Readability is reading efficiency over continuous text
Aliases: reading efficiency · continuous reading · prose reading
What it is
The same three characters, dropped into an article that still has two pages to go, change the question: how long to finish, how much was understood, whether the next line has to be hunted, whether the later pages grate. That is readability. Its object is continuous text; its pass or fail is efficiency — speed, comprehension, stamina — not whether a particular l can be named.
Help, terms, news, long chat bubbles in an interface live on this layer. A label that passes discrimination does not answer whether this screen of explanation can be finished.
Why it happens
Continuous reading is a chain of fixations and saccades, plus working memory assembling words into propositions. Efficiency usually drops not because one character failed to split, but because some joint of that pipeline is paying tax: size shrinks how many letters a fixation can take (visual span), measure turns the return into a bet, leading lets rows steal features, tracking breaks the word as a clump, all-caps pinches word-shape. These are layout variables. They can make a passage slow and tiring while every character is still discriminable.
So the main causes of readability are often not “a more recognisable alphabet” but whether size, measure, leading, contrast, and weight-texture leave the paragraph a track that can be walked to the end. Tinker’s print work measured those as joint conditions of page speed; Dyson on screens kept asking how measure and scrolling change efficiency. Letter design does enter — extremely condensed, decorative, or heavy body type leaves less structure a fixation can absorb — but it is supporting; layout is the lead.
Chinese continuous reading adds a further tax: no word spaces, so segmentation is entirely on the reader, and measure and punctuation rhythm press working memory differently from Latin. Readability parameters tuned on English paragraphs cannot be assumed to transfer onto Han.
Studying it
Materials have to be coherent paragraphs, long enough to accumulate fixations and returns, followed by comprehension items. Independents: size, measure (character count), leading, alignment, face, contrast, scrolling or not. Dependents: speed (words or characters per minute), comprehension accuracy, regressions, rated effort, whether the passage was finished. Speed alone biases: some people scan fast and miss every question.
Eye tracking can give mean fixation duration, visual span, return landings. Without a tracker, pauses and skips in oral reading, plus timed comprehension, still measure efficiency. Do not use random letter strings for readability — those are legibility materials, and they strip the contextual advantage of continuous reading.
Where it stops holding
A three-word button, a nav item, a badge has no continuous text; talking readability is the wrong layer — go back to discrimination and size. Exhaustion, glare, a shaking carriage set a floor that viewing conditions, not layout, decide. Skilled readers take a word-shape route on familiar words and can “scan past” poor setting — novices, long legal sentences, and second-language readers expose readability failure more readily. Shallow comprehension items let high speed impersonate readability; the questions have to ask for information the paragraph actually carries.
Applying it
- For any explanation, terms, or article longer than a few lines, accept it with a reading task: finish under a time limit, answer two or three questions that cannot be guessed without reading. If they cannot, inspect size, measure, and leading before swapping in a decorative face.
- Treat readability knobs as a set. Change size, recount characters per line and leading. Do not turn one knob alone.
- Measure English and Chinese body separately. A measure and leading tuned on English do not automatically transfer.
- Mark where people start reading and where they quit. Quits clustered on one kind of page (unbroken blocks, overlong lines, tight leading) are readability dying on that page, not impatient users.