An electronic capacitance meter was used to estimate pasture yield in a grazing experiment of two years' duration. One day each month a yield measurement and meter reading were made at ten sites within each of four paddocks. The regression of pasture dry weight on meter reading was calculated for each of 84 paddocks and only about half were found significant at the 5% level. Significant differences in regression coeficients or intercepts existed between individual paddock regressions for 19 of the 21 months. Pred'iction of yield was least reliable in autumn. It is concluded that factors causing variations in yield-meter reading relationships need identification and their effects reduced before the meter can be used widely in grazing trials