This is not to suggest that gender identity is simply binary — male or female — or that gender identity is inflexible for everyone. Nor does it mean that conventional gender roles always feel right; the sheer number of people who experience varying degrees of mismatch between their preferred gender and their body makes this very clear.
In fact, recent neuroscience research suggests that gender identity may exist on a spectrum and that gender dysphoria fits well within the range of human biological variation. For example, Georg S. Kranz at the Medical University of Vienna and colleagues elsewhere reported in a 2014 study in The Journal of Neuroscience that individuals who identified as transsexuals — those who wanted sex reassignment — had structural differences in their brains that were between their desired gender and their genetic sex.
Dr. Kranz studied four different groups: female-to-male transsexuals; male-to-female transsexuals; and controls who were born female or male and identify as such. Since hormones can have a direct effect on the brain, both transsexual groups were studied before they took any sex hormones, so observed differences in brain function and structure would not be affected by the treatment. He used a high-resolution technique called diffusion tensor imaging, a special type of M.R.I., to examine the white matter microstructure of subjects’ brains.
What Dr. Kranz found was intriguing: In several brain regions, people born female with a female gender identity had the highest level of something called mean diffusivity, followed by female-to-male transsexuals. Next came male-to-female transsexuals, and then the males with a male gender identity, who had the lowest levels.
In other words, it seems that Dr. Kranz may have found a neural signature of the transgender experience: a mismatch between one’s gender identity and physical sex. Transgender people have a brain that is structurally different than the brain of a nontransgender male or female — someplace in between men and women.
This gradient of structural brain differences, from females to males, with transgender people in between, suggests that gender identity has a neural basis and that it exists on a spectrum, like so much of human behavior.
Some theorize that the transgender experience might arise, in part, from a quirk of brain development. It turns out that the sexual differentiation of the brain happens during the second half of pregnancy, later than sexual differentiation of the genitals and body, which begins during the first two months of pregnancy. And since these two processes can be influenced independently of each other, it may be possible to have a mismatch between gender-specific brain development and that of the body.
Is it really so surprising that gender identity might, like sexual orientation, be on a spectrum? After all, one can be exclusively straight or exclusively gay — or anything in between. But variability in a behavior shouldn’t be confused with its malleability. There is little evidence, for example, that you really can change your sexual orientation. Sure, you can change your sexual behavior, but your inner sexual fantasies endure.
In fact, attempts to change a person’s sexual orientation, through so-called reparative therapy, have been debunked as quackery and rightly condemned by the psychiatric profession as potentially harmful.
Of course, people should have the freedom to assume whatever gender role makes them comfortable and refer to themselves with whatever pronoun they choose; we should encourage people to be who they really feel they are, not who or what society would like them to be. I wonder, if we were a more tolerant society that welcomed all types of gender identity, what the impact might be on gender dysphoria. How many transgender individuals would feel the need to physically change gender, if they truly felt accepted with whatever gender role they choose?
At the same time, we have to acknowledge that gender identity is a complex phenomenon, involving a mix of genes, hormones and social influence. And there is no getting around the fact that biology places constraints on our capacity to reimagine ourselves and to change, and it’s important to understand those limitations.
The critical question is not whether there is a range of gender identity — it seems clear that there is. Rather, it is to what extent and in which populations gender identity is malleable, and to what extent various strategies to change one’s body and behavior to match a preferred gender will give people the psychological satisfaction they seek.
Although transsexualism (defined as those who want to change or do change their body) is very rare — a recent meta-analysis estimated the prevalence at about 5 per 100,000 — it garners much media attention. What do we really know about how these individuals feel and function in their new role?
The data are all over the map. One meta-analysis published in 2010 of follow-up studies suggested that about 80 percent of transgender individuals reported subjective improvement in terms of gender dysphoria and quality of life. But the review emphasized that many of the studies were suboptimal: All of them were observational and most lacked controls.
Dr. Cecilia Dhejne and colleagues at the Karolinska Institute in Sweden have done one of the largest follow-up studies of transsexuals, published in PLOS One in 2011. They compared a group of 324 Swedish transsexuals for an average of more than 10 years after gender reassignment with controls and found that transsexuals had 19 times the rate of suicide and about three times the mortality rate compared with controls. When the researchers controlled for baseline rates of depression and suicide, which are known to be higher in transsexuals, they still found elevated rates of depression and suicide after sex reassignment.
