first and second language patterns of brain activation in ... - CiteSeerX

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION IN KOREAN LAYRNGEAL CONTRASTS Haeil Park1 and Gregory K. Iverson1,2 1

University of Wisconsin-Milwaukee University of Maryland Center for Advanced Study of Language

2

Abstract. This study aims to localize the brain regions involved in the apprehension of Korean laryngeal contrasts and to investigate whether the Internal Model advanced by Callan et al. (2004) extends to first versus second language perception of these unique three-way laryngeal distinctions. The results show that there is a significant difference in activation between native and secondlanguage speakers, consistent with the findings of Callan et al. Specific activities unique to younger native speakers of Korean relative to native speakers of English were seen in the cuneus (occipital lobe) and the right middle frontal gyrus (Brodmann Area [BA] 10), areas of the brain associated with pitch perception. The current findings uphold Silva’s (2006) conclusion that the laryngeal contrasts of Korean are increasingly distinguished less by VOT differences than by their effect on pitch in the following vowel. A subsequent experiment was conducted to establish whether more traditional, older native speakers of Korean who still make clear VOT distinctions also activate both the cuneus and BA 10 in the same task. Preliminary results indicate that they do not, whereas speakers with overlapping VOT distinctions do show intersecting activations in these areas, thus corroborating Silva’s claim of emergent pitch sensitivity in the Korean laryngeal system. Keywords: Second language acquisition, Korean laryngeal contrast, Internal Model, superior temporal gyrus, cuneus, VOT, pitch.

1. INTRODUCTION It has long been observed that adult speakers have considerable difficulty distinguishing nonnative sounds that correspond to a single phoneme or phonetic 1 Korean Linguistics, Volume 14,1-19. © 2008 International Circle of Korean Linguistics.

2

H. PARK AND G. IVERSON

category in their own languages (as first reported by Lado 1957, followed up experimentally by Eckman et al. 2003). In confirmation of these observations, event-related brain potential (ERP) studies have exhibited different mismatch negativity (MMN) between native and nonnative, hard-to-perceive categories (Winkler et al. 1999; Sharma and Dorman 2000), while fMRI (functional Magnetic Resonance Imaging) studies have demonstrated a measure of reorganization in brain regions involving adult learning speech sounds, specifically, of Hindi dental-retroflex contrasts (Golestani and Zatorre 2004) and Mandarin lexical tone distinctions (Wang et al. 2003). Conversely, Golestani and Zatorre (2004) found that, after learning nonnative sounds, second-language speakers produced no significant difference in brain activation regions compared to native-language speakers when both groups performed a perception task. Recently, however, Callan et al. (2004) (cf. also Callan et al. 2003) compared brain activation patterns involving perception of the English /r/-/l/ contrast by native speakers of English with native speakers of Japanese who had learned the English contrast over a period of one month. In contrast to Golestani and Zatorre, the findings of Callan et al. support the ‘Internal Model’ of neurolinguistic organization in that the Japanese learners, through training, produced an activation pattern involving certain mappings in the brain which were different from those of native English speakers. The Internal Model as advanced by Callan et al. is a mechanism that simulates how articulatory-auditory and articulatory-orosensory mappings between different brain regions are used for identifying difficult second-language phonetic contrasts. The present study contributes to this discussion by comparing brain activation patterns between native speakers of Korean and adult native speakers of English who have learned the three-way laryngeal contrasts of the Korean stop system: tense /p’ t’ k’/ vs. lax /p t k/ vs. aspirated /ph th kh/. The study aims to localize the brain regions involved in the learning of these famously difficult phonetic categories, and investigate whether there is support from Korean for the Internal Model that Callan et al. have put forward. The hypothesis associated with the Internal Model is that second language learners, in processing learned sounds that are difficult for them, will produce more activations in areas of articulatory-auditory and articulatory-orosensory mappings, whereas native speakers will rely more on the regions of auditory-phonetic representation. Preliminary results lend support to the Internal Model, showing that nonnative phoneme perception activates brain regions of articulatory-auditory and articulatory-orosensory mappings more than in native phoneme perception. The results of the experiment also indicate that there is a significant difference in activation between native and second-language speakers, contradicting the findings of Golestani and Zatorre (2004). The most significant difference in native

