Although I have not conducted a separate usability test, I have used subjective comments of the controlled user study to determine possible errors and source of problems. Of primary concern were the usability of the operations such as resize, move/copy, open/close, and maximize/return. Some of the subjects had also concerns about the visibility of the hierarchical nesting. This problem has been relieved to some degree by iterative design of borders. Shadowed border visualization with gradual content background coloring works satisfactorily.
Most of the users had no problems with the resize operation. However, I would assume people less comfortable with the mouse might encounter problems due to the border width. The resize operation requires the user to click on the border and drag it. Thus, it requires effective handling of the mouse. The Elastic Windows interface allows users to change window borders from the window menu providing two versions: thick and thin borders 15 and 5 pixels, respectively. Thus, users having problems might adjust border widths for their usage.
A similar problem is observed for the move/copy operation, where the user is required to drop the objects on the window border. Selectable border widths also relieves this problem.
Since the maximize and return operations are frequent operations special widgets are used on the top border of a window for these operations. However, to overcome possible problems, especially with thin window borders, these operations are also listed in the window menu along with the other operations. As the number of items on the window menu increase, it is possible that the window menu might become too long and error prone.
While the effect of operations on the layout was predictable for most of the users an improvement might prove to be useful. When operations such as resize are being performed the layout changes considerably. In order to reflect the changes as they occur the outlines of the window borders are drawn. However, I believe drawing the updated window contents would further improve predictability of the operations. Another important factor is the extent of the operations. It is argued that operations affecting the whole layout are more disorienting, and less predictable. For this reason, the extent of the operations is limited to the current window group.
The current implementation requires a 3-button mouse. While the usage of a 3-button mouse might be harder for a novice user, the functionality of the buttons are grouped consistently, thus requiring little training. The left mouse button is used for select and resize operations. The middle mouse button is for drag and move, where both operations involve a movement of an object (e.g. icon, window, etc.). The right mouse button is used for open and menu operations, where both allow the user to invoke an operation on an object.