[en] We review the data on long-term changes in the solar diurnal variation of cosmic rays, observed at Deep River with a neutron monitor and at Embudo with an underground muon telescope, during the period 1965-1976. We show that the model proposed by Erdos and Kota (1979, Sixteenth ICCR, Kyoto. Conference Papers 4, 45). cannot explain the systematic changes seen in our data during 1971-1976. An insight gained from the diffusion-convection model enables us to obtain a unique solution for K parallel to Gsub(r) in terms of the parameters applicable to the solar wind and the diurnal anisotropy. This helps us calculate the rigidity (R) dependence of the diffusion coefficent K parallel. We find that K parallel = 1.37 x 10¹⁸ Rsup(0.6) m²s^-1, where R is in gigavolts, and K perpendicular 1.08 x 10¹⁸ m²s^-1. Our results are discussed. (author)