1 Five temperate floras were studied to assess to what extent seed siz
e correlations with other plant attributes are consistent across flora
s. The floras were from three continents: Australia (semiarid woodland
s of western New South Wales, arid woodlands of Central Australia, and
the Sydney region), North America (Indiana Dunes) and Europe (Sheffie
ld region, UK). The plant attributes used were growth form, plant heig
ht, perenniality and dispersal mode. We used general linear models to
consider not only the primary correlations between seed size and each
other attribute, but also the overlap patterns among correlations to d
etermine if each correlation could be interpreted as a secondary effec
t via a third variable. 2 Plant height and growth form were consistent
ly correlated with the largest proportion of log seed mass variation (
up to 37% in Central Australia). Although there was strong overlap in
the amount of log seed mass variation explained by the two attributes
(6-22%), each could explain small but significant variation after the
other in all floras. The strong association between growth form/plant
height and seed size was found not only among unassisted and wind-adap
ted species, but also among species dispersed by other means. 3 In all
floras, dispersal mode was also able to account for significant varia
tion in log seed mass independently of growth form and plant height. T
he association between plant perenniality and seed size could be expla
ined as a secondary correlation of growth form and plant height with b
oth seed size and perenniality. 4 There were significant differences i
n log seed mass among the five floras. However, seed size ranged over
at least five orders of magnitude in each flora. Differences between f
loras could account for relatively little (4%) of the variation in see
d size between species, compared to the attributes growth form (20%),
plant height (20%) and dispersal mode (29%), despite the quite differe
nt soils and climates of the five floras. This suggests that seed size
is more strongly associated with other plant attributes than with the
environmental conditions for establishment. It appears that within an
y one community, plants have found a diversity of possible solutions t
o the problems of seedling establishment, resulting in a wide range of
log seed mass.