Spanish–English speech perception in children and adults: Developmental trends

Brice, A., Gorman, B., & Leung, C. (2013). Spanish-English speech perception in children and adults: Developmental trends. Clinical Linguistics and Phonetics,27(3) 220-234, doi:10.3109/02699206.2012.757805 Abstract

This study explored the developmental trends and phonetic category formation in bilingual children and adults. Participants included 30 fluent Spanish–English bilingual children, aged 8–11, and bilingual adults, aged 18– 40. All completed gating tasks that incorporated code-mixed Spanish–English stimuli. There were significant differences in performance according to phonotactic construction of the stimuli, with fastest word recognition on words with voiceless initial consonants. Analysis of developmental trends revealed significant differences in children’s performance by grade level and fastest recognition on English voiceless initial consonants than Spanish voiceless initial consonants. Differences in voice onset time between English and Spanish may have contributed to quicker recognition of English voiceless consonants than Spanish voiceless consonants. It is also possible that increased exposure to both spoken and written English may account for faster recognition of English voiceless words than Spanish voiceless words. In conclusion, multiple factors may influence perception of a second language.

Keywords: phonetics, speech–speech production measurement, speech-multilingualism, language Introduction

Introduction

The study of speech perception skills and their critical contributions to overall linguistic development has long been of research interest (Jusczyk, 1997; Kuhl, 2000; Tsao, Liu, & Kuhl, 2004; Werker & Tees, 1999). Research on speech perception in bilingual or multilingual speakers, however, remains relatively limited. According to Leather (1999), “...the scientific study of structures and processes of speech in second [or multiple] language learning, once quite prominent, now seems comparatively underrepresented...” (p. 1).

Extant research has indicated striking differences in speech perception development between monolinguals and bilinguals (Fennel, Byers-Heinlein, & Werker, 2007). Some differences, for example, relate to the developmental timeline for establishing speech sound categories and crowding within phonetic space (Bosch & Sebastián-Gallés, 2003). Bilinguals are unique in that they activate and use phonetic detail relevant to the language of input of both languages at a given moment; thus,

they likely face increased computational and processing demands (e.g. Paradis & Genessee, 1996). In addition, it appears that second-language or multiple-language learners tend to map what they hear in their second language (L2/L3) onto the existing sound system of their first language (L1; Leather, 1999). Consequently, the study of bilingual speech recognition is largely determined by how the two languages interact with each other; specifically, whether there is transference of skill from one language to another, or interference between the languages. Questions related to the developmental progression of phonetic mapping, what sounds are mapped easily (i.e. transference) and what sounds are difficult to map (i.e. interference) warrant further investigation.

Clearly, enhanced understanding of bilinguals’ unique development is needed, which was the primary motivation of the current study. Such research may have important implications for our understanding of bilinguals’ development in speech perception, and subsequently, of numerous skills impacted by speech perception including phonological development (Escudero, 2007), literacy- related skills including phonological awareness (McBride-Chang, 1995) and lexical development (Fennel et al., 2007).

Infant and child speech perception research

Fennel et al. (2007) examined the development of speech sound categories in bilingual infants. They were interested in determining if bilingual infants demonstrated a temporary delay in the establishment of native speech sound categories as compared with monolinguals, as suggested previously (Paradis & Genessee, 1996). They found that the bilingual infants used phonetic details (e.g. “bih” vs. “dih”) to guide their word learning at 20 months of age, as compared with 17 months in monolinguals. They interpreted these findings to suggest that the tight phonetic space between sounds in bilingual children’s two systems may result in temporary perceptual confusion or interference. Consequently, they suggested that bilinguals needed more time than monolinguals to fully establish their speech sound categories in both languages, possibly due to limited cognitive resources.

Chiappe, Glaeser, and Ferko (2007) analyzed the speech perception skills of monolingual English and Korean–English bilingual first graders. They analyzed children’s ability to identify minimal pair words which differed by initial voiced or voiceless consonants (i.e. /pæt/ vs. /bæt/; /fæt/ vs. /væt/ and, /sɪp/ vs. /zɪp/). It should be noted that both the /v/ and /z/ sounds do not occur in Korean. Results revealed that the Korean–English bilinguals had greater difficulty differentiating /s/-/z/ contrasts than the monolinguals, suggesting some phonetic interference in acquiring English (L2). However, they did not demonstrate difficulty differentiating /f/-/v/ contrasts, which suggests that other factors beyond native language phonemic inventories may contribute to transference or interference between speech sounds in L1 and L2.

Adult speech perception research

One prominent method for assessing speech perception abilities in both monolingual and bilingual populations has been the gating methodology (Brice & Brice, 2008; Cotton & Grosjean, 1984; Gros- jean, 1988, 1996; Li, 1996; Sebastián-Galles, Echeverría, & Bosch, 2005; Tyler & Wessels, 1985). Gating is useful for ascertaining the amount of phonetic information a listener needs in order to correctly recognize and identify a word (Li, 1996). Specifically, the gating task involves recognizing a word after increasing segments of the auditory word stimuli are presented. The listener is instructed to give a confidence rating after each stimulus item presentation (Grosjean, 1996). The following three studies all incorporated the gating methodology in investigating bilingual speech perception.

Li (1996) conducted a study to examine Chinese–English bilingual adults’ spoken word recognition skills. Due to the frequent use of code-switching in the participants’ community, the researcher was particularly interested in examining their word recognition of code-switched stimuli. Results indicated that the bilingual participants identified the code-switched words presented within con- straining contexts as quickly and with the same amount of acoustic information as English monolinguals. Li interpreted this finding as evidence of parallel activation of and interactive processes between the two languages in fluent bilingual adults.

Sebastián-Galles et al. (2005) investigated the speech recognition and word identification abilities of Spanish–Catalan and Catalan–Spanish adults, all of whom learned L2 before age 4 and were fluent in both languages. The participants were asked to discriminate words based on tense /e/ and lax /ɛ/ vowel differences. Catalan contains both vowels, while Spanish only contains the tense /e/ vowel. Despite fluency in both languages, the Spanish L1 bilinguals failed to distinguish the tense–lax vowel contrast. According to the authors, these results appear to suggest that early language exposure may have a strong and long-lasting impact on how individuals perceive first- and second-language sounds.

