Litterfall is an important nutrient flux in forests, but separating litter by species and collecting fresh litter for nutrient analysis for multiple years is time-consuming and expensive.  Understanding the sources of uncertainty in litterfall nutrient flux estimates would help to optimize allocation of research effort.  We analyzed data on litterfall mass (6 years) and nutrient concentration (4 years) for nine tree species in 13 northern hardwood sites differing in stand age and other factors.  Total leaf litterfall mass varied dramatically from year to year, ranging from 201 g/m2 to 317 g/ m², averaged over 13 stands.  This interannual variation was greater than the variation among stands: five young stands (20 to 37 y-old) averaged 236 g/m² over the 6 years of measurement, while older stands (46 to 130 y-old) averaged 265 g/ m².  Species composition of leaf litter differed across stands, but was relatively consistent within stands over time, and could be predicted by allometric equations.  Concentrations of nutrients in litter were more consistent over time for Ca (CV across years averaged 14%) and Mg (20%) than for K (33%) or P (30%) across the 9 tree species. Variation in nutrient concentration across stands varied by species, with ash and aspen being most variable.  With the exception of K, variation in nutrient concentration was higher across sites than years (mean CVs across sites were 23% for Ca, 32% for Mg, 26% for K, and 53% for P).  Based on these observations, we suggest that to estimate litterfall nutrient fluxes accurately in forests of this type, measuring total mass over multiple years is more important than measuring litter chemistry every year.  Analysis of nutrient concentrations should be site-specific as well as species-specific, but separating litter by species to estimate mass-weighted nutrient concentrations may not be necessary if allometric equations are available.