Inconsistent avatar shapes slow list scanning

What it is

If a column of avatars mixes circles, squares, and squircles, the eye cannot ride one silhouette down the list; every row has to re-acquire the outline. Shape unity is not tidiness. It is a predictable left edge for the scan. Jitter that edge and both name-finding and photo-checking slow. Mixed shapes are also read as type signals: square for an organisation, circle for a person — even when the product never said so.

Why it happens

List scanning leans on repeating units. The avatar column is the leftmost, most stable repeat. Once the outline is constant, attention can sample only the face or letter inside. Change the outline every row and the visual system handles "what shape is this" before "who is this"; that extra step scales across dozens of rows. A squircle and a circle look close as singles and become a column of corners that swell and shrink. People also assign categories to shapes, so a column that should distinguish persons is read as grouping types. The scan cost therefore has two layers: the beat breaks, and the category is misread.

Studying it

Take one roster in three masks — all circle, all square, mixed — and time "find this person" plus "scan down to this row". Independent: whether shape is constant. Dependents: scan time and misses. Eye tracking can show whether fixations sit on the outline or on interior features — mixed rows raise outline fixations. If the product truly encodes type in shape (person vs group vs bot), that is a separate treatment and must not be pooled with uncoded mixing.

Where it stops holding

• When type really is encoded in shape, unity yields to decodability — but one shape per type, still unified inside the type. Not a different squareness per group chat.
• A page with two or three avatars will barely show a time delta.
• Distinguishable placeholders and initial collisions are not fixed by changing shape. Making some people squares and some circles turns an identification problem into a scan problem.
• A crop that loses the face is internal to the photo. The outer outline should still be one shape for the set.

Applying it

• Pick one outline (circle or squircle). Distinguish people, placeholders, groups, and bots with interior marks or badges, not by changing the outer mask.
• Fit imported square brand marks into the same outline rather than inserting them in their native shape.
• How to check: turn the column into silhouettes. If several outlines can still be counted at a glance, the scan beat is not locked.

Related

• Same group: F6.09.1 Avatars identify people; defaults need distinguishable placeholders · F6.09.2 Solid-colour initial placeholders collide on same name or same colour · F6.09.4 Photo crops have to keep the face and other recognition-critical regions
• Nearby: F3.04 Scan patterns · E4.02 List items
• Search terms: avatar shape · scan efficiency · squircle