A paradigm of cohesive sediment transport research is that erosion and deposition are mutually exclusive. Many laboratory studies have shown that there is a velocity/stress threshold below which erosion does not occur and a lower threshold above which deposition does not occur. In contrast, a deposition threshold is not included in standard noncohesive sediment transport models, allowing erosion and deposition to occur simultaneously. Several researchers have also modeled erosion and deposition of mud without a deposition threshold. This distinction can have important implications for suspended sediment transport predictions and for data interpretation. Model-data comparisons based on observations of in situ erosion and deposition of upper Chesapeake Bay mud indicate poor agreement when the sediments are modeled as a single resuspended particle class and mutually exclusive erosion and deposition is assumed. The total resuspended sediment load increases in conjunction with increasing bottom shear stress as anticipated, but deposition is initiated soon after the shear stress begins to decrease and long before the stress falls below the value at which erosion had previously begun. Models assuming no critical stress for deposition, with continuous deposition proportional to the near bottom resuspended sediment concentration, describe the data better. Empirical parameter values estimated from these model fits are similar to other published values for estuarine cohesive sediments, indicating significantly greater erodability for higher water content surface sediments and settling velocities appropriate for large estuarine flocs. The apparent failure of the cohesive paradigm when applied to in situ data does not mean that the concept of a critical stress for deposition is wrong. Two possibilities for explaining the observed discrepancies are that certain aspects of in situ conditions have not been replicated in the laboratory experiments underlying the cohesive paradigm, and that in situ sediment behavior is better described as a sequence of particle classes than as the single particle class modeled here. However, the in situ measurements needed to resolve these questions are very difficult and data generally are not available. For practical modeling purposes, allowing continuous deposition of a single resuspended particle class may often give quite satisfactory results. ?? 1993.