Chapter 4 Results and Findings
4.2 English Vowel Perception
4.2.3 General Perceptual Performance
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
types of errors are all highlighted, and all of the null types are left blank in Table 4-9.
4.2.3 General Perceptual Performance
The results of both experiment III and IV show that the participants in this study were gradually establishing a vowel grid that is similar to that of the American English native speakers, for in Experiment III, the elementary learners did poorly in picking up words with the sound mapping to the orthographic words. The correct rate of the participants in elementary levels in Experiment III was poorer than that of all of the other groups.
Also, from the results of the types of error in Experiment III, we can find that the elementary learners are still guessing about the sounds, for their choice of results is scattered among many different vowels and shows inconsistency in the ascription of a particular sound to particular word. In the ABX (Experiment IV), however, the elementary group performed better in comparison to in Experiment III. This may imply that the elementary levels of learners may be somewhat capable of distinguishing different phonemes, while they may not be able to precisely map each phoneme to a precise word.
The intermediate group, on the other hand, started to misperceive [e] as [ɛ] and [æ ], [ɛ] as [e] and [æ ] as [e] in a systematic way. This behavior may be a clue as to the nature of verb assimilation, as they showed a relatively high percentage of being confounded in discriminating between [æ ]-[ɛ] and [ɛ] -[e] in their decisions in the ABX discrimination task. As for the back vowels, however, the intermediate learners were quite similar in that they misperceived [o] as [a] more than the elementary group. In fact, they made more errors on diphthongs, but not on monophthongs, and the most obvious type of error was misperceiving [aɪ] as [ɛ]. Generally, this group of participants
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
did not yield significantly better results in the discrimination tasks at first glance than those of the previous group. Even when we took a close look at the types of error that the participants made, we couldn’t find any significant evidence of an improvement in the vowel perception of the intermediate level participants. Also, when we further investigated the variance of the type of error in the intermediate male groups, the value of the variance was even larger than that of the elementary male group.
We found that the values among the high-intermediate group of the types of error in the ABX discrimination task and of the error rates in Experiment III were way lower than those of the other two groups. However, [i] and [ɪ]; [ɛ] and [ɪ]; [e] and [ɛ]; [ʌ] and [u]; and [æ] and [ɛ] remained as five pairs which were frequently misperceived by the participants in the high intermediate group. What is noteworthy here is that, while to fail to distinguish [æ] and [ɛ] pair is not a relatively salient type of error, [e] & [ɛ] still remained as a salient group in this group of participants.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Chapter 5 Discussion
In this study, we hypothesize that Taiwan Mandarin EFL learners may experience a process of development in the learning of English vowels. The developmental process should encompass aspects including forms of perception and production where the L1 and the L2 are both involved and will interfere with each other. We constructed L1 production models and L2 perceptual and production models in the previous chapter, while question about the mobility of vowels among all of the groups, the asymmetry and symmetry between production and perception and the interferences among the different levels of language proficiency and the nature of the L1 are still pending to be solved.
We employed the Contrastive Analysis Hypothesis to first explain the possible output of the results and further adopted the SLM, the PAM and the U-Shaped Learning Theory together to help explain our findings. Most of our findings were consistent with those hypothesized in the above-mentioned learning models, while one inconsistency with the SLM and the PAM was that the instability of certain L1 vowels may indirectly affect the process of the learning of the L2.
An overall discussion of the results from Experiments I to IV will be provided in this chapter in accordance with the results we revealed in the previous chapter. The results from Experiment I showed that the vowels produced by the Taiwan EFL learners still follow most of our expectations, and those from Experiment II show the vowels produced by the Mandarin speakers. The results from both of the production tasks gave us a clear hint that the production of L2 vowels performed by the EFL learners may improve as they improve their proficiency and that their L1 may play a key role in
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
influencing that of their L2, while it would not be the case in the other way around. It is clear that the vowels spaces of the three different levels of learners are distinct from each other, and that there is a progression in the process of the assimilation and dissimilation among different L2 vowels which makes them distinguishable from those of their L1. The results from Experiments III and IV may give us hints about how L2 learners perceive English vowels and the process of the development of their capacity for vowel perception. Through analysis of the results from both of the experiments, we are able to give an explanation as to the reasons for the possible symmetry and asymmetry between learners’ perception and production.
