Table 1.

Direct effects of adjacent habitat (factor: crop or SNH), management (factor: organic or conventional), proportion of cropland in 1 km radius (% cropland, continuous), and insecticide intensity in the landscape (continuous) on visits of honey and bumble bees and direct and indirect effects of them on pollen delivery (hypothesised causal structure see Fig. 1).

Response	Mediated by	Predictor	Estimate	Std.Err	z-value	P	R2
Honey bee visits		~					0.33
		Adjacent SNH	12.1	21.2	0.6	0.57
		Organic	−19.8	10.2	−1.9	0.053
		% Cropland	0.1	0.4	0.3	0.76
		Insecticide intensity	−4.2	3.8	−1.1	0.26
Bumble bee visits		~					0.61
		Adjacent SNH	4.6	11.1	0.4	0.68
		Organic	8.5	5.0	1.7	0.090
		% Cropland	−0.64	0.19	−3.5	0.001
		Insecticide intensity	0.3	1.8	0.2	0.88
Pollen delivery		~					0.67
		Honey bee visits	22	25	0.9	0.38
		Bumble bee visits	183	40	4.5	<0.001
	Honey bee visits	Adjacent SNH	260	571	0.4	0.65
		Organic	−425	523	−0.8	0.43
		% Cropland	2	8	0.3	0.77
		Insecticide intensity	−91	118	−0.8	0.45
	Bumble bee visits	Adjacent SNH	829	2059	0.4	0.68
		Organic	1566	973	1.6	0.11
		% Cropland	−118	39	−3.0	0.003
		Insecticide intensity	52	339	0.2	0.88

Indirect effects on pollen delivery are split in effects mediated by bumble bee visits or by honey bee visits. Results from the structural equation model (number of observations = 18, minimum generalized least-squares chi-square statistic = 9.3, df = 11) are displayed. For all predictors estimates, standard errors, z-values and p-values are given. R² is given per response.