Language Learning Roundtable*
5 September 2012
Interdisciplinary perspectives on the acquisition of second language phonology
Catherine T. Best & Michael D. Tyler (University of Western Sydney, Australia; Haskins Laboratories, USA)
Contributions of L1 perceptual assimilation to L2 speech perception and perceptual learning
Grzegorz Dogil (University of Stuttgart, Germany)
How does language learning influence brain activity: Real-time fMRI study of processing of prosody
Wander Lowie (University of Groningen, Netherlands)
Acquiring a second language sound system: a developmental perspective
Ineke Mennen (Bangor University, Wales, UK)
Beyond segments: a look at new or neglected approaches in L2 speech research
Ocke-Schwen Bohn (Arhus University, Denmark)
Katarzyna Dziubalska-Kołaczyk (Adam Mickiewicz University in Poznań)
Magdalena Wrembel (Adam Mickiewicz University in Poznań)
* The event will be held due to a grant by the Language Learning journal.
Catherine T. Best and Michael D. Tyler
Contributions of L1 perceptual assimilation to L2 speech perception and perceptual learning
Language experience systematically “tunes up” our perception of speech. Perceptual attunement to the native language (L1) is evident quite early, well before the end of the infant’s first year, and continues to be refined during the preschool years and into childhood. The primary benefit of this attunement is that it optimizes the speed and accuracy of word recognition in the native language. However, it also has a well-known drawback: L1 attunement places severe limits on perceptual sensitivity to the consonant and vowel contrasts of other languages that deviate phonologically and/or phonetically from native segments and contrasts. These perceptual constraints are especially obvious in adults who are completely naïve to the language of the target stimuli. However, L1 perceptual biases are also striking in the perception of many non-native L2 segmental contrasts by late-onset (post-puberty) second language (L2) learners, whose speech perception and comprehension are hindered not only during the initial phase of L2 acquisition but even after years of L2 use. Moreover, the L1 perceptual bias appears to contribute substantially to L1-accented production of L2 words, even in speakers with years of L2 experience. At the same time, however, it is important to note that certain non-native contrasts nonetheless remain quite easy for naïve and late-L2 listeners to perceive and/or learn.
Our presentation will address theories and empirical findings on perceptual learning of L2 speech contrasts, particularly with respect to the wide variations in degree of difficulty for perceiving and learning different types of non-native contrasts. We will begin by considering how research on non-native speech perception by completely naïve listeners may illuminate phonetic and phonological aspects of speech perception at the beginning of late L2 learning. One common assumption has been that naïve adults’ perception of non-native phonemic contrasts can directly account for the relative difficulty or ease that late L2 learners have with specific segments and contrasts in their adopted language. Another common assumption has been that developmental changes in non-native speech perception underlie the observed onset-age differences that have been reported in the literature on the acquisition of native-like speech perception and production in an L2. However, as we have argued previously (Best & Tyler, 2007), evaluation of these assumptions must take into account that models of non-native speech perception such as the Perceptual Assimilation Model (PAM) have focused primarily on naïve listeners’ perception of unfamiliar speech contrasts, whereas models of L2 speech acquisition such as the Speech Learning Model (SLM) have focused on experienced listeners who vary in onset age and/or level of exposure to those contrasts.
Our presentation at EUROSLA 2012 will build from that foundation to further probe the factors that may influence naïve listeners and L2 learners in similar versus different ways. We will review and discuss recent empirical findings from our own and others’ laboratories, in order to critically assess and expand upon our earlier hypotheses (Best & Tyler, 2007) regarding the commonalities and complementarities between inexperienced listeners and those learning an L2, as viewed from the most-frequently cited models of experiential effects on speech perception: PAM and SLM. Among the issues we will address are:
• How active L2 learning may affect perception of phonetic versus phonological information;
• How early- versus vs. late-onset L2 acquisition may impact L2 speech perception:
• How L2 vocabulary development may interact with perceptual adjustment in L2 speech perception;
• What these considerations imply for optimal timing and conditions for perceptual “re-tuning” to L2 speech contrasts and maximal skill development for recognition of spoken words in the L2.
How does language learning influence brain activity: Real-time fMRI study of processing of prosody
Brain and language have lived in a symbiosis for thousands of years. However, mechanisms of cortical reorganization that underlie enhancement of language and speech processing have been poorly investigated. Recently, methods (real time functional magnetic resonance imaging rt-fMRI) and devices (brain computer interfaces) have been developed that actually allow us the insight into the language processing brain in vivo and at real time. Here, we addressed changes in functional connectivity induced in subjects who learnt to deliberately increase activation in the right inferior frontal gyrus (rIFG) and improved their ability to identify emotional intonations. At the beginning of their training process, we observed a massive connectivity of this region to a widespread network of temporo-parietal areas, which decreased and lateralized to the right hemisphere with practice. Volitional control of activation strengthened connectivity of the rIFG to the right PFC whereas practice increased its connectivity to pre-central gyri bilaterally. These findings suggest that the mechanisms that underlie enhancement of speech processing cannot fully be described in terms of activation patterns but should take into consideration functional connectivity of brain networks.
