Stable isotope analyses have been widely used to estimate terrestrial carbon contribution to zooplankton secondary production. In such studies, the δ13C values of the two potential C sources for zooplankton, i.e. bulk phytoplankton and terrestrial carbon, are assessed and subsequently introduced in two-source mixing models. Subsequently, two assumptions are implicitly made: (1) zooplankton do not feed selectively on particular algae-classes within bulk phytoplankton, or (2) if zooplankton do, the variability of the δ13C values amongst the different algae classes is small compared to the variability between the δ13C values of bulk phytoplankton and terrestrial end-members. In a field study of six coastal lakes of British Columbia, we investigated these assumptions, at two different seasons (July and October), performing stable isotope and carotenoids pigment analyses on particulate organic matter and zooplankton. Analyses of carotenoid pigments revealed that herbivorous cladocerans, in these lakes, consistently avoided diatoms at these seasons and positively selected cryptophytes in July, thus rejecting the first assumption. In addition, cryptophytes were shown to be substantially 13C-enriched compared to bulk phytoplankton and even to the terrestrial end-member, thus refuting the second assumption. In July, the actual phytoplanktonic food source for herbivorous cladocerans was significantly 13C-enriched compared to bulk phytoplankton. We conclude that applying bulk phytoplankton δ13C as an end-point in a two-source mixing model in these lakes would overestimate the contribution of terrestrial carbon to zooplankton secondary production.