Bilingual and Monolingual Infant Word Learning (2014) PDF

International Journal of Behavioral Development http://jbd.sagepub.com/ You sound like Mommy: Bilingual and monolingual infants learn words best from speakers typical of their language environments Christoph...

International Journal of Behavioral Development http://jbd.sagepub.com/ You sound like Mommy: Bilingual and monolingual infants learn words best from speakers typical of their language environments Christopher Fennell and Krista Byers-Heinlein International Journal of Behavioral Development 2014 38: 309 DOI: 10.1177/0165025414530631 The online version of this article can be found at: http://jbd.sagepub.com/content/38/4/309 Published by: http://www.sagepublications.com On behalf of: International Society for the Study of Behavioral Development Additional services and information for International Journal of Behavioral Development can be found at: Email Alerts: http://jbd.sagepub.com/cgi/alerts Subscriptions: http://jbd.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav >> Version of Record - Jun 4, 2014 What is This? Downloaded from jbd.sagepub.com at CONCORDIA UNIV LIBRARY on June 4, 2014 Special section: Language development in multilingual environments International Journal of Behavioral Development You sound like Mommy: Bilingual and 2014, Vol. 38(4) 309–316 ª The Author(s) 2014 Reprints and permissions: monolingual infants learn words best sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0165025414530631 from speakers typical of their ijbd.sagepub.com language environments Christopher Fennell1 and Krista Byers-Heinlein2 Abstract Previous research indicates that monolingual infants have difficulty learning minimal pairs (i.e., words differing by one phoneme) produced by a speaker uncharacteristic of their language environment and that bilinguals might share this difficulty. To clearly reveal infants’ underlying phonological representations, we minimized task demands by embedding target words in naming phrases, using a fully crossed, between-subjects experimental design. We tested 17-month-old French-English bilinguals’ (N ¼ 30) and English monolinguals’ (N ¼ 31) learning of a minimal pair (/kεm/ – /gεm/) produced by an adult bilingual or monolingual. Infants learned the minimal pair only when the speaker matched their language environment. This vulnerability to subtle changes in word pronunciation reveals that neither monolingual nor bilingual 17-month-olds possess fully generalizable phonological representations. Keywords bilingualism, infancy, phonological development, word learning Infants are surprisingly precocious in their word learning, showing monolingual (Mackenzie, Curtin, & Graham, 2012; Werker et al., evidence of word knowledge from as young as age 6 months (Ber- 1998) and bilingual (Byers-Heinlein, Fennell, & Werker, 2012) gelson & Swingley, 2013), and this early lexical acquisition appears infants successfully learn dissimilar-sounding words (e.g. lif – to proceed in tandem with the development of related perceptual neem) in this procedure. When taught minimal pair words that dif- abilities. Each instance of a word is unique (e.g. differing speaker fer by one phoneme (e.g. bin – din), which directly tests infants’ use characteristics, emotional valence), leading to different realizations of detailed phonetic information in word learning, monolingual of its constituent sounds. This variability, however, does not affect infants typically do not succeed until the age of 17 months (e.g. a word’s meaning and should be downplayed in infants’ represen- Werker, Fennell, Corcoran, & Stager, 2002). Bilingual infants may tations of a word’s form. Instead, learners need to selectively focus succeed at the same age as monolinguals (Mattock, Polka, on differences that are phonemic in the native-language, for exam- Rvachew, & Krehm, 2010), but other research suggests that this ple the distinction between /b/ and /g/ that gives ‘‘boat’’ and group may have difficulty with minimal pairs up until 20 months ‘‘goat’’ different meanings in English. During the first year of life, of age (Fennell, Byers-Heinlein, & Werker, 2007). This suggests infants’ ability to discriminate phonetic differences that are mean- that young infants do not always apply their phonetic sensitivities ingful (i.e., phonemic) in the native language is maintained or to word learning. enhanced (e.g. Kuhl et al., 2006; Polka, Colontonio, & Sundara, Significant theoretical work has attempted to reconcile young 2001), while their ability to discriminate phonetic differences that infants’ ability to perceive relevant phonetic differences with their are irrelevant in the native language attenuates (e.g. Werker & apparent inability to use this information in some tasks. One pro- Tees, 1984). When can infants apply these phonetic sensitivities posal is that word learning initially entails high task demands (Wer- to word learning in a mature way, by selectively attending to, ker & Fennell, 2004). Infants must concurrently learn about objects, encoding, and retrieving information that is phonemic? The words, and the connection between them, all while appropriately answer to this question is far from straight-forward: infants show storing and retrieving the rich phonetic detail of the target words. complex patterns of success and failure across a variety of ages This might be especially difficult for novice word learners. Thus, and tasks (Werker & Curtin, 2005). task demands and infants’ developmental level together influence One important method used to investigate infants’ use of pho- netic information in word learning is the Switch task (Werker, Cohen, Lloyd, Casasola, & Stager, 1998). In the classic version 1 University of Ottawa, Canada of the task, infants are habituated to two novel word–object pair- 2 Concordia University, Canada ings. At test, infants experience a Same trial where a habituated pairing is presented, and a Switch trial where a pairing violation Corresponding author: is presented (e.g. Object A with Word B). Infants demonstrate suc- Christopher Fennell, University of Ottawa, 136 Jean-Jacques-Lussier, cessful word learning if they look longer to the novel Switch trial Ottawa, K1N6N5, Canada. than to the familiar Same trial. Around 12–14 months of age, Email: [email protected] Downloaded from jbd.sagepub.com at CONCORDIA UNIV LIBRARY on June 4, 2014 310 International