The frequencies of floral morphs in populations of tristylous Eichhornia paniculata often deviate from the theoretical expectation of equality. This variation is associated with the breakdown of tristyly and the evolution of self-fertilization. Differences in morph frequencies could result from selection pressures due to variable levels of insect visitation to populations and contrasting foraging behavior among the floral morphs. We estimated pollinator densities in 16 populations and quantified visitation sequences to morphs in five populations of E. paniculata in northeastern Brazil. Foraging behavior among floral morphs was measured as the frequency of visits to morphs relative to their frequency in the population (preference) and number of flights between inflorescences of the same versus different morphs (constancy). Pollinator density (number/m2/minute) was not correlated with population size, plant density or morph diversity. Pollinator densities varied most among populations of less than 200 plants. Whether pollinators discriminated among the morphs, depended on whether they primarily collected nectar or pollen. In four populations, nectar-feeding bees (Ancyloscelis and Florilegus spp.) and butterflies showed no consistent preference or constancy among the morphs. In contrast, pollen-collecting bees (Trigona sp.) visited a lower proportion of longstyled inflorescences than expected and tended to visit more mid-and short-styled inflorescences in succession, once they were encountered. Pollinator constancy for morphs did not result from differences in inflorescence production or spatial patchiness among the morphs. Although non-random pollinator visitation to morphs in heterostylous populations could potentially affect mating and hence morph frequencies, the observed visitation patterns in this study do not provide evidence that pollinators play a major role in influencing floral morph frequencies.