Taiwan Journal of Linguistics Vol. 14.2, 1-31, 2016 DOI: 10.6519/TJL.2016.14(2).1 PERCEPTION OF ENGLISH LEXICAL STRESS WITH A MARKED PITCH ACCENT BY NATIVE SPEAKERS OF MANDARIN* Shu-chen Ou National Sun Yat-sen University ABSTRACT Three perceptual experiments were conducted to investigate the perception of English lexical stress with a marked nuclear pitch accent by native speakers of Taiwan Mandarin at the phonological and phonetic levels of processing. The stimuli were English disyllabic word pairs differing only in the position of the stress and in the morphosyntactic categories (e.g., PERmit (n.) vs. perMIT (v.)), presented for identification or discrimination. The general finding is that Mandarin-speaking learners of English performed in a native-like manner in the unmarked pitch accent context but not in the marked one. Further examination of their performance in the latter context revealed that while experienced learners failed to match the stress patterns with their morphosyntactic categories in the identification task, they could categorize them in the ABX task, which suggests that they were not phonologically deaf to stress. In contrast, inexperienced learners could not do the same, presumably deaf to stress at the phonological level. Yet, their sensitivity to both within- and across-category stress differences in the AX task indicates that they did not experience stress deafness at the phonetic level. Key words: English prosody, stress deafness, interaction between stress and intonation patterns * The author would like to thank two anonymous reviewers for their valuable and constructive suggestions; the author is responsible for all errors or omissions in this article. The author is also grateful to the Ministry of Science and Technology of Taiwan for sponsoring the research (Project number: NSC 98-2410-H-110-039). 1 Shu-chen Ou 1. INTRODUCTION It is well-documented that the phonetic categories of one’s first-language (L1) have persistent effects on speech perception. The L1 phonological system, which is developed as early as the first year after birth (Werker and Tees 1984), may come to hinder or assist the discrimination of sounds that are non-native to the listener. Observations of this phenomenon have been reported in many second language (L2) studies. For example, it is a challenge for speakers of Japanese to distinguish between the English /r/-/l/, as the two sounds do not contrast with each other phonologically in Japanese (e.g., Goto 1971; Miyawaki et al. 1975). Spanish listeners experience difficulty in discrimination of the English /i/ and /ɪ/ due to the lack of such a contrast in Spanish (e.g., Flege, Bohn, and Jang, 1997; Morrison 2002), whereas German listeners have less difficulty with this vowel distinction because a similar contrast between /i/-/ɪ/ is found in German (e.g., Bohn and Flege 1990; Flege et al. 1997). These examples illustrate how perception is crucially affected negatively or positively by the established native phonetic categories. Languages differ not only in segmental contrasts but also in the phonological use of suprasegmental features, such as duration, pitch, and intensity, to provide contrast in the meaning of words, phrases, or utterances. Parallel to the findings from segmental studies, the perception of non-native prosodic contrasts is also found to be impeded or assisted by L1. Take the cross-linguistic use of pitch as an example. Some languages do not generally contrast lexical items using the variations of pitch height and pitch contour. Consequently, speakers of such languages show considerable problems in discriminating or interpreting Mandarin lexical tones (Gandour 1978; Wang, Sereno, and Jongman 2006). The problems are found, for instance, in French speakers, who are reported to perceive the synthesized tonal continua of Mandarin Chinese in a less categorical way than Mandarin natives do, presumably because pitch variations are not implemented in French to differentiate word meanings (Hallé, Chang, and Best 2004). The inaccurate interpretation of tones is also observed in native speakers of American English, who tend to misinterpret Mandarin lexical tones (which are used to distinguish word meanings) as English intonation patterns (which are used to attach 2 English Stress Perceived by Speakers of Mandarin syntactic or pragmatic meanings to utterance) (Chen 1997; Juffs 1990). These two examples demonstrate the interference of L1 prosodic contrasts with the perception of L2 prosodic contrasts. Nevertheless, speakers of tone languages are shown to be able to perceive words or non-words that contrast in the position of stress (e.g., PERmit vs. perMIT) just as well as speakers of stress languages like English do (Lin, Wang, Idsardi, and Xu 2014; Lukyanchenko, Idsardi, and Jiang 2010). The former’s success can be attributed to the fact that stress has multiple phonetic correlates—pitch, duration, intensity, vowel quality, and spectral tilt—among which pitch also serves as a crucial cue to lexical