Brice and Brice (2008) were also interested in examining potential transference and interference effects in bilingual Spanish–English adults, particularly related to specific phonotactic features of words and participants’ age of exposure to L2. Relative to age of exposure, the subgroup of bilinguals with later exposure to English displayed faster word recognition in Spanish. The subgroup of bilinguals with earlier exposure to English

displayed faster recognition in English; moreover, all bilinguals in this study had no difficulty identifying tense of lax vowels in Spanish or English. The middle bilinguals (age of exposure between 9 and 15 years of age) performed equally well in both languages, which the investigators interpreted as evidence of skill transfer from Spanish to English. Moreover, the middle bilingual groups’ performance also provided further support for Thomas and Collier’s (2001) finding that bilingual individuals require at least 4–5 years of language exposure with both languages for positive forward transfer of skills from Spanish to English to occur.

Language alternations: code switching and code mixing

Another frequent means for bilinguals to demonstrate transference of skill between languages is through their use of code switching or code mixing (Poplack, 1980). Grosjean (1982) defined code switching as “the alternation of two or more languages in the same utterance or conversation” (p. 145). Code switching and code mixing are natural behaviors exhibited by bilingual speakers (Brice & Brice, 2009; Goldstein & Kohnert, 2005; Gutierrez-Clellen, 1999; Kohnert, Yim, Nett, Duran, & Duran, 2005). Some bilingual individuals may occasionally experience some brief difficulty in perceiving one alternated word as another word in the other language. However, most bilinguals have no difficulty in understanding what was communicated when code switching or code mixing occur (Brice & Brice, 2009; Goldstein & Kohnert, 2005; Gutierrez-Clellen, 1999; Kohnert et al., 2005). Li (1996) stated that bilinguals primarily use phonemic and phonological cues to distinguish language alternated words. According to the speech learning model, phonologic cues tend to be more recognizable when they are specific only to one language than when the cues are common to both languages (Flege, 1995).

A distinction between two kinds of language alternations (intersentential alternations or code switching and intrasentential alternations or code mixing) has emerged (Brice & Rosa-Lugo, 2000). Consequently, intersentential code switching occurs when the alternation occurs across sentence boundaries. An example of code switching is when the speaker says, “Ya, se acabó. Siéntate. The time is up.” (“It’s finished. Sit down. The time is up”). Language alternation that occurs within a sentence (intrasentential alternation) is often referred to as code mixing (Grosjean, 1982; Torres, 1989). Embedded words and phrases from at least two languages are conjoined within a sentence. An example is when the speaker communicating with bilingual Spanish–English students says, “Sabes qué quieres hacer este weekend?” (Do you know what you want to do this weekend?).

Purpose

The primary goal of this study was to enhance our understanding of the unique characteristics of bilingual speakers’ speech perception development and skills. The extant literature on bilinguals has typically addressed the speech perception of either infants and young children or adults. The authors are unaware of any study that has compared the performance of bilingual children and adults or examined developmental trends in children. Therefore, the current study was designed to examine whether there were any differences in bilinguals’ speech perception performance according to

phonotactic construction, age or language, and also to examine the potential developmental trends in bilingual children’s performance.

González-López (2012) stated that, “Code switching [and code-mixing] then offers an opportunity to study L2 production and L1-L2 interaction while both languages are being assessed by the perception and production systems...” Consequently, we were interested in examining perception and category formation in a bilingual context utilizing code mixing. Such language alternation offers a glimpse into bilingual phonetic category formation and greater delineation of phonetic categories within and between languages unavailable through monolingual phonetic research.

Specifically, there were two main objectives of the current study:

(1) To examine bilinguals’ speech perception skills according to phonotactic construction (CCV+ voiced; CCV− voiceless; CV tense; CV lax), age group (child vs. adult) and language (Spanish, English).

(2) To examine the potential developmental trends in the subgroup of child participants by analyzing whether there were any performance differences in speech perception by grade level (3rd, 4th and 5th grade).

Methods

Participants

Thirty fluent Spanish–English bilinguals with no reported history of speech, language or hearing disorders participated in this study. They included 15 children between 8 and 11 years of age, and 15 adults ranging from 18 to 40 years of age. At the time of the study, all participants were at- tending elementary school or universities in large metropolitan cities in the southeastern US. All had come to the US between the ages of 3 and 8 years and were classified as early bilinguals (Brice & Brice, 2008). The study was approved by the Institutional Review Board at the first author’s institution.

Oral language ability participant ratings

Speaking fluency in both Spanish and English was a prerequisite and was determined by: (a) self- report and (b) an oral language proficiency level of 3 or higher in both languages on the Inter- national Second Language Proficiency Rating tool (ISLPR; Wylie & Ingram, 1999). The ISLPR is designed to evaluate the language proficiency in speaking, listening, reading and writing using a Likert scale rating (i.e. 0 indicates no proficiency, 5 indicates absolute or native-like proficiency). For this study, participants’ speaking skills during a structured oral language interview were measured (Murray, 2005) based on the ISLPR’s general purpose model. The interviews were conducted and scored by the first author, a highly proficient and balanced Spanish–English speaker, a certified and licensed speech–language pathologist with 25 years of clinical experience, and formally trained using the ISLPR. All participants were rated on at least three occasions over the course of the Spanish and English interview.

The oral ratings took place during a single interview of approximately 3–5 min. The interview consisted of conversations in Spanish, English, and code mixed Spanish–English, based on topics chosen by the interviewee and interviewer. Intra-rater reliability was achieved by rating each of the 30 participants three times during the interview (i.e. at the beginning, middle and end of the interview). The intra-rater reliability was 98% (89/90) for Spanish and 96% (87/90) for English.

Gating task

Grosjean (1996), who developed gating, defined it as a task in which a spoken language stimulus is presented in several segments of increasing duration. Participants are asked to identify what the word is after each segment and to give a confidence rating of their accuracy. He stated that gated segments can vary from 20 to 100 msec. The gates used in the current study consisted of 70 msec, replicating Li (1996). Grosjean referred to the percentage of the stimulus needed to identify the word correctly as the listener’s “isolation point.” For example, if a listener identifies the word after 8 gates of 13 total possible gates, the isolation point would be 62%. He termed the point at which the listener correctly identified the word with confidence as the “recognition point.” If the listener is confident in her/his accuracy, then typically the isolation points and recognition points occur at the same interval.

