When we are dealing with the situation in which we are randomly assigning treatments to the experimental units and then observing the responses, it is a requirement of the inference procedures discussed in this book that these observations are independent.