Personality Assessment System Foundation


The Personality Assessment System
A radical hypothesis

C. J. Krauskopf - Ohio State University

Introduction

The radical hypothesis described here, along with a theory based on it, is that differential aptitudes are a "cause" of much of the preferential behavior we call personality. Other things being equal, people would rather use the aptitudes they feel they are better at and avoid using those in which they feel weaker. This hypothesis is apparently radical, because it has not been taken seriously. Some have come close. Witkin's (Witkin, Dyk, Faterson. Goodenough & Karp, 1962) concept of Field Independence/Dependence (FI/D) comes close. Baron (1994) does say that people would rather do what they are good at, but does not develop the idea. Cantor (Cantor & Kihlstrom 1987) gives intelligence a central place. She posits intelligence as an aptitude that can be applied to developing social skills and behaviors that we might consider a part of personality. She appears to consider intelligence a general concept and differential aptitudes not very important.

"This hypothesis is apparently radical, because it has not been taken seriously. "

Many have made connections between personality variables and intelligence, but when they consider causative direction at all, it is that personality "causes" differential aptitudes. (Mathews & Dorn 1995;Barat, 1995; Goff & Ackerman, 1992) Connecting personality and differential aptitude has a long history going back, at least, to Pressey (1918) who grouped different kinds of items on the Stanford-Binet test and tried to make personality inferences. The most elaborate of these tries at application was Rapaport, Gill & Schafer (1945), inferring various forms of abnormal behavior from patterns on the Wechsler scales. After the disappointing results of Rapaport, Gill & Schafer such efforts nearly ceased, other than occasional application of Wechsler's sign approach (1958).

The essence of the scientist-practitioner model of applied psychology is that applications should be based, insofar as possible, on good science and that practice should inform science in the generation of hypotheses and information about what seems to work. I want to tell two stories here. The first is about a practitioner developing an assessment method beginning with a connection of test "signs" with behavior, and proceeding to a descriptive system which could mystify observers. The second is about the application of scientific methods to try to answer the question, "How does he do that?"

The First Story

Shortly after World War II John Gittinger was appointed the Chief (and only) psychologist at the Norman State Hospital in Oklahoma. Psychologists were expected to give diagnostic tests in those days and one of the most popular was the Rorschach. Gittinger found that in that hospital population responses to the Rorschach were quite sparse, giving little information to use. In contrast, when he gave the Wechsler-Bellevue they would try hard. They seemed to have the attitude that if they could show they had some intelligence, they might get out of the hospital.

In the days before the dominance of franchised fast food, small diners were common, and one source of jobs for marginal workers. The common jobs were fry cooks and dishwashers. Gittinger's first test-behavior connection was that dishwashers did relatively poorly on the Digit Span subtest (D) and that fry cooks did relatively well. This became the first variable in his descriptive system. He reasoned that a dishwasher could function even in a distracting environment and that the fry cooks had to keep orders straight in kind and time, having to have some way of avoiding the distractions. He calls this Internalization-Externalization using the Jungian notion of "direction of libido". Some people are able to store and retrieve information in their internal world easily, some less able to do this are more able to perceive external information and detect changes there. This difference makes the internal world more "real" for some and the external world more for others.

"This difference makes the internal world more "real" for some and the external world more for others."

After Gittinger moved to New York and was working with a much better educated population, the connections between Digit Span, distractibility and the other behaviors he thought related disappeared. The key to the disappearance seemed to be the Arithmetic subtest (A). When Arithmetic was high relative to all the other subtests, the original relationship was partially reversed. There was an interaction between the subtests such that when D was low and A was low the original relationship obtained. When D was low and A high, the person behaved more like the original high D person. Gittinger called this apparent reversal of behaviors compensation. Learning is compensating for an originally weaker ability by providing what now would be called procedural knowledge (Anderson, 1982), learning acquired over long periods of time.

Over the next several years the system took on more variables and more interactions until there were three clusters of subtests, three which are called "primitive" modified by two additional, interacting levels, the second level called basic and the third surface, primarily of increasing specific content of the subtests.

