Ten extractants were used for measuring the available phosphorus (P) in 24 tropical soils, and the extractable P values were correlated against plant parameters obtained from a 170‐day continuous cropping experiment and a 50‐day response experiment. The Olsen method was the ‘best’ for the continuous cropping experiment and the Hislop resin method the ‘best’ for the response experiment. When the 24 soils were separated into groups on the basis of organic carbon levels, soil classification, or mineralogy, a higher set of correlation coefficients was obtained than when a single extractant was used for all the soils. The increased precision was thought to justify the extra work involved.