The size distribution of fruit on an apple tree on any potential harvest da
te can be estimated by measuring a sample of fruit early in the season and
projecting forward in time using standard growth curves. However, 'Royal Ga
la' apples grown in New Zealand are picked in multiple harvests spread over
about three weeks, so the overall size distribution of harvested fruit can
differ significantly from that on any single harvest date. On each havest
date, fruit are selected for picking according to background colour, not si
ze. By measuring the sizes at harvest of all fruit on selected limbs of eig
ht 'Royal Gala' trees in Hawkes Bay, New Zealand, we show that the relation
ship between the size distributions of picked and unpicked fruit can be des
cribed by a single-parameter model. Although size is not used directly in f
ruit selection, we find that larger fruit still tend to be picked first, wi
th the odds of a fruit being selected approximately doubling for each 20 g
increase in size. The mean harvest fruit size is largest for the first harv
est, and decreases in later harvests, despite ongoing fruit growth. Because
the smaller fruit left on the trees continue to grow, the standard deviati
on of the overall fruit size distribution from multiple harvests can be exp
ected to be significantly smaller than it would have been had all fruit bee
n harvested on the same date. The model is tested using size and background
colour measurements collected on tagged fruit at three sites in two years,
including a range of shading and crop load treatments. For most combinatio
ns of site, year and treatment, the model predicted the size distribution o
f picked fruit well, but in a few cases a different parameter value was nee
d to obtain a good fit.