Content uploaded by Jeffrey T Hancock
Author content
All content in this area was uploaded by Jeffrey T Hancock on Jan 18, 2014
Content may be subject to copyright.
COMMUNICATION RESEARCH • June 2001Hancock, Dunham • Impression Formation
JEFFREY T. HANCOCK1
PHILIP J. DUNHAM
Impression Formation in Computer-
Mediated Communication Revisited
An Analysis of the Breadth and
Intensity of Impressions
Following either a text-based, synchronous computer-mediated conversation
(CMC) or a face-to-face dyadic interaction, 80 participants rated their part-
ners’ personality profile. Impressions were assessed in terms of both their
breadth (the comprehensiveness of the impression) and intensity (the magni-
tude of the attributions). Results indicated that impressions formed in the
CMC environment were less detailed but more intense than those formed
face-to-face. These data provide support for theories that, in addition to
acknowledging the unique constraints and characteristics of CMC, consider
the cognitive strategies and heuristics involved in the impression formation
process. The differential impact of a text-based medium on trait-specific
impressions (e.g., extraversion, neuroticism) is also discussed in the context of
a cross-modal approach to impression formation.
Whether our interactions are long or short, task-oriented or casual, awkward
or comfortable, as humans we seem to exit most of our social encounters with
some general impression of the other person’s characteristics and disposi-
tions. When forming impressions about others, two general factors are
involved. First and most obvious are the various forms of direct and indirect
information available. Such information involves autonomous cues such as
physical appearance (Jones, 1990), descriptions of a target’s behavior or per-
sonality traits (Anderson, 1981), and other more subtle cues that emerge dur-
ing social interactions. The latter can include social markers in speech
(Scherer, 1979), reciprocation and verbal elaboration (Wyer, Swan, &
Gruenfeld, 1995), and other aspects of language use (Semin, 1989). In addi-
tion, once we have gathered data about our social partners, we employ a
325
COMMUNICATION RESEARCH, Vol. 28 No. 3, June 2001 325-347
© 2001 Sage Publications, Inc.
variety of inferential heuristics or strategies that also influence the impres-
sion formation process (Fiske & Taylor, 1991). Although an extensive review
of these heuristics is beyond the scope of the present article, examples of such
cognitive strategies include schema development (Cantor & Mischel, 1977),
attitudes and stereotyping (Fiske & Taylor, 1991; Hogg & Abrams, 1988), per-
son memory (Wyer & Srull, 1989), and other systematic biasing mechanisms
that influence our attributions (Bem, 1972; Funder, 1987).
Questions about the factors influencing impression formation are cur-
rently being revisited with renewed vigor in the context of recent technologi-
cal developments generally described as text-based, computer-mediated com-
munication (CMC) (see Jacobson, 1999; Lea & Spears, 1992, 1995; Sproull &
Kiesler, 1986; Walther, 1996, 1997). Early analyses of these text-based social
exchanges emphasized the extent to which the CMC environment modifies or
eliminates many of the cues and sources of information that have been identi-
fied as important in traditional impression formation research (for reviews,
see Johansen, Vallee, & Spangler, 1979; Kiesler, Siegel, & McGuire, 1984;
Walther, Anderson, & Park, 1994). Initial data tended to suggest, in general,
that the impoverished CMC environment supports at best impersonal and
task-oriented forms of communication (Hiltz, Johnson, & Turoff, 1986; Siegel,
Dubrovsky, Kiesler, & McGuire, 1986; Sproull & Kiesler, 1986).
Theories that evolved from or were applied to these early findings tended
to focus on CMC’s diminished capacity for conveying emotional and personal
information. These accounts include, among others, social presence theory
(Short, Williams, & Christie, 1976) and reduced social context cues (Siegel
et al., 1986; Sproull & Kiesler, 1986). Collectively, these theoretical
approaches have been described by Culnan and Markus (1987) as defining a
cues filtered-out (CFO) perspective. The unifying theme central to these
approaches is that the reduction of nonverbal social and relational cues in
CMC produces a depersonalized form of communication and decreased
awareness of others, inhibiting interpersonal relations (for reviews and dis-
cussion of this perspective, see Garton & Wellman, 1995; Parks & Floyd,1996;
Spears & Lea, 1992; Walther, 1996).
The fundamental implication of the CFO perspective for impression for-
mation in CMC is clear. The restriction and elimination of impression-rele-
vant signals should lead to the development of relatively amorphous impres-
sions when compared to a face-to-face (FtF) environment (Walther, 1993).
Consider, for example, Dubrovsky, Kiesler, and Sethna’s (1991) suggestion
that the reduced cues in CMC “may lead people to forget that messages are
communications, not just soliloquies to a computer. People can forget the
nature and size of their audience or even that their communications will be
read” (p. 124).
326
COMMUNICATION RESEARCH • June 2001
Whereas the CFO approach focuses on the reduction in sources of informa-
tion available for impression formation, more recent theories have ques-
tioned the technological determinism implicit in this approach and have
begun to focus on some of the social heuristics and inferential processes that
contribute to impression formation. Two early and important challenges to
the CFO perspective are the Social Identification/Deindividuation (SIDE)
model (Lea & Spears, 1991, 1995; Spears & Lea, 1992, 1994) and the social
information-processing theory (Walther, 1993; Walther & Burgoon, 1992).
The SIDE model acknowledges the lack of cues afforded by CMC but shifts
the focus to the social identity variables that frame CMC interactions and to
the cognitive processes by which humans make significant inferences and
overattributions about others on the basis of minimal information (Lea &
Spears, 1992, 1995; Spears & Lea, 1992, 1994). Essentially, Spears and Lea
(1994) argue that the lack of individuating cues in CMC compared with FtF
interactions (e.g., physical appearance, vocal cues, etc.) renders zero-history
participants relatively anonymous. The major consequence of this
deindividuation, defined by visual anonymity and physical isolation, is an
increased reliance on the few remaining social cues (e.g., cues to role, status,
etc.) on which to form impressions of a CMC partner (Lea & Spears, 1995;
Spears & Lea, 1994).
