Abstract Relationships between tree size and water use indicate how soil water is partitioned between differently sized individuals, and hence competition for water. These relationships are rarely examined, let alone whether there is consistency in shape across populations. Competition for water among plants is often assumed to be size-symmetric, i.e., exponents (b1) of power functions (water use ∝ biomassb1) equal to 1, with all sizes using the same amount of water proportionally to their size. We tested the hypothesis that b1 actually varies greatly, and based on allometric theory, that b1 is only centered around 1 when size is quantified as basal area or sapwood area (not diameter). We also examined whether b1 varies spatially and temporally in relation to stand structure (height and density) and climate. Tree water use ∝ sizeb1 power functions were fitted for 80 species and 103 sites using the global SAPFLUXNET database. The b1 were centered around 1 when tree size was given as basal area or sapwood area, but not as diameter. The 95% confidence intervals of b1 included the theoretical predictions for the scaling of plant vascular networks. b1 changed through time within a given stand for the species with the longest time series, such that larger trees gained an advantage during warmer and wetter conditions. Spatial comparisons across the entire dataset showed that b1 correlated only weakly (R2 < 12%) with stand structure or climate, suggesting that inter-specific variability in b1 and hence the symmetry of competition for water may be largely related to inter-specific differences in tree architecture or physiology rather than to climate or stand structure. In conclusion, size-symmetric competition for water (b1 ≈ 1) may only be assumed when size is quantified as basal area or sapwood area, and when describing a general pattern across forest types and species. There is substantial deviation in b1 between individual stands and species.