A deep-rooted concept in the dendrosciences is to infer climate or ecosystem variability from statistical relationships established between tree-ring and instrumental data. By extent, these relationships can then be extrapolated into past or future time frames, for example, to anticipate the impacts of anthropogenic climate change on forest ecosystems. Problematically, evidence is mounting that matters are not so simple and that the uncertainties associated with temporal and spatial scaling of tree-ring data are substantial. At the example of recent research, I will discuss some of the challenges associated with i) climate-growth relationships that change through time, ii) inter-specific competition that modulates the abundance and climate response of tree species, and iii) the changing potential for calibrating climate reconstructions in a warming world. In doing so, I aim to stimulate a discussion within the extended tree-ring community and associated disciplines on how to embrace and address non-linearity and non-stationarity in tree-ring data. This will likely require a better understanding of the fundamental ecosystem processes that underly the formation of growth rings. So, back to the roots to predict the future.