Classics in the History of Psychology

An internet resource developed by
Christopher D. Green
York University, Toronto, Ontario

(Return to index)


Neal E. Miller (1941)
(with the collaboration of Robert R. Sears, O.H. Mowrer, Leonard W. Doob & John Dollard)
Institute of Human Relations, Yale University

First published in Psychological Review, 48, 337-342.

The frustration-aggression hypothesis is an attempt to state a relationship believed to be important in many different fields of research. It is intended to suggest to the student of human nature that when he sees aggression he should turn a suspicious eye on possibilities that the organism or group is confronted with frustration; and that when he views interference with individual or group habits, he should be on the look-out for, among other things, aggression. This hypothesis is induced from commonsense observation, from clinical case histories, from a few experimental investigations, from sociological studies and from the results of anthropological field work. The systematic formulation of this hypothesis enables one to call sharp attention to certain command characteristics in a number of observations from all of these historically distinct fields of knowledge and thus to take one modest first step toward the unification of these fields.

A number of tentative statements about the frustration-aggression hypothesis have recently been made by us in a book. [2] Unfortunately one of these statements, which was conspicuous because it appeared on the first page, was unclear and misleading as has been objectively demonstrated by the behavior of reviewers and other readers. In order to avoid any further confusion it seems advisable to rephrase this statement, changing it to one which conveys a truer impression of the authors' ideas. The objectionable phrase is the [p. 338] last half of the proposition: "that the occurrence of aggression always presupposes the existence of frustration and, contrariwise, that the existence of frustration always leads to some form of aggression."

The first half of this statement, the assertion that the occurrence of aggression always presupposes frustration, is in our opinion defensible and useful as a first approximation, or working hypothesis. The second half of the statement, namely, the assertion "that the existence of frustration always leads to some form of aggression" is unfortunate from two points of view. In the first place it suggests, though it by no means logically demands, that frustration has no consequences other than aggression. This suggestion seems to have been strong enough to override statements appearing later in the text which specifically rule out any such implication. [3] A second objection to the assertion in question is that it fails to distinguish between instigation to aggression and the actual occurrence of aggression. Thus it omits the possibility that other responses may be dominant and inhibit the occurrence of acts of aggression. In this respect it is inconsistent with later portions of the exposition which make a distinction between the instigation to a response and the actual presence of that response and state that punishment can inhibit the occurrence of acts of aggression. [4]

Both of these unfortunate aspects of the former statement may be avoided by the following rephrasing: Frustration produces investigations to a number of different types of response, one of which is an instigation to some form of aggression.

This rephrasing of the hypothesis states the assumption that was actually used throughout the main body of the text. Instigation to aggression may occupy any one of a number of positions in the hierarchy of instigations aroused by a specific situation which is frustrating. If the instigation [p. 339] to aggression is the strongest member of this hierarchy, then acts of aggression will be the first response to occur. If the instigations to other responses incompatible with aggression are stronger than the instigation to aggression, then these other responses will occur at first and prevent, at least temporarily, the occurrence of acts of aggression. This opens up two further possibilities. If these other responses lead to a reduction in the instigation to the originally frustrated response, then the strength of the instigation to aggression is also reduced so that acts of aggression may not occur at all in the situation in question. If, on the other hand, the first responses do not lead to a reduction in the original instigation, then the instigations to them will tend to become weakened through extinction so that the next most dominant responses, which may or may not be aggression, will tend to occur. From this analysis it follows that the more successive responses of non-aggression are extinguished by continued frustration, the greater is the probability that the instigation to aggression eventually fail become dominant so that some response of aggression actually will occur. Whether or not the successive extinction of responses of non-aggression must inevitably lead to the dominance of the instigation to aggression depends, as was clearly stated in later pages of the book, upon quantitative assumptions beyond the scope of our present knowledge. [5] [6]

Frustration produces instigation to aggression but this is not the only type of instigation that it may produce. Responses incompatible with aggression may, if sufficiently instigated, prevent the actual occurrence of acts of aggression. In our society punishment of acts of aggression is a frequent source of instigation to acts incompatible with aggression.

