Bit and beat are heard as the same: Mapping the vowel perceptual patterns of Greek-English bilingual children

Highlights

• Several pairs of English vowels were assimilated to a single Greek vowel category.

• There was poor to moderate discrimination for specific English vowel contrasts.

• Sensitivity in acoustical differences was not affected by proficiency level.

Abstract

It has been previously proposed that small native vowel inventories impede the acquisition of foreign language vowels that come from larger inventories. The present study aims to investigate how the already-formed Greek phonological system is modified after the perception of the English vowels in a non-naturalistic environment and to what extent Greek speakers are able to discriminate challenging English vowel contrasts. Also, it aims to examine if learning experience enhances sensitivity to acoustical information of the foreign language vowels. The study relies on the theoretical framework of the Perceptual Assimilation Model-Second Language (PAM-L2). For developing predictions about the perception of the English vowels by Greek speakers, the vowel spaces of both Greek and English native speakers were investigated; 3 native speakers of Greek and 3 native speakers of English produced vowels in their native language. 26 Greek learners of English as a foreign language in Cyprus with an age that varied from 8 to 12 years participated in the perceptual study. Learners were divided into two groups according to their proficiency level in English. The participants took an identification test in order to determine how English vowels were assimilated to the listeners’ native phonological categories and then they were tested in an AXB discrimination task in order to investigate to what extent they are able to discriminate English vowel contrasts. The results showed that several pairs of two or more English vowels were assimilated to a single Greek phonological category. Furthermore, the discrimination test showed poor to moderate discrimination accuracy for both groups regarding the English vowel contrasts, yet differences in the discrimination accuracy of the contrasts between the novice and the more advanced group of learners were minimal; only the English/e/-/ɜː/contrast was discriminated slightly better by the advanced learners. Thus, learners with a higher proficiency level did not generally perceive the English vowels better than learners with a lower proficiency level, signifying that perception of foreign language vowels is not merely a matter of amount of exposure to the foreign stimuli. Conclusions are drawn about the interference of native Greek with the learning of English vowels and the acquisition of the foreign language stimuli in a classroom environment.

Introduction

It is widely recognized that infants are able to distinguish a great number of phonetic contrasts at least until the first 6 months of life (Best et al., 1988, Eimas et al., 1971). At a later stage of life, the establishment of the first language (L1) speech sound inventory affects the acquisition of unfamiliar sounds often creating difficulties in both perception and production of the foreign speech sounds (Georgiou, 2018c). Specifically, contrasts of similar sounds that are not contrastive in the native language of an L2/foreign language learner are usually difficult to distinguish (Escudero, 2005, Morrison, 2003, Flege et al., 1997). This is more evident for learners whose L1 vowel inventory is smaller in size than the inventory of the L2 (Lengeris, 2009a, Lengeris, 2009b) because there is an increased probability that the two unfamiliar sounds will be perceived as exemplars of the same category. For instance, native speakers of Spanish, whose vowel system has just five vowels, do not discriminate the English/iː/-/ɪ/vowel contrast well (Sakai, 2016); English has many more vowels than Spanish. Similarly, according to Kartushina and Frauenfelder (2013), Spanish learners of French discriminate the French/e/-/ε/contrast poorly, perceiving both vowels as instances of a single Spanish vowel category (/e/). By contrast, some studies indicate that native speakers of languages that have a larger vowel system than English, such as German and Norwegian, are able to identify the English vowels more consistently than native speakers of Spanish (Alispahic et al., 2017, Iverson and Evans, 2009, Iverson and Evans, 2007).

Several models predict the ability of learners to identify vowel contrasts in an unfamiliar language or an L2. One of the most influential models is the Perceptual Assimilation Model (PAM) (Best, 1995). In PAM, the way in which unfamiliar pairs of speech sounds are assimilated to the listeners' native sound inventory determines how accurately they are discriminated by naïve learners of a language (i.e., learners who have minimal or no contact with the unfamiliar language). PAM proposes six types of assimilation for unfamiliar L2 contrasts to the L1 phonological categories of the L2 learners; the assimilation types determine the discrimination accuracy of the contrasts, which can vary from poor to very good (Best, 1995). The assimilation types are the following: 1) Two Category (TC) assimilation in which two sounds are assimilated to two different L1 phonological categories; this predicts excellent discrimination, 2) Single Category (SC) assimilation in which the two sounds are assimilated to the same L1 phonological category as equally good or poor exemplars of the L1 segment; poor discrimination is predicted, 3) Category Goodness difference (CG) assimilation in which the two sounds are again assimilated to the same L1 category, but, in this case, one is a good exemplar of that category while the other is poor; discrimination ranges from moderate to good, 4) Uncategorized-Categorized (UC) assimilation in which a very good discrimination is predicted because one sound is assimilated to an L1 phonological category while the other falls into a part of the phonetic space that is not occupied by a native category, 5) Uncategorized-Uncategorized (UU) in which both L2 sounds are not assimilated to any L1 category; this predicts a poor to moderate discrimination and 6) Non-assimilated (NA) in which the two unfamiliar sounds are not perceived as speech sounds; this predicts poor to very good discrimination depending on the similarity of the sounds. The relationship between the predictions of the assimilation types have been tested in several studies. Georgiou (2018b) investigated the perceptual patterns of 15 adult Greek-Arabic bilinguals who performed both identification and discrimination tasks in order to determine how the L2 vowels assimilated to the listeners' L1 vowel system and to relate those patterns to listeners' performance discriminating specific vowel pairs. He concluded that Greek contrasts which fell into the TC assimilation type could be discriminated better than CG contrasts; confirming PAM's predictions regarding the relationship of the aforementioned assimilation types. Tyler, Best, Faber and Levitt (2014) investigated the perceptual abilities of American English listeners in three languages in which listeners had no experience with (French, Norwegian and Thai). Accordingly, they identified and discriminated minimal pair contrasts in these languages. The results indicated that contrasts of the TC assimilation type were more distinguishable than the pairs of the CG and SC types. Pairs of the SC assimilation type were the least distinguishable among the three types (TC > CG > SC).

