Subvocalization is the voice you hear in your head when you read silently. Almost every speed-reading course calls it a bad habit and promises to remove it. The research points the other way: the sound of a word is part of how a skilled reader recognises it, and when experimenters block inner speech, understanding of anything difficult gets worse. You cannot switch the voice off, and you would not want to.
What you can do is stop over-articulating easy text, where the voice lingers on every word for no benefit. That is a smaller goal than the one usually advertised, and it is one you can actually reach.
What is subvocalization?
The term covers two things that are often mixed up. One is inner speech: the experience of hearing the words as you read. Rayner et al. (2016) describe it as an abbreviated form of speech that nobody else can hear and that may involve no visible movement of the mouth. The other is faint activity in the speech muscles, the lips, tongue and larynx, which sometimes accompanies silent reading and at its most obvious becomes lip movement or whispering.
Underneath both is what researchers call phonological coding: the mind works out how a written word sounds as part of recognising it. This is not a leftover from learning to read aloud. It happens automatically in skilled adult readers, including readers of Chinese, a script that does not spell out sounds letter by letter.
The evidence comes from mistakes people make with words that sound alike:
- Asked whether a briefly shown word belonged to a category such as food, skilled readers wrongly said yes to meet about 19 percent of the time, against 3 percent for the look-alike melt. The sound of meat had been activated.
- Reading silently and judging quickly whether sentences made sense, people accepted "She has blonde hare" about 12 percent of the time, compared with about 5 percent for "She has blonde harm".
If sound were an optional extra, readers under time pressure would simply drop it. They do not. A review of this literature (Leinenger, 2014) concludes that sound codes come into play early, before a word is fully recognised, and may go on to support memory and comprehension afterwards.
Does subvocalization slow you down?
The argument against the inner voice is simple: you can only speak at a certain rate, so if you "say" every word, you can only read at speaking speed. The measurements do not fit that argument. Adults read English aloud at about 183 words per minute on average, but read silently at about 238 words per minute for non-fiction and 260 for fiction (Brysbaert, 2019). Silent readers already outpace their own reading aloud with the inner voice running, which fits the description of inner speech as abbreviated and not a full performance of each word. See average reading speed for more on those figures.
The same meta-analysis notes that silent reading rates are in line with the fastest speech people can follow by ear. That points to the real limit. Reading speed is capped by how fast you can understand language, not by the eyes and not by the voice.
What happens when researchers suppress the inner voice?
It is hard to stop someone thinking a sound, but three classic experiments tried, each in a different way.
| Study | How inner speech was reduced | Result |
|---|---|---|
| Hardyck & Petrinovich (1970) | A tone sounded whenever sensors picked up speech-muscle activity; readers tried to keep it off. | Comprehension was unaffected for easy text and impaired for more difficult text. Monitoring an arm muscle the same way did no harm. |
| Slowiaczek & Clifton (1980) | Readers repeated irrelevant speech aloud while reading silently. | Questions about individual concepts were answered well; questions that needed ideas to be combined or inferences drawn were answered worse. |
| Daneman & Newson (1992) | Readers counted aloud or repeated a word while reading; a comparison group tapped a finger sequence. | Comprehension fell with the speaking task but not with tapping, so the cost came from blocking speech and not from doing two things at once. |
The pattern is consistent. On simple material you can get by with less inner speech. As soon as a text asks you to hold a sentence in mind, connect it to the previous one or work out what is implied, the voice is doing useful work. Rayner et al. (2016) sum it up by saying that inner speech is not a nuisance to be eliminated: turning print into sound helps readers understand it.
Suppressing inner speech costs nothing on easy text and costs comprehension on hard text. The inner voice is part of how you understand, not a brake on it.
Can you stop subvocalizing?
Not completely, and the popular methods for trying are worth a second look. The usual advice is to count "one, two, three" under your breath, hum, or repeat a syllable while you read. That is the same procedure the laboratory studies used to block inner speech, and the outcome there was lower comprehension. If one of these tricks seems to make you faster, check what you remember afterwards. People who read at several times the normal rate are skimming, as covered in does speed reading work?
There is one version of the habit that is worth dropping:
- Lip movement and whispering. Visibly mouthing words is reading aloud with the volume turned down, and reading aloud is the slower mode. Resting a finger on your lips while you read is a simple way to notice when it happens.
- Sounding out every word of easy text with equal weight. Giving the, of and and the same full inner pronunciation as the words that carry the meaning is slower than it needs to be.
