Cogito, ergo sum: An electrophysiological investigation of inner speech properties in a sensory attenuation paradigm

Sensory attenuation describes the decrease in the intensity of sensations we produce ourselves compared to externally-generated sensations, even when elicited by physically identical stimuli. This phenomenon is thought to result from the comparison of sensory predictions to sensory feedback. To predict the sensory consequences of self-initiated actions, duplicates of outgoing motor commands are generated by the motor cortex, known as efference copies. Sensory attenuation is hypothesised to occur if the sensory prediction matches the perceived sensation and has been studied in the context of voluntary actions across many modalities, including auditory. This provides an explanation for why our own speech sounds quieter to us than when we hear someone else speaking the same words. The effect is generated by the suppression of the auditory cortex when producing and hearing self- generated speech compared to when passively listening to a recording of the same speech. Auditory cortex suppression can be measured as an amplitude reduction of the N1 component of the auditory evoked potential. Indeed, a large body of literature shows that smaller N1 amplitudes are elicited by self-generated speech compared to external feedback of the same sound. Analogous sensory attenuation has also been found for the silent production of vowels or words in one's mind–inner speech. This subjective experience of language without any overt articulation plays a fundamental role in various cognitive domains and can be seen as an attenuated version of overt speech. Furthermore, inner speech, as a cognitive phenomenon, engages us in thinking through words and is the foundation for introspection, intricately connected to the philosophical realm of consciousness, ultimately epitomised in René Descartes’ axiom: Cogito, ergo sum ('I think, therefore I am.'). However, the neural processes underlying inner speech and its sensory suppression mechanisms remain largely unexplored. This warrants further investigation of the precision of sensory attenuation elicited by inner speech and whether it carries specific acoustic properties, such as sex, loudness, accent, tempo, rhythm, timbre, etc. Therefore, we first provided a conceptual framework for the subjective and objective assessment of inner speech (Chapter 2). This is followed by an electrophysiological experiment in which participants silently produce two different types of inner phonemes, concurrently with audible phonemes of female and male sex. These overt phonemes either match or do not match the internally produced phoneme of the participant in content (/BA/ versus /BI/) or sex (Chapter 3). We hypothesised that increased acoustic overlap between inner and audible phoneme (identical content or gender) mediated sensory suppression. Our results showed a general N1 amplitude reduction when producing inner speech (active conditions) compared to passive listening of the sample phonemes. However, neither the same content nor the same sex of inner and audible phoneme led to more sensory suppression, compared to non-matching content or sex, respectively. This suggests that inner speech might not precisely encode the content or sex of the perceived voice. Furthermore, we demonstrated in the same sample that other measures of acoustic overlap between inner and audible phonemes, such as peak frequencies in the sound spectrum, known as formants, do not predict the degree of sensory suppression. More specifically, similar formant values of the participant’s individual phoneme (recorded as their overtly uttered syllable) and the audible phoneme did not mediate N1 amplitude reduction (Chapter 4). These two studies suggest that inner speech may contain more impoverished acoustic properties than overt speech that cannot be captured with techniques commonly used in overt speech. Finally, we showed that the early auditory-evoked gamma-band response, another measure of neural activity, did not effectively track sensory attenuation in our data set and experimental paradigm. More specifically, both measured amplitude and inter-trial phase consistency did not differ between the active inner speech and passive listening conditions (Chapter 5). Our findings provide insight into the mechanisms underlying speech-induced sensory attenuation and pave the way for further research of the neural underpinnings of dysfunctional sensory attenuation and auditory-verbal hallucinations in schizophrenia