Measurement error in speech timing studies Free

In recent years, studies of speech timing have begun to move from the purely descriptive into more hypothesis testing and model theoretic paradigms. Since a careful accounting for variance components is often necessary for such analyses, intuitively derived guesses as to the magnitude of measurement error are no longer adequate. This paper discusses various sources of measurement error, from the speaker's head to the experimenter's pencil, and suggests how such errors may have perverted some recent investigations. Methods for calibrating and controlling measurement error are described, and implications for future research are drawn.