The Myth of Mars and Venus

One linguist, Deborah Cameron (2009), interprets Hyde’s (2005) findings as meaning that women and men, in truth, have more in common than pop psychologists’ books might have us believe. She writes to a popular audience and tries to convince laypeople that “evolutionary biology” is mere conjecture, and that writers like Tannen (1991) and Brizendine (2006), who emphasize the differences, are doing more harm than good. However, Cameron (2009) relies heavily on James & Drakich’s (1993) findings on amount of talk and on Hyde’s (2005) finding that the overall effect of gender on differences in communication is small. While Cameron is not fabricating numbers ala Brizendine (2006), she conveniently ignores all of Leaper’s studies that found significant differences in male and female styles of speech.

With the publication of Cameron’s The Myth of Mars and Venus (2009) as a popular/linguistic rival to Tannen’s You Just Don’t Understand (1991), two camps were created. They use different sets of data to justify different conclusions about gender’s effect on amount of talk. Only time will tell which school of thought will fade into history and which will prove lasting, or if, like Universal Grammar, one camp digs in, establishes its own universities, and refuses to listen to others in the field.

The Gender Similarities Hypothesis.

Hyde (2005) conducted a meta-analysis of the most up-to-date and methodologically sound 46 meta-analyses that examined psychological gender differences. She examined Leaper & Smith (2004) and Hyde & Linn (1988); she did not include any other study examined in this paper in her meta-analysis. Hyde found that “despite Tannen’s (1991) assertions, gender differences in most aspects of communication are small” (p. 586). The five meta-analyses concerning communication differences in gender, with a total of 13 different areas, had an averaged effect size (Cohen’s d ) of -.213, which according to Cohen’s benchmarks, is a small difference. Two of these communication meta-analyses did not study language: one examined studies on smiling and the other studied facial expression processing. The three meta-analyses concerned with language examined interruptions, talkativeness and whether the talk was affiliative or assertive (a different meta-analysis than Leaper and Sanders’ meta-analysis, discussed above), and self-disclosure. I computed that the average Cohen’s d of these three meta-analyses and their combined eight areas was -.039, which is decidedly insignificant. Hyde found some areas where the differences were statistically significant, but these areas were not related to amount of talk (they included throwing distance and amount of masturbation).

Gender differences in verbal ability: a meta-analysis.

Hyde & Linn (1988) set out to examine an adage in psychology that has been oft-repeated and seldom examined: that girls have better verbal abilities than boys. Their study included dissertations and unpublished studies, which are accessible via ERIC; all told, they examined 165 studies. The authors identified five different skills that might be measured in tests of verbal ability: definition retrieval, retrieval of names from a picture stimulus, creating relations between words, filtering for relevant information, and measures of verbal production. Unfortunately, Hyde & Linn (1988) did not report running tests without statistical outliers, instead simply excluding one study based on, apparently, their own judgment. They conclude, “that there are no gender differences in verbal ability, at least at this time, in American culture, in the standard ways that verbal ability has been measured” (p. 62). The verbal production subtests indicated that females performed better on tests that measure quality of production, but that males produced more in terms of “total talking time” (p. 64).

Race and gender in current American politics: A discourse-analysis perspective.

At least one study combined qualitative rhetorical analysis with quantitative word counts: Suleiman & O’Connell (2008) examined political guests on Larry King’s talk show. They found that the race of guests influenced the amount of non-standard speech they felt comfortable using, with Barack Obama, Condoleeza Rice, Hilary Clinton, and Colin Powell all using fewer non-standard forms than Bill Clinton. I mention this to note that researchers are beginning to take other sociological variables into account when examining discourse; I found this to be rare even in current studies. Suleiman and O’Connell’s quantitative analysis found that “Bill Clinton consistently talks the most, and Hillary Clinton the least” (p. 381). They note that this is consistent with Mehl’s finding that, contrary to the popular myth, women do not speak more than men. By comparing the amount of talk generated by political guests on a talk show to Mehl’s findings, the authors imply that the political interviewees verbal production can be generalized to the larger American population.

