MisMatch Negativity-study showing pre-lexical sensitivity to both primary Final Accent and secondary Initial Accent in French

In French, accentuation is not lexically distinctive and tightly intertwined with intonation. This has led to the language being described of as ‘a language without accent’ and to French listeners being alleged ‘deaf to stress’. However, if one considers Di Cristo’s model in which the metrical structure of speech plays a central role, it becomes possible to envision stress templates underlying the cognitive representation of words. This event-related potential (erp) study examined whether French listeners are sensitive to the French primary final accent (fa) and secondary initial accent (ia), and whether the accents are part of the French phonologically expected stress pattern. Two oddball studies were carried out. In the first study, in one condition, deviants were presented without (−fa) and standards with final accent (+fa), while in another condition, these positions were switched. We obtained asymmetric mmn waveforms, such that deviants −fa elicited a larger mmn than deviants +fa (which did not elicit an mmn), pointing toward a preference for stress patterns with fa. Additionally, the difference waveforms between identical stimuli in different positions within the oddball paradigms indicated −fa stimuli to be disfavored whether they were the deviants or the standards. In the second study, standards were always presented with both the initial and final accent, while deviants were presented either without final accent (−fa) or without initial accent (−ia). Here, we obtained mmns both to deviants −fa and to deviants −ia, although −fa deviants elicited a more ample mmn. Nevertheless, the results show that French listeners are not deaf to the initial and final accents, pointing instead to an abstract phonological representation for both accents. In sum, the results argue against the notion of stress deafness for French and instead suggest accentuation to play a more important role in French speech comprehension than is currently acknowledged.

listeners can still readily decode speech, despite their supposed 'phonological deafness', it-according to these 48 scholars-stood to reason that accentuation is unlikely to play an important function in French comprehension 49 processes. Consequently, French accentuation has attracted rather little interest in the linguistic field. As mentioned above, in a study addressing the phonological status of the French initial accent (I A), Aguilera et al. (2014) showed that I A is not only perceived, but anticipated by listeners as belonging to the abstract 117 representation of the word (see also Astésano et al., in pr ep). The authors manipulated the phonetic realization 118 of I A on trisyllabic words in an oddball paradigm. Participants either listened to a version of the oddball-task 119 wherein the stimulus +I A was in the standard position and the word −I A in the deviant position, or a version 120 wherein ±I A positions were reversed. All listeners completed two tasks, one passive task during which they 121 listened to the stimuli while attending a silent movie, and one active task during which the listeners were asked 122 to respond as quickly and accurately as possible when they detected the deviant stimulus. Results indicated that 123 the listeners clearly distinguished between the trisyllabic words carrying I A and those that did not. This again 124 indicates that French listeners are in fact not deaf to stress, but readily perceive the accentual manipulation. the initial accent. Indeed, it is plausible that, if I A is part of a preferred stress template, only rarely presenting 130 the template might make it the deviant within the experiment, but it does not make the template improbable.

131
In other words, in the condition in which the oddball was presented with I A, while atypical in the context of 132 the test, the oddball was still the expected stress template. Therefore, no M M N emerged.

133
In order to further ascertain that the observed M M Ns were independent from differences in acoustic process-134 ing, Aguilera and colleagues carried out an additional analysis wherein they compared the M M Ns resulting from 135 the difference wave between −I A-deviants and −I A-standards to the difference wave between +I A-deviants 136 and +I A-standards (i.e. between participants comparison). Again, results indicated that the difference be-137 tween stimuli without initial accent was significantly larger than the difference between stimuli with initial 138 accent, allowing for the purely acoustic interpretation of the results to be ruled out. Finally, the behavioral 139 results from the active task confirmed the interpretation of the E R P results. That is, the deviant stimuli −I A 140 were slower to detect than the deviant stimuli +I A, and generated more detection errors. Overall, Aguilera and 141 colleagues thus not only show that stimuli without I A are noticed by listeners, but also that I A is anticipated 142 and attached to the metrical template underlying the representation of words.

143
In the current study, we set out to build on these findings and investigated the phonological representation 144 of the French final accent in an oddball paradigm (F A). Following Di Cristo, we argue words to be encoded 145 with bipolar stress templates underlying their representation, marking not only the left (I A) but also the right 146 (F A) lexical boundary. Here we sought to determine whether F A is phonologically represented, similar as I A, 147 and manipulated the presence of F A on trisyllabic words in an auditory oddball paradigm. In a first study  conditions, durations were equalized between ±F A stimuli by shortening the first two syllables of +F A 176 stimuli. To additionally avoid confounds from shortening the two initial syllables, these first two syllables 177 were shortened below the perceptual threshold following Rossi (1972) and Klatt (1976). To verify that the 178 durational modulations on the first two syllables were not perceptible, two independent French phonetic experts 179 completed an XO-task wherein they listened to word pairs that were either both manipulated on the first two 180 syllables (25%), both without the durational manipulation (25%), or one with and the other without (50%).

181
The listeners judged whether the two words were identical or different. Only stimuli with accuracy rates that     presented +I A. This discrepancy potentially indicates differential processing between I A and F A, which is 254 elaborated upon in the main discussion of the two experiments.  In the current studies, we sought to investigate the phonological representation of French accentuation. We  Below, we will discuss our findings in turn: In section 2.1, we present the results obtained in the first oddball 328 study that show F A to not only be readily perceived, but also to be expected by the listener and phonologically 329 natural. In section 2.2, we discuss the differential processing of −F A and −I A deviants. We will interpret the 330 results from an acoustic, exogenous point of view, as well as inspect the possibility for this difference to reflect 331 more substantial, endogenous differences in the functions of the respective accents during word processing. anticipate words to be marked with final accent, most likely due to its established phonological representation.

348
The comparison between standards and deviants presented with final accent points to the same conclusion.

383
In the exogenous interpretation, the dissimilar M M N amplitudes between I A stimuli and F A stimuli reflect 384 differences in acoustic processing. Indeed, the acoustic manipulations had not been the same between our ±I A 385 and ±F A stimuli, the former involving exclusively a manipulation of the f 0 rise, and the latter involving mainly between the accents and the stages in speech perception, which we will turn to next.  to not only mismatch with the anticipated phonological stress template, but change the predicted lexical item 453 because it was missing its right boundary mark (e.g. "paradis" → "paradigmatique"). That is, listeners may