Abstract: A software-implemented bridging routine is provided for a full duplex audio telephone system comprising units A, B and C. of which only unit A need have at least two lines and be full duplex. Unit A receives and speaker-reproduces an audio signal B+C that echoes with a time delay dt.sub.1 and is detected by unit A's microphone as a signal k.sub.b Bdt.sub.1 +k.sub.c Cdt.sub.1, where k.sub.b and k.sub.c are co-efficients. Unit A's host software determines time dt.sub.1 and generates a delay dt.sub.2 .apprxeq.dt.sub.1. Unit A's host processor sums its own microphone output with a signal generated by a non-destination unit that is delayed by time dt.sub.2. Thus, when communicating with unit B, unit A's software generates a host processor output signal A+k.sub.b Bdt.sub.1 +k.sub.c Cdt.sub.1 +Cdt.sub.2. The k.sub.b Bdt.sub.1 echo component is removed by unit A's AEC, which is coupled between units A and B. This causes unit B to receive the signal A+Cdt.sub.2 +k.sub.c Cdt.sub.1 from unit A. Since dt.sub.2 .apprxeq.