Content uploaded by Gwyneth Doherty-sneddon
Author content
All content in this area was uploaded by Gwyneth Doherty-sneddon on Oct 16, 2020
Content may be subject to copyright.
Memory & Cognition
2005, 33 (4), 727-733
During difficult cognitive activity—for example, re-
membering information, thinking of an answer to a ques-
tion, planning what we are going to say, or speaking—we
often close our eyes, look up at the sky, or look away from
the person we are in conversation with (Doherty-Sneddon,
Bruce, Bonner, Longbotham, & Doyle, 2002; Glenberg,
Schroeder, & Robertson, 1998). A number of studies
have reported ways in which adults switch off from en-
vironmental stimulation (both live faces and other sorts
of visual displays) in order to concentrate on cognitive
tasks (e.g., Beattie, 1981; Glenberg et al., 1998). Our own
work has shown that older children (8-year-olds) also
look away more when answering difficult questions, as
compared with easy ones, when questioned face to face
(FTF; Doherty-Sneddon et al., 2002; Phelps, Doherty-
Sneddon, & Warnock, in press). Gaze aversion is, there-
fore, potentially a useful cue during pedagogical interac-
tions, since it gives a nonverbal indication of a child’s
level of understanding and concentration (see Doherty-
Sneddon et al., 2002). This article reports a study of gaze
aversion by 8-year-olds when they were asked questions
of increasing difficulty. The children were questioned by
the same interviewer either across a live video link (LVL)
or FTF, to ascertain what underlies our tendency to avert
our gaze when we need to concentrate.
Considerable research effort has been expended on ex-
amining the role played by visual communication signals
(e.g., eye gaze, gestures, and facial expressions) in human
interaction. There is much evidence that these cues are
often important sources of information, and many re-
searchers have proposed that they play a facilitatory role
in human communication (e.g., Clark & Brennan, 1991;
Goldin-Meadow, Wein, & Chang, 1992; McNeill, 1985).
However, the fact that such signals are informative means
that they carry a cognitive load. The processing costs of
visual signals have been documented. Excessive eye gaze
between speakers is associated with increased cognitive
load, evidenced by, for example, less fluent speech (Beat-
tie, 1981). In addition, the cognitive difficulty of a task
relates to the likelihood that people will avert their gaze
from other people’s faces (Doherty-Sneddon et al., 2002;
Ellyson, Dovidio, & Corson, 1981; Glenberg et al., 1998;
Phelps et al., in press). So one explanation of the link be-
tween cognitive difficulty and gaze aversion is that the
interlocutor’s face, an information-rich aspect of the en-
vironment, requires cognitive resources to monitor (Glen-
berg, 1997; Glenberg et al., 1998). When people avert their
gaze, they can deploy additional cognitive resources to
the task at hand and, hence, improve their performance—
the cognitive load hypothesis (cf. Glenberg, 1997).
In addition to the cognitive influences on gaze behav-
ior, social psychologists have drawn attention to the social
constraints on gaze in conversation. Gaze serves a num-
ber of social functions, including being a signal of inti-
macy, dominance, and social competence (e.g., Argyle,
1996; Argyle & Dean, 1965; Burgoon, Manusov, Mineo,
& Hale, 1985). Argyle suggests that gaze indicates inti-
macy, and levels must be negotiated to maintain a suitable
equilibrium in conversation. For example, when speakers
are physically close, this leads to a higher level of inti-
macy that is compensated for by gaze being averted from
727 Copyright 2005 Psychonomic Society, Inc.
This work was supported by ESRC Grant R000239930, held by G.D.-S.
We thank the children who participated in this research and their teach-
ers and parents for their cooperation and consent. Correspondence re-
garding this article should be addressed to G. Doherty-Sneddon, De-
partment of Psychology, University of Stirling, Stirling FK9 4LA,
Scotland (e-mail: gds1@stir.ac.uk).
Gaze aversion: A response to cognitive
or social difficulty?
G. DOHERTY-SNEDDON and F. G. PHELPS
University of Stirling, Stirling, Scotland
When asked questions, adults and children often avert their gaze at certain points within the inter-
action, especially when questions are difficult (Doherty-Sneddon, Bruce, Bonner, Longbotham, &
Doyle, 2002; Glenberg, Schroeder, & Robertson, 1998). Gaze aversion may be a way of managing the
cognitive load associated with the processing of visual environmental information, or it may serve to
alleviate a negative social-emotional experience, such as the self-consciousness associated with, for ex-
ample, a fear of failure. In the present study, thirty-six 8-year-olds were questioned either face to face
or across a live video link. Questions varied in type (arithmetic, verbal reasoning, and autobiographi-
cal and episodic memory) and in difficulty. Children averted their gaze more during face-to-face ques-
tioning than during video-mediated questioning; however, question difficulty had a very strong influ-
ence on aversion in both interview conditions. It is concluded that although social factors play a role
in children’s gaze aversion during pedagogical question–answer sequences, the primary function of
averting gaze is to manage the cognitive load involved in the processing of environmental information.
