Phonetic Accommodation in Non-native Directed Speech Supports L2 Word Learning and Pronunciation

doi:10.21203/rs.3.rs-3242565/v1

Download PDF

Article

Phonetic Accommodation in Non-native Directed Speech Supports L2 Word Learning and Pronunciation

https://doi.org/10.21203/rs.3.rs-3242565/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 02 Dec, 2023

Read the published version in Scientific Reports →

You are reading this latest preprint version

This study assessed whether Non-native Directed Speech (NNDS) facilitates second language (L2) learning, specifically L2 word learning and production. Spanish participants (N=50) learned novel English words, presented either in NNDS or Native-Directed Speech (NDS), in two tasks: Recognition and Production. Recognition involved matching novel objects to their labels produced in NNDS or NDS. Production required participants to pronounce these objects’ labels. The novel words contained English vowel contrasts, which approximated Spanish vowel categories more (/ɪ-i/) or less (/æ-ɛ/). Participants assigned to the NNDS group recognized novel words faster and learned and pronounced the /ɪ-i/ contrast words better compared to participants in the NDS group. Participants’ ability to discriminate the target vowel contrasts was also assessed before and after the tasks, with no improvement detected in the two groups. These findings support the didactic assumption of NNDS, indicating the relevance of the phonetic adaptations in this register for successful L2 acquisition.

Biological sciences/Psychology

Biological sciences/Psychology/Human behaviour

Non-native Directed Speech (NNDS) is a clear speech register that native speakers use to address second language (L2) learners of their own language. NNDS is often studied in comparison with Native Directed Speech (NDS), which is the register used between native speakers without the intention of enhancing intelligibility (Ferguson & Kewley-Port, 2002). NNDS has also been referred to as “L2 speech accommodation” because it is assumed to be the result of the speaker’s accommodation to the listener’s low L2 proficiency and learning needs (see Giles, 2016; Lindblom, 1990 for theoretical frameworks). In line with this, NNDS results in clear speech, and it is proposed to serve a didactic function by assisting learners to better understand, perceive, and pronounce their L2 (Piazza, Martin, et al., 2022; Scarborough et al., 2007; Uther et al., 2007). Piazza et al. (2022) proposed that the didactic function of NNDS comprises two related aspects: a didactic purpose and a didactic impact. The former is the function of producing clear speech to support L2 teaching, reflected in the acoustic features of NNDS, whereas the latter is the actual effect on L2 learning, perception, and production. While there is evidence for the didactic purpose indicating that speakers systematically adjust their speech production, resulting in clearer speech (Uther et al., 2007), so far, the didactic impact of NNDS has never been directly explored. In the present study, we investigated whether L2 learners benefit from being exposed to NNDS by testing its didactic impact on perceiving, learning, and pronouncing L2 words.

From high clarity to the didactic impact of NNDS

NNDS is characterised by speech adaptations to the non-native listener. Compared with NDS, such adaptations lead to the production of several acoustic features that enhance NNDS clarity and potentially support L2 learning. The most typically studied features of NNDS are speech rate reduction and acoustic exaggeration of vowels, i.e., vowel hyperarticulation (Piazza et al., 2022). Vowel hyperarticulation is assumed to be the key acoustic feature that serves a didactic function in NNDS because it results in a clearer and more distinctive representation of vowel categories (Kuhl, 1997). These features together are proposed to support speech perception, comprehension, and even production (Bradlow & Bent, 2002; Ferguson & Kewley-Port, 2002; Piazza et al., 2022; Smiljanić & Bradlow, 2005, 2009).

Indirect evidence that NNDS supports speech comprehension is provided by studies showing that listeners rate the intelligibility of NNDS higher than that of NDS. For instance, Knoll et al., (2009) and Uther et al., (2007) asked naïve native listeners to rate NNDS and NDS audio samples. They found that NNDS was rated as clearer than NDS but less so than other clear speech registers, like Lombard speech, which is a register produced to contrast background noise during native-native interactions (Garnier & Henrich, 2014; Hazan et al., 2015; Lombard, 1911). Conversely, L2 learners have been reported to understand NNDS better than both NDS and Lombard speech (Bobb et al., 2019a; Kangatharan, 2015). Given that Lombard speech is not oriented to L2 learners, these seemingly conflicting results could be due to the lack of a didactic function (both purpose and impact) in Lombard speech (Piazza et al., 2022).

These rating findings indicate that the perception of NNDS and its enhancement of clarity differ between native and L2 learners, suggesting that there may be differences in how helpful NNDS may be for the two populations (Rothermich et al., 2019). However, direct evidence showing that NNDS supports spoken word learning, recognition, or pronunciation in L2 learners is still missing. Few experiments have tested the efficacy of clear speech registers for word learning in adults. For instance, (Golinkoff & Alioto, 1995; Ma et al., 2020) found that Chinese Infant Directed Speech (IDS) helps non-native adult participants to learn words. IDS shares various acoustic features (including vowel hyperarticulation, Uther et al., 2007) and proposed didactic function with NNDS, although these registers are intended for different addressees. Thus, one could expect that NNDS is particularly suited to support adults’ L2 learning. To test this assumption here, we investigated how L2 learners acquire perception and pronunciation of L2 words and phonemes when exposed to NNDS. In the following section we introduce the most relevant aspects of L2 word learning and the difficulties that novice L2 learners can face during this process.

Aspects of auditory L2 word learning

Perception and assimilation. Initial L2 word learning is primarily mediated by the perception of novel phonemes (Escudero, 2005, 2006; Flege, 1995; Melnik-Leroy et al., 2022; Van Leussen & Escudero, 2015). L2 learners often have difficulties in discriminating phonetic contrasts that are not present in their L1 (both vowels and consonants). The relative difficulty in distinguishing L2 phonemes depends on the perceptual assimilation to the listener’s L1 phonology (Best & Tyler, 2007; Flege, 1995; Mora et al., 2022). According to the Perceptual Assimilation Model for L2 (PAM-L2, Best, 1991; Best & Tyler, 2007), the most difficult situation for L2 perception is when the two L2 phonemes map onto a single native category (see also Flege, 1995; Flege & Bohn, 2021; Kramsch, 2007 for alternative frameworks). In this case, the two L2 phonemes can either map equally poorly to the native category (Single Category), or one phoneme can be a better fit than the other (Category Goodness). For Spanish learners of English, an example of Single Category is the vowel contrast /ʌ-æ/ (contained in words like cap/cup), comprised by two vowels that are not present in the Spanish phonemic inventory. In this case the pair of L2 vowels fall within the perceptual space of a single L1 vowel category (/a/), which makes it difficult to perceive the phonetic differences between the vowels (Baigorri et al., 2019; Escudero, 2001; Rallo Fabra & Romero, 2012). Conversely, an example of Category Goodness for Spanish listeners is the vowel contrast /i-ɪ/ (contained in words like sheep/ship), in which the /i/ of sheep is a better instance of the Spanish /i/ than /ɪ/ (which is not present in the Spanish phonemic inventory). According to PAM-L2 (Best, 1991; Best & Tyler, 2007), instances of Category Goodness are relatively easier to perceive than Single Category. To test this, Baigorri et al. (2019) investigated late Spanish-English bilinguals‘ categorical perception of English vowel contrasts. Participants had particular difficulties recognizing /æ - ɑ/, /ʌ - ɑ/, and /ʌ-æ/ contrasts, whereas discrimination accuracy was higher for /ɪ - ɛ/ and /i-ɪ/ (see also Boomershine, 2013for similar findings on perceptual ratings). Although accuracy was higher for /i-ɪ/, other studies found that late bilinguals tend to perceive /i-ɪ/ vowels in a less categorical way than native listeners (Escudero, 2001; Peng et al., 2010).

There is broad consensus that experience (re)shapes L2 learners’ phoneme perception (Aoyama & Flege, 2011; Escudero, 2006; Flege et al., 1997; Flege & Liu, 2001) Flege et al. (1997) tested experienced and inexperienced L2 learners of various languages on synthetic /i-ɪ/ and /æ - ɛ/ continua and reported that the experienced group was more accurate than the inexperienced group at both perceiving and producing the vowel contrasts. This suggests that the perceptual system adapts to learning novel vowel contrasts and that perception can be changed with training (Iverson et al., 2008; Logan et al., 1991; Tremblay et al., 1998; Wang, 1999). However, it is not clear how much training is needed to observe such a change in the perception of phonological boundaries in L2 learners. Some studies found that perceptual change happens only in mid-to-high proficiency L2 learners (Reinisch et al., 2013), whereas others found changes in low proficiency L2 learners within the duration of an experimental session (O. A. Drozdova et al., 2015; P. Drozdova et al., 2016). Nevertheless, it is currently unknown whether such a perceptual change occurs in L2 learners after exposure to NNDS. L2 learners’ perceptual change of L2 phonemes is likely an important step in the learning process. Testing learning in the context of NNDS will also shed new light on the adaptation of phonological boundaries after short training in the L2.