We have also constructed a vowel inventory model for three different levels of Taiwan Mandarin EFL learners in Taiwan. We found a significant asymmetry between the production and perception of the participants and also that the production of the L1 may influence that of the L2 to some extent among the four experiments.
The following sections will discuss the development of vowel perception and production. In 5.1, we found the L2 learners’ production of English vowels. The production models for Mandarin vowels will be shown in 5.2. The perception of English vowels will be discussed in 5.3. The symmetry and asymmetry of the perception and production models will be discussed in 5.4. A discussion of the inter-lingual influences on Taiwan Mandarin EFL learners will be provided in 5.5.
5.1 Development of the Production of English Vowels (Experiment I)
The participants in this study encompass a wide range of levels from elementary to high intermediate Taiwan Mandarin speakers who were studying English as a foreign language and who had little or no linguistics background. From the results of
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Experiment I, we were able to identify that our high intermediate participants may be able to differentiate most of the vowels in monophthongs, and that our elementary level participants may be confounded when called upon to utter a variety of phonemes.
These findings are partially consistent with those of former studies. That is, [i]-[ɪ]
and [ɛ]-[e] are two pairs that bear very similar formant structures. However, the results are contrary to C.-Y. Lin (2013), which indicates that the pairs [i]-[ɪ] and [ɛ]-[æ] are two most difficult pairs to perceive and produce. We may see that for the high-intermediate groups, [ɛ]-[e] should be more difficult for the Taiwan Mandarin speakers rather than [ɛ]-[æ] because there is a distinct gap between the production of [ɛ]-[æ]
rather than between [ɛ]-[e]. More detailed issues will be later discussed later in the following sections.
The development of the space of [ɛ]-[e]- [æ] indicates that the learners may have already shown a tendency to both dissimilate and assimilate their intra-L2 (with-in) phonemes. The results where the participants in the high-intermediate levels were successful in distinguishing the three vowels from each other further strengthens our belief in the presence of developmental process by L2 learners.
The previous part of this section mainly discussed the performance of English vowels by Taiwan Mandarin EFL learners. We still need to make a comparison of the vowels produced by Taiwan Mandarin EFL leaners and native speakers. First, we will deal with the front vowels, and then we will tackle the back ones, and then the mid vowels.
The vowels [i] and [ɪ] made by the Taiwan Mandarin EFL learners differed mainly in respect to the F2 values. The American English native speakers, on the other hand,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
showed a significant difference between the F1 and the F2 values. The difference in the F1 values between these two vowels among the native speakers was greater than the difference of the values between [i] to [e] and [e] to [ɪ]. The gap was wide enough to the extent that [e] was to be believed to reside between them two vowels. In comparison, the values of [e] did not differ between the Taiwan group and the American group.
When we attend to the vowels [ɛ] and [æ], we find that the Taiwan learners seem to have struggled to make them into different categories, since their space in the vowels changed greatly among the different level. The American speakers, in contrast, made a clear distinctness among these three vowels. The vowel [e] was the highest one, then [ɛ], and then, finally, [æ]. The tongue advancement of [e]- [ɛ] and [æ] were the same, where [e] was at the most front and [æ] at the most rear in comparison.
The [u], [ɑ] and [o] produced by the American group and the Taiwan group were quite similar, while the position of [ʊ] and [ɔ] deviated from the Taiwan group to the American group. The [u] vowel made by both groups was a high back vowel, while the vowel [ʊ] was not the same case. The American group tended to make the vowel [ʊ] a high central vowel, but not a high back one which form was adopted by the Taiwan groups. This trend can be seen in the results from Y. Chen et al. (2001), while it is contradicted in many other studies, such as Hillenbrand et al. (1995), Peterson and Barney (1952) and among many others. We suspect that the differences in the results between the studies may reflect a possible vowel shift in American English since the data collection done by Y. Chen et al. (2001) and Hillenbrand et al. (1995), Peterson and Barney (1952) differs over a wide range of time points. The vowel [ɔ], on the contrary, was at a much lower point by the American group than by the Taiwan groups.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The vowel [ɔ] did not change much in comparison to the results given by Hillenbrand et al. (1995), while it has been mentioned that [ɔ] was not a vowel that was pronounced steadily at that time. We may also suggest that the instruction in the vowel in Taiwan may produce an output that is different from that of native speakers, and the possible transfer may result from such reason.
