Women and men in language

Ann Weatherall;

Abstract:

Empirical and theoretical research on gender and language has demonstrated how issues of power and identity are communicated by the way in which women and men are referred to in language. To broaden and update past work on sex bias in language, the present study investigated references to men and women in prompted but impromptu discussions of a popular television program.

Empirical and theoretical research on gender and language has demonstrated how issues of power and identity are communicated by the way in which women and men are referred to in language. To broaden and update past work on sex bias in language, the present study investigated references to men and women in prompted but impromptu discussions of a popular television program. The content of the descriptions was examined for evidence of sex bias. There was some indication that the form and content of descriptions varied depending on the sex of the referent. However, the language used was not as overtly or pervasively sexist as had been expected. It is suggested that an important task is to document how linguistic bias is routinized or legitimated in different conversational contexts. In that way, language use can lead to a better understanding of the creation and maintenance of power hierarchies in social groups.

The study of language and communication has contributed significantly to understanding how gender relations are reflected, created, and sustained through everyday interaction (see Kalbfleisch & Cody, 1995; Kramarae, 1990, for recent overviews). In a now classic work, Lakoff (1973) suggested that women's position of low power compared to that of men was reflected both in the ways that women talk and the way that they are talked about. A
considerable literature has developed that considers how women's style of communication can be understood in terms of power (Lakoff,l990) and social identity (Maltz & Borker, 1982). However, there has been less empirical investigation and theoretical development on how power and identity are communicated through the way women and men are referred to (Crawford, 1995).

Language About Women and Men

The ways in which the English language differentially encodes information about the sexes has attracted considerable comment (Henley, Miller, & Beazley, 1995; Martyna, 1980; Miller & Swift, 1976; Poynton, 1989; Spender, 1980). Many forms of sexist language have been identified, but Henley (1987) suggests that sexism in language may be reduced to three types-language that ignores women, language that defines women narrowly, and language that depreciates women.

Women may be ignored in language simply by not being the topics of discourse (Graham, 1975). However, it has been the use of masculine generics for referring to sex-indefinite referents that has been used as the primary example of how language conceals women and girls (Silveria, 1980). The narrow definition of women in language relates to the observation that women are more often discussed in terms of their relationships, whereas men are more often discussed in terms of what they do (Key, 1975). The power to define women through
language is seen in the tradition of women losing their names through marriage. Scholars such as Allport (1 963) emphasize the link between name and identity-hence, the cultural practice of name changing has' been criticized for obscuring women's individuality (Stannard, 1977). The secondary social status of women may be marked additionally by the personal name form used in everyday interaction. Men's more common use of women's first names or other familiar forms of address can be interpreted as reflecting men's general lack of respect
for women (Lakoff, 1973; Poynton, 1989).

Language that depreciates women includes all the unfavorable terms that refer to them. In a study of female and male terms listed in dictionaries, Stanley (1 977) found many more negative, sexual, and trivializing terms to describe women than to describe men. Words that are parallel in meaning (e.g., masterhnistress, bachelorhpinster) often cast women in a less favorable light (Lakoff, 1973; Schulz, 1975). Even once-neutral words may become
derogatory when applied to women (Miller & Swift, 1976; Ng, Chan, Weatherall, & Moody, 1993). Thus, the connotations of words associated with the female linguistic category tend to be downgraded in meaning compared to words associated with the male linguistic category.

Consequences of Linguistic Sexism

Scholars of language and communication generally agree which linguistic forms constitute sex bias. There is less agreement about how significant sexist language is for thought and social interaction. Lakoff (1 973) suggested that linguistic biases are worthy of study because they reflect real-world inequities. Theoretically, she was opposed to the view that sexist language forms could influence people's perceptions of women and argued that linguistic change follows social change and not vice versa.

Communication researchers influenced by a Whorfian view of language advocate language change as a strategy for social change. Strong empirical support for the influence of language on thought has been provided by experimental studies on the impact of masculine generic language. For example, a proactive inhibition procedure was used by Ng (1990a) to show that masculine generic forms are coded in the masculine but not the feminine linguistic categories. The pervasive androcentric bias of masculine generic language demonstrated by numerous studies has lead to the conclusion that the use of such forms functions to conceal the existence and importance of women as a social group (Martyna, 1980; Moulton, Robinson, & Elias, 1978; Wilson & Ng, 1988). The experimental evidence has been used to support policies encouraging the use of nonsexist language (e.g., American Psychological Association, 1975; British Psychological Society, 199 1 ; State Services Commission, 1988).

Recent theoretical developments on language, power, and social relationships suggest more subtle and complex associations between language and social behavior than those initially used to describe the significance of sex bias in language. For example, discourse analysts explore how the production and organization of social categories in conversation are influenced by the context and function of the particular interaction (Potter & Wetherell, 1987). Using sexist language as the main reference, Ng (1 990b) outlined how language facilitates the routinization of control. He argued that when sexist language is accepted as part of everyday interaction, then the subordination of women becomes routinized in daily discourse. The routine use of forms that conceal or trivialize women reduces the salience and significance of women as a social category. The routinization of sex bias in language makes
sexism more covert and hence a legitimate way of communicating discrimination (van Dijk, 1987). Thus, articulation of social categories in communicative contexts provides an important way of understanding the creation and maintenance of power hierarchies in social groups.

