Why Your Second Reading Speed Test Is Higher

You took the test two weeks ago. You took it again yesterday. The number climbed by about fifty words per minute, and it looks good on screen.

Here's the bad news up front, so we don't waste time: some of that gain would have shown up even if you'd done nothing at all for those two weeks. Not all of it. But enough of it that it's worth subtracting before you celebrate.

Screenshots that say "from 200 to 500 in two weeks" are rarely deliberate lies. Far more often, they're badly run measurements that a well-meaning person believed. Below are the five places the false gain sneaks in, and in the second half, what a test looks like when you can trust the celebration at the end of it.

The Number Went Up, But You Didn't Change in Two Weeks

Reading speed doesn't live in your eyes. It lives in how fast you recognize words, and that recognition thickens through exposure — through months of reading, not fourteen days (Kuperman & Van Dyke, 2011).

That puts a common-sense ceiling on any jump. The adult average for non-fiction, from a meta-analysis of 190 studies, is 238 words per minute, with a typical range of 175 to 300 (Brysbaert, 2019). A hundred-word-per-minute jump in two weeks would mean you crossed a quarter of the adult distribution in the time it takes other people to finish a book.

What does real training actually produce, measured? In a recent controlled study, a commercial speed-reading app added around 42 words per minute over the control group (Klimovich, Tiffin-Richards & Richter, 2023). That's it. If you found 120 at home, the rest isn't coming from training. It's coming from how you measured.

The First Time You Learned the Test. The Second Time You Actually Took It.

The first measurement carries a stowaway. While you were reading, part of your head was busy with something else: where's the timer, how hard should I push, what counts as "done," am I allowed to go back to a line.

All of that learning happens inside the timed minute and gets deducted from your speed. The second time, you pay none of it. The procedure is already familiar, so your eyes just do their job.

That produces a higher number with no training behind it at all. The effect is well-known enough that serious studies don't look at the pre-test-to-post-test difference within a single group: the control group also takes the test twice, and it also improves. The training effect is the difference between groups, not the difference between your own two numbers.

How much this piece weighs shows up in that same 2023 study. A group that only listened to a one-hour lecture on how reading works, with no drills at all, gained 29 words per minute — and the difference from the group trained with the app wasn't statistically significant.

You can't run a control group of your own at home. But you do have a poor man's fix, and it works: burn the first measurement. Treat it as a rehearsal, and let your baseline be the second one.

What Text You Picked Matters More Than What You Trained

The material moves the number harder than any technique does. Fiction and non-fiction differ by about twenty words per minute even at the level of averages, and the gap between a detective novel and a chapter on tax law is much wider than that.

There's a subtler trap too: text you've already seen. If you retest on the same passage, or on one from the same book, part of the vocabulary and the sentence patterns are already warmed up. You're not building meaning from scratch anymore, you're recognizing it. Speed goes up, and the gain is real — it just belongs to the text, not to you. On the next page, the unread one, it disappears.

The rule that fixes both problems is short: new text, at the same difficulty. Not "the same length" — the same density.

Speed Without a Check Isn't Reading Speed

If the test didn't ask you anything about what you just went through, you didn't measure reading. You measured how fast you can move your eyes over lines, and that number can double any time, at the right price.

Just, Masson, and Carpenter showed as far back as 1980 that speed reading and skimming split apart exactly at the questions: on the stopwatch, they look identical. Carver (1990), who gathered the research on reading rate into a single synthesis, reaches the same uncomfortable conclusion: past a certain threshold, what you gain in speed you pay for in comprehension.

That's why the number that matters isn't raw speed, it's speed multiplied by comprehension percentage. A test that doesn't give you the second term gives you half a result, and the missing half is exactly the one that breaks first when you rush.

You Wanted a Higher Number, and You Had a Good Day

The fourth source is in your own head, and there's nothing shameful about it. You know what the number is supposed to show. You worked for it. So you push a bit harder than you would have on the couch on a Saturday, no timer running. You're not cheating at anything — you're just no longer measuring your reading speed, you're measuring your reading speed when something is at stake.

How badly we judge our own comprehension shows up in a report done for NASA Ames. A photoreading instructor with years of practice rated her own comprehension of a physiology text at 4.5 out of 5. On the actual quiz she scored 35%, where guessing alone gets you 25%, and afterward marked her own score down to 2 (McNamara, 2000). The feeling of having understood is the weakest measuring tool in the kit.

On top of that sits a selection bias you don't notice. You don't retest on the evening you're wiped out, you retest when you feel sharp, coffee in hand, phone on silent. The first measurement landed wherever it landed. The second is picked from the peak of your good days.

There's also something purely arithmetic that hits motivated people specifically. If you retested because the first number seemed low, chances are you caught a bad day the first time around. Any measurement that includes a bad day is followed, on average, by one closer to your real mean. Nothing improved. The dice just got rolled a second time.

What a Test You Can Actually Trust Looks Like

The good news is you don't need a lab. You need five decisions, made before the test, not after.

New text, the same difficulty. Same genre, same register, same density of technical terms — but unread. If you're picking your own passages, keep two or three in reserve from the same book, from chapters you haven't opened yet.

A comprehension check in the same test, with no going back to the text. A few questions about ideas, not about the number that appeared in paragraph two. Without them you don't have a measurement, you have a stopwatch.

Same time of day, same state. Note the time, how much sleep you got, whether you'd eaten. Not to run a scientific study, just so that a month from now you know what you're comparing against what.

Several measurements, and their median. Three, on different days. The average gets dragged around by a single bad session, the median doesn't. How much your number swings from day to day isn't written in any meta-analysis, because it depends on you. There's only one way to find out — measure a few times.

A threshold declared in advance. From your three baseline measurements, take the highest and write it down somewhere. That's the line. Any future number below it doesn't count, no matter how good it looks on the chart. Written down in advance, because after the fact, any result finds its own justification.

The full protocol for that first measurement — what text to choose and what invalidates it — is in the guide to measuring at home. And for where the noise ends and the trend begins, once you have four or five numbers, see how to read your progress chart without lying to yourself.

The Last Check: Where Comprehension Went

One move closes the discussion. Put the two pairs of numbers side by side — speed and comprehension, first measurement and second.

If speed went up and comprehension held steady, you made progress. If both went up, you made more progress than you thought. If speed went up and comprehension went down, you didn't progress: you changed the task. You started doing something different with the page and kept counting it with the same unit. The speed-accuracy trade-off is real and well documented, and the field's most authoritative synthesis says plainly that there's no silver bullet (Rayner et al., 2016).

The good news, at the end of all these filters, is smaller than a screenshot and worth more than one. Gains that survive a properly run measurement are counted in tens of words per minute, not hundreds. But they're actually yours.

→ Measure your speed and comprehension, free