Online persuasion: an examination of gender differences in computer mediates interpersonal influence
1. Group Dynamics: Theory, Research, and Practice Copyright 2002 by the Educational Publishing Foundation
2002, Vol. 6, No. 1, 38 –51 1089-2699/02/$5.00 DOI: 10.1037//1089-2699.6.1.38
Online Persuasion: An Examination of Gender Differences in
Computer-Mediated Interpersonal Influence
Rosanna E. Guadagno and Robert B. Cialdini
Arizona State University
The purpose of this research was to investigate how computer-mediated communica-
tion affects persuasion in dyadic interactions. Two studies compared participants’
attitudes after hearing a series of arguments from a same-gender communicator via
either e-mail or face-to-face interaction. In Study 1, women showed less message
agreement in response to e-mail versus face-to-face messages, whereas men showed no
difference between communication modes. Study 2 replicated this finding and exam-
ined the impact of prior interaction with the communicator. For women, the condition
that provided the least social interaction led to the least message agreement. For men,
the condition that provided the most social challenge led to the least message agree-
ment. Results are interpreted in terms of gender differences in interaction style.
A mounting body of evidence indicates that It is noteworthy that social psychological re-
communication modality influences the charac- search on persuasion has rarely examined either
ter and effectiveness of the communication pro- computer-mediated or face-to-face interaction,
cess (see Chaiken & Eagly, 1983, for a review). preferring to use other modalities for reasons of
Although there are many ways in which com- methodological control and ease. One goal of
munication modes differ, one dimension may be the present research was to redress this disparity
particularly relevant to current thinking about by assessing the impact of persuasive appeals
interpersonal processes: the extent to which the delivered in the ecologically frequent but
medium makes prominent (or merely available) grossly understudied contexts of face-to-face
various personal and social factors not related to and computer-mediated exchanges.
the message itself (Chaiken & Trope, 1999). According to dual-processing models of per-
Communication modalities that restrict the avail- suasion (Chaiken, 1980; Chaiken & Trope,
ability of factors such as those mentioned above 1999; Petty & Cacioppo, 1984), circumstances
may be referred to as socially constrained, that direct the attention of communication re-
whereas those that provide access to such cues cipients toward or away from features of the
may be termed socially unconstrained. message can have decidedly different persua-
We argue that this dimension extends from
sive consequences. Thus, socially constrained
written, entirely text-based modes (e.g., essays,
and unconstrained communication modes may
e-mail, newspaper articles) on the socially con-
produce different persuasion patterns among in-
strained side, to voice-based modes (e.g., radio
or intercom transmissions), through visually dividuals focused differentially on message-re-
based modes (e.g., televised or videotaped pre- lated or interpersonal aspects of the communi-
sentations), and finally to face-to-face interac- cation setting.
tions (e.g., workplace meetings, corridor con- For example, Chaiken and Eagly (1976) ex-
versations) on the socially unconstrained pole. amined how mode of communication affected
message processing as well as subsequent atti-
tudes. In their study, participants received either
Rosanna E. Guadagno and Robert B. Cialdini, Depart-
a difficult or an easy to comprehend persuasive
ment of Psychology, Arizona State University. message through one of three communication
This research was supported by a National Science Foun- modalities: videotape, audiotape, or written.
dation Graduate Fellowship. The easy message was more effective in the
Correspondence concerning this article should be ad-
dressed to Rosanna E. Guadagno, Department of Psychol-
videotape condition, the socially unconstrained
ogy, Arizona State University, Box 1104, Tempe, Arizona communication modality in which the speaker’s
85257-1104. E-mail: rosanna@asu.edu cues were most salient. Conversely, the difficult
38
2. SPECIAL ISSUE: ONLINE PERSUASION 39
message was more effective in the written com- implication of these results is that social con-
munication condition, the socially constrained straint of the communication modality has an
mode in which source cues were minimized. impact on the persuasive factors at work in a
This study provided clear evidence that differ- negotiation process (Morley & Stephenson,
ent types of persuasive messages produce dif- 1977). With more social cues available, the
ferential degrees of attitude change as a function research participants were less swayed on the
of communication medium. quality of their opponent’s position.
Chaiken and Eagly (1983) conducted a fol- In sum, the results of the previously reviewed
low-up study in which they manipulated likabil- studies (Chaiken & Eagly, 1976, 1983; Morley
ity of the communicator. As in the initial study, & Stephenson, 1977) suggest that the persua-
participants received a persuasive message sive impact of different types of messages is
through one of three communication modalities. moderated by the extent to which the commu-
However, in this case, participants also read a nication modality makes salient message-rele-
personal statement from the speaker that made vant versus non-message-relevant (e.g., social)
him sound likable or unlikable. When the cues.
speaker was likable, participants in both video-
tape and audiotape conditions evidenced greater Computer-Mediated Communication:
attitude change than participants in the written A New Communication Mode
communication condition. When the speaker
was not likable, attitude change was greatest for More recently, a newer communication mo-
participants who received the written commu- dality has emerged— computer-mediated com-
nication. These results suggest that in the vid- munication (CMC)—which stands to become
eotape and audiotape conditions (the less so- increasingly important as a medium for com-
cially constrained conditions), the personal cues munication. A recent survey reported that 71
of the communicator were salient and partici- million people in the United States have access
pants engaged in heuristic processing of the to the Internet (Iconocast, 1999), and the num-
persuasive message. Conversely, in the written bers are increasing. People use the Internet to
communication condition, in which source cues send e-mail, participate in real-time interactive
were less salient, participants processed the group discussions, download software, partici-
message systematically. pate in noninteractive discussion (e.g., Usenet),
Similarly, Morley and Stephenson (1977) use a remote computer, conduct business trans-
conducted a series of studies that investigated actions, and engage in real-time audio or video
the influence of formality of communication conversations (Jones, 1995).
system on negotiation. These studies primarily To date, CMC has been highly socially con-
investigated the persuasive factors involved in a strained, restricted for the most part to text-
two-person negotiation that took place either based, impersonal forms. Therefore, we would
over the phone or face-to-face. According to our expect that persuasive messages delivered in
terminology, because nonverbal feedback (e.g., this fashion would produce response patterns
eye contact, body language, facial expression) similar to those of other socially constrained
was not available to participants in the phone communication modalities.