This study doesn’t prove that gender reassignment per se was the cause of the excess morbidity and mortality in transsexual people; to answer that, you would have to compare transgender people who were randomly assigned to reassignment to those who were not. Still, even if hormone replacement and surgery relieve gender dysphoria, the overall outcome with gender reassignment doesn’t look so good — a fact that only underscores the need for better medical treatments in general for transgender individuals and better psychiatric care after reassignment.
Alarmingly, 41 percent of transgender and gender nonconforming individuals attempt suicide at some point in their lifetime compared with 4.6 percent of the general public, according to a joint study by the American Foundation for Suicide Prevention and the Williams Institute. The disturbingly high rate of suicide attempts among transgender people likely reflects a complex interaction of mental health factors and experiences of harassment, discrimination and violence. The study analyzed data from the National Transgender Discrimination Survey, which documents the bullying, harassment, rejection by family and other assorted horrors.
On a broader level, the outcome studies suggest that gender reassignment doesn’t necessarily give everyone what they really want or make them happier.
Nowhere is this issue more contentious than in children and adolescents who experience gender dysphoria or the sense that their desired gender mismatches their body. In fact, there are few areas of medicine or psychiatry where the debate has become so heated. I was surprised to discover how many professional colleagues in this area either warned me to be careful about what I wrote or were reluctant to talk with me on the record for fear of reprisal from the transgender community.
If gender identity were a fixed and stable phenomenon in all young people, there would be little to argue about. But we have learned over the past two decades that, like so much else in child and adolescent behavior, the experience of gender dysphoria is itself often characterized by flux.
Several studies have tracked the persistence of gender dysphoria in children as they grow. For example, Dr. Richard Green’s study of young boys with gender dysphoria in the 1980s found that only one of the 44 boys was gender dysphoric by adolescence or adulthood. And a 2008 study by Madeleine S. C. Wallein, at the VU University Medical Center in the Netherlands, reported that in a group of 77 young people, ages 5 to 12, who all had gender dysphoria at the start of the study, 70 percent of the boys and 36 percent of the girls were no longer gender dysphoric after an average of 10 years’ follow-up.
THIS strongly suggests that gender dysphoria in young children is highly unstable and likely to change. Whether the loss of gender dysphoria is spontaneous or the result of parental or social influence is anyone’s guess. Moreover, we can’t predict reliably which gender dysphoric children will be “persisters” and which will be “desisters.”
So if you were a parent of, say, an 8-year-old boy who said he really wanted to be a girl, you might not immediately accede to your child’s wish, knowing that there is a high probability — 80 percent, in some studies — that that desire will disappear with time.
The counterargument is that to delay treatment is to consign this child to psychological suffering of potentially unknown duration. This is a disturbing possibility, though much can be done to help alleviate depression or anxiety without necessarily embarking on gender change. But rather than managing these psychological symptoms and watchfully waiting, some groups recommend pharmacologically delaying the onset of puberty in gender dysphoric children until age 16, before proceeding to reassignment. Puberty suppression is presumed reversible, and can be stopped if the adolescent’s gender dysphoria desists. But the risks of this treatment are not fully understood. Even more troubling, some doctors appear to be starting reassignment earlier. Some argue that the medical and psychiatric professions have a responsibility to respond to the child as he or she really is.
But if anything marks what a child really is, it is experimentation and flux. Why, then, would one subject a child to hormones and gender reassignment if there is a high likelihood that the gender dysphoria will resolve?
With adolescents, the story is very different: About three quarters of gender dysphoric teens may be “persisters,” which makes decisions about gender reassignment at this age more secure.
Clinicians who take an agnostic watch-and-wait approach in children with gender dysphoria have been accused by some in the transgender community of imposing societal values — that boys should remain boys and girls remain girls — on their patients and have compared them to clinicians who practice reparative therapy for gays.