We have also observed some interesting individual differences among the speakers that we investigated. In particular, we were interested in the neural patterns that are used by so called “phonetically talented” subjects. In the presentation we will discuss the details of our conceptualization of talent. We will present all the factors which seem to determine phonetic talent and we will have a look of its neural underpinnings. The functional brain connections used by talented speakers differ qualitatively and quantitatively from the neural networks used by average language learners. Real time brain imaging coupled with intelligent brain computer interfaces gives us, for the first time, the possibility to attempt a bio-feed-back solution directly at the neural level and in this way “learn from the gifted”. Implications of these findings for language learning and language teaching will be discussed in the presentation.
Acquiring a second language sound system: A developmental perspective
The acquisition of a second language sound system has often been regarded as the result of a linear relationship between several factors, like the learner’s age, the L2 sound system, the motivation to learn, the phonological aptitude, and the learning context. In these views, acquisition is usually seen as the development from strong L1-like pronunciation to native-like pronunciation, during which most learners stop learning somewhere along the way. Research that has investigated the impact of these individual factors on the acquisition of L2 sound systems has yielded contradictory results about, for instance, the influence of perception on production and the influence of age related factors on L2 sound development.
The factor that is most frequently mentioned in relation to the development of the L2 phonological system is ‘age’. The general observation is that young starters of second language learning stand a greater chance of attaining a native-like pronunciation than older starters. Whether or not this is due to a critical period for language acquisition is a matter of debate. While some scholars have argued that ultimate L2 attainment is impossible for learners who have started learning the second language at a late age (Scovel, 1988), others have claimed that even late learners can reach the highest levels of L2 proficiency in certain aspects of the language system, provided that the learning situation is ideal (in terms of input, language instruction, motivation, etc.) (Bongaerts, 1999). Fact is that numerous studies have reported that the age of onset of L2 acquisition is the most successful predictor of the learner’s ultimate attainment, most obviously when it concerns the acquisition of the L2 sound system. To account for the advantage for younger starters, several studies have associated this observation with the influence of the learner’s native language. While the role of crosslinguistic influence in syntax and semantics is subject to debate, it is generally recognized that learners transfer properties of their first language’s sound system into the foreign language. Earlier research has shown that the role of crosslinguistic influence on L2 sound systems correlates with the age of acquisition (AoA): the lower the AoA, the less important crosslinguistic influence is. The most familiar models accounting for the acquisition of L2 sound systems (Best’s (1994) Perceptual assimilation Model, Kuhl’s (1992) Native Language Magnet Model and Flege’s (1995) Speech Learning Model) have all related age to the influence of the native language sound system. All of these models address the issue of categorical perception and the mechanisms of creating language specific phoneme boundaries. With the coming of age, it is argued, it becomes increasingly difficult to reset phoneme boundaries. Although this account provides a strong explanation for the way in which native speakers acquire the sound system of their first language, is also raises several questions that are left unanswered. While some researchers have shown that phonemic vowel categories have become established at around 6 months (Kuhl, 1986), it has also been argued that establishment of phonemic categories is not complete until young adulthood (Flege, 1995). These seemingly contradictory views cannot be overcome by traditional product-oriented research.
Recent studies in other areas of L2 research have clearly demonstrated that longitudinal, variability-based approaches can reveal remarkable facts about the learning process. These studies have used a Dynamic Systems Theory (DST) approach to language development, and have focused on the development of L2 writing and on L2 vocabulary (Verspoor, Lowie & van Dijk, 2007; Caspi, 2010; Verspoor & Spoelman, 2011S) and L2 sound systems (Lowie, 2011). A number of longitudinal case studies have shown that a DST-approach can contribute to our understanding of the dynamic interaction of factors affecting L2 phonology in important ways. For instance, longitudinal analyses of perception and production data of both children and adults show that the perception and production affect each other differently at different moments in time, and in interaction with the situational context. This means that for the individual language learner perception does not straightforwardly precede production. These studies also point to the large amount of exposure that is required for the sound system to develop or change.
In this contribution, I will present an alternative to the traditional product-oriented view of the development of L2 sound systems. Referring to data from children and adults, I will demonstrate how a nonlinear, process-focused approach is able to account for learner data by truly appreciating the complex dynamic interaction of the forces shaping the L2 sound system.
