The Measurement of Value
Louis L. Thurstone
In this paper I shall try to summarize briefly the attempts of several investigators to extend the concepts of measurement to the subjective domain. While this work is admittedly crude and exploratory, the results do look promising, so that this field should be challenging for further study. Here we shall have time only for brief statements of the fundamental ideas without details of theory or experimental procedure. Our purpose here is only to sketch the nature of this field of research.
When we propose to measure human values, colleagues in the humanities may shudder at the very idea. When I wrote a paper entitled "Attitudes Can Be Measured," some of my colleagues did shudder. They were sure that social attitudes contain some essence that could not be identified and measured. They were sure that, in making the attempt, we would measure only the trivial.
Human values are essentially subjective. They can certainly not be adequately represented by physical objects. Their intensities or magnitudes cannot be represented by physical measurement. At the very start we are faced with the problem of establishing a subjective metric. This is the central theme in modern psychophysics in its many applications to the measurement of social values, moral values, and aesthetic values. Exactly the same problem reappears in the measurement of utility in economics.
In order to establish a subjective metric, we must have a subjective unit of measurement. Before we can accept a subjective metric, it must satisfy the logical requirements of measurement as distinguished from rank order. These objectives have been approximated in the equation of comparative judgment and its variants.
Before proceeding to discuss the many applications of the subjective metric, we shall review briefly the principal psychophysical concepts by which a subjective metric can be established.
Let us consider these concepts in terms of a rather simple example, namely, the judgment of excellence of handwriting. When we look at several specimens of handwriting, it is fairly easy to select some that are considered to be excellent and others that are judged to be poor. In general, there is good agreement in such judgments. If we were asked to equate our judgments of excellence in a handwriting specimen to some physical measurements on the script, we would find it difficult. One of the main requirements of a truly subjective metric is that it shall be entirely independent of all physical measurement. In freeing ourselves completely from physical measurement, we are also free to experiment with aesthetic objects and with many other types of stimuli to which there does not correspond any known physical measurement.
If we present a single handwriting specimen to a subject with the request that he tell us how good he thinks it is, then he must try to convey the degree of excellence in terms of words. It is well known that people vary tremendously in their use of superlatives in appraisals of experience, and, consequently, it is preferable to avoid such
( 48) a direct procedure. Next we proceed to pairs of stimuli. We can ask the subject to judge which is the better of two specimens. In so doing, the subject gives his comparative judgment for each pair, and he is not asked to give any verbal description of excellence.
The degree of excellence of a handwriting specimen is experienced by the subject in terms of some subjective process or quale. Since nothing is known about the neurological correlates of judgments of excellence of handwriting, we shall dodge all such terminology by merely referring to the discriminal processes by which the subject does, in fact, discriminate between the different specimens. These processes may be assumed to be physical or truly subjective according to the preferences of the investigator. His preference on this point has nothing to do with the subsequent development of the law of comparative judgment.
When the subject makes a judgment that one specimen seems to him to be better than another specimen, we postulate discriminal processes which differ in some manner in terms of which the percipient does make the discrimination. The more excellent specimen has some quale which differs from that of the poorer specimen. Imagine that the discriminal processes which correspond to different values are arranged in a spectrum from those discriminal processes in terms of which the percipient experiences the good specimens to the other end of the spectrum with discriminal processes in terms of which he experiences what he calls the poorer specimens.
Consider next the phenomena of dispersion. If one subject were to examine the same specimen in comparative situations on a large number of occasions it is not to be expected that he would always experience a particular specimen with the same discriminal process. It can be assumed that the same specimen will be experienced in terms of discriminal processes in the same general region of the subjective continuum that has been postulated. So far we have no metric.