I think that criticism is misguided. First, there is abundant evidence that reparative therapy is both ineffective and often harmful, while there is no comparable data in the area of gender dysphoria. Second, unlike sexual orientation, which tends to be stable, gender dysphoria in many young people clearly isn’t. Finally, when it comes to gender dysphoria, the evidence for therapeutics are simply poor to start with: There are no randomized clinical trials and very few comparative studies examining different approaches for this population.
Given the absence of good treatment-outcome data, how can anyone — whether transgender activist, parent or clinician — be sure of the best course of action?
There is obviously a huge gap between rapidly shifting cultural attitudes about gender identity and our scientific understanding of them. Until we have better data, what’s wrong with a little skepticism? After all, medical and psychological treatments should be driven by the best available scientific evidence — not political pressure or cherished beliefs.Continue reading the main story
1While in references to human adults and inanimates gender selection almost always corresponds to sex (he / she) or to the absence of sex (it), it has been well established since Sweet (1898: 42) that in references to animals there was no such close correspondence. For instance Corbett (1991: 12) notes ‘a high degree of variability for animals’, while Swan (1997: 219) does not include them in his introductory account of gender selection: ‘Usually people are he or she and things are it.’ The more recent analyses (Huddleston and Pullum 2002: 488, Gardelle 2006: 109, Siemund 2008: 1) show that even when the antecedent noun specifies the sex, the co-referring anaphor might be a neuter pronoun. This is true both in contemporary English (1) and in earlier modern English (2), in which the animate genders were more common in references to animals:1
(Siemund 2008: 1)(ICE-GB) The cow is either admitted for milking, or it may be turned away.
(Gardelle 2006: 109) (Evelyn 1661: 190) Sir Jo Finch told us of an exquisite poyson of the D: of Florences that kill’d with a drop: That drawing a threit and needle dipt in it thro a hens thigh it perish’d immediately.
2This lack of correlation between gender and sex in references to animals has led to the formulation of a specific criterion for these referents, alternatively defined as projection of personality (Sweet 1898: 42, Curme 1931: 551, Quirk et al. 1985: 341, Leech and Svartvik 1994: 56), familiarity (Zandvoort 1965: 132, Biber 1999: 317), animal thought of in its individual aspect (Kruisinga and Erades 1960: 445, Morris 1991: 158), greater degree of interest, involvement or empathy (Jespersen 1942: 209, Joly 1987: 234, Biber 1999: 317, Huddleston and Pullum 2002: 489) and relevance or importance of sex (Roggero 1988: 202, Sinclair et al. 1990: 29).
3The aim of the present article is to contribute to a better understanding of ‘point of view’ in relation to the contexts of use of gendered pronouns. It seeks to determine whether some measurable trends can be made out – in other words, whether some contexts or configurations favour the animate genders or conversely the neuter, which no study (to our knowledge) has sought to establish. This is achieved through a statistical analysis of gender use at the scale of a multi-million word corpus, the Corpus of Contemporary American English (COCA).
4Because in a majority of utterances it is not possible to ascertain whether the sex of the referent is known to the speaker or not, it was decided to exclude this variable and to focus specifically on cases in which the sex of the animal is known to the speaker. This can typically be ascertained only when the antecedent noun specifies the sex of the animal (e.g. mare, cock-pheasant or gobbler)2; consequently the present study will only consider this configuration. Six variables will be analysed in turn: position of the anaphor relative to the antecedent (same clause, different clause within the same sentence, different sentence), type of utterance (generic, nonspecific or specific), genre, definiteness of the antecedent, syntactic function of the antecedent (more specifically whether it is a grammatical subject or not) and syntactic function of the pronoun. The reasons for selecting these variables will be developed in the analyses below3. Section I details the method used to collect the data. Sections II to VI present the findings for each variable. These were obtained using Pearson’s chi square test complemented by the odds ratio estimate. All the analyses were carried out with R version 2.14.1. P values lower than 0.05 (P < 0.05) were considered to be statistically significant.
I. Data collection procedure
5The corpus used for the study is the COCA, which when the data was collected (March to May 2011) contained 176,389 texts (425 million words) spanning the period 1990-2011 (20 million words per year). The corpus is designed to be balanced and representative of English language use in the United States. It is divided equally among five genres: spoken (90 million words; the occurrences are taken from TV and radio programmes and do not include free conversation), fiction (85 million words, from short stories to movie scripts), popular magazines (90 million words from nearly 100 different publications in a variety of domains such as news, health and hunting), newspapers (87 million words from more than 10 different titles) and academic texts (86 million words from about 100 peer-reviewed journals).