tone. In other words, the shared phonetic cue somehow aids L1 lexical tone language speakers in their perception of lexical stress, despite the typological differences between their L1 and stress languages. However, it should be noted that the non-native perception of lexical stress reviewed above is primarily investigated when the test words are carried in the unmarked pitch accent context, where the stressed syllable is relatively more prominent than the unstressed syllable in terms of all phonetic correlates—higher pitch, longer duration, and greater intensity. Not much work has been done on how stress is perceived by non-native speakers in the marked pitch accent context (e.g., when stress is not primarily realized in a higher pitch), and there are several reasons to investigate this issue in, for example, English. First, stress is an abstract metrical notion, which dictates the relative prominence relationship between two prosodic constituents (e.g., syllables) (Hayes 1981, 1995; Liberman 1975), but the phonetic realization of the stressed syllable varies with the pitch accent pattern it takes (Beckman and Pierrehumbert, 1986; Ladd 1996). Specifically, in English, a stressed syllable may be associated with either a high or a low pitch. The association of the stressed syllable with a low nuclear pitch accent is exemplified by two of the five common pitch accent patterns in English, as illustrated by Ladefoged (2006). This suggests that the low pitch realization of stressed syllables in English is not uncommon. Therefore, there is a need to investigate the perception of English lexical stress in the marked pitch accent context. 3 Shu-chen Ou Second, the investigation of the perception of stress in the marked pitch accent pattern might contribute to our understanding of specific problems in relation to L2 English pronunciation as well. For instance, in language teaching and learning, the technique of recast, in which the instructor reformulates a student’s utterance by replacing a spotted wrong pronunciation with the correct one (Lyster and Ranta 1997), is sometimes used to help rectify learners’ pronunciation errors. Some studies have indicated that the method is effective in helping the acquisition of the English /r/-/l/ contrasts by L1 Japanese learners (Saito 2014). When recast is adopted in the correction of students’ error in the placement of stress, stress factors such as an instructor’s ability to pronounce the stress in different pitch accent contexts and students’ ability to perceive the stress in these contexts would determine whether recast works. An anecdotal datum collected from the author’s classroom is given below, where the stressed syllable of the word in question, important, is capitalized, and its pitch movement is indicated by impressionistic lines above it. (1) Teacher: What does VIP stand for? Student: Very IMpotant person. [ˋɪmpɔtənt] Teacher: You mean very imPORtant person? [ɪmˋpɔɹtənt] Student: Very IMportant person. [ˋɪmpɔɹtənt] It is obvious that the student exhibited misplacement of stress in the word important, accompanied by the dropping of the post-vocalic /r/, as well. The teacher attempted to recast the errors by uttering the right stress placement in the rising pitch accent pattern, with a more marked pitch accent (a low tone) on the stressed syllable. The recast worked in the case of the omitted post-vocalic /r/, but it was not successful for the 4 English Stress Perceived by Speakers of Mandarin misplacement of the stress because the student might not have been able to capture the instructor’s corrected form. In summary, this paper intends to investigate the perception of English lexical stress in the marked pitch accent context. The following section presents a review of the literature on the perception of stress by non-native speakers, based on which the questions of the current study are generated. Then, the method, the results, and the discussion are presented. Finally, the conclusion ends the paper. 2. LITERATURE REVIEW The present study is motivated by the observation that there is interplay between word stress and pitch accent patterns. According to Beckman (1986), English typologically belongs to a family of languages in which prominent syllables are marked with lexical stress. Stress, which has a number of phonetic correlates including duration, spectral tilt, and segmental quality, interacts with pitch accent patterns in the formation of the pitch contour of the utterance. Specifically, the stressed syllable is to be associated with the nuclear pitch accent, which may be followed by other pitch accents that indicate phrasal or utterance boundaries. The pitch accents are underlying representations for pitch height that can be categorically labeled as high or low tone, and the surface realization of a certain pitch pattern is assumed to be an interpolation of these underlying target tones (Beckman and Pierrehumbert 1986; Beckman, Hirschberg, and Shattuck-Hufnagel 2005). The examples in (2), presented with