Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

ON SOCIAL COGNITION1

FRITZ HEIDER
University of Kansas

T HIS talk will deal with social cognition and


with the way we perceive and think about
other people—about their attitudes, their in-
tentions, their emotions and actions, and about
different places in the building. But this class dis-
order has made possible a different and much more
complex kind of structure. Instead of having one
relation, namely similarity of parts, determining
their relations to us and to each other. And it will the order there is now a manifold of very different
deal mainly with one aspect of this thinking about relations which nevertheless link the different parts
other people: namely, the way in which several to make a whole adapted to a particular use.
items of information combine and make up orderly This example refers only to static structures and
structures. does not take into account changes that occur
True enough—we also find structure and order, within structures. But similar differences can be
implications and patterns in our thinking about the observed when we consider cases in which one
nonsocial aspects of the world. But it would not element causes a change in others.
be unreasonable to suggest that the most com- Ashby (1960), in his book, Design for a Brain,
plexly interwoven cognitive patterns occur when talks about "poorly joined" and "richly joined"
people try to give an account of social situations to environments. A poorly joined environment is one
themselves or to others. in which "most variables have an immediate effect
This, then, is what I want to discuss, order and on only a few of the totality of variables [Ch. 16]."
the relation of components to other components Ashby is probably correct in stating that our
and the relation of parts to an encompassing struc- ordinary nonsocial environment is "poorly joined."
ture in our thinking about social events. He gives the examples of dipping the pen into an
Of course, one can speak of order or structure ink well or opening a door—acts which affect only
in two senses. One of these is concerned with the the variables that are directly concerned but which
way we group things into concepts, and how we have no influence on the chairs, the electric light,
establish a hierarchy of superordinate and sub- the books on the shelves, and a host of other parts
ordinate concepts. In the other sense, we con- of the environment.
sider how components are linked together into a On the other hand, a "richly joined" environ-
more complex structure, as they are, for example, ment is one in which "every variable is joined
in a syllogism or a story. To use an analogy: When directly to every other variable so that what hap-
a house is being built, materials may be assembled pens at each variable is conditioned . . . on the
in advance on the site—bricks here, sand for values of all the other variables in the system." As
cement there, etc. The components are thus an example of a richly joined environment, Ashby
grouped according to similarity. This grouping is uses a set of simultaneous equations in which the
a low-level structure. It is like a classification ac- values of the unknowns depend on the whole set
cording to concepts: Things that belong to one and where changing any part of the set implies a
class are put together. change in the solution. Logical syllogisms have
By the time the house is built, the order accord- the same character: The conclusion depends on the
ing to classes has been completely broken up: Parts whole set of premises and not on single parts.
that belong to the same class appear in many Now I would like to submit that in opposition
1
This research was supported by Grant 05-224 from the to the environment of solid, everyday things (and
National Institute of Mental Health. I also want to here we would have to exclude some of the more
acknowledge the assistance of the Computation Center of complicated things like radio receivers or com-
the University of Kansas which is supported in part by
the Public Health Service and the National Science Founda- puters) our social environment consisting of other
tion. persons with ourselves among them is usually a
25
26 AMERICAN PSYCHOLOGIST

