Sex Roles‚ Vol. 39 ‚ Nos. 3/4 ‚ 1998 Meta-Analyses of Gender Effects on Conversational 1 Interrup tion : Who‚ What ‚ When ‚ Where ‚ an d How Kristin J. An derson an d Cam pbell Leap er 2 University of California‚ Santa Cruz Meta-an alyses of 43 published studies com parin g adult wom en ’s an d m en ’s in terruption s durin g con versation s were con du cted. Com bin ed sign ificance levels an d com bin ed effect sizes were an alyzed. Across studies ‚ m en were sign ifican tly m ore likely than wom en to use in terruption s. This differen ce‚ however‚ was asso ciated with a negligible effect size (d = .15). A m ore substan tial effect size (d = .33) was fou nd when studies lookin g specifically at intrusive types of interruption were an alyzed separately. Other m oderator variables were foun d to be related to gender effects on the use of in trusive in terru ptio n s. Most n otably‚ repo rts of gen d er d ifferen ces in in tru sive interru ptions were m ore likely an d larger in m agnitude when either wom en (versus m en) were first au thors ‚ participan ts were observed in natu ralistic (versus laboratory) settings ‚ or participants were observed interactin g in grou ps of three or m ore person s (versus in dyads). These results lend support to a con textual-in teractive m od el of gen der that em ph asizes the im portan ce of situation al m oderators on gender-related variation s in social behavior. O ne of the most widely conteste d areas of ge nde r and language is whether men inte rrupt their conve rsation partne rs more often than do wome n (e .g. ‚ see Arie s‚ 1996; Crawford ‚ 1995; Tanne n ‚ 1983 ‚ 1994) . Z imme rman and West (Zimme rman & We st‚ 1975; We st & Zimmerman ‚ 1983) were among 1 This research was supported by research grants from the Academic Senate (No. 503035-19900) and the Social Sciences Division (No. 443060-09523) of the University of California‚ Santa Cruz to the second author. Jennifer Jipson‚ Jill Denne r‚ and the anonymous revie wer are thanked for their comments on earlier drafts. 2 To whom correspondence should be addre ssed at Departme nt of Psychology‚ Unive rsity of California‚ Santa Cruz‚ CA 95064; e-mail: cam@ cats.ucsc.edu 225 0360-0025/98/0800-022 5$15.00/0 Ó 1998 Plenum Publishing Corporation 226 An derson an d Leaper the first re se arche rs to inve stigate the topic by obse rving casual conve rsation be tween same and mixed-ge nde r pairs. Their work followe d the mode l put forth by Sacks ‚ Sche gloff ‚ and Jeffe rson (1974) that conve rsations are orderly‚ whereby one person talks at a time and transitions betwee n speakers occur at the pote ntial end of a unit type (e .g.‚ clause ). That is ‚ when a speake r has appe are d to reach a possible comple tion point ‚ a change in spe ake r can le gitim ate ly occur. An inte rruption occurs whe n a se cond spe aker be gins to speak before a pote ntial transition point occurs (se e Z immerman & West‚ 1975) . Although fe w interruptions occurred in Z immerman and West’s (1975) recorded conve rsations ‚ the patte rn of interruptions betwee n same - and mixe d-ge nde r interactants was differe nt. In same -gender inte ractions ‚ the distribution of inte rruptions was fairly e qually divide d be twe e n spe ake rs. In contrast ‚ in mixe d-ge nde r inte ractions men made nearly all of the inte rruptions of wome n’s spe ech. Z immerman and We st (1975) conclude d that men ’s dominance in conve rsation via inte rruption mirrors the ir dominance in conte mporary western culture . Inte rruption is “ a de vice for e xe rcising powe r and control in conve rsation ” because it involve s “ violations of speake rs’ turns at talk” (We st & Z imme rman ‚ 1983 ‚ p. 103) . Inasmuch as men typically e njoy greate r status and powe r than do wome n in most socie tie s‚ the inference is that men are more like ly than wome n to assume they are e ntitle d to take the conve rsational floor. Since We st and Z immerman’s e arly work‚ many studie s have re plicate d their findings (e.g.‚ Bohn & Stutman ‚ 1983; Brooks ‚ 1982; Case ‚ 1988) . However‚ many othe r studie s e ithe r have found no ge nde r diffe re nces (e.g.‚ Carli ‚ 1990; Dindia ‚ 1987; Johnson ‚ 1994) or have found that women inte rrupt more than men (e .g. ‚ Kenne dy & Camden ‚ 1983; Nohara ‚ 1992) . In a recent narrative re view of article s publishe d betwee n 1965 and 1991 ‚ Jame s and Clarke (1993) conclude d that the re is little e vide nce that men inte rrupt more than wome n in either same - or mixed-ge nde r inte ractions. Aries (1996) dre w similar conclusions in her narrative revie w of the literature . James and Clarke furthe r speculate d that women ’s and men ’s inte rruptions may differ in the ir function (se e also Tanne n ‚ 1994 ‚ for a similar point). O ne helpful strate gy has bee n to distinguish be tween interruptions and ove rlaps. Tanne n (1994) de fine d an inte rruption as when a se cond spe aker usurps anothe r spe aker’s right to continue spe aking by taking the conve rsational floor in the abse nce of any evide nce that the othe r spe aker inte nde d to re linquish the turn. In contrast ‚ an ove rlap is whe n a se cond spe aker begins speaking at what could be a transition-re le vant place such as the e nd of a clause . Women and members of cultural communitie s she describe s as “ high involve ment ” often overlap with each othe r in spee ch Meta-An alyses of Gen der Effects on In terr uption 227 as a way of demonstrating coope ration and enthusiasm. Tanne n propose s that “ coope rative overlapping [is] supportive rathe r than obstructive ‚ evide nce not of dom ination bu t of participation ‚ not powe r ‚ bu t the paradoxically re late d dime nsion ‚ solidarity ” (p. 62) . She argue s that by assuming that inte rruption is a monolithic conve rsational de vice “ we are forced into a position that claims that high involve ment speake rs‚ such as blacks and Jews and ‚ in many circumstance s‚ wome n ‚ are pushy‚ aggre ssive ‚ or inconside rate or foolishly noisy” (p. 73) . A META-ANALYTIC APPROACH TO THE LITERATURE ON INTERRUPTIONS In the pre se nt revie w‚ meta-analyse s were conducte d in orde r to add r e s s t h e c on t r o ve r s y o ve r wo m e n ’s a n d m e n ’s i n te r r u p t io n s in conve rsations. Meta-analysis is the statistical integration of results of inde pe nde nt studie s. It provide s a single set of numbe rs that de scribe and summarize the re sults of inde pe nde nt pie ces of rese arch. Although narrative lite rature re views are useful in summarizing the re sults of a rese arch domain ‚ meta-analytic revie ws are a useful tool be cause the y simultane ously take into account the significance le vel‚ the sample size ‚ and the e ffect size of e ach individual study in order to produce combine d significance le ve ls and effe ct sizes. Me ta-analysis also allows for the statistical analysis of pote ntial moderator variable s. O ne of the pote ntial mode rators of ge nde r e ffects on the use of inte rruptions is how inte rruption is operationally de fine d in individual studie s. The pre sent meta-analyse s distinguishe d betwee n three definitions of inte rruptions: (1) those that were either unde fined or broadly defined in the original study; (2) those that e xplicitly e xclude d back channe ls and minimal listening response s (e .g. ‚ “ uh-huh ” ); and (3) those de fine d as intrusive ‚ and sugge st a dominating motivation on the part of the interrupte r. O ne type of interruption often defined as intrusive is the “ successful interruption ” whereby the interrupting speake r succe ssfully take s ove r the conve rsational floor. Pre vious work has associate d successful interruptions with the manife station of dominance (e .g. ‚ Aries‚ 1996; Kollock ‚ Blumste in ‚ & Schwartz ‚ 1985; Natale ‚ Entin ‚ & Jaffe ‚ 1979; Smith-Lovin & Brody‚ 1989; se e James & Clarke ‚ 1993 for a contrasting vie w). Succe ssful inte rruptions ‚ for example ‚ have bee n rate d by pe ople as more domine e ring than unsucce ssful one s (McLaughlin ‚ 1984) . Because men have more ofte n bee n associate d with dominanc e in conve rsation al inte rruption s‚ we hypothe size d that me n would be found to make more intrusive inte rruptions than women. 