Sexist Language in Everyday Communicative Contexts

Previous research has documented various forms of sexist language and provided evidence of the negative cognitive consequences of linguistic sexism. However, much of that work has used data from sources such as books (e.g., Nilsen, 1977) and newspapers (e.g., Henley et al., 1995) that may emphasize stereotypical portrayals of men and women. Such investigations have diverted attention away from documenting the ways in which male and female linguistic categories are spontaneously expressed in conversation. The lack of systematic investigation into the way language is used spontaneously to categorize people on the basis of sex seems a serious omission. One of the aims of the present research was to examine claims about sexism in language by looking at how men and women were referred to in impromptu language use. Hence, the present study complements the existing effects research by substantiating the forms of linguistic sex bias used in a conversational context. Understanding the way that gender categories are articulated in conversation should enhance understandings of how social biases are communicated in social interaction.

Ideally, an investigation into the everyday use of sexist language forms would examine a range of naturally occurring language. A representative sample of written and spoken forms recorded in a variety of different contexts would be desirable. In this way, one could be more confident of the validity and generalizability of any patterns of language use found. Although preferable, the use of naturally occurring language is not problem free. Involved with spoken language are ethical issues concerning the making of surreptitious recordings. A further
difficulty with studying natural language use is that the linguistic forms of interest may not occur frequently enough to be subjected to quantitative analysis.

Fictional contexts have been used in the past to overcome various difficulties associated with collecting and analyzing actual language use. Kramer (1 979, for instance, investigated gender bias in systems of address using novels and short stories written by prize-winning American authors. In a similar vein, Weatherall (1996) examined sexist language use in a British soap opera. The assumption that such research makes is that the language used in literary and popular texts has some validity because the authors reproduce, with some accuracy, actual language norms. However, the use of fictional works may still emphasize stereotypical portrayals of gender; thus, there is a need to examine spontaneous language. To build on research that has examined language use in preexisting texts, the research reported here recorded and analyzed prompted but impromptu spoken descriptions of characters from a popular television soap opera.

The decision to use prompted descriptions of soap opera characters as data was influenced by mass communication literature. Research on the social function of soap opera has shown that talk about such shows is a spontaneous, enjoyable activity that serves to verify interpretations of the program and to solidify social relationships (Morley, 1986; Wober & Reardon, 1989). Thus, asking viewers to talk about the program was mimicking an everyday, enjoyable activity. The particular soap opera, Coronation Street, was chosen as a stimulus
topic because it has been the most consistently popular television program in Britain since it was first broadcast in 1960. Approximately 17 million British people watch each episode, and it has appeared in the top 20 ratings for nearly 30 years (Wober, 1992). Finally, a content analysis of the program found virtually no evidence of a pervasive linguistic bias against the female characters (Weatherall, 1996). Hence, any sexism in the participants' descriptions would be an indication of a bias in the language norms available for describing men and women.

The aim of the study was to explore a conversational corpus for evidence of how sex bias is
articulated in spontaneous descriptions of people. The primary research question was:

RQ 1 : Can the language used to describe women and men in seminaturalistic conversations be understood as being biased against women?

The three types of sexist language identified by Henley (1 987) were used to guide the analysis. The issue of whether women get ignored by simply not being topics of discourse is not relevant for the present study. Participants were shown both male and female stimulus prompts, so it was expected that descriptions would be rich in both female and male gender reference terms. The use of masculine generic terms will be of interest in so far as they will provide some indication about norms of their use.

Personal names are an interesting reference form in that they have been implicated in issues related to both identity and power. The use of titles (Mrs. /Miss) and name changing upon marriage have both been criticized by feminists as cultural naming practices that serve to conceal women's independent identity (Spender, 1980). The choice of name form (e.g., first name vs. full name or title and last name) can be used to infer information about perceived
status or solidarity in social relationships (e.g., Brown & Ford, 1961; Wolfson & Manes, 1981). The potential that names have to mark status differences between the sexes led to the following hypothesis.

HI : Male participants will use more first-name-only forms when describing female characters than when describing male characters and more full-name forms when describing male characters than when describing female characters.

The issue of whether language defines women in relationship terms more than it does men in the conversations was operationalized by focusing on the use of kinship terms, job titles, and references to others in relation to the character prompts. Hence, the following set of hypotheses were postulated.

H2a: Female characters will be referred to more often using kin terms than will male characters.

H2b: Descriptions of female characters will include more references to others than will descriptions of male characters.

H2c: Job titles will be used more often in descriptions of male characters than in descriptions of female characters.

Finally, the way language demeans women was examined by looking at other gender reference terms. It was expected that:

H3: There will be more positive lexical items used to refer to male characters than to refer to female characters.

METHOD

Participant Selection

A television-viewing behavior questionnaire was distributed to students enrolled in a first-year Culture and Communication course at a university in England. Students were asked to rate their viewing behavior of Coronation Street in one of the following categories: never, less than once a month, about once a fortnight, 1 to 2 times per week, 2 to 3 times per week, 3 to 4 times per week, or other.