Journal of Behavioral Development 38(4) whether the phonetic detail distinguishing native phonemes will be their everyday language environments, but that monolinguals could accessed in a particular task (Curtin, Byers-Heinlein, & Werker, not efficiently process the unfamiliar variation in the targets 2011; Werker & Curtin, 2005). Indeed, 14-month-old infants suc- because they are not normally exposed to such variability. How- cessfully learn minimal pair words when supported by contextual ever, Mattock et al. did not test bilingual infants on monolingual- cues that reduce task demands, such as familiarizing infants with only stimuli. As such, while it appears that monolingual infants the target objects prior to word learning (Fennell, 2012), adding were not using abstracted phonemes to guide their word learning, indexical variability while holding the relevant phonetic contrast the claim of enhanced flexibility in bilinguals’ phonological repre- stable (Rost & McMurray, 2010), and training on lexical contexts sentations remains an open question. that highlight the minimal pair (Thiessen & Yee, 2010). Indeed, the results of another Switch study indicate that bilin- One task demand inherent to the classic Switch procedure itself gual infants’ word learning might similarly be disrupted when they is its lack of referential cues. Referential information signals to the are presented with phonetic information that does not match their infant that the target word should not only be associated with a par- language-learning environment. Using the exact same minimal pair ticular object, but that it should be associated with its abstract con- stimuli (/bI/ - /dI/) that English monolinguals learn by age 17 ceptual representation (Waxman & Gelman, 2009). In the standard months (Werker et al., 2002), Fennell et al. (2007) found that bilin- version of the Switch task, to-be-learned words are presented as iso- gual infants learning English and another language did not accu- lated tokens (e.g. ‘‘Bin!’’), and thus are stripped of the rich referen- rately learn the minimal pair until 20 months. Fennell and tial information characteristic of natural language. One way to colleagues attributed the difference in results to weaker phoneme enhance the referential nature of the target words is to embed them representations in bilinguals, perhaps due to their reported difficul- in naming phrases, which is more typical of parental speech, and ties in early phonetic perception (for reviews, see Byers-Heinlein & provides syntactic information that indicates the target word refers Fennell, 2013; Werker, 2012), or to less overall experience with to the object (i.e., it is a noun). When such referential information is each language in comparison to monolinguals (Curtin et al., provided (e.g. ‘‘Look, it’s the bin!’’), 14-month-old infants success- 2011; Werker, Byers-Heinlein, & Fennell, 2009). However, an fully learn minimal pairs (e.g. Fennell & Waxman, 2010). English monolingual produced these stimuli, which specifically led In sum, 14-month-old infants only learn minimal pair words to monolingual success and bilingual failure at 17 months. Thus, when task demands are alleviated, while older infants successfully another interpretation of these results is that bilingual infants’ dif- learn minimal pair words across many situations. What does this ficulty was not related to their bilingualism per se, but rather to developmental pattern indicate about infants’ underlying represen- an issue common to monolinguals and bilinguals: vulnerability to tations? Under the PRIMIR framework of infant speech perception less familiar pronunciations of native phonemes. and word learning (Processing Rich Information from Multidimen- Within a given language, what kinds of pronunciations would be sional Interactive Representations; Curtin et al., 2011; Werker & familiar to monolingual as compared to bilingual infants? Monolin- Curtin, 2005), minimal pair word learning is initially supported gual infants are usually raised by parents who are monolingual, by perceptually-based phonetic categories. However, the phonetic while bilingual infants are often raised by at least one parent who information critical to phoneme categories is stored and processed is bilingual (Byers-Heinlein, 2013). The majority of bilinguals who together with other types of acoustic information irrelevant to word acquire a second language in adulthood speak that language with an meaning (e.g. indexical information like speaker gender). Young accent (Piske, MacKay, & Flege, 2001), but even bilingual adults word learners do not always attend to the relevant information, and who acquired both languages early in life produce phonemes that thus can fail in more demanding minimal pair word learning tasks. differ in small ways from monolingual speakers (e.g. mean voice With experience, infants develop abstract phonemic categories, onset time, greater acoustic variation; Antoniou, Best, Tyler, & which summarize across phonetic variation and direct infants’ Kroos, 2010; Flege & Eefting, 1987; Hazan & Boulakia, 1993). attention to relevant phonological features of a word. PRIMIR pre- One published study specifically examined productions of bilingual dicts that, once infants have developed phonemes, they should mothers. Bosch and Ramon-Casas (2011) reported that even when robustly succeed in minimal pair word learning tasks as long as the the mothers were highly proficient Spanish-Catalan bilinguals, their critical phonetic variation falls within the acceptable range of their vowel productions differed as a function of their own early lan- native phoneme categories. guage environments. This slight ‘‘accent’’ in simultaneous and Surprisingly, infants sometimes fail to learn minimal pairs even early bilinguals has long been attested in the literature, including for at 17 months, an age at which it has been hypothesized that they the Canadian French–English bilingual population involved in the have developed phonemes (Werker & Curtin, 2005). Using the clas- current research (Caramazza, Yeni-Komshian, Zurif, & Carbone, sic Switch task, Mattock et al. (2010) tested monolingual (English 1973; MacLeod & Stoel-Gammon, 2009; Sundara, Polka, & Baum, and French) and bilingual (English–French) infants of 17 months on 2006). Together, these studies suggest that there might be systema- three realizations of a minimal pair valid in