In view of the large variation found in plant cell wall digestibilities with ruminants, an attempt was made to group 124 feeds into different lignification classes (clusters) on the basis of chemical characteristics. Each feed cluster was described using a structural coefficient [Formula: see text] that related the potentially digestible fiber (PDF, %) to the ratio between lignin and cell wall volume. The optimum number of clusters was determined iteratively by performing a regression of the apparent digestibility of dry matter at maintenance level (DDM1, %) against the PDF and cell soluble (SOL, %) contents of feeds. The [Formula: see text] coefficients varied from 0.05 (grains, N = 13) to 1.85 (corn silage, N = 3) and increased with the maturity of the grasses from 0.88 (legumes, vegetative cool season grasses, N = 26) to 1.33 (mature, cool season grasses, N = 19). Predicted PDF were closely correlated (r > 0.9, P < 0.01) to in vitro cell wall disappearances (IVCWD). Apparently digestible cell wall in four grasses and four legumes increased linearly with 96-h IVCWD and standard error (SE) was similar to the SE of predicted apparent digestible SOL from SOL concentrations. Assuming that similarity between SE could be also observed in larger samples, PDF and SOL were used in summative equations to predict apparent dry matter digestibility. DDM1 discounted for intake (DDM1 – 4, %) was regressed against SOL and PDF concentrations of 87 feeds:[Formula: see text]with ds and df, the true digestibilities of SOL and PDF. Estimates of ds and df were 0.98 and 0.95 for a zero-production (maintenance) level of intake, and 0.91 and 0.79 for an intake level four times maintenance. Since the true digestibility of the PDF component was only 4% – 13% lower than that of the cell soluble component, the concentration of PDF in cell wall was the major determinant in the variation in apparent digestibility of forages. Key words: lignin, neutral detergent fiber, true digestibility, cluster analysis, feeds