This study investigated how L1 and L2 sounds of different degrees of similarity are perceived by L2 speakers. Twenty-six participants were tested with six Mandarin-English C-/i/ syllable pairs varying in degrees of similarity in the initial consonant. Two competing models were examined, the speech learning model (Flege, 1995) and the perceptual assimilation model (Best, 1995). The former proposes that the greater the perceived dissimilarity between an L2 sound and its closest L1 counterpart, the easier it is for L2 speakers to discern the difference, while the latter proposes that L2-to-L1 assimilation corresponds to their articulatory distance. As for two L2 sounds corresponding to the same L1 target, perceptual performance may still differ depending on how similar the L2 sounds are to the target. Results showed that perceptual performance was independent of L2 proficiency. The SLM is supported in that the accuracy for discerning individual consonants of ‘‘dissimilar’’ pairs was higher than that for ‘‘similar’’ pairs, which in turn was higher than that for ‘‘identical’’ pairs. However, the accuracy for discerning ‘‘identical’’ pairs was still significantly higher than the chance level, indicating that L2 speakers are still able to tell the difference, which can be better explained by the PAM.

One of the aims of cochlear implantation is to improve communicative abilities in perception or understanding and in production or intelligibility. Estimates of those abilities should take account of confounding variables within the child and their environment. A postal questionnaire survey was completed by the parents of 1063 profoundly hearing-impaired children who did not have cochlear implants, and the parents of 468 implanted children. Teachers also completed questionnaires for 798 profoundly hearing-impaired non-implanted children and 383 implanted children. Data included a rating for perception and production abilities in three domains: British Sign Language (BSL), sign-supported English (SSE) or signed English (SE), and spoken English. Analyses of the reported ratings controlled for duration of use of a cochlear implant, age of onset of hearing impairment, presence of other disabilities, parental skill level as a marker of socio-economic status (Standard Occupational Classification, 2000), hearing level and age. Analyses of each variable, comparing children similar in all other respects, identified the significant positive differences in ratings shown in Table 1. In summary, higher levels of ability in production and perception in all three domains were reported for children who were older and who had no other disabilities. Higher levels of ability in speech production and perception were reported for children who had an onset of impairment after 3 years of age, who had better average hearing levels and whose parents were classified in a higher skill level. In addition, parents reported higher levels of ability in BSL production for children of families with higher skill levels and in SSE/SE and speech production for children who had used an implant for longer than four years. Higher levels of ability in BSL perception were reported by parents for children who had used implants longer than four years and in speech perception for children who had used implants for longer than two years. These results demonstrate that possession of an implant is

One of the major tenets of the Speech Learning Model (SLM) is that “if two L2 sounds differ in perceived dissimilarity from the closest sound in the L1 inventory, the more dissimilar of the L2 sounds will manifest the greater amount of learning” (Aoyama et al. 2004:248). Given that certain studies have provided “evidence of greater learning for [®] than [l] by N[ative]J[apanese] learners of English” (2004:246), the SLM hypothesis can only be upheld if English [l] is more similar to Japanese [R] than English [®] is. However, this is clearly counterintuitive since, by most accounts, [R] represents a central flap, [l] a lateral approximant, and [®] a central approximant. In this study, it will be argued that English laterals cannot be more similar to Japanese /r/ than English rhotics are, as the SLM would have it, unless the Japanese sound contains a lateral component such as that which is found in the flap [‰]. As it happens, a number of phoneticians and phonologists have argued that this is indeed the case with Japanese /r/, as will be shown, and this is something that the proponents of the SLM would need to acknowledge if their theoretical stance is to be maintained.

