Fritz Heider (1946)
First published in The Journal of Psychology, 21, 107-112.
Attitudes towards persons and causal unit formations influence each other. An attitude towards an event can alter the attitude towards the person who caused the event, and, if the attitudes towards a person and an event are similar, the event is easily ascribed to the person. A balanced configuration exists if the attitudes towards the parts of a causal unit are similar. 
It is tempting to generalize from this statement and to omit the restriction to causal unit formation. Do units in general interact with attitudes in a similar way?
In trying out this hypothesis we shall understand by attitude the positive or negative relationship of a person p to another person o or to an impersonal entity x which may be a situation, an event, an idea, or a thing, etc. Examples are: to like, to love, to esteem, to value, and their opposites. A positive relation of this kind will be written L, a negative one ~L. Thus, pLo means p likes, loves, or values o, or, expressed differently, o is positive for p.
The relation "unit" will be written U. Examples are: similarity, proximity, causality, membership, possession, or belonging. pUx can mean, for instance, p owns x, or p made x; p ~ Ux means p does not own x, etc. Other relations which, in many ways, seem to function like units are: p is familiar with, used to, or knows well o or x, and p is in situation x. In lumping together all these relations we are, of course, aware of the dissimiarities between them. Only in a first approximation can they be treated as belonging to one class.
The hypothesis may be stated in greater detail thus: (a) A balanced state exists if an entity has the same dynamic character in all possible respects (e.g., if p admires and at the same time likes o) ; in other words, if pLo or p ~ Lo is true for all meanings of L. (We may anticipate here that the analogous statement for pUo does not seem to hold in a general way.) (b) A balanced state exists if all parts of a unit have the same dynamic character (i.e., if all are positive, or all are negative), and if entities with different dynamic character are segregated from each other. If no balanced [p. 108] state exists, then forces towards this state will arise. Either the dynamic characters will change, or the unit relations will be changed through action or through cognitive reorganization. If a change is not possible, the state of imbalance will produce tension.
The first part of the hypothesis refers to influence of dynamic relations or attitudes on each other. Since the different dynamic relations are not included in each other logically ("p likes o" does not imply "p admires o"), the same o or x can be positive in one respect and negative in another. An example in point is the conflict between duty and inclination. A tendency exists to make the different dynamic relations agree with each other by means of cognitive restructuring (excuses or rationalizations). Another example would be the tendency to admire loved persons and to love admired persons.
More numerous are the possibilities to which the second part of the hypothesis refers. They can be grouped according to the entities making up the configurations: (a) person and non-person (p, x) ; (b) two persons (p, o); (c) two persons and a non-person (p, o, x); (d) three persons (p, o, q). Many of the examples seem to substantiate the hypothesis. Examples which do not fit may eventually lead to greater insight into the nature of the dynamic characters and of the unit relations. All examples refer to p's life space. This is true even of oLp which therefore means: p thinks that o likes or admires p.
(a) p and x. Since the own person (p) is usually positive, a balanced state will exist if p likes what he is united with in any way, or if he dislikes the x he is segregated from. The cases (pLx) + (pUx) and (p ~ Lx) + (p ~ Ux) are balanced. Examples: p likes the things he made; p wants to own the things he likes; p values what he is accustomed to.
(b) p and o. Analogously, the two balanced states for p and o will be: (pLo) + (pUo) and (P ~ Lo) + (p ~ Uo). Examples: p likes his children, people similar to him; p is uneasy if he has to live with people he does not like; p tends to imitate admired persons; p likes to think that loved persons are similar to him.
pUo is a symmetrical relation, i.e., pUo implies oUp. That they belong to a unit is true for p and o in the same way, though their roles in the unit may be different (for instance, if U is a causal unit). However, pLo is non-symmetrical since it does not imply oLp. It is in line with the general hypothesis to assume that a balanced state exists if pLo and oLp (or p ~ Lo and o ~ Lp) are true at the same time. Attraction or repulsion between p and ~o are then two-way affairs; the relation is in symmetrical harmony. pLo is a non-symmetrical relation logically, but psychologically it [p. 109] tends to become symmetrical. Examples: p wants to be loved by an admired o; p dislikes People who despise him. oLp is similar to pLo admire him to pUo. Examples: p likes to meet people who, he is told, admire him.
(c)p, o, and x. The combinations become more numerous with three entities making up the configurations. Only a few possibilities can be mentioned. We shall always give the balanced state in symbols before stating the examples which refer to it.
(pLo) + (pLx) + (oUx). Both o and x are positive and parts of a unit. Examples: p admires clothes of loved o; p wants to benefit his friend o; p likes to think that his friend benefits him. A seeming exception is the case of envy. If o owns x (oUx) and p likes x (pLx), p ~ Lo may often follow. This exception can be derived from the fact that ownership is a one-many relation. A person can own many things but each thing can, ordinarily, be owned only by one person. Therefore "o owns x"excludes "p owns x," or oUx implies p ~ Ux. Since pLx may tend toward pUx, conflict is introduced.
Implications between unit relations often lead to conflict. Lewin's three cases of inner conflict rest on implications. Approach to a positive valence may imply withdrawal from another positive valence. Withdrawal from a negative valence may imply approach to another negative valence. Finally, approach to a positive valence may imply approach to a negative valence if both are located in the same region. Analogously, one can talk of three cases of outer conflict between persons. pUx may imply o ~ Ux (for instance, if U means ownership), and if both want x, conflict (competition) will arise. In the same way conflict appears if p and o want to get away from x but only one of them can do so (if p ~ Ux implies oUx and vice versa). Lastly, it may happen that p likes x and o hates it, but p and o have to move together (pUx implies oUx, e.g., in marriage). They either can both have x, or both not have it.