system that is richly joined. In a group of people by interactions among several factors. In a typical
the change of a single variable, e.g., one person case, the pleasantness or unpleasantness of a situa-
behaving in a certain way, may very well bring tion depends on a set of three propositions. In
about changes in every one of the other members the p-o-x triad as Jordan (1953) used it, we were
of the group, and, what is more, not the same dealing, for example, with the propositions: Person
changes in each one. The kind of change that will P likes Person O. P likes Thing x. O likes x.
occur in each case will depend on many factors, The pattern can be varied by making the com-
e.g., preexisting attitudes of the interacting per- ponent sentences either positive or negative. What
sons, the relations of these persons to third per- is especially significant in this work on balance is
sons, etc. Not only do we find that there are in- that it shows how often the total set of proposi-
fluences from every part to every other part, but tions rather than any single item is involved in
also that influences are specific to parts in such a determining the preference for the triad, though
way that the whole may become very complex. certainly single items also play a role in this kind
But, with all this complexity—and here is a of cognition.
point of major importance for the problem of social It is not hard to find further examples of triads
cognition—these wholes are easily grasped by any- of sentences which are linked together in a way
one who is familiar with the group. This means that reminds us of syllogisms and which play a role
that not only the systems of the objective social in our thinking about the behavior of other people.
world are often richly joined, but that the thought For instance, you will agree that if a person can do
systems that we use when we deal with them also something, and he tries to do it, then he will suc-
have that character. Let us say we have arrived ceed in doing it (Heider, 1958). If he can stand
at a definite idea of a social situation, and this idea on his head, and he tried to do it, he will do it.
may contain quite a few variables; for example, per- This is a kind of conceptual truth—we do not have
sonality traits of the people involved, relations be- to make an experiment to find out whether it is
tween them, mutual attitudes, etc., etc. In many really so. The relation between can, try, and
cases these factors are closely connected and a piece succeed is determined by the way in which these
of new information could easily affect a great concepts are defined, and when one states it in the
number of them. Also, these variables can usually way I did just now, the relation is obvious. But we
be expressed in sentences, and just as sentences should not overlook it and forget it because it is
are combined in, a paragraph or a story, so these obvious. Obvious relationships often appear in a
variables are often the components of an encom- disguised, embedded form in which we do not recog-
passing structure. George Miller (1965) has said, nize them at first glance.
"The meaning of an utterance is not a linear sum These are some examples of the patterns of so-
of the words that compose it [p. 18]." One can cial cognition that I am trying to describe. I want
go further and say, "The meaning of a paragraph to spend the rest of my time talking about two
or story is not a linear function of the meanings of different experiments that I am making at the
its sentences." University of Kansas with the collaboration of
To be sure, the way the meanings of the sen- Grace Heider under a grant from the National In-
tences interact to produce the meaning of the stitute of Mental Health. The first of these con-
bigger whole is not any more a matter of language cerns the cognitive structure of a story. We used
forms, but of thought forms; and these thought stories in this investigation because stories are
forms can sometimes be considered a kind of logic, about people and in understanding and interpreting
and at other times a psycho-logic, as Abelson and them we employ a wide range of the schemata of
Rosenberg (1958) have put it. Demonstrations of social cognition. In many ways a story is one of
these interfacings of causal connections in inter- the best examples of a richly joined structure. One
personal relations are given in many recent studies. of the most complex cognitive achievements of
A good example is the monograph by E. E. Jones everyday life is the recall of the general features of
(1964) on Ingratiation. Other illustrations are a play or a movie. It is an astounding feat if we
found in studies of balance that show that some of consider the intricacies of the stimulus pattern to
our responses to a social situation are determined which we are exposed for 2 or 3 hours, with the
ON SOCIAL COGNITION 27
Dotted Line Indicates Path of Movement

Jj
C -\

LT
u

cc/
201 /,,' 2f

FIG. 1. Outline of story in cartoon form.

continuous changes that take place in a profusion a single factor in a photographic sequence without
of visual and auditory stimuli. Yet it is something also introducing other changes, e.g., in lighting or
that all of us do, and with an amazing degree of tempo. Therefore, we searched for other material
accuracy. that would serve a similar purpose. In one such
But to describe our experiment: We were led attempt the original story of the motion picture
to work on this problem of the cognitive organiza- was presented in a series of stills—-the cartoons
tion of longer meaningful passages by an experi- which you see in Figure 1. (This experiment was
ment that I made some years ago with Marianne performed by Richard Lawless, graduate student
Simmel (Heider & Simmel, 1944). In this experi- at Kansas University.) Subjects were well able
ment, subjects were shown a short moving picture to follow the story in this still-frame form. Their
in which three geometrical figures moved about in reports were similar in general structure to those
different ways and were asked to write a short that we had obtained with the moving picture,
account of what they had seen. In most cases though they were less rich and less subtle on the
they described the movements of the geometrical whole. This result seemed to indicate that, for our
figures in terms of actions of people or animals, purposes, it did not matter very much what medium
and they tied the parts of their stories together to is used for the presentation, and in our next at-
make units with meaningful sequences of events. tempt we tried a verbal form of the same plot.
In other words, most of our subjects wrote stories This instrument consisted of 18 short sentences
that we might call richly joined. (averaging S words each) as follows:
This moving picture has been used in a number
Jim is standing in front of his house.
of experiments and has been a fruitful source of
Bob and Mary come along.
suggestions for further study, but it has serious Jim approaches them.
limitations as an experimental tool, especially in its Bob goes toward him.
lack of flexibility. It is technically difficult to vary Jim and Bob begin to quarrel.
28 AMERICAN PSYCHOLOGIST

