Abstract
The present study focuses on word order patterns in English personal binomials and argues that conjunct order in this binomial type is partly subject to other factors than those shown in earlier research on binomials in general. On the theoretical level, mixed-gender personal binomials are discussed as linguistic instantiations of dominance and difference thinking in relation to gender. On the empirical level, the article presents an in-depth study of personal binomials in the written component of the BNC. The factor with the highest impact on conjunct order is found to be lexical gender. Moreover, the modifying influence of a range of other factors is tested. Among these are factors that have proven relevant in earlier research on word order in binomials (phonology, orthography, conjunct frequency) as well as factors that have so far not or only sporadically been tested (lexical field, morphology, sex of author, target audience sex). Finally, the findings are related to more recent theoretical discussions of the relationship between language and gender.
The research literature on English binomials, that is, coordinative structures that combine two items of the same word class within a complex phrase, has so far only sporadically touched upon structures that conjoin two personal reference forms. However, the ordering of conjuncts in personal binomials may be subject to mechanisms quite different from those that are relevant for structures like pots and pans, wash and go, nice and slow, or up and down. This difference results from the connection of personal nouns and pronouns to issues of social identity and the fact that the coordinative order of these forms also constructs a relationship between the social groups denoted by the conjuncts. As earlier research has highlighted, a central aspect that has a bearing on conjunct order in personal binomials is lexical gender. Its influence, however, is not consistent, because it is evident that the widely attested macro-pattern “male before female” does not hold for a substantial number of binomials (with ladies and gentlemen being the most prominent exception). Although personal binomials are generally nonidiomatic, in the sense that a reversal of the order does not affect their denotation, it is obvious that certain order preferences have become institutionalized (Gustafsson 1976:624), and that actually occurring coordination orders can therefore be related to discussions of syntactic markedness (Waugh & Lafford 2006:494), that is, the commonly found unmarked order contrasts with the less frequently occurring marked word order (Hellinger 1990:98-99). It is the aim of this article to present the first in-depth corpus study of personal binomials and their gendered order patterns.
The article starts with a section on previous, mainly corpus-based research on gendered linguistic structures in English. This is followed by a section outlining the theoretical background of the study, which locates mixed-sex personal binomials at the interface of dominance and difference thinking. Following a section on the data and method employed, the empirical part of the present study tests the impact of a set of factors on conjunct order in personal binomials. Among these factors are some that have traditionally figured prominently in research on binomials in general (e.g., Malkiel 1959; Cooper & Ross 1975; Benor & Levy 2006; Mollin 2012; see also Landsberg 1996 for an overview of studies), namely phonology, orthography, and conjunct frequency. Others have so far received little or no scholarly attention, maybe because they are less or not relevant for nonpersonal binomials. This group of factors includes lexical field, morphology, sex of author, and sex of target audience. The concluding section contextualizes the findings within more recent theorizations of the relationship between language and gender.
Previous Research
Most previous corpus-based research on the linguistic representation of women and men has so far not dealt with conjunct order in personal binomials but has rather focused on the frequency, collocations, and usage patterns of female and male personal nouns and pronouns (e.g., Holmes 2000; Romaine 2001; Sigley & Holmes 2002; Pearce 2008; Holmes, Sigley & Terraschke 2009; Baker 2010; Caldas-Coulthard & Moon 2010). A recurrent finding in these studies is that female forms generally occur less frequently than male forms, although the former seem to be on the increase (e.g., Romaine 2001; Sigley & Holmes 2002). Collocational analyses have shown that female terms are more likely to be combined with negative concepts or that the adjectival collocates of female and male forms draw a stereotypical picture (for example, collocates originating from the lexical fields of “appearance” and “importance” respectively; e.g., Pearce 2008; Caldas-Coulthard & Moon 2010).
Research that is most directly related to the present study is work that has analyzed order preferences in coordinated female and male personal names (Wright & Hay 2002; Wright, Hay & Bent 2005; Hegarty et al. 2011). Wright, Hay and Bent (2005), for example, have shown that both phonology and conjunct frequency may have an impact on the ordering of female and male personal names in coordination. However, they have also demonstrated that these influences are not usually strong enough to erase the substantial influence of gender on these structures. In fact, they seem to reinforce that influence, because male names are used more frequently and are more likely to possess what Wright et al. call “first position phonology” (Wright, Hay & Bent 2005:539). Hegarty et al. (2011), on the other hand, were interested in how far gender stereotypes affect binomial order, for example, when subjects have to give names to an imaginary couple. They found that traditional heterosexual couples (i.e., couples that follow traditional gender scripts) were more often identified with the male name first—an effect that was absent for the naming of nontraditional heterosexual couples. Moreover, female subjects positioned male names more often in first position when the imaginary couple was said to live in the 1920s and 1950s (compared to the 1980s and the twenty-first century). This suggests that social developments also manifest themselves in binomial order patterns and that these changes are more likely to be led by women.
Among the few studies to date that concentrate on conjunct order in mixed-sex coordinated noun phrases is one conducted by Vefalı and Erdentuğ (2010). However, the authors merely focus on the use of coordinations involving the lexemes woman and man in a corpus of New Age talks and compare it to usage patterns in the British National Corpus (BNC; see the Data and Method section) and the Corpus of Contemporary American English (COCA). Not surprisingly, the findings are consistent across the three corpora: the lexeme man is clearly used more often in first position than the lexeme woman. For the present study, a greater number of mixed-gender binomials will be tested in order to find out in which subfields of the personal lexicon an M1 (male first) predominance is verifiable and which other factors may have an influence on conjunct order in these structures.
A cursory study conducted by Liberman (2009) tests the order of ten mixed-sex binomials in four corpora, namely the BNC and three more recent American English corpora: telephone conversations from the LDC (Linguistic Data Consortium, University of Pennsylvania), newspaper data from the LDC, and the COCA. The results prove to be remarkably similar across the four corpora. It is only for the pair niece/nephew that the American English data show a preference for the reverse order. 1
Finally, the lack of substantial changes in the word order patterns of personal binomials as evident from previous studies is to be judged as less remarkable if one compares American and British English with other varieties of English. Holmes, Sigley and Terraschke (2009:202), for example, found that southern hemisphere varieties (Australian and New Zealand English) are more progressive in terms of avoiding sexist language use.
