1. A standardized protocol for analysing the behaviour of pollinators foraging on more than one plant type (species or morph) is needed.
2. A protocol is presented in which the first step is to test whether foraging trips are homogeneous in the frequency of visits to each plant type, or whether there are two or more groups of pollinators with different foraging preferences.
3. Tests for foraging preference and constancy in the sequence of plants visited then should be made separately for each homogeneous group.
4. A hypothetical example is given in which ignoring heterogeneity of preference would lead to the acceptance of a false null hypothesis of random behaviour.
5. A real example is given in which heterogeneity of preference was the only non-random aspect of pollinator behaviour (no significant positive constancy or overall preference), while transfer of powdered dye particles (pollen analogues) was assortative among floral morphs.
6. Heterogeneity of preference is probable, as many factors may cause individual pollinators to forage differently on the same patch of plants, and may be sufficient to produce non-random pollen transfer among plant morphs.