CHUNSHENG YANG

University of Connecticut, USA

AcquisitionofMandarinLexicalTones:TheEffectofGlobalPitchTendency

CHUNSHENG YANG

University of Connecticut, USA

This paper discusses the acquisition of Mandarin lexical tones by American second language (L2) learners. The analysis of the tone errors in the reading passage and the spontaneous speech showed that the surface Tone 4 error is the most frequent at the prosodic word (PW)-initial position in both the intermediate and the advanced learner groups. Attempts were made to account for this tone error pattern within some second language acquisition (SLA) frameworks, but in vain. It was argued that this tone error pattern may be due to the high pitch tendency at the phrase- or sentence-initial position in speech. Although this is more of a general pattern across languages, the production of such a tendency by L2 learners superseded the underlying tones, hence the surface Tone 4 errors. The special status of Tone 4 in L2 speech corresponds to the special status of the same tone in the first language (L1) speech (Wan, 2007). However, it was cautioned that such similarity should not be taken as evidence for the similarity in L1 and L2 acquisition in that different processes are involved in L1 and L2 production.

Tone acquisition, Mandarin Chinese, Tone 4 (falling tone), prosodic word (PW)-initial position

SECOND LANGUAGE PHONOLOGICALACQUISITION

Studies on second language (L2) acquisition have shown that L2 segments (i.e., vowels and consonants) may pose difficulty for L2 learners (Best, 1995; Eckman, 1977; Flege, 1987; Flege, 1995; Flege, Schirru, & MacKay, 2003; Major, 1992; Strange, 1995). In contrast with the abundant studies on L2 segments, studies on L2 suprasegmentals are rather limited (see Zampini, 2008, for a brief review of studies on L2 segments and suprasegmentals). Among the limited studies on L2 suprasegmentals, the production and perception of lexical tones, as in Mandarin Chinese, Taiwanese and Thai, have received great attention and have proved to be difficult for L2 learners of non-tonal languages (Bent, 2005; Chen, 1997, 2000; Miracle, 1989; Shen, 1989; Sun, 1998; Wangetal., 1999, 2003; Wayland & Guion, 2004; White, 1981; Winke, 2007; Zhang, 2007, 2010).

L2 acquisition of lexical tones is different from that of segments. Depending on the similarity between L1 and L2 segments (i.e., an L2 sound can be new, similar or identical to an L1 sound), L2 learners may establish a new phonetic category if the L2 sound differs from the closest L1 sounds and the bilinguals discern at least some of the phonetic difference (as predicted by the Speech Learning Model, SLM) (Flege, 1995; Flegeetal., 2003). However, category formation may be blocked by equivalence classification by which a single phonetic category is used for both L1 and L2 sounds. When it comes to lexical tones, L2 learners whose native languages do not have tonal contrast usually do not directly assimilate lexical tones to L1 prosodic categories (Bent, 2005). Within the framework of Perceptual Assimilation Model (PAM) (Best, 1995), Bent found that Mandarin lexical tone pairs were perceived as “uncategorizable” by naïve L2 learners, the perception of which depended on the acoustic similarity between the categories and their proximity to native categories. It was also found that the English listeners processed lexical tones primarily in the acoustic mode, while the native Mandarin listeners processed tones mainly in the linguistic and phonetic mode (i.e., categorical perception) (Bent, 2005).

Most studies on both L1 and L2 tone acquisition have shown that lexical tones were acquired in similar orders. For example, the high level tone and the falling tone in Mandarin Chinese were acquired earlier than the rising tone and the low dipping tones (Jeng, 1985; Li & Thompson, 1977; Sun, 1998; Wong, Schwartz, & Jenkins, 2005; Zhang, 2007). Just as the examination of L1 tone errors can shed light on the phonological organization of tones (Wan, 2007), the examination of L2 tone errors can show how L2 learners process lexical tones, and how tone production differs between L1 and L2, which will, in turn, contribute to our understanding of the interaction of L1, L2, and linguistic universals and specifics.

MANDARIN TONES AND ENGLISH PROSODY

Mandarin Chinese is a tone language, in which tones are lexically specified, namely the fundamental frequency (F0) or pitch pattern over a syllable can be used to distinguish the lexical meaning of words. There are four lexical tones in Mandarin Chinese, namely, the high level tone (Tone 1 or T1), the rising tone (Tone 2 or T2), the low dipping tone (Tone 3 or T3), and the falling tone (Tone 4 or T4). It should be pointed out that Tone 3 often surfaces as a low tone except in isolation (Duanmu, 2000, p. 95). In Chao’s (1930) five-level scale, Tone 1 through Tone 4 are represented as “55”, “35”, “214”, and “51”, respectively.

Within the framework of Autosegmental-Metrical (AM) phonology (Beckman, 1996; Pierrehumbert & Beckman, 1988; Ladd, 1996, 2008), there are two levels of phrases in English, namely, intermediate phrase (ip) and intonational phrase (IP), and English prosody can be described with three pitch events: pitch accent, phrase accent, and boundary tone. Pitch accent falls on stressed syllables, marking the prominence of the syllables (H*, L*, L+H*, etc.). An intermediate phrase has one pitch accent or more, together with a phrase accent (H-, L-). The phrase accent has a scope over the entire intermediate phrase (Pierrehumbert & Hirschberg, 1990). The phrase accents are associated with coherence and cohesion in the discourse. There are two types of phrase accents in English. The high phrase accent (H-) indicates that “the current [prosodic] phrase is to be taken as forming part of a larger composite interpretive unit with the following phrase,” whereas the low phrase accent (L-) emphasizes the separation of the current phrase from the subsequent phrase. An intonation phrase may have one or more intermediate phrases, and ends with a boundary tone, which can be either high (H%) or low (L%). A series of pitch accent, phrase accent, and boundary tone forms a tune in English. Tunes in English are closest to lexical tones in Mandarin in terms of pitch contours (Bent, 2005, p. 21). Bent showed that all lexical tones in Mandarin may correspond to tunes in English. However, the details of pitch realization, especially the pitch range, the alignment with stressed syllables, and the amount of syllabic materials over which the contour can be placed, differ drastically in these two languages (Bent, 2005, p. 31). Maybe because of these differences, Bent found that naïve English listeners did not assimilate lexical tones to English sound categories.