In both types of phonetic assimilation discussed above, problems with the correct mapping of L2 phonemes hinders L2 learners from creating two distinct vowel categories. This determines the difficulty in perceiving and producing these vowels in a distinct manner (Flege, 1995; Mora et al., 2022). So far, there is little evidence regarding the effectiveness of phonetic training for improving such mappings and phonological representations (Lee et al., 2015), and, to our knowledge, there is no research on the effectiveness of NNDS in improving L2 perception. Therefore, in this study we focused on the learning process of both perception and production of L2 vowels and words in NNDS. We focused on the NNDS didactic impact for learning the two types of assimilation categories, Single Category and Category Goodness respectively, as realised by the /ʌ-æ/ and /i-ɪ/ English vowel contrasts. By doing this we aim to provide a well-rounded research approach for the study of the didactic impact of NNDS with a simulation of L2 learning of English words.

Production. The studies reviewed above all focused on L2 phoneme perception. However, this is just one, although fundamental, aspect of learning an L2, which also include production (Flege, 1995). L2 learners must deal with the challenge of correctly pronouncing novel words, and most adult L2 learners do not reach native-like pronunciation. Instead, speaking with a non-native accent, dependent on their L1, is common (Derwing & Munro, 2009; Flege, 1984). It is worth underlining that L2 learners’ most important objective is reaching comprehensible speech, rather than sounding like a native speaker (see Cook, 2016; Rothman et al., 2023 for a discussion). Although L2 learners’ non-native pronunciation is expected, most naïve learners also have issues in distinguishing the pronunciation of L2 vowel contrasts (O’Brien, 2021). This makes the two vowels difficult to distinguish, lowers intelligibility, and possibly leads to miscommunication. Thus, to accurately pronounce L2 vowels, phonetic differences between vowel categories must be learned. For this reason, we are also interested in investigating whether exposure to NNDS confers advantages in learning to pronounce words and vowels.

The present study

L2 learners perceive NNDS to be clearer than NDS (Bobb et al., 2019), but to date, research assessing the impact of NNDS on L2 learning is not available (Piazza et al., 2022). To disclose the didactic impact of NNDS in the L2 learning process, there is need for research on the effect of exposure to NNDS on learning, perceiving, and producing L2 words and vowels. For this purpose, we recruited Spanish native listeners who were novice learners of English to participate in an online experiment. Participants were randomly assigned to a register group (NDS, NNDS) and asked to learn a set of 24 English pseudowords. All participants learned three types of novel words: (1) minimal pairs containing the /ʌ-æ/ contrast (like guck/gack), (2) minimal pairs containing the /i-ɪ/ contrast (like deest/dist), and (3) non-minimal pairs containing the /a/ and /u/ vowels (like parg/phoon), which were included as fillers to increase item variability. Participants were auditorily taught the associated label for each object in either NNDS or NDS. After this brief learning phase, participants completed three tasks to test word learning, word production, and vowel perception. Participants completed multiple blocks in each task, so that these tasks were part test and part training.

1) Recognition task. Participants were tested on the association between the labels and novel objects. Accuracy and response times across blocks (Block factor) were compared between the NNDS and NDS groups.

2) Production task. Participants were presented with the previously learned objects, one by one, and were asked to pronounce their names. Accuracy and response latencies across blocks (Block factor) were compared between the NNDS and NDS groups. Also, we computed phonetic accuracy as means of a) Euclidean distance between the participants’ vowel production and the vowels in the stimuli. In addition, we computed the b) Euclidean distance within each vowel contrast (/i-ɪ/ and /ʌ-æ/) in participants’ productions. This latter measure enabled us to investigate phonetic accuracy from the perspective of vowel distinctiveness.

3) Continuum discrimination task. Participants were administered two continuum categorical perception tests of the /i-ɪ/ and the /ʌ-æ/ contrasts embedded in familiar real words. We tested participants before the learning phase (pre-test) and after they completed both the Recognition and the Pronunciation tasks (post-test). That is, we investigated potential changes in participants’ ability to discriminate these vowels as a result of exposure to the sounds in NNDS or NDS registers.

Using these tasks, we were interested in answering the following questions:

1) Does NNDS enhance word learning as compared to NDS?

2) Does exposure to NNDS improve L2 vowel pronunciation accuracy and distinction as compared to NDS?

3) Does exposure to NNDS as compared to NDS shape L2 vowel perception?

The Recognition and Production tasks aimed to answer the first question. In line with the assumption that NNDS yields a didactic impact on the process of L2 learning, we expected the NNDS group to learn the words and vowel sounds better than the NDS group. This would be revealed by a steeper learning curve across blocks and faster responses in the Recognition task. NNDS is also assumed to deliver articulatory information by providing L2 learners with exaggerated phonetic contrasts, which is not the case for NDS. Thus, in the Production task, we expected higher accuracy (i.e., production closer to target reference) and faster responses with a steeper learning curve in the NNDS group compared to the NDS group. In addition, for Spanish participants, the /ʌ-æ/ contrast (Single Category, henceforth Single) is expected to be more difficult to produce than the /i-ɪ/ contrast (Category Goodness, henceforth Goodness) (Best & Tyler, 2007; Escudero, 2006; Flege et al., 1997). Thus, we expected participants to have lower accuracy and slower response times, in both tasks, for the Single than Goodness contrast.

The Production task also aimed to answer the second research question. As NNDS provides enhanced articulatory information, the NNDS group was predicted to produce vowels that are closer to what they heard as compared to the NDS group. This should be reflected by a reduced Euclidean distance between participants’ production and the target reference – and steeper improvement across blocks – in the NNDS as compared to the NDS group. We also expected NNDS participants to pronounce vowel contrasts (Single and Goodness) in a more distinctive way than the NDS group, reflected by greater Euclidean distance between vowels of both contrasts.

Lastly, the Continuum discrimination task aimed to answer the third research question. Previous research suggests that Spanish speakers struggle differentiating the vowel pairs used in this study. Native perception of vowels is quasi-categorical (Altmann et al., 2014), but non-native perception is not. Thus, both NDS and NNDS participants, with low levels of English knowledge, were not expected to show a clear perceptual boundary between the two target vowels in the pre-test. However, if NNDS enhances vowel discrimination, this may also transfer to previously known words. So, in the post-test, we expected only the NNDS group to show more native-like perception of the two contrasts. This would suggest that NNDS induces adaptation in the listener’s L2 perceptual system after short training.

Recognition task

This task aimed to investigate whether NNDS promotes L2 novel word learning. Accuracy. The final model indicated a significant effect of the Block factor’s linear term (β = 0.561, z = 4.019, p < .001) but not quadratic term (β = -0.0655, z = -0.809, p = .419). Participants improved in accuracy linearly from 56.75% on average in Block 1 to 71.63% in Block 6. The main effects of Register (β = -0.277, z = -0.785, p = .432) and Contrast (β = -0.261, z = -1.866, p = .062) were not significant but their interaction was (β = 0.533, z = 4.007, p < .001). The NNDS and NDS groups did not differ for the Single (z = 0.028, p = 1) or the Goodness accuracy (z = 0.502, p = 0.436). However, within contrasts, NNDS participants were more accurate in recognizing novel words containing the Goodness contrast than the Single contrast (z = 3.373, p = .004; see Fig. 1). Conversely in the NDS group this difference was not significant (z = -0.034, p = 1). No other interactions were significant (see Data Availability).

INSERT FIGURE 1 ABOUT HERE

Response latencies. The final model showed significant effects of linear (β = -0.077, t = -13.754, p < .001) and quadratic terms (β = 0.033, t = 6.854, p < .001). This was due to a decrease in reaction time, from 6338ms on average in the 1st block to 4834ms in the 5th block, and then reached plateau performance in the 6th block (4892ms on average). Also, the effect of Register was significant (β = 0.051, t = 7.562, p < .001) with the NNDS group (3792ms) responding overall faster than the NDS group (4551ms; see Fig. 2). Conversely, the effect of Contrast (β = -1.525e-04, t = -0.019, p = .984) and any interaction did not reach significance.

