To the Editor:Thirty-seven of the 38 graduate students of psychology enrolled at the University of Tennessee during the Kail Quarter of the aeademic year 1948-49 took both the Miller Analogies Test, Form G, and the I. E. R. Intelligence Scale CAVD, Levels JVI, N, O, P, and Q, Form 3.Two criterion measures were used to check the validities of these tests. The first was the grade-point average in all courses taken while enrolled as a graduate student at the University of Tennessee. This was computed by assigning the grade "A" a point-value of 4; "B", 3; "C", 2; "D", 1; and "F", 0. The grade-point values were averaged, weighting each by the number of credit-hours, and keeping one decimal.The second criterion measure was an overall rating completed by each member of the department who was teaching graduate courses in psychology. The instructions for rating were as follows:"The list, which accompanies this form includes the names of all the present graduate students in psychology. You are requested to rate all of them whom you know well enough, in terms of their overall abilities as graduate students."1. Cross out the names of all the students you do not feel able to rate: fairly and with reasonable accuracy. Do not feel called on to rate all the students whom you may lie teaching in class. The class situation does not always provide close enough personal acquaintance for rating purposes."2. Consider the remaining names. On the blank spaces at the right, write these remaining names in order of ability. On the first line write the name of the ablest student among those you are rating, on the second line the name of Ihe next ablest, etc."3. Draw a heavy line under the name of the last student in your list whom you consider a suitable candidate for the Master's degree. If you believe all of the students in your list are suitable candidates, draw this line under the last name."4. Draw another heavy line under the name of the last student in your list whom you consider a suitable candidate for the Doctor's degree. If you do not believe any of the students in your list are suitable candidates, draw this line over the first name."The distribution of the number of ratings per student was as follows:NUMHKR OF RATINGS

No abstract

Different applicatiions require different systems of cluster analysis. The ways in which systems differ are pointed out. The present system was designed originally; to identify, in a homogeneous callection of questionnaire or inventory items or of tests, groups of items which can be scored as subtests, or groups of tests which can be combined to yield globa1-trait scores, but it may well be applicable to other problems.

No abstract