The consequences of dose reduction of three new herbicide formulations were studied for the control of annual broad-leaved weeds in fields of spring barley (Hordeum vulgare L.) and spring wheat (Triticum aestivum L.). The herbicide formulations were MCPA/mecoprop-P, MCPA/dichlorprop-P and MCPA/fluroxypyr. The efficacy of the lowest recommended dose and a 30% lower rate were tested and compared with the reference herbicide tribenuron-methyl. Trials were conducted at seven sites for three years. Considerable annual fluctuations in weed infestation were recorded. Although the dose reduction occasionally caused considerable decline in control (on %-scale), suppression of weed biomass was still satisfactory in most of the trials. On average, a 75% reduction of weed biomass in spring barley and an 83% reduction in spring wheat were achieved with reduced herbicide doses. Use of reduced herbicide doses for three years in the same fields caused neither a significant increase in weed infestation nor changes in the species composition of weed populations compared with treatments at recommended rates of application. There was a significant difference in biomass production between weed species. Consequently, the total biomass production of annual dicotyledonous weeds correlated only weakly (r=0.48) with the total weed density. Even in untreated plots the weed biomass at harvest constituted, on average, only 3.1-3.6% of the total vegetative biomass of crop stands. Thus, the crop yield responses to chemical weed control remained low.