Resorption of nutrients from senescing tissues is an important aspect of the nutrient economy of perennial plants, and a major influence on ecosystem nutrient dynamics. Attempts to interpret the adaptive significance of variation in nutrient resorption efficiency have met with little success, leading some scientists to emphasize the need for more work on the biochemical and physiological basis of resorption. One hypothesis suggests than variation in resorption is controlled by sink-source effects within the plant. According to this hypothesis, we might expect resorption to be positively correlated with plant growth rates. This could be expected not only for comparisons among species or populations, but also when comparing conspecific individuals growing at different rates, or the same plant during different phases of the annual growth cycle. We measured seasonal variation in leaf litter nitrogen and phosphorous concentrations of four evergreen tree species of the Valdivian rainforest, and tested predictions of the sink hypothesis. Relatively low mean N:P ratios of litter (ranging from 5.5 to 9.9 among the four species) suggest nitrogen limitation of tree growth. N and P concentrations showed nearly twofold seasonal variation in all four species. All species showed low resorption proficiency (high litter [N]) in winter, consistent with the sink-source hypothesis. However, most species showed their lowest litter concentrations of both nutrients (highest resorption proficiency) in autumn. As growth sink strength should be low in autumn, this finding appears to be inconsistent with the sink hypothesis