A single glance away from the road has a duration cap

What it is

One look away from the forward roadway cannot be stretched at will. That is a cap on single glance duration: how long this departure lasts, not how often the eyes leave during a task or how much time they spend off-road in total. The “about two seconds” figure that appears in human-factors guidance is a design ceiling used to stop a glance from running on. It is not a constant written into nature, and it is not total eyes-off-road time. Short glances may be repeated. A single drawn-out glance is dangerous because of that one blank.

Why it happens

A road event can appear and finish inside one glance. The vehicle keeps its speed, so the empty interval maps to more space as speed rises. A pedestrian, a lead-vehicle brake, an object in the lane do not wait for a line of text to be read. Visual memory does not store what changed while the eyes were gone; on return the driver rebuilds lane and headway rather than resuming a paused tape. A very long glance therefore cannot be offset by “looking again later”—the later look is after the fact. An interface that forces reading, search, and confirmation inside one departure is a design for one long glance.

Studying it

Eye tracking treats each departure from the forward road as its own event and inspects the duration distribution, not only the mean. Visual-manual evaluation protocols usually report both the presence of stretched single glances and total off-road time; the two have to be read apart. Simulators can drop a lead-vehicle brake or a crossing pedestrian into the glance window and compare detection after short versus long glances.

Independent variables: how much must be read in one look, target size, whether the eyes may return to the road mid-task. Dependent variables: single glance duration, counts of glances past a commonly used design ceiling, detection and braking for events inside the window.

A mean washes out a few very long glances. The tail is the object: those stretched glances are the blanks in which something can happen. Lab participants expect events and shorten glances; naturalistic long glances more often appear when the interface stalls, a tap misses, or a target cannot be found.

Where it stops holding

Brief samples of the mirrors or the speedometer are driving, not the same cap as reading a center-stack menu. At low speed on an empty closed course the spatial cost of the same glance is smaller; that tolerance does not transfer to mixed urban traffic. A head-up display shortens downward travel, so the geometry of “off-road” changes, but time spent reading symbols is still a visual occupation and still has a single-glance cap. The constraint does not apply when the vehicle is stopped.

Applying it

• Compress what one glance must read and find so it can end quickly: one action per view, state coded by shape or place rather than a paragraph.
• Do not throw long copy, multi-column tables, or unpaged search results that must be finished on the spot while moving.
• If the task is not done, allow the eyes back to the road. Do not pin people to the display with a countdown or an unfinished modal.
• Verify from the glance-duration distribution, not the average. The long glances in the tail point to the step that stretched them; change that step until a single departure is no longer held by the interface.

Related

• Within the group: K6.02.2 Total eyes-off-road time is a separate constraint · K6.02.3 Glance metrics constrain design; they are not acceptance tests
• Adjacent: K6.01 Types of Driver Distraction · K6.03 Total Task Time · K6.05 Head-up Display
• Search terms: single glance duration · long glance · eyes-off-road