A survey of management practices on wheat farms in northern New South Wales was carried out on 50 farms between 1983 and 1985 and was supplemented by a questionnaire mailed to 750 growers in 1985. Information was collected on crop rotation, tillage practice, fertiliser use and weed control practices. Data were collected from 1 paddock on each farm and included: wheat grain yield and quality, available soil water and nutrients at sowing, wild oat density, and incidence of soil-borne diseases. The 3-year average grain yield in survey paddocks was 2.2 t/ha. Multiple regression analysis was used to identify factors affecting grain yield and protein in 1985. Of the variation in wheat grain yield, 74% was explained by variation in available soil water at sowing, available soil nitrate at sowing, sowing date and wild oat density. Grain protein content declined with increasing available soil water and phosphate at sowing and with earlier sowing, but increased with available nitrate at sowing. Agronomic practices aimed at maximising wheat grain yield, in the presence of a deficiency ofavailable soil nitrate, are likely to result in a reduction of grain protein content. Likewise, responses to application of nitrogenous fertiliser are likely to be inversely related to available soil water at sowing. The mean gross margin for 1984 and 1985, based on $100/t of wheat grain, was $128. The mean gross margin for the least profitable 20% of paddocks was $37, and $253 for the top 20%. New varieties of wheat and herbicides were readily adopted by farmers. On the other hand, adoption of nitrogenous fertiliser use was slow, considering the widespread and long-standing deficiencies of nitrogen in cropping soils of the region. Crop rotation and tillage practices have changed only marginally since the late 1940s. The results of this survey indicate that the usefulness of soil testing for predicting fertiliser requirements could be improved by taking into account levels of available soil water, weed competition and sowing date and by using multiple regression analysis.