Each move is one cell, so distance is keypress count
Aliases: keypress distance · Manhattan distance · discrete Fitts
What it is
A direction key moves focus one neighboring cell. Cost from A to B is not centimetres of arm sweep; it is how many presses the focus graph requires. That D-pad path length equals step count, usually columns plus rows. In continuous pointing, a larger target saves time. On a TV grid, a larger target is still one node: you walk cell by cell. This is a distance–time relation on a discrete lattice, not a restatement of Fitts’ law.
Why it happens
Each D-pad press takes roughly constant time: short finger travel, no mid-course aiming, no continuous velocity. Total time therefore grows approximately linearly with steps, not with the logarithm of distance over width. Width does not help—focus lands on a node, not on the node’s area. Drawing a huge card only stretches the pixel gap between neighbors; the step count stays put.
Linearity has a second consequence. Distant targets are not “a bit farther but about as fast”; each extra cell is another press, another visual check, another chance to overshoot. Overshooting then walks the path back. People switch to key-repeat; once repeat outruns the eye, they cannot stop on the target.
So “near” on television is graph-near: shortest steps from the default focus to the target, not pixel distance and not visual salience.
Studying it
Count D-pad path length directly. Fix a start (usually the screen’s default focus) and a target, allow only a four-way pad, and record presses, time, and overshoot-and-return. Run a continuous-pointer control of the same task to show when the step model displaces the logarithmic one.
Independent variables: grid rows and columns, start position, key-repeat on or off, degree of the target node. Dependent variables: keypresses, completion time, overshoot, rated effort.
Do not treat pixel distance as the independent variable in a Fitts equation—the fit may look linear, but there is no width to interpret. With key-repeat on, time is no longer strictly proportional to steps; log repeat delay and rate separately. Lab “go to this poster” undercounts living-room cost: people decide which poster while moving, so decision and locomotion intertwine.
Where it stops holding
Trackpad flicks, air mice, and a phone used as a touchpad restore continuous movement; the step model fails and width matters again. Voice that jumps to a title zeroes path length, given a correct recognition. On a short single-column list of five or six settings, step differences drown in the fixed cost of pressing anything. A gamepad stick is analogue: neither one-cell nor pixel pointer, so path length has to be counted in stick displacement.
Applying it
- Budget layout in steps, not pixels. Write the shortest path from default focus to each primary task as a number; anything beyond three or four steps needs to move closer or get a shortcut.
- Do not expect a larger drawing of the target to shorten arrival; it helps visibility, not keypresses.
- Give key-repeat a perceptible ramp, and a stronger focus change as the target is about to arrive, so stopping is possible.
- Verify by recording the key sequence for every primary task. Annotate the layout with step counts and remeasure the same sequences after a change, rather than trusting task time, which animation and loading contaminate.