Using a highly resolved Long Term Ecological Research (LTER) database collected near Palmer Station, Antarctica, from 1991 to 1994, the variability in the column photosynthetic cross section (&PSgr;*,m2 g Chl a-1) was analyzed. The relationship between the daily integrated primary production rates versus the product of surface irradiance (QPAR)(0+)) and the integrated chlorophyll content (down to 0.1% QPAR(0+)) gave a &PSgr;* value of 0.0695 m2 g¿hl a-1 (r2=0.85, p<0.001, n=151) which is similar to those determined for temperate and tropical seas. However, the average value of single &PSgr;* estimates is higher (0.109¿0.075 m2 g Chl a-1) with extreme values extending over a fiftyfold range (0.009-0.488 m2 g Chl a-1). The possible drivers of this variability are analyzed in detail, considering variables which are presently used in biooptical models (e.g., surface irradiance and chlorophyll content) and those which are not (taxonomic composition). A sixfold variation in &PSgr;* was observed with time of year and strongly associated with the high seasonality in incident irradiance characteristic of these polar sampling sites. Variability in daily incident irradiance as influenced by cloudiness and variation in chlorophyll content were responsible for an additional twofold variation in &PSgr;*. Finally, the taxonomic dependency of &PSgr;* was demonstrated for the first time. For identical chlorophyll content and surface irradiance, mean &PSgr;* values of 0.114¿0.051 m2 g Chl a-1 were recorded for diatom blooms and 0.053¿0.011 m2 g Chl a-1 for cryptophyte-dominated populations. Results illustrate the validity of &PSgr;*-based approaches for estimating primary production for the Southern Ocean but emphasize the need to address taxon-specific photophysiology to better estimate primary production on smaller spatio-temporal scales. ¿ 1997 American Geophysical Union