Variation in the incidence of invasive pneumococcal disease across South and West England, in 1995, was measured through a survey of microbiology laboratories. A 100% response rate was achieved. The incidence by laboratory varied between 5·2 and 20·4 per 100000 catchment population (P<0·001). Adjusting for pneumococcal vaccine uptake rate in over 65 year olds, hospital admission rates, blood culture system used and for the age and sex structure of the population, did not account for this variation. When blood culture sampling rates were included in a logistic regression model, the variation between laboratories was much less and of lower statistical significance (P=0·019). Higher rates of blood culture sampling were associated with a higher incidence of invasive pneumococcal disease. Consistently high sampling should be encouraged because a higher diagnostic rate should result in more selective prescribing of antibiotics, and secondly because improved ascertainment of severe pneumococcal infections is a prerequisite for the evaluation of new pneumococcal conjugate vaccines.