Phoneme discrimination of an unrelated language: Evidence for a narrow transfer but not a broad-based bilingual advantage

Abstract

This study examines monolingual and multilingual individuals’ discrimination of stop consonants in a language to which they had never been exposed: Korean. If bilingualism leads to increased flexibility in phonological categorization, we may see a broad-based bilingual advantage for phoneme discrimination. Using a Korean phoneme discrimination task, we compared 56 adults in four groups: monolingual English, bilingual Spanish, bilingual Armenian, and trilingual. Findings indicate that Spanish–English bilingual individuals scored no better than English monolinguals, and lower than Armenian–English bilingual individuals. In this case, the advantage from early childhood non-English exposure or current bilingualism was found to be specific only to languages with similar phonemic categories. This supports a narrow first/second language to third language transfer view of phoneme discrimination skills.

Keywords

Phoneme discrimination Korean bilingual advantage unrelated language monolingual trilingual narrow transfer

Adult language learners often experience difficulty distinguishing among the speech sounds, or phonemes, of a new language (Johnson & Newport, 1989; Long, 1990). This is especially true if the second or third language (L2/L3) has phoneme contrasts that do not exist in the native language (L1). There is considerable individual variability in adult language learners’ success at learning the phonemes of a new language. In particular, there seems to be a bilingual advantage in phoneme discrimination for unfamiliar languages (for a review see Cenoz, 2003). There is some question, however, whether this advantage is general to all bilingual individuals, reflecting a broad-based advantage, or specific only to those with exposure to similar kinds of phoneme distinctions, implying a narrow language transfer. Some evidence suggests a broad-based bilingual advantage that could be a result of increased flexibility in phoneme categorization. For example, Kaushanskaya and Marian (2009) found that bilingual individuals performed better than monolinguals in learning phonetically unfamiliar words. Previous research on bilingual advantages in phoneme discrimination in an unrelated target language has been mixed on if there is a broad-based advantage. The research has varied, including finding a bilingual advantage (e.g., Enomoto, 1994), marginal trends (Lambert & MacNamara, 1969), mixed results (Beach, Burnham, & Kitamura, 2001), and no bilingual advantage (e.g., Werker, 1986). The current study investigates whether there is an advantage and if so the breadth of this advantage, and possible mechanisms, by comparing monolingual, bilingual, and trilingual individuals in a stop consonant discrimination test in an unfamiliar language (Korean).

In this introduction, we first very briefly review the research on general bilingual advantages in languages that are related to participant’s L1, and on phonological tasks. By “related” we mean the languages have a common known root language. For example, English and Spanish are related in that they have a common root language of Latin. This research provides context and suggests possible hypotheses for our main literature review—that of bilingual advantages in discriminating phonemes in a language not related to the participant’s own languages.

General bilingual advantages in learning a new language

Evidence is mixed as to whether bilingual individuals outperform monolingual individuals in the acquisition of skills in a new language. Research in Spain, with English as the target language, has shown that bilingual individuals have advantages in the oral and written aspects of the target language (Sanz, 2000). Research in Switzerland, Canada, and the United States, with French as the target language, all showed similar bilingual advantages (Bild & Swain, 1989; Brohy, 2001; Thomas, 1988). Other studies in Europe, however, have shown no significant differences between monolingual and bilingual groups in L3 proficiency in grammar, writing, vocabulary, dictation, reading, and comprehension (Jaspaert & Lemmens, 1990; Sanders & Meijers, 1995). These results indicate that bilingual individuals sometimes show an advantage when learning a language that is related to the participant’s L1(s). Later, we review research on bilingual advantages when the target language is not related to the L1(s), and we show that the evidence is mixed.

For more on the general advantages of bilingualism on L3 acquisition see the review by Cenoz (2003), which suggests that bilingualism is advantageous during L3 acquisition in the domains of linguistic communication and cognitive skills, general proficiency, phonetics, syntax, and pragmatics—and that this positive transfer between languages is stronger if languages are similar typologically. See also a review by Bialystok, Craik, Green, and Gollan (2009) for a broad discussion of bilingual benefits (such as in selective attention and inhibition, task switching, and even delayed onset of dementia) and disadvantages (such as in semantic fluency and vocabulary size in a given language).

Phonology-related bilingual advantage

In particular, there is some evidence of bilingual advantages in phonological tasks. Bialystok, Luk, and Kwan (2005) studied first graders (six years old) and found that Hebrew–English and Spanish–English bilingual children performed better than English monolinguals on a phoneme counting task and non-word decoding task (tasks were performed in the language(s) participants were fluent in). It should be noted, though, that Chinese–English bilingual children performed about the same as English monolinguals, showing that the bilingual advantages were not found in all bilingual groups.

