The present investigation involves 600 personality questionnaire items. The 300 Guilford items comprise 78 marker clusters for 15 Guilford factors; the 300 Cattell items represent marker items for 17 Cattell factors. The study involved two major analyses. In the first, the 600 x 600 matrix was factor analyzed by the Principal Factor Method, extracting 18 factors for one rotation, and 15 factors for a second. The second analysis involved two steps. Step one consisted of separate factor analyses and rotations of the 300 x 300 item Guilford and Cattell matrices, in each case including factor loadings for the "other" 300 items. Step two was designed to determine the factors in each matrix after removing the effects of the factor loadings of the "other" set of items in that matrix. All rotations were made by both Varimax and Promax computer programs. Coefficients of congruance were computed between all rotated factors in all analyses. The correlations among items and the factored and rotated results demonstrated, beyond question, that analysis at the item level is highly destructive to the factors previously assembled with inadequate concern for their loadings in large matrices in which a wide range of factors is known to exist. If the results are accepted, the indications for reclassification of at least 400 of the 600 items included in the study are obvious. (IM)