ABSTRACT. In Spanish, vowels across word boundaries (V#V) are characterized as being resolved by native speakers in a smooth manner with full linking. In English, however, glottal stop tends to intervene, which textbooks warn to not transfer and instead to produce “smooth linking”, the putative native norm. Additional types of linking have been observed (diphthong formation, merged/hybrid vowels, replaced/eliminated vowels and degrees of shortening when V1=V2), but have received little empirical attention. For L2 Spanish, while some of these types are treated in pedagogical texts, strategies such as merged and replaced vowels are rarely mentioned. The current study investigates V#V production of native and non-native speakers, examining the distribution of continuous (smoothly linked) versus discontinuous (separated; glottal stop or other strategy) speech. (See visuals below.) Two learner groups were studied and compared to documented native norms from Veracruz, Mexico. Learner data come from a simple picture narrative (40-50 token sequences) and a short reading passage (59 contexts). All speakers were recorded in a quiet environment and acoustic analysis was conducted with Praat on formant structure and duration as cues, as well as employing auditory impressions. Results from an initial phase of this study indicate that NSs do not categorically produce continuous sequences, but rather produce a variable proportion of discontinuous tokens (7%) separated by glottal phonation and in a few cases, consonantal insertion (y/w). Across both tasks, NNSs (graduate students) and upper intermediate (undergraduate) learners produced discontinutous/separated tokens 28% and 68% of the time, respectively. Thus, native-like behavior significantly increases as target language proficiency increases. NSs had the most linked and merged forms; followed by advanced learners, who linked most with diphthongs; then lower learners, who linked most with hiatus, though with greater durations, but who also separated the sequences more. Both learner groups linked more in the reading task. The current phase of the study analyzes the impact of vowel quality (high/low, same/different, etc.), stress, type of word, word frequency, and constituent position on the distribution of linking vs. separation and of linking resolution strategies. It further will analyze preliminary results of a pilot study that explores native and learner perceptions of V#V productions. We will elaborate all results in the presentation, discussing them considering the theoretical, descriptive and pedagogical implications we believe our study has for native and L2 Spanish.

ABSTRACT. Although it is established and well known that a dialect or an L2 accent in pronunciation can cause a negative evaluation of speakers in the dimensions of social status and solidarity (cf. e.g. Lambert et al. 1960; Cunningham-Andersson 1997; Derwing 2003; Settinieri 2011; cf. Dahmen 2022 for an overview), the social acceptability of different pronunciation variants, in contrast to their intelligibility and accentedness, is still relatively little researched. In the context of the quantitative online questionnaire study described here, a total of 18 speakers (6 with L1 German in a standard variety, 4 with L1 German with a Bavarian accent, and 8 with German as L2 and several different L1s) each recorded three short sentences containing segmental pronunciation difficulties typical for L2 speakers that were submitted to 102 raters for evaluation. The raters judged the randomized utterances with respect to their social acceptability on a six-point Likert scale from 1 = "I don't like this voice at all." to 6 = "I like this voice a lot.". In addition, the raters were asked about aspects of the voices that they perceived as particularly positive or negative. By means of inferential statistical analyses, the extent to which the raters' evaluations differ with respect to the groups of standard L1 German, German with a Bavarian dialect, and with L2 accent is examined. Subsequently, the results are tested for a possible influence of the factors gender, speaking rate as well as experience as a foreign language teacher. The results show that the listeners rated the voices of the standard speakers of German significantly higher than the voices of the speakers with dialect or accent. With regard to the speaking rate, the evaluation of the voices of speakers with standard German as L1 for the group of L2 raters depends significantly on the speaking rate: The slower their speech, the better the L1 speakers are rated by the L2 raters. Furthermore, men rate significantly more severely than women. Whether someone has experience as a foreign language teacher, on the other hand, turns out to be unrelated. Methodological challenges and implications for sociophonetic theory building are put up for discussion.

