Results of speech recognition tests with competing sound obtained by the use of standard audiometric equipment will suffer from considerable systematic errors because of normal calibration variability. To avoid this, and to keep the reliability high, not only in investigative situations but also in clinical practice, it is highly recommended that test materials be used with speech and noise mixed in the recording. Considering this, a test material was developed consisting of six Swedish PB word lists and speech-weighted noise pre-mixed with a fixed speech-to-noise ratio and recorded on compact disc. This material was investigated for list equality, and normative recognition data were obtained. The material was found to be reliable and suitable for clinical use.