The measurement of δ18O in cellulose of tree rings is a potential means to reconstruct δ18O variations of precipitation and climate. We present δ18O data from cellulose of 3 beech trees (Fagus silvatica) growing on a relatively dry site in Switzerland where the roots do not have access to ground water, as well as data of 2 other groups of 4 beech trees each, one from a dry site and the other from a semi-dry site, respectively. The measurements cover the time period from 1934 to 1987 in 3-year-groups for the 1st site and 1965 to 1992 with a 1-year resolution for the other 2 sites. We find a high degree of common variance (61%) between the δ18O variations of the 3 trees from the 1st location suggesting a common external cause. The comparison with climate data indicates that spring temperature (April/May/June) is the main influence for the long-term isotope variations, with a temperature coefficient of 0.33‰ per °C, whereas the short-term variations are mainly influenced by the relative humidity, with a coefficient of – 0.13‰ per %. This latter value is about ⅓ of the expected model value which points to leaf water pools with different δ18O values influencing the isotopic composition of synthesised carbohydrates or to oxygen exchange with stem water. Tree ring δ18O is correlated with δ18O in precipitation, and the slope of the linear regression for different months most probably yield information about the growth rate function of the particular tree group. Our results confirm the potential of δ18O measurements in tree ring cellulose for climate reconstruction, in particular information about δ18O in precipitation can be gained.