Emotion In Man And Animal: An Analysis of the Intuitive Processes of Recognition

D. O. Hebb[1]
Yerkes Laboratories of Primate Biology


There has been a marked and necessary scientific reaction against the mentalistic extravagances of earlier writing on animal behavior. There is little justification and less explanatory value in ascribing man's elaborate conscious processes to animals, and discussing emotions in such terms would be futile. At the same time, a rigid refusal to 'anthropomorphize' may have its scientific disadvantages. Obviously, the true objection to anthropomorphism is not to discovering a similarity of mechanism in human and animal behavior, but to inventing similarities that do not exist. A complete rejection of all concepts derived from experience with man would leave a vacuum in animal psychology, for the closer we come to man in the phylogenetic scale the more evident it is that some quite complex modes of human behavior occur in animals and one cannot help recognizing that the overt pattern is the same in each case. A discussion of jealousy in the earthworm is obvious nonsense, but not in primates.

The difficulty extends throughout the field of emotion and temperament. Among the chimpanzees of the Yerkes Laboratories there are marked, stable differences of behavior which simulate all the varieties of human temperament. There are no names for these differing modes of behavior except anthropomorphic ones. In spite of its mentalistic flavor and connotation of reference to conscious processes, the anthropomorphic terminology in this field may have another and more valuable significance as a classification of overt behavior.

A thoroughgoing attempt to avoid anthropomorphic description in the study of temperament was made over a two-year period at the Yerkes Laboratories. A formal experiment was set up to provide records of the actual behavior of the adult chimpanzees, and from these records to get an objective statement of the differences from animal to animal. All that resulted was an almost endless series of specific acts in which no order or meaning could be found. On the other hand, by the use of frankly anthropomorphic concepts of emotion and attitude one could quickly and easily describe the peculiarities of the individual animals, and with this information a newcomer to the staff could handle the animals as he could not safely otherwise. Whatever the anthropomorphic terminology may seem to imply about conscious states in the chimpanzee, it provides an intelligible and practical guide to behavior. The objective categorization therefore missed something in the behavior of the chimpanzee that the ill-defined categories of emotion and the like did not—some order, or relationship between isolated acts that is essential to comprehension of the behavior. The question then is raised whether one might start with the intuitive categories of emotion and discover by analysis what behavioral relationships they are based on. In this way one might objectify the categorization and make it suitable for the purposes of a scientific comparative psychology.

Such an analysis would have value in two other respects. It is worth while

( 89) to show what is involved in the recognition of human emotions, since it appears that failure to understand the real nature of the process is responsible for much of the confusion in the field of emotional theory. It is also worth while to throw some light on the intuitive process itself. Intuitive judgments are to be found in use as scientific data in fields as divergent as abnormal psychology (in the recognition of psychopathic syndromes) and cytoarchitectonics (where an adequate definition of the characteristics of any of Brodmann's areas, except 4 and 17, has never been achieved). Whenever an intuitive judgment can be objectified it will be improved, for in so doing one decreases the likelihood of an illicit confusion between interpretation and the fact interpreted.

It is proposed therefore (1) to analyze the way in which human beings recognize chimpanzee emotions; (2) to show that the recognition of human emotions is on a similar basis; and (3) to show that references to emotion, attitude and so on, whatever they may seem to imply about conscious entities that may not exist, even in man, have value as summary descriptions and predictions of behavior.


Several terms will be used in discussion which can easily be misunderstood, and it is necessary to define them explicitly.

Intuitive is used to refer to judgments which follow premises or steps of inference of which the judge is unaware and especially those which he cannot put into words. The word has undesirable connotations but there is no good substitute for it. The fact is evident that one frequently reaches a right conclusion without being able to state the evidence which really determines it. An equally evident fact is that intuitive judgments are often wrong, and as long as they re'main intuitive and unverbalized their flaws are not demonstrable.

Emotion is a term with more than one meaning. It is often used to refer to some distinctive mental state or conscious content, but sometimes it refers to more vaguely conceived states of excitation without any definite implications about consciousnesss. Now it happens, as we shall see, that competent students are agreed that there is no peculiar mental content referred to by the term emotion. In its most common use, therefore, the term has become meaningless. The second significance of the term seems more useful: at the time when an emotion is said to occur there is definitely a state of changed excitability or limen, with a selective effect on behavior: It used to be thought that the change of limen was due to a conscious event. This idea must now be rejected, but the fact of the changed responsiveness remains. It seems that 'emotion' could be used to refer to it. For this discussion, therefore, emotion explicitly does not mean any conscious process, nor any pattern of motor or glandular activity, but designates certain neurophysiological states, inferred from behavior, about which little is known except that by definition they predispose toward certain specific kinds of action.

Recognition of emotion then is the inference that one of these postulated states exists in the subject. It will be seen that there are two different conditions of the inference, which are mentioned here only to bring matters of terminology together. (1) In one condition, a deviation of behavior from a more usual pattern is directly classified as due to a particular emotion by the nature of the deviation itself. This is the recognition of 'primary' emotional behavior. (2) In the second condition, classification is determined by the fact

( 90) that the observer has recognized an accompaniment of the primary pattern: some act which would not in itself be classed as due to a particular emotion has become, by association, a sign of that emotion. The behavior in question is called 'associated' emotional behavior. The basis for this distinction will be shown later.


The data upon which this study is based are the accumulated diary records of chimpanzees in the Yerkes Laboratories; descriptions of individual animals, made in their own words by four members of the scientific and care-taking staffs of the Laboratories; and systematic records (made in detail by a system of abbreviations) of the behavior of the animals—particularly at the times when a member of the staff was ascribing some emotion to the animal he was observing.

At the time of this analysis the Yerkes colony included 30 adolescent and adult chimpanzees, all but two of which had been under daily observation for periods of 6 to 19 years. For each animal a diary record is kept in considerable detail. It records behavior from the date of birth or of arrival at the colony, and often the staff's interpretation of the behavior worded in terms of emotion or attitude. Many such entries are based on intimate knowledge of the subjects.

It will appear from the subsequent discussion that this familiarity with the animals is particularly important. Current ideas about the recognition of emotion are based on the laboratory studies of the last 30 years, which have used human subjects and judgments based on a short period of observation. These studies have led to the conclusion that an emotion cannot be accurately identified by another observer. However, it is possible that such a conclusion is the product of a particular experimental method. The chimpanzee data can give us information about judgments of emotion made in other circumstances, which would be hard to control experimentally with man but which in their time relationships are comparable to the recognition of human emotion in social situations.

