Subvocalization: what your inner reading voice does

Subvocalization is the silent speech you hear in your head as you read. Many speed-reading books tell you to eliminate it, on the claim that it slows you down (Rayner et al., 2016). A review of the research by Rayner and colleagues (2016) found that inner speech helps you identify and understand words, and that attempts to suppress it hurt comprehension when a text was difficult and required inferences.

This guide covers what subvocalization is and what happened in the studies that tried to suppress it. It also covers the cases where relying on it less is reasonable, with steps you can check against your own numbers in the free reading speed test.

What is subvocalization?

Subvocalization, also called inner speech, is an abbreviated form of speech that the reader experiences and other people cannot hear, and it may involve no visible movement of the mouth (Rayner et al., 2016). Researchers study it at two levels. The first is muscle activity: in a 1966 study, electrodes on the skin over the larynx recorded the speech muscles during silent reading, and some readers showed more activity than when they were relaxed (Hardyck et al., 1966). The second is the inner voice itself, a sound code that, in a review by Leinenger (2014), appears to be separable from the movement of the speech muscles.

The inner voice also resembles your own. Filik and Barber (2011) tracked the eye movements of 26 readers from the north and south of England as they read limericks whose final word rhymed in one accent and not in the other. Reading was disrupted when the final word failed to rhyme in the reader’s own accent, which suggests inner speech carries the reader’s pronunciation (Filik & Barber, 2011).

Does everyone subvocalize?

Readers process the sounds of words even when that leads them into errors. In a study reviewed by Rayner et al. (2016), people judging at speed whether sentences made sense accepted “She has blonde hare” about 12% of the time, against about 5% for “She has blonde harm”. The authors read this as a sign that readers cannot switch the sound of words off, and that phonology is fundamental to reading (Rayner et al., 2016).

The conscious experience varies more. In 160 posts from a question-and-answer website analysed by Vilhauer (2016), many people described an inner reading voice with qualities of speech such as pitch and emotional tone. Some recognized the voice as their own and others as someone else’s (Vilhauer, 2016).

Can subvocalization be detected?

The muscle side can be measured. Electromyography (EMG) records it through electrodes on the skin over the throat (Hardyck et al., 1966). The inner voice is harder to record, so researchers infer it from behavior, such as the eye-movement disruptions in the limerick study (Filik & Barber, 2011).

Why speed-reading courses target it

Speed-reading proponents describe inner speech as a habit carried over from learning to read aloud, and as a drag on reading speed (Rayner et al., 2016). Many of the speed-reading books that Rayner and colleagues surveyed recommend eliminating it. Their authors are vague about what that would mean (Rayner et al., 2016).

Early lab work seemed to back the idea: in the 1966 study, readers who showed speech-muscle activity were given auditory feedback of that activity, and the feedback produced an “immediate and long-lasting” stop to it (Hardyck et al., 1966). The authors expected the method to help treat some reading problems (Hardyck et al., 1966).

A related idea holds that the inner voice limits silent reading to the speed of speech. Reading rates do not fit that limit. A meta-analysis of 190 studies by Brysbaert (2019) puts the average adult silent reading rate in English at 238 words per minute (WPM) for non-fiction, against 183 WPM for reading aloud. Leinenger (2014) also reviews evidence that inner speech runs faster than overt speech, which suggests it can run ahead of the speech muscles.

What the research says about suppressing it

Researchers have tried to block inner speech in two ways. One trains readers to quiet their speech muscles, and the other keeps the mouth busy with another sound. Both methods lowered comprehension, with the clearest losses on difficult text and on questions that required connecting ideas.

Muscle feedback studies

In a 1970 follow-up described by Rayner et al. (2016), Hardyck and Petrinovich played a tone whenever a monitor picked up speech-muscle activity and asked readers to keep the tone off. Readers managed this to some extent. Their scores on a later comprehension test were unaffected for easy text and lower for more difficult text (Rayner et al., 2016). A comparison group whose arm muscles were monitored the same way showed no loss, which suggests the cost did not come from the extra task of listening for a tone (Rayner et al., 2016).

Leinenger (2014) adds two caveats: the participants were students in a remedial English class who may have struggled with the texts to begin with, and it is unclear whether comprehension fell because readers could not subvocalize or because holding it back took effort.

Articulatory suppression studies

The second method, articulatory suppression, has readers count aloud or repeat a word while they read (Leinenger, 2014). In experiments by Slowiaczek and Clifton (1980), counting or saying “cola-cola-cola” aloud impaired reading comprehension but, in most tests, not listening comprehension. The damage was specific to tests that required integrating concepts within or across sentences, while memory for individual word concepts held up (Slowiaczek & Clifton, 1980).

