Natural-stream designs are commonly based on the dimensions of the bankfull channel, which is capable of conveying discharges that transport sediment without excessive erosion or deposition. Regional curves relate bankfull-channel geometry and discharge to drainage area in watersheds with similar runoff characteristics and commonly are utilized by practitioners of natural-stream design to confirm or refute selection of the field-identified bankfull channel. Data collected from 66 streamflow-gaging stations and associated stream reaches between December 1999 and December 2003 were used in one-variable ordinary least-squares regression analyses to develop regional curves relating drainage area to cross-sectional area, discharge, width, and mean depth of the bankfull channel. Watersheds draining to these stations are predominantly within the Piedmont, Ridge and Valley, and Appalachian Plateaus Physiographic Provinces of Pennsylvania and northern Maryland. Statistical analyses of physiography, percentage of watershed area underlain by carbonate bedrock, and percentage of watershed area that is glaciated indicate that carbonate bedrock, not physiography or glaciation, has a controlling influence on the slope of regional curves. Regional curves developed from stations in watersheds underlain by 30 percent or less carbonate bedrock generally had steeper slopes than the corresponding relations developed from watersheds underlain by greater than 30 percent carbonate bedrock. In contrast, there is little evidence to suggest that regional curves developed from stations in the Piedmont or Ridge and Valley Physiographic Province are different from the corresponding relations developed from stations in the Appalachian Plateaus Physiographic Province. On the basis of these findings, regional curves are presented to represent two settings that are independent of physiography: (1) noncarbonate settings characterized by watersheds with carbonate bedrock underlying 30 percent or less of watershed area, and (2) carbonate settings characterized by watersheds with carbonate bedrock underlying greater than 30 percent of watershed area. All regional curves presented in this report have slopes that are significantly different from zero and normally distributed residuals that vary randomly with drainage area. Drainage area explains the most variability in bankfull cross-sectional area and bankfull discharge in the noncarbonate setting (R2 = 0.92 for both). Less variability is explained in bankfull width and depth (R2 = 0.81 and 0.72, respectively). Regional curves representing the carbonate setting are generally not as statistically robust as the corresponding noncarbonate relations because there were only 11 stations available to develop these curves and drainage area cannot explain variance resulting from karst features. The carbonate regional curves generally are characterized by less confidence, lower R2 values, and higher residual standard errors. Poor representation by watersheds less than 40 mi2 causes the carbonate regional curves for bankfull discharge, cross-sectional area, and mean depth to be disproportionately influenced by the smallest watershed (values of Cook's Distance range from 3.6 to 8.4). Additional bankfull discharge and channel-geometry data from small watersheds might reduce this influence, increase confidence, and generally improve regional curves representing the carbonate setting. Limitations associated with streamflow-gaging station selection and development of the curves result in some constraints for the application of regional curves presented in this report. These curves apply only to streams within the study area in watersheds having land use, streamflow regulation, and drainage areas that are consistent with the criteria used for station selection. Regardless of the setting, the regional curves presented here are not intended for use as the sole method for estimation of bankfull characteristics; however, th