The window width varies with how sensitive the task is to causality

What it is

The acceptable synchrony offset is not a constant. The more a task depends on judging "my action caused this result," the narrower the tolerance; when the task does not interrogate causality, larger offsets pass. A single millisecond threshold for every context is therefore wrong.

Why it happens

The difference follows from how much the task leans on causal attribution. In precise control, users keep correcting their action against feedback, so an offset enters the control loop and amplifies into a noticeable error; in tasks that only need to know something happened, feedback is a one-off confirmation and slight misalignment changes nothing. Sensitivity also depends on whether the offset is stable: a fixed offset is easy to adapt to, while random jitter is harder to tolerate even at the same mean.

Studying it

Compare tolerance across task types: have users perform continuous manipulation, discrete tapping, and pure notification tasks while the same set of offsets is injected, and measure detection thresholds and preference. Variables include task type, whether offsets are random, and whether users were told the offset exists. Report both thresholds and task performance, since they can disagree.

Where it stops holding

When users know about the offset and attribute it externally—a known network transfer, say—tolerance rises, but design cannot rely on that situation. If a task spans both continuous control and discrete confirmation, the stricter part sets the design baseline. Across devices the lower bound is set by hardware latency, and the design goal becomes internal consistency rather than absolute synchrony.

Applying it

• Define synchrony targets per task group by causal sensitivity rather than a single threshold.
• Set the strictest target for continuous manipulation and relax it for discrete confirmation and notification.
• Prioritize removing random jitter, since stable offsets are easier to adapt to than random ones.
• Verification: measure detection thresholds in each task class, confirm the targets match measurements, and check that jitter is held within an acceptable range.

Related

• Within the group: D3.04.1 Haptics leading vision is more acceptable than lagging it · D3.04.2 De-synchronization weakens the sense of direct manipulation
• Adjacent: D4.06.3 Integration window width differs by sensory pairing · D1.01.1 The delay between trigger and feedback decides whether causality holds
• Search terms: synchrony window · task sensitivity · latency tolerance