INSTERT FIGURE 2 ABOUT HERE

Production task

This task investigated whether NNDS promotes learning of novel words for production. One participant was excluded from the analyses due to very low production accuracy ( ̴ 8%).

Accuracy. Results indicated significant improvement in model fit on the linear term (β = -0.469, z = -2.775, p = .005), but not the quadratic term (β = 0.086, z = 1.043, p = .297). This is reflected in the Aline distance scores, which, on average, went from 0.332 in the 1st block to 0.235 in the 6th block. Effects of Register (β = 0.231, z = 1.128, p = .259) and Contrast (β = 0.074, z = 0.446, p = 0.656) were not significant, but their interaction was (β = -0.315, z = -2.298, p = .022). This interaction was likely driven by the NDS group’s difference between Single and Goodness contrasts, but post-hoc analyses failed to reveal any differences between Single and Goodness in both the NDS (t = 1.436, p = .477) and NNDS groups (t = − 0.446, p = .971; see Fig. 3). No other interactions were significant.

INSTERT FIGURE ABOUT HERE

Response latencies. The final model yielded a significant quadratic term (β = 0.048, t = 2.844, p = .004) but not linear term (β = 0.014, z = 0.817, p = .414), indicating that participants’ response latencies across blocks best fitted a parabola shape. The Contrast factor showed a significant effect (β = -0.049, t = -2.093, p = .037) reflecting overall shorter latencies in producing the Goodness contrast (1946ms) than the Single contrast (2408ms; see Fig. 4), particularly form Block 3 onwards. Conversely, the Register factor (β = -0.029, t = -0.390, p = .698) was not significant and neither were any interactions.

INSERT FIGURE 4 ABOUT HERE

Euclidean distances. The analysis of Euclidean distances assessed whether the exposure to NNDS improved participants’ vowel production accuracy and category distinction as compared to NDS. For this purpose, we computed two measures of Euclidean distance:

a) Euclidean distance between participants’ vowel production and training vowel. For this variable, the Goodness (/i-ɪ/) and Single (/ʌ-æ/) contrasts were analysed in two separate models, each including the factors Vowel (i and ɪ or ʌ and æ) and Register. The results for the Goodness model did not reveal significant improvement in model fit on the linear (β = 0.0375, t = 0.481, p = .630) or quadratic terms (β = 0.381, t = 0.504, p = .614). The main effects of Register (β = 0.387, t = 5.448, p < .001), Vowel (β = 0.417, t = 2.461, p = .038) and their interaction were significant (β = 0.219, t = 3.320, p < .001). Post-hoc analyses indicated that the NDS group produced /ɪ/ further from the target than /i/ (t = 3.269, p = .031), whereas this was not true for the NNDS group (z = 2.144, p = .198). Also, NDS participants pronounced /ɪ/ with greater distance from the target vowel than the NNDS participants (t = -5.302, p < .001; see Fig. 5). This was not true for the /i/ vowel (t = -2.219, p = .127). This result suggests lower phonetic accuracy for the NDS than the NNDS group for this vowel. The model for the Single contrast did not reveal any significant effects (see Figure A of the Appendix 2) (Register β = -0.257, t = -0.455, p = .650, Vowel β = -0.012, t = -0.146, p = .887, Block linear term: β = -0.031, t = -0.538, p = .591; Block quadratic t.: β = -0.057, t = -0.980, p = .327) or interactions.

INSERT FIGURE 5 ABOUT HERE

b) Euclidean distance within vowel contrasts. The two contrasts, Goodness (/i-ɪ/) and Single (/ʌ-æ/), were separately investigated in two models, which included the Block and Register factors. The final model for the Goodness contrast indicated a main effect of Register (β = 0.166, t = -2.520, p = .016) but no effect of linear (β = -0.061, t = -1.396, p = .163) or quadratic terms (β = -0.005, t = -0.114, p = .909). The NNDS group produced the vowels in this contrast more distinctly than the NDS group (Euclidean distance NNDS = 0.732 Hz²; NDS = 0.557 Hz²), without substantial changes across the 6 blocks (see Fig. 6). Instead, the model for the Single contrast did not show any significant effects or interactions (Register: β = 0.021, t = 0.518, p = .607; linear term: β = − 0.035, t = -1.178, p = .239; quadratic term: β = -0.026, t = -0.853, p = .394).

INSERT FIGURE 6 ABOUT HERE

3.4.3. Continuum discrimination task

This task assessed whether the brief exposure to the target vowel sounds in NNDS or NDS induced changes in the participants’ L2 perceptual system that transferred to real English words. No significant main effects or interactions were found in the final model for either the sheep-ship continuum (Register, β = 0.419, z = 1.287, p = .198, Exposure, β = 0.105, z = 0.800, p = .424) or the cup-cap continuum (Register, β = 0.264, z = 0.942, p = .346; Exposure, β = -0.085, z = -0.924, p = .355, see Fig. 7).

INSERT FIGURE 7 ABOUT HERE

Previous literature has assumed that NNDS is endowed with a didactic purpose – reflected in the acoustic features of NNDS – and a didactic impact (Piazza, Martin, et al., 2022; Piazza et al., 2023; Scarborough et al., 2007; Uther et al., 2007). Such a didactic impact would support L2 learners both in comprehension and production. However, so far, whether L2 learners’ perceptual and production learning is promoted by exposure to NNDS remained unknown. We addressed these questions by conducting an online experiment where two groups of L2 learners of English (Spanish L1) learned the association between novel objects and novel English words pronounced in either NNDS or NDS. Perception and learning of English vowel contrasts (/i-ɪ/ = Goodness, /ʌ-æ/ = Single), which are absent in the Spanish phonological inventory, was assessed. In order to investigate whether NNDS yields learning benefits in the production of novel words and vowels, participants’ production accuracy and latency were also measured. We predicted that the group exposed to NNDS would learn to perceive and pronounce novel words and vowel contrasts more successfully and faster than the NDS group.

NNDS Benefits

The present study provides the first evidence for the benefits of NNDS on L2 learning. That is, NNDS participants were better at perceiving L2 novel words as compared to NDS participants. Such a benefit was mainly shown in the Recognition task results, which indicated that the NNDS group responded faster than the NDS group in recognising novel words (both vowel contrasts). This represents evidence in support of the didactic function hypothesis of NNDS and speech accommodation theories (Bobb et al., 2019; Giles, 2016; Lindblom, 1990, 2009; Piazza et al., 2022; Uther et al., 2007).

NNDS benefit depends on properties of the speech contrasts to be learned

Our results also provide evidence that NNDS effects are qualified by the properties of the speech contrasts to be learned, and how they relate to listeners’ L1 (Best & Tyler, 2007). That is, NNDS benefits were particularly pronounced for the Goodness contrast (/i-ɪ/). For example, Recognition accuracy of the NNDS group was higher for the Goodness than the Single contrast, whereas there was no such improvement in the NDS group. This suggests that even though the NNDS group did not show overall better accuracy than the NDS group for both contrasts, their exposure to NNDS promoted recognition of words including the Goodness contrast. On the other hand, Production results showed that NNDS delivered articulatory information that improved L2 pronunciation learning, but for the Goodness contrast only (in line with Best & Tyler, 2007). The NDS group produced /ɪ/ vowels further away from the target than /i/ vowels, whereas this was not true for the NNDS group. More importantly, NNDS participants produced /ɪ/ closer to the target than NDS participants. This means that the NNDS group was more precise in producing the /ɪ/ target vowel than the NDS group. Moreover, the NNDS group was better than the NDS group at separating the vowels of the Goodness contrast (larger /i-ɪ/ distance in their production). These results suggest that NNDS provides articulatory information to the listeners, who use such cues to achieve the target pronunciation.

These findings are probably due to the acoustic features of NNDS, which enhance the differences between contrast vowels. The NNDS novel words containing the Goodness contrast were produced (by a native speaker who recorded the stimuli) with greater /i-ɪ/ duration differences and reduced Euclidean distance than the same novel words pronounced in NDS (see Material and Appendix 2). Participants’ performance was in line with previous literature that reported Spanish listeners to be particularly sensitive to duration differences between L2 vowels (Escudero, 2001, 2006). This also indicates that NNDS duration cues (directed to Spanish listeners) are particularly suited to enhance L2 learners’ discrimination of the /i-ɪ/ vowel contrast rather than contrasts signalled by formant value information. Research has suggested that such cues are intuitively produced by native speakers to support communication with L2 learners (Giles, 2016; Piazza et al., 2022, 2023; Uther et al., 2007; see also Appendix 2). Here, we show that these duration cues also support word learning, bearing a didactic impact for L2 learners. Our results do not enable us to exclude a beneficial role of NNDS in supporting learning of Single contrast vowels, but suggest that such effect may be weaker and require more exposure to lead to detectable improvements. Lastly, production latency results showed that participants were faster to respond to Goodness than Single contrast words, regardless of the Register group. This finding does not relate to our focus on differences between NNDS and NDS, but it is still interesting because it confirms that the Goodness contrast used here is easier to discriminate than the Single contrast for our participants, as we discuss below.

