PSYCHOMI~rRIKA~VOL. ] 2, NO. 3 S~SER, 1947 CONFIRMATION AND CLARIFICATION OF PRIMARY PERSONALITY FACTORS RAYMOND B,. CATTELL UNIVERSITY OF ILLINOIS In connection with a study bridging rating, questionnaire, and objective test factors, confirmation was sought with respect to the twelve personality factors previously found for young adult men. Variables were chosen to clarify and discriminate the nature of re- lated factors. Ratings of and by 373 students were obtained, and the present study describes the separate factorization for the 133 men among them. Factorization yielded eleven factors, of which, on "blind" rotation for simple structure, 9 or 10 proved to be identical with those of the previous study. A new factor M is described. 1. Perspective on Plan of Integrating Research This study is one of a coordinated trio directed to (a) clearing up the number and nature of personality factors in three fields: 1. Behavior rating data (behavior in real life situations), 2. Ques- tionnaire data, and 3. Objective test data; (b) determining the rela- tionships among the factors in these three fields, to test the hypothe- sis that they are actually the same factors outcropping in different media. This hypothesis supposes that the factors so far discovered (3) are in some sense functionally unitary personality traits, which mani- fest themselves alike in the total behavior (Behavior ratings or BR factors), in laboratory test situations (Objective tests or T factors) and in questionnaire self ratings (Q factors). The Q factors may be considered "mental interiors" -- the view of the external behavior (BR and T) factor patterns as seen, or rather "felt," by the subject. The bridging of BR, Q, and T factors, which has not previously been at- tempted*, demands that the same subjects be used for all three series of testings, a requirement which, because of its exactions in time, is not easily attained. In the present research we started with 400 stu- dentt subjects for the BR and Q studies, who dwindled somewhat * There have, of course, been numerous studies bridging in terms of simple BR, Q, and T variables, but not in terms of matching definite unitary traits. t Thuse were students in a state university, taking a rather large section of the population, swollen still further by the first returning veterans; so that the study is, fortunately, not open to the usual criticism of the use of student sub- jects, namely that they represent an extremely small and biased sample of the general population. Certainly. in spite of the comments of teachers, these subjects are somewhat selected in intelligence and probably to a lesser extent in character; but not enough to do more than alter the relative variance due to different factors. 197198 PSYCHOMETRIKA before the T studies were finally completed. The present article de- scribes the BR factor research only, the remaining researches being in press elsewhere. 2. The P,rimacy of Behavior Rating Studies The behavior rating study is considered the keystone of the present research design. It is of no avail to establish factors in ques- tionnaires or laboratory test variables unless such factors are rela- ted to the total life situation and the gross behavior patterns known to be of importance in every*day life. The choice of variables in this latter realm can be made truly representative, through use of the personality sphere concept*, whereas the choice of variables in tests may lead to general factors which are anything but general factors in the ~otal behavior, occupying, indeed, some extremely narrow seg- ment. Some psychologists have regarded the use of ratings as suspect, and rightly so in the uncontrolled fields of applied psychology in which they have been most used. But no apology is necessary for their use here, because (a) As indicated above, only ratings on total behavior i/a situ will give us the true frame of reference we desire as a founda- tion for studying the total personality. (b) We propose, as soon as the rough scaffolding of factors in ratings is obtained, to use this as a means to set up and proceed to factors in objective tests measuring these same factors. (c) Granted proper planning of the rating situa- tion, the conventional objections ¢o ratings, i.e., assessments of be- havior in real life situations, in the writer's opinion, no longer apply with any force. The proper planning of rating, as discussed elsewhere (1, 4, 8), requires such essentials as that the raters live with the ratees; have no official relation to them; are not asked to rate more than about twenty people, because more cannot be known well; deal with all sub- jects at one time with respect to one trait and not all traits with re- spect to one subject; have traits behaviorally defined and illustrated; have the ratings pooled for at least ten judges to cancel individual perspectives. When these are faithfully observed, the only remaining source of systematic error is halo or stereotype. This effect, though much * For discussion of this concept see (3, p. 215), whence the following quotation: "The personality sphere may be considered to represent a complete surface, con- stituted by many small 'trait areas' . . . the whole constituting an endless but finite continuum of. behavior meaning... It is part of our assumption, from the study of language, that no large area remains completely neglected by existing (trait name) vocabulary." The sphere has the space dimensionality of the prin- cipal factors required to describe personality.RAYMOND B. CATTELL 199 talked about, has never been demonstrated in trained raters, and in even untrained raters the alleged halo is open to the alternative in- terpretation that desirable qualities really do go together. Sociolo- gists have demonstrated an occasional stereotype (but not in person- ality, apart from superficial cultural, class, or national patterns) ; yet they have begged the question as to how the stereotype arises if it is not based on experience. The stereotype of