The effect of background noise on speech production is an important issue, both from the practical standpoint of developing speech recognition algorithms and from the theoretical standpoint of understanding how speech is tuned to the environment in which it is spoken. Summers et al. [J. Acoust. Soc. Am. 84, 917-928 (1988]) address this issue by experimentally manipulating the level of noise delivered through headphones to two talkers and making several kinds of acoustic measurements on the resulting speech. They indicate that they have replicated effects on amplitude, duration, and pitch and have found effects on spectral tilt and first-formant frequency (F1). The authors regard these acoustic changes as effects in themselves rather than as consequences of a change in vocal effort, and thus treat equally the change in spectral tilt and the change in F1. In fact, the change in spectral tilt is a well-documented and understood consequence of the change in the glottal waveform, which is known to occur with increased effort. The situation with F1 is less clear and is made difficult by measurement problems. The bias in linear predictive coding (LPC) techniques related to two of the other changes-fundamental frequency and spectral tilt-is discussed.