Você está na página 1de 14

Sex Roles, Vol. 37, Nos.

7/8, 1997

Gender Differences in Self-Reported Response to Troubles Talk


Shari L. Michaud and Rebecca M. Warner1
University of New Hampshire

Tannen [(1990) You Just Don't Understand: Women and Men in Conversation, New York: William Morrow] and others have argued that women and men differ in communication style, and particularly in the way they respond to "troubles talk." However, her research on gender differences in communication style has primarily made use of qualitative analysis of naturally occurring discourse. A Communication Styles Survey was developed to assess self-reported behavioral and emotional responses in "troubles talk" situations. This survey was administered to a total of 384 participants, the majority of whom were Caucasian college students. Statistically significant gender differences consistent with Tannen's predictions were found but the effect sizes tended to be small. Self-reported responses to troubles talk (such as giving advice) showed consistent individual differences among persons; for example, some persons tend to report themselves as "advice givers" or "jokers" across a wide range of troubles talk situations. Women and men have often been described as having quite different communication styles. Women are typically described as more expressive, more relationship oriented, and more concerned with creating and maintaining intimacy; whereas men are described as more instrumental, more task oriented, and more concerned with gathering information, or with establishing and maintaining social status or power (e.g., Bernard, 1972; Indvik & Fitzpatrick, 1982; Parsons & Bales, 1955). This idea has seized the popular imagination and has become the theme of many best-selling books on gender differences in communication style, e.g., Gray(1992), Kingma (1993), and Tannen (1986, 1990).
1To

whom correspondence should be addressed at Department of Psychology, Conant Hall, University of New Hampshire, Durham, NH 03824.

527
0360MX)25/97/1000.0527SJ2.50/0 1997 Plenum Publishing Corporation

528

Michaud and Warner

In her recent analysis of gender differences in communication style, Tannen (1990, p. 77) stated that "For most women, the language of conversation is primarily a language of rapport; a way of establishing connections and negotiating relationships. Emphasis is placed on displaying similarities and matching experiences." Conversely, "For most men, talk is primarily a means to preserve independence and negotiate and maintain status in a hierarchical social order. This is done by exhibiting knowledge and skill, and by holding center stage through verbal performance such as storytelling, joking, or imparting information" (Tannen, 1990, p. 77). Tannen went on to argue that, as a consequence of these general differences in communication style, women and men may tend to choose different behavioral responses (such as giving advice, giving sympathy) when confronted with "troubles talk." Women and men may also interpret and evaluate these behaviors differently, and therefore feel different emotional responses to advice and sympathy. These gender differences can lead to problems and misunderstandings in communication between women and men. Tannen suggested that when confronted with troubles talk, men are more likely than women to give advice, tell a joke, change the subject, or remain silent (ways of exhibiting expertise or avoiding emotional expression). Women are more likely than men to respond to troubles talk by sharing a similar problem or by openly expressing sympathy (ways of expressing solidarity). In addition to gender differences in the likelihood of different communication strategies (e.g., telling a joke, offering sympathy), there may also be gender differences in the way these communication strategies are evaluated. There are many possible emotional responses to receiving sympathy, for instance: a person might feel comforted, or angry at being condescended to. Gender differences in the responses that are judged to be most appropriate can lead to communication problems when people receive (or do not receive) the responses that they feel are most appropriate. According to Tannen (1990), women give sympathy and expect sympathy in response to troubles talk. When men respond to women's troubles talk by offering advice, women tend to feel that their feelings are being invalidated, their problems are being minimized, or that their partner is being condescending by telling them how to "fix" the problem. Conversely, when women offer sympathy to men, men may feel that they are being placed in a one-down or lower status position and being condescended to. Summarizing Tannen's description of gender differences in the frequency of use of communication styles, it appears that men are more likely than women to respond to troubles talk by giving advice, joking, changing the subject, or giving no response. Women are more likely than men to respond to troubles talk by sharing a similar problem or expressing sympathy. Summarizing her description of the emotional or evaluative responses

