Reliability of objective voice measures of normal speaking voices.

OBJECTIVE: To determine the reliability of objective voice measures used commonly in clinical practice.

SUBJECTS: Eighteen healthy volunteers (nine males and nine females).

METHODS: Objective voice measures were performed on 18 healthy volunteers on 10 occasions under similar conditions over a 30-day period. Consistency of measures was analyzed to determine reliability.

RESULTS: Using currently accepted normative values, intraclass correlation coefficients were moderate (>0.6) for consistency over the 10 testing sessions for most acoustic measures that do not depend on intensity, measures of laryngeal efficiency, and perturbation measures of fundamental frequency (F0) for both genders. For females, cepstral peak prominence (CPP) had moderate reliability, whereas for males, the smoothed CPP was reliable. Other than F0, none of the perturbation measures are reliable for females. However, jitter, relative average perturbation, and standard deviation of F0 are reliable for males. Noise-to-harmonic ratios (NHRs) had the lowest consistency of all measures over the course of the 10 sessions.

CONCLUSIONS: Clinicians should be cautious in their use of acoustic voice measures that depend on the intensity and in their use of most perturbation measures. NHR was found to be the least reliable measure. Additionally, the reliability of CPP measure varies by gender. Understanding the degree of within-person variability on some objective voice measures and whether that variation is due to biological differences or measurement error will lead clinicians to consider the need for a more standardized testing protocol. Additional research is needed to investigate what factors within the testing protocol and/or changes to the measurement instruments may lead to more consistent test results.