The questionnaire was completed by 146 students; 59 males and 87 females. A chi-square analysis showed that there was no difference between men's and women's responses about viewing habits, x2(5, N = 146 ) = 3.9, ns. A total of 80 students, all English and between 18 and 21 years old, participated in the main study with the understanding that they would be required to describe characters in a popular television program.

Design

A median-split method was used to classifL viewing behavior for design purposes. Responses of never and less than once a month were used to classify individuals as irregular Coronation Street viewers (about half of the respondents were thus classified), and once a fortnight or more responses were used to classify individuals as regular viewers.

A total of 32 regular and 32 irregular viewers participated in the conversations, with equal numbers of males and females in each viewing category. A male or female regular viewer was paired with a male or female irregular viewer, resulting in eight of each of four dyad types: male regular and male irregular, male regular and female irregular, female regular and male irregular, and female regular and female irregular. The dyads were constructed in this way for two reasons: to promote detailed discussions of the program and to control for any
sex differences in the use of personal references.

Materials

A set of six images was used as stimuli. Included in this set were pictures of all the characters appearing in a single episode of Coronation Street (broadcast on April 22,199 1). The images included the following female characters: Deirdre Barlow, Mavis Wilton, Stephanie Barnes, Phyllis Pearce, Liz MacDonald, Rita Fairclough, Audrey Roberts, and Mrs. Shaw (temporary character). The male characters featured were Phil Jennings, Des Barnes,
Andy MacDonald, Jim MacDonald, Steve MacDonald, Martin Platt, and Alf Roberts. An equal number of regular female and male characters appeared in the images. The pictures were taken from stills of the program so that characters appeared together as they did in the show itself.

Procedure

Each dyad met in a room fitted with video cameras, and the participants were given a folder containing the stills taken from the Coronation Street episode. The pictures were organized in a random order that was different for each dyad. The dyads were instructed to talk about each of the pictures in as much detail as possible. The resulting interactions were videorecorded and the conversations later transcribed.

After finishing their descriptions, participants were told that the purpose of the research was to investigate the everyday use of sexist language. Feedback from the participants about the task was positive.

Stages of Analysis

The analysis began with a general summary of the content of the descriptions, including the number of words and the frequency of personal reference terms used.1 However, the major focus was on how characters were named, labeled, and described by viewers. During the course of the task, viewers not only referred to the characters depicted in the set of pictures, as they had been instructed, but also made reference to other characters in Coronation Street as well as references to characters from other programs. Terms and phrases used to reference the depicted characters can be understood as being prompted directly, whereas others were not. Prompted references were the focus of the following analysis. Results that report other labels and descriptors include all personal references that were made.

Two coders counted and categorized the data. Each reference to an individual character was coded into one of the following categories: first name only, full name, kinship term (e.g., mother, father), job title, and other. Kappa was used to calculate the reliability between the two coders (Cohen, 1960). The resulting reliability indicated good agreement in the way references were identified and coded (K = .89).

A preliminary analysis found that reference rate was not influenced by sex of participants or sex composition of the dyad.

RESULTS

The spoken descriptions took between 20 and 60 minutes to complete. A total of 62,692 words were transcribed from the spoken dialogues. Of the words and phrases uttered by the regular viewers (n = 48,894), 6% (n = 2,811) were coded as personal references. Excluding pronouns, there were more coded references to female (n = 750) than to male (n = 468) prompts. The most common reference types in the spoken descriptions (excluding pronouns) were: first name only (44%), kinship term (17%), full name (13%), and job title (So/a). A total of 18% of reference terms were coded in an ''other'' category.

Personal Names

A clear indicator of the importance of personal names was that they were the most frequent forms of personal reference (57%). Viewers clearly believed that recalling a character's name was the most important way of identifling a character, despite the fact that the task was not specifically to name the prompts. Naming the characters or admitting to not knowing the names was a common way of starting the conversations. The following two examples are typical of how conversations began.

Female viewer: Oh right, this is Deirdre Barlow and that is Phil Jennings. Male viewer: Right, that's Rita Fairclough.

An admission of not knowing a name could even be considered grounds for an apology. For example:

Male viewer: What's her name? Female viewer: Mrs. Roberts, but I can't remember her first name, sorry umm, I do know it, I just can't think of it at the moment.

The cultural practice whereby a woman changes her name upon marriage is closely adhered to in Coronation Street. An interesting phenomenon was that women's changes in surnames seemed to cause some confusion for viewers when recalling names. For example:

Male viewer: I can't remember her surname now, it was Riley before. . . the husband, that's it, it's Derek Wilton, so her name is Mavis Wilton now. Male viewer: Don, second husband, that's right, she's not called Ivy Tilsely any more, she's called Ivy Brennan because her husband is called Don Brennan.

The female title "Mrs." was used as a substitute if a female character's first name could not be recalled. Examples of this included:

Male viewer: Mrs. Duckworth, who's married to Jack Duckworth. Female viewer: Alf Robert's shop and that's his wife . . . Mrs. Roberts, I can't remember her first name.

In contrast, when a male name could not be recalled, the use of a surname alone appeared to be sufficient for identifying the character.

Male viewer: His name was something Fairclough, I can't remember his first name.