Stimuli

There were a total of 20 target stimuli ranging from 4 to 13 gates. A range of values for the gates (i.e. from 60 to 70 msec) was used so that the segments were cut at zero crossings (Brice & Brice, 2008; Li, 1996). In mathematics, zero crossing is the point at which the sign of a function changes from positive to negative (or vice versa); thus, the zero crossing is the point where the waveform crosses the zero axis. Segments are cut at this zero axis in order to minimize waveform artifacts (i.e. noise). Each stimulus was presented within a sentence of three to four words. The last word in each sentence was the targeted gated stimulus, which was analyzed for the purposes of this study. Real words were chosen to approximate real speech recognitions skills (Grosjean, 1988, 1996; Li, 1996). Fourteen sentences contained Spanish–English code-mixing, beginning in one language and ending with a final word in the other language (e.g. “Is that my plata?,” “Dame el brush”). Six sentences were monolingual; three contained all Spanish words and three contained all English words. These monolingual sentences were randomly interspersed among the code-mixed utterances to prevent the participants from predicting which sentences would contain code-mixed targets. Different carrier sentences were used in order not to fatigue the participants.

The phonotactic construction of the stimuli contained consonants common to both languages: (a) comprising clusters with initial voiced consonants (CCV + ; e.g. /br/, /gr/); (b) initial voiceless consonants (CCV−; e.g. /pl/, /pr/); (c) CV tense phonemes (all in Spanish) and (d) CV lax phonemes. Please note that lax vowels in Spanish only occur in certain contexts and are not phonologically distinctive; therefore, only English words were utilized for contrasts with voiced consonants, voiceless consonants and tense vowels. Three words were repeated in order to establish participant intra-rater reliability.

A one-way ANOVA was conducted to determine whether there were differences in scores between the three repeated items. Results indicated that no significant differences were noted for all participants, indicating adequate intra-rater reliability (F = 0.004, df = 1, p = 0.953). The study thereby yielded 20 test sentences. All words used were common and of high-frequency use (Castellon-Perez, 2001).

The stimuli sentences were randomly distributed so that the presentation order did not influence the speech perception task. The intervening variables consisted of:

(1) Languages: The carrier sentence and target word stimuli consisted of: (a) Spanish–English 6; (b) English–Spanish 8; (c) English–English 3 and (d) Spanish–Spanish 3. (2) Beginning Phonetic Constructions: The beginning consonant cluster (CCV+ voiced and CCV− voiceless) or beginning consonant tense and lax vowel (CV) phonetic constructions were used. These word constructions were chosen in order to determine whether faster recognition points of Spanish or English code mixed words existed.

A fluent Spanish–English speaker, of Costa Rican background, carefully articulated all words and phonemes intentionally speaking a neutral Spanish dialect. This particular speaker was selected because the dialect of Costa Rica is considered one of the most neutral of Latin America (Bjarkman, 1989; Hammond, 1989; Terrel, 1989; Whitley, 2002). For example, the word “lista” was articulated precisely as [lista], i.e. the speaker did not delete any weak syllable final /s/ consonants as is seen in certain Spanish dialects (Goldstein & Iglesias, 1996). In addition, the sentences were read as “true” code mixes (i.e. each word was articulated with careful pronunciation of English or Spanish phonetics). These stimuli have been used in several other peer-reviewed studies (Brice & Brice, 2008; Brice, Castellon-Perez, & Ryalls, 2004; Brice & Ryalls, 2004).

Instructions

Child participants were evaluated in a quiet office within their school, and adult participants were evaluated in a quiet university lab. Testing for children and adults lasted on average between 20 and 30 min. All participants were noted to be attentive. The first author gave explanations to all participants in Spanish and/or English. Each participant listened to the first segment (first gate of 60–70 msec) of the stimulus and was asked to: (a) identify the language; (b) identify the target word and (c) indicate if they were 100% certain of their answer. The researcher recorded all responses. The participant then listened to the next segment consisting of the first two gates (e.g. 120–140 msec of the word). This process continued until the participant correctly ascertained the word with confidence (i.e. recognition point of 100%, as in most gating studies) or the entire word was presented. When the participant correctly identified the word and indicated 100% confidence in their response on two successive gated presentations, then the examiner presented the next sentence. When a participant indicated confidence but did not identify the target word correctly, the next gated segment of the target word was presented. When the listener was unable to identify the word correctly with 100% confidence, the entire gated word was presented before moving on to the next sentence.

Equipment

A Dell desktop computer with a Kay Elemetrics Speech Model 4300B computerized speech lab (CSL) was used to create the gated stimuli. An AKG Micromic II C420 headset microphone was placed according to the manufacturer’s suggestions when recording the stimuli. A CSL Kay Elemetrics program (i.e. version 5.05) was used to time and digitally record the stimuli sentences with the targeted gated word. The recordings were then copied to an Apple iBook computer (500 MHz PowerPC G3 processor) or a Dell Mini 9 Inspiron computer (a 1.6-GHz processor). The Apple computer soundcard (found on the motherboard) samples sound at 44.1 kHz and 16-bit quantization. The Dell Mini 9 Inspiron Computer comes with two-channel high-definition audio with a Realtek ALC268 audio controller (i.e. 44.1 kHz sample rate and 16-bit quantization). The adult participants heard the stimuli sentences through Sony Dynamic Stereo Professional MDR-7506 linear headphones. The students heard the stimuli at 60 dB through JBL speakers with a frequency response range of 40 Hz to 20 kHz and signal-to-noise ratio greater than 80 dB. The researchers found that not all children were tolerant of the headphones for prolonged periods of time; therefore, they listened to the responses through the speakers. Despite this difference, the researchers believe that all participants heard a high-quality signal in a quiet environment. In addition, no outliers indicating a compromised condition were found within the data. No children were distracted during the testing. Response forms were used to record the participants’ responses (i.e. the identified word and whether the participants were 100% sure in their choices).

Results

Statistical analysis

This study used a repeated measures multivariate analysis of variance. The independent variables for the study consisted of: (a) phonotactic construction (CCV+ voiced; CCV− voiceless; CV tense; CV lax); (b) child vs. adult groups and (c) third, fourth and fifth grade student groups. The dependent variable was calculated to be a percentage of the total number of gates the participants needed in order to be 100% positive on two occasions concurrently in correctly recalling the word (i.e. recognition point). Participants’ scores are presented in Table 1.

