In short routine experiments involving small samples from a normal distribution, it is often found to be more convenient to use the sample mean absolute deviation (m) rather than the sample standard deviation (S) in forming confidence intervals. Approximate and exact methods for correctly usingmin confidence intervals are given. The expected lengths of the intervals usingmare then compared with the expected lengths of the intervals usingS. The results indicate that the increase in expected length usingmrather thanSis insignificant, whereas there is a decrease in expected length usingmwhen one can take an additional observation.