What Is Subvocalization? Why You Hear a Voice When You Read (and How to Reduce It)

What Is Subvocalization? (Quick Definition)

Subvocalization is the habit of silently sounding out words in your head as you read them. It's sometimes called inner speech, silent speech, or subvocal reading, and it recruits the same phonological loop, the part of working memory that handles verbal and auditory information, that you use when you talk or listen. Your vocal cords and tongue may even make tiny, measurable movements during it, even though no sound comes out.

Nearly everyone does this, including fluent, fast readers. It isn't a bad habit picked up in a classroom or a sign of reading below grade level. It's a byproduct of how most people learned to read in the first place: by sounding words out loud, then gradually internalizing that sounding-out until it became silent and automatic. The inner voice you notice while reading this sentence is the residue of that process, still running in the background decades later.

Is Subvocalization Bad? What the Research Actually Says

This is where most advice on subvocalization goes wrong. The typical blog post treats it as a bug: something to suppress, override, or train out of yourself as fast as possible. That framing skips over what the actual research on subvocalization and comprehension found.

Slowiaczek and Clifton's 1980 study, "Subvocalization and Reading for Meaning," published in the Journal of Verbal Learning and Verbal Behavior, tested what happens to comprehension when subvocalization is experimentally suppressed. The finding: blocking subvocalization measurably hurt comprehension of semantically complex or syntactically ambiguous sentences. In other words, your inner voice isn't just narrating passively. For difficult material, it's doing real cognitive work, helping you hold a sentence's structure in working memory long enough to parse what it actually means.

That doesn't mean subvocalization is uniformly helpful. On straightforward, familiar text, it mostly just constrains your speed without adding much comprehension benefit, since you'd understand the sentence just as well without narrating every word. The honest, evidence-backed position sits between the two extremes competitors usually offer: subvocalization isn't a defect to eliminate, and it isn't a habit you should leave completely alone either. It's a tool that helps with hard text and limits you on easy text, which is exactly why the right move is reducing reliance on it at speed, not treating it as an enemy.

Why You Can't (and Shouldn't Try to) Eliminate It Completely

Most pages on this topic promise you can "eliminate" subvocalization with enough practice. That promise doesn't survive contact with the research. Studies that suppress subvocalization, through techniques like articulatory suppression (occupying your mouth and tongue with an unrelated task, such as repeating "the-the-the"), don't switch it off. They interfere with it temporarily, and comprehension on harder material drops while that interference is active.

There are two honest reasons full elimination is the wrong goal:

It's mechanically difficult to fully suppress. The phonological loop is deeply wired into how your brain processes language, spoken or written. Willpower alone doesn't disconnect it, which is why gum-chewing and humming produce modest, inconsistent effects rather than a clean off switch.

It would cost you comprehension on hard text if you could. Per Slowiaczek and Clifton, subvocalization aids understanding of complex or ambiguous sentences. Fully eliminating it wouldn't just be difficult, it would make you worse at understanding exactly the material where you need the most help.

The realistic target isn't silence. It's reducing how much you lean on subvocalization as your reading speed increases, so it stops being the bottleneck on straightforward material while still being available when a sentence actually needs the extra processing. That's a mechanism-level goal, not a willpower goal, and it's why the next two sections focus on pace rather than suppression.

How Subvocalization Caps Your Reading Speed

Here's the mechanical reason subvocalization limits speed, stated plainly: you can't silently "say" words faster than you can actually say them out loud. Average speech rate runs around 150 words per minute in conversational speech. If your inner voice is narrating every word you read, your reading speed is tied to that same rough ceiling, plus a bit of headroom, since silent narration runs somewhat faster than speaking aloud.

Marc Brysbaert's 2019 analysis of reading rate studies puts average adult silent reading speed at approximately 238 words per minute for non-fiction text, closely matching the commonly cited 200 to 250 WPM range. That number isn't a coincidence: it sits just above normal speech rate, which is exactly what you'd expect if most readers' silent reading speed is being paced by an inner voice running close to talking speed.

This is also why simply "trying to read faster" rarely works on its own. If your eyes move faster across the page but your inner voice is still narrating every word, something has to give, usually comprehension, as your narration falls further and further behind what your eyes have already seen. The fix isn't forcing your eyes to move faster while your inner voice tries to keep up. It's changing the input method so your inner voice has less material to fully narrate in the first place, which is the mechanism behind the techniques in the next section.