When the occurrence of acts of aggression is prevented by more strongly instigated incompatible responses, how is the existence of instigation to aggression to be determined? If only the more direct and overt acts of aggression have been [p. 340] inhibited, as is apt to be the case because such acts are the most likely to be punished, then the instigation to aggression may be detected by observing either indirect or less overt acts of aggression. If even such acts of aggression are inhibited, then a different procedure must be employed. Two such procedures are at least theoretically possible. One is to reduce the competing instigations, such as fear of punishment, and observe whether or not acts of aggression then occur. The other is to confront the subject with an additional frustration which previous experiments have demonstrated would by itself be too weak to arouse an instigation strong enough to override the competing responses inhibiting the aggression in question. If the instigation from this additional frustration now results in an act of aggression, then it must have gained its strength to do so by summating with an already present but inhibited instigation to aggression. The presence of the originally inhibited instigation to aggression would be demonstrated by the effects of such summation. Thus the fact that an instigation may be inhibited does not eliminate all possibility of experimentally demonstrating its presence.

At this point two important and related qualifications of the hypothesis may be repeated for emphasis though they have already been stated in the book. It is not certain how early in the infancy of the individual the frustration-aggression hypothesis is applicable, and no assumptions are made as to whether the frustration-aggression relationship is of innate or of learned origin.

Now that an attempt has been made to clarify and to qualify the hypothesis, four of the chief lines of investigation which it suggests may be briefly considered. [7]

1. An attempt may be made to apply the hypothesis to the integration and elucidation of clinical and social data. Here the fact that certain forms of aggression are spectacu- [p. 341] larly dangerous to society and to the individual is relevant. This means that acute personality conflicts are apt to arise from the problem of handling aggression and that the problem of aggression is apt to play an important rôle in shaping certain great social institutions such as the in-group as an organization against the out-group.

2. An attempt may be made to formulate more exactly the laws determining the different ways in which instigation to aggression will be expressed under specified circumstances. Some of the problems in this field are suggested by the phenomena of displacement of the object of aggression, change in the form of aggression, and catharsis of aggression.

3. An attempt may be made to secure more information concerning the other consequences which frustration may produce in addition to the instigation to aggression. Such an attempt would lead into studies of rational thought and problem solution as suggested in the classical work of John Dewey, and into studies of experimental extinction, trial-and-error learning, substitute response and regression. [8] Work along this line of investigation may deal either with the clinical and social significance of these other consequences of frustration or with the discovery of the laws governing them.

4. An attempt may be made to improve or to reformulate the basic frustration-aggression hypothesis itself. The determination of the laws which allow one to predict exactly under which circumstances instigation to aggression may be expected to occupy the dominant, the second, the third, or some other position in the hierarchy of instigations aroused by a frustrating situation is a most important problem of this type. Another problem is the reduction of the frustration-aggression hypothesis to more fundamental principles and the more accurate restatement of the hypothesis in terms of these more basic principles. One of the steps in this direction would be to scrutinize any exceptions to the hypothesis as now formulated. Another step would involve a careful study of the early stages of the socialization of the individual in an [p. 342] attempt to analyze the interlocking rôles of three factors: first, innate physiological reaction patterns; second, learning mechanisms; and third, the structure of the social maze which poses the learning dilemmas and contains the rewards and punishments. An empirical and theoretical analysis along these lines might lead to a fundamental reformulation getting a closer approximation of the socially and scientifically useful truths imperfectly expressed in the present frustration-aggression hypothesis.

[1] This and the following six articles are revisions of the papers read at the Symposium on Effects of Frustration at the meeting of the Eastern Psychological Association at Atlantic City, April 5, 1940. A. H. Maslow's, the Chairman's, paper was not read at the meeting because of lack of time.

[2] J. Dollard, L. W. Doob, N. E. Miller, O. H. Mowrer, and R. R. Sears. Frustration and aggression. New Haven: Yale University Freer, 1939.

[3] Op. cit., pp. 8-9, 19, 58, 101-102.

[4] Ibid., pp. 32-38 also 27, 39-50 75-87, III, 166. In this later exposition a distinction is made not only between instigation to aggression and acts of aggression but also between conspicuous acts of overt aggression and inconspicuous acts of non-overt aggression. It is assumed that the former are more apt to be culturally inhibited by strong punishments than the latter

[5] Op. cit., p. 40.

[6] "The notions used here are similar to those employed by Professor Hull in describing trial-and-error learning. See Hull, C. L. Simple trial-and-error learning -- an empirical investigation. J. comp. Psychol., 1939, 27, 233-258.

[7] Both of the first two of these chief lines of investigation have been developed at length in Frustration and Aggression. No attempt was made there to elaborate upon either the third or the fourth. Thus that first effort does not purport to be a complete systematization of all principles within a single field, but rather, an exploratory attempt to apply a strictly limited number of principles to several different fields. Op. cit., pp. 18, 26.

[8] These problems are discussed in more detail by Dr. Sears in the next paper of this series, 'Non-aggressive responses to frustration.'