When it comes to criticism it's best to confront the issue directly by loosely following Eysenck's extensive objections raised in The Structure of the Human Personality.

The first objection concerns Sheldon's mathematics. Sheldon did not use factor analysis, which was still developing when he wrote, but a less refined method called cluster analysis. Later investigators attempted to put his data in more mathematically sophisticated factorial forms. Eysenck cites Adcock (1948) and Lubin (1950) who found difficulties in understanding how Sheldon arrived at his conclusions from a mathematical point of view. A treatment of Sheldon's data led Ekman (1951) to the conclusion that it would be possible to represent Sheldon's three factors more parsimoniously by two orthogonal factors. And Hammond, in "The Status of Physical Types" (1957), reviews the two basic approaches to the study of body types, that is, both Sheldon's and the factor analytic approach, and describes the difficulties that he feels are found in each. For example, the elements in the somatotypes are complexly related and factor types offer no generally agreed solution. And then he reviews several studies of somatotypes by means of factor analysis in which Howells (1952), and Lauren Fields (1954) found two basic factors. One was endomorphy-ectomorphy with the endomorphs and the ectomorphs at the opposite ends of one continuum, and the second factor differentiated the mesomorphs from the endomorphs or ectomorphs. Hammond, himself, suggests a compromise based on Parnell's length, muscle and fat, "expressed by positive and negative numbers whose magnitude indicates general size and whose pattern shows the type distinction". (p. 240)

These lines of mathematical critique were pursued vociferously by Humphreys (1957) in his "Characteristics of Type Concepts with Special Reference to Sheldon's Typology". He was upset with Sheldon's way of discovering the three factors:

... with regard to establishing the physical types, it is clear that the procedure was not empirically sound. The types originated in the arm chair. Sheldon did have large numbers of photographs spread out before him when he selected the types, but that hardly makes the procedure empirical." (p. 219)

He goes on to attempt to demonstrate on mathematical grounds alone "that Sheldon has evidence for no more than two independent (not necessarily valid) types of human physique." (p. 221) And he makes an interesting point meant as a thoroughgoing criticism, but which can be interpreted in a wider context, as we will do later. "Measurement entered later as a means of differentiating objectively the subjectively determined types. It was also shown that the choice of types to describe human physique and temperament automatically restricts the data in predictable ways." (p. 227)

Eysenck adds another category of criticism by accusing Sheldon of being derivative. According to him, Sheldon got his principle ideas from Kretschmer, his embryological notions from Bessonet-Favre (1910), Bauer (1923) and Castellino (1927), and his idea of separate factors from Plattner (1938). What Sheldon added was his photographic method. (p. 321-24). All this leaves the impression of Sheldon furtively scurry" ing about picking up bits and pieces and cobbling them together into a system. The facts are rather different. Eysenck glosses over the differences between Kretschmer and Sheldon, and ignores Sheldon's extensive familiarity with the history of typology. For example, Sheldon's chart of his predecessors is more extensive than the one Eysenck himself gives. Sheldon does mention Bauer and Plattner, as well as dozens of other people (but not Bessonet-Favre and Castellino) and is careful to indicate the origin of his ideas.

But there are more important objections in the offing. Do somatotypes. really change? Eysenck cites Lasker (1947) who found differences in somatotype ratings before and after partial starvation, and Newman (1952) who found changes with age. He also cites Tanner (1956) who found little relationship between birth measurements and later measurements, but did find that childhood measurements and adult measurements were more strongly correlated. Hammond (1953) found some constancy in children over a period of several years, though a percentage of both the boys and the girls in his study had, according to him, changed their physical type. In a study by Zuk (1958) the somatotypes of male and female subjects were measured at 12, 17, and 33. Most of the correlations were fairly high, but different problems, for example, the estimation of endomorphy, were noted. Eysenck concludes "the evidence suggests that there is considerable stability of somatotypes but this is by no means as perfect as Sheldon would have us believe." (p. 329)

