We describe a series of sampling sensitivity experiments to examine potential errors due to data scatter around expendable bathythermograph (XBT) transects in the tropical Pacific. We use a linear, multiple vertical mode model forced with three different monthly mean wind stress sets for the period 1979--1983. The model is sampled along approximately straight lines of grid points corresponding to the mean positions of XBT tracks in the eastern, central, and western Pacific and then sampled again at the dates and locations of actual XBT casts for 1979--1983. Model dynamic heights are calculated with a resolution of 1¿ of latitude and 1 month, then processed to a monthly mean seasonal cycle and anomalies associated with the 1982--1983 El Ni¿o. When results are compared for the two methods of sampling, the model indicates that data scattered zonally around XBT transects in general can lead to about 2 dyn cm error in dynamic height (equivalent to a 10-m error in model pycnocline displacement) in composite sections of XBT data. This magnitude of error generally does not obsure anomalies associated with the 1982--1983 El Ni¿o or the annual and semiannual harmonics of the mean seasonal cycle in the model, though frequencies higher than the the semiannual can be adversely affected. Errors larger than 2 dyn cm occur in regions where XBT sample spacing in the zonal direction is insufficient to resolve Rossby wave variations in the model (for example, for 16¿N to 20¿N in the central Pacific and from 8¿S to 20¿S in the eastern Pacific). These conclusions are insensitive to the choice of monthly mean wind stress used to force the model.