underlying tones and impressionistic lines that indicate surface pitch movement, show surface pitch patterns that may occur upon the stressed syllable of a word. In the pitch accent pattern that syntactically denotes “declarative” or “affirmative”, the stressed syllable is realized with a high nuclear pitch accent (or a high word tone; H*), which may be optionally followed by a low phrasal pitch accent (or a low phrasal tone; L-) and a low boundary tone (or a low tone indicating the end of the utterance; L%), as shown in Example (2a). In the pitch accent pattern that denotes interrogation, the stressed syllable bears a low nuclear pitch accent (or a low word tone; 5 Shu-chen Ou L*), optionally followed by a high phrasal pitch accent (or a high phrasal tone; H-) and a high boundary tone (or a high tone indication the end of the utterance; H%).1 Among the five common pitch accent patterns in English, as shown in (2) on next page, cited from Ladefoged (2006:126), two have a low nuclear pitch accent (i.e., (2b) and (2c)). In other words, the stressed syllable may coincide with a high pitch, but that is not always the case. Although these examples illustrate that high pitch is not a consistently reliable phonetic cue to stress, it is of considerable interest to investigate whether speakers of a lexical tone language (e.g., Mandarin), who exploit pitch in making lexical distinctions, will be “deaf” to stress when the pitch accent context is relatively marked (e.g., when the stress is applied in the context of a low nuclear pitch accent; L%). The inability of non-native speakers’ to perceive stress contrasts is known as “stress deafness,” a term originally coined to refer to French speakers’ failure to discriminate non-word pairs that differ only in the position of stress (Dupoux, Pallier, Sebastián, and Mehler 1997). Using an ABX task, Dupoux and her colleagues found that French speakers were unable to ascribe a stimulus X (e.g., bópelo) to another stimulus of the same stress pattern, which would be either stimulus A or B (e.g., bópelo or bopélo). However, after the manipulation of the level of the difficulty of the tasks, French speakers were found to be able to distinguish between the stress minimal pairs in the same-different AX task (e.g., responding whether a heard X (e.g., bópelo) is the same as or different from the previously presented A stimulus (e.g., bopélo)). Compatible findings have been further confirmed by the series of studies they have conducted (Dupoux, Sebastián-Gallés, Narrete, and Peperkamp 2008; Peperkamp, Dupoux, and Sebastián-Gallés 1999; Peperkamp and Dupoux 2002). These studies suggest that the stress deafness of French speakers occurs only at the phonological level but not at the phonetic level. The inability of French speakers to phonologically 1 Exceptions have been found in a few varieties of English dialects. For instance, in the Brighton area of the United Kingdom, the pitch accent that indicates interrogation is a falling intonation. However, this should not influence our study because the non-native speakers in this study have acquired major dialects of English (e.g., General American English or Received Pronunciation). 6 English Stress Perceived by Speakers of Mandarin categorize stress is attributed to the lack of phonological representation of lexical stress in French. (2) Common English pitch accent patterns (cited from Ladefoged (2006:126)) a. A simple statement in response to What is her name? impressionistic pitch realization A mé li a | | | H*L- L% underlying pitch accent pattern b. A yes/no question, equivalent to Did you say Amelia? impressionistic pitch realization A mé li a | | | L* H-H% underlying pitch accent pattern c. Addressing Amelia, indicating that it is her turn to speak impressionistic pitch realization A mé li a | | | L*L-H% underlying pitch accent pattern d. A question indicating surprise impressionistic pitch realization A mé li a | | | L+H*L-H% underlying pitch accent pattern e. A strong reaction, reprimanding Amelia impressionistic pitch realization A mé li a | | | L+H* L- L% underlying pitch accent pattern 7 Shu-chen Ou Unlike French, Mandarin is a tone language in which pitch height and pitch contour shapes serve as the most important cues in distinguishing the tones and thus word meanings (Chao 1968; Cheng 1973). This might assist native Mandarin speakers’ perception of English stress in some cases. For example, when the stressed syllable is phonetically realized in a high pitch, L1 Mandarin speakers are able to perceive English stress contrasts (Lin et al. 2014; Lukyanchenko et al. 2010). Nevertheless, when the stressed syllable is associated with a low pitch, the way in which it is perceived by L1 Mandarin speakers is less well investigated. This lack of investigation is notable, since Mandarin speakers are found to experience a rather different type of deafness, as reported in Ou (2010). The study involved two groups of L1-Mandarin learners of English: One included learners who had learned