Table 1.

A summary of the design of the generalized linear models used to analyse the data in open orchards.

response variable	explanatory variables	random variables	error structure
proportion flights across rows	non-Apis bees present	orchard/day	binomial
pollen tube at base of style?	non-Apis bees present	orchard	binomial
no. pollen tubes initiating growth	non-Apis bees present	orchard, subjecta	Poisson
no. pollen tubes at base of style	non-Apis bees present	orchard	Poisson
proportion fruit set	non-Apis bees present, rate of flower visitationb	orchard	binomial

^aA subject-level random variable was used to account for over-dispersion [37].

^bVisits by all bees, taken from standard observations of flower visitors in the orchards [7].