Mixed-Sex Binomials and the Dominance-Difference Interface
The following section locates mixed-sex binomials at the interface of two fundamental mechanisms: dominance and difference thinking. These two perspectives originate from well-known approaches within the field of language and gender.
As Nöth (1993:26) has noted, word order regularities in (personal) binomials represent a “diagrammatic icon of cognitive, affective, social or cultural order.” An in-depth analysis and discussion of them, therefore, sheds light on the (competing) gendered messages that they convey. Such a conceptualization can easily be linked to how early work in feminist linguistics (e.g., Lakoff 1975; Spender 1980) has theorized the relationship between language and gender, namely in terms of dominance. Almost needless to say, “dominance” invariably was taken to mean male dominance over women. This early work must be credited for establishing language and gender studies as a legitimate field of study and as raising public awareness for fundamental gendered asymmetries in language structure and use. Lakoff’s (1975) work especially spawned discussions revolving around sexist language structures and, therefore, functioned as a central catalyst for attempts at nonsexist language reform. Aspects like the generic use of male forms (he, man) or the morphological markedness (author-ess, usher-ette) and (often concomitant) semantic derogation (governess, mistress) of female personal nouns vis-à-vis their male counterparts were documented in this research (Baron 1986; Cameron 1992; Hellinger & Bußmann 2001).
Conjunct order and agreement patterns in mixed-sex coordination are also structural aspects that have raised the concern of feminist linguists (e.g., Cheshire 1985:24; Hardman 1999). In languages with grammatical gender, the masculine form usually predominates in the sense that a mixed-gender coordinated phrase, in most cases, triggers masculine agreement (e.g., Latin Romeo et Julia beati sunt ‘Romeo.
The feminist critique of conjunct order in mixed-gender binomials crucially hinges on a reading of the M1 position in binomials as a matter of social iconicity, that is, women are more often named in second position, and this is perceived as indicating a power-related social order—a thought already expressed by Simone de Beauvoir’s description of women as “the second sex” (de Beauvoir 1949/1953). This has been claimed not just by feminist linguists, but also by some researchers studying binomials, who generally treat gender under the factor “power” (e.g., Allan 1987:57-58; Benor & Levy 2006; Mollin 2012).
The relationship of conjunct order and power as manifest in personal binomials can be tested by looking at the order frequencies in pairs that show a clear power asymmetry. An example would be master/servant, where one conjunct denotes a person of higher authority than the one denoted by the other conjunct. Table 1 lists the frequencies of some such pairs in the written component of the BNC. These were tested in two orders (using the Simple Query Syntax as outlined in the Data and Method section below): once with the noun denoting the more powerful function in first position (P-p) and once with the noun denoting the less powerful function in first position (p-P).
Order of Personal Conjuncts in Relation to Power.
Even though this kind of binomial has, so far, not been systematically studied, the (introspectively based) selection of pairs in Table 1 illustrates that there is a general trend for the more authoritative term to be the first conjunct (see also Boers & Lindstromberg 2008:341). All differences are highly significant. 2 This indicates that, by extension, gendered asymmetries in binomial order can be interpreted as connected to a power differential that causes one gender to be named more often in first position. However, this does not automatically mean that men are invariably named in first position and women in second position. In fact, the empirical parts of the present study will show that binomials exhibit some variance concerning the position of female and male forms.
The placement of a personal noun in the first position of a binomial can also be interpreted as a kind of dominance from a conceptual point of view. As Cooper and Ross (1975) have pointed out, binomial order is, in many cases, a matter of semantic ordering principles, which they subsume under the “me-first” rule, that is, the first conjunct in binomials often has semantic features like “male,” “adult,” “positive,” “animate,” “friendly,” “front,” “agentive,” or “power source” (Cooper & Ross 1975:65-66; see also Landsberg 1995:71-72). It is interesting to quote them at length, as they describe the semantic features of first conjuncts in such constructions in a (today maybe outdated) hegemonically masculine way. More specifically, these are said to
conspire to provide an approximate portrait of a current American hero, Archie Bunker. Archie, by his own admission, is Here, Now, Adult, Male, Positive, Singular, Living, Friendly, Solid, Agentive, Powerful, At Home, and Patriotic, among other things. In addition, he is General because he is a stereotype, and he is a count noun. (Cooper & Ross 1975:67)
Nöth (1993:30-31) adds some more semantic dimensions such as “more dynamic before less dynamic,” “more active before less active,” “useful before useless,” “sympathy before antipathy,” “good before bad,” “beautiful before ugly,” “interesting before uninteresting,” “valuable before worthless,” and “superior before subordinate.” All of these order patterns indicate that second conjuncts in binomials often denote concepts that are less valued. In a similar vein, Allan (1987:52) proposes a superordinate empathy-related hierarchy of “more familiar” before “less familiar,” which indicates that, if men are more frequently named in first position, the world is predominantly viewed from a male perspective.
Other tools that have been used to explain binomial order are the markedness (Waugh & Lafford 2006) and naturalness theories (e.g., Birdsong 1995; Andersen 2008; Mollin 2012:88). Such approaches are based on the premise that more prototypical representatives of a certain category or more frequent concepts are more likely to occur in first position because this facilitates processing. First conjuncts, therefore, possess a higher degree of cognitive salience and perceptual accessibility and are less marked than second conjuncts.
Mixed-sex binomials are not just associated with power structures. As has been noted by Uchida (1992), for example, difference and dominance are often two sides of the same coin (see also Cameron 1992:84). In other words, the highlighting of gender differences may in fact be a decisive factor on which dominance is based. The specification of women and men in coordinating constructions is likely to be perceived as a construction of women and men as different. The underlying motivation is the belief that women and men are so fundamentally different that they have to be named separately. There are also semantic reasons for conceptualizing coordinations of gendered personal nouns as a matter not just of difference but of gender polarization. As has been shown by some linguists (e.g., Malkiel 1959; Fill 1977; Masini 2006:220; Fiedler 2007:41; Davies 2012:67-68), the conjuncts in coordinating constructions are often in a semantic relation of complementarity, opposition, or incompatibility (cf. rain or shine, food and drink, up and down, black and white), and it is not surprising that the gender dimension is also generally treated in this way (e.g., Gustafsson 1975:89; Dury 1996:25). Still it has to be observed that conjuncts, even when they are antonymous, generally share a number of semantic features (for example, woman and man share [adult] and [human]). In other words, the semantic relation between conjuncts is usually one of “difference in unity” (Dury 1996:26; see also Davies 2012:43). For gendered personal nouns in coordination in particular, this constructs a relatively stereotypical picture of opposites that attract each other.