Bob goes away. whole in which many of the parts fitted specific
Jim goes toward Mary. constraints, and into which only a limited number
Mary goes into the house.
Jim goes into the house.
of substitutions could have been introduced.
Bob returns. For the purpose of this report I will not attempt
He looks through the door into the house. a systematic review of the results as related to our
Mary runs out and joins Bob. two independent variables, the sex of the subjects and
Bob and Mary run away together. the word, "sad" or "glad." But I want to mention
Jim comes out of the house.
He looks around for Bob and Mary. a few of the integrating factors that showed them-
He cannot find them. selves. There were some that had obvious relation
He is sad. (Or, with another experimental group: He is to something like a balance hypothesis: The char-
glad.) acters that were described in the plot as moving
Jim goes back into the house. together (Bob and Mary) were described as liking
Except for the word "sad" or "glad," the 18 state- each other, and were also rated as similar to each
ments of the plot consist almost entirely of descrip- other on our valuative dimensions. When Jim was
tions of overt actions: stand, come along, approach, sad at being unable to find Bob and Mary, he was
etc. The principal independent variable was the seen as liking them more than when he was glad.
two words, "sad" and "glad," which referred to Of course the content of the story also depended
Jim's feelings at the end when he was unable to on the role suggested by the plot for the cast of
find the other two persons of the story. characters. In preliminary experiments where the
The sex of the subjects was treated as a second names of the characters did not so clearly suggest
independent variable. There were 200 subjects 2 in a girl and two rivals for her affections, there were
all, divided into four experimental groups accord- a number of stories about a father and mother and
ing to sex and the critical word of the plot. child, or a father or brother in relation to a girl and
The plot was presented as part of a mimeo- boy friend. These, in brief survey, are some of the
graphed booklet containing instructions and allow- relations that made for the fitting together or the
ing space for information regarding a number of meanings of the sentences.
dependent variables. In this report I shall discuss An especially interesting outcome of several dif-
only two: 7-point ratings of each character on three ferent analyses of these ratings is the difference in
dimensions that are heavily loaded on Osgood's stability between, overall structure and single ele-
(Osgood, Suci, & Tannenbaum, 1957) evaluative ments within the complex created by our subjects
factors (good-bad, kind-cruel, and wise-foolish), and from the elementary plot material. For example,
ratings on a 3-point scale of Jim's attitude toward judgments of Jim varied from group to group—he
the other two characters with points designated as might be seen as better or worse than the other
like, neutral, and dislike. two characters of the story, but in spite of that
We may say that this procedure consisted of giv- they were treated in each group as a pair and Jim
ing the subjects a sequence of 18 sentences, poorly as the outsider. This showed itself in correlations
joined in the sense that each one could have been between ratings: There were consistent negative
replaced by a great variety of very different sen- correlations between ratings of Jim and Bob and
tences; there were no constraints that required the positive correlations between ratings of Bob and
Mary.
presence of the actual sentences that were given.
The task of the subjects was, essentially, to produce In another approach to the problem of stability
from this a richly joined unit by providing reasons of structure as compared with that of details, our
for the actions that occurred and by introducing statistical collaborator, Juliet P. Shaffer of the
evaluative information regarding the characters. University of Kansas Psychology Department, used
On the whole, the subjects were well able to fol- an adaptation of the Euclidean distance measures
low the instructions and to produce an integrated previously described by Cronbach and Gleser
2
(19S3), and by others, to deal with the structure
We are grateful to undergraduate students from Duke
of the similarity relationships among the three
University and from the University of North Carolina in
Chapel Hill, Greensboro, and Raleigh who acted as sub- characters as they were rated by our subjects. The
jects in this series of experiments. sum of the ratings on the three evaluative scales
ON SOCIAL COGNITION

