Open-loop error comes from programming precision, not execution correction

What it is

Whether an open-loop movement lands near the target or off to one side depends entirely on how accurate the force and direction parameters set before launch already were — not on how much correction happens during execution, because an open-loop movement has no execution-time correction step at all. That means making an open-loop movement more accurate can't come from "pulling it back on course mid-flight" — it can only come from making that one pre-launch setting more precise. This is the fundamental difference between where open-loop error and closed-loop error come from.

Why it happens

The force a muscle produces, and the timing control needed to sustain it, are never perfectly reproducible: issuing the same "how much force, how long" instruction repeatedly still produces output that scatters around the intended value, and that scatter grows larger as the required force or speed increases — this is impulse variability: the faster and harder a single burst of force has to be, the harder it is to reproduce its magnitude and timing precisely at the output end. Because an open-loop movement has no feedback channel after it's launched to correct that scatter, the scatter shows up unchanged in the final landing point — the size of the landing error is essentially a direct projection of that initial output variability, not the result of "correction that wasn't done well."

Studying it

A common way to verify this has participants repeatedly perform the same target movement, measuring the actual peak force or peak velocity each time, and observing how the spread of these values around the intended target changes as required speed or force increases. If the spread does grow systematically with increased speed demands, and that spread reliably predicts the size of the final landing error, this supports the explanation that error mainly comes from output-end variability rather than insufficient in-flight correction. Another approach artificially removes all feedback channels right after the movement launches and compares landing error against a feedback-available condition — if the two are essentially the same, the movement was never relying on in-flight correction to begin with.

Where it stops holding

This holds only for movements that are genuinely open-loop-dominated; once a movement's duration is long enough to accommodate in-flight feedback correction, landing error is no longer a simple projection of initial output variability — it also picks up new error introduced by the correction phase itself, plus whatever improvement correction contributes, and the net effect of the two is more complex and can no longer be explained by this simple correspondence. Impulse variability itself also shows clear individual differences and practice effects: the same movement, after extensive practice, may show reduced output variability even with the force and speed requirements unchanged.

Applying it

• For fast, open-loop-dominated operations with a high error rate, don't add visual or haptic guidance during execution — that feedback simply can't be used — instead check the pre-launch target setup: relax a hard speed requirement, widen the tolerance on force or angle, or make the target bigger, since the root of the error is the scatter in the output itself, and that scatter shrinks as the speed requirement drops.
• For open-loop operations requiring precise force or precise extent (a gesture that has to draw a fixed length or apply a fixed force in one go), rather than requiring users to hit it precisely on the first try, design the acceptance criterion as a tolerance band whose width is set by the typical output variability at that speed requirement, rather than an idealized single value.
• How to check: have the same users repeat the same gesture many times, record the actual extent or force each time, and compute the standard deviation. If that standard deviation is already close to or exceeds the tolerance width the interface currently requires, the current acceptance criterion exceeds what the body can reliably reproduce at that speed, and it should be widened rather than prompting users to "be more precise."

Related

• Same group: A8.06.1 An open-loop action does not draw on feedback once it has launched · A8.06.2 Fast, short movements are dominated by an open-loop component · A8.06.4 Without any feedback, control has to stay open-loop, and accuracy has a fixed ceiling
• Nearby: A8.09 Speed-Accuracy Tradeoff · A8.11 Graded Force Control
• Search terms: impulse variability · motor output variability · open-loop error