Assessment of phonotactic features for children and adults

A general linear model multivariate repeated measures or randomized blocks analysis of variance (RBNOVA) using the Wilks’ Lambda F-test was performed in order to answer the questions regarding word identification and phonotactic constructions (i.e. CCV+ voiced; CCV− voiceless; CV tense; CV lax). Mauchly’s test of sphericity was not significant (p = 0.752); therefore, the sphericity assumption was kept. Given that sphericity was assumed, the results indicated significant differences for phonotactic construction [F(3,26) = 9.660, p < 0.001, η2

p = 0.256] with partial η2 indicating a small effect size (Cohen, 1988). There was no significant difference between groups (i.e. children vs. adults) [F(1,28) = 1.152, p = 0.292] and no significant interaction of phonotactic construction by Group [F(3,26) = 1.283, p = 0.286].

Post hoc analyses for children and adults

Using the Bonferroni correction controlling for type I errors, a p value of <0.008 (0.05/6 = 0.008) was required for the significance for the paired t-test analyses. The paired samples t-tests indicated that the following phonotactic contrasts were significant: (a) CCV+ voiced vs. CV tense [t(29) = −3.371; p = 0.002, η2

p = 0.272] with partial η2

demonstrating a small effect size); (b) CCV+ voiced vs. CV lax [t(29) = −4.807; p < 0.001, η2

p = 0.341] with a medium effect size and (c) CCV− voiceless vs. CV lax [t(29)

= −4.025; p = 0.003, η2p = 0.360] with a medium effect size. The CCV− voiceless vs.

CV tense [t(29) = −2.790; p = 0.009, η2p = 0.276] comparison demonstrated a trend

toward significance (Table 2). Participants recognized the English CCV+ voiced stimuli faster than the Spanish CCV+ voiced (English 66.33% recognition point vs. Spanish 80.49% recognition point); English CCV− voiceless (58.48% recognition point) faster than Spanish CCV− voiceless words (78.76% recognition point) and CV tense words (79.96% recognition point; all in Spanish) as quickly as the CV lax words (81.12% recognition point; all in English) (Table 3).

Table 1 Descriptive Phonetic Category Scores for Children and Adults Spanish

CC+ Voiced

English CC+ Voiced

Spanish CV-‐ Voiceless

English CV-‐ Voiceless

Spanish CV Tense

English CV Lax

Children 83.216 69.177 79.386 58.300 82.044 81.583 Adults 77.775 63.488 78.146 58.666 77.888 80.675 Total 80.495 66.333 78.766 58.483 79.966 81.129 Note: Lower scores indicate faster "recognition points"

Table 2 Post-‐Hoc Analyses: Paired Samples T-‐test for Children and Adults Phonotactic Contrasts

DF t Significance

Partial Eta Squared (Effect Size)

CCV+ Voiced vs. CCV-‐ Voiceless

29 .483 .633 0.0432

CC Voiced vs. CV Tense

29 3.371 .002* 0.2729

CCV+ Voiced vs. CV Lax

29 4.807 .000* 0.3417

CCV-‐ Voiceless vs. CV Lax

29 4.025 .003* 0.3604

CCV-‐ Voiceless vs. CV Tense

29 2.790 .009 0.2760

CV Tense vs. CV Lax

29 .793 .633 0.0632

* Significant with Bonferroni correction of p value of less than .008 (.05/6= .008)

Developmental trends for children

The Friedman test, a non-parametric RBNOVA, was chosen because of the small sample size and unequal number of participants employed in this portion of the study (3rd grade n = 4; 4th grade n = 6; 5th grade n = 5; total n = 15; Lehmkuhl, 1996; Tomkins, 2006). With parametric statistics, a small sample may be problematic, affecting the power of a test and consequently type II errors (Steelman & Levy, 2006). However, there is no technique for estimating the sample size and power analysis for non-parametric measures (McDonald, 2009; Mumby, 2002). Mumby (2002) stated that, “Even the ‘bible’ of power analysis (Cohen, 1988) does not describe how to assess non-parametric power” (p. 85). Yet, non-parametric tests can be as powerful as their parametric equivalents if assumptions are met, e.g. independent samples, homogeneity of variance (Cardone, 2010; Mumby, 2002; Peres-Neto & Olden, 2001; Siegel, 1957). The children sample was obtained independently, i.e. each child was unrelated to the other children in the sample. The Brown–Forsythe (i.e. non-parametric Levine) test for homogeneity of variance indicated non-significant results for all phonotactic and language measures by group (3rd, 4th and 5th grade students; p ≥ 0.05). Hence, the homogeneity of variance was assumed. In addition, an increase in the number of comparisons conducted (i.e. each sample should contain five or more measures) can also increase the power of the test (Cardone, 2010; Mumby, 2002). The repeated measures children sample contained 510 data cells (i.e. 34 separate measures). Consequently, the use of non-parametric statistics appeared justified with sufficient power.

Alpha or significance level was set at p < 0.05. Analysis for the gated word identification task indicated significant differences from the ranking of the measures for the three student groups [χ2(3) = 9.800, p = 0.020]. Subsequently, a Kruskal–Wallis ANOVA test revealed phonotactic constructions to be significantly different among the third, fourth and fifth grade students for the CCV− voiceless phonotactic construction [χ2(2) = 6.416; p = 0.040] (Table 4). Effect size differences cannot be computed for the Kruskal–Wallis test; however, Green and Salkind (2008) stated that differences in mean ranks between the three groups can serve as an effect size index. The mean score for CCV− voiceless for third grade was 76.41; for fourth grade was 76.91 and for fifth grade was 68.50. All other phonotactic constructions were deemed to be non-significant: (a) CCV+ voiced (p = 0.075); (b) CV tense (p = 0.284) and (c) CV lax (p = 0.418).