What to do instead of eliminating it
These are practical ways to keep the inner voice light on easy material. None has been shown to change inner speech itself; they work, as far as they work, by keeping your pace and attention steady so the voice has less room to dawdle.
- Sort your reading by difficulty. News, email, familiar topics and light fiction are where a lighter voice is safe. Contracts, textbooks, poetry and anything you must reason about are not. Let the voice run there.
- Read in phrases. Aim your attention at groups of two or three words and let the inner voice stress the content words, the way you would when speaking quickly. More on this in chunking.
- Use a pacer. A moving highlight, or a finger under the line, sets a tempo slightly above your habit and stops you lingering.
- Try word-at-a-time display at a moderate speed. RSVP fixes the pace for you. You will still hear the words, and that is fine. Keep the speed where you follow nearly everything, because RSVP also takes away the chance to glance back.
- Build vocabulary. Words you know well are recognised faster, sound included. This is the slow route, and the one the research supports most clearly.
A two-week practice plan
Ten minutes a day is enough. Use easy text only: articles on subjects you know, or a novel you are reading for pleasure. The plan is a practical suggestion and has not been tested in a study.
- Day 1: baseline. Time yourself on about 500 words and work out your words per minute. Write a three-sentence summary from memory. Notice whether your lips or tongue move.
- Days 2–5: paced phrases. Read for ten minutes with a pacer set at your baseline, rounded to the nearest 25 WPM, showing two or three words at a time. Let the voice emphasise only the important words.
- Days 6–10: small steps up. Raise the pace by 25 WPM. Stay there for two days. Go up again only if you can still summarise each session in three sentences; otherwise go back down.
- Days 11–13: mixed text. Alternate five minutes of easy text at your new pace with five minutes of something demanding at whatever speed it needs, voice and all. Practise the switch.
- Day 14: re-measure. Repeat the Day 1 test on a new passage of similar difficulty and compare both the speed and the quality of the summary.
Expect a small change: a somewhat quicker, steadier pace on easy reading and a clearer sense of when to slow down. If your summaries get worse as the number goes up, the number is not telling you anything useful. The full method, including vocabulary and skimming, is in how to read faster, and related terms are defined in the glossary.
Frequently asked questions
Is subvocalization bad?
No. It is a normal part of skilled silent reading, not a fault left over from childhood. Studies that interfere with inner speech find that readers understand difficult passages less well, particularly when they have to connect ideas or draw inferences. Only the visible forms, such as moving your lips or whispering, are worth unlearning.
Do fast readers subvocalize?
As far as research can tell, yes. Experiments show skilled readers activating the sounds of words automatically, even when it causes mistakes with words that sound alike, which suggests it cannot simply be turned off. People who read at several times the normal rate are generally skimming, and they give up detail to do it.
Does chewing gum or humming stop subvocalization?
Occupying your mouth or voice is meant to get in the way of inner speech; that is the idea behind these tips. But laboratory studies that had readers count or repeat a word aloud while reading found lower comprehension, not better reading. If you try it, test yourself on what you read afterwards before you trust the extra speed.
Why do I move my lips when I read?
Lip movement is the most visible end of subvocalization: inner speech with the speech muscles joining in. It is not harmful, but mouthing every word is closer to reading aloud, which is slower than silent reading on average. Resting a finger on your lips while you read easy text is a simple way to notice the habit and ease off.
Does RSVP reading stop subvocalization?
There is no evidence that it does, and you will normally still hear the words. What a word-at-a-time display changes is the pace, which is set by the app and not by your habit. Studies find that comprehension holds up at ordinary reading rates and falls as the stream is sped up, so use it at a moderate speed.
Sources
- Rayner, Schotter, Masson, Potter & Treiman (2016). So much to read, so little time: How do we read, and can speed reading help? Psychological Science in the Public Interest, 17(1), 4–34
- Leinenger (2014). Phonological coding during reading. Psychological Bulletin, 140(6), 1534–1555
- Hardyck & Petrinovich (1970). Subvocal speech and comprehension level as a function of the difficulty level of reading material. Journal of Verbal Learning and Verbal Behavior, 9(6), 647–652
- Slowiaczek & Clifton (1980). Subvocalization and reading for meaning. Journal of Verbal Learning and Verbal Behavior, 19(5), 573–582
- Daneman & Newson (1992). Assessing the importance of subvocalization during normal silent reading. Reading and Writing, 4, 55–77
- Brysbaert (2019). How many words do we read per minute? A review and meta-analysis of reading rate. Journal of Memory and Language, 109, 104047