Theories of Second Language Acquisition that explain the NNDS benefit

The asymmetrical benefit we observed between Goodness and Single contrasts is in line with PAM-L2, which claims that Goodness contrast phonemes are more easily recognized and pronounced than Single contrast phonemes (Best, 1991; Best & Tyler, 2007). Other second language acquisition accounts also provide explanations for this learning asymmetry, such as the Speech Learning Model (Flege, 1995; Flege & Bohn, 2021) the Contrastive Analysis Hypothesis (Foley & Flynn, 2013; Kramsch, 2007), and the Input Hypothesis (part of the Monitor Model; (Krashen, 1989,1982, 1985). For instance, the Input Hypothesis assumes that L2 learners acquire language when they are exposed to comprehensible input. This refers to language input containing previously acquired elements and new instances that are slightly beyond L2 learners’ current level of proficiency. Accordingly, when participants learned the Goodness contrast, /i/ represented the known element and /ɪ/ the new instance to be acquired. Conversely, the Single contrast was far beyond participants’ proficiency level for substantial improvement. Thus, participants in both NNDS and NDS groups may have received comprehensible input that facilitated some level of learning.

However, participants showed greater improvement in learning the Goodness contrast than the Single contrast if exposed to NNDS rather than NDS. But the above-described accounts, including the Input Hypothesis, do not consider such an interaction between Register and Contrast. For instance, the Input Hypothesis does not specifically address the learning of phonetic contrasts or pronunciation and does not fully explain the observed benefit in NNDS compared to NDS. The significant improvement in learning observed in NNDS compared to NDS suggests that there may be additional factors at play.

To provide a more comprehensive explanation, it may be necessary to incorporate additional theories and factors. The interaction between Register and Contrast can be further explained by adding a complementary socio-cognitive factor (Atkinson, 2014) to the previous models, which would provide a combined framework that can explain this advantage for the Goodness contrast in NNDS (Kuhl, 1997; Piazza et al., 2022; Uther et al., 2007). The socio-cognitive theory of second language acquisition claims that L2 learning is a natural and adaptive process of ecological alignment (Atkinson, 2002, 2014; Atkinson et al., 2007). In fact, our results reveal that learners adapt their perception and production of L2 novel words to the social environment (i.e., learning differs depending on speech adaptation of the speaker/teacher). This suggests that NNDS is a socially mediated promoter of phoneme category distinction and acquisition.

NNDS benefit depends on the modality and task demand

NNDS seems to be a suitable tool for teaching a second language, which supports L2 learners’ performance, both in recognition and production. However, the present study also revealed that this overall L2 support differs depending on the modality (i.e., word recognition vs. production). We observed better recognition and production performance in the NNDS than NDS group (as for the Goodness contrast), but the production benefit was limited to higher accuracy in target vowel pronunciation (Euclidean distance measures). Contrary to our expectation, accuracy (Aline distance) and production latency results did not differ across registers. Even though the model for the Aline distance yielded a significant interaction between Register and Contrast, post-hoc analyses failed to highlight any statistically significant differences between conditions. We argue that this interaction was driven by the NDS group’s higher Aline distance (hence lower phonemic accuracy) in the Single contrast than the Goodness contrast (see Fig. 3). However, this remained an anecdotal observation as no significant differences between conditions were observed. In sum, the NNDS benefit was visible in faster word recognition but not in faster word production. It could be that NNDS is beneficial for word production accuracy and speed as well, but that longer training would be needed to observe those effects on production. This assumption is in line with previous literature reporting that, when learning linguistic elements, comprehension precedes production learning (Childers & Tomasello, 2002; Gershkoff-Stowe & Hahn, 2013; Hendriks & Koster, 2010).

However, it is worth noting that the Production task was always carried out after the Recognition task. Participants first learned to perceive the differences between vowels and novel words, and only afterwards were asked to produce them. We argue that this could be the main cause of the observed disadvantage in the production of /ɪ/ of the NDS participants. During the Recognition task, NDS participants were exposed to novel words containing the /i-ɪ/ contrast in which the duration cue (/ɪ/ shorter than /i/) was reduced as compared to the NNDS group. We think this absence of clear duration cues might have impaired accurate perception (and thus learning) of the Goodness contrast. Thus, NDS participants carried over this disadvantage to the Production task, where they could not improve their production (Flege, 1995). Nevertheless, NNDS participants, who were exposed to reduced /i-ɪ/ Euclidean distance as compared to NDS, instead produced wider /i-ɪ/ Euclidean distance than NDS participants. This leads to two important observations: a) the NNDS production benefit does not simply derive from mirroring production of target phonemes; b) for Spanish listeners, duration cues are particularly relevant for learning the /i-ɪ/ contrast, and this affects vowel formant production as well.

An important consideration is that the present study used an online method to collect participants’ responses. Several studies have addressed the question of whether online experiments provide reliable results and revealed that chronometric experiments for speech production can be implemented online without information loss (Anwyl-Irvine et al., 2020; Bridges et al., 2020; Fairs & Strijkers, 2021; Piazza, Kartushina, et al., 2022; Vogt et al., 2021). Thus, we are confident in sustaining that the differences we found between speech register groups were genuine and not driven by the online setting. However, future research should run similar experiments in a laboratory to dispel any doubts that the benefit derived from the exposure to NNDS differs online and onsite.

NNDS does not induce changes in L2 sound phonetic boundaries (after short training)

Lastly, we found that the effects of NNDS exposure did not – at least in this study – change participants’ phonetic boundaries of the /i-ɪ/ and /ʌ-æ/ contrasts: phonetic boundaries did not become more native-like despite the improvement in both word recognition and production. In the Continuum discrimination task, we expected to find an adaptation of the phonetic boundaries for both continua (sheep-ship and cup-cap) in the NNDS group’s post-test. However, we did not find any difference between the two groups, nor between pre-test and post-test in both vowel continua. This means that the two groups did not significantly differ for initial perception of the two vowel contrasts, and that neither of the two changed their phonetic boundaries in the post-test. Previous research suggests that adaptation of phonetic boundaries can happen within a single experimental session (O. A. Drozdova et al., 2015; P. Drozdova et al., 2016), whereas other research points that longer exposure and experience is needed (Reinisch et al., 2013). Our result aligns with the latter proposal. However, research reported that phonetic adaptation within a single experimental session is visible at the neurophysiological level (Grimaldi et al., 2014). We cannot exclude, therefore, that NNDS induces phonetic adaptation after short training, but it is not detectable at the behavioural level, with the particular task and stimuli we used. Thus, further research (both using behavioural and neurophysiological methods) is needed to address this point.

To summarise, this study provides new insights on the process of learning an L2 after exposure to NNDS and makes a step forward to understanding the precise mechanisms involved in L2 teaching and learning. We found that NNDS has an impact on learning L2 words for recognition and production, but (especially) production improvements depend on the relationships between the phonemes to be learned and learner’s L1 phonemic categories. It is important to underline that, in this study, participants were exposed to NNDS (or NDS) for a very short period (< 2 hours); hence, it is probable that more benefits would derive from extended exposure to NNDS (e.g., classroom teaching). These findings and future research on more prolonged exposure to NNDS are fundamental to building models of L2 communication and learning. This research is particularly relevant given that communication between native and non-native speakers is becoming ever more frequent in our increasingly multicultural and multilingual societies.

Participants

We recruited 50 native Spanish participants with a low-to-mid level of English knowledge, aged 18–40. Participants were recruited following an individual interview with an expert linguist, who assessed their English level and assigned marks from 1.0 to 5.0 (1.0 = low; 5.0 = native-like). In the interview, fluency, vocabulary, grammar, and pronunciation were evaluated, and then combined into an overall mark. We only recruited participants who obtained an overall mark between 1.0 and 3.0 (NDS group: M_mark = 1.8, SD = 0.45, NNDS group: M_mark = 1.9, SD = 0.32). The participants were randomly assigned to one of two groups (25 participants each), exposed to either NDS or NNDS (NDS group: M_age = 26.76 years, SD = 6.55, Male = 3; NNDS group: M_age = 27.36 years, SD = 6.48, Male = 3). In addition, at the end of the experimental session, participants were asked to carry out a Raven matrices test and a pseudoword repetition task in Spanish, used as indices of participants’ non-verbal IQ and phonological memory (Clark et al., 2012; Kaufman & Kaufman, 2014; see Appendix 4 for a description of these tasks). All participants signed an informed consent form before starting the experimental procedure, and the study was approved by the Basque Center on Cognition, Brain and Language (BCBL) Ethics Committee and conducted in accordance with the relevant guidelines and regulations. Participants were paid 20 euros for taking part in the study.