High or low refers not to a high or low scaled score on the Wechsler, but to a concept Gittinger calls Normal Level (NL). None of the usual measures of central tendency capture his use of NL to determine what are high or low subtests. Saunders (Krauskopf & Saunders, 1994) has been able to generate a formula to predict what Gittinger would say is the NL of a profile which correlates +.89 with Gittinger's judgment.

The Clusters

Internalization-Externalization (I-E)

The first cluster is the three subtests of the Wechsler which describe Internalization-Externalization. The first is D which Gittinger considered "Primitive". Primitive means that it is governed more by neural structure than by learning, although some learning seems necessary. Without extensive learning, development reduces to simple maturation. The second subtest is A which seemed to require both internal and external capacity. Gittinger calls this level basic. Basic level requires very long term learning and is determined by the demands and rewards in the environment. The basic level stabilizes in late adolescence. Up to this point learning has been primarily directed by others. The third level is called surface, but has also been called ideal. At this point awareness and volition of the individual begin to play a larger part in development. The Information (I) subtest is used as an estimate of development at this level. Failure of either the external capacity to notice things or the internal capacity to store information in a retrievable form will result in a low score on this subtest.

Regulated - Flexible (R-F)

The second cluster is called Regulated-Flexible (R-F) and refers to style of learning. Regulated people are more able to learn by rote, by bits and pieces. Flexible people are more able to see relations among things. Regulated people have been described as needing to learn in order to understand and Flexible people as needing to understand in order to learn. Flexible people are more likely to see the forest and regulated people the trees. This dimension is closely related to Witkin's (Witkin, Dyk Faterson, Goodenough & Karp, (1962) dimension of Field Independence/Dependence. R-F is related to emotional sensitivity as well as learning. F people more readily see connections between people and their environments. R people, seeing one element at a time are more likely to accept the world around them as it is. Block Designs (BD) is the primitive indicator for this dimension. R people seem to begin the test by identifying a block's place in the design and proceeding to the next one rapidly. F people seem to lose speed by needing to consider the whole design determining what to do with an individual block. For PAS purposes the Wechsler Bellevue I was probably better than subsequent versions because F's spent time wondering what to do with the additional colors on the original Kohs blocks. Similarities (S) is the basic level subtest. For an R it represents a learned capacity for seeing relationships. For an F a good score is more easily obtained. For the F a low score represents learning the behavior more typical of the R to the point of damaging their capacity to see relations. The Surface level subtest is Comprehension (C).

"Flexible people are more likely to see the forest and regulated people the trees. "

Role Adaptable - Role Uniform (A-U)

The third cluster, Role Adaptable (A) and its opposite, Role Uniform (U), refers to a persons skill in meeting demands others make of him. It is an output dimension and the behavior is done without awareness. At one extreme is the A, charming, moving easily in many situations, always making a good first impression. At the other end is the U, at best able to handle only a few social roles and often socially inept. The behavior is exhibited mostly in new social situations for in familiar situations the U may appear very comfortable and accepted. The primitive measure in this cluster is the Picture Arrangement (PA) subtest. The basic level measure is Picture Completion (PC). These subtests have been connected with social intelligence, but PA has more often been called planning ability. The surface level subtest is Object Assembly (OA), which has been interpreted as a measure of social anxiety.

"The behavior is exhibited mostly in new social situations"

Gittinger assigned a special role to the Digit Symbol (DS) subtest. It has carried several labels over the years, none of them really satisfactory. Most often it has been called "Activity Level", or some kind of available psychological energy.

Interpretations, or personality descriptions, were worked out for each of the 512 patterns resulting from combinations of the three clusters. One can be either I or E at the primitive level, can continue or reverse the direction at the basic level and at the surface level giving 8 possible arrangements (2x2x2=8) for each cluster. Combining with the other clusters yields the 512 patterns (8x8x8=512). DS is treated as a moderator variable. The interpretations (Gittinger, 1964) were written as a combination of Gittinger's insight and experience, aided by sorting a data bank of profiles by pattern and searching for common behaviors. The data bank at the time the descriptions were written was approaching 10,000 protocols. Associated behaviors ranged from complete case histories to nothing. The cases came from Gittinger's files, those who worked with him, and many people doing research with Wechsler scales who were willing to share data. The descriptive Atlas contains, in addition to pattern descriptions, a description of behaviors which ought to be associated with each trait.

