ashi krishnan

Subvocalization is the imperceptibly tiny activations of the muscles involved in speech. We subvocalize when we read or talk to ourselves, imperceptibly moving our tongue and larynx as though we’re speaking the words as we read them, even though we’re not making audible sounds.

This is true for sign language users too! Instead of tongue and larynx activation, we see activation of the fingers and forearms. I suspect other muscles heavily implicated in signing, like those in the face, activate as well, but I can’t find any studies on it.

(There really aren’t enough studies of sign language users in linguistics or cogsci. It took until the 60s for mainstream linguistics to accept what deaf folks long knew: ASL is a true language, with recursive grammar just as expressive as, but different from, spoken language.

This is surely connected to ableist theories that deaf people aren’t as intelligent as hearing people, or that sign language users can’t do <X> thing that spoken language users can. Unsurprisingly, these theories have repeatedly been shown to be wrong.)

Most of us learn to read by reading aloud or signing visibly. It seems this imprints a connection between text and speech acts that never goes away. You’ll often notice little kids narrating what they’re doing. Turns out, adults do too! We just learn to do it much less obviously.

If this is happening all the time, you should be able to detect it, right? And even do speech recognition on it? Yep. Here’s some NASA work on subvocal speech recognition: https://www.nasa.gov/centers/ames/news/releases/2004/subvocal/subvocal.html …

Subvocalization appears to be important to reading comprehension. Not recognizing individual words, but understanding the structure of sentences and making connections within the text. Some speed readers suppress (but can’t eliminate) subvocalization, but comprehension suffers.

In fact, this is one of the pieces of evidence for subvocalization: if you force people to keep their tongues pressed to the roof of their mouth when they read, their comprehension suffers vs people who were made to do some other annoying thing that doesn’t interfere with subvoc.

Which makes a lot of sense when you consider the relationship between language and movement. To move, your brain needs to plan a sequence of movements—it needs to compose a script. To comprehend the movement of others, your brain needs to derive their script from sensory data.

Movement planning and comprehension force us to serialize our internal states into a sequence of efferents (action commands), and also, to reduce sequences of perceived efferents into internal states. This is exactly the kind of stream processing necessary for language.