Bayesian analyses showed that the two groups did not significantly differ in age, English proficiency, non-verbal IQ, and phonological memory. Two-tailed analyses with Cauchy prior distribution (scale of γ = 0.707) revealed that age, proficiency, IQ, and phonological memory of the two groups were respectively (Bayes factors, BF₀₁) 3.39, 2.25, 3.05, and 3.53 times more likely under the null than the alternative hypothesis.

Material

Empirical evidence on the realisation of vowels other than /a/,/i/,/u/ (e.g., /ɪ/, /ʌ/, /æ/) in NNDS is limited in the literature. For this reason, we first ran a pilot study to assess matrices of NNDS adaptation on /ɪ/, /ʌ/, /æ/ vowels. We recruited five native speakers of English (British accent), who were (or had been) teachers of English with Spanish speaking students. We report the results and description of this preliminary study in Appendix 1. Below, the materials used in the three tasks are described.

Recognition task and Production task. For the present study, we created 16 novel words containing the /i-ɪ/ (e.g., [di:st - dɪst]) and /ʌ-æ/ contrasts (e.g., [gʌk – gæk]). The novel words for both vowel contrasts were minimal pairs, so that participants had to rely on the target vowels to distinguish the words. To increase item variability, we also created 8 novel words containing the /a/ and /u/ vowels (not forming minimal pairs) that served as fillers (e.g., [p^ha:g – fu:n]; see Appendix 2 for the full list of experimental stimuli). A set of 24 novel objects was selected to match the 16 target novel words and 8 filler words. The images were taken from the Horst & Hout (2016) novel object database and represented unknown objects and unfamiliar tools. To create the object-word pairings while avoiding any effects derived from specific relations between words and objects in our stimuli, we created 3 lists of pseudo-random associations, and the presentation of these word-object lists was counterbalanced across participants.

The stimuli were recorded by a female native speaker of British English. This speaker was chosen from the 5 speakers who participated in the pilot study as best representing the observed preliminary results (see Appendix 2; wider vocalic area, longer sentence duration, larger /ʌ-æ/ Euclidean distance and /i-ɪ/ duration difference). This speaker produced novel words in NNDS with wider vocalic area (+ 187%), longer sentence duration (M_NNDS= 3640ms, M_NDS= 3561ms), greater /ʌ-æ/ Euclidean distance (M_NNDS= 358.10 Hz², M_NDS= 161.96 Hz², and larger /i-ɪ/ duration difference (M_NNDS =15ms, M_NDS= 4ms) than in NDS. Conversely, she produced smaller /i-ɪ/ Euclidean distance in NNDS than NDS (M_NNDS= 933.88 Hz², M_NDS= 1169.25 Hz²). All stimuli were normalised for intensity and used in both the Recognition and the Production task.

Continuum discrimination task. A female native speaker of British English, who did not record the stimuli for the other tasks, was recorded while producing the words sheep, ship, cup, cap. These recordings were used to create two seven-step continua. The sheep-ship continuum was created by gradually changing the formants and the length of the target vowels. The cup-cap continuum was created by solely changing the formants of the target vowels as this contrast is not marked by vowel duration (Escudero, 2001, 2006). Based on the continua, we created 7 isolated instances of words from sheep to ship and from cup to cap that were used in this task.

Procedure

The experiment was administered online via PennController for Ibex (Zehr & Schwarz, 2018), which is a JavaScript-based platform. During the session, participants remained connected with the experimenter via Zoom™, but video streaming was always disabled. This allowed the experimenter to verify that participants’ microphone worked properly and that they stayed focused on the task, without the participants feeling observed during the session. We asked participants to wear headphones and a head-mounted microphone if available, but any type of microphone with acceptable quality was allowed. Before the start of the experiment, participants recorded and played back their own voice to self-check audio quality. Participants’ compliance was confirmed using a screening test (Woods et al., 2017). After that, the experimental session followed this order: Continuum discrimination task (pre-test), Familiarisation phase, Recognition task, Production task, Continuum discrimination task (post-test), Raven matrices test, Pseudoword repetition task.

Continuum discrimination. The task began by displaying two images on the screen, one at a time (either a sheep and a ship or a cup and a cap, in counterbalanced order across participants). For each image, participants were presented with an auditory recording of the image’s name pronounced in NDS. Then, the task started, and participants used the mouse to click a button on the centre of the screen to listen to the stimuli. They were presented with one sound of the continuum at a time (in a random order). The two pictures previously displayed (a sheep and a ship or a cup and a cap) were presented on the screen and participants were asked to click on the picture corresponding to the word they heard. Each endpoint and mid-step word (7 in total) were repeated 6 times (42 trials per contrast). After completing the block corresponding to the first two images (e.g., sheep and ship), the same procedure was followed for the other minimal pair (e.g., cup and cap). Both pre-test and post-test followed the exact same procedure.

Familiarisation phase. The object-word pairs were presented once during this phase. Participants were exposed to the novel objects presented together with the auditory version of their name, embedded in a carrier phrase (e.g., “this is a deest”). The images of the objects were presented and after 250ms the phrase containing the label was played. Next, a button appeared on the screen and the participants clicked on it to proceed to the next object. Each sentence was pronounced in either NNDS or NDS, depending on the participants’ group allocation. Target and filler novel words were presented in a random order and no response was required by participants (passive learning task).

Recognition task. Participants saw images of 4 objects on the screen and heard a sentence used in the familiarisation phase (e.g., “this is a deest”). The 4 objects comprised the target object (e.g., the referent of deest), a competitor (e.g., which served as a referent of dist on another trial) and two distractors (e.g., which served as referents of gack and phoon on other trials). Participants used the mouse to click a button on the centre of the screen to hear the cue-sentence. Then, the objects were displayed on the screen until participants provided a response by clicking on one of the 4 objects. As soon as they did so, all the objects disappeared and the correct one was displayed on the centre of the screen for 2500ms. This provided feedback on the correct answer to participants. Each block included 24 trials (16 experimental trials + 8 fillers) and participants were exposed to 6 blocks in a row (total number: 96 target trials + 48 fillers). In this way, each block served both as a test and further training of the novel words. Stimuli presentation was pseudorandomized to prevent the same target vowel appearing more than twice in a row.

Production task. Participants were presented with the same 24 objects from the recognition task. The objects were displayed one at a time on the screen and the participants were asked to name each of them. As soon as an object was displayed on the screen, the browser started recording participants’ oral responses. The object remained on the screen until participants clicked the button ‘Send your response’. The microphone continued recording for 500ms after the response was sent to avoid any responses being trimmed by an early button press. After sending their response, participants heard the novel word embedded in the carrier phrase, as in the recognition task, which served as feedback. Then, the next trial began by displaying a new object on the screen. This procedure was repeated until all the object-word pairs were presented (in random order) and repeated in 6 consecutive blocks (96 target trials + 48 fillers).

Measures and statistical analysis

Recognition task. For this task we extracted 1) response accuracy across the 6 blocks. Offline, scores of 0 and 1 were assigned respectively to incorrect and correct responses. We also measured 2) response latencies across blocks. Latencies were measured from the moment the cue-sentence finished playing to the moment participants provided an answer. Only correct answers were included in the latency analysis.

Production task. We measured 1) response accuracy across blocks. As a measure of production accuracy, we computed the Aline distance between each participant’s phonetic production and the target pronunciation of the stimuli with the alineR package (Downey et al., 2008, 2017) in R (R Core Team, 2021). This measure was the result of feature-weighted linguistic distance calculations. Scores resulted in finite values from 0 to 1, where 0 represented perfect production-target match and 1 represented answers containing sound categories unrelated to the target. We also calculated 2) response latencies across blocks, measured from the object presentation until participants orally responded. Furthermore, based on the values of the first (F1) and second (F2) formants, we computed 3) the Euclidean distance between participants’ vowel productions and the stimulus speaker’s pronunciation of the vowel stimuli (henceforth just vowel stimuli), and 4) the Euclidean distance within participants’ vowel contrast productions (/ʌ-æ/ and /i-ɪ/). In addition, to test the two Euclidean distance variables, formant values of participants’ vowel productions (and vowel stimuli) were normalised using the Lobanov method (Lobanov, 1971). This method uses a log-mean method to normalise the formant values and computes a single grand mean for all participants. This was done to prevent participants’ physiological differences from driving the observed effects.