Techniques to Reduce Subvocalization (Ranked by Evidence, Not Folklore)

Most lists of subvocalization fixes read like folk remedies: chew gum, hum a tune, touch your lips. Some of these have a real, if modest, mechanism behind them. Others are closer to a distraction than a fix. Here's how they actually stack up.

Finger pacing and gum-chewing aren't useless, but they work indirectly and inconsistently: gum-chewing occupies your articulatory muscles, which mildly interferes with subvocalization, but the effect is small and doesn't scale well as material gets harder. Finger pacing mostly helps by reducing regressions and keeping your eyes moving forward, a separate mechanism from subvocalization itself.

RSVP works differently and more directly. It doesn't ask your inner voice to compete with a wandering visual pace. It sets the pace itself, word by word, at a speed your inner voice either keeps up with or doesn't. That's a testable, falsifiable claim, and it's exactly what the next section lets you check for yourself.

Try It Yourself: How RSVP Reduces Reliance on Subvocalization

Here's the part every other page on this topic skips, because they don't have a working tool to back it up. Paste any paragraph, this article, an email, a news story, into the RSVP reader and run it at three speeds.

At 250 WPM, subvocalization keeps pace easily. You'll likely feel like you're narrating every word in your head, more or less in sync with what's on screen, and comprehension feels completely normal. This is close to average silent reading speed, so there's no mismatch yet between your eyes and your inner voice.

At 400 WPM, most readers notice their inner voice starting to compress or "skip." You're no longer narrating every word in full; syllables get clipped, function words like "the" and "of" blur past without full articulation, and you're relying more on rapid recognition than narration. Comprehension usually holds up well here, but the experience of reading has noticeably changed.

By 600 WPM, most readers can no longer subvocalize every word. What tends to happen instead is a shift: content words (nouns, verbs, the words that carry the actual meaning) still get some inner narration, while function words get processed visually without being "said" at all. This is the mechanism-level effect this article has been describing rather than just asserting: RSVP doesn't switch subvocalization off, it structurally outpaces the parts of it that aren't essential to meaning, at a speed you can test on your own eyes in the next thirty seconds.

Try it with a paragraph you actually care about understanding, not filler text. The effect is more noticeable on real content than on throwaway sentences.

How Fast Can You Realistically Get?

The honest ceiling matters as much as the honest floor. Research reviewed by Rayner et al. (2016) in Psychological Science in the Public Interest found that claims of reading at 1,000 or more words per minute with full comprehension don't hold up once you account for how eyes and brains actually process language. Comprehension depends on constructing meaning, and that process has its own bottlenecks that no amount of subvocalization reduction removes entirely.

What does hold up: readers using RSVP can reasonably sustain 400 to 700 WPM on straightforward material with comprehension staying close to baseline, roughly double to triple the average 200 to 250 WPM starting point. That's the realistic range for reduced-subvocalization reading, not a marketing number. Push much past it and comprehension starts to decline for nearly everyone, because you're no longer just outrunning your inner voice, you're outrunning the rate at which the brain builds meaning from language.

Where you land in that 400 to 700 WPM range depends heavily on the material. Familiar, low-density text gives you more room; dense or unfamiliar material narrows it, since the bottleneck shifts from "how fast can I recognize the words" to "how fast can I actually process new ideas." For a fuller picture of what's realistic for your reading level and how it compares across contexts, see the reading speed benchmarks page, and for the deeper research on eye movement and comprehension trade-offs behind all of this, see the science of speed reading.

If you haven't measured your own baseline yet, a quick reading speed test gives you a real number to compare against the ranges above, rather than a guess. And if subvocalization is only one piece of what's capping your speed, the speed reading techniques guide covers chunking, pacing, and the rest of the toolkit that works alongside it. Both start from the same page: the speed reading hub, which lays out what the research supports across the whole practice, not just this one mechanism.

Subvocalization isn't the enemy. It's a default setting built for speech-paced reading, and RSVP is the most direct, testable way to push past it. Run a paragraph through the reader above at three speeds and see exactly where your own ceiling sits.