For lack of space, the full details of recorded behavior upon which the following discussion rests cannot be given, except in a few instances which perhaps will illustrate all. The discussion really depends for conviction on the reader's social experience and upon his finding such an evident parallel between usage of emotional terms with chimpanzees and with human subjects that he will not need detailed illustrations in every instance. A much larger body of behavioral records and controlled observation has been analyzed than can be reported here.


It has already been said that there are two conditions in which chimpanzee emotions may be named. One is the classification of a particular form of behavior as due to some emotion when the pattern of behavior in itself is enough to determine the naming (though the necessary temporal duration of the period of observation, of the pattern of behavior, may be long). The second condition is when the naming depends on the observer's having learned that the behavior he sees is frequently accompanied by intrinsically diagnostic behavior of the first type. The justification for the dichotomy appears in the following section; here we are dealing with the classification of behavior which is observed as forming a characteristic pattern.

The first fact to establish is that the

( 91) naming of chimpanzee emotional behavior is rarely based on the immediate behavior only, or even on immediate behavior plus a knowledge of the stimulus. The first most comprehensive example is found in the use of three terms relating to avoidance: fear, nervousness and shyness. The context of the diary references, and observations of behavior to which these terms were applied, show that the terms in general have definitely distinguishable references. Fear is used when a specific object or region is avoided and implies that the response may be a strong one. Nervousness, quite apart from an alternative reference to a long-term characteristic of the animal, is used to refer to a temporary lowering of the avoidance limen, but not to the point of constant or extreme avoidance. Sometimes this state of affairs is detected when a sudden movement or noise produces startle more easily than at other times. Shyness specifically implies that the object avoided is man or another chimpanzee, and that more familiar persons are not as strongly avoided. Both nervousness and shyness imply that there is no adequate stimulus to avoidance (that is, the objects avoided are not such that one need expect them to cause avoidance); and that the avoidance is not indefinitely prolonged, for otherwise the observer speaks of an unexplained 'fear.'

Two points emerge here. One is the definite importance of the kind of stimulus in the choice of terms by which to designate the response: The same thing has 'been emphasized in the recognition of human emotions. But this is not to say that the stimulus alone determines the choice, as some writers have concluded. The response itself, as an avoidance, has a definite part to play. The second point is just as important. The choice of terms is affected by familiarity with the animal: his past behavior, the amount of his exposure to various stimuli, and his responses to them. Saying that an animal is being shy, for example, not only says that he is avoiding an unfamiliar person, but also that he does not avoid familiar ones. Otherwise his avoidance would be called nervousness, or fear of man. 'Shy' is a summary reference to a variability of behavior with stimuli which are of the same kind except with respect to the subject's past experience. This is a further qualification of the idea that the stimulus is what determines the choice of names for an emotional state. In some instances, knowledge of the stimulus is important only because the observer's interpretation takes in the animal's experience and behavior with similar stimuli. In short, the distinction between shyness, nervousness and fear is impossible without knowing something about the animal; and the term chosen, in these instances, is affected by knowledge of the stimulus ofthe response, and ofthe subject's experience and behavior in. the past.

Let me repeat that what is discussed here is the designation of transient emotional states, although some of the terms mentioned may also refer to habitual or lasting modes of behavior. Obviously one would have to be familiar with an animal—to observe his behavior repeatedly—before one could state that he always avoided strangers. But the point is that shyness might be observed on a single occasion only, and naming it as shyness would still demand a knowledge of the animal's lack of avoidance of persons he knows. The same term, 'shy' or 'nervous,' can refer to an immediate emotional condition, or to the fact that the animal is in such states often. The distinction is the same as that between a nervous person and one who is nervous as he gets up to make a speech. In chimpanzees, the designation even of the temporary emo-

( 92) -tional states requires knowledge of the animal's past behavior.

Another example is found in the naming of rage or anger. Bimba, one of the chimpanzees of the colony, is said to be friendly to man, but quick-tempered; while another chimpanzee, Pati, is said to hate man. Bimba's anger, and Pati's hate, are actually manifested in at'tacks which cannot be reliably distinguished from one another. But when Bimba is not attacking an observer, she behaves very differently from Pati. She is always responsive to man, and acts in a way which promotes contact and petting by the attendants (except at the times when she is said to be angry), while Pati has a long history of vicious attacks with few efforts to be friendly. It is in this difference that the real distinction of rage from hate must lie.

This was not evident at first. It seemed that there actually was a real difference in the way in which these two animals made their attacks. Due presumably to a kind of halo effect, Bimba's attacks seemed to occur only after some movement by the attendant that might appear like a threat, or like teasing; and it seemed that her attacks were always open, with the frank violence of anger, while Pati's were stealthy. But this was a fallacy of memory. When a record was kept of the circumstances of the attacks and their form, it appeared that Bimba, the friendly, made more attacks without cause than Pati ; and that fairly often she attacked deliberately after getting the observer within reach by an appearance of friendliness. (The greater frequency of Bimba's attacks, however, is not significant; because she is so friendly she gets many more chances.)

There was then no distinction between some of Bimba's attacks and some of Pati's. Yet there was perfect agreement by the staff in regarding all of Bimba's as due to anger and not to some fitful malice. The name given to a transient emotional state was deter'mined in part by the immediate behavior (that is, by the fact that it was, aggressive) but also by behavioral relationships over a period of years. It should be noted that the judgment was: an intuitive one, the observers not, realizing how much entered into their choice of terms. The person who does, the identifying seems to feel instead' that in some way (which he cannot specify) the immediate behavior in itself is really distinctive. In these circumstances, we can understand the skepticism of psychologists unfamiliar with the animals, when they are told that the same behavior is an expression now of one emotion, now of another.