Unlike the traditional words-and-rules approach to language processing (Pinker, 1999), usage-based models of language have emphasised the role of multi-word sequences (Christiansen & Chater, 2016b; Ellis, 2002). Various psycholinguistic experiments have demonstrated that multi-word sequences (MWS) are processed quantitatively faster than novel phrases by both L1 and L2 speakers (e.g. Arnon & Snider, 2010; Wolter & Yamashita, 2018). Collocations, a specific type of MWS, hold a prominent position in psycholinguistics, corpus linguistics and language pedagogy research. (Gablasova, Brezina, McEnery, 2017a). In this dissertation, I explored the processing of adjective-noun collocations in Turkish and English by L1 speakers of these languages through a corpus-based study and psycholinguistic experiments. Turkish is an agglutinating language with a rich morphology, it is therefore valid to ask if agglutinating structure of Turkish affects collocational processing in L1 Turkish and whether the same factors affect the processing of collocations in English and Turkish. In addition, this study looked at L1 and L2 processing of collocations in English. This thesis firstly has investigated the frequency counts and associations statistics of English and Turkish adjective-noun collocations through a corpus-based analysis of general reference corpora of English and Turkish. The corpus study showed that unlemmatised collocations, which does not take into account the inflected forms of the collocations, have similar mean frequency and association counts in the both languages. This suggests that the base forms – uninflected forms of the collocations in English and Turkish do not appear to have notably different frequency and association counts from each other. To test the effect of agglutinating structure of Turkish on the collocability of adjectives and nouns, the lemmatised forms of the collocations in the both languages were examined. In other words, collocations in the two languages were lemmatised. The lemmatisation brings the benefit of including the frequency counts of both the base and inflected forms of the collocations. The findings indicated that the vast majority (%75) of the lemmatised Turkish adjective-noun combinations occur at a higher-frequency than their English equivalents. In addition, agglutinating structure of Turkish appears to increase adjective-noun collocations’ association scores in the both frequency bands since the vast majority of Turkish collocations reach higher scores of collocational strengths than their unlemmatised forms. After the corpus study, I designed psycholinguistic experiments to explore the sensitivity of speakers of these languages to the frequency of adjectives, nouns and whole collocations in acceptability judgment tasks in English and Turkish. Mixed-effects regression modelling revealed that collocations which have similar collocational frequency and association scores are processed at comparable speeds in English and Turkish by L1 speakers of these languages. That is to say, both Turkish and English speakers are sensitive to the collocation frequency counts. This finding is in line with many previous empirical studies that language users process MWS quantitively faster than control phrases (e.g. Arnon & Snider, 2010; McDonald & Shillcock, 2003; Vilkaite, 2016). However, lemmatised collocation frequency counts affected the processing of Turkish and English collocations differently, and Turkish speakers appeared to attend to word-level frequency counts of collocations to a lesser extent than English speakers. These findings suggest that different mechanisms underlie L1 processing of English and Turkish collocations. The present study also looked at the sensitivity of L1 and L2 advanced speakers to the frequency of adjectives, nouns and whole collocations in English. Mixed-effects regression modelling revealed that L2 advanced speakers are sensitive to the collocation frequency counts like L1 English speakers because as the collocation frequency counts increased, L1 Turkish-English L2 speakers responded to the collocations in English more quickly, as L1 English speakers did. The results indicated that both groups showed sensitivity to noun frequency counts, and L2 English advanced speakers did not appear to rely on the noun frequency scores more heavily than the L1 English group while processing adjective-noun collocations. These findings are in conflict with the claims that L2 speakers process MWS differently than L1 speakers (Wray, 2002).

The grammaticality of adverb positions varies by language. Consequently, L2 and L1 speakers may differ from each other in their acquisition of adverb positions. Given that L2 Welsh speakers outnumber L1 Welsh speakers in Wales, differences in acquisition may change which adverb positions occur in contemporary Welsh. This study compares which adverb positions L2 and L1 speakers produce in the spoken data from Cardiff in the CorCenCC corpus (Knight et al., 2020) in order to identify any differences in acquisition. Comparisons of L2 and L1 English speakers find that L2 speakers consistently acquire novel adverb positions yet they frequently use ungrammatical adverb positions. They also do not acquire additional constraints on adverb positions. This study largely reinforces these findings. First, L2 Welsh speakers produce every adverb position that L1 speakers produce. Secondly, although the definiteness constraint that Borsley et al. (2009: 50) describe is not productive in the sample of Cardiff Welsh speakers, L1 speakers exhibit a heaviness constraint on V-Adv-O that L2 speakers do not. Therefore, L1 transfer neither inhibits the acquisition of adverb positions nor facilitates the acquisition of additional constraints. However, unlike L2 English speakers, L2 Welsh speakers do not produce ungrammatical adverb positions. This likely derives from the lack of transferable adverb positions between Welsh and English rather than a lack of transfer. Therefore, this sample of Cardiff Welsh reinforces the crosslinguistic consistency of L2 speakers’ acquisition of adverb positions. It also suggests that L2 Welsh speakers most likely diverge from L1 speakers in the contexts in which they use adverb positions rather than the adverb positions that they use.