6As there is no exhaustive list of sex-specific nouns to date4, in order to establish such a list a systematic search for the words male and female was carried out in the definition field of the Oxford English Dictionary (2011 online edition). This yielded 3040 occurrences (1115 for male,1925 for female), from which the relevant headwords (sex-specific nouns for animals) were selected. Most sex-specific headwords denoting human beings, this process only yielded 63 nouns. This figure does not include compounds with he, she, lady or man as their first element, which were not retained. In the case of lady and man (e.g. lady elephant), the possible personification involved in the use of the noun could bias the results; as for he and she (e.g. he-goat), the OED does not provide a list of entries, so that exhaustiveness could not be achieved. A distinct study of the influence of those modifiers on pronominal gender would have to be carried out. All nouns that could be sex-neutral (e.g. duck, which can denote the species as well as the female of the species) were also discarded and are therefore not included in this count.
7The 63 nouns obtained were then searched for in the COCA. 30 of them (such as cock-lobster, spawner or vixen) returned no hits, either for the noun itself or for occurrences of the noun with a co-referential pronoun. The final number of nouns under study is thus 33: 16 for males and 17 for females, of which 24 are species-specific and 9 can denote sex for more than one species (buck / doe, bull / cow, cock / hen, male / female and sow). The latter were retained because of their relevance to the study. For each noun, all the hits were read in their context of use in order to determine whether there was a co-referential pronoun. The COCA interface enables an automatic search for collocates, but within nine words to the left or right only, which could have restricted the study. The procedure was therefore carried out manually, and although the mean distance between the anaphoric pronoun and its antecedent was eventually found to be 6.16 words, there were indeed 165 relevant occurrences of pronouns further away than 9 words. For reasons of feasibility, when there were well over 1,000 occurrences (which was the case for 19 nouns, for instance 8,563 hits for buck), the search was limited to a random 1,000, as permitted by the COCA interface. A further restriction was imposed on male and female. Their use as heads of noun phrases being extremely uncommon compared to compound uses, two samples of 1,000 occurrences of each noun failed to return any hits for co-referential pronouns. An automatic search for the two nouns with collocating pronouns was then carried out, but in references to animals none were found to be co-referential. It was therefore decided to restrict the search by imposing a determiner. Asample with a male / a female only showed results in compounds or for human beings, so the search eventually had to be restricted to the male / the female. The figures obtained for these antecedents were included (except for the specific variable of the definiteness of the antecedent) but only once a comparison of the corpus data with and without these two nouns showed that the restriction to determiner the did not bias the results. Finally occurrences of pronouns for dead animals, which occurred in the context of hunting or cooking, were discarded, as the /-animate/ feature could have had an additional influence on gender selection.
8What can be concluded from the data collection is that in a majority of utterances the nouns under study do not present co-referential pronouns in the COCA, either because the referent is not mentioned later on or because subsequent reference does not involve a personal pronoun. Out of the 23,519 utterances examined, only 734 pronominal references were found with sex-denoting antecedent nouns for living animals: 413 for males (including 29 for the male) and 321 for females (including 18 for the female). These occurrences, however, can be regarded as representative of gender use for animals in American English and the figure is high enough for reliable statistical analyses.
9The data shows that the neuter is far from marginal in the corpus: it is found in 130 occurrences out of 734 (17.7 %). This finding confirms that in references to animals gender does not correlate with sex and that point of view has a major influence on gender selection.
