We investigated feeding by phototrophic red-tide dinoflagellates on the ubiquitous diatom Skeletonema costatum to explore whether dinoflagellates are able to feed on S. costatum, inside the protoplasm of target dinoflagellate cells observed under compound microscope, confocal microscope, epifluorescence microscope, and transmission electron microscope (TEM) after adding living and fluorescently labeled S. costatum (FLSc). To explore effects of dinoflagellate predator size on ingestion rates of S. costatum, we measured ingestion rates of seven dinoflagellates at a single prey concentration. In addition, we measured ingestion rates of the common phototrophic dinoflagellates Prorocentrum micans and Gonyaulax polygramma on S. costatum as a function of prey concentration. We calculated grazing coefficients by combining field data on abundances of P. micans and G. polygramma on co-occurring S. costatum with laboratory data on ingestion rates obtained in the present study. All phototrophic dinoflagellate predators tested (i.e. Akashiwo sanguinea, Amphidinium carterae, Alexandrium catenella, Alexandrium tamarense, Cochlodinium polykrikoides, G. polygramma, Gymnodinium catenatum, Gymnodinium impudicum, Heterocapsa rotundata, Heterocapsa triquetra, Lingulodinium polyedrum, Prorocentrum donghaiense, P. micans, Prorocentrum minimum, Prorocentrum triestinum, and Scrippsiella trochoidea) were able to ingest S. costatum. When mean prey concentrations were 170-260 ng C/ml (i.e. 6,500-10,000 cells/ml), the ingestion rates of G. polygramma, H. rotundata, H. triquetra, L. polyedrum, P. donghaiense, P. micans, and P. triestinum on S. costatum (0.007-0.081 ng C/dinoflagellate/d [0.2-3.0 cells/dinoflagellate/d]) were positively correlated with predator size. With increasing mean prey concentration of ca 1-3,440 ng C/ml (40-132,200 cells/ml), the ingestion rates of P. micans and G. polygramma on S. costatum continuously increased. At the given prey concentrations, the maximum ingestion rates of P. micans and G. polygramma on S. costatum (0.344-0.345 ng C/grazer/d; 13 cells/grazer/d) were almost the same. The maximum clearance rates of P. micans and G. polygramma on S. costatum were 0.165 and 0.020 microl/grazer/h, respectively. The calculated grazing coefficients of P. micans and G. polygramma on co-occurring S. costatum were up to 0.100 and 0.222 h, respectively (i.e. up to 10% and 20% of S. costatum populations were removed by P. micans and G. polygramma populations in 1 h, respectively). Our results suggest that P. micans and G. polygramma sometimes have a considerable grazing impact on populations of S. costatum.