The thickness and uniformity of TiNx layers, which are widely used as diffusion barriers, is routinely determined using sheet resistance measurements. However, resistivity is sensitive to stoichiometry as well as to film thickness. In this paper it is shown that : (1) the thickness of TiNx derived from resistivity measurements may be incorrect due to composition variations; (2) good resistivity uniformity does not guarantee that the thickness uniformity is also good since these characteristics are sensitive to different deposition parameters; and (3) uniformities of both resistivity and thickness of a TiNx layer can be balanced by optimization of deposition parameters, Such optimization requires simultaneous and independent measurement of the thickness and composition variations in the film. This requirement identifies ellipsometry as a natural choice for TiNx characterization. Use of multiple angle-of-incidence (MAI) ellipsometry with a blue Ar laser (458 nm) allows thickness and composition effects to be decoupled and ensures accurate measurements over 1000 Angstrom. Comparison of MAI ellipsometry, spectroscopic ellipsometry (SE), profilometry, and resistivity measurements of TiNx films on specially prepared wafers confirms the reliability and accuracy of MAI ellipsometry measurements. Variations in TiNx composition and thickness across the wafers were found to be consistent for MAI ellipsometry, SE, and profilometry measurements.