Troubles Talk

529

to communication styles: she suggested that when they receive sympathy, women feel more comforted and closer to their partners than men; men who receive sympathy tend to feel more angry or put down. According to Tannen, neither men nor women evaluate advice giving very positively, but her examples suggest that men are more likely to feel "put down" by receiving advice than women. Tannen presented numerous detailed examples of conversations of friends and excerpts from novels and plays that illustrate these gender differences in communication styles. Virtually all of her examples were consistent with this general pattern. What other evidence, beyond these qualitative discourse analyses, do we have that these gender differences in the use and evaluation of communication style really exist? How strong are these gender differences? A potential problem with qualitative data of the sort presented by Tannen is that the examples reported may not be representative; examples are selected to make a point. Surely there are situations in which women give advice or men give sympathy. Furthermore, qualitative studies do not readily lend themselves to precise estimates of effect size. The examples have a vividness that may give an impression of far more extreme gender differences than may actually exist. A more systematic approach to observation and data gathering may provide more easily generalizable finding on the magnitude of gender differences in communication style. Goals of the Present Study The present study evaluates gender differences in communication style through a systematic survey assessment with a large sample of respondents. The intention of the study is to assess whether Tannen's findings of gender differences in troubles talk can be replicated using a different methodology (that has different weaknesses and different strengths than Tannen's research methods). The Communication Styles Survey that was developed for use in this study translates Tannen's general claims about communication into a closed-ended questionnaire. The first part of the survey addressed differences in the tendency to use different communication styles. Respondents read vignettes that briefly present hypothetical "troubles talk" situations (e.g., "A friend tells you that he/she may be failing a course"). They were then asked to rate how likely they would be to use each of the responses that lannen described in her book (e.g., tell a joke, give advice, give sympathy, change the subject). We decided not to require a forced choice among these alternate responses; in real life situations a person might decide to do more than one of these things. Instead, we asked participants to rate how likely they would be to engage in each of the behaviors. A separate scale was created for each response; the "joking" scale

530

Michaud and Warner

summarizes the likelihood of choosing to tell a joke across each of the 6 hypothetical situations described in the survey. The second part of the survey dealt with emotional or evaluative reactions to two of these communication styles: giving advice and giving sympathy. These two behaviors were chosen because Tannen focused on them so prominently in describing gender differences in the evaluation of communication style. Respondents were again presented with hypothetical situations, but in this second set of questions, they were asked to evaluate how they would feel in response to receiving advice (e.g., "You tell a friend that you may be failing a course. Your friend tells you a way to solve this problem. How would you respond?") The emotional responses (e.g., angry, "put down," comforted) were taken from Tannen's accounts. The main goal of the study was to assess whether there are consistent gender differences in communication style, of the type that Tannen (1990) reported. In addition, however, examination of scores on our survey also allows us to assess whether there are consistent individual differences in these communicative behaviors: do some individuals tend to "joke" across many different types of troubles talk situations? Do some individuals tend to be advice givers across many situations? Looking at the interitem agreement across hypothetical situations addresses this question. If some people are "jokers," then they should report a high likelihood of joking across a variety of troubles talk situations. If there is poor interitem reliability across situations, then this would suggest that people's choice of communication style is highly situation specific and does not represent a characteristic individual communication style.

METHOD Participants The participants were 145 men and 239 women who ranged in age from 14 to 79. The majority of these participants (278) were recruited through an introductory psychology participant pool at a state university in northern New England. In addition, some participants were obtained by mailing surveys to randomly selected residents of the local area (with a 14% return rate) and from a local high school (71% return rate). There were not sufficient numbers of older adults to do statistical comparisons among age groups, and overall response patterns were consistent enough across groups to warrant pooling the data from these three sources. Data on ethnicity were not obtained, but fewer than 2% of the students at this university are non white or Hispanic, therefore potential differences associated with race or ethnicity could not be examined.