Female viewer: And that is Platt, what's his first name, though?

Overall, 47% (n = 356) of references to female prompts used first names only, and 10% (n = 78) used full names. Of all references made to male prompts, 40% (n = 188) were first names only and 16% (n = 76) were full names. To test the first hypothesis, a proportion of first and full name to all other references was calculated for each character. The proportions were averaged separately for the female and male characters, respectively, to control for the difference in the total number of female and male references.

The influence of sex of regular viewer, sex of irregular viewer, and sex of reference on the name type used (first, full) by the regular viewer was analyzed using a single four-way mixed ANOVA. There were three significant effects. First, there was a main effect due to sex of the character referenced. A greater proportion of name references were to female (M = 0.29) rather than to male prompts (M = 0.22), F(l, 28) = 18.78, MSE = 0.009, p < .01. There was also a main effect due to name. A greater proportion of references to the prompts used first
names only (M = 0.37) rather than full names (M = 0.13), F(l, 28) = 65.92, MSE = 0.028, p < .O 1. The two main effects were modified by a significant interaction between sex of reference and reference type. There was a larger difference between the use of female and male first names as opposed to full names, F(l, 28) = 12.25, MSE = 0.18, p < .01. A Tukey's post hoc test indicated that the difference between the use of female first (M = 0.45) and full names (M = 0.13) and the difference between male first (M = 0.30) and hll names (M = 0.14) were both significant. In addition, the difference between the use of female and the use of male first names was significant. However, there was no difference found between the use of female and the use of male fill names.

Hence, support was not found for HI. Sex of speaker had no effect on the form of name used, and male prompts were not referred to more often than were female prompts with full names.

Roles and Relationships

The most commonly used terms in the kinship category were: wife, husband, daughter, son, mother, and father. There was a greater lexical diversity in the female (n = 19) as opposed to the male (n = 14) kinship categories. Kinship terms accounted for 15% (n = 110) and 19% (n = 88) of references made by viewers to female and male prompts, respectively.

To test for differences in the use of male or female kinship terms (H2a), the influence of sex of regular viewer, sex of irregular viewer, and kinship term used (male, female) on the proportion of kin reference terms used was tested using a single, three-way mixed ANOVA. The analysis of kinship references showed one significant main effect. On average, a greater proportion of kin terms were used to reference male (M = 0.20) rather than female (M = 0.13) prompts, F( 1, 28) = 6.40, MSE = 0.01 1, p < .05. Note that the opposite pattern of results was postulated by H2a. Here, male prompts were described more often in terms of their family roles than were female prompts.

The next analysis examined the number of other characters talked about in response to each of the character prompts (H2b). Rita was talked about in relation to the most distinct others. Her relationships with eight other female and six other male characters were mentioned in response to her image. The following is an example of how Rita was talked about; in this extract, she is talked about in relation to four men and one woman.

Female viewer: Rita Fairclough, she was managed by Alec, who now runs the pub and is married to Bet, and she's also been married to Len Fairclough, who dies in a car crash, and she's being pursued by Reg Holdsworth, the manager over at SuperBuy, and she went out with this guy called Allen who turned out to be a nasty piece of work.

For each character prompt, the average number of other characters talked about in relation to them was calculated. A four-way mixed ANOVA was used to test whether there was a relationship between the sex of regular viewer, sex of irregular viewer, sex of character prompt, and sex of associated character on the average number of others linked to the prompt. There were two significant effects. There was a main effect due to sex of prompt. Viewers talked about a greater number of associations in response to a female prompt (M = 3.92) rather than a male prompt (M = 3.10), F(1, 44) = 4.10, MSE = 4.25, p < .OS). Consistent with H2b, viewers did tend to respond to female characters by talking about other people.

The main effect was qualified by a significant interaction between sex of character prompt and sex of associated character. Viewers tended to respond to female characters by talking more about male (M = 4.43) than about female characters (M = 3.42), F(1, 44) = 6.50, MSE = 1.52, p < .OS).

The final part of the analysis, examining the issue of whether language defines women in relationship terms more than it does men, focused on the use of job titles. A total of 8% of references made by regular viewers to male (n = 37) and female (n = 62) character prompts were job titles. If references to character prompts and other characters are included, a greater diversity of job titles was used with reference to males (n = 38) than with reference to females (n = 19). Of the job titles used to reference female characters, six were marked linguistically as feminine. These were: actress, barmaid, cleaning lady, landlady, papergirl, and woman solicitor. The terms actor and cleaner were used with reference to female characters, indicating that neutral forms are in everyday use. A total of 10 job titles would be classified conventionally as masculine generics (e.g., salesman, businessman). All of the masculine generic terms were used to reference male characters, suggesting that those terms were being used in a sex-specific rather than a sex-inclusive sense.

Differences in the use of job titles to refer to character prompts could not be tested statistically because of the small proportion of such references to each character. However, the greater diversity of jobs provides an indication that male characters were described more in relation to what they did than were female characters.