All incorrect responses or that – despite some similarity with the target – clearly pointed at a distractor were excluded from the analyses of 2), 3), and 4). For example, if a participant said [pi:fəl] for the object associated with the novel word [pi:v], they would have obtained an Aline score of 0.31, but their response was considered incorrect and excluded from analyses of latency and the two Euclidean distances because it pointed at the distractor [bi:fəl]. The excluded trials represented 39.58% of the total responses. A total of 2900 trials were kept for analyses: 1559 in NNDS and 1341 in NDS (BF₀₁ = 1.89, anecdotal evidence for H₀).

The dependent variables of the Recognition and Production tasks were independently analysed using growth curve analysis (GCA) models (Mirman, Dixon, et al., 2008; Mirman, Magnuson, et al., 2008) fitted in R (lme4 package; Bates et al., 2015; see Appendix 3 for a list of the models). This technique is explicitly designed to assess changes over time at group and individual levels. GCA allowed us to add to the models the linear and quadratic polynomial terms to account for the overall slope change and the curvature of the observed effects. The linear term reflects the overall slope, and the quadratic term reflects the curvature (i.e., change in slope across learning blocks). Thus, the 6 blocks were added to the model as Block factor, including linear and/or quadratic terms depending on the best model fit. The Register (NNDS and NDS) and Contrast (Single and Goodness contrasts) factors, together with the Block factor, were added to the models as fixed effects (unless otherwise specified). Subject and novel words were included as random effects. Other predictors were considered only if they improved the model fit (see Appendix 3 for a list of the final models). Starting with the minimal structure, various models were created; the final models were chosen according to the best fit indicated by the Performance package in R (Lüdecke et al., 2021). For all models, we set a priori sum contrasts so that within Register, -0.5 was assigned to NDS and + 0.5 to NNDS, whereas within the Contrast factor, -0.5 was assigned to Category Goodness and + 0.5 to Single Category (Schad et al., 2020). Response latencies were transformed using the Box-Cox method (Box & Cox, 1964). Conversely, accuracy of the Recognition task was tested by fitting GCA with generalised linear mixed-effects (glmer) models (binomial family). Given the distribution of the Aline distance, results were fitted in a glmer model of the beta family. Both measures of Euclidean distance were tested in two separate models (one for each contrast: Single and Goodness).

Continuum discrimination. For this task, we used a generalised linear mixed effect model (binomial family) to compare vowel discrimination between the pre-test and the post-test (Exposure factor) and between the two speech register groups. We did not include polynomial terms because GCA did not apply for this variable. Ship/sheep and cup/cap continua were tested in separate models.

For all tasks, model significance was tested with the lmerTest Package (Kuznetsova et al., 2017) and interactions between main effects were explored by running post-hoc analyses in the emmeans package (Lenth et al., 2019) with Tukey HSD correction for multiple comparisons. Given the number of interactions tested in each model, below we report only significant interactions; all results, including non-significant results are reported in the Data Availability.

Acknowledgments

This research was supported by a Doctoral Fellowship (LCF/BQ/DI19/11730045) from “La Caixa” Foundation (ID 100010434) to G.P., and by the Spanish Ministry of Science and Innovation through the Ramon y Cajal Research Fellowship (RYC2018-024284-I) to M.K. This research was supported by the Basque Government through the BERC 2022-2025 program and by the Spanish State Research Agency through BCBL Severo Ochoa excellence accreditation CEX2020-001010-S. The research was also supported by the Spanish Ministry of Economy and Competitiveness (PID2020-113926GB-I00 to C.D.M.), and the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 819093 to C.D.M.).

Author contributions

G.P.: Conceptualisation, Formal analysis, Investigation, Data Curation, Visualisation, Writing - Original Draft; M.K.: Conceptualisation, Supervision, Writing - Original Draft, Writing: Reviewing and Editing, Funding acquisition; C.D.M.: Conceptualisation, Supervision, Writing - Original Draft , Writing: Reviewing and Editing, Funding acquisition.

Data Availability

Material, data, experiment script, analysis code, and non-significant results can be found at https://osf.io/xtky5/?view_only=4ec02c26bd084296b088780811ebbb07. A list of the material, supplementary images, and statistical formula can be found in Appendix 1, 2 and 3.

Competing Interests Statement

The authors declare that they have no known competing financial interests or personal relationships that would influence the work reported in this paper.

