Experimental data were used to test the hypothesis that the effective fungicide dose (ED) ± the dose required to achieve a given level of disease suppression ± varies in a predictable manner according to the pattern of development of the wheat canopy. Replicated and randomized field plots received a single systemic fungicide spray at either zero (control), 0´25, 0´5, 0´75 or 1´0 dose (the recommended dose), at one of eight timings from April to June. Wheat cultivars and locations for experiments were selected to promote epidemics of septoria tritici spot and yellow rust caused by Septoria tritici (anamorph of Mycosphaerella graminicola) and Puccinia striiformis, respectively. Logistic or exponential disease progress curves were fitted to disease severity data and used to estimate the date of disease onset (t 0 ) and relative epidemic growth rate (r) on each leaf layer for each treatment. Area under the disease progress curve (AUDPC) values were used to construct fungicide dose by spray timing response surfaces for each of the upper four leaves. A parsimonious function, with an exponential form in the dose±response dimension and a normal distribution in the timing dimension described a high proportion of the variation in AUDPC (R 2 values ranging from 0´73 to 0´97). Consistent patterns of treatment effect were noted across pathogen species, leaf layers, sites and seasons. Fungicide applications that coincided with full leaf emergence delayed t 0 on that leaf layer. Treatments applied after full leaf emergence did not delay t 0 , but reduced r. Progressively earlier or later treatments, or lower doses, had decreasing effects. AUDPC was affected more by t 0 than r. AUDPC response surface parameter estimates showed that curvature of the dose±response was not affected by spray timing, but appeared to be a characteristic of the fungicide±pathogen combination. However, the lower asymptote of the dose±response curve, and hence the ED, varied substantially with spray timing. The pattern of change in ED with spray timing was consistent across a range of leaf layers, pathosystems and seasons, and the spray timing at which the ED was minimized varied only within a small range, around the time of leaf emergence. In contrast, variation in untreated disease severity, resulting from variation in initial inoculum and weather, was large. It was concluded that the main value of disease forecasting schemes may be in their capacity to predict the level of untreated disease, to which the economic optimum, or appropriate', dose relates. Spray timing determines the part of the canopy where disease will be efficiently controlled and hence the green leaf area saved. Timing decisions should relate to observations of emergence of those leaf layers important to yield.