At this point we recall one of the fundamental restrictions on the problem of establishing a subjective metric. The discriminal processes must be assumed to be of such a character that they do not necessarily have intensities or magnitudes which can be in any sense measured. This is an old problem that was discussed many years ago in psychophysical theory. For theoretical considerations, imagine that the discriminal processes could actually be identified on each occasion when the subject makes a comparative judgment. The repeated observations of the same specimen can be assumed to produce an error variation from one occasion to the next. If we consider the relative frequencies of these discriminal processes as responses to the same stimulus, then we can postulate a Gaussian error distribution for the responses to the repeated observations of the same stimulus. Let us now assume that the spectrum of discriminal processes is stretched or contracted in different parts in such a way that the frequency distribution of these processes is Gaussian in terms of any given stimulus. Now we have a metric, but it is so far an entirely arbitrary metric. Imagine, at least in theory, that the same procedure can be repeated for many different stimuli which cover the whole range of discriminal processes in terms of which degree of excellence is experienced. It is now a question of experimental fact whether the metrics determined for the separate stimuli will be the same when all of the stimuli are considered together. It has been found in many experiments that such is the case.
If we represent in the same model the
( 49) comparative judgment of two stimuli in which the subject says for each presentation which of the pair is the better, then we can observe the proportion of attempts in which the subject judges specimen j to be better than specimen k. If we have a whole table of such proportions, it is possible to infer the spatial separations of the different distributions of discriminal processes. Each stimulus is then assumed to project a Gaussian distribution on the subjective continuum with a mean and a discriminal dispersion. An ambiguous stimulus will project a wider dispersion on the subjective continuum than a sharply defined or relatively unambiguous stimulus. Each stimulus will then be defined in the subjective continuum by its mean position, which is called a scale value, and by the standard deviation of its dispersion of discriminal processes. Each stimulus is then defined by two parameters in the subjective continuum.
Before we can put numbers into these parameters, we must define an arbitrary origin which may be taken as the mean value that one of the stimuli projects on the continuum. As a unit of measurement we may choose arbitrarily the standard deviation of the dispersion which that stimulus projects on the subjective continuum. When that has been done, similar numerical values can be assigned to all of the other specimens that have entered into the comparative judgments. Further, we can test for the internal consistency of this theoretical model.
It should be carefully noted that we have not assumed that the discriminal processes have magnitudes of any kind. They have been dealt with merely as subjective qualia, and we have assumed only that in principle their relative frequency of association with any given stimulus can be ascertained. While this cannot be done directly, these frequencies can be inferred indirectly from the observed comparative data. It should also be noted that we have not postulated the existence of any physical measures of any kind for the stimuli that have entered into the comparative judgments.
With this formulation of the law of comparative judgment, we are free to proceed with comparative studies of all kinds of stimuli which have no physical measure whatever. Hence we can turn to a wide array of interesting psychological problems involving value judgments. The freedom from any postulated physical measurement is the key that makes studies of this kind possible.
The method of comparative judgment turns out to be a rather general experimental procedure, and the well-known constant method in psycho-physics is a special case in which one of the stimuli is arbitrarily taken as the standard which is compared with all of the other stimuli. Classical psycho-physics was concerned with the more restricted problem of limen determinations.
We turn next to a brief review of some of the classical psychophysical methods because some of them have application in modern problems which transcend the determination of limens. In the method of equal-appearing intervals, the subject is asked to sort a large number of stimuli into a specified number of successive categories, say, six or eight or ten. He is instructed to sort them in such a way that the intervals represented by the categories seem to him to be equal. This method is useful for rough survey purposes, but it can be shown that, even when the subject attempts to do this, he actually does not succeed in making the intervals subjectively equal. The method is, however, useful for coarse scaling, such as the construction of attitude scales. The old method of equal-appearing intervals has been modified into what we call the method of successive intervals, in which
( 50) the intervals are defined by descriptive phrases or by sample specimens. This method has been found to be very useful in various types of surveys to be discussed.
One of the old psychophysical methods was to ask the subject to sort a number of specimens into rank order. It has been found that rank orders can be analyzed in such a way as to obtain data approximately equivalent to those of the method of paired comparison. The method of successive intervals can even be analyzed as a variant of the method of single stimuli.
