Reverberation mapping of active galactic nucleus (AGN) emission-line regions requires the numerical deconvolution of two time series. We suggest the application of a new method, regularized linear inversion, to the solution of this problem. This method possesses many good features; it imposes no restrictions on the sign of the response function; it can provide clearly defined uncertainty estimates; it involves no guesswork about unmeasured data; it can give a clear indication of when the underlying convolution model is inadequate; and it is computationally very efficient. Using simulated data, we find the minimum S/N and length of the time series in order for this method to work satisfactorily. We also define guidelines for choosing the principal tunable parameter of the method and for interpreting the results. Finally, we reanalyze published data from the 1989 NGC 5548 campaign using this new method and compare the results to those previously obtained by maximum entropy analysis. For some lines we find good agreement, but for others, especially C III lambda(1909) and Si IV lambda(1400), we find significant differences. These can be attributed to the inability of the maximum entropy method to find negative values of the response function, but also illustrate the nonuniqueness of any deconvolution technique. We also find evidence that certain line light curves (e.g., C IV lambda(1549)) cannot be fully described by the simple linear convolution model.