People have a limited ability to grade output force

What it is

People can apply varying amounts of force as needed, but the number of reliably distinguishable, reliably reproducible force levels they can grade is limited — it isn't a matter of dividing force into as many levels as one likes. This is separate from the ceiling on how much force a person can produce: someone able to lift something heavy isn't thereby able to reliably distinguish a dozen intermediate levels within that force range. Fine control has a ceiling of its own, independent of maximum strength.

Why it happens

Controlling how much force to apply depends on the central drive signal sent to the muscles, and that signal's own noise scales in proportion to its strength — telling two adjacent force levels apart reliably requires the difference between the drive signals to exceed this noise, and the more finely levels are packed together, the more easily adjacent levels get swallowed by the noise and become indistinguishable. This sets a rough ceiling on how many reliably distinguishable levels exist within a given force range — a ceiling that doesn't rise just because a user wants finer control.

Studying it

Common paradigms are force-matching or force-reproduction tasks: participants first feel a target force, then are asked to reproduce it without feedback, relying only on the current muscular sensation, or are given absolute-identification trials among a set of preset force levels (a level is presented and the participant judges which category it belongs to). The independent variables are the number of levels to be distinguished and the force range; the dependent variables are reproduction error or identification accuracy.

Methodological note: results are highly sensitive to the body part applying the force (finger, wrist, forearm) and to where the force falls within the absolute range (near maximum versus well below it) — a level count measured for one action can't simply be carried over to a different way of applying force or a different force range.

Where it stops holding

This ceiling is a reasonably stable value measured under controlled conditions, but a common interface practice is to map continuous pressure input into far more levels than this ceiling allows — for example splitting a pressure value into dozens of bins tied to different response intensities. Such a design assumes a level of controllable precision that exceeds what a human hand can reliably deliver; what the user can actually perceive and reproduce may only be a handful of coarse levels, and the rest of the fine gradation never really exists for the user, no matter how finely the underlying value is divided.

Applying it

• Before designing pressure- or force-based interactions (pressure-sensitive keys, stylus pressure for writing, force-controlled game controllers), confirm that the number of levels being mapped falls within a range humans can reliably distinguish, rather than copying the sensor's raw resolution directly.
• For scenarios that need multi-level force feedback, default to a coarse, small number of levels (three to five, say), each paired with a perceptibly different response, rather than aiming for a continuous, finely graded mapping.
• To verify: have users repeatedly try to hit a specified level without seeing a numeric readout, and tabulate the rate at which they land on that level or an adjacent one. If the hit rate drops sharply as the number of levels increases, the current level count already exceeds what users can actually distinguish, and the count needs to be reduced.

Related

• Same group: A8.11.2 Relative force-control precision worsens as the fraction of maximum force rises · A8.11.3 Force reproduction error grows sharply without feedback · A8.11.4 The usable number of pressure levels falls far short of sensor resolution · A8.11.5 Sustaining a given force level is harder than reaching it
• Nearby: C10.12 Pressure input and force grading · A8.13 Finger independence and strength
• Search terms: force grading · force discrimination · motor noise