Morphometric, discharge, and sediment data were gathered for 51 streamflow gaging locations in the Piedmont and nontidal coastal plain of Virginia and North Carolina. The Piedmont data are compared to those obtained in the Midwestern United States to calibrate a variable exponent channel geometry model. Least squares regression of the Midwestern data demonstrates that estimated discharge-width exponents for high silt and clay, gravel, or cobble bed channels deviate significantly from those estimated for alluvial beds with 30% or greater sand content. However, no significant trend is apparent for the estimated exponents among other sand-silt-clay channel categories. Analyses of the Piedmont data indicate no significant departure of the discharge-width relations if sediment categories are used to group stream types. This may be due in part to the relative homogeneity of the sediments that compose the bed and banks of the Piedmont stream channels sampled for this study. Least squares regression of the Midwestern and Piedmont data indicate that estimated width-discharge exponents for groups of streams with width:depth ratios less than 45 range from 0.35 to 0.46 but are statistically indistinguishable. For groups of streams having width:depth ratios greater than 45, the width-discharge exponents decrease toward values less than 0.15. This trend suggests systematic variation in the exponents and a diminished influence of channel shape on width-discharge functions for streams with width:depth ratios greater than 45. These findings do not support the results of a previous investigation that used envelope curves to estimate width-discharge exponents for Midwestern streams, and they suggest that the assumptions regarding the variable exponent channel geometry model should be reevaluated. ¿ American Geophysical Union 1995