The classification of Bimba's attacks as due to anger is also a particularly clear case of the way in which naming an emotion is occasionally quite independent of a knowledge of the stimulus. On repeated occasions there was no provocation at which one could even guess—that is, there was no stimulus evident. Yet one thinks of anger as precipitated by some event, real or imaginary. It. is therefore significant that one of the caretakers said that Bimba is 'quick to resent slights, or a fancied lack of attention.' Another said that she 'gets mad over nothing at all,' meaning that the causes were trivial and not such that they should lead to anger. A psychologist would hardly dare to make formal use of such language but it can be said that the scientific staff acquiesce in the description, and if the statements are interpreted as they would be socially, with human behavior, they are a reasonable summary of the observed facts. Obviously the classification of the response is achieved first, and a thoroughly hypothetical idea of the cause, or stimulus, is set up to fit in with it.

( 93)

If such constructs were made at random they would mean nothing. In or'der to show that the persons who diagnose Bimba's 'anger without apparent cause' are not making an entirely wanton use of language, several instances may be given in which the same persons are unable to interpret aggressive behavior, and refuse to name an underlying emotion.

Kambi is an animal who has been in the colony as long as Bimba, but the caretaking staff refuse in general to explain her occasional aggressions. Something necessary to a diagnosis is missing. There is a lack of the consistency in over-all behavior that leads to agreement in Bimba's or Pati's case. The staff simply regard Kambi as a 'screw-ball,' a moody psychopathic whose behavior is past accounting for.

Frank has generally been regarded as an exceptionally friendly and trust-worthy animal. He consequently has unusually good opportunities to injure the attendants. Recently he made a deliberate and vicious attack, for the first time, taking two or three minutes to get a good grip on an attendant's forearm outside the wire mesh of his cage, in an apparently friendly way, before suddenly beginning to bite and gnaw at the flesh pressed within reach of his teeth. The incident has been repeatedly discussed by the staff, and circumstances have been mentioned which could possibly account for the attack; but the point here is that no one has definitely settled on an explanation, or categorized the behavior by specifying an underlying emotional state.

Finally, an adult named Shorty has been in the colony for two years. He is frequently aggressive, frequently friendly, and always responsive to man. No one, however, is yet willing to give a name to his attitude. Bimba's superficially similar behavior fits the conception of a friendly animal who is easily annoyed; Shorty's does not. His aggressive behavior is too regular, and not even an imaginary provocation can be assigned. He becomes violent as an observer approaches and at these times may scratch or bite, but almost at once will change to a non-aggressive behavior if he is allowed to examine the observer's skin or clothing. He is tentatively thought to be 'friendly,' a term implying that in most circumstances one might enter his cage without fearing injury. But staff members are not at all sure of this with Shorty, as they are with some other animals, and no ore is willing to classify his aggressive behavior by saying definitely that it is only done to attract attention.

These are examples of refusals to anthropomorphize. In general, the staff will refuse to designate emotional states in the animals more often than not. The instances which have been chosen for discussion, in which an emotion is named, might have suggested that such diagnoses were made indiscriminately. By making it clear this is not so, I am trying to show that there are definite prerequisites in naming animal emotions, even by the psychologically naive. The names are applied only when familiarity with the animal reveals a long-term pattern of behavior with which the observer is already familiar in man, and some evidence of validity in the categorization is found in its practical value in predicting the outcome of a behavioral sequence.


The preceding discussion has maintained that there are temporal patterns of behavior that lead directly to the naming of emotions by human observers. These are referred to as 'primary' emotional behavior, meaning that from the observer's point of view the patterns are classifiable as soon as they

( 94) are clearly perceived. We come now to the way in which behavioral patterns which might mean nothing in themselves can determine a diagnosis of emotion, from their association with a primary emotional pattern.

In the chimpanzee, as in man, an emotional stimulus is not always fully effective but may produce a low level of excitation which is capable of summation, or of facilitating a later response. A single stimulation perhaps has a trivial effect. Prolonging or repeating it produces primary, characteristic emotional behavior, qualitatively changed from the apparently trivial response to first stimulation. The first stimulation may even have no evident effect, and yet have a facilitating action. These facts require the concept of subliminal emotional excitation, by precisely the same logic that allows us to deduce the various phases of excitability in a nerve cell.

Several examples can be given in which a stimulus, at first ineffective, gives rise to emotional behavior when it is repeated or prolonged. Cuba, caged next to Dita, begged loudly for some pieces of food which Dita had left uneaten. Dita gave her several pieces, then went away and sat in the sun; when Cuba persisted in her noisy begging, Dita at first sat watching her, then suddenly got up belligerently and beat violently on the wire separating them. In an experiment on avoidance a mounted snake was carried in the experimenter's hands up to the part of the enclosure where Shorty sat. He moved calmly away to another part of the enclosure, with no hint of excitement; but when the snake was again brought near him in his new position, he sprang up with hair erect and screaming, and hurled a large piece of timber at the experimenter with excellent aim. The same kind of summation effect is often seen in discrimination training, where a single failure may have no apparent effect but repeated failures lead to sulking, temper tantrums or destructive attacks on the apparatus. In such instances the existence of a subliminal excitation is purely a matter of inference.

More often, however, the emotional excitation is subliminal only in a restricted sense; it does not produce a primary emotional response, but does affect the animal's behavior. A single experimental frustration may for a given animal produce the response of scratching, nothing else. Scratching is not intrinsically a sign of emotion; and in this sense the emotional excitation is subliminal. When repeated frustrations produce temper tantrums in the same animal, the scratching becomes an indication of a changed responsiveness, of an increased probability of temper tantrums, and it is assumed that the animal's emotional state is not the same as before stimulation.

In another animal, the preliminary signs of temper tantrums might be restlessness, or a noise like moaning. Neither behavior is in itself diagnostic, but knowledge of the animal and of the stimulus would make it diagnostic, by its association with temper tantrums when frustration is repeated. Individual signs of the development of frustration responses are quite* varied. They include the animal's scratching himself, ducking his head, moving away from the apparatus quietly and becoming unresponsive to the experimenter, restlessness, moaning, whimpering, and erection of hair.

