Visual feedback of voice parameters has been shown to be useful in singing training, but most current systems are designed for speech applications. Extraction of parameters such as the fundamental (F0) and formant frequencies can be problematic in singing because of the overlap between F0 and the formants over much of the pitch range. This has limited the application of visual feedback tools in the teaching of singing. This experiment aimed to obtain empirical data on the boundaries within which standard speech-oriented visual feedback systems can be usefully utilized for teaching singing. Four singing teachers learned to operate a simplified set of visual feedback modes on the Computer Speech Laboratory (Kay Elemetics), and then used the system in several lessons, after which they were interviewed about the utility of the system. They found that feedback of F0 and vowel formants were the most useful modes, in spite of limitations in the precision of pitch feedback and the reliability of formant extraction at high F0. The results suggest that current speech-based visual feedback systems can be usefully integrated into singing pedagogy, but that even greater benefits would accrue if more precise F0 and reliable formant-identification algorithms were developed.