Abstract
In some crops of winter wheat selected from a range monitored in Western Europe during 1981‐83,Septoria nodovumand S. triticiwere spatially very uniformly distributed from the beginning of the growing season onwards. A cube root transformation produced a constant variance in lesion numbers per leaf, similar for both pathogens at about 0.5. This permits the sample size needed for a given accuracy to be estimated. Counts of conidia washed from leaves by a standard procedure had a constant coefficient of variation, independent of disease level. Large samples would be necessary for accurate counts, particularly if leaves from different layers were examined separately. The pattern of lesion numbers on leaves was best described by a negative binomial distribution: this predicts an incidence‐severity curve to which the data conformed closely. Hence incidence estimates can be used to estimate severity, which may be more economical of sampling effort. Correlations between lesion counts on different leaves of the same tiller were negative and highly significant forS. nodorumin 1982, but positive and significant forS. triticiin 1983. The causes of this difference are unknown.