228 An derson an d Leaper Other Possible Moderators of Interruption s In addition to the ope rational de finition use d ‚ seve ral othe r variable s may moderate the like lihood and magnitude of ge nde r difference s in inte rruptions. Jame s and Clarke (1992) conclude d their narrative re view of inte rruptions by stating ‚ “ Lastly‚ the ways in which the results of studie s may have be en affe cted by such subje ct and situational variable s as age ‚ degre e of intimacy ‚ size of group ‚ and type of interactional conte xt remain uncle ar.” (p. 295) O the r publication characte ristics such as the year of the study or the author ’s ge nde r may also mode rate the like lihood of gende r effe cts (Le ape r‚ Ande rson ‚ & Sande rs‚ 1998) . Publication Characteristics. Year of publication and the first author ’s ge nde r are two publication characte ristics that were e xamine d as possible mode rator variable s. Publication ye ar may act as a mode rator variable to the extent that either historical change s in gende r equality or change s in how rese archers conduct their re se arch have had an impact on the like lihood of finding ge nde r differe nces. Some prior meta-analyse s have found a de crease over time in the numbe r of studie s finding gende r differe nces on measure s such as mathe matics (Hyde ‚ Fenne ma‚ & Lamon ‚ 1990) and verbal ability (Hyde & Linn ‚ 1988) . O ve r the ye ars ‚ the pe rformance of wome n and men has be come more similar in these areas — pe rhaps the result of increased opportunitie s for women. In their narrative re view of inte rruptions ‚ James and Clarke (1993) re port that the gende r gap be tween wome n and men in initiating interruptions has de creased or possibly reversed ove r the ye ars be tween 1965 and 1991. In the pre se nt meta-analysis ‚ we include d studie s that span over a thre e decade pe riod during which many political and cultural change s challe nge d traditional ge nde r role s. There fore ‚ we expe cted that ge nde r differe nces in interruptions would decrease ove r time . Howe ve r ‚ we also note the pote ntial counte rvailing influe nce of methodological advance s in observational research that have occurred over the years (e.g. ‚ Bake man & Gottman ‚ 1986; Bakeman & Q ue ra‚ 1995) . For e xample ‚ Hall (1978) found that ge nde r diffe re nces in nonve rbal de coding were more common in more recent studie s. She proposed that recent improve ments in measuring te chnique s may account for this patte rn rathe r than historical change s in gende r role s. Anothe r publication characteristic that we conside red was the author ’s gender. James and Clarke (1993) speculated that the author ’s gender may have an influence on whether one gender is more likely to interrupt than the other. However‚ they did not detect any corresponding patterns in the studies they reviewed. In contrast ‚ meta-analyses on other topics have found a significant relationship betwee n author ge nder and the magnitude of the gender difference (Eagly & Carli ‚ 1981; Leaper et al.‚ 1998). For instance ‚ Leaper et al. (1998) Meta-An alyses of Gen der Effects on In terr uption 229 examine d parents ’ talk to their childre n and found a significant relationship between author gender and the magnitude of the effect size with some measures. When author gender did act as a moderator ‚ men authors found gender differences more often than women authors. Thus‚ the author ’s own gender may reflect some sort of researcher bias (Beall‚ 1993). In the present metaanalyses‚ we explored whether or not author gender moderate s the like lihood ‚ the magnitude ‚ or the direction of gender effects on interruption. Aspects of the Interactional Setting as Possible Moderators of Interruptions We examine d several aspe cts of the interactive conte xt as pote ntial moderators of interruptions. Recent conte xtual-inte ractive mode ls of ge nder-typing (e.g.‚ Beall ‚ 1993; Deaux & Major ‚ 1987; Leaper et al.‚ 1998) sugge st that the incide nce and magnitude of gende r effects may large ly depend on aspects of the particular situation. In contrast‚ essentialist models of gender argue for the existence of inhere nt differences between women and men that are responsible for observe d variations in behavior. A contextual-interactive model of gende r differences in interruption would be supporte d if aspects of the interactive setting — such as the characte ristics of participants ‚ the task‚ or the setting — were found to moderate the like lihood of gender effects. Characteristics of the Participan ts. First‚ we conside red factors associate d with the relationship be tween the inte ractants such as ge nder composition ‚ group size ‚ and the familiarity of the interactants. James and Clarke (1993) found that there is a tendency for men to inte rrupt more often in mixed-ge nder than in same-gender interactions. The y spe culate that if the major de terminant of interruption is simply having more status or power than one ’s conve rsational partne rs‚ then gende r effe cts on interruptions should be large r in mixed-ge nde r than same -gende r inte ractions. In contrast ‚ Aries (1996) inferre d from her review that there was no patte rn of gender difference in interruption related to the ge nder composition of the group. We sought to clarify this matter in the prese nt meta-analyse s. Group size was anothe r pote ntial moderator variable e xamine d here . We expe cted that men would be more like ly to inte rrupt in large r groups than in dyads. If interrupting is a demonstration of dominance ‚ the ne ed to display dominance would be gre ater in a more public situation with many witne sse s than in one -to-one interactions in which pressure to act more stere otype d may be le ssene d. In the ir narrative revie w of inte rruptions ‚ James and Clarke (1993) inferred a slight tende ncy for men to interrupt more than women in mixe d-ge nde r groups than in dyads. We sought to confirm this interpre tation with the pre se nt meta-analyse s. 230 An derson an d Leaper Anothe r possible mode rator variable that we inve stigate d was the familiarity of the inte ractants. Prior work has found that unacquainte d pe ople are more like ly than acquainte d pe ople to rely on ge nde r-ste reotype d e xpe ctations to guide the ir be havior ‚ whe re as acquainte d pe rsons such as close friends and intimate s are more apt to re ly on individual characte ristics (Drass ‚ 1986; Wood & Karten ‚ 1986) . Exte nding this rationale to interruptions ‚ men may be more like ly than wome n to interrupt most e spe cially in unacquainte d inte ractions be cause more dominating behavior may be e xpected of the m. Conve rse ly‚ to the exte nt that strange rs may fee l more social pre ssure to be polite ‚ there might be fewer ge nde r difference s in inte rruption be tween strange rs than betwee n frie nds or romantic partne rs. These two views were teste d in the pre se nt revie w whe n we compare d inte rruptions be tween strange rs‚ frie nds ‚ and romantic partne rs. Characteristics of the Activity Setting. In addition to examining characteristics of the interactants ‚ we also examine d the nature of the task and the setting as potential moderators of women’s and men’s interruptions. First‚ we compare d whe ther the obse rvation took place in a laboratory or in a naturalistic setting. Although James and Clarke (1993) saw no consiste nt gender difference in inte rruptions based on observational se tting whe n the y reviewed the literature ‚ an effect could emerge in a meta analysis. We expected that if men were found to make more intrusive interruptions than women‚ they would be more like ly to occur in naturalistic settings. In laboratory settings ‚ polite ness norms may be more salie nt and thereby reduce the like lihood of domine ering behaviors such as intrusive interruptions. Second‚ we compare d studies that examined instrumental ‚ expressive ‚ or unstructure d topics in order to determine whether the type of activity moderated interruptions. James and Clarke (1993) suggeste d that to the extent that a given topic is perceived to be women’s or men’s presumed area of expertise ‚ either the woman or the man may feel more of an “ authority” in that area and ‚ consequently ‚ may fee l more justifie d in making inte rruptions. Aries (1996) proposed that during unstructure d discussions ‚ women frequently made affiliative overlaps in their conve rsations with each other. Following these views‚ we hypothesize d that women would make more interruptions during expressive topics ‚ while men would make more interruptions during instrumental topics. Based on Aries’ interpretation ‚ we expected that women would make more interruptions during unstructured than structured discussions. Finally‚ we e xamine d the le ngth of observation as a pote ntial moderator. Prior re se arch has sugge ste d that longe r observation time s tend to be a more valid measure of social interaction qualitie s (see Aries‚ 1996) . With longe r obse rvations ‚ an observer is more apt to see pe ople ’s stylistic variations. Therefore ‚ we e xpe cted that gende r diffe re nces in interruptions would be large r as the le ngth of the observation increase d. Meta-An alyses of Gen der Effects on In terr uption 231 To summarize ‚ our meta analysis addre ssed the “ Who ‚ What ‚ Where ‚ When ‚ and How? ” of ge nde r e ffects on conve rsational inte rruption. First ‚ the most pressing issue in the re search literature has be en the “ Who? ” que stion: Do men and wome n diffe r in the ir like lihood of inte rruption? Howe ver‚ the answer to this que stion may de pe nd on the “ How? ” que stion. Dete cting ge nde r diffe rence s in interruption de pe nds large ly on how the inte rruption is de fine d. Additionally ‚ gende r differe nces in inte rruption may depend on the “ What? ‚” “ When? ‚” and “ Whe re ? ” aspe cts of the inte ractive context. What activitie s are most like ly to be associate d with ge nde r difference s in inte rruption? For instance ‚ is there a differe nce betwee n structure d and unstructure d tasks? When are inte rruptions like ly to occur? In othe r words ‚ do e ithe r the year of the study or the length of observation mode rate the like lihood of ge nde r effe cts? Finally ‚ where are gende r diffe rence s like ly to take place ? Do ge nde r effe cts on interruption diffe r in naturalistic ve rsus research laboratory se ttings? METHOD Literatu re search Forty-thre e publishe d studie s e xamining women ’s and men ’s use of inte rruptions were colle cted through a varie ty of sources. Ten studie s were counte d twice be cause they had more than one useable analysis resulting 3 in a total of 53 hypothe sis te sts. Most of the studie s we re ide ntifie d through compute rized searches of the Psychological Abstracts. We also reviewed relevant studie s cited in the se article s and in Jame s and Clarke ’s (1993) and Arie s (1996) re views. The date s of publication for the colle cted studie s range d from 1965 to 1996. Three selection criteria were used: (1) O nly studie s that tested for ge nde r effe cts on interruptions were include d. (2) Only studie s using quantitative obse rvational measure s were include d. (3) Only studie s publishe d in e ithe r re se arch journals or books were include d. Although publishe d studie s may be more biase d than unpublishe d studie s toward reporting significant e ffe cts‚ this was not indicate d with our sample of studie s (see description of fail-safe te st in Results). 