Deciphering speech: a two-piece puzzle

How is language processed by the brain? What mechanisms enable it to extract meaning from the sounds emitted by speech? These are the questions addressed by a research team from Inserm and Aix Marseille Université, in collaboration with Cambridge University. Their work shows that, to understand speech, the human auditory cortex simultaneously follows two essential rhythms: that of syllables (slow) and that of phonemes (fast). They also show that these rhythms characterize speech across languages, and that they are processed simultaneously by the brain's auditory regions. These results pave the way for a better understanding of language processing and related disorders.

Reading time: 5 minutes

Key facts:

Scientists have studied the brain's language-processing mechanisms and the way it analyzes syllables and phonemes to reconstruct comprehensible information from the continuous sound signal associated with speech. As soon as the auditory cortex processes the speech signal, the slow rhythms of syllables and the fast rhythms of phonemes are tracked in parallel by the brain.
This ability to simultaneously detect, isolate and decode the 2 types of "puzzle pieces" enables it to segment the continuous speech signal.
By analyzing the way speech is constructed in 17 different languages, the scientists found that, whatever the language, the speech signal systematically presented the same slow and fast rhythms, assimilated to those of syllables and phonemes.

During a discussion, what are the mechanisms that enable the brain to extract and reconstitute information from the continuous flow of speech? Cognitive science is still trying to decipher them today.

Understanding language as a puzzle

To better understand them, they invite us to imagine speech as a puzzle. Syllables, such as "ma" in "mom", are the big pieces, while phonemes, the individual sounds that make up syllables, such as "m" and "a", are the little pieces. The large parts (syllables) are produced at a slow rate, and are mainly reflected by variations in the sound amplitude (volume) of the vocal signal over time. Small parts (phonemes), on the other hand, occur at a much faster rate and are reflected in abrupt changes in the spectral ^content1 of the speech signal. To understand speech and "translate" it into successive, recognizable and meaningful isolated words, the brain must analyze both types of parts at the same time.

A research team led by Benjamin Morillon, Inserm Research Director at the Institut de neurosciences des systèmes (Inserm/Aix Marseille Université), has been investigating the brain's language-processing mechanisms, and how the brain is able to analyze syllables and phonemes to reconstruct comprehensible information from the continuous sound signal associated with speech. The scientists analyzed the brain activity of 11 people (already fitted, for clinical reasons, with electrodes that record neuronal activity in different regions of the brain), while listening to 315 sentences in French, played at different speeds and containing a fixed number of words, but varying numbers of syllables and phonemes.

Their results show that as soon as the vocal signal is processed by the auditory cortex (a region of the brain responsible for processing sounds), the slow rhythms of syllables and the fast rhythms of phonemes are tracked in parallel by the brain. This ability to simultaneously detect, isolate and decode the 2 types of "puzzle pieces" enables it to segment the continuous speech signal.

A universality in the way speech is constructed

In addition, by analyzing the way speech is constructed in 17 different languages, the scientists found that, whatever the language, the vocal signal systematically presented the same slow and fast rhythms, assimilable to those of syllables and phonemes."This universality suggests the existence of a biological mechanism for the production and perception of speech common to all humans", says Benjamin Morillon, and the research team hopes that this breakthrough will lead to a better understanding of the complex and fascinating process of speech processing and its associated disorders.

"These discoveries open up new perspectives in the understanding of language disorders such as dyslexia," adds Benjamin Morillon. "A deficit in the brain's ability to follow these two rhythms could, for example, explain certain difficulties in oral comprehension", he concludes.

1] The spectral content of the vocal signal could be defined as the "sound texture" of a voice; it is what makes each voice unique. It is made up of three characteristics: the base frequency, linked to the vibration of the vocal cords, which determines the pitch of the sound (low or high); the harmonics, which are multiple frequencies of the base frequency and enrich the sound; and lastly, the formants, which correspond to peaks in the harmonics and are linked to the resonances of vocal tract structures such as the mouth, throat, glottis, vocal cords or nostrils. As the anatomy of these structures is unique to each individual, they contribute to the specificity of a voice and the sounds it emits.