Signals beyond the window are perceived as two separate events

What it is

Signals beyond the integration window are perceived as two separate events. If a user sees a change and then feels a vibration, they read it as two things that happened in sequence rather than as one action's feedback. What the designer meant as one action becomes two in experience.

Why it happens

The consequence is not merely "not synchronized enough." Once split, each event must be explained separately: the earlier one is read as an unprompted system change and the later as a response to the user's action, or the reverse. The causal chain is broken and direct manipulation weakens. Splitting also miscounts events—one operation's feedback is recorded as two—leading users to double-check or trigger again. The splitting threshold matches the window width, so tighter timing is the only direct fix.

Studying it

Measure event count and causal attribution: at progressively larger offsets, ask how many things happened and which was caused by the user's action, recording the offset at which splitting begins. Variables include channel pairing, event complexity, and whether attention is on the action. Outcomes include event-count accuracy and attribution accuracy. The splitting threshold usually coincides with the simultaneity boundary, providing cross-validation.

Where it stops holding

When two channels are designed to express different content—vision showing state, haptics confirming receipt—they are not two faces of one event and splitting is not a problem. Under high load the splitting threshold can drop, so feedback is more easily perceived as two things. If one channel's signal is very weak, users may not notice it at all, and splitting shows up as a miss rather than a miscount.

Applying it

• Schedule the multichannel feedback of one action on a shared time base rather than timing each independently.
• If a channel cannot fire in time, design it explicitly as a later state update rather than pretending it is the same feedback.
• Avoid firing two temporally separated identical cues for one action, since they will be counted as two.
• Verification: at increasing offsets, ask users how many events they perceived, find where splitting begins, and confirm actual latency sits below the threshold.

Related

• Within the group: D4.06.1 Signals from different senses must fall inside one narrow window to fuse · D4.06.4 The window must be designed around the slowest channel
• Adjacent: D1.11.4 Long-running tasks need a transitional state between the two feedbacks · D3.04.2 De-synchronization weakens the sense of direct manipulation
• Search terms: event splitting · causal attribution · multisensory binding