Under these conditions, participants are assumed to construct more stereo-
typed and exaggerated representations of their partners, based on the mini-
mal cues emerging primarily from the relevant contextual signals provided
in text-based discourse (e.g., cues to gender, status, categorical membership,
etc.; Lea & Spears, 1991). Impression-relevant information may also emerge
from the partner’s communication style (e.g., word choice, paralinguistic
cues, typographic information, etc.; Lea & Spears, 1992). As Lea and Spears
(1992) note,
In CMC, the presence of typing errors in a message may indicate that
the sender was in a hurry when writing, but if errors are consistently
observed over a series of messages they may be interpreted to mean
that the person is careless or incompetent. (p. 324)
Within the original SIDE model formulation, any impressions formed are
presumed to reflect the operation of underlying social categorization pro-
cesses (Lea & Spears, 1992).For example, when deindividuated CMC partici-
pants perceive themselves as part of a group, this group identity is intensified
and becomes more salient than their individual self-identities. The primary
consequences of this social perception are strong positive feelings toward the
partner and intensified attributions of similarity. The opposite outcome is
327
Hancock, Dunham • Impression Formation
assumed in equally deindividuated conditions where the individual identi-
ties of participants are made more salient. In this latter case,unique aspects
of a partner’s communication style can accentuate negative feelings toward
the partner (Lea & Spears, 1992; Spears & Lea, 1992). Note that, in sharp
contrast to the CFO perspective, the scarcity of social and interpersonal
information in CMC is assumed, paradoxically, to produce more intense and
exaggerated positive or negative impressions of communicative partners,
depending on the social context.
Social information-processing theory challenged the CFO perspective on
several issues primarily concerned with the development of social impres-
sions in CMC over time (Walther, 1993; Walther & Burgoon, 1992). The major
thrust of social information-processing theory is that CMC retards the rate at
which impression-relevant cues are exchanged during social interaction,
rather than simply reducing or eliminating the amount of such information.
Communicators are assumed to take an active role in forming impressions
through text-based information. Initial impressions are presumed to be
incomplete relative to FtF, but they become more developed and comprehen-
sive over time as participants seek out relevant information about their part-
ners (Walther, 1993, 1996).In one test of this hypothesis, Walther (1993) com-
pared the depth of impressions over the course of three interactive sessions in
CMC and FtF conditions. As predicted,after the first session, the impressions
of partners formed by CMC participants were incomplete relative to those
formed FtF, but by the third session, impression development in CMC
approximated the development observed in the FtF condition.
A more recent theory proposed by Walther (1996,1997), the hyperpersonal
model, extends the developmental constraints outlined by social informa-
tion-processing theory within a larger framework that recognizes both the
cognitive processes highlighted by the SIDE model and some of the unique
communicative features of the CMC environment. The hyperpersonal model,
for example, incorporates the SIDE model’s fundamental assumption that
deindividuated participants in this social context will tend to form more ste-
reotyped (i.e., hyperpersonal) impressions based on the limited social and
interpersonal cues available. As described earlier, these stereotypic represen-
tations lead participants to overattribute in the absence of individuating or
contraindicating information (Walther, 1997).
Although the SIDE model emphasizes the influence of social categoriza-
tion processes on the intensity of positive and negative attributions emerging
during CMC, Walther (1996) describes a wider array of interpersonal and
social-communicative mechanisms that may operate over time. Participants
are also presumed, for example,to take advantage of the communicative limi-
tations of CMC to engage in what Walther (1996) has termed selective
328
COMMUNICATION RESEARCH • June 2001
self-presentation. Because CMC provides only textual information, partici-
pants can intentionally select positive and desirable cues (e.g.,witty rapport)
to present to their partners while masking or minimizing physical and
behavioral cues not normally under the participants’ control (e.g., physical
appearance). In addition, whereas the lack of physical cues in CMC is typi-
cally portrayed by the CFO perspective as a hindrance to communication, the
hyperpersonal model notes that this lack of cues may allow the user to reallo-
cate cognitive resources normally applied to active involvement in FtF inter-
action, such as providing timely back-channel responses (Clark,1996), to lan-
guage selection, message construction, and impression management
(Walther, 1996, 1997).
When these processes are considered together, the hyperpersonal model
presents several experiential, cognitive,and behavioral factors that may pro-
duce interpersonal impressions in CMC that are more intense, rather than
less so. Receivers of CMC messages tend to make overattributions regarding
their partners’ personality; senders of CMC messages,freed from some of the
demands and constraints of FtF interaction, may engage in selective self-pre-
sentation and construct more thoughtful and carefully created messages.
Given this general overview, it is perhaps informative to contrast, in a par-
ticular social context, the predictions generated by the hyperpersonal model
with predictions generated by the traditional CFO perspective. Suppose, for
the moment, that two people meet for the first time in a text-based, CMC
environment (e.g., a chat room, a technical support forum, a distributed
learning environment, etc.) and converse about a particular topic. After the
conversation, assume the participants are asked to rate each other on various
personal characteristics measured on a Likert scale.
The CFO perspective suggests that impressions should be poorly devel-
oped in CMC relative to those formed during a similar interaction situated in
an FtF context. The basic implications of the CFO model at the level of mea-
surement operations are that (a) the impoverished CMC social environment
will reduce the number of personal characteristics on which participants are
confident in rating each other (i.e., participants will make fewer attributions
in this impoverished environment) and (b) those characteristics that are
rated will tend to cluster in the neutral range of a Likert scale. Note that
these predictions address the impact of the impoverished environment in
terms of both the breadth (i.e., number of characteristics rated) and the inten-
sity (i.e., the magnitude of the attributions) of the impressions formed during
the exchange.
In contrast, the hyperpersonal model acknowledges the impoverished
nature of this communicative environment but implies that zero-history,
deindividuated participants should initially be willing to form exaggerated
329
Hancock, Dunham • Impression Formation
and stereotypic impressions enhanced by the unique communicative fea-
tures provided by the CMC environment (e.g., selective self-presentation).
The social information-processing perspective adopted by the hyperpersonal
model suggests that the relatively slow rate of social information exchange
presumed to occur in CMC interactions will tend to reduce the number of
individual characteristics that participants are willing to rate after a
one-time interaction (Walther, 1993). With regard to the intensity issue, how-
ever, the hyperpersonal model suggests an increase in the intensity of the
more stereotypic and exaggerated aspects of the impressions that are formed.
