Aerosol particles can be withdrawn from industrial process streams and separated into two or more aerodynamic size fractions with a commercially-available inertial impactor or cyclone. However, with this currently-available equipment, the process gas must be sampled at a constant flow rate to provide constant size cut(s). Thus, traversing of a duct in one sampling run while maintaining isokinetic flow rate as in EPA Reference Methods 5 or 17, cannot be performed, because the flow rate cannot be adjusted from point to point to match stream and sampling nozzle velocities. Sampling theory was analyzed to develop a protocol for existing hardware which limits anisokinetic sampling error to acceptable levels, requires traversing, and is practical enough for source sampling. Anisokinetic sampling error was characterized for aerosol size distributions measured at sources and for hypothetical distributions. A protocol was developed and tested in which net sampling error is always less than 13% even if anisokinetic error is 20%. In this investigation, sampling at typical sources yielded much smaller errors. At most sources, a full traverse is permitted with one sampling run and one nozzle size per run.