A substantial body of research has examined the role of English in South Africans’ linguistic repertoires. Many of these studies have investigated whether a language shift towards English might be underway among first-language (L1) speakers of the indigenous languages. At the same time, the role of English in the repertoires of L1 English speakers has received little attention, the implicit assumption being that English remains dominant for them. This paper presents an empirically informed comparison of English experience within and across L1 and non-L1 (Ln) English speakers in order to shed light on the importance of English in these individuals’ repertoires. Cluster analysis is employed to analyze language background data from bi-/multilinguals (n = 200). The analysis produces two clusters, which differ in extent of English exposure and preference for English use. All but one of the L1 English speakers belong to the ‘higher exposure, higher preference’ cluster, while the vast majority of the Ln speakers fall into the ‘lower exposure, lower preference’ cluster. The results indicate that English experience is relatively homogeneous across L1 speakers but differs, for the most part, across L1 and Ln speakers. The findings are relevant to our understanding of language use in multilingual South Africa.

Many studies of speech accommodation focus on native speakers with different dialects, whereas only a limited number of studies work on L2 speakers' accommodation and discuss theories for second language (L2) accommodation. This paper aimed to fill the theoretical gap by integrating the revised speech learning model (SLM) with the exemplar-based models for L2 speech accommodation. A total of 19 Cantonese-English bilingual speakers completed map tasks with English speakers of Received Pronunciation and General American English in two separate experiments. Their pronunciations of THOUGHT and PATH vowels, and fricatives [z] and [θ] were examined before, during, and after the map tasks. The role of phonetic dissimilarity in L2 accommodation and L2 category formation in the revised SLM (SLM-r) were tested. First, the results suggested that global phonetic dissimilarity cannot predict Hong Kong English (HKE) speakers' accommodation patterns. Instead, the segment-specific phonetic dissimilarity between participants and interlocutors was found to be positively correlated with the participants' degree of accommodation. In addition, HKE speakers who did not form a new L2 category of [z] were found to significantly accommodate toward their interlocutor, suggesting that L2 accommodation might not be constrained by phonological category. An integrated exemplar model for L2 accommodation is proposed to explain these findings.

This article explores the applications of research on native spoken language into second language learning in the concept of subordination. Second language (L2) learners’ ability to integrate subordinate clauses is considered an indication of higher proficiency (e.g., Ellis & Barkhuizen, 2005; Tarone & Swierzbin, 2009). However, the notion of subordination is challenged in the analysis of spoken syntax due to the potential problems in identifying clausal relationships. The study compares the production of if‐clauses of first language (L1; N = 20) and L2 (N = 20) speakers of English in informal conversations and a map task. Results show that L2 speakers exhibited a strong preference for pre‐posed if‐clauses consisting of one if‐clause; compared to L1 users, they used few multiclausal if‐clause complexes and postscript and semiformulaic post‐posed clauses in the conversations. As for the map task, L2 speakers did not use single if‐clause directives, frequently used by L1 speakers. The findings seem to indicate that L2 speakers were constrained in their structural and hence functional repertoire. The article concludes that using subordination as a measure of complexity may not be straightforward, because clausal relations typical of spoken language need to be taken into account.

