Neural signature of metrical stress processing in French

The purpose of the current dissertation was to determine whether there is metrical stress in French and what role it plays in French speech comprehension. Metrical stress is well-known to have an invaluable role during speech processing. It holds a central position within the prosodic organization which structures the speech flow in that it both serves as the anchor point of the intonation contour and is the regulating force behind the speech rhythm. Moreover, modulations in the acoustic characteristics of metrical stress inform listeners on the depth of the prosodic hierarchy. Furthermore, metrical stress underlies the abstract representation of the word the smallest informational unit such that it readily contributes to lexical, word-level processing. Finally, metrical stress interplays with attention, concurrently through its acoustic prominence at the surface level and through its underlying temporal predictability. That is, metrical stress both harnesses attention from bottom-up through its attention-grasping acoustic salience, as well as top-down guides attention by providing a metrical framework, which the listener can use to bounce attention from one stressed syllable to the next. The role of metrical stress in French speech processing has, however, been less straightforward. Amidst languages, French, the language under investigation in the current dissertation, may be seen as eccentric, because the language, allegedly, has no stress (Rossi, 1980). That is, French accentuation is commonly attributed a low phonological and post-lexical status because it is not lexically distinctive and frequently overlaps with intonation. The French primary accent (FA) is held to exclusively be located at the right phrasal boundary, such that it is perceptually overshadowed by higher level constituent marking. That is, phrase final accents acoustically merge with the intonation contour such that their phonetic parameters are spread and diluted over nearby syllables (i.e. « acoustic chiasm », Foenagy, 1980), and phrase internal accents may be reduced in order to favor syntactic and semantic coherence within the phrase. Similarly, the French secondary and optional accent ( IA ) is held to serve only post-lexical functions in pragmatically expressing emotional or para-linguistic emphasis, or in the rhythmic structuration of the utterance, yielding to F A when not rhythmically necessary. This means that, in French, the post-lexical re-structuration of the speech stream not only blurs local accents phonetically (i.e. at the surface level), but also functionally, which has given way to the notion that French listeners have no representation of stress in their mental lexicon, and, worse still, are unable to hear stress. As a result, French accentuation has attracted rather little interest in the linguistic field, with most descriptions of French prosody focusing on the tonal and intonational organization of the language, while leaving aside the characteristics of stress other than pitch movements as well as ignoring the possibility for lexical functions for stress during speech comprehension. In the current dissertation, we sought to address this gap in the academic domain. We argued that the initial and final accents carry metrical weight and are represented in bipolar, cognitive stress templates underlying the lexical word (cf. Di Cristo, 1999). Such a metrical perspective allowed us to imagine a phonological role for stress in French speech processing. That is, if I A and F A are metrically strong and mark both boundaries of the word, the accents are more readily integrated in theoretical frameworks on speech processing, according to which metrical stress is crucially involved in the analysis of speech. In our interdisciplinary investigation, we took a functional, yet metrically rooted, approach. We used the method of Event-Related Potentials ( E R P ), which provided us with a highly sensitive and temporally precise measure and allowed us to determine whether there is metrical stress in French, and to what extent metrical stress aids the listener in speech comprehension. More specifically, we combined the method of Event-Related Potentials E R P with three paradigms (i.e. the oddball paradigm, the lexical decision paradigm, and the semantic anomaly paradigm) wherein we carefully manipulated the acoustic manifestation of the stress patterns underlying the words. We were especially interested in four ERP-components : the MMN, PMN, N325 and N4. These components have in common that they all reflect a mismatch between a prediction based on long-term memory representations or established phonological representations and the violation in an experimental setting. The components therefore allow for inferences on the time course (i.e. processing stage) and obstructed linguistic process (e.g. word recognition, access to meaning) that could result from presenting French listeners with stress or metrical information that does not agree with their anticipation. As such, these components and the method of E R P in general are exceptionally well-suited for our investigation wherein we examined three main properties of metrical stress in French : 1. The phonological representation of French metrical stress 2. The function of metrical stress in word recognition 3. The interaction of French stress with lexico-semantic processing The results of our studies indicate (1) that French listeners have metrical expectations when processing speech, (2) that they expect stressed syllables at the level of the word, and (3) that speech processing is hindered when expectations are not met. In other words, the results showed that F A plays a role other than its collaborative function in the marking of the phrase, and that I A is more than a rhythmic counterweight or heavy emphatic stress, but, above all, structural in nature (cf. Astesano et al., 2007 ; Astesano & Bertrand, 2016 ; Astesano, 2017). Specifically, the results in this investigation show that listeners consistently expected words to be marked with both accents at their lexical boundaries. The listener naturally and automatically extracts the metrical information, which s/he then uses during the pre-lexical, lexical and post-lexical processing stages of speech comprehension. The results therefore point to a cognitive representation and phonological anticipation for the French accents at the level of the word and, therein, demonstrate the value of metrical stress processing in French. We conclude that both French accents have a metrical status and appeal for metrical stress to be given a more prominent place in the descriptions of French prosody. Indeed, while until recently, the existence of metrical stress and its role in lexical processing was poorly defined in French, we hope that the present work will lead to a new appreciation for French metrical stress in the academic field, as it clearly lays way to compelling future work which will most certainly advance our understanding on how French listeners use prosodic information to comprehend speech.