In addition to this evidence, Kauchanskaya and Marian (2009) found that both English–Spanish and English–Mandarin bilinguals outperformed English monolinguals in learning novel words. The novel words contained non-English phonemes that were not familiar to Spanish or Mandarin speakers. Kauchanskaya and Marian’s (2009) findings could suggest a broad-based bilingual advantage in phoneme categorization flexibility. Given this, we might expect to see consistent bilingual advantages in perceiving speech sounds of other languages,

However, there are not always consistent bilingual advantages in phoneme discrimination. Gallardo del Puerto (2007) found that even when the participants’ and the target languages were phonologically similar, bilingual proficiency did not lead to any advantage on L3 speech sound discrimination. It remains unclear why these advantages are sometimes found, but other times not—something we attempt to address in this article. We now look specifically at the current study’s area of research—phoneme discrimination in unrelated languages.

Bilingual advantages in phoneme discrimination of unrelated languages

Bilingual advantages in phoneme discrimination have been investigated in unrelated target languages. This is the main focus of the current study. Some prior studies have shown no statistically significant differences between bilingual and monolingual groups on their discrimination of the phonemes of an unrelated language (e.g., Russian: Lambert & MacNamara, 1969). Werker (1986) found no bilingual advantage in the discrimination of speech sounds from an unrelated language, in particular Hindi and Thompson (a Salishan North American language). Participants were fluent in languages unrelated to Hindi or Thompson, such as English, French, and German. The results showed no differences in performance between the monolingual, bilingual, and trilingual individuals on their discrimination of the unrelated language phonemes. Werker concluded that broad, non-specific bilingual language experience is insufficient in itself to maintain the ability to discriminate unrelated language sounds. Rather, Werker hypothesized that specific listening experience of the speech sounds under investigation may be required.

Beach et al. (2001) also found no broad bilingual advantage, but they did find one specific advantage. They compared English monolinguals, Greek–English bilingual individuals, and a Thai control group on their discrimination of Thai bilabial-stops. The results showed no difference between the English monolinguals and Greek–English bilinguals on the discrimination of some Thai bilabial stop consonants (/ba/ versus /p^ha/ and /pa/ versus /p^ha/), but Greek–English bilinguals were better than English monolinguals at discriminating the Greek-like prevoiced versus unaspirated contrast (/ba/ versus /pa/). Although it is unclear if the latter was due to narrow transfer or a general bilingual advantage, the researchers found that those Greek bilinguals who produced extreme voice onset times (VOTs) for /b/ and /p/ discriminated best of all. This indicated that the advantage could be due to a narrow transfer of a specifically practiced skill. In this study, bilinguals showed some advantages in distinguishing among unfamiliar speech sounds, but this was the case for only certain phoneme distinctions similar to those found in their L1/L2.

In contrast, some studies have suggested an unambiguous broad-based advantage of bilingualism in the discrimination of an unrelated language’s speech sounds. Cohen, Tucker, and Lambert (1967) found that English–French bilinguals were better than English monolinguals at accurately producing phonemes not present in either language after hearing them. Of note is that, unlike the present study and other studies reviewed here, Cohen et al. (1967) did not use phonemes from any one specific unfamiliar language. Instead, the phonemes were a mixed set from a variety of languages, none of which were the participants’ L1s.

In a study of British college students who were either English monolinguals or bilingual in English and another European language (Italian, French, German, or Spanish; one participant was trilingual), with Japanese as the unrelated target language, Enomoto (1994) found that the bilingual participants scored reliably higher on the speech sound discrimination task than the monolingual group. Enomoto concluded, contrary to Werker (1986), that non-specific broad language experience does improve speech sound discrimination in an unrelated language. In particular, Enomoto theorized that bilingualism helps cognitive flexibility, and leads to a greater tendency to look for structure in perceptual situations.

To further examine the nature of a bilingual advantage in phoneme discrimination, the current study compared monolingual, two bilingual, and a trilingual group of adults on their ability to discriminate phonemes of an unfamiliar language (Korean). For one bilingual comparison group, neither of their languages (Spanish and English) was phonologically similar to the target language in the contrasts we studied. This group therefore served as a comparison group for a broad-based bilingual advantage. In the other bilingual group, one of their languages (Armenian) was phonologically similar to Korean, at least for the stop consonants we used in the study. This Armenian–English bilingual group therefore served as the comparison group for a specific exposure advantage. Our research question therefore involves whether the mechanism of a possible bilingual advantage is either (a) language-general flexibility in changing phoneme category boundaries, or (b) narrow L1/L2 language transfer to L3, facilitated by exposure and practice of specific phoneme contrasts similar or identical to those in the target language. On the one hand, Kaushanskaya and Marian (2009) and Enomoto’s (1994) results suggest that phoneme distinction boundaries may be generally more flexible due to bilingual exposure. On the other hand, Werker’s (1986) results would suggest that there is no evidence of increased flexibility in the phoneme distinction boundaries of bilinguals, compared to monolinguals. If we find that only the Armenian–English bilingual group has a phoneme discrimination advantage, and not the Spanish–English bilingual group, that would support the narrow transfer hypothesis, which would suggest that phoneme distinction boundaries are not necessarily more flexible, but are set in a way that allows for better discrimination of relevant contrasts in their languages. If this is the case, this advantage would transfer to novel languages that have similar phoneme category boundaries. On the other hand, if the Spanish–English bilingual group shows a phoneme discrimination advantage over English monolinguals, this would suggest increased flexibility in phoneme distinction boundaries.

