Summary: The determination of the volume of remaining oil in a reservoir is a key ingredient in planning future oil recovery operations. Nuclear Magnetism Logging (NML™) is an openhole wireline logging method that can provide accurate estimates of remaining-oil volumes. The signal processing required to extract these estimates from the logging data and the NML logging procedures used in this application are the subjects of this paper. The physics of the NML method and the wellsite log-inject-log procedures used for remaining-oil determination is reviewed. The latter procedures include the doping of the mud with magnetite to kill the borehole signal and the injection of paramagnetic ions into regions sensed by the NML measurement to kill the formation-water signal. A new NML signal-processing algorithm for determining remaining oil is developed. The algorithm is an improvement over conventional envelope processing in low SNR environments. The algorithm explicitly accounts for the statistical properties of bandwidth-limited Gaussian thermal noise in extracting signal parameters from signal-plus-noise measurements. This is especially important for remaining oil determinations for which accurate estimates of both the fractional bulk volume of remaining oil, Vbro, and the uncertainties in these estimates are required. The method is applied to residual-oil data from a Shell well. The data consist of repeat waveforms at 44 stations. The FFFs are mostly in the range from 1.0 to 4.0 porosity units (p.u.) and the signal decay times are less than 50 msec. Thus, the data represent an abnormally difficult signal-processing problem as a result of the low SNR's. The results of processing these data are compared with results obtained independently by Shell with their own processing methods, which are widely regarded in the industry as standards. The volumes of residual oil and standard deviations estimated with the new signal-processing method developed in this paper are in excellent agreement with the results of Shell's processing methods.