Table 3 Post-‐Hoc Analyses: Word Construction Comparisons Between Spanish and English for Children and Adults Phonotactic Construction Mean Standard Deviation N English CCV+ Voiced 66.3333 12.5708 30 Spanish CCV+ Voiced 80.4958 10.2849 30 English CCV-‐ Voiceless 58.4833 12.7113 30 Spanish CCV-‐ Voiceless 78.7667 6.1845 30 Spanish CV Tense 79.9667 9.7226 30 English CV Lax 81.1292 8.5891 30

Descriptive analyses of children

Descriptive analyses indicated that English CCV+ voiced consonants were recognized faster than the Spanish consonants for 3rd grade (English 82.08% recognition point vs. Spanish 85.87% recognition point), 4th grade (English 75.16% recognition point vs. Spanish 84.37% recognition point) and 5th grade (English 62.73% recognition point vs. Spanish 75.50% recognition point). A developmental trend was noted for the English CCV+ voiced words, i.e. a consistently faster recognition of words by increasing the grade level (3rd grade 72.08% recognition point vs. 4th grade 75.16% recognition point vs. 5th grade 62.73% recognition point). This same developmental trend was noted for Spanish CCV+ voiced Spanish consonants (3rd grade 85.87% recognition point vs. 4th grade 84.37% recognition point vs. 5th grade 75.50% recognition point). See Table 4.

It was observed that all CCV− voiceless consonants were recognized faster than all CCV+ voiced consonants (75.42% recognition point vs. 77.20% recognition point) for all grade levels. In addition, descriptive analysis indicated that English CCV− voiceless stimuli were recognized faster than the Spanish CCV− voiceless (English 58.30% recognition point vs. Spanish 78.45% recognition point). A developmental trend was noted for the Spanish CCV− voiceless words, i.e. a consistently faster recognition of words by increasing the grade level. Third-grade students recognition of the Spanish

CCV− voiceless was 80.61% recognition point; 4th grade recognition was 79.68% recognition point; while 5th grade recognition was 75.24% recognition point. Such a noticeable trend was not as Spanish–English speech recognition 229 evident with the English CCV− voiceless words (3rd grade 55.75% recognition point; 4th grade 63.08% recognition point; 5th grade 54.60% recognition point; Table 4).

Table 4 Post-‐Hoc Analyses: Word Construction Comparisons for Children Phonotactic Construction

Grade Mean (percentage recognition point)

Standard Deviation

English CCV+ Voiced All Grades (n=15) 72.8667 16.9231 3rd (n=4) 82.0833 10.7750 4th (n=6) 75.1667 14.4212 5th (n=5) 62.7333 17.7394 Spanish CCV+ Voiced All Grades (n=15) 81.8167 8.6903 3rd (n=4) 85.8750 2.3496 4th (n=6) 84.3750 6.6777 5th (n=5) 75.5000 11.3261 All CCV+ Voiced All Grades 77.2000 8.7358 English CCV-‐ Voiceless All Grades (n=15) 58.3000 14.5460 3rd (n=4) 55.7500 16.6458 4th (n=6) 63.0833 17.6278 5th (n=5) 54.6000 9.2965 Spanish CCV-‐ Voiceless All Grades (n=15) 78.4519 5.7644 3rd (n=4) 80.6111 3.8065 4th (n=6) 79.6852 7.6198 5th (n=5) 75.2444 3.6127 All CCV-‐ Voiceless All Grades 75.4222 6.9667

Additional analyses: child language

The Friedman test was also chosen to analyze Spanish vs. English word identification. However, the results indicated non-significant differences under all conditions (p = 0.531) (Spanish–English, English–English, English–Spanish and Spanish–Spanish) for the three groups.

Discussion

The primary objective of this study was to enhance our understanding of speech perception skills in bilinguals by examining their performance on a gating task according

to phonotactic construction, age, group and language. Because the limited previous research has focused on either infants and young children or adults, we were particularly interested in analyzing the performance of both children and adults and also examining the potential developmental trends in children.

Addressing our first research question, results did indeed indicate significant differences of the entire groups of bilingual participants according to the different phonotactic constructions. Both the Spanish–English children and adult listeners appeared to identify words fastest when their initial consonants were voiceless (i.e. CCV− voiceless), indicating that the voiceless feature facilitated their speech recognition. Comparing performance by language, participants recognized English CCV− voiceless stimuli more quickly than the Spanish CCV− voiceless (English 58.48% recognition point vs. Spanish 78.76% recognition point). The voice onset time (VOT) for initial voiceless stop consonants is longer in English than Spanish (Thornburgh & Ryalls, 1998; Zampini, 1998). That participants identified the English CCV− voiceless stimuli more quickly may perhaps be attributed to the longer VOT in English than Spanish (e.g. +60 msec vs. +20 msec) and also the overlap between the Spanish CCV− voiceless stimuli (e.g. +20 msec) and the English CCV+ voiced stimuli (e.g. +30 msec; Zampini, 1998). This overlap may have caused perceptual interference between English and Spanish, i.e. confusion over similar VOTs. However, participants recognized CV tense words (79.96% recognition point; all in Spanish) nearly as quickly as the CV lax words (81.12 recognition point; all in English).

At the time of this study, the participants were living in the US and attended English speaking schools (i.e. elementary schools or university). Therefore, it is possible that their frequent exposure to both spoken and written English may account for their faster recognition of English CCV− voiceless words than Spanish CCV− voiceless words. It would be of interest to examine participants with more frequent exposure to Spanish in both oral and written contexts. In addition, it was found that differences occurred when comparing CCV+ voiced and CCV− voiceless consonants vs. CV tense vowels and CV lax vowels. Both CV tense and CV lax vowel words took longer to recognize than either the CCV+ voiced or CCV− voiceless words. These results indicated that consonants seem to offer less ambiguity in recognition and identification. It was observed on several occasions that some bilingual students had difficulty differentiating Spanish vs. English vowels /o/ in “problema” vs. the /ɒ/ in “problem.” Bilingual speakers seemed to show greater difficulties with English vowels (Brice, Carson, & O’Brien, 2009; Cotton & Sharp, 1988). It also appeared that the bilingual students may not yet have a full representation of Spanish vs. English vowels compared with the bilingual adult speakers.

The participants had no difficulty in differentiating tense vs. lax vowels. Some students experienced difficulty in differentiating the /o/ vs. /ɒ/ vowels; however, both of these vowels are tense vowels. This latter finding is interesting because of the vast differences in Spanish and English vowel systems (Manrique, 1979; Strange, Verbrugge, Shankweiler, & Edman, 1976; Treiman, 1991). It would be expected from previous research that differences among tense and lax vowels would be present (Sebastián-Galles et al., 2005). However, these findings may result from the fact that all bilingual participants in this study had spoken and been exposed to English for a minimum of 6

years. This time frame may have been a sufficient period to learn tense–lax vowel distinctions. Nevertheless, this aspect merits further investigation.

