In seeking better predictions of grain yield under light-limited conditions, shading was applied to field-grown winter wheat cv. Slejpner during each of five consecutive phases (canopy expansion, ear expansion, pre-flowering, grain expansion and grain filling). Absolute measures were taken of solar radiation and its effects on growth in three seasons, at a site where water and nutrient supplies were not limiting. Replicate mobile shades automatically occluded 0·80 of incident light when mean total solar radiation exceeded 250 J/m2 per s. Mean effects over seasons of shading on incident total solar radiation were −296, −139, −78, −157 and −357 MJ/m2 for the five phases respectively, and corresponding effects on shoot dry weight were −236, −184, −58, −122 and −105 g/m2. Estimated efficiency of radiation use after flowering was 1·2 g/MJ unshaded, tending to increase with shading. Shading in all phases reduced grain dry matter yield: mean effects over seasons were −106, −64, −61, −93 and −281 g/m2 for the five consecutive shading periods. Shading from GS31–39 increased mean maximum area of the two top leaves from 2700 to 3100 mm2 per leaf but, with fewer stems, canopy size remained unaffected. This and the next shading, from GS39–55, reduced specific leaf weight from 42 g/m2 by 4 and 3 g/m2 respectively, but effects on shoot dry weight were largely due to stem and ear. By flowering, stem weights, and especially their reserves of water-soluble carbohydrates, had partially recovered. Effects on yield of shading from GS31–39 were explained by a reduction in grains/m2 of 3070 from 26109. Shading from GS39–55 reduced grains/m2 by 4211 due to fewer grains per ear, whilst mean weight per grain increased in compensation. Shading from GS55–61 decreased grains/ear by 2·5. Shading from GS61–71 decreased ear growth and reduced stem weight, and at harvest resulted in 4·3 less grains/ear. Effects of the final shading from GS71–87 were fully explained by a reduction in mean dry weight/grain of 10·3 mg. Except for shading from GS71–87, source- and sink-based explanations of grain yield both proved feasible, within the precision of the measurements. Constraints to accurate comparison of source- and sink-based approaches are identified, and the implications for yield forecasting are discussed.