Prediction intervals provide bounds to contain the results of a future sample, based upon the results of a past sample from the same population. Such intervals are frequently required in quality control. This paper deals with sample size requirements for obtaining a prediction interval to contain a single future observation, and more generally the mean of a future sample, from a normal population. Charts, tables and a numerical example are provided. The extension of the concepts to other situations, such as a prediction interval to containallofkfuture observations and one to contain the standard deviation of a future sample, is indicated.