With regard to the second research question, it was asked if developmental differences in bilingual word identification skills in children at various grade levels would be found. No significant group differences were found between the children and the adults. However, it should be noted that the children took longer to perceive words under all phonotactic constructions when compared with adults. These results seem to indicate a developmental learning trend. However, phonotactic constructions were found to be significant indicating some developmental trends in Spanish for the third, fourth and fifth grade children.

The children across all grade levels recognized the English CCV+ voiced consonants faster than the Spanish CCV+ voiced consonants. There are numerous potential explanations for this finding. It seems possible that students are exposed to more English in the school environment than Spanish in the home. Consequently, their increased English exposure may have had an effect on their perceptual abilities. However, recognition of English voiced consonants may have also been due to the functional load (i.e. how often the sound occurs in word positions) and phonetic frequency of consonants (i.e. how often the sound occurs in spoken language) in each language, as English has more voiced consonants than Spanish (Ingram, 2011; Stockwell & Bowen, 1965). In addition, developmental trends were noted for both the English and Spanish CCV+ voiced consonants suggesting that perceptual skills improve with age even in the upper elementary aged students.

The English CCV− voiceless consonant words were recognized faster than the Spanish CCV− voiceless consonant words. Again, it is hypothesized that English may be more dominant in exposure and that results reflect lags in their Spanish development. Again, functional load and phonetic frequency of English voiceless sounds may have affected the results, with English having more voiced consonant sounds than Spanish. For Spanish CCV− voiceless words, it was found that faster recognition occurred for the fifth grade students followed by fourth grade and then third grade students, suggesting a developmental learning trend. Developmental trends seemed to occur when comparing: (a) recognition of voiceless consonants to voiced consonants (e.g. [p] sounds are developmentally easier than [b] sounds) and (b) the child subgroups (3rd vs. 4th vs. 5th grade students with differences on the CCV+ voiced and CCV− voiceless consonant words).

In addition, it was noted that some students had difficulty differentiating between CCV− voiceless consonants vs. CCV+ voiced consonants, particularly the /p/ vs. /b/ distinctions in the Spanish words “plata” and “problema” (i.e. perceiving the words as “blata” and “broblema”).

Spanish voiceless stop consonants /p, t., k/ are unaspirated stops with a short VOT value (e.g. +20 msec; Thornburgh & Ryalls, 1998; Zampini, 1998). The Spanish voiced stop consonants /b, d, g/ are typically prevoiced (e.g. −20 or 0 msec) or with a shorter lag (e.g. +20 msec; Zampini, 1998). Consequently, it appeared that the Spanish students had

difficulty discerning short lags in /p/ vs. the sometimes short lag of /b/ (Brice & Brice, 2008).

Limitations

It is noted that the three repeated words used for reliability checks were limited. However, a one-way ANOVA resulted in no significant differences for word comparisons; thus, indicating that adequate intra-rater reliability was obtained. In addition, the different listening conditions of the children and adults were a noted limitation. However, no outliers were found within the data and the children did not seem distracted during the data collection. Consequently, it is recommended that future studies Spanish–English speech recognition 231 address these limitations by expanding the number of repeated items and also having all test conditions be equal across participants.

Future research

Due to the relationship between speech perception skills and other skills including phonological development (Escudero, 2007), literacy-related skills including phonological awareness (McBride-Chang, 1995) and lexical development (Fennel et al., 2007), future research should examine if ease or difficulty of speech perception according to phonotactic characteristics impacts children’s phonological, lexical or literacy skills (e.g. phonological awareness, word reading, spelling). It may be that children require explicit instruction by a speech–language pathologist or teacher to address voiceless consonant recognition followed by voiced consonant recognition in phonological and phonemic awareness instruction, even with 3rd, 4th and 5th grade bilingual students. It also appeared that the bilingual students were still acquiring both their Spanish and English phonemic and lexical skills that are both crucial indicators of reading development.

Researchers have illustrated that speech perception is both a top-down and bottom-up process (Burgess & Chiarello, 1996; Marlsen-Wilson, 1975; Marlsen-Wilson &Welsh, 1978). The participants were assessed for their speech perception skills (i.e. bottom-up skills); therefore, phonemic awareness, phonological awareness and phonics skills seem to play an important role in decoding words. However, literacy skills (which are top-down skills) may also impact speech perceptions skills. Differences in literacy skills and speech perception skills among children and adult participants are a topic for further research.

In conclusion, certain phonotactic features appear to be dependent on perception in each language and age of development (e.g. CCV− voiceless consonant words for children). Sebastián-Galles et al. (2005) found that the Spanish–Catalan (i.e. Spanish L1 speakers) early bilingual speakers failed to distinguish the contrast /e/-/ɛ/ vowel. However, all children and adult speakers in this study were able to differentiate all contrasts between Spanish and English, indicating that identification errors may be language specific or specific only for certain sounds. Recall that Fennel et al. (2007) found that language interference occurred for the Korean–English first-grade students differentiating the /s/-/z/ contrasts; whereas, the /z/ sound does not occur in Korean. The Spanish–English

speakers in this study showed almost equal recognition times for the CV tense vs. CV lax vowels words; thus, the non-Spanish lax vowels did not pose recognition difficulties. Consequently, this study does not completely support all the findings from Sebastián-Galles et al. (2005) nor those of Chiappe et al. (2007). Fluent bilinguals (both children and adults) in this study did not appear to have difficulty identifying words in each language when simultaneously exposed to both languages.

Conclusion

Overall, it appears that multiple factors such as similarity in languages, age of arrival, age of development, the language environment and/or developmental factors may influence one’s perception of the second language. In conclusion, it was noted that both children and adult bilingual speakers in this repeated measures study were able to recognize all consonants and vowels under all conditions. Bullock, Toribio, González, and Dalola (2006) stated that, “Such research contends that bilinguals’ language use is malleable in that they may behave differently according to which language they are producing or perceiving at a given time” (p. 9). Hence, language abilities in bilinguals appear to be one that varies according to language and specific task.

