The grazing pressure of the macrofaunal invertebrates associated with the biomass of Ulva rigida in the Venice lagoon, their species composition and relationship with environmental variables have been studied. Fifteen sampling campaigns were carried out during different seasons both in the central basin of the lagoon, dominated by macroalgae (especially U. rigida C. Ag.) and in the southern basin, prevalently populated by seagrasses (especially Zostera marina L.). Replicate experiments were conducted in the field by exposing Ulva fronds in net cages of 10 and 1 mm (control) mesh-sizes to allow or prevent grazer entrance. The grazing pressure was determined as Ulva growth rate difference in the cages. In the absence of invertebrate herbivores, Ulva exhibited per cent relative growth rates (%RGRs) ranging from 1.5 to 9.5% day(-1), whereas in their presence the %RGRs were significantly lower (from -2.5 to 3.4% day(-1)) and frequently negative, especially in the station dominated by macroalgae. In this area, peak grazing rates and macrofauna biomasses of up to 8.6% day(-1) and 1,480 g m(-2) fwt (84.4 ash-free dry weight), respectively, were found. On the whole, during in field experiments in the Ulva-dominated station, herbivores removed an amount of biomass whose percentage ranged from 59 to 165% (mean: 103%) of the biomass yield (grazers excluded) found in the cages. These results suggest the possibility that grazers could act as an important factor affecting Ulva production in the Venice lagoon. Macrofauna populations were analysed by means of multivaliate techniques applied to biological variables only and biological and environmental variables together. Data of individual abundance, after a log(x + 1) transformation and the calculation of the Bray-Curtis matrix, were classified using the Cluster Analysis and ordinated by means of the Non-Metric Multidimensional Scaling (MDS) technique, in accordance with the strategies used in the study of multispecies distributions. Finally, biological and environmental variables were analysed together by means of correlation matrices and the Principal Component Analysis.