Abstract: Soil coring is one of the most labor-saving and accurate sampling methods in obtaining representative data to estimate the total root value, but lacks a corresponding accuracy analysis. Sampling at three holes has been widely adopted in the literature, but our experiment employed a higher-density coring method to acquire cotton (Gossypium hirsutum) roots. Root dry weight was chosen to study the accuracy of three, five and seven holes respectively. Five holes give better estimates of total root dry weight with less work while three holes could be representative of the soil surface (<30 cm), but had a wide deviation below 30 cm depth, especially in the bare rows, which had low coefficients of determination and failed to reach a level of significance. The rational sampling method is to collect samples from two holes, one hole, one hole at bare, narrow and wide rows respectively when considering both precision and time consuming labour. The root dry weights in a vertical direction were single-peak curves, and dropped sharply above the 40 cm soil layer, but were flat below 40 cm although there was a small increase in the 70-90 cm layer. Root dry weights were heavier in wide rows than in bared rows with respect to horizontal distribution.