English for more than ten years and the other included learners who had learned the language for less than three years. In a two-forced choice identification task, the participants were generally able to correctly perceive English non-word stress minimal pairs (e.g., FERcept vs. ferCEPT) when the stressed syllable was phonetically realized in higher pitch along with longer duration and slightly greater intensity, but were unable to do so when the higher pitch was replaced by a low nuclear pitch accent (L*). In addition, when lexical items were embedded in the marked rising pitch accent pattern, the more experienced learners tended to respond randomly, but the less experienced group showed a bias toward items with iambic stress. These findings suggest that Mandarin speakers experienced a type of stress deafness that arises from the marked pitch accent context. The issue of stress deafness from the perspective of types of pitch accents, however, is not fully explored in Ou (2010). This is due to the fact that the experiment in the study uses an identification task, in which only a single sound stimulus was presented in each trial, and then a categorization response was required. To perform the task, the non-native speakers had to resort to internal phonological categories stored in the long-term memory. Such a task is argued to be very demanding in the testing of the formation of L2 phonetic categories (Strange and Shafer 2008). A more appropriate approach to investigating the formation of sound categories in a non-native language is suggested 8 English Stress Perceived by Speakers of Mandarin to be an ABX discrimination task, which can accompany an AX discrimination task that assesses non-native listeners’ phonetic sensitivity to non-native phones when the phonetic mode of processing is likely to be adopted. Thus, there is a need to re-examine the perception of English lexical stress by L1 Mandarin speakers using different experimental paradigms. Table 1. The stress deafness constructed based on modes of processing, types of pitch accents and evidence found in previous studies Context Unmarked falling (where the Marked rising (where the stressed syllable is associated stressed syllable is asso- Mode of with a high pitch accent) ciated with a low pitch processing accent) Found in L1 French speakers. (Dupoux et al. 1997; Dupoux Not studied in L1 French et al. 2008; Peperkamp et speakers yet al. 1999; Peperkamp and Phonetic mode Dupoux 2002) Not studied in L1 Mandarin Not studied in L1 speakers yet. Mandarin speakers yet Found in L1 French speakers. (Dupoux et al. 1997; Dupoux Not studied in L1 French et al. 2008; Peperkamp et al. speakers yet 1999; Peperkamp and Dupoux Phonological 2002) mode Not found and not fully Found but not fully explored in L1 Mandarin explored in L1 Mandarin speakers speakers (Lukyanchenko et al. 2010; (e.g., Ou 2010) Lin et al. 2014; Ou 2010) 9 Shu-chen Ou In summary, previous studies have reported cases of stress deafness experienced by native speakers of Mandarin Chinese and French, which are summarized in Table 1 on the previous page. This line of research points to two factors in stress perception: (i) the pitch accent context in which test items are embedded and (ii) the mode of processing that the perceptual experiments require. With these factors in mind, this study aims to investigate L1 Mandarin speakers’ perception of English lexical stress in two pitch accent patterns (i.e., unmarked and marked) and by using two tasks (i.e., ABX and AX) that tap into different types of knowledge (i.e., phonological and phonetic). The questions to be pursued are shaded in grey in Table 1. 3. METHOD 3.1 Materials The stress minimal pairs used in this study were five pairs of disyllabic nouns and verbs with primary stress on the first (i.e., trochaic) or the second (i.e., iambic) syllables: PERmit (n.) vs. perMIT (v.), SURvey (n.) vs. surVEY (v.), IMpact (n.) vs. imPACT (v.), REsearch (n.) vs. reSERACH (v.), and IMport (n.) vs. imPORT (v.), where the stressed syllables are spelled in capital letters.2 All of these words met the criteria established to ensure that they occurred frequently in language use and thus our non-native participants could be presumed to be familiar with them: That is, they each occurred at least one million times in the CELEX database (Baayen, Piepenbrock, and van Rijn 1993) and appeared in the 7000 wordlist for Taiwan High School Students published by the Ministry of Education in Taiwan. Token stimuli of the words were recorded by a trained phonetician, a female native English speaker with a North American accent, by using a Sony Hi-MD recorder. Each of the words was read three times in two carrier sentences: One was an affirmative statement (i.e., I said ______.) to elicit the 2 Pairs differing not only in the position of stress but also in the quality of the vowels (e.g., REcord vs. reCORD) were not included as our main interest was to investigate how suprasegmental features affect the perception of stress. 10