The functional response of T. raschii feeding on monocultures of small phytoplankton at 0.4–14 μg pigment liter–1 was determined using a time-series method in a large volume flow-through grazing system. A model-free robust regression procedure aided by graphical and statistical methods was used to compare Ivlev, Michaelis-Menten, linear, Disk equation and Holling type III models. Ivlev, Michaelis-Menten and Disk equation models were less preferable to a linear model because their parameters were highly correlated and could not be uniquely determined, although they appear to fit the data graphically. Holling type III model did not fit the data. A linear model both matched the model-free robust regression, and avoided problems of correlated parameters. No feeding threshold was detectable for the range of concentrations examined. T. raschii ingested the diatom Chaetoceros gracilis and both toxic and nontoxic clones of the dinoflagellate Gonyaulax tamarensis at the same rate at low concentrations (0.4–1.6 μg pigment liter–1).