3 Spe cifically‚ Carli (1990) ‚ Jones e t al. (1995) ‚ Natale et al. (1979) ‚ Roge r & Schumacher (1983) ‚ and Smeltze r and Watson (1986) separately analyzed two differe nt measures of interruption. Kollock e t al. (1985) analyzed both friends and romantic partners. Bilous and Kraus (1988) ‚ Nohara (1992) ‚ Rede ker and Maes (1996) and Simkins-Bullock & Wildman (1991) separately examined interruptions used in same- and mixed-gende r pairs. This information is detailed in Table I. 232 An derson an d Leaper There were se ve ral categorie s of studie s that had to be e xclude d from the se meta-analyse s: First‚ studie s with only qualitative analyse s and no infe rential statistics could not be include d (Goldbe rg‚ 1990; Murata ‚ 1994; Thomas ‚ Roge r‚ & Bull ‚ 1983; Woods ‚ 1989) . Second ‚ studie s that did not report a sample size were e xclude d (Be attie ‚ 1977; Murray & Cove lli ‚ 1988; Willis & Williams ‚ 1976) . Third ‚ studie s that include d only me n (Rim ‚ 1977; Thimm ‚ Rademacher‚ & Kruse ‚ 1995) ‚ only wome n (Fe rguson ‚ 1977) ‚ or didn ’t compare women and men (Drass‚ 1986; Hawkins ‚ 1991) were e xclude d. Fourth ‚ studie s e xamining pe rceptions of othe rs who inte rrupt were e xclude d (Chambliss & Fee ny‚ 1992; Hawkins ‚ 1991; Robinson & Reis ‚ 1989) . Finally‚ studie s that did not obse rve face -to-face inte ractions were exclude d (e .g.‚ talking on the te lephone ; Mott & Pe trie ‚ 1995) . Moderator Variables Several variable s that may moderate the magnitude of e ffects associated with women’s and men’s interruptions were examined. The characte ristics for each mode rator variable associate d with e ach study are presented in Table I. Participan t Demograph ic Characteristics. Studie s including sample s othe r than middle -class ‚ Europe an Americans were too few to permit te sting for e thnicity or e conomic status as pote ntial mode rator variable s. Also ‚ although the re was variation in the geographical re gions of the differe nt studie s‚ the e ffects in the prese nt study did not de monstrate any consiste nt or meaningful patte rns across the differe nt meta-analyse s. Conse que ntly‚ the se re sults are not pre se nte d. Interruption Classification. The operational definitions of interruptions in the 53 hypothe sis tests varied widely. When considering all studie s collapsing across operational definitions ‚ we will use either the term total interruptions or overall interruptions. Otherwise ‚ interruptions were divide d into three categories indicating increasing specificity of the definition: The first category of interruptions were general interruptions and consisted of studies in which interruptions were either not specifically defined or the operational definitions include d broade r criteria such as affiliative overlaps ‚ unsuccessful interruptions ‚ and ‚ in some instances ‚ back channels and minimal response s (e.g.‚ “uh-huh ”). Aries (1996) reports that one way to classify interruptions is to make explicit whether the definition has exclude d or included back channe ls and minimal response s. Correspondingly‚ the second cate gory of inte rruptions we re those in which back channels were excluded in the definition of an interruption ‚ but were still broadly defined. For instance ‚ the definitions include d in this category may have allowed for affiliative overlaps. The third category include d intrusive interruptions which Meta-An alyses of Gen der Effects on In terr uption 233 indicate d an attempt on one speaker to usurp the other spe aker’s conversational turn. Some of the studie s in this category measured what were calle d “successful” interruptions whereby one speake r stops talking as a result of another speaker’s incursion. The two authors were able to classify interruption categories with high reliability ( k = .90). According to Bakeman and Gottman (1986) ‚ kappa value s above .75 reflect “excellent” levels of agreement. Other Moderator Variables. The following 8 moderator variable s were also examine d: (1) The year of the study re fe rs to the year the study was publishe d. (2) Author gen der refers to whe the r the first author of the study was a woman or a man. (3) G ender com position refers to whether the participants were observed in same- or mixe d-ge nde r inte ractions ‚ or both (the latte r case include s studie s that did not analyze same- and mixe d-ge nde r groups se parate ly) . (4) G rou p size re fe rs to whethe r participants were observed e ithe r in pairs or in groups of 3 or more . (5) Fam iliarity pertains to whether the interactants were strange rs‚ friends‚ romantic partne rs‚ or anothe r type of familiar re lationship. (6) The observation al setting contrasted whether obse rvations took place in either a re se arch laboratory or a naturalistic se tting. (7) The activity structu re distinguishe d betwee n situations that were e ithe r unstructure d ‚ instrume ntal (e .g. ‚ a proble m-solving task) ‚ e xpre ssive (e .g.‚ a se lf-disclosure task) ‚ or othe rwise unclassifiable (e.g.‚ including a combination of tasks). (8) Finally‚ the length of observation (in minute s) was take n into account. RESULTS Statistical Analyses We employe d Mulle n’s (1989) meta-analysis software to carry out the statistical analyse s. Mulle n ’s program provide s the following information for the meta-analysis of effe ct sizes: diffuse comparisons ‚ combinations ‚ focused comparisons ‚ and blocking by leve ls of a moderator. Diffuse Com parison of Significan ce Levels and Effect Sizes. This procedure tests for the ove rall variability around the ave rage study outcome . Significant tests for the diffuse comparisons of significance levels or e ffect sizes indicate that the significance le ve ls of the include d studie s or the strength of e ffects were significantly heteroge neous and may be thought of as having be en sample d from diffe rent populations (Mulle n ‚ 1989). Additional ly‚ the diffu se com parison te sts compu te a fail-safe nu m be r of unpublishe d studie s with null results that would be ne eded to counte ract any obse rve d e ffects. This numbe r is use ful give n the possible bias against publishing nonsignificant re sults. Beattie (1981) b Bilous & Kraus (1988) b Bilous & Kraus (1988) Bohn & Stutman (1983) Brooks (1982) Campbell e t al. (1992) c Carli (1990) c Carli (1990) Case (1988) Craig & Pitts (1990) Dabbs & Ruback (1984) De Boe r (1987) Dindia (1987) Eakins & Eakins (1976) Fallon & Guo (1994) Frances (1979) Johnson (1994) c Jones et al. (1995) c Jones et al. (1995) Kennedy & Camde n (1981) d Kollock et al. (1985) d Kollock et al. (1985) LaFrance & Carmen (1980) Leet-Pellegrini (1980) Leffle r et al. (1982) Makri-Tsilipakou (1994) Marche & Peterson (1993) Margolin & Wampold (1981) Martin & Craig (1983) McLachlan (1991) McMillan et al. (1977) c Natale et al. (1979) Author p = .5 t = 1.30 t = 6.18 2 c = 16.02 p < .001 t = 2.83 p = .5 t = 1.48 p < .05 F = 2.36 t = 1.44 2 c = 7.65 p = .5 p = .5 p = .5 p = .5 p = .5 p = .5 p = .5 2 c = 10.21 p < .05 p = .5 F = 2.20 p = .5 p = .5 p < .0001 p = .5 F = 3.07 p = .5 p = .5 t = 6.34 p = .03 Stat 0 .34 ¯ 1.62 2.14 .37 .66 0 .40 1.22 ¯.89 ¯.68 1.91 0 0 0 0 0 0 0 ¯ 1.28 ¯.54 0 .37 0 0 1.07 0 .58 0 0 1.78 .45 d 45 60 60 30 294 104 118 118 10 38 50 16 60 9 20 88 141 50 50 35 40 30 72 140 56 62 30 78 40 44 98 72 N a I B B G G G I B G I G I I G G B G I G B B B I B G I I G B B I B M U U M W M W W W M M U W W M W W W W W M M W W W W W W W M W M M M S S M M B B M M S S B M M B B S S M S M S B B M B M B S B B 5 2 2 2 18 8 2 2 10 4 5 2 2 9 2 2 3 2 2 3 2 2 2 2 2 3 2 2 2 2 6 2 M S S S M S S S M M S F S M M S S S S F P F S S S M F P S M S S N L L L L L L L L O L N L N L L L L L L N N L L L N L L L L L L I I I I I I I I I I U U U I I U I E E I E E M I I U U E U I I U 45 10 10 10 2880 n/a 10 10 180 n/a 20 60 30 n/a 3 14 9 9 9 60 n/a n/a 7 10 30 585 20 20 4 n/a 30 30 Oper. De f. Author Partne r Size Familiar Setting Activity Length Table I. Summary of Studie s Testing for Women ’s and Men’s Interruptions c p = F = p = F = p = p = Z = p = p = p = t= p < p = p = 2 c = 2 c = p < F = p = p = p < .5 .23 .5 2.54 .5 .5 2.0 .5 .5 .5 1.36 .001 .5 .5 2.94 3.46 .05 5.15 .5 .5 .03 0 .07 0 .85 ¯ 0 0 1.63 0 0 0 .47 ¯ 1.20 0 0 ¯.41 ¯.45 .79 ¯.74 0 0 1.48 72 168 24 16 8 10 10 56 36 36 36 36 39 39 72 72 20 40 36 64 10 I B G I I B B I B I B G G G G I B G B G I M M M W W W W M M M M M W W M M M M W W W B M B S M M S M S S S M S M M M S S B B M 2 6 4 2 2 5 5 2 2 2 2 2 2 2 4 4 5 2 2 2 2 S S S S S M M S S S S M S S M M S S S S S L L L L L O O L L L L L L L L L L L L L L U I I I I I I I I I I I I I I I U U M U U 30 120 20 10 10 n/a n/a 8 8 8 15 20 15 15 75 75 120 24 2 20 12 a Stat = Statistic. d = Cohen ’s d. N = sample size. Oper. Def. = Ope rational De fintion (G = Ge neral‚ B = Explicitly e xcludes backchanne ls‚ I = Intrusive ). Author = First author ’s gender (W = woman ‚ M = man‚ U = unclear). Partne r = Gende r Composition (S = same-gender ‚ M = Mixe d-ge nde r ‚ B = both). Size = Group size. Familiar = Familiarity ( S = strangers ‚ F = friends ‚ P = dating/partne rs ‚ M = mixed/other). Setting = Observational setting (L = laboratory‚ N = naturalistic ‚ O = other). Activity = Activity type (I = instrumental‚ E = e xpressive ‚ M = mixe d‚ U = unstructured). Le ngth = le ngth of obse rvation in minutes (n/a = information not available ). b Study is include d in meta-analysis more than once because it reported results for the same sample in different conditions (see Method section). An adjusted N was used in the analyse s (see Method section). c Study is included in meta-analysis more than once because it reported results for the same sample using more than one operational definition of interruptions. An adjusted N was use d in the analyses. d Study is include d in meta-analysis more than once. Because it reported results for more than one sample ‚ it is treated as two separate samples. Natale et al. (1979) Neme th et al. (1976) Ng e t al. (1995) d Nohara (1992) d Nohara (1992) b Redeker & Maes (1996) b Redeker & Maes (1996) Roger & Ne sshoever (1987) c Roger & Schumacher (1983) c Roger & Schumacher (1983) Rogers & Jones (1980) Shaw & Sadle r (1965) d Simkins-Bullock & Wildman (1991) d Simkins-Bullock & Wildman (1991) c Smeltzer & Watson (1986) c Smeltzer & Watson (1986) Smith (1977) Street & Murphy (1987) Trimboli & Walker (1984) Welkowitz et al. (1984) West & Z imme rman (1983) 236 An derson an d Leaper Combinations of Significance Levels and Effect Sizes. Combinations of significance levels and effect sizes provide information on the typical study outcome across studie s. Combinations of significance levels e stimate the probability that the p value s of the sample d studies would be obtaine d if the null hypothesis were true. This procedure uses the standard normal deviate Z (i.e.‚ mean = 0; standard deviation = 1) for significance level. Combinations of effect sizes estimate the overall magnitude of the effect size across the sample d studie s. Both unweighte d and weighted (by sample size) combinations of significance levels and effect sizes are reported. Fisher’s z and Cohen’s d are two indices of effect size that are reported here. Cohen (1977) characte rized effect sizes as “ small” when d = .2‚ “medium ” when d = .5‚ and “large ” when d = .8. Thus‚ an effect size of d = .2 and above may be viewed as meaningful ‚ whereas an effect size below d = .2 is conside red trivial in magnitude . Focused Comparison s of Significance Levels and Effect Sizes. Focused comparison tests estimate how well a particular mode rator explains variability across studie s. Focused comparisons of significance le vels and effect sizes test for the re lationship be tween the mode rator and variability across studie s in e ithe r statistical probabilitie s or effect size s‚ re spectively. Focused comparison tests for categorical moderator variable s were carried out using the Z statistic. The effects of continuous moderator variable s (ye ar of study and length of obse rvation) were tested by using regression analyse s with study sample size as a weighte d factor. Focused comparison te sts are calculate d only for unweighted scores (Mulle n ‚ 1989) . The re is no accepted method for computing this type of test for weighte d score s. Blockin g within Levels of a Moderator. This technique classifies or blocks studies by levels of a moderator variable allowing for combinations of significance levels and effect sizes (described above) at each specific level of a moderator. Comparison tests between each level are also computed. Blocking was carried out for weighted as well as unweighted scores. Results using both types of scores are presented in the tables. However‚ when different effects occurred using unweighted versus weighted scores‚ the results from the weighted scores will be given priority in the text because they adjust for the sample size in each study. The results from the analyse s are summarize d in Table II for overall interruptions (i.e.‚ collapsing across operational de finitions) and in Table III for intrusive inte rruptions. Each table breaks down the numbe r of studie s (k) ‚ each study’s sample size (N) ‚ significance levels and effect sizes (Fisher’s z and Cohen’s d) for each of the categorical moderator variable s in re lation to gender e ffects on the use of interruptions. Effe ct size s with a positive sign indicate that men interrupte d more than women. The results te sting for the correlations between the continuous moderator variable s (year of study and length of observation) are reporte d in the text below. Meta-An alyses of Gen der Effects on In terr uption 237 OVERALL INTERRUPTIONS Diffuse comparisons were significant for the combine d significance 2 2 le ve ls ‚ c (52) = 175.92 ‚ p < .001 ‚ and the combine d e ffect size s‚ c (52) = 232.15 ‚ p < .001 ‚ indicating a significant amount of variability among the 52 studie s (hypothe sis tests). The fail-safe numbe r was 31 ‚ indicating that it would take that many unpublishe d studie s with null re sults to re ve rse any obse rve d effe cts. Thus ‚ the use of only publishe d studie s did not appe ar to bias the overall dire ction of findings. As see n in Table II‚ the combination of significance le vels was significant using either unwe ighte d or weighte d score s. Across studie s‚ men were significantly more like ly than wome n to inte rrupt. Howeve r‚ the combination of e ffect sizes re vealed that the magnitude of the difference was negligible (d < .2) using e ithe r unweighte d or weighte d scores. Operation al Definition Focused Comparison Tests. Focused comparison tests using unweighted scores were carried out to test the effect of operational definition (general vs. back channels excluded vs. intrusive). The results we re not significant for either combined significance levels‚ Z = .99 ‚ n.s.‚ or combine d effect sizes‚ Z = .99 ‚ n.s. Blocking. Although the focused comparison tests did not reveal a significant overall e ffect for operational definition as a moderator‚ the blocking analyses suggested a different story. As seen in Table II‚ statistically significant gender differences with meaningful effect sizes appe ared when studies looked specifically at intrusive interruptions. In contrast ‚ measures of combined significance levels and combine d effect sizes were negligible when studies looke d at interruptions that either were generally defined or were limited to excluding only back channels. (However‚ the combine d significance level for general interruptions was significant using weighted but not unweighte d scores.) To reite rate ‚ the intrusive inte rruption cate gory was the only measure that was associated with a significant combined significance level and a meaningful combined effect size using e ither unweighte d or weighted scores. Conse quently‚ subsequent analyses were carried out to test the effects of the other moderator variables on gender differences in the use of total interruptions (i.e.‚ collapsing across operational definitions) as well as in the use of intrusive interruptions in particular. The effects of the moderator variables on total interruptions are summarized in Table II but are not mentioned any further in the text. The results from the combinations of significance levels and effect sizes for the 17 studies specifically examining intrusive interruptions are presented in Table III and are described below. Dyad Group (3 or more) Group Size Same -gende r Mixed-ge nder Both 35 18 17 21 15 22 Me n Ge nder Composition 28 Women First Author ’s Gender Gene ral Inte rruptions Non-Back Channel Interruption Intrusive Interruptions 18 18 17 53 Ove rall Ope rational Definition k Moderator Variable 1796 1262 645 1256 1157 1141 1781 1156 1043 859 3058 N c 2.63 bd .67 a .27 a 1.15 a 2.26 ac 3.34 ae ¯ .26 b 1.14 ab ¯ .03 a 2.53 b 2.07 Z a .04 a .10 a .07 a .03 a .08 a ¯.01b .11 a .05 a .02 a .12 a .06 Fishe r’s Z Unweighted .08 .20 .14 .07 .16 ¯ .02 .23 .09 .04 .23 .12 Cohen ’s d e .73 e 4.20 ¯.32 e 3.36 d 2.40 .46 4.41 d e 3.00 .09 e 4.05 3.85 Z .03 .13 .01 .18 .10 .00 .12 .08 0 .16 .07 Weighte d Fisher ’s Z Combinations of Significance Levels and Effe ct Sizes Table II. Effects of Mode rator V ariables on Women ’s and Men’s Ove rall Inte rruptions .06 .26 .02 .16 .20 00 .25 .15 ¯.01 .33 .15 Cohen’s d Other 3 58 2798 202 111 118 722 2107 1.98 a .33 a c 1.47 ab 2.25 ac ¯ .21 a .04 a .65 a .10 a .04 a .18 a .06 a .01 a .05 a .07 a .21 .08 .37 .12 .02 .09 .14 d ¯ 1.22 3.18 .01 e ¯ 1.18 .77 d 2.75 2.84 ¯.30 .15 .01 .08 ¯.07 .10 .07 ¯.60 .30 .03 .17 ¯.14 .20 .13 Unweighted scores with diffe re nt subscripts are significantly different (p < .05). (There is not an accepted comparison test for weighted scores.) The corre lation betwee n the effe ct size d and year of study was not significant for total interruptions (r = ¯ .01‚ n.s.). The correlation be tween the e ffect size d and length of observation was not significant (r = .11‚ n.s.). b p < .10. c p < .05. d p < .01. e p < .001. a Lab Naturalistic 44 6 4 2 13 Friends Romantic Partners Other Obse rvational Setting 34 Strangers Familiarity 240 An derson an d Leaper INTRUSIVE INTERRUPTIONS First Author’s G ender Focused Com parison Tests. The focuse d comparison te sts using unweighte d score s re ve ale d that first author ge nde r acte d as a significant mode rator in relation to both combine d significance le ve l‚ Z = 3.19 ‚ p < .001 ‚ and e ffect size ‚ Z = 3.00 ‚ p < .01. Gende r differe nces were more like ly in studie s with wome n as first authors. Blockin g. The blocking analyse s for author gende r are summarize d in Table III. The combine d signific ance le ve l was significant in wome nauthore d studie s. Wome n-authore d studie