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION

3

Korean speakers over native English speakers involved the Koreans’ activation of the cuneus in the occipital lobe and the right middle frontal gyrus, areas which are known to be involved with pitch or tone perception. As will be described, this points toward a special organization of the brain particularly among younger speakers of Korean with respect to laryngeal processing. 2. BACKGROUND In this section we review in some detail a breakthrough study by Callan et al. (2004) which bears importantly on the current investigation. Callan et al. investigated the neural processes underlying perceptual identification of the same phonemic contrasts by native and second-language speakers. They proposed a model of how perceptual identification is facilitated under conditions in which the phonetic distinction is ambiguous, predicting that audition-based phonetic representations will play a greater role in phoneme identification for native speakers than for second-language speakers. Specifically, then, the Internal Model is a mechanism that simulates how various mappings of different brain regions are used for identifying difficult second-language phonetic contrasts, as illustrated in Figure 1. The model maintains that second language learners, in processing learned sounds that are difficult for them, will use articulatoryauditory (the yellow arrow in Figure 1) and articulatory-orosensory mappings (red arrow) more, whereas native speakers will rely more on the regions of auditory-phonetic representation (blue circle). (The magenta arrow in Figure 1 represents mapping of the induced auditory activity into an orosensory representational space, while the cyan arrows indicate reciprocal connections between the cerebellum and cortical areas involved with articulatory, auditory, and orosensory aspects of speech). The claim is thus that non-native speakers implement articulatory-auditory and articulatory-orosensory mappings as they internally “mouthe” target phonemes in order to perceive them better, 1 while native speakers, having deeper familiarity with the sounds, activate the areas of auditory-phonetic representation directly. Specifically, Callan et al. demonstrated that, in perception of English, native speakers of English uniquely activated the audition-focused areas of the superior and middle temporal gyrus. The lower brain images in Figure 2 show activations unique to native speakers of English for the English /r-l/ perception. As

1

As one reviewer noted, this is then reminiscent of the “Motor Theory of Speech Perception” as proposed by Liberman et al. (1967), according to which one perceives speech with reference to how it is produced and accesses knowledge of phonemes by how they are articulated.

4

H. PARK AND G. IVERSON

Figure 1. Internal Model of brain regions and connections (Callan et al. 2004:1183).

Figure 2. Brain activity for English /r/ and /l/ vs. vowel identification; Japanese-unique activations (upper) and English-unique activations (lower) (Callan et al. 2004:1187).

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION

5

the Internal Model predicted, it was also shown that second-language speakers (i.e., native speakers of Japanese) during perceptual identification of /r/ and /l/ uniquely activated the brain regions involved with executing articulatoryorosensory and articulatory-auditory mappings, namely, Broca’s area (speech production planning), the anterior superior temporal gyrus (acoustic phonetic perception), the posterior superior temporal gyrus (articulatory gestural representaton), the supra marginal gyrus (orosensory speech representation), and the cerebellum (internal model instantiation). The upper brain images in Figure 2 exhibit brain activation regions unique to native Japanese speakers for the English /r-l/ perception. We turn now to the methodology underlying the present fMRI study of the perception of laryngeal distinctions by first and second language speakers of Korean. 3. METHODOLOGY 3.1. Stimuli and procedure Facilities for the fMRI experiment were made available by the Laboratory of Molecular Neuroimaging Technology and the Department of Diagnostic Radiology at Yonsei University College of Medicine in Seoul. Stimuli consisted of 168 monosyllabic words beginning with one of the 3-way laryngeally contrastive Korean stops (lax, tense, aspirated) followed by one of five different vowels: /a/, /e/, /i/, /o/, /u/. In addition, as a control, 56 words beginning with vowels were included. All of the stimuli were pronounced by a 30-year old female native speaker of Seoul Korean, and recorded in a silent room at the rate of 44,100 Hz. Her speech is not the subject of this investigation, of course, but her productions showed a three-way VOT contrast across the stop types. For the three coronal stops exemplified in Table 1, the values were: tense 12ms, lax 65ms, aspirated 92ms. There were five test conditions, as identified in Table 1, with 56 words in each. The first condition consisted of lax consonant-initial words, the second tense consonant-initial words, and the third aspirated consonant-initial words. The fourth condition consisted of vowel-initial words and the fifth was the baseline condition (silence). During the baseline condition, the subjects listened

H. PARK AND G. IVERSON

6

passively to background scanner noise (which existed in the other conditions as well), and did not respond to that trial 2 . Table 1. Design of experiment CONDITION 1 2 3 4 5

DESCRIPTION Lax-initial Tense-initial Aspirated-initial Vowel-initial Baseline

EXAMPLE tal t’al thal al

GLOSS ‘moon’ ‘daughter’ ‘mask’ ‘egg’