condition, the phone condition was more so- And in fact, such patterns were found in
cially constrained than the face-to-face condi- the research of Kiesler, her colleagues, and
tion. In each negotiation, one participant was others in studies of group decision making
given a strong case (i.e., a large number of (Dubrovsky, Kiesler, & Sethna, 1991; Hiltz,
high-quality arguments) whereas the other was Johnson, & Turoff, 1986; Kiesler, Siegel, &
given a weak case to argue. The overall results McGuire, 1984; Siegel, Dubrovsky, Kiesler, &
of this series of studies indicted that, as pre- McGuire, 1986). These investigators found that
dicted, the strong case was more successful in compared with face-to-face participants in
the (more socially constrained) phone condition decision groups, individuals communicating
than in the (less socially constrained) face-to- through a computer were more likely to violate
face condition. Conversely, the weak case argu- social norms of politeness and to be focused
ment was more successful in the face-to-face more uniformly on the task. Similarly, a study
condition than in the phone condition. A clear by Matheson and Zanna (1989) on face-to-face
3. 40 GUADAGNO AND CIALDINI
versus CMC persuasion revealed a direct link Overview of the Present Experiments
between social cues and attitude change only in
the face-to-face condition. Several studies have As more and more people gain access to the
additionally reported that participants interact- Internet, a greater amount of interpersonal com-
ing via face-to-face like their discussion part- munication is taking place through this medium.
ners more than those interacting via CMC Moreover, an increasing number of business
(Kiesler, Zubrow, Moses, & Geller, 1985; decisions are being made primarily through
Weisband & Atwater, 1999). Additionally, text-based messaging, such as e-mail. The pur-
these measures of partner liking were predictive pose of these experiments was to examine (a)
of decisions in face-to-face interaction but not the ways in which this relatively new and very
in CMC. As with the previously reviewed stud- socially constrained communication modality
ies on communication modality, it appears that influences the process of a persuasive appeal
individuals who communicate through this so- and (b) the impact that this modality has on
cially constrained mode are less focused on male versus female message recipients, who
their partners and more focused on the assigned tend to focus differentially on the interpersonal
task. aspects of an exchange.
Gender Differences in Persuasion Study 1
Social roles, especially gender-based roles, The purpose of our first study was to examine
are another factor that can influence responses the way a communicator could persuade a dis-
to persuasion attempts. In a meta-analysis of cussion partner to favorably evaluate a counter-
148 studies, Eagly and Carli (1981) found a attitudinal message, depending on strength of
tendency for women to be more persuadable argument, communication modality, and gender
than men, but this effect was moderated by the of the target of persuasion. In our design, a
extent to which social factors, such as group confederate attempted to induce agreement in a
pressures and publicness of response, were same-gender research participant on the merits
present. Eagly (1987) and others (Tannen, of instituting comprehensive exams as a new
1990; Wood & Stagner, 1994) attribute these graduation requirement. The confederate was
findings to different social role expectations for interviewed by each research participant using
men and women. Men are said to be oriented one of two communication modality conditions:
toward agency, which often manifests in at- e-mail or face-to-face. During the course of the
tempts to demonstrate one’s independence from interview, the confederate used either a strong
others in successful performances. Women, on or a weak set of arguments in favor of the
the other hand, are said to be more communally proposal in an attempt to persuade the research
oriented, which often manifests in activities de- participant.
signed to foster interpersonal cooperation and A novel aspect of this paradigm was the use
relationship formation and maintenance. of a confederate to administer the persuasive
According to linguist Deborah Tannen (1990), arguments in an interactive exchange rather
men’s communication style in interpersonal in- than having research participants read a written
teraction is based on the perception that in message on a computer screen or watch a vid-
interactions, a man must strive to achieve in- eotaped message. This allowed us to investigate
dependence and avoid failure. Women, on the the impact of persuasive messages in a bilateral
other hand, have a slightly different percep- exchange context rather than in a unilateral
tion of their world. This perspective is one in persuasion agent-to-persuasion target context,
which conversations are meant to achieve close- which is typical of persuasion research but
ness and consensus. Thus, according to Tan- might not be representative of the way persua-
nen’s view, when interacting with others, men sion is accomplished in many nonexperimental
are interested in establishing independence settings.
through assertiveness or mastery of their envi- Given that men and women differ in their
ronment, whereas women are interested in mak- motivational goals, we expected that women in
ing connections with other individuals through the face-to-face condition would express more
cooperation. agreement with the confederate than would
4. SPECIAL ISSUE: ONLINE PERSUASION 41
women in the e-mail condition because the so- Procedure
cial constraint of e-mail does not easily allow
for the establishment of a connection or bond. On arrival, research participants were in-
For men, however, we expected communication formed that they were to take part in a two-
mode not to make as much of a difference in person interview– discussion and opinion sur-
evaluation of the persuasive message because vey on behalf of the university administration.
men are more likely to enter interactions with a They were told that they would be paired with a
desire for independence rather than cooperation partner (the confederate) with whom they would
or bonding. Because communicator characteris- discuss the merits of instituting senior compre-
tics matter more for women and are more salient hensive exams as an additional graduation re-
in face-to-face interactions, we also predicted quirement and that this discussion would be
that personality trait ratings of the confederate structured like an interview. One partner was
would be related to attitude toward the compre- assigned the task of “interviewing” the other
hensive exams only among women in the face- partner (always the confederate). At this point,
to-face condition. participants were given an informational para-
We also predicted that across communication graph describing the comprehensive exam pro-
mode and gender, strong arguments would elicit posal and the goal of the interview discussion.
more message agreement than weak arguments. Next, participants were given a personal
Finally, because e-mail is a highly socially con- statement handwritten by the confederate. All
strained communication mode, we expected that participants read the same statement from the
participants in this condition would generate confederate, which included information such
more message-oriented cognitive responses to as his or her year in school, favorite food, and
the persuasive interaction. Conversely, because hobbies and an indistinct description of his or
face-to-face interaction is a socially uncon- her personality. This information was provided
strained communication mode, we expected to create a uniform initial impression of the
participants to be more focused on the commu- confederate.
nicator and therefore to record more communi- Prior to engaging in the interview– discus-
cator-focused thoughts than participants in the sion, participants in the e-mail condition re-
email conditions. ceived training on how to use the computer
Method 1
An additional set of participants were run but excluded
from the analyses: 12 expressed suspicion, 5 did not under-
Participants stand the task, and 4 failed the relevance manipulation
check. Analyses of the primary dependent measures with
Research participants were 159 (80 fe- these participants included in the data set did not yield
different results than reported.
male, 79 male) undergraduate psychology stu- 2
Personal relevance (high vs. low) was also manipulated,
dents.1 Only those with computer experience but it did not yield any significant effects on the primary
were eligible for this study. dependent measures. Thus, this variable is not discussed
further. However, relevance was still included in all the
analyses reported here for appropriate partitioning of the
variance.