Anwyl-Irvine, A., Dalmaijer, E. S., Hodges, N., & Evershed, J. K. (2020). Realistic precision and accuracy of online experiment platforms, web browsers, and devices. Behavior Research Methods. https://doi.org/10.3758/s13428-020-01501-5
Aoyama, K., & Flege, J. E. (2011). Effects of L2 experience on perception of English /r/ and /l/ by native Japanese speakers. Journal of the Phonetic Society of Japan, 15 (3), 5–13. https://doi.org/10.24467/ONSEIKENKYU.15.3_5
Atkinson, D. (2002). Toward a Sociocognitive Approach to Second Language Acquisition. The Modern Language Journal, 86(4), Article 4.
Atkinson, D. (2014). Language Learning in Mindbodyworld: A Sociocognitive Approach to Second Language Acquisition. Language Teaching, 47(4), Article 4. https://doi.org/10.1017/S0261444813000153
Atkinson, D., Churchill, E., Nishino, T., & Okada, H. (2007). Alignment and Interaction in a Sociocognitive Approach to Second Language Acquisition. The Modern Language Journal, 91(2), Article 2. https://doi.org/10.1111/j.1540-4781.2007.00539.x
Baigorri, M., Campanelli, L., & Levy, E. S. (2019). Perception of American–English Vowels by Early and Late Spanish–English Bilinguals. Language and Speech, 62(4), Article 4. https://doi.org/10.1177/0023830918806933
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software, 67(1), Article 1. https://doi.org/10.18637/jss.v067.i01
Best, C. T. (1991). The Emergence of Native-Language Phonological Influences in Infants: A Perceptual Assimilation Model. Haskins Laboratories Status & Speech Research, SR·107/108, 1–30.
Best, C. T., & Tyler, M. D. (2007). Nonnative and second-language speech perception: Commonalities and complementarities. Language Experience in Second Language Speech Learning: In Honor of James Emil Flege. https://doi.org/10.1121/1.1332378.
Bobb, S. C., Mello, K., Turco, E., Lemes, L., Fernandez, E., & Rothermich, K. (2019a). Second Language Learners’ Listener Impressions of Foreigner-Directed Speech. Journal of Speech, Language, and Hearing Research: JSLHR, 62(9), 3135–3148. https://doi.org/10.1044/2019_JSLHR-S-18-0392
Bobb, S. C., Mello, K., Turco, E., Lemes, L., Fernandez, E., & Rothermich, K. (2019b). Second Language Learners’ Listener Impressions of Foreigner-Directed Speech. Journal of Speech, Language, and Hearing Research: JSLHR, 62(9), Article 9. https://doi.org/10.1044/2019_JSLHR-S-18-0392
Boomershine, A. (2013). The Perception of English Vowels by Monolingual, Bilingual, and Heritage Speakers of Spanish and English. In C. Howe, S. E. Blackwell, & M. L. Quesada (Eds.), Selected Proceedings of the 15th Hispanic Linguistics Symposium (pp. 103–118). Cascadilla Proceedings Project. http://www.lingref.com/cpp/hls/15/abstract2879.html
Box, G. E. P., & Cox, D. R. (1964). An Analysis of Transformations. Journal of the Royal Statistical Society. Series B (Methodological), Vol. 26(N° 2), Article N° 2.
Bridges, D., Pitiot, A., MacAskill, M. R., & Peirce, J. W. (2020). The timing mega-study: Comparing a range of experiment generators, both lab-based and online. PeerJ, 8, e9414. https://doi.org/10.7717/peerj.9414
Childers, J. B., & Tomasello, M. (2002). Two-year-olds learn novel nouns, verbs, and conventional actions from massed or distributed exposures. Developmental Psychology, 38(6), Article 6. https://doi.org/10.1037/0012-1649.38.6.967
Cook, V. (2016). Where Is the Native Speaker Now? TESOL Quarterly, 50(1), Article 1.
Derwing, T. M., & Munro, M. J. (2009). Putting accent in its place: Rethinking obstacles to communication. Language Teaching, 42(4), Article 4. https://doi.org/10.1017/S026144480800551X
Drozdova, O. A., Zamyatina, E. V., Volodina, D. N., Zakharova, E. O., Ruchina, A. V., & Nepryakhin, A. F. (2015). Situational Communication in Teaching Russian as a Foreign Language to Beginner Learners. Procedia - Social and Behavioral Sciences, 215, 118–126. https://doi.org/10.1016/j.sbspro.2015.11.584
Drozdova, P., Hout, R. V., & Scharenborg, O. (2016). Lexically-guided perceptual learning in non-native listening*. Bilingualism: Language and Cognition, 19(5), 914–920. https://doi.org/10.1017/S136672891600002X
Escudero, P. R. (2005). Linguistic perception and second language acquisition: Explaining the attainment of optimal phonological categorization. LOT.
Escudero, P. R. (2006). The phonological and phonetic development of new vowel contrasts in Spanish learners of English. English with a Latin Beat, 41–55.
Escudero, P. R. (2001). The Role of the Input in the Development of L1 and L2 Sound Contrasts: Language-Specific Cue Weighting for Vowels. In Proceedings of the 25th Annual Boston University Conference on Language Development, Vols. 1–2 (pp. 250–261). Cascadilla Press.
Fairs, A., & Strijkers, K. (2021). Can we use the internet to study speech production? Yes we can! Evidence contrasting online versus laboratory naming latencies and errors. PsyArXiv. https://doi.org/10.31234/osf.io/2bu4c
Ferguson, S. H., & Kewley-Port, D. (2002). Vowel intelligibility in clear and conversational speech for normal-hearing and hearing-impaired listeners. J. Acoust. Soc. Am., 112(1), 13.
Flege, J. E. (1984). The detection of French accent by American listeners. The Journal of the Acoustical Society of America, 76(3), Article 3. https://doi.org/10.1121/1.391256
Flege, J. E. (1995). Second Language Speech Learning. Theory, Findings, and Problems. In Winifred Strange (CditOf), Speech Perception and Linguistic Experience: Issues in Cross-Language Research. Timonium.
Flege, J. E., & Bohn, O.-S. (2021). The revised Speech Learning Model. Second Language Speech Learning: Theoretical and Empirical Progress,. https://doi.org/10(9781108886901.002).
Flege, J. E., Bohn, O.-S., & Jang, S. (1997). Effects of experience on non-native speakers’ production and perception of English vowels. Journal of Phonetics, 25(4), Article 4. https://doi.org/10.1006/jpho.1997.0052
Flege, J. E., & Liu, S. (2001). THE EFFECT OF EXPERIENCE ON ADULTS’ ACQUISITION OF A SECONDLANGUAGE. Studies in Second Language Acquisition, 23(4), Article 4. https://doi.org/10.1017/S0272263101004041
Foley, C., & Flynn, S. (2013). The role of the native language. In J. Herschensohn & M. Young-Scholten (Eds.), The Cambridge Handbook of Second Language Acquisition (pp. 97–113). Cambridge University Press. https://doi.org/10.1017/CBO9781139051729.008
Garnier, M., & Henrich, N. (2014). Speaking in noise: How does the Lombard effect improve acoustic contrasts between speech and ambient noise? Computer Speech & Language, 28(2), 580–597. https://doi.org/10.1016/j.csl.2013.07.005
Gershkoff-Stowe, L., & Hahn, E. R. (2013). Word comprehension and production asymmetries in children and adults. Journal of Experimental Child Psychology, 114(4), Article 4. https://doi.org/10.1016/j.jecp.2012.11.005
Giles, H. (2016). Communication Accommodation Theory. In The International Encyclopedia of Communication Theory and Philosophy (pp. 1–7). American Cancer Society. https://doi.org/10.1002/9781118766804.wbiect056
Golinkoff, R. M., & Alioto, A. (1995). Infant-directed speech facilitates lexical learning in adults hearing Chinese: Implications for language acquisition*. Journal of Child Language, 22(3), 703–726. https://doi.org/10.1017/S0305000900010011
Grimaldi, M., Sisinni, B., Gili Fivela, B., Invitto, S., Resta, D., Alku, P., & Brattico, E. (2014). Assimilation of L2 vowels to L1 phonemes governs L2 learning in adulthood: A behavioral and ERP study. Frontiers in Human Neuroscience, 8. https://www.frontiersin.org/articles/10.3389/fnhum.2014.00279
Hazan, V., Uther, M., & Granlund, S. (2015). How does foreigner-directed speech differ from other forms of listener-directed clear speaking styles? ICPhS 2015.
Hendriks, P., & Koster, C. (2010). Production/comprehension asymmetries in language acquisition. Lingua, 120(8), Article 8. https://doi.org/10.1016/j.lingua.2010.02.002
Iverson, P., Ekanayake, D., Hamann, S., Sennema, A., & Evans, B. (2008). Category and Perceptual Interference in Second-Language Phoneme Learning: An Examination of English /w/-/v / Learning by Sinhala. Journal of Experimental Psychology. Human Perception and Performance, 34, 1305–1316. https://doi.org/10.1037/0096-1523.34.5.1305
Kangatharan, J. (2015). The role of vowel hyperarticulation in clear speech to foreigners and infants [Brunel University London]. https://bura.brunel.ac.uk/bitstream/2438/10519/1/FulltextThesis.pdf
Knoll, M. A., Scharrer, L., & Costall, A. (2009). Are actresses better simulators than female students? The effects of simulation on prosodic modifications of infant- and foreigner-directed speech. Speech Communication, 51(3), 296–305. https://doi.org/10.1016/j.specom.2008.10.001
Kramsch, C. (2007). Re-reading Robert Lado, 1957, Linguistics across Cultures. Applied linguistics for language teachers. International Journal of Applied Linguistics, 17(2), Article 2. https://doi.org/10.1111/j.1473-4192.2007.00149.x
Krashen, S. (1989). We Acquire Vocabulary and Spelling by Reading: Additional Evidence for the Input Hypothesis. The Modern Language Journal, 73(4), Article 4. https://doi.org/10.1111/j.1540-4781.1989.tb05325.x
Krashen, S. D. (1982). Principles and practice in second language acquisition (1st ed). Pergamon.
Krashen, S., D. (1985). The input hypothesis: Issues and implications. Addison-Wesley Longman Limited.
Kuhl, P. K. (1997). Cross-Language Analysis of Phonetic Units in Language Addressed to Infants. Science, 277(5326), 684–686. https://doi.org/10.1126/science.277.5326.684
Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest Package: Tests in Linear Mixed Effects Models. Journal of Statistical Software, 82, 1–26. https://doi.org/10.18637/jss.v082.i13
Lee, J., Jang, J., & Plonsky, L. (2015). The Effectiveness of Second Language Pronunciation Instruction: A Meta-Analysis. Applied Linguistics, 36(3), Article 3. https://doi.org/10.1093/applin/amu040
Lenth, R., Singmann, H., Love, J., Buerkner, P., & Herve, M. (2019). Package ‘emmeans’. Package ‘emmeans’. https://github.com/rvlenth/emmeans
Lindblom, B. (1990). Explaining Phonetic Variation: A Sketch of the H&H Theory. In W. J. Hardcastle & A. Marchal (Eds.), Speech Production and Speech Modelling (pp. 403–439). Springer Netherlands. https://doi.org/10.1007/978-94-009-2037-8_16
Lindblom, B. (2009). On the Communication Process: Speaker-Listener Interaction and the Development of Speech. Augmentative and Alternative Communication, 6, 220–230. https://doi.org/10.1080/07434619012331275504
Lobanov, B. M. (1971). Classification of Russian vowels spoken by different listeners. Journal of the Acoustical Society of America, 49:606–08.
Logan, J. S., Lively, S. E., & Pisoni, D. B. (1991). Training Japanese listeners to identify English /r/ and /l/: A first report. The Journal of the Acoustical Society of America, 89(2), Article 2. https://doi.org/10.1121/1.1894649
Lombard, E. (1911). Le signe de l’élévation de la voix. Annales Des Maladies de L’Oreille et Du Larynx., 37, 101–119.
Lüdecke, D., Ben-Shachar, M. S., Patil, I., Waggoner, P., & Makowski, D. (2021). performance: An R Package for Assessment, Comparison and Testing of Statistical Models. Journal of Open Source Software, 6(60), Article 60. https://doi.org/10.21105/joss.03139
Ma, W., Fiveash, A., Margulis, E. H., Behrend, D., & Thompson, W. F. (2020). Song and infant-directed speech facilitate word learning. Quarterly Journal of Experimental Psychology (2006), 73(7), 1036–1054. https://doi.org/10.1177/1747021819888982
Melnik-Leroy, G. A., Turnbull, R., & Peperkamp, S. (2022). On the relationship between perception and production of L2 sounds: Evidence from Anglophones’ processing of the French /u/–/y/ contrast. Second Language Research, 38(3), Article 3. https://doi.org/10.1177/0267658320988061
Mirman, D., Dixon, J. A., & Magnuson, J. S. (2008). Statistical and computational models of the visual world paradigm: Growth curves and individual differences. Journal of Memory and Language, 59(4), Article 4. https://doi.org/10.1016/j.jml.2007.11.006
Mirman, D., Magnuson, J. S., Estes, K. G., & Dixon, J. A. (2008). The link between statistical segmentation and word learning in adults. Cognition, 108(1), Article 1. https://doi.org/10.1016/j.cognition.2008.02.003
Mora, J. C., Ortega, M., Mora-Plaza, I., & Aliaga-García, C. (2022). Training the pronunciation of L2 vowels under different conditions: The use of non-lexical materials and masking noise. Phonetica, 79(1), Article 1. https://doi.org/10.1515/phon-2022-2018
O’Brien, M. G. (2021). Ease and Difficulty in L2 Pronunciation Teaching: A Mini-Review. Frontiers in Communication, 5. https://www.frontiersin.org/articles/10.3389/fcomm.2020.626985
Peng, G., Zheng, H.-Y., Gong, T., Yang, R.-X., Kong, J.-P., & Wang, W. S.-Y. (2010). The influence of language experience on categorical perception of pitch contours. Journal of Phonetics, 38(4), Article 4. https://doi.org/10.1016/j.wocn.2010.09.003
Piazza, G., Kalashnikova, M., Fernández-Merino, L., & Martin, C. (2023). Speakers’ communicative intentions lead to acoustic adjustments in native and non-native directed speech. PsyArXiv. https://doi.org/10.31234/osf.io/kz72c
Piazza, G., Kartushina, N., Flege, J. E., & Martin, C. D. (2022). Comparison of Acoustic Features in Speech Production Studies Run Online and in the Lab. 27, 232. https://cdn.ymaws.com/www.psychonomic.org/resource/resmgr/annual_meeting/2022_meeting/ps22_abstract__book_10.27.22.pdf
Piazza, G., Martin, C. D., & Kalashnikova, M. (2022). The Acoustic Features and Didactic Function of Foreigner-Directed Speech: A Scoping Review. Journal of Speech, Language, and Hearing Research. https://doi.org/10.1044/2022_JSLHR-21-00609
Rallo Fabra, L., & Romero, J. (2012). Native Catalan learners’ perception and production of English vowels. Journal of Phonetics, 40(3), Article 3. https://doi.org/10.1016/j.wocn.2012.01.001
Reinisch, E., Weber, A., & Mitterer, H. (2013). Listeners retune phoneme categories across languages. Journal of Experimental Psychology: Human Perception and Performance, 39, 75–86. https://doi.org/10.1037/a0027979
Rothermich, K., Harris, H. L., Sewell, K., & Bobb, S. C. (2019). Listener impressions of foreigner-directed speech: A systematic review. Speech Communication, 112, 22–29. https://doi.org/10.1016/j.specom.2019.07.002
Rothman, J., Bayram, F., DeLuca, V., Pisa, G. D., Duñabeitia, J. A., Gharibi, K., Hao, J., Kolb, N., Kubota, M., Kupisch, T., Laméris, T., Luque, A., Osch, B. van, Soares, S. M. P., Prystauka, Y., Tat, D., Tomić, A., Voits, T., & Wulff, S. (2023). Monolingual comparative normativity in bilingualism research is out of “control”: Arguments and alternatives. Applied Psycholinguistics, 44(3), Article 3. https://doi.org/10.1017/S0142716422000315
Scarborough, R., Dmitrieva, O., Hall-Lew, L., Zhao, Y., & Brenier, J. (2007). An acoustic study of real and imagined foreigner‐directed speech. The Journal of the Acoustical Society of America, 121(5), Article 5. https://doi.org/10.1121/1.4781735
Schad, D. J., Vasishth, S., Hohenstein, S., & Kliegl, R. (2020). How to capitalize on a priori contrasts in linear (mixed) models: A tutorial. Journal of Memory and Language, 110, 104038. https://doi.org/10.1016/j.jml.2019.104038
Smiljanić, R., & Bradlow, A. R. (2005). Production and perception of clear speech in Croatian and English. The Journal of the Acoustical Society of America, 118(3), 1677–1688. https://doi.org/10.1121/1.2000788
Smiljanić, R., & Bradlow, A. R. (2009). Speaking and Hearing Clearly: Talker and Listener Factors in Speaking Style Changes. Language and Linguistics Compass, 3(1), 236–264. https://doi.org/10.1111/j.1749-818X.2008.00112.x
Tremblay, R. E., Schaal, B., Boulerice, B., Arseneault, L., Soussignan, R. G., Paquette, D., & Laurent, D. (1998). Testosterone, Physical Aggression, Dominance, and Physical Development in Early Adolescence. International Journal of Behavioral Development, 22(4), Article 4. https://doi.org/10.1080/016502598384153
Uther, M., Knoll, M. A., & Burnham, D. (2007). Do you speak E-NG-L-I-SH? A comparison of foreigner- and infant-directed speech. Speech Communication, 49(1), Article 1. https://doi.org/10.1016/j.specom.2006.10.003
Van Leussen, J.-W., & Escudero, P. (2015). Learning to perceive and recognize a second language: The L2LP model revised. Frontiers in Psychology, 6. https://www.frontiersin.org/articles/10.3389/fpsyg.2015.01000
Vogt, A., Hauber, R., Kuhlen, A. K., & Rahman, R. A. (2021). Internet-based language production research with overt articulation: Proof of concept, challenges, and practical advice. Behavior Research Methods. https://doi.org/10.3758/s13428-021-01686-3
Wang, W. (1999). Age and Second Language Acquisition in Adulthood: The Learning Experiences and Perceptions of Women Immigrants. TESL Canada Journal, 01–19. https://doi.org/10.18806/tesl.v16i2.715
Woods, K. J. P., Siegel, M. H., Traer, J., & McDermott, J. H. (2017). Headphone screening to facilitate web-based auditory experiments. Attention, Perception, & Psychophysics, 79(7), Article 7. https://doi.org/10.3758/s13414-017-1361-2
Zehr, J., & Schwarz, F. (2018). PennController for Internet Based Experiments (IBEX). https://doi.org/10.17605/OSF.IO/MD832