Since Weber's law and Fechner's law have figured so prominently in the history of psychophysics, we shall make a few comments about these two laws in relation to the modern setting. These two laws are frequently referred to as the Weber-Fechner law with the implication that they are the same law, but that is an error. It is possible to set up experiments with rather simple stimuli in which one of these laws will be verified when the other one is not verified. It would be useful to set up such experiments in order to show clearly the separation between the two laws. Weber's law states that the proportion of judgments R > kR is a constant. R signifies here the physical magnitude of the stimulus, and k represents another constant. Weber's law
is concerned solely with physical measurements. It does not explicitly refer to the subjective continuum. On the other hand, Fechner's law states frankly the relation between the subjective continuum and the physical stimulus continuum. Fechner's law states that this relation is generally logarithmic, and it should be taken as a rough approximation to the relation between the subjective and the physical continua. Further, it can be seen that Fechner's law is applicable only to those stimuli which have a physical magnitude as well as an experienced intensity. The law of comparative judgment is completely independent of any physical stimulus magnitudes. The problem of the stimulus error is not ordinarily of serious concern to our problem. It deals with the ambiguity in the mind of the subject when he is asked to judge a stimulus as to the intensity of the subjective experience. Sometimes he attempts instead to judge the physical magnitude. A good example is that of a grocery clerk who can judge the weight of a bag of sugar. If he were asked to serve as a subject in the method of mean gradation, he would probably commit what Titchener would have called the stimulus error. In the measurement of social values, we are not interested in physical measurements because in general they do not exist for such values.
A very important advance in the application of psychophysical methods was accomplished by Richardson when he devised the triad method for studying the dimensionality of a domain. Instead of asking a subject to judge whether one stimulus is x'er than some other stimulus, where x is any specified attribute, he set up the discrimination experiment in such a way that no attribute was specified. In the method of triads, the subject would be shown three patches of color, for example, and he would be asked to indicate which is the odd one, with the implication that the remaining pair are more alike than any other of the three pairs. In this way the subject can make judgments of the degree of similarity or difference without having any specified attribute. Data collected in this manner can be transformed into the equation of comparative judgment, and the dimensionality of the domain can then be ascertained by the Young-Householder theorem. Such a method can be used experimentally to determine the dimen-
( 51) -sionality of the various sensory modalities.
Perhaps the best-known application of these experimental methods for the study of values is in the measurement of social attitudes. The most sensitive experimental procedure is to present the subject with pairs about which he is asked to make certain judgments. For example, he may be presented with pairs of nationalities, and he may be asked to judge for each pair which he would rather associate with. That type of experiment has been carried out in several ways. The judgments that are made by the subject depend, of course, partly on his own preferences, which are closely related to his own nationality, and the judgments are also determined by the nationalities that are judged. If two groups of subjects are asked to make judgments of this kind, one can say on the basis of objective evidence which of the two groups is more tolerant of other nationalities. At one extreme we would have people who are completely tolerant toward all nationalities. They would then also, of course, be completely indifferent about their own. Such people would have no national loyalty or identification. At the other extreme we would have people who are said to be strongly prejudiced or biased. They would have extreme loyalties to some nationalities and extreme dislikes for others. I doubt whether we should consider either of these two extremes to be ideal.
Some years ago Professor Eggan wrote a master's thesis in psychology before he went into the field of anthropology. In that master's thesis he wanted to know the effect of different forms of question with reference to nationalities. He had five different questions representing different degrees of intimacy. All five groups of subjects were given the same lists of pairs of nationalities, but there were different questions. One group had the question, Which of each pair of nationalities would you rather associate with? Another group had the same nationality lists, but they were given the question, Which would you rather have as a fellow student? Another group had the question, Which nationality would you rather have your sister marry? The proportions were superficially quite different, but the rank orders of the nationalities were essentially the same. In this case, we would probably find that the form of the question has a tremendous effect on the discriminal dispersion but relatively little effect on the order of the nationalities. The effectiveness of comparative judgment for studies of this type should be exploited further.