Design 3
Pretesting of the introductory psychology subject pool
indicated the mean favorability rating of this issue was 3.14
The experimental design was a 2 (communi- on a scale ranging from 1 (extremely unfavor) to 9 (ex-
cation mode: face-to-face vs. e-mail) 2 (ar- tremely favor).
4
gument strength: strong vs. weak) 2 (gender The arguments used in this study were adapted from
Petty, Harkins, and Williams (1980). Examples of the
of dyad: male vs. female) factorial.2 From evi- strong arguments emitted by the confederate are as follows:
dence that most undergraduates would not sup- “The quality of teaching is better at schools with exams”
port the institution of senior comprehensive ex- and “Average starting salaries are higher for graduates for
ams,3 the confederate was instructed to present schools with exams.” Examples of the weak arguments are
as follows: “Companies that develop these exams wouldn’t
either a set of strong or weak arguments (adapted market to schools unless they worked” and “Graduate stu-
from Petty, Harkins, & Williams, 1980)4 in an dents have to take comprehensive exams and if undergrads
attempt to change the participant’s attitude. don’t have to, that’s discrimination.”
5. 42 GUADAGNO AND CIALDINI
program.5 All participants were given a list of Participants in the strong argument condition
questions to ask their discussion partner. Once rated the comprehensive exam more favorably
participants were prepared, they were intro- (M 5.56) than participants in the weak argu-
duced to the confederate and the interview– ment condition (M 4.76).
discussion began. As depicted in Table 1, there was also a
For each question, the confederate was pre- significant communication mode by gender in-
pared to emit a scripted response that contained teraction, F(1, 143) 6.58, p .01, 2 .044.
an assigned (strong or weak) argument. The A test of simple effects indicated that this in-
confederate was instructed that during the inter- teraction was due to the fact that women in the
view– discussion he or she should discuss only face-to-face condition (M 5.54) were more
the comprehensive exam proposal and provide favorable toward comprehensive exams than
only statements based on the assigned set of women in the e-mail condition (M 4.68),
arguments. However, the confederate was told F(1, 156) 3.87, p .05, whereas men
that if the participant went off-task, he or she showed a nonsignificant trend, F(1, 156)
should try to reveal only information consistent 2.47, p .12, in the opposite direction.
with the information on the personal statement
or to state additional opinions on the compre-
hensive exams that were consistent with the Partner Ratings
arguments and overall cover story. At the end of
the interview– discussion, each participant com- A principal-axis factor analysis with oblique
pleted the dependent measures. Next, a suspi- rotation was conducted on the partner trait rat-
cion check was conducted, and the participant ings. This analysis indicated that the 13 traits on
was debriefed and excused. which the confederate was rated could be re-
duced to form three distinct factors. Factor 1,
labeled “Congenial,” accounted for 49% of the
Dependent Variables variance and contained the following traits:
The main measure was participants’ attitude approachable, confident, likable, interesting,
toward the comprehensive exam proposal, mea- friendly, sincere, and warm. Factor 2, labeled
sured using a scale ranging from 1 (extremely “Knowledgeable,” accounted for 7% of the
un ) to 9 (extremely ) on the following variance and contained the following traits:
dimensions: workable, valuable, needed, and competent, informed, and credible. Finally,
favorable. Factor 3, labeled “Sincerity,” accounted for 6%
Next, participants’ cognitive responses to the of the variance and contained the following
interview– discussion were measured using a traits: modest, honest, and trustworthy. The fac-
thought-listing exercise. Finally, a series of ad- tor loadings for Congenial ranged from .573 to
ditional measures assessed participants’ impres- .879; for Knowledgeable, from .737 to .804; and
sion of the confederate on a scale ranging for Sincerity, from .323 to .851.
from 1 (not at all ) to 9 (very ) on A 2 (face-to-face vs. e-mail) 2 (strong vs.
the 13 different dimensions. weak argument) 2 (male vs. female dyad)
ANOVA on the Knowledgeable composite re-
vealed a significant main effect for argument
Results strength, F(1, 143) 20.82, p .01, 2
.127, indicating that participants perceived the
Attitude Measure
confederates emitting the strong arguments as
Participants’ ratings of the comprehensive more knowledgeable (M 7.35) than confed-
exam proposal on the attitude measures were erates emitting the weak arguments (M 6.34).
averaged to form one composite of overall atti- Similar ANOVAs on the Congenial and Sincer-
tude toward the exam proposal ( .92). A 2 ity composites revealed no significant differ-
(face-to-face vs. e-mail) 2 (strong vs. weak ences by condition.
argument) 2 (male vs. female dyad) analysis
of variance (ANOVA) revealed two significant 5
Both the participant’s and the confederate’s names ap-
effects. The first was a main effect for argument peared on the screen, creating a nonanonymous CMC
strength, F(1, 143) 6.64, p .01, 2 .044. environment.
6. SPECIAL ISSUE: ONLINE PERSUASION 43
Table 1 cator thoughts (M 2.02) than participants
Mean Attitude Toward the Comprehensive Exam in the e-mail condition (M 1.44),
(and Standard Deviations) F(1, 143) 4.83, p .03, 2 .033. This
Communication mode analysis revealed an additional significant main
Dyad and effect for argument strength, indicating that par-
attitude Face-to-face E-mail
ticipants in the weak argument condition
Study 1 (M 2.01) recorded more communicator
Female 5.54 (1.85) 4.68 (2.10) thoughts than participants in the strong argu-
Male 4.86 (1.99) 4.68 (1.99) ment condition (M 1.45), F(1, 143) 4.38,
Study 2 p .04, 2 .03.