728 DOHERTY-SNEDDON AND PHELPS
an interlocutor’s face. Question topic also influences in-
timacy distance. Exline, Gray, and Schuette (1965) found
that people engaged in less eye contact when asked em-
barrassing questions than when asked innocuous ques-
tions. Therefore, if a question is difficult or personally
salient, this may increase self-consciousness or embar-
rassment, thereby leading to increased gaze aversion to
compensate for this. Similarly, Stanley and Martin (1968)
proposed that eye contact is associated with an increase
in anxiety and that gaze aversion plays a role in anxiety
reduction. Part of the load associated with looking at
faces, especially when difficult questions are answered,
may, therefore, be associated with self-consciousness. It
is, therefore, important to understand the relative impact
of social and cognitive influences on gaze aversion, since
one set of reasons why children avert their gaze during
difficult questioning may be because they feel embar-
rassed or self-conscious and want to avoid looking at the
questioner. The question addressed by the present article
is not whether gaze aversion per se is influenced by so-
cial or cognitive factors, since both clearly contribute to
patterns of gaze. The key issue is whether the increase in
aversion associated with increasing difficulty of ques-
tions is driven primarily by a need to reduce environ-
mental stimulation, especially visual stimulation from
the face (faces are a particularly salient type of visual
form; Russell & Lavie, 2001), or to alleviate the self-
consciousness that may accompany being “put on the
spot” by difficult questioning.
The study of gaze aversion has important practical ap-
plications. For example, Vrij (2002) cited gaze aversion
as a nervous behavior commonly believed to be associ-
ated with deception. He proposed that people who are in-
tentionally trying to deceive others often feel nervous
about being found out and will also have to “think hard”
in order to make their lies convincing. In addition, he
predicted that younger children are likely to exhibit more
signs of hard thinking and nervousness (including more
gaze aversion) while lying—in part, because they are
less able to control their behavior. However, a number of
studies have clearly shown that gaze aversion is not a re-
liable cue to deception (e.g., Zuckerman & Driver, 1985).
Recent work in which gaze aversion in children has been
investigated has shown that young children (5 years and
younger) were inconsistent in their use of gaze aversion
as a response to cognitive difficulty and often, for ex-
ample, “stared the adult out” when they experienced dif-
ficulties (Doherty-Sneddon et al., 2002). Understanding
gaze aversion is, therefore, crucial in order to success-
fully inform applications of knowledge.
Glenberg et al. (1998) have reported that adults avert
their gaze when answering difficult questions even when
these are presented on a computer screen, suggesting
that gaze aversion functions to reduce environmental
stimulation per se, rather than to reduce social factors,
such as feeling embarrassed. In the present study, we ma-
nipulate social factors by comparing FTF questioning
with video-mediated questioning. Evidence that video-
mediated communication (VMC) has important social ef-
fects on users comes from research by Doherty-Sneddon
and colleagues. She investigated how children of differ-
ent ages use verbal and nonverbal signals in FTF and
video-mediated interactions (e.g., Doherty-Sneddon &
Kent, 1996; Doherty-Sneddon & McAuley, 2000). In-
cluded in this work was a study of the impact of video
links on children when they were questioned as witnesses.
In a number of countries, including the United States,
Canada, and the United Kingdom, when children are
called as witnesses in criminal trials, they are sometimes
given the option of being cross-examined via LVLs rather
than in open court (Davies, 1999). In a study of real-life
court testimony of children, Davies and Noon (1991)
found that LVLs reduced stress in child witnesses and
improved the quality of their evidence: LVL children
were rated as more fluent and audible; they were rated as
happier and were more likely to be judged as competent
to swear the oath. Doherty-Sneddon and McAuley (2000)
found that children experienced an increase in social dis-
tance between themselves and the interviewer in VMC
that influenced many aspects of their verbal and nonverbal
communication (e.g., an increased resistance to misleading
questions and a more relaxed demeanor). In other words,
the children appeared to be less self-conscious and more
socially confident when interviewed across the video
link.
These results are consistent with findings from studies
of adults using live video links. Overgazing by adult users
of VMC has been reported by a number of authors (e.g.,
Abel, 1990). Indeed, our own research has shown that
adults look at one another around 56% more when com-
municating via LVLs, as compared with FTF (Doherty-
Sneddon et al., 1997). This suggests that the social norms
relating to gaze are different when interactions are me-
diated, even for adults. The present study is novel in that
we investigate whether the use of gaze aversion as a cog-
nitive management strategy changes when gaze is video
mediated versus FTF.
Of course, as well as increased social distance, there
can be an attenuation of visual cues in VMC (see Doherty-
Sneddon et al., 1997). In the present study, we used a
life-sized head and upper body shot in the mediated con-
dition, so that the size of the mediated face was equiva-
lent to that of the live face, allowing good discrimination
of facial features and movement. Furthermore, the me-
diated condition allowed for mutual eye contact, a very
important part of FTF interaction. Although it is possi-
ble that physical aspects of the display (e.g., loss of depth
cues in a two-dimensional image) meant that the medi-
ated face was a less potent source of visual information,
this is unlikely, since a number of researchers have shown
that high-quality VMC (e.g., one with no delay in the
audio signal) results in the transmission of amounts of
information equivalent to those found in FTF interaction
(e.g., Anderson et al., 1997). In addition, Bruce (1995)
CHILDREN’S GAZE AVERSION 729
reported that perception of facial expressions remains in-
tact in video-mediated interactions.
