Scale testing must include the system’s largest type bucket, not only the usual range

What it is

System Dynamic Type has more buckets than the “regular” cluster in the middle. Accessibility sizes sit above that cluster, sometimes in a separate control. Tapping through default and “a bit larger” never visits the buckets people who need enlargement actually live in. Test Dynamic Type at maximum is a claim about the sample: the largest bucket is an acceptance point, not a flourish.

Why it happens

Type buckets are not a gentle “one more step.” Past the regular maximum, line boxes, wrapping, and control heights enter a steeper slope: a list row becomes three lines, a toolbar wraps, a badge that just fit starts stacking glyphs. Testing only the regular range sees the shallow slope; the failures live on the steep one. Test habit follows the designer’s own device: default, maybe two steps up, never the accessibility maximum. Then “we tested enlargement” is true in the report and false for people at the max. The error is not sloppy execution. It is dropping the tail of the population from the sample.

Where it stops holding

Desktop pages without system Dynamic Type still have a usable upper bound — browser zoom or OS text size — and that bound still has to be tested, not “one hundred and twenty-five percent looked fine.” Television and in-car reading distances are fixed; users often cannot reach phone-like accessibility sizes, so “maximum” is device-defined. A locked brand wordmark or a legally frozen seal is not supposed to follow Dynamic Type; do not accept those on reflow at max.

Applying it

• Crank system type to the largest accessibility bucket the platform documents and walk the main paths: reading, forms, checkout, settings. A pass at regular sizes does not stand in for this walk.
• Keep a gallery of max-bucket screenshots beside default. Differences should be wrapping and height, not “nobody had seen this page.”
• If automation only runs default, add a max-bucket job. Do not silence failures with “extreme users are rare.” That tail is the sample this claim keeps.

Related

• Same group: F4.16.2 Icon-plus-label controls must keep their alignment after type scales · F4.16.3 Large type may be on together with other assistive settings · F4.16.4 Interfaces that ignore Dynamic Type lose low-vision and older users
• Nearby: F4.09 Dynamic Type · J2.11 Text resize and reflow
• Search terms: Dynamic Type · largest accessibility size · font scale testing