Declaration of Interest: The authors report no declaration of interest. The authors have no financial interest, direct or indirect, in the subject matter or materials discussed in the manuscript. This project was a non-funded study.

References

AKG Acoustics. (2001). AKG micromic II C420 headworn microphones. Nashville, TN: Harman. Apple. (2001). Apple iBook. Cupertino, CA: Author. Bjarkman, P. C. (1989). Abstract and concrete approaches to phonological strength and weakening chains: Implications for Spanish syllable structure. In P. C. Bjarkman & R. H. Hammond (Eds.), American Spanish pronunciation (pp. 106–136). Washington, DC: Georgetown University Press.

Bosch, L., & Sebastián-Gallés, N. (2003). Simultaneous bilingualism and the perception of a language-specific vowel contrast in the first year of life. Language and Speech, 46, 217–243. Brice, A., & Brice, R. (2008). Examination of the critical period hypothesis and ultimate attainment among Spanish–English bilinguals and English-speaking monolinguals. Asia Pacific Journal of Speech, Language and Hearing, 11(3), 143–160.

Brice, A., & Brice, R. (2009) (Eds.). Language development: Monolingual and bilingual acquisition. Old Tappan, NJ: Merrill/Prentice Hall.

Brice, A., Carson, C., & O’Brien, J. (2009). Spanish–English articulation and phonology of four and five-year old preschool children: An initial investigation. Communication Disorders Quarterly, 31(1), 3–14. doi: 10.1177/1525740108327447

Brice, A., Castellon-Perez, Y., & Ryalls, J. (2004). Speech recognition of code switched words by proficient Spanish–English bilinguals. Journal of Distinguished Language Studies, 2, 13–22.

Brice, A., & Rosa-Lugo, L. (2000). Code switching: A bridge or barrier between two languages? Multiple Voices for Ethnically Diverse Exceptional Learners, 4(1), 1–12.

Brice, A., & Ryalls, J. (2004). Recognition of code-mixed words by Spanish/English bilinguals: A focus on proficiency levels. Conference Proceedings of the 26th World Congress of the International Association of Logopedics and Phoniatrics, 6, 53–56.

Bullock, B. E., Toribio, A. J., González, V., & Dalola, A. (2006). Language dominance and performance outcomes in bilingual pronunciation. In M. G. O’Brien, C. Shea, & J. Archibald (Eds.), Proceedings of the 8th Generative Approaches to Second Language Acquisition Conference (pp. 9–16). Somerville, MA: Cascadilla Proceedings Project.

Burgess, C., & Chiarello, C. (1996). Neurocognitive mechanisms underlying metaphor comprehension and other figurative

language. Metaphor and Symbolic Activity, 11, 67–84. Cardone, R. (2010). Nonparametric: Distribution-free, not assumption free. Six Sigma Newsletter. Retrieved October 8, 2010 from http://www.isixsigma.com/tools-templates/hypothesis-testing/nonparametric-distribution-free-not-assumption-free/

Castellon-Perez, y. (2001). Recognition of code-switched words by Spanish/English bilinguals. (Unpublished master’s thesis). University of Central Florida, Orlando, FL.

Chiappe, P., Glaeser, B., & Ferko, D. (2007). Speech perception, vocabulary, and the development of reading skills in English among Korean- and English-speaking children. Journal of Educational Psychology, 99(1), 154–166. doi: 10.1037/0022- 0663.99.1.154

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.

Cotton, S., & Grosjean, F. (1984). The gating paradigm: A comparison of successive and individual presentation formats. Perception and Psychophysics, 35, 41–48.

Cotton, F., & Sharp, J. (1988). Spanish in the Americas. Washington, DC: Georgetown University Press.

Dell. (2008). Inspiron mini 9 netbook. Round Rock, TX: Author.

Escudero, P. (2007). Second language phonology: The role of perception. In M. Pennington (Ed.), Phonology in context (pp. 109–134). Oxford, UK: Macmillan.

Fennel, C. T., Byers-Heinlein, K., & Werker, J. F. (2007). Using speech sounds to guide word learning: The case of bilingual infants. Child Development, 78(5), 1510–1525. doi: 10.1111/j.1467-8624.2007.01080.x

Flege, J. (1995). Second language speech learning: Theory, findings and problems. In W. Strange (Ed.), Speech perception and linguistic experience: Theoretical and methodological issues (pp. 233–277). Baltimore: York Press.

Goldstein, B., & Iglesias, A. (1996). Phonological patterns in normally developing Spanish-speaking 3- and 4-year-olds of Puerto Rican descent. Language, Speech, and Hearing Services in Schools, 27, 82–89.

Goldstein, B., & Kohnert, K. (2005). Speech, language and hearing in developing bilingual children: Current findings and future directions. Language, Speech and Hearing Services in Schools, 36(3), 264–267. doi: 10.1044/0161-1461(2005/026)

González-López, V. (2012). Spanish and English word-initial voiceless stop production in code-switched vs. monolingual structures. Second Language Research, 28(2), 243–263. doi: 10.1177/02676583124398321

Green, S. B., & Salkind, N. J. (2008). Using SPSS for Window and Macintosh: Analyzing and understanding data (5th ed.). Upper Saddle River, NJ: Pearson Prentice Hall.

Grosjean, F. (1982). Life with two languages: An introduction to bilingualism. Cambridge, MA: Harvard University Press.

Grosjean, F. (1988). Exploring the recognition of guest words in bilingual speech. Language and Cognitive Processes, 3, 233–274.

Grosjean, F. (1996). Gating. Language and Cognitive Processes, 11(6), 597–604. doi: 10.1080/016909696386999

Gutierrez-Clellen, V. (1999). Language choice in intervention with bilingual children. American Journal of Speech-Language Pathology, 8, 291–302.

Hammond, R. M. (1989). American Spanish dialectology and phonology from current theoretical perspectives. In P. C. Bjark-man & R. H. Hammond (Eds.), American Spanish pronunciation (pp. 137–169). Washington, DC: Georgetown University Press.

Harman Multimedia. (2004). JBL duet multimedia speakers. Northridge, CA: Author.

Ingram, D. (2011). Prologue: Cross-linguistic and multilingual aspects of speech sound disorders in children. In S. McLeod & B. Goldstein (Eds.), Multilingual aspects of speech sound disorders in children (pp. 3–12). Bristol, UK: Multilingual Matters.