The Perceptual Assimilation Model-L2 (PAM-L2) extents the predictions of PAM to the success of L2 learners in discriminating L2 phonetic segment pairs, in addition to the novice listeners addressed in PAM (Antoniou et al., 2012, Best and Tyler, 2007). In this model, the distinction between phonetic and phonological categories is important. Phonetic categories refer to individual sounds which differ in a gradient manner; phonological categories might include one or more phonetic categories that together serve a contrastive phonological function in the language. One reason to make the distinction is that L2 learners may distinguish an L2 phonetic segment from an L1 segment based on their gradient differences, but they may recognize that both are members of the phonological category/p/, for example, that contrasts with/b/in L1 and L2 (see. e.g., Antoniou et al., 2012). As in PAM, L2 speech sounds might be assimilated to existing L1 categories. But for L2 speech sounds that are sufficiently different from any existing category in memory, listeners may instead establish new L2 categories. Assimilations take place, first, on a phonetic and later they expand also to the phonological level. With the expansion of learners' vocabulary, L2 learners learn which phonetic distinctions are contrastive, and discrimination of L2 sounds then can be based on lexically relevant categorical differences in the L2 (Antoniou et al., 2012). So, as a learner's lexicon grows in size, discrimination of L2 segments improves because learners become attuned to phonological contrasts as well as to the phonetic and articulatory differences that were not available before (Antoniou, 2010).

Another important model in the area of L2 acquisition is the Speech Learning Model (SLM) (Flege, 1995, Flege, 2005) which deals with experienced L2 learners. According to SLM, during L2 learning, new sound categories are formed if L2 sounds differ sufficiently from L1 category members. However, if an L2 sound is perceived as similar to an existing L1 sound, merged L1-L2 categories are created. This can lead to inaccurate production of the L2 sound. So, SLM provides a link between perception and production (Flege, 2002) supporting that perception precedes production and that the former can shape the latter. A central position in the philosophical background of the model is the existence of a mechanism called equivalence classification. This mechanism underlies merging of L2 sounds with L1 phonetic categories, with the result that listeners may confuse an L2 sound with a similar one in their L1 inventory. In order to make testable predictions, the model considers the parameter of experience because, according to the model, perception of L2 sounds differs depending on the listener's experience with L2. For SLM, if much phonetic dissimilarity is perceived between an L2 sound and its closest L1 sound, then it is likely that the phonetic differences between these sounds will be discerned.

The aforementioned models have the same purpose: to reveal the way that L2/unfamiliar contrasts are perceived by learners/listeners. Nevertheless, there are differences. One difference that is not relevant here is that PAM/PAM-L2 adopts the direct realist claims that listeners' directly perceive speech gestures. SLM does not adopt that claim. More relevant to the present research, SLM addresses highly experienced learners of a language while PAM focuses on listeners with no previous (or very little) experience in a language (e.g., discrimination of Zulu clicks by English speakers) while PAM-L2 focuses on the learning process itself. Also, SLM is mainly a production model. Although it seeks to explain perceptual reorganization of speech sound with perceptual experience, it does not provide a detailed picture of the learners' perceptual mechanisms. In contrast, PAM/PAM-L2 proposes the assimilation types described earlier. Hypotheses based on these assimilation types have been tested with success in many studies in the literature. PAM-L2 is the most appropriate model for use in the present study because the participants are young learners actively acquiring an L2. Furthermore, the assimilation types that emerge from the model's theoretical framework promise to provide a clear picture of L2 learners' perceptual patterns.

