We develop a new methodology which compares quantitatively styles of folding from seismic reflection data. The goal of the fold style inversion (FSI) method is to provide an objective choice of the most appropriate model used when solving for the shape of an unimaged blind fault from folded layer geometry. FSI is a discretization of the dip isogon fold classification scheme reformulated as simple vector transformations. A data set's goodness of fit to parallel (class 1c) or similar (class 2) fold geometry is assessed by calculating misfit between the predicted and observed bed geometries through a grid search of the parameter space specific to each transformation; the two fold types correspond to the constant bed length and arbitrarily inclined simple shear (AISS) fault solution routines, respectively. For seismic reflection data, confidence estimates may be placed on the preference of fold style and its corresponding fault solution by Monte Carlo simulations of depth correlative, spatially limited depth conversion errors. For synthetic geometric examples FSI determines fold style preference and parameters exactly. At low fold limb dips (<~15¿) the actual geometric difference between parallel and similar folds is very small, and the difference between fold styles cannot be resolved, highlighting a general difficulty in the analyses of young blind thrust structures. For a synthetic seismic line of an AISS fault-related fold the method chooses the correct folding style and leads to the correct fault geometry at depth. In examples of real data from the Barrancas/Lunlunta-Carrizal anticlinal complex in Mendoza, Argentina, FSI analysis determines 71% and 54% probability of similar preference for two seismic lines on separate structures. The corresponding fault solutions for the first example are well constrained, whereas for the second example the solutions are widely variant. This analysis helps to quantify the relationship between the predicted sub surface fault trajectories and hypocenter and aftershock data of the 1985 Mw 5.9 Mendoza earthquake, showing that the earthquake and the fault causing the Barrancas/Lunlunta-Carrizal anticlinorium are most likely unrelated. ¿ 2000 American Geophysical Union