Weekly incidence data for influenza-like illness, routinely collected in sentinel general practices in England and Wales and in the Netherlands over 10 winter periods (week 37 in one year to week 20 in the next, 1987/1988-1996/1997) were examined in conjunction with matching virus isolate data to define epidemic periods of influenza in the two countries. We first defined the background rates of recording influenza-like illness which occurred at times when only sporadic or no isolations of virus were reported. The background rates were similar in the two networks with mean weekly incidence in England and Wales of 28.1 per 100,000 (all ages) and in the Netherlands 29.8. Epidemic periods defined as lying above the upper 95% confidence level of the background rate lasted on an average of about 10 weeks. Once epidemics were recognised, peak incidence was generally achieved within 4 weeks. The excess population (all ages) consulting general practitioners during influenza epidemic periods was calculated from the difference between the observed and background incidence rates, and expressed as a percentage of the total population. In the 10 periods surveyed, the percentage of the population consulting and diagnosed with influenza-like illness in England and Wales ranged from 0.4% in 1991/1992 to 1.7% in 1989/1990 and in the Netherlands from 0.5% in 1990/ 1991 to 2.1% in 1989/1990. The duration and epidemic periods were broadly similar in the two countries though the excess consulting population during the 10 epidemics studied averaged 0.85% in England and Wales compared with 1.39% in the Netherlands. There were substantial differences between the two countries in the impact of influenza in individual years, as measured in the consulting population even though the predominant virus (sub)types were similar.