Jusczyk, P. W. (1997). The discovery of spoken language. Cambridge, MA: MIT Press.

KayPentax Medical Company. (2001). Kay Elemetrics speech model 4300B computerized speech lab (CSL). Lincoln Park, NJ: Author.

Kohnert, K., Yim, D., Nett, K., Duran, P. F., & Duran, L. (2005). Intervention with linguistically diverse preschool children: A focus on developing home languages(s). Language, Speech and Hearing Services in Schools, 36(3), 251–263. doi: 10.1044/0161-1461(2005/025)

Kuhl, P. K. (2000). A new view of language acquisition. Proceedings of the National Academy of Sciences of the United States of America, 97, 11850–11857. Leather, J. (1999). Second-language speech research: An introduction. Language Learning, 49(1), 1–56.

Lehmkuhl, L. D. (1996). Nonparametric statistics: Methods for analyzing data not meeting assumptions required for the application of parametric tests. Journal of Prosthetics & Orthotics, 8(3), 105–113.

Li, P. (1996). Spoken word recognition of code switched words by Chinese-English bilinguals. Journal of Memory and Language, 35, 757–774. doi: 10.1006/jmla.1996.0039

Manrique, A. M. B. (1979). On the recognition of isolated Spanish vowels. In H. Hollien & P. Hollien (Eds.), Current issues in the phonetic sciences (pp. 677–681). Amsterdam, the Netherlands: John Benjamin B.V.

Marlsen-Wilson, W. D. (1975). Sentence perception as an interactive parallel process. Science, 189, 226–228.

Marlsen-Wilson, W. D., & Welsh, A. (1978). Processing interactions and lexical access during word recognition in continuous speech. Cognitive Psychology, 10, 29–63.

McBride-Chang, C. (1995). What is phonological awareness? Journal of Educational Psychology, 87(2) 179–192. McDonald, J. (2009). Handbook of biological statistics (2nd ed.). Baltimore, MD: Sparky House Publishing.

Mumby, P. J. (2002). Statistical power of non-parametric tests: A quick guide for designing sampling strategies. Marine Pollution Bulletin, 44, 85–87. Murray, D. (2005) (Ed.). First language support in adult ESL in Australia. Sydney, Australia: National Centre for English Language Teaching and Research.

Paradis, J., & Genessee, F. (1996). Syntactic acquisition in bilingual children: Autonomous or independent? Studies in Second Language Acquisition, 18, 1-25.

Peres-Neto, P. R., & Olden, J. D. (2001). Assessing the robustness of randomization tests: Examples from behavioural studies. Animal Behaviour, 61, 79–86. doi: 10.1006/anbe.2000.1576

Poplack, S. (1980). Sometimes I’ll start a sentence in Spanish y termino en español: Toward a typology of code-switching. Linguistics, 18, 581–618.

Sebastián-Galles, N., Echeverría, S., & Bosch, L. (2005). The influence of initial exposure on lexical representation: Comparing early and simultaneous bilinguals. Journal of Memory and Language, 52, 240–255.

Siegel, S. (1957). Nonparametric statistics. The American Statistician, 11, 13–19.

Sony. (2001). Sony dynamic stereo professional MDR-7506 professional headphones. New York: Author.

Steelman, L. A., & Levy, P. E. (2006). Using advanced statistics. In F. T. L. Leong & J. T. Austin (Eds.), The psychology research handbook (2nd ed., pp. 306–315). Thousand Oaks, CA: Sage.

Stockwell, R. P., & Bowen, J. D. (1965). The sounds of English and Spanish. Chicago: University of Chicago Press. Strange, W., Verbrugge, R. R., Shankweiler, D., & Edman, T. R. (1976). Consonant environment specifies vowel identity. Journal of the Acoustical Society of America, 60, 213–224.

Terrel, T. (1989). Teaching Spanish pronunciation in a communicative approach. In P. C. Bjarkman & R. H. Hammond (Eds.), American Spanish pronunciation (pp. 196–214). Washington, DC: Georgetown University Press.

Tomkins, C. C. (2006). An introduction to non-parametric statistics for health scientists. University of Alberta Health Sciences Journal, 3(1), pp. 20, 21, 24–26.

Treiman, R. (1991). Children’s spelling errors on syllable-initial consonant clusters. Journal of Educational Psychology, 83, 346–360.

Thomas, W. P., & Collier, V. P. (2001). A national study of school effectiveness for language minority students long-term academic achievement. Berkeley, CA: Center for Education Research, Diversity and Excellence.

Thornburgh, D. F., & Ryalls, J. H. (1998). Voice onset time in Spanish–English bilinguals: Early versus late learners of English. Journal of Communication Disorders, 31, 215–229.

Torres, L. (1989). Code-mixing and borrowing in a New York Puerto Rican community; A cross generational study. World Englishes, 8(3), 419–432.

Tsao, F.-M., Liu, H.-M., & Kuhl, P. K. (2004). Speech perception in infancy predicts language development in the second year of life: A longitudinal study. Child Development, 75(4), 1067–1084.

Tyler, L., & Wessels, J. (1985). Is gating an on-line task? Evidence from naming latency data. Perception and Psychophysics, 38, 217–222.

Werker, J. F., & Tees, R. C. (1999). Influences on infant speech processing: Toward a new synthesis. Annual Review of Psychology, 50, 509–535.

Whitley, M. S. (2002). Spanish–English contrasts (2nd ed.). Washington, DC: Georgetown University Press.

Wylie, E., & Ingram, D. E. (1999). International second language proficiency ratings. Brisbane, Australia: Griffith University.

Zampini, M. L. (1998). The relationship between the production on perception of L2 Spanish stops. Texas Papers in Foreign Language Education, 33, 85–99.

Appendix 1: Stimulus words

(1) cartoon

(2) problema (Spanish)

(3) green

(4) carta (Spanish)

(5) postre (Spanish)

(6) plata (Spanish)

(7) brush

(8) card

(9) grande (Spanish)

(10) brillo (Spanish)

(11) carro (Spanish)

(12) plate

(13) bruja (Spanish)

(14) poster

(15) gratis (Spanish)

(16) great

Spanish–English speech perception in children and adults: Developmental trends

Documents