was used, and the distances were computed in a tive or a negative meaning. (We omitted, at this
space formed by taking each subject's score for point, possible gradations between the two poles.)
each character as a value on an orthogonal dimen- The design of the experiment was such as to use a
sion. The space has thus as many dimensions as full array of the possible combinations in each
there are subjects. The distances measured in this series—thus with a manifold of 3 statements, each
way showed the following order: Bob-Mary the of which could be either positive or negative, there
smallest; then Jim-Mary; and finally Jim-Bob the were 2 X 2 X 2 or 8 combinations. In other words,
largest distance. This ordering held for each of each proposition could be treated as a two-level
our four experimental groups, men-women, glad- factor in an analysis of variance design. The dif-
sad, and continued in the same direction when ference between these experiments and the usual
elevation and scatter differences were removed. balance experiment is that the propositions were not
This seemed a convincing demonstration of the restricted to liking and unit relations, and that the
relative constancy of structure as compared to the dependent variable could be changed from experi-
variability of details in this material. ment to experiment. We were interested both in
These results parallel what Bartlett found long commonalities of responses and in locating indi-
ago in his studies of Remembering in which single vidual differences that might be related to other
subjects were asked to recall story material again factors.
and again over a long period. Bartlett (1932) con- In this particular series of four experiments we
cluded from his work that in spite of changes in were dealing with factors that cause an action or an
details that occurred with time "the form, plan, agent to be judged as good or bad when his action
type, or scheme of the story seems, in fact, for was one that affected another person in some way.
the ordinary adult to be the most dominant and In describing these experiments I will designate the
persistent factor in this kind of material [p. 83]." persons as A and B. The factors considered in-
We had expected subjects to organize the informa- cluded the power relationship between the persons,
tion that was given in terms of the characters of A's liking for B, A's behavior toward B (he might
our story, and in a sense that was what they did. harm or benefit B), or the fact that A and B ap-
However, the core of the organization seems to have proach each other or withdraw from each other.
been the larger configuration of the three charac- I will not attempt, in this brief report, to describe
ters taken together. Their individual traits were the four-experiment design that we; used, but will
the details that were filled in with considerable only say that each complex of three or four state-
variation, but in most cases in such a way as to con- ments used a different combination of letter desig-
form to the whole structure. nations for the persons and that the 40 complexes
I now want to describe a series of experiments, were presented in a different randomized order to
that we carried out for the purpose of pinpointing each subject. The subject's task was to make a
more of the assumptions that are included in this good-bad judgment on a 7-point scale of a given
network that underlies our social cognition. These aspect of each complex. (For two series the evalu-
experiments, in one respect, are exactly opposite in ation was of one of the persons, in the other two
design to the one with the story. In that experi-
of the act performed by person A in relation to
ment we used a single independent variable, the word
person B.)
"sad" or "glad" that described Jim's feelings (or
two if we also count the sex of the subjects). In The actual experimentation was carried out by
this series we have used several independent varia- Sara Jane Spaulding, a University of Kansas stu-
bles and a single dependent variable—an evalua- dent. There were 100 subjects s in all, SO men and
tive judgment made on a 7-point scale. (In each SO women, but since we found rather few significant
case a rating of "1" refers to the favorable pole sex differences I will give results for the group as a
of the scale, a rating of "7" to the unfavorable whole, and because of limitations of time will take
pole.) Like some of the balance experiments, these up only two of the four experiments, Series I and
used sentences as elements, presenting them in com- Series IV.
plexes of three or four. Each sentence is a proposi- 8
Undergraduate students of the University of Kansas
tion that can be phrased to convey either a posi- and Baker University.
30 AMERICAN PSYCHOLOGIST
7 really a case of balance, though appearing here in
a somewhat unusual and unexpected guise.
6 Harm * * There is also a significant interaction between
the benefit-harm factor and the power factor.
5 When the agent has more power than the recipient,
o>
c
the effect of the difference between benefit and
14 harm on the judged value of the agent is much
greater than when he has less power. One might
Benefit consider the hypothesis that the greater his power,
Benefit the more the agent will be held responsible for his
action and the smaller his power, the more the
outcome of the action is seen as depending on other
_L I
factors.
o> e
5
OQ
o
I
Series IV used the triad:
Powerful
Main Effect Interactions A approaches B.
B knows that A likes B.
N = IOO **p«.01 B approaches A. How would you rate Person B?
SERIES I
FIG. 2. Main effect and interactions of Benefit-Power-Good
In this case, the first factor is A's act of ap-
experiment. proaching B or withdrawing from B. The second
factor tells us whether A likes or dislikes B, and
Series I consisted of three statements with the that B is aware of this. The third factor is B's re-
first arrangement as follows: action—he approaches A or withdraws from him.
Figure 3 shows the results of this series. There
A is more powerful than B. is again a strong main effect with B rated more
B is a good person.
A benefits B. How would you rate person A? favorably when he approaches than when he with-
draws. Approach as such has the connotation of
In other words, the first factor is the power rela- a friendly, positive act, in spite of the fact that it
tion between A and B: A is more powerful than B is sometimes connected with aggression.
or less powerful. Power was explained to the sub-
jects as having a broad meaning, referring to
status, money, or bodily strength, etc. The second 1 1 I I I I
factor is the value of B—good or bad—and the
third is A's action, benefiting B or harming B.
Figure 2 gives the results for this series, based
on mean ratings of Person A made by our 100 •<»* Reactor-^\
• ** ^^-« Withdraws \
subjects. We see that benefit versus harm showed ** N, _
a significant main effect, actually significant beyond
the .01 level. It is obvious that a person who
^Reactor ^ f -
benefits another will be judged more favorably than ^^— -• Approaches /
one who harms another. Neither the power rela-
tionship nor the goodness of the recipient as such
had a consistent influence on the evaluation of the 1 1 1 1 1 1
agent. However, there is a big interaction (again IS)
«
</)
* ^£ *
o'o fe a ° £ <£
far beyond the .01 level) between benefit versus °s
tj e
si,
<-»Jc o 1 .5^ ~% ~=
sa cc* s--s
harm and the goodness of the recipient. Our sub- cr<t !< Is z5 '.is
jects imply that it is better to benefit a good person Main Effect Interactions
N = 100
than a bad one, and worse to harm a good person
than a bad one. This is one of those cases of a SERIES H
combination of positive and negative relations, Fie. 3. Main effect and interactions of Approach-Like-
and a closer examination could show that it is Approach experiment.
ON SOCIAL COGNITION 31