II. Variable 1: position of the anaphor relative to the antecedent
10The hypothesis to be tested here is whether gender selection is influenced by the position of the anaphor relative to its antecedent – whether they are both in the same clause, or in different clauses within the same sentence, and so on. There are two reasons for this variable to be considered. The first one is a syntactic one. In same-clause contexts the anaphoric expression bears a stronger grammatical relationship to its antecedent than in other contexts. It is part of the same constituent (the clause) and it is typically bound by its antecedent (in the sense given by Binding Theory: it is c-commanded by the antecedent and co-referential with it). The hypothesis to be tested is whether this has consequences on gender selection, possibly by favouring the mention of sex in the anaphor. The second reason is a semantic one. Relevance Theory (Sperber and Wilson 1986) has shown that mentioning an item of information in context implies that it is relevant to the speaker. Choice of lexical information in discourse is guided by the Optimal Relevance principle, which is the result of a balance between processing cost and contextual effect (viz. the triggering of implications); as a result, any mention of information implies that the information is deemed worth the processing cost at the point when it is uttered. If this is applied to gender, the fact that a speaker chooses a sex-specific noun in the antecedent, at least in all the cases in which a sex-neutral word is also available, implies that sex is deemed relevant and important at that stage. When the anaphor is in the same clause as its antecedent, because it belongs to the same event description, the point of view on the referent cannot have evolved since the mention of the antecedent, so that sex still has to be relevant and important to the speaker when the pronoun’s gender is selected. The neuter might therefore be expected to be disfavoured in this context.
The distribution of genders according to position is as follows:
Table 1 – Gender distribution according to the position of the anaphor relative to its antecedent5
Different clause in same sentence
11The neuter is found in 17% of anaphors in same-clause contexts, 20.4% in different clause (same sentence) contexts and 13.4% of anaphors whose antecedent is in the previous sentence. These differences in proportions, however, are not found to be statistically significant (P = 0.111).
12One notes a slightly lower proportion of it for the ‘next sentence’ context compared to the other two. An additional statistical analysis along the sole criterion of sentence boundaries was therefore carried out to establish whether occurrence in a different sentence favoured the neuter:
Table 2 – Gender distribution according to the position of the anaphor
relative to the antecedent – same/different sentence
13The neuter is found in 19.4% of pronouns in same sentence contexts against 13.2% in different sentence contexts. This difference is slightly above 0.05 (P [uncorrected] = 0.059, P [corrected6] = 0.064), so that there is no statistically significant evidence that this variable has any influence on gender selection.
14Finally, an analysis of same-clause contexts against all others was carried out in order to establish whether belonging to the same clause played a role in gender selection:
Table 3 – Gender distribution according to the position of the anaphor relative to the antecedent – same/different clause
15The neuter is found in 17% of pronouns in same clause contexts against 17.9% in other contexts. This difference in proportion is not statistically significant (P = 0.816).
16In order to test the possible influence of Relevance, which would only apply when a generic term is available to the speaker, an extraction was made of pronouns in same-clause contexts whose antecedent nouns had a sex-neutral counterpart (e.g. [sheep] – ewe). Although the number of occurrences is fairly low (26 occurrences) it suggests that the neuter in same-clause contexts is not any less favoured here than in the corpus as a whole: it is found in 26.9% of occurrences (7 out of 26) against 17% in the corpus as a whole.
17As a conclusion on variable 1, there is no evidence that the position of the anaphor relative to its antecedent (same clause / ...) has any influence on gender selection.
III. Variable 2: type of utterance
18Given that animate genders in animal references imply such points of view as an animal thought of in its individual aspect (Kruisinga and Erades 1960, Morris 1991) or a greater degree of interest or empathy (Jespersen 1942, Huddleston and Pullum 2002), the hypothesis to be tested here is whether he and she are more common in specific utterances than in generic contexts.
19A specific NP is one that refers to a limited amount of elements in a given situation (Larreya and Rivière 2005: 171). In other words the NP has a specific referent in the extralinguistic world. The pronouns in utterances such as (3) were therefore considered specific:
Fifteen feet ahead of us, the ram hit the solid canyon wall dead on and then it flew
20Conversely an NP is generic when it applies to a whole class of individuals (Ali-Bouacha 1993: 2, Burton-Roberts 1981 in Galmiche 1985: 34, Larreya and Rivière 2005: 171). Utterances such as (4), which is about the prototypical male of a given species, were therefore considered as displaying a generic use of pronouns:
To construct its monumental tail, the male must divert valuable resources toward a body part that contributes nothing to its ability to compete for food.