3.2. Subjects 6 English and 7 Korean speakers participated in the study. The English speakers consisted of 5 males and one female, whereas the Korean speakers consisted of 5 males and 2 females. Their ages ranged from 20 to 35. All subjects were right-handed as measured by the Edinburgh Handedness Inventory, and reported neither neurological/psychiatric illness nor speech/hearing difficulties. They were paid for their participation and gave written informed consent in accordance with the Declaration of Helsinki. All of the English-speaking subjects received Korean instruction over six to ten months in Korea, and all were at an intermediate level of proficiency. 3.3. Experimental design An event-related design was used in order for the participants not to be able to predict the subsequent stimulus. The stimuli were presented in a pseudorandom order made using the OptSeq mechanism. Stimuli were presented every 2.5 seconds through an MR-compatible headphone set (Figure 3). Using the left thumb, the task given to the subjects was to indicate which among the three consonant types they heard, or a vowel, by pressing one of four buttons. Inside the scanner, they were able to see four choices presented on the screen above their head in both English 3 ([1] lax, [2] tense, [3] aspirated, [4] vowel) and Korean ([1] ㄱ, ㄷ, ㅂ, ㅈ; [2] ㄲ, ㄸ, ㅃ, ㅉ; [3] ㅋ, ㅌ, ㅍ, ㅊ; [4] ㅇ. Even 2

This condition was designed to identify brain regions uniquely activated for listening to speech as opposed to background noise, but we did not include the brain images for that distinction because of space limitations. 3 Since the English-speaking subjects received instruction about which category of Korean sounds each of the four labels refers to (lax, tense, aspirated, and vowel), and they were familiar with the letters of the Korean alphabet, they had little problem understanding the four choices.

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION

7

though they would not see the buttons they press inside the scanner, they learned through instruction which finger to use in correspondence to the numbers [1] through [4], as coded above. By having the participants use their left hand to press the button, and because of hemispheric cross-over, we can see how the motor areas in the left hemisphere of the brain are activated without being affected by artifacts induced by button-press motor reactions. The subjects were required to respond as quickly as possible to minimize the difference in hemodynamic responses. Null trials where only silence occurs were included to be used as a baseline condition.

Figure 3. Protocol of Korean laryngeal contrast perception experiment.

All participants had a practice session outside the scanner in order to familiarize themselves with the experiment. This experiment was one of several that the subjects participated in while in an MRI scanner. The duration of the time that subjects stayed in a scanner ranged from 30 minutes to 1 hour, depending on how many experiments they participated in. For this experiment, the session lasted 13 minutes. 3.4. FMRI data collection and pre-processing Brain activity was measured using a Philips 3T MRI system (IntraAchieva, Phillips Medical Systems, Best, the Netherlands) for the acquisition of a T2weighted gradient echo planar imaging (EPI) sequence sensitive to the BOLD contrast [TR 4 = 2500ms, TE = 35ms, flip angle 90° 5 , slice thickness 6 = 4.5mm,

4

TR and TE are two important factors governing the time at which MR images are collected. TR, known as repetition time, is the time interval between successive excitation pulses. TE, known as echo time, is the time interval between excitation and data acquisition. Excitation denotes the introduction of energy into the ‘spinning’ nuclei via radio frequency pulse, thereby inducing the nuclei into a higher energy state.

8

H. PARK AND G. IVERSON

scan image matrix 7 of 80×80 and field of view 8 of 220mm, voxel 9 unit of 2.75×2.78x3 mm3]. In order to facilitate later spatial normalization, a highresolution T1-weighted structural MRI volume data set was also obtained from all subjects with a SENSE head coil configured with the following acquisition parameters: axial acquisition with a 256×256 matrix; 240mm field of view; 0.9375×0.9375×1.5mm3 voxels; TE 4.6ms; TR 20ms; flip angle 25°; slice gap 0mm; 1 averaging 10 per slice. 4. RESULTS 4.1. Behavioral performance We computed the accuracy scores for each subject’s responses to the given stimuli inside the scanner by calculating how many responses were correct among all responses, while excluding several cases of no responses in the calculation. The scores of all subjects included in the study were significantly higher than chance performance when evaluated at 50% by an independent onesample t test (t(12) = 6.78, P < 0.05, two tails) for the lax-tense-aspirated contrast. The native Korean-speaking participants performed significantly better than the native English speakers for all stimulus conditions (t(11) = 11.38, P < 0.05, two tails).

