Upper limit of experimental coefficient between the measured transfer of latent heat and the estimated vapour flux in the frame of the conventional model of latent heat transfer in soil was examined by analysing the measured latent heat transfer and temperature gradient in soil under steady-state temperature gradient. To exclude the temperature gradient as an uncertainty factor from the experimental coefficient, the temperature gradients of overall soil and soil pore were included into the vapour fluxes in the atmosphere. The estimated experimental coefficient did not exceed unity, which indicated that both the latent heat transfer and the vapour fluxes in the soil were smaller than those in the atmosphere. The gap that appeared between the experimental coefficient and the product of the tortuosity factor and air-filled porosity implied the existence of an unidentified parameter relevant to characteristic of the circulation of water in soil which is the main mechanism of latent heat transfer in soil. By quantifying this characteristic with simultaneous measurements of the latent heat transfer, distributions of temperature, water content and solute content in various soils under the steady-state condition, the conventional model would be modified, or an alternative model being independent of the conventional model would be developed.