People found it uncanny to listen to Gittinger interpret a Wechsler profile, often using only the subtest scores, age and sex. After the descriptive Atlas was published, it was also sometimes startling to read the pattern descriptions and compare them to characteristics of known people. In fact, this is the process by which some people of a more scientific turn of mind got interested in the PAS. This brings us to the next story.

"People found it uncanny to listen to Gittinger interpret a Wechsler profile, often using only the subtest scores, age and sex"

The Second Story

Many psychologists who come into contact with the PAS dismiss it immediately. For example, Eysenck (199 ) says he could find no reason why anyone should believe it. Kaufman (1990) merely says he does not like it. Robinson (1985) refers to its near miraculous correspondence with the WAIS, then proceeds to apply the interpretation of uncompensated I's and E's to a highly compensated sample of medical students, before saying it does not work. Many who dismiss the PAS quickly mention Cohen's (1952; 1957) factor analysis which yielded 3 factors and point out that this interpretation of Wechsler scales would require 9 or 10 factors.

Some psychologists, however, did respond to the challenge of the question, "How does he do that?" This response was always after seeing Gittinger interpret profiles, usually from persons who were not known to Gittinger, but were well known to others. D. R. Saunders was one of the early ones who responded. He was aware of Cohen and the common belief that there were only 3 factors in Wechsler scales, but intrigued by the interpretations. He then gave several tests himself to people Gittinger could not possibly have had contact with and was sort of captured by the resulting session where Gittinger said things about these people that there was no way he could have known other than getting some clue from the tests.

Dimensionality

One of the first things Saunders did was consider what might be the problem with the factor analyses of Cohen (1952; 1957) and others. If Gittinger were doing what he said, there must be many more than 3 factors. Using Thurstone's (1947) criteria one always can obtain a perfect fit between factors and correlations in 7 factors or fewer when starting with an 11x11 matrix. There was one factor analysis of Wechsler scales which did obtain more factors. Davis (1956) expanded the matrix by adding variables, for example the MMPI scales. There are other ways of expanding a matrix from Wechsler scales. One can use split half scoring, or item level factoring. Saunders employed both of these methods (1959; 1960a; 1960b). His results were in reasonably good agreement with the PAS.

Table 1 is the result of a factor analysis done as an example. The sample is 46 high school students, 16 and 17 years old, tested with the WISC III as part of an evaluation for emotional or learning disabilities. To expand the matrix the simple device of subtracting an average of the subtests from each subtest score was used and the results put in as additional variables. Digits forward and digits backward and the identification number were also entered. An orthogonal extraction with quartimax rotation was used. The result is 8 factors which meet the common eigenvalue criterion and a scree plot suggests 10 or 11. The factors are in good agreement with the PAS use of the test. Arithmetic is an exception and may result from sample dependent variance with this sample generally scoring low with small variance. Other factor analyses show Arithmetic dominating a factor.

Table 1
A Wechsler Factor Analysis Rotated Factor Matrix
Variable F1 F2 F3 F4 F5 F6 F7 F8
D 95 04 04 13 01 09 07 03
A 68 26 06 15 40 -07 -30 -26
I 14 93 18 08 13 01 -14 05
BD 36 20 24 80 13 14 06 24
S 43 50 55 -32 -16 12 08 18
C 15 20 92 14 17 00 -05 07
PA 12 02 03 18 20 92 01 14
PC 21 28 28 20 77 19 05 25
OA 28 24 19 31 33 11 -09 74
DS 01 -05 07 13 14 13 93 -11
D-X 70 -37 -35 -11 -29 -12 14 -17
A-X 43 -04 -26 -05 25 -32 -35 -53
I-X -25 88 -09 -13 -08 -19 14 -11
BD-X 06 -09 -01 93 -08 -01 14 16
S-X 05 21 31 -67 -50 -09 16 01
C-X -28 -13 88 -07 -06 -22 -03 -12
PA-X -25 -31 -26 -01 -01 86 05 -01
PC-X -30 -09 -04 -03 82 00 15 12
OA-X -16 -12 -15 15 15 -11 -08 87
DS-X -26 -27 -14 -02 -03 -01 88 -11
ID 35 11 01 -08 -11 51 16 -27
Dbak 80 -10 01 -08 -11 03 -33 09

Note: Verbal subtests are indicated by first letter, Performance subtests by first letters of each word. ID is the case identification number. Dbak is digits backward. X is an average of all subtests.