Message thoughts. Message thoughts refer
Female to any comments about the comprehensive ex-
Competitive 5.33 (1.90) 5.27 (2.10)
ams and the arguments emitted by the confed-
Independent 5.52 (2.06) 4.25 (2.08)
Cooperative 5.21 (0.99) 5.09 (1.64) erate. A 2 (face-to-face vs. e-mail) 2 (strong
Male vs. weak argument) 2 (male vs. female dyad)
Competitive 4.08 (1.95) 5.57 (1.85) ANOVA on the total number of message
Independent 6.12 (1.47) 5.89 (1.78) thoughts recorded by participants did not reveal
Cooperative 5.57 (1.42) 5.50 (1.26) any significant effects. However, an examina-
tion of the negative message thoughts revealed
a significant main effect for communication
mode. Participants in the e-mail condition re-
Correlation coefficients between the partner corded relatively more negative message
evaluation composites and the attitude measure thoughts (M 1.12) than participants in
indicated that partner trait ratings were signifi- the face-to-face condition (M 0.83),
cantly related to attitude in only one condition. F(1, 143) 3.98, p .05, 2 .027.
For women in the face-to-face condition, atti-
tude was significantly positively correlated with Unscripted Comments
all three factors (for Congenial, r .406, p
.01; for Knowledgeable, r .448, p .01; and The interview– discussion transcripts were
for Sincerity, r .317, p .05). For women in content coded.7 The total number of unscripted
the e-mail condition and for men in both com- comments emitted by participants in each ses-
munication modalities, there were no significant sion was counted. A 2 (face-to-face vs. e-mail)
correlations between partner trait ratings and 2 (strong vs. weak argument) 2 (male vs.
attitude. female dyad) ANOVA on the total number of
unscripted comments revealed a significant
Cognitive Responses main effect for communication mode, indicat-
ing that participants in the face-to-face condi-
Two independent judges who were blind to tion were more likely to deviate from the
experimental condition rated participants’ cog- scripted questions (M 6.91) than were partic-
nitive responses as recorded on the thought list- ipants in the e-mail condition (M 1.38),
ing measure for valence (positive, negative, or F(1, 132) 35.74, p .01, 2 .213. This
neutral) as well as focus (communicator, mes- analysis also revealed a main effect for gender,
sage, or irrelevant). The judges’ ratings were indicating that men were more likely to deviate
averaged to form a more reliable measure.6 from the scripted questions (M 5.49) than
Communicator thoughts. Communicator
thoughts refer to any comments about the con- 6
The interrater reliabilities for each type of cognitive
federate that were recorded by participants on response were as follows: for total, r .88; communicator,
the thought-listing measure. A 2 (face-to-face r .99; positive communicator, r .87; negative commu-
vs. e-mail) 2 (strong vs. weak argument) 2 nicator, r .48; message, r .73; positive message, r
(male vs. female dyad) ANOVA on the total .78; negative message, r .73; irrelevant, r .67. Note that
the lower correlation coefficients occurred in cells in which
number of communicator thoughts recorded by there was a restricted range of responses.
participants indicated that participants in the 7
Eleven transcripts (10 face-to-face, 1 e-mail) were lost
face-to-face condition recorded more communi- through a recording error.
7. 44 GUADAGNO AND CIALDINI
women (M 2.55), F(1, 132) 10.38, p .01, Perhaps the most intriguing finding in this
2
.073. These two main effects were quali- study is that men and women differed in mes-
fied by a significant two-way communication sage-favorable attitudes depending on which
mode by gender interaction, F(1, 132) 3.92, communication mode they used to interact
p .05, 2 .029. A test of the simple effects with the same-gender confederate. Specifically,
indicated that for participants in the face-to-face women in the face-to-face condition reported
condition, men were more likely to make un- more agreement than did women in the e-mail
scripted comments (M 9.29) than were condition. However, there was no significant
women (M 4.54), F(1, 145) 9.29, p .01. difference between men in the e-mail and men
However, in the e-mail condition, there was no in the face-to-face condition. These findings do
significant gender difference in number of un- not appear to be due to a tendency for women to
scripted comments. be more persuadable than men, as the women
never exhibited significantly higher levels of
agreement than men in comparable conditions.
Discussion
We believe that participants responded in a
This study provides new information on the manner consistent with gender-stereotypical ex-
impact of strength of argument, participant gen- pectations. For men, there was no difference in
der, and communication mode on interpersonal agreement with the message between e-mail and
influence. Our results suggest that the most ef- face-to-face perhaps because the differences in
fective way to persuade an individual differs social constraint between the two conditions
according to the strength of the argument as were relatively unimportant to the men, whose
well as his or her gender and the mode of social roles focus more on independence and
communication. agency than on relationships. We believe that
A number of conclusions can be drawn. First, women, on the other hand, whose roles focused
in this study, as in previous research (see Petty them more on relationship formation and coop-
& Cacioppo, 1986), the strength of the argu- eration, aligned their attitudes more with the
ment had a significant impact on agreement persuader’s position in the face-to-face condi-
with the message, indicating that strong argu- tion because it was in that condition that their
mentation is more persuasive than weak argu- relationship goals were salient and attainable.
mentation. Additionally, participants were more This interpretation receives support from the
focused on the communicator in the weak argu- strong positive correlations between attitude to-
ment condition as compared with the strong ward the exams and persuader personality trait
argument condition, as indicated by the finding ratings that occurred only for women in the
that participants in the weak argument condition face-to-face condition.8
recorded more communicator thoughts than did An alternative explanation of these results is
participants in the strong argument condition. that male and female confederates differed in
As predicted, participants in the face-to- their persuasiveness and that these differences
face condition recorded more communicator led participants to evaluate the arguments dif-
thoughts than did participants in the e-mail con- ferently. To explore this possibility, a pair of
dition, suggesting that source cues were more raters coded the e-mail transcripts for persua-
salient in the face-to-face condition than in the siveness. All references to participant gender
e-mail condition. In addition, more negative were removed. An analysis of these data re-
message thoughts were generated in the e-mail vealed no significant gender differences. How-
condition as compared with the face-to-face ever, a main effect for argument strength was
condition, suggesting a greater message focus in revealed, as in the attitude measure, F(1, 74)
the e-mail condition as compared with the face- 153.60, p .01, 2 1.0.
to-face condition. The fact that the communica-
tion mode difference occurred only for negative
8
message thoughts suggests that e-mail partici- The previous research that reported greater liking in
face-to-face interactions as compared with e-mail did not
pants may have responded to the message with include a detailed breakdown of the means by gender com-
counter argumentation that was suppressed for position of the dyad, so it is difficult to compare these
those in face-to-face interactions. results with the previous findings.
