Computational methods have been developed to model the effects of constrained or restricted amplitude uniaxial rotational diffusion (URD) on saturation transfer electron paramagnetic resonance (ST-EPR) signals observed from nitroxide spin labels. These methods, which have been developed to model the global rotational motion of intrinsic membrane proteins that can interact with the cytoskeleton or other peripheral proteins, are an extension of previous work that described computationally efficient algorithms for calculating ST-EPR spectra for unconstrained URD (Hustedt and Beth, 1995, Biophys. J. 69:1409-1423). Calculations are presented that demonstrate the dependence of the ST-EPR signal (V'(2)) on the width (Delta) of a square-well potential as a function of the microwave frequency, the correlation time for URD, and the orientation of the spin-label with respect to the URD axis. At a correlation time of 10 micros, the V'(2) signal is very sensitive to Delta in the range from 0 to 60 degrees, marginally sensitive from 60 degrees to 90 degrees, and insensitive beyond 90 degrees. Sensitivity to Delta depends on the correlation time for URD with higher sensitivity to large values of Delta at the shorter correlation times, on the microwave frequency, and on the orientation of the spin-label relative to the URD axis. The computational algorithm has been incorporated into a global nonlinear least-squares analysis approach, based upon the Marquardt-Levenberg method (Blackman et al., 2001, Biophys. J. 81:3363-3376). This has permitted determination of the correlation time for URD and the width of the square-well potential by automated fitting of experimental ST-EPR data sets obtained from a spin-labeled membrane protein and provided a new automated method for analysis of data obtained from any system that exhibits restricted amplitude URD.