Phonological research has demonstrated that English intonation, variably referred to as prosody, is a multidimensional and multilayered system situated at the interface of information structure, morphosyntactic structure, phonological phenomena, and pragmatic functions. The structural and functional complexity of the intonational system, however, is largely under-addressed in L2 pronunciation teaching, leading to a lack of spontaneous use of intonation despite successful imitation in classrooms. Focusing on contrastive and implicational sentence stress, this study explored the complexity of the English intonation system by investigating how L1 English and Mandarin-English L2 speakers use multiple acoustic features (i.e., pitch range, pitch level, duration, and intensity) in signaling contrastive and implicational information and how one acoustic feature (maximum pitch level) is affected by information structure (contrast), morphosyntactic structure (phrasal boundary), and a phonological phenomenon (declination) in L1 English and Mandarin-English L2 speakers' speech. Using eye-tracking technology, we also investigated (1) L1 English and Mandarin-English L2 speakers' real-time processing of lexical items that carry information structure (i.e., contrast) and typically receive stress in L1 speakers' speech; (2) the influence of visual enhancement (italics and bold) on L1 English and Mandarin-English L2 speakers' processing of contrastive information; and (3) L1 English and Mandarin-English L2 speakers' processing of pictures with contrastive information. Statistical analysis using linear mixed-effects models showed that L1 English speakers and Mandarin-English L2 speakers differed in their use of acoustic cues in signaling contrastive and implicational information. They also differed in the use of maximum pitch level in signaling sentence stress influenced by contrast, phrasal boundary, and declination. We did not find differences in L1 English and Mandarin-English L2 speakers' processing of contrastive and implicational information at the sentence level, but the two groups of participants differ in their processing of contrastive information in passages and pictures. These results suggest that processing limitations may be the reason why L2 speakers did not use English intonation spontaneously. The findings of this study also suggest that Complexity Theory (CT), which emphasizes the complex and dynamic nature of intonation, is a theoretical framework that has the potential of bridging the gap between L2 phonology and L2 pronunciation teaching.

Non-linguistic factors leave a distinct thumbprint on our speech production that is perceptible to listeners. A steadily growing line of research demonstrates that listeners can perceive a contrast between native and non-native (L2) speakers based on accents and further classify these speakers according to dialectal variation, even when they are not native speakers of a language. Most of these studies have focused on dialectal variation within US English speakers, a combination of US and International English dialects, or L2 speakers representing a wide range of languages. Most have also featured listeners who are monolingual native speakers of the target language coming from a homogenous background, or a contrast between these and a targeted set of L2 speakers. We therefore lack knowledge of how exposure to, or familiarity with, diverse accents and languages, or specific native language competence of the native language of L2 speakers, can guide listeners’ accent perception and categorization. In this research, we employed a free classification task, presenting listeners with speech samples of native speakers with accents representing multiple English dialects, and L2 speakers of nine Asian languages across three geographic regions speaking Asian-accented English. There were six groups of listeners: monolingual US English listeners in a diverse linguistic context, monolingual US English listeners in a homogeneous linguistic context, native speakers of a non-Asian language and English (bilinguals), and native speakers of each of the three target Asian language groups who are L2 speakers of English. The results reveal that nearly all listeners are sensitive to accents capturing native/L2 contrasts and dialectal variation in English. While regular exposure to a diversity of accents results in increased classification accuracy, classification of Asian L2-accented English speakers is best performed when there is alignment of similar language family and geographic area, as demonstrated by South Asian listeners.

This paper explores whether Korean EFL listeners perception of English vowels and their English vowel category development can be estimated from cross-language labelling between English and Korean. The paper also investigates whether cross-language labelling makes different predictions for General American English (GAE) vowels and British English (Received Pronunciation, RP) vowels, given that vowel category variation exists between the two accents. Thirty-six university students in Seoul completed cross-language labelling and English vowel identification. Fit indices were calculated based on cross-language labelling for both GAE and RP vowels in order to answer the questions posed. Specifically, the paper tested the assumption of the Speech Learning Model (SLM) that L2 category development is closely related to the perceived phonetic distance between L1 and L2 sounds by calculating fit indices. The results reveal that the fit indices computed had limitations in accounting for the identification accuracy of L2 vowels. The fit indices also made similar predictions for GAE and RP vowels, but there were accent and vowel category variations between GAE and RP. Thus, the overall results suggest that L2 vowel perception or L2 vowel category learning depends on factors such as L2 learners’ target language accent and their overall interlanguage phonological system, as well as the perceived phonetic distance between an L2 sound and its closest L1 counterpart.