In studying the measurement of social attitudes, the attempt is sometimes made to validate such experiments in terms of overt behavior, but that is an error. Professor Sam Stouffer at Harvard wrote a doctor's dissertation some years ago at the University of Chicago on this problem. He investigated social attitudes by means of statement scales in reference to the prohibition issue. He obtained data about his subjects as to their actual behavior on prohibition. He found that there was pretty fair agreement between what the subjects said on the attitude scales and how they actually behaved. I should like to point out that, while such a comparison is of considerable interest, it is not a validation of the attitude scale. A man may be entirely consistent in what he says and in what he does about a controversial issue, and yet both of these indexes may be dead wrong in reflecting his attitude. In order to determine a man's attitudes in the sense of affective disposition about a controversial issue, it will be necessary for his friends to ask him privately when he is free to speak his mind and when he is not likely to be quoted. His personal atti-
( 52) -tudes may or may not agree with what he says and what he does. Here again, attitudes are essentially subjective experiences which may or may not conform with overt action.
Another distinction in the study of social attitudes which is sometimes lost sight of is that the cognitive and the affective appraisals may be entirely independent. For example, a group of subjects may agree in their strong dislike of communism. Someone might give them an examination in order to show that the subjects actually do not know what they are talking about. That might very well be true, but the psychological fact is nevertheless in-escapable that the affective attitudes may be strongly for or against a stimulus even if there is a great deal of confusion about its cognitive description.
The statement scale is not so sensitive as the paired-comparison procedure. It consists in a set of statements to which the subject responds by acceptance or rejection for each statement. In constructing such a scale, one presents a large number of statements to a group of subjects whose principal qualification is that they can read English. These subjects are asked to indicate for pairs of statements which represents the stronger attitude for or against x, where x represents the psychological object to which the attitude scale refers. For rough survey purposes the attitude scales are useful.
An interesting application of these methods of studying values is to appraise the effects of propaganda. We made a large number of experiments on the effects of motion-picture films on the social attitudes of high school children. Statement scales and paired-comparison schedules of various kinds were given before and after the showing of a motion picture. By this method we were able to ascertain whether a given picture had a significant effect and in what direction it did affect the children's social attitudes.
The method has also been applied in the study of international tensions by noting newspaper editorials. In one of those investigations a study was made with Chinese and Japanese newspaper editorials concerning each other, and it was shown, by treating key statements from the newspaper editorials, that the tensions increased at a very great rate before the two countries were at war. Professor Quincy Wright has suggested in his political science studies that such applications of psychophysical methods might be useful in studying international tensions before they become very marked.
An application of these subjective measurement methods which has not yet been made will be in the definition of the morale of a group. In general, the morale of a group is described by newspaper reporters and by others who mix their own value judgments with the characteristics of the group to be described. For scientific work we should have a definition of morale which is entirely independent of the value judgments of the observer. Such a definition could be stated in terms of the dispersions of all of the debatable issues within the group. Other applications would be in the comparison of cultural and nationality differences as to the values that are considered to be essential. It is unfortunate that most students of social psychology and political science are too descriptively minded to adapt the quantitative methods that may be available.
Let us turn next to the experimental study of moral values. We have carried out several experiments in which a group of subjects was given a list of offenses that were presented in pairs. For each pair the subjects were asked to indicate which of the pair they considered to be the more serious. On the basis of data of this kind and with the aid
( 53) of the equation of comparative judgment, we ascertain the scale values and dispersions for these offenses. In one case we gave a group of high school students such a list of offenses, and we determined the scale values and dispersions for these stimuli for three occasions. The first presentation was a day or two before they saw a film that described the life of a gambler. A few days after seeing the film they were given the second similar schedule. About six months later they were given the third schedule. The film described the life of a gambler, and we wanted to know whether this film had an appreciable effect on the attitudes of the high school youngsters toward gambling. We found that they considered gambling to be a much more serious offense after seeing this film than they did before seeing the film. In a number of experiments of this type, we also found that the motion pictures had much more lasting effects than is ordinarily supposed. In many cases we found that only half of the effect of the film wore off in six months. It should be said, however, that these experiments were carried out in small towns in Illinois, where the children do not see so many movies as in the large cities. We carried out a similar experiment in the Hyde Park High School in Chicago, where the children were given free tickets to a movie at the Tower Theater, a few blocks away. There we found that the effect was very slight. Our interpretation was that one movie more or less for children in a large city high school makes very little difference in their attitudes. These methods of studying moral values could be used very effectively in the comparison of different groups in a large city. The groups might represent different nationality backgrounds and different religious backgrounds. It would be interesting to ascertain what these differences would be. Such social psychological studies would help us to understand the problems of the extremely heterogeneous populations in the large cities. In a similar manner we have investigated experimentally the summation effect in propaganda, where the effect of a single stimulus does not show a statistically significant effect.