L1 AND L2 ACQUISITION OF MANDARINLEXICAL TONES

In L1 acquisition of Mandarin tones, studies have shown that the four lexical tones are acquired in this order: Tone 1, Tone 4, Tone 2, and Tone 3, in which Tone 3 is the most difficult and acquired the latest (Jeng, 1985; Li & Thompson, 1977; Wong, Schwartz, & Jenkins, 2005). The acquisition order of the four tones by L2 learners follows a similar pattern, namely, Tone 2 and Tone 3 are the most difficult for L2 learners, and accordingly are the last to be acquired (see Sun, 1998; Zhang, 2007 for two detailed reviews). Bent (2005) categorized tone errors in previous studies into different groups: (1) pitch range errors, namely learners’ pitch range was smaller than that of that native Mandarin speakers; (2) pitch register errors, namely the tone targets in learners’ production fell short of the targets, both high and low targets (Chen, 1997; Miracle, 1989; Wangetal., 2003); (3) tone contour errors, namely the wrong contour was produced in the form of incorrect direction of pitch contour or substituting a static tone for the dynamic tones (Wangetal., 2003). Bent (2005) asked her English speakers who had no prior experience with Mandarin Chinese to imitate Mandarin lexical tones in isolation in the T1_T4 frame. She found that tone production/imitation had very high degree of accuracy, although acoustical analysis showed that there were differences, such as the compressed pitch range in Tone 2 and Tone 4, etc.

In terms of the frequent surfaced tone errors in L1, Wan (2007) found that the falling tone (i.e., Tone 4) was the most frequent tone to be replaced by other tones (28.6%), and also the most frequent tone to replace other tones (32.5%). Wan argued that the special status of the falling tone in tone errors was due to the frequency of the falling tone in both written and oral data in Mandarin (35.444%, much higher than the other three tones) as well as the less physiological efforts required in the production of the falling tone (Ohala & Ewan, 1973; Sundberg, 1973). Interestingly, studies on L2 tone acquisition shows that the falling tone also has a special status. Zhang (2010) found in her disyllabic elicited data that high level tone—falling tone pair (Tone 1-Tone 4 in Mandarin) were the most frequently produced tone sequence (note that Zhang did not use Tone 3 in her stimuli). In contrast, Yang (2011) found that there were frequent surfaced Tone 2 and Tone 3 errors in the elicited L2 conversations. More specifically, Tone 2 tended to occur at the prosodic phrase-initial position, whereas Tone 3 tended to occur at the prosodic phrase-medial position. It was argued that the frequent surfaced Tone 2 and Tone 3 were attributed to the transfer of English rising intonation and the stress pattern.

Most studies on tone acquisition by American L2 learners have attributed tone errors to the interference of English intonation, at least to some extent. However, Winke (2007) argued that, since L2 learners of different L1 backgrounds demonstrated similar sensitivity to tone contour differences and displayed similar patterns of tone errors, the difficulty in tone production is due to “the overall novelty of the tone system to the learners”. Taking the perspective of phonological universals (Chomsky, 1980, 1981, 1999), Zhang (2010) claimed that the tone errors in her study were not directly derived from English intonation system, nor from the Chinese grammar, implying the existence of universals in tone acquisition.

The above discussion shows that previous studies have resulted in different findings, especially with respect to pattern of the surfaced tone errors (i.e., the focus of this study). One possible reason for the different findings may be the different data collected in different studies. For example, Winke (2007) and Zhang (2010) used disyllabic words in their study, while Yang (2011) used read conversations. Meanwhile, different approaches of error analysis may lead to different patterns. For example, Yang (2011) focused on the positional effect of the surfaced tone errors in the intermediate and advanced learners, while Zhang (2010) did not take the position where tone errors occurred into account. More importantly, although some previous studies (such as Chen, 2000) examined the tone errors in spontaneous speech, little has been done in identifying the patterns of the surfaced tone errors.

To fill in the gaps in previous research, the present study examines the surfaced tone errors in the reading passage and the spontaneous speech by two levels of American L2 learners. The following questions are addressed:

(1) What are the patterns of the surfaced tone errors in the reading passage and the spontaneous L2 speech?

(2) Of the four lexical tones, which tone(s) are frequently produced to replace other underlying tones?

(3) What are the differences in the patterns of the surfaced tone errors in the reading passage and the spontaneous speech?

(4) How can we go about explaining the patterns of the surfaced tone errors in L2 speech?

Considering the scarcity of previous studies on tone acquisition in reading passage and spontaneous speech and the exploratory nature of this study (i.e., to examine the patterns of the surfaced tone errors), it would be difficult to make specific hypotheses/predictions for the above questions. Based on the review of previous studies above, it is predicted that Tone 2 and Tone 3 may pose more difficulty for L2 learners than Tone 1 and Tone 4. Meanwhile, if L2 acquisition is similar to L1 acquisition, Tone 4 may be frequently produced to replace other tones in L2 speech and there will be more tone errors in the spontaneous speech than in the reading passage.

METHODOLOGY

RecordingStimuli

In this study, two types of data were collected, namely, passage reading and spontaneous speech. The reading passage was adapted from the Chinese textbook for the intermediate learners to guarantee that they knew all the characters (see Appendix). The spontaneous speech was elicited by presenting subjects with four pictures related to the story ofTheNorthWindandtheSun. Subjects were instructed to tell the story according to the pictures.

Participants

Ten intermediate-level (2nd-year and 3rd-year learners) and ten advanced-level (4th-year or above learners) American L2 learners were recruited to participate in this study. The L2 learners were recruited from the L2 learners at the Department of East Asian Languages and Literatures in an American mid-western public university. All subjects reported no speech or hearing problems. They were paid10 for their participation in the recording. Table 1 presents the background information of the subjects in this study, including gender, further categorization, age, duration of Chinese learning (years), and duration of study abroad in China or Taiwan (years).

Table 1 Detailed information of subjects in the three groups

Procedure

The recording was conducted in a sound-attenuated studio at the same university. Prior to recording, the researcher gave instructions to the subjects. Then the subjects signed the consent form. The subjects could ask any question related to the study. They could also withdraw from the recording any time without any penalty.

