Journal of Feisonality and Social Psychology
1977, Vol 35, No. 4, 250-256
The Halo Effect: Evidence for Unconscious
Alteration of Judgments
Richard E. Nisbett and Timothy DeCamp Wilson
University of Michigan
Two different videotaped interviews were staged with the same individual—a
college instructor who spoke English with a European accent. In one of the
interviews the instructor was warm and friendly, in the other, cold and distant.
The subjects who saw the warm instructor rated his appearance, mannerisms,
and accent as appealing, whereas those who saw the cold instructor rated these
attributes as irritating These results indicate that global evaluations of a person
can induce altered evaluations of the person's attributes, even when there is
sufficient information to allow for independent assessments of them. Furthermore,
the subjects were unaware of this influence of global evaluations on
ratings of attributes. In fact, the subjects who saw the cold instructor actually
believed that the direction of influence was opposite to the true direction. They
reported that their dislike of the instructor had no effect on their ratings of
his attributes but that their dislike of his attributes had lowered their global
evaluations of him.
Although the halo effect is one of the oldest
and most widely known of psychological phenomena,
surprisingly little is known about its
nature. The halo effect is generally defined as
the influence of a global evaluation on evaluations
of individual attributes of a person, but
this definition is imprecise with respect to the
strength and character of the influence. At
one extreme, the halo effect might be due
simply to an extrapolation from a general
impression to unknown attributes. Global
evaluations might color presumptions about
specific traits or influence interpretation of
the meaning or affective value of ambiguous
trait information. Thus, if we like a person, we
often assume that those attributes of the person
about which we know little are also favorable.
(Politicians often seem to capitalize on
this tendency by appearing warm and friendly
but saying little about the issues.) Such a
phenomenon could best be described as a deThe
present research was supported by National
Science Foundation Grant BNS75-23191.
We are indebted to Eugene Borgida and Hersz
Zukier for their assistance and advice and to Lee
Ross for criticism of an early draft of the manuscript.
Requests for reprints should be sent to Richard E
Nisbett, Research Center for Group Dynamics, Institute
for Social Research, University of Michigan, Ann
Arbor, Michigan 48106.
duction from an implicit personality theory
holding that "nice people tend to have nice
attributes and less nice people have less nice
attributes." Similarly, the global evaluation
may alter the interpretation of the meaning
or evaluation of ambiguous attributes. Thus,
if one is told that a warm and friendly person
is impetuous, a quite different set of behaviors
come to mind from those that occur when one
is told that the impetuous person is angry
and hostile. The evaluation of the attribute
impetuous may differ in the two cases, but
the difference could be a mere interpretive
one arising from the ambiguity inherent in the
term impetuous.
It is possible, however, that the halo effect
is a much more powerful phenomenon. Global
evaluations may be capable of altering perceptions
of even relatively unambiguous stimuli,
about which the individual has sufficient information
to render a confident judgment. For
example, a person's appearance may be perceived
as more attractive if we like the person
than if we do not. If so, then the halo effect
involves much more than influence on presumptions
about attributes or interpretations
of the meaning of attributes, but extends to
alteration of judgments about attributes for
which we generally assume we are capable of
rendering independent assessments.
250
HALO EFFECT 251
The literature to date on the halo effect
does not provide a clear picture of either the
strength or the nature of the phenomenon. Although
correlational studies are plentiful, they
do not serve to demonstrate even that the
phenomenon exists. In any array of trait ratings,
correlations tend to be positive, that is,
affectively consistent, but it is always possible
that the correlations are based on veridical
social perception. For example, people with
more pleasant personalities may actually be
better looking on the average than people with
less pleasant personalities. Alternatively, people's
criteria for likability and good looks may
differ in such a way as to produce positive
correlations in the absence of any influence of
one judgment or the other. For example, a
college football player and a drama major
may have different criteria for both likability
and good looks in other males. The football
player may prize a hail-fellow-well-met style
and may regard brawn as essential to masculine
good looks. The drama major may favor
verbal wit and a lean, ascetic appearance. If
an open personal style and brawniness are
associated and if a witty style and lean, ascetic
looks are associated, then both the football
player and the drama major might show
positive correlations between their judgments
of likability and looks, even though they give
high scores on both dimensions to utterly different
people and even though neither judgment
has in any way influenced the other.