So gaze aversion, from the face of an interlocutor, oc-
curs more often during cognitively demanding tasks, to
reduce cognitive load by reducing visual environmental
stimulation and/or to alleviate self-consciousness. In this
study, we attempted to tease apart these aspects. We did
this by comparing gaze aversion in video-mediated in-
teractions with that in FTF encounters. In the LVL con-
dition, it was assumed (on the basis of previous research)
that there would be an increase in social distance be-
tween the child and the questioner, reducing the social
impact of the questioner’s face and changing the norms
relating to acceptable levels of gaze. We therefore pre-
dicted that the children would avert their gaze less dur-
ing the video-mediated interactions. Furthermore, if the
tendency to avert during difficult questions is due to the
self-consciousness associated with being asked difficult
questions, we would predict that the difficulty of the
questions would have less of an influence on gaze aver-
sion when social presence was alleviated. If, however,
children avert their gaze during difficult questions in
order to reduce incoming environmental information,
question difficulty should influence gaze aversion re-
gardless of whether the interaction was mediated or not.
In this study, we investigated 8-year-olds’ gaze aver-
sion. This age group was chosen because it was one that
we had previously found to respond to increasing ques-
tion difficulty by increasing the use of gaze aversion in
a way that was equivalent to that for adults. For example,
Doherty-Sneddon et al. (2002) found that, like adults,
8-year-olds looked away more when asked difficult ver-
bal and arithmetic questions. In addition, their gaze aver-
sion peaked while the children were thinking, followed
by when they were speaking, with the lowest amount of
aversion occurring during listening. This is the same pat-
tern as that found with adults. In contrast, younger chil-
dren (5-year-olds) apply gaze aversion much less con-
sistently and to a lesser extent. The authors’ most recent
work in progress suggests that there is an important de-
velopment of gaze aversion behaviors between 5 and
6 years of age (Phelps et al., in press). By 8 years of age,
adult patterns are firmly established. Furthermore, by 8
years, children have acquired considerable knowledge of
the social rules about gazing and are likely to use gaze
and gaze aversion in response to social situations in ways
beginning to be similar to those of adults (Abramovitch
& Daly, 1978). The study of 8-year-olds, therefore, will
extrapolate well to older children and adults.
We questioned 8-year-old children either FTF or across
an LVL. Questions varied in both type (arithmetic, ver-
bal reasoning, autobiographical memory, or episodic
memory) and difficulty (easy, medium, or hard). The hy-
potheses of this study were the following: (1) We pre-
dicted that children would avert their gaze less when in-
terviewed across an LVL, as compared with FTF, and
(2) on the basis of our own previous work (Doherty-
Sneddon et al., 2002; Phelps et al., in press), we also ex-
pected that gaze aversion would be more frequent when
the children were asked moderate and very difficult ques-
tions, as compared with easy ones. If this functions pri-
marily to alleviate self-consciousness, we would expect
to see a greater effect of task difficulty in FTF interaction
than in video-mediated questioning. If however, children
increase their aversion primarily to switch off from visual
environmental stimulation, we would expect an effect of
question difficulty in both interview conditions.
METHOD
Participants
Thirty-six children (16 boys and 20 girls) took part in the study;
their mean age was 8 years 5 months (range ⫽8 years 3 months to
9 years 5 months). Children were randomly allocated to one of two
interview conditions: 17 were interviewed via an LVL (7 boys and
10 girls; Mage ⫽8 years 5 months, range ⫽8 years 4 months to 9
years 4 months) and 19 were interviewed FTF (9 boys and 10 girls;
Mage ⫽8 years 4 months, range ⫽8 years 3 month to 9 years 5
months). A further 2 children were allocated to the LVL condition.
Of these, 1 chose not to participate; the data for the other child were
lost, owing to equipment failure. Unpaired comparisons showed
that the groups were comparable in terms of their age [t(34) ⫽0.56,
p⫽.28]. Inclusion criteria specified English as the f irst language.
None had previously participated in related experiments.
Stimuli
Four different types of questions were posed to the children: ver-
bal (n⫽36), mental arithmetic (n⫽36), episodic memory (n⫽
18), and autobiographical memory (n⫽18). Both the verbal ques-
tions and the arithmetic questions were based on stimuli described
in the Wechsler Preschool and Primary Scale of Intelligence, where
verbal questions required the children to define words (n⫽9), spell
words (n⫽9), repeat word lists (n⫽9), and give information about
concepts (n⫽9). The mental arithmetic questions involved addi-
tion (n⫽9), subtraction (n⫽9), multiplication/division (n⫽9),
and number use (n⫽9). The autobiographical questions involved
asking the children to recollect events that they had personally ex-
perienced, such as information pertaining to recent classroom ac-
tivities. The episodic memory questions were based on stimuli de-
scribed in the Wechsler Memory Scale–Revised (WMS–R). This
involved instructing the children to remember stimuli paired in time
(e.g., car–red or ball–green). The experimenter then provided one
half of each pairing (e.g., “what color was the car?”), and the child’s
task was to correctly recollect the event with which it had previously
been paired (i.e., red ).
