Between 1989 and 1999, 33 experiments tested the effects of potassium (K) fertilizer on the yield and quality of potatoes. The experiments were done on a range of soil types and used varieties and management conditions common to modern commercial production. The average yield in these experiments was 48 t/ha. Nearly half of the experiments were done on soils that had exchangeable K values < 120 mg/l (MAFF Indices 0–1) but use of K fertilizer resulted in statistically significant increases in fresh weight yield in only seven experiments. Generally, soil exchangeable K was a poor predictor of the probability of a yield response. Potassium fertilizer caused an increase in dry weight yield in only four experiments and these experiments were characterized by the absence of irrigation, soils with small amounts of exchangeable K and use of determinate varieties. Re-examination of published data supported the findings in the current work: potatoes are not particularly responsive to K fertilizer and the optimal K application rate is rarely > 170–210 kg K/ha. When applied at the optimal rate, the effects of K fertilizer on tuber dry matter concentration were nonsignificant. Exceeding the optimal K application rate caused occasional reductions in tuber dry matter concentrations particularly if potassium chloride (KCl) was used. In the two experiments where it was tested, application rate and form of K had no effect on crisp fry-colour. The effect of K fertilizer on tuber K concentration was measured in 21 experiments and on average each tonne of fresh weight yield was associated with 4·2 kg K. The range in values was large, 2·8–5·7 and related to soil exchangeable K.

For fertilizer recommendations based solely on the probability of a significant yield response to K fertilizer it is suggested that no more than 210 kg K/ha be applied even on soils with < 120 mg exchangeable K/l. For fertilizer recommendations based on crop K removal, an uptake value of 4·8 kg K/t fresh weight (FW), as has been suggested, would be adequate, although errors in the estimation of yield may lead to over or under application of K. Since there was little evidence to support fertilizer policies that apply more K than is removed by the crop a fertilizer recommendation system based primarily on the probability of a yield response would be more than sufficient.