The use of indices of speededness in tests is reviewed and three possible indices which derive from a model which assumes that rate of work is normally distributed is proposed. Each of these indices is seen as limited by the failure to adequately consider the correlation between speed and power, but they have the advantage that they are derivable from a single administration of a test. The plausibility of an assumption of a normal distribution of work rates was tested on empirical data from seven tests, and in six cases the assumption was not unrealistic although the fit was only approximate. The analysis of tests for which the assumption is not completely tenable will often be instructive for those who construct instruments. Specific item material seems to govern departures from normality. The goal of a single, intuitively satisfying index of speededness in tests is important. Without some sort of metric or scale with which to assess degrees of speededness, the evaluation of tests in this important area remains inordinately subjective. (Author/BJG)