For each type of question, there were three levels of difficulty:
easy, medium, and hard. These levels of difficulty were deemed to
be appropriate through consultation with the participants’ teachers
(n⫽3), who rated how difficult an average 8-year-old would find
each question. Rating instructions stressed that the questions would
be read aloud to the child, who, in turn, would give a verbal re-
sponse. Teachers rated the difficulty of an initial set of 232 ques-
tions, using a 7-point scale, with 1 labeled very easy, through 4 la-
beled moderately easy, through 7 labeled very hard. These ratings
were used to establish the mean difficulty value for each question,
and enabled the compilation of a controlled set of questions (n⫽
108), with questions falling under three distinct levels of difficulty:
easy (mean difficulty score ranging from 1.0 to 2.0), medium (mean
difficulty score ranging from 3.5 to 4.5), and hard (mean difficulty
score ranging from 6.0 to 7.0). Teachers’ ratings for the difficulty
levels of critical stimuli correlated highly [Teacher 1 vs. Teacher 2,
r(107) ⫽.86; Teacher 1 vs. Teacher 3, r(107) ⫽.92; Teacher 2 vs.
Teacher 3, r(107) ⫽.91]. Examples of easy, medium, and hard
verbal-vocabulary questions were What is a photograph? What is
730 DOHERTY-SNEDDON AND PHELPS
a menu? and What is humor? respectively. Examples of the differ-
ent levels of arithmetic questions were: 9⫺6⫽__, 30⫺12 ⫽__ ,
and 265⫺34 ⫽__. Easy, medium, and hard autobiographical ques-
tions were, for example, “How many points did your class group
get: this fortnight; last fortnight; two fortnights ago?” Easy episodic
questions involved the child’s remembering two different pairings
of information, medium questions involved three pairings, and hard
questions involved four.
Apparatus
Questions were asked either FTF or via LVL. In the case of the
LVL, images of the face and upper body were transmitted through
the use of video tunnels (Smith, O’Shea, O’Malley, Scanlon, & Tay-
lor, 1991). Each of these consisted of a color monitor (JVC TM-
90PSN, 35 cm diagonally) mounted in a wooden box behind a half-
silvered mirror. A second, fully silvered mirror was fixed parallel
above the half-silvered mirror, so that light reflected from the first
mirror would reflect off the second into a video camcorder (Pana-
sonic M10) located behind the box and above the TV monitor. Be-
cause of the half-silvered nature of the first mirror, the child, look-
ing directly into the video tunnel, was able to see an image of the
experimenter on the monitor. The child’s image was also reflected
off the mirror into a second camcorder, from which it was trans-
mitted to the monitor in the experimenter’s identically arranged
video tunnel. With correct positioning of the camera, the child and
the experimenter were able to make direct eye contact with each
other. Both the child and the experimenter were seated directly in
front of the video tunnels.
Procedure
The children were asked the questions individually in a quiet lo-
cation separate from their classroom. The ordering of question type,
question difficulty, and the individual questions was fully counter-
balanced across participants. Questioning took approximately 35 min
for each child. The children interviewed FTF were seated directly
facing the questioner across a table (at a distance of approximately
5 ft) and were, therefore, able to see the face and upper body of the
questioner. The children interviewed via an LVL were also seated
directly facing the questioner (again at a distance of approximately
5 ft), but in this instance, they saw one another on a TV monitor.
The image also showed a head and upper body shot, with the face
life-sized. The amount of the questioner the child could see was,
therefore, roughly equivalent across conditions. In both conditions,
the questioner maintained her gaze on the child’s face throughout
each question episode—from when she began speaking the ques-
tion until the child finished responding.
To enable quantification of the children’s direction of eye gaze
during each stage of the question–answer interaction (listening,
thinking, and speaking), a front-on view of each child’s head and
shoulders was video recorded throughout testing, using a digital
camcorder. Although all the children were fully aware that they
were being filmed, they were not aware that their gaze behavior was
of specific interest.
The dependent measures were (1) response accuracy for each
type of question and (2) the mean percentage of time that gaze was
averted during the thinking stage of the interaction for each type of
question. The amount of time gaze was averted was measured from
when the experimenter stopped speaking the question until the child
started the response and included pauses, hesitations, and requests
for the question to be repeated. This time was converted to per-
centage of time spent averting gaze by dividing the gaze avert time
by the total thinking time and multiplying by 100. The children did
avert their gaze while speaking their responses, but this was the case
for less than half the amount of time, as compared with thinking
(mean thinking gaze aversion ⫽61.2%, mean speaking gaze aver-
sion ⫽29.0%). Thinking time was of particular interest in this
study, since our earlier work had shown this to be the period of high-
est cognitive load and that in which gaze aversion is most frequent
(Doherty-Sneddon et al., 2002), and was the period chosen for
analysis.
Interjudge reliability as to whether gaze aversion had occurred
was calculated for a random sample of 11.11% of the participants.
This calculation included all of the listening, thinking, and speaking
aversion scores for every question asked. In total, 1,296 episodes
were coded by two judges, for which there was 87.04% interjudge
agreement. Furthermore, the coders’ scoring for the duration of gaze
aversion correlated significantly [r(1,295) ⫽.62, p⬍.01].
RESULTS
The effects of question difficulty and interview status
on both response accuracy and the percentage of time
that gaze was averted during the thinking stage of the
interaction (the period of time from when the questioner
stopped speaking until the child began to speak) were an-
alyzed as described in the following sections.
