To compare mono- and bi-exponential relaxation model equations to discriminate between normal and fatty liver disease. Six rats on a choline deficient amino acid modified (CDAA) diet and six on normal chow were studied. Multiple spin echo images with increasing echo times (TEs) were collected at 9.4T. Pixel-wise T2 maps were generated using mono-exponential decay function to calculate T2M , and a bi-exponential to calculate, short T2 component (T2S ), long T2 component (T2L ), and fractions of these components (ρS , ρL ), respectively. Statistical F-tests and Akaike's information criterion (AIC) were used to assess the relative performance of the two models. F-test and AIC showed that in the CDAA group, T2 bi-exponential model described the signal of T2 weighted imaging of the liver better than the mono-exponential model. Controls were best described by the mono-exponential model. Mean values for T2M , T2L , T2S , ρS , ρL were 31.2 ± 0.7 ms, 72.8 ± 3.3 ms, 8.2 ± 0.6 ms,71.2 ± 2.1%, 30.4 ± 1.3%, respectively, in CDAA rats, compared with 18.8 ± 0.5 ms, 32.3 ± 0.7 ms, 9.2 ± 1.8 ms, 79 ± 2%, 21.0 ± 1.1% in controls. In the fatty liver of CDAA rats we have shown that T2 weighted images fit the bi-exponential model better than mono-exponential decays thus providing a better description of the data. 1 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018;47:468-476.