Measuring and modelling the soil shrinkage curve (ShC) has generated increased interest, but uncertainty in shrinkage curve parameters has not been documented, and the most recent models have not been compared. This study was performed on 11 core samples collected in the same horizon of a vertisol. The objectives of the study were to (i) characterize the anisotropy of soil samples shrinkage by comparison of vertical, horizontal and volume shrinkage, (ii) analyse the impact of shrinkage anisotropy on fitted shrinkage parameters, (iii) determine uncertainty in shrinkage parameters, and (iv) compare the plasma pore volume as modelled by the clay-paste model (CP) to the plasma pore volume as estimated with the soil shrinkage XP model. The coefficients of variation of the different ShC parameters were smaller than 12% and 3% for water contents and volumes, respectively. Cracking led us to exclude some vertical ShC for XP model fitting, as the ShC departed strongly from the S-Shape model. The vertical geometry factor was smaller than 3 and the horizontal geometry factor was larger than 3 in average. The geometry factor values were changing with water content; they deviated widely from 3 at water saturation, but remained constant and close to 3 in between air entry (AE) and maximum swelling (MS). Cracking and anisotropy had no impact on the XP fitted shrinkage parameters. They were not significantly different between one-dimensional (1-D) and volume measurements. The results of our study suggest that parameters from 1-D shrinkage might be adequate to describe three-dimensional shrinkage. The XP plasma pore volumes at shrinkage limit (SL) were in the range of mercury porosimetry values, the average fitted plasma volume was only 1.7% larger than the median measured value. Direct calibration of CP model from independently measured parameters gave water contents at SL and AE one order of magnitude smaller and two times larger than the fitted values, respectively. But when the AE point water content was set equal in the CP and the XP model, the two modelled plasma ShC were almost identical, with differences depending only on the small difference between fitted and measured plasma volume at SL. Thus, the CP model could be used instead of XP-plasma model to describe the plasma ShC, provided that AE points were identical. This is, however, contradictory with the proposed independent calibration method. The present results are encouraging regarding ShC determination and the comparability of shrinkage modelling approaches. Further comparison of the models requires additional experimental information on the soil plasma shrinkage, especially on MS point.