Troubles Talk

531

Procedures Informed consent was obtained from all participants prior to participation in the study. Written informed consent was obtained from all participants; in addition, for participants under the age of 18, parental consent was also obtained. Participants recruited through the introductory psychology participant pool filled out the questionnaire packet in small groups in a classroom setting; those recruited through high schools or random mailing completed the packet in their homes and returned them by mail. The questionnaire packet took about 30 minutes to complete. Materials Each participant was asked to fill out the Communication Styles Survey (CSS) that was developed by the authors for use in this research and to provide minimal demographic information. The CSS consisted of vignettes that described hypothetical "troubles talk" situations with a friend. The first section of the CSS asked participants to rate the likelihood that they would make specific behavioral responses to a friend who presented a problem. A typical vignette is as follows: "A friend tells you he/she is upset because he/she may fail a course. How would you respond?" Several possible behavioral responses are then listed (e.g., "tell a joke to cheer him/her up," "change the subject," "tell him/her not to worry") and the participant was asked to rate the likelihood that they would make each of these possible behavioral responses on a 1-5 Likert scale (with the anchor points labeled 1 = never, 5 = always). Sample items from the CSS are shown in Table I. Pilot testing indicated that specifying male or female friends in the vignette did not elicit systematically different responses, and so gender of the friend was left open (he/she) in this version of the survey (Michaud, 1991). Factor analysis of the pilot test data also suggested that the "give advice" behavior might be ambiguous; the likelihood of giving advice correlated highly with the likelihood of giving sympathy. Therefore in the revised CSS the wording that described advice giving behavior was changed to "tell him/her a way to solve the problem." In order to assess whether the tendency to engage in a particular communication style (e.g., joke, change subject) was consistent across hypothetical situations, 6 different troubles talk situations were included in the survey. The content of these situations was tailored to the concerns of the college aged population: these included problems with dating partners, coursework, drinking, co-workers, roommates, and depression.

532

Michaud and Warner

Table I. Sample Items and Responses from the Three Sections of the Communication Styles Survey Part I: Behavioral responses For each of the following responses, the respondent rates the likelihood of making that response on a 5-point Likert scale. A friend tells you that he/she is upset because he/she may fail a course. How would you respond? would would would would would would tell him/her how to resolve this problem ("give advice" response). say something sympathetic. change the subject. tell him/her about a similar problem I have had. tell a joke to cheer him/her up. tell him/her that he/she shouldn't worry. Part II: Emotional response (Each emotional response rated on a 1-4 scale) You tell a friend that you may break up with your dating partner. Your friend {tells you a way to solve this problem} {gives you sympathy}. How would you respond? Comforted Angry Grateful That he/she had "put me down" That this was helpful That we were "close friends" Hurt

The second part of the CSS dealt with emotional responses to these communication styles. These vignettes involved situations in which the participant is asked to imagine that he or she has presented a problem to a friend and received one of the specific behavioral responses described earlier; the participant is then asked to rate the extent to which he or she would feel various emotions. A typical item stem is, "You tell a friend that you may break up with your dating partner. Your friend offers you sympathy. How would you respond?" The emotional responses that are listed include "comforted," "angry," "grateful," and other feelings; the participant was asked to rate the extent to which he mor she would experience each of these emotions, on a 4 point Likert scale with the anchors labeled 1 = not at all and 4 = a lot. A sample item is shown in Table I. In part I of the CSS, all 6 of the hypothetical troubles talk situations were presented; in part 2, a subset of 3 of these situations (dating, coursework, and drinking problems) were presented in combination with 2 of the behavioral responses (advice and sympathy). That is, participants rated their feelings in response to receiving advice and receiving sympathy in response to each of these three kinds of troubles talk.

Troubles Talk Table II. Reliability Coefficients for the Six Behavioral Scalesa Scale Advice Sympathy Change Subject Similar Situation Joking Don't Worry Cronbach a

533

.85 .87 .84 .76 .89 .86

"Each scale consisted of 6 items, each describing a different "troubles talk" situation.

The decisions to leave the gender of the friend unspecified in these vignettes and to use only a subset of the situations in part 2 of the survey were made in order to keep the survey a manageable length. The original longer version of the survey (that repeated all situations with both a same gender and an opposite gender friend in the vignette) was very time-consuming and tedious, and did not reveal any of the expected differences for gender of friend. Scoring of the CSS was done in the following manner. Six communication behavior scales were created from the first section of the CSS. For instance, the "joking" scale was formed by summing the likelihood that the participant would tell a joke in response to the six hypothetical troubles talk situations; similar scales were created for each of the six communication behaviors. Internal consistency reliabilities for these scales, as assessed by Cronbach alpha, ranged from .76 to .89 (see Table II). Separate scales were also created for each of the possible emotional responses to advice and to sympathy in part II of the survey. For instance, the scale that assessed whether a person felt comforted by advice was formed by summing the ratings of "comforted" across the three hypothetical situations in which the person received advice in response to troubles talk. Although each of these 14 scales consisted of only 3 items, the internal consistency reliabilities for these scales ranged from .64 to .84 (see Table III). These emotional responses could be roughly grouped into positive and negative emotional responses, but for these analyses, the scale scores of each individual emotional responses were examined separately.