ABSTRACT. During a very long time, intonation has chronically (see De Bot 1986: 81) been labelled as the «neglected oral skill» (Mansfield 2001: 81). This marginality is the result of a combination of causes: the enduring dominance of written language in linguistic studies (see Voghera 1993: 10); the difficulties in clearly defining terms and methods of intonational phonology (see Giordano 2008; De Dominicis 2003, 2010) and the unavailability – until few decades ago – of instruments allowing its objective analysis and representations. As a domino effect, these limitations in studying the multi-functionality of intonation (see Mansfield 2001: 33 et seq.) also extended to those fields of inquiry that borrowed from theoretic linguistics their concepts and instruments, as in the case of second language acquisition and teaching, in which suprasegmentals have been addressed only recently (see Chun 2002). Existing studies in both fields present similar problems. Among others, the fact that, in current linguistic studies, «the signifier has been studied mostly without the signified» (Albano Leoni 2009: 23, our translation) particularly stands out. This approach results in an almost complete lack of interest in the pragmatic functions of intonation in interaction, and has consequently determined a general recourse to data prepared ad hoc to verify hypothesis. This process makes it impossible to observe concrete and plural intonation uses in discourse, i.e. where suprasegmentals more strongly contribute to express the illocutive values of the message (see Nencioni 1975) and guarantee its pragmatic interpretability (see Moneglia-Raso 2014: 469; Cresti 2000: 63). Spontaneous acquisition is therefore rarely considered a useful target in intonational phonology, but could be instead really useful in shedding light on the principles underlying natural patterns of L2 development and processing, as already attested for morpho-syntax (see Giacalone Ramat 1993: 343 et seq.). This research presents the results of a corpus-driven study – part of a larger ongoing PhD research – focusing on the emergence of significant uses of intonation in context. It involves adult migrants arrived in Italy for different purposes, with different linguistic backgrounds, having acquired Italian mainly in natural contexts. Synchronic speech data have been recorded in spontaneous conversations with the author and then labelled and analyzed in order to determine the emergence of intonation as an integrant part of communicative competence (see Hymes 1972; Desideri 1984), considering both the context of the conversation and the illocutive values of detected acts (see Cresti 2000). The paper aims at proposing preliminary answers to the following questions: a) How can we determine what is correct in L2 intonation? Is the dichotomy correct/incorrect adequate for evaluating a performance, or appropriateness to the context – relating to pragmatics – might be a better solution, both in L2 assessment and teaching (see for example Council of Europe 2007)? b) Can we hypothesize that intonation patterns have an autonomous development in spontaneous acquisition, or in the learner interlanguage (Selinker 1972) priority is given to segmental features, in order to “economize” resources in the cognitive effort to de-construct the input autonomously? c) What do spontaneous speech data on intonation tell us about meaning construction in interaction, and how can these results be converted in materials suitable for class uses?

ABSTRACT. Description of the phonetic acquisition based on the ELEI (Español Lengua Extranjera en Italia) corpus, an interlinguistic corpus of native Italian speakers learning Spanish in a guided teaching context at the University of Salerno. The main purpose is to investigate the steps involved in the acquisition of the phonetic phenomena of Spanish as a foreign language (ELE.) by Italian speakers.