It is, therefore, not surprising that the lexeme woman has the lexeme man among its most frequent collocates in the written component of the BNC, and vice versa. This can be shown by a search for the two lemmas {woman} and {man} (alone). The collocation facility of BNCweb allows the analyst to specify various collocational window spans to the left and right of the node. This procedure produces a ranking list with the strongest collocates for the window span in question on top. Table 2 presents how the two lexemes rank in varying collocation window spans: four and three words to the left, both four words to the left and right, and finally three and four words to the right of the node. 3 For the production of this list, collocations across sentence boundaries were excluded.
Collocation Rankings of {woman} and {man} for Various Window Spans in the Written BNC.
Taking the largest window span as a starting point (four left to four right), one finds that {man} is the most frequent collocate of {woman}, and {woman} is the third most frequent collocate of {man} (after {who} and {young}, and before {a}). It is interesting to note that the collocational behavior of the two lemmas varies on the two sides of the node. Whereas {man} is the most frequent collocate of {woman} to the left side of the node, it ranks only sixth and fourth on the right side. This indicates already that collocations like men and women are more frequent than women and men. Similarly, {woman} is the second most frequent collocate to the right side of {man}, but ranks only ninth and fourteenth on the left side of the node (where it is a weaker collocate of {woman} than {child}).
From the point of view of semantic markedness (Waugh & Lafford 2006:495), it is apparent that coordinations of personal nouns often represent structures that conjoin two semantically marked items, that is, a female and a male term, none of which can function generically (e.g., girl and boy, prince and princess). However, even when one of the conjuncts is semantically unmarked, the markedness of the other conjunct contextually restricts the meaning potential of the unmarked item to a female or male meaning. For example in the pair actors and actresses, only the female form is, per se, semantically marked and carries a female-specific meaning. In coordination with the unmarked form actor, which oscillates semantically between generic and male meanings, the generic meaning is ruled out. One can, therefore, see that coordination structures may have a polarizing effect because the presence of a female (or, more rarely, male) form blocks a generic reading of the other conjunct (see also Zwarts et al. 2009). This shows that female and male meanings as conveyed by linguistic forms are, to some extent, relational: The terms masculine and feminine do not refer to “essences”, definite, unchanging qualities which exist independently. If there were no concept of femininity, there could be no concept of masculinity either. In fact . . ., the two terms behave exactly like a classical Saussurean sign . . ., defined not by their essence but by their difference. (Cameron 1992:83; italics in original)
The specification of both women and men in the shape of coordinated noun phrases has been a common strategy recommended by feminist linguists for languages with grammatical gender in order to counter the male bias conveyed by the use of generic masculine forms (cf. the UNESCO guidelines for French and German: Desprez-Bouanchaud, Doolaege & Ruprecht 1999; Hellinger & Bierbach 1993). In English this strategy is less relevant, but it has also been suggested as an alternative to generic he (he or she) and generic uses of man-compounds (spokeswomen and -men; Doyle 1995:92). The main aim of splitting is to counter linguistic sexism by increasing female linguistic visibility vis-à-vis men. Therefore this strategy corresponds to the spirit of the difference paradigm within language and gender studies, which is associated with a positive reevaluation of women’s linguistic practices and the claim that female and male experiences are fundamentally different (e.g., Maltz & Borker 1982; Coates 1989; Tannen 1990).
More recent language and gender research has taken a more critical stance on the notions of dominance and difference, mainly by pointing out that their theorization of the relationship between gender and language is too essentialist, because either male dominance or female-male difference is assumed to be omnipresent. Still, it must be acknowledged that dominance and difference are, in fact, two central mechanisms through which gender is constructed in many contexts. Therefore they possess a high degree of psychological reality for most people, even though their invariable relevance cannot be taken for granted.
Data and Method
For the present study, the written component of the BNC was used via the online interface BNCweb. The spoken part of the BNC was excluded in order to make the data under study more homogeneous and because it is much smaller than the written component (the latter amounting to approximately 88 million words). Moreover, it was expected that noun phrase coordination in general occurs more frequently in written language use (see Meyer 1996:38).
The same can be assumed for less traditional patterns of conjunct order in mixed-gender binomials. Since guidelines for gender-fair language focus more on reforming public written language use, their repercussions (if there are any) should first of all be found in written texts. As the latest material in the BNC is constituted by texts dating from the first half of the 1990s, it may be objected that it does not reflect current British English usage and that the BNC can be considered a (largely) historical corpus today. This would certainly be true if one wanted to study lexical innovation patterns. However, for syntactic features like coordination language change is generally much slower, and therefore the BNC can safely be considered an adequate source for the investigation of coordination patterns in present-day British English (see also Hoffmann et al. 2008:45). In the written component of the BNC, 89.2 percent of the material stems from 1975 to 1993 (with 8.5 percent being undated and only 2.3 percent originating from a time before 1975; Pearce 2008:6). This means that the majority of the corpus data originate from a time that coincides with a twenty-year period in which the subject matter of Lakoff’s (1975) famous book Language and Woman’s Place was popularized. The findings of diachronic corpus studies (e.g., Baker 2010; Holmes, Sigley & Terraschke 2009) suggest that more recent language use has indeed become less sexist as a consequence of significant changes in language attitudes (see also McGuire & McGuire 1992:231). On the other hand, the probability that the pervasiveness of the M1 predominance in mixed-sex binomials has been erased is extremely low. As the studies by Vefalı and Erdentuğ (2010) and Liberman (2009) outlined above demonstrate, more recent corpora do not show substantial changes with respect to conjunct order in mixed-gender binomials, despite today’s higher public awareness of linguistic sexism.