Next the issue on Eysenck's agenda was the close relationship Sheldon found between somatotype and temperament. There is a mainly negative study by Fiske (1944) and one with more positive results by Child (1950), which we should look at in more detail. A group of Yale students were somatotyped by Sheldon and some of them were given a questionnaire during a course, while the bulk of the subjects filled out the questionnaire after it was mailed to them. This procedure resulted in 414 actual subjects. Child had developed a whole series of predictions based on Sheldon's system, and "of 96 predictions based on Sheldon's views 77 per cent are confirmed by the direction of relationship found in these data. Of the 74 correlations which are in the predicted direction, 20 reached significance at the 5 per cent level, whereas only I of the 21 correlations, contrary to prediction is significant at the 5 per cent level. The most marked contrast between the results which conform to prediction and those which do not is found in the number of instances of results significant at the I per cent level, where the contrast is between 10 and none." (p. 445)

Child concludes:

"It is thus possible, but not certain, that appropriate measures based on self-ratings such as were used here have a quite sizable relationship with dimensions of physique. It is reasonably certain that this relationship does not at all approach the magnitude of the relationships reported by Sheldon between dimensions of physique and his measures of temperament." (p. 447)

Results in the same direction can be seen in Davidson et al. (1957) who found a relationship between ectomorphy and the traits meticulous, fussy and conscientious, and Smith (1957) who compared somatotypes and MMPI scales and found many significant correlations, most of them as predicted from Sheldon, but the correlations were much lower than the ones Sheldon found.

A more positive study by Sheldon and Wittman of 155 psychotic male patients found higher correlations between the various psychiatric components and the components of physique. Eysenck found these results of considerable interest, and this is a study we will return to later in another connection. There are also studies by Fiske (1942), Smith (1949) and Janoff et al. (1950) where somatotypes were compared with objective tests of temperament, and these found extremely low correlations.

Eysenck concludes his examination by saying that Sheldon's contributions "probably contain sufficient truth and insight to be worthy of proper scientific investigation.

"Sheldon's analysis is essentially subjective; his types, as are those of Kretschmer and his predecessors, are derived from intuition and theory. It is obviously advantageous to have a more objective approach based on empirical data..." (p. 334)

Have we, then, reached the point where Sheldon has been exposed as an armchair theorist, as Humphreys would have it, or a derivative dabbler whose work must be redone at a proper scientific level? No. This is hardly the case. We are once again confronted with the problem of method in psychology, and this delicate theoretical problem is exacerbated by the attitudes of the combatants. Sheldon does employ various mathematical techniques, but he does it in such a way that at times it appears, as Eysenck says, "en passant", and has something of an off-hand sense of condescension for the less perceptive. This, of course, would be infuriating to those on the other side of the methodological fence, but they, in their turn, have their weak points. Humphreys, for example, wants to limit the empirical to what his mathematical methods can perceive, and thus relegates Sheldon to being an armchair theorist, which is not correct. Eysenck, himself, appears to picture his role as the difficult and laborious one of reconstructing psychology on more secure scientific foundations. This has a leveling effect. The intricate and fascinating typological edifice that Jung constructed on the basis of empirical material is limited to the ideas of introversion and extraversion. The four functions are ignored, as well as the dynamic relationship between the conscious and unconscious attitudes. The same process can be seen in relationship to Sheldon. Sheldon's highly perceptive insights, which he systematized into a new high point of the understanding of physique and behavior, are subjected to a similar process of leveling, which eliminates almost all of what is most interesting in Sheldon. It amounts to saying since our methods cannot achieve the results that Sheldon says he achieved, therefore Sheldon is wrong. No one is particularly interested in following Sheldon's methods. Eysenck says: "One difficulty in the way of a ready acceptance of Sheldon's work must be the difficulty of repetition. Few psychologists can observe their subjects for a whole year and give each one 20 or more analytic sessions before making a rating." (p. 141)

Sheldon's method is very different from that of his critics. At first glance he may appear to be employing exact measurement and mathematical analysis, but this is misleading. He is best understood when he is placed in Jung's company in terms of scientific methodology. L.L. Thurstone gives us an interesting criterion for trying to clarify this point, when in a talk to the Psychometric Society in 1936 on the establishment of psychology as a quantitative and rational science, he distinguishes the use of mathematics as an aid or tool from its use as the very language in which the psychologist thinks. No one would make the mistake of imagining that Jung thought in the language of mathematics, and while Sheldon used mathematics, a case could be made that he did not think in the language of mathematics. For example, even though he worked with Thurstone on his original attempt to mathematically represent his constitutional conclusions, they did not find favor with later factor analysis advocates. Further, Sheldon was apparently in no rush to try to objectify his somatotyping procedures. He moved from observation to measurement and back again. It was only with the urging from his long-standing associates Eugene McDermott and C. Wesley Dupertuis that he produced his final objective method.