Thus, those characteristics that are rated by participants should tend to clus-
ter toward the extreme (either positive or negative) ranges of the Likert scale.
Again, we note that the impact of the CMC environment on impression for-
mation is considered in terms of both the breadth and intensity dimensions of
the process.
If one examines the existing literature for data relevant to these issues,
there are, to the best of our knowledge,no experiments comparing impression
formation in CMC and FtF environments in which both the breadth and the
intensity of impressions are measured independently in the same experi-
mental context. Procedures that assess the breadth of the impression formed
are particularly rare. In most studies, participants are required to provide
intensity ratings on a fixed number of traits identified as relevant by the
experimenter and are not given the option of omitting an item if they are not
confident in making the attribution. Indeed, as described above, in the first
and only study in which participants were given the option of not rating par-
ticipants, Walther (1993) reported that CMC participants were willing to rate
their conversational partners on a smaller array of attributes than FtF par-
ticipants following an initial interaction. As indicated earlier, this observa-
tion that impressions are less broadly formed after an initial interaction is
consistent with predictions flowing from both the CFO perspective and the
hyperpersonal model.
Research measuring the intensity dimension of impression formation has
yielded inconsistent data. Several studies directly comparing CMC
(deindividuating) and FtF (individuating) conditions have failed to yield sig-
nificant intensity differences in the impressions formed in these two differ-
ent communicative environments. For example, in an early study primarily
assessing self-awareness in CMC, Matheson and Zanna (1988) found no dif-
ference across CMC and FtF conditions for participants’ average evaluations
of their partners on a 13-item, bipolar adjective scale following a short inter-
action. Similarly, in a study that was primarily concerned with group
330
COMMUNICATION RESEARCH • June 2001
decision-making processes, Lea and Spears (1991) failed to observe a signifi-
cant difference across CMC and FtF conditions when participants were also
asked to complete an ancillary measure assessing interpersonal evaluation
of their partners (see also Lea & Spears, 1992, Study 2).
These failures to observe differences in the intensity of impressions
formed across FtF and CMC settings contrast somewhat with more recent
results reported by Walther (1997).Although this study did not directly com-
pare across FtF and CMC conditions, significant differences in partner rat-
ings of attractiveness (e.g., social and physical), productivity, and affection
were observed within CMC conditions. Identity factors (individual versus
group-salient identity) were factorially crossed with short-term versus
long-term group membership. Impressions formed by partners in both
group-identity conditions (short term and long term) were significantly more
intense than those formed by partners in the two individual identity condi-
tions (short term and long term). Of particular interest in the present context
is that these differences in the intensity of impressions emerged from indi-
viduation manipulations within a CMC environment. Previous failures to
observe differences across deindividuated CMC and, by default, individuated
FtF conditions are therefore somewhat surprising. Note also that the more
intense ratings that emerged on these interpersonal traits are consistent
with the hyperpersonal model’s assumption that personal attributions can
be exaggerated in CMC.
The specific goal of the present study was to measure both the breadth and
intensity dimensions of impression formation independently in the same
experimental context. Following a short-term interaction between zero-
history dyads in either a CMC or FtF environment, participants were asked
to rate the personality characteristics of their conversational partners on a
well-established, standardized measure of personality attributes that was
modified to assess separately both the breadth and intensity of their
impressions.
As discussed above, both the CFO and hyperpersonal perspectives suggest
that, given the option, CMC interactants will be willing to rate fewer charac-
teristics of their partners than FtF interactants. However, the predictions of
the CFO and hyperpersonal models differ with respect to the intensity of the
participants’ ratings. The CFO model assumes that CMC participants will
make relatively neutral attributions (i.e., attributions in the neutral range of
a Likert scale) in this deindividuated environment, whereas the
hyperpersonal approach argues that CMC participants will make more
intense attributions based on the limited cues available in CMC.
331
Hancock, Dunham • Impression Formation
Method
Participants
Participants were 84 English-speaking (12 male and 72 female) members
of the university community, who received course credit or token remunera-
tion for participating in the experiment. Participants were randomly paired
to form 42 dyads. The members of each dyad were not acquainted with each
other prior to the experiment. Participants knew when they volunteered that
a partner would be involved. More females volunteered to participate in the
study; consequently, random pairings of participants produced 12 male-
female and 30 female-female combinations. Two dyads (four participants)
were excluded from analysis because one of the participants in the dyad mis-
understood the instructions and failed to complete the task.
Assessments
To assess participants’ perceptions of their partners, each participant
completed an observer form of the NEO-Five Factor Inventory (NEO-FFI;
Costa & McCrae, 1991).This instrument is a shortened form of the NEO-Per-
sonality Inventory (NEO-PI), which taps the five-factor model of personality
(Neuroticism, Extraversion, Openness, Agreeableness, and Conscientious-
ness), a well-known and generalized model of personality (Costa & McCrae,
1992; McCrae & John, 1992). Cronbach’s alpha reliabilities from the present
administration, generated from cases that completed all 12 items of the rele-
vant factor, were estimated to be Neuroticism (n= 20), .71; Extraversion (n=
26), .83; Openness (n= 18), .74; Agreeableness (n= 28), .79; and Conscien-
tiousness (n= 32), .88. These reliability estimates are generally consistent
with previous assessments of this scale (Foltz,Morse, Calvo, & Barber, 1997).
In addition to its well-established psychometric properties, this instru-
ment was selected in an effort to assess a much wider array (breadth) of per-
sonal characteristics than has typically been measured in previous research.
The observer form of the NEO-FFI consists of 60 items, 12 for each of the five
traits. Each item or statement (e.g., This person is not a worrier) contained
six possible responses: strongly disagree,disagree,neutral,agree,strongly
agree and “cannot make judgment.” The instructions described this last
option as the appropriate response to the item if the participant felt he or she
had insufficient evidence about the partner to make that particular judg-
ment. The neutral option, in contrast,was described as the average response.
Participants were instructed to select one response for each item.
