Objectives of the experiment were to evaluate the interactions between undigested ash-free NDF (uNDFom) measurements and parameters of 2 and 3-pool models, and determine if precision of long-term fermentations (>72h) justifies their measurement. On a DM basis, the 35 alfalfa samples ranged from 21.4 to 54.7% amylase-treated ash-free NDFom. In vitro uNDFom was measured after 6, 12, 18, 24, 72, 120 and 240 h of fermentation. The mean aNDFom at 0 h was 36.8%. The mean and duplicate variation was 28.1 ± 1.57, 23.8 ± 0.95, 22.6 ± 1.18, 20.9 ± 0.82, 19.2 ± 0.98, 18.9 ± 0.70 and 19.0 ± 0.56% of DM at 6, 12, 18, 24, 72, 120 and 240 h, respectively. Using Solver in MS Excel, individual uNDF for each sample was fitted to a 2-pool model of digestion: uNDFom(t) = pdNDF*exp(-kd*(time-lag)) + iNDF2; where pdNDF = potentially digestible NDF, kd = fractional digestion rate, lag = discrete time before digestion begins and iNDF2 = indigestible NDF for a 2-pool model at time = infinity. When uNDFom at 0 to 72 h (uNDF0–72) were used, average kinetic parameters were: 0.30 h, 0.129/h, 47.4% of NDFom and 52.6% of NDFom for lag, kd, pdNDF and iNDF2, respectively. When uNDFom measured for 0 to 240 h (uNDF0–240) were used, average kinetic parameters were: 0.04 h, 0.113/h, 48.6% of NDFom and 51.4% of NDFom for lag, kd, pdNDF and iNDF2, respectively. In this study, 90% of pdNDF in alfalfa digested in 20–22h suggesting that fermentations > 24h may underestimate kd. Attempts to derive parameters for a 3-pool model were unsuccessful with the exception of indigestible NDF at time = infinity (iNDF3). Typically, fast kd were >1.0/h and fractions of fast and slow pools were variable. Duplicate variation for uNDFom was small, but even with low variability it was difficult to detect differences among uNDF72, uNDF120 and uNDF240. Adding uNDF120 and uNDF240 measurements to estimate kinetic parameters resulted in small changes in kd and iNDF for a 2-pool model of NDF digestion and did not affect RumNDFD [ = pdNDF%NDF*kd/(kd+0.03)]. The cost:benefit of fermentations longer than 72 h are not justified for a 2-pool model, and more measurements between 24 and 120 h are needed to parameterize a 3-pool model.