Labels and Descriptors

A total of 19% (n = 143) of female prompted references and 17% (n = 79) of male prompted references were coded in the "other" category. This category was examined in relation to H3, concerning the number of positive and negative lexical items used to describe the character prompts. To explore more closely the types of words coded as "other," the terms were further subdivided into six subcategories. Table 1 shows the lexical diversity in each of the subcategories with examples of the terms coded in each. The most striking results were: there were more animal references made to female than to male characters, there was a similar number of sexual references to both female and male characters, and there was a greater lexical diversity in the male as opposed to the female ''other" descriptors and gender noun categories. However, it is interesting to note that many of the terms coded in the male "other" category had negative connotations.

Generic references formed a large category of other labels and descriptors. Overall, the words coded in the other labels and descriptors category provided no evidence for H3. There was a rich diversity of negative terms to refer to the male characters as well as to the female characters.

DISCUSSION

The prompted but impromptu descriptions of a popular television show proved a rich source of data on the spontaneous language used to refer to people. The viewers' accounts of the program contained a wide variety of personal reference forms, most of which were gender specific. The sex of the characters appearing in the show could be seen from the stimulus pictures yet, despite this, characters rarely were described as persons, lovers, parents, children, or soap stars, but rather as women, men, fathers, mothers, sons, barmaids, and tradesmen. Could the pervasive identification of the characters as male or female be understood as coding and communicating a bias against women? There were some indications that the form and content of descriptions varied depending on the sex of the referent. However, the language used was not as overtly or pervasively sexist as had been expected. There are at least two reasons why the conversational data contained fewer examples of sexism than was anticipated. As was mentioned in the beginning of the article, previous research on sex bias in language largely relied on language sources that emphasize stereotypical portrayals of gender. Hence, predictions of sex bias in language based on such work may be inflated. A second explanation is that in everyday conversation, there are social norms against uninhibited discriminatory talk (van Dijk, 1987). Therefore, blatantly stereotyped and negative views about women were not articulated. However, some subtle evidence of sexism was found in qualitative examples as well as in the quantitative analysis.

Personal names were the most common linguistic form used for identiQing characters. The close link between personal name and identity (Allport, 1963) and address form and power (Brown & Ford, 1961) combined with the frequency with which they were used meant they were a linguistic form that had potential for communicating sexism. The cultural practices of women changing their names and female titles (Mrs./Miss) to mark marital status have been criticized by feminist scholars for obscuring women's independent identities. There was some evidence that those practices were functioning in that way. If the viewers were having difficulty recalling a female character's name, they would refer to her husband (e.g., "I can't remember her name now, it was Riley before. . . the husband, that's it, it's Derek Wilton, so her name is Mavis Wilton") or merely use "Mrs." (e.g., "Mrs. Duckworth,who's married to Jack Duckworth"). These examples on their own seem trivial, but they are an example of routine linguistic practices that define a woman in terms of her relationship with a man. Thus, according to Ng's (1990b) routinization of control model, they function to reduce the salience of a woman's individual identity. In addition, the form of the personal name that a speaker chooses can function as a socially acceptable strategy for expressing a negative opinion about someone (van Dijk, 1987).

A quantitative analysis tested whether there was any evidence of viewers using name form (first vs. full name or title and last name) to mark status differences between the sexes. In contrast to what was predicted by the first hypothesis, there was no evidence that names were used to mark a general lack of respect for women; full-name forms were not used more frequently to refer to male characters than to female characters. However, differences
were found in the use of names. First names were used more frequently than were full names, and female first names were used more frequently than were male first names. The result can be understood best as a direct consequence of the topic for discussion. Defining characteristics of soap opera are that female characters play strong and central roles, and viewers form an intimate knowledge about the characters' lives (Morley 1986). The centrality of the female characters in the program and viewers' familiarity with them were evident in
their descriptions. Examples included:

Male viewer A: That's Rita in the corner shop. Male viewer B: What sort of shop?

Male viewer A: Just a newsagents, she runs it with, you've heard of Mavis?2 Male viewer B: Yeah.

Male viewer A: Everybody's heard of Mavis, she runs it with Mavis. Female viewer A: Who are her friends?

Female viewer B: Mavis. Female viewer A: I've heard of Mavis. Female viewer B: Everyone has.

Although no evidence was found in the data analyzed here of a differential use of personal name forms to refer to men and women, other studies using different data sets have found a bias. For example, Wolfson and Manes (198 1) reported that in service encounters, women were more likely to be addressed using familiar forms than were men. They suggested that women may find familiar forms annoying because the use of them suggests that women are not worthy of respect. The potential that names have as markers of power differential is undeniable (seeBrown & Ford, 1961). However, what is not clear are the contexts in which that potential is realized. One difference between the present study and Wolfson and Manes' (1981) research was that here, names were used to refer to absent third persons, whereas Wolfson and Manes examined face-toface address forms. An important avenue for future work may be to identify the characteristics of contexts in which names may be used to mark power differences. Such work could be useful in establishing areas in which language reform needs to be promoted.