Using this same sample with the usual 11x11 matrix, a principal components extraction and varimax rotation yields three factors by the eigenvalue criterion and a scree plot also suggests three factors, resembling the usual published factor analytic results from Wechsler scales.

Saunders also factor analyzed the Japanese version of the WAIS (Kodama, Shinagawa & Indo, 1964) with results similar to his factor analyses of American samples (Saunders (1959). The most extensive factor analytic study was by Klingler and Saunders (1975). They used item level factoring of nine subtests (Digit Symbol and Vocabulary were omitted) with a sample of 980. They found a 15 factor solution necessary to account for the item intercorrelations. Factor analyzing an 11x11 matrix of Wechsler scores just does not take seriously the hypothesis that there may be more than three factors.

All of these factor analyses have some degree of fit with the PAS arrangement of the subtests. The larger and more varied samples have the best fit. What the PAS calls primitive subtests are generally factorially more clear. Information, a surface level subtest has items that load on 5 or 6 factors. The conclusion must be that we have the dimensional complexity to support a scheme like the PAS, and that its more fundamental dimensions have good support.

Early Validity Studies

William Thetford and Helen Schucman at the College of Physicians and Surgeons, Columbia University were early researchers interested in the PAS. They initiated a program of research more like orthodox validity studies, developing hypotheses from the PAS for testing. An early study (Schucman & Thetford, 1968) reasoning from the fundamental hypothesis was that conversion symptoms should interfere with preferred modes of cognitive functioning. They studied two symptom groups, pronounced psychomotor symptoms and severe headaches. The two psychomotor symptom groups each had significantly more Externalizers than control groups. The two severe headache groups contained no Externalizers.

Another interesting study by Thetford & Schucman (1969) concerned the acceptance of the PAS personality descriptions by the people tested. They concentrated on the R-F dimension and using the PAS idea that compensated people generally are not aware of such compensation and are more likely to think of themselves as having the characteristics of their basic pattern, hypothesized that those testing R without compensation would be more likely to select a description of themselves containing theoretical R characteristics, but those whose tests showed compensation toward F would select F descriptions as being more like them. Uncompensated F should select F and compensated toward R should choose R. There were not enough F compensated toward R in the sample, but otherwise the selections were as predicted. These results were replicated by Mojonnier (1971). The studies are, at least, some indication that people are not just seeing themselves as a result of the "Barnum effect".

Another idea that seems reasonable is that if a test predicts behavior, then the reverse should also obtain. York (1963; 1994) has provided a study on this point. He recruited a sample of people he had some chance to observe and who were willing to provide Wechsler scores or to be tested. His next step was to predict the direction of score deviations from NL for each subject. He then compared his predictions to the tests with significant results. He also speculated on his mistakes, which were more on the A-U dimension than the others. He concluded that he had been reluctant to assign the U direction both to females and to people he liked.

The PAS assigns no value to any of the patterns, but many people do. Psychologists value interpersonal skills. Many years of training applied psychologists has shown me that they place enough positive value on the A descriptions that they are reluctant to accept U descriptions of themselves. Some go so far as to believe that there is something "wrong" with people who test U.

If some forms of abnormal behavior are exaggerations of normal characteristics, then some abnormal behaviors should be concentrated in appropriate PAS patterns. At least one variety of schizophrenia is characterized by behavior which the PAS would describe as extreme IRU. That is they are withdrawn from their environment, perseverative and non-responsive to people. Goodnow (1961), using a version of the Wechsler standardized for Chinese, compared a sample of Hong Kong Chinese diagnosed as schizophrenic with a sample of Chinese workers and a sample of Chinese students. Both of the normal samples were significantly different from the schizophrenics. 100% of the schizophrenic sample were I, 95% were R and 70% U.