Within limits, associated behavior may have diagnostic value for a species as a whole. The erection of hair in a chimpanzee is in general a danger signal to anyone within his reach. Screaming on the other hand may be only a welcome for approaching food, and has not always the connotations it would have

( 95) with man. A sputtering noise with the lips is not derisive but usually a sign of friendliness and characteristically accompanies the grooming of another animal. Thus a knowledge of the species has considerable importance in the interpretation of the associated behavior of emotional excitation. Very often, however, any single item of behavior is not associated alone with a single emotion, but with several: erect hair occurs equally with rage and fear, and a particular posture by the female may precede copulation or be a signal of submission to another female. It is this fact, I think, that accounts for such concepts as 'nervous tension' and 'emotional excitement'—states in which it is evident, from associated signs, that there is emotional activity but with no definite indication of the specific form it will take. The significance of these phrases is apparently not invariable, but they seem to be much less frequently used when the observer has reason to expect a particular form of emotional behavior as the resultant of further stimulation. A more specific prediction of course usually depends on familiarity with individual characteristics. Important as a knowledge of the species is in interpreting the preliminary signs of emotion, the great variety of chimpanzee behavior appears to make a knowledge of individual animals more important.

The complexity of relationship which a neutral form of behavior may have to primary emotional behavior is illustrated by the unusual response pattern developed by the chimpanzee Soda in the course of an experiment with a number of avoidance-producing objects. The objects were brought close to the front of the animal's cage in a presentation box with a hinged cover which was opened for 30 seconds to expose the stimulus objects. Before and after exposure each animal was fed, with the intention of bringing him close to the stimulus so that definite avoidance or lack of avoidance could be observed. In the early trials Soda showed a strong avoidance of objects which also produced avoidance in other animals, and was less excited by others which had not so much avoidance value. As the experiment went on, however, a peculiar stereotyped behavior developed. Soda always has a good appetite, and came regularly to get the food. As she got it, and while the presentation box was being opened, she would move promptly but without apparent haste to a point about 10 feet away and sit watching the infant quarters next door, her gaze at ninety degrees from the direction of the stimulus object. Once this mode of response appeared, only those objects which were very exciting for other animals could produce a clear-cut avoidance in Soda.

How was her later behavior with the less stimulating objects to be interpreted? For the purposes of the experiment, 'avoidance' was defined as moving directly away from the stimulus in such a way as to indicate that the response was produced by that stimulus: in the promptness and precipitancy of response, or by signs such as vocalization and erection of hair on the appearance of the stimulus, or by the combination of moving away from the stimulus and keeping the gaze fixed on it. Soda showed none of these signs clearly, except with the most stimulating objects, and in itself her stereotyped behavior on any one trial might as well be called a sign of boredom as of fear.

However, there are these points to be considered. The peculiar pattern of behavior occurred only in an exciting, and often a 'fear-provoking' situation; and Soda is known to be easily scared. The peculiarity of the response lay in the recognizable repetition of an otherwise trivial series of movements in special

( 96) circumstances, and in the alternation of these movements with another recognizable pattern of definitive avoidance. On a few trials, there were indications of haste in the stereotyped response itself. The stereotyped behavior, then, was associated with avoidance and with a special situation which produced avoidance in a number of animals. Furthermore, it was significant that Soda's avoidance limen was lowered even during her apparently calm observation of the infants next door. On two occasions the hinged cover of the presentation box was dropped by accident, producing a sudden but not loud noise. Each time Soda jumped back in a way that clearly conformed to the definition of avoidance. Such noises did not produce as strong a response at other times. In brief, Soda showed a kind of behavior which was distinctive only in the regularity of its occurrence in a particular situation; this situation was one which at times produced a violent avoidance in Soda, and frequently in other chimpanzees; and finally, at the times when she exhibited the stereotyped behavior her limen for startle responses was lowered. Her stereotyped behavior was associated with avoidance and an avoidance-producing situation, and by that association can be interpreted as a sign of fear.

In a somewhat similar experiment two young males, Tom and Dick, could see the experimenter approaching before their own behavior could be observed. The single stimulus used was repeated on a number of different days. At first it produced definite avoidance in these two young animals; later, however, whenthe experimenter arrived at their cage they were always found sitting quite calmly near the back wall of the cage. The later behavior did not fit the definition of avoidance at all; but its consistency in a situation which at first produced avoidance in these two animals, and which continued to produce avoidance in a number of others, was well beyond chance. A single observation could not give a diagnosis of fear, but the relationships observed over a longer period can.

These are instances of an apparent suppression or obviating of open emotional behavior. The chimpanzee is a very excitable animal, but it is not uncommon to observe such an 'inhibition' of fear responses among adult animals. The apparent inhibition occurs also with rage, but to a less degree, and is rarely observed with other emotions. It can be plausibly considered to constitute a conflict of motivations, in which behavioral associations sometimes give a clue to the presence of 'unexpressed' emotional tendencies. Far more often than not one can merely guess at the nature of an unexpressed emotion, however. Only in favorable circumstances can a definite interpretation be made.


Let us now summarize the factors which have been shown to affect the choice of names for emotion in a chimpanzee, and ask how useful and acceptable such terms are for the purposes of comparative psychology. It is evident that the named emotion has a detectable objective significance, and that the apparent mental reference of the term should be disregarded (in view especially of the fact that emotional entities cannot be found even in human consciousness) ; but while these facts can be recognized and made use of, it would be wrong to leave the impression that nothing more is necessary to obtain a satisfactory categorization of emotional behavior.

Classifying emotional behavior '(that is, naming emotions) is based on a complex set of cues. There may be in-

( 97) -volved a knowledge of (1) the stimulus, (2) the subject's experience with this and related stimuli, (3) the response, (4) various aspects of the subject's other behavior and (5) behavioral characteristics of the species. Every judgment is not affected by all of these things, and the most important are (3) and (4), or the relation of the response to behavior at other times.

The primary emotional pattern is essentially a deviation from a base line which may differ with individuals. It is upon this that intuitive judgment puts weight rather than on the details of the act itself. Also, the intuitive judgment treats as equivalent all acts, however variable, which tend to have the same end result, implicitly postulating a constant central process with a variable manifestation.

Such a conception is not intrinsically improbable. Lashley (14) has shown that a highly variable motor pattern may be controlled by a single physiologically coherent mechanism, and his reasoning is directly applicable to emotional responses. The absence of a constant motor pattern in rage or fear behavior therefore is not evidence against the existence of psychological entities 'rage' and 'fear.' On the other hand, there is not sufficient evidence that our present classification of these entities as discrete, independently functioning processes is necessarily satisfactory. There is every reason to believe that there are innately established emotional entities, but the existing evidence does not take us much farther. It is one thing to show that a variable series of acts is under unitary control, particularly when one can demonstrate, as in sexual behavior, a correlation with a definable physiological condition; but it is quite different to show from behavioral evidence alone that a series A and a series B, occurring at different times or in different subjects, are the product of a single mechanism and that this mechanism is discrete from others.

