Six open-top chamber experiments with field-grown wheat Triticum aestivum L. (five with spring wheat and one with winter wheat) were combined to test which of the two ozone exposure indices, AOT40 and CFO(3), that provided the most consistent relationship between relative yield loss and ozone exposure. AOT40 is the accumulated exposure over a threshold ozone concentration of 40 nl l(-1), while CFO(3) is the cumulative flux of ozone (uptake) to the flag leaves. The ozone uptake of the flag leaves was estimated using a stomatal conductance model, sensitive to phenology, light, vapour pressure deficit (VPD) and temperature in combination with measurements of the boundary layer conductance in the open-top chambers. Both indices were calculated for the grain-filling period, defined as the time from anthesis until 2 weeks before harvest. The duration of the grain-filling period was shown to be closely related to the rate of accumulation of thermal time above a base temperature of 0 degrees C. The CFO(3) index provided a much more consistent pattern in terms of ozone effects compared to the AOT40 index. This was especially the case for spring wheat, for which a linear regression between relative yield and CFO(3) using all five data sets is presented. According to the stomatal conductance model, VPD limited daytime stomatal conductance in warm and dry years, while temperature was the most important limiting factor during daytime in cool and humid years. The effect of light was mainly to delimit the time period of the day during which substantial uptake of ozone took place. It is concluded that, compared to the AOT40 index, the more mechanistically relevant flux-based index CFO(3) will estimate larger yield loss in the relatively humid parts of western and northern Europe, while smaller yield loss will be estimated for the dry summer climates in south and central Europe. The use of an ozone flux threshold, similar to the cut-off concentration 40 nl l(-1) in AOT40, did not improve the performance of the CFO(3) index.