In the present study, a range of mixed-gender coordination pairs was tested using the so-called Simple Query Syntax of the BNCweb interface (Hoffmann et al. 2008:93-118). Semantically, the conjuncts in these pairs only differ in terms of their lexical specification as female or male. This means that asymmetrical pairs like mother and son or father and daughter were excluded. The list of pairs used for the quantification was generated by testing common female-male pairs of personal nouns and pronouns. As English has relatively few lexically gendered noun types, the number of pairs tested was per se limited. For pronouns and address terms, for example, a finite set of pairs can easily be generated. For other lexical fields, an effort was made to test the most common pairs that have also received the greatest attention in the research literature on gendered binomials. In the field of kinship terms, for instance, pairs of lexically gendered terms spanning across four generations were tested (e.g., granddaughter/-son, daughter/son, mother/father, grandmother/-father). The following query was used to determine the frequencies of these pairings: (1) {woman} (and | or) (a | the)? {man}
Curly brackets represent a lemma search, that is, a search that covers all inflected forms of the word in question. For example, the query {woman} would find the forms woman and women. The second bracket in the query (and | or) ensures that structures are found that contain either one of the two most common and prototypical English coordinating conjunctions used for noun phrase coordination: and or or (Meyer 1996:30-31). The following bracket (a | the)? allows for articles in between the conjunction and the second conjunct. For pairs in which the second conjunct starts with a vowel (e.g., actress/actor, aunt/uncle), the form of the indefinite article was adapted to an in the query. The question mark after the bracket means that these articles are optional. Accordingly, the query under (1) would find all of the following structures: (2) woman and man – woman or man
a woman and a man – a woman or a man – a woman or man
the woman and the man – the woman or the man – the woman or man
women and men – women or men
the women and the men – the women or the men – the women or men
In principle, asymmetrical structures like the woman and a man or a woman and the man would also be found, but a glance at the random order list of hits shows that there are hardly any such instances in the data. Note that structures involving the genitive forms woman’s, women’s, man’s, and men’s are not retrieved by this query because the inflectional endings are in such cases treated as separate tokens by the BNC search engine (cf. Hoffmann et al. 2008:95-96).
For some word pairs, the query had to be further adapted. For example, the British and American spellings of the word mum/mom were integrated by the following query: (3) {m[o,u]m} (and | or) (a | the)? {dad}
In order to reflect current language use, coordinations of compounds were not tested in their full form but with the repeated component omitted in the second conjunct, as can be seen in the following example: (4) {policewoman} (and | or) {man}
Note that such coordinations do not allow for the insertion of articles before the second conjunct. For the word pair female/male, only the nominal (nonadjectival) usages were of interest to the present study. Therefore a part-of-speech tag was included in the query in this case: (5) {female/N} (and | or) (a | the)? {male/N}
Similarly, only pronouns that can replace a noun phrase were included in the study (i.e., personal, reflexive, and predicatively used possessive pronouns as opposed to attributively used possessives): (6) she (and | or) he her (and | or) him herself (and | or) himself her (and | or) himself - him (and | or) herself hers (and | or) his
The respective queries were performed in two orders: once with the female and once with the male form in first position. In order to be able to make statements about statistical significance, only those word pairs in which at least one of the two orders (female-male or male-female) resulted in ten or more hits were included in the quantitative study. This led to the exclusion of pairs such as granddaughter/-son, grandmother/-father, or waitress/waiter, which are only infrequently coordinated in the BNC. This procedure ensures that the thirty most frequent female-male binomial types are included in the study, while pairs with low frequencies are ignored. All pairs of personal nouns and pronouns included are listed in Table 3.
Gendered Order Preferences in Personal Binomials in Relation to Lexical Subfield.
Empirical Analyses of Factors Affecting Conjunct Order in Personal Binomials
In the following, the written BNC data will be used to test the influence of a range of factors on conjunct order in mixed-sex personal binomials. Gender, the central aspect in this study, will be analyzed not just as a monolithic category but in relation to lexical subfields within the larger field of personal nouns, namely those that are particularly likely to include lexically gendered field members. Afterwards, the influence of three structural factors will be tested, namely phonology, orthography, and morphology. The remaining three factors to be discussed are usage-based: frequency, sex of author, and sex of target audience.
Gender and Lexical Field
Although most personal nouns are not lexically gendered in English, there are certain subareas within the larger lexical field of personal nouns that are more likely to contain lexically gendered members. These will be illustrated in the following. Word pairs in brackets are those that have been included in this study (with at least one of the orders found ten times or more in the written BNC component). Beside general terms denoting women and men (woman/man, girl/boy, female/male [n.]), lexical genderization is common for address terms (Mrs/Mr, lady/gentleman, madam/sir), nobility titles (queen/king, princess/prince, duchess/duke, lady/lord, countess/earl), (traditionally) heterosexual role names (wife/husband, bride/(bride)groom, widow/widower), and kinship terms (mother/father, m[o,u]m/dad, mummy/daddy, aunt/uncle, niece/nephew, daughter/son, sister/brother). In the subfield of occupational and function terms, lexical genderization is the exception in English. For the present analysis, three such exceptional cases were included: hostess/host, actress/actor, policewoman/policeman. The forms host and actor are not lexically gendered per se. In coordination with the female forms, however, their meaning potential is contextually restricted to a male reading. Coordinations of other woman/man-compounds apart from policewoman/man were tested collectively with the query {+man} (and | or) {woman}, in which the plus sign stands for one or more characters.
Table 3 sheds light on the gendered ordering preferences (f-m, m-f) of coordinated personal nouns and pronouns in relation to lexical subfield.
Table 3 shows that for the forms that are most general in their meaning, that is, the categories of general nouns and pronouns, the M1 rate hovers around 90 percent. Lexical subfields with more specific meanings show percentage shifts in both directions. While the M1 rate is categorical or virtually categorical for nobility titles and occupational terms, address terms (77.7 percent M1), heterosexual roles (73.8 percent M1), and kinship terms (56.5 percent M1) show a somewhat weaker M1 predominance, due to individual field members exhibiting a higher F1 (female first) rate. These latter exceptions merit greater attention because they show that the M1 predominance in personal binomials is not absolute. They will therefore be discussed in more detail.