The various expressions of rage and fear, for example, might be so arranged as to represent a continuum, and only traditional thought can clearly dichotomize them. In other words, we have at present no good criterion for emotional classification. Certainly many of the distinctions of the traditional classification must be psychologically meaningless, referring, as they evidently do, not to a distinction of mechanism but to the circumstances in which the mechanism is activated or the degree of activation. One might guess that there are at least three different modes of emotional behavior—aggression, flight, and nonviolent adience—and perhaps more, but this must be only a guess until the actual physiological control of the various forms of behavior is understood.

The contribution of this paper is in showing just what objective facts underlie the traditional categorization, presumably facilitating the discovery of its elements of value, and in revealing the elaborate conceptual framework involved. It has been seen, for example, that the distinction of Bimba's anger from Pati's hate is not independent of the idea that Bimba is a friendly animal, and requires the conception of imagined causes of anger. It has also been seen that there is involved the postulate of subliminal emotional excitations; and frequently the variability of manifestation of an emotion is explicitly attributed to a conflict between emotional mechanisms. Such concepts may be right or wrong—the concept of a subliminal activation, for example, seems entirely justified—but there is no definite basis at present for separating bad from good. The traditional classification evidently implies an elaborate

( 98) theory. If there is some truth in it, as its practical value in predicting behavior suggests, the present analysis is a first step toward discovering what that truth is.

The unfortunate fact is that one's observations of emotional behavior are likely to be intuitive and a function of preconceived ideas, but following the direction of the present discussion it should be possible at least to retrace one's steps in the intuitive judgment and discover just what the facts are, as distinct from the concealed interpretation. In trying to isolate the cues which determine one's recognition of an emotion, the question to ask is not, "What is the distinctive feature of this behavior," but "What is the nature of the deviation from the animal's behavior at other times?" Then, "Is the deviation in itself enough to identify a particular emotion, or is it diagnostic only for the particular animal or species concerned?"This latter question makes the distinction between primary and associated emotional behavior. Finally, one may ask whether the same emotion would have been recognized if the stimulus had been different. With such an approach one can often determine with a good deal of confidence the cues which have affected one's judgment, and thus be in a better position to evaluate its significance. Before the thesis of this paper was clearly formulated, there were frequent occasions on which I could only say that a certain chimpanzee was afraid although he made no definable fear response—as in the case of Soda, mentioned earlier. When the distinction between primary and associated emotional behavior is made, and the course of events over an extended period considered, the evidence can be reported in a form which others can evaluate and not merely as an unanalyzable personal opinion.


Here and in the following section I shall try to make the hypothesis plausible that naming man's emotions is much the same process as naming the chimpanzee's, and that the hypothetical entities designated are inferred from human behavior and not from events in consciousness.

If this should be true, it would at once remove a number of contradictions which plague the student of emotion and emotional recognition. The current opinion that human emotion is recognizable only from the nature of the stimulus conflicts with common experience. Also, there is a frequent objection to naming animals' emotions, as something anthropomorphic and unjustified. The attitude is often met in psychologists who do not object in the same way to naming man's emotions. Yet if we accept authoritative opinion, which holds that the name of an emotion is determined either by the stimulus alone or by stimulus and strength of response, naming an animal's emotion ought to be precisely as valid as naming a man's. The original source of such confusion is to be found, I believe, in the general assumption that an emotion is recognized, if at all, from momentary observation; and in the failure to distinguish intrinsically recognizable patterns, such as a temper tantrum, from associated signs such as facial expression which may be diagnostic only when the judge is familiar with the subject.

The persistent idea that the recognition of emotions can be fully tested by allowing the judge to see the subject's behavior for a moment or so, even by means of a picture showing the subject as he was at a single instant, has dominated every experimental investigation of the problem. The results show a low level of agreement among

( 99) judges as to what emotion is expressed. Together with a failure to find characteristic visceral patterns for the various emotions, this has led to the conclusion that the subject's actual response has little to do with the name given to his emotion. But the conclusion obviously deals only with spatial or nontemporal patterns of response. All that has been shown is that there is no characteristic pattern of coexistent motor activity for

a particular emotional state. A further limitation on the conclusion is imposed by the nature of the conditions in which human emotions can be studied experimentally. Uninhibited rage cannot safely be provoked in adults, nor uninhibited fear. It is also hard to obtain emotional behavior that is not suppressed and distorted by social factors. This is not true with infant subjects, but emotional mechanisms at birth are not differentiated as they are later, as is shown by the late appearance of distinctive rage (4) or by the changes of fear susceptibilities with maturation (9, 10, 16, 17, 19). Consequently a failure to recognize infant emotions must be inconclusive. Thus a lack of recognition of emotion has been established only for momentary response patterns, without regard for temporal sequences of integrated action, and has not really dealt with the full, uninhibited expression of human emotion as it is known to occur socially and as one can study it in chimpanzees.

The idea that human emotions are verbal fictions, and that the name given to one is mainly determined by the stimulus situation, seems to have been generally accepted as a result of the failure to find distinctive muscular patterns of response. Sherman's work with infants (18) put emphasis on the stimulus as a factor in 'recognizing' emotions. Landis had drawn attention to this factor in an early paper (11), and more recently has stated his position as follows: "In every case it seems that the name [given to an emotion] is one which is assigned to some particular configuration or type of situation" (12, p. 411). Dunlap (6) shares this opinion; emotions are nothing more than 'estimation of the situations in which they arise,' either by an observer or by the subject himself. Duffy (5) includes a reference to the strength (but not the form) of response; stimulus and strength of response are the only factors affecting the identification of emotion. Harlow and Stagner (8, p. 191) adopt a similar position; rage is excitement plus the perception of a 'situation calling for attack,' fear the same thing in a situation calling for retreat.

