Accurate grass organic matter digestibility (OMD) values are required for intake prediction, for calculating metabolisable protein and ultimately to accurately formulate diets. The three objectives of this study were: (i) to compare the neutral detergent cellulase (NDC)- and pepsin cellulase (PC)-predicted OMD to in vivo OMD; (ii) to compare the NDC method to the PC method for accurately estimating grazed grass OMD and (iii) to compare the effect of sample preparation on the NDC method results, specifically oven drying samples at 60 degrees C compared to freeze drying at 55 degrees C. Data consisted of 34 forage grass samples on which OMD was measured in vivo using the total faecal collection method in sheep (OMDINVIVO). The grass samples were also analysed in vitro in triplicate. For the NDC method the samples were oven-dried at 60 degrees C for 48 h or were freeze-dried at 55 degrees C for 1 week. The PC method samples were oven-dried at 60 degrees C for 72 h. The NDC organic matter solubility (OMS) results were inserted into the equation developed by Stakelum et al., (1988. Irish J. Agric. Res. 27, 104-105) to give OMDNDC. The OMS results from the PC method were inserted into the equation developed by Aufrere et al., (2007. INRA Prod. Anim. 20, 129-136) to give OMDPC. The comparison between OMDPC (60) (mean +/- standard error) (0.708 +/- 0.0361; n=34) and OMDINVIVO (0.719 +/- 0.0361) showed that there was a tendency for the two methods to give different results (P=0.06). The comparison between OMDNDC (FD) (0.740 +/- 0.0049; n=18) and OMDINVIVO (0.764 +/- 0.0063) showed that the two methods gave different results (P < 0.01). The comparison between OMDNDC (60) (0.680 +/- 0.0265; n=34) and OMDINVIVO (0.719 +/- 0.0361) showed that the two methods gave different results (P < 0.05). The positive correlation with OMDINVIVO suggested that OMDNDC (FD). OMDNDC (60) and OMDPC (60) were able to predict OMDINVIVO of perennial ryegrass grazed swards. The comparison between OMDNDC (FD) (0.740 +/- 0.0049) and OMDPC (60) (0.746 +/- 0.0109) showed that the two methods gave similar results (P > 0.05). The comparison between OMDNDC (60) (0.680 +/- 0.0230) and OMDPC (60) (0.708 +/- 0.0230) showed that the two methods tended to give different results (P=0.06). To improve the relationship between OMDINVIVO and OMDNDC (FD), and between OMDINVIVO and OMDNDC (60), future work should expand the number of grass samples, the months of the year over which samples are collected and the range of chemical composition and OMDINVIVO of the samples. Sample preparation method had an effect on the NDC-predicted OMD results as OMSNDC (FD) differed to OMSNDC (60) and OMDNDC (FD) differed to OMDNDC (60). This indicated that specific correction (regression) equations must be used for specific sample preparation methods.