Long-Distance Transmission of Facial Affect

Signals
Joseph C. Hager and Paul Ekman
University of Culifornia, San Francisco

This study examined the distance at which certain facial source a signal remains effective. Little is
expressions can transmit affect messages. A man and a known about how far facial expressions transmit
woman assumedfacial expressions that were selected information, although Hall (1966) claimed that
carefully to represent six affects. These expressions were at 30 ft (9.15 m) and beyond the details of facial
shown in still photographs and in live portrayals to 49 expression disappear. Our study was designed
observers who composed four groups which were 30,
to show that facial expressions can send affect
35, 40, and 45 meters away from the stimuli. Photo-
messages much farther, well beyond the dis-
graphs and live portrayals produced comparable re-
sults. Every observer was able to label the expressions tances at which intimate, face-to-face interac-
accurately although accuracy declined as distance in- tions occur.
creased. Extrapolation from the data suggested that Studying the transmission of facial signals is
some messages may be sent far beyond the distances complicated by the many variables which can
used in this study. These results raise important issues influence the sending and receiving of visual
about the transmission of facial signals over distance information in the face: expressor variables (fa-
and suggest that the face is a long-distance transmitter cial size and physiognomy, age, sex, degree and
of affect signals. location of muscle contractions, etc.), observer
variables (visual acuity, skill, and knowledge,
Key Words: Facial expression.
etc.), and environmental variables (ambient
light, lighting of the face, stimulus background,
orientation of the observer and expressor, etc.).
INTRODUCTION Our experimental procedures controlled or
measured the most significant variables affecting
Research has shown that some facial expres-
the distance range of facial signals. Some of the
sions function as signals of affect and that these
conditions in which communication via facial
signals can have definite, universal meanings
expression naturally occurs were approximated
(for reviews see Ekman et al., 1972: Ekman,
by having unpracticed observers judge the live
1973). Facial expressions share certain impor-
expressions of models. Artificiality was intro-
tant properties with other communicative sig-
duced by using a prescribed list of verbal re-
nals (e.g., Altmann. 1967: Marler, 1968). One of
sponses and a restricted number of posed
these properties is the “strength of transmis-
expressions which were not seen in their usual
sion” (Moynihan, 1970) or how far from its
context. Photographs as well as live portrayals
were used as stimuli to determine whether the
judgments of each would be similar enough to
Received June I?. IY7X
allow the use of photographs alone in future
Addre\\ repnnl requea to: Paul Ekman. Human Inrer-
action L;thoratory. Department of Psychiatry. UniverGty of research. Four groups of observers, each at a
California. San Francisco. CA Y4143. different distance, judged these stimuli.

Ethology and Sociobiology I: 77-82 (lY7Y) 77

@ Elsevirr North Holland. Inc.. lY7Y 0162-309Y79/010077-06/502.25
78 J. C. Hager and P. Ekman

METHOD been found in previous studies to yield high

agreement among observers who judged affect
(Ekman and Friesen, 1976b). Next, these pic-
Selection of Facial Expressions tures were scored with the Facial Action Coding
System (Ekman and Friesen, 1976a. 1978) to
A stepwise procedure was devised to select identify the muscle movements which expres-
expressions which would obviously, rather than sors would have to perform to reproduce each
subtly, signal six affects: happiness, sadness, expression. Movements of the body or head and
fear, anger, surprise, and disgust. These affects changes in gaze direction which typically ac-
were chosen because previous research had company affect were ignored. The most frequent
shown that they have the most reliable associ- score in each sample of affect pictures was cho-
ations with facial expressions and that these as- sen to comprise a set of instructions to expres-
sociations are universal (Ekman et al., 1972). sors for performing the six expressions. Because
Only one expression was selected for each affect stronger muscle contractions produce clearer.
to avoid too many judgments for one session more visible expressions, maximum contraction
and to make the expressors’ task easier. In order of the muscles producing the expressions was
to explore the maximum distances over which specified in these instructions. Finally, the
the face can transmit information, muscle move- mouth opening was made equal across all
ments were identified which would create the expressions so that the configuration of the
most visible expressions with the least ambigu- mouth was the signal rather than whether it was
ous meanings. open or closed. These steps produced stimulus
The first step to identify muscle movements expressions which looked extreme but natural,
was to select pictures for each affect which had as exemplified in Fig. 1.

