Neural signatures of phonetic learning in adulthood: A magnetoencephalography study

Abstract

The present study used magnetoencephalography (MEG) to examine perceptual learning of American English /r/ and /l/ categories by Japanese adults who had limited English exposure. A training software program was developed based on the principles of infant phonetic learning, featuring systematic acoustic exaggeration, multi-talker variability, visible articulation, and adaptive listening. The program was designed to help Japanese listeners utilize an acoustic dimension relevant for phonemic categorization of /r–l/ in English. Although training did not produce native-like phonetic boundary along the /r–l/ synthetic continuum in the second language learners, success was seen in highly significant identification improvement over twelve training sessions and transfer of learning to novel stimuli. Consistent with behavioral results, pre–post MEG measures showed not only enhanced neural sensitivity to the /r–l/ distinction in the left-hemisphere mismatch field (MMF) response but also bilateral decreases in equivalent current dipole (ECD) cluster and duration measures for stimulus coding in the inferior parietal region. The learning-induced increases in neural sensitivity and efficiency were also found in distributed source analysis using Minimum Current Estimates (MCE). Furthermore, the pre–post changes exhibited significant brain-behavior correlations between speech discrimination scores and MMF amplitudes as well as between the behavioral scores and ECD measures of neural efficiency. Together, the data provide corroborating evidence that substantial neural plasticity for second-language learning in adulthood can be induced with adaptive and enriched linguistic exposure. Like the MMF, the ECD cluster and duration measures are sensitive neural markers of phonetic learning.

Introduction

A fundamental question in cognitive neuroscience is the degree of neural plasticity as a function of age and experience. Classic studies and arguments on the putative “critical” or “sensitive” period for language acquisition highlight the superiority of learning a second language prior to puberty, and data support both maturation and experience as mechanistic explanations for the effect (Flege et al., 1999, Hernandez and Li, 2007, Johnson and Newport, 1989, Kuhl et al., 2008, Lenneberg, 1967, Mayberry and Lock, 2003). In the phonetic domain, there is clear evidence that early language learning does not involve a permanent loss of perceptual sensitivity to all the nonnative distinctions (Best et al., 2001, Werker and Tees, 2005). Furthermore, adults' perception of nonnative speech can be improved by using a variety of short-term intensive training methods (Akahane-Yamada et al., 1997, Bradlow et al., 1999, Hazan et al., 2006, Iverson et al., 2005, Jamieson and Morosan, 1986, Logan et al., 1991, McCandliss et al., 2002, Pruitt et al., 2006, Strange and Dittmann, 1984, Tremblay et al., 1997, Wang et al., 2003, Zhang et al., 2000). These training studies, among others, have not only provided important empirical data for reevaluating the “critical period” hypothesis but also revealed key factors that facilitate second language learning independent of age. However, as epitomized by the classic problem of the /r–l/ phonemic contrast for adult Japanese speakers, neither intensive training nor prolonged naturalistic exposure has led to native-like mastery (Callan et al., 2003, Iverson et al., 2005, McCandliss et al., 2002, Takagi, 2002, Takagi and Mann, 1995). The experiential mechanisms that enhance or limit neural plasticity in adulthood are not well understood.

In our language acquisition model, adults' difficulty with nonnative languages stems from an early strong neural commitment to the statistical and spectral patterns in the language input during infancy (Kuhl et al., 2008). The effects of native language neural commitment (NLNC) are self-reinforcing and bidirectional — it enhances the detection of higher-order linguistic patterns, such as words, that utilize learned phonetic patterns, while at the same time hindering the detection of non-conforming patterns contained in foreign languages, as shown behaviorally (Iverson et al., 2003) and neurally (Zhang et al., 2005). We further theorize that second language acquisition in adulthood can be improved by manipulating the language input to incorporate the basic principles underlying infants' acquisition of the sound patterns of their native language (Kuhl et al., 2001, Zhang et al., 2005).

To address the underlying mechanisms of brain plasticity for phonetic learning in adulthood, we designed a training software program in a preliminary single-subject MEG study to test its success (Zhang et al., 2000). The program incorporated features that were motivated by studies of infant-directed speech (IDS) or “motherese” (Burnham et al., 2002, Fernald and Kuhl, 1987, Kuhl et al., 1997, Liu et al., 2003), including adaptive signal enhancement, visible articulation cues, a large stimulus set with high variability, and self-initiated selection. The preliminary results suggested that rapid improvement could be achieved on the difficult nonnative phonemic contrast. Approximately 12 h of training for the Japanese adult subject showed an overall 22% improvement in identification accuracy with remarkable transfer of learning — there was a 27% improvement in recognizing the /r–l/ tokens by untrained voices. The training effect was also shown in enhanced neural sensitivity for the /r–l/ distinction, particularly in the left auditory cortex. Compared with other /r–l/ training studies with equivalent amounts of behavioral improvement, transfer of learning and sustained effect tested six months after training (e.g., Bradlow et al., 1999, Callan et al., 2003), our program reduced the total training hours by over 70%. As the preliminary results were based on a single subject, more subjects needed to be tested in order to evaluate the training methodology and investigate the neural mechanisms that reflect phonetic learning at both the individual and group levels.

There are two main objectives in the present training study: (a) to test the efficacy of our IDS-motivated training program in adults' learning of second language phonetic categories, and (b) to examine two hypothetical neural markers of learning in terms of brain-behavior correlates: neural sensitivity, as measured by the mismatch field response for phonetic discrimination (Näätänen et al., 1997), and neural efficiency, as measured by the focal degree and duration of brain activation during phonetic perception in terms of equivalent current dipole (ECD) clusters (Zhang et al., 2005). Previous neurophysiological studies have shown strong evidence of learning-induced enhancement in neural sensitivity to support phonetic categorization in adults as well as in children (Cheour et al., 1998, Imaizumi et al., 1999, Kraus et al., 1995, Menning et al., 2002, Näätänen et al., 1997, Nenonen et al., 2005, Rivera-Gaxiola et al., 2000, Tremblay et al., 1997, Winkler, 1999, Zhang et al., 2000). There is also evidence for learning-induced shift toward left hemisphere dominance in terms of enhanced neural sensitivity for linguistic processing (see Näätänen et al., 2007 for a review.). In line with the neural efficiency idea, fMRI studies have reported more focal activation for learned auditory stimuli particularly in native speakers or more advanced learners (Callan et al., 2004, Guenther et al., 2004, Wang et al., 2003). Cross-language MEG data have additionally indicated a shorter duration of bilateral activation for native speech processing in specific brain regions — the superior temporal and inferior parietal cortices (Zhang et al., 2005).

The central question of our study is whether substantial behavioral improvement in second language phonetic learning can be achieved in adulthood and simultaneously reflected by the spatiotemporal markers of neural sensitivity and neural efficiency, resulting in native-like perception and native-like brain activation patterns for learning the difficult speech contrasts in a second language. To cross-validate the brain activation patterns shown by the ECD cluster analysis approach and investigate the relationship between neural sensitivity and efficiency, we also employ distributed source analysis using minimum current estimates with fundamentally different assumptions about the source activity (Uutela et al., 1999, Zhang et al., 2005). We predict that our IDS-motivated training program would help circumvent interference from neural networks that have been shaped by native language experience, yielding significant brain-behavior correlations in both domains of sensitivity and efficiency.