In recording, a microphone was placed between the subject and the laptop. The subject sat comfortably behind the microphone. The subject first read the passage presented through Powerpoint slides on a laptop. Then the subject was shown two pictures of the tale,TheNorthWindandtheSunfirst, and then retold the story according to the pictures. The recording input was digitized at 44.1 kHz with a 24-bit resolution. The whole recording lasted less than 30 minutes.

JudgmentofToneErrors

The judgment of tone errors in this study was done in context. The author judged all the tones, with a month’s interval between the two judgments. The agreement rate of tone correctness judgment was 97.5%. For the cases of judgment discrepancy, the tones were listened to again and a final judgment was made. Afterwards, another judge, a Chinese Ph. D student in phonetics, was hired to judge the correctness of the tone data. The tone judgment agreement rate between the hired judge and the author was 96.3%. In the cases of discrepancy in judgment, the two judges listened to the original tones in the utterances again and made a final judgment. All the surface tone errors were classified into one of the following categories: high tone (i.e., Tone 1), low tone (i.e., Tone 3), rising tone (i.e., Tone 2), and falling tone (i.e., Tone 4).

In order to examine whether the surface tone errors display any positional effect, the positions where tone errors occurred were also included in the analysis. With respect to prosodic structure, English and Chinese share similar prosodic constituents: the syllable, the foot, the prosodic word, the prosodic phrase/intermediate phrase, and the intonation phrase/breath group (Chu & Qian, 2001; Nespor & Vogel, 1986; Pierrehumbert & Beckman, 1988; Selkirk, 1984; Speeretal., 1989; Tsengetal., 2005; Wangetal., 2003). Only prosodic word was considered in this study due to its particular status in Mandarin Chinese (Yip, 1980). The prosodic word is the minimal unit of articulation in normal communication (Levelt, 1989). A prosodic word often contains a single lexical word together with associated unstressed functional words (i.e., auxiliaries, determiners, conjunctions, or prepositions). In Mandarin Chinese, prosodic words are mostly disyllabic or trisyllabic. Prosodic words in Mandarin usually start with a stressed, full-toned syllable and function as the rhythmic unit (Yip, 1980), and the domains for morphological and phonological processes, such as tone sandhi (Dai, 1998; Feng, 1996, 1997, 2001). Prosodic words are usually accompanied by acoustic cues (Pengetal., 2005; Tseng & Chou, 1999; Tsengetal., 2005; Yang & Wang, 2002; Yang, 2011). Considering the nature of L2 speech, such as slow speech tempo and disfluency, this study focused on prosodic words when the position of tone errors was examined.

RESULTS

ToneErrorsintheReadingPassage

Tone errors①in the reading passage were first discussed.

Repeated measure ANOVA was conducted on the count of different types of tone errors with the group as the between-subject variable and the position and the error type as within-subject variables. ANOVA results show that there are main effects for both the position (F3,54=20.84,p<0.000) and the error type (F11,198=11.21,p<0.000), and interaction between the position and the error type (F33,594=14.29,p<0.000). However, no significant difference was found for the group factor. Namely, the intermediate and the advanced learner groups performed similarly in their error patterns. Tukey post-hoc tests on the tone errors at different prosodic positions show that there were more tone errors at the PW-initial position than at the PW-medial, the PW-final and standalone position (pis less than 0.000 in all cases). Tukey post-hoc tests on different types of tone errors show that the tone error pattern “T1-T4” was significantly more than any other tone error pattern, namely Tone 1 tended to be produced as Tone 4 very frequently. The interaction of the prosodic position and error patterns can be shown in Figure 1. As shown in Figure 1, there were much more tone errors at the PW-initial position and the error pattern “T1-T4” was the most frequent at this position.

Figure 1 Interaction of prosodic position and tone error patterns

The repeated measure ANOVA above did not show the significance for the group variable. However, it was observed that some fixed learners in either group produced the majority of tone errors in each group. Therefore, the subjects in each learner group were further categorized into two sub-groups, namely, the higher and lower subgroups, as shown in Table 1. A t-test on the error counts across the two subgroups was conducted to see whether there was any significance in the counts of tone errors. The t-test result showed that the learners in the lower subgroup produced significantly more tone errors than those in the higher subgroup (t608=4.74,p<0.000).

Although there was only significance between the tone error pattern “T1-T4” and the other tone error patterns, but no significance between the other tone error patterns, it was worth examining what were the other tone error patterns in the two learner groups. Table 2 shows the counts and percentages of each type of tone errors in the reading passage by the two learner groups.

Table 2 Tone errors in the reading passage by intermediate and advanced learners

(*In each tone error, the first tone is the intended tone, whereas the second one is the produced tone, namely, the tone error)

It can be seen from Table 2 that “T1-T4” is the predominant error type in both the intermediate and the advanced learner groups, as shown in the statistical analysis above. Following “T1-T4”, is the “T2-T4” error pattern. Other relatively less frequent tone error patterns include “T3-T4”, “T3-T2” and “T2-T3”, which might occur only in one learner group, though. Put together, it can be seen that Tone 4②(i.e., the falling tone) is the most frequently surfaced tone error in the reading passage by both learner groups. As a matter of fact, all other tones could be produced as a Tone 4. Besides the Tone 4 errors, T3’s were frequently produced as Tone 2 and the other way around, implying the confusion between these two tones.

Due to the similarity between Tone 2 and Tone 3 (Huang, 2001; Shen & Lin, 1991), the confusion between these tones is no surprise. It was also observed from Table 2 that Tone 3 tended to be produced as Tone 2 in both learner groups, whereas Tone 2 tended to be produced as Tone 3 more frequently in the advanced group only. The error patterns of Tone 2 and Tone 3 suggest that the confusion between Tone 2 and Tone 3 may be one-directional, namely Tone 3 tended to be produced as Tone 2, not the other way around. However, it is interesting to find that Tone 4 is the most frequent surfaced tone error, differing from previous studies (Yang, 2011), which find that Tone 2 and Tone 3 are the most frequent surface tone errors in the elicited L2 conversations.

As discussed above, Tone 4 (or the falling tone) enjoys a special status in both Mandarin Chinese and other tonal languages. Actually, the L2 tone errors found in the reading passage in this study are similar to those found in L1 Taiwan Mandarin (Wan, 2007). A general discussion on this will be provided later.