Method

Participants

Fifty-six undergraduates participated in the study for course credit (age range 18–37; 70% female, 30% male). Participants included English monolingual (n = 14, age range 18–20; 79% female, 21% male), Spanish–English bilingual (n = 14, age range 18–21; 50% female, 50% male) Armenian–English bilingual (n = 14, age range 18–22; 79% female, 21% male) and trilingual (at least three languages; n = 14, age range 18–37; 71% female, 29% male) speakers. Four of the trilingual speakers were fluent in a language with aspirated versus non-aspirated stop consonant distinctions (three Armenian, one Thai). See Table 1 for participant characteristics by language group, and Table 2 for a full list of the languages in which the trilingual participants were fluent.

Table 1.

Participant characteristics by language group.

Characteristic	Monolingual English	Bilingual Spanish–English	Bilingual Armenian–English	Trilingual
Average age (M years)	18.6	19.2	19.1	21.3
Percent female (male)	78.6 (21.4)	50.0 (50.0)	78.6 (21.4)	71.4 (28.6)
First language	English (100%)	Spanish (78.6%)	Armenian (92.9%)	Armenian (21.4%)
		English (14.3%)	Both (7.1%)	English (14.3%)
		Both (7.1%)		Spanish (7.1%)
If fluent: How old when first learned Spanish or Armenian (M years)	n/a	1.7 (Spanish)	1.5 (Armenian)	1.3 (Armenian, n = 3)8.8 (Spanish, n = 4)
Other languages: How old when began speaking (M years)	8.5 (Spanish, n = 3)	6.4 (English, n = 12)	4.7 (English, n = 6)12.0 (Spanish, n = 2)	Overall 5.7 (6.6, French, n = 3; 12.0, Spanish, n =2; 4.0, Farsi, n = 2; 3.0, Hebrew, n = 2)
How old (M years) when started learning English	1.0	6.0	3.7	4.5
M age became fluent in English	3.0	7.9	6.8	6.9
Percent of participants who spoke non-English language at home	7.1% (Spanish)7.1% (Tagalog)85.7% None	100% (Spanish)	100% (Armenian)	100% (various) (21.4% Armenian, 7.1% Thai, 7.1% Spanish, & 64.3% other)
Percent of parents who spoke non-English language at home	14.3% (Spanish)7.1% (Tagalog)78.6% None	100% (Spanish)^a	100% (Armenian)	100% (various) (21.4% Armenian, 7.1% Thai, 7.1% Spanish, & 64.3% other)

Note. ^aExcludes one Spanish–English bilingual participant who marked on the questionnaire that he had never spoken Spanish at home, but since this does not fit with his interview and other responses, this could be an error.

Table 2.

Trilingual participant languages.

Participant number	First most familiar non-English language	Second most familiar non-English language	Third most familiar non-English language
401	French	Spanish	–
402	Armenian^a	Farsi	–
403	Farsi	Hebrew	–
404	Hebrew	Yiddish	–
405	Ukrainian	Spanish	–
406	Creole	French	–
407	Armenian^a	Farsi	–
408	Spanish	Q’anjob’al	–
409	Farsi	Spanish	–
410	Ukrainian	Russian	–
411	Visayan	Tagalog	–
412	Hindi	Thai^a	French
413	Armenian^a	Russian	–
414	Farsi	French	–

Indicates fluency in a language with an aspirated versus non-aspirated distinction in stop consonants.

Note. All participants were fluent in English. Participant 404 was the only participant who had English as their first language. Trilingual in this study is used to mean participants with three or more fluent languages.

Materials and procedure

Participants came into a quiet windowless lab room, one at a time, and completed a language background questionnaire (10 minutes) and interview (5 minutes), and then a Korean phoneme discrimination test (15 minutes), which was programmed in E-Prime 2.0 software (Schneider, Eschmann, & Zuccolotto, 2002).

Language background

A detailed questionnaire asked about the participant’s general language background (adapted from Au, Knightly, Jun, & Oh, 2002). The questions inquired about their first language, parents’ spoken languages, number of languages spoken, and so on. There were specific questions about how much they heard, were spoken to, and conversed in their language(s) during various periods of their lives. A follow-up interview confirmed and clarified participants’ responses on the language background questionnaire.

Phoneme discrimination

Stimuli utilized Korean stop consonants as the target phonemes (materials adapted from Oh, Au, & Jun, 2010). Korean utilizes a three-way distinction in stop consonants (lenis, tense, aspirated) that includes a non-aspirated and aspirated distinction. While an easy task for native Korean speakers, distinguishing these phonemes is challenging for speakers of other languages (Oh, Jun, Knightly, & Au, 2003). Similar to Korean, Armenian dialects have a two- or three-way aspirated versus non-aspirated stop consonant distinction (e.g., voiced, voiceless, and voiceless aspirated; Vaux, 1998, p. 9). Thai also has a three-way distinction in stop consonants. In contrast, Spanish and English do not include the aspirated versus non-aspirated distinction. The other languages spoken by participants, not mentioned above, also did not have this distinction.