5 The flip angle is usually defined as the angle of excitation for a field echo pulse sequence, i.e., the angle to which the net magnetization is rotated or tipped relative to the main magnetic field direction. Flip angles between 0° and 90° are used in the gradient echo sequences adopted in our study. 6 The thickness of an imaging slice is determined by the distance between points at half the sensitivity of the maximum. Selecting the best fitting slice thickness is important for image quality. 7 A matrix is an array of numbers in rows and columns. A matrix with m rows and n columns is an m × n matrix with m and n as its dimensions. The matrix used in MRI determines the resolution of the scan. 8 Field of View (FOV), usually defined in units of mm, is the size of the two or three dimensional spatial encoding area of the image. In other words, it is the square image area that contains the object of interest to be measured; thus, the smaller the FOV, the higher the resolution and the smaller the voxel size but the lower the measured signal. 9 A voxel is a volume element which represents a value in three dimensional space, corresponding to a pixel for a given slice thickness. It is often used for the visualization and analysis of medical data. 10 Averaging, often called signal averaging, is a signal-to-noise improvement method to suppress the effects of random variations or random artifacts. This is commonly used to increase the signalto-noise ratio (SNR) by averaging several measurements of the signal. The number of averages is also seen as the number of excitations or the number of acquisitions. Using multiple averages can reduce respiratory motion in the same way it increases the SNR.

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION

9

4.2. Statistical image analysis A random-effects one-sample t test 11 of the {(/lax/ + /tense/ + /aspirated/) – 3 × vowel} contrast for both native English and native Korean-speaking participants was conducted to locate brain regions involving the processing of Korean laryngeal contrasts by native speakers of English. This is different from vowel processing, and is not due to general differences among lax, tense, and aspirated consonants or of vowels as processed by native Korean speakers. The results of the random-effect one-sample t test of the {(/lax/ + /tense/ + /aspirated/) – 3 × vowel} contrast for the native English speakers are shown in

Figure 4. Native English (left) and native Korean speakers (right) for /lax/, /tense/, and /aspirated/ perceptual identification relative to vowel perceptual identification. Upper images are from a side surface view, lower images from a sagittal view.

11 A subject’s response tends to vary from trial to trial and this response is variable from subject to subject. These two sources of variability must be considered when inferring about the population (Penny and Holmes 2003). In our fMRI experiment, 13 subjects were randomly drawn from the population at large. Thus, the subject variable is a random effect, allowing us to take sampling variability into account and to make inferences about the population from which the subjects were drawn. A one-sample t-test was used in this experiment because it tests the null hypothesis that the mean of one group or sample of observations for a given contrast is identical to zero. In our experiment, the null hypothesis is that the tested group or sample (either native English speakers or native Korean speakers) does not significantly activate any brain regions for the particular contrast {(/lax/ + /tense/ + /aspirated/) – 3 × vowel}.

H. PARK AND G. IVERSON

10

Table 2. Activation areas for {/lax/ + /tense/ + /aspirated/ – 3 × vowel} in native English speakers, P < 0.005 (T = 2.85) and cluster size > 20.

Side Left

Right

12

Region Inferior parietal lobule (Supramarginal gyrus & Angular gyrus) Superior parietal lobule Cerebellum, Anterior lobe, Culmen Medial frontal gyrus (SMA) Medial frontal gyrus (ACC) Middle frontal gyrus (DLPFC) Inferior frontal gyrus, triangular part (Broca’s area) Superior parietal lobule Cingulate gyrus Parietal lobe, Precuneus 14 Cerebellum, Anterior culmen Middle frontal gyrus (premotor & SMA) Insula

Brodmann BA 40 BA 40

MNI x,y,z 12 (mm) -40, -42, 44 -36, -37, 44

BA 7

4.35 3.51

Cluster size 445 118

-32, -50, 49 -4, -61, -10

3.62 3.38

118 56

BA 6

0, -5, 52

3.88

338

BA 32

0, 10, 46

4.48

338

BA 9

-36, 19, 29

3.21

90

BA 47 BA 47 BA 9 BA 7 BA 24 BA 7

-30, 31, -5 -34, 25, -8 -46, 17, 23 24, -60, 44 10, -1, 50 18, -58, 53 10, -66, 47 23, -45, -21 0, -58, -22 40, 4, 48 32, 10, 51 42, 16, 5

3.62 3.47 3.80 3.79 3.70 3.47 3.54 2.96 2.72 3.21 3.14 4.66

46 46 90 142 338 142 142 41 56 20 20 35

BA 6 BA 13

Zmax 13

The MNI coordinates, named for the Montreal Neurological Institute, refer to a position in the brain identified by using a large series of MRI scans on normal controls in an attempt to define a standard brain that represents the population more generally. 13 Zmax is a maximum Z value. The SPM (Statistical Parametric Mapping) method used in this study for analyzing fMRI data provides Z scores by converting the t statistics to Z scores. A higher Zmax corresponds to greater activation areas or a larger cluster size. 14 Our result that the precuneus in the Parietal Lobe is activated in the right hemisphere for native English speakers, whereas it is activated in the left hemisphere for native Korean speakers, appears to be consistent with the findings of Berman et al. (2003) that the right Precuneus shows increased blood flow during detection of (foreign) accents inasmuch as Korean functions as a foreign accent (indeed, foreign language) to native speakers of English.