To demonstrate that global evaluations alter
evaluations of particular attributes requires
an experimental approach. There have
been two sorts of experimental studies in this
area. One type employs adjectives rather than
persons as the stimuli (e.g., Anderson, 1974;
Asch, 1946). This research supports the generalization
that the evaluative rating given to
a trait adjective is a function of the total
adjective set in which it is embedded. In a
sense, then, the global evaluation alters the
evaluation of a particular attribute. But this
work suffices only to show that the total adjective
set may influence the meaning or evaluation
of a particular adjective whose meaning
or value is somewhat ambiguous when considered
in isolation, as in the example above
of the interpretation of the adjective im-
Petuous.
Another type of experimental study has
employed actual persons as the stimuli (Clifford
& Walster, 1973; Dion, 1972; Dion,
Berscheid, & Walster, 1972; Miller, 1970),
but these studies suffice to show only the weak
"altered presumption" version of the halo effect
because they provide extremely impoverished
information about the attributes to be
rated. For example, Miller (1970) gave photographs
of attractive and unattractive target
persons to subjects and asked them to rate the
personalities of the target persons. Attractive
people were ascribed more favorable traits
than unattractive people, but this merely
constitutes a demonstration that people's implicit
personality theories include the assumption
that physically attractive people have
more attractive personalities. A convincing
demonstration of the strong version of the
halo effect phenomenon would require giving
subjects sufficient information on the attributes
to be rated to allow for an independent
assessment of the attributes.
One study in the above tradition went substantially
further. Landy and Sigall (1974)
showed that evaluations of an essay given by
male college students were substantially
higher when the alleged writer of the essay
was an attractive woman than when the alleged
writer was an unattractive woman. The
effect was very pronounced, especially when
the essay was of relatively poor quality. This
study supports the stronger interpretation of
the halo phenomenon because, unlike the
attributes in the above studies, the subjects
had information about the woman's essay that
was entirely sufficient to allow for an independent
assessment of its quality. The Landy
and Sigall study, however, was not precisely
a study of the halo effect because it was a
single attribute rather than a global evaluation
that was manipulated and a person's
product rather than an attribute that was
measured.
The present experiment was designed to
address an additional issue—the question of
people's awareness of the halo effect. Nisbett
and Wilson (1977) have recently argued that
people have little awareness of the nature or
even the existence of the cognitive processes
that mediate judgments, inferences, and the
production of complex social behavior. The
252 RICHARD E NISBETT AND TIMOTHY DECAMP WILSON
halo effect would appear particularly likely
to be such a subterranean, unrecognized
process. If people were capable of rendering
independent assessments of attributes, they
would surely do so. The halo effect, particularly
if it is the strong, distorting influence
postulated by some theorists, would appear
to depend upon lack of awareness for its
existence. On the other hand, it is not inconceivable
that subjects might actually be able
to report correctly on its operation if they
were directly questioned about it (e.g., Oh
yes, now that you mention it, maybe part of
the reason I liked his looks was because I
liked his personality). The present experiment
directly assessed people's reports about the
nature and direction of the influence of global
evaluations on evaluations of attributes.
Method
Overview
College student subjects were asked to evaluate a
psychology instructor, whom they viewed in a videotaped
interview, on several dimensions The subjects
were told that the investigators were studying the
possibility that ratings of an instructor presented in
such a brief fashion might resemble ratings by students
who had taken an entire course with the instructor
The subjects were shown one of two different
interviews with the same instructor, a native
French-speaking Belgian who spoke English with a
fairly pronounced accent. In one interview, the instructor
presented himself as a likable teacher, respectful
of his students' intelligence and motives,
flexible in his approach to teaching, and enthusiastic
about his subject matter In the other interview, he
appeared to be quite unlikable, cold and distrustful
toward his students, rigid and doctrinaire in his
teaching style. After viewing the videotaped interview,
the subjects rated the instructor's likability,
as well as the attractiveness of his physical appearance,
his mannerisms, and his accent In order to
explore the question of subject awareness, some subjects
were then asked whether the instructor's likability
had affected their ratings of his appearance, mannerisms,
and accent. Others were asked the reverse
question, namely, whether their evaluation of his
appearance, mannerisms, or accent had affected their
ratings of his likability. It was anticipated that the
subjects would rate the instructor as having a more
attractive physical appearance, more attractive mannerisms,
and a more attractive accent when he was
likable than when he was unlikable It was also anticipated
that the subjects would be unaware of these
effects on their judgments and would deny that their
global evaluation of the instructor had affected their
other ratings.