Targets in the phoneme discrimination task were the Korean denti-alveolar (lenis: /t/, tense: /t*/, aspirated /t^h/), bilabial (/p/, /p*/, /p^h/), and velar (/k/, /k*/, /k^h/) consonants. We used nine minimal triplets of Korean words (three triplets per place of articulation) that varied only on the target consonant (e.g., /pang/, /p*ang/, /p^hang/). Using the ABX technique (Harris, 1952; Liberman, Harris, Hoffman, & Griffith, 1957), participants heard one speaker say two different words (A and B), and then a second speaker say one of the first two words (X). The participant then had to indicate whether X matched A or B. An interstimulus interval (ISI) of 1000 ms separated each word or phrase. Within each ABX triplet the phoneme contrast was presented with phrase-initial or phrase-medial placement (in separate trial blocks). Participants were randomly assigned to conditions. Condition 1 first presented the phrase-initial phonemes, followed by the phrase-medial trials. Condition 2 presented the blocks in the opposite order. Each participant completed 108 trials: 54 trials with the target phoneme in phrase-initial position and 54 trials in phrase-medial position.

Results

In the results that follow we seek to investigate whether early language exposure and current fluency in the participants’ L1s predict better discrimination of the Korean phonemes. We then investigate whether the effects we found are due to general bilingualism or specific exposure to an aspirated versus non-aspirated distinction in stop consonants. We also test for possible confounds.

Early language exposure

In this section, we investigate whether early exposure to non-English language predicts better discrimination of Korean stop consonants, and seek to determine whether any advantage found is due to exposure to similar phonemes (as in Armenian), or generally true of all bilingual exposure. We found a positive relationship between the number of hours of early exposure (before age five) to Armenian, a language that has aspirated versus non-aspirated stop consonant distinctions, and accuracy on the Korean phoneme discrimination task, r = .439, p = .001 (N = 56; correlations are bivariate Pearson’s r throughout). However, among those participants whose languages do not have aspirated versus non-aspirated stop consonant distinctions (n = 38), we found no significant correlation between the number of hours of exposure to non-English languages during early childhood and accuracy on the Korean phoneme discrimination task, r = .030, p = .859. This suggests that the advantage may be due to specific exposure to similar phonetic categories rather than general bilingual exposure.

To investigate the same question from a different angle, we analyzed the effect of the number of fluent languages on phoneme discrimination. The number of languages the participants had heard during early childhood correlated positively with their overall accuracy on the Korean phoneme discrimination task (N = 56). This was true whether the criterion was the number of languages before age one, r = .311, p = .019, age three, r = .427, p = .001, or age five, r = .384, p = .003. However, when we excluded data from participants whose languages distinguish between aspirated and plain stop consonants (i.e., 17 fluent in Armenian and 1 fluent in Thai), we found no significant correlations between the amount of exposure to non-English languages and accuracy in discriminating the Korean stop consonants (n = 38; see Table 3). As before, this suggests a narrow transfer and provides little evidence for a general bilingual advantage.

Table 3.

Excluding languages that distinguish aspirated from non-aspirated stop consonants: The correlation between childhood exposure to non-English languages and overall accuracy on the Korean stop consonant discrimination task (N = 38).

Correlation between discrimination score and…	r	p
Hours heard non-English language(s) before age 5	.030	.859
Hours talked to in non-English language(s) before age 5	.080	.632
Hours spoke non-English language(s) before age 5	.009	.957
Total hours non-English exposure before age 5	.045	.789
Combined hours of non-English exposure birth–present	−.053	.750

We found no statistically significant correlation between the amount of cumulative non-English language exposure and the accuracy distinguishing the Korean stop consonants, r = .251, p = .062, N = 56. This is a marginal result so we conducted a hierarchical regression analysis to investigate whether this relationship (if any) holds when adjusted for whether the non-English language had aspirated versus non-aspirated phoneme distinctions in stop consonants (and when adjusting for other variables; see Table 4). In Model 1 we adjust for gender and age to investigate whether exposure to non-English languages before the age of five predicts phoneme discrimination. In Model 2 we also adjust for exposure to non-English languages at other times during development. The regression analysis found that the participants’ self-report of the amount of time they heard non-English languages before the age of five was a reliable predictor of phoneme discrimination in Korean, adjusting for age, gender, and exposure during other developmental periods (see Models 1 and 2). However, Model 3 shows that this relationship does not hold when adjusting for whether the participants are fluent in languages that make an aspirated distinction in their stop consonants (e.g., Armenian, coded 1) or not (e.g., Spanish, English, coded 0). Consistent with the earlier analyses, this suggests that early exposure to more than one language only seems to produce a statistically significant advantage when the language heard also has similar phonemic distinctions.

Table 4.

Predictors of accuracy in Korean phoneme discrimination. Examining hours of hearing non-English Language (NEL) at various stages of development (N = 56).