ABSTRACT. The acquisition of French liaison is a widely studied topic, from both L1 and L2 perspectives and in different theoretical frameworks. However, a description within the framework of Opti-mality Theory (OT) is still missing, although an OT analysis could model the acquisition process in more detail and trace it back to universal phonological principles. I will analyse the acquisition of liaison by German learners of French as a foreign language based on data from a cross-sectional study with pupils in Germany. French liaison is an external sandhi process in which a liaison consonant appears at the boundary between two words (W1 and W2). The liaison consonant is usually only realised before a vowel- or glide-initial word as in les oranges ‘the oranges [lezɔʁɑ͂ʒ] and mute in every other context, especially in isolation or in preconsonantal position as in les pommes ‘the apples’ [lepɔm]. In the case of liaison, normally a resyllabification takes place and the liaison consonant becomes the onset of the following syllable ([le.zɔ.ʁɑ͂ʒ]). Depending on the context and the nature of W1 and W2, liaison can be categorical (or obligatory), variable (or optional), or impossible (or forbidden) (Delattre 1947; Durand et al. 2011). In this study, I will focus on categorical liaison whose acquisition is crucial for L2 learners. Several studies on learners with different L1s showed that less advanced learners tend to omit the obligatory liaison or to realise the liaison consonant without resyllabification, while more advanced learners and especially university students of French produce high rates of correct liaison in obligatory contexts (e.g. Mastromonaco 1999; Thomas 2004; Barreca 2015; Pustka et al. 2022). Especially studies with learners from different Germanic languages, in which the congruence between lexical and syllabic boundaries is much stronger than in French, indicate that this phonological difference leads to great difficulties for learners in the perception and production of liaison (cf. Wauquier 2009). The corpus on which the analysis will be based currently includes recordings from 42 pupils performing a specifically designed task of simplified picture description. The pupils belong to three different groups depending on the number of years of French lessons (respectively one, three and five) in the German secondary school. Further recordings are planned for summer 2023. Confirming the findings in other studies on learners with German as L1 (especially Pustka et al. 2022), the data show that correct liaison productions increase over time. Nevertheless, liaison presents a difficulty and is not completely acquired in the first five years of French instruction in secondary school. Furthermore, there are important inter-individual differences, and the learner variety is unstable over the whole period. Different phases in the acquisition of liaison can be identified. My analysis is based on two theoretical claims. First, I argue for the utility of OT to model French liaison as it can explain the process by recurring on universal principles for syllable structure and syllabification. I will propose a slightly simplified adaptation of the analysis by Smolensky & Goldrick (2016) and Smolensky et al. (2020). Second, I will demonstrate that OT is especially suited for modelling the acquisition process of liaison. Compared to existing usage-based accounts (e.g. Bybee 2005, Chevrot et al. 2009; Dugua & Chevrot 2015) and auto-segmental accounts (e.g. Wauquier 2009) of L1 and L2 acquisition of liaison, OT can easily account for inter- and intra-learner variation. Particularly, the L2 acquisition process from lear-ners with one specific L1 (here: German) can be described more adequately. I will state that the differences between the German and the French constraint ranking are a possible explanation for the different types of errors observed and the typical acquisition process of liaison.

ABSTRACT. This paper aims to report on international students’ attitudes to a great variety of English native and non-native accents in a highly diversified cultural community at United World College East Africa, Tanzania. In the school year 2022-2023, on two campuses in Arusha and Moshi, there are altogether 661 students representing 109 nationalities. The geographical composition of the class of 2023 is 38% African, 27% European, 20% Asian, 9% North American and 7% South American. One of the International Baccalaureate World School values is “an education in a global context that promotes an appreciation and understanding of multiple perspectives and interdependence of individuals, societies and environments.” The study intends to examine the influence of diversified accents on UWCEA students’ production and perception in English.

The questionnaire includes forty-two items: mostly statements (on 5-point Likert scale) and some open and close-ended questions. It is divided into three parts: 1) general opinion about international and intercultural communication in English (nativeness versus intelligibility principles, comprehensibility of other accents); 2) the insights into informants’ own English (preferred model, attention to their own accent, the effect of other accents on their speech, the techniques applied for being understood and negotiation of meaning); 3) accent perception (familiarity with and recognition of accents, ease of understanding native versus non-native accents, the relationship between accent versus intelligibility and comprehensibility).

The study is still in progress, therefore, final results cannot be presented at this stage. However, it is hypothesized that the notion of intelligibility rather than nativeness will come to the fore. It is expected that the exposure to diversity of accents will result in a positive impact on the informants’ understanding of native and non-native Englishes as well as on their own enriched English accent.

ABSTRACT. In the past three decades, the number of first-generation non-native speakers of Icelandic has increased considerably, thus bringing about new demographic and sociolinguistic circumstances. For a long time, Iceland has been a monoethnic and monolingual society, in which the Icelandic language has served as a main element in the construction and maintenance of national identity (Árnason 1999; Skaptadóttir & Innes 2017). Considering that non-native accents in today's speech community are a novelty, it is interesting to see how representatives of the speech community approach different accents and their geographic origin against the background of both a stable evaluation system concerning the native variety and newly-emerged challenges from swift demographic changes in the past years. Drawing on methods rooted in perceptual dialectology and established by folk linguistics (Niedzielski and Preston 2003: 41ff.), five focus groups with thirty-two native-speaker participants were conducted, employing a semi-structured interview guide. Participants in focus groups were both presented with a voice-placing task involving six verbal guises and asked to freely elaborate on their choices, thus providing emic themes, i.e. meaningful categories contributed by the folk representatives themselves (Lett 1990:130), that are connected to their voice-placing strategies. Verbal guises (e.g. Lindemann 2003) were produced by one native speaker of Icelandic and five non-native speakers with L1 American English, Danish, Filipino/Tagalog, Lithuanian, and Polish. Consistent with the folk linguistic approach taken in this study, the analysis of data employed concepts from grounded theory (Glaser & Strauss; Charmaz 2006). Results from the voice-placing task’s map data indicate that participants place the native speaker of Icelandic with great certainty, whereas, for the non-native speakers, it seems that participants tend to place some guises with more security than others. Four main categories influencing voice-placing decisions emerged from emic themes as stemming from participants’ free elaboration: familiarity with an accent, cultural stereotypes, perceived phonological features in L2 speech, and significance of visual cues, with each of the four main categories comprising further sub-themes, including perceptions and evaluations of speaker effort, L2 competence, and L2 mimicking ability.

