An experiment was held from 1978 to 1983 to know how soil phosphorus extractors compare when samples are colected at different periods after application of triple superphosphate, Mg-thermophosphate or rock phosphate from Araxá. The fertilizers were broadcasted to the soil at rates of 0 - 145 - 290 - 435 and 580 kg/ha of total P(2)0(5) and then incorporated with a heavy horrow. Each 120 days, in a period of 48 months, a composite soil sample was collected from 15 sites of each plot and analyzed for P by extraction with 0,05 N H2SO4, Bray-l and Olsen extractors. The results showed that all extractors were efficient in recovering phosphorus applied either as triple superphosphate or Mg-termophosphate; when applied as phosphate from Araxá the best result was obtained with Bray-l method. Olsen method gave the most consistent results for samples collected at different times after application of triple superphosphate or Mg-thermophosphate ; when phosphate from Araxá was applied results were most consistent whith Bray-l and Olsen extractors. At the 32nd month after application of phosphate from Araxá and thereafter it was observed a tendency for the extraction methods to be similar in respect to phosphorus extraction.