Another interesting field of application is in experimental semantics. It would be useful, for example, to have an index of affective intensity for adjectives in a dictionary. Two adjectives may be equivalent as to cognitive meaning and yet differ widely in affective meaning. The words "famous" and "notorious" might be examples. So are the words "pleasant," "gay," and "hilarious." Such affective indexes would be useful in translating a foreign language.
We turn now to another type of psychophysical problem. In the psycho-physical methods that we have considered so far, the main problem was to allocate each idea or object to a subjective continuum which may be unidimensional or multidimensional, depending on the nature of the problem. In most problems it is unidimensional. For example, if we ask subjects to judge the relative seriousness of offenses, we are dealing frankly with a unidimensional continuum, even though the discriminations may take place in a multidimensional continuum. We have here an obverse psychophysical problem. Having determined the subjective space which describes a group of subjects as to their attitudes in some field, we now inquire whether we can predict in any way what these people will do. When we turn the psycho-physical problem in this manner, we find some exceedingly interesting psychophysical theorems of a new kind. I shall give a few examples.
Consider two political candidates for an election. Let one of them have a
( 54) wide dispersion on the affective continuum. By this we mean that some people are very enthusiastic about this candidate, whereas others actually hate him. Let the other candidate have the same average popularity, but assume that he has a narrow dispersion, so that very few people are enthusiastic about him and very few people strongly dislike him. If these two candidates come to an election, we should expect them to split the vote evenly. However, the more variable of these two candidates might introduce a third candidate of approximately equal popularity and who also has a narrow dispersion. Then we would have three candidates, one with wide dispersion on the affective continuum, and two candidates of narrow dispersion, and all three of them would be equally popular on the average. In such a situation the more variable of the candidates would draw half the votes, and the other two candidates would get 25 per cent each. These proportions would be altered somewhat, depending on intercorrelations between the attitudes toward the candidates, but the principle can be illustrated in the general case for zero correlation. This principle is no doubt well known among politicians, but I doubt whether any of them have ever thought of this principle as a psycho-physical theorem.
Let us turn to another simple example from the field of market research. Consider a mail-order house or a retail store which carries a limited number of neckties. They desire to please the majority of their clientele. The manufacturers offer many hundreds or thousands of necktie patterns. If you turn to market research people with this problem, they may ascertain the twenty or thirty or perhaps fifty of the most popular designs, and they may suggest that these be the designs that should be carried. But that is the wrong answer. Suppose that several hundred necktie patterns were submitted to a sample of the clientele. With such records one could rather easily determine not only which patterns should be carried but also the number of patterns that should be carried in order to satisfy a specified proportion of the clientele. We would start with the most popular design and set that aside to be included. In the sample population we would then eliminate all who chose that popular pattern. Then we would inquire about the most popular pattern in the remainder of the sample population. That pattern would be set aside as the second design to be accepted. Eliminating those who chose that pattern, we would ascertain the most popular pattern in the remainder of the sample population. Proceeding in this way, we would come to the point where an additional pattern would increase the selection by only a very small percentage of the population, and that would be the time to stop. In such a procedure we could determine the number of patterns as well as the designs which should be used in order to satisfy a specified proportion of the clientele. The ordinary solution of selecting the most popular designs would lead to a situation where some customers are confused by having many patterns which are equally acceptable, while other customers find nothing to please them. The maximum satisfaction will be derived by proceeding in some such way as I have outlined. There is nothing profound about this procedure, and yet it would probably be novel in market research. There are situations where problems of this sort can be of national importance. If it should be necessary to restrict the manufacture of civilian goods, then it might be important to encourage the manufacture of a limited number of designs for all sorts of things and to select those designs in such a manner as
( 55) to please the majority of the civilian population. In this manner the psychophysical methods may be important in contributing toward national morale.