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION

11

Figure 4 and Table 2. Activity for this contrast in the English speakers was found bilaterally in the supplementary motor area, somatosensory association cortex (Brodmann Area 15 7) and the cerebellum as well as in Broca’s area, the left anterior cingulate cortex, left cingulate gyrus, right premotor cortex, right insula, and left supra marginal gyrus. Results of the random-effect one-sample t test for the native Korean speakers are also shown in Figure 4 and Table 3. Activity for the same contrast in native speakers of Korean was found bilaterally in the cuneus in the occipital lobe as well as in the left supramarginal gyrus, the left precuneus (BA 7), the left angular gyrus and the right middle frontal gyrus. Table 3. Activation areas for {/lax/ + /tense/ + /aspirated/ – 3 × vowel} in native Korean speakers, P < 0.005 (T = 2.85) and cluster size > 20.

Side

Region

Left

Occipital lobe, Cuneus Superior parietal lobule (Angular gyrus) Parietal lobe, Precuneus Supramarginal gyrus Occipital lobe, Cuneus Frontal lobe, Middle frontal gyrus

Right

Brodmann BA 18 BA 39

MNI x,y,z (mm) 0, -88, 17 -12, -69, 53

Zmax 3.36 4.35

Cluster size 16 21 46

BA 7

-18, -60, 49

3.15

46

BA 40 BA 18 BA 10 BA 10

-42, -37, 33 8, -94, 18 32, 49, 18 32, 57, 10

3.21 3.67 3.63 2.81

39 21 58 58

5. DISCUSSION The results of the current experiment show that there is a substantial difference in brain activation between native speakers and second-language speakers, consistent with the findings of Callan et al. (2004) but in conflict with those of Golestani and Zatorre (2004) 17 . Brain activity reflecting differences in the percep15

A ‘Brodmann area’ is a region in the brain cortex defined on the basis of its cytoarchitectonic makeup, i.e., a different neuron type or distribution of neurons. Each has a different anatomical location in the cerebrum. Brodmann areas were originally defined by Korbinian Brodmann, who referred to them by numbers from 1 to 52. 16 This refers to the size of the cluster of significantly activated voxels in a brain region. 17 The contradiction between the results reported here and those of Golestani and Zatorre (2004) may be due in part to their participants having received 5 hours of non-native phoneme perception training right before the scanning, and in part to the fact that their study did not use the same phonetic contrast to compare differences in brain activity between native and non-native phoneme

12

H. PARK AND G. IVERSON

tion of lax, tense, and aspirated phonemes relative to that of vowels by native English and native Korean speakers is shown in Figure 4 and Tables 2 and 3. There was expected significant differential activity in crus VI of the cerebellum 18 (a region known to be involved with lip and tongue motor representation and speech production; Grodd et al., 2001; Wildgruber et al., 2001), in the left supra marginal gyrus (involved with orosensory-articulatory speech representation; Guenther and Perkell, in press), and in Boca’s area as well as the premotor cortex (involved with speech production planning). Notably, however, differential activation in the acoustic-phonetic area, i.e., the anterior superior temporal gyrus, and the articulatory-gestural speech representation area, i.e., the posterior superior temporal gyrus and planum temporale, was not found in native English perceptions of Korean stops. That the Internal Model thus does not apply completely to Korean laryngeal perceptions by native speakers of English suggests that the aforementioned areas may be involved only with oral gesture representation of non-native phonetic contrasts, as opposed to laryngeal gestures. Our results on the task of identifying the Korean three-way laryngeal contrasts relative to vowels show that some brain regions that were not incorporated into the Internal Model have greater differential activation for native English than for native Korean speakers. Regions of activation unique to English speakers are associated with speech production planning (both Broca’s area and the pre-motor area), with sensory-motor integration (the cerebellum), and with attentional control and selection of responses (the supplementary motor area, the right dorsolateral pre-frontal cortex and the right cingulate gyrus; Desmond et al., 1998; Milham et al., 2003). This is consistent with the expectation that there should be greater attentional control in selecting a perceived phonetic contrast for second language users than for native speakers. It is interesting to note, moreover, that activation areas unique to our relatively young native speakers of Korean, aged 20-35, were found in the cuneus in the occipital lobe and the right middle frontal gyrus (Brodmann Area [BA] 10). Wong et al. (2004) and Zatorre et al. (1992), among others 19 , have shown that the cuneus in the occipital lobe and BA 10 are involved with tone or pitch perception, and so the activation of both of these would appear to dovetail with perceptions (Hindi retroflex phonemes were used to investigate non-native perception but English alveolar phonemes to assess native perception). 18 Callan et al. (2003; 2004) maintain that the cerebellum is involved with instantiating articulatoryauditory and articulatory-orosensory internal models, but it may be instead that the cerebellum is involved with controlling the larynx along with the orosensory area and somatosensory association cortex (BA7), as the cerebellum is known to be a region of the brain that plays a key role in the integration of sensory perception and motor output (Grodd et al., 2001; Wildgruber et al., 2001). 19 Alain et al. (2002) also found that the cuneus region is implicated in the identification of pitch.