The name form used to identify the female characters was one way in which they tended to be talked about in terms of their relationships. The content of the descriptions also was examined for evidence that women were discussed more in terms of their family and interpersonal roles, whereas men were described more in terms of what they do. This was examined by investigating the use of kinship terms and the use of job titles as well as by
quantifying the number of different individuals discussed in relation to each of the prompts. In direct contrast to what was expected, male kinship terms were used more frequently than female kinship terms. More consistent with the claim that women are more often described in terms of others was the finding that viewers tended to respond to female character prompts by describing others, specifically other men. These contradictory findings suggest that sexism in everyday interactive conversations may not be as straightforward as some of the
literature suggests (e.g., Spender, 1980). Rather, the way the bias is communicated varies according to the function of the interaction being examined and the context of the conversation. For example, it would be expected that linguistic bias would be manifested differently in conversations between a group of men drinking in a pub and in sports commentators reporting the performance of athletes. However, even in those two contexts, sex bias may be carefully presented, perhaps in the form of jokes in the pub or as a signal of solidarity to the "girls"' teams.

The lexical level of the descriptions was also examined for evidence of sex bias. A greater lexical diversity was evident in the female kinship term and the male job title categories. The lexical resources used by the participants in the study can be understood as promoting a view that assumes the centrality of family-to-female roles and work-to-male roles. Further evidence of bias was found when looking at the forms of job titles. The use of adjuncts (e.g., woman doctor) and suffixes (e.g., actress) were evident in descriptions of the female character prompts. Female markers have been criticized for having a weakening, diminishing, and trivializing effect (Spender, 1980). In addition, supposedly generic masculine forms were used solely to reference men. It appears that despite more than two decades of feminist efforts to discourage gender-marked terms, they are still spontaneously used by the British student participants. In this study then, it was at the lexical level that the routinization of sex bias was most obvious.

If the language policies of the American (APA) and British Psychological (BPS) Associations are an indication of general awareness of sex bias in language, then differences in the use of sexist language forms may be expected. APA published guidelines for the nonsexist use of language more than two decades ago; hence, American college students may be less likely to use those forms. The equivalent British guidelines were published more recently (BPS, 1991). An interesting study would be to test an equivalent sample of British and American college students on their use of male- and female-marked terms.

The final hypothesis addressed by the analysis compared the vocabulary used to describe female as opposed to male characters. Female characters were likened to an impressive variety of animals from the domesticated (e.g., cow) to the exotic (e.g., camel) to the supernatural (e.g., monster). There were fewer animal metaphors used with reference to male characters, but those that were used were unflattering (e.g., rat, pest). Many of the descriptors used with reference to male characters as well as female characters were used to convey sexual and negative connotations (e.g., toyboy, fancy stuff, pillock). The range of terms used by viewers in the present study did not strongly support the view that women and not men were described narrowly and negatively. Although the labels used may be understood as reflecting the range of character types typical of a soap opera (Morley, 1986), the result raises doubts about the validity of research that relies on dictionaries to document vocabulary relating to men and women. Those sources may be reflecting the bias of the lexicographers in selecting words for inclusion rather than accurately reflecting the rich vocabulary available to demean and trivialize both men and women in everyday parlance.

An analysis of prompted descriptions of a soap opera has several limitations, which restrict generalizations about sex bias in language in real-life conversations. The stimulus topic was very narrow, so the full range of language norms for referring to people may not have been used. Giving stills of Coronation Street to dyads was a successful method of getting two previously unacquainted individuals to converse. However, the degree to which the task
elicited the kind of linguistic sexism that would be found in the field is not certain. The descriptions were recorded in the context of psychological research, which would have affected participants' responses. A further shortcoming was the restricted sample who participated in the research. The viewers were all relatively young, White students attending a university in the Northwest of England. Viewers from other sectors may have differed in their descriptions of the program. T. A. van Dijk (1987), among others, has noted that the methodological complications associated with accessing how, when, where, and with whom people talk about particular social groups are practically insurmountable. The method that was used in this study provides one innovative example, albeit limited, of how to collect a suitable source of talk to analyze for references to men and women. Interviews and focus group discussions are alternative methods with their own limitations.

Although the specific results of the present study necessarily have limited generalizability, some general points can be made from the analysis presented here. One point is that it is important to examine spontaneous speech rather than stereotyped representations of language to understand how bias may become routinized by the use of particular language forms to describe members of different social groups. A considerable body of theoretical and empirical research has demonstrated the significance of language in regard to thought and behavior. Thus, it is vital that there are studies that document and monitor language biases, whether they be sexist, racist, homophobic, or ageist, in order to understand contemporary discriminatory beliefs and practices. A further point that can be made fiom the present study is that sex bias in language is not pervasive but appears in different and often subtle forms in different contexts. Further theoretical and empirical work is needed to refine our understanding of how and when language is used in the reproduction of sexism. Descriptions of a popular television program provided one example of a linguistic context in which sex bias was not pervasive. However, in other contexts in which women are not so visible, such as in newspapers (Stirling, 1987) and sports programs (Messner, Duncan, & Jensen, 1993), reliable evidence of linguistic sexism has been found. Women's negative linguistic position
relative to men is not universal, but rather is restricted to particular forms and domains. An up-to-date identification of the forms of sexism and the contexts in which it is prevalent could potentially provide useful information for the targeting of social interventions.

NOTES

1. An initial examination of the dialogues revealed that irregular viewers often merely repeated information
disclosed to them by their partners. To avoid complexities associated with this feedback factor, analyses
testing for sex differences used only those personal references made by regular viewers.
2. Mavis (n = 108) was the most frequently used name in the spoken descriptions.