both English and French tic differences in the adult pronunciations typical in monolingual (/bos/ – /gos/): monolingual English, monolingual French, and and bilingual infants’ early language environments, which could bilingual productions. These latter tokens were produced by a bilin- in turn affect their ability process phonetic information from differ- gual speaker, and included a mixture of English- and French- ent types of speakers. pronounced tokens. Infants succeeded only when the stimuli The aim of the current study was to better understand infants’ matched their language background: bilingual infants succeeded use of phonetic information during word learning at 17 months. Our on the bilingual stimuli, French infants on the French stimuli, and study therefore tested monolingual and bilingual infants on both English infants on the English stimuli. English and French monolin- monolingual- and bilingual-produced target words. This fully- gual infants failed on the bilingual stimuli, and French infants failed crossed design has not been undertaken before. Further, our design on the English stimuli. Mattock and colleagues hypothesized that aims to put monolingual and bilingual infants on equal footing by bilingual infants could flexibly process the more variable set of embedding the tokens in language-specific naming phrases that will bilingual tokens because they experience phonetic variability in clarify language information for the infants (e.g. placing the target Downloaded from jbd.sagepub.com at CONCORDIA UNIV LIBRARY on June 4, 2014 Fennell and Byers-Heinlein 311 Table 1. Mean acoustic measurements of the target consonant contrast (voice onset time) and vowels (formants) present in the /gεm/ and /kεm/ tokens with standard deviation in parentheses. Voice onset time (ms) First formant at midpoint (Hz) Second formant at midpoint (Hz) Speaker and language /gεm/ /kεm/ /gεm/ /kεm/ /gεm/ /kεm/ Monolingual English 14.33 (4.97) 82.89 (22.52) 852.33 (86.83) 970.22 (135.25) 1526.67 (198.37) 1575.78 (255.79) Bilingual English 20.38 (4.78) 96.38 (20.46) 739.12 (48.18) 787.50 (85.26) 2041.50 (202.20) 2016.62 (132.66) Bilingual French 0.14 (29.97) 74.71 (11.5) 741.71 (70.43) 793.71 (65.22) 2094.57 (186.93) 2241.57 (69.35) Note. Each measurement is based on the nine auditory tokens of each target word presented in habituation and test trials (per speaker condition). in an English sentence highlights that the novel word is English). 95% exposure to English, as reported by parents. Sixteen infants Past research has shown that naming phrases facilitated monolin- (seven females) heard English bilingual tokens and 15 infants (eight gual infants’ word learning in the Switch task (Fennell & Waxman, females) heard English monolingual tokens. An additional 10 2010). This clarification might be particularly important for bilin- infants were tested but not included in the analyses due to fussiness gual infants, who could need to determine a word’s language before (9; 8 in the bilingual token condition) or being off-camera at test (1 appropriately interpreting phonetic information (see Gonzalez & in the bilingual token condition). Lotto, 2013, for related adult work). If infants have developed sta- ble, abstract phonemes by this age (Curtin et al., 2011; Werker & Curtin, 2005), then minimal pair word learning should be reason- Auditory stimuli ably robust to non-criterial phonetic variation, such as the slight dif- Auditory stimuli consisted of three novel nonsense words presented ferences in the phoneme productions of adult bilingual and in the context of the seven different naming phrases used in Fennell monolingual speakers. However, if infants’ representations are still and Waxman (2010) or their French equivalents. Three versions of immature at 17 months, they might only succeed when tokens the stimulus set were created: bilingual tokens recorded from a match the prevalent phonetic characteristics of their language- bilingual speaker in French and English, and monolingual tokens learning environment. A final possibility is that our study will recorded from a monolingual speaker in English. The words used reveal overall differences in monolinguals’ and bilinguals’ ability during the habituation and test phases formed a minimal pair differ- to learn minimal pair words, with bilinguals either showing greater ing only in the voicing of the initial velar consonant, /kεm/ and / flexibility (Mattock et al., 2010) or greater difficulty than monolin- gεm/. These words were chosen because they violate no French guals (Fennell et al., 2007). or English phonotactic rules and all constituent phonemes are pro- duced in a similar manner across the two languages, although there are differences in voice onset time (VOT) for the velars across the Method languages. The word /nib/ was used during the pre- and post-test trials, and was chosen because all of its phonemes differ from the Participants target words. Infants included in the study were, according to parental report, Stimuli were initially recorded in a soundproof booth, and edited typically-developing, without any apparent health or hearing prob- using the PRAAT computer program (Boersma & Weenik, 2012). lems, and at least 37 weeks gestation. Infants were from the metro- Matched kem, gem, and neeb versions of each naming phrase were politan area of Ottawa, Canada. Within each language group, created by excising the /nib/ tokens from the /nib/ sentences (e.g. infants were randomly assigned to speaker condition (bilingual or ‘‘Look at the ’’), and splicing /gεm/ and /kεm/ tokens from monolingual speaker). the corresponding sentences into their place. This strategy ensured that infants could not use any slight sound differences in the naming Bilingual infants. Thirty bilingual infants successfully completed phrases to help them discriminate the target words. Each trial the study (mean age ¼ 17.48 months; range ¼ 16.13–18.62 included nine instances of the novel word: an isolated token at the months). Infants were exposed to English and French from birth and beginning, seven instances within naming phrases, and another iso- had a maximum of 80% exposure to one language and a minimum lated token at the end. See Table 1 for acoustic measurements. of 20% exposure to the other, with 50% mean exposure to English and 50% to French, as assessed by the Language Exposure Ques- Monolingual tokens. A native-English