based on Koppel et al. (2002) and Mulac and Lundell (1994).  However, it is possible that these differences among male/female language users may be genre/register specific. Genre is likely to influence the use of gendered language since the author must use the language features appropriate for and constitutive of the target genre, which may limit the range of 'gendered' forms they can use (Sterkel, 1988;Swales, 1990). Numerous studies investigating gender-preferential features have been conducted in the context of literary works (e.g., Fischer-Stracke, 2010;Holmes, 1998;Stubbs, 2005) and orallyproduced discourse (e.g., Furumo & Pearson, 2007;Hyde & Linn, 1988;Leaper & Ayres, 2007;Schirmer et al., 2005). An example of a contrastive study across registers is that of Argamon et al. (2003) who explored the role of gender across non-fiction and fiction texts in the British National Corpus. The results showed that gender-preferential features in fiction documents were more prevalent compared to the non-fiction documents in the corpus. In a study exploring gender-related structural and rhetorical styles involving 100 fiction and non-fiction books, Hiatt (1976) also found that female fiction authors used more "feminine verbs" associated with feelings, perceptions, and emotions than male fiction authors. In addition, gendered language forms were more likely to be observed in fiction than in non-fiction. Gendered language preferences are believed to more frequently be exhibited in genres such as literature, where the author is afforded the space to express themselves freely. This has been found in drama texts (Culpeper, 2009), novels (Fischer-Stracke, 2009, novellas (Stubbs, 2005), and poetry (Enkvist, 1964).
However, in academic argumentative writing, the use of gendered language forms appears to be reduced due to the formality and standardised structures of this register, where both males and females are restricted to the same standards (Mulac & Lundell, 1994). This is also seen in Smeltzer and Werbel's (1986) study exploring samples of business texts. This study correlated differences in the language features used across male and female writers with that of writing quality, finding no significant differences among the genders. Likewise, Sterkel (1988) conducted a study investigating twenty gender-linked text attributes, including qualifiers, superlatives, politeness words, and a coding scheme of direct/indirectness in business letters. Male and female authors did not differ significantly in their use of any of the twenty language features analysed.
Moreover, while there is a dearth of studies on gendered language features in academic writing, there are fewer still for L2 academic writing. Previous investigations of academic writing and gender features have been typically limited to monolingual sources, where researchers have focused on scientific articles (Argamon et al., 2003;Sarawgi et al., 2011;Koppel et al., 2002), essays (Engelhard et al., 1992;Jones & Myhill, 2007;Mulac & Lundell, 1994;Rubin and Greene, 1992), and web blogs (Sarawgi et al., 2011). Previous studies comparing the L1 and L2 use of gendered language features are as yet rare. Studies that have been done include Samar and Shirazizadeh (2010). They found that the role of gender was more evident in the authors' native language rather than in their second language, as L2 learners lacked the L2 vocabulary to express their stance in the manner predicted by the writers' gender. Argamon et al. (2003), exploring the British National Corpus, also found that gendered language is more noticeable in L1 than L2 production. However, there is still a great need for contrastive L1/L2 studies on gendered language use, particularly covering academic writing.
This proposed study explores the use of genderpreferential features in Indonesian L1 and L2 English argumentative academic writing under the Contrastive Interlanguage Analysis (CIA 2 ) framework (Granger, 2015). Argumentative academic writing represents a dynamic use of language representing real-life experiences while at the same time offering assertions and evaluations of the presented evidence. This type of writing allows the writers to express their interpersonal voice in providing argumentative viewpoints and building mutual consensus with the readers (Hyland, 2005). The question remains as to whether this academic, interpersonal voice stifles the use of genderpreferential language features by gender across L1 and L2 texts. Moreover, it remains to be seen whether (in fact) the notion of gender-preferential language features can be supported by corpus evidence. This study was, therefore, conducted to find out the extent of variation in the use of 'gendered language features' between male and female-produced L1 and L2 texts. Then, it also seeks whether the use of male/female 'gendered-language features' across male/female produced L1/L2 texts match their suggested gender preference. Finally, it determines to what extent L1's preference for 'gender language features' affects male and female learners' use of such language in L2.