s were also associate d with a mode rate combine d e ffect size whe n weighte d score s were used (d = .54) . In contrast ‚ men-authore d studie s were associate d with a nonsignificant combine d significance leve l and a small combine d effe ct size when we ighte d score s were used (d = ¯.21) . The ne gative direction of the combine d e ffect size for men-authore d studie s indicate s there was actually a tende ncy in the se re ports for wome n to use intrusive inte rruptions more than men. Publication Date There was a nonsignificant corre lation associate d with intrusive interruptions and the publication date ‚ r(17) = ¯.28 ‚ n.s. The small magnitude and ne gative dire ction of the corre lation sugge st a slight te nde ncy for smalle r gende r diffe rence s in more re cent studie s. G ender Com position Focused Com parison Tests. Contrary to e xpe ctation ‚ ge nde r composition did not act as a significant moderator. The focuse d comparison te sts were not significant when e ithe r combine d significance leve ls or combine d effe ct size s were analyze d. Blockin g. As se en in Table III‚ when same- and mixe d-ge nde r interactions we re analyze d se parate ly‚ combine d significanc e leve ls were all nonsignificant. When weighte d scores were use d ‚ the re was a small combine d e ffe ct size indicating men used more intrusive inte rruptions than wome n during same-ge nde r interactions (d = .24) . The magnitude of diffe rence during mixed-ge nde r inte ractions was negligible (d = .11) . 12 5 11 6 13 3 1 Group Size Dyad Group (3 or more) Familiarity Strangers Familiar Observational Setting Lab Naturalistic Othe r 250 487 50 72 698 123 38 596 263 544 315 c 3.42a d .21b .00b .00b 1.50a b 3.74b d .00c 2.19a b 1.30a 1.33a b 2.61abc 1.22a .87a 2.42a c 3.47a d 1.35 b ¯ 2.55 Z .34a ¯.03b .00b .00b .08a .45a .00a .11a .12a .11a .13a .12a .08a .16a .18a .11 ¯ b .12 Fisher ’s Z Unwe ighted .70 ¯.06 .00 .00 .16 .94 .00 .23 .24 .22 .27 .25 .16 .32 .36 .22 ¯ .23 Cohen’s d b d d 3.59 .63 .00 .00 d d 3.17 d 3.51 .00 d 3.67 b 1.55 1.00 d 4.74 1.36 1.17 d 3.86 5.01 1.25 ¯ 3.96 Z .36 .03 .00 .00 .15 .37 .00 .19 .09 .06 .31 .12 .06 .28 .27 .11 ¯ .16 Fisher ’s Z We ighted .73 .05 .00 .00 .31 .76 .00 .38 .19 .13 .63 .24 .11 .56 .54 .21 ¯ .32 Cohen ’s d b Unweighted scores with different subscripts are significantly different (p < .05). (There is not an accepted comparison test for weighted scores.) p < .05. c p < .01. d p < .001. a 6 9 1 1 5 7 5 Gende r Composition Same-ge nder Mixe d-gende r Both Activity Unstructure d Instrumental Expressive Mixe d/Other 524 319 10 6 First Author ’s Gender Wome n Men 190 291 378 859 17 Overall N k Mode rator V ariable a Combinations of Significance Le vels and Effect Sizes Table III. Effects of Moderator Variables on Women’s and Me n’s Intrusive Interruptions 242 An derson an d Leaper G rou p Size Focused Comparison Tests. The focused comparison test for combined significance levels was marginally significant ‚ Z = 1.47 ‚ p < .10. As see n in Table III‚ gender differe nces in intrusive inte rruptions tended to be more likely when studies observed groups (3 or more persons) than dyads. The focused comparison test for combined effect sizes was not significant ‚ Z = .25 ‚ n.s Blockin g. As se en in Table II‚ the Z for combine d significance le ve ls was significant when groups (of 3 or more persons) were analyze d and nonsignificant whe n dyads were studie d. When weighte d score s were use d for combine d e ffect sizes‚ a small but meaningful of diffe re nce occurre d in groups (d = .31) and a ne gligible differe nce occurred in dyads (d = .13) . Fam iliarity Due to the limite d range of studie s examining participants that were not strange rs‚ it was necessary to combine friends ‚ romantic partne rs‚ and other types of familiar relationships into a single familiar category. This allowe d for 11 studie s looking at strange rs and 6 studie s looking at participants who were familiar with one anothe r (see Table I for further breakdown). Focused Com parison Tests. The resulting focuse d comparison te sts using unwe ighte d scores were nonsignificant for combine d significance leve ls ‚ Z = .25 ‚ n.s.‚ as well as for combine d effe ct size s Z = .04 ‚ n.s. Blocking. As seen in Table III‚ when studie s were blocke d into the strange rs versus familiar leve ls‚ the gender effects appe ared especially strong using weighte d scores when studie s looke d at inte ractions be tween strange rs. The combine d significance levels test was significant and the combine d effe ct size was in the small-to-mode rate range (d = .38) . In contrast ‚ when studie s looke d at interactions be tween familiar partne rs‚ the combine d significance levels only approache d significance and the combine d effect size was much smalle r (d = .19) . Thus ‚ gender e ffects on intrusive interruptions may be more like ly between strange rs than familiar persons. Observational Setting Focused Com pariso n Tests. Focuse d com par iso n te sts using unweighte d scores were significant for both combine d significance leve ls ‚ Z = 2.73 ‚ p < .01 ‚ and combine d e ffect sizes‚ Z = 2.91 ‚ p < .01. The like lihood and the magnitude of ge nde r diffe re nces in the use intrusive inte ractions was greate r in naturalistic than laboratory settings. Meta-An alyses of Gen der Effects on In terr uption 243 Blockin g. As shown in Table III‚ when weighte d scores were used‚ the combine d significance level was statistically significant in both naturalistic and lab se ttings. Howe ver‚ the combine d effect size was substantially large r in studies of naturalistic settings (d = .76) than lab settings (d = .31). Thus‚ although weighted scores indicate that gender differences tended to occur in either setting ‚ they were large r in magnitude during naturalistic settings. Activity Structure The analyse s of activity structure as a moderator of gende r e ffects on intrusive interruptions were base d on 6 studie s of unstructure d activitie s and 9 studie s of instrume ntal activitie s. Only one study conside red an e xpre ssive activity and anothe r study looke d at a mixed-task activity. Focused Com parison Tests. The comparison of instrume ntal and unstructure d activitie s was significant for combine d significance leve l‚ Z = 2.52 ‚ p < .01 ‚ as well as for combine d effe ct size ‚ Z = 3.18 ‚ p < .001. As pre dicted ‚ ge nde r diffe re nces in intrusive inte rruptions were more like ly and of greate r magnitude in unstructure d activitie s. Blockin g. As se en in Table III‚ unstructure d activitie s — but not instrume nta l activitie s — we re associat e d with a sign ifican t com bine d significance le vel. Similarly ‚ substantial e ffect sizes occurred when unstructured activitie s were analyze d (d = .73) but not when instrume ntal activitie s were studie d (d = .05) . Length of Observation There was a nonsignificant corre lation be tween weighte d e ffect sizes and length of obse rvation ‚ r(17) = .24 ‚ n.s. The positive dire ction of the corre lation sugge sts a slight tende ncy for gende r diffe rence s to be large r with longe r observation time s. DISCUSSION Meta-analyses of 43 published studie s were carried out to addre ss the controversial issue of whether women or men are more like ly to interrupt their conve rsational partne rs. Although some investigations have replicated Zimmerman and West’s (1975) often cited finding that men tend to interrupt more often than women‚ there have also been contradictory results indicating either an absence of gender difference or even that women interrupted more than men (see Table I). Furthermore ‚ two recent narrative reviews (Aries‚ 1996; James & Clarke ‚ 1993) concluded that the re is no consistent evidence that men do inde ed interrupt more than women. 244 An derson an d Leaper When all 43 studie s (yie lding 53 hypothe sis tests) comparing women ’s and men ’s use of interruption were combine d ‚ the pre se nt meta-analysis both contradicte d and confirme d the conclusion re ache d in Arie s’ (1996) and Jame s and Clarke ’s (1993) narrative revie ws. First ‚ a significant combine d significance le ve l indicate d that men were more like ly than women to initiate inte rruptions. However‚ the corre sponding combine d effe ct size was negligible (weighte d d = .15) ‚ indicating that the magnitude of gende r 4 differe nce was in substan tial (Cohen ‚ 1977) . Looking beyond the ove rall analysis of studie s‚ the meta-analysis revealed some factors that may mode rate the like lihood and magnitude of ge nde r diffe rence s in inte rruptions. One of the se factors was the operational definition that is use d to measure inte rruption. As othe r write rs (Arie s‚ 1996; James & Clarke ‚ 1993; Tanne n ‚ 1994) have note d‚ there are multiple ways in which interruptions have bee n de fine d. When inte rpre ting interruption as a form of domine ering behavior ‚ the type we re fe r to as intrusive inte rruptions may be most re le vant. Intrusive inte rruptions function to usurp the spe aker’s turn at talk with the intent of demonstrating dominanc e . In contr ast ‚ inte rruption s that in clude bac k-chan ne l liste ning response s or affiliative ove rlaps may de monstrate e nthusiasm ‚ agre e ment ‚ or rapport. There fore ‚ any analysis of conve rsational interruptions should take into account the multiple me anings of interruptions. Both Arie s (1996) and Jame s and Clarke (1993) pointe d out the inconsiste ncy in researchers’ definitions of interruptions. Inde e d‚ some studie s we surve ye d did not provide a specific de finition of inte rruption. Thus‚ any te nde ncy for gende r differe nces in conve rsational inte rruptions may be more like ly to be dete cted when the more narrowly de fine d intrusive inte rruption cate gory is use d. When interruptions in this category were analyze d separate ly‚ a small but substantial effe ct size emerged (we ighte d d = .33) ‚ indicating that men inte rrupte d more than did wome n. Subse quent analyse s using spe cifically intrusive interruptions re ve aled 5 othe r factors that may moderate gende r e ffects on inte rruption. The moderator variable s that were inve stigate d include d publication characte ristics ‚ aspe cts of the participants ‚ length of observation ‚ and aspe cts of the interactive setting. Whe n discussing the analyses of effe ct sizes in the text‚ the we ighted effect size s will be cited because these scores adjusted for the sample size of e ach study. 