8. SPECIAL ISSUE: ONLINE PERSUASION 45
Finally, although the interview– discussion interaction. In addition, we gave some partici-
transcripts were coded only for the quantity of pants a competitive experience with the com-
unscripted comments, in informal observations municator before the persuasion attempt, and
during this coding we noted that when the men we gave other participants a cooperative expe-
in face-to-face conditions went off script, they rience with the communicator before the per-
seemed to be attempting to establish domi- suasion attempt.
nance, whereas women who went off script According to a social roles perspective, one
tended to do more bonding. This observation is would expect a prior competitive interaction to
similar to an analysis of gender differences in have the most negative effect on the men and
the content of Internet newsgroup postings con- that this would be the case primarily in the
ducted by Herring (1993) and greatly influenced face-to-face communication mode, where social
the design of Study 2. and personal cues are prominent. By this same
account, however, one would expect the most
Study 2 negative effect on the women’s levels of mes-
sage agreement to occur when there had been
We conducted a second study to (a) replicate the least amount of prior interaction and that
the gender by communication mode interaction this would be the case primarily in the e-mail
of Study 1 (so as to increase confidence in its condition, where social and personal cues are
reliability) and (b) modify the pattern through most prominent.
additional manipulations designed to shed light
on the conceptual mediation on this basic effect. Method
That is, if it is the case that men are more likely
to see their interactions with others in term of Participants
competition whereas women are more likely to
see such interactions in terms of cooperation Research participants were 237 (139 fe-
(Eagly, 1987; Tannen, 1990), we wondered male, 98 male) undergraduate psychology stu-
whether it would be possible to influence male dents selected in the same way as in Study 1.9
and female responses to a communicator by
varying the nature of their prior (competitive or Design
cooperative) interaction. In the socially uncon-
strained environment of face-to-face communi- The experimental design was a 2 (communi-
cation, men who have had a prior competitive cation mode: face-to-face vs. e-mail) 2 (gen-
interaction with the communicator should re- der: male vs. female dyad) 3 (prior interac-
spond competitively by rejecting the communi- tion: competitive vs. cooperative vs. indepen-
cator’s argument. This should not be the case dent) factorial. The confederate was instructed
for women, however, for whom prior interac- to present the set of strong arguments used in
tion may serve to establish a relationship in Study 1.
which cooperation and harmony are sought.
Thus, for women, it might be the case that Procedure
various forms of prior interaction would set the
stage for future agreement with the other Participants were told that they would partic-
by way of relationship-building attempts. For ipate in two separate studies, one examining the
women, then, it would not be a competitive way individuals put together numbers and one
prior interaction that would lead to rejection of providing feedback on proposed changes to ac-
a communicator’s argument but rather a lack of ademic policy.
meaningful prior interaction. As in the previous study, the participant and
To examine these possibilities, we used the the confederate arrived at the same place and
face-to-face and e-mail procedures of Study 1 to time. This time, however, the experimenter
replicate the basic finding of that study (that
women showed less message agreement in e- 9
An additional 20 participants were excluded from the
mail versus face-to-face modes, whereas men’s data analyses because they expressed suspicion. Analyses of
levels of agreement did not differ) when we the primary dependent measures with these participants
provided participants with no meaningful prior included did not yield different results.
9. 46 GUADAGNO AND CIALDINI
brought both participant and confederate to the participants to attend to the task at the same
same lab room at the same time. level as participants in the other conditions.
The experimenter then proceeded to present All participants were given a sheet to track
instructions for the first study, a number game their performance. Participants were told that
that was designed to manipulate prior interac- this tracking sheet would be used to compute
tion. There were three versions of the game: one their performance scores. Finally, the experi-
designed to induce competition, one designed menter set a timer for 5 min and left the con-
to induce cooperation, and one that provided federate and participant alone in the room to
equivalent interpersonal exposure without any play the game.
explicit cooperative or competitive interaction. Once the timer rang, the experimenter reen-
All participants were presented with the same tered the room and transitioned to the discussion
introductory statement and were told that they of the comprehensive exam. The experimenter
would take turns building their own puzzle us- explained that the participant and confederate
ing three-sided dominos. The instructions then would discuss a potential change to academic
deviated depending on condition. policy. The participant was asked to pick out of
Participants in the competitive prior interac- an envelope one of four possible topics. The
tion condition were told that the purpose of this confederate was asked to pick out of an enve-
study was to compare the performance of intro- lope one of two possible roles: the interviewer
ductory psychology students with that of edu- (the one who asks the questions) or the respon-
cation students (such as the confederate).10 dent (the one who answers the questions). In
They were each instructed to take turns playing reality, each envelope contained multiple copies
a piece, then take a piece from the other’s pile of the same choice: comprehensive exam for the
of pieces. Confederates were instructed to try to topic of discussion and respondent as the role
take pieces that the participants were likely to for the confederate. After these assignments
were made, the experimenter moved the partic-
use. Finally, participants in this condition were
ipant to a room with a computer and the rest of
told that the person who performed the best
the experiment replicated the procedure from
would receive a $25 prize. The prize was of-
Study 1.
fered to strengthen the competitive environment
and to motivate participants to attend to the
task.
Dependent Variables
Participants in the cooperative prior interac- Attitude toward the comprehensive exam,
tion condition were told that the purpose of this cognitive responses, and partner trait ratings
study was to examine the performance of intro- were assessed using the same measures as in
ductory psychology students partnered with ed- Study 1.
ucation students. They were each instructed to As manipulation checks for the nature of the
take turns playing a piece and then offer a piece prior interaction (game check), participants
to their partner. Confederates were instructed to were asked three questions. First, they were
try to offer pieces to the participants that they asked to rate the nature of the number-matching
could use. Finally, participants in this condition game on a Likert-type scale ranging from 1
were told that the pair who performed the best (cooperative) to 9 (competitive) with 5 (neither
would receive a joint $25 prize. In this case, the competitive or cooperative) as the scale mid-
prize was offered to strengthen the cooperative point. Next, participants assessed their partner’s
environment and to motivate participants to at- competitiveness and cooperativeness using a
tend to the task. Likert-type scale ranging from 1 not at
Participants in the independent prior interac- all to 9 very .
tion condition received only the introductory To assess computer experience, we asked
statement and were then instructed to take turns participants to indicate the number of hours
playing without exchanging any pieces. They
were also told that the best performing intro- 10
The confederate was also introduced as an education
ductory psychology participant would win a student in Study 1. This was done to reduce suspicion due
$25 prize, as would the best performing educa- to the fact that two participants arrived for an experiment for
tion student. The prize was offered to motivate which only one was scheduled.
