Perception of Brazilian Portuguese Vowels by Australian English and Spanish Listeners

Many studies on non‐native speech and second language (L2) perception suggest that a second language learner hears with a foreign accent when listening to or perceiving the sounds of the new language (for a review see Strange, 1995, 2007; Escudero, 2005). It is well‐established that the learners’ native or first language influences how they hear and categorise L2 sounds (e.g., Flege, 1995; Flege, Bohn & Jang, 1997; Escudero, 2005; Best & Tyler, 2007). The present study examined group differences in the discrimination of Brazilian Portuguese (BP) vowels by Australian English monolinguals and Spanish listeners who learned English as their L2. It also investigated whether vowel inventory and vowel acoustic properties are equally good predictors of cross‐ language perception difficulty. Acquiring the sounds of a new language is a challenging task for many adult second language (L2) learners. They struggle to attain native‐like proficiency even when surrounded by the second language (Escudero, 2005). Failure to attain native‐like pronunciation (i.e., foreign accented speech) is often related to how learners perceive the sounds of their target language (Best, 1994; 1995; Flege, 1995; Escudero, 2005). However, not all Jaydene Elvin and Paola Escudero 146 learners have the same degree of difficulty when perceiving and producing new L2 sounds. Theoretical models such as the Perceptual Assimilation Model (PAM; Best, 1995), which was extended to PAM‐L2 to incorporate L2 development (Best & Tyler 2007), and the Second Language Linguistic Perception (L2LP) model (Escudero, 2005) propose that the learners’ linguistic background predicts difficulty in learning new L2 sounds. The models explain that when we are introduced to a new speech sound, we filter and categorise it according to the sounds already present in our native language. The present study investigates whether Australian English and Spanish listeners have the same problems in their perception of Brazilian Portuguese vowels. It also aims to determine whether vowel inventory and vowel acoustic properties are equally good predictors of listeners’ cross‐language perception difficulty or lack thereof. The languages investigated in this study vary in the size of their vowel inventories: Spanish has the smallest inventory with only five monophthongs /i, e, a, o, u/, Brazilian Portuguese (BP) has a slightly larger inventory of seven oral monopthongs, /i, e, ɛ, a, o, ɔ, u/, and Australian English (AusE) has the largest vowel inventory with 12 monophthongs, /iː, ɪ, e, eː, ɜ, ɐ, ɐː, æ o, ɔ, ʊ, ʉː/. BP was chosen as the target language as it contains more vowels than Spanish yet less than AusE, which allows us to compare cross‐ language perception in listeners with a smaller versus a larger vowel inventory. Research on the acquisition of target languages with vowel inventories larger than that of the learner’s native language is abundant. One of the most common problems in this scenario is Single category assimilation (SC, PAM; Best, 1995), which occurs when two non‐native sounds are perceived as equally good or poor examples of the same native category (Best, 1994, 1995). Numerous studies (Escudero, 2000; 2005; Flege, 1997; Morrison, 2009) report SC assimilation for Spanish learners’ perception of the English vowel contrast /i‐ɪ/, since they map the two English sounds on to their single native category /i/. It has also been shown that Brazilian Portuguese learners of English have difficulty discriminating the vowels /æ, ɑ and ʊ/ in the contrasts /ɛ‐æ/, /ɔ‐ɑ/ and /u‐ʊ/ (Rauber, et al., 2005), likely because they map them to their native vowels /ɛ/, /a/, and /u/, respectively. Spanish learners of Dutch have also been reported to map /a‐ ɑ/ to their Spanish /a/ and to have difficulty when distinguishing this Dutch contrast (Escudero & Wanrooij, 2010; Escudero & Williams, 2011). Fewer studies have investigated the acquisition of a target language with a smaller vowel inventory than that of the learner’s native language Perception of Portuguese Vowels 147 (Escudero & Boersma, 2002; Morrison, 2003; Vasiliev & Escudero, 2011; Gordon, 2011). When learners have to shift from a larger to a smaller vowel inventory, they commonly exhibit Multiple Category Assimilation (MCA, L2LP; Escudero & Boersma, 2002). MCA occurs when two vowels in a binary contrast are perceived as more than two categories in the L1. For example, Dutch learners of Spanish exhibit MCA for the Spanish front vowels /i/ and /e/. As Dutch has three short vowels /i, ɪ and ɛ/ they will sometimes perceive the Spanish /i/ as the Dutch /i/ or /ɪ/. For the Spanish /e/ it can be perceived by Dutch learners as /ɪ/ or /ɛ/ (Escudero & Boersma, 2002). However, there are competing views as to whether or not MCA is problematic for L2 learners. Escudero and Boersma (2002) suggest that MCA is problematic because it leads to a subset problem where the learner needs to realize on the basis of positive evidence alone that some features or vowels of their own language do not exist in the target language. Additionally, even if the learner overcomes this subset problem, they may have difficulty to stop perceiving the false category (Escudero & Boersma, 2002; Gordon, 2011). Morrison (2003) and Gordon (2011) also found MCA in the perception of Spanish vowels by Canadian and American English learners, respectively. However, unlike Escudero and Boersma (2002) whose Dutch learners had difficulty in categorizing the Spanish front vowels in an identification task, Morrison (2003) and Gordon’s (2011) learners did not exhibit such a difficulty. It has been suggested that the most effective way of predicting discrimination difficulties is through the administration of a perceptual assimilation task, where listeners are asked to classify the vowels of the target language as their own native vowels (Strange, 2007). While this is an effective means of predicting discrimination difficulty, it can only be applied after a perceptual assimilation task has been conducted. Alternatively, one could compare the vowel inventories of the two languages or the vowels’ acoustic properties prior to conducting a perceptual assimilation task. In the present study, we first compared vowel inventories and acoustic properties in order to predict difficulty in the discrimination of BP vowels by Spanish and AusE listeners. Iverson and Evans (2007) explained that learning an L2 vowel system may be fundamentally different for individuals with a larger and more complex vowel system than for those with a smaller and simple vowel system. Although most studies have examined the L2 acquisition of pairwise contrasts (Flege & Bohn, 1997; Morrison, 2009), Iverson and Evans (2007) investigate broader implications of L1 phonetic categories on the learning of entire vowel systems. The authors tested German and Jaydene Elvin and Paola Escudero 148 Norwegian learners who possess a larger vowel inventory and compared their identification of English vowels to that of French and Spanish learners who have a smaller vowel system. While their results demonstrated that there was no fundamental differences in the way the individuals learned the new language, it was shown that listeners with a larger and more complex native vowel inventory (German and Norwegian) were more accurate at identifying English vowels than those with smaller and less complex vowel inventories (Spanish and French). Table 1 shows that AusE has more vowels than BP and Spanish If having a larger vowel inventory is indeed advantageous for vowel perception as demonstrated in Iverson and Evans (2007), AusE listeners should find the discrimination of BP vowels easier than Spanish listeners, since AusE vowels include, among others, most BP vowels, while Spanish lacks /ɛ/ and /ᴐ/. Despite the fact that Spanish and Portuguese belong to the same language family, AusE listeners should be better than Spanish listeners when learning BP vowels if having a larger vowel inventory is the sole predictor. Table 1. A Comparison of the Portuguese, Australian English and Spanish