Trying out variations of the triad (pLo) + (pLx) + (oUx), we find that (pLo) + (pLx) + (oLx) also represents a balanced case. Examples: P likes what his friend o likes; p likes people with same attitudes. This case is not covered by the hypothesis unless we treat L as equivalent to U. Actually, in many cases the effects of L and U in these configurations seem to be the same. Furthermore, this case shows the psychological transitivity of the L relation. A relation R is transitive if aRb and bRc imply aRc. Thus, p tends to like x if pLo and oLx hold. As in the case of the symmetry of the pLo relation, we again have to stress the difference between logical and psychological aspect. Logically, L is not transitive but there exists a psychological tendency to make it transitive when implications between [p. 110]U relations do not interfere with transitivity. The relation U, too, seems (oUx) can lead to be in this sense psychologically transitive. (PUO) + pUx; p feels responsible for what people belonging to him do.
Taking into account these considerations, we can reformulate the hypothesis: (a) In the case of two entities, a balanced state exists if the relation between them is positive (or negative) in all respects, i.e., in regard to all meanings of L and U. (b) In the case of three entities, a balanced state exists if all three possible relations are positive in all respects.
The question arises whether, with a triad, one can make any generalizations about balanced cases with negative relations. For instance, (pLo) + (o ~ Ux) + (p ~ Lx) is balanced. Examples: p likes o because o got rid of something p dislikes. In this case two entities, p and o are related positively to each other, while both are related negatively to the third entity x. This holds generally: the triad of relations is in balance. if two relations are negative and one positive. This statement can be derived from the assumption that L and U are, in a balanced configuration, exchangeable, symmetrical, and transitive. L and U can then be treated as formally analogous to an identity relation. The "balanced" cases with three terms are for this relation: a = b, b = c, a = c; a = b, b ¹ c, a ¹c; a ¹ b, b¹ c, a ¹ c. By substituting L or U for the identity sign one obtains the balanced cases for these relations, though the case with three negative relations does not seem to constitute a good psychological balance, since it is too indetermined.
Therefore, the second part of the hypothesis must be stated as follows: ~(b) In the case of three entities, a balanced state exists if all three relations are positive in all respects, or if two are negative and one positive.
(pLo) + (oLx) + (pUx). Examples: p likes o because o admired p's action; p wants his friend o to like p's productions; p wants to do what his friends admire.
(pUo) + (pLx) + (oLx). Examples: p wants his son to like what ,ie likes; p likes x because his son likes it.
(d) p, o, and q. Among the many possible cases we shall only consider one (pLo) + (oLq) + (pLq). Examples: p wants his two friends to like each other. This example shows, as the parallel case with x instead of q, the psychological transitivity of the L relation.
However, the transitivity of the L relation is here restricted by implications between unit relations when L represents a one-one love relation. p does not want his girl friend to fall in love with his boy friend q because oLq in this case implies oLp, which conflicts with pLo. Jealousy, as well [p. 111] as envy and competition, is derived from implications between unit relations.
After this discussion of the different possibilities there are several more points worth mentioning which refer to exa~mples of different groups. One is the problem of self evaluation. High self regard of p can be expressed by pLp, low self regard by p ~ Lp (though the two p's in these expressions are not strictly equivalent). All of the examples so far considered presupposed pLp. However. one also has to take into account the possibility of p ~ Lp. As to be expected, it plays a role contrary to that of pLp. Examples: if p has low self regard he might reject a positive x as too good for him; if p has guilt feelings he will think he ought to be punished; if his friend admires his product he will think it only politeness. A negative action attributed to himself will produce p ~ Lp, etc.
The equivalence of the L and U relations seems to be limited by the fact that often the U relation is weaker than the L relation. One can assume, that pLx brings about pUx (p wants to have a thing he likes) more often than pUx produces pLx (p gets to like a thing which belongs to him). Again (pLo) + (oLx) usually will lead to pLx (transitivity), but (pUo) + (oUx) will not do so if there holds at the same time p ~ Lo.
We saw that one can derive forces towards actions, or goals, from the configurations. It can also happen, that the choice of means to a goal is determined by these patterns. If p wants to produce oLx, and he knows that oLp holds, he can do so by demonstrating to o the relation pLx, because (oLp) + (pLx) will lead to oLx. If p wants to bring about oLp, and he knows that oLx holds, he can produce pUx, for instance, he will perform an act o approves of.
An examination of the discussed examples suggests the conclusion that
a good deal of inter-personal behavior and social preception [sic] is determined
or at least co-determined by simple cognitive configurations. This fact
also throws light on the problem of the understanding of behavior. Students
of this problem often mentioned the aspect of rationality which enters
into it. Max Weber and others pointed out one kind of rationality in behavior,
namely, the rationality of the means-end relation. Choosing the appropriate
means to gain an end makes for a "good," a "rational" action, and we can
understand it. In Lewin's concept of methodological space this kind of
rationality is elaborated. However, understandable human behavior often
is not of this sort, but is based on the simple configurations of U
and L relations. Since they
determine both behavior and perception we can understand social behavior
of this kind.
1. HEIDER, F. Social perception and phenomenal causality. Psychol.
Rev., 1944, 51, 358-374.