The mid vowels [ɝ] and [ʌ] yielded two different comparisons from the American group to the Taiwan groups. The Taiwan groups made a relatively similar [ɝ] to that of the American group. However, the American group tended to make the vowel [ʌ] mid and central, but the Taiwan group tended to put it too far to the back for it to be a high back vowel.
5.2 Development of Production of Mandarin Vowels (Experiment II)
We did not find much relevant evidence to show vowels shifts from group to group among the male participants. The female groups, in contrast, showed inconsistency in the vowel grids among the three learner groups. The back vowels, in this study, appeared to be comparably more unstable than the front vowels, which is in contradiction to the previous literature (Locke, 1980).
In the case of the female groups, we may suggest that it could be consistent to finding in most studies that the female EFL learners may give up their own accent easily and tend to adopt a relatively dominant language that they go through the learning stages more easily (Norton, 1997; Pavlenko & Norton, 2007; Polat, 2011). Their performance of their L1 may, therefore, be more easily influenced by their L2. Some strong evidence for this may be that the elementary female learners’ vowel grid deviates from the other two levels distinctly. It may also explain the reason why we did not find
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
a great shift on in the case of their [o] sound. We may only find that the intermediate group has once mapped them to a former position, but soon mapped them back to the back position, and it could be explained by the fact that they were having problems in differentiating the new sound in the L2.
5.3 Development of Perception of English Vowels (Experiment III & IV)
This section will discuss the two perception tasks, Experiments III and IV. Since the two perception tasks were both designed to examine the learners’ perceptual development, we decided not to separate the report of the two sets of findings, but to make a general discussion.
W We may conclude that there is a process of development in the perception of vowels by the Taiwan Mandarin EFL learners in this study which is evidenced by the results of Experiments III and IV. The elementary EFL learners may start in a status of a chaos in perceiving different phonemes, since their error types were relatively inconsistent and unsystematic. We may assume that they should be capable of differentiating most of them, but, also, that they fail to match them into pairs. After a series of practices, they gradually learn to differentiate each sound and map them into different slots. Thus, it is very likely that the learners indeed linked their L2 vowels to their L1 category, and, according to the PAM, from an uncategorizable speech sound to a native sound (Best, 1995).
Though most of our results were congruent to PAM, the performance of the participants at the intermediate levels did not meet our hypothesis. Their performance was poorer (by means of error types) than that of the elementary group. Despite of the possible anomaly contingent, the U-Shape Learning Theory may give us a way to
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
explain why the intermediate level participants show a relatively poor perceptual performance. We suggest that there are several stages for learners to go through to assimilate their L2 vowels to L1 vowels and to dissimilate (discern) the certain features that differentiate the L2 vowels and L1 vowels. In this case, they may still be under a process of dissimilating the different vowels, and the clue to this phenomenon may be the finding that intermediate levels systematically misperceive [e] as [ɛ] and [æ], [ɛ] as [e] and [æ] as [e]. Their conception of the distinctness and features of each vowel may be linked in an in correct pattern, which may probably be explained by the fifth hypothesis (H5) in the SLM raised by Flege (1995). If we see the burst of mistakes as evidence for the linking of different sounds to different categories, the emergent stage should be the situation where the learners were experiencing (Brown, 2007). According to the U-shape learning theory, all of the phenomenon discussed above may imply the finding of the low performance and low curve in the U-shape learning process.
The results also showed that the word ‘hudd’ ([ʌ]) was frequently perceived wrongly.
In Taiwan group, [ʌ] is a sound that the Taiwan Mandarin EFL learners do not frequently perform well, and this could be cross referent to their performance in the production of the [ʌ] sound in Experiment I. If we take this phenomenon into account for the first hypothesis (H1) in the SLM (Flege, 1995), we may find that learners should assimilate [ʌ] into a Mandarin vowel such as [i], [ɪ], [æ], [u], [o] or [ʊ] at first. As the learners’ proficiency grows, they may be better in discerning [ʌ]. We could find that the errors become more concentrated. The only consistent mistake made by the high-intermediate group was in perceiving [ʌ] as [ɑ]. This could be explained by the fact that the learners had been through the stages from making a sound as an allophone of one
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
in the L1, and then succeeded in discerning the phoneme into a new category, but may have failed to do so sometimes, for they may have faced a fossilization effect, which blocks the mechanism of establishing a stable enough phonemic feature in the L2.