No competing interests reported.

SupplementaryMaterialfexpo.docx

Download PDF

Journal Publication

published 02 Dec, 2023

Read the published version in Scientific Reports →

Editorial decision: Major revision
30 Aug, 2023
Reviews received at journal
25 Aug, 2023
Reviewers agreed at journal
14 Aug, 2023
Reviewers agreed at journal
14 Aug, 2023
Reviewers invited by journal
14 Aug, 2023
Editor assigned by journal
14 Aug, 2023
Editor invited by journal
10 Aug, 2023
Submission checks completed at journal
10 Aug, 2023
First submitted to journal
07 Aug, 2023

You are reading this latest preprint version

Phonetic Accommodation in Non-native Directed Speech Supports L2 Word Learning and Pronunciation

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

From high clarity to the didactic impact of NNDS

Aspects of auditory L2 word learning

The present study

Results

Recognition task

INSERT FIGURE 1 ABOUT HERE

INSTERT FIGURE 2 ABOUT HERE

Production task

INSTERT FIGURE ABOUT HERE

INSERT FIGURE 4 ABOUT HERE

INSERT FIGURE 5 ABOUT HERE

INSERT FIGURE 6 ABOUT HERE

3.4.3. Continuum discrimination task

INSERT FIGURE 7 ABOUT HERE

Discussion

NNDS Benefits

NNDS benefit depends on properties of the speech contrasts to be learned

Theories of Second Language Acquisition that explain the NNDS benefit

NNDS benefit depends on the modality and task demand

NNDS does not induce changes in L2 sound phonetic boundaries (after short training)

Method

Participants

Material

Procedure

Measures and statistical analysis

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1