The subfield of address terms is constituted by three word pairs that can potentially be used for directly addressing people. A look at the BNC data reveals that coordinations of lady/gentleman and madam/sir, in most cases, indeed function as address forms (cf. also Mollin 2013:30 on American English). This is somewhat different in the case of Mrs/Mr, for which purely referential functions are also common. Within the subfield address terms, the pair lady/gentleman, which overwhelmingly occurs in the plural in the corpus, constitutes a substantial exception to the M1 rule. Historical evidence suggests that the strong F1 predominance that characterizes this binomial today did not exist in Middle English times, when it exhibited a normative M1 tendency (Potter 1972:314 as cited in Mollin 2013:175). The F1 predominance of this binomial has also been noted in earlier research, which has tended to see the F1 predominance in this pair as a matter of politeness (e.g., Cooper & Ross 1975:105; cf. Mollin 2013:198). However, a closer look at this argument reveals that it is not particularly convincing. For one, it begs the question why this politeness principle is only operative for this particular pair and not for other gendered coordination types. The address terms madam/sir also qualify as polite and are associated with highly formal contexts (Miller & Swift 2001:141-142), but here one finds categorically M1. Another aspect to be witnessed in this subfield is that the M1 rate is highest for the pair Mrs/Mr, in which the male form stands categorically in first position. In contrast to this, the other two pairs in this subfield are not necessarily marriage-related. This suggests that the context of heterosexual marriage may actually increase M1 predominance.
A further aspect that may play a role is that while Mrs/Mr and madam/sir are exclusively coordinated as singular forms in the corpus, only 9 of the 166 cases of lady/gentleman coordination involve singular nouns. This indicates that when two specific or individual people are addressed or talked about, it may be more important to mention the male person first, whereas more general references to larger groups of people allow more easily for women to be named first. One could even speculate whether this constitutes a more general pattern, according to which women are more likely to be named first in coordinations of pluralized conjuncts.
In order to test this, the lemma-based searches illustrated in the Data and Method section above were transformed into token searches by replacing the lemmas with concrete singular forms. By means of this procedure, one can find out how often a certain pair is coordinated in the singular. Table 4 gives the singular frequencies for all pairs whose conjuncts are pluralizable (thereby excluding, for example, pronouns).
Frequencies of Singular Binomials in the Written BNC.
The singular frequency was then divided by the number of all tokens found for a certain binomial, resulting in a percentage of singular usage. The percentages range from 100 percent (i.e., invariable singular use) in the case of madam/sir, lady/lord, and hostess/host down to around 5 percent for lady/gentleman, policewoman/-man, and other woman/man-compounds, which are overwhelmingly coordinated in the plural. A closer look at Table 4 reveals that no clear gender pattern can be identified. Pairs with F1 predominance can be found among mainly singular binomials as well as among mainly plural binomials. The pattern identified for lady/gentleman is, therefore, not applicable to other pairs.
Apart from the field of address terms, the influence of heterosexuality can be further tested in the subfield of heterosexual roles. Although the field as a whole shows an M1 predominance, this is due to only one pair, namely wife/husband. The pairs bride/(bride)groom and widow/widower exhibit a robust F1 predominance. What distinguishes the pair wife/husband from the other two pairs is that it describes the situation within a marriage. The terms widow/widower are applicable only after marriage, when one of the spouses has died. Bride and (bride)groom, on the other hand, describe a situation that only pertains to the time before the actual marriage and the day on which the wedding takes place. In other words, one can see, once again, that the context of heterosexual marriage is associated with an increase in M1 rates. The following traditional gender discourses may be retrieved from this distribution: men occupy the dominant position within marriage; women seem to be perceptually more salient when they are not yet or no longer married.
The lexical subfield of kinship terms shows the highest degree of variance with respect to gendered conjunct order. While the pairs mother/father, m[o,u]m/dad, mummy/daddy, and aunt/uncle show higher frequencies of F1, the pairs niece/nephew, sister/brother, and daughter/son show higher frequencies of M1. All differences apart from aunt/uncle and niece/nephew are statistically significant. It is apparent from these two lists that F1 is at work in the parental generation, in which a higher value seems to be attributed to mothers than to fathers (maybe in relation to child rearing; cf. Allan 1987:70; Benor & Levy 2006:240). Conversely, the higher M1 rates in the offspring generation suggest an old patriarchal pattern, according to which sons are more important than daughters.
Judging from the data at hand, it can be claimed that mixed-gender binomials in English exhibit domain-specific patterns of syntactic markedness, that is, in some lexical subfields female-male is the unmarked sequence, and in others male-female is the unmarked order. Moreover, one finds that the coordinated constructions studied do not just construct people binarily as women and men. They also perpetuate harmful, difference-oriented discourses about the “appropriate” societal roles of the two sexes. It is unsurprising that these discourses echo gender images that now seem outdated (men dominating in general and in the areas of nobility, profession, heterosexual marriage; men as the preferred kind of offspring; women as notable when unmarried and as better parents). The following sections will test how far the gendered patterns identified may be modified by other factors.
Phonological Factor: Number of Syllables
One factor that may influence the ordering in binomials is phonology. Although a whole range of phonological features has been shown to have an effect on conjunct order (e.g., Gustafsson 1974; Cooper & Ross 1975:71; Benor & Levy 2006; or studies on coordinated personal names such as Wright & Hay 2002; Wright, Hay & Bent 2005), the present study will concentrate on the phonological feature that has been demonstrated to have the most robust effect on binomial ordering across studies, that is, the number of syllables in a conjunct (Cooper & Ross 1975:78-79; Pinker & Birdsong 1979:500; Fiedler 2007:42). More generally, it has been shown that metrical factors like syllable number play a stronger role in predicting conjunct order than other, nonmetrical phonological factors (cf. Mollin 2012:92-93). Across languages, a trend has been documented that shows a prevalence of shorter forms preceding longer forms in coordination (a rule that is also known as “Panini’s law” after the famous Sanskrit scholar or as Behaghel’s “law of increasing terms”; e.g., Sullivan, Casagrande & Belyayeva 1995). However, there is also one study in which no effect of phonology on binomial order could be verified (McDonald, Bock & Kelly 1993).