Recently we made an experiment on the prediction of choice with regard to menus. In this problem we were concerned with the simplification of psychophysical methods to the point where they would be practically useful for survey purposes. The psychophysical methods of the laboratory are often too laborious to be used in practical surveys. It was decided to adapt the method of successive intervals for this problem. We presented a list of forty foods on a successive interval schedule in which each subject was asked to indicate by a single check mark his relative degree of like or dislike for each food item. There were nine short descriptive phrases which represented degrees of like and dislike for foods. This schedule of forty items required less than five minutes for each of several hundred adult men subjects. In addition to this short survey schedule, we also presented them with sixteen menus in which they were asked to indicate what they would be likely to choose from each menu. For example, there were four lists of desserts, several lists of entrees, other lists of vegetables, and the like. For each menu the subjects were asked merely to check which they would select from a given list. Vanilla ice cream occurred in several of the desert menus. The proportion of the subjects who selected vanilla ice cream for dessert depends, of course, in part on their relative like or dislike for this dessert, but the selections would also depend on the competing items in the dessert list. By the application of the method of successive intervals and some theorems in psychophysics, we predicted the proportion of the subjects who would select each one of the items, and there were fifty-six such predictions. These predictions were based entirely on the short, five-minute schedule for the whole' list of forty foods. We compared these predictions with the actual choices that the subjects made when they were confronted with the actual menus. The agreement was remarkable. The maximum discrepancy was between 3 and 4 per cent, with one conspicuous exception for a dichotomy, namely, roast beef and fried chicken. The ratings for these two items were both in the upper two categories, and the discrepancy there was 8 per cent, which was probably due to the effect of coarse grouping. The experiment demonstrated quite adequately that the prediction of choice can be effectively made with very simple survey schedules if these schedules are properly analyzed.
Some of these experiments deal with rather trivial values, while others deal with socially more important values, but our principal concern here is in the development of those scientific methods which can be adapted over a wide range of values whether they be socially important or trivial.
We turn next to the application of psychophysical theory to some experimental problems in economics. For a long time there has been consider-able interest in the measurement of utility, but the measurements have generally been indirect. Psychologists have been able to measure utility experimentally for over two decades, but economists have not until very recently expressed interest in these methods. In the last few years there seems to have been a marked change in the attitude of economists to these problems. In principle, utilities can be measured for an individual subject, but it is easier experimentally to apply these methods to the measurement of utility for a group of subjects. Psychophysical theory lends itself well to a number of variations in the measurement of utility. For exam-
( 56) -ple, the utility of a purchase can be described as the algebraic sum of the utilities of the object and of the price. In this case, the utility of the object would presumably be positive, whereas the utility of the price would be negative. The question then arises about the location of a rational zero point for the scale of utility. An experiment is now in progress to demonstrate an experimental procedure for locating the zero point in the scale of utility. It seems reasonable that the prices of various competing objects should be checked with their utilities to ascertain for any specified population to what extent some objects are over-priced or underpriced. Survey methods are available for doing these things. In determining the zero point for the scale of utility, we are asking several hundred subjects to express their preferences among various objects that might be given to them as birthday presents. Each of these single objects will then be given a value on the scale of utility. In addition to these judgments, we also asked the subjects to make a number of different judgments. We asked them whether they would prefer to receive gifts A and B or C. In this case they must judge whether the satisfaction from A and B is greater or less than the anticipated satisfaction from the single birthday present C. By judgments of this sort we expect to be able to locate the zero point of utility, because the sum of the affective values of A and B combined should equal the utilities for these two objects taken separately. Within the range of the experiment with a small number of different objects to be selected, an additive theorem can be assumed to hold reasonably well. Diminishing returns would probably not be noticeable within the choice of four or five different objects.
