A recent paper, Inner speech captures the perception of external speech, by M Scott and others in JASA Letters (see citation below), opens with the observation: Throughout the day most of us engage in a nearly ceaseless internal banter. This stream of inner speech is a core aspect of our mental lives, and is linked to a wide array of psychological functions. Despite this centrality, inner speech has received little scientific attention.

The reason for this, I would guess, is because many people still think of inner speech as being in-and-of consciousness and therefore not requiring explanation – simply being what it appears to be. Because the processes that create conscious experience are transparent to us, it is easy to assume there are no such processes.

This group does investigate the source of inner speech. They start with the hypothesis the it is an effect of corollary discharge.

Corollary discharge is a neural signal generated by the motor system that serves to prevent confusion between self-caused and externally-caused sensations. When an animal performs an action, its motor system uses a forward model (an internal model of the body) to predict the sensory consequences that will result. This prediction is corollary discharge. Corollary discharge is relayed to sensory areas where it serves to segregate out incoming sensations that match the prediction since these are likely to be caused by the animal’s action. Corollary discharge is what allows you to speak without confusing your voice with other voices/sounds in the environment.Auditory corollary discharge (for speech) is therefore an internal prediction of the sound of one’s own voice.

They looked for the tell-tail signs of corollary discharge. One is that corollary discharge can affect perception. When in doubt we perceive what we expect to perceive. This is called perceptual capture.

Perceptual capture is a shift in perception caused by the fact that corollary discharge is an anticipation and as such can pull ambiguous stimuli into alignment with the anticipated percept.

So if inner speech is a corollary discharge phenomenon (the auditory content of inner speech is provided by corollary discharge) than it should show perceptual capture. Their experiments tested this.

In the spectrum from normal speech, through silent speech, to imagined speech, they looked at both ‘mouthed speech’ and entirely silent speech. Both types showed perceptual capture with the mouthed speech having a stronger capture then the silent. Ambiguous speech sounds were heard as similar to the imagined phoneme. /ɑˈbɑ/ and /ɑˈvɑ/ were used with the inner voice of one of them affecting the hearing of the other.

For both mouthed and pure inner speech, participants were more likely to hear an ambiguous sound as matching the content of their imagery. The two-way directionality of the effect (mouthing/imagining /ɑˈbɑ/ pulling perception in one direction but mouthing/imagining /ɑˈvɑ/ pulling in the opposite direction) demonstrates that it is the content of the inner speech that is responsible for the effect, not some extraneous factor. This experiment also shows a distinction between Mouthing and Imagining: For both /ɑˈbɑ/ and /ɑˈvɑ/, the Mouth conditions were significantly different from the corresponding Imagine conditions. This is predicted under the assumption that greater articulator engagement triggers more corollary discharge engagement.

To rule out phoneme priming as the cause of the effect, a second experiment was done.

Experiment 2 demonstrates that inner speech can make the perception of an external sound match the subphonemic aspects of an imagined sound, without there being phonemic identity between imagined sound and percept. By extension, this experiment also tests the claim that subphonemic content exists in inner speech…. The mouthed/imagined sounds were /ɑˈfɑ/ and /ɑˈpɑ/. /ɑˈfɑ/ is similar to /ɑˈvɑ/ at a subphonemic level (both are labiodental fricatives) and /ɑˈpɑ/ is similar to /ɑˈbɑ/ (both are bilabial stops). If imagery of /ɑˈfɑ/ can cause an /ɑˈbɑ/ ∼ /ɑˈvɑ/ ambiguous target to be perceived as /ɑˈvɑ/ (and the converse for imagery of /ɑˈpɑ/) that would indicate the presence of perceptual capture at the subphonemic level. … this experiment succeeded in showing subphonemic influences from pure inner speech in addition to mouthed inner speech.

So this look at inner speech implies that we are not thinking in this inner speech but at some pre-motor non-conscious level and this unconscious thinking goes through a motor program and then becomes a sensory input into the content of consciousness. That seems plausible.

M Scott, H Yeung, B Gick, J Werker (2013). Inner speech captures the perception of external speech Journal of the Acoustic Society of America Letters, 133 (4)