Accuracy of Response
We made no predictions about accuracy of responses
to questions across conditions. Analyses of accuracy
were done to show that the three levels of difficulty were
set appropriately. For each type of question, a 4 (question
type: autobiographical memory, arithmetic, verbal, or
episodic memory) ⫻3 (question difficulty: easy, medium,
or hard) ⫻2 (interview status: FTF or LVL) mixed design
analysis of variance (ANOVA) was employed, with ques-
tion type and question difficulty the within-groups vari-
ables and interview status the between-groups variable.
Means for each condition are displayed in Table 1.
There was a significant effect of question type
[F(3,99) ⫽7.41, p⬍.0001]. Paired ttests showed that
there was greater accuracy for the episodic memory ques-
tions and the arithmetic questions than for both the verbal
questions [t(34) ⫽3.63, p⬍.001, and t(35) ⫽4.07, p⬍
.001, respectively] and the autobiographical questions
[t(34) ⫽3.38, p⬍.01, and t(35) ⫽2.69, p⬍.01, respec-
tively; means: episodic ⫽69.84%, arithmetic ⫽67.70%,
autobiographical ⫽60.77%, and verbal ⫽60.15%]. There
was a significant effect of question difficulty [F(2,66) ⫽
401.33, p⬍.0001]. Paired ttests showed that there was
greater accuracy for easy questions than for both medium
questions [t(35) ⫽12.42, p⬍.0001] and hard questions
[t(35) ⫽29.02, p⬍.0001] and for medium questions
than for hard questions [t(35) ⫽16.84, p⬍.0001; means:
easy ⫽88.15%, medium ⫽64.54%, and hard ⫽41.20%].
There was no effect of interview status. There was a sig-
nificant interaction between question type and question
difficulty [F(6,198) ⫽6.53, p⬍.0001]. Simple effects
analyses showed that when questions were easy, there
was greater accuracy for arithmetic questions than for all
other types of questions (autobiographical memory ques-
tions, p⬍.0001; verbal questions, p⬍.05; episodic mem-
ory questions, p⬍.01]. When questions were medium,
there was greater accuracy for arithmetic questions than
for autobiographical memory questions ( p⬍.05) and
verbal questions ( p⬍.01 only). When questions were
hard, there was greater accuracy for episodic memory
questions than for all other types of questions (autobio-
CHILDREN’S GAZE AVERSION 731
graphical memory questions, p⬍.0001; arithmetic ques-
tions, p⬍.0001; verbal questions, p⬍.0001). No fur-
ther interactions were found.
Gaze Aversion During Thinking Stage
A 4 (question type: autobiographical memory, arith-
metic, verbal, or episodic memory) ⫻3 (question diffi-
culty: easy, medium, or hard) ⫻2 (interview status: FTF
or LVL) mixed design ANOVA was employed, with ques-
tion type, question difficulty, and episode the within-
groups variables and interview status the between-groups
variable. Means for each condition are displayed in Table 2.
There was a significant effect of question type
[F(3,99) ⫽30.09, p⬍.0001]. Paired ttests showed that
the children used significantly more aversion for autobi-
ographical memory questions than for all other types of
questions [arithmetic [t(35) ⫽4.48, p⬍.0001; verbal
[t(35) ⫽9.64, p⬍.0001; episodic memory [t(34) ⫽6.84,
p⬍.0001]. The children also used significantly more
aversion for arithmetic questions than for verbal ques-
tions [t(35) ⫽5.40, p⬍.0001] and episodic memory
questions [t(34) ⫽2.67, p⬍.01] and for episodic mem-
ory questions than for verbal questions [t(34) ⫽2.00,
p⫽.054; means: autobiographical memory ⫽71.76%,
arithmetic ⫽62.77%, episodic memory ⫽57.86%, and
verbal ⫽53.60%]. There was a significant effect of ques-
tion difficulty [F(2,66) ⫽156.04, p⬍.0001]. Paired
ttests showed that children used significantly more aver-
sion for both hard questions [t(35) ⫽14.31, p⬍.0001]
and medium questions [t(35) ⫽14.38, p⬍.0001] than
for easy questions and for hard questions than for medium
questions [t(35) ⫽4.35, p⬍.0001; means: easy ⫽46.10%,
medium ⫽67.03%, hard ⫽72.26%]. There was a signif-
icant effect of interview status [F(1,33) ⫽4.58, p⬍.05],
with the children using significantly more aversion when
interviewed FTF (means: FTF ⫽65.86%, LVL ⫽56.59%).
There was a significant interaction between question
type and question difficulty [F(6,198) ⫽11.17, p⬍
.0001]. Simple effects analyses showed that although
there was significantly more aversion for both hard ques-
tions and medium questions than for easy questions for
each type of question (i.e., autobiographical memory,
arithmetic, verbal, and episodic memory), it was only for
the verbal questions and episodic memory questions that
there was also significantly more aversion for hard ques-
tions than for medium questions (all ps⬍.0001). No fur-
ther interactions were found. Importantly, the interaction
between question difficulty and interview status was not
significant [F(2,66) ⫽0.082, p⫽.92; mean FTF, easy ⫽
50.3, medium ⫽70.6, hard ⫽76.7; mean LVL, easy ⫽
40.7, medium ⫽62.1, hard ⫽67.0]. In the FTF condi-
tion, the 95% confidence interval (CI) for the easy–hard
difference was 26.37 ⫾5.13, and in the LVL condition,
it was 25.94 ⫾5.09. These differences are large relative
to their CIs, and it follows that the only interactions that
could plausibly be present are substantially smaller than
the differences between the easy and the hard tasks.
