P-, Q-, AND R-TECHNIQUES 197

the best order.1 Similar methods have been used by Schonfeld for
assessing and factorizing the ' sense of humour,' by Dewar for
pictorial appreciation, by Eysenck for other forms of visual art, by
Wright and Wing for musical ability, by Williams and Wood for
literary appreciation. In every one of these investigations it seemed
necessary to correlate persons rather than tests or traits.

What, then, in general terms, are the " conditions under
which it is necessary " to adopt this particular procedure ?
Since Dr. Stephenson thinks that I have overlooked
them, may I briefly refer to earlier passages (some of
them in the two articles he himself cites, e.g. [101],
pp. 61-6, [114], pp. 176-7; cf. also [130], pp. 416 f.) in
which I attempted to summarize them, both in general
and in detail ? A comparison of the common features in the
inquiries just enumerated will indicate the general nature of
the problems and purposes for which, as I argued, we seem
" compelled " to adopt this alternative procedure.

Grounds for Correlating by Persons.—Broadly speaking,
we correlate persons rather than tests when we are con-
cerned with the complex resemblances between total
personalities (or aspects of those personalities) rather than
with the more limited resemblances between particular
traits or their tests. And, whenever we desire to investi-
gate what I have called the * general factor for persons,'
these are the correlations that it is necessary (and not merely
convenient) to factorize. In psychological work this
general factor may be a cognitive, affective, or conative
factor, a combination of the three, or a sheer artefact; it
may be a factor of c reliability/ of c preference,' or * judge-
ment/ or of efficiency measured by any of these three, or a
mere ' halo-effect' ; in short, it may be any abstract
characteristic which is ' subjective' 2 in the sense that it

1 c A Judgement Test for Measuring Intelligence/ Mental Welfare, XX,
pp. 45-48. Here the correlation with independent assessments of the
examinees was not so high as Stephenson's (-67 to -73).

2 Stephenson takes exception to my " limiting inverted factor-technique
to problems of * subjective ' judgement." He prefers, as we have seen, to
use the more general term * significance.' Elsewhere he says " only the
psychologist can decide upon matters of significance ; but that by no means
implies that appraisal for significance is subjective " : for (i) the psychologist