A source-to-tap evaluation of the origin and fate of chloramination N-nitrosodimethylamine (NDMA) precursors at 21 full-scale drinking water plants was conducted. Upstream wastewater discharges accounted for (on a median basis) similar to 16 ng L-1 NDMA formation potential (FP). A correlation between concentrations of the artificial sweetener sucralose (wastewater tracer) and NDMA FP was found within certain watersheds, with increased river flow decreasing sucralose and NDMA FP concentrations by diluting the wastewater discharges. The polymers polydiallyldimethylammonium chloride (polyDADMAC) and polyamine contributed (median) 6 and 14 ng L-1 of NDMA FP to coagulated water (which was pre-chloraminated in two cases), respectively. Biofiltration increased NDMA FP by (median) 6 ng L-1; biofiltration tended to increase precursor loading rather than reduce it. Although wastewater and polymers are known sources of precursors, biofilters as a source of precursors was an important finding. Ozonation of raw or settled water was effective at destroying NDMA precursors (median 34%). As free chlorine exposure increased (from <= 3 min to >1 h), NDMA formation in the chloraminated distribution system decreased (from median removal of 21% to 90% of the NDMA FP sampled prior to chlorination). For either oxidant, precursor abatement was typically higher at pH similar to 8-9 than at 7. Riverbank filtration, and powdered and granular activated carbon removed (median) 64, 47, and 64% of watershed-derived NDMA precursors, respectively. Each was able to remove NDMA FP better than that of the bulk total organic carbon. Granular activated carbon did not appear to be effective at removing polyDADMAC-derived precursors added during coagulation. The contribution of different watershed or in-plant sources to NDMA precursors varies by plant and over time and, depending upon precursor sources, different in-plant treatment strategies can effectively control NDMA formation.