The accuracy of two newer generation theoretical intraocular lens (IOL) power calculation formulas and of the empirical SRK I and II formulas was evaluated in a series of 500 IOL implantations including a series of unusually long and short eyes. The prediction error of the theoretical formulas was found to be largely unaffected by the variation in axial length and corneal power, while the prediction of the SRK I formula was less accurate in the short and long eyes. The prediction of the SRK II formula was more accurate than the SRK I in that no systematic offset error with axial length could be demonstrated. However, because of a relatively larger scatter in the long eyes and a significant bias with the corneal power, the absolute error of the SRK II formula was higher than that of the theoretical formulas in the long eyes. The higher accuracy of the newer generation theoretical formulas was attributed to their improved prediction of the pseudophakic anterior chamber depth.