Speaker sex and perceived apportionment of talk

It is a widely held belief that women talk more than men; but experimental evidence has suggested that this belief is mistaken. The present study investigated whether listener bias contributes to this mistake. Dialogues were recorded in mixed-sex and single-sex versions, and male and female listeners judged the proportions of talk contributed to the dialogues by each participant. Female contributions to mixed-sex dialogues were rated as greater than male contributions by both male and female listeners. Female contributions were more likely to be overestimated when they were speaking a dialogue part perceived as probably female than when they were speaking a dialogue part perceived as probably male. It is suggested that the misestimates are due to a complex of factors that may involve both perceptual effects such as misjudgment of rates of speech and sociological effects such as attitudes to social roles and perception of power relations. According to proverbial wisdom, women talk more than men. The English proverb "Women's tongues are like lambs' tails never still" has parallels in many cultures (Swacker, 1975). But experimental evidence does not justify this belief. Some evidence seems at first to favor it: Preston and Gardner (1967) conducted a factor analysis across a large number of language performance measures and found significant sex differences on a word productivity factor, such that women produced more words than men, paused less often than men, and had a larger vocabulary than men. Similarly, Gall, Hobby, and Craik (1969) found that women produced more words to describe visual displays than men. But in a similar study by Swacker (1975), men produced more words to describe pictures than women, while Brotherton and Penman (1977) found no significant difference between the sexes in a similar task. In each of these investigations, however, the speakers produced monologues, and monologues are not the most frequent type of speech performance. The picture from research on conversations is clearer. Hilpert, Kramer, and Clark (1975) measured the relative contributions of ©1990 Cambridge University Press 0142-7164/90 $5.00 + .00 Applied Psycholinguistics 11:3 254 Cutler & Scott: Speaker sex and perception of talk each partner in a mixed-sex dyad to an unstructured conversation; more often than not, the male partner spoke more. Argyle, Lalljee, and Cook (1968) also found that males in mixed-sex dyads spoke more than females. Markel, Long, and Saine (1976), on the other hand, looked at sameand mixed-sex dyads, and reported no effects for overall proportion of contribution to the conversation; but they did find that female turns were longer. Interestingly, they found that listener sex was a significant factor; everyone spoke more to a female listener. By implication, this tells us that there was more inequality in mixed-sex dyads with males speaking more than females than in same-sex dyads. In the dyads studied by Duncan and Fiske (1977), it was the men who tended to take longer speaking turns. In larger groups, the evidence even more clearly suggests that men tend to speak more than women. Eakins and Eakins (1976) found that men spoke longer and more often in faculty meetings than their female colleagues. Spender (1979, 1980) taperecorded spontaneous group discussions and found that male participants often contributed a greater proportion of the discussion even when they were in a minority. On balance, then, there is actually better evidence for men speaking more than women than vice versa. There is certainly no evidence to support the widespread folk belief that women are overwhelmingly the more garrulous sex. Perhaps the existence of such a belief itself constitutes evidence that people are very bad at judging how much is spoken. In fact, Hilpert, Kramer, and Clark (1975) found that the perceptions of their conversational participants were in general fairly accurate about who had spoken most, although the men's perceptions were slightly less accurate than the women's. Spender (1979), however, reported very inaccurate perceptions by her group discussants; in a group in which almost twice as many men as women had spoken, participants reported that the majority of speakers had been female. Extraordinarily, we do not know how listeners actually assess how much is spoken. Common sense tells us that someone who drawls a sentence slowly is not considered to have said more than another person who gabbles the same sentence twice in less time than the first person took to say it once. That is, we normally make allowance for speaking rate in judging who says most; amount of linguistic material produced is what really counts. Although speech scientists have not directly investigated how listeners assess amount of talk, there is a large literature on how listeners judge rate of speech. This literature shows that listeners are in fact quite poor at estimating rate. Their estimates take into account both articulation rate (words, syllables, or segments per unit time) and the number and duration of pauses (Grosjean & Lane, 1974,1976, 1981), but they are also affected by irrelevant factors either higher intensity or higher fundamental frequency, for example, can cause one of two speech signals spoken at identical rates to be perceived as having a faster rate (Bond & Feldstein, 1982; Feldstein & Bond, 1981), as can absence of intonational phrasing (Rietveld & Gussenhoven, 1987). The tendency for speech rate to be misjudged as a function of fundamenApplied Psycholinguistics 11:3 255 Cutler & Scott: Speaker sex and perception of talk tal frequency is explained as due to our frequent experience of withinspeaker concomitant rises in speech rate and voice pitch. But