ToneErrorsintheSpontaneousSpeech

The tone errors in the spontaneous speech were analyzed in this section. Ideally, statistical analysis should be conducted to compare the tone errors across subject groups, prosodic position and error types, as in the reading passage. However, subjects in the two learner groups varied significantly in their retelling the story, in terms of both word choice and length. As a result, it would be meaningless to compare the counts of different types of tone errors. Therefore, no statistical analysis was conducted here. Instead, the counts and percentages of each type of tone errors between the two learner groups were compared and contrasted.

Table 3 presents the counts and percentages of each type of tone errors in the spontaneous speech by the two learner groups. Similar to the tone errors in the reading passage, Table 3 shows that “T1-T4” is the most frequent tone error pattern in both learner groups, followed by “T3-T4”. Other less frequent tone error patterns include “T3-T2”, “T4-T2”, and “T4-T3”, which only occurred in one learner group. Similarly to the error pattern “T2-T3” in the reading passage, Tone 2 was seldom produced as Tone 3 in the spontaneous speech. However, the error pattern “T3-T2” was relatively more frequent in the spontaneous speech than in the reading passage, especially by the intermediate learner group. Thus, the error patterns between Tone 2 and Tone 3 again indicate that the confusion between Tone 2 and Tone 3 is one-directional. Another difference in tone error patterns between the reading passage and the spontaneous speech lies in the errors related to Tone 4. In the reading passage, Tone 1, Tone 2, and Tone 3 tended to be produced as a Tone 4 in either learner group, whereas, in the spontaneous speech, only Tone 1 and Tone 3 tended to be produced as Tone 4. Furthermore, Tone 4 was seldom produced wrongly in the reading passage (<10%). However, in the spontaneous speech, Tone 4 was frequently produced as Tone 3 by the intermediate learner group and was frequently produced as Tone 2 by the advanced learner group. Thus, apart from the less frequent “T2-T4” error pattern in the spontaneous speech, it seems that the tone error patterns in the spontaneous speech are more similar to the tone errors in L1 speech (Wan, 2007).

Table 3 Tone errors in spontaneous speech by intermediate and advanced learners

Similarly, the distribution of tone errors at different prosodic positions in the spontaneous speech was examined with results being presented in Table 4.

As shown in Table 4, the majority of tone errors in the spontaneous speech occurred at the PW-initial position and a considerable amount of errors occurred at the PW-final position.

Table 4 Tone errors in spontaneous speech by prosodic position

Figure 2 Tone error distribution at the PW-initial position in spontaneous speech by the two learner groups

Also, the distribution of tone errors at the PW-initial position in the spontaneous speech was examined, as shown in Figure 2. Similar to the distribution of tone errors at the PW-initial position in the reading passage, the error pattern “T1-T4” is the most frequent. However, we also observed some differences in tone errors from those in the reading passage. The error pattern “T3-T2” in the intermediate group and the error pattern “T3-T4” in both learner groups were also relatively more frequent in the spontaneous speech that those in the reading passage.

DISCUSSIONS

SummaryoftheFindings

Statistical analyses on the error counts in the two learner groups showed that there were main effects for the prosodic position and the error type, as well as the interaction between the position and the error type. Further post-hoc tests revealed that the error pattern “T1-T4” was the most frequent in both the reading passage and the spontaneous speech. Meanwhile, the error pattern “T2-T4” in the reading passage, and the error pattern “T3-T4” in the spontaneous speech were also relatively frequent, although not necessarily statistically significant. It was also found that the error pattern “T3-T2” were more frequent than the pattern “T2-T3” in both groups, suggesting the one-directionality of the confusion between these two tones. Considering that both the low-dipping variant of Tone 3 and Tone 2 had the rising part, it was easy to understand why such errors occurred. The main reason for this error pattern was that the learner did not drop their pitch to a low point before raising their pitch, namely their turning point of pitch in Tone 3 is similar to that in Tone 2, thus producing Tone 3 as Tone 2, at least perceptually. Furthermore, except for the “T2-T4” error pattern in the spontaneous speech, it seemed that the error patterns related to Tone 4 in the spontaneous speech were more similar to those in the L1 speech (Wan, 2007), namely, Tone 4 tended to be produced as other tones and tended to replace other tones as well.

Furthermore, the analysis of tone errors showed that most tone errors occurred at the PW-initial position in both the reading passage and the spontaneous speech for both learner groups. Further analysis of the tone errors at the PW-initial position in both types of speech indicated that “T1-T4” was the most frequent error pattern for both learner groups.

Although there was no statistical difference in the tone errors between the intermediate and advanced learner groups, the further categorization of the two groups into lower and higher subgroups revealed that there was significant difference between the lower and the higher subgroups. It seemed that the level of courses that the subjects were taking did not constitute a reliable indicator of the actual level of learners’ proficiency. Therefore, caution should be made when the findings in this study are generalized to a specific learner group.

WhyDifferentToneErrorsinDifferentTasksandData?

Yang (2011) analyzed the tone errors in different tone sequences in the elicited conversations and found that Tone 2 and Tone 3 were the most frequent tone errors. Tone 2 tended to occur at the prosodic phrase-initial position, whereas Tone 3 tended to occur at the prosodic phrase-medial position. Therefore, it would be interesting to compare the findings in Yang (2011) and the findings in this study. In comparison with Yang (2011), the findings in this study were completely different. But why? Although the same subjects participated in Yang (2011) and the present study, the tasks involved were different. In Yang (2011), two subjects worked together in recording the conversation. Although instructed to read as naturally as possible, the learners were rather passive in completing the task. Also, the alternation between two subjects somewhat prevented the subjects from using more natural speech. More importantly, thepinyinRomanization, which was meant to assist them, may have adversely affected the data collected. It may happen that learners just read thepinyinwithout processing the content itself. As a result, their tone production did not tap into their L2 phonology and the error patterns found in such data were sure to differ from those in the spontaneous speech. Even so, Yang (2011) sheds new light on the acquisition of tones by L2 learners, in the reading conversations. In contrast, when L2 learners were reading a passage, presented in Chinese characters only, the task tapped into their L2 phonology sincepinyinRomanization was not provided. However, different from the spontaneous speech, the reading passage may not have been very natural, which, in turn, may have hidden some elements in more natural speech. In the spontaneous speech, learners produced more natural speech, the task of which definitely tapped into L2 phonology and could reveal a better and clearer picture of L2 phonology. That may be one of the reasons why we found the tone error patterns in the spontaneous L2 speech were more similar to those found in the native speech (Wan, 2007).