	Model 1			Model 2			Model 3
	b	SE(b)	β	b	SE(b)	β	b	SE(b)	β
Age	−.27	.50	−.07	−1.62	.93	−.44	−1.01	.89	−.28
Gender	−.78	3.01	−.04	−.80	3.00	−.04	−1.18	2.80	−.05
Weeks heard NEL before age 5	.12	.06	.27^*	.18	.08	.40^*	.15	.08	.15
Weeks heard NEL elem. School (5–11)				−.18	.09	−.52	−.16	.09	−.16
Weeks heard NEL middle Sch. (11–12)				.13	.16	.19	.02	.15	.02
Weeks heard NEL high School (14–17)				.08	.10	.17	.08	.09	.08
Weeks heard NEL college (18–present)				.14	.09	.44	.07	.08	.07
Aspirated distinction /t/p/k/ in Language							9.29	3.19	.42^**
Constant	70.22^***	10.03		95.71^***	18.10		85.24^***	17.21
F (df)	1.48 (3, 52)			1.58 (7, 48)			2.65^* (8, 47)
ΔR ²	.08			.11			.12
R ² _adjusted	.03			.07			.19

p < .05; ^**p < .01; ^***p < .001. Statistically significant statistics in boldface.

Note. NEL = non-English language. Variance inflation factors (VIF) collinearity statistics ranged from 1.01 to 4.61.

Current fluency

In this section, we investigate the question of whether the languages the participants are fluent in today predicts their performance on the Korean discrimination task. Figure 1 shows the mean phoneme discrimination scores and standard errors for English monolinguals, Spanish–English bilinguals, Armenian–English bilinguals, and trilinguals. On the right we dichotomize the sample into those who are fluent in a language that has similar stop consonant distinctions to Korean and those who are not. A one-way analysis of variance (ANOVA) revealed a main effect of language group in overall accuracy on the Korean phoneme discrimination task, F(3, 52) = 5.23, p = .003, η_p² = .232 (Figure 1, left). Tukey honestly significant difference (HSD) post hoc analyses indicate that Spanish–English bilingual participants (M = 62.76, SD = 5.32) scored lower than both Armenian–English bilingual (M = 74.01, SD = 10.50, p = .012) and trilingual participants (M = 74.21, SD = 10.86, p = .010). There were no other statistically significant differences. To test for an early exposure confound, we controlled for early childhood exposure to a L2/L3 and still found an Armenian–English over Spanish–English advantage, F(1, 27) = 12.77, p = .001, η_p² = .329. Just being fluent in multiple languages does not seem to result in a sizable general advantage in the discrimination of phonemes of an unfamiliar language.

Figure 1.

Discrimination of Korean stop consonants by language group (left; n = 14 per group, N = 56) and by categorizing individuals according to whether they are fluent in a language that distinguishes between aspirated and non-aspirated stop consonants /k/, /p/, and /t/ (right; no aspirated n = 38, aspirated n = 18). Scale shown with chance performance at origin. Error bars represent standard errors.

Those participants fluent in a language that distinguishes between aspirated and non-aspirated stop consonants scored higher on the Korean phoneme discrimination task (M = 76.23, SD = 10.22; n = 18) compared to those participants not fluent in such a language (M = 66.11, SD = 8.74; n = 38), t(54) = 3.84, p < .001, r_pb² = .214 (Figure 1, right).

Specific phoneme contrasts

We investigated whether bilingual advantages were general to all the phoneme distinctions we tested, or specific to just some. Table 5 shows the results of several ANOVAs examining how monolingual, bilingual, and trilingual individuals compare on their discrimination of specific phoneme target types (such as aspirated, lenis, velar, denti-alveolar; N = 56, n = 14 per group). Tukey post hoc analysis showed that the Armenian–English bilinguals’ advantage over Spanish–English bilinguals held for most of the phoneme contrasts in the study. As can be seen, trilingual individuals usually scored higher than Spanish bilingual individuals, but not always.

Table 5.

Analysis comparing monolingual, bilingual, and trilingual individuals’ ability to distinguish various types of Korean phoneme contrast (N = 56, n =14 per group).

Phoneme type of target	F (3, 55)	p	η_p²
Aspirated stop consonant	4.07	.011	.190
Post hoc: Armenian bilingual and trilingual > Spanish bilingual
Lenis stop consonant	4.04	.012	.189
Post hoc: Armenian bilingual and trilingual > Spanish bilingual
Tense stop consonant	4.58	.006	.209
Post hoc: Armenian bilingual > monolingual and Spanish bilingual
Velar stop consonant	3.94	.013	.185
Post hoc: Trilingual > monolingual and Spanish bilingual
Denti-alveolar stop consonant	3.16	.032	.154
Post hoc: Armenian bilingual > Spanish bilingual
Bilabial stop consonant	5.53	.002	.242
Post hoc: Armenian bilingual and trilingual > Spanish bilingual

Note. All post hoc analysis in this table used Tukey’s honestly significant difference (HSD) test. All other differences were not significant.

