Electrofishing is widely used to survey fish communities in freshwater, but uncertainties remain about electro-taxis (i.e. swimming toward the electrode) and the efficiency with which it catches decapod species. This three-part study was performed to define effective electrofishing settings for mixed communities of freshwater deca-pods and fish species. First, we studied the behaviour of five species of swimming decapods (Caridae) when exposed to four waveforms (direct current (DC) or pulsed direct current (PDC) at 2, 5 or 40 Hz) and two power density settings (low (30 mu W.cm(-3)) or high (80 mu W.cm(-3))) in the laboratory. DC at low power density and PDC-2 Hz at high power density induced significantly more electrotaxis than the other settings. We then tested three electrofishing settings (low power DC (DClp), high power DC (DChp) and high power 40 Hz PDC (PDC40hp)) under natural conditions using the depletion method (with three passes) in enclosed sections of a shallow stream, followed by complete removal of bottom substrate and aquatic vegetation to catch the remaining individuals. DClp consistently detected all shrimp (nine species) and fish (seven species) after two electrofishing passes, but the two other settings did not. DC performed better than PDC for estimating the abundance of shrimp (80% vs. 70% of individuals caught, respectively) and fish (90% vs. 70%, respectively) after three electrofishing passes. Using depletion models, DClp gave the most accurate estimates of abundance for shrimp and fish species. Finally, we assessed the ability of three-pass DClp electrofishing to estimate abundances in nine shallow and coarse -substrate Caribbean streams. More than 10,000 shrimp from 10 species and 1600 fish from 8 species were caught and examined. We observed a correlation between estimated and total abundances after complete removal of bottom substrate and aquatic vegetation in the enclosed stream sections, except for small (<= 20 mm long) shrimp. The fish mortality rate due to electrofishing was 0.22%, and only one species (Eleotris perniger) showed evidence of spinal damage (4.4% of individuals) based on ultrasound analysis of the spinal column. Thus, we recommend using DClp (30 mu W.cm(-3)) to survey sympatric communities of shrimp and fish species in shallow, coarse-substrate tropical streams.