The consensus, then, is that in any recognition of emotion the only variable, or the most important one, is the nature' of the stimulus. This is simply not in accord with common social experience. Human emotions are identified socially without perception of the cause. A wife knows that her husband is annoyed but not what he is annoyed about; a child is seen to be frightened when the observer was not expecting it, although he could perceive the stimulus (the thing he later calls the cause of fear) in advance. If emotions are recognized by situation plus excitement, no one could distinguish fear from rage when .a child is punished—the situation is said to provoke both emotions. This difficulty for current theory comes up whenever different emotions are identified in the same situation; when, for example, bawdy pictures produce sexual excitement in one person and disgust in another, or when a practical joke causes fear first and anger when it is repeated. The difficulty cannot be removed by postulating errors in identification. If emotions are only 'estimation of the situations in which they arise,' any one observer would be consistent in attributing one emotion to all

( 100) subjects in what he considers the same situation. There are a number of such facts which do not fit the conclusion that the apparent stimulus is the main determinant of recognition. Yet no other conclusion seems possible if man names emotions by a momentary observation without regard to temporal sequences of response: if the classical laboratory experiments provide an adequate test of the recognition of emotion, we are involved in a set of hopeless contradictions.

If, however, it is assumed that human emotions are recognized as the chimpanzee's are, the contradictions disappear. The failure of recognition in the laboratory tests would be accounted for, since human emotions could be named only (1) by perception of a deviation from the base line of ordinary individual behavior, or (2) by the fact that the observer has associated the subject's neutral acts with the intrinsically nameable deviations. Either condition requires an extended period of observation of the subject. The stimulus would still affect the naming: in experimental conditions the observer, lacking all other cues to the nature of the subject's disturbance, would fall back on his knowledge of the species Homo and name the emotion such a stimulus would be most likely to produce. Socially, with extended observation of the subject, his judgments would be more confident, less dependent on a perception of the stimulus, and more likely to agree with those of other judges and of the subject himself (although the agreement between observer and subject has certain limitations which are discussed below) . Finally, the phenomenon of a named emotion without apparent cause or with an unusual cause would become comprehensible, since the nature of the stimulus would play only a subsidiary role in recognition.

Thus the assumption that man's emotion is recognized in the same way as the chimpanzee's has theoretical advantages. The main objection to it will be found in the common idea that emotions are directly known in consciousness, only indirectly from behavior. This objection must be considered in detail.


It is proposed here that the hypothesis of a primary objective recognition can clarify even the subject's recognition of his own emotion. The hypothesis is to the effect that, with human as with chimpanzee subjects, naming emotions is based essentially on' observed deviations of behavior. Whenever the subject detects the direction of his own subsequent behavior before an observer does, he can name his emotion quicker and more accurately. He may also have access to 'associated signs' in the form of imagery and so on which the other observer does not have, and yet these signs may originally have acquired meaning from their relation to behavior. The hypothesis therefore does not deny the obvious fact that in some circumstances the subject's recognition of his emotion is superior (quicker, of better predictive value) to recognition by another observer. It does deny that subjective naming is always superior and that it is in essence opposed to objective naming.

The points in support of this view can be listed as follows:

(1) In some circumstances it is accepted socially that the observer's classification of emotions is superior to the subject's.

(2) Authoritative opinion unanimously denies that there is any simple event in consciousness which corresponds to a named emotion, and is clearly in the dark as to the actual

(101) cues used by the subject in naming his emotion.

(3) There is convincing reason to believe that the meaning of emotional terms must have been learned on the basis of distinctions of behavior, hence that these terms have ultimately a behavioral reference.

(4) When the distinction between subliminal and openly expressed emotional excitations is made, it is significant that the superiority of subjective recognition applies only to the first kind of excitation; hence the superiority may lie only in an ability to predict the nature of eventual behavior before observers can, and does not imply that the subject has different criteria of emotion per se.

Let us consider these points in order.

(1) Socially, the subject's classification of his emotional excitations is not always trusted. Preliminary signs of sexual excitation in the socially naive (though not necessarily inexperienced) subject may be better evaluated by a sophisticated observer. Jealousy is more apt to be classified as such by others than by the one 'experiencing' it, for to him it is likely to appear as indignation, disapproval on moral grounds, simple annoyance or the like. Subject and observers in a social situation often agree in diagnosis of emotion; but when they disagree do we always regard the subject's report as superior, as a better guide to his future behavior? The answer may be yes, when the observers' reports are variable and disagree; but when observers agree among themselves and the subject disagrees it is often accepted that the observers' report is more trustworthy, even when the subject is not lying. It certainly is not assumed socially that a subjective evaluation is always more reliable than one based on the observation of behavior.

(2) Writers on the topic of emotion insist that there is no characteristic conscious event that can be called an emotion, nor any simple conscious index of an emotion. Dashiell (3), Duffy (5), Dunlap (6), Harlow and Stagner (8), Landis (12), Lashley (15), and Young (21) all deny explicitly or by inference

that emotions have any intrinsic conscious content that can identify them. Furthermore, Dashiell, Duffy, Dunlap, Harlow and Stagner, and Landis believe that the main determinant of recognition, even when the subject is naming his own emotion, is the nature of the stimulus. This theory is refuted by the not infrequent reports of fear without cause, subjectively evaluated (cf., for example, Cantril and Hunt, 2), but it means that those who accept the theory see no essential difference in the subjective and the objective recognition of emotion. No other theory of subjective recognition is at present maintained by psychologists. Landis and Hunt (13, p. 484) conclude only that "emotional experience . . . is a highly variable state" and "often partakes of the complicated nature of a judgment." The statement fits closely with the hypothesis of the present paper; taken with Cantril and Hunt's (2) work, in which the injection of adrenalin in normal subjects produced widely varying subjective reports, it indicates that the criteria of emotion subjectively recognized are not constant and possibly casts doubt on the reliability of subjective report. The 'complicated judgment' and variability of report become comprehensible if the hypothesis is accepted that the primary recognition of emotion is in the perception of actual overt behavior; apart from this, it can be seen what doubtful ground we venture on when subjective recognition is made the sine qua non of an emotion.

(3) Lashley[2] has pointed out that as

( 102) children we can only have learned what names to apply to our own emotions if the emotions can be recognized objectively. Either adults must recognize the child's emotion from his behavior, and supply him with a name for it; or he must first connect the name with something seen in others and then transfer this to his own behavior. In either case accuracy in naming an emotion subjectively is in the first place dependent on a good objective recognition.