METHOD
In this study, the use of gendered language across the seventeen categories of female linguistic features by Rubin and Greene (1992) and three categories of male linguistic features by Koppel et al. (2002) and Mulac and Lundell (1994) are compared across two corpora of academic essays taken from L1 Indonesian and L1 Indonesian L2 English learners respectively. In particular, this research seeks to test the validity of these gender-preferential lists within academic writing and across gender and L1/L2 dimensions.

Contrastive Interlanguage Analysis
Contrastive Interlanguage Analysis (CIA) was proposed by Granger (1996) and refined in Granger (2015) as a methodology used to compare distinctive aspects of language use through language corpora. CIA studies have focused on native language vs. interlanguage varieties (e.g. Breckle & Zinsmeister, 2012;Chen, 2010;Hyland & Milton, 1997) as well as interlanguage vs. interlanguage varieties (e.g. Snape, 2008;Crosthwaite, 2016). The linguistic phenomena investigated under a CIA framework can cover grammatical and/or lexical analyses (Granger, 2015). The most recent version of the CIA framework is shown in Figure 1.

Figure 1
CIA 2 (Granger, 2015) A previous version of CIA (Granger, 1996) sought only to compare L1/L2 and L2/L2 texts but was revised in Granger (2015) to focus on 'varieties', after claims that CIA resulted in the 'comparative fallacy' (Bley-Vroman, 1989). The latest iteration eschews an L1/L2 distinction, focusing instead on Reference Language Varieties (RLV) and Interlanguage Varieties (ILV). RLVs and ILVs can take many forms, e.g. novice L1 writers can be realised as an ILV, with professional L2 writers as an RLV. By carefully controlling for task and learner variables, a meaningful comparison of corpus data produced by different language groups can be conducted across dialectic and diatypic variables, in this case, argumentative academic essays. Our RLVs in this study include male-and female-produced essays from L1 Indonesian, while our ILVs in this study include male-and female-produced essays from L1 Indonesian L2 English learners. However, due to differences in the presence/frequency of particular grammatical categories between Indonesian/English (e.g. use of determiners is far more frequent in English), we do not directly compare ILV and RLV corpora quantitatively, instead of inferring ILV/RLV differences from our interpretation of the separate analyses.

Corpus data -ILV
The L1 Indonesian L2 English ILV data was sourced from the International Corpus Network of Asian Learners of English (ICNALE, Ishikawa, 2011), a freely available downloadable collection of L2 learners' writing and speaking production compiled to facilitate CIA research (Ishikawa, 2011). The ICNALE written data is comprised of argumentative essays produced by both males and females across only two topics: For the purposes of the present study, our data sample is taken only from the ICNALE's B1-2 (corresponding to a mid-point between CEFR B1-B2 levels). The total number of texts written under the PTJ prompt numbered 48 texts from male writers and 35 texts from female writers, with an equal number from both groups for texts produced under the SMK prompt. Texts from lower levels were not considered as the texts did not contain enough of the gendered language features for analysis, while there were only six texts at the highest ICNALE B2 level from writers from L2 Indonesian language backgrounds. Due to the poor quality of some of the ICNALE texts (e.g. some texts appeared to show the use of translation software, or were duplicates), the researchers handpicked 80 essays using purposive sampling across both PTJ/SMK prompts, with 40 produced by male writers and 40 by female writers. In total, the observed data involved 10,666 and 11,144 tokens from the male and female groups, respectively.

Corpus data -RLV
To generate an equivalent L1 Indonesian data set as an RLV, 40 Indonesian EFL students (20 females and 20 males) were recruited using random sampling from an undergraduate academic discussion course at a university in Indonesia. All were native L1 Indonesian speakers. The L1 Indonesian texts were collected online under the same conditions as that of the ICNALE data to ensure comparability. The two ICNALE writing prompts (PTJ/SMK) were translated into Indonesian, and participants were informed that each essay should be comprised of between 200-300 words and meet the requirement of a good argumentative essay with a clear thesis and supporting statements. The sample of writing collected totalled 80 Indonesian argumentative essays with a 50/50 split across PTJ/SMK prompts. With approximately 250 words submitted for each essay, in total, the data totalled 9,605 and 9,268 tokens from the male and female groups, respectively.
The overall structure of corpora used in this study is summarised for the reader in Figure 2.