WhysoManyTone4Errors?It is worth examining why Tone 4 is the most frequent tone error to replace other tones, especially Tone 1, at the PW-initial position. Wan (2007) attributed the frequent surfaced Tone 4 errors to the high frequency of Tone 4 in Chinese lexicon. Although it is tempting to attribute the frequent surfaced Tone 4 errors in L2 Chinese to the high frequency of Tone 4 in the lexicon as well, caution should be made in that L2 learners may not have been exposed to the same language experience as L1 speakers. Meanwhile, there are no statistics of the frequency of the four tones in the L2 Chinese textbooks. In addition, an analysis of the distribution of the tones in the reading passage showed that T1 accounted for 26.1% of all the tones, T2 25.6%, T3 16.1%, Tone 4 19.1%, and neutral tones 13.1%. Therefore, at least Tone 4 was not the most frequent tone in the reading passage. Therefore, frequency may not be able to explain the frequent surfaced Tone 4 errors in the L2 speech. However, it should be pointed out that Tone 1 has the highest percentage in the reading passage, which may be argued to explain why Tone 1 was produced wrong most frequently. However, there is still no explanation for the frequent surfaced Tone 4 errors. Then, can second language acquisition models explain the frequent surfaced Tone 4 errors?

According to the tone markedness scale, Tone 2 is more marked than Tone 4, which, in turn, is more marked than Tone 1 and Tone 3. Also, as the literature review above showed, Tone 4 has been shown to be the easiest to be acquired in both L1 and L2. It is interesting to see that the tone in the middle of the markedness scale tended to be produced to replace all other tones, both more marked and less marked ones. Both the Markedness Differential Hypothesis (MDH, which predicts the L2 sounds which are more marked than L1 counterparts are more difficult to acquire) (Eckman, 1977) and the Speech Learning Model (SLM) requires the comparison between the L1 and L2 sounds, namely, the lexical tones in Mandarin and the counterparts in English (i.e., tunes), with respect to the markedness and similarity. However, as mentioned earlier, even though the pitch contours equivalent to the four lexical tones in Mandarin Chinese can be found in English, there are so many differences between them, especially the domains on which the lexical tones and the English tunes occur. In Chinese, tones fall on single syllables, whereas the pitch contours or tunes in English fall on a span of several syllables. As a result, listeners’ tone production and perception are inevitably influenced by English phrase- and utterance-level prosody, as shown in Chen (1997, 2000), Chiang (1979), and White (1981). Thus, it is difficult, if not impossible, for the L2 Mandarin learners to map lexical tones to English tunes. This hypothesis was borne out in Bent’s discussion of tone perception. Therefore, it seems that both MDH and SLM cannot make predictions for the potential tone error patterns. Furthermore, even if predictions could be made based on MDH, SLM or other SLA models, the predictions only concern the difficulty of different lexical tones, which says nothing about the surfaced tone error patterns, namely, which tones tended to be produced as which tones more frequently.

Since SLM, PAM and MDH do not account for the frequent error patterns in this study, can the error patterns be attributed to the transfer or interference of English prosody? Considering the fact that the surfaced Tone 4 errors were stressed in most cases, it seems that the carrying syllable has a high pitch accent followed by a low phrase accent (i.e., H* L-). However, H*L- usually spans more than one syllable in English, while in L2 Mandarin, the tone events occur on a single syllable. Also it is worth noting that at the beginning of both an intermediate phrase and an intonation phrase in English, H*L- does not occur frequently, actually very rarely in that H*L- usually indicates a declarative statement and, thus, tends to occur in the middle or toward the end of an utterance (Pierrehumbert & Hirschberg, 1990). Thus, the surface Tone 4 errors at the PW-initial position could not be attributed to the transfer of English prosody, either. It seems that it is difficult to explain the frequent Tone 4 errors in any phonological or second language acquisition models.

Xu (1994) showed that the focused syllable in both Mandarin Chinese and English usually are accompanied with expanded pitch range and post-focus pitch compression. It has been found that the beginning phrase in English has higher F0 and a larger pitch range than the following phrases, even when the focus is not on the first syllable/phrase (Tsengetal., 2005). Even though there is no evident higher pitch at the beginning of a Chinese sentence, the existence of declination and the raised pitch register associated with the discourse-initial topic in Mandarin Chinese suggests that Mandarin Chinese is similar to English in this aspect, namely, both languages tend to have sentence-initial higher pitch register and larger pitch range. However, in Mandarin Chinese speech, even if there is higher pitch register at the phrase-initial position, the underlying pitch contours are still maintained. As for the L2 learners, when they produce the discourse-initial pitch event, they produce the higher onset pitch, followed by an immediate pitch drop. In this way, the underlying Mandarin tones are superseded by the global pitch patterns, and the production of such global pitch pattern over the PW-initial syllable results in the perception of the surface Tone 4 error. To illustrate the above point, two utterances containing the Tone 4 errors are provided in Figures 3 and 4.

Figures 3 and 4 show that the PW-initial Tone 4’s (i.e., the syllables in the red circles) were stressed. In Figure 3, both surfaced Tone 4’s were produced with raised pitch height and pitch range, and were followed by compressed pitch range, patterning similarly to the focus in English, Chinese, and other languages. In Figure 4, the first PW-initial Tone 4 was produced with raised pitch height, followed by compressed pitch range, while the other PW-initial Tone 4’s were stressed, but were not focused.

In this sense, one possible interpretation for the frequent Tone 4 errors at the PW-initial position is that L2 learners tend to place a narrow focus on the PW-initial syllable, due to the global pitch tendency across languages. However, the production of the global pitch pattern over the first syllable supersedes the tone on that syllable, hence the production of the surface Tone 4 error.

