An apparatus and method for detecting and controlling audible errors in a sound communication system at the receiver utilizes channel quality data and also iterative synthesis. Errors occurring in synthesized speech are detected by searching for atypical sound with a stringency dependent upon channel quality. The greater the channel quality deficiency is, the higher the typicality standards will be. Errors are controlled by either re-synthesizing the signal in an iterative way using typicality standards which vary with channel quality deficiency, or by modifying the output signal using typicality standards which vary with channel quality deficiency, or both.