Analysis
The researchers generated a set of wordlists of gender-preferential language features by expanding the existing lists created by Koppel et al. (2002), Mulac and Lundell (1994) and Rubin and Greene (1992) in both English and Indonesian languages (Tables 3-6). The examples for each category were checked by two native speakers from each language background to ensure accuracy and completeness.    , ke, atas, dalam, dekat, sana, sini, situ, bawah, tengah, sudut, depan, belakang, sekitar [In, to, above, in, near, there, here, there, under, middle, corner  To determine the frequencies of the features contained in Tables 3-6, the corpus query tool CoQuery (Version 0.10.0, Kunter, 2017) was used. Rather than searching for each word individually, CoQuery allows the user to derive the frequencies of all query terms simultaneously from a .csv file. For certain language features that may fulfil multiple functions (e.g. modals), the researchers doublechecked the results for these by consulting concordance lines for each word/phrase and reducing the total frequency for each term identified as serving an alternative function, or where the function was not clear from the concordance output. The output frequency lists were exported to Excel sheets to facilitate further statistical analysis using the Log-Likelihood Calculator (Rayson & Garside, 2000) to determine whether any gender-preferential features or categories were significantly over-or under-used across ILVs and RLVs using the Log-likelihood Test (G 2 ), with effect-size values calculated under the Effect Size for Log-Likelihood (ELL) criterion (Johnston et al., 2006).

Male-preferential features
Referring to the works by Koppel et al. (2002) and Mulac and Lundell (1994), the male writing style is characterised by frequent use of 'informational' linguistic features including determiners, locatives, and quantifiers. This section presents the malepreferential features used by male and female authors in the L1 Indonesian argumentative essays (RLV). Male authors tended to use locatives, quantifiers, and determiners more frequently than female authors ( Figure 3).

Figure 3 The Frequency Counts of Male-Preferential Features in L1 Writing
The data in Table 7 suggests that all three malepreferential feature categories were used more frequently by male writers, but only significantly so for locatives (G 2 = 9.58; p < 0.01). The differences found in other features were not significant as their Log-likelihood values were lower than the significance threshold of 3.84. The effect size (ELL) shown for locatives, however, was not large (ELL = 0.00009).

Female-preferential features
The 'involved' female-produced data has seventeen distinct categories as classified by Rubin and Greene (1992). Figure 4, which compares the frequency counts for female-preferential features in the L1 essays, indicates that most of the female-preferential features were used more frequently by females than male writers, including additive connectives, secondperson pronouns, modal adjuncts, egocentric sequences, refusals, illustrators, auxiliaries of possibility, intensifiers, perceptual verbs, conditional connectives, illative connectives, and de-intensifiers.
The other five features that were hypothesised to be more frequently used by the female authors were relatively underused, including first-person pronouns, proximals, causal connectives, temporal connectives, and adversative connectives. Table 9 shows that the use of additive connectives, second-person pronouns, and modal adjuncts are significantly linked to female writing. However, first-person pronouns and proximalssupposedly female-preferential features-were significantly more likely to be used by male authors. Furthermore, no significant differences were observed across the genders for the other twelve categories, showing that there appears to be little impact of gender on the use of female-preferential features in argumentative writing.   ). There were no significant differences observed in the use of other seventy femalepreferential words/phrases found in the corpus across the genders, while thirty-six words/phrases were absent from either sub-corpus.