10. SPECIAL ISSUE: ONLINE PERSUASION 47
spent using a computer and the Internet for cooperative condition, F(1, 225) 52.01, p
multiple purposes. .01, 2 .316. This analysis also indicated that
men viewed the game as more competitive than
Predictions did women (M 5.32 vs. M 4.93),
F(1, 225) 10.84, p .01, 2 .046.
We predicted that the communication mode Competitive trait rating. A 2 (face-to-face
by participant gender interaction reported in vs. e-mail) 2 (male vs. female dyad) 3
Study 1 would be replicated in the independent (competitive vs. cooperative vs. independent
prior interaction conditions in this study. Spe- prior interaction) ANOVA revealed a signifi-
cifically, we expected the women in the inde- cant main effect for prior interaction condition,
pendent prior interaction, e-mail condition to indicating that participants perceived their part-
show less message agreement than women in ners in the competitive (M 5.47) and inde-
the face-to-face conditions. In addition, we ex- pendent (M 5.52) prior interaction conditions
pected to find no communication mode differ- as more competitive than those in the coopera-
ence between men across conditions. tive prior interaction condition (M 4.78),
The general social role prediction for men F(1, 225) 4.80, p .01, 2 .041. In addi-
was that there would be no communication tion, this analysis revealed a significant main
mode or prior interaction difference in agree- effect for participant gender, indicating that
ment toward the message except in the face-to- men perceived their partners as more competi-
face, competitive prior interaction condition. In tive than did women (M 5.32 vs. M 4.93),
this condition, we expected less message agree- F(1, 225) 11.55, p .01, 2 .049.
ment than in the other male conditions, resulting Cooperative trait rating. A 2 (face-to-face
in a 1 versus 5 pattern of results. vs. e-mail) 2 (male vs. female dyad) 3
For women, the general social role prediction (competitive vs. cooperative vs. independent
was for a different 1 versus 5 pattern of results. prior interaction) ANOVA revealed a signifi-
Because motivation for cooperation and bond- cant main effect for participant gender, indicat-
ing should override the competitive nature of ing that across all conditions, female confeder-
the prior interaction, we expected women to ates were rated as more cooperative than male
report more message agreement in all condi- confederates (M 8.33 vs. M 7.82),
tions in which some type of interaction oc- F(1, 225) 9.41, p .01, 2 .04. Thus,
curred. Thus, we predicted a 1 versus 5 pattern analyses of these three manipulation checks in-
of results, with women in the e-mail, indepen- dicated that the situation and the confederate
dent prior interaction condition showing less were perceived accurately in each condition. In
message agreement as compared with all other addition, men perceived the confederate and
conditions. situation as more competitive than did women.
Finally, we did not expect to find any gender
differences in computer experience among our
participants.
Attitude Measure
The reliability of the attitude composite was
Results .91.11 The 2 (face-to-face vs. e-mail) 2
(male vs. female dyad) 3 (competitive vs.
Manipulation Checks cooperative vs. independent prior interaction)
ANOVA yielded a significant two-way interac-
Game check. A 2 (face-to-face vs. e-mail) tion between communication mode and prior
2 (male vs. female dyad) 3 (competitive interaction, F(1, 225) 3.54, p .03, 2
vs. cooperative vs. independent prior interac-
tion) ANOVA revealed a significant main effect
for prior interaction condition, indicating that 11
Prior attitudes toward the comprehensive exam pro-
participants perceived the competitive game posal were available for 181 participants (76% of the sam-
condition (M 6.16) as more competitive than ple). An analysis of covariance on the attitude measure
using the pretest attitude as the covariate revealed the same
the independent (M 5.52) and the cooperative pattern of results as without the covariate. Consequently,
(M 3.00) conditions and that the independent data on the full sample without the covariance analysis are
condition was seen as more competitive than the reported hereafter.
11. 48 GUADAGNO AND CIALDINI
.031; a significant two-way interaction between prior interaction was cooperative or competi-
prior interaction and gender, F(1, 225) 4.14, tive, F(1, 225) 0.032 and F(1, 225) 0.019,
p .02, 2 .036; and a marginal two-way respectively. The hypothesized 1 versus 5 con-
communication mode by gender interaction, trast between the independent, face-to-face con-
F(1, 225) 3.05, p .08, 2 .013. See Table 1 dition and all other female conditions was sig-
for a breakdown of means by condition. nificant, F(1, 225) 6.94, p .01. Thus, for
In addition, we conducted a more focused set women, the absence of any type of prior inter-
of analyses relative to our specific predictions. action minimized their willingness to agree with
Our first major prediction was that the commu- their discussion partner. See Table 1 for a pre-
nication mode by gender interaction of Study 1 sentation of the means for this measure.
would be replicated in the independent interac-
tion condition of Study 2. An analysis of the Partner Ratings
independent prior interaction cells revealed a
nearly significant interaction of communication The partner trait ratings were analyzed using
mode and participant gender, F(1, 235) 3.57, a confirmatory factor analysis to test whether
p .06. Simple effects tests indicated that, as in the factors established by the exploratory factor
Study 1, women in the e-mail condition re- analysis in Study 1 generalized to this sample.
ported less agreement than women in the face- The model fit reasonably well, CFI .928,
2
to-face condition (M 4.25 vs. M 5.52), (1, N 62) 161.34, p .01.
