Max Planck Institute for Empirical Aesthetics
Lecture by Philippe Martin
Intonation, prosodic structure and brain waves
The accent phrase, sequence of syllables with only one stressed syllable (excluding emphatic stress), is considered as the minimal unit of prosody in both the Autosegmental-Metrical (AM) and Dependency Incremental (DI) approaches. The hierarchical organization of accent phrases, mapped from the syntactic structure (AM view) or incrementally built during the sentence production and perception (DI view) constitute the sentence prosodic structure.
In lexically stressed languages such as Italian or English, every content word (verb, adverb, adjective or adverb) bears a unique stressed syllable, usually marking some morphological boundary (for example between a stem and a suffix). Therefore, accent phrases, containing one stressed syllable, assemble one content word and its depending grammatical words (which are normally not stressed).
In languages such as French, the function of lexical stress progressively disappeared and accent phrases can merge more than one content word, keeping the sequence of syllables with only one stress to a limit defined by the time needed to pronounced them in connected speech. The upper limit reached by very fast speakers is about 1250 ms or slightly more for some subjects. On the other side of the scale, the minimum gap between two consecutive stressed syllables is about 250 ms. A shorter gap will make the first syllable perceived as unstressed, so that one stressed syllable must be preceded by a sufficient gap to form an accent phrase. Therefore, accent phrase duration spans between 250 ms and 1250 ms.
The analysis of speech production tells us as well that there is a tendency for eurhythmy, i.e. for speakers and readers to balance the duration of successive accent phrases, either by adjusting the average duration of syllables (spontaneous speech), or by selecting in the speech chain the words to be included in successive accent phrases (read speech).
The fact that the duration range between consecutive stressed syllables, from 250 ms to about 1250 ms, is comparable to the range of delta brain oscillations, suggests that delta waves may be synchronized by stressed syllables, and indirectly trigger the accent phrase perception process.
Four observations support this view: 1) if two consecutive acoustically stressed syllables are separated by less than 250 ms, the first one ceased to be perceived as stressed, as the 250 ms delta cycle could not be accomplished in the stress perception process; 2) if two consecutive syllables are separated by more than 1250 ms in connected speech (generated by speech synthesis for instance), extra syllables, although not acoustically stressed, are perceived as stressed by listeners so that the gap between stressed syllables is maintained below 1250 ms (due to a free running delta spike non synchronized by a stressed syllable); 3) accent phrases in silent reading cannot exceed the size of those found in oral reading, but can be read faster, down to 250 ms per accent phrase (the minimum duration of delta oscillations); 4) eurhythmicity and the speech rate chosen by the speaker or the listener at the beginning of the process determines which potential accent phrases will be defined by these extra stressed syllables.
These observations lead to consider a mechanism for delta waves that would be both bottom-up (for synchronization by stressed syllables) and top-down (for the identification of accent phrases in the lexicon).