As a part of a series of construct validity studies this study was designed to assess the convergent validity of the cognitive preference construct. Three measures of cognitive preference were administered to 71 college students and to a small sample of tenth graders. One instrument was a conventional cognitive preference test, the other two were designed to assess cognitive preference in common learning settings using student responses to a science reading and teacher ratings of relevant classroom behavior respectively. Correlations between the science reading and the conventional test showed strong convergence of the Q (questioning) mode of cognitive preference but rather weak convergence of the R (recall), P (principles), and A (applications) modes. A subgroup of students with significantly discrepant R, P, Q, and A score patterns, thus a distinct cognitive preference, showed higher correlations between instruments than the total sample, as one would expect in case of construct validity. Apparently the cognitive preference construct is a more valid descriptor for the cognitive style of some people than for others.