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION

13

Silva’s (2006) finding that the laryngeal contrasts of Korean are increasingly distinguished via their effect on pitch in the following vowel rather than by VOT differences, at least among younger speakers. In order to explore further whether Silva’s claim bears on the current fMRI results (cf. also Park 2007), we conducted a subsequent experiment to see if older speakers of Korean who still have clear VOT distinctions also activate both the cuneus and BA 10 (areas of the brain associated with pitch) in the same task. When three older Koreans aged 43-53 were scanned while performing the same task as younger Koreans, i.e., identifying whether the word that they hear begins with a lax, tense or aspirated consonant, or a vowel, preliminary results indicate that they do not activate these two areas together, whereas (typically younger) speakers with overlapping VOT distinctions do, thus corroborating Silva’s claim. 20 More specifically, if an older speaker’s VOT values overlap between the lax and aspirated phonation types (indicating that the subject employs pitch differences to discern phonation types in stops), the perception task should result in activation of pitch-associated areas of the brain. Following Silva’s methodology, then, we used as targets 9 three-syllable words beginning with a consonant in the sentence frame, i ken ____ i-la-ko haci-yo ‘this is called ___.’ Each sentence was written out in Korean on index cards. Subjects were asked to read each sentence at a normal and comfortable rate of speech in three randomized orders. In order to measure mean VOT, we averaged the values of three tokens of the first consonant of each target word, a listing of which is given in Table 4. Figure 5 illustrates the relationship between overlapping VOT and strong activation in both the cuneus and BA 10 for one older speaker, while Figures 6 and 7 show the relationship between distinct VOT and activation in only one of the two pitch areas for two other older speakers. The results for these two, then, indicate that it is not pitch differences but rather the substantial VOT distinctions that play a primary role in laryngeal contrast perception, whereas for younger (and the first of the older) speakers, the significant activation of both pitch-related areas, the cuneus and BA 10, suggests that pitch differences in the following vowel are the primary distinguisher between lax and aspirated stops.

20 One reviewer remarks that Silva’s finding that the VOT distinction between lax and aspirated consonants in younger speakers is being reduced in favor of f0 differences in the following vowel relates to speakers, not to listeners, and that therefore our fMRI results do not bear on the role of pitch in speech production. To the extent that perception mirrors production, however — indeed, arguably leads it (cf. Eckman and Iverson 2008) — we see the fMRI results as striking confirmation of the mental category distinctions that are implied by Silva’s acoustic observations.

14

H. PARK AND G. IVERSON Table 4. Target words used to measure mean VOT values (Silva 2006).

Lax Tense Aspirated Labial panucil ‘sewing’ ppatutuk ‘grinding sound’ phanamul ‘scallion salad’ Alveolar tamocak ‘cropping’ ttalikun ‘flatterer’ thakocang ‘forgn. village’ Velar kalodung ‘street lamp’ kkamaduk ‘far away time’ khanibal ‘carnival’

Figure 5. Mean VOT values for one 44 year-old Korean speaker (YG) for each phonation type with corresponding activation in both pitch-related areas of the brain, viz., the cuneus and BA 10. Brain activation patterns for younger speakers showing VOT overlap (or near overlap) between aspirated and lax stops were similar.

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION

15

Figure 6. Mean VOT values for one 50 year-old Korean speaker (MB) for each phonation type corresponding with activation in only the right middle frontal gyrus (BA 10).