adult female who grew up in tionnaire (Bosch & Sebastián-Gallés, 2001). Sixteen infants the local Ottawa area produced the monolingual tokens. As with the (7 females) heard bilingual tokens of the target words (half the majority of Canadians, she was exposed to French throughout her infants in English, half in French) and 14 infants (9 females) heard schooling; however, she had attended only English schools, did not English monolingual tokens. An additional 14 infants were tested use French on a daily basis, and had always lived in an English but not included in the analyses due to fussiness (10; 3 in the bilin- monolingual household. The velar consonants conformed to gual token condition), parental interference (3; 1 in the bilingual English VOT measures (Lisker & Abramson, 1967). For the /ε/ token condition), or distraction during testing (1 infant dropped his vowel common to both words, the first and second formants were toy in the bilingual token condition). closer to each other than reports of female North American English speakers, but were similar to male values (Hillenbrand, Getty, English monolingual infants. Thirty-one monolingual infants suc- Clark, & Wheeler, 1995). However, her vowels were perceptible cessfully completed the study (mean age ¼ 17.51 months; as /ε/ and the formant values did not differ across /gεm/ and /kεm/ range ¼ 16.30–18.38 months). All monolinguals had greater than tokens. Downloaded from jbd.sagepub.com at CONCORDIA UNIV LIBRARY on June 4, 2014 312 International Journal of Behavioral Development 38(4) Test Pretest Habituation Same Switch Post-test /nib/ /kεm/ /gεm/ /kεm/ /gεm/ /nib/ Figure 1. Outline of the experimental procedure and the audio-visual stimuli. The auditory tokens were presented in naming phrases. Bilingual tokens. The bilingual tokens were produced by an adult loudspeakers, located below the monitor. The lens of the digital female raised from birth in a bilingual French–English household video camera peeked out of a hole in the cloth 10 cm below the in a Canadian city where both French and English are spoken in monitor. A Macbook Pro ran the Habit X computer program, which everyday life. She had both French and English schooling, and used ordered stimuli presentation and computed looking time data. In a French and English regularly in her daily life. Although she was not nearby testing room, the experimenter monitored infants’ looking originally from the same region as the infants tested, she had spent time via a closed circuit television system. The researcher coded the previous 4 years in the Ottawa area and possessed a generic infant looking by pressing a designated key whenever infants Canadian accent in both languages. This speaker produced two sets looked at the screen. The testing setup for the monolingual condi- of stimuli: one with French naming phrases and one with English tion was highly similar and only differences are highlighted here. naming phrases. Her French productions of /k/ were similar in VOT Testing took place in a 2.89 m by 1.30 m quiet room. A NEC Duo- to previous reports of Canadian French (Ryalls, Cliché, Fortier- com LT280 projector presented the images on to a SmartBoard Blanc, Coulombe, & Prud’hommeaux, 1997). Her French produc- screen (1.62 m diagonal) 1.5 m in front of infants. The digital video tions of /g/ had less prevoicing than some reports (Ryalls et al., camera was hidden under a table draped with black cloth below the 1997), but were similar to others (Caramazza & Yeni-Komshian, screen, with the lens peeking out of a hole in the cloth 30 cm below 1974). Her English productions of the /k/ and /g/ were similar in the screen. VOT to the monolingual’s productions (as were the standard devia- tions), although she had longer lag for both consonants. Her French and English productions of /ε/ were typical (Hillenbrand et al., Procedure 1995; Martin, 2002) and she had less variability than the monolin- After participants arrived, the experimenter explained the proce- gual speaker. dure, obtained consent, and gave parents the Language Exposure Mattock et al. (2010) had created a bilingual condition by inter- Questionnaire. In the testing room, infants sat on parents’ laps fac- spersing a bilingual’s typical English and typical French produc- ing the screen. Parents listened to female vocal music over head- tions of target words, whereas we had two separate language sets. phones to mask the stimuli. From the observation room, the Our bilingual condition was solely based on the speaker’s back- experimenter initiated the first trial by pressing a designated com- ground to explore whether infants were sensitive to the subtly dif- puter key when infants looked at the screen. An animated oval pre- ferent pronunciations of a bilingual versus a monolingual speaker ceded the first trial, and all subsequent trials, to draw infants’ (see Bosch & Ramon-Casas, 2011). attention to the display. Monolingual and bilingual versions of the stimuli were vali- Figure 1 outlines the procedure. Infants were tested using either dated by having three adult monolingual speakers of each language the bilingual or monolingual tokens. For the bilingual token condi- and three adult bilingual speakers identify the target phonemes /g/ tion, all English monolinguals heard English tokens, whereas half and /k/. All adults correctly identified all target consonants. the bilingual infants heard French tokens and half heard English tokens. Habituation trials occurred in a quasi-random order such that each trial type was presented twice within a 4-trial block, with Visual stimuli no more than three consecutive trials of the same type. Trial length Figure 1 shows the three distinctive objects used during testing. The was fixed at 20 s. When average looking time across a four-trial crown and molecule objects moved back and forth across the screen block decreased to 65%, the habituation phase ended. Infants com- at a slow and constant speed. The water wheel remained in the cen- pleted a minimum of 8 and a maximum of 24 habituation trials. tre of the screen with its arms rotating. Immediately after habituation, infants moved to the test phase, where they viewed one Same and one Switch trial in one of eight testing orders that counterbalanced trial order (Same–Switch/ Apparatus Switch–Same) and the particular pairings presented. The monolingual and bilingual token conditions were tested in