ABSTRACT: Habitat patchiness is known to alter the relation of a population to both its predators and its prey. We developed a biologically interpretable measure of habitat complexity to test whether juvenile fish density depends on degree of patchiness. In Newfoundland coastal waters, juvenile fish species, including Atlantic cod Gadus morhua, are associated with eelgrass Zostera marina. However, the association of density with eelgrass cover is dependent upon site and the scale of analysis. We tested for non-monotonic relationships (with an intermediate optimum) between 5 spatial characteristics of eelgrass and density of 3 juvenile (Age-0) fish species: Atlantic cod, Greenland cod G. ogac, and white hake Urophycis tenuis. We used aerial photography to determine eelgrass perimeter and area measurements at multiple scales, fractal dimensions of perimeter (D_P) and area (D_A), and a measure that combines perimeter and area complexity at these scales (β_P/A)_. Fish densities were estimated at each site using a seine net. We found parabolic relationships between β_P/A and density for all 3 species, indicating highest fish densities at sites of intermediate patchiness and edge regularity. Furthermore, we determined that β_P/A provided a less ambiguous estimate of spatial configuration than other measures. This intermediate maximum may reflect a trade-off, whereby eelgrass sites of intermediate spatial complexity provide juvenile fish with both optimal protective cover and opportunity to feed. This pattern may apply to any species requiring open areas in which to forage as well as shelter to offset predation risk.