REFERENCES

Allport, G. W. (1963). Pattern and growth in personality. New York: Holt, Rinehart & Winston. American
Psychological Association. (1975). Guidelines for nonsexist use of language: Task force on issues of sexual bias in graduate education. American Psychologist, 30, 682-684. British Psychological Society. (1991). Code of conduct and ethical principles and guidelines. Leicester: Author.
Brown, R, & Ford, M. (1961). Address in American English. Journal of Abnormal and Social Psychology,
Cohen, 3. A. (1960). Acoefficient of agreement for nominal scales. Educational and Psychological
Measurement, 20, 37-46.
Coronation Street. (1991, April 22). Manchester, UK: Granada Television. Crawford, M. (1995). Talking
difference. On gender and language. London: Sage Ltd.
Graham, A. (1975). The making of a non sexist dictionary. In 8. Thorne & N. M. Henley (Eds.), Language
and sex: Difference and dominance (pp. 57-63). Rowley, MA: Newbury House.
Henley, N. M. (1987). This species that seeks a new language: On sexism in language and language
change. In 3. Penfold (Ed.), Women and language in transition (pp. 3-27). New York: State University of
New York Press.
Henley, N. M., Miller, M., & Beazley, 1. (1995). Syntax, semantics and sexual violence. Agency and
passive voice. Journal of Language and Social Psychology, 14, 60-84. Kalbfleisch, P J., & Cody, M. 3.
(1995). Gender, power and communication in human relationships. Hillsdale, N1: Lawrence Erlbaum.
62, 375-385.
Key, M. R (1975). Male/female language. New Jersey: Scarecrow. Kramarae, C. (1990). Changing the
complexion of gender and language research. In H. Giles & W. P Robinson (Eds.), Handbook of language
and social psychology (pp. 345-361). New York: John Wiley.
Kramer, C. (1975). Sex-related differences in address systems. Anthropological Linguistics, 17,
Lakoff, R (1973). Language and women's place. Language and Society, 2, 45-80. Lakoff, R. (1990).
Talking power: The politics of language in our lives. New York: Basic Books.
Maltz, D. N., & Borker, R. A. (1982). A cultural approach to male-female communication. In J. Gumperez
(Ed.), Language and social identity (pp.196-216). Cambridge, UK: Cambridge University Press.
Martyna, W. (1980). The psychology of the masculine generic. In S. McConnell-Ginet, R. Borker, & N.
Fulman (Eds.), Women and language in literature and society (pp. 60-78). New York: Praeger.
198-210.
Messner, M. M., Duncan, M. C., & Jensen, K. (1993). Separating the men from the girls: The gendered
language of televised sports. Gender and Society, 7,121-137.
Miller, C., & Swift, K. (1976). Words and women: New language in new times. New York: Anchor.