Procedure
The subjects were 118 University of Michigan students
enrolled in introductory psychology; 62 were
males and 56 were females. The results for the two
sexes were entirely similar and therefore were combined
for presentation. The subjects participated in
the experiment in groups of 6-17. Sessions were conducted
in such a way as to insure that no overt
communication between subjects took place. (But in
any case, conclusions are not affected when the group
rather than the individual was used as the unit of
analysis.) On arrival, the subjects were ushered into
a well-appointed conference room with chairs oriented
toward a 19-inch (.48-meter) television monitor.
The subjects were told that the psychology department
had recently conducted a large-scale evaluation
of teachers of introductory psychology:
At the end of the term teachers were evaluated by
their students on a number of dimensions, ranging
from traditional questions of teaching effectiveness
to questions about matters such as physical appearance
and mannerisms We are interested in
finding out whether evaluations of teachers who
are seen in brief videotaped interviews at all resemble
the evaluations of these teachers by students
who were exposed to them for a whole semester.
We will ask you to look at these videotapes and
then rate the teachers on some of the same dimensions
employed in the original study.
The subjects were then shown two videotaped interviews
All subjects were first shown a 7-minute
filler interview with a bright and pleasant instructor
who answered such questions as, What is your general
educational philosophy? Do you encourage student
discussion? How do you handle evaluations of
student work? Then half the subjects saw the Belgian
instructor answer the same questions in his
warm guise and half saw him answer the questions
in his cold guise Each of these interviews also lasted
about 7 minutes.
The warmness or coldness of the instructor was apparent
in his answers to virtually all the questions
For example, in his warm guise the instructor answered
the question about student discussion by saying
that, yes, he encouraged discussion and that he
was stimulated by the give and take it provided and
felt that it enhanced student interest in the lecture
material. In his cold guise he answered the question
by saying that he didn't allow much discussion because
"there's a time to be a student and a time to
be a professor" and he wouldn't be the professor if
he didn't know more than they did. In his warm
guise the instructor answered the question about
evaluation of student work by saying that he gave
paper assignments that the students seemed to like
and profit from and gave exams that were a mixture
of objective questions and thought questions. In his
cold guise he said he gave weekly multiple-choice
quizzes because otherwise "y°u
can't trust them to
do the reading."
HALO EFFECT 253
Dependent Measures
After seeing each interview, all subjects were asked,
"How much do you think you would like this
teacher?," and were asked to rate him on an 8-point
scale ranging from "like extremely" to "dislike extremely."
The questionnaire then read: "Listed below
are some characteristics of the teacher you saw.
Please indicate how you think you would feel about
each of the characteristics if you were to take a
course from the teacher" The characteristics included
physical appearance, mannerisms, and for the
Belgian instructor, accent. The subjects rated each
characteristic on the scale below:
— extremely appealing
— very appealing
— somewhat appealing
— slightly appealing
— slightly irritating
— somewhat irritating
— very irritating
— extremely irritating
In addition, 34 subjects were told that the investigators
were interested in knowing "how much, if at
all, your liking for the teacher influenced the ratings
you just made." The subjects indicated their answers
on scales of the following type:
My liking for the teacher caused me to rate his
physical appearance:
much higher
higher
slightly higher
had no effect
slightly lower
lower
much lower
Another 56 subjects were asked to indicate "how
much, if at all, the characteristics you just rated influenced
your liking of the teacher." The subjects
indicated their answers on scales of the following
type:
Physical appearance made me like him:
much more
more
slightly more
had no effect
slightly less
less
much less
After responding to the questionnaire, the subjects
were quizzed intensively about their reactions to the
videotapes and to the questionnaire items, and then
were debriefed.
