In this paper some shrunken and pretest shrunken estimators are suggested for the scale parameter of an exponential distribution, when observations become available from life test experiments. These estimators are shown to be more efficient than the usual estimator when a guessed value is nearer to the true value.