RESULTS Prior to computing scale scores on the CSS, a factor analysis on each group of CSS items was performed to assess whether the items that asked about a common response behavior (such as "joking") all loaded on one

534

Michaud and Warner Table III. Reliability Coefficients for the Emotional Response Scales Emotion Scales Angry Hurt Put down Comforted Helped Grateful "Close friends"
aEach

a for Response to Advice

a for Response to Sympathy

.67 .70 .71 .73


.76 .74

.64 .77 .70 .81


.78 .81

.84

.84

of these fourteen scales consisted of three items.

factor, which would indicate that the responses to items were consistent enough across problem situations to warrant summing items to form scales. The factor analysis results are not reported in detail here. A factor analysis was performed for the group of items forming each section of the CSS, using Principal Axis Factoring with varimax rotation in SPSS. The factor analysis of part I (the behavioral responses to troubles talk) on the CSS yielded 6 factors with eigenvalues greater than 1.0. The rotated factor loadings showed a simple and interpretable pattern: all the items that referred to each specific behavioral response (e.g., the likelihood of "joking" in response to six different kinds of troubles talk) had large positive loadings (>.5) on the same factor. There were no instances of items that had large loadings on more than one factor. This pattern of factor analysis results suggested that the a priori scoring method (for instance, combining the six joking items to form a "joking" scale) made sense. Scale scores were calculated for each of the 6 behaviors and 14 emotional responses as described in the previous sections, and correlations among these scores were examined to see if they represented distinguishable responses. Correlations among the six behavioral response scales are displayed in Table IV These correlations ranged from essentially zero to moderate and positive, and overall it appeared that the six behaviors tapped by these 6 scales were relatively independent of each other.
Table IV. Correlations Among the Behavioral Response Scales

1
1. Advice 2. Sympathy 3. Joke 4. Similar Situation 5. Change Subject 6. Don't Worry
ap

.36a .34" .39 -.01 .37a

.31a .32a -.13 .24a

.37 .22a .55a

.21a .32"

.27"

< .05.

Troubles Talk Table V. Correlations Among the Emotional Response Scales (Responses to Advice)

535

1
1. Angry 2. Hurt 3. Put down 4. Comforted 5. Helped 6. Grateful 7. Close Friends

.68a .60a -.30a -.23 a -.27a -.14"

.65" -.25" -.17a -.18" -.08

-.24a -.17" -.21" -.14"

.76" .77" .49a

.79a .63"

.54"

Correlations Among the Emotional Response Scales (Reactions to Sympathy)

1
1. Angry 2. Hurt 3. Put Down 4. Comforted 5. Helped 6. Grateful 7. Close Friends
ap

.74" .65" -.16" -.15" -.13" -.09

.63"

-.14"
-.05 -.08 -.04

-.16" -.08 -.10 -.07

.75" .80" .63"

.80a .69a

.62"

< .05.

A factor analysis was also performed on part II of the CSS (the emotional response items). A two-factor solution emerged with separate factors for positive and negative affect. Although the emotional response items did tend to be grouped together (all the positive emotional responses had relatively large loadings on factor 1, for instance) separate scores were retained for each specific emotional response. Correlations among the 14 emotional response scales are presented in Table V. Negative affect scales (e.g., angry, put down, hurt) tended to show positive intercorrelations with other negative affect scales (r"s on the order of .60 to .74). Positive affect scales (e.g., comforted, helped, grateful) also tended to be positively correlated with other positive affect scales. Correlations between positive and negative affect scales (e.g., comforted with hurt) tended to be small and negative. We decided to examine the individual emotional response items separately in order to test more specifically the claims made by Tannen (1990) about gender differences in emotional response to troubles talk situations. Gender Differences The primary goal of this study was to assess whether this survey methodology would find differences between men and women in their behavioral and emotional responses to troubles talk, comparable to the differences that have been reported in studies that involve discourse analysis and direct observation of marital interaction. This question was addressed by performing

536

Michaud and Warner Table VI. Means, / Tests, and Effect Sizes for Gender Differences on the Behavioral Response Scales Scale Advice Sympathy Joking Change Subject Similar Situation Don't Worry
ap

