Paging costs more than scrolling
Aliases: watch pagination · page-flip cost · crown paging
What it is
On a watch, cutting content into full-screen pages that replace each other is more expensive than scrolling continuously along one list. Each page-turn is a discrete commitment: the crown or a tap must cross a boundary, the previous page leaves the field, spatial position resets, and people have to re-find where they are. Scrolling keeps the spatial string—the item just seen still sits at the edge, the next arrives from the same direction. This entry compares the cost of those two ways of moving. It is not about how many foci belong on a screen, and not about how short the words on each page should be.
Why it happens
A microdisplay has no periphery large enough to anchor “the last page is still there.” A full-screen replace clears the spatial map in working memory; people can only join on a page number or the last item they remember. Continuous crown rotation could have mapped to continuous displacement; once it is quantized into pages, rotation must hit a page boundary to do anything, and ticks in between become dead zones or accidental skips. Comparison tasks (which of seven days has rain, which of several unread mails is urgent) need two items present at once, or at least just-present. Paging forces a remember-then-flip, and glance memory will not hold it. Scrolling’s cost is overshoot, but overshoot can be undone immediately with space still intact. A missed page needs another commitment, and the wrist often drops between the two.
Studying it
Build the same set as full-screen pages (one item or one group per page) and as one scrollable list. Run a find task (“get to Thursday”) and a compare task (“which two days have rain”).
Independent variables: movement (full-screen page / continuous scroll), set length, task type (find / compare). Dependent variables: completion time, page-turns or overshoots, second raises, accuracy on compare items.
Unlimited lab time lets people buy back paging’s memory cost. Add walking to expose wrist-drops in the gaps between pages. Page-turn animation confounds “paging is slow” with “the animation is slow”; report animation time separately, or include a no-animation control.
Where it stops holding
Wizard decisions (pick a workout type, then a goal) change the question at each step; paging is changing questions, not walking one set, and should not be scored against scrolling. Two very short pages (agree / details) can afford one turn. Accessibility item-by-item browsing (one item spoken at a time) is another kind of paging, built for audition; it does not argue that vision should also full-screen-flip. Very long scrolling without a position anchor causes lostness; that is a missing position cue, not paging suddenly becoming cheaper.
Applying it
- Default any one set (messages, events, forecast, tracks) to a scrollable list the crown moves continuously. Do not put one item per page.
- When paging is mandatory, share a visible position cue across neighbors (current / total, a sliver of the previous page) and make sure one turn cannot skip two pages.
- Do not put comparison on paging alone: at least two items’ key fields must be able to appear together, or back-to-back, inside one scroll.
- Verify with “find two items in the set that match a condition.” If paging drops accuracy and raises second-raises, while scrolling finishes in one session, keep scrolling. Do not replace that result with a more ceremonial page-flip.