Sheldon resisted developing completely objective techniques, and part of the resistance has to be seen against our earlier discussion of the deadends that an overreliance on measurement and statistics had led him to early in his career. Perhaps he overreacted. He was annoyed at his colleague C.W. Dupertuis when in the course of their examinations of the thousands of mental patients Dupertuis wanted to take anthropometric measurements, for he felt he had tried them at one time but they didn't show what he was looking for.

Eventually he came around to the necessity of showing that his somatotyping could be done on a completely objective basis. But this came late in his career and went virtually unnoticed. We will examine this contribution a little later, but now let's turn to see what Eysenck and his factor analytic colleagues would put as the more objective structure that should replace Sheldon's. The various factor analytic studies of physique can be briefly summarized as having yielded two factors: one of body size and another of height and width. According to Eysenck:

"The major outcome of factorial analyses of body build appears to be that we may regard the body as a rectangle which can be described with fair accuracy in terms of two independent dimensions, to wit, height and width." (p. 339)

On this basis physique can be divided into three classes: the eurymorph, equivalent to Sheldon's endomorph and mesomorph, or the lateral physique, and the leptomorph, equivalent to the ectomorph, or linear physique, and a mesomorph which is not Sheldon's mesomorph, but which is simply someone in the middle from a statistical point of view. We have returned to the old dichotomy which originated with Hippocrates, but now in a more sophisticated form.

This is no place to take up a mathematical defense of Sheldon. Perhaps some day a highly impartial and mathematically gifted researcher will unravel this question, and I don't think Sheldon would object to this. Quite the contrary. C.J. Adcock, for example, in his factorial examination of Sheldon's types analyzes Sheldon's figures and concludes that a halo effect must be the explanation for the uniformity of Sheldon's correlations. "It's much too perfect. Nevertheless we feel that there is a sound basis to it all and that with all allowance made for the halo there is still something substantial left." (p. 318)

When Adcock writes "these correlations between S (somatotonia) and C (cerebrotonia) are too high which makes impossible a normal analysis of this data", Sheldon, in a margin of an offprint of his article responds, "Yes. The original error, of course, lay in the insufficient thickness of the assumed 3dimensional plot of the S-types (76 S-types). With the current parameters and greater spread of S-types this difficulty vanishes and the 3 components are more nearly orthogonal". And he felt he corrected this original error with his later article "Psychotic Patterns".

It is important to note, however, that Sheldon's mathematics are built on the foundation of his observations. As Humphrey noted in another connection, this might automatically effect how the data is analyzed. Sheldon is trying to quantify something he has already seen, and so there is a constraint placed on the measurements he will choose and their analysis. The factor analysts are taking a series of measurements and seeing what emerges. Their mathematics suffers from fewer constraints, and therefore is clearer and more lucid. But at the same time is it not possible that they are flattening out what Sheldon saw, since their techniques cannot grasp it? Thus, skeletal measurements readily show a height and width factor, but have trouble dealing with the question of muscularity and fat, which stand out more sharply from an observational point of view. This does not mean there are no highly distinctive mesomorphs so that mesomorphy becomes just a statistical classification. It can also mean that different methods produce different results, somewhat of a truism which will take on flesh when we look at the work of Parnell, Dupertuis and others.

Let's return for a moment, though, to the results of factor analysis in other areas comparable to Sheldon's work. The eurymorph body build is connected with extraversion and manic-depressive disease, while the leptomorph is connected with introversion and schizophrenia. To this can be added the leptomorph's inclination to what Eysenck calls "dysthymic" kinds of neurosis, and the eurymorph to more hysterical disorders. All this parallels insights found both in Sheldon and Jung.

But what of the factor of size? Eysenck in a 1940 study compared 156 microsomatic, 156 macrosomatic and 688 mesosomatic male neurotic soldiers, that is, he compared the small-bodied to the large and medium-bodied. The small-bodied had the following characteristics: poor physical health, weak and dependent, inert, depressed, poor muscular tone, and so forth. "Altogether both mentally and physically he is what is popularly called a poor specimen." (p. 345-6) The reader of Sheldon would clearly recognize the parallel between the microsomatic and Sheldon's asthenic characteristics. The asthenic person was characterized by Sheldon, following Hooton, as "poor protoplasm, poorly put together".

