Crops of winter wheat (Triticum aestivum L. cv. Hereward) were grown in the field in four consecutive seasons from 1991/1992 to 1994/1995 at Reading, UK, within polyethylene-covered tunnels along which a temperature gradient was super-imposed on the ambient temperature variation at normal atmospheric (ca. 370) or an increased [CO 2 ] (ca. 700 μmol CO 2 mol - 1 air), producing many environments from one sowing date in each season at one location. Mean seasonal temperatures varied by up to 4°C along the temperature gradient. Increased [CO 2 ] had no effect on crop duration, or on the rate of reproductive development, which had the same temperature sensitivity across all years. A 2°C warming, on the 4-year ambient mean temperature (10°C), reduced crop duration by 42 days (from 254), and reduced the reproductive phase by 16 days (from 130). Crop biomass generally declined with increase in mean temperature, and was greater at increased [CO 2 ], with the effect of increased [CO 2 ] varying with temperature and between years (6-34% range in relative stimulation by increased [CO 2 ]). Grain yield was substantially reduced by warmer temperatures, and increased by doubling [CO 2 ], but the effect varied greatly between years and with temperature (7-168% range). There were both positive and negative interactions of temperature and increased [CO 2 ] on biomass and grain yield. In all 4 years, the increase in grain yield from doubling [CO 2 ] was negated by an increase in mean seasonal temperature of only 1.0-2.0°C. Year-to-year variation in the responses of biomass and grain yield to [CO 2 ] and temperature resulted from differences in environmental conditions, influencing biomass partitioning and altering the role of different yield components.