In 1980 a field of winter wheat at Little Raveley, Cambridgeshire (U.K.) suffered a bad attack of take-all, which was confined mainly to areas dominated by Ragdale series, one of five soil types on the field. Take-all and yield were assessed on experimental areas within the field in the three subsequent years (1981–3). On a strip 50 m wide, which was sown with wheat in each of these years, take-all was in decline and although slight differences in take-all occurred between some of the soil types, they had no effect on grain yield. Following a break crop of beans in 1981, the remainder of the field carried wheat in 1982 and 1983. Here take-all was generally less and yields generally greater in the area that suffered the 1980 attack, probably because it had developed a natural partial immunity to the disease. However, measurements of takeall in relation to soil characteristics on 1 m2 plots suggested that the disease was becoming more prevalent on soils that are less well drained in winter, are more deeply decalcified, and contain less extractable phosphorus or more exchangeable potassium.