Mental imagery of speech: linking motor and perceptual systems through internal simulation and estimation

Abstract

The neural basis of mental imagery has been investigated by localizing the underlying neural networks, mostly in motor and perceptual systems, separately. However, how modality-specific representations are top-down induced and how the action and perception systems interact in the context of mental imagery is not well understood. Imagined speech production ("articulation imagery"), which induces the kinesthetic feeling of articulator movement and its auditory consequences, provides a new angle because of the concurrent involvement of motor and perceptual systems. On the basis of previous findings in mental imagery of speech, we argue for the following regarding the induction mechanisms of mental imagery and the interaction between motor and perceptual systems: (1) Two distinct top-down mechanisms, memory retrieval and motor simulation, exist to induce estimation in perceptual systems. (2) Motor simulation is sufficient to internally induce the representation of perceptual changes that would be caused by actual movement (perceptual associations); however, this simulation process only has modulatory effects on the perception of external stimuli, which critically depends on context and task demands. Considering the proposed simulation-estimation processes as common mechanisms for interaction between motor and perceptual systems, we outline how mental imagery (of speech) relates to perception and production, and how these hypothesized mechanisms might underpin certain neural disorders.

Keywords: auditory hallucination; corollary discharge; efference copy; internal forward model; mirror neurons; phantom limb; sensory-motor integration; stuttering.

Figure 1

Model of speech processing and its implication for mental imagery of speech. The internal simulation and estimation model proposed as a second route to generate mental images. The motor systems that mediate action preparation carry out the same functions in mental imagery of speech, but only perform motor simulation, in the sense that the planned motor commands are truncated along the path to primary motor cortex and are not executed (the red cross over external outputs). A copy of such planned motor commands (motor efference copy) is processed internally and is used to estimate the associated somatosensory consequences. A copy of the somatosensory estimation is further sent to modality-specific areas, and the associated perceptual consequences that would be produced by the overt action are estimated. The quasi-perceptual experience during mental imagery (the feeling of movement of the articulators and the feeling of auditory perception in the case of articulation imagery) is the result of residual activity from these internal estimation processes, because of the absence of cancellation from the external feedback (the red crosses over external somatosensory and perceptual feedback).

Figure 2

Results from Tian and Poeppel (2010). The sequential estimation during articulation imagery revealed by MEG recordings. All plots are MEG topographies (response patterns) when participants actually speak (lower row), imagine hearing (middle row), and imagine speaking (top row). The activity patterns in the first column are temporally aligned with the onset of articulation movement. At a similar latency, bilateral frontal, bilateral temporal, and left lateralized parietal activity patterns are observed in articulation, hearing imagery, and articulation imagery conditions. In articulation imagery, about 150–170 ms later after the parietal activity, bilateral temporal activity is also observed. All the bilateral temporal activity patterns in hearing imagery and articulation imagery resemble the topography of the auditory response during actual hearing (highlighted in a blue box, response pattern when participants listen to the same auditory stimuli as in other conditions).

Figure 3

Sufficiency and necessity between motor simulation and perceptual estimation. The characteristics of the proposed motor simulation and perceptual estimation processes, and the nature of motor involvement during perceptual tasks. The internal motor simulation can take a similar path as motor preparation to derive a corresponding motor representation that in turn derives associated perceptual representations in a one-to-one fashion. Such one-to-one mapping is the same as the one in the external connections between the similar motor action and perceptual consequences. In the other direction, when the perceptual representation is needed, different paths can be taken. It can rely on memory retrieval to directly recreate the perceptual representation. It can also take another less demanding path that relies on the motor simulation to derive the associated perceptual representation.