Cohen (1955) had a group of experienced clinicians attempt to differentiate the broad categories of schizophrenia, neurosis and brain damage using Wechsler scales. In this study the clinicians were using whatever their experience or their own theoretical positions taught them, and the results were very discouraging to those interested in Wechsler pattern analysis. Saunders wanted to see if a more coherent theoretical position, such as the PAS, could do better. He took a separate sample of 125 schizophrenics and rank ordered the 64 possible PAS basic patterns by the number of schizophrenics in them. Then the 200 schizophrenic and neurotic cases of the Cohen sample were analyzed for PAS patterns, and ranked according to the ranks of the new sample of 125 schizophrenics. After this ranking the top 100 of Cohen's cases were found to have twice as many schizophrenics as neurotics. Beyond the ranking no attempt was made to optimize indicators (Saunders & Gittinger, 1968). It would seem that there is enough information in Wechsler profiles to make valid classification into such a rough dichotomy.

My own introduction to the PAS was similar to Saunders experience. Perhaps I was prepared to listen because I had been speculating on aptitude-personality connections. However, I too had read Cohen (1952; 1957). Saunders offered to send me interpretations if I would send WAIS protocols. The interpretations he sent were from Gittinger's Atlas (1964). They were impressive. For example, more than half mentioned alcoholism as a potential problem. The cases were, in fact, hospitalized alcoholics. The only information available for interpretation was the scaled scores of the subtests, age and sex. There was no information about who the people were or where they came from.

Reliability

In addition to the dimensionality problem discussed above, there was another fundamental problem to be addressed, reliability of the subtests. Using the published reliabilities and the significance testing procedures commonly recommended (McNemar, 1957; Silverstein, 1982) a Wechsler profile does not often have more than two or three subtests that should be interpretable. this is a problem for both the clinician and the scientist. The case for interpretation has been argued more extensively elsewhere (Krauskopf, 1991), and is only briefly presented here. Both Type I and Type II errors deserve consideration. The conventional view considers only Type I error.

For the clinician balancing the two results in a difference of a subtest from an average of all subtests approximately that recommended by Wechsler (1958). Cahan (1988; 1989) has argued that the probability of the null hypothesis being true when analyzing a WAIS profile is so low that all such testing should be abandoned. My argument is that the clinician is better off basing hypotheses about a client on a 10:1 probability of a real test difference than to act as if there were no information at all.

For a scientist happy ignorance is not a desirable state, yet indiscriminant tolerance for Type II errors facilitates that state.

Life Circumstance

An embarrassment of research directions seemed needed. More validity studies like the ones Shucman and Thetford did were needed. Making connections to other psychological research and theory seemed highly desirable. And, many of these were done, but another direction made possible by the large data bank that had been collected was what I call the Life Circumstance hypothesis.

If there is preferential choice by personality pattern, and if people have a reasonable degree of free choice in what they do, or alternatively if there is forced distribution based on behavior, then there should be higher proportions of patterns which are appropriate to the dominant activities, or to the sanctioned behaviors. An example is in Krauskopf & Saunders (1994) in the tables based on the standardization sample of the WAIS (Wechsler,1955) and another table based on college students. The differences are highly significant. C hi-square is more than 30 times the degrees of freedom, and this without using IQ or NL. One of the differences that should appear is that uncompensated and unmodified E, being a very unintellectual pattern, should be, and is, very rare in the college students. This pattern is much less frequent in college students than in either the standardization sample or in prisoners. The Life Circumstance hypothesis is not really a new idea. For example, Holland (1985) uses it to describe the "personality" of occupations.

The schizophrenic studies above are also Life Circumstance studies. If people are called schizophrenic, then there should be some restricted range of patterns on any multiscale measure. Another example in Krauskopf & Saunders (1994) is imprisoned felons. The data bank has several samples and they are individually and collectively different from the standardization sample. What PAS would call psychopathic derives from a primitive IRA pattern, is largely uncompensated, especially on the A-U dimension. This pattern is heavily overrepresented in imprisoned populations.

