This study dealt with the extent and dynamics of a thin zone of soil that interacts with rainfall and overland flow in releasing soil chemicals to runoff. A relatively immobile tracer, 32P, was applied at 0.0 (soil surface), 0.5, 1.0, 1.5, or 2.0 cm depths in duplicate soil boxes of three different soils. Simulated rainfall of 6.5 cm/h was applied to each soil box for two separate 30‐min periods. The degree of interaction decreased very rapidly, more or less exponentially, with depth below the surface. An effective average depth or zone of interaction, within which the degree of interaction equals that of the soil surface, was assumed to exist. The effective average depth calculated from the data ranged between 0.2 and 0.3 cm, depending more upon the period of rainfall than upon the type of soil. These average depths were used, along with values of total desorbable P and the fractions of the 32P applied on the soil surface that appeared in runoff, to predict the P concentrations in runoff, which agreed rather well with the measured values. The assumption of an effective average depth was thus valid for P. Transient changes in the effective average depth of interaction during a rainfall period were calculated by using simultaneous P and 32P concentrations in runoff, where the 32P was applied at the soil surface. The effective average depth increased somewhat with time during a rainfall period, especially during the first 30‐min rain. The 30‐min mean effective average depth of interaction calculated by this method agreed well with that obtained by the first method described above.