Now it is conceivable that the adult is right in naming the child's emotion only 75 per cent of the time, and that the child attaches the name to the conscious process which most frequently coincides with the word spoken by the adult. Because of the adult's inaccuracy in objective recognition, the spoken name would not coincide always with the 'true' conscious processes of emotion, but would do so oftener than with any other mental event. Once this selective association is formed, the child himself would tend always to use the right name for his emotion. Thereafter, his subjective recognition would be more accurate than the original objective recognition by the adult. On this assumption, it does not follow, merely from the way in which the emotional names must be learned by the child, that subjective accuracy in naming is no higher than objective accuracy.

But other facts make nonsense of such an assumption. There is first the extreme difficulty of sophisticated adults in finding any mental concomitant of emotional behavior—shown in the denial that there is one, in the unreliability of subjective report, already stressed, and in the conclusion of Landis and Hunt that the conscious aspect of emotion is complex, elusive and variable. Think now of the freedom with which children express their emotions at the stage at which they would be learning the meaning of the words 'fear,' 'love' and so on. Would a child first associate the word 'fear' with his own vague, complex and variable conscious events or with his relatively clear-cut behavior? The first meaning of fear to a child must be physical avoidance; of anger, biting and striking; of love, close bodily contact and kissing; and so on. The development of a more subjective and sophisticated meaning will be discussed in a moment, but the facts already dealt with make it necessary to assume that the fundamental meaning of emotional terms must be in their reference to patterns of actual overt behavior.

(4) Finally, it seems that an emotion is better named by the subject than by an observer only when its expression is suppressed and distorted. When fear, anger, jealousy, sexual excitement, affection or disgust is openly expressed, without regard to social standards, the diagnosis is evident to any observer who has a little knowledge of the person displaying the emotion. If it is supposed that man's emotions are named as the chimpanzee's are, the diagnosis of a suppressed emotion, however, would require much more familiarity with the subject's habitual modes of behavior; and the one who is most familiar with his habits is the subject himself. If there were no other way of naming one's own emotion but from overt behavior, the subject would still tend to be better at diagnosing his emotion than another observer. Consequently, the fact that self-diagnosis in certain circumstances is apt to be quicker and more accurately predictive of future behavior does not in any way mean that the recognition of emotion is primarily subjective.

However, common experience makes it quite clear that more is involved in the recognition of one's own emotion. Psychologists agree that there is no special conscious quale of an emotion; but

( 103) this does not imply that there is no awareness of a tendency toward a certain kind of action, any more than the experimental failure to find specific patterns of muscular action is a denial that a frightened person may run away, or that an angry one may strike. We have then the fact that the subject may be aware of a tendency (desire, intention) to act in a particular way. Obviously he might be aware of such a tendency, when the emotion is socially suppressed, before another observer could be. How such awarenesses are to be explained, psychologically, I do not pretend to know; but it is clear that they exist. Now for the present argument, it does not matter whether one assumes that the awareness of a tendency to run away is perceptually equivalent to running away, or whether one assumes only that the awareness is bound to be associated with the act of running. On all those occasions on which flight is first suppressed and then actually follows, the perceived tendency would be closely followed by the act, and the two would be associated.

Thus the awareness of a tendency to act can be treated merely as an 'associated sign' of emotion, though it may be more. There must be other associated signs which the subject (as he develops from the infantile stage at which anger means only a sudden overt attack) comes to notice in himself as concomitants either of gross behavioral deviations or tendencies thereto. As he grows he is obliged more and more to suppress the frank expression of emotion; and the tendency to action, and other associated events, subjectively perceived, must become more and more closely related to the essential meaning of emotional terms. It is not likely that the original meaning, and the relation to overt behavior, is ever lost, but the adult in this culture who is able to keep out of jail must become used to recognizing his own emotions from perceived tendencies instead of from uninhibited action. Once the overt emotional behavior has come under the full strength of social control, an emotional state is practically always one of conflict between a tendency to action and the processes of repression. In English-speaking countries at least there is no emotion whose expression is unaffected by social standards. The adult subject might easily come to think of emotion as typically being what has been treated here as an incompletely expressed and distorted emotion, as a state of internal conflict between opposed tendencies. Obviously the one observer who knows these conflicting tendencies best is the subject himself, who would also be able to utilize associated signs of imagery and bodily sensation that would be available to no one else. In the civilized expression of emotion, therefore, the person who experiences the emotion would often be in the best position to diagnose and name it even though the assumption of this paper is correct, that the real meaning of emotional terms is derived from overt behavior.

Blatz (1) observed that none of his subjects reported fear in an unexpected fall who had not made abortive movements toward escape at the moment of falling. This directly supports the idea that the subjective recognition of emotion involves the perception of a tendency to action, more or less suppressed. The reports of cold emotion in Cantril and Hunt's experiment (2) show that there are also subjective associated signs of emotion. The reports distinguish clearly between an emotion and its subjective accompaniments. The reports of genuine emotion in the same experiment cannot be evaluated from the data given. The subject's conviction that he actually experienced fear

( 104) means either that the injection of adrenalin actually facilitated avoidance responses (experienced as a desire to escape from something, without perceiving the something), or that it produced associated signs which were so strong and compelling that the subject found them a convincing evidence of fear. If one were testing the validity of such reports by the subject, the essential question would be whether the adrenalin actually facilitates avoidance behavior. For the hypothesis of this paper, the fear would be genuine even if actual avoidance were not produced, since emotion is regarded as a state in which a particular deviation of behavior is facilitated, whether liminally or subliminally. The existence of such states could be reliably determined only by a method of summation, and not by subjective report.

To sum up the discussion of the diagnosis of emotion by the subject himself: existing psychological opinion as to the determinants of emotional recognition, and the common assumption that subjective recognition is always more direct and trustworthy than the objective, are both contradicted by fact. All the facts, however, can be comprehended by the assumption that the original reference of an emotional term is to a deviation of observable behavior, and that the growing child learns more slowly, with the possibility of continuing error, how to diagnose his own incompletely expressed emotions. The self-diagnosis is regarded as depending on an ability to detect one's own tendencies toward action before it occurs, and on the association of certain physiological and psychological changes in oneself with overt emotional behavior. In this way the adult could often evaluate his own emotion more quickly and accurately than another observer, but the superiority of subjective recognition would not hold in all circumstances.


The main conclusions of this paper can be summarized as follows:

(1) The recognition of a full, characteristic expression of emotion is the classification of a deviation of behavior from an habitual base line. It is not a discrimination of the momentary behavior itself, but of the direction of the deviation, so that both present and past behavior affect the observer's judgment.