In making these adaptations of psychological measurement theory to economics, one naturally wonders whether economics could be developed as an experimental science. Although I am not an economist, it has seemed to me entirely feasible that economics should be developed as an experimental science. In discussing this question with some of my friends in economics, I find that they are divided. Some of them insist emphatically that economics can never be an experimental science, while others are equally certain that this is possible. As an example we might consider the indifference function in economic theory. An indifference curve can be considered as a curve showing the combinations of two commodities X and Y which have the same utility value. If the amounts of the two commodities are considered to be the x and y axes in a three-dimensional model, then utility can be considered as the ordinates which are perpendicular to the x-y plane. An indifference curve would then be a horizontal section parallel to the x-y plane which represents constant utility. For different values of utility we would then have sections at different elevations which give a family of in-difference curves. It has been shown that these indifference curves can be determined experimentally. There are many situations of controlled economies where the shapes of these functions can be studied experimentally. Such situations are in occupied countries or in prisons and in other situations with central control of prices. By altering the price of a commodity, the changes in the indifference curves can be noted experimentally.
As a final example of the adaptation of psychophysical theory in the measurement of values, we shall consider the field of aesthetics. If aesthetics were to be regarded as a purely normative science, then we should expect the aesthetic value of an object to be determined by its physical properties. Such an inter-
( 57) -pretation seems well-nigh hopeless. It seems much more fruitful to recognize that the aesthetic value of an object is determined entirely by what goes on in the mind of the percipient. In this manner of looking at the problem, we deal again with values that are subjective experiences and which may vary from one person to another and certainly from one culture to an-other. An esthetic object symbolizes human emotional experience and its resolution in a conceptual and abstract manner. Except in extreme cases the aesthetic experience is not itself emotional. It is essentially an abstraction. There is nothing absolute about the value of an aesthetic object. The aesthetic value is determined by the experience and the attitudes of the observer.
Some time ago I attended a series of seminars on aesthetics at the home of one of my colleagues. Most of the participants in that seminar were from the humanities and the arts. The seminars were devoted to discussions about the theory of aesthetics. In some of those discussions it occurred to me that the question at issue could be treated as a question of experimental fact, and I ventured to suggest how the psychophysical methods could be adapted to obtain an empirical answer to the question at issue. It was an illuminating experience to discover that some of my friends in the humanities were hostile to the very idea of subjecting questions of aesthetic theory to empirical inquiry. On one of those occasions a friend showed me a quotation from Aristotle that settled the matter for him. It was heresy when I suggested that we knew more about this problem than Aristotle. Artists are sometimes suspicious of the experimental study of artistic preferences and perhaps with some reason. Sometimes experimental studies are made in aesthetics when the investigator is interested in secondary effects rather than in the aesthetic experience. On the other hand, I have found some artists who are very much interested in such inquiry. A friend who is a portrait painter frequently encouraged experimental studies of this kind at the Art Institute in Chicago. Unfortunately, I have not been able to induce many students of psychology to study experimental aesthetics.
In closing, I should like to comment briefly on the social studies as science. It is unfortunate that the social studies have rather low prestige among the sciences. I believe that this is what we should expect because so far the social studies have not adopted the impartial, objective, and intellectual attitudes of science. Quite generally in these fields the writers argue for social action of some kind, about the right and wrong ways of life, about what is good and what is evil in the opinions of the writers, about the good and the bad names and categories for describing their political friends and enemies. It is the exception when a social scientist studies social phenomena as science to identify the forces at work without name-calling and without injecting his own value judgments into what he is describing. As long as social scientists fail to distinguish between propaganda and science, they will have low prestige among the sciences.
SUMMARY
This paper has been concerned with the problems of a subjective metric. Social studies do not need to be quantitative in order to qualify as science. Some of the most important experiments in science deal first of all with the description of basic phenomena in a qualitative way. It usually happens that quantitative methods appear with more intensive study. Here we have con-
( 58) -sidered some exploratory attempts to establish a subjective metric for the measurement of values. I have not succeeded in persuading social science students about the fascinating challenge to develop their field as science. To do so, we must free ourselves from the impulse for social action which has no place here. We should avoid problems in which we have an ax to grind. As citizens we have the privilege and the duty to participate in political elections. But when we work as scientists, we should be aloof from the issues of the moment and the chatter of the market place. Only in scientific detachment and objectivity can we eventually be helpful in developing the social studies as science.