Gender-preferential features in English (L2) writing
Male-preferential features Figure 5 and Table 11 compare the results of the occurrences of the three categories across male and female writing in the L2 English dataset. As shown in Table 11, the Log-likelihood value for determiners was 4.84, showing a statistically significant difference at the level of p < 0.05. On another note, the use of locatives and quantifiers across both genders were not seen as significantly different despite a higher raw frequency in the male data than the female data. Table 12 describes which particular expressions (across determiners, locatives and quantifiers) were more likely to be used by male writers as compared with females. Due to limited space, only the words and phrases which exhibited significant differences in use across male/female subcorpora are provided. The data shows that out of the forty-eight expressions listed as male-preferential, only three were significantly overused by males, namely the locative at, the quantifier many, and the determiner the. Note. ****p < 0.0001; critical value = 15.13 ***p < 0.001; critical value = 10.83 **p < 0.01; critical value = 6.63 *p < 0.05; critical value = 3.84 Note. ****p < 0.0001; critical value = 15.13 ***p < 0.001; critical value = 10.83 **p < 0.01; critical value = 6.63 *p < 0.05; critical value = 3.84   Figure 6 compares the frequency of occurrence of the female-preferential features across the male-and female-produced L2 essays. Ten features were seen to be more prevalent in the female-produced subcorpus, including second-person pronouns, deintensifiers, additive connectives, adversative connectives, temporal connectives, proximals, auxiliaries of possibility, conditional connectives, first-person pronouns, and egocentric sequences. However, against predictions, the remaining seven features (causal connectives, modal adjuncts, refusals, perceptual verbs, illustrators, intensifiers, and illative connectives) occurred more often in the male-produced essays. Table 13 depicts the Log-likelihood and ELL results across the male-and female-produced data for the use of female-preferential features. Only four features seem to be significantly more likely to be used by females; namely second-person pronouns, de-intensifiers, additive connectives, and adversative connectives. There were no significant differences observed in the use of the other listed femalepreferential features across the male-and femaleproduced data. Table 14 describes the over/underuse of individual female-preferential features across all categories in the male-and female-produced data. The data shows the words you, just, your, still, therefore, also, I believe, however, might, first, and will were significantly more likely to be used by female L2 writers, in descending order of significance. However, against predictions, the words auxiliary of possibility must and causal connective because were significantly more likely to be used by male L2 writers. There were no other significant differences in male/female use across other seventysix female-preferential expressions that were present in the L2 corpus. Seventy-nine additional femalepreferential expressions were not found in the L2 corpus and were therefore excluded from the statistical analysis.