Results
Effect of Global Evaluation on
Evaluations of Attributes
The manipulation had a very pronounced
effect on liking of the teacher. The t value
contrasting the warm condition mean of 5.48
with the cold condition mean of 3.18 is 8.62
(df = 1/117, p < .00001 ) . 1
Figure 1 presents
ratings of physical appearance, mannerisms,
and accent as a function of the liking manipulation.
For ease of presentation, all of the
positive (appealing) ratings are grouped together
and contrasted with the negative (irritating)
ratings. A small number of subjects
checked the point intermediate between appealing
and irritating, and these subjects are
included with the irritating ratings. It may be
seen that the effect of the manipulation on attribute
ratings was very pronounced. A substantial
majority of the subjects who saw the
teacher in his warm guise rated his physical
appearance as appealing, whereas a substantial
majority of those who saw him in his cold
guise rated his appearance as irritating (t =
4.71, df - 1/H7, p < .00001). A majority of
warm condition subjects rated the teacher's
mannerisms as appealing, whereas a majority
of cold condition subjects rated bis mannerisms
as irritating (I = 4.06, df - 1/117, p <
.0001). About half the warm condition subjects
rated the teacher's accent as appealing
80
60
PERCENT
OF 4 0
SUBJECTS
20
0
80
60
PERCEUT
OF 4 0
SUBJECTS
20
0
80
60
PERCENT
Of 4 0
SUBJECTS
20
WARM TEACHER I COLD TEACHER
Physical Appearance
Mannerisms
Accent
APPEAUW IRMTHIIC
• •1APfOLIK IRRITATIHC
Figure 1. Ratings of Teacher's Physical Appearance,
Mannerisms, and Accent as a Function of Manipulated
Warmth versus Coldness.
1
All p values are based on two-tailed tests.
254 RICHARD E. NISBETT AND TIMOTHY DECAMP WILSON
and half as irritating, whereas the overwhelming
majority of subjects in the cold condition
rated his accent as irritating (* = 3.78, df —
1/117, p< .0002).
The results demonstrate that global assessment
of a person can powerfully alter evaluations
of particular attributes. That is, this is
the case if one is willing to make the assumption
that the attributes manifested by the
confederate were in fact constant across experimental
conditions. Although this is a reasonable
assumption, it is by no means indisputable.
It might have been the case, for
example, that the teacher smiled a lot in his
warm guise, thus making his appearance more
appealing and/or frowned a lot in his cold
guise, thus making it less appealing. Similarly,
his mannerisms, both physical and verbal,
might have differed when he was playing the
two roles. It is even conceivable that his accent
might have been less pronounced and
harsh when he was playing the warm role
than when playing the cold role.
In order to demonstrate that the teacher's
physical appearance did not in fact differ
across the two conditions, a follow-up study
was conducted with 34 subjects from the same
introductory psychology pool. These subjects
80
60
Kucat
Of 4 0
SUBJECTS
20
80
60
PERCENT
or 40
SUBJECTS
20
80
60
PERCENT
OF 4 0
SUBJECTS
20
WARM TEACHER | COLO TEACHER
Physical Appearonc*
DECREASED »0 INOEASED
RATH* EFFECT M T K
DECREASED NO INCREASED
RATUK EFFECT RATHX
60
PERCENT 4 0
Of
SUBJECTS 2 0
PERCENT
OF 4 0
SUBJECTS
20
60
PERCEUT
OF 4 0
SUBJECTS
20
WARM TEACHER | COLD TEACHER
Physical Appearance
DECREASED M MOUSED
RATNK EFFECT R O M
DECEASED M INCREASED
RATHK EFFECT BATIK
Figure 2. Reported Effect of Teacher Likability on
Ratings of Physical Appearance, Mannerisms, and
Accent.