What all this amounts to is that Sheldon is guilty of having rated his initial 200 subjects on both somatotype and temperament, and therefore could have let the results of one level influence the other. This does not mean that anyone has demonstrated that he did allow them to influence each other. Further, his mathematical analysis of the data is not ideal, a fact he would admit and worked to rectify. At the same time he was working at a time when the current factor techniques were still being developed. And he was determined not to let mathematical technique obscure the deeper realities he was trying to draw out.

There is a final issue that became a point of contention. What are the biological underpinnings both to Sheldon's work and to factor analysis? Eysenck finds support for the two factor theory in Lindegard's work in measuring the length of the long shaft bones and their thickness.

"The very extensive correlations reported by him suggest strongly that these two factors, derived from physiological and anatomical considerations, correspond closely to the length and width factors of the factor analyst, which thus find strong biological backing." (p. 341)

But when it comes to Sheldon's assumption that there is a connection between his three factors and the three layers of the embryo, Eysenck is convinced of the importance such a relationship would have if it could be proven, for "the observed intercorrelations among the types would be a cheap price to pay for the gain in understanding obtained through this relationship." (p. 326) But he has various reservations about this connection actually existing. He asserts that Sheldon fails to account for the complexity of the development of the germ layers, makes no deductions from his hypotheses, does not take into account the "mesenchyme", or a sort of intermediate layer between the other layers, and he cites an article by Hunt (1949) as evidence to the contrary for a relationship between the layers of the embryo and the different physiques. This is an issue of considerable importance which outweighs, as Eysenck makes clear, the various debates about two or three factors, and we will return to it in a later section.

All in all, there is a rather remarkable parallelism between the basic results of the two methods. If there were not such methodological bad blood, this common ground would be more to the forefront and more collaborative efforts would be taking place between people involved in both methods. But this excursion through Eysenck can still leave the impression that Sheldon's work has to be satisfied with faint praise begrudgingly given. It is important, then to spend some time looking at a variety of studies in this chapter and the next which will help redress the balance.

Among these studies pride of place should go to Sheldon's final paper entitled, "Psychotic Patterns in Physical Constitution", a 30-year follow-up of 3,800 psychiatric patients in New York State, which was given at a symposium on schizophrenia in New York in November, 1968. Unfortunately, by then, his critics had dominated the field for so long that it must have appeared to the psychological community that Sheldon had nothing to say in his own defense, and this paper did little to change that state of affairs. His critics, if they heard about it, which was not too likely, ignored it. This is unfortunate because it is a brilliant piece of work that is very much Sheldon at his best, sharp and perceptive. It contains two sections: one, a preliminary report on the psychiatric studies that he and his colleagues had carried out over the course of 30 years in various hospitals around New York State. The other part was the final objectification of somatotyping. Sheldon reviews his previous work, and then turns to the objections brought against it, admitting that in his Varieties of Human Physique, "a number of major problems were doubtless oversimplified and some were left unexplored or even worse, unmentioned." (p. 844) He summarizes the objections under four headings: (1) the somatotype changes (2) somatotyping is not objective (3) there are only two, not three, primary components (4) somatotyping omits the factor of size.

In considering the first objection he describes his early attempts during the time of his Ph.D. thesis to identify the primary factors in physical constitution: "Thurstone decided that 4 primary factors were demonstrated and that one of these factors was size. When we expressed all the measurements as ratio to stature, thus eliminating size, three factors remained." (p. 844) These Sheldon took to be expressions of the three components he had discovered by his inspection of the somatotype photographs. If there were no relationship between the three, there should have been 343 somatotypes. If the correlations were highly negative there would not be three components, but only two, which, as we have seen, was a conclusion many of his critics arrived at. But Sheldon felt that this was not a good conclusion to come to when dealing with a three-dimensional organism. So he tried to steer a middle path. After a series of discussions with Thurstone he decided there was no way, a priori, to determine how many somatotypes there would be. It was going to be a process of "try it out and see". In his first book he identified 76 somatotypes. By 1954, in The Atlas of Men, there were 88, and the search still went on for those measurements which would more and more objectively determine the somatototype which Sheldon felt remained stubbornly constant through life. Finally, he discovered what he called the trunk index, which is the ratio of the thoracic trunk over the abdominal trunk. In other words, the trunk is divided at the waist into two sections and the area of each is measured by means of a planimeter on the standard somatotype photographs. Its discovery paved the way for solving three of the objections to somatotyping. It distinguished between endomorphy and mesomorphy; it remained constant from about the third year of life to old age, despite variations in fatness or leanness, and it opened the way for the complete objectification of somatotyping.

