Maria Pombo¹, Minjung Kim², Denis G. Pelli¹; ¹New York University, ²Meta Platforms, Inc.

Crowding and reading speed vary more than two-fold across participants and also vary across fonts. Here, we try many fonts and examine crowding and reading speed. Seventy-four online participants performed three objective tasks — visual crowding, ordinary reading, RSVP reading — and two subjective rating tasks — comfort and beauty. Each participant did these for 3 of 12 fonts. Fonts ranged from common text fonts (e.g., Times New Roman) to display fonts (e.g., Zapfino, an intricate script font, and Omfug, a bubbly, graffiti-like font). In the crowding task, participants identified the middle of three letters presented at ±5º horizontal eccentricity. QUEST measured the threshold spacing. In the ordinary reading task, participants read from a short story and answered reading retention questions. For RSVP, on each trial, 3 words were presented, one at a time. Participants then identified the three words among foils. QUEST measured the threshold word duration. Participants also rated (on a 7-point scale) their comfort after a fixed-rate RSVP reading task. Lastly, to assess the beauty of the font, participants rated how much beauty they felt from looking at a page of Latin Lorem Ipsum text. This emphasizes looking without language processing. Results reveal strong correlations between crowding and both reading comfort (r = –0.93) and speed (r = –0.89 for RSVP, r = –0.58 for ordinary reading). A mixed-effects linear model with font as a fixed effect and participant as a random effect explains the variance in log crowding distance (89%), log reading speed (84%), and log comfort (43%). Text fonts generally produce low crowding and faster, more comfortable reading, while display fonts with high crowding and slower reading are better suited for titles and ads. In sum, measuring a person’s crowding online predicts their reading performance and comfort across fonts.