Female Mean

Male Mean

t
-0.84 -5.22" 1.72 5.96" 1.70 2.30"

3.47 3.83 2.83 1.50 3.16 2.89

3.40 3.40 3.02 1.93 3.30 3.12

I2 .00 .07 .01 .12 .01 .02

< .05.

t tests to compare the mean scores of male and female participants on the six behavioral response scales and the fourteen emotional response scales. Results of the t tests and means for the female and male respondents on the six behavioral communication styles for men and women appear in Table VI. Statistically significant differences (p <.05, two tailed) were found for three of the behaviors. As predicted by Tannen (1990), women reported a higher likelihood of "offering sympathy" in response to troubles talks than men; men reported higher likelihoods of "changing the subject" and "telling the person not to worry" than did women. However, the effect sizes (as indexed by t]2) were quite modest for all three of these differences, with T)2 less than .12. There were also several significant differences between men and women in the emotional response scales. One set of the emotional response items asked participants how they would feel if they received advice from a friend in response to talking about troubles. Results of t tests for these scales are reported in Table VII. Gender differences were statistically significant for all 7 of the emotional response scales that reported reactions to advice. Men reported higher likelihoods than women that they would feel "hurt," "put down," and "angry" in response to receiving advice about
Table VII. Means, t Tests, and Effect Sizes for Emotional Response Scales (Responses to Receiving Advice) Scale Angry Hurt Put Down Comforted Grateful Helped Close Friends
ap

Females

Males

t
2.09a 3.69a 2.32a -3.60a -3.46" -2.96a -2.90a

4.00 3.59 3.77 2.97 3.10 3.08 2.99

4.41 4.25 4.19 2.68 2.82 2.85 2.73

n2 .01 .02 .01 .01 .01 .01 .01

< .05.

Troubles Talk
Table VIII. Means, t Tests, and Effect Sizes for Emotional Response Scales (Responses to Receiving Sympathy) Scale Angry Hurt Put Down Comforted Grateful Helped Close Friends
"p < . 05.

537

Females

Males

t
0.67 1.80

n2
.00 .01
.01

3.96 3.73
3.60 2.89 2.75

4.08 4.07 3.97


2.70

2.60
2.45 2.58

2.62
2.85

2.30" -2.17" -1.72 -1.91 -2.90"

.01 .00 .00


.01

their troubles. Women reported higher likelihoods that they would feel "comforted," "helped," "grateful," and as if they were "close friends" to the other person in response to advice. Overall, men reported more negative feelings in response to advice from a friend; women reported more positive feelings in response to advice. A second set of emotional response items assessed how the participant would feel if he or she received sympathy from a friend in response to troubles talk. Results of these t tests are reported in Table VIII. Three of these items showed significant gender differences. Men rated the likelihood of feeling "put down" in response to sympathy higher than did women; women rated the likelihood of feeling "comforted" and "close friends" to the other in response to sympathy higher than did men. The effect sizes for these emotional response scales were also very small, with eta squared values of .01.

DISCUSSION The Communication Styles Survey that was developed to assess the kinds of communication styles described qualitatively by Tannen (1990) and others does appear to tap several behavioral and emotional responses with good internal consistency reliability. Do men and women differ in these behavioral and emotional responses, in the manner predicted by Tannen (1990) and others? Many statistically significant differences between men and women were found in this study, and all were in the direction predicted by Tannen's work. The large number of significance tests conducted may lead to an inflated risk of Type I error. However, it was clear that the number of significant differences greatly exceeded the 5% significant outcomes that would have been expected by chance if the null hypothesis were true; of the 20 t tests corn-

