Particle trajectories of 6.4μm Latex spheres were recorded by video, both near an isolated blue mussel, Mytilus edulis, gill filament and, in place of an intact interfilamentary canal, in a model canal of width 200, 100 or 70μm, formed by a transparent plate positioned next to a gill filament. Each arrangement was placed in a 2 x 10 x 10 cm test vessel filled with seawater. Serotonin (nerve-transmitter) stimulation was used to activate lateral cilia and to either lock latero-frontal cirri at the end of an active stroke (10-5 M), or to activate them (10-6 M), yielding lateral cilia beat frequencies of 19 and 16 Hz, respectively. With latero-frontal cirri locked, image analysis of particle tracks gave maximum velocities of ca. 2.9±0.2mm s-1 close to the tips of lateral cilia, for both isolated filament and model canal cases. Experimental velocity profiles along the 200-μm wide model canal were recorded and used as good approximations to the fluid velocity because of the low Reynolds number. A two-dimensional steady model was proposed for the gill pump, assumed to only comprise lateral cilia. This model was solved numerically for the experimental model, canal in the vessel and the results showed satisfactory agreement with experimental volocity profiles from particle tracks. The numerical approach was also applied to a model of a single interfilamentary canal in the vessel. The resulting mean velocity in the canal was 1.70 mm s-1, but the resistance to flow in the model was less than that in an intact mussel gill. Video graphs of particle tracks indicated that active latero-frontal cirri play a role in the transfer of particles from through current to frontal current, probably by means of a strong interaction through the motion of intervening fluid rather than through a direct physical contact. M. edulis specimens used in the present study were collected in 1990 at Helsingør and in 1991 at Kerteminde, Denmark.