Figure 3 Tone 4 error produced by an intermediate learner

Figure 4 Tone 4 error produced by an advanced learner

As for why the error pattern“T1-T4” is the most frequent in both reading passage and the spontaneous speech, it might be due to the shared high targets in these two tones. For L2 learners, to maintain the high flat pitch contour over Tone 1 is not easy. And, since Tone 1 starts at high pitch, it is natural for the pitch to drop, hence the production of “T1-T4” errors. As mentioned above, it may be argued that the frequent “T1-T4” tone error was related to the frequency of Tone 1, at least in the reading passage. However, Tone 2 has almost equal frequency as Tone 1 in the reading passage, but do not lead to the frequent surfaced Tone 4 errors. Also in the spontaneous speech, Tone 1’s may or may not have such frequency advantage; therefore the frequent tone error “T1-T4” cannot be attributed to the relatively more Tone 1’s in the reading passage.

The frequent tone error patterns in this study suggested that the most frequent tone errors were due to the global pitch tendency across languages, namely linguistic universals. Wan (2007) found that in the tone errors of L1 Taiwan Mandarin, not only was Tone 4 frequently replaced by other tones, Tone 4 was also the tone to replace other tones most frequently. Our findings here show that Tone 4 also frequently replaced other tones at the PW-initial position in L2. It may be argued the patterns of Tone 4 errors in L1 and L2 provide evidence for the similarity between L1 acquisition and L2 acquisition. Actually, the analysis of tone errors in the reading passage and the spontaneous speech have shown that the tone error patterns in the spontaneous speech are more similar to the error patterns in the native speech, as shown in Wan (2007). However, although the similar tone error patterns in L1 (Wan, 2007) and in the spontaneous speech in this study suggest that L1 and L2 acquisition are similar, caution should be made before any generalizations between L1 and L2 acquisition are to be made. The surface similarity of two things may not mean that they are derived from the same internal mechanisms. Actually, the frequent Tone 4 errors in L1 and L2 may have different motivations. For native Mandarin speakers, tones are part of their phonology and the production of tone errors is mainly the result of slips of tongue, which reflects the phonological processes in their L1 phonological system. However, for L2 speakers, tone production has not been internalized in their phonological system. Thus, their tone production was influenced by various factors, such as transfer and interference of L1 prosody, linguistic universals, attention, working memory in speech production, and even their mastery of the tones. Therefore, the similarity in tone errors in L1 and L2 does not necessarily provide evidence for the similarity of L1 and L2 acquisition.

ContextualityandDirectionalityChen (2000), Maddieson (1978), Wan (2007), and Yip (2002) found that in both Thai and Mandarin Chinese, most tone errors are contextual and have directionality, namely, most tone errors have preservative preferences to produce previously occurring tones. A careful examination of the tone errors in both reading passage and the spontaneous speech showed that most tone errors in this study are non-contextual, which means that the tone errors in L2 speech may be motivated differently from the tone errors in L1. For native Mandarin speakers, their tone errors may be just “slips of tongue”, which are usually context-dependent. However, for L2 learners, their tone errors can be slips of tongue, can be influenced by L1 prosody, acquisitional universals, and articulatory mechanism or other factors. Thus, the tone errors in L2 are not contextual and do not have directionality.

ResearchImplications

As discussed above, studies on the acquisition of Mandarin lexical tones have found different patterns of tone errors. The divergence in research design, especially with respect to the type of data collected and the approach of data collection, could explain many of the differences. For example, tone errors in the monosyllabic reading task and in the spontaneous speech are very likely to be different. Meanwhile, the proficiency level of the participants is another factor that contributes to the difference in the patterns of tone errors. At the beginning stage, leaners’ tone production was still inconsistent and had not been internalized, thus displaying different patterns than learners at higher levels. Another factor which may play a role is the fossilization of tones of some learners, especially some advanced learners. Therefore, caution should be made to distinguish tone errors which were made due to the influence of L1 prosody, from those which were due to the linguistic universals, and those which had been internalized in the learners’ phonology. Furthermore, learners’ background, such as language learning aptitude, native language, music background and imitation/performance capability may also correlate with learners’ tone production performance.

The divergent difference in studies on tone acquisition suggests that tone acquisition should be investigated in a more systematic and complex design, namely the longitudinal design. To date, nearly no longitudinal study on the acquisition of L2 lexical tones have been conducted (Sun (1998) is close to a longitudinal study, but only spans one academic semester). It is expected that the longitudinal design on the tone acquisition can help identify the developmental pattern of lexical tone acquisition by L2 learners. At the same time, the comparison and contrast of tone acquisition by learners from different L1 backgrounds in different types of tasks/data in the longitudinal design can help distinguish language universals, language-specific acquisition patterns, and individual difference. Therefore, the longitudinal design of tone acquisition should be implemented in further studies.

It should also be pointed out that most of the tone errors in both the intermediate and the advanced learner groups were made by some fixed learners, namely, the lower level learners in either learner group. That is to say, the lower level learners in either learner group had great difficulty in tone production, whereas other learners in either group only produced very few tone errors. Therefore, both the “intermediate” and the “advanced” learner groups in the discussion above mainly referred to the lower subgroup in either group, whereas the higher subgroups in both learners groups performed equally well. Future studies should take into account the difference of participants in a group and further divide them into subgroups, if necessary.

PedagogicalImplications

This study has pedagogical implications. It seems that tone acquisition is a long process, at least to some learners. The Tone 4 errors found in this study are very prevalent in learners’ speech. Instead of avoiding talking about this, the instructors should talk about these tone errors in class so that learners are aware of such tendency in their pitch production. Meanwhile, tone production practice is not merely related to beginners. Learners at the intermediate level, even at the advanced levels, should be trained on their tone production, without which their tone production cannot be further improved and may be fossilized. Considering the positional effect of tone errors, tone training should be conducted at the phrase and sentence level as well. That is to say, tone training should be done at different levels, at the mono-syllabic, di-syllabic, and phrase level, and equally important if not more, at the utterance level.