332
COMMUNICATION RESEARCH • June 2001
Two dependent measures corresponding with our hypotheses were
extracted from the observer form of the NEO-FFI. First, the number of the
NEO-FFI questions answered provided an index of breadth (i.e., how detailed
participants’ impression of their partners was). Similar to the operational
definition in Walther (1993),the more questions answered,or the fewer “can-
not make judgment” responses, the more comprehensively developed the
impression. In addition, the NEO-FFI permitted an analysis of impression
development across the five personality traits by examining the number of
questions answered within each trait domain.
Second, the intensity of each response was scored in terms of deviation
from the neutral point on the Likert scale. Scores for each question could,
therefore, range in either direction from 0 (average) to 2 (extreme).For exam-
ple, both strongly agree and strongly disagree responses were scored as a 2
(extreme); both agree and disagree responses were scored as a 1; neutral
responses were scored as 0 (neutral). Mean attribution ratings for each indi-
vidual were obviously based on only those questions answered by the
participant.
Procedure
On arrival at the laboratory, participants were led to separate rooms,
avoiding any contact between the members of a dyad prior to the task. At this
point, participants were told that they would collaborate to complete a figure-
matching task. This task was adapted from Schober and Clark (1989). One
participant, randomly designated the director, described a series of 12 tan-
grams (abstract shapes) to another participant, the matcher, and the latter
attempted to identify the tangrams being described from an array of 16 tan-
grams. The director was instructed to describe each tangram, one at a time,to
the matcher. Each tangram was discussed until the matcher claimed to have
identified it. The matcher was encouraged to ask questions during this pro-
cess and was instructed to notify the director when he or she believed the tan-
gram had been correctly matched. Note, however, that to minimize individu-
ating cues, participants were instructed not to reveal their name, age, or
gender. Finally, no time limit was imposed on the dyads.
Two additional points about the above-described procedures should, per-
haps, be emphasized. First, our decision to employ a task-oriented social
interaction (i.e., the tangram-task) was an effort to avoid imposing ceiling
effects on the breadth and/or intensity measures employed in the study. Intu-
ition suggests that a task-oriented social environment would support more
moderate levels of impression formation than a purely social exchange (e.g.,
333
Hancock, Dunham • Impression Formation
discussing an attitude-based topic), especially in the FtF condition. The sec-
ond point concerns our decision to instruct participants in the CMC condition
not to reveal their names, gender, or age. There are obviously a large number
of specific ways in which these global CMC and FtF environments differ (e.g.,
information differs regarding a partner’s race, age, gender, dress, hairstyle,
hygiene, prosody, affect, etc). These differences in available information con-
tribute to the deindividuation process that is emphasized in current theories
of CMC. Although this may threaten ecological validity to some degree, our
goal in including explicit instructions about name, age, and gender in the
CMC condition was to optimize deindividuation in this initial study. Presum-
ably, if reliable effects are observed under these optimal conditions, subse-
quent research can proceed to restore specific properties to the CMC environ-
ment and/or strip away specific characteristics of the FtF environment in an
effort to identify the roles played by the different elements of individuating
information that define each of these two more global conditions.
Participants assigned to the CMC condition performed the task at isolated
computer terminals. Participants used one of two desktop computer stations
while the experimenter monitored and recorded the interaction from a third
station. The three computers were networked such that signal transmission
was virtually instantaneous between stations. Each terminal was equipped
with the Windows95™ operating system and Mirabilis ICQ software (v.
Beta1.113). Partners could send and receive messages simultaneously, with
messages displayed on a What You See Is What I See (WYSIWIS), charac-
ter-by-character basis. Once participants were situated at their terminals,
the experimenter briefly demonstrated the use of the computer interface.
Participants assigned to the FtF condition performed the task in an inter-
action room where they sat at a table across from each other. The table had a
36 cm vertical barrier that prevented the participants from viewing each
other’s tangrams during the task.
Once the task was completed, participants in the FtF condition returned
to separate rooms to complete the NEO-FFI about their partner; participants
in the CMC condition remained at their computer station to complete the
measure. After completing the NEO-FFI observer form, each member of the
dyad was introduced to his or her partner and debriefed. Participants did not
know in advance that they would be meeting their partner.
Results
Recall that the number of responses on the observer form NEO-FFI is
assumed to index the breadth of impression development regarding the par-
ticipant’s partner. A 2 (communicative environment) ´5 (trait) mixed General
334
COMMUNICATION RESEARCH • June 2001
Linear Model (GLM) with trait as the repeated measure was conducted on
the number of items answered (see Table 1). Consistent with both the CFO
and hyperpersonal perspectives, a main effect of condition was obtained, F(1,
78) = 5.48, p< .05, h2= .07, with participants in the CMC environment
responding to significantly fewer questions per trait (M= 7.19, SD = 3.79)
than participants in the FtF environment (M= 8.95, SD = 3.54).
This difference, however, was not uniform across the five traits,as revealed
by a significant interaction, F(4, 312) = 2.54,p< .05, h2= .04. Planned compar-
isons on each trait across conditions revealed that the groups differed in
response frequency for three of the five traits, Extraversion,t(78) = 2.43, p<
.05; Agreeableness, t(78) = 2.33, p< .05; and particularly Neuroticism, t(78) =
2.45, p< .01. In contrast, the differences in response rate for Openness, t(78) =
1.68, n.s.; and Conscientiousness t(78) = 1.36, n.s.; were not statistically sig-
nificant and suggest that participants were prepared to make judgments
regarding their partner’s openness and conscientiousness in as much detail
in the CMC condition as in the FtF condition (see Table 1). It should also be
noted that the interaction effect obtained because high mean scores on some
dimensions (e.g., 8.28 for FtF Neuroticism) were lesser than low mean scores
on other dimensions (e.g., 9.23 for CMC Conscientiousness). Within each
dimension, directional relationships between FtF and CMC were consistent,
if not individually significant; the interaction is, therefore, ordinal and does
not challenge the main effects.
The next question of interest was whether attributions in the CMC envi-
ronment would be more intense than attributions made FtF. To address this
question, a second 2 (condition) ´5 (trait) mixed GLM was conducted on the
intensity scores derived from responses to the observer form of the NEO-FFI.