Sheldon had extensive opportunities to test out this new index. He examined the photographs of 400 subjects who were part of the Berkeley Growth Study, for whom somatotype photographs were taken since they were children until they were fully grown. The trunk index remained constant. He looked at the somatotype photographs taken at the University of Minnesota during a starvation experiment. Starvation had not changed the trunk index. He also examined West Pointers upon entering and leaving the Academy after closely supervised body-conditioning programs, Columbia University students who had been photographed as freshmen and then had been somatotyped 40 years later, a series of 46 identical twins, and men and women in a weight-reduction clinic, which included women who had lost as much as 150 pounds. In all these series the trunk index remained unchanged from early childhood onwards.

He finally had an objective way to distinguish between endomorphy and mesomorphy. He had a second parameter in the maximum weight the subject had reached, and he found the third in stature. This would allow him to measure the three-dimensional organism: "First we have to have a measure of massiveness; then a separator for the two kinds of mass (endomorphy and mesomorphy); finally a measure of the degree of stretched-outness into space." (p. 848) The Trunk Index has zero correlations with stature and ectomorphy. The correlations between endomorphy and mesomorphy approach zero in both sexes, and the correlations between ectomorphy and endomorphy and ectomorphy and mesomorphy average about -.40. Sheldon realized that in his earlier more visually oriented methods he had been doing the equivalent of taking into account both stature and the trunk index, but indirectly. Now he could simply measure maximum weight, height and the trunk index and calculate the somatotypes, and he could answer the objections against his somatotyping:

"1) The somatotype changes. It cannot change, since maximal stature and maximal massiveness are simply items of historical fact, and TI (Trunk Index) is constant through life.

2) Is not objective. Now it is completely objective. Can be derived on a computer as a function of three parameters, thus providing an operational definition of the procedure.

3) There are only two, not three primary components. This difficulty no longer exists. It arose from the fact that the negative correlations among the primary components were too high. That condition has been corrected, thus permitting us to live and operate in three spatial dimensions again.

4) Omits the factor of size. Size has been restored by using stature as a determining parameter." (p. 848-9)

It gives one a strange feeling to read Sheldon's culminating contribution against the background of all the criticism his work provoked, and to realize it was met with a thundering silence. Here was a completely objective method that could be replicated by anyone, but virtually no one bothered. That this would happen is a good illustration of how much the scientific process is wrapped up in human nature with its passions and weaknesses. Sheldon resists objectification, antagonizes his critics, finally comes up with an objective method, and no one seems to have the time to examine it. But its implications are enormous. If this objective method is viable, and there is no reason to think that it is not, then it sheds an interesting light on Sheldon's less objective methods and on the question of the relationship between somatotype and temperament. If Sheldon was substantially correct in his choice of the three factors, was he also correct in finding a close relationship between somatotype and temperament? Let's look at two other studies, one which makes use of Sheldon's latest method, and another which studies the relationship between physique and temperament.

Richard Walker and James Tanner in their "Prediction of Adult Sheldon Somatotype I and II from Ratings and Measurements at Childhood Ages" (1980), examined the photographs of 82 boys at ages 5, 8, 11, 14 and 18, using both Sheldon's anthroposcopic method and his objective method. They found interjudged correlations for the anthroposcopic ratings ranging from .79-.93, while the correlations for the Trunk Index method ranged from .94-.99. The mean somatotype changed little with age by either method, and the correlations between the two ratings were in the low .80s. In the case of the anthroposcopic ratings, nearly 40% of the ratings were identical, and nearly 80% within a half point. With the Trunk Index method, there was nearly complete agreement for ectomorphy, and 87-96% of the ratings for mesomorphy and endomorphy fell within a half a point.