dif- Coding. Videos of all test trials were coded frame-by-frame ferent locations at the same university due to a laboratory move. For (1 frame ¼ 30 ms) by an experienced coder, blind to condition. the bilingual token condition, testing took place in a 2.38 m by A second coder then recoded 25% of the videos frame-by-frame, 1.82 m room, dimly lit by a 60 W lamp situated 80 cm to the left with high reliability (r ¼.997, p <.001). These frame-by-frame of the infant. Infants sat on parents’ laps facing a 21-inch video measurements were used for all analyses, except for one infant monitor approximately 85 cm away, which was surrounded by for whom online coding values were used because the video was black cloth. Audio stimuli played at 65 dB, þ/ 5 dB, over two corrupted. Downloaded from jbd.sagepub.com at CONCORDIA UNIV LIBRARY on June 4, 2014 Fennell and Byers-Heinlein 313 Table 2. Mean looking times in seconds to habituation and test trials by group. Standard deviations are indicated in parentheses. Monolingual infants Bilingual infants Habituation and test trials Monolingual tokens: Match Bilingual tokens: Mismatch Monolingual tokens: Mismatch Bilingual tokens: Match Pre-test trial 18.32 (2.32) 18.29 (3.48) 15.71 (3.96) 16.83 (4.40) First habituation block 17.09 (2.35) 16.23 (2.10) 16.02 (3.18) 15.24 (3.23) Last habituation block 8.94 (2.99) 9.56 (2.68) 8.57 (2.94) 9.01 (3.33) Same test trial 9.04 (4.94) 13.66 (5.23) 8.60 (4.76) 8.13 (3.29) Switch test trial 11.80 (3.71) 14.75 (3.70) 9.78 (3.51) 12.49 (5.58) Post-test trial 15.98 (4.06) 19.33 (1.38) 14.06 (5.18) 18.23 (3.11) Note. Monolingual infants: n ¼ 31 (15 in monolingual token condition, 16 in bilingual token condition). Bilingual infants: n ¼ 30 (14 in monolingual token condition, 16 in bilingual token condition). All trials were a maximum of 20 seconds. Table 3. Main ANOVA. Nevertheless, to explore the significant interaction, we examined infants’ test trial performance separately for each possible group- Source F p partial 2 ing, using the Bonferroni corrected alpha of.012. Infants who heard Between subjects stimuli from a speaker that matched their language learning envi- Stimuli match (M): Yes vs. No 1.86.18.03 ronment looked longer during the Switch than during the Same Language environment (L): Mono- vs. Bilingual 6.86.01.11 trial, showing that they detected the phoneme change; Monolin- ML 6.29.02.10 guals: t(14) ¼ 2.91, p ¼.011, d ¼ 0.63; Bilinguals: t(15) ¼ Within subjects 4.02, p ¼.001, d ¼ 0.95. However, infants who heard stimuli Test trial (T): Same vs. Switch 16.65.00.23 that did not match their language background looked equivalently TM 4.43.04.07 during both test trials; Monolinguals: t(15) ¼ 0.80, p ¼.44; Bilin- TL.54.47.01 guals: t(13) ¼ 1.06, p ¼.31 (see Figure 2). TLM.42.52.01 For between-subjects effects in the main ANOVA, there was no Note. n ¼ 61. The df for all statistics presented in the table are 1, 57. main effect of match; however, there was a main effect of language, moderated by a language by match interaction. Monolinguals looked longer to test trials overall, particularly in the condition where they heard bilingual-produced tokens. Results Finally, because monolingual-produced stimuli were only in Table 2 shows infants’ looking times to key trials. Preliminary anal- English, we investigated whether bilingual infants with more Eng- yses revealed no gender effects in any of the following statistical lish exposure outperformed those with less English exposure. We tests. Also, for bilingual infants in the bilingual token condition, split bilingual infants by language dominance: those with 50% or there were no effects related to whether they heard French versus greater exposure to English (n ¼ 6) and those with less (n ¼ 8). English stimuli. Our research question concerned whether infants’ A 2 (trial type: Same vs. Switch)  2 (English dominance: yes performance differed as a function of whether the stimuli matched vs. no) mixed ANOVA revealed a significant interaction between (e.g. monolingual infants hearing monolingual-produced tokens) or trial type and English dominance, F(1, 12) ¼ 4.98, p ¼.046, partial did not match (e.g. monolingual infants hearing bilingual-produced 2 ¼.29. No other effects were significant. English dominant tokens) their prevalent language experience. Infants in all conditions infants looked longer to the Switch trial (M ¼ 10.15 s; SD ¼ habituated to the stimuli, showing a significant decrease in attention 2.39) than to the Same trial (M ¼ 6.44 s; SD ¼ 3.11), whereas in the last habituation block compared to the first (ps <.001). The French dominant infants did not (Switch: M ¼ 9.51 s; SD ¼ 4.31; number of habituation trials across conditions was statistically equiv- Same: M ¼ 10.22 s; SD ¼ 5.31). However, due to low power engen- alent (ps >.10), thus any test trial differences cannot be due to differ- dered by the small subgroup sample sizes, the difference in looking ences in perceptual adaptation to auditory stimuli. Further, all groups times for the English-dominant subgroup did not achieve statistical of infants also showed significant recovery during the post-test as significance (p ¼.096). However, there was a significant positive compared to the last habituation block (ps <.002); infants were not correlation, r(12) ¼.57, p ¼.03, between amount of English expo- fatigued or generally disinterested in the task. sure from all sources (e.g. mother, father, grandparents, etc.) and A 2 (trial type: Same vs. Switch)  2 (stimuli match: yes vs. no) task performance (magnitude of the difference in looking time in  2 (infant language environment: bilingual vs. monolingual) favor of Switch), which further supports the above findings. Greater mixed ANOVA tested whether infants showed different looking exposure to English facilitated bilinguals’ detection of the phoneme times to the Same and Switch trials; see Table 3 for the ANOVA change in the monolingual English stimuli. results. We found a significant main effect of trial type, moderated To investigate whether this pattern might be explained by the by a significant interaction between trial type and stimuli match. nature of the English spoken by parents to their infants, we classi- However, there was no interaction between trial type and language fied each parent according to whether or not they grew up as Eng- environment and no three-way interaction between trial