Hydropathy is a dominant force in protein folding and has been measured with numerous methods. Several hydropathy scales are widely used in sequence-based predictions, however, without knowledge about their reliability. We investigated the prediction accuracy of 56 hydropathy scales by correlating predicted values with the accessible surface area in known 3-D structures of proteins. The correlations for the best scales are in the order of - 0.26, whereas the weakest have on average merely - 0.11. Results for different amino acids vary greatly within the scales, but are more consistent between the scales. One of the most common applications of hydropathy scales is to predict antigenic regions. Our analysis indicated that some epitopes are located among the most exposed regions. Despite poor overall correlation, hydropathy predictions can still be used in certain applications where qualitative analysis is sufficient.