Figure 1. Two of the stimulus photographs (not to scale) used in this study: (a) the woman’s happy
expression, (b) the man’s fear expression.
Transmission of Facial Affect Signals 79

There were 4 trials of each of the 6 expressions

Expressors, Photographs, and Observers by each expressor for a total of 48 live perform-
Two expressors were chosen who differed sub- ances. These combinations were duplicated by
stantially in many respects including sex. head the display of photographs. All trials were mixed
size, physiognomic features, and age (46 years randomly with respect to expressor. expression,
for the male. 34 for the female). They could and presentation mode. except for minor con-
accurately and reliably perform the required straints (e.g., not more than two consecutive
muscle movements because they had previously presentations of the same expression).
received extensive training in voluntary con-
traction of their facial muscles. Their portrayals
of the selected expressions were photographed. RESULTS
printed on black and white matte paper. and
enlarged with equal print densities io the life The observers’ visual acuity, as measured by
size of each expressor. the lowest line of the Snellen chart completed
Observers were 49 adult volunteers (30 fe- correctly, ranged from 20/40 to 2O/lO with a
male) who were assigned randomly to groups mean of N/18. The groups of observers who
and paid $2.50 each. There were 13. I I. I?, and were seated at the different viewing distances
13 observers nearest to farthest from the stimuli. did not differ in their mean acuity, f(3. 45) =
respectively. 0.59, p = 0.62.
Even at the farthest distance every observer
Procedure judged the stimuli significantly better than
The experiment was conducted in one session chance (p < 0.01, exact binomial, with the ex-
in the largest indoor area available, a campus pectation that d of the 96 trials would be correct
parking structure. Observers’ visual acuity was by chance alone). This measure does not reveal
measured individually. aided by their usual cor- which of the six expressions observers could
rection (if any). using a Snellen wall chart. Each identify at each distance, but this issue is ad-
group of observers saw the stimuli from only dressed below. To estimate the farthest distance
one distance and was seated either 30. 35. 40, each expression could be identified. a straight
or 45 m away from the stimuli, the maximum line was fitted to observers’ scores as a function
separation permitted by the building’s dimen- of the four experimental distances. Figure 2
sions. A pilot study had shown that accurate shows the observed mean scores at each dis-
labeling of most expressions would decline rap- tance for each expression and expressor com-
idly over this distance range. The expressors bination. A line was fitted for each of these
and photographs were in a portable studio that combinations Bnd was used to predict the dis-
recreated the same lighting and background con- tances at which accuracy would decline to
ditions in which the photographs had been chance levels. Unfortunately, these levels can-
taken. A panel hid the expressor’s body so that not be known exactly because the expectations
only the head could be seen. Looking through of chance success for any one expression de-
the open end of the studio, the observers colild pends upon how well the other expressions
clearly see the faces and photographs without could be identified and eliminated as choices. In
glaring reflections. They were told the purpose this experiment the nonnegative affect expres-
of the experiment, that they might see the same sions. happiness and surprise, were identified so
expression or the same photograph more than well that they probably were eliminated as
once, and that they should judge each stimulus choices when the other expressions were pre-
without trying to remember previous judgments. sented. Therefore, the expectation of chance
On each trial observers had 2 set to focus on success was set conservatively at one-fourth of
the stimulus location and accomodate to the the trials for the four negative affect expressions
lighting before seeing the stimulus expression. and at one-sixth for happiness and surprise. The
During a trial, either an experimenter displayed predicted distances where accuracy would de-
a photograph or an expressor performed an cline to these chance levels were checked by
expression once. In both cases the stimulus comparing them to the accuracies observed at
could be seen for 3 sec. Observers had up to IO the experimental distances to insure that the
set to circle one label from a list of six affects. predictions were not overestimates. Goodness
80 J. C. Hager and P. Ekman

I I
dm 3Am 40m 45m

Distance DiStOllC.
a. b.