CONCLUSION

This paper discusses the acquisition of Mandarin lexical tones by English-speaking L2 learners. The analysis of the tone errors in the reading passage and the spontaneous speech showed that the surface Tone 4 error is the most frequent at the PW-initial position in both the intermediate and the advanced learner groups. Attempts were made to account for this tone error pattern within SLA frameworks, but in vain. It was argued that this tone error pattern may be due to the high pitch tendency at the phrase- or sentence-initial position across languages. Even though this is a general pitch pattern across language, the production of such a tendency by L2 learners superseded the underlying tones, hence the surface Tone 4 errors. The special status of Tone 4 in L2 speech corresponds to the special status of the same tone in L1 speech (Wan, 2007). However, it was argued that such similarity should not be taken as evidence for the similarity in L1 and L2 acquisition in that different processes are involved in L1 and L2 production. It was also found that the L2 tone errors are not contextual nor have directionality. Some theoretical and pedagogical implications were provided as well. Further studies should adopt the longitudinal design so that the developmental pattern of tone acquisition in L2 Mandarin can be identified, and linguistic universals, language-specific patterns, and individual differences can be distinguished from each other.

NOTES

1 The tone errors here involved one tone produced as another tone, which is self-evident. Coupled with the limits of pages, no specific examples of each type of tone error, except for some typical ones, are given in this paper.

2 It should be pointed out that this study focuses on the surfaced tone error patterns, instead of on the underlying tones. Therefore, the Tone 4 error in this study referred to the production of Tone 4 to replace other underlying tones in L2 speech, instead of Tone 4 being produced as other tones.

Beckman, M.E. (1996). The parsing of prosody.LanguageandCognitiveProcesses, 11(1/2), 17-67.

Bent, T. (2005).Perceptionandproductionofnon-nativeprosodiccategories. Unpublished doctoral dissertation, Northwestern University.

Best, C. T. (1995). A direct realist view of cross-language speech perception. In W. Strange (Ed.),Speechperceptionandlinguisticexperience:Issuesincross-languageresearch(pp. 171-204). Baltimore: MD: York Press.

Chao, Y.R. (1930). A system of tone letters.LaMaitrePhonetique30, 24-27. (Reprinted in English orthography in Fangyan 1980, 2, 81-83.)

Chen, M. (2000).Tonesandhi. Cambridge: Cambridge University Press.

Chen, Q. (1997). Toward a sequential approach for tonal error analysis.JournalofChineseLanguageTeachersAssociation, 32(1), 21-39.

Chen, Q. (2000).AnanalysisofmandarintonalerrorsinspontaneousspeechbyEnglish-speakingAmericanadultlearners:Astudyatandabovethewordlevel. Unpublished doctoral dissertation, Brigham Young University.

Chiang, T. (1979). Some interferences of English intonation with Chinese tones.IRAL, 17(3), 245-250.

Chomsky, N. (1980).Rulesandrepresentations. Oxford: Blackwell.

Chomsky, N. (1981). Principles and parameters in syntactic theory. In N. Hornstein & D. Lightfoot (Eds.),Explanationinlinguistics:Thelogicalproblemoflanguageacquisition(pp. 32-75). London: Longman.

Chomsky, N. (1999). On the nature, use and acquisition of language. In T. Bhatia & W. Richie (Eds.),Handbookofchildlanguageacquisition(pp. 33-54). San Diego: Academic Press.

Chu, M. & Qian, Y. (2001). Locating boundary for prosodic constituents in unrestricted Mandarin texts.ComputationalLinguisticsandChineseLanguageProcessing, 6(1), 61-82.Dai, J. X. -L. (1998). Syntactic, phonological, and morphological words in Chinese. In Jerome Ⅰ. Packard (Ed.),NewapproachestoChinesewordformation:Wordformation:morphology,phonologyandthelexiconinmodernandancientChinese(pp. 103-134). Berlin: Mouton de Gruyter.

Eckman, F. (1977). Markedness and the contrastive analysis hypothesis.LanguageLearning, 27, 315-330.

Feng, S. (1996). On the Chinese prosodic words.ChineseSocialSciences, 1, 161-176.Feng, S. (1997).OnChineseprosody,syntaxandlexicon. Beijing: Beijing University Press.

Feng, S. (2001). Prosodic words and the construction of scientific theory.ChineseTeachingintheWorld, 1, 53-63.

Flege, J. E. (1987). The production of ‘new’ and ‘similar’ phones in a foreign language: Evidence for the effect of equivalence classification.JournalofPhonetics, 15, 47-65.Flege, J. E. (1995). Second language speech learning: Theory, findings, and problems. In W. Strange (Ed.),Speechperceptionandlinguisticsexperience:Issuesincross-languageresearch(pp. 233-272). Baltimore, MD: York Press.

Flege, J. E., Schirru,C., & MacKay, I. R. A. (2003). Interaction between the natie and second language phonetic subsystems.SpeechCommunication, 40, 467-491.Huang, T. (2001). The interplay of perception and phonology in Tone 3 sandhi in Chinese Putonghua.OSUWorkingPapersinLinguistics, 55, 23-42.

Jeng, H. -H. (1985). A developmentalist view of child phonology.StudiesinLanguageandLiterature, 1, 1-25.

Ladd, D. R. (1996).Intonationalphonology. Cambridge: Cambridge University Press.

Ladd, D. R. (2008).Intonationalphonology. New York: Cambridge University Press.

Levelt, W. J. M. (1989).Speaking:Fromintentiontoarticulation. Cambridge: MIT Press.

Li, C. N. & Thompson, S. A. (1977). The acquisition of tone in Mandarin-speaking children.JournalofChineseLanguages, 4, 185-1999.

Maddieson, I. (1978). Universals of tone. In J. Greenberg (Ed.),Universalsofhumanlanguage(pp. 335-365). Stanford University.

Major, R. C. (1992). Losing English as a first language.TheModernLanguageJournal, 76, 190-208.Miracle, W. C. (1989). Tone production of American students of Chinese: A preliminary acoustic study.JournalofChineseLanguageTeachersAssociation, 24(3), 163-176.

Nespor, Marina & Vogel, I. (1986).Prosodicphonology. Foris Publications Holland: Netherlands.

Ohala, J. J. (1972). The physiology of tone. In L. M. Hyman (Ed.),Consonanttypesandtone(pp. 1-4). Southern California Occasional Papers in Linguistics 1.

Ohala, J. J. & Ewan, W. (1973). Speed of pitch change.JournaloftheAcousticalSocietyofAmerica, 53, 345.

