Start by defining the reading task
A search task, a story read for pleasure and a chapter studied for an examination ask different things of a reader. Before timing, specify what counts as success. Otherwise a fast skim may be compared with careful learning, producing a dramatic number but an unclear conclusion.
Write down the conditions: the kind and length of material, the purpose, whether the passage remains visible during questions and whether rereading is allowed. Keep those conditions similar when measuring change. If one session uses a phone and another a wide monitor, record that too.
Research methods can themselves alter the task. A reader looking freely at a page, advancing through self-paced segments and watching a timed word stream does not have the same control in each case. Methods research helps explain why their timing measures should not be treated as interchangeable. [1]
Calculate words per minute transparently
For a fixed passage, divide the number of words by the reading time in minutes. Equivalently, multiply the word count by 60 and divide by seconds. Record what your system counts as a word and whether headings, captions and notes are included.
600 words in 150 seconds
600 × 60 ÷ 150 = 240 words per minute. This describes the timed exposure to that passage. It does not, by itself, establish how much the reader understood or remembered.
Start the clock when the reader can begin and stop it at the defined finish. Do not quietly omit time spent on difficult parts while including all of the words. If the task permits pausing or looking up terms, say whether that time is included. For study, the whole-task time may be more relevant than the time spent looking at the passage alone.
Across writing systems and languages, word boundaries and word lengths can differ. A character-based estimate and a counted English-word rate are different measures. Label them clearly before comparing them.
Use questions that test the intended understanding
A reading test needs questions that depend on the passage. If an answer can be guessed from the title or common knowledge, it provides little evidence that the text was processed. A useful set samples more than one aspect of the message.
| Question type | What it can reveal | Example |
|---|---|---|
| Literal | Access to a stated detail. | Which condition changed? |
| Inference | A connection implied by the passage. | Why would the second proposal cost less? |
| Explanation | A coherent account in the reader’s own words. | How does the evidence support the conclusion? |
| Application | Using the idea beyond its original wording. | Would the rule apply to this new example? |
| Delayed recall | What remains accessible later. | Explain the main process tomorrow. |
Keep access to the original passage consistent. An open-text question may measure finding and interpreting information; a closed-text question also places demands on memory. Neither is inherently invalid, but they answer different questions.
Keep speed and comprehension as separate results
Suppose the same reader tries three comparable 600-word passages under the same instructions. These invented results illustrate the arithmetic, not an experiment or a prediction.
| Session | Time | Rate | Questions correct |
|---|---|---|---|
| A | 180 seconds | 200 WPM | 8 of 10 |
| B | 120 seconds | 300 WPM | 5 of 10 |
| C | 150 seconds | 240 WPM | 8 of 10 |
B is the fastest reading rate, but it does not preserve A’s question performance. C looks more promising for a task requiring that level of understanding. Even so, one passage per condition is not enough to establish a stable improvement: material difficulty and ordinary variation may explain part of the difference.
A combined “effective speed” score can hide this distinction. Multiplying rate by a percentage creates a convenient index, but it does not make a missed critical condition equivalent to a small timing change. Show the underlying measures as well as any summary.
Small tests produce jumpy percentages
With five equally weighted questions, one answer changes the score by 20 percentage points. With ten, it changes the score by ten points. That makes a short quiz useful for feedback, but a fragile basis for a precise claim about a reader’s ability.
Question difficulty matters too. Five direct recognition questions are not equivalent to five questions requiring explanations of unfamiliar relationships. If multiple-choice options are easy to eliminate, correct answers may overstate what the reader could explain unaided.
Use several passages and sessions. Look at the pattern, the kinds of errors and whether the result persists. A very short passage also magnifies small start/stop timing differences. Report an ordinary result and its variation, rather than choosing only the best trial.
For important learning, include a later check. Prose-learning experiments show that immediate and delayed performance can favor different study activities. A good immediate result is not a substitute for measuring the retention you actually need. [2]
Build a baseline that cannot be improved by familiarity alone
Use previously unread material for a progress comparison. Reusing the same passage allows memory of its content, layout or questions to contribute. That may be useful for learning the passage, but it does not isolate an improvement in general reading.
Choose passages of roughly comparable length, language demands and topic familiarity. Record anything unusual: a difficult term, a distracting interruption, a known subject or a mistaken timer start. Do not discard an inconvenient result without a reason you would also apply to a flattering one.
If you want to evaluate a program rather than your own routine, a comparison group becomes important. Repeated testing, ordinary reading and expectation can all contribute to change. Random allocation and consistent testing help distinguish those alternatives, while a personal before/after record remains a personal progress check.
A presentation rate is not a measured reading rate
A display set to 600 WPM is scheduling words at that rate. It has not established that the reader understood them at that rate. Pauses, repetitions, word lengths, punctuation delays and the opportunity to look back can change the experience substantially.
Research comparing rapid sequential presentation and skimming has treated comprehension as an outcome of its own. The method and its pauses are part of what is being tested. [3]
When reporting a result, distinguish the setting, the actual timed passage and the question performance. If the reader pauses to think, that can be a sensible part of reading; simply label how it affects the reported rate. A useful interface makes the distinction clear instead of making the largest number the headline.
Use reference numbers with context
Brysbaert’s large synthesis estimated adult English silent-reading averages of 238 WPM for nonfiction and 260 WPM for fiction. The estimates describe populations and tasks, not a universal standard everyone should reach. [4]
For your own record, the most useful comparison is usually with your earlier performance on a similar task. If the goal is reading work reports, a test on easy fiction may not be the best reference. If the goal is enjoyment, a slower pace may be a choice rather than a limitation.
Keep the date, passage or topic, word count, time, comprehension result, type of questions, access to the text and one sentence about what was difficult. Add a later recall check when retention matters. This is enough to start asking more reliable questions of your results.
Common questions
What is a good WPM?
There is no single target for every task. Interpret the rate alongside the material, purpose and comprehension outcome. A useful improvement is less time for the understanding you need, or better understanding within the time available.
Can I compare two different app tests?
Only after checking how they count words, time the task, permit rereading and assess understanding. A presentation setting, a character estimate and a timed reading rate may all be labeled WPM while measuring different things.
How many passages should I use?
Use more than one and repeat the comparison across sessions. There is no universal minimum that makes every personal test conclusive; the aim is to see whether a pattern survives changes in passage and ordinary day-to-day variation.
Sources for this page
These links lead to the original paper, publisher or author archive. For annotations and limitations, visit the research library.
- Paradigms and processes in reading comprehensionMarcel Just, Patricia Carpenter and Jacqueline Woolley (1982). Methods and theory
- Test-enhanced learning: Taking memory tests improves long-term retentionHenry Roediger and Jeffrey Karpicke (2006). Experiments
- Conceptual processing of text during skimming and rapid sequential readingMichael Masson (1983). Experiments
- How many words do we read per minute? A review and meta-analysis of reading rateMarc Brysbaert (2019). Meta-analysis