Figure 2. Graphs of the mean number of correct responses at each distance for each expressor and
expression combination: (a) woman, (b) man: Ha = happiness, Su = surprise, Fe = fear, Di = disgust, Sa
= sadness. An = anger. Eight correct = IO@%.

of fit tests indicated that lines of a higher degree error terms are given in Table 1. Duncan’s Mul-
would not add to predictive accuracy except for tiple Range Tests on the means in Table 2
the woman’s expression of anger. This analysis showed that accuracy decreased significantly
indicated that the limit of accurate transmission with increasing distance except between 40 m
of the woman’s anger had been reached at 35 m; and 45 m, where there was no significant differ-
expressions of her sadness, fear, and disgust at ence. Simple main effects and simple interaction
40 m; expression of the man’s disgust at 45 m; effects of two- and three-way interactions in-
and expressions of his sadness and fear at 54 m. volving the presentation mode, expressor, and
Predictions for the limit of accurate transmission expression variables helped to clarify the effects
of happiness, surprise, and the man’s anger were of these variables. In general, judgments about
from 100 to 220 m. too far beyond the experi- the man were more accurate than judgments
mental distances to maintain confidence in their about the woman except that accuracy was the
reliability. All these predictions should be inter-
preted as rough estimates only for the particular
Table 1. Summary of Omnibus Analysis of Variance
expressions, expressors, and conditions of this
experiment. They suggest that equally extreme Source MS df F
expressions signalling different affects transmit A (distance) 38.01 3 5.02"
to different distances, and that while most of S/A (error) 7.58 45
these affect signals transmit only as far as the C (expressor) 128.22 177.16c
range of distances uied in this study, some can AXC 2.42 : 3.35"
C X SIA 0.72 45
transmit much farther.’ D (expression) 151.41 5 92.38'
A mixed design, 4(distance) x Z(presentation D X S/A 1.64 225
mode) x Z(expressor) x 6(expression) analysis B (presentation mode) 6.86 5 7.11c
of variance with repeated measures oil the last XD
B X D X S/A 0.96 225
three factors was conducted with the number of CXD 39.02 5 40.23c
correct responses as the dependent variable. C X D X S/A 0.97 225
The significant sources of variance and their BXCXD 3.58 5 5.82'
BXCXDXS/A 0.62 225

ap < 0.05
’ Preliminary findings from subsequent research using
only the mate’s pictures indicate that expert observers
‘p < 0.01