For the purposes of the present study, the gendered word pairs were divided up into three groups: one in which the female term contains more syllables (e.g., woman/man), one in which both terms contain an equal number of syllables (e.g., girl/boy), and one in which the male term contains more syllables (e.g., wife/husband). Some word pairs were excluded from this quantification, because syllable status was either unclear (e.g., Mrs/Mr) or variable between singular and plural. For example, in niece/nephew the male form has one syllable more; in nieces/nephews both forms have two syllables. The results are presented in Table 5.
Conjunct Order in Relation to Syllable Number.
One can find higher M1 rates in all three phonological subgroups. The highest value (90.6 percent) occurs when the female form is longer. When the female form is equally long, the M1 rate is reduced to 69.5 percent. When the male term has more syllables, the M1 rate is still at 66.1 percent. However, this latter figure is only due to one pair in the field, namely wife/husband. If this pair is ignored, the M1 rate drops to 17.6 percent, corresponding to an F1 rate of 82.4 percent. This shows that phonology, indeed, has an influence on binomial order in personal binomials.
Orthography
Another factor that may be thought to have an effect on binomial ordering is the spelling of the conjuncts. Such an effect has been documented for English in a study by Sullivan, Casagrande, and Belyayeva (1995). According to this study, one may expect that forms that come first in the alphabetical order regularly precede those that come later. This aspect is not as trivial as it appears at a glance. We have seen that phonology may have a modifying effect on binomial order even though written language data were used. An influence of the spelling would, therefore, be even more plausible for these data.
For the quantification, the word pairs were divided up into (1) pairs in which the female form starts with a letter that stands earlier in the alphabet and (2) pairs in which the male form starts with a letter that stands earlier in the alphabet. All word pairs that start with the same letter were excluded from this quantification. The results are presented in Table 6.
Conjunct Order in Relation to Orthography.
As can be seen in Table 6, the spelling of words does not seem to have any influence on the order of conjuncts in personal binomials. In both groups, the M1 rate is around 75 percent.
Morphological Complexity
Morphology has, so far, not received any attention as an explanatory factor for binomial order. This seems surprising and may be an outcome of a conflation of morphological complexity with word length, which has traditionally been tested in terms of syllable number. However, it is obvious that syllable number and morpheme number, in many cases, do not match. Another reason why morphology has not yet been tested may be that studies on binomials in general are less likely to encounter binomial pairs in which one conjunct is a derivation of the other. This phenomenon is much more likely to occur with personal nouns (e.g., host/-ess, widow/-er), with female nouns being particularly likely to be derived. The fact that, for example, the female suffix -ess can be attached to lexically male (prince/princess) and generic terms (author/authoress) shows that this suffix has two different meanings of personal relevance. With generic bases it adds the meaning [female] (an authoress is a female author), but with male bases it rather carries the meaning [female counterpart of] (a princess is not a female prince; see Bobaljik & Zocca 2011:152).
Table 7 provides the order frequencies for three morphological subgroups: (1) word pairs in which the female form is morphologically more complex, (2) word pairs in which both forms are morphologically equally complex, and (3) word pairs in which the male form is morphologically more complex. Some word pairs were excluded from this quantification, either because the number of morphemes is opaque (e.g., Mrs/Mr) or because differences in morpheme number were due to the dropping of repetitive elements in one conjunct (e.g., policewoman/-man, him-/herself).
Conjunct Order in Relation to Morphological Complexity.
The results for equally complex pairs and pairs in which the female form is morphologically more complex are similar to the results obtained for syllable number. The M1 rate is highest (90.7 percent) when the female form is more complex. When the morphological complexity is equal, the M1 rate still amounts to 73.4 percent. A different picture evolves when one looks at the pairs in which the male form is morphologically more complex. Here one finds a clear F1 predominance of 95.9 percent. This in turn means that the morphologically more complex form is generally found in second position, irrespective of gender. Compared to phonology, morphology turns out to have an effect that is potentially stronger than gender. Still two points need to be noted. First, the M1 predominance persists when the two forms are equally complex and is eradicated only when the male form is more complex. Second, there are only a few word pairs in which the male form is morphologically more complex—an outcome of the fact that female forms are generally more likely to be morphologically marked.
A concentration on pairs in which one conjunct is morphologically derived from the other does not lead to great changes in the patterns observed. For pairs involving a female derived form, M1 is categorical, whereas for pairs involving a male derived form the F1 predominance amounts to 90 percent. Moreover, it is of no consequence whether the coordination involves a semantically marked and a semantically unmarked item (actress/actor, hostess/host) or two semantically marked items without any generic meaning potential (princess/prince, duchess/duke, widow/widower, bride/bridegroom). The morphologically less complex form predominates in first position, independently of morphological relatedness or the semantics of the derivational base.
Conjunct Frequency
The frequency of the individual conjuncts can also potentially influence word order in binomials. More frequent words are regularly positioned first in coordination (e.g., Fenk-Oczlon 1989). To test the impact of this factor on personal binomials, the individual lexemes were first searched in the written component of the BNC to determine which of the two occurs more frequently. Certain word pairs were excluded from this quantification, either because one conjunct is not shown in its full form (e.g., policewoman/-man, her-/himself) or because the isolated search showed that a substantial share of the hits did not have the meaning that the form has in the binomial. For example, the pairs lady/gentleman and lady/lord had to be excluded because a search for lady alone does not distinguish between its two senses. Similarly, bride/groom was excluded because a search for the noun groom revealed that it was mostly used not in parallel to bride but in the sense of ‘person employed to look after horses.’ Mummy/daddy was excluded because the first item in this pair may also denote a preserved dead body. The remaining word pairs were then divided into two groups: (1) pairs in which the female form occurs more frequently and (2) pairs in which the male form is more frequent. Table 8 shows the results for the two groups.
Conjunct Order in Relation to Word Frequency.
The results indicate that word frequency, as an influencing factor, is not strong enough to override M1 predominance and rather has a modifying impact. However, it must be noted that the difference in percentage between the two conditions is relatively large. When the male form is more frequent, the M1 rate is 93 percent. By contrast, when the female form is more frequent, the M1 rate drops to 57.8 percent.