16

H. PARK AND G. IVERSON

Figure 7. Mean VOT values for one 57 year-old Korean speaker (LJ) for each phonation type corresponding with brain activation only in the cuneus.

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION

17

6. CONCLUSION Using fMRI techniques, the present study investigated the systematic but variable neural correlates of first versus second language perception of Korean laryngeal manner contrasts in stops. We tested specifically the Internal Model hypothesis that it is dominantly articulatory-auditory and articulatoryorosensory mappings between different brain regions that are employed in identifying difficult second language phonetic contrasts, whereas native phoneme perception involves more the regions of auditory-phonetic representation. Nonnative perception also points toward differential activation according to whether the phonemes of interest are distinguished in terms of oral or laryngeal gestures, the latter perhaps via involvement of the cerebellum. Moreover, speakers of Seoul Korean for whom VOT is no longer the chief differentiator between lax and aspirated stops were found to be activating two pitch-related areas of the brain, consistent with Silva’s conclusion that Korean may be undergoing a shift in the properties that distinguish its system of laryngeal contrasts. Specifically, increased f0 at the beginning of the vowel following an initial stop in the aspirated relative to the lax series derives from the stiffer vocal folds employed in articulation of the aspirated series (Cho et al. 2002); but for (generally younger) speakers now bringing together the VOT values of initial lax stops and initial aspirates, this difference in f0 is becoming contrastive, serving as the chief cue distinguishing lax from aspirated stops. As Silva suggests, for speakers who have reduced the heretofore substantial VOT difference between lax and aspirated stops, this may call for a reinterpretation (or even replacement) of the features which distinguish the stop types phonologically. Whatever the particulars of the phonological rearrangement, however (cf. Silva 2006 for one possibility), support for it now lies not just in the acoustic signal itself, but in the mental representation of the perception of these sounds as revealed by the present fMRI study. Thus, speakers who reduce the VOT difference between lax and aspirated stops show activation of two pitch-related regions of the brain, indicating that for them the f0 differences in the following vowel between the two stop types are significant, or phonemic. On the other hand, speakers who continue to maintain a substantial VOT difference between lax and aspirated stops, and who also show similar pitch differentiation in the vowel following, do not activate both pitch-related areas of the brain, indicating that for them the f0 variation is still redundant, or non-phonemic. The fMRI results reported here thus confirm the mental category distinctions implied by Silva’s acoustic observations and undergird the emergent development of pitch sensitivity in the Korean laryngeal system.

18

H. PARK AND G. IVERSON

REFERENCES Alain, Claude, Stephen R. Arnott, Stephanie Hevenor, Simon Graham and Cheryl L. Grady. 2001. “What” and “where” in the human auditory system. Proceedings of the National Academy of Sciences 98: 12301-12306. Berman, Steven M., Mark A. Mandelkern, Hao Phan and Eran Zaldel. 2003. Complementary hemispheric specialization for word and accent detection. NeuroImage 19: 319-331. Callan, Daniel, Keiichi Tajima, Akiko Callan, Rieko Kubo, Shinobu Masaki and Reiko Akahane Yamada. 2003. Learning-induced neural plasticity associated with improved identification performance after training of a difficult second language phonetic contrast. NeuroImage 19: 113124. Callan, Daniel, Jeffrey Jones, Akiko Callan and Reiko Akahane-Yamada. 2004. Phonetic perceptual identification by native- and second-language speakers differentially activates brain regions involved with acoustic phonetic processing and those involved with articulatoryauditory/orosensory internal models. NeuroImage 22: 1182-1194. Cho, Taehong, Sun-Ah Jun and Peter Ladefoged. 2002. Acoustic and aerodynamic correlates of Korean stops and fricatives. Journal of Phonetics 30: 193-228. Eckman, Fred R., Abdullah Elreyes and Gregory K. Iverson. 2003. Some principles of second language phonology. Second Language Research 19: 169-208. Eckman, Fred R. and Gregory K. Iverson. 2008. Perception versus production in the acquisition of L2 phonemic contrasts. Paper presented at the 18th International Congress of Linguists (Workshop on Second Language Acquisition and Applied Linguistics), Seoul, Korea. Golestani, Narly and Robert J. Zatorre. 2004. Learning new sounds of speech: reallocation of neural states. NeuroImage 21: 494-506. Grodd, Wolfgang, Ernst Hulsmann, Martin Lotze, Dirk Wildgruber and Martin Erb. 2001. Sensorimotor mapping of the human cerebellum: fMRI evidence of somatotopic organization. Human Brain Mapping 13: 55-73. Guenther, Frank H. and Joseph S. Perkell. in press. A neural model of speech production and its application to studies of the role of auditory feedback in speech. B. Maassen, R. Kent, H. Peters, P. Van Lieshout and W. Hulstijn, eds., Speech Motor Control in Normal and Disordered Speech. Oxford: Oxford University Press. Halle, Morris and Kenneth Stevens. 1971. A note on laryngeal features. MIT Quarterly Progress Report 101: 198-212. Han, Jeong-Im. 1996. The Phonetics and Phonology of Tense and Plain Consonants in Korean. Ph.D. dissertation, Cornell University. Han, Mieko S. and Raymond S. Weitzman. 1970. Acoustic features of Korean /P, T, K/, /p, t, k/ and /ph, th, kh/. Phonetica 22: 112-128. Hardcastle, William J. 1973. Some observations on the tense-lax distinction in initial stops in Korean. Journal of Phonetics 1: 263-272. Iverson, Gregory K. and Joseph C. Salmons. 1995. Aspiration and laryngeal representation in Germanic. Phonology 12: 369-396. Kim, Chin-Woo. 1970. A theory of aspiration. Phonetica 21: 107-116. Kim, Mi-Ryoung and San Duanmu. 2004. Tense and lax stops in Korean. Journal of East Asian Linguistics 13: 59-104. Lado, Robert. 1957. Linguistics Across Cultures: Applied Linguistics for Language Teachers. Ann Arbor: University of Michigan Press.