21Putting the NP the male into the plural does not change the truth value of the utterance: it applies to the species as a whole.
22The data in the corpus, however, calls for a third type of utterance, in which the NP does not have a specific referent in the extralinguistic world, yet has a virtual one in a given virtual situation, so that the context cannot be construed as being truly generic either. This is illustrated by (5):
You need to grid-pattern the landscape. ‘Only fools rush in when they spot a big buck,’ Ward stresses. ’I’ve spent over an hour planning a stalk. You have to consider everything: terrain, wind direction, thermals, where the buck is bedded, where he might go. If I spot him in the evening, I may just wait and try him in the morning.’
23The plural would change the truth value of the utterance: in the virtual situation there is only one buck at a time. This type of utterance is termed ‘nonspecific’ by Burton-Roberts (1981) and constitutes a third type in his classification, distinct from specific or generic utterances. In the present study this idea of a threefold classification will be retained. The data falls down as follows:
Table 4 – Gender distribution according to context type
24This table shows that references to animals by means of personal pronouns are much more common in specific contexts than in generic or nonspecific ones: they represent 74.9% of all references (550/734). What remains to be determined is whether there are any statistically significant differences in gender distribution across types. The neuter is used in 15.3% of generic references, 12.3% of nonspecific references and 18.9% of specific references. These differences in proportion are not found to be statistically significant (P = 0.297). If nonspecific references are removed and one considers only truly generic and truly specific references, the difference in proportion is not statistically significant either (P = 0.372). As a conclusion on variable 2, there is no evidence that utterance type has any influence on gender selection.
IV. Variable 3: genre
25The genres retained for analysis are those predefined in the COCA: magazines, fiction, newspapers, academic writing and spoken English. The spoken component is regarded by COCA’s designer, Mark Davies, as being representative of overall spoken English, but the fact that there is no free conversation might demand caution.
26Pronominal references to animals are found predominantly in magazines and fiction. Gender distribution according to genre is given in table 5.
Table 5 – Gender distribution across genres
27The neuter is found in 12.5% of occurrences for magazines, 23.6% for fiction, 12.5% for newspapers, 20% for academic texts and only 4.2% for spoken utterances. Pearson’s chi square test shows that the differences in proportion are statistically significant (P = 0.002). Exactly what differences are concerned? There is no significant difference in proportion between magazines and newspapers (P = 0.993), magazines and academic texts (P = 0.195), fiction and newspapers (P = 0.084), fiction and academic texts (P = 0.612), newspapers and academic texts (P = 0.382) or spoken English and all of written English (P = 0.078)7. The only significant difference concerns magazines and fiction (P < 0.001). More specifically, the odds ratio estimate shows that in the corpus the neuter is 2.1 times more likely to occur in fiction than in magazines (OR = 2.152, 95% CI8 = 1.398 to 3.310, P < 0.001). This finding could be related to point of view. The magazines in which pronominal references to animals were found were mainly hunting and outdoor magazines and most of the articles concerned were shooting reports or hunting tips, which might favour a closer relationship to the animal than other types of writing. Fiction, on the other hand, presents animals that are not necessarily close to the speaker.
V. Variables 4 and 5: syntactic properties of the antecedent
28The first hypothesis to be checked here is as follows. If the animate genders are favoured when an animal is thought of in its individual aspect (Kruisinga and Erades 1960, Morris 1991) or when the speaker shows a greater degree of interest or empathy (Jespersen 1942, Huddleston and Pullum 2002), they might be expected to be more common when the antecedent is a definite NP than when it introduces the referent in discourse as indefinite. The distribution of genders according to the status of the antecedent is given in table 6. Occurrences of the male and the female were excluded for this analysis, as were three occurrences with determinative each, which is not marked for definiteness or indefiniteness.
Table 6 – Gender distribution according to the status of the antecedent
29The animate genders occur in 78.7% of anaphors with indefinite antecedents against 82.8% with definite antecedents. This difference in proportion is not found to be statistically significant (P = 0.193).
30The second hypothesis to be tested is whether the animate genders are favoured when the antecedent is a grammatical subject – regardless of whether it is definite or indefinite. Indeed, although the properties of a subject are notoriously difficult to define (e.g. Falk 2006) and although it can be associated with various thematic roles and various statuses in the information structure, it is prototypically associated with topichood (ibid. 6). As such, anaphors whose antecedent is a grammatical subject might be expected to favour animate genders.
