CRITICISMS OF METHODS 373

As is inevitable in a statistical survey, I have laid myself under a
deep obligation to the many persons who have co-operated. I am
indebted first of all to the London County Council for permission to
make use of material gathered while I was still in their service. To
the numerous teachers, social workers, and research students, who
have assisted me in collecting or in analysing the data, I owe a
heavy debt of thanks. For help with the arithmetical calculations
I am especially grateful to Mr. A. F. Roberts and Miss L. J. Carter,
to my secretary Miss G. Bruce, and to Dr. A. J. Marshall, Research
Assistant in the psychological laboratory.

Most of all, perhaps, I am indebted to my former colleague,
Dr. W. Stephenson, whose questions and criticisms are representa-
tive of those put forward by many other readers, and at the same
time possess the additional advantage of being based on an unusually
wide experience in similar fields of research. A frank expression of
disagreement is always more fruitful than the mere expression of
assent ; and Dr. Stephenson has been good enough, not only to
spend many hours in discussing these problems, but also to set down
his views on paper in the form of detailed letters and memoranda
which could be studied and discussed at length. In another
publication [138] we have endeavoured jointly to summarize the
several points on which we agree and disagree. Here it will be my
purpose to deal chiefly with outstanding points of disagreement
there disclosed (loc. cit., pp. 274-80), and, so far as I am able, to
defend more fully the position I there briefly outlined.

This, too, is an appropriate place to express my thanks to
Professor Godfrey Thomson for his sympathetic criticisms of my
method in his invaluable book on ?"he Factorial Analysis of Human
Ability. He has since been generous enough to suggest that the
differences between us may possibly be due, not so much to an
actual disagreement over fundamental principles, but rather to the
gaps that were almost inevitable in a brief preliminary description.
In the following pages the more glaring of these gaps have, I hope,
been filled in. Actually, the text of this paper was completed and
typed before his book appeared; and consequently a few of his
points may still remain unanswered. However, in the first of the
three papers in this volume (which was written last of all) I have
taken the opportunity to discuss what seem the more important
and more general of his criticisms (e.g. those relating to negative
saturations or loadings, the assumption of objective differences in
the variances, and the use of unstandardized scores). Hence there
is less need to take up these more theoretical questions here.

On the concrete issue—the distribution of temperamental types—