The objective of this study was to compare the application of different in vitro and in situ methods in empirical and mechanistic predictions of in vivo OM digestibility (OMD) and their associations to near-infrared reflectance spectroscopy spectra for a variety of forages. Apparent in vivo OMD of silages made from alfalfa (n = 2), corn (n = 9), corn stover (n = 2), grass (n = 11), whole crops of wheat and barley (n = 8) and red clover (n = 7), and fresh alfalfa (n = 1), grass hays (n = 5), and wheat straws (n = 5) had previously been determined in sheep. Concentrations of indigestible NDF (iNDF) in all forage samples were determined by a 288-h ruminal in situ incubation. Gas production of isolated forage NDF was measured by in vitro incubations for 72 h. In vitro pepsin-cellulase OM solubility (OMS) of the forages was determined by a 2-step gravimetric digestion method. Samples were also subjected to a 2-step determination of in vitro OMD based on buffered rumen fluid and pepsin. Further, rumen fluid digestible OM was determined from a single 96-h incubation at 38°C. Digestibility of OM from the in situ and the in vitro incubations was calculated according to published empirical equations, which were either forage specific or general (1 equation for all forages) within method. Indigestible NDF was also used in a mechanistic model to predict OMD. Predictions of OMD were evaluated by residual analysis using the GLM procedure in SAS. In vitro OMS in a general prediction equation of OMD did not display a significant forage-type effect on the residuals (observed - predicted OMD; P = 0.10). Predictions of OMD within forage types were consistent between iNDF and the 2-step in vitro method based on rumen fluid. Root mean square error of OMD was least (0.032) when the prediction was based on a general forage equation of OMS. However, regenerating a simple regression for iNDF by omitting alfalfa and wheat straw reduced the root mean square error of OMD to 0.025. Indigestible NDF in a general forage equation predicted OMD without any bias (P ≥ 0.16), and root mean square error of prediction was smallest among all methods when alfalfa and wheat straw samples were excluded. Our study suggests that compared with the in vitro laboratory methods, iNDF used in forage-specific equations will improve overall predictions of forage in vivo OMD. The in vitro and in situ methods performed equally well in calibrations of iNDF or OMD by near-infrared reflectance spectroscopy.