As predicted by the hyperpersonal model, a significant effect of condition was
obtained, F(1, 67) = 3.92, p< .05, h2= .06. The magnitude of the responses was
greater in the CMC condition (M= .92, SD = .23) than in FtF (M= .84, SD =
.26; see Table 2). The condition by trait interaction for attribution magnitude
was not significant, indicating that the increased intensity of judgments in
CMC was consistent across the different traits.2
Given that some dyads consisted of males and females (dyads = 11)
whereas the majority of dyads consisted of only females (dyads = 29), addi-
tional analyses were conducted to determine whether the gender of partici-
pants had an impact on either the breadth or intensity of impression develop-
ment. A 2 (condition) ´2 (male-female vs. female-female dyads) ´5 (trait)
mixed GLM was conducted first on the frequency scores and then on the
intensity scores. Neither analysis revealed any significant patterns across
conditions. A second analysis was conducted to examine whether the gender
of the perceiver influenced either dimension of impression development. A 2
335
Hancock, Dunham • Impression Formation
Table 1
Mean Frequency (Standard Deviation) of Responses to Items by Personality Factor
Neuroticism Extraversion Openness Agreeableness Conscientiousness Total
M (SD) M (SD) M (SD) M (SD) M (SD) M (SD)
Computer-mediated
conversation 5.78 (4.11) 7.30 (4.03) 6.30 (3.90) 7.32 (4.02) 9.23 (2.91) 7.19 (3.79)
Face-to-face
conversation 8.28 (4.03) 9.33 (3.34) 7.78 (3.97) 9.25 (3.33) 10.13 (3.01) 8.95 (3.54)
Mean difference 2.50 2.03 1.48 2.02 .90 1.76
Total 7.03 (4.24) 8.31 (3.84) 7.04 (3.98) 8.29 (3.80) 9.68 (2.98)
Note. Frequency scores could range from 0 to 12 for each trait.
336
Table 2
Mean Intensity (Standard Deviation) Scores by Personality Factor
Neuroticism Extraversion Openness Agreeableness Conscientiousness Total
M (SD) M (SD) M (SD) M (SD) M (SD) M (SD)
Computer-mediated
conversation .85 (.19) .85 (.25) .95 (.24) .97 (.24) .95 (.21) .92 (.23)
Face-to-face
communication .81 (.21) .77 (.27) .78 (.26) .91 (.28) .92 (.28) .84 (.26)
Mean difference .04 .08 .17 .06 .03 .08
Total .83 (.20) .82 (.26) .86 (.26) .94 (.23) .94 (.24)
Note. Intensity scores could range from 0 (Neutral) to 2 (Extreme).
337
(condition) ´2 (male vs. female perceiver) ´5 (trait) mixed GLM again
revealed no significant results. These results should be interpreted with
some caution, however, given that the low number of males and male-female
dyads in the sample reduced the power to detect any differences.
Finally, although the tangram task is not the focus of the present article, it
should be noted that no differences in task performance were observed across
communicative settings, t(38) = .40, n.s. This result is consistent with other
studies that have failed to observe task performance differences across CMC
and FtF conditions (Williams, 1977). Dyads in the FtF condition made as
many errors (M= 2.85, SD = 2.35) as did dyads in the CMC condition (M=
2.60, SD = 1.47). Furthermore,individual differences in task performance did
not correlate with any of the measures of impression formation (i.e., breadth
of impression, intensity of attributions).
As would be expected, significant differences were also observed between
conditions for task completion times, t(20.73, corrected for unequal vari-
ances) = 6.78, p< .001. Dyads communicating FtF completed the task more
quickly (M= 7.69 minutes, SD = 3.52) than dyads communicating via com-
puter completed the task (M= 36.22 minutes, SD = 18.49).As a consequence,
CMC participants were exposed to their conversational partners for a signifi-
cantly longer period of time than their FtF counterparts.
Discussion
The present research revisited the issue of impression formation in an
effort to examine both the breadth and intensity of impressions, across a
range of personality traits, that developed during a one-shot interaction in
synchronous CMC and FtF conditions. The present data speak to several
issues.
Consider, first, the breadth or comprehensiveness of impression formation
in the CMC and FtF environments. An examination of response frequency on
the observer form of the NEO-FFI revealed that, after a one-time interaction,
zero-history participants in a text-based communicative environment made
fewer attributions about their partners relative to participants communicat-
ing FtF. Indeed, setting aside specific traits,interactants in the CMC environ-
ment responded to only 59.9% of the 60 items; interactants in the FtF condi-
tion responded to 74.6%. This difference is consistent with Walther’s (1993)
observation that initial impressions formed in a CMC environment are rela-
tively incomplete relative to those formed during an FtF interaction.
This finding supports predictions flowing from both the CFO and
hyperpersonal perspectives. The broad reduction of nonverbal data about a
partner in a one-time CMC interaction apparently constrains the range of
338
COMMUNICATION RESEARCH • June 2001
information normally relied on to form more robust impressions. It is impor-
tant to note, however, that although CMC participants made fewer attribu-
tions about their partners, CMC participants nonetheless indicated that, in
absolute terms, sufficient information was available to rate their partners on
more than half of the 60 items (59.9%). We should also note that a ceiling
effect was not observed in the FtF condition, with participants responding to
substantially less than 100% of the items (74.6%).
Consider next the intensity of the impressions formed in each communica-
tive environment. An analysis of the magnitude of the attributions on the
observer form of the NEO-FFI revealed that, rather than developing neutral
impressions as suggested by the CFO perspective, participants in the CMC
condition made more extreme attributions than did partners communicating
FtF. That is, compared with impressions developed after an FtF interaction,
impressions of a partner’s personality profile after a CMC interaction were
more intense.
These data contrast with previous studies that have failed to observe sig-
nificant differences in the intensity of impressions across FtF and CMC con-
ditions (e.g., Lea & Spears, 1991; Matheson & Zanna, 1988). Our tentative
explanation for this discrepancy is that the wider array of personal charac-
teristics assessed in the present study, and the fact that participants were not
required to make intensity ratings on every item, may constitute a more sen-
sitive measure of the intensity dimension than those employed in previous
research.
