The production-perception link in the acquisition of L2 prosodic marking of information structure: evidence from Italian and French.

The relationship between perception and production in second language (L2) acquisition has been widely debated, as accurate perception does not always entail accurate production (see Best, 1995; Best & Tyler, 2007; Munro & Bohn, 2007). This mismatch has been documented for both segmental contrasts and suprasegmental features, including the prosodic marking of information structure (Chen, Gussenhoven & Rietveld, 2004; Zárate-Sández, 2015). Learners may correctly identify prosodic cues in perception yet fail to reproduce them, suggesting that production depends not only on perceptual competence but also on other cognitive factors or universal tendencies in L2 acquisition. For Italian-French learners, research at the segmental level has addressed perception and production separately, so the relationship between the two remains unexplored, even if partial evidence exists in each domain. In the prosodic domain, the evidence is even scarcer: aside from L1 baselines and a few L2 production studies (e.g., Benazzo et al., 2021; Turco et al., 2010, 2011, with learners from Germanic L1 backgrounds), no study, to our knowledge, has examined the perception-production link in the prosodic marking of information structure for this language pair. Our study addresses this gap through a combined production and perception experiment. In the production task, we collected task-elicited speech from 60 participants (15+15 native speakers of Italian and French and 15+15 L2 learners in each learning direction). Italian L1 and L2 data were collected in Turin, and all Italian participants therefore produced or were exposed to the Piedmont-based variety of Italian. French L1 and L2 data were collected in Paris, and all French participants likewise produced or were exposed to the Paris variety of European French. Participants were prompted to produce utterances with different target constituents (subjects, accusatives, datives, adverbials) under three focus conditions: broad focus, narrow identification focus, and narrow correction focus (Krifka, 2008). Although native speakers do not invariably employ systematic prosodic marking of narrow focus, the prosodic analysis reveals consistent patterns: focus constituents are characterised by specific melodic contours, which are clearly distinguished from those of background constituents or topics (see Fig. 1). By contrast, learner productions often fail to distinguish focus from background in their realisations, with narrow-focus constituents frequently remaining “unmarked”, and uttered with the same prosodic profile used for background or non-contrastive topic constituents (Fig. 1). Whether this pattern reflects a misperception of the input should be assessed through the auditory task. Stimuli for the perception experiment were drawn from a subset of native-speaker utterances elicited in the production experiment. Two criteria guided the selection: sentences had to feature canonical word order (thus excluding syntactic focus-marking cues), and their focus/background articulation had to be judged as unambiguous by the first author and two trained student assistants (Inter-Annotator Agreement measures were performed). In the perception experiment, participants (distinct from those in the production study) listened to the utterances and reconstructed, on the basis of auditory cues only, the corresponding Question Under Discussion (QUD; van Kuppevelt, 1995; Roberts, 2012). The test comprised 10 critical items and 20 fillers, plus 2 control items to ensure that participants understood the task. Fig. 2 illustrates the formulation of the test questions. Results from native speakers confirm that comprehension is near ceiling in both languages: Italian L1 listeners (n = 44) reach very high accuracy (≈99%), as well as French L1 listeners (n = 47, ≈93%), with limited variability across participants. This shows that, even if prosodic marking of focus is not always produced consistently, its interpretation remains clear and unambiguous for native listeners, at least when the stimuli display a prototypical prosodic focus. Figure 3 also reports the perception results of Italian learners of L2 French (n = 37), whose accuracy is lower (≈89%) and more variable than that of native French listeners. Perception data from French learners of L2 Italian are still being collected and analysed. Findings will be interpreted in relation to current models of L2 prosodic acquisition (Mennen, 2015) and to broader mechanisms of second language learning, considering learning factors such as markedness, consistency of the input, and its degree of systematicity. In addition, since the production study makes it possible to compare prosodic and syntactic focus-marking strategies, we will consider whether learners show a preference for more explicit syntactic devices (such as clefts) over prosodic cues when signalling and/or processing information structure (De Paolis et al., 2022; Yan & Calhoun, 2022).