Is Your Progress Chart Climbing, or Just Zigzagging? How to Tell a Trend From Noise

You've got twelve points on a chart. The last three climb, the one before that dips, and somewhere in the middle sits a flat week — the exact week you worked hardest.

You stare at the shape and try to wring a verdict out of it. Working or not working.

The problem is that a chart tells you nothing until you know how much it moves on its own.

Two Different Numbers, Nothing In Between

Time yourself today. Time yourself tomorrow, on a text of the same difficulty, at the same hour. Don't train anything in between. The two numbers will differ — sometimes by twenty or thirty words per minute.

That's not a stopwatch malfunction. The time you spend on a word depends on how familiar it is, how long the sentence is, and how much you already knew about the subject. Just and Carpenter put this in writing back in 1980: the eye stays on a word for roughly as long as it takes to process it, so the text sets the pace more than your intent does. Rayner, synthesizing two decades of eye-tracking research in 1998, describes fixation durations scattered across a wide range even within the same reader.

Add last night's sleep, the time of day, and whether you managed to eat. Out comes a number that jumps around on its own, with nothing about your actual capacity having changed.

Brysbaert pulled together almost two hundred studies in 2019 and found a wide spread of reading speeds across adults — and those numbers are worth reading on their own. What matters here is different: that same spread lives, at a smaller scale, inside a single person too, from one day to the next.

So the first question you ask the chart isn't "how much have I improved." It's "how much do I vary anyway."

How to Find Your Spread, in Four Evenings

Four measurements, on four different days, on comparable passages from the same book, with no training in between. What a correct measurement actually looks like is its own discussion; here it's enough that all four are done the same way.

Then you look only at the smallest and the largest. The gap between them is your noise band.

A made-up example, so the arithmetic is visible. Say you get 236, 228, 254, and 241 words per minute. The minimum is 228, the maximum 254, the gap 26. Against the midpoint of the series, somewhere around 238, that's nearly 11 percent.

You've learned two things, one useful and one uncomfortable. The useful one: any gain under 26 words per minute is, for you, a good day, not progress. The uncomfortable one: your best day with zero training was already 254. To feel good about a peak of 256 after a month of work, you'd have to forget you hit almost the same number doing nothing at all.

The band gets measured before the program starts, not after. And watch the first two points: the second time around, you're simply better at the test itself, which inflates the number without you having changed. The effect is strong enough that it's worth taking out of the calculation, so if you can, take the band from your last three measurements, not all four.

The Newest Point Is Your Worst Witness

You look at the chart precisely because a new point just landed on it. And that point is a single draw from your band, carrying the luck or bad luck of exactly one day.

The peak is even more misleading. You remember it because it's the highest, which means you picked it out using exactly the criterion that guarantees that day also had the best alignment: more familiar text, a rested mind, phone on silent. A peak isn't your level. It's your level plus everything that happened to line up that day.

This is where a disappointment comes from that you'll almost certainly live through. An exceptional day is usually followed by a weaker one. You haven't lost anything overnight — you've just drifted back toward the middle of your own band. The phenomenon has a name, regression to the mean, and it's a property of any series of numbers with variation in it, not a flaw in you.

What you read instead of the last point: the median of your last three measurements, compared against the median of your first three. Median, not average — so one catastrophic day or one euphoric day can't drag the whole result along with it.

Again on made-up numbers: the first three points are 251, 268, 249, so the median is 251. The last three are 275, 262, 281, so the median is 275. The difference is 24 words per minute — under the 26-word band you measured for yourself. The honest verdict isn't "I gained 30" (281 minus 251), it's "maybe I gained a little, but it doesn't stand out from noise yet, so I'll keep measuring for two more weeks."

A Curve That Flattens Isn't a Curve That Stopped

The typical shape looks the same for almost everyone. Two or three weeks of visible climb, then something much flatter.

The explanation isn't a romantic one. At first, you're not learning to read faster — you're stopping doing things that cost you for nothing: pausing on filler words, reflexive backtracks that fix nothing, a gaze that lands on the first letter of a word instead of the middle. That waste is a finite quantity. You wring it out in a few weeks, then there's nothing left to wring.

Underneath that sits the slow piece. Kuperman and Van Dyke showed in 2011 that differences in vocabulary and word-recognition speed show up directly in the pattern of eye movements. That doesn't train with two weeks of drills. It's built by reading, over years.

So a plateau after the first month is the expected shape of your curve, not proof that you've failed.

The distinction that matters on the chart is between a plateau and a decline. A plateau is flat with the same spread as before: the points jump up and down, but around the same level. A real decline shifts the whole band lower, for several measurements in a row. When you see that, look first for the boring explanations: you switched to a harder book, you're measuring in the evening after two hours of reading, you had a cold.

A Chart With Only One Line Is Half a Measurement

This is the flaw that ruins most progress charts, including the nicely drawn ones.

If your chart has a single line and that line is speed, you can push it up tomorrow with zero training. Move your eyes faster, skip whatever's hard, done. Carver gathered the evidence on the relationship between speed and comprehension back in 1990, and the synthesis by Rayner and colleagues in 2016 is blunt about it: large speed gains almost always come from skimming. Duggan and Payne measured in 2009 exactly what gets lost: under time pressure, the main ideas survive and the details disappear.

That's why the chart needs two lines: speed, and a comprehension score taken the same way every time.

The arithmetic is simple and unpleasant. Speed climbs from 240 to 290 words per minute, which looks fantastic on screen. Comprehension drops over the same period from 70 to 55 percent. The product of the two falls from 168 to roughly 160. You read worse, and the speed line congratulates you for it.

One warning about that second line: it has its own noise, and it's coarser than the first. If you check comprehension with ten questions, the smallest possible step is 10 percent, so a single different answer shows up on the chart as a collapse. Read the comprehension line even more conservatively than the speed line: what matters is the direction across three or four points, not the jump from one to the next.

What Tells You to Keep Going, What Tells You to Change Something, and What Tells You to Stop

You write the threshold down before you look at the data. Otherwise you'll move it exactly as much as you need to make it flattering — and everyone does that, without any bad faith involved.

Keep going if the median of your last three points has beaten the median of your first three by more than your band, and the comprehension line has held steady or risen. That's it. You don't need more than that.

Change something in two situations. When speed rises and comprehension falls, you haven't gotten faster — you've swapped tools: you've started skimming without noticing, and the fix is to redo the measurements with the explicit instruction to read to remember. When both lines sit still for four or five weeks, with the routine done conscientiously, the problem is no longer how much effort you're putting in but what exactly you're practicing. Change the type of exercise or the material you're working with.

You stop when, after a month and a half, everything is still inside the band. That's not a defeat — it's information that few people get this cheaply. That time goes further invested in pages read, because that's where the slow piece that's actually limiting you grows.

There's one more outcome the curve never shows you. In these months, you read more books than you did the year before. The chart doesn't have a line for that, even though it was, probably, the reason you started.

See what reading training with two measurement lines looks like