In order to test for a broad-based general bilingual advantage we also investigated if being currently fluent in Spanish gave any advantage in distinguishing any of the specific Korean phonemes studied. A series of planned t-tests showed that Spanish–English bilingual individuals (n = 14) showed no advantage compared to English monolingual individuals (n = 14) on any of the 47 different phoneme contrasts we analyzed (all p values > .05; we examined various combinations of lenis, tense, and aspirated phonemes in initial or medial position for the stop consonants /p/, /t/, or /k/).

Those participants who were fluent in Armenian (n = 17; 14 bilingual, 3 trilingual) were compared to all other participants (n = 39, a mix of 14 monolingual, 14 bilingual, and 11 trilingual individuals) on distinguishing the Korean phonemes. Overall, those fluent in Armenian (M = 75.65, SD = 10.22) scored higher than those who were not (M = 66.62, SD = 9.20) on overall accuracy of distinguishing the Korean phonemes, F (1, 55) = 10.68, p = .002. Independent two-sample t-tests showed that those fluent in Armenian showed significantly higher accuracy compared to those not fluent in Armenian on 29 of the 47 different phoneme contrasts analyzed. No clear pattern emerged of which particular types of phonemes were most likely to show significant differences.

Controlling for early L2 exposure

To investigate whether the amount of early experience accounted for Armenian–English bilingual individuals’ better discrimination (compared to Spanish–English bilinguals), we ran an analysis of covariance (ANCOVA) controlling for the amount of early exposure by setting the estimated number of days of exposure to a non-English language before the age of five as the covariate. We found Armenian–English bilingual individuals (n = 14) still scored significantly higher than Spanish–English bilingual individuals (n = 14) on overall accuracy of distinguishing the Korean phonemes, F(1, 27) = 12.77, p = .001, η_p² = .329.

Aspirated versus non-aspirated analysis

Detailed phoneme contrasts

We analyzed accuracy in distinguishing many of the different phoneme combinations by comparing those participants fluent in a language that does distinguish between aspirated and plain stop consonants (n = 18, e.g., Armenian or Thai) and those participants not fluent in such a language (n = 38). As seen in Table 6, there was consistency in the results: 9 of the 12 phoneme-specific comparisons show that those who have a language that distinguishes aspirated from non-aspirated stop consonants had an advantage discriminating the Korean phonemes used in this study. With multiple comparisons there is a risk of finding a significant difference just by chance when in fact none exist in the population (a type I error), so we performed a Bonferroni correction to adjust for this probability and still found that 5 of the 12 comparisons showed that those fluent in a language with the aspirated distinction had an advantage (shown in boldface in Table 6).

Table 6.

Planned t-tests comparing individuals with a language that distinguishes aspirated from non-aspirated stop consonants (n = 18) to individuals without (n = 38) on various target phoneme discrimination accuracy scores (total N = 56).

Target phoneme	M_accuracy with aspirated^a	M_accuracy without aspirated^b	t(54)	r _pb ²
Aspirated /k/	78.24	65.79	2.30^*	.089
Aspirated /p/	78.70	69.96	2.02^*	.070
Aspirated /t/	73.61	66.89	1.51	.041
Lenis /k/	67.59	66.45	.25	.001
Lenis /p/	71.30	60.96	2.76^**	.124
Lenis /t/	72.22	64.91	1.79	.056
Tense /k/	78.70	64.47	3.82^***	.213
Tense /p/	85.19	67.76	4.36^***	.261
Tense /t/	80.56	67.76	3.48^**	.183
Initial placement	75.00	63.16	3.46^**	.182
Medial placement	77.47	69.05	3.44^**	.180
Accuracy overall	76.23	66.11	3.84^***	.214

p < .05; ^**p < .01; ^***p < .001.

Note. A Bonferroni correction of α = .05/12 yields α = .004. Those comparisons that are significantly different at α = .004 are shown in boldface.

Participants who are fluent in a language with an aspirated versus non-aspirated distinction in stop consonants (e.g., Armenian).

Participants who are not fluent in a language with an aspirated versus non-aspirated distinction.

Is there a broad-based bilingual advantage?

To investigate whether there is a general benefit for bilingualism above and beyond the specific effect for the languages that distinguish between aspirated and non-aspirated stop consonants (i.e., Armenian, Thai), we excluded data from participants whose languages make that distinction. We compared monolingual (n = 14), bilingual (n = 14), and trilingual (n = 10) individuals. We found no significant difference between monolingual, bilingual, or trilingual participants on their performance distinguishing Korean stop consonants, F(2, 35) = 2.33, p = .112, η_p² = .118.

Trilingual comparisons

We found no statistically significant difference between trilingual individuals without an aspirated versus non-aspirated distinction (M = 70.28, SD = 10.36; n = 10) and English monolinguals (M = 66.47, SD = 9.44; n = 14), t(22) = .936, p = .359, r_pb² = .038. In contrast, the trilingual individuals with an aspirated versus non aspirated distinction (M = 84.03, SD = 3.06; n = 4) scored higher than those trilingual individuals who do not have that distinction, equal variances not assumed, t(11.685) = 3.80, p = .003, r_pb² = .352.