can accurately label expressions of happiness and surprise at cp < 0.001

100 m.
Transmission of Facial Affect Signals 81

Table 2. Mean Number of Correct Responses Judgments of pictures and live expressions
to All the Man’s and Woman’s Stimuli at appear to be similar enough to justify using only
Each Distance pictures in the future when experimental con-
Distance ditions are comparable to those in this study.
Expressor 30 m 35 m 40 m 45 m The absence of interactions between the dis-
Man” 36.69 34.36 30.08 28.92 tance and the presentation mode variables indi-
Woman* 31.38 25.54 19.67 21.69 cates that the effects of distance are similar for
a Error MS = 3.%
both types of stimuli. Also, there was no main
effect of presentation mode despite the obvious
b Error MS = 4.33
differences between pictures and live portrayals.
Note. Maximum score = 48.
However, as experimental conditions diverge
from those of this study (e.g., a greater distance
same for the happy expression of each expres- range or more ambiguous, quicker, and less vis-
sor, and the woman’s live disgust expression ible expressions), the comparability of pictures
was better judged than the man’s. Usually, there and live portrayals becomes more uncertain.
were no significant differences between judg- Our demonstration of the face’s capacity to
ments of live expressions and pictures, but send affect signals over long distances is an in-
sometimes either the live or photographic stim- itial exploration of relationships between dis-
uli were judged significantly better, depending tance and the transmission of facial information.
unpredictably on expressor and expression com- Such parameters of facial expression as the mus-
binations. Expressions were not judged with cles contracted, the number recruited, the de-
equal accuracy, and their rank order varied gree of contraction, and the areas of the face
slightly depending on the expressor and pres- involved might have systematic effects on the
entation mode. Multiple Range Tests showed visibility of facial messages. Future studies
that happiness and surprise were judged invari- could also examine the’ effects produced by
ably better than the other expressions except for characteristics of the model (e.g., sex, age, head
the man’s anger which was judged*equally well. size, skin color, facial hair, physiognomic qual-
ities, use of cosmetics), different environmental
circumstances, and other types of observers
(e.g., expert). Comparing the ranges of different
DISCUSSION
affect signals, other facial signals such as iden-
These results show that the face is a long-dis- tity, age, sex, and ethnicity (Ekman, 1978). and
tance transmitter which can signal affect much signals in vocal and other visual channels would
farther than the boundaries of intimate interac- show the relative strength of transmission of
tion. Observers can receive some affect mes- these signals. Research on these issues could
sages 45 m away from the face. Extrapolations help us to understand the evolutionary signifi-
based on the data suggest that some expressions cance of communicative movements and certain
can probably be identified at more than 100 m. physical features (e.g., the sclera and facial
To appreciate how far these distances are in hair), the patterns of control on muscle move-
relation to contexts for social interaction, con- ments in deception, and the use of certain mus-
sider that no spectator watching a Shakespear- cles in deliberate communication (e.g., the brow
ean play in the Globe theater could have been raise in greetings). Theories of the function, de-
more than 27 m from any player. The farthest velopment, and evolution of the face and its
a member of the House of Representatives can signals should consider the role of the face in
sit from the speaker’s rostrum in the House distant communication (e.g., between unfamiliar
Chamber is 25 m. As another index to the power groups) as well as in intimate interactions (e.g.,
of facial signals, consider the world’s record between mother and infant).
javelin throw, 94.58 m, which is approximately
how far professional baseball players can throw.
If this distance represents an upper limit to the The authors thank Michael Lewis for raising a ques-
range of hand-propelled weapons, then some tion about distance which instigated our research,
affect messages probably can be recognized be- Harriet Oster and John Johnson for their indispensible
yond primitive combat ranges. service as models, and Mary McCarthy for assistance
a2 J. C. Hager and P. Ekman

during the study. This research was supported in part Ekman. P., Friesen. W.V. Pic,clrres of FucicJ/
by a grant from the Harry F. Guggenheim Foundation. Afjecr. Palo Alto, CA: Consulting Psycholo-
gists’ Press, 1976b.
Ekman, P., Friesen, W.V. The Fuciul Acriotl
Coding S~srcm. Palo Alto, CA: Consulting
REFERENCES Psychologists’ Press, 1978.
Ekman. P.. Friesen. W.V.. Ellsworth.
Altmann, S.A. The structure of primate social P. Emotion i/l thcl Humat, Fac.1~: Glliclrlitlt,.s
communication. In S~cicrl Co/?f,?lJt/lict~lio/f ,for Rrscwrch i~tit/ LJU Itltc~grcJIiorl of‘Fitditlgs.
u,rlo,lg Primatc>s, S.A. Altmann (Ed.). Chi- New York: Pergamon. 1972.
cago: University of Chicago Press. 1967, pp. Hall. E.T. The Hicldol L)i,~lr/~si~,r. Garden
375-362. City. NY: Doubleday, 1966.
Ekman. P. &Jrufll cltld F~JcitJl t$rrssiorJ. New Marler. P. Visual systems. In A/rimtr/ Commrr-
York: Academic, 1973. nictr tiorl: Trch/r;c/rtr,s oJ’ Strtt/,v cJllt/ Rrs/tlt.s qf‘
Ekman. P. Facial signs: Facts. fantasies. and Rrsrcrri./f. T.A. Sebeok (Ed.). Bloomington.
possibilities. In SighI. SOJ/~/ a/it/ Sruw. ID: Indiana University Press. 1968. pp. lO3-
T.A. Sebeok (Ed.). Bloomington. ID: Indiana I/h.
University Press, 1978. pp. 124-156. Moynihan. M. Control, suppression, decay,
Ekman, P.. Friesen. W.V. Measuring facial disappearance and replacement of displays.
movement. ~~i~~ir~~~l~?lr~lrtJ~ P.s~cholog~~ ~rtit/ Jorlr/lrJ/ of 77lr,wl~r;cYJ/ Biology ‘9: X5- I 12
Norll~rrbal Behu,,ior I: 56-75 ( 1976a). (1970).