It is difficult to find a truly representative area of any city, since its districts vary greatly in socio-economic terms. Nevertheless, we made a careful selection in Rawalpindi of a recently established housing area, with poor public utilities, of middle to lower socio-economic status. As presented in our paper, the socio-economic findings confirmed our choice and revealed a wide social spectrum. For example, among men, 31% had had no formal education yet 38% had been educated to tertiary college level. Over half the house-holds had an income of less than 5000 rupees (currently worth £55) per month.

As it happens, the socio-economic status of the nearby rural population in our Gujar Khan study (Reference Mumford, Saeed and AhmadMumford et al, 1997) was quite similar to that in Rawalpindi in terms of education and income, and in fact they reported greater ownership of most electrical appliances. So financial prosperity alone is not a plausible explanation for the very striking difference we found in psychiatric morbidity (i.e. less than half) in urban Rawalpindi compared with a rural village in the Punjab.

Whether urban populations in Pakistan indeed have more “strong and varied social networks” than rural populations, as Dr Mirza suggests, remains to be investigated, but this is doubtful. We are planning further studies to determine to what extent the quality of life in the city, as opposed to selective migration, can account for the enormous rural—urban differences in psychiatric morbidity. Replication of our study in other cities in Pakistan would be very useful.

However, the truly remarkable finding is not the prevalence of common mental disorders in urban Rawalpindi, which is more in line with rates reported elsewhere in the world. It is the exceptionally high rate of psychiatric morbidity in rural villages in Pakistan, recently confirmed by other investigators in another village near Gujar Khan (Reference Hussain, Creed and TomensonHussain et al, 2000); this cries out for further research.

Regarding the use of male doctors to conduct the second-stage interviews of female subjects, we were obliged to do the same in the previous study in rural Chitral (Reference Mumford, Nazir and JilaniMumford et al, 1996). However we have found no psychometric inconsistencies between the three epidemiological surveys (in Chitral, Gujar Khan and Rawalpindi) to suggest that this was a source of bias while making psychiatric diagnoses according to ICD—10 criteria for research.