Beyond segments: A look at new or neglected approaches in L2 speech research
The field of L2 speech learning has gained a fairly good understanding of segmental aspects of language differentiation. For example, we know that languages may differ in their phoneme inventories or phonological features, but also in the phonetic detail of segments. We also know that it is this level of phonetic detail of cross-language differences that is most important in predicting which sounds will pose difficulties in L2 production and perception of segments (Strange 2007). Our understanding of the exact nature of cross-language phonetic similarity/dissimilarity is, however, still rather limited as it has only been examined in very restricted phonetic/phonotactic and prosodic contexts and findings cannot be generalized beyond these (Strange 2007). Current theories of L2 speech learning (e.g. Flege’s Speech Learning Model [Flege 1995, 2003] and Best’s Perceptual Assimilation Model [Best 1995, Best and Tyler, 2007]) rely on this poorly understood concept of phonetic similarity between native language (L1) and L2 speech to predict the relative difficulty/ease of production and perception of non-native speech.
Although the above models have been in existence for almost two decades, they have not yet attempted to account for any non-segmental aspects of L2 speech learning, and there are only very few studies that address aspects of L2 speech learning that go beyond segments. The focus of my talk will be on two neglected aspects in L2 speech research: intonation and articulatory settings. There are several reasons for the apparent gap in empirical research in these two areas: the fact that (i) they are difficult to separate from other influences, in particular segmental ones; (ii) until relatively recently it was almost impossible to measure suprasegmental features in real time or provide direct access to the articulators and current technological advances have not been exploited fully; (iii) our understanding of intonational primitives and of articulation and its relation to acoustics is still rudimentary; and (iv) we are faced with the challenge of considerable between-speaker variability which may mask crosslanguage differences.
My talk will give an overview of L2 research in these two areas. It will discuss some of the theoretical and practical problems that arise in analysing L2 intonation and articulatory settings and it will demonstrate some of the theoretical and technological advances that have been made in these areas that have made the study of these issues more feasible. In my talk, I will be drawing on some of the existing research as well as on various recent research projects to illustrate these problems and show examples of what is currently possible. Finally, I will be discussing some ways to take research into these area further forward.
Discussion of contributions to the 2012 Roundtable on “Interdisciplinary perspectives on the acquisition of second language phonology”
The contributors to this Roundtable highlight conspicuous gaps in our knowledge of second language speech, and they present current and future research activities which address hot topics in our field.
The presentations by Dogil and Mennen show most clearly how methodological and technological advances now make it possible to address questions which, until quite recently, most colleagues shunned away from because they were deemed “too difficult” (given the tools at our disposal), “too speculative”, or both. One of these topics consists of issues surrounding language learning “talent”. Previous studies have shown that even if the most important variables in second language speech learning, age of learning and language use patterns, are accounted for, a fairly large amount of variation remains unaccounted for (e.g., Flege et al. 1995). The research presented by Dogil aims to account for important causes of individual variation among learners whose learning environments and conditions are very similar, and it provides exciting perspectives on improving instructed second language learning. The Dynamic Systems approach advocated by Lowie also holds the promise of a better understanding of individual variation in speech learning abilities. While surely everyone in our field would agree that longitudinal studies of speech learning are highly desirable, the very vast majority of L2 speech research has been (and probably will be) based on cross-sectional research for purely research-pragmatic reasons (i.e., time & money). Lowie presents longitudinal data which are needed to address many open questions in our field, among these the complex relation of speech production and perception at various stages in L2 speech learning (e.g., Bohn & Flege 1997).
Mennen’s presentation is probably the most courageous one because she focuses on two topics that others have stayed away from for very different but perhaps understandable reasons. The reason why the most important models of L2 speech learning, Flege’s SLM (1995) and Best’s PAM/PAM-L2 (Best 1995, Best & Tyler 2007) deal with segments rather than segments plus suprasegmentals is not that segments are easy, but that segments are relatively easy to describe, to analyze, and to test in speech learning compared to anything happening above the level of the segment. Mennen shows that technological advances make this important but much neglected aspect of L2 speech learning now more accessible for research. These advances make it also possible to investigate the topic of “articulatory setting” in a professional manner. This topic has a fairly long history of largely fruitless, impressionistic speculation in foreign language speech pedagogy, while speech scientists have stayed away from it until quite recently (e.g., Gick et al. 2004). Both topics, intonation and articulatory settings, open up very much needed perspectives on a complete view of second language speech learning beyond the segment.
In equally important step beyond the segmental speech level is presented by Best & Tyler, whose PAM-L2 (Best & Tyler 2007) is the first theoretically based attempt in L2 speech research to link speech development to lexical development. When one considers the by now very large body of research on L2 speech, one could easily get the impression that the ultimate aim of L2 speech learning is, well, to learn L2 speech sounds. Best & Tyler’s presentation is a highly appropriate reminder that we communicate in words (and sentences), of which phonetic segments are “merely” the building blocks, and that at least parts of L2 speech learning are highly likely to be lexically driven. Their model provides interesting suggestions for examining how vocabulary development interacts with L2 speech learning.