Peng, S-H, Chan, M.K.M., Huang, T.T., Lee, O.J., & Beckman, M.E. (2005). Toward a Pan-Mandarin system for prosodic transcription. In S. Sun (Ed.),Prosodictypology:Thephonologyofintonationandphrasing(pp. 230-270). Oxford: Oxford University Press.Pierrehumbert, J. & Beckman, M.E. (1988).Japanesetonestructure. Cambridge, MA: The MIT Press.

Pierrehumbert, J. & Hirschberg, J. (1990). The meaning of intonational contours in the interpretation of discourse. In P. Cohen, J. Morgan, & M. Pollack (Eds.),Intentionsincommunication(pp. 271-311). Cambridge, MA: The MIT Press.

Selkirk, E. (1984).Phonologyandsyntax:Therelationbetweensoundandstructure. Cambridge, MA: The MIT Press.

Shen, X.S. (1989). Toward a register approach in teaching Mandarin tones.JournalofChineseLanguageTeachersAssociation, 24(3), 27-47.

Shen, X. S. & Lin, M. C. (1991). A perceptual study of Mandarin tones 2 and 3.LanguageandSpeech, 34, 145-156.Speer, S., Shih, C-L., & Slowiaczek, M. L. (1989). Prosodic structure in language understanding: Evidence form tone sandhi in Mandarin.LanguageandSpeech, 32(4), 337-354.

Stockwell, R. & Bowen, J. (1965).ThesoundsofEnglishandSpanish. Chicago: University of Chicago Press.

Strange, W. (1995). Cross-language studies of speech perception: A historic review. In W. Stragen (Ed.),Speechperceptionandlinguisticsexperience:Issuesincross-languageresearch(pp. 3-45). Baltimore, MD: York Press.Sun, S. (1998).ThedevelopmentofalexicaltonephonologyinAmericanadultlearnersofstandardMandarinChinese. University of Hawaii Press.

Sundberg, J. (1973). Data on maximum speech of pitch changes.QuarterlyProgressStatusReport, 4, 39-47.

Tseng, C. -Y., Pin, S-H, Lee, Y-L., Wang, H-M., & Chen, Y-C. (2005). Fluent speech prosody: Framework and modeling.SpeechCommunication, 46, 284-309.

Tseng, C. -Y. & Chou, F-C. (1999). A prosodic labeling system for Mandarin speech database. Proceedings of thethinternationalcongressofphoneticscience(pp. 2379-2382). San Francisco, USa.Wan, I-P. (2007). On the phonological organization of Mandarin tones.Lingua, 117, 1715-1738.Wang, J. (2002).Rhythmicgrouping,toneSandhiandstressinBeijingMandarin. Unpublished doctoral dissertation. Macquarie University, Sydney, Australia.

Wang, Y., Spence, M. M., Jongman, A., & Sereno, J. A. (1999). Training American listeners to perceive Mandarin tones.JournaloftheAcousticalSocietyofAmerica, 106, 3649-3658.

Wang, Y., Jongman, A., & Sereno, J., 2003. Acoustic and perceptual evaluation of Mandarin tone productions before and after training.JournaloftheAcousticalSocietyofAmerica, 113, 1033-1043.Wayland, R. & Guion, S. (2004). Training English and Chinese listeners to perceive Thai tones: A prelimnary Report.LanguageLearning, 54, 681-712.Wong, P.C. M. & Diehl, R. L. (2003). Perceptual normalization for inter- and intra-talker variation in Cantonese level tones.JournalofSpeech,LanguageandHearingResearch, 46(2), 413-421.Wong, P, Schwartz, R. G. & Jenkins, J. J. (2005). Perception and production of lexical tones by 3-year-old, Mandarin-speaking children.JournalofSpeech,LanguageandHearingResearch, 48, 1065-1079.White, C. (1981). Tonal pronunciation errors and interference from English intonation.JournalofChineseLanguageTeachersAssociation, 16(2), 27-56.

Winke, P. (2007). Turning into tones: The effect of L1 background on L2 Chinese learners’ tonal production.JournaloftheChineseLanguageTeachersAssociation, 42(3), 21-55.

Xu, Y. (1994). Production and perception of co-articulated tones.JournaloftheAcousticalSocietyofAmerica, 95, 2240-2253.Yang, S. (2011). A preliminary study on the perceptual center of tones in standard Chinese.ActaPsychologicaSinica, 3, 247-253.

Yang, Y. & Wang, B. (2002). The acoustic correlates of prosodic boundaries.Proceedingsofinternationalconferenceonprosody(pp. 707-710). France.Yip, M. (1980).ThetonalphonologyofChinese. Unpublished doctoral dissertation. MIT.

Yip, M. (2002).Tone. Cambridge: Cambridge University Press.

Zampini, M. L. (2008). L2 speech production research. In J. G. H. Edwards & M. Zampini (Eds.),PhonologyandSecondLanguageAcquisition(pp. 219-249). Amsterdam/Philadelphia: John Benjamins.

Zhang, H. (2007).AphonologicalstudyofsecondlanguageacquisitionofmandarinChinesetones. Unpublished M.A. Thesis. The University of North Carolina at Chapel Hill.Zhang, H. (2010). Phonological universals and tone acquisition.JournaloftheChineseLanguageTeachersAssociation, 45(1), 39-65.

Appendix:Readingpassage

星期天早上刚五点钟,王国明家里人都已经起来了。因为王国明要坐早上八点钟的火车到北京去,所以他家人都要到火车站去送他。 王先生帮着小王收拾行李,王太太特别给王国明做了很多吃的东西。 王太太想这两天天气热,火车上的东西恐怕不干净。他们把行李收拾好了的时候,已经七点钟了。王国明和家人坐出租车到火车站。他们到火车站的时候,别的同学都在那儿等着他呢 王国明把他的同学介绍给他家人。 过了一会儿,王国明和他的同学开始上车。王家人和王国明和他的同学说再见。

10.3969/j.issn.1674-8921.2014.12.007

Correspondence should be addressed to Chunsheng Yang, Department of Literature, Cultures and Languages, University of Connecticut, Oak 207, 365 Fairfield, Storrs, CT 06269, USA. Email: ycsgeorge@gmail.com