All 684 post-orthotopic liver transplantation (OLT) liver biopsies performed at the Royal Free Hospital (RFH) between 1988 and 1993, from 120 patients, were reviewed in order to try to define the relative importance of the histological features of immunosuppression-responsive cellular rejection. Twenty histological features considered to be possible contributors to the diagnosis of cellular rejection were documented in a binary (present/absent) fashion. These features in 106 biopsy specimens obtained 1 to 8 days after OLT were analyzed using stepwise logistic discriminant analysis. All clinical and treatment records were reviewed, and each biopsy specimen was assigned to a diagnostic category depending on these records and follow-up information. Important determinants of the histological diagnosis of cellular rejection (which occurred in 84 of the 106 cases) were moderate/severe mixed portal inflammation, eosinophils, endotheliitis, and bile duct damage. When these all occurred together, the odds of rejection increased 3.6-fold. The original histological diagnosis was recorded, and each biopsy specimen showing cellular rejection was regarded according to the specific criteria of Snover et al., Demetris et al., and a novel RFH scoring system. The latter consists of evaluating portal inflammation, endotheliitis, eosinophils, and bile duct damage, each on a 0 to 3 scale (none, mild, moderate, or severe, respectively) and summation. The resulting cellular rejection score thus can range from 0 to 12. The agreement between the different scoring systems was analyzed using K statistics, and there was good concordance (K, 0.64 to 0.78), despite different histological criteria being used to derive each score. Each system showed a similar degree of sensitivity (87% to 96%). The specificity ranged from 59% to 77%. We conclude that the histological diagnosis of cellular rejection relies mainly on the previously described features of mixed portal inflammation, endotheliitis, eosinophils, and duct damage. There is scope for unification and simplification of the existing grading systems, which depend on differing criteria, and we suggest one such scheme.