Research in the use of physical measures to evaluate speech intelligibility is showing promising results, and a device for measuring the speech transmission index has been developed [Technical Notes and Research Briefs, J. Acoust. Soc. Am.71, 1612 (1982)]. For many important applications of voice testing, the state of the art has not yet reached a point where such a device can be substituted for standard intelligibility tests using listeners. The two major problems are: (1) The standard error of estimate is too large to make reliable comparisons in the range that would be of interest for selecting among competing methods; and (2) it is not clear that any but the simplest adjustments to voice systems that improve the STI also improve intelligibility, or that a change that improves intelligibility necessarily improves the STI.