This study investigated the performance of a spectral similarity index and a multivariate partial least-squares regression technique for detecting the presence of the gyroxanthin-diester-carrying toxic dinoflagellate Karenia mikimotoi from measurements of spectral light absorption. The methods were applied to fourth-derivative absorption spectra of K. mikimotoi mixed with the toxic dinoflagellate Prorocentrum minimum, which does not contain gyroxanthin-diester. Acclimating the cultures to different light and nutrient conditions allowed us to evaluate the sensitivity of the methods to changes in pigmentation and intracellular light absorption (pigment packaging) of the algae. Both methods were able to determine the fraction of K. mikimotoi and the gyroxanthin-diester concentration. However, whereas the partial least-squares predictions were almost insensitive to the induced variability in optical properties of the algae, predictions based on the similarity index differed significantly depending on the acclimation of the algae. Furthermore, discriminating K. mikimotoi from natural phytoplankton assemblages indicated a significant influence of cell-size composition, through pigment packaging, on the accuracy of the similarity index. Our results suggest that optical discrimination of phytoplankton species from spectral absorption signatures will improve significantly by applying the partial least-squares regression technique.