We also find highly significant interactions be- First—a change in one part should produce
tween the action of B and the two other factors. changes in a number of other parts, and that was
Strongest is the interaction with like and dislike, tested in the story experiment with one inde-
and we have to remember that it is not the attitude pendent variable (sad-glad) and the many de-
of the person who is being rated that is involved. pendent variables.
He is judged according to whether he is approach- Second—a change in one part might be a func-
ing or withdrawing from someone who likes or tion of a combination of changes in other parts.
dislikes him. That was tested in the experiments that were made
If one spells out the interaction, it says that a with three or four independent variables and a
person who approaches another who likes him is single dependent variable. Both these kinds of
rated more favorably than one who approaches experiments show the relation between one part and
someone who dislikes him. The opposite is true for the other parts, but in one the changes in one part
withdrawal: The person who withdraws from some- are cause, in the other, effect.
one who dislikes him is rated more favorably than In the end, the second kind of experiment may be
one who withdraws from a person who likes him. the more fruitful for the study of the tissue of
It seems that in this case, the agent is approved social cognition.
of if he acts in conformity with the other's at-
titude toward him, and disapproved of if he acts in REFERENCES
opposition, to it. The one signifies a sympathetic ABELSON, R. P., & ROSENBERG, M. J. Symbolic psycho-
action, the other an antagonistic one. logic, a model of attitudinal cognition. Behavioral
The second interaction, between the action of Science, 1958, 3, 1-13.
the rated Person, B, and the previous action of Per- ASHBY, W. R. Design for a brain. New York: Wiley,
son A can be explained in the same way. B's ap- 1960.
BARTKETT, F. C. Remembering. Cambridge: Cambridge
proach or withdrawal can be in harmony with or University Press, 1932.
in opposition to the approach or withdrawal of A. CRONBACH, L. J., & GLESER, G. C. Assessing similarity be-
When B responds to A's opening move in kind he tween profiles. Psychological Bulletin, 1953, SO, 456-
is rated more favorably than when he responds 473.
antagonistically. HEIDEK, F. The psychology of interpersonal relations.
These two experiments deal with the problem of New York: Wiley, 1958.
HEIDER, F., & SIMMEL, M. An experimental study of ap-
person perception in a broad sense. That is, they
parent behavior. Journal of Psychology, 1944, 57, 243-
serve to demonstrate factors that are responsible 259.
for our image of another person. And they make JONES, E. E. Ingratiation. New York: Appleton-Century-
clear that this image is often not a simple function Crofts, 1964.
of a particular datum, but rather a function of the JORDAN, N. Behavioral forces that are a function of at-
interactions of a number of data. titudes and of cognitive organization. Human Relations,
1953, 6, 273-287.
I will conclude briefly: This talk has been con-
MILLER, G. A. Some preliminaries to psycholinguistics.
cerned with the interrelatedness of parts in struc- American Psychologist, 1965, 20, 15-20.
tures of social cognition. If these structures are OSGOOD, C. E., Suci, G. T., & TANNENBAUM, P. H. The
richly joined, to use Ashby's term, then we may measure oj meaning. Urbana: University of Illinois Press,
say: 1957.

You might also like