Name                                                 Age                      Date

Below are some statements that we would like to have you complete about yourself. Fill in each blank with a word from the suggested list following each statement. For any blank, three in each statement, you may select any word from the list of twelve immediately below. An exact word to fit you may not be in the list, but select the words that seem to fit most closely the way you are.

6. Underline one word out of the three in each of the following lines which most closely describes the way you are.

a) assertive, relaxed, tense
b) hot-tempered, cool, warm
c) withdrawn, sociable, active
d) confident, tactful, kind
e) dependent, dominant, detached
f) enterprising, affable, anxious

The first group consisted of 73 boys who were high school seniors, and clearly endomorphs, mesomorphs or ectomorphs, and the authors found a strong relationship between somatotype and the self-rating of temperament: endomorphy to viscerotonia +.51, mesomorphy to somatotonia +.69 and ectomorphy to cerebrotonia +.43, all of which were significant beyond the .01 level. And there were negative correlations between each physical component and the other two components of temperament. The correlations in their very extreme types were even higher: viscerotonia and endomorphy, +.66, somatotonia and mesomorphy +.74 and cerebrotonia and ectomorphy +.59.

In their second group of 100 college girls "there is perfect correspondence in all groups between the components of physique and the number of traits selected. In every instance, as the girls were higher or lower in each component of physique, they selected a higher or lower number of traits of the corresponding component of temperament." (p. 436) The correlations were: viscerotonia and endomorphy +.59, somatotonia and mesomorphy +. 57 and cerebrotonia and ectomorphy +.60. In their third group they tested 20 subjects who had committed serious crimes. All were high in mesomorphy, and they all selected more temperamental traits from the second group. The authors conclude: "All these findings are very positive. It has been shown that physique and self-description of. temperament are intimately associated." (p. 438) They also point out the diversity of the subjects they tested, the possibility of distortions by self-flattering as the subjects took the quiz, and the difficulty the subjects could have in understanding themselves as well as the meaning of the traits, all of which would have worked against positive correlations. Though they consider the possibility of various environmental factors at work, they state: "It seems reasonable to assume that physique, through the organs of the body, glandular secretions, and the particular chemotype, limits in individuals the range of temperamental traits and predisposes, together with other variables, towards some traits more than towards others." (p. 438)

It is safe to say, then, that while Sheldon has been neglected in recent years, he has not been discredited. There is something there and it is strongly there in Sheldon's somatotyping and its connection with temperament. Lindzey and Hall in an extensive review of Sheldon's constitutional psychology, pleasing in its sense of balance, state:

"One may quarrel over the degree of relation between physique and personality, or even over the factors that mediate this relation, but present evidence leaves little doubt, at least in the authors' minds, that something important is afoot here. Prior to Sheldon's forceful appearance on the scene, such a relationship was customarily dismissed in this country as representing little more than superstition or speculation." (Theories of Personality, p. 372)

Lindzey in a later article, "Behavior and Morphological Variation", perceptively adds:

"In his research and writing Sheldon is much more the sensitive naturalist, observer, and categorizer and much less the hard, quantitative, and objective scientist than would be optimal to assure a good press from our colleagues. Moreover, in his writings he has proven to be singularly adept at ridiculing or parody" ing just those aspects of the scientific posture of psychologists that are most sensitively, rigidly, and humorlessly maintained. One might argue convincingly that, if Sheldon had conducted the same research but had reported it in an appropriately dull, constricted, and affectless manner (consistent, let us say, with Journal of Experimental Psychology standards), its impact upon the discipline of psychology might have been much greater." (p. 228)

His comments on the high correlations Sheldon found between physique and temperament are equally well worth quoting. "If, as the climate of current opinion urges, the magnitude of association reported by Sheldon represents in large measure covariance attributable to experimental error, this fact remains to be demonstrated unequivocally. Moreover, in view of the extensive criticism of his study, it does seem odd that there has not been one single effort at a careful replication eliminating the major defects in Sheldon's study, while at the same time attempting to preserve other relevant conditions as exactly as possible. What we have witnessed, instead, has been the complacent dismissal of a potentially important set of results with no serious attempt at an empirical resolution." (p. 234)

In the 20 years since Lindzey wrote those words we still cannot point to any study that has tried to duplicate Sheldon's temperament work using the procedures that Sheldon advocated.