type, lan- lish monolinguals, regardless of their current language usage.1 We guage environment, and match. Thus, infants behaved differently expected that bilingual infants with more exposure to speech from across the test trials depending on whether the stimuli matched their these parents might more readily detect the phoneme changes in our learning environment or not, but this did not differ by language English monolingual stimuli. Indeed, there was a significant posi- background: monolinguals and bilinguals showed the same pattern. tive correlation between number of hours per week the infants Downloaded from jbd.sagepub.com at CONCORDIA UNIV LIBRARY on June 4, 2014 314 International Journal of Behavioral Development 38(4) Same Switch 18 16 * ** 14 12 Looking time in s 10 8 6 4 2 0 Monolingual tokens Bilingual tokens Monolingual tokens Bilingual tokens (Match) (Mismatch) (Mismatch) (Match) Monolingual infants Bilingual infants Figure 2. Infants’ mean looking times in seconds to the test trials across the conditions. Note. Monolingual infants: n ¼ 31 (15 in monolingual token condition, 16 in bilingual token condition). Bilingual infants: n ¼ 30 (14 in monolingual token condition, 16 in bilingual token condition). All trials were a maximum of 20 seconds. * p <.05; ** p <.01. listened to English produced by these parents and their task perfor- that neither group possessed fully generalizable phonemes. Further, mance, r(11) ¼.58, p ¼.04. Importantly, there was no relationship we can reject two hypotheses raised in the introduction: bilinguals between their task performance and the number of hours per week this age do not appear to possess more flexible phonological repre- they listened to English produced by parents raised bilingually or sentations than monolinguals (Mattock et al., 2010) and monolin- raised as French monolinguals, r(11) ¼ .05, p ¼.88. We also guals do not appear to have more solid representations than examined if exposure to bilingual speakers’ productions and to bilinguals (Fennell et al., 2007). monolingual speakers’ productions affected bilingual infants’ per- Our findings are consistent with previous results where mono- formance in the bilingual token condition. No significant effects lingual infants failed to learn minimal pair words whose produc- were found (all p values >.4). However, the lack of power due to tions did not match their language-learning environment (i.e., missing information complicated these analyses. Importantly, our French-learning infants failed with English-pronounced stimuli; assumption that bilingual infants were hearing bilingual speakers Mattock et al., 2010). The current study reveals that 17-month- was supported. Twenty-one of 25 bilinguals for whom we had this old infants are vulnerable to even more subtle phonetic differences: information were hearing speech from parents raised as bilinguals bilingual and monolingual productions within the same language. or from parents raised as monolinguals in the opposite language As monolingual infants tend to have monolingual parents and (e.g. hearing French from a parent raised as an English monolin- bilingual infants tend to have one or more bilingual parents gual). In other words, only four infants had parents adhering to the (Byers-Heinlein, 2013), our results are concordant with other evi- one parent-one language approach, wherein parents only address dence suggesting that infants’ early discrimination of phonemes their infant in the language in which they themselves were raised is tailored to the specific phonetic properties present in their care- as monolinguals. givers’ productions (Cristiá, 2012). The influence of language environment on infants’ performance was further highlighted by individual differences amongst the bilin- Discussion gual infants hearing English monolingual tokens: English-dominant This study investigated monolingual and bilingual 17-month-old bilinguals outperformed French-dominant bilinguals in this condi- infants’ ability to learn minimal pairs produced by an adult repre- tion. Additionally, there was a significant positive correlation sentative or unrepresentative of their learning environment. Placing between the amount of English infants heard from monolingual Eng- target words in naming phrases should minimize the cognitive load lish parents and their task performance, yet there was no significant in the task for both monolingual and bilingual infants, thus allowing correlation between their task performance and the amount of for a clearer picture of infants’ phonological representations. We English they heard from parents who were not raised as English found that all infants showed vulnerability to stimuli that did not monolinguals. Further evidence of the importance of environment- match their language-learning environment: monolingual infants speaker concordance in infant word learning was found via a succeeded with a monolingual speaker, bilingual infants with a re-analysis of 17-month-old bilinguals’ performance in Fennell bilingual speaker, but each group failed with the opposite speaker. et al. (2007), which also tested bilingual infants on English monolin- Seventeen-month-old infants’ overall difficulty with unfamiliar gual stimuli. Splitting infants into English-dominant (n ¼ 19) and productions of native phonemes in the word-learning task revealed other-dominant (n ¼ 18) groups, we found that the groups differed Downloaded from jbd.sagepub.com at CONCORDIA UNIV LIBRARY on June 4, 2014 Fennell and Byers-Heinlein 315 in their behaviour across Same and Switch trials, interaction: F(1, 35) development is the same. Bilingual infants are neither advantaged ¼ 4.22, p ¼.047, partial 2 ¼.11. English-dominant infants showed nor disadvantaged relative to monolinguals, and the inclusion of some evidence of learning the minimal pair, Same: M ¼ 6.2 s; these two groups illuminates phonological development across all Switch: M ¼ 7.8 s; t(18) ¼ 1.79, p ¼.09, while other-dominant infants. Our results further suggest that using identical stimuli with infants did not, Same: M ¼ 6.98 s; Switch: M ¼ 5.66 s; t(17) ¼ monolingual and bilingual infants can disadvantage one group or 1.18, p ¼.26. Once again, infants whose language environments bet- the other, a factor that should be considered in future studies. Mat- ter matched the stimuli were better able to learn minimal pair words. tock et al. stated that ‘‘the first