DISCUSSION
For all types of questions, the children averted their
gaze less when interviewed across an LVL, as compared
with FTF, even though the actual physical distance be-
tween the experimenter and the child was held constant
across interview conditions and the amount of the ques-
tioner visible was equivalent. Furthermore, the children
Tab l e 1
Mean Percentages for Accuracy of Responses to Different Question Types for the Live Video Link (LVL) Group
and the Face-to-Face (FTF) Group (With Standard Deviations)
Question Difficulty
LVL Group FTF Group
Easy Medium Hard Easy Medium Hard
Question Type MSD MSD MSD MSD MSD MSD
Autobiographical memory 78.96 15.48 61.25 16.64 38.96 11.66 87.89 11.86 59.56 17.94 36.05 16.20
Arithmetic 97.40 15.02 70.17 17.70 40.10 13.68 91.15 17.69 70.14 17.37 36.36 22.49
Verbal 89.32 20.69 65.68 30.97 39.84 22.67 87.94 19.71 53.29 18.30 26.10 14.82
Episodic memory 82.29 22.13 68.40 16.57 55.21 16.56 88.16 13.12 68.71 15.85 55.70 14.38
Tab l e 2
Mean Percentages of Time Spent in Gaze Aversion While Thinking by the Live Video Link (LVL) Group
and the Face-to-Face (FTF) Group (With Standard Deviations)
Question Difficulty
LVL Group FTF Group
Easy Medium Hard Easy Medium Hard
Question Type MSD MSD MSD MSD MSD MSD
Autobiographical memory 53.69 17.09 71.11 14.87 72.06 16.36 68.25 15.55 81.21 11.18 80.45 11.70
Arithmetic 35.04 17.09 68.47 19.39 69.56 23.73 48.09 24.80 72.97 17.24 81.67 10.01
Verbal 40.98 14.97 50.87 15.59 57.86 13.25 43.42 18.08 60.79 18.89 68.03 15.64
Episodic memory 33.17 23.02 57.83 17.88 68.47 12.05 41.46 21.22 67.46 14.41 76.55 11.78
732 DOHERTY-SNEDDON AND PHELPS
averted their gaze most when being asked personally rel-
evant questions—the autobiographical memory ques-
tions. This could in part be because the autobiographical
questions were more difficult (shown by the decreased
accuracy of response, as compared with episodic and
arithmetic questions). However, verbal and autobiograph-
ical questions had equivalent accuracy levels, and still
the autobiographical questions elicited more gaze aver-
sion than did the verbal. Clearly, there is a significant so-
cial component driving interpersonal gaze aversion dur-
ing questioning. Part of the reason we look away in order
to concentrate could be because we experience a nega-
tive social emotion, such as self-consciousness or em-
barrassment (Stanley & Martin, 1968). An alternative
social explanation is that we use gaze aversion to delib-
erately signal to the other person that we are in the pro-
cess of thinking. Some recent doctoral work suggests
that children can deliberately use gaze aversion as a sig-
nal of thinking. For example, children and adults are
more likely to look up while thinking if their questioner
is facing them but to look down if the questioner is turned
away from them (McCarthy, 2004). In a similar vein,
Kendon (1967) reported that adults use looking away as a
deliberate turn-holding cue—in other words, as a way of
saying “I’m about to say something”/“I’m still thinking
about what I’m going to say.”
The present results make sense in terms of our own
earlier work. We have shown that the reduction in social
co-presence produced by video-mediated interviews re-
sults in children being more confident and less intimi-
dated.This had an important impact on many aspects of
their verbal and nonverbal behavior; for example, they
were more relaxed and happy and better able to disagree
with an adult questioner when asked misleading ques-
tions (Doherty-Sneddon & McAuley, 2000). The data are
also consistent with our research with adults who look at
one another around 56% more when communicating via
LVLs, as compared with FTF (Doherty-Sneddon et al.,
1997). So children and adults reduce the baseline amount
that they avert their gaze from the face of an interlocutor
when they communicate across LVLs. Our next question
was whether cognitive difficulty of questions would in-
fluence gaze aversion in both mediated and FTF condi-
tions equally.
Earlier work (Doherty-Sneddon et al., 2002; Phelps
et al., in press) has shown that gaze aversion is more fre-
quent when children are asked moderately difficult ques-
tions than when asked easy ones. If this is due primarily
to the self-consciousness associated with dealing with
cognitively difficult material, we would expect to see a
greater effect of task difficulty in FTF interaction and
little effect in video-mediated questioning. If, however,
children increase their aversion to avoid environmental
information (in this case, the face of the questioner), we
would expect an effect of question difficulty in both in-
terview conditions. Our results support the latter possi-
bility. Question difficulty had an enormous impact on
the amount that the children averted their gaze from the
experimenter’s face in both interview contexts. Further-
more, this was consistently the case across all question
types. The children averted their gaze, on average, 52%
more when answering hard questions, as compared with
their easy counterparts, in FTF interviews. Similarly,
they increased their gaze aversion by 64% for hard ques-
tions in the LVL condition. Even when social factors
were greatly reduced (LVLs), the children looked away
from their interviewer more when they were trying to
process increasingly difficult information. This gives
considerable weight to the cognitive load hypothesis.