5 The re were no major findings that occurre d in the analyses of overall interruptions (i.e.‚ collapsing across all ope rational de finitions) that did not occur in the analyses of intrusive interruptions. The re fore ‚ only the results associated with the analyses of intrusive interruptions are subse quently discusse d. 4 Meta-An alyses of Gen der Effects on In terr uption 245 The year of pu blication and the first au thor’s gender were two publication characte ristics that we hypothe size d might moderate ge nde r e ffects on the use of inte rruptions. Year of study did not prove to be a statistically significant mode rator. Howeve r‚ consist with our hypothe sis ‚ a small negative corre lation sugge ste d a slight te nde ncy for ge nde r diffe re nce s in intrusive inte rruptions to de crease ove r time. It might be that both researche rs and research participants have be come le ss ge nde r-stereotype d ove r the three de cade period of the studie s that were include d ‚ although this inte rpretation warrants furthe r inve stigation. The first author ’s gender was a significant moderator. Significant gender effects and large r effect sizes were more like ly when the first author was a woman. Specifically‚ women authors were more like ly than men authors to report that men interrupted more than women. When effect sizes were analyzed‚ there was even a tendency for men authors to report the opposite finding; that is‚ men authors tended to report women interrupting more than men. In their narrative review of the interruptions literature ‚ James and Clarke (1993) intimated that author gender might influe nce whether one gender is more likely to interrupt than the other‚ but the y did not speculate on any particular pattern. Our finding that author gender acted as a significant moderator is consiste nt with meta-analyse s on other topics (e.g.‚ Eagly & Carli‚ 1981; Leaper et al.‚ 1998). The difference between women and men authors potentially reflects some form of researcher bias: Some women researchers may be biase d toward identifying men as more dominant than women; also‚ some men researchers may be biased against identifying men as more dominant than women. Turning to specific aspects of the research procedures as possible moderators‚ one factor that we investigated was the length of the observation. Presumably‚ detecting stylistic differences in people ’s behavior is more apt to occur when longer observation periods are used. Accordingly‚ we expected effect sizes would be positively correlated with observation length. There was a nonsignificant positive correlation between interaction time and gender differences in intrusive interactions suggesting a slight tendency for gender differences to increase as the length of interaction increased. The strength of the correlation was likely restricted by the narrow range of observation lengths in the studies sampled. Most studies were limited to a relatively short interaction time (around 10-15 minutes). There were se ve ral aspe cts pertaining to the interactive se tting that were examine d as pote ntial mode rators. Contrary to expe ctations ‚ the gender com position of the group was not a significant moderator of intrusive inte rruptions. In their narrative re views of conve rsational inte rruptions ‚ James and Clarke (1993) found a te nde ncy for men to interrupt more in mixe d-ge nde r than same -gende r inte ractions. In contrast ‚ Arie s (1996) infe rre d no patte rn of ge nde r diffe rence relate d to ge nde r composition. Our results are consiste nt with Aries’ (1996) conclusion. 246 An derson an d Leaper G roup size was anothe r possible mode rating factor that was e xplore d. Consiste nt with our expe ctation ‚ significant ge nde r diffe re nces in intrusive inte rruptions occurre d in groups of thre e or more but not in dyads. When weighte d effe ct sizes were analyze d ‚ the magnitude of diffe re nce was moderate in groups (d = .63) but was negligible in dyads (d = .13) . To the exte nt that intrusive inte rruption is a manife station of dominance ‚ large r and more public group settings may be especially like ly to e licit this behavior in men. During childhood and adole scence ‚ boys have bee n found to be more like ly than girls to e mphasize group dominance in the ir pe er relationships. In contrast ‚ girls are more like ly to emphasize close ne ss in one -on-one interactions (Le ape r‚ 1994). Thus‚ inte racting in dyads may be a way for men to mitigate the ir domine e ring be havior. Anothe r aspect of the social relationship that may affect whether or not gender diffe rences in inte rruption are found is the fam iliarity between interactants. O ne of the limitations of most conve rsation research is that it is based on observations between strange rs (usually re cruited through introductory unive rsity courses). Relative ly few studie s have examine d interactions betwee n friends or romantic partne rs (Leape r & Ande rson‚ 1997). In our analyse s of intrusive interruptions ‚ 6 of the identifie d studie s looke d at participants who knew e ach othe r in one way or anothe r‚ the remaining 11 studie s examine d interactions be tween strange rs. We hypothe sized that unacquainte d people would be more like ly than acquainte d people to rely on ge nde r-ste reotype d e xpe ctations to guide the ir be havior. In contrast ‚ acquainte d persons such as close frie nds and intimate s have been found more like ly to rely on individual characte ristics (Drass ‚ 1986; Wood & Karten ‚ 1986). Howeve r‚ the familiarity betwee n the interactants did not prove to be a significant moderator variable in our analyse s. Pe rhaps the abse nce of any difference between strange rs and familiar interactants is related to the short observation periods in most studie s. As previously note d‚ stylistic differences related to pe ople ’s prefere nces‚ pe rsonalitie s‚ or personal relationships may be more apt to emerge over longe r periods of time. The observation setting was yet anothe r aspect of the interactive context we conside red. Interactions observed in naturalistic settings were contraste d with those occurring in a research laboratory or office. As hypothe sized‚ gender differences in intrusive interruptions were greater in naturalistic than lab settings. The combine d effect sizes associate d with naturalistic settings were the large st seen in the study (weighte d d = .76). To the extent that intrusive interruptions are generally conside red rude conve rsational practice ‚ perhaps people feel more inhibite d inte rrupting in laboratory than naturalistic se ttings. However‚ the issue deserves further exploration. Only three of the 17 studie s examining intrusive interruptions were in naturalistic settings. Therefore ‚ the finding should be conside red with caution. Meta-An alyses of Gen der Effects on In terr uption 247 Finally‚ the activity structure was anothe r obse rve d moderator of gender e ffects on intrusive inte rruptions. Spe cifically‚ significant gende r e ffects with large r effe ct sizes were more like ly when participants were observed in unstructure d activitie s than during instrume ntal tasks. A moderate-tolarge combine d e ffe cts size was associate d with unstructure d activitie s (weighte d d = .73) . In contrast ‚ there was virtually no diffe rence associate d with instrume ntal activitie s (weighte d d = .05) . These findings contradict our expe ctations as well as Jame s and Clarke ’s (1993) speculation that men would make more inte rruptions during instrume ntal activitie s since men are suppose dly more “ e xpe rt” in those tasks. Inste ad ‚ ope n-e nde d and unstructure d situations se emed to be whe re gende r diffe re nces occurred ‚ with men inte rrupting more than wome n. The obse rve d finding is actually consiste nt with a re cent meta-analysis of gende r effe cts on pare nts ’ spe ech to the ir childre n (Le ape r e t al.‚ 1998). Diffe rence s in mothe rs’ and fathe rs’ spe ech as well as differe nces in spee ch dire cted to daughte rs ve rsus sons were more like ly during unstructure d than structure d activitie s. The e ffect was interpreted in re lation to e cological-conte xtual models of ge nde r. According to this vie w‚ many ge nde r e ffects on social behavior are mediate d through the type s of activitie s that are se lected. By controlling for the activity‚ one constrains the type s of behaviors that may follow. Thus‚ whether or not men prefer instrume ntal activitie s more than wome n ‚ both women and men may act similarly whe n participating in instrume ntal tasks. Conclu sion The findings from the se meta-analyse s are consiste nt with a conte xtual-inte ractive mode l of gende r (e .g.‚ Be all ‚ 1993; Deaux & Major ‚ 1987; Leape r e t al.