steps to word learning are easier These findings have important ramifications for studies of bilingual when the phonetic shoes fit’’ (2010, p. 241). The current study infants’ speech perception abilities. While the role of language dom- clearly demonstrates just how tightly those shoes need to fit at 17 inance has been investigated in the context of some speech percep- months of age. tion tasks (Sebastián-Gallés & Bosch, 2002), many studies in the bilingual infant literature have not reported analyses with respect Acknowledgments to language dominance. Researchers should take both the production We would like to thank Monika Nazair, Meghan Spring, Tamara of their stimuli and infants’ language dominance into account when Hudon, and Thyra Driver for their work on this project. Finally, designing studies and interpreting results. Failing to do so may result we would like to all the parents and infants who participated. in conclusions that bilingual infants generally fail at a perceptual task when, in reality, certain sub-groups of bilinguals can succeed while Funding others are having difficulty. The research was funded by Natural Sciences and Engineering The current findings recall monolingual infants’ reported diffi- Research Council of Canada Grants to Christopher Fennell and culties in processing phonetic variation in word learning and recog- Krista Byers-Heinlein, and a Fonds de recherche du Québec - Société nition tasks (Rost & McMurray, 2010), particularly variation et culture grant to Krista Byers-Heinlein. associated with accented speech. While monolinguals can recog- nize familiar words in their own dialect at age 6 months (Bergelson Note & Swingley, 2012), they do not recognize familiar words pro- nounced in a different dialect until 19 months of age (Best, Tyler, 1. We are missing information regarding the linguistic background Gooding, Orlando, & Quann, 2009; Mulak, Best, Tyler, Kitamura, of one bilingual infant’s parents in the monolingual token con- & Irwin, 2013). Further, infants have difficulty learning a novel dition and four in the bilingual token condition. word produced in a foreign accent up until 30 months of age (Schmale, Hollich, & Seidl, 2011). Similar to the idea of generaliz- References able phonological representations (Werker & Curtin, 2005), Best Antoniou, M., Best, C. T., Tyler, M. D., & Kroos, C. (2011). et al. used the term phonological constancy to describe 19- Inter-language interference in VOT production by L2-dominant month-old infants’ ability to recognize a word’s identity across bilinguals: Asymmetries in phonetic code-switching. Journal of these natural phonetic variations. Our results add to the accruing Phonetics, 39, 558–570. doi:10.1016/j.wocn.2011.03.001 data that infants under 19 months do not possess this phonological Bergelson, E., & Swingley, D. (2012). At 6 to 9 months, human infants constancy. know the meanings of many common nouns. Proceedings of the An important question remains with respect to our findings: National Academy of Sciences of the USA, 109, 3253–3258. doi: what specific phonetic information contributed to the difficulties 10.1073/pnas.1113380109 for each infant group? What cues aided monolinguals but impaired Best, C. T., Tyler, M. D., Gooding, T. N., Orlando, C. B., & Quann, C. bilinguals, and vice-versa? Acoustic measurements of our stimuli A. (2009). Development of phonological constancy: Toddlers’ per- did not point to any obvious cause as the bilingual and monolingual ception of native- and Jamaican- accented words. Psychological speaker differed only slightly, but infants may be tuned to that fine Science, 20, 539–542. doi: 10.1111/j.1467-9280.2009.02327.x level. Perhaps infants’ difficulties stemmed from one or more Boersma, P., & Weenik, D. (2012). Praat: Doing phonetics by computer aspect of our stimuli that we were not able to measure, or a combi- [Computer program]. Version 5.1. Retrieved from http://www.fon. nation of differences. The discovery of which specific phonetic hum.uva.nl/praat/ cues aid monolinguals and bilinguals could be a rich area for future Bosch, L., & Ramon-Casas, M. (2011). Variability in vowel production study. One further possibility is that infants’ perception of speakers’ by bilingual speakers: Can input properties hinder the early stabili- monolingual and bilingual ‘‘accents’’ may have been exacerbated zation of contrastive categories? Journal of Phonetics, 39, 514–526. by large amount of phonetic information in the naming phrases. doi:10.1016/j.wocn.2011.02.001 Although we hypothesized that naming phrases would be facilita- Bosch, L., & Sebastián-Gallés, N. (2001). Evidence of early language tive, they might have provided infants with further evidence that the discrimination abilities in infants from bilingual environments. speaker was not typical of their language environments. A language Infancy, 2, 29–49. doi:10.1207/S15327078IN0201_3 clarification manipulation that does not involve naming phrases Byers-Heinlein, K. (2013). Parental language mixing: Its measurement could test this explanation. For example, one could include a train- and the relation of mixed input to young bilingual children’s voca- ing phase prior to the word-learning phase wherein infants hear a bulary size. Bilingualism: Language and Cognition, 16, 32–48. doi: few isolated familiar words in one of their languages, prior to their 10.1017/S1366728911000010 exposure to the novel minimal pair. Byers-Heinlein, K., & Fennell, C. T. (2013). Perceptual narrowing in The present study confirmed that both bilingual and monolin- the context of increased variation: Insights from bilingual infants. gual infants have difficulty processing novel minimal-pair words Developmental Psychobiology, 56, 274–291. doi:10.1002/dev. when listening to a speaker of their language who does not match 21167 their language-learning environment. Although the optimal stimuli Byers-Heinlein, K., Fennell, C. T., & Werker, J. F. (2012). The devel- differ for monolingual and bilingual infants, the general pattern of opment of associative word learning in monolingual and bilingual Downloaded from jbd.sagepub.com at CONCORDIA UNIV LIBRARY on June 4, 2014 316 International Journal of Behavioral Development 38(4) infants. Bilingualism: Language and Cognition, 16, 198–205. doi: but not 15-month-olds, identify familiar words spoken in a 10.1017/S1366728912000417 non-native regional accent. Child