Citation
The Effect of Implicit and Explicit Feedback

Material Information

Title:
The Effect of Implicit and Explicit Feedback A Study on the Acquisition of Mandarin Classifiers by Chinese Heritage and Non-Heritage Language Learners
Creator:
Han, Ye
Place of Publication:
[Gainesville, Fla.]
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (239 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Linguistics
Committee Chair:
Antes, Theresa A.
Committee Members:
Henderson, Brent Mykel
Boxer, Diana
De Jong, Ester J.
Graduation Date:
8/7/2010

Subjects

Subjects / Keywords:
Classrooms ( jstor )
Cultural preservation ( jstor )
English language learners ( jstor )
Foreign language learning ( jstor )
Language ( jstor )
Learning ( jstor )
Linguistics ( jstor )
Mandarins ( jstor )
Nonnative languages ( jstor )
Perceptual learning ( jstor )
Linguistics -- Dissertations, Academic -- UF
chinese, explicit, feedback, heritage, implicit, meta, recast
Genre:
Electronic Thesis or Dissertation
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
Linguistics thesis, Ph.D.

Notes

Abstract:
THE EFFECT OF IMPLICIT AND EXPLICIT FEEDBACK: A STUDY ON THE ACQUISITION OF MANDARIN CLASSIFIERS BY CHINESE HERITAGE AND NON-HERITAGE LANGUAGE LEARNERS Previous studies revealed mixed results in terms of the relative effects of implicit and explicit feedback: some found that explicit feedback worked more efficiently than implicit feedback; others found no difference between the two feedback types. These contrasting results called for further investigations into this issue, particularly examining those underexplored learner groups who have developed their target language (TL) in a different path from traditional foreign/ second language learners investigated in previous studies, such as heritage language (HL) learners. Therefore the present study aims at contributing to the literature by investigating the relative effects of implicit feedback in the form of recasts and explicit feedback in the form of meta-linguistic feedback on the acquisition of Mandarin classifiers by Chinese heritage language (CHL) and non heritage language (non-CHL) learners. This study employed a pre-test-treatment-post-test research design, in which 64 participants (CHL=35, non-CHL=29) were randomly divided into four experimental groups and two control groups. Feedback was provided on learners? errors by a native Chinese interlocutor during the two treatment sessions. Learners? perceptions of feedback were measured by their verbal comments provided during the stimulated recall, and their acquisition of classifiers were assessed by their test scores. The data were both quantitatively and qualitatively analyzed. The results showed that both feedback types were effective in facilitating learners? acquisition. In addition, learners? language background was also found to affect their perceptions of feedback, as well as their acquisition. The findings of this study expand our knowledge about implicit and explicit feedback. In addition, they also provide invaluable information particularly for educators and administrators who are involved in HL instruction. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2010.
Local:
Adviser: Antes, Theresa A.
Statement of Responsibility:
by Ye Han.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Embargo Date:
10/9/2010
Resource Identifier:
004979720 ( ALEPH )
705931024 ( OCLC )
Classification:
LD1780 2010 ( lcc )

Downloads

This item has the following downloads:


Full Text






Day 1














Day 2









Day 3


CHL learner


Interview (15-20 min)


Non-CHL learner


Exit


Figure 4-2. Experimental procedure and approximate time

The 17 Learners in the control group took the pre-test on Day 1. They also first

took the oral test, then the written test. On day 2, they completed the oral and written

post-tests in the same order as they did in pre-test.


138


Consent form, background questionnaire (5 min)




Pre-tests: Oral test (5-15 min)
Written test (5-15 min)





Treatment 1 (10-20 min)


Treatment 2 (10-20 min)




Post-tests: Oral test (5-15 min)
Written test (5-15 min)




Simulated Recall (30-40 min)









determining the variability of both HL and non-HL learners. HL learners showed similar

processing strategies as L2 learners when their dominant language was the same as

the L2 learners' L1 (Kim, 2006). Thus it would be interesting to compare the linguistic

behavior of members in these two learner groups.

Learners at high-intermediate and advanced levels were chosen due to the

following concerns: first of all, as discussed in Chapter 2, HL learners usually tend to

have strong speaking and listening skills even with very little formal classroom

instruction. In order to ensure the communicative activities adopted in the treatment

appropriate for learners in both groups-not too easy for CHL learners on the one hand,

but also not too challenging for non-CHL learners on the other hand-non-CHL learners

need to reach a proficiency level that allows them to carry on communication with the

NS. Secondly, since the majority of the target classifiers had already been introduced in

the textbook at the beginning and intermediate level, all the non-CHL learners should

have already acquired, or at least encountered all or the majority of the target classifiers

from their textbook through formal classroom learning. For CHL learners, although they

may not have acquired all the classifiers in the classroom as some of them skipped the

beginning level, they were more likely to have heard all of these classifiers that are

frequently used in daily life. Third, for HL learners, after at least one year of formal

instruction at the university level, their explicit knowledge of Mandarin, which started to

grow at earlier years of their life through language exposure at home, was re-acquired.