‚ 1998) . According to this pe rspective ‚ gende r-re late d variations in be havior are influe nce d more by situation al factors than by inhe rent individual diffe re nces betwee n wome n and men. Rele vant situational factors include characte ristics about the inte ractions such as the numbe r of pe rsons‚ their re spe ctive gende rs‚ and their relationship to one anothe r. Also ‚ the activity structure is anothe r pote ntially important situational influe nce . Of these specific factors‚ we found that group size was a significant moderator of ge nde r effe cts on interruption. Studie s also indicate that gende r-re late d variations in be havior are reduced or disappe ar when the type of activity is take n into account (se e Leape r et al. ‚ 1998) . If girls and women are apt to se le ct expre ssive activitie s‚ they may act in a more affiliative manne r. Conve rsely‚ if boys and men are more like ly to sele ct task-orie nte d activitie s‚ they may act in a more instrume ntal way. We found that ge nde r e ffects were large r when unstructure d activitie s were 248 An derson an d Leaper obse rve d. It may be that whe n participants were assigne d specific tasks ‚ the ir behavior adapte d to the de mand characte ristics of the situation and the re by reduce d the like lihood of gende r diffe rence s in interruption. To the extent that ge nde r variations in social be havior can be found to depend on situation al factors ‚ we se e more e vide nce for conte xtual-inte ractive mode l of ge nde r as oppose d to the e sse ntialist mode l that e mphasize s the existe nce of inhe re nt ‚ immutable differe nces be tween wome n and men. Lim itations an d Direction s for Future Research Despite the contributions of the pre se nt analysis ‚ the re are limitations worth highlighting . First ‚ measure s of inte rruption in the re se arch lite rature are ofte n vague ly de fine d. In the ir narrative re vie ws both Arie s (1996) and Jame s and Clarke (1993) e xpre sse d frustration at the lack of a consiste nt de finition of an inte rruption across studie s. As our analyse s sugge st‚ the type of in terruption is not a trivial matte r in the e xam ination of possible ge nde r diffe re nce s. Some type s of ove rlapping spe e ch se e m to de monstrate conve rsational dominance (the “ intrusive ” inte rruption) ‚ while othe r forms indicate affiliative e ngage ment (the “ back channe l ” liste ning re sponse ). The se two type s of ove rlap paralle l traditional ge nde rrole and status diffe re nce s in communication style ( se e e .g. ‚ Tanne n ‚ 1983 ‚ 1994; We st & Z immerman ‚ 1983) . In an e ffort to be gin conside ring how the type of inte rruption (or ove rlapping spe e ch) might mode rate the like lihood or magnitude of ge nde r diffe re nce s‚ we organize d studie s into thre e cate gorie s of inte rruptions be ginning with the most ge ne ral de finition ‚ which like ly include d affiliative ove rlaps and minim al liste ning re sponse s. In the most spe cific cate gory‚ we include d studie s that use d the re lative ly narrow form of intrusive inte rruptions. Howe ve r‚ we were unable to go be yond the se thre e some what unre fine d cate gorie s of ope ration al de finit ion. For ins tan ce ‚ fe w stud ie s in clud e d in the pre se nt analyse s analyze d affiliative ove rlaps as a se parate cate gory of inte rruptions. With e nough available studie s ‚ we would have te ste d the hypothe sis that wome n use affiliative ove rlaps more than did me n. Another limitation was our inability to conside r possible sociocultural moderators of gender effects on the use of interruption. Writers (e.g.‚ Crawford‚ 1995) have criticize d researchers for focusing on mostly on middle -class‚ Euro-American sample s. Give n the predominantly homogene ous nature of the sample s reflected in the studie s used in the meta-analysis‚ we were unable to consider the possible influence s of factors such as cultural background ‚ educational level‚ or socioe conomic status. Also‚ individual psychological factors such as gende r self-conce pt‚ gender-role ideology‚ or personality are worth Meta-An alyses of Gen der Effects on In terr uption 249 inve stigating. For example ‚ two studies (Roger & Schumache r‚ 1983; Roger & Nesshoeve r‚ 1987) found that successful interruptions were associate d with personalitie s high in dominance more than those with personalities low in dominance. Finally‚ relationship qualities were barely addre ssed due to the overwhelming numbe r of studie s based on interactions between strangers. Besides the type of relationship (e.g.‚ friendship ‚ dating ‚ married‚ etc.) which we did examine ‚ relationship qualitie s such as relative dominance or power might moderate the use of intrusive interruptions and other power-assertive speech forms. For instance ‚ Kollock‚ Blumstein & Schwartz (1985) studie d homose xual and heterosexual couple s’ decision making and found that those partne rs that had more power over day-to-day decision making made more interruptions than those partne rs more equal in power. Unfortunate ly‚ there were not enough studie s that include d measures of sociocultural factors‚ individual-psychological factors‚ or relationship qualitie s to include in our me ta-analysis. We encourage researchers to explore these factors in future studie s on gender and communication style. REFERENCES6 Aries ‚ E. (1996). Men and wom en in interaction: Reconsidering the difference. New York: Oxford University Press. Bakeman ‚ R.‚ & Que ra ‚ V . (1995) . Analyzing interaction: Sequential analysis with SDIS and G SEQ. New York: Cambridge University. B ake m an ‚ R. ‚ & G ottman ‚ J. M. ( 1986) . Observing interaction . Cambridge : Cambridge University Press. Beall ‚ A. E. (1993) . A social constructionist view of ge nder. In A. E. Beall & R. J. Sternberg (Eds.) ‚ The psychology of gender. New York: Guilford Pre ss. Beattie ‚ G. (1977). The dynamics of interruption and the filled pause. British Journal of Social and Clinical Psychology‚ 16 ‚ 283-284. *Beattie ‚ G. W. (1981) . Interruption in conversational interaction and its re lation to the sex and status of the interactants. Linguistics ‚ 19 ‚ 5-35. *Bilous‚ F. R.‚ & Krauss ‚ R. M. (1988). Dominance and accommodation in the conversation al behaviors of same - and mixe d-gende r dyads. Language and Com munication ‚ 8 ‚ 183-194. *Bohn ‚ E.‚ & Stutman ‚ R. (1983) . Se x role differences in the relational control dime nsion of dyadic interaction. Wom en ’s Studies in Com munication ‚ 6‚ 965-104. *Brooks ‚ V . R. (1982) . Sex differences in student dominance behavior in female and male professors ’ classrooms. Sex Roles‚ 8 ‚ 683-690. Camp be ll ‚ K. E .‚ Kle im ‚ D. M.‚ & O lson ‚ K. R. ( 1992) . Conve rsational activity an d interruptions among men and women. The Journal of Social Psychology‚ 132 ‚ 419-421. *Carli ‚ L. L. (1990) . Gender ‚ language ‚ and influence. Journal of Personality an d Social Psychology‚ 59 ‚ 941-951. *Case ‚ S. S. (1988) . Cultural differences ‚ not de ficiencies: An analysis of managerial women ’s language. In S. Rose & L. Larwood (Eds.) ‚ Wom en ’s careers: Pathways and pitfalls. New York: Prae ger. 6 Asterisked references are studies used in meta-analyse s. 250 An derson an d Leaper Chambliss‚ C. A.‚ & Fe eny‚ N. (1992). Effect of se x of subject‚ sex of interrupter‚ and topic of conversat ion on the perce ptions of interruptions. Perceptual an d Motor Skills‚ 75 ‚ 1235-1241. Cohen‚ J. (1977) . Statistical power analysis for the behavioral sciences (re v. ed.). Ne w York: Academic Press. *C raig ‚ D. ‚ & Pitts‚ M. K. (1990) . The dyn amics of dominance in tutorial discussion. Linguistics ‚ 28 ‚ 125-138. Crawford‚ M. (1995). Talking difference: On gender and language. London: Sage. *Dabbs ‚ J. M.‚ Jr. ‚ & Ruback‚ R. B. (1984) . V ocal patterns in male and fe male groups. Personality and Social Psychology Bulletin ‚ 10 ‚ 518-525. *De B oe r ‚ M. (1987) . Se x difference s in language : O bse rvations of dyadic conversations betwee n me mbers of the same se x. In D. Brouwe r & D. de Haan (Eds.) ‚ Wom en ’s language ‚ socialization and self-image. Providence ‚ RI: Foris. De aux‚ K. ‚ & Major ‚ B . ( 1987) . Putting ge nde r into conte xt: An interactive mode l of gende r-relate d be havior. Psychological Review ‚ 94 ‚ 369-389. *Dindia ‚ K. (1987) . The effects of se x of subject and se x of partne r on interruptions. Hum an Com m unication Research ‚ 13 ‚ 345-371. Drass ‚ K. A. (1986) . The effe ct of ge nder identity on conve rsation. Social Psychology Quarterly‚ 49 ‚ 294-301. Eagly‚ A. H.‚ & Carli‚ L. L. (1981) . Se x of rese arche rs and sex-typed communications as determinants of sex differe nce s in influenceability: A meta-analysis of social influence studies. Psychological Bulletin ‚ 90 ‚ 1-20. *Eakins ‚ B.‚ & Eakins‚ R. G. (1976). V erbal turn-taking and e xchanges in faculty dialogue . In B. L. Dubois & I. M. Crouch (Eds.) ‚ The sociology of the languages of Am erican wom en. San Antonio‚ TX: Trinity University. *Fallon ‚ J.‚ & Guo ‚ D. (1994) . The relationship betwe en topic familiarity and conversational dominance. Psychology‚ a Journal of Hum an Behavior ‚ 31 ‚ 53-57. Fe rguson ‚ N. (1977). Simultaneous speech ‚ interruptions and dominance. British Journal of Clinical Psychology‚ 16 ‚ 295-302. *Frances ‚ S. J. (1979) . Se x differences in nonverbal be havior. Sex Roles‚ 5‚ 519-535. Goldbe rg ‚ J. A. (1990). Interrupting the discourse on interruptions: An analysis in terms of re lationally neutral ‚ power- and rapport-oriented acts. Journal of Pragm atics ‚ 14 ‚ 883-903. Hall ‚ J. A. (1978) . Ge nde r e ffects in decoding nonverbal cues. Psychological Bulletin ‚ 85 ‚ 845-857. Hawkins ‚ K. (1991) . Some conse quences of deep interruption in task-oriented communication. Journal of Language & Social Psychology‚ 10 ‚ 185-203. Hyde ‚ J. S.