For high-intermediate learners, they could both differentiate different sounds and put them into pairs and also identify the correct stimuli when they were mixed with others.
However, in Experiment IV, we still see that the participant made some mistakes on some of the stimuli. Since these mistakes were not quite systematic, we may suggest that such type of mistake might have been an inadvertent one due to the rapidity of the speed of the recording or that one might have been distracted by something the natural context of the test environment.
In a general review, we also found that the vowel [ɪ] was misperceived among all levels, and that the vowel sounds that were presented in this case were like lax vowels.
The duration of the vowels may be the clue to the existence of this phenomenon. As discussed in Hillenbrand et al. (1995), Hung (2012) and many others the duration may be an essential key to influencing listeners’ perception, and, especially in the case of L2 learners, once they rely upon a probable variant, they may build up their phoneme categories based on it. Therefore, we suggest that the Taiwan Mandarin EFL learners may experience a shift from relying mainly on vowel duration to somewhat relying on the formant features.
The consistent errors of perceiving [o] as [ɑ] may be understood as a dialectal problem, since some native speakers, also, made errors in producing them. Though we tried to neutralize the differences among the American English accents of our American English native speaker group, the wide variety of vowels may still be a hard nut to crack,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
and this issue remains open for further research.
5.4 Symmetry and Asymmetry between Learners’ Perception and Production Model
In this section, we will first review the errors that the participants made in their perception tasks, and cross-referent such errors to their production of the relevant vowels. In so doing, we may find if their perception and production of a particular vowel is symmetrical or not. Next, we will go back and scrutinize their production patterns and see if they made similar errors on their perception task, so as to examine if there is any asymmetry in their vowel perception and production.
5.4.1 Symmetry between Vowel Perception and Production
As discussed in 5-3, we may realize that the elementary learners were confused by the front vowels [ɪ], [ɛ], [e] and [æ] and the back vowels [ɑ], [ɔ] and [o]. Also, [ɪ] and [e] were frequently misperceived as [ɛ], while [e] may also have been perceived as [ɛ]
and [æ]. Thereafter, we look further into the [ɪ], [e], [ɛ] and [æ] sounds of the Taiwan Mandarin EFL learners.
In rescreening the performance of the Taiwan Mandarin EFL learners in the production experiment of [ɪ], we found that the participants may be able to differentiate [ɪ] owing to the fact that their [i] and [ɪ] were pronounced in different positions.
Nevertheless, the [ɪ] as uttered by all of the levels of the participants were all very different from that produced by the American English native speakers. Especially, when we compare the L2 results and the L1 results (as shown in Figure 4-5 and Figure 4-6), we can better understand that the place of [ɪ] as pronounced by native speakers is actually quite close to that of [e]. Due to the fact that the pronunciation in the recording
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
that we adopted in this study was American English, it is no surprise that they made an error on distinguishing it from [e] or [ɛ]. Thus, we may reveal the symmetry between the perception and the production of the vowel by certain of the participants.
One may question why, if the participants tended to place the vowel [ɪ] incorrectly in the wrong position as [e], why, then, would the participants choose [ɛ] as an answer.
This may be explained by reference to the previous studies where it was found that L2 learners rely on vowel duration to a significant degree, and this phenomenon could also be evident in the results where [e] is misperceived as [ɛ]. When pronouncing [e], the participants may also think of the sound of [ɛ], while they may dissimilate these two vowels by their duration. Unfortunately, we did not measure vowel duration in this study, and such task remained for future studies. The same problem exists in the case of [ɛ], [e] and [æ]. These all suggest the symmetry of one’s vowel perception and
This may be explained by reference to the previous studies where it was found that L2 learners rely on vowel duration to a significant degree, and this phenomenon could also be evident in the results where [e] is misperceived as [ɛ]. When pronouncing [e], the participants may also think of the sound of [ɛ], while they may dissimilate these two vowels by their duration. Unfortunately, we did not measure vowel duration in this study, and such task remained for future studies. The same problem exists in the case of [ɛ], [e] and [æ]. These all suggest the symmetry of one’s vowel perception and