(2) The recognition of emotion otherwise is a discrimination of a state of changed responsiveness detected from 'associated signs': acts which would not have a definite emotional significance in themselves, but which have been observed as the accompaniment of more openly emotional behavior.

(3) The emotions thus detected are inferred special states which facilitate or actually produce the primary emotional behavior of (1), although little is known of these states or of a satisfactory classification for them.

(4) The recognition of emotion in man and animal is not fundamentally different. Conceptually, the states of changed responsiveness discriminated are the same in both cases and have the same relationship to overt behavior.

(5) Even when the subject himself does the recognizing, the ultimate criteria of the various emotions are found in distinctions of overt behavior. In subjective recognition cues are available which a second observer cannot utilize, but these cues (of imagery and so on) must be essentially of the nature of associated signs which the growing child learns gradually to interpret after first learning the meaning of emotional terms in relation to actual overt behavior.

(6) Finally, it may be concluded that the failure to obtain a reliable recognition of emotions in the laboratory experiments of the last thirty years

( 105) was the result of a particular experimental procedure (the use of too short a period of observation), and does not show that emotion cannot be recognized socially. Also, the conclusion of some writers that emotions are nothing but figments of the imagination stems directly from the apparent unreliability of recognition; if recognition is reliable socially and does not depend mainly on knowledge of the stimulus, this conclusion is unjustified.

A strong argument in favor of the hypothesis presented is that it removes the contradictions inherent in current discussions of emotion and emotional recognition. A stronger argument perhaps is found in the original datum of this paper: the fact that psychologically sophisticated and unsophisticated alike experience an overwhelming tendency to name the chimpanzee's emotions, with a human terminology. Yet the details of emotional expression are quite different in the two species. Recognition of emotion from the chimpanzee's facial expression is even worse than with man's (Foley, 7), and most of the incidental signs of chimpanzee excitement have a significance that must be learned by prolonged observation. Yerkes (20) makes it clear that close familiarity with the chimpanzee is necessary for the discrimination of many emotional states, and that the individual components of emotional behavior do not have the same significance in man and chimpanzee. It is only in the higher-order units that the identity of behavior in the two species becomes evident. Consequently the tendency of any observer to recognize and name chimpanzee emotions is not due to a mere superficial similarity. From my own observations I should say that the recognition of identity really begins only after several weeks of observation, and that the strength of one's conviction, that the identity is real, increases thereafter for months.

Such a fact might be evidence only of suggestibility, but this cannot be the whole explanation. The behavior of the dog is complex enough, but lifelong familiarity with this animal does not produce the degree of 'anthropomorphism' in psychologically sophisticated persons that six months of exposure to the chimpanzee will produce. The tendency to identify human emotional patterns in the chimpanzee is the more convincing since, as I have shown, the identification is not indiscriminate. The naming of attitude or emotion is much less frequent than the reader might suppose, and occurs only when the behavior in its long-term significance and intercorrelations is something already nameable in man. When the elements of behavior have a relation that is not known in man, no name is given to the underlying emotion. The naming, therefore, is not an irresponsible form of animism, but a classification of deviations of behavior. The facts of behavior must constitute the ultimate reference of emotional terms.


1. BLATZ, W. E. The cardiac, respiratory, and electrical phenomena involved in the emotion of fear. J. exp. Psychol., 1925, 8, 109-132.

2. CANTIL, H., & HUNT, W. A. Emotional effects produced by the injection of adrenalin. Amer. J. Psychol., 1932, 44, 300-307.

3. DASHIELL, J. F. Are there any native emotions? PSYCHOL. REV., 1928, 35, 319-327.

4. DENNIS, W. Infant reaction to restraint: an evaluation of Watson's theory. Trans. N. Y. Acad. Sci., 1940, Ser. 2, 2, No. 8, 202-218.

5. DUFFY, E. Emotion: an example of the need for reorientation in psychology. PSYCHOL. REV., 1934, 41, 184-198.

6. DUNLAP, K. Are emotions teleological constructs? Amer. J. Psychol., 1932, 44, 572-576.


7. FOLEY, J.P. judgment of facial expression of emotion in the chimpanzee. J. soc. Psychol., 1935, 6, 31-67.

8. HARLOW, H. F., & STAGNER, R. Psychology of feelings and emotions. II. Theory of emotions. PSYCHOL. REV., 1933, 40, 184-195.

9. JERSILD, A. T., & HOLMES, F. B. Children's fears. New York: Teachers College Bureau of Publications, 1935.

10. Jonas, H. E., & JONES, M. C. A study of fear. Childhood Educ., 1928, 5, 136-143.

11. LANDIS, C. Studies of emotional reactions: II. General behavior and facial expression. J. comp. Psychol., 1924, 4, 447-501.

12. —. Emotion. In Psychology (E. G. Boring, H. S. Langfeld, H. P. Weld, Eds.). New York: Wiley, 1935, Chap. 16.

13. —, & HUNT, W. A. Adrenalin and emotion. PSYCHOL. REV., 1932, 39, 467-485.

14 LASHLEY, K. S. Experimental analysis of instinctive behavior. PSYCHOL. REV., 1938, 45, 445-471.

15. —. The thalamus and emotion. PSYCHOL. REV., 1938, 45, 42-61.

16. MCCULLOCH, T. L., & HASLERUD, C. M. Affective responses of an infant chimpanzee raised in isolation from its kind. J. comp. Psychol., 1939, 28, 437-445.

17. RICHARDSON, W. B. Reaction toward snakes as shown by the wood rat, Neotoma albigula. J. comp. Psychol., 1942, 34, 1-10.

18. SHERMAN, M. The differentiation of emotional responses in infants. I. judgments of emotional responses from motion picture views and from actual observation. J. comp. Psychol., 1927, 7, 265-284.

19. VALENTINE, C. W. The innate bases of fear. J. genet. Psychol., 1930, 37, 394-419.

20. YERKES, R. M. Chimpanzees: a laboratory colony. New Haven: Yale Univ. Press, 1943.

21. YOUNG, P. T. Emotion in man and animal. New York: Wiley, 1943.


  1. Research Fellow, Harvard.
  2. Personal communication.

Valid HTML 4.01 Strict Valid CSS2