We sampled variably altered tidal creeks to determine community structure in a developing coastal (USA) landscape. Throw trapping collected smaller and juvenile nekton in the vegetated marsh while minnow trapping in unvegetated channels targeted relatively larger fishes. Non-metric multidimensional scaling ordinations were used to assess community structure. Environmental factors and species most rank-correlated with community dissimilarities in ordinations were determined through the nonparametric BIOENV analysis. We found differences in community composition among creeks that were largely consistent in 2 years. Juveniles of the dominant salt marsh fish Fundulus heteroclitus from throw trapping had a pattern opposite of amphipods and associated with creeks with marsh downstream while larger F. heteroclitus (from minnow trapping) associated with creeks with high percentages of marsh coverage and lower watershed imperviousness. A transient fish Lagodon rhomboides was associated with creeks with lower marsh percentages. Results indicate that loss of marsh, and breaks in marsh connectivity to areas downstream of tidal creeks, can lead to reduced abundances of a dominant resident marsh fish. In order to maintain production of marsh fishes, planners should prioritize the maintenance of intact salt marsh habitats and natural landforms.