The signal of diffusion-weighted imaging of the human kidney differs from the signal in brain examinations due to the different microscopic structure of the tissue. In the kidney, the deviation of the signal behavior of monoexponential characteristics is pronounced. The aim of the study was to analyze whether a mono- or biexponential or a distribution function model fits best to describe diffusion characteristics in the kidney. To determine the best regression, different statistical parameters were utilized: correlation coefficient (R(2)), Akaike's information criterion, Schwarz criterion, and F-test (F(ratio)). Additionally, simulations were performed to analyze the relation between the different models and their dependency on signal noise. Statistical tests showed that the biexponential model describes the signal of diffusion-weighted imaging in the kidney better than the distribution function model. The monoexponential model fits the diffusion-weighted imaging data the least but is the most robust against signal noise. From a statistical point of view, diffusion-weighted imaging of the kidney should be modeled biexponentially under the precondition of sufficient signal to noise.