Six species of atyid prawns, representing five genera, occur in streams on the West Indian island of Dominica (figures 1-0). The ecology and habits of each are described and the relation of features of gross morphology to ways of life noted. Xiphocaris elongata, the most primitive living atyid, is a lightly built prawn whose adult habits are related to life in quiet pools in streams. An agile species and an excellent swimmer, it picks up individual small food particles with specialized chelipeds (figures 18 and 19) that differ from those of all other atyids and manipulates them with mouthparts (figure 77) which, while highly complex, are more primitive than those described for any other member of the family. Atya innocous and A. scabra, representing perhaps the most specialized atyid genus, are very similar in gross morphology and are robustly built ambulatory species. A. innocous is common in a variety of situations: A. scabra is rare and has been found only in fast-flowing water. Both have chelipeds whose three distal segments are extremely specialized (figure 30) and whose propus and dactylus are armed with an exceedingly complex array of long, slender bristles. These can be used either as brushes for collecting finely particulate detritus (figures 58-00) or as filtering fans (figures 08 and 69) which, held passively in flowing water, extract suspended particles. The Atyidae is unique among the Malacostraca in having representatives that filter passively by means of the chelipeds. The bristles (figure 40) are extended (figure 49), not by muscles, of which there are none in the distal parts of the propus and none anywhere in the dactylus, but by hydraulic forces. The return of the bristles to rest is by means of a cuticular spring. Some of the bristles of A. innocous are armed distally with minute denticles (figures 41 and 42) that facilitate scraping and sweeping: no such are present in A. scabra. The difference is related to the relative importance of scraping in the two species' A. innocous scrapes frequently, A. scabra seldom. Finely particulate food is transferred and manipulated by the extremely complex oral machinery (figure 78). One of the most elaborate parts of this is a teaselling device in which components of the maxillae and first maxillipeds participate (figures 80 and 81). The feeding mechanism is described. Morphologically and functionally Micratga poeyi can be regarded as a miniature version of Atya. It can both sweep and filter. Potimirim glabra is rare in Dominica and its habits but little known. Morphologically it is similar to, but more primitive than, Micratya. Its cheliped bristles are clearly specialized for sweeping and show few signs of being used for passive filtration. Jonga serrei occupies a separate and well-defined niche in the quieter parts of streams. For this it shows many morphological specializations and lacks such attributes as stout claws and robust walking legs that are the hallmark of its relatives living in fast-flowing waters. Its chelipeds are armed with distal scrapers that bear a remarkable similarity to those already described for African species of Caridina, to which animal it bears a general overall similarity. These are used for collecting food from substrates. It is incapable of passive filtration. Atyid mandibles, while specialized- greatly so in the case of Atya (figure 83)retain primitive features both of the skeleton and musculature such as are found even in the Branchiopoda. Retention of a large molar process is clearly associated with microphagous habits. Specialization for such habits has led to end points very different from those of advanced decapods such as crabs and crayfishes whose mandibles often slice material from large food masses. Specialization has been achieved by additions to and refinements of the primitive crustacean mandible whose salient features are still retained. Comparison of Xiphovaris (primitive) and Atya (advanced) reveals many functional trends in the evolution of food manipulation. These two genera do not, however, simply represent primitive and derived conditions respectively but also end points of divergent specialisation. Likewise, Jonga, Potimirim and Micratya, while to some extent indicating stages on the route to an Atya-like condition, also indicate divergent specialization along that route. All Dominican atyids have a complex two-chambered fore-gut. The most specialized, that of Atya (figures 91 and 92), is described in detail. The complex systems of ossicles developed in its walls are specialized, not for the crushing and grinding of large items of food as they are in many 'higher' decapods but for propelling fine particles posteriorly. Elaborately guarded channels (figures 114-119) from the gland filter that run along each side of the cardio-pyloric valve are presumed to be the route along which enzymes from the hepatopancreas pass into the cardiac chamber. Within the pyloric chamber a system of tubules made up from a delicate convoluted membrane (figures 111 and 112) ensures that there are wide spaces between the several strings of food particles thus separated (figures 105 and 106), thereby facilitating the digestive processes. The membrane is analogous to, and perhaps homologous with, a peritrophic membrane, but is permanent and not continuously renewed. Jonga, Potimirim and Micratya have a fore-gut similar to that of Atya. That of Xiphocaris is different. Apart from a somewhat different arrangement of ossicles the method of spreading food particles so as to expose a large surface area for efficient digestion and absorption is not by means of a reticulated membrane but by a spinule-covered projection that confines particles to a narrow tunnel, crescentic in transverse section (figures 93 and 94). A decapod fore-gut lacking heavy teeth and grinding ossicles is not, as has been suggested, less efficient than one that employs such structures, nor is a large molar region of the mandible necessarily indicative of a crushing function, and non-crushing mandibles are certainly not inefficient. Different foods require different treatments by both mouthparts and fore-gut, both of which show appropriate specializations. Atyids, of which there are no known marine species, have probably had a long history as freshwater animals. Fossils are known from freshwater deposits of Cretaceous age in Brazil. The occurrence of some species in brackish waters is not necessarily indicative of colonization of such habitats from the sea. Many continental species certainly reproduce in freshwater and it is likely that some of those living on islands also do so. Dispersal by sea as adults is virtually impossible and such physiological evidence as is available gives little confidence in the ability of larvae to serve as agents of dispersal across large tracts of ocean. Paradoxically atyids have colonized several remote oceanic islands, some of which have not been reached by truly freshwater fishes. Thus, while dispersal appears to have presented few problems in the past, how it was, or is, effected is uncertain. In the West Indies there is no correlation- such as holds good for example in birds-between island size and number of epigean atyids. To these prawns each stream is the analogue of an island and it seems that a small island like Dominica, with many streams, may present a similar array of habitats to a large island such as Cuba. When ecological opportunities are increased by the provision of additional habitats, such as caves, the number of species of atyids is increased.