Butler, Dora and Florence L. Geis. 1990. "Nonverbal Affect Responses to Male and Female Leaders:
Implications for Leadership Evaluations." Journal of Personality and Social Psychology 58:48-59. Carli,
Linda. 1990. "Gender, Language, and Influence." Journal of Personality and Social Psychology 59:94151.
Corrigan, Roberta. 1993. "Causal Attributions to States and Events Described by Different Classes of
Verbs." British Journal of Social Psychology 32:335-348. Craig, J.M. and Carolyn W. Sherif. 1986. "The
Effectiveness of Men and Women in Problem Solving Groups as a Function of Group Gender Composition."
Sex Roles 14:453-66.
Dabbs, James M. and R.B. Ruback. 1984. "Vocal Patterns in Male and Female Groups." Personality and
Social Psychology Bulletin 10:518-25. Erber, R. and Susan T. Fiske. 1984. "Outcome Dependency and
Attention to Inconsistent Information." Journal of Personality and Social Psychology 47:709-26.
Fiedler, Klaus and Gun R. Semin. 1988. "On the Causal Information Conveyed by Different Interpersonal
Verbs: The Role of Implicit Sentence Context." Social Cognition 6:21-39.
Franco, F. and L. Arcuri. 1990. "The Effect of Semantic Valence on Implicit Causality of Verbs." British
Journal of Social Psychology 29: 161-70. Garvey, Catherine and Alfonso Caramazza. 1974. "Implicit
Causality in Verbs." Linguistic Inquiry 5:459-64.
Gollob, Harry F. 1974. "The Subject-Verb-Object Approach to Social Cognition." Psychological Review
Gollob, Harry F. and G.W. Fisher. 1973. "Some Relationships between Social Inference, Cognitive
Balance, and Change in Impression." Journal of Personality and Social Psychology 26: 16-22. Gollob,
Harry . and B.B. Rossman. 1973. "Judgments of an Actor's Power and Ability to Influence Others."
Journal of Experimental Social Psychology 9:391-406. Gruber, K.J. and J.W. White. 1986. "Gender
Differences in the Perceptions of Self's and Others' Use of Power Strategies." Sex Roles 15:109-18.
Haas, A. 1979. "Male and Female Spoken Language Differences: Stereotypes and Evidence."
Psychological Bulletin 86:616-26. Heider, Fritz. 1958. The Psychology of Interpersonal Relations. New
York: Wiley.
Heise, David R. 1969. "Affectual Dynamics in Simple Sentences." Journal of Personality and Social
1989. "Effects of Emotion Displays on Social Identification." Social Psychology Quarterly 52: 10-21.
Heise, David R. and Neil MacKinnon. 1988. "Affective Bases of Likelihood Judgments." In Analyzing
Social Interaction: Advances in Affect Control Theory, edited by Lynn Smith-Lovin and David R. Heise.
New York: Gordon and Breach.
Heise, David R. and Lynn Smith-Lovin. 1981. "Impressions of Goodness, Powerfulness, and Liveliness
from Discerned Social Events." Social
81 :286-321. Psychology 11:204-13.
Psychology Quarterly 44:93-106. Heise, David R. and L. Thomas. 1989. "Predicting Impressions
Created by Combinations of Emotion and Social Identity." Social Psychology Quarterly 52: 14148.
Hilton, Denis and Ben Slugoski. 1986. "KnowledgeBased Causal Attriubtion: The Abnormal Condition
Focus Model." Psychological Review 93:75-88. Hoffman, Curtis and M.A. Tchir. 1990. "Interpersonal
Verbs and Dispositional Adjectives: The Psychology of Causality Embodied in Language." Journal of
Personality and Social Psychology 58:765-78. Kimble, Charles E., J.C. Yoshikawa, and H.D. Zehr. 1981.
"Vocal and Verbal Assertiveness in Same-Sex and Mixed-Sex Groups." Journal of Personality and Social
Psychology 40: 1047-54. LaFrance, Marianne and Nancy M. Henley. 1994. "On Oppressing Hypotheses
or Sex Differences in Nonverbal Sensitivity Revisited." Pp.287-311 in Power and Gender, edited by L.
Radtke and H. Stam. London: Sage.
MacKinnon, Neil J. 1994. Symbolic Interactionism as Affect Control. Albany: SUNY Press. MacKinnon,
Neil 3. and David R. Heise. 1993. "Affect Control Theory: Delineation and Development." Pp. 64-103 in
Theoretical Research Programs: Studies in the Growth of Theory, edited by Joseph Berger and Morris
Zelditch Jr. Stanford: Stanford University Press.
Mannetti, L. and E. De Grada. 1991. "Interpersonal Verbs: Implicit Causality of Action Verbs and
Contextual Factors." European Journal of Social Psychology 21 :429-43.
McArthur, Leslie Z. 1972. "The How and What of Why: Some Determinants and Consequents of Causal
Attribution." Journal of Personality and Social Psychology 22: 171-93.
Meeker, B.E and PA. Weitzel-O'Neill. 1977. "Sex Roles and Interpersonal Behavior in Task-Oriented
Groups." American Sociological Review 42:91-104. Mulac, Anthony, R.L. Lundell, and James J. Bradac.
1986. "Male/Female Language Differences and Attributional Consequences in a Public Speaking
Situation: Toward an Explanation of the Gender-Linked Language Effect." Communication Monographs
53:115-29. Ridgeway, Cecilia L. 1981. "Nonconformity, Competence, and Influence in Groups: A Test of
Two Theories." American Sociological Review 46:333-47. Ridgeway, Cecilia L. and D. Diekema. 1992.
"Are Gender Differences Status Differences?" Pp. 157-180 in Gender, Interaction and Inequality, edited
by Cecilia L. Ridgeway. New York: Springer-Verlag. Robinson, James, and Leslie Z. McArthur. 1982.
"Impact of Salient Vocal Qualities on Causal Attribution for a Speaker's Behavior." Journal of Personality
and Social Psychology 43:236-47.
Ruble, Thomas L. 1983. "Sex Stereotypes: Issues of
Change in the 1970's." Sex Roles 9:397-402. Smith-Lovin, Lynn. 1988. "Impressions from Events." Pp.
35-70 in Analyzing Social Interaction: Advances
in Affect Control Theory, edited by Lynn Smith-Lovin and David R. Heise. New York: Gordon and Breach.
Steckler, N.A. and Robert Rosenthal. 1985. "Sex Differences in Nonverbal and Verbal Communication
With Bosses, Peers, and Subordinates." Journal of Applied Psychology 70: 157-63.
Thome, Barrie, Cheris Kramarae, and Nancy M. Henley
eds. 1983. Language, Gender and Society. Rowley, MA: Newbury House.
Van Kleeck, M.H., L.A. Hillger, and Roger Brown. 1988. "Pitting Verbal Schemas Against Information
Variables in Attribution." Social Cognition 6:89-106.

Marianne LaFrance is professor of social psychology at Boston College. Her research interests center on
the intersections of gender power and communication.
Hiram Brownell is an associate professor of psychology at Boston College and an adjunct assistant
research professor of neurology (neuropsychology) at the Boston University School of Medicine. His
research interests include discourse ability in normal and brain-injured people.
Eugene Hahn is a doctoral student in the School of Business and Public Management at the George
Washington University. His current interests involve the augmentation of management decision making
processes via quantitative methods.

return to Schedule