The more intense partner ratings observed are, however, consistent with
Walther’s (1997) findings indicating that, within a CMC environment in
which individual identities are not made salient, participants can make
exaggerated attributions regarding a partner’s personal qualities (e.g.,phys-
ical and social attractiveness, productivity, etc.). In addition, although
Walther did not explicitly manipulate FtF and CMC conditions in this study,
the attractiveness ratings of collocated participants, who potentially used
alternate methods of communication over the course of the study (e.g., FtF,
telephone), were compared with the ratings of distributed partners,who com-
municated only by CMC. Congruent with the present data,the ratings of col-
located (CMC/FtF) participants were more neutral than the ratings of the
distributed (CMC only) partners, who produced more intense attributions
(e.g., more or less socially attractive).
Indeed, the pattern that emerges from the breadth and intensity data in
the present study provides general support at the initial stage of impression
formation for Walther’s (1996, 1997) hyperpersonal model. The less broadly
developed impressions in the CMC condition are consistent with the tempo-
ral constraints associated with the differential rates of social information
339
Hancock, Dunham • Impression Formation
processing specified within the hyperpersonal framework, whereas the
increased intensity observed in the CMC partner ratings is in line with an
underlying assumption of the hyperpersonal model that “CMC partners pro-
duce overattributions of each other based on the minimal social cues con-
veyed by the medium, in the absence of contraindicating information”
(Walther, 1997, p. 350).
The hyperpersonal model also suggests that as the interaction proceeds,
participants begin to actively generate adaptive strategies for reducing
uncertainty about a partner’s personal qualities in text-based communica-
tion (Walther, 1996). Presumably, given this relatively short period of interac-
tion, CMC participants were unable to seek out and gather sufficient impres-
sion-relevant information about their partners to match the breadth of the
impressions formed by FtF communicators, although the short duration of
their interactions did not prevent participants from producing exaggerated
attributions.
An important question remains, therefore, as to how these differences in
breadth and intensity will change over time. Our reading of the
hyperpersonal model suggests that with more interactions over a longer
period of time, CMC/FtF differences in the breadth of impressions should dis-
appear as CMC participants learn more about each other. Predictions are less
clear for the intensity measure. On one hand,as more individuating informa-
tion is learned about a partner, impressions of that partner should be based
more on interpersonal information and less so on exaggerated, social iden-
tity-derived stereotypes. With the increased individuating information,
impressions should become less intense (Spears & Lea, 1992; Walther,1997).
On the other hand, the unique affordances of the CMC environment outlined
by the hyperpersonal model indicate that impressions can become more
intensified over time as participants engage in selective self-presentation
and cognitive reallocation and as intensification processes such as behavioral
confirmation begin to operate (Walther, 1997). Future longitudinal research,
in which both breadth and intensity of measures are assessed over time, is
obviously an important next step in addressing these questions adequately.
Several other factors may also have played a role in the impression forma-
tion process in the present context. These include (a) the synchronous format
of the CMC condition, (b) the mechanical nature of the tangram task, and (c)
the instruction set. The synchronous format employed here is a departure
from the asynchronous formats (e.g., e-mail, newsgroup) employed in previ-
ous research on the hyperpersonal model and its predecessor, the social infor-
mation-processing theory (e.g.,Walther,1993,1997).Although the use of this
format extends the conditions for which the hyperpersonal model has
received support, it also presents the possibility that the array of active
340
COMMUNICATION RESEARCH • June 2001
strategies and cognitive mechanisms described by the hyperpersonal model
to facilitate overattributions was more difficult to implement. Cognitive real-
location, for example, may have been less feasible given the online, simulta-
neous nature of the synchronous format. Similarly, both the instructions to
not reveal one’s name, age,or gender and the task-oriented nature of the tan-
gram problem-solving conversations may have dissuaded or discouraged
participants from explicitly seeking information about their partners.
Indeed, an examination of the transcripts indicates that active interpersonal
inquiry, defined as questions focused on learning about a partner’s personal
qualities, were not observed in the CMC condition.
These factors did not, however, eliminate interpersonal information and
signals from being presented in the CMC condition. A partner’s choice of
descriptive devices (e.g., geometric vs. analogic descriptions), communicative
style, and paralanguage (e.g., use of emoticons, punctuation, capitalization,
etc.) all provided potentially impression-relevant information. Consider the
following example:
Director: OK, the next one looks like an altar boy. He is holding something
out.
Matcher: What exactly does an altar boy look like?
Director: Sorry, I am not too religious myself.
In this example, the director infers from the matcher’s question that the
matcher is not religious, and this piece of information may factor into the
director’s judgments of his partner. In the director’s response, subtle selective
self-presentation is also evident, with the director describing himself as simi-
larly not religious.
In the next example, the style of the communication, rather than its con-
tent, can be seen as potentially influential in the formation of an impression.
Matcher: Thanks.
Director: No problemo :)
Matcher: Um . . . the body — kind of like a
Director: A parallelogram.
Matcher: —thanks, that’s the word I was looking for. :-)
Director: :)
The use of ellipses, emoticons, hyphens, and the collaborative nature of the
interaction in this example all demonstrate some of the additional nonverbal
forms of information about a partner’s personality available in text-based
discourse.
341
Hancock, Dunham • Impression Formation
As described by the SIDE model, and incorporated into the hyperpersonal
framework, the salience of these social and interpersonal cues is enhanced
within the CMC condition’s deindividuating context (Lea & Spears, 1992;
Spears & Lea, 1992; Walther, 1996). The exaggerated, stereotypical impres-
sions formed in this condition were presumably based on, and biased by, these
minimal cues, in the absence of other individuating information.
It is also of some interest to note that these data have some implications
for cross-modal approaches to the examination of impression formation. Pre-
vious research has compared trait-related information presented in the
visual and audio modalities (Borkenau & Liebler, 1992; Maxwell, Cook, &
Burr, 1985). Borkenau and Liebler (1992), for example, demonstrated that
some traits, such as extraversion, are consistently expressed in both the
visual (e.g., clothing style, extent of smiling, etc.) and audio (e.g., loud, power-
ful voice, etc.) modalities. The use of the NEO-FFI in the present experimen-
tal context raises similar questions about what type of trait-specific informa-
tion is expressed in text-based verbal exchanges.
The significant interaction between communicative environment and
trait domain observed for response frequencies revealed that the influence of
a textual communication format on the breadth of impression formation fol-
lowing a single interaction was not equivalent across the five trait domains.
