Almond ( Prunus dulcis) trees of cultivars ‘Nonpareil’, ‘Chellaston’ and ‘Johnstons Prolific’ in selected parts of 10 orchards in South Australia were assessed for tree size and production of nuts in 1978–1980. Those data were used to select meaningful measures of yield by comparing the yields of almond trees of different ages and sizes, and by examining the inter-relationships between several measured variables. Tree size, measured as butt cross-sectional area, was found to be linearly related to tree age, and deviated less from a linear function than did butt circumference. Number of nuts per tree was significantly correlated with butt cross-sectional area in some cultivar-year combinations, and so the measure “number of nuts per square centimetre of butt cross-sectional area” may allow a direct comparison between trees of different sizes with respect to their yielding ability. Significant negative correlations were found between mean weight-per-kernel per tree versus yield per tree. For almond, we suggest that yield comparisons between individual trees should take into account differences in tree size as well as variation in nut set and number of flowers. When nut set and number of flowers are constant, we suggest that a meaningful measure of yield per tree is nuts per square centimetre of butt cross-sectional area. This is in contrast to the situation with fruit crops like peach and apple where fruit size is important in yield comparisons. We also suggest that terms such as weight of nuts per hectare are comparable between orchards only if the tree densities and other measurements are known.