Apart from L1 experience, many other factors may affect the acquisition of foreign speech sounds. These factors include the age of acquisition of L2 (Flege, MacKay & Meador, 1999a), the length of residence in the country where the L2 is dominant (Flege and Liu, 2001), L1-L2 use (Flege et al., 1997) and motivation (Flege, Munro & MacKay, 1995b). Some studies pay attention to the impact of proficiency level in the perception of foreign language sounds. For example, Al-Abdely et al. (2016) examined the effect of proficiency level on perception of the English monophthongal vowels by Iraqi EFL learners in Malaysia. They found that vowel perception differed significantly across learners of four different proficiency levels: elementary (A2), lower intermediate (B1), upper intermediate (B2) and advanced (C1). Zhang et al. (2015) reported that highly proficient Chinese EFL learners perceived English mid vowels better than listeners of lesser proficiency, and that high language proficiency resulted in a more categorical perception of the front vowels than the back ones. Romig (2013) examined the perception of ten English vowels by 14 adult native speakers of Brazilian Portuguese in Canada. Participants were divided into two groups according to their proficiency level in English. The findings showed that proficiency level predicted perception and production accuracy of the English vowels; however, neither proficiency group discriminated contrasts in a native-like manner. Many other studies examine the effects of related factors such as “years of instruction” or “formal instruction” (e.g., Georgiou, 2018a, Piske et al., 2002, Piske et al., 2001) on the acquisition of the L2 vowels, indicating only a minimal positive effect (Flege and Fletcher, 1992, Flege et al., 1995). This small effect is attributed to the facts that pronunciation is not cultivated in most foreign classrooms and that the lack of any training on sound perception and production prevents learners from acquiring a more native-like pronunciation in the L2/foreign language.

The nature of linguistic input also affects perception of L2 sounds. In the case of the foreign language learning, language is acquired through formal instruction and, most of time, this learning is confined to the classroom (Best and Tyler, 2007). Several studies (e.g., Piske et al., 2001, Flege and Liu, 2001) underlined that input quality and quantity are important in order for L2 learners to develop native-like pronunciation. Flege and Liu (2001) argued that, if an L2 learner perceives high quality native-like input that is devoid of foreign accent, they may learn to perceive and produce the sounds of an L2 accurately. But, according to Best and Tyler (2007), most foreign language classrooms have teachers who are not native speakers of the language they are teaching or they are speakers of nonstandard dialects of the language being taught; thus, they do not produce phonetically appropriate variants of the speech sounds. Teachers who speak non-natively accented English are also found in Cyprus (Yazgin, 2007). Accented speech deprives learners of the opportunity to come in contact with native pronunciations of a language. Karakaş (2012) stated that Turkish teachers of English do not produce native-like vowels such as/iː/,/a/or consonants such as/l/and/w/.

The present study seeks to answer two main questions: (a) how does L1 experience of young learners interfere with the acquisition of a foreign language (English), and to what extent can this interference be predicted by comparing the acoustical characteristics of Greek and English vowels? and (b) what is the effect of learning experience on the perception of English vowels by young native Greek speakers? Greek and English languages were chosen because the former language has a smaller vowel inventory than the latter and, in the light of the aforementioned literature, native speakers of Greek are predicted to face difficulties distinguishing specific English vowel contrasts. Few studies investigate the perception of the English vowels by native speakers of Greek. Lengeris, 2009a, Lengeris, 2009b) investigated the assimilation and discrimination of English vowels by adult Greek speakers. He found that Greek speakers tended to assimilate more than one English vowel to the same Greek phonological category. In particular, the English vowels/iː/and/ɪ/were both assimilated to the Greek category/i/, English/ε/and/ɜː/to Greek/e/, English/æ/and/ʌ/to Greek/a/, English/ɑː/,/ɒ/and/ɔː/to Greek/o/and lastly English vowels/ʊ/and/u:/to the Greek phonological category/u/. Furthermore, most of the contrasts showed low to moderate discrimination scores. Also, Lengeris, 2009a, Lengeris, 2009b: 53) argues that English/iː/-/ɪ/and/æ/-/ʌ/contrasts are probably the most challenging ones for native speakers of languages with small vowel inventories, such as Greek.

While a vast number of studies that investigate the perception of vowels by native speakers of languages with smaller vowel systems employ adult participants (e.g, Alispahic et al., 2017, Lengeris, 2009a, Lengeris, 2009b) this study examines the perceptual skills of younger learners who learn English as a foreign language mainly in a controlled classroom environment in Cyprus. Young learners were chosen in order to examine if these learners, who are believed to acquire a foreign language with less effort than adults (Snow, 1987), also face challenges in regard to the discrimination of unfamiliar vowel contrasts. Specifically, the participants consisted of two groups of young learners (below 12 years old); the members of the two groups differed in their experience with English. This allowed us to test the effect of experience on L2 vowel discrimination. Furthermore, the teachers of these learners were native speakers of Greek, not of English. To develop hypotheses regarding the perception of the English vowels by native speakers of Greek, an acoustical comparison of Greek and English vowels as produced by native speakers was carried out.