Marsh terracing is used to restore coastal wetlands by converting shallow nonvegetated bottom to intertidal marsh. Terraces are constructed from excavated bottom sediments, and are commonly arranged in a checkerboard pattern of square cells with open corners to form terrace fields. In 1999, terrace cells of three sizes (large = 122 m sides with 1.30 ha ponds; medium = 61 m sides with 0.29 ha ponds; small = 30 m sides with 0.06 ha ponds) were incorporated into a restoration project constructed at Galveston Island State Park, Texas, USA. This restoration project provided an opportunity to examine how nekton populations and the cost effectiveness of terracing projects vary with cell size. We compared nekton density and biomass (as measures of habitat value) in marsh and open water habitat types among the three cell sizes of the terrace fields. We also compared the habitat value of these terrace fields with the area before project construction, with nearby nonvegetated bottom, and with natural marsh habitat. Nekton abundance and biomass increased substantially in the project area following restoration by marsh terracing. An analysis of post-construction samples detected few statistically significant differences in animal density and biomass among cell sizes or between the terraced areas and adjacent natural habitats. Within terrace cells, density, biomass, and species richness were generally higher in marsh vegetation than over nonvegetated bottom. Using these post-construction density data, GIS, and population models for selected fishery species, we show that populations of most fishery species increase as cell size decreases. However, as cell size decreases, the cost of terrace construction increases much faster than population size. Therefore, terrace fields constructed of medium or large cells would be more cost effective in providing fishery habitat than would terraces composed of small cells.