Legumes and grasses were analysed from two series of trials (a) pure species trials (b) mixed pastures from which legumes and grasses were dissected out for separate analysis.In the pasture trials on Horotiu sandy loam under cow grazing, yield responses to potassium fertilisers commenced in September, reached a maximum in summer (November to January), declined in autumn and were at a minimum in winter. The appearance of deficiency symptoms in white and red clovers corresponded to this pattern of growth responses which followed the seasonal growth cycle. Symptoms were not detected in winter.On deficient soils, uptake responses to potassium fertilisers were appreciable over the whole range from severe deficiency to adequacy, but were small or not detectable on soils with high available potassium.Large differences in analytical levels were found in different tissues of plants well supplied with potassium. Differences were much smaller in potassium-deficient plants.Seasonal decline in the potassium content of pasture species in autumn may have been due mainly to differences in maturity of tissues, associated with longer intervals between grazings, and to "carbohydrate dilution".Clover leaves (leaflets plus petioles), with deficiency symptoms, showed potassium levels below 0.7 %.The critical level, or minimum level for near-maximum growth, was approximately 1.8% K for whole leaves (immature plus mature) of white and red clover at grazing height. The value for ryegrass was approximately 1.6% K and similar levels were indicated for cocksfoot and Yorkshire fog. Sweet vernal and especially paspalum appear to have much lower needs. The gap between the maximum level in clover leaves showing symptoms and the apparent critical level in relation to yield may be partly due to (a) differences in maturity and type of tissues analysed, (b) soil heterogeneity, (c) other limiting growth factors.In pastures on potassium-deficient soils, grasses are more efficient at securing their potassium needs than associated clovers. When ryegrass-dominant pastures are analysed for the purpose of predicting responses of the clover component to potassium fertiliser, higher reference levels may be required than for pure clover stands.