Figure 6
The Frequency Counts of Female-Preferential Features in L2 Writing

What is the extent of variation in the use of 'gendered language features' between male and female-produced L1 and L2 texts?
The results of our analyses suggest that there are instances of significant variation in the use of 'gender-preferential' language features across maleand female-produced L1/L2 essays, although these are far less prevalent than was predicted by the literature. Of the so-called 'male-preferential' features, one out of three (locatives) exhibited significant differences across male-and femaleproduced writing in L1 Indonesian, with only determiners in the L2 writings. While this partly supports the findings of Koppel et al., (2002), Mulac and Lundell (1994), and Sterkel (1998) about the use of determiners as an indicator of male writing and Argamon et al. (2003) about locatives, overall, our data, in fact, indicates that male and female authors of argumentative essays tend to use 'informational' language features equally. This trend is also found at the word/phrase level, with only eight out of thirtytwo word/phrases showing significant male/female differences in the L1 data, and only three out of thirty-three word/phrases in L2 data, indicating that male and female authors used the majority of the 'male-preferential features' at similar frequencies.
However, the absence of any significant differences for quantifiers is in opposition to the outcomes of previous studies (Biber et al., 1998;Koppel et al., 2002;Mulac & Lundell, 1994;Scates, 1981;Sterkel, 1998;Swacker, 1975) that claim such forms are more likely to be found in male-produced language.
Regarding female-preferential features, out of seventeen female-preferential categories listed for both L1/L2, only three out of seventeen features showed significant male/female differences in use (additive connectives, second-person pronouns, and modal adjuncts), with only four in the L2 data (second-person pronouns, de-intensifiers, additive connectives, and adversative connectives). This trend is also seen at the word and phrase-level, where only 9 out of 79 features present in the corpus showed differences in male/female use in the L1 data, with 13 out of 79 in the L2 data. To some degree, the increased presence of second-person pronouns, deintensifiers, additive connectives, modal adjuncts, and adversative connectives in a female language is in line with the studies conducted by Mulac and Lundell (1994) and Sterkel (1988). However, most of the features exhibited similar frequencies of use between men and women, both in L1 and L2, which is more in line with the findings of Sarawgi et al. (2011).
As the results for L1/L2 groups appear to be similar, we mainly attribute our findings to the genre in which the texts were produced. Our results indicate that argumentative essays tend to be gender-neutral to a certain extent, confirming the studies of Samar and Shirazizadeh (2010) and Mulac and Lundell (1994). Of course, we would require a follow-up study comparing male/female-gendered language use across multiple genres to confirm this hypothesis.
Does the use of male/female 'gendered-language features' across male/female produced L1/L2 texts match their suggested gender preference? As mentioned in the previous section, there was some positive evidence that certain gender-preferential features were in fact more likely to be used by their hypothesised gender group, as seen in the use of locatives (L1) and determiners (L2) by men, as well as forms including second person pronouns, deintensifiers, additive connectives, etc. by women. There were no significant differences between male/female produced texts in L1 or L2 for the majority of 'gendered' linguistic categories and individual word/phrase features in the data.
In fact, certain 'female' preferential features were shown to be more frequently used by men in L1 Indonesian, and vice-versa. ke, a male-preferential locative (which translates to to in English) was found to be more frequently used by women at the p<.05 level. First-person pronouns and proximalsconsidered 'female-preferential'-were more likely to be used by men, while the expressions kita [we], sekitar [around], and agar [so that] were also more likely to be used by male writers. In L2 English, must and because, considered as female-preferential forms, were more likely to be used by male writers. These findings seriously call into question the reliability of any list of so-called 'genderpreferential' features if corpus data suggests that certain linguistic categories or forms are in fact more likely to be used by the opposite gender for which they have been categorised.
To what extent do L1 preferences for 'gendered language features' influence male and female learners' use of such language in L2? This research has explored the use of genderpreferential features in Indonesian L1 and Indonesian L2 English varieties at the upper-intermediate level of proficiency. While we cannot directly statistically compare L1/L2 data due to differences in the presence of certain gendered language features in the grammars of the L1/L2, we can infer from the individual results that the hypothesised preferences for gendered language use were not more likely to be observed in the L1 Indonesian texts than they were in the L2. While male authors used locatives more frequently in L1, this was not seen the L2 data, where determiners occurred more frequently in the maleproduced texts. For the female-preferential features, additive connectives, second-person pronouns, and modal adjuncts were indeed more likely to be used by women than men in L1, but only additive connectives and second-person pronouns were used more by women in L2. De-intensifiers and adversative connectives, which were not overused by female writers in L1, were more likely to be used by women in L2. Considering these findings, we conclude that the preference for gendered language features is generally not more prominent in L1 than in L2, unlike the findings of Argamon et al. (2003) and Samar and Shirazizadeh (2010).

CONCLUSION
The present study has investigated the presence of socalled 'gender-preferential language' in L1 Indonesian and L2 English argumentative essays. Motivated by the lack of investigation of gendered language features in academic texts, we determined that lists of such features need to be treated with caution, as their predicted use is not always realised in either L1 or L2 data, and there is little evidence that any L1 preferential use carries over to L2. We also (partially) conclude that argumentative essays appear to be 'gender-neutral' to some degree, in that this genre does not appear to provide writers with enough opportunity to demonstrate their gender-preferential identity through language.
Two main limitations observed in this study dealing with the sizes of the corpora and the wordlist used for comparison. Firstly, the sizes of the corpora from which to generalise the results of the statistical analysis were relatively small. As a result, a larger corpus would allow for more reliable statistical comparison, although large-scale Indonesian L1/L2 corpora are still rare. Secondly, both Indonesian and English wordlists were compiled by the 1 st author. It is possible that certain other words related to specific categories have not been included in the wordlists, although two native speakers of both languages were used to confirm the coverage of the wordlists used for the present study. We invite other researchers working on L1/L2 texts produced by speakers of Indonesian, or other languages, to address the limitations of this study in future research.