Edges are unbounded in the direction of motion, so pointing cost is lowest
Aliases: Fitts edge · screen edge · ballistic pointing
What it is
Fling the mouse at the top of the screen and the menu bar will catch it — motion past the top is physically stopped, so the target is infinitely tall in that direction. Mac menus against the top, Windows taskbars against an edge, both spend this: an infinite-width target. An interior button of the same pixel height cannot; the pointer can overshoot it.
The cheapness holds only for a throw along the edge’s normal. Slide parallel to the edge to hit a narrow icon and width is finite again; the discount vanishes.
Why it happens
Fitts’s index of difficulty is $\mathrm{ID} = \log_2(D/W+1)$. For a target against the display edge, overshoot is eaten by the bezel, $W$ in the approach direction tends to infinity, and ID collapses. People can then fire a ballistic movement: accelerate, hit the wall, stop, without a deceleration aim over the target. It is one of the rare pointing channels that can be used without looking.
The engineering cost is almost zero: no larger drawing, only a hit region that runs to the pixel boundary, and a window that cannot paint past it. Matching that ID with an interior button means inflating visible height to a cartoon.
Studying it
A classic Fitts pointing task places the target on the screen edge, a few pixels in, and in the interior, and compares movement time and errors. The edge condition should let the pointer leave window coordinates, to simulate the wall.
Independent variables: whether the target is flush with the edge, approach direction (normal vs tangential), width along the tangent. Dependent variables: movement time, corrections after overshoot, whether the target can be hit without looking at it.
On touch there is no pointer inertia into a wall, so the edge advantage shrinks; report those runs separately.
Where it stops holding
A canvas with no physical boundary (infinite scroll, a panning map) has no wall; sitting against the “viewport edge” is not infinite width. The gap between monitors is not a wall either — the pointer leaks onto the other display, and a menu parked on the seam is a miss. Touch is coarse and has no pointer inertia, so the infinite-width dividend is much smaller than with a mouse. Once a whole edge is made clickable, tangential width is still finite; picking one of a row of small icons along the edge is not cheap.
Applying it
- On desktop, park global, frequent, “fling and you’re there” entries against the screen edge: menu bar, taskbar, the window’s drag strip.
- Run the hit region to the pixel boundary. A dead gap between bezel and control turns infinite width back into finite width.
- Targets arrayed along an edge still need tangential width, or grouping gaps; the infinite normal does not aim for them along the tangent.
- How to check: from screen centre, fire the pointer at that edge without looking. If a corrective wiggle is still required on the target, the edge is not spending its width.