Number of fluent current languages

The number of languages in which the participant was currently fluent was positively correlated with accuracy on the Korean phoneme discrimination task, r = .294, p = .028, N = 56. However, excluding languages that distinguish between aspirated versus non-aspirated stop consonants, the correlation becomes not statistically significant, r = .142, p = .395, N = 38.

Secondary analysis: phoneme position, age, and gender

We found no significant difference between Condition 1 (who first were presented with initially placed phonemes) and Condition 2 (medial position phoneme trials first) overall and on any of the phoneme types tested (n = 28 in each condition; all p values > .1). This indicates that the order of initial or medial presentation of the phonemes had no effect. In addition, neither age (r = –.055, p = .688) nor gender (r = –.051, p = .709) was associated with accuracy discriminating Korean phonemes (N = 56).

Discussion

We conclude that in the case of this study’s target phonemes that there is little or no evidence for a general bilingual or trilingual advantage in the ability to distinguish among the phonemes of an unfamiliar language. Most or all of the advantage is specific to experience with languages that are similar phonologically to the target language contrasts (such as with Armenian, with Korean as the target). In this case, fluency in languages that have an aspirated versus non-aspirated distinction gives a specific advantage when perceiving similar distinctions in an unfamiliar language. We found that both early non-English exposure and current fluency in a non-English language was associated with better Korean stop consonant discrimination, but only when those non-English languages had an aspirated versus non-aspirated stop consonant distinction. These results, in sum, support a narrow L1/L2 transfer advantage in the discrimination of phonemes.

Given the possible limitation of a lack of statistical power due to sample sizes, we also looked at the direction of the effects regardless of p values. Even when results were not statistically significant, we found no consistent direction of effect for a broad-based bilingual advantage, which suggests that small effects would not necessarily become statistically significant with more power (for one example see Figure 1: Spanish–English bilinguals versus monolinguals). Similarly, previous research has failed to find a consistent broad-based bilingual advantage (Beach et al., 2001; Davine, Tucker, & Lambert, 1971; Lambert & MacNamara, 1969; Werker, 1986). Enomoto’s (1994) conclusion that there is a general broad-based advantage could be an exception, but that particular study had potential sample selection problems (7 of the 10 participants were selected from Enomoto’s own class) and a small sample size compared to the other studies.

We found support for a narrow transfer view of the transmission of phoneme discrimination abilities from L1/L2 to L3. In other words, the set of phoneme contrasts within an individual’s L1 will help that individual distinguish between similar contrasts in other languages. The fact that we found no broad transfer advantage suggests that the bilingual and trilingual individuals did not necessarily have better phoneme discrimination than monolinguals, and the hypothesis that flexibility in categorizing phonemes could explain a broad advantage in discrimination did not hold.

There may be moderators and mediators that merit further investigation. Such variables could be anything from the nature of the task or the specific language features, to individual differences in participant samples. All of these factors are likely going to vary from study to study, so they may be useful in future research to uncover mechanisms and boundary conditions.

Our findings could cast some new light on Beach et al.’s (2001) findings that Greek–English bilingual individuals had some advantage in perceiving some Thai phonemes. The researchers themselves suggested that it depended on how much the individual speakers emphasized the distinction in their language, which is congruent with what we found. Given our current findings, we would predict that if Beach et al. (2001) had been able to find another bilingual group who had even less phonemic similarities to Thai than Greek, then they would have found a reduced bilingual advantage in that group.

Our study adds to the limited number of articles on phoneme discrimination in unrelated languages by exploring different participant languages, more than one bilingual group, a trilingual group, and different target languages phonemes. Our main findings are important in that they can inform society about what strategies should be used to maximize new language learning ability in children and adults. Exposing children to many languages in order to increase general phonemic categorization flexibility may not be as effective as specific exposure to similar phonemic distinctions. However, the narrowly transferred bilingual advantages of phoneme discrimination we identified in this paper will help some language groups when learning a new language.

Our primary research question was to also investigate whether the mechanism behind a bilingual advantage is greater flexibility in changing phoneme category boundaries, or due to narrow L1/L2 language transfer, facilitated by exposure and practice of specific phoneme contrasts similar or identical to those in the target language. Our results did not suggest that phoneme distinction boundaries are flexible due to bilingual exposure. Instead, our results support a narrow transfer mechanism where the phoneme distinction boundaries may be relatively stable, and are set in a way that can allow for better discrimination of similar contrasts in unrelated languages.

Footnotes

Acknowledgements

We would like to thank Professor Sun-Ah Jun, PhD and Sahyang Kim, PhD, of the Department of Linguistics at UCLA for their help with developing the stimuli for this study, and members of the Linguistic Minority Development Research Laboratory at California State University, Northridge, who contributed to the project in a number of ways: Bertha Nash, Kyungwon Kang, Mariko Iwabuchi, Gloria Chanyang Lee, Jonathan Zeledon, and Martin Nolasco.

Funding

This work was supported in part by a National Science Foundation research award (BCS-0844106) to the second author, and by a National Science Foundation Graduate Research Fellowship awarded to the first author.

Author biographies

Lawrence Patihis, MA, is a National Science Foundation Graduate Research Fellow and doctoral student in the Department of Psychology and Social Behavior at the University of California, Irvine.