Gender differences across correlated corpora: Preliminary results.

Sabin, Goodwin, Goldstein-Stewart, & Pereira (2008) examined six different genres on the internet and found, consistent with Mehl (2007), no difference in amount of words generated by male and female authors, except “males generated significantly larger amounts of text than females on the topic of the legalization of marijuana” (p. 209).

Symbolic Capital in a Virtual Heterosexual Market.

Susan Herring studies computer-mediated communication and found that Italian women sending text message to a TV program sent longer text messages and used more non-standard typed methods of expressions, like emoticons and non-standard punctuation (Herring & Zelenkauskaite, 2009).

Are women really more talkative than men?

One study regarding the talkativeness of males and females has attracted the same news outlets as Brizendine’s book. This study has been highly influential, in both academic and in non-academic spheres. Mehl (2007) paid university students to wear an unobtrusive recording device for several days. The students removed the device only when activities might damage it (during showers, playing sports, etc.). Mehl found no statistically significant difference in the amount of words spoken by men and women. He readily admits that the study is limited because all the participants were university students.

Of all the studies examined in this paper, Mehl’s seems most like the type of measurement that is most likely to answer the question of whether men or women talk more. The study is not without its problems: participants were able to take off the recorder at any time or to erase any speech they did not wish the researchers to hear; also, the researchers apparently used a convenient sample of undergraduates. However, it took samples of speech in the most natural of environments: the participants’ daily lives. This is a great improvement over controlled, laboratory experiments measuring amounts of speech or even observed behavior in homes because the unobtrusiveness of the device is the closest thing we have yet to the proverbial “fly on the wall.”

More studies like Mehl’s are needed to determine if his results are generalizable to the larger population. The recording device he used sampled the actual ambient noise, allowing him to record naturalistic data without a constructed context; the device was also unobtrusive, which removed problems related to the “observer’s paradox” of trying to observe a natural behavior which changes when it is observed. When Mehl’s study is replicated with other, non-undergraduate populations, the amount of talk from each sub-population can be meta-analyzed and we can determine who talks more (while wearing the recording device). Until then, the academic community cannot consider the issue completely resolved.

The results of Mehl’s 2007 study were published in various popular news sources, like Reuters, ABC News, Associated Content, and National Public Radio. Particularly interesting are the comments at the end of the stories, which can erupt into heated arguments. At the end of Scientific American’s online summary of Mehl’s findings, Kenji1960 (2009) wrote: “’There are lies, Damn lies, and then there are statistics.’ Smart people can make statistics say anything they want. I know how long my wife talks on the phone to her friends.”

Kenji1960’s comment shows that the cultural myth bears more weight than empirical studies, which he obviously does not trust. Despite empirical evidence to the contrary, judging form other comments posted on other popular news sites, the cultural myth that women speak more than men appears to be alive and well. In order to combat the stereotype and this persistent cultural myth, we’ll need more than just solid methodology and publication in Scientific American. I don’t know what exactly Kenji1960 will find convincing, but if the academic community wants to do work which benefits society by dispelling stereotypes, we must reach out to people distrustful of us.

A Meta-Analytic Review of Gender Variations in Children’s Language Use: Talkativeness, Affiliative Speech, and Assertive Speech

Leaper & Smith (2004) conducted a meta-analysis similar to the one done in Leaper & Ayres, but concentrating on children’s talkativeness and affiliative versus assertive speech. They did not use unpublished studies and ran trimming tests, which determined that the outliers did not bias the results. A meta-analysis of 73 studies investigating talkativeness found that “girls were significantly more talkative than boys” (p. 994). Although this finding was determined to be statistically significant, “it was relatively negligible in magnitude,” meaning that the effect size was too small to be generalizable. The most important moderator was interaction partner. When the child spoke to an adult, girls were far more talkative than boys and this finding was significant both in effect size and statistically. Girls used more affiliative speech than did boys, although this finding also had a small magnitude and effect sizes were greatly increased depending on contextual factors (structured or unstructured activities, presence of video recorder, etc.). Boys were slightly more likely than girls to use assertive speech, and this effect increased depending on group size, whether the interaction took place in the home, and other contextual variables.