Figure 3. Reported Effect of Reactions to Physical
Appearance, Mannerisms, and Accent on Ratings of
Likability of Teacher.
were shown the interviews minus the audio
portions (purportedly for a study of nonverbal
communication) and were then asked to rate
the physical appearance of the teacher, as well
as his mannerisms. The subjects shown the
warm version rated the teacher's physical appearance
and mannerisms only trivially higher
than the subjects shown the cold version (for
both, t < 1). This indicates that the physical
appearance and nonverbal mannerisms of the
teacher did not in fact differ across conditions.
Thus, it can be assumed that the ratings of
physical appearance for the original subjects
differed solely because of the global affective
difference induced by the audio portion of the
interview.
Awareness of the Influence of
Global Evaluation
The subjects were apparently unaware of
the nature of the influence of global evaluation
on their ratings. Thirty-four of the subjects
were asked if their liking for the teacher
had affected their ratings of his attributes.
Figure 2 presents these subjects' answers to
this question. For ease of presentation, the
data are divided into three categories: (a)
decreased rating (representing 3 scale points),
(b) no effect (representing 1 scale point), and
(c) increased rating (representing 3 scale
points). It may be seen that a majority of
HALO EFFECT 255
subjects, in both the warm and cold conditions,
believed that their overall liking for
the teacher had not influenced their ratings.
Moreover, for all ratings in both conditions,
as many subjects felt that their overall evaluation
had increased their ratings as believed
it had decreased their ratings. None of the
distributions differs significantly from the
zero, no effect point (all six one-sample ts
not significant).
Fifty-six subjects were asked if their evaluations
of the teacher's appearance, mannerisms,
or accent had affected their liking of the
teacher. Figure 3 presents the results. It may
be seen that the subjects who saw the warm
teacher were divided on the question of
whether the ratings of particular attributes
affected liking. Only for the physical appearance
question did the majority of subjects
express the opinion that appearance had had
no effect on liking. For the other two questions,
approximately equal numbers of subjects
believed that the ratings had increased,
decreased, or had no effect on liking. None of
the three distributions differs significantly
from a mean of zero.
For the subjects who saw the cold teacher,
the results were quite different. On balance,
the subjects believed that their evaluations of
each of the three attributes had decreased
their liking for the teacher. The mean of all
three distributions differs from zero (t for
physical appearance = 3.28, p < .01; t for
mannerisms = 2.80, p < .01; / for accent =
2.74, p < .01).
Thus, the subjects appear to have gotten
matters precisely backwards. Their liking for
the teacher was manipulated, and this affected
their ratings of particular attributes. Yet the
subjects did not acknowledge this effect, and
the subjects who saw the cold teacher actually
reported the opposite effect. These subjects
reported that their evaluations of the teacher's
attributes lowered their liking for him, although
they denied that their liking for the
teacher had affected their ratings of his attributes.
(Logically, of course, it is possible that
the subjects who saw the cold teacher are
partly right. It is conceivable that the
teacher's appearance, mannerisms, and accent
we
re all intrinsically unattractive. In that
ev
ent, the subjects who saw the cold teacher
might have been accurate when they reported
that their liking for the teacher was decreased
because of their evaluations of these attributes.
If that were the case, then the subjects
in the warm condition erred in believing that
the teacher's appearance, mannerisms, and
accent were neutral and had had no effect on
their liking of him.)
Conversations with the subjects following
the experiment served to reinforce the implications
of Figures 2 and 3. The subjects were
asked if they were quite sure that their global
evaluations had not influenced their ratings of
the teacher's attributes. Most subjects persisted
in their denial—some warily, perhaps
because the}' suspected the experimenter knew
something they didn't, and some heatedly.
One young woman said, in a somewhat exasperated
tone, "Before he even began talking,
I made my judgment about his appearance
and then I stuck with it." The
conversations gave reason to suspect, in fact,
that some of the subjects in Figure 2 who reported
that their liking for the teacher had
influenced their ratings did not actually believe
this. One of the subjects who had indicated
this said, on being probed about his
answer: "Actually, I just turned the question
around I disliked his accent a little, so that
made me dislike him more."
