Key message <p>A multi-parameter-study of quantitative wood anatomy of oak reveals site-dependent climate-growth relations. Many of these are unstable, making climate reconstructions based on oak questionable. Winter temperatures may be an exception.</p> Abstract <p>In Europe, oak tree-ring chronologies spanning millennia offer considerable potential for dendroclimatology. However, site conditions might influence the climate-growth relationships, particularly for species such as lowland European oak (Quercus robur/petraea) that grow under a range of soil moisture conditions. The inclusion of sub-fossil or archaeological samples in chronologies often introduces uncertainty regarding the soil moisture regime in which the trees grew, potentially impacting the contained signal and the suitability of millennia-long oak chronologies as paleoclimate proxies. Here, we present a rare example of a multi-parameter study on the quantitative wood anatomy of oak. We analysed the climate-growth relationships of earlywood vessel sizes, earlywood width and latewood width in trees from five nearby stands situated on wet and dry soils in north-eastern Switzerland that experienced the same atmospheric conditions. We observed a large variety of climate-growth relationships. And we observed both signal plasticity (variability of climate growth relationships between site categories) and classical signal instability over time. In fact, the only stable relationships are between earlywood vessel sizes in trees on wet soils and winter/early spring temperatures. The relationship between late spring/early summer precipitation and latewood width on dry soils is probably stable as well but less reliable. In both cases, the correlation coefficient reaches values as high as <i>r</i> = 0.6. We suggest that signal instability needs to be explored with further and larger datasets before oak can be reliably used as a paleoclimate proxy.</p>