Abstract: A conference circuit in a digital telephone switching system includes a first code converter (CCL), which converts nonlinearly coded speech samples into linearly coded speech samples, and adder (ADD), which sums the converted speech samples originating from the participants in the conference, and a second code converter (LCC), which reconverts the sum values into nonlinearly coded speech samples. To avoid distortion without increasing the insertion loss, the speech samples of the individual participants are integrated with respect to time and compared in an evaluation circuit (EVC). The speech samples of the loudest participant are fed to the adder unattenuated, while those of all other participants are attenuated by a predetermined value before being added.