The processing costs of visual communication signals
(such as facial expressions, eye gaze, and lip movements)
are documented. For example, if one is forced to look
constantly at a listener while speaking, one’s speech will
become less fluent and contain more “erms” and “uhms”
(Beattie, 1981). We have found that asking children to
maintain gaze with a speaker’s face while listening to de-
scriptions of abstract shapes interferes with their abili-
ties to understand these descriptions. In addition, we
found that children are less able to retain visuospatial in-
formation when they have to monitor a face during re-
tention than if they are allowed to look away (Doherty-
Sneddon, Bonner, & Bruce, 2001). So gaze aversion is
an overt strategy for shifting one’s attention from envi-
ronmental stimulation (such as faces) to processing of
other internal information. The present study shows that
this is true for both live and video-mediated faces.
This work has clear educational implications. It ap-
pears that children use gaze aversion to control their own
cognitive load. There are social factors at work, and
teachers and others working with children would be well
advised to allow children to set a comfortable intimacy
distance and avoid intimidating questioning. Another
practical application of this work is with child witnesses.
Earlier work has shown that children are more confident
and less suggestible to misleading questions when cross-
examined across LVLs (e.g., Davies, 1999). The fact that
the children in the present study averted their gaze from
the questioner less during video link interactions sug-
gests that they too were less intimidated. It is yet to be es-
tablished whether the gazing behavior of child witnesses,
either in open court or across LVLs, influences how cred-
ible jurors perceive them to be, an extremely important
issue for the court process (Goodman et al., 1998).
Cultural rules about gaze behavior also help deter-
mine our response to gaze and gaze aversion. For exam-
ple, in Britain, children are typically told to “Look at me
when I’m speaking to you,” whereas in America, chil-
dren are encouraged “Not to stare.” In addition, Black
Americans interpret gaze aversion as a sign of respect,
rather than as a sign of disinterest (Hanna, 1984). Cul-
tures differ significantly in terms of how much gaze be-
tween individuals is acceptable or desirable. For exam-
ple, with contact cultures (such as Latin America and
Southern Europe), people engage in more gaze, overall,
than do people coming from noncontact cultures, such
as Northern Europe (Halberstadt, 1985; Watson, 1970).
CHILDREN’S GAZE AVERSION 733
However, social influences aside, we must also be
aware of the important cognitive function that gaze aver-
sion plays in any conversation, especially when the ma-
terial being discussed is demanding. Gaze aversion is
often a sign of concentration and inner effort. Teachers
and other professionals must be proficient in reading
children’s gaze aversion in conjunction with other cues, in
order to determine, for example, when children are en-
gaged in learning, are just about to understand, or per-
haps are switching off from a task (Doherty-Sneddon
et al., 2002). Indeed, in our ongoing research, we are in-
vestigating whether teachers do, in fact, use information
from children’s gaze (Doherty-Sneddon & Phelps, 2005).
REFERENCES
Abel, M. (1990). Experiences in an exploratory distributed organiza-
tion. In J. Gallagher, R. E. Kraut, & C. Egido (Eds.), Intellectual
teamwork: Social and technological foundations of cooperative work
(pp. 489-510). Hillsdale, NJ: Erlbaum.
Abramovitch, R., & Daly, E. M. (1978). Children’s use of head ori-
entation and eye contact in making attributions of affiliation. Child
Development, 49, 519-522.
Anderson, A. H., O’Malley, C., Doherty-Sneddon, G., Langton, S.,
Newlands, A., Mullin, J., Fleming, A. M., & Van der Velden, J.
(1997). The impact of VMC on collaborative problem solving: An
analysis of task performance, communicative process, and user sat-
isfaction. In K. E. Finn, A. J. Sellen, & S. B. Wilbur (Eds.), Video-
mediated communication (pp. 133-156). Mahwah, NJ: Erlbaum.
Argyle, M. (1996). Bodily communication. London: Routledge.
Argyle, M., & Dean, J. (1965). Eye-contact, distance and affiliation.
Sociometry, 6, 289-304.
Beattie, G. W. (1981). A further investigation of the cognitive inter-
ference hypothesis of gaze patterns during conversation. British Jour-
nal of Social Psychology, 20, 243-248.
Bruce, V. (1995). The role of the face in face-to-face video communi-
cation. In S. J. Emmott (Ed.), Information superhighways: Multime-
dia users and futures (pp. 227-238). London: Academic Press.
Burgoon, J. K., Manusov, V., Mineo, P., & Hale, J. L. (1985). Effects
of gaze on hiring, credibility, attraction and relational message inter-
pretation. Journal of Nonverbal Behavior, 9, 133-146.
Clark, H. H., & Brennan, S. E. (1991). Grounding in communication.
In L. B. Resnick, J. Levine, & S. D. Teasley (Eds.), Perspectives on
socially shared cognition (pp. 127-149). Washington, DC: American
Psychological Association.
