correspondence between case reports and fatality data. These data also establish that mortality rates are not affected by epidemic phase 24. Further confirmation of these results is provided by an analysis of the Aberdeen data (N.B.M-B., P.R. and B.T.G., manuscript in preparation). Concerning infection-induced mortality rates, classic work by Butler 24 , Bartlett 25 , Creighton 5 and others indicates significant mortality due to measles and whooping cough during these periods. Estimates of case fatality rates for measles vary widely, from 1-2% in the postwar era up to 46% prewar 14,26,27 , whereas estimates for whooping cough are in the 3-15% range 24. Data analysis These time series contain a substantial annual component and are further complicated by increasing population sizes over the two periods examined. Hence, analyses of the relationship between measles and whooping cough outbreaks were carried out on de-trended data. We used three separate methods. First, Pearson correlation coefficients were estimated for data aggregated over each epidemic year (October to October). Second, we carried out a linear regression of annual counts of measles against whooping cough and used the slope as a measure of synchrony. The results of this technique were qualitatively identical to those of the Pearson correlation, so we present only those. Finally, we also used Wavelet spectra to explore phase differences between filtered time series 28,29. Further information can be found in the Supplementary Information.

No abstract