=+item, is not considered to be a valid form of replication. Instead, several different items should each be measured once. What is the danger in using repeated measurements of the same item instead of truly replicating an experimental result? What do you expect the effect on the F statistic to be if repeated measurements of a single item are used at each level of a factor?