Davies, G. (1999). The impact of television on the presentation and re-
ception of children’s testimony. International Journal of Law & Psy-
chiatry, 22, 241-256.
Davies, G., & Noon, E. (1991). An evaluation of the live link for child
witnesses [Home Office report]. London.
Doherty-Sneddon, G., Anderson, A. H., O’Malley, C., Langton, S.,
Garrod, S., & Bruce, V. (1997). Face-to-face and video mediated
communication: A comparison of dialogue structure and task per-
formance. Journal of Experimental Psychology: Applied, 3, 105-125.
Doherty-Sneddon, G., Bonner, L., & Bruce, V. (2001). Cognitive
demands of face monitoring: Evidence for visuospatial overload.
Memory & Cognition, 29, 909-919.
Doherty-Sneddon, G., Bruce, V., Bonner, L., Longbotham, S., &
Doyle, C. (2002). Development of gaze aversion as disengagement
from visual information. Developmental Psychology, 38, 438-445.
Doherty-Sneddon, G., & Kent, G
.
(1996). Visual signals and the
communication abilities of children. Journal of Child Psychology &
Psychiatry, 37, 949-959.
Doherty-Sneddon, G., & McAuley, S. (2000). Influence of video
mediation on adult–child interviews: Implications for the use of the
live link with child witnesses. Applied Cognitive Psychology, 14,
379-392.
Doherty-Sneddon, G., & Phelps, F. (2005). Teachers’ responses to
children’s gaze aversion during classroom interactions. Manuscript
in preparation.
Ellyson, S. L., Dovidio, J. F., & Corson, R. L. (1981). Visual behav-
ior differences in females as a function of self-perceived expertise.
Journal of Nonverbal Behavior, 5, 164-171.
Exline, R., Gray, D., & Schuette, D. (1965). Visual behavior in a
dyad as affected by interview content and sex of respondent. Journal
of Personality & Social Psychology, 1, 201-209.
Glenberg, A. M. (1997). What memory is for. Behavioral & Brain Sci-
ences, 20, 1-19.
Glenberg, A. M., Schroeder, J. L., & Robertson, D. A. (1998).
Averting the gaze disengages the environment and facilitates re-
membering. Memory & Cognition, 26, 651-658.
Goldin-Meadow, S., Wein, D., & Chang, C. (1992). Assessing knowl-
edge through gesture: Using children’s hands to read their minds.
Cognition & Instruction, 9, 201-219.
Goodman, G. S., Tobey, A. E., Batterman-Faunce, J. M., Orcutt, H.,
Thomas, S., Shapiro, C., & Sachsenmaier, T. (1998). Face-to-face
confrontation: Effects of closed-circuit technology on children’s eye-
witness testimony and jurors’ decisions. Law & Human Behavior,
22, 165-203.
Halberstadt, A. G. (1985). Differences between blacks and whites in
nonverbal communication. In A. W. Siegman & S. Feldstein (Eds.),
Multichannel integrations of nonverbal behavior (pp. 229-259).
Hillsdale, NJ: Erlbaum.
Hanna, J. L. (1984). Black/white non-verbal differences, dance and
dissonance. In A. Wolfgang (Ed.), Nonverbal behavior: Perspectives,
applications, intercultural insights (pp. 373-409). Lewiston, NY:
Hogrefe.
Kendon, A. (1967). Some functions of gaze direction in social inter-
action. Acta Psychologica, 26, 22-63.
McCarthy, A. (2004, January). Are the eyes the windows to the soul?
Talk given at University of Kyoto, Japan.
McNeill, D. (1985). So you think gestures are nonverbal? Psycholog-
ical Review, 92, 350-371.
Phelps, F., Doherty-Sneddon, G., & Warnock, H. (in press). Func-
tional benefits of children’s gaze aversion during questioning. British
Journal of Developmental Psychology.
Russell C., & Lavie, N. (2001). Changing faces: A detection advan-
tage in the flicker paradigm. Psychological Science, 12, 94-99.
Smith, R., O’Shea, T., O’Malley, C., Scanlon, E., & Taylor, J.
(1991). Preliminary experiments with a distributed, multi-media,
problem-solving environment. In J. Bowers & S. Benford (Eds.),
Studies in computer-supported co-operative work: Theory, practice
and design (pp. 31-48). Amsterdam: Elsevier.
Stanley, G., & Martin, D. S. (1968). Eye-contact and the recall of
material involving competitive and noncompetitive associations. Psy-
chonomic Science, 13, 337-338.
Vrij, A. (2002). Deception in children: A literature review and implica-
tions for children’s testimony. In H. L. Westcott, G. M. Davies, &
R. H. C. Bull (Eds.), Children’s testimony (pp. 175-194). London:
Wiley.
Watson, O. M. (1970). Proxemic behaviour: A cross-cultural study.
The Hague: Mouton.
Zuckerman, M., & Driver, R. E. (1985). Telling lies: Verbal and non-
verbal correlates of deception. In A. W. Siegman & S. Feldstein
(Eds.), Multichannel integrations of nonverbal behavior (pp. 129-
147). Hillsdale, NJ: Erlbaum.
(Manuscript received March 5, 2004;
revision accepted for publication July 22, 2004.)
