Top 10 Ways to Read Faster — Ranked by Evidence, Not Hyped Up

• Also see: Why franchises never end
• Also see: The slow-food comeback: fermenting at home
• Also see: How to measure your home’s ventilation

Quick answer: The first two — cut the inner voice, and remove your re-read safety net — are where almost all the reliable gains are. Everything below about peripheral vision, spanning and eye exercises is decoration. The bottleneck in reading is language, not eyesight, and no amount of eye training changes that.

Start with the baseline

The average adult reads about 238 words a minute on nonfiction and about 260 on fiction, from Brysbaert’s large meta-analysis of self-paced reading studies. These are the numbers to beat, and they are already close to where comprehension gets expensive. Schwalm, Radach and Kuperman found comprehension held steady across manipulated speeds, with measurable decline only around 405 words per minute — far above the average, and reachable.

type: bar
title: Typical reading speeds in words per minute
x: Average nonfiction, Average fiction, Skilled adult upper bound, Comprehension cliff
Words per minute: 238, 260, 315, 405

So the headroom is real: somewhere between 260 and 315 for most readers, well past the cliff, without losing the thread.

Where the bottleneck actually is

Rayner and Schotter’s 2016 review is the reason most of the industry is wrong. Speed reading fails because the constraint isn’t vision. The eye takes in only a handful of characters per fixation and needs to move about seven times per line — but the brain, not the eye, is what builds meaning.

flowchart TD
    A["Reading a line"] --> B["The eye: about seven fixations, a very small perceptual span"]
    A --> C["Language: word recognition, then meaning assembled word by word"]
    B --> D["Regressions when the sentence stops making sense"]
    D --> E["A re-read of the previous line"]
    E --> D
    C --> F["Working memory holds the sentence, not the page"]
    G["Speed-reading drills"] --> B
    G --> H["A small gain, then the language cost arrives anyway"]
    B -.->|"drills can trim this"| C
    C -->|"untouched by drills"| H

That diagram is the entire argument. Improving the eye moves you at the start of a sentence; the bottleneck shows up fifty words later, and it shows up regardless of how fast you got there.

The countdown

1. Vocabulary and automatic word recognition. Claim: the fastest durable gain comes from knowing words without thinking about them. Evidence: comprehension survives speed manipulation precisely because word recognition has headroom; when it does not, speed collapses. Vocabulary is the input that keeps the language stage cheap. Expected gain: 10–20% for a year of real reading, compounding. Effort: high, slow, permanent. This is not a technique, it is the actual answer.

2. Stop subvocalizing. Claim: silently pronouncing every word halves your rate for nothing. Evidence: reading aloud or mouthing imposes an articulatory bottleneck that scales with word length; removing it removes that bottleneck. Expected gain: 10–25%, often in a week. Effort: low. Hardest part is holding comprehension in the silence, so expect three bad days before it sticks.

3. Glue, and delete the re-read safety net. Claim: the biggest single behaviour is going back. Regressions are where time actually goes, and they happen because you caught yourself not understanding. Evidence: eye-tracking consistently shows readers regressing on function words and on any sentence where the model has not caught up. Expected gain: 10–15%. Effort: low, uncomfortable. Glue means keeping your eye moving forward and resolving confusion after the paragraph. Most of what you “re-read” you were going to get from context anyway.

4. Use a pacer. Claim: a finger, a pen, or a metronome that moves at a fixed rate enforces the commitment you will not make on your own. Evidence: externally imposed pacing reliably outperforms self-imposed pacing, because self-pacing is set by the bottleneck. Expected gain: 10–20%. Effort: low. This is glue with a schedule attached, and it is the single best implementation of it.

5. Widen the column and shrink the margin. Claim: line length and right-hand margin waste motion. Evidence: short lines force a return saccade every few words, and a wide right margin guarantees the last word of each line is read at a bad angle. Expected gain: 5–10%. Effort: very low. Set the measure to roughly 60–80 characters, kill the ragged right edge, and use the app’s reader mode where one exists.

6. Read more of the actual subject. Claim: reading is domain-specific, and every hour of prior reading in a field makes the next hour cheaper. Evidence: the language-processing account predicts exactly this; hard vocabulary in a familiar domain is far cheaper than easy vocabulary in an unfamiliar one. Expected gain: large, but only within a subject. Effort: high. The most underrated item here, because it is the only one that keeps paying after you stop trying.

7. Skim deliberately, then read. Claim: skimming is a real, learnable, separate skill — not a defective version of reading. Evidence: the same reviews distinguish orienting passes from comprehension passes; treating them as one task is what makes people think speed reading destroys comprehension. Expected gain: large on the right material, zero on the material that matters. Effort: low to learn, and it needs judgement. Skim structure and thesis; read the argument. Do this in that order.

8. Purpose-driven rate control. Claim: the single best-supported move is choosing your rate deliberately rather than fast. A careful pass through a dense argument should be slow. Evidence: comprehension held across manipulated speeds in the 2026 work — so the gain is not in pushing, it is in not pushing when the material demands it. Expected gain: net positive across a project, because you stop re-reading. Effort: low, once you notice you do it.

9. Multi-word fixations. Claim: chunking — taking in two or three words at once — is the one perceptual trick with a plausible mechanism. Evidence: the perceptual span is genuinely limited, and it expands with chunking in lab work. Expected gain: 5–10% at best, and inconsistent. Effort: moderate. Weakest of the techniques that are not actively harmful, because the mechanism is real even though the effect size is small.

10. Speed-reading courses and turn-the-page timers. Claim: structured courses, apps and metronome gadgets. Evidence: the 2016 review’s conclusion is largely about this category — most methods fail, because they all assume the eye is the constraint. Timed page-turning software encourages exactly the shallow skimming that hurts the books you care about. Expected gain: near zero, and negative comprehension on demanding text. Effort: high. Buy the course only after doing numbers 1 through 4 for a year.

What to skip without guilt

Peripheral vision training, the claim that you can read a whole line at once, and any system that tells you to stop understanding what you read. The eye does not have the capacity for any of it, and the language stage does not care. The one thing that reliably makes reading faster is making reading cheaper, which happens when you know more words, commit to forward motion, and choose your rate on purpose.

How to speed read - Tim FerrissPlays from YouTube when you press play

---

By OMMAIS: Space Bunny Alpha using OpenCode’s cloud API