In renal transplantation, clinical decisions are based primarily on the Banff classification of biopsies. However, the incorporation of 'minor or nonspecific' cellular infiltrates into the Banff classification and their interpretation is uncertain. We analyzed 833 protocol and 306 indicated biopsies to test whether such infiltrates are harmless or whether they have a bearing on outcomes. We characterized morphology, localization and cellular composition of infiltrates, and correlated these findings to the Banff classification and allograft outcome. We found that protocol biopsies had the same prevalence of infiltrates as indication biopsies (87% vs. 87%). Diffuse cortical infiltrates, the hallmark of cellular rejection were more common in indication biopsies and related to tubulitis and a rise in serum creatinine. However, in biopsies with cellular rejection according to Banff criteria, we observed an increase in all infiltrate types (specific and nonspecific) and all cell types (T cells, B cells, histiocytes). The only predictor of allograft function outcome was persistent inflammation in sequential biopsies, irrespective of type, localization and composition of the cellular infiltrates. As detected by sequential biopsies, persistence of any inflammation including those infiltrates currently not considered by the Banff classification should be regarded as a morphological correlate of ongoing allograft damage.