B. S. Everitt's excellent article on cluster analysis (Journal, February 1972, p. 143) provides a much needed warning of the difficulties involved in the use of this technique. One way, however, of avoiding at least some of the pitfalls of this method needs to be stressed. This is the application of what is perhaps the universal panacea for scientific flights of fancy, namely, common sense. Common sense is particularly applicable to the choice of variables to be measured, and, of course, to the problem of naming the groups once they have been found. The use of mathematical techniques without the concurrent application of common sense is one of the greatest traps for the unwary and the overenthusiastic.

The reply of Morgan and Compton in this issue of the Journal (pp. 433–6), is based on a misunderstanding of our results and of the problem investigated. This leads them into a refutation of ‘claims' never made and they buttress it with a statistical exercise of great naivety. Our findings were:

(a) ‘… in certain important respects in-patients and out-patients are derived from different though overlapping populations.’ The most marked differences (dismissed by Morgan and Compton as 'slight’) were found among the elderly. ‘The admission rate for the over 65s of both sexes was 4 · 90 per 1,000. In contrast, increasing age was associated with a gradual fall of out-patient referral rate to 1 · 60 for the over 65s.’ We did not claim to have demonstrated the cause of these differences, but mentioned possible reasons for them.

(b) In a district general hospital-centied psychiatric service we observed a 34 per cent increase of new out-patient referrals while hospital admissions remained static; this occurred over a period of three years, when nationally hospital admissions were still rising. We quoted in illustration some extreme figures from official statistics and the figures for the Sheffield and the Liverpool regions were extreme whether one refers to Tables 3, 11 or 13 (1). This prompted the hypothesis that an increase in new out-patient referrals can prevent admission or prove an alternative to it. No ‘claim’ was made, and we advisedly used the term 'suggestion’ to emphasize that these are ‘no more than tentative and incomplete incursions into this difficult and relatively unexplored field’ (2).