Development, 84, 2064–2078. Caramazza, A., & Yeni-Komshian, G. H. (1974). Voice onset time in doi:10.1111/cdev.12087 two French dialects. Journal of Phonetics, 2, 239–245. Piske, T., MacKay, I. R., & Flege, J. E. (2001). Factors affecting degree Caramazza, A., Yeni-Komshian, G. H., Zurif, E. B., & Carbone, E. of foreign accent in an L2: A review. Journal of Phonetics, 29, (1973). The acquisition of a new phonological contrast: The case 191–215. doi:10.1006/jpho.2001.0134 of stop consonants in French–English bilinguals. Journal of the Polka, L., Colantonio, C., & Sundara, M. (2001). A crosslanguage com- Acoustical Society of America, 54, 421–428. parison of / d / – / ð / perception: Evidence for a new developmental Cristià, A. (2011). Fine-grained variation in caregivers’ /s/ predicts pattern. Journal of the Acoustical Society of America, 109, their infants’ /s/ category. The Journal of the Acoustical Society 2190–2201. doi:10.1121/1.1362689 of America, 129, 3271–3280. Rost, G. C., & McMurray, B. (2010). Finding the signal by adding Curtin, S., Byers-Heinlein, K., & Werker, J. F. (2011). Bilingual begin- noise: The role of noncontrastive phonetic variability in early word nings as a lens for theory development: PRIMIR in focus. Journal of learning. Infancy, 15, 608–635. doi:10.1111/j.1532-7078.2010. Phonetics, 39, 492–504. doi:10.1016/j.wocn.2010.12.002 00033.x Fennell, C. T. (2012). Object familiarity enhances infants’ use of pho- Ryalls, J., Cliche, A., Fortier-Blanc, J., Coulombe, I., & Prud’hommeaux, netic detail in novel words. Infancy, 17, 339–353. doi:10.1111/j. A. (1997). Voice-onset time in younger and older French-speaking 1532-7078.2011.00080.x Canadians. Clinical Linguistics & Phonetics, 11, 205–212. Fennell, C. T., Byers-Heinlein, K., & Werker, J. F. (2007). Using speech Schmale, R., Hollich, G., & Seidl, A. (2011). Contending with foreign sounds to guide word learning: The case of bilingual infants. Child accent in early word learning. Journal of Child Language, 38, 1096. Development, 78, 1510–1525. doi:10.1111/j.1467-8624.2007.01080.x 1108. doi:10.1017/S0305000910000619 Fennell, C. T., & Waxman, S. R. (2010). What paradox? Referential Sebastián-Gallés, N., & Bosch, L. (2002). Building phonotactic knowl- cues allow for infant use of phonetic detail in word learning. Child edge in bilinguals: Role of early exposure. Journal of Experimental Development, 81, 1376–1383. Psychology: Human Perception and Performance, 28, 974–989. Flege, J. E., & Eefting, W. (1987). Production and perception of English doi:10.1037/0096-1523.28.4.974 stops by native Spanish speakers. Journal of Phonetics, 15, 67–83. Sundara, M., Polka, L., & Baum, S. (2006). Production of coronal stops Gonzales, K., & Lotto, A. J. (2013). A bafri, un pafri: Bilinguals’ pseudo- by simultaneous bilingual adults. Bilingualism: Language and Cog- word identifications support language-specific phonetic systems. Psy- nition, 9, 97–114. chological Science, 24, 2135–2142. doi:10.1177/0956797613486485 Thiessen, E., & Yee, M. (2010). Dogs, bogs, labs, and lads: What pho- Hazan, V. L., & Boulakia, G. (1993). Perception and production of a voi- nemic generalizations indicate about the nature of children’s early cing contrast by French–English bilinguals. Language and Speech, word-form representations. Child Development, 81, 1287–1303. 36, 17–38. Waxman, S. R., & Gelman, S. A. (2009). Early word-learning entails Hillenbrand, J., Getty, L. A., Clark, M. J., & Wheeler, K. (1995). reference, not merely associations. Trends in Cognitive Sciences, Acoustic characteristics of American English vowels. Journal of the 13, 258–263. doi:10.1016/j.tics.2009.03.006 Acoustical Society of America, 97, 3099–3111. Werker, J. F. (2012). Perceptual foundations of bilingual acquisition in Kuhl, P. K., Stevens, E., Hayashi, A., Deguchi, T., Kiritani, S., & Iver- infancy. Annals of the New York Academy of Sciences, 1251, 50–61. son, P. (2006). Infants show a facilitation effect for native language doi:10.1111/j.1749-6632.2012.06484.x phonetic perception between 6 and 12 months. Developmental Sci- Werker, J. F., Byers-Heinlein, K., & Fennell, C. T. (2009). Bilingual ence, 9, F13–F21. beginnings to learning words. Philosophical Transactions of the Lisker, L., & Abramson, A. S (1967). Some effects of context on voice Royal Society B: Biological Sciences, 364(1536), 3649–3663. doi: onset time in English stops. Language and Speech, 10, 1–28. 10.1098/rstb.2009.0105 Mackenzie, H., Curtin, S., & Graham, S. A. (2012). 12-month-olds’ pho- Werker, J. F., Cohen, L., Lloyd, V., Casasola, M., & Stager, C. (1998). notactic knowledge guides their word-object mappings. Child Devel- Acquisition of word-object associations by 14-month-old infants. opment, 83, 1129–1136. doi:10.1111/j.1467-8624.2012.01764.x Developmental Psychology, 34, 1289–1309. MacLeod, A. A. N., & Stoel-Gammon, C. (2009). The use of voice Werker, J. F., & Curtin, S. (2005). PRIMIR: A developmental frame- onset time by early bilinguals to distinguish homorganic stops in work of infant speech processing. Language Learning and Develop- Canadian English and Canadian French. Applied Psycholinguistics, ment, 1, 197–234. 30, 53–77. doi:10.1017/S0142716408090036 Werker, J. F., & Fennell, C. T. (2004). From listening to sounds to lis- Martin, P. (2002). La système vocalique du français du Québec de tening to words: Early steps in word learning. In D. G. Hall & S. R. l’acoustique à la phonologie [The vowel system of Québec French Waxman (Eds.), Weaving a lexicon (pp. 79–109). Cambridge, MA: from acoustics to phonology.] La Linguistique, 38, 71–88. MIT Press. Mattock, K., Polka, L., Rvachew, S., & Krehm, M. (2010). The first Werker, J. F., Fennell, C. T., Corcoran, K. M., & Stager, C. L. (2002). steps in word learning are easier when the shoes fit: Comparing Infants’ ability to learn phonetically similar words: Effects of age monolingual and bilingual infants. Developmental Science, 13, and vocabulary size. Infancy, 3, 1–30. 229–243. doi:10.1111/j.1467-7687.2009.00891.x Werker, J. F., & Tees, R. C. (1984). Cross-language speech perception: Mulak, K. E., Best, C. T., Tyler, M. D., Kitamura, C., & Irwin, J. Evidence for perceptual reorganization during the first year of life. (2013). Development of phonological constancy: 19-month-olds, Infant Behavior & Development, 7, 49–63. Downloaded from jbd.sagepub.com at CONCORDIA UNIV LIBRARY on June 4, 2014