Janet S. Oh holds a PhD in Developmental Psychology from UCLA and is currently associate professor of Psychology at California State University, Northridge. She conducts research on the development of individuals who come from backgrounds in which a heritage language is spoken. Her research is currently supported by a National Science Foundation grant.

Tayopa Mogilner obtained an MA in cognitive science from the University of California, Irvine, in the Brains, Language, and Bilingualism laboratory led by Professor Mary Louise Kean. She is currently pursuing her Ph.D. in the Department of Cognitive Sciences at the University of California, Irvine.

References

T. K.

Knightly

L. M.

Jun

S.-A.

J. S.

(2002). Overhearing a language during childhood. Psychological Science, 13, 238–243.

Beach

E. F.

Burnham

Kitamura

(2001). Bilingualism and the relationship between perception and production: Greek/English bilinguals and Thai bilabial stops. International Journal of Bilingualism, 5, 221–235.

Bialystok

Craik

F. I. M.

Green

D. W.

Gollan

T. H.

(2009). Bilingual minds. Psychological Science in the Public Interest, 10, 89–129.

Bialystok

Luk

Kwan

(2005). Bilingualism, biliteracy, and learning to read: Interactions among languages and writing systems. Scientific Studies of Reading, 9, 43–61.

Bild

E. R.

Swain

(1989). Minority language students in a French immersion programme: Their French proficiency. Journal of Multilingual and Multicultural Development, 10, 255–274.

Brohy

(2001). Generic and/or specific advantages of bilingualism in a dynamic plurilingual situation: The case of French as official L3 in the school of Samedan (Switzerland). International Journal of Bilingual Education and Bilingualism, 4, 39–49.

Cenoz

(2003). The additive effect of bilingualism on third language acquisition: A review. International Journal of Bilingualism, 7, 71–83.

Cohen

S. P.

Tucker

G. R.

Lambert

W. E.

(1967). The comparative skills of monolinguals and bilinguals in perceiving phoneme sequences. Language and Speech, 10, 159–168.

Davine

Tucker

Lambert

(1971). The perception of phoneme sequences by monolingual and bilingual elementary school children. Canadian Journal of Behavioural Science, 3, 72–76.

10.

Enomoto

(1994). L2 perception acquisition: The effect of multilingual linguistic experience on the perception of a less novel contrast. Edinburgh Working Papers in Applied Linguistics, 5, 15–29.

11.

Gallardo del Puerto

(2007). Is L3 phonological competence affected by the learner’s level of bilingualism? International Journal of Multilingualism, 4, 1–16.

12.

Harris

J. D.

(1952). Remarks on the determination of a differential threshold by the so-called ABX technique. Journal of the Acoustical Society of America, 24, 417.

13.

Jaspaert

Lemmens.

(1990). Linguistic evaluation of Dutch as a third language. In Byram

Leman

(Eds.), Bicultural and trilingual education: The Foyer model in Brussels (pp. 30–56). Clevedon, UK: Multilingual Matters.

14.

Johnson

J. S.

Newport

E. L.

(1989). Critical period effects in second language learning: The influence of maturational state on the acquisition of English as a second language. Cognitive Psychology, 21, 60–99.

15.

Kaushanskaya

Marian

(2009). The bilingual advantage in novel word learning. Psychonomic Bulletin & Review, 16, 705–710.

16.

Lambert

MacNamara

(1969). Some cognitive consequences of following a first-grade curriculum in a second language. Journal of Educational Psychology, 60, 86–96.

17.

Liberman

A. M.

Harris

K. S.

Hoffman

H. S.

Griffith

B. C.

(1957). The discrimination of speech sounds within and across phoneme boundaries. Journal of Experimental Psychology, 54, 358–368.

18.

Long

M. H.

(1990). Maturational constraints on language development. Studies in Second Language Acquisition, 12, 251–285.

19.

J. S.

T. K.

Jun

S.-A.

(2010). Early childhood language memory in the speech perception of international adoptees. Journal of Child Language, 37, 1123–1132.

20.

J. S.

Jun

S.-A.

Knightly

L. M.

T. K.

(2003). Holding on to childhood language memory. Cognition, 86, B53–B64.

21.

Sanders

Meijers

(1995). English as L3 in the elementary school. ITL: Review of Applied Linguistics, 107 / 108, 59–78.

22.

Sanz

(2000). Bilingual education enhances third language acquisition: Evidence from Catalonia. Applied Psycholinguistics, 21, 23–44.

23.

Schneider

Eschmann

Zuccolotto

(2002). E-Prime user’s guide. Pittsburgh, PA: Psychology Software Tools, Inc.

24.

Thomas

(1988). The role played by metalinguistic awareness in second and third language learning. Journal of Multilingual and Multicultural Development, 9, 235–246.

25.

Vaux

(1998). The Phonology of Armenian. Oxford, UK: Oxford University Press.

26.

Werker

(1986). The effects of multilingualism on phonetic flexibility. Applied Psycholinguistics, 7, 141–156.