Data collected with acoustic Doppler current profilers installed on CTD rosettes and lowered through the water column [lowered ADCP (LADCP) systems] are routinely used to derive full-depth profiles of ocean velocity. In addition to the uncertainties arising from random noise in the along-beam velocity measurements, LADCP-derived velocities are commonly contaminated by bias errors due to imperfectly measured instrument attitude (heading, pitch, and roll). Of particular concern are the heading measurements, because it is not usually feasible to calibrate the internal ADCP compasses with the instruments installed on a CTD rosette, away from the magnetic disturbances of the ship. Heading data from dual-headed LADCP systems, which consist of upward- and downward-pointing ADCPs installed on the same rosette, commonly indicate heading-dependent compass errors with amplitudes exceeding 10°. In an attempt to reduce LADCP velocity errors, several dozen profiles of simultaneous LADCP and magnetometer/accelerometer data were collected in the Gulf of Mexico. Agreement between the LADCP profiles and simultaneous shipboard velocity measurements improves significantly when the former are processed with external attitude measurements. Another set of LADCP profiles with external attitude data was collected in a region of the Arctic Ocean where the horizontal geomagnetic field is too weak for the ADCP compasses to work reliably. Good agreement between shipboard velocity measurements and Arctic LADCP profiles collected at magnetic dip angles exceeding and processed with external attitude measurements indicate that high-quality velocity profiles can be obtained close to the magnetic poles.