F(1, 234) 6.49, p .01. For the men in the Congenial. A 2 (face-to-face vs. e-mail)
independent condition, the difference between 2 (male vs. female dyad) 3 (competitive
e-mail and face-to-face was not significant, vs. cooperative vs. independent prior interac-
F(1, 234) 0.01, ns. Thus, the results of tion) ANOVA revealed a significant main effect
Study 1 were replicated. for gender, indicating that women rated the con-
The second major prediction was that mes- federate as more congenial than did men (M
sage agreement would not differ among men 7.03 vs. M 6.30), F(1, 225) 21.57, p
except for those in the competitive prior inter- .01, 2 .087.
action, face-to-face condition, which should In addition, women in the independent, e-
show the least agreement. A 1 versus 5 contrast mail condition rated their discussion partner as
testing this hypothesis proved significant, less likable than did women in any other con-
F(1, 225) 11.65, p .01. An additional ex- dition, F(1, 225) 10.53, p .01. Thus, the
amination of the attitude measure for men same 1 versus 5 pattern that appeared in the
across condition revealed that in addition to the attitude measure also appeared in ratings of
above results, there was no communication partner congeniality. For men, there were no
mode difference between men within the coop- differences in partner ratings by condition.
erative condition, F(1, 225) 0.008, ns, or in Knowledgeable. A 2 (face-to-face vs. e-mail)
the independent condition, F(1, 225) 0.17, 2 (male vs. female dyad) 3 (competitive
ns. For men in the competitive condition, how- vs. cooperative vs. independent prior interac-
ever, there was less agreement in the face-to- tion) ANOVA revealed a significant main effect
face condition as compared with men in the for gender, indicating that women rated the con-
e-mail condition (M 4.08 vs. M 5.57), federate as more knowledgeable than did men
F(1, 225) 5.91, p .02. Thus, for men, the (M 7.50 vs. M 6.95), F(1, 225) 6.27,
type of interaction did not have an impact on p .01, 2 .027.
agreement unless they were initially forced to Sincerity. A 2 (face-to-face vs. e-mail) 2
compete, then later placed in a face-to-face in- (male vs. female dyad) 3 (competitive vs.
teraction where their prior competitor espoused cooperative vs. independent prior interaction)
his views. ANOVA revealed a significant main effect for
For women, a different picture emerged in gender, indicating that women rated the confed-
keeping with the third major prediction. The erate as more sincere than did men (M 7.50
least agreement occurred among those in the vs. M 6.95), F(1, 225) 16.21, p .01,
2
independent prior interaction condition. The .067.
difference in attitude toward the exams between In addition, women in the independent, e-
e-mail and face-to-face was not significant if the mail condition rated their discussion partner as
12. SPECIAL ISSUE: ONLINE PERSUASION 49
less sincere than did women in any other con- stead this difference looks to be associated with
dition, F(1, 225) 7.03, p .01. Thus, the gender differences in interaction style: women
same 1 versus 5 pattern that appeared in the are motivated to form bonds, whereas men are
attitude measure also appeared in ratings of motivated to compete if necessary to achieve
partner sincerity. For men, there were no differ- independence. The finding that women reported
ences in partner ratings by condition. the same level of message agreement in all
Finally, an examination of the correlations conditions involving prior interaction with the
between attitude toward the comprehensive ex- confederate, even when this prior interaction
ams and communicator trait ratings did not re- was competitive in nature, bolsters this interpre-
veal any significant differences by condition, tation. This interpretation is additionally sup-
contrary to the results reported in Study 1. ported by the predicted 1 versus 5 pattern
wherein women in the e-mail, independent con-
Cognitive Responses dition reported lower ratings on attitude toward
An analysis of the message and communica- the comprehensive exam, partner congeniality,
tor thoughts did not reveal the communication partner sincerity, and positive thoughts about
mode differences found in Study 1—that face- the partner as compared with women in the
to-face interaction produced more communica- other conditions. It is not surprising that women
tor thoughts than did CMC interaction, F(1, chose to bond rather than compete, in that
225) 0.14, ns. This suggests that the prior women feel more comfortable cooperating, even
interaction eliminated the differences in mes- in a competitive environment (Anderson &
sage processing typically found in the persua- Morrow, 1995). Additionally, women will choose
sion literature. to bond with other women, especially in times
An examination on the positive communica- of stress (Taylor et al., 2000). Finally, previous
tor thoughts showed that women in the inde- research on gender-stereotypical behavior indi-
pendent, e-mail condition recorded fewer posi- cates that women will reject imposed roles if
tive thoughts about their discussion partner they do not agree with them (Cialdini, Wosin-
than did women in any other condition, ska, Dabul, Whetstone-Dion, & Heszen, 1998).
F(1, 225) 6.94, p .01. Thus, the same 1 Men, on the other hand, did not appear to be
versus 5 pattern that appeared in the attitude focused on establishing a cooperative bond with
measure also appeared in ratings of positive the confederate. Instead, they evaluated the ar-
thoughts about their partner. For men, there guments for what they were and showed no
were no differences in positive communicator differences in attitude toward the exams unless
thoughts by condition. they had competed previously, and then took
part in the face-to-face discussion with the prior
Computer Experience rival. Although men showed no universal ten-
dency for competition, it appears that they can
The items on the computer experience mea- be pushed to compete and that the competitive,
sure were summed to form one composite ( face-to-face condition spurred them to do so,
.68). An ANOVA on this measure revealed no decreasing their willingness to align their atti-
significant effects for gender, communication tudes with their competitor.
mode, or prior interaction. Thus, it appears that
the gender differences in persuasion cannot be
explained by gender differences in computer use. General Discussion
Discussion Taken together, these two studies shed light
on the impact of interactive CMC on interper-
The results of this study replicated the finding sonal influence. For women, having any prior
in Study 1 that without a prior meaningful in- interaction with a communicator enhances the
teraction, women taking part in a persuasive level of agreement relative to that occurring in
exchange via e-mail agreed with a communica- impersonal e-mail interactions. For men, only
tor less than women taking part in the same an intensely competitive environment led to less
exchange in a face-to-face setting. In addition, agreement.
we demonstrated that this result was not due to However, certain unanswered questions re-
gender differences in computer experience. In- main and deserve further investigation. First, it
13. 50 GUADAGNO AND CIALDINI
is possible that the communication mode differ- References
ence in message processing typically found in
the persuasive communication literature (that Anderson, C. A., & Morrow, M. (1995). Competitive
aggression without interaction: Effects of compet-
face-to-face interaction produces more commu- itive versus cooperative instructions on aggressive
nicator-relevant thoughts than CMC interac- behavior in video games. Personality and Social
tion) may extend only to short-term interactions Psychology Bulletin, 21, 1020 –1030.
with strangers. The fact that there was no dif- Carli, L. L. (1989). Gender differences in interaction
ference in cognitive responses in Study 2 sug- style and influence. Journal of Personality and
gests that a prior relationship with the commu- Social Psychology, 56, 565–576.
nicator superseded the communication mode. In Chaiken, S. (1980). Heuristic versus systematic in-
formation processing and the use of source versus
addition, Walther and Burgoon (1992) found
message cues in persuasion. Journal of Personality
that modality effects in impression formation and Social Psychology, 39, 752–766.
were eliminated after a group interacted via Chaiken, S., & Eagly, A. H. (1976). Communication
CMC over an extended period of time. Thus, in modality as a determinant of message persuasive-
situations where an individual attempts to per- ness and message comprehensibility. Journal of
suade a person he or she knows, there may be Personality and Social Psychology, 34, 606 – 614.
no difference in the amount of systematic or Chaiken, S., & Eagly, A. H. (1983). Communication
heuristic processing. Future research should ex- modality as a determinant of persuasion: The role
of communicator salience. Journal of Personality
amine this phenomenon in real-world settings.12 and Social Psychology, 45, 241–265.
