Feedback inside a very short delay is perceived as direct causation

What it is

Press a switch and the paddle flips at once: people do not experience “the system responded”. They experience “I flipped it”. Inside that very short window, action and consequence are coded as one event. That coding is instantaneous causal binding: feedback arriving on the order of a tenth of a second is taken as part of the act, not as a later reply. The usual anchor is 0.1 s (~100 ms), a measured bound on what counts as “instant” in human–computer dialogue, not an aesthetic preference.

The question it answers is “does this still count as something I did”, not “has waiting become annoying”. Annoyance lives on a longer timescale.

Why it happens

A motor event and its visual (or auditory, haptic) consequence are separated by a temporal gap. If the gap is narrow, perception binds the two ends into a single event — the same pathway as “tap the glass, hear the ring”. Once the gap opens, binding fails and the brain switches to causal inference: first the action, then a result, with “the system is working” in between. Binding is about interval, not wording. A control can be labelled as direct manipulation and still feel like a reply if the highlight arrives late.

Window width depends on channel. Visual feedback sits around a hundred milliseconds; haptics are tighter, often tens of milliseconds before the hand feels empty. Networked operations almost always miss this seam, so a local, network-independent change has to land first or causality breaks on the opening beat.

Studying it

The classic move is to inject controlled delay into pointing or keying and ask whether the change “was caused by my press” or “was given afterwards by the system”. Card, Moran and Newell marked ~100 ms as the border between perceptual processing and an instantaneous response; Nielsen later wrote 0.1 s as the usability ceiling for feeling instant. Finer work uses staircases on delay detection: same action, delay stepped up from 0 ms, looking for the turn where it stops feeling self-caused.

Independent variables: injected delay, feedback channel (visual / auditory / haptic), whether the task is continuous tracking. Dependent variables: delay detection threshold, ratings of “I caused that”, misattribution of system actions to the self.

Lab thresholds come from repetitive, low-semantics keying. A “Send” in a product mixes network, animation and permission sheets; an 80 ms lab number is not a product SLA. Measure from the input event to the first relevant pixel change, not to a function return.

Where it stops holding

The window is not one line for every person. Age, fatigue, and whether the eyes are already on the object push the detection threshold up or down. Command lines, batch jobs and “submit then see a result” dialogues are already modelled as two events; instantaneous binding is not the design goal. Deliberately stretched motion (a 200–300 ms state transition) is buying orientation, a different ledger — the transition still has to start on the press, or the first frame has already left the window. In remote collaboration the other person’s cursor is necessarily late; the local cursor must still be instant. Those two causal chains are not one metric.

Applying it

• Land evidence that “the hand already acted” inside 100 ms: pressed state, focus ring, row selection, switch paddle — all from local state, none from the network.
• Networked results may follow; the first visual change must appear in the same frame as the input event, or the next one.
• Measure from pointer/touch timestamp to the first related pixel change. Interface duration is not a substitute.
• How to check: set the network to a 1 s delay and tap a switch or select a row. If nothing moves on the press itself, instantaneous causality is already broken.

Related

• Same group: I1.01.2 Beyond that threshold, action and result are perceived as two events · I1.01.3 Direct manipulation depends on that threshold holding
• Nearby: I1.02 Continuity-of-thought threshold · I1.04 Input latency and tracking
• Search terms: instantaneous causal binding · 0.1 second · perceived causality