Regression

For the same predictor value, why is an interval for a single future response generally wider than an interval for the mean response?