538

Michaud and Warner

paring female and male behavioral and emotional responses in troubles talk situations that were conducted, 13 were significant. In general, women were more likely to offer sympathy than men; men were more likely to change the subject or to tell a joke than women. This is consistent with the generally accepted stereotype about the communication styles of women and men and the predictions made by Tannen (1990) and others, suggesting that women are more likely to be supportive and men are more likely to be avoidant. However, we did not obtain significant gender differences in the tendency to give advice (Tannen's anecdotal examples suggest that men are more prone to do this than women) or in talking about similar experiences in response to troubles talk (Tannen's accounts suggested that women do this more than men.) Furthermore, the effect sizes were very small, even for the differences that were statistically significant. The methodology in this study involved presenting very brief written vignettes. Murphy et al. (1986) pointed out that studies using such vignettes ("paper people") typically generate larger effect sizes than studies that present more lifelike stimuli (such as videotapes), probably because there is less irrelevant or distracting information. Therefore it seems likely that, if anything, our methodology may lead us to overestimate effect sizes. Given this caveat, then, we suggest that while there may be real gender differences of the sort identified by Tannen and others, that these differences may in reality be quite small; there is very substantial overlap in the behaviors of men and women. Not all men are advice givers; not all women are sympathy givers. The Cronbach a's that were obtained for these scales tended to be reasonably high, even though the individual scales were composed of very few items (6 items in each behavioral scales, 3 items in each emotional response scale). The fact that the internal consistency reliabilities are so high suggests that the self-reported tendency to engage in a particular behavior, such as advice-giving, in response to troubles talk is quite consistent. That is: some persons consistently reported a high likelihood of giving advice, in response to all 6 types of troubles talk vignettes, while other people consistently reported low likelihoods of giving advice. Being an "advice giver" may therefore be a fairly consistent response style. Of course, we do not know whether the actual tendency to give advice shows as high a level of consistency as we found for these self-reports of likelihood of behavior. Our results suggest that some persons view themselves as advice givers, or sympathy givers, or jokers, across a variety of situations, while other persons do not view themselves this way. The relatively modest intercorrelations among the behavioral scales suggest that these scales tap fairly distinct behavior patterns. The highest correlation between scales, the .55 correlation between the likelihood of

Troubles Talk

539

telling the person not to worry and the likelihood of telling a joke, suggests that people who engage in one form of avoidant behavior are likely to also engage in the other; but neither of these behaviors correlated highly with "changing the subject," which seems like an even more clearly avoidant response. Similarly, high internal consistency reliabilities were also found for the emotional response scales, suggesting that the tendency to report feeling "hurt" or "angry" in response to advice or to sympathy may also be a fairly consistent individual difference. Intercorrelations among the emotional response scales indicates two somewhat separate groups of items; ratings of the likelihood of all the negative emotions (angry, hurt, put down) were positively intercorrelated; ratings of the likelihood of all the positive emotions (comforted, helped, grateful, close friends) were positively intercorrelated; and correlations between positive and negative emotion items tended to be negative, but rather small. Some of the correlations between positive emotion items were fairly large (.79 for the correlation between "grateful" and "helped") and so it might be useful in future research to combine these emotions into overall positive and negative emotion scales; however, for purpose of this exploratory analysis, these items were kept separate in order to obtain a more detailed description of gender differences in emotional response. Of course, it is questionable whether people's actual behaviors are predictable from self-reported likelihoods of engaging in behaviors. However, when these self-report data are considered along with the discourse analyses of Tannen and others, they provide another kind of evidence about the existence and strength of gender differences in communication style. To the extent that similar conclusions are reached using different methodologies (discourse analysis versus self report) we can have greater confidence in the findings. These two methodologies detect similar qualitative differences between the communication styles of women and men; however, the effect sizes appear to be much smaller in the self report survey study than one might have expected from the examples of discourse analysis.

REFERENCES
Bernard, J. (1972). Early gender differences in the functional usage of language. Sex Roles, 12, 909-915. Gray, J. (1992). Men are from Mars, women are from Venus, New York: Harper Collins. Indvik, J., & Fitzpatrick, M. A. (1982). Sex differences in imagined interactions. Sex Roles, 21, 263-272. Kingma, D. R. (1993). The men we never knew. Berkeley, CA: Conari Press.

540

Michaud and Warner

Michaud, S. L. (1991, August). Gender and communication styles. Paper presented at the McNair Graduate Opportunities Program Conference, University of New Hampshire, Durham, NH. Murphy, K. R., Herr, B. M., Lockhart, M. C, & Maguire, E. (1986). Evaluating the performance of paper people. Journal of Applied Psychology, 71, 654-661. (use this in discussion, under limitations of our methodology) Parsons, T, & Bales, R. E (Eds.) (1955). Family, socialization, and interaction process. Glencoe, IL: Free Press. Tannen, D. (1986). That's not what I meant: How conversational style makes or breaks relationships. New York: Ballantine. Tannen, D. (1990). You just don't understand: Women and men in conversation. New York: William Morrow.

Você também pode gostar