Significant differences existed between conditions for Neuroticism,
Extraversion, and Agreeableness,with participants in the CMC setting mak-
ing fewer attributions for these three trait domains than participants that
interacted FtF. In contrast, participants rated Conscientiousness and Open-
ness in equivalent detail across the CMC and FtF environments.
Why is it that participants interacting in CMC would be less willing to
rate their partners on the traits of Neuroticism, Extraversion, and Agree-
ableness than participants interacting FtF? Although any attempt to answer
this question is, at this point,admittedly speculative, there are relevant anal-
yses in the traditional FtF literature to suggest that the specific nature of the
cues present in CMC may selectively undermine impressions of certain traits
and provide adequate information about others. Traits have been suggested
to vary, for example, in their general observability (Kenrick & Stringfield,
1980) or visibility to an observer (Funder & Dobroth, 1987; John & Robins,
1993; Kenrick, McCreath, Govern, King, & Bordin, 1990). From this perspec-
tive, the impressions of highly visual traits, such as Extraversion, should be
undermined in a text-based CMC interaction whereas impressions of less
visually observable traits, such as Neuroticism and Openness, should be less
affected (see Table 3 for example items and their level of visibility).Although
such an analysis accounts for some of the patterns observed in the present
data, this approach seems to be an oversimplification. According to our data,
342
COMMUNICATION RESEARCH • June 2001
as would be expected, impressions of highly visible traits (e.g., Extraversion)
were undermined in the CMC condition relative to the FtF condition whereas
impressions of less visible traits (e.g.,Openness) were equally detailed across
conditions. However, impressions of the least observable trait, Neuroticism,
were impoverished in CMC relative to FtF. Participants were more hesitant
to rate this trait in CMC, suggesting that factors other than visibility must be
involved in ratings of Neuroticism, factors that are available in a single FtF
interaction but not in a single CMC interaction.
We did not anticipate these trait-specific effects, and a more definitive
understanding of the mechanisms underlying them will depend on addi-
tional research. Studies that systematically enrich the text-based CMC envi-
ronment by adding various forms of visual and auditory information should
help us to assess the impact of these different modalities on trait-specific
impressions formed across various task settings.
Finally, although the CMC environment is a unique and potentially
important addition to the cross-modal approach to impression formation, the
present research highlights and confirms the limitations inherent in a tech-
nologically deterministic, cues filtered-out approach to interpersonal pro-
cesses in mediated communication (Garton & Wellman, 1995; Spears & Lea,
1992, 1994; Walther, 1996). Although the reduction of interpersonal and
social cues in a CMC environment is an important factor in impression for-
mation, the impact of the cognitive processes involved must also be taken into
account to predict accurately both the breadth and intensity of impressions.
Indeed, the explanatory advantage of the hyperpersonal, social information-
343
Hancock, Dunham • Impression Formation
Table 3
Example NEO-Five Factor Inventory (NEO-FFI) Items Demonstrating the Visibility
of Traits
Visibility
Trait High Low
Neuroticism Sometimes this person feels
completely worthless
Extraversion This person laughs easily
Openness Poetry has little or no effect on
this person
Agreeableness This person tries to be
courteous to everyone
Conscientiousness This person is not a very
methodical person
processing, and SIDE models is derived directly from their consideration of
the cognitive strategies (e.g., information seeking, selective self-presentation)
and heuristics (e.g., social identity and stereotyping) involved in the forma-
tion of impressions. The current research also underlines the importance of
considering both the breadth and intensity dimensions of impression forma-
tion. Procedures that require participants to respond to all items on a scale
may conceal important differences in the breadth of impression development
or the rater’s confidence associated with each attribution.
Notes
1. This research was supported by grants from the Natural Sciences and Engi-
neering Research Council of Canada and the Social Sciences and Humanities Research
Council, awarded to Philip J. Dunham. A Social Sciences and Humanities Research
Council fellowship and a Killam doctoral scholarship supported Jeffrey T. Hancock.
The authors thank Diana Chard for assistance in the collection and analysis of these
data, Frances Dunham for technical support, and an anonymous reviewer for com-
ments on earlier drafts of this manuscript.
2. A directional analysis of the intensity scores revealed that there was no system-
atic low- or high-score bias in the trait attributions across conditions. That is,the more
intense scores observed in the CMC condition were bidirectional and represented both
higher and lower attributions than those observed in the FtF condition. Note that this
observation is consistent with a hyperpersonal interpretation of the results, in which
partners are assumed to come to know each other more intensely but not necessarily
unidirectionally (i.e., only positive or only negative impression).
References
Anderson, N. H. (1981). Foundations of information integration theory. New
York: Academic Press.
Bem, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Advances in
experimental social psychology (Vol. 6, pp. 1-62). New York: Academic
Press.
Borkenau, P., & Liebler, A. (1992). The cross-modal consistency of personal-
ity: Inferring strangers’ traits from visual or acoustic information. Jour-
nal of Research in Personality,26, 183-204.
Cantor, N., & Mischel, W. (1977). Traits as prototypes: Effects on recognition
memory. Journal of Personality and Social Psychology,35, 38-48.
Clark, H. H. (1996). Using language. Cambridge, UK: Cambridge University
Press.
Costa, P. T.,& McCrae, R. R. (1991). NEO-FFI Form S.Odessa, FL: Psychologi-
cal Assessment Resources.
Costa, P. T., & McCrae, R. R. (1992). NEO-PI/FFI manual supplement.
Odessa, FL: Psychological Assessment Resources.
344
COMMUNICATION RESEARCH • June 2001
Culnan, M. J., & Markus, M. L. (1987). Information technologies. In F. Jablin,
L. L. Putnam, K. Roberts, & L. Porter (Eds.), Handbook of organizational
communication (pp. 420-443). Newbury Park, CA: Sage.
Dubrovsky, V.J., Kiesler, S., & Sethna, B. N. (1991). The equalization phenom-
enon: Status effects in computer-mediated and face-to-face decision- mak-
ing groups. Human Computer Interaction,6, 119-146.
Fiske, S. T., & Taylor, S. E. (1991). Social cognition (2nd ed.). New York:
McGraw-Hill.
