The sound of the welding process is a consequence of the modulation in amplitude of the current for the voltage of the electric arc. Previous experiments demonstrated that an experienced welder has an absolute dependence on the acoustics in the control of the welding process. In this work, a new technique for the detection of welding defects is presented, based on the stability of the airborne acoustics of the MIG/MAG process for the short-circuit transfer mode. Statistical parameters of the sound pressure and of the sound pressure level were determined for welding without defects from multiple tests; an algorithm was developed from these results for detecting defects based on a moving window that moves through the statistical signals calculated from the acoustical process. Finally, a group of test plates with and without defects was tested; the algorithm based on the airborne acoustics of the developed process showed satisfactory results for the detection of defects. © 2010 Taylor & Francis.