It may well be the case that this effect is stronger for pairs that show a greater difference in their conjunct frequencies. To test this aspect, an index score was calculated for each word pair in Table 8 by dividing the absolute frequency of the less frequent conjunct by that of the more frequent conjunct. For example, the absolute frequency of the lemma {sister} in the written component of the BNC is 8,272, while {brother} occurs 10,873 times. The index score for this pair is 0.76 (i.e., the frequency of {sister} amounts to 76 percent of the frequency of {brother}). Table 9 lists those pairs that have an index score of less than 0.50 (and therefore show a greater difference) .
Conjunct Order in Relation to Word Frequency Revisited: Pairs with Greater Differences in Conjunct Frequency.
The results show that such an effect can indeed be verified: the greater the difference in conjunct frequency, the higher the impact of frequency as a factor influencing conjunct order.
Sex of Author
Studies on the coordination of personal names (Wright & Hay 2002; Wright, Hay & Bent 2005; Hegarty et al. 2011) have found that speaker sex exerts an influence on binomial order and that, for example, male speakers are more likely than female speakers to place men first in such constructions. A similar own-sex effect has been described by McGuire and McGuire (1992:232) in a study on binomials involving kinship terms. The written BNC component certainly does not provide the best kind of data to test for such an effect because author sex is known for only a relatively small portion of the corpus (53 % Hoffmann et al. 2008:31). Moreover, the quantification of linguistic features for two gendered macro-groups in large corpora must be met with some reservation, because such a procedure essentializes women and men and does not attend to the substantial intracategorial heterogeneity that the two groups exhibit. Nevertheless, such an analysis was conducted in order to see whether the trends of earlier studies can be replicated here. For this purpose, only the most frequent word pairs with M1 or F1 predominance were tested: woman/man and mother/father. The results can be found in Tables 10 and 11.
Order of the Conjuncts {woman} and {man} in Relation to Sex of Author.
Order of the Conjuncts {mother} and {father} in Relation to Sex of Author.
The results show that there is indeed such an effect, but its impact is relatively small. As expected, the M1 rate is highest for male authors (96.5 percent), closely followed by mixed-sex authors (92.8 percent). Even though the M1 rate is smallest with female authors, it is still remarkably high under this condition (81.1 percent).
The same procedure was also applied to the word pair mother/father, which exhibits a predominance of F1. Here it is interesting to note that the greatest F1 bias is found not with female authors (68.6 percent), but with mixed-sex authors (86.2 percent). As expected, male authors show the lowest F1 rate (51.1 percent). This rate is in fact so low that there is no F1 predominance in this condition.
The sex of the author can be considered only a moderate factor influencing binomial order. However, the two quantified word pairs demonstrate that male authors exhibit the most extreme patterns: they show the highest M1 rate for woman/man and do not adhere to the common F1 predominance in mother/father. If initial position in personal coordination is taken as an index of power differentials, this linguistic behavior indicates that men tend to follow their own interests in how they construct the world: they keep up their dominant position and do not grant it to women, even in domains that are stereotypically female-dominated. Women, on the other hand, seem to make only limited use of the linguistic means that place them in the center, which can be seen as evidence for the fact that their linguistic agency is restricted by normative forces. Of course, such patterns are very general and may therefore possess only a limited value for the description of local linguistic practices, which may depart considerably from these.
Target Audience Sex
A factor that has, so far, not been tested in research on binomial order is target audience sex. As the mark-up of the written BNC component allows for testing this option (77 percent of the corpus is annotated for target audience sex; Hoffmann et al. 2008:31), it was decided to perform two additional queries for woman/man and mother/father in relation to target audience sex (see Tables 12 and 13). Again it must be pointed out that such a procedure is not unproblematic, in this case because the coding procedure for target audience sex in the written BNC is vague. The BNC reference guide does not provide information on how the categories “male target audience” or “female target audience” were assigned. A look at the corpus, however, indicates that this classification has been carried out along highly stereotypical lines. For example, in the periodical section, male target audience was assigned to articles from magazines such as The Guitarist or Autocar and Motor, that is, it was assumed that titles pertaining to these socially gendered domains are not read by women. Similarly, among the texts classified as female-targeted one finds book titles such as Charles and Diana (apparently a biography of the royal couple) or The Art of Starvation.
Order of the Conjuncts {woman} and {man} in Relation to Target Audience Sex.
Order of the Conjuncts {mother} and {father} in Relation to Target Audience Sex.
The results for woman/man in relation to target audience sex are similar to those for author sex. The M1 rate is highest for the male target audience (97.4 percent) and smallest for the female target audience (84.6 percent). Taken together, there is no sign that the sex of the target audience can erase M1 predominance.
For the pair mother/father, the F1 rate is the highest for the female target audience (81 percent). Interestingly, the F1 rate for the male target audience is slightly higher (64.3 percent) than those for mixed-sex (58.8 percent) and unknown target groups (60 percent). However, the F1 predominance for male targets and unknown targets is not statistically significant. In other words, the significantly higher F1 rates are found only where it is clear that women form (a part of) the audience.
Conclusion
This study of conjunct order in mixed-gender binomials has shown that personal binomials are partly affected by different factors than nonpersonal ones. While earlier studies on binomials have highlighted the role of lexical gender in conjunct order and noted a general tendency for M1, the present study has refined this description by relating it to specific lexical subfields of personal nouns. It turns out that the M1 predominance is far from absolute and exhibits varying degrees of intensity. Moreover, certain domains that are more likely to exhibit an F1 dominance were identified. Traditional factors such as phonology (syllable number) and conjunct frequency proved to be relevant for conjunct order in personal binomials as modifiers of F1 or M1 predominance. Alphabetical order, by contrast, had no modifying effect. Other influential factors that have not, or only sporadically, been tested in previous studies are morphological complexity, sex of author, and target audience sex. The finding that lexical gender is, in most cases, stronger than the other factors replicates a common finding of earlier research, namely that (conceptually more salient) semantic factors prevail over other factors as far as conjunct order is concerned (cf. Cooper & Ross 1975; McDonald, Bock & Kelly 1993). Still it is noteworthy that another study on English, French, and Russian binomials by Sullivan, Casagrande, and Belyayeva (1995) found that gender was the weakest factor tested (after syllable number and alphabetical order). This, however, can be explained in terms of the data used, as that study did not concentrate on personal binomials. As the present study shows, personal binomials need special treatment and should not be lumped together with other binomial types.