Daneman and Newson (1992) applied the method to long passages of natural prose. Counting aloud while reading hurt comprehension and recall of both the details and the gist, with an average drop of 10% to 12% in comprehension performance (Daneman & Newson, 1992). Tapping a sequence with the fingers did not produce the same loss (Rayner et al., 2016), which Daneman and Newson took as evidence that the cost came from competition for speech-related resources (Daneman & Newson, 1992). The size of the effect differed a great deal from one reader to the next (Daneman & Newson, 1992).

Suppression also makes errors harder to catch. In studies by Baddeley and colleagues (1981), reviewed by Leinenger (2014), readers who repeated a sound aloud were more likely to accept a sentence with an anomalous word as meaningful, and missed more swapped word pairs. Leinenger concludes that the inner voice may help readers keep track of exact word order (Leinenger, 2014).

These experiments have limits. Counting aloud adds a second task, so part of the cost may come from the extra load, and the inner voice may persist while the mouth is busy (Leinenger, 2014). Even so, Rayner et al. (2016) conclude that attempts to eliminate inner speech have impaired comprehension when texts were difficult and required readers to make inferences.

Is subvocalization bad?

For understanding what you read, subvocalization helps: inner speech plays an important role in identifying words and understanding text during silent reading, and translating what you see into sound helps readers understand it (Rayner et al., 2016).

The idea that it is a reading problem has a history. In 1969, Hardyck and Petrinovich suggested that the inner voice hinders fluent reading and proposed treatment programs for children who subvocalized (Leinenger, 2014). Leinenger cites substantial evidence of a link between processing the sounds of words and success in learning to read, across several languages (Leinenger, 2014).

Rayner et al. (2016) put language skill at the heart of reading speed, and conclude that going from around 250 WPM to 500 or 750 WPM while understanding a text as well as before is unlikely. The guide to reading faster covers the habits that do raise speed.

When reducing it is reasonable

Relying less on the inner voice makes sense when full comprehension is not the goal, as when you need an overview of a text or a single fact from it. Reading faster than your normal rate means not reading the whole text, which is skimming (Rayner et al., 2016). Skimming can run two to four times faster than typical silent reading, at the price of lower comprehension (Rayner et al., 2016).

Familiar material is the best candidate. Rayner et al. (2016) describe effective speed readers as people who already know a great deal about the topic and accept the lower comprehension that comes with skimming at rapid rates. In the 1970 feedback study, comprehension of easy text held up while readers kept their speech muscles quiet (Rayner et al., 2016), and Leinenger (2014) reads that result as a sign that readers can understand easier texts without them.

How to stop subvocalizing, and what to try instead

A drill that has you count or repeat a word while you read resembles the lab method: counting aloud or repeating “cola” is how Slowiaczek and Clifton (1980) and Daneman and Newson (1992) blocked inner speech, and comprehension fell in both studies. Biofeedback quieted the speech muscles in 1966 (Hardyck et al., 1966), and the 1970 follow-up found lower comprehension of difficult text once readers kept them quiet (Rayner et al., 2016).

A better plan measures your reading first and saves speed for the material that can take it:

1. Measure on easy and hard text. Take the reading speed test once at Easy difficulty and once at Hard, and note your words per minute and comprehension score each time. If a new habit raises your speed while your Hard score falls, that habit has turned reading into skimming.
2. Skim on purpose. For familiar, low-stakes material, read the headings and the first sentence of each paragraph. Slow down where you find what you need (Rayner et al., 2016).
3. Pace short texts with the speed reader. It shows one word or a small group of words at a time, at a pace you set. Set it near the WPM from your test. The display is unlikely to silence your inner voice: readers who saw sentences one word at a time at 720 WPM still appeared to generate sound-based codes for the words (Petrick, 1981, reviewed by Rayner et al., 2016).
4. Let the voice run on hard text. Read difficult material at your normal pace, and reread a sentence when its meaning breaks down.
5. Build vocabulary through practice. Rayner et al. (2016) conclude that the way to get through text faster while keeping comprehension high is to practice reading and become a more skilled language user, for example by growing your vocabulary. The how to read faster guide turns that into a routine.

Next steps

Keep your inner voice for anything you need to understand in full, and rely on it less when you skim material you already know. Get a baseline from the reading speed test and compare your Easy and Hard scores. Use the speed reader for short texts at your measured pace, and read how to read faster for habits that raise speed without the comprehension cost.