note that we also experience voice pitch variation between speakers. In particular, female vocal tracts are generally shorter than male vocal tracts, which makes female voices in general higher pitched than male voices. The tendency for higher pitch to be judged as signaling faster rate suggests that one might find female rates of speech to be systematically judged to be faster than they actually are. Indeed, Kramer (1977), who studied stereotypes relating to speech, found that not only was "talking a lot" reported to be a characteristic of women, but so was 'talking fast." There is no evidence that the second stereotype is any truer than the first; none of the studies reviewed above found any evidence of sex-related differences in the rate of production of words (Brotherton & Penman, 1977; Markel, Long, & Saine, 1976; Swacker, 1975, all explicitly measured this factor). Possibly, then, women are judged to speak faster simply because they have higher pitched voices. Now what will happen as a result of rate misjudgments if, as we suggested earlier, listeners usually make allowances for rate of speech in judging how much people talk? Obviously, allowances will sometimes be made without justification. One might argue, then, that the folk belief that women talk more than men is just due to rate misjudgment: women do not actually talk more, but they are judged to be talking more because they are judged to be speaking faster, and faster speakers must be talking more per given unit of speaking time. An alternative suggestion is more complex and may rely on a difference in content between men's and women's speech. Kramer (1975) and Spender (1980) suggested that women are undervalued in society, and as a consequence women's speech is undervalued female contributions to conversation are overestimated because they are held to have gone on "too long" relative to what female speakers are held to deserve. Preisler (1986) similarly argued that evaluation of women's speech is a function of (under)evaluation of the social roles most usually fulfilled by women. The former explanation suggests that overestimation of women's conversational contributions is a perceptual bias effect that should be reproducible in the laboratory simply by asking listeners to judge amount of talk produced by male and female speakers, even if content of the talk is controlled. The latter explanation would predict that the content of conversation should make a strong contribution to misjudgments; if content is strictly controlled, there may be no overestimation of women's contributions, but if content is biased, judgments of relative contribution should be similarly biased. The present study provides an initial test of these predictions. We recorded a number of identical two-party conversations and systematically varied speaker sex. We then had listeners judge the relative proportional contribution of each speaker to the conversation. We could thus ascertain whether male or female speakers were judged as having spoken for a greater proportion of the conversation than was actually the case. If underlying the folk Applied Psycholinguistics 11:3 256 Cutler & Scott: Speaker sex and perception of talk belief is a perceptual bias that we can reproduce in the laboratory when conversational content is controlled, then we should find that in a conversation in which women and men contribute equally, the women are perceived as talking more than the men. We also separately estimated the degree to which the textual content of our conversations was judged to be typical of female or of male speech. If the folk belief rests crucially on perception of speech in terms of social roles, then we should find that the degree to which a given contribution is overestimated should be related to how "female" the text of the contribution is judged to be. It would be desirable to produce speech in which the actual contributions of male and female speakers were exactly matched. This can be done by using synthetic speech, or natural speech that has been digitized and then appropriately compressed or expanded. In either of these cases, however, the speech produced does not sound fully natural. This opens up the possibility that listeners may react to such speech differently from the way they react to speech of "real" men and women speakers. As the present study represents the first direct examination of whether content-controlled speech is misjudged, we preferred to avoid such a possible confounding effect of unnaturalness, and hence we did not digitally manipulate our recorded conversations. By controlling conversational content, we factored out effects specific to personal participation; our listeners were mere observers of others' interaction, not participants themselves. If they are systematically inaccurate in their perceptions of relative amount of talk, their bias is a general one, not one based on their evaluation of their own participation. We also tested whether male and female listeners varied in the accuracy of their judgments. At least one study (Hilpert, Kramer, & Clark, 1975) suggested that women's perceptions of relative amount of talk could be more accurate than men's. Finally, we varied in addition the sex of the experimenter. Our subjects were tested in groups; in such situations the experimenter stands at the front of the room to give instructions and implicitly wields a certain amount of authority. It is possible that the sex of the person giving instructions could affect the operation of any general sex-related bias in conversational perception by affecting transitory perceptions of power relationships. Both authors of this study are female; we therefore enlisted two male colleagues to serve as experimenters for half of the group testing sessions.