Mesobenthic amphipods (2-15 mm) constitute an important faunal component of coastal ecosystems, and often represent a major trophic link between primary producers and fishes. However, obtaining basic demographic data on these taxa is often hindered by a lack of suitable non-destructive, quantitative, collection techniques. Although artificial substrates have been advocated for collecting data at these lower trophic levels, recent studies suggest that they may not accurately represent the natural benthic assemblage. This study evaluated how a structurally-complex artificial substrate (3-dimensional) simulated the natural substrate of a shallow Acropora reef environment. Assemblage development was followed on a series of artificial substrate units (ASUs) at 2-week intervals over a 14-week period; adjacent natural substrate was sampled monthly. Multivariate analyses were applied to determine the (dis)similarity of assemblages between the two substrate types. Statistical analyses showed significant differences in assemblage structure between artificial and natural substrates. However, no significant differences were found between sites or habitats within substrate type. Samples from the ASUs were dominated by Elasmopus balkomanus, Bemlos kunkelae, Neomegamphopus kalanii, and Ericthonius punctatus. Taxa abundant on the natural substrate, but numerically poor on the ASUs, included Chevalia carpenteri, Apolochus sp., Gammaropsis atlantica, and Globosolembos smithi. Assemblage dissimilarities depended chiefly on differences in species composition and relative abundance largely contributed to the dissimilarity between the two substrates.