We tested the performance of 2 types of glass fiber filters (GF/F: 0.7 mu m; GF/C: 1.2 mu m) and 2 membrane filters (PC0.2: polycarbonate 0.2 mu m; CE0.22: mixed cellulose esters 0.22 mu m) in estimating chlorophyll a and primary production with the C-14 technique. Four experiments were carried out with water samples from the NW Mediterranean, the NE Atlantic and the Antarctic Ocean. The first experiment compared measurements of particulate organic carbon (POC) production whereas the other 3 also considered total (TOC) and dissolved (DOC) carbon fixation. No significant differences among filters were found regarding chlorophyll a retention but large discrepancies existed in the amount of labelled organic carbon retained in all the experiments. Both types of glass fiber filters, especially GF/F, yielded higher values of apparent pO(14)C recovery than the membrane filters. The GF/F-derived POC production rates were up to twice the PC0.2-derived rates and 63% higher than CE0.22-derived ones. Accordingly, the estimated rates of phytoplanktonic DOC production were higher with the membrane filters in comparison to the GF/F ones. This discrepancy was attributed to a high (DOC)-C-14 adsorption to the glass fibers of GF filters. Due to uncertainties in the magnitude of this process in other samples, we conclude that GF filters are not suitable when particulate primary production must be measured without interference of released dissolved products. and that membrane filters should be used instead.