Table 7 – Gender distribution according to the syntactic function of the antecedent
Antecedent = subject
Antecedent = other than subject
31The animate genders are found in 84.2% of anaphors whose antecedent is a subject, against 79.6% when the antecedent has another syntactic function. This difference in proportion is not statistically significant (P = 0.105).
As a conclusion on variables 4 and 5, there is no evidence of an influence of the antecedent (definiteness or subject status) on gender selection.
VI. Variable 6: syntactic function of the pronoun
32Although the hypothesis that subject function favours an animate gender has not been validated when applied to the antecedent, it needs to be tested on the anaphors themselves. This section first examines the overall distribution across syntactic functions, and then the specific subject vs. non-subject distinction:
Table 8 – Gender distribution according to syntactic function9
Complement of a preposition
33The animate genders are used in 87.6% of anaphors occurring as subjects, 80.8% for direct objects, 78.1% for prepositional complements and 79.4% for possessive determiners. These differences in proportion are not statistically significant (P = 0.062). If possessive determiners are redistributed to the other three syntactic functions (e.g. if a possessive determiner in a subject NP is treated as being in subject position), however, the figures are as follows:
Table 9 – Gender distribution reduced to three syntactic functions
Complement of a preposition
34This time, the differences in proportions are statistically significant (P = 0.036). More specifically, table 10 shows that the subject function favours the animate genders (P = 0.036):
Table 10 – Gender distribution along the subject/non-subject distinction
Other function (incl. possessive in subject NP)
35The animate genders are 1.8 times more likely in subject position than in other syntactic functions (OR = 1.80, 95% CI = 1.166 to 2.790. P = 0.008).
If possessives in subject NPs are included as ‘subjects’, the difference in proportions remains statistically significant (P = 0.012):
Table 11 – Gender distribution along the subject/non-subject distinction – possessives in subject NPs treated as subjects
Subject (incl. possessive in subject NP)
36In this configuration, the animate genders are found to be nearly 1.7 times more likely to appear in subject position than in other functions (OR = 1.67, 95% CI = 1.118 to 2.506, P = 0.012).
37This statistical analysis of gender use in references to animals whose sex is made explicit by the antecedent noun has enabled to establish a number of facts. First of all it confirms that point of view plays a decisive role in gender selection: the neuter is found in over 1/6 of occurrences (17.7%). Furthermore two variables were found to have a statistically significant influence on gender selection in discourse. In the corpus the animate genders are more likely to occur when the anaphor is in subject position (1.67 or 1.8 times more likely according to how possessives are considered) and 2.1 times more likely in magazines than in fiction. These factors might be directly related to point of view, namely topic status of the referent at the point when the anaphor is uttered, and hunting contexts, which favour closeness to the animal considered. Conversely there is no evidence that type of utterance (generic, nonspecific, specific), position of the anaphor relative to its antecedent (same clause, ...), genre distinctions other than magazines vs. fiction, or syntactic properties of the antecedent (definiteness or subject function) influence gender selection. These findings hold at the scale of the COCA, which is regarded as representative of American English in general. Corpus-based studies for other varieties of English would enable to establish whether they can be extended to English as a whole.
38Finally it can be noted, although this is not relevant to the study of gender, that pronominal references to animals strongly favour specific utterances (74.9%) and two genres (magazines and fiction as opposed to the other genres, 84.7%). They were also found predominantly within the same sentence as their antecedent (73.2%). This last fact confirms Ariel (1990, 1994, 1996)’s conclusion that personal pronouns are used for highly accessible referents. It also complements a sample study by Ariel (1990: 18) which seeks to determine in what contexts personal pronouns are favoured. In her sample of four texts of 2,200 words each personal pronouns are found to be used mostly in next-sentence contexts (60.5%) against 20.8% only for same-sentence environments. The larger corpus used in the present study, although it only concerns a subset of animal references and not all uses of personal pronouns, suggests that same-sentence contexts might be more highly favoured than predicted from Ariel’s sample.