Once the stopping for unknown words has gone, the largest remaining cost is going back. The eye returns to a clause it has already crossed, re-reads it, and carries on, and the reader has no memory of having done it. Regressions are invisible from the inside and they are where the minutes are.
The reason they are invisible is that the eye does not move smoothly across a line at all. It jumps and stops, jumps and stops, taking in a chunk at each stop, and a backwards jump feels no different from a forwards one. You cannot notice a regression by paying attention, because paying attention is what causes several of them.
Eight facts about reading faster
- slow reading is usually a vocabulary problem, not a technique problem, and no amount of eye training fixes an unknown word.
- two hundred words a minute is normal for a competent adult reader in their own language, and claims far above that involve skimming.
- subvocalising slows you to speaking speed, which is a real ceiling, and it is very hard to suppress in a second language.
- read in phrases, not words. The eye takes in a group at a time, and forcing single words is what makes reading feel like decoding.
- do not reread automatically. Going back over a line you understood is a habit rather than a need.
- preview the structure first, reading headings and first lines, which makes the full pass considerably faster.
- speed varies with purpose. Finding one fact, following an argument and enjoying a novel are three different speeds.
- a fixed time beats a fixed length as a daily target, because it is the only one you can honestly keep.
Reading in phrases is what fewer stops looks like
A reader who takes in three or four words at each stop crosses a line in three or four jumps. A reader taking in one word at a time makes twelve, and each stop has a fixed cost regardless of how much was collected in it. That difference is the whole of what reading in phrases means, and it is a description of what fluent reading looks like from the outside rather than a technique you can adopt.
Which is why the advice cannot be applied directly. Nobody widens their chunk by deciding to. The chunk widens when the words inside it are recognised fast enough to arrive together, so it is a consequence of familiarity, and the only lever available is how much you have read.
The one lever that does work directly is preventing the return, because a regression is a movement rather than a recognition. Cover the line above with a card, or move a finger along at a constant pace and follow it, and the eye cannot go back. That is a physical constraint on a physical habit, and it is the only speed technique on this page that does anything.
Where it stops: re-reading as diligence
The stall is treating the return as care. A sentence goes slightly out of focus, so the reader goes back and reads it again, and it feels responsible: better to be sure than to press on half understanding.
What actually happens is that the second reading is not much better than the first, because the difficulty was usually a word rather than the syntax, and the same word is still there. Meanwhile the thread of the paragraph has been dropped and picked up again, which costs more than the clause was worth. Readers who press on recover the sense two sentences later from what follows, and readers who go back arrive there anyway, later.
The habit is reinforced by every reading test anybody has ever sat, where accuracy was scored and time was not.
A clean column, which is not cosmetic
The reader mode built into every browser strips a page to text in a single column of moderate width, free and with one click. That width is the point rather than the tidiness: a line long enough to fill a wide screen makes the sweep back to the start of the next line unreliable, so the eye lands on the wrong line and has to search, and that search is a regression caused entirely by layout.
It replaced printing a web page in order to read it, which is what people used to do to get a clean column and is a strange thing to have needed. Its limitation is real: it strips images, tables and captions, and on some pages those were the content rather than the decoration.
What it looks like when it goes wrong
The recognisable failure is finishing a page and being unable to say what was on it. That is not a memory problem. It is usually a page read in a large number of small backwards steps, where no continuous run was ever long enough to build a thread.
The second failure is buying speed by skipping. A reader pushed to go faster starts dropping the small words, and the small words are where the grammar is, so the pace goes up and the comprehension falls off a cliff. The measure to watch is not words per minute on its own; it is words per minute at a comprehension you would be willing to be tested on.
Whether any of this is your bottleneck at all depends on how many words per page you are stopping for, which is settled in getting through more of it without losing the meaning, and the supply of pages at a density where the eye can run is books written to your level, and how to pick one.
What none of this touches is the decision you make at the unknown word you did not stop for. Pressing on is only useful if something arrives to fill the hole, and there is a procedure for that which is more specific than hoping: getting the sense without reaching for a dictionary.