FIRST AND SECOND LANGUAGE PATTERNS OF BRAIN ACTIVATION

19

Liberman, Alvin M., Franklin S. Cooper, Donald P. Shankweiler and Michael Studdert-Kennedy. 1967. Perception of the speech code. Psychological Review 74: 431-461. Lisker, Leigh and Arthur S. Abramson. 1964. A cross-language study of voicing in initial stops: Acoustical measurements. Word 20: 384-422. Park, Haeil. 2007. The perception of Korean laryngeal contrasts by native speakers of English. LSO Working Papers in Linguistics 7: 125-133. Linguistics Student Organization, University of Wisconsin–Madison. Penny, Will D. and Andrew J. Holmes. 2003. Random-effect analysis. R.S.J. Frachowiak, K.J. Friston, C. Frith, R. Dolan, C. J. Pice, S. Zeki, J. Ashburner and W.D. Penny, eds., Human Brain Function (2nd ed.). 843-850. London: Academic Press. Sharma, Anu and Monica F. Dorman. 2000. Neurophysiologic correlates of cross-language phonetic perception. Journal of the Acoustical Society of America 107: 2697– 2703. Silva, David J. 2006. Acoustic evidence for the emergence of tonal contrast in contemporary Korean. Phonology 23: 287-308. Wang, Yue, Joan A. Sereno, Allard Jongman and Joy Hirsch. 2003. fMRI evidence for cortical modification during learning of mandarin lexical tone. Journal of Cognitive Neuroscience 15: 1019-1027. Wildgruber, Dirk, Hermann Ackermann and Wolfgang Grodd. 2001. Differential contributions of motor cortex, basal ganglia, and cerebellum to speech motor control: effects of syllable repetition rate evaluated by fMRI. NeuroImage 13: 101-109. Winkler, István, Teija Kujala, Hannu Tiitinen, Päivi Sivonen, Paavu Alku, Anne Lehtokoski, István Czigler, Valéria Csépe, Risto J. Ilmoniemi and Risto Näätänen. 1999. Brain responses reveal the learning of foreign language phonemes. Psychophysiology 36: 638–642. Wong, Patrick C.M., Lawrence M. Parsons, Michael Martinez and Randy L. Diehl. 2004. The role of the insular cortex in pitch pattern perception: The effect of linguistic contexts. The Journal of Neuroscience 24: 9154-9160. Zatorre, Robert J., Alan C. Evans, Ernst Meyer and Albert Gjedde. 1992. Lateralization of phonetic and pitch processing in speech perception. Science 256: 846-849. [Material in this paper was presented at the 13th Annual Meeting of the Organization for Human Brain Mapping, Chicago, June 2007 and at the 5th Workshop in General Linguistics, University of Wisconsin–Madison, March 2007. We thank these audiences and two anonymous reviewers for their helpful comments.] Haeil Park Department of English University of Wisconsin–Milwaukee Milwaukee, WI 53201-0413 Gregory K. Iverson Center for Advanced Study of Language University of Maryland College Park, MD 20742-5121

Received 9 March 2008; revision received 16 May 2008; accepted 16 May 2008