Saunders has proposed a more sophisticated, and perhaps more persuasive, form of the hypothesis. That is if two or more people show test patterns that are highly alike, then there should be some salient behaviors that are alike. We agree with Sines (1964) proposed that this should be the direction of the best validity studies, form homogeneous groups with variables thought to be important and then look for the similar behaviors. Sines did a d2 sorting of MMPI patterns and discovered a group of highly homogeneous profiles for which the salient behavior, for 80% of the sample, was violence. Saunders calls his method Reference Grouping (Saunders 1989; Krauskopf & Saunders, 1994). This method takes some advantage of the large collection in our data bank. Briefly, the method is to factor analyze a group of profiles which have some degree of resemblance, using a factor score "key", compute the distance of each profile from the group centroid. The farthest cases are discarded and the process is repeated until a highly homogeneous group is left. Then a larger collection can be scanned for possible additional members.

One of the first groups was called ERA Psychopath. There were 84 members, half of which were imprisoned felons. Others were from a juvenile detention center, five mental hospitals and several other places. Hospitalized cases contained notations such as, sociopath, character disorder, larceny, drunkenness and so forth.

Other groups that have been discovered include, Applied Scientist, Rapists, Professional Generalists, several kinds of alcoholics, and Performing Artists among many others. The proportion of people with the pattern, and not in the originally formed group, showing the common salient behavior can be turned into an approximate correlation by taking the square root of the proportion. The Applied Scientist group yields a correlation of +.75.

Before 1980 Saunders had at least one reference group for each of the 64 possible basic level PAS patterns. Some of the groups showed prominent abnormal behavior, some normal. A few had both. Gittinger had said very early that it was difficult to look at a Wechsler protocol and tell if someone would be diagnosed as abnormal or disturbed, but if it were known that there was a problem, he could tell what kind of behavior it would involve. The reference groups that contained both normal and abnormal cases seemed like an empirical demonstration. One of the groups was labeled Professional/Paranoid. It was not that these people made their living being paranoid, but that the group seemed either to have responsible, professional occupations, or to be in some sort of difficulty over paranoid behavior. Many of us think that at sometime we worked for one of these. These dual reference groups are examples of diathesis-stress.

Finding at least one group representing each of the basic PAS patterns, combined with the observation that frequencies of people across PAS patterns is very uneven suggests that PAS patterns are not any form of multivariate normal distribution, or the result of any random process. Such a lumpy distribution suggested to Saunders the possibility of creating a Reference Group scheme in which every profile could be placed. He has been working on this problem for several years with some success. One indication of progress is the hypothesis that in such a lumpy, multivariate distribution, randomly generated profiles should be harder to place than those generated by a real person. Therefore, it should be possible to tell the difference between real profiles and randomly generated profiles. With another person generating 150 randomly rearranged profiles, he undertook to differentiate real people's profiles from an unrestricted random arrangements of subtests. That is the subtests were randomly relabeled. This was done with 80+% accuracy. Then a sample was generated with the restriction that the variances of the subtests should be preserved. In this Saunders achieved 67%. When the intercorrelations among the subtests were preserved it became harder to identify the otherwise randomly arranged profiles (Krauskopf & Saunders, 1994).

This is interesting evidence. First, it would not have been possible to make the Reference Group arrangement for all cases without a good place to start, which the theory provided. The theory also provides behavioral expectations for resulting groups. Second, the way of identifying the random profiles would not have been possible without the sorting scheme for real profiles to compare it to.

Predictions

Any good theory should be able to make predictions. With the PAS this is complicated because with the explicit consideration of interactions, it should be a rare behavior which is influenced by one variable sufficiently to test many of the variables directly. One possible method we have used is to make predictions using the whole profile. An early study (Krauskopf & Bielefeld, 1981) was to test an entire class in psychological testing and to predict the rank order of their grades in the class. The rank order correlation of predictions with grades was +.51 (p<.05). Numbers were too small to try to generate actuarial rules for the predictions, but the correlation of IQ with grades was +.03.

Another whole profile study was Fallahi (1989). She predicted the most frequent problems students would pick on the Mooney Problem Check List as applying to them. There was no interest in the Mooney other than it was a list of real problems that real people would acknowledge. Two people made the predictions for a sample of 160. One did an additional 52 for which the data was less complete. Both predictors did significantly better than chance. They also were more accurate than identifying the three most frequent categories college students endorse and always selecting them.