Previous research has indicated that perception and production ability are significantly related among L2 learners in natural learning settings. The present study focused on the relation between these two abilities during speech training, specifically examining whether perception or production training alone led to improvements of production or perception ability, and whether the order of perception and production training influenced the improvement of the abilities. Forty-five Japanese university students received thirteen-week perception and production training on six American English contrasts (e.g., b/v, l/r). One group received perception training first, followed by production training, while the order of training was reversed for the other group. It was found that the initial perception training in the former group resulted in a significant but relatively weak transfer to improvements of production ability, while the initial production training in the latter group showed a significant and strong transfer to those of perception ability. It was also shown that the improvements of perception and production ability during the speech training were significantly related in both groups. The order of training, however, did not significantly influence the overall degree of improvements in either of the abilities. Implications for speech learning models will be discussed.

Human neuronal encoding of English syntactic violations as revealed by both L1 and L2 speakers

Second language (L2) learners often encounter difficulties caused by the interference of their native language (L1). The aim of this study is to examine how the Javanese and Sundanese vowel systems hinder the perception of ten English vowels. Thirty Javanese, thirty Sundanese, and twenty English native speakers participated in a mouse-tracking experiment. Participants were required to identify English vowels corresponding to an auditory token by clicking on one of two word strings presented on a computer screen. According to the Speech Learning Model (SLM) hypothesis, the Javanese and Sundanese speakers were predicted to have higher error rates and show a larger Area Under the Curve (AUC) for similar vowels (same IPA symbols, but different diacritics between L1 and the target vowels) than the native English speakers. For new vowels (no same IPA symbols found between L1 and the target vowels), the L2 speakers were predicted to have lower error rates and a smaller AUC than the native English speakers. According to the Second Language Linguistic Perception (L2LP), however, the prediction is stated in the reverse. Repeated measures of ANOVAs found that: 1) the Javanese and Sundanese speakers were less accurate in perceiving the new vowels /ɑː/, /ʌ/, /ae/, /e/, /ɪ/, and /ʊ/ and similar vowels /iː/ and /uː/. 2) The Javanese speakers showed a larger AUC than native speakers for new vowels /ɑː/, /ɜː/, /ɔː/, and /ʌ/ and for similar vowels /iː/ and /uː/. The Sundanese speakers showed a greater attraction to the incorrect alternatives than the native speakers for new vowels /ɑː/, /ɜː/, /ɔː/, /ʌ/, /ae/, /e/, /ɪ/, and /ʊ/ and similar vowels /iː/ and /uː/. Our findings partially support the L2LP hypothesis that the Javanese and Sundanese listeners are likely to show high error rates and a large attraction towards the incorrect alternatives of new vowels. The results confirmed that perceptual difficulties varied significantly according to the influence of L1 vowel inventories.

Grounded on the theoretical constructs of cross-language speech learning, this research investigates English L1 speakers’ production of Arabic pharyngeal and pharyngealized consonants (/ħ, ʕ/ and /tˤ, sˤ, dˤ, ðˤ/, respectively). Central to this study is how English speakers’ proficiency in Arabic affects their ability to produce these sounds differently. In particular, it examines the effect of Arabic proficiency on the production of pharyngealized versus non-pharyngealized consonants; pharyngeals versus non-pharyngeals; pharyngeal versus pharyngealized sounds; intra-category differences among the target consonants; and pharyngeal and pharyngealized consonants which differ in the adjacent vowel. The study adopts the posttest-only control group design in which Arabic learners and non-learners constitute the research groups. This thesis exploits Best’s (1995) perceptual assimilation model (PAM); Flege’s (1995) speech learning model (SLM) and principles of gestural phonology in interpreting the split-plot ANOVA results. Theoretical and practical implications are discussed based on the study results.