Foltz, C., Morse, J. Q., Calvo, N., & Barber, J. P. (1997). Self- and observer rat-
ings on the NEO-FFI in couples: Initial evidence of the psychometric prop-
erties of an observer form. Assessment,4, 287-295.
Funder, D. C. (1987). Errors and mistakes: Evaluating the accuracy of social
judgment. Psychological Bulletin,101, 75-90.
Funder, D. C., & Dobroth, K. M. (1987). Differences between traits: Properties
associated with inter-judge agreement. Journal of Personality and Social
Psychology,52, 409-418.
Garton, L., & Wellman, B. (1995).Social impacts of electronic mail in organi-
zations: A review of the research literature. In B. R. Burleson (Ed.), Com-
munication yearbook (Vol. 18, pp. 434-453). Thousand Oaks, CA: Sage.
Hiltz, R.S., Johnson, K., & Turoff, M. (1986). Experiments in group deci-
sion-making: Communication process and outcome in face-to-face versus
computerized conferences. Human Communication Research,13,
225-252.
Hogg, M. A., & Abrams, D. (1988). Social identification. London: Routledge.
Jacobson, D. (1999). Impression formation in cyberspace: Online expectations
and offline experiences in text-based virtual communities. Journal of
Computer-Mediated Communication,5[Online]. Available: www.
ascusc.org/jcmc/vol5/issue1/jacobson.html
Johansen, R., Vallee, J., & Spangler, K. (1979). Electronic meetings: Technical
alternatives and social choices. London: Addison-Wesley.
John, O. P., & Robins, R. W. (1993). Determinants of interjudge agreement on
personality traits: The big five domains, observability, evaluativeness, and
the unique perspective of the self. Journal of Personality,61, 521-551.
Jones, E. E. (1990). Interpersonal perception. New York: W. H. Freeman.
Kenrick, D. T., McCreath, H. E., Govern, J., King, R., & Bordin, J. (1990). Per-
son-environment intersections: Everyday settings and common trait
dimensions. Journal of Personality and Social Psychology,58, 685-698.
Kenrick, D.T.,& Stringfield,D. O. (1980).Personality traits and the eye of the
beholder: Crossing some traditional philosophical boundaries in the
search for consistency in all of the people. Psychological Review,87,
88-104.
Kiesler, S., Siegel, J., & McGuire, T. W. (1984). Social psychological aspects of
computer-mediated communication. American Psychologist,39,
1123-1134.
Lea, M., & Spears, R. (1991). Computer-mediated communication, de-
individuation, and group decision-making. International Journal of
Man-Machine Studies,34, 283-301.
345
Hancock, Dunham • Impression Formation
Lea, M., & Spears, R. (1992). Paralanguage and social perception in com-
puter-mediated communication. Journal of Organizational Computing,2,
321-342.
Lea, M., & Spears, R. (1995). Love at first byte? Building personal relation-
ships over computer networks. In J. T. Wood & S. Duck (Eds.), Under-
studied relationships: Off the beaten track (pp. 197-233). Thousand Oaks,
CA: Sage.
Matheson, K., & Zanna, M. P. (1988). The impact of computer-mediated com-
munication on self-awareness. Computers in Behavior,4, 221-233.
Maxwell, G. M., Cook, M. W., & Burr, R. (1985). The encoding and decoding of
liking from behavioral cues in both auditory and visual channels. Journal
of Nonverbal Behavior,9, 239-263.
McCrae, R. R., & John, O. P. (1992). An introduction to the five-factor model
and its applications. Journal of Personality,60, 175-215.
Parks, M. R., & Floyd, K. (1996). Making friends in cyberspace. Journal of
Communication,46, 80-96.
Scherer, K. (1979). Personality markers in speech. In K. Scherer & H. Giles
(Eds.), Social markers in speech. Cambridge, UK: Cambridge University
Press.
Schober, M. F., & Clark, H. H. (1989). Understanding by addressees and over-
hearers. Cognitive Psychology,21, 211-232.
Semin, (1989). Impressions of personality revisited: The contribution of lin-
guistic factors to attribute inferences. European Journal of Social Psychol-
ogy,19, 85-101.
Short, J., Williams, E., & Christie, B. (1976). The social psychology of telecom-
munications. London: John Wiley.
Siegel, J., Dubrovsky, V., Kiesler, S., & McGuire, T. (1986). Group processes in
computer-mediated communication. Organizational Behavior and Human
Decision Processes,37, 157-187.
Spears, R.,& Lea,M.(1992).Social influence and the influence of the ‘social’ in
computer-mediated communication. In M. Lea (Ed.), Contexts of computer-
mediated communication (pp. 30-65). London: Harvester-Wheatsheaf.
Spears, R., & Lea, M. (1994). Panacea or panopticon? The hidden power in
computer-mediated communication. Communication Research,21, 427-459.
Sproull, L., & Kiesler, S. (1986). Reducing social context cues: Electronic mail
in organizational communication. Management Science,32, 1492-1512.
Walther, J.B. (1993). Impression development in computer-mediated interac-
tion. Western Journal of Communication,57, 381-398.
Walther, J. B. (1996). Computer-mediated communication: Impersonal, inter-
personal, and hyperpersonal interaction. Communication Research,23,
3-43.
Walther, J. B. (1997).Group and interpersonal effects in international com-
puter-mediated collaboration. Human Communication Research,23,
342-369.
Walther, J. B., Anderson, J. F., & Park, D. (1994). Interpersonal effects in
computer-mediated interaction: A meta-analysis of social and anti-social
communication. Communication Research,21, 460-487.
346
COMMUNICATION RESEARCH • June 2001
Walther, J. B., & Burgoon, J. K. (1992). Relational communication in computer-
mediated interaction. Human Communication Research, 19, 50-88.
Williams, E. (1977). Experimental comparisons of face-to-face and mediated
communication: A review. Psychological Bulletin,84, 963-976.
Wyer, R. S., & Srull, T. K. (1989). Memory and cognition in its social context.
Hillsdale, NJ: Lawrence Erlbaum.
Wyer, R. S., Swan, S., & Gruenfeld, D. H. (1995). Impression formation in
informal conversations. Social Cognition,13, 243-272.
347
Hancock, Dunham • Impression Formation