Leaper & Smith (2004) note that girls are more talkative at all age levels except between five and nine years, when the difference is negligible (d=.06). Both genders engaged in more assertive speech in groups while more affiliative speech was observed in dyads. Older children were more likely to use gendered speech (affiliative for girls and assertive for boys) in single-sex groups than in mixed-sex groups. Boys used more assertive speech at home than in labs. Leaper & Smith’s (2004) meta-analysis is comparable to James & Drakich’s (1993) meta-analysis in that gender as a variable cannot account for the variation in generally male (dominant or assertive) versus generally female (cooperative or affiliative) styles of speech.

Moderators of Gender Effects on Parent’s Talk to Their Children: A Meta-Analysis

Another meta-analysis examined the specialized context of parents’ language with their children. They did not include self-report studies in their meta-analysis, and as a result had significantly different findings from earlier parent-child speech meta-analyses. Leaper, Anderson, & Sanders (1998) identified two types of speech: supportive speech, which they defined as “positively responsive,” was associated with mothers’ speech and is comparable to Leaper & Ayres’ affiliative speech; and directive speech, which the authors associated with fathers’ speech and is comparable to assertive speech (although Leaper & Ayres’ “assertive speech” included criticism and giving information, which were not included in Leaper, Anderson, & Sanders’ “directive speech”). The authors’ findings included: 1) “mothers tended to use more supportive language strategies with their children than did fathers;” 2) fathers used more directive and informative language as well as asked more questions; 3) mothers were more talkative with daughters than with sons (p. 12).

Significantly for the focus of this paper, Leaper, Anderson, & Sanders (1998) found that “In general, mothers were more likely to demonstrate higher amounts of verbal interaction” (p. 21): mothers spoke to their children more. This is in line with Leaper & Ayres’ (2007) findings that the presence of a child was a significant moderator in women’s talkativeness.

A Meta-Analytic Review of Gender Variations in Adults’ Language Use: Talkativeness, Affiliative Speech, and Assertive Speech.

Leaper and Ayres’ 2007 meta-analysis examined 63 published studies that were methodologically sound (e.g., actually sampled data instead of relying on intuition) and peer-reviewed in either journals or books. The authors examined not only the amount of words spoken by each gender but also categorized the amount of words spoken into two categories: affiliative speech, which is primarily concerned with the sociocultural relationship of the interlocutors (e.g., compliments, back channeling, and probing questions); and assertive speech, which is persuasive or controlling (e.g., criticism, explanations, and, in one case, a verbal threat). These two categories are similar to James & Drakich’s (1993) “cooperative” and “dominant” styles of speech. The authors ran separate tests for single-sex versus mixed-group results and also removed the outliers at 10% and 20%. Leaper & Ayres concluded that, overall, “men were significantly more talkative than women”; however, “as predicted, women used significantly more affiliative speech than men” and “men used significantly more assertive speech” (p. 351). These findings do not corrorborate James & Drakich’s (1993) findings. In certain contexts, Leaper & Ayres (2007) found women to be more talkative than men, notably when a child was present or children were the topic of conversation, and during self-disclosure. The authors note that these are stereotypically envisioned as female domains, and so women’s talkativeness fit into our society’s gendered norms.

Understanding gender differences in amount of talk

To my knowledge, the earliest meta-analysis to examine which gender speaks more was James & Drakich’s chapter in Gender and Conversational Interaction (1993). They found that most studies found that men speak more than women, and their findings did not support the widelyheld view that men’s speech is “dominant” and women’s speech is “cooperative.” This was probably the most widely cited meta-analysis, although it only surveyed 56 studies; meta-analyses published since have examined more and do not support James & Drakich’s findings.