ABSTRACT. This study examines the impact of audio vs. visual stimuli on attitudes towards different varieties of English. While there has been a great deal of research on attitudes towards different varieties of English (see for example Hansen Edwards, 2018; Hansen Edwards, Zampini, & Cunningham, 2019; Tan & Castelli, 2013), previous research has primarily employed either survey data or Verbal or Matched Guise Techniques (V/MGT) to explore how listeners react to different accents of English. These approaches, however, offer a limited view into language attitudes as the participants either do not have exposure to any audio or visual stimuli (surveys) or have exposure to only audio stimuli. This does not reflect real life communication, however, which increasingly incorporates visual stimuli given the preponderance of Zoom and Facetime as well as other communications technologies. To date, it is not clear how visual stimuli impacts language attitudes. The current study attempts to address this gap through a study conducted with audio and video speech extracts across three modes: 1. Audio only, 2. Audio and Visual mode, and 3. Audio and Visual with face mask. The study attempted to answer several questions: 1. Do ratings of speakers’ English accentedness, comprehensibility and the acceptability vary across the three modes – audio only, audio visual, and audio visual with face mask?; 2. Do listener’s ability to identify the speaker’s background and variety of English vary across modes – audio, audio visual, and audio visual with face mask? Six speakers of three varieties of English (two each of Indian, Malaysian, and Indonesian English) were recorded in a naturalistic conversational style answering general questions including: What are your favourite hobbies? What childhood games did you play? What is your favourite movie or TV series? Excerpts of approximately 20 seconds in duration judged to be devoid of culture specific information were extracted from each of the speakers’ conversations in the three modes and embedded into an online data task. Listeners were asked to rate each of the 18 speech extracts (6 speakers x 3 modes) on a 1-9 Likert Scale for level of accentedness (from 1 for ‘very strong accent’ to 9 a ‘mild or no accent’), comprehensibility (from 1 ‘very difficult to understand’ to 9 ‘very easy to understand’) and acceptability (from 1 ‘their English is not acceptable’ to 9 ‘their English is very acceptable’). Through two open answer questions, the listeners were also asked to indicate where they thought the speaker was from and which variety of English they thought the speaker spoke. To control for listener background variables, only listeners from Hong Kong were recruited for the study. A total of 100 listeners participated in language attitude tasks using a counter-balanced design to counter any task effects. Additionally, the speech samples were randomized within each task. Results indicate that for all three modes, ratings of accentedness were the lowest of the three ratings for all six speakers across all three tasks, followed by ratings of comprehensibility. Across all speakers and tasks, ratings of acceptability were the highest. This indicates that while listeners felt that while the speakers’ speech was accented, they felt that it was comprehensible (understandable) and acceptable. Listeners rated the speech of the Malaysian and Indonesian speakers as more accented, less comprehensible, and less acceptable in the visual stimuli tasks than in the audio only task. In contrast, the speakers of Indian English were generally rated less accented, more comprehensible, and more acceptable in the visual modes than in the audio mode. These variety-specific results may, in part, be explainable through the open-ended background and variety data: The speakers of Indian English were identified more accurately than both the Malaysian and Indonesian speakers, indicating that listeners’ ability to accurately identify the speaker by background and variety may impact their ratings. These results will be discussed in detail along with future directions for language attitudes research.