It is also worthwhile pointing out the limitations of the present study. Although the thirty most frequent mixed-gender binomial types were incorporated in this study, future research should aim at a higher comprehensiveness by also including less frequent pairings. Moreover, it may prove fruitful to use methodologies that allow for a more precise weighting of the factors influencing conjunct order in personal binomials. Comparisons across English varieties, languages, and time periods would also be desirable. For example, on a language typological level, future research could address the question of whether the patterns of personal binomial order detected for English also hold for other languages and cultures. Kikvidze (2001) has shown that in Georgian dvanda compounds, which are similar to the coordinative binomial structures studied here, female forms generally (and irreversibly) precede male forms. Basque co-compounds, on the other hand, are more likely to show M1 (see Motschenbacher 2010:69-70). This points to a cross-linguistic variance of gendered ordering principles that may be interesting to explore.
Mixed-gender binomials are located at the interface of dominance and difference with respect to gender construction. Whereas difference theorists view the specification of both women and men as positive, namely as a strategy for overcoming female invisibility, dominance theorists have taken a more critical stance on mixed-gender binomials, namely by highlighting the role they play in the construction of gendered power asymmetries. In terms of supplementing the difference view on gender with a more critical perspective, the present study has shown that the specification of both women and men in connection with the order principles detected often conveys harmful discourses about normative positions and roles of women and men in society. For example, it was pointed out that the binary gender constructions of mixed-gender binomials contribute to the materialization of domain-specific female (and, to a lesser extent, male) subordination.
It is not the aim of this article to make suggestions for language reform. One reason for this is that judgments of what has to be replaced cannot operate on a general level and have to take note of the interactional context and the linguistic forms involved (for example, for some pairs such as aunt/uncle or niece/nephew, nonbinary or lexically gender-neutral alternatives are hard to find). Different attitudes would result in different kinds of recommendations, and it is unfeasible to claim that any of them is, per se, less political or unbiased (see also Motschenbacher, forthcoming).
However, there is some value in raising awareness of the fact that all linguistic choices we make, even if traditional, are not neutral, and that language users need to make up their minds as to which messages they want to convey (Cameron 1992:125-127). The aim cannot be to change the language system by setting up (another system of) normative rules. It seems more reasonable to empower individual speakers as agents in language use and change by pointing out that their usage contributes to the ongoing competition of discourses—and this competition may indeed shift in certain directions if more speakers choose to adhere to less traditional usage patterns.
The conceptualization of the relationship between language and gender as described in the preceding paragraph is well compatible with more recent developments in the field of language and gender. Later work inspired by linguistic anthropology, Foucauldian and Derridean poststructuralist thinking, and Butler’s (1990) performative gender concept has generally adopted social constructionist or discursive stances on the relationship between language and gender. A central aspect of this line of research is exploring ways of going “beyond binary thinking” (Bing & Bergvall 1996)—a kind of reasoning that is today epitomized by Queer Linguistics, which emerged as a more coherent field of study in the 2000s (e.g., Barrett 2002; Motschenbacher 2010, 2011). Queer Linguistics takes a critical view of the role language plays in the formation of the dominant discourses of heteronormativity and gender binarism. Following Butler (1990), these two mechanisms are conceptualized as mutually reinforcing, that is, dividing humanity up into two macro-groups women and men is seen as strengthening a perception of heterosexuality as the norm and reinforces male-female power differentials. Concomitant with this effect is a highlighting of intergender differences and a downplaying of intragender differences.
The poststructuralist impetus on this more recent work can further be recognized in the fact that it tends to view what has traditionally been considered as the language system as the result of discursive materialization processes—a conceptualization that is compatible with usage-based theorizations of language structures as emerging in linguistic performances (e.g., Hopper 1998). Accordingly, mixed-gender binomials constitute a structure type in which gendered power differentials, gender binarism, and gender difference have become firmly materialized. As a consequence, they cannot be interpreted as natural but rather as discursively and, more specifically, linguistically mediated social constructions that may be criticized and changed (Cameron 1992:89). Even though individual language users cannot be said to have initiated such discourses (because as ideological formations they work over-individually; see Black & Coward 1981/1998), they are certainly not helpless victims of these dominant discourses. In actual linguistic performances, they may choose to use forms that revolt against such discourses or at least do not further entrench them.
A positive aspect about poststructuralist theorizations of language and gender is that they are based on a notion of ongoing linguistic change via concrete linguistic performances. This idea appears to be particularly relevant for gendered personal binomials. Recent research on the historical development of binomial reversibility shows that this type of binomial is particularly likely to be subject to changes in reversibility status (Mollin 2013:195-198). Even though a complete reversal from M1 to F1, as in the case of ladies and gentlemen, forms the clear exception, it is evident that most mixed-gender binomials with a traditional M1 preponderance have shown a development since the 1950s that Mollin calls “unfreezing,” that is, their M1 predominance is slowly becoming less absolute (Mollin 2013:196). This aspect runs counter to the overall development in binomials, for which freezing (i.e., stabilizing the exhibited order preference) is the more common tendency (see Mollin 2013:189). The few binomials with F1 preponderance (lady/gentleman, mom/dad, mother/father), by contrast, show exactly such a freezing trend, that is, their F1 tendency is becoming more pronounced (Mollin 2013:195-196). While the unfreezing associated with M1 preponderance binomials may be considered a positive development, it is evident that contemporary usage patterns are still far away from gender parity. The freezing of F1 preponderance binomials can also not automatically be evaluated as positive, because it entrenches domain-specific female stereotypes even further and causes a shift away from gender parity. The asymmetries of the ordering of conjuncts in personal binomials demonstrate that binary gender specification is almost invariably connected to gender inequality. From this point of view, it is legitimate to question the usefulness of gender-binary specification in many contexts where it is practiced in a relatively unmotivated fashion (for example, when addressing an audience as ladies and gentlemen).
Footnotes
Acknowledgements
I would like to thank the editors of the Journal of English Linguistics, Matthew Gordon and Peter Grund, and two anonymous reviewers for their constructive comments on earlier versions of this article. All remaining errors and insufficiencies are, of course, my own.
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.
Notes
Author Biography
.