Connections with other Differential Psychology

Two studies have been done connecting the PAS to Witkin's concept of FI/D, Frank, (1969) and Johnson, (1970). Factor analyses including Rod and Frame and Embedded Figures tasks from Witkin and the PAS suggest that the PAS R-F dimension is closely related to FI/D. Witkin had recognized this and indicated willingness to use Block Designs as an alternate measure. There is some suggestion that the PAS principle of compensation might help sort out some of the confusing relations of FI/D with personality.

I-E has been studied in relation to several other ideas of extraversion, Eysenck's measure (Sell, 1971), Myers-Briggs Type Indicator (Austin, 1983; Saunders, 1982), Cattell's 16PF (Saunders, 1982; Turner, Willerman & Horn, 1976). Wagner (1967) used several I-E measures including the PAS. All are positively related, but all are somewhat different from each other.

Jackson & Krauskopf (1986) factor analyzed two social self instruments, Fenigstein, Schier & Buss's (1975) Self-Consciousness Scale and Snyder's (1974) Self-Consciousness Scale along with the PAS and found some expected relations with the A-U dimension. Primitive A is negatively related to social anxiety and unrelated to self-consciousness. Self-consciousness is more related to primitive F. A-U is related to Self-Monitoring, primarily in the quality and quantity of social relationships (Gabrenya & Arkin, 1980) and to using others behavior as a reference point.

Structural Speculation

The above mentioned studies are a sampling of studies that have been done. Over many years we have done some speculation about what it means to have a "primitive trait". Saunders suggested that if we separated abilities into aptitudes and achievements, we could make some hypotheses about distributions in large samples. After O'Conner, (1928) an aptitude is defined as an ability whose development is mostly controlled by maturation and later by deterioration beyond the control of the individual and the environment. An achievement is an ability that is dependent on the experience of the individual, either the striving of the individual for some goal or the press of the environment forcing particular sorts of learning. The distribution hypotheses were that the more primitive aptitudes would be closer to normal distributions and that achievements would be mostly negatively skewed because an "overachievement" in one area would be at the cost of disproportionate underachievement in some other area, or areas. Using a sample of 12111 he found that departure from normality was less on those subtests the PAS considers primitive, D, BD, PA and DS (Saunders & Gittinger, 1968).

Until recently the psychology of learning was not very helpful because there were very few studies of learning that extended over more than a few hours. Cognitive psychology has proposed some useful ideas. What we are considering the basic level of personality organization very much like what they call procedural knowledge. Learning related content, not only results in the learning of the content, but the experience teaches methods of learning more related things (Anderson, 1982). PAS theory suggests an even deeper layer of rule learning than that studied to this point. Saunders and I have predicted that as cognitive psychology develops they will discover such a level and it will contain more general problem solving rules than what is now called procedural knowledge (Krauskopf & Saunders, 1994, p. 84)

We have not systematically searched, but we have noted some work in psychobiology which relates to PAS. For example Mirsky (1987) has located brain pathways active in digit span tasks in the reticular formation and forebrain. Using Mundy-Castle's data (1955), then a cross validation sample Saunders (1960) found a positive correlation between alpha frequency and Digit Span score.

Frequently asked Questions

  1. You are talking about genetics and genetics cannot be changed. How can I believe in such a hopeless theory?

    The principle of compensation is how the effects of genetics are changed, even if one cannot change the genetic endowment. Behavior genetics does seem to be enjoying increased popularity. Buss' (1991) review of recent activity in this area does show how person-situation interactions can be explained. If we were to speculate in a way similar to the way Buss talks about sexual selection, we would think about how PAS characteristics could have survival value in a group living, hunter-gatherer animal. Variety on the dimensions we propose should have survival value.

  2. PAS seems tied to the Wechsler scales in a "near miraculous" fashion. Is the theory tied to the Wechsler scales?

    No. Wechsler's wide ranging idea of intelligence provided subtests which do seem to provide measures of important individual differences, and were instrumental in the beginnings of the theory. However, they also provide information on possible expansion of the variables considered (cf Klingler & Saunders, 1975) which might also be used to construct a better intelligence test. Saunders (Krauskopf & Saunders, 1994) has used factor analytic data to search for other possible variables to fill "empty space". He has proposed adding a version of the Stroop procedure (MacLeod, 1991) and a time estimation task to PAS measures. We have (Miller & Krauskopf 1986) found a pattern of these two measures and Digit Symbol that relates to "Type A " personality. Jackson's Multiple Aptitude Batter, which is keyed to the Wechsler does provide an alternative for compensation and modification variables, and possibly two of the primitive variables. The Woodcock-Johnson is also contains a wide variety of aptitude and achievement measures and might be tried. Ideally, we would be able to construct a test specific to the PAS.