Discussion
The present results support the strong interpretation
of the halo effect phenomenon.
They indicate that global evaluations alter
evaluations of attributes about which the individual
has information fully sufficient to
allow for an independent assessment. These
results, it should be noted, are consistent with
the very earliest theorizing about the phenomenon.
Thorndike (1920), who gave the
phenomenon its name, clearly believed that
it represented far more than an effect on
presumptions about or interpretations of the
evaluative meaning of attributes, but rather
that it represented a fundamental inability
to resist the affective influence of global
evaluation on evaluation of specific attributes.
Subsequent theorists have tended to
share this view, though in the absence of
strong evidence to support it.
256 RICHARD E NISBETT AND TIMOTHY DECAMP WILSON
The layman has probably never doubted
the existence of an inability to separate evaluations
of attributes from global evaluations,
at least at the affective extremes (and so long
as he was referring to the foibles of others).
The strong interpretation of the halo effect is
embedded in such sayings as "love is blind"
and "a face only a mother could love." The
present study, however, suggests that global
evaluations may have a pronounced impact on
evaluations of specific attributes even when
the global evaluations in question are less extreme
than love and hate.
The present results, and those of Landy and
Sigall (1974), also provide prima facie evidence
that people lack awareness of the influence
of one evaluation on another. Indeed,
it would appear that the altered judgments
require the absence of awareness. The subjects
in the present study would surely have
been disconcerted to know that they were incapable
of rendering an independent evaluation
of such characteristics as physical appearance,
mannerisms, and accent. They would
have been disconcerted, that is, if they could
have been persuaded that they had been influenced
by their global evaluation. The subjects
were quite unfazed by the debriefing,
probably because each subject exempted himself
from the accusation that he had been
prone to such an influence. The subjects in
the Landy and Sigall (1974) study would undoubtedly
have been even more horrified if
they could have been convinced that they had
upgraded their evaluations of the essay of an
attractive woman or downgraded their evaluations
of the essay of a plain woman.
But the prima facie case is not the only
one that can be made. The subjects in the
present study were directly quizzed about
their beliefs concerning the nature and direction
of influence. They overwhelmingly reported
that their ratings of particular attributes
had not been affected by their global
evaluations. The subjects who saw the cold
instructor actually believed that the influence
ran in the opposite direction—that their dislike
of the teacher's attributes prompted their
unfavorable global evaluation.
When considering the establishment of
rules concerning blind review, conflict of interest,
nepotism, and the like, it would therefore
seem advisable to consider more than the
possibility that some individuals in the system
may be venal and corrupt. The protestations
of even the most virtuous and disinterested
participants that they are capable of independent
judgments should be considered sus-
pect.
References
Anderson, N. H. Information integration theory: A
brief survey. In D H. Krantz, R. C. Atkinson,
R. D. Luce, & P Suppes (Eds.), Contemporary
developments in mathematical psychology (Vol. 2).
San Francisco: Freeman, 1974.
Asch, S. E. Forming impressions of personality. Journal
of Abnormal and Social Psychology, 1946, 41,
258-290.
Clifford, M. M., & Walster, E. The effect of physical
attractiveness on teacher expectations. Sociology of
Education, 1973, 46, 248-258.
Dion, K K Physical attractiveness and evaluation of
children's transgressions Journal of Personality
and Social Psychology, 1972, 24, 207-213
Dion, K. K., Berscheid, E., & Walster E. What is
beautiful is good. Journal of Personality and Social
Psychology 1972, 24, 285-290
Landy, D., & Sigall, H. Beauty is talent: Task evaluation
as a function of the performer's physical attractiveness.
Journal of Personality and Social
Psychology, 1974, 29, 299-304.
Miller, A G. Role of physical attractiveness in impression
formation. Psychonomic Science, 1970, IP,
241-242
Nisbett, R. E., & Wilson, T. D. Telling more than we
can know: Verbal reports on mental processes.
Psychological Review, 1977, 7, 231-259.
Thorndike, E. L A constant error in psychological
ratings. Journal of Applied Psychology, 1920, 4,
25-29.
Received August 30, 1976 •