In vivo NMR spectroscopy is often complicated with problems of low signal-to-noise, poor resolution, undefined peak shapes, and nonlinear baselines despite the efforts of investigators to optimize their experiments. Several data processing options are available to spectroscopists to enhance resolution and signal-to-noise and/or to flatten baselines. There is some question about how these processing protocols affect quantitative information. This paper evaluates five different processing protocols for their ability to extract quantitative information from a set of nonideal spectra. Three of the protocols involve recently developed statistical signal processing methods, maximum entropy Fourier spectral deconvolution, linear prediction singular value decomposition, and baseline deconvolution. These protocols are compared with the conventional processing methods of convolution difference and zeroing initial data points of the FID. The methods are evaluated by use of a quantitative 31P model sample and also are demonstrated on surface coil 31P data.