    Other measures have been tried. The most promising at the moment is Jackson's Multi-aptitude Battery (1989). Questionnaire measures have been tried, none of which are satisfactory. They violate the basic assumption of the theory, which may be why they do not work well.

  3. Can I do diagnosis with PAS? Can I determine DSM categories?

    No. PAS dimensions are related to the rich variety of normal variation. Some DSM categories may be exaggerations of normal characteristics and may relate well to PAS. Some other kinds of abnormal behavior may be reflected in aptitude and ability patterns, but we have had only a little success in trying to match DSM. For example, Konar (1985) found little relationship with official diagnosis in his sample, but clear relations with the problematic behaviors that brought the children to the hospital in the first place.

  4. PAS seems to violate the law of parsimony. How do you justify that?

    The law of parsimony actually says that one should not unnecessarily complicate explanations. People are complex creatures and need a complex theory to explain their behavior. We would like to keep explanations as simple as possible, but so far we have not found anything to throw out.

  5. Doesn't the "Big 5" idea show that you don't need so many dimensions?

    The Big 5 theory (Digman; Goldberg; Costa & McRae) summarizes the factors that account for a large proportion of the variance found in self report instruments. There are some resemblances with PAS. There is an I-E cluster. The added stress dimension resembles neuroticism. However, the main difference is self report. While people are able to report much about themselves, it is likely that there are aspects that are not available to conscious awareness. Gough has reported on the difficulty of finding subtle items for personality inventories. The studies by Thetford and Schucman (1969) and Mojonnier (1975) suggest that people sometimes pick what they would like to be.

  6. PAS is a trait theory. Hasn't trait theory been discredited?

    We think the ills of trait theory are more related to researchers tendency to postulate a trait for about anything that might be measured. Trait theory seems to us to be the theoretical style most consistent with biology which is at the base of all behavior. We do believe that situations (the environment) must be considered when making behavioral predictions, but it is desirable to distinguish person characteristics from situation characteristics.

Some Implications

If PAS theorizing has some validity to it, there are implications for assessment, treatment, education, child rearing and self improvement at a minimum. For example, beyond assessment the strategy for treatment of behavior disorders would pay more attention to the surface level, or what a person might be striving for. Our ideas of connecting this kind of learning to current cognitive psychology should change cognitive therapy from how one talks to oneself to how one adds to long term learning and problem solving styles. In education we think that the several changes in mathematics education over the past 50 years have been making math easier for one kind of child and making it more difficult for those who found the former method easier. That is, 50 years ago R children did not have much trouble with the normal teaching style, but did have difficulty when understanding was stressed early. Trying to teach understanding benefited the F children. For child rearing, since we know from our life circumstances studies that people in the more advantageous places in our society, and groups judged to be psychologically better off, are compensated on the I-E dimension, we would suggest efforts to make this happen. Krauskopf (1995) has written about the implications for career counseling.

For assessment, we believe that we could now create a better strategy for making an intelligence test. For one "g" would be less important as a goal and a variety of aptitudes would be more deliberately assessed.

Summary

This is not really a summary, nor is it conclusions. I hope I have provided two interesting stories, neither of which has an ending. For those who wish to learn more about the PAS I recommend Bem's (1982) summary in the Nebraska Symposium for Motivation; Winne and Gittinger's (1973) Clinical Psychology Monograph; Saunders and Gittinger (1968); and Krauskopf and Saunders (1994).

The hypothesis that differential aptitudes are causative of much of the variety of behavior we call personality does seem a bit radical. But, by taking it seriously we have been able to gain insights into both intelligence and personality assessment, make some predictions and better understand people and the circumstances we find them in.

Copyright Sage Publications. Used with permission of the publisher