The experiment is a within-subject counterbalanced design with the twelve subjects in four groups of 3 subjects. Subjects performed a total of 18 tasks in both of the interfaces.
Independent variables and their treatments are as follows:
Each subject is tested on both of the interfaces with the order of counterbalanced for half of the users. To reduce the chance of performance improvement for the second interface, a parallel set of tasks are used on the second interface. The tasks in both of the tasks sets are matched in the same sequence. The order of the task set is also counterbalanced for half of the subjects starting with one set of tasks, the rest with the other one. Thus, effectively there are 4 groups of subjects (i.e. ), where indicates that first Elastic Windows with task set 1 and then independent overlapping windows with task set 2. Three subjects were assigned to each group in an order (Subjects 1, 2 and 3 to ; Subjects 4, 5 and 6 to ; ... )
The design requires each subject to test both of the interfaces. This is beneficial as it allowed subjects to examine both of the interfaces and make subjective comments and comparisons. Since all four permutations are included, statistical results are presented for aggregated groups.