‚ Fe nne ma ‚ E .‚ & Lamon ‚ S. J. ( 1990) . Ge nde r differe nce s in mathe matics performance : A meta-analysis. Psychological Bulletin ‚ 107 ‚ 139-155. Hyde ‚ J. S.‚ & Linn‚ M. C. (1988) . Gender differences in verbal ability: A meta-analysis. Psychological Bulletin ‚ 104 ‚ 53-69. James ‚ D.‚ & Clarke ‚ S. (1992) . Interruptions ‚ gender ‚ and power: A critical review of the literature . In K. Hall ‚ M. Bucholtz‚ & B. Moonwoman (Eds.) ‚ Locating power: Proceedings of the Second Berkeley Wom en and Language Conference. Berke ley‚ CA: Berke ley Wome n and Language Group. James ‚ D.‚ & Clarke ‚ S. (19930. Women ‚ me n‚ and interruptions: A critical re view. In D. Tannen (Ed.) ‚ G ender and conversational interaction. New York: Oxford University Pre ss. *Johnson ‚ C. (1994) . Ge nde r ‚ legitimate authority‚ and leader-subordinate conve rsations. American Sociological Review ‚ 59 ‚ 122-135. *Jones ‚ E. S.‚ & Gallois‚ C.‚ Callan‚ V . J.‚ & Barker ‚ M. (1995) . Language and power in an academic context: The effects of status‚ ethnicity‚ and se x. Journal of Language and Social Psychology‚ 14 ‚ 434-461. *Kennedy ‚ C. W.‚ & Camden ‚ C. (1983) . A new look at interruptions. Western Journal of Speech Com munication ‚ 47 ‚ 45-58. *Ko llock ‚ P . ‚ B lum ste in ‚ P. ‚ & Sch wart z ‚ P . ( 19 85) . Se x an d powe r in inte racti on: Conve rsational privileges and duties. American Sociological Review ‚ 50 ‚ 34-46. Meta-An alyses of Gen der Effects on In terr uption 251 *LaFrance ‚ M.‚ & Carmen ‚ B. (1980) . The nonverbal display of psychological androgyny. Journal of Personality & Social Psychology‚ 38 ‚ 36-49. Leaper ‚ C. (1994) . Exploring the conse quences of gender segre gation on social relationships. In C. Le ap e r ( E d.) ‚ Ch ildhoo d gen der segregation : Cau ses and consequ ences ( Ne w Directions for child Deve lopment‚ No. 65). San Francisco: Jossey-Bass. Le ape r ‚ C.‚ & Ande rson ‚ K. J. (1997) . Ge nde r de ve lopme nt and he te rose xual romantic re lationships during adolescence . In W. Damon (Series Ed.) & S. Shulman & W. A. Collins (Issue Eds.) ‚ Rom antic relationships in adolescenc e: Developm ental perspectives (New Directions for Child Developm ent‚ No. 78) . San Francisco: Jossey Bass. Leaper ‚ C.‚ Anderson ‚ K. J.‚ & Sanders‚ P. (1998) . Mode rators of gender e ffects on parents ’ talk to their children: A meta-analysis. Developm ental Psychology‚ 34 ‚ 3-27. *Le e t-Pellegrini ‚ H. M. ( 1980) . Conve rsational dominance as a function of ge nde r and e xp e rtise . In H . Gi les ‚ W . P. Robinson ‚ & P . M. Smith ( E ds.) ‚ Language: Social psychological perspectives. Ne w York: Pergamon Press. *Leffle r‚ A.‚ Gillespie ‚ D.‚ & Conaty‚ J. C. (1982) . The effects of status differentiation on nonverbal behavior. Social Psychology Quarterly‚ 45 ‚ 153-161. *Makri-Tsilipakou‚ M. (1994) . Interruption revisited: Affiliative vs. disaffiliative intervention. Journal of Pragm atics‚ 21 ‚ 401-426. *Marche ‚ T. A.‚ & Pe terson‚ C. (1993) . The deve lopment and sex-relate d use of interruption behavior. Hum an Com munication Research ‚ 19 ‚ 388-408. *Margolin ‚ G.‚ & Wampold ‚ B. E . (1981) . Se que ntial analysis of conflict and accord in distressed and nondistressed marital partners. Journal of Consulting & Clinical Psychology‚ 49 ‚ 554-567. *Martin ‚ J. N. ‚ & Craig‚ R. T. (1983) . Selecte d linguistic sex differe nce s during initial social interactions of same-se x and mixe d-sex student dyads. The Western Journal of Speech Com m unication ‚ 47 ‚ 16-28. *McLachlan ‚ A. (1991) . The effects of agre ement ‚ disagreeme nt‚ ge nder and familiarity on patte rns of dyadic interaction. Journal of Language & Social Psychology‚ 10 ‚ 199-213. McLaughlin ‚ M. L. (1984) . Conversation: How talk is organized. Beve rly Hills‚ CA: Sage . *McMillan ‚ J. R.‚ Clifton‚ A. K.‚ McGrath ‚ D.‚ & Gale ‚ W. S. (1977) . Women ’s language: Uncertainty ‚ or interpersonal sensitivity and emotionality? Sex Roles ‚ 3 ‚ 545-559. Mott‚ H.‚ & Petrie ‚ H. (1995) . Workplace interactions: Women ’s linguistic behavior. Journal of Language and Social Psychology‚ 14 ‚ 324-336. Mullen‚ B. (1989). Advanced BASIC meta-analysis. Hillsdale ‚ NJ: Lawrence Erlbaum. Murata ‚ K. (1994) . Intrusive or co-operative? A cross-cultural study of interruption. Journal of Pragm atics ‚ 21 ‚ 385-400. Murray‚ S. O.‚ & Covelli‚ L. H. (1988) . Wome n and men spe aking at the same time. Journal of Pragm atics ‚ 12 ‚ 103-111. *Natale ‚ M.‚ Entin‚ E.‚ & Jaffe ‚ J. (1979) . Vocal interruptions in dyadic communication as a function of spee ch and social anxie ty. Journ al of Personality & Social Psychology‚ 37 ‚ 865-878. *Neme th ‚ C.‚ Endicott‚ J.‚ & Wachtler ‚ J. (1976) . From the ‘50s to the ‘70s: Women in jury deliberations. Sociom etry‚ 39 ‚ 293-304. *Ng ‚ S. H. ‚ Brook‚ M.‚ & Dunne ‚ M. (1995) . Interruption and influence in discussion groups. Journal of Language and Social Psychology‚ 14 ‚ 369-381. *Nohara ‚ M. (1992) . Sex differe nce s in interruption. An experimen tal reevaluation. Journal of Psycholinguistic Research ‚ 21 ‚ 127-146. *Redeker ‚ G.‚ & Mae s‚ A. (1996) . Gende r differences in interruptions. In D. I. Slobin‚ J. Gerhardt ‚ A. Kyratzis ‚ & J. Guo (Eds.) ‚ Social interaction ‚ social context‚ and language: Essays in honor of Susan Ervin-Tripp. Mahwah ‚ NJ: Lawrence Erlbaum Associates. Rim ‚ Y. (1977) . Personality variable s and interruptions in small discussion. European Journal of Social Psychology‚ 7 ‚ 247-251. Robinson‚ L. F.‚ & Re is‚ H. T. (1989) . The effects on interruption‚ gender and status on interpersonal perce ptions. Journal of Nonverbal Behavior ‚ 13 ‚ 141-153. *Roger ‚ D.‚ Nesshoe ver ‚ W. (1987) . Individual differe nce s in dyadic conversational strategies: A further study. British Journal of Social Psychology‚ 26 ‚ 247-255. 252 An derson an d Leaper *Ro ge r ‚ D. B .‚ & Sch umache r ‚ A. ( 1983) . E ffects of individual diffe re nce s on dyadic conve rsational strategies. Journal of Personality and Social Psychology‚ 45 ‚ 700-705. *Rogers ‚ W. T.‚ & Jones ‚ S. E. (1980). Effects of dominance tendencies on floor holding and interruption behavior in dyadic interaction. Hum an Com munication Research ‚ 1 ‚ 113-122. Sacks‚ H. ‚ Schegloff ‚ E. A.‚ & Je fferson‚ G. (1974). A simplest systematics for the organization of turn-taking for conversation. Language ‚ 50 ‚ 696-735. *Shaw‚ M. E. ‚ & Sadle r‚ O. W. (1965). Interaction patterns in heterose xual dyads varying in degree of intimacy. Journal of Social Psychology‚ 66 ‚ 345-351. *Simkins-Bullock‚ J. ‚ & Wildman ‚ B. G. (1991) . An inve stigation into the re lationships betwee n gende r and language. Sex Roles ‚ 24 ‚ 149-160. *Smeltzer ‚ L. R.‚ & Watson ‚ K. W. (1986) . Gender differe nce s in verbal communication during negotiations. Com munication Research Reports ‚ 3 ‚ 74-79. *Smith ‚ H. W. (1977) . Small group interaction at various ages: Simultaneous talking and interruptions of others. Sm all G roup Behavior ‚ 8 ‚ 65-74. Smith-Lovin ‚ L.‚ & Brody‚ C. (1989). Interruptions in group discussions: The effe cts of ge nder and group composition. American Sociological Review ‚ 54 ‚ 424-435. *Street ‚ R. L.‚ Jr.‚ & Murphy‚ T. L. (1987) . Interpersonal orientation and speech behavior. Com m unication Monographs ‚ 54 ‚ 42-62. Tanne n‚ D. (1983) . When is an ove rlap not an interruption? O ne component of conversational style. In R. J. DePie tro‚ W. Frawley‚ & A. Wedel (Eds.) ‚ The first Delaware Sym posium on Language Studies. Newark ‚ DE: University of Delaware Press. Tanne n‚ D. (1994) . Interpre ting interruption in conversation. In D. Tannen (Ed.) ‚ G ender and discourse. New York: Oxford. Thimm ‚ C.‚ Rademache r‚ U.‚ & Kruse ‚ L. (1995) . “Power-re lated talk”: Control in verbal interaction. Journal of Language and Social Psychology‚ 14 ‚ 382-407. Thomas ‚ A. P.‚ Roge r ‚ D. ‚ & B ull‚ P. ( 1983) . A se que ntial analysis of informal dyadic conve rsation using Markov chains. British Journal of Social Psychology‚ 22 ‚ 177-188. *Trimboli ‚ C.‚ & Walke r ‚ M. B. (1984) . Switching pauses in coope rative and competitive conve rsations. Journal of Experim ental Social Psychology‚ 20 ‚ 297-311. *W e lkowitz ‚ J. ‚ Bon d ‚ R. N. ‚ & Fe ldstein ‚ S. ( 1984) . Conve rsational time patte rns of Japane se -Ame rican adults and children in same and mixed-ge nder dyads. Journal of Language and Social Psychology‚ 3‚ 127-138. *West ‚ C.‚ & Z imme rman ‚ D. (1983) . Small insults: A study of interruptions in cross-sec conve rsations betwee n unacquainted person. In B. Thorne ‚ C. Kramarae ‚ & N. He nley (Eds.) ‚ Language ‚ gender and society. Rowley‚ MA: Newbury House . Willis‚ F. N.‚ & Williams ‚ S. J. (1976) . Simultane ous talking in conve rsation and se x of speakers. Perceptual and Motor Skills‚ 43 ‚ 1067-1070. Wood ‚ W. ‚ & Karten ‚ S. J. (1986). Se x differe nce s in interaction style as a product of perce ived sex differences in competence. Journal of Personality & Social Psychology‚ 50 ‚ 341-347. Woods‚ N. (1989) . Talking shop: Sex and status as determinants of floor apportionment in a work se tting. In J. Coates & D. Came ron (Eds.) ‚ Wom en in their speech com munities. New York: Longman. Zimmerman ‚ D. H.‚ & West ‚ C. (1975) . Sex roles‚ interruptions and silences in conversation. In B. Thorne & N. Henle y (Eds.) ‚ Language and sex: Difference and dom inance. Rowley‚ MA: Ne wbury House.
© Copyright 2024 Paperzz