It is additionally possible that women in the face- Chaiken, S., & Trope, Y. (Eds.). (1999). Dual-pro-
to-face conditions agreed more with the persua- cess theories in social psychology. New York:
sive message because it facilitated bonding and a Guilford Press.
comfortable interaction environment. It would be Cialdini, R. B., Wosinska, W., Dabul, A. J., Whet-
interesting to test the duration of their attitude stone-Dion, R., & Heszen, I. (1998). When role
toward the exams. If their reported opinions were salience leads to role rejection: Modest self-pre-
sentation among women and men in two cultures.
just a function of public conformity, then we would
Personality and Social Psychology Bulletin, 24,
expect that their agreement with the message 473– 481.
would fade faster over time than it would for men. Deaux, K., & Major, B. (1987). Putting gender into
Finally, it is possible that our results may context: An interactive model of gender-related
hold true only for same-gender pairings. Same- behavior. Psychological Review, 94, 369 –389.
gender pairs were used in the present research to Dubrovsky, V. J., Kiesler, S., & Sethna, B. N.
reduce additional error variance that may have (1991). The equalization phenomenon: Effects in
occurred as the result of mixed-gender pairings. computer-mediated and face-to-face decision-
making groups. Human-Computer Interaction, 6,
Our results might not replicate as strongly in 119 –146.
other contexts such as a mixed-gender situation, Eagly, A. H. (1987). Sex differences in social behav-
because mixed-gender pairs display less gen- ior: A social-role interpretation. Hillsdale, NJ:
der-stereotypical behavior than do same-gender Erlbaum.
pairs (Carli, 1989; Deaux & Major, 1987) and Eagly, A. H., & Carli, L. L. (1981). Sex of research-
evaluate each other differently (see Kiesler et ers and sex-typed communications as determinants
al., 1985). Future research on this phenomenon of sex differences in influenceability: A meta-anal-
ysis of social influence studies. Psychological Bul-
should be conducted on mixed-gender dyads.
letin, 90, 1–20.
12
Herring, S. C. (1993). Gender and democracy in
In addition, these results may not generalize to con- computer-mediated communication. Electronic
texts in which the CMC is completely anonymous. Research
Journal of Communication, 3(2). Retrieved 2000
indicates that in-group identity becomes more salient when
from http://www.cios.org/www/ejcmain.htm
the CMC is anonymous (see Postmes, Spears, & Lea, 1998,
for a review). Social identities such as gender become more Hiltz, S. R., Johnson, K., & Turoff, M. (1986). Ex-
salient and may serve as a heuristic cue and may lead to periments in group decision making: Communica-
more agreement with in-group members and less agreement tion process and outcome in face-to-face versus
with out-group members. However, because participants in computerized conferences. Human Communica-
this study were not anonymous, an increased salience of tion Research, 13, 225–252.
social categories was not an important feature of the CMC Iconocast. (1999). Internet users at a glance. Re-
environment we created. trieved from http://www.iconocast.com
14. SPECIAL ISSUE: ONLINE PERSUASION 51
Jones, S. G. (Ed.). (1995). Cybersociety: Computer- view. Journal of Personality and Social Psychol-
mediated communication and community. Thou- ogy, 38, 81–92.
sand Oaks, CA: Sage. Postmes, T., Spears, R., & Lea, M. (1998). Breaching
Kiesler, S., Siegel, J., & McGuire, T. W. (1984). or building social boundaries? SIDE effects of
Social psychological aspects of computer-medi- computer-mediated communication. Communica-
ated communication. American Psychologist, 39, tion Research, 25, 689 –715.
1123–1134. Siegel, J., Dubrovsky, V., Kiesler, S., & McGuire,
Kiesler, S., Zubrow, D., Moses, A. M., & Geller, V. T. W. (1986). Group processes in computer-medi-
(1985). Affect in computer-mediated communica- ated communication. Organizational Behavior and
tion: An experiment in synchronous terminal-to- Human Decision Processes, 37, 157–187.
terminal discussion. Human-Computer Interac- Tannen, D. (1990). You just don’t understand: Men
tion, 1, 77–107. and women in conversation. New York: Ballantine
Matheson, K., & Zanna, M. P. (1989). Persuasion as Books.
a function of self-awareness in computer-mediated
Taylor, S. E., Klein, L. C., Lewis, B. P., Gruenewald,
communication. Social Behaviour, 4, 99 –111.
T. L., Gurung, R. A. R., & Updegraff, J. A. (2000).
Morley, I. E., & Stephenson, G. M. (1977). The
Biobehavioral responses to stress in females:
social psychology of bargaining. London: Allen &
Unwin. Tend-and-befriend, not fight-or-flight. Psycholog-
Petty, R. E., & Cacioppo, J. T. (1984). The effects of ical Review, 107, 411– 429.
involvement on responses to argument quantity Walther, J. B., & Burgoon, J. K. (1992). Relational
and quality: Central and peripheral approaches to communication in computer-mediated communi-
persuasion. Journal of Personality and Social Psy- cation. Human Communication Research, 19, 50 –
chology, 46, 69 – 81. 88.
Petty, R. E., & Cacioppo, J. T. (1986). Communica- Weisband, S., & Atwater, L. (1999). Evaluating self
tion and persuasion: Central and peripheral and others in electronic and face-to-face groups.
routes to attitude change. New York: Springer- Journal of Applied Psychology, 84, 632– 639.
Verlag. Wood, W., & Stagner, B. (1994). Why are some
Petty, R. E., Harkins, S. G., & Williams, K. D. people easier to influence than others? In S. Shavitt
(1980). The effects of group diffusion of cognitive & T. C. Brock (Eds.), Persuasion (pp. 149 –174).
effort on attitudes: An information-processing Boston: Allyn & Bacon.