When a Normal probability plot shows data markers close to the straight diagonal line, we have evidence that the error terms (the difference between the predicted and the observed values) are indeed normally distributed.