Thus, the difference in the performance between them and the non-HL learners was

more likely due to their HL experience, which the current study was mostly eager to

explore.


113









APPENDIX D
WRITTEN PRE-TEST

Please complete the following sentences. You can write either in Chinese characters or
in Pin Yin.
1. 1i. ___________o
I have three cats and two horses.
2. R ___ ___
I ate a chicken for dinner.
3. m)______o
My little sister drew a dragon.
4. ___n, ____o
My little brother's pet is a snake.
5. 17 T_______ o____Vo
There are a pair of chopsticks and two pairs of scissors on the table.
6. 11]XWJ____ MR7____I o
Tomorrow is my birthday, my mom bought a pair of shoes for me.
7. Fi_____Jii',i,, XT ______,-ETo
I lost one of my gloves yesterday when I went to watch a movie.
8. 4A'AXI____, W____to
He is really weird, he only wears one shoe on his foot.
9. an I I___(. n____ o
There is one toothbrush and one pen in her bag.
10. x4_____7 ____,, ____I RR

Yesterday I went to a shopping center and bought a lot of stuff: two shirts, one sweater,
and one belt.
11. W]__________ _______*_n
There is a piano in Xiao Gao's living room.
12. RT-- ___T __ o
I spent $15000 on buying a car.
13. f_____, '.":,, $

The day after tomorrow is my dad's birthday, I want to give him a tie and a hat.
14. %lt+_____ tp^____ _n

I didn't get up until 10 o'clock. I only had one slice of bread and three grapes.
15. 'ttA [ ;X, %^ TO______ T L o
On Valentine's day, I gave by girlfriend eighteen red roses.
16. l ____ -_ 'l
There are eleven pearls in my bracelet.
17. j, n o


222










Group
OMHL
SRHL












Figure 5-1. Feedback episodes provided for CHL group.

Table 5-2. Feedback episodes provided for non-CHL learner group
Group N M SD Percentage
RNHL 427 35.58 12.81 55.89
MNHL 337 28.08 5.78 44.11
Note. N= raw frequency; M = mean frequency; SD = standard deviation (frequency).
MNHL= non-CHL meta-linguistic group; RNHL= non-CHL recast group.
Group
1IMNHL
MRNHL











Figure 5-2. Feedback episodes provided for non-CHL group.

Of the 428 feedback episodes that were provided for CHL learners, 93.93%

(n=402) of them were provided with introspective comments by learners. Of the 224

recast episodes, 96.88% (n=217) were commented on by learners. Similarly, of the 204

meta-linguistic feedback episodes, 90.69% (n=185) were commented on by learners.

Table 5-3 summarizes these results. All these introspective comment data were

collected during the stimulated recall.


149









CHAPTER 6
DISCUSSION AND CONCLUSION

In this chapter, I will discuss the findings of the study with regards to the research

questions, followed by a consideration of the theoretical and pedagogical implications.

Then I will address some limitations of the study, and suggest areas for future research.

Finally, I will conclude the chapter by addressing how the current study extends our

knowledge about the effects of implicit and explicit feedback.

6.1 Research Question 1

Research question 1 asked whether CHL learners perceive both implicit and

explicit feedback on Chinese classifiers more accurately than non-CHL learners.

In this study of learners' perceptions about target classifiers in conversational

interaction, learners with CHL background did not show any statistically significant

advantage over those without CHL background in the same feedback condition. The

only significant difference between the two learner groups was seen between the CHL

learners in the meta-linguistic feedback group and the non-CHL learners in the recast

group. On the other hand, learners in the meta-linguistic feedback group all significantly

outperformed those in the recast group, regardless of their language background.

Comparing the findings of current study with those of previous research (e.g.,

Lyster & Ranta, 1997; Lyster 1998a, 1998b; Panova & Lyster, 2002), it is noticeable that

the percentage of learners' accurate perception of feedback was higher, particularly in

the case of recasts (57%~88%). The discrepancy of results between the current study

and previous studies could be due to the following factors: the learning context; the

learners' orientation, and their learning experience.


169









Three. One is red, Also have (one) black.
Recall data
Again when I was doing that, I know I didn't know the measure words. It
makes, kind of, my personality, I like to know what I talk about. I knew I
just kept saying "ge" because I didn't know what the actual measure word
was.

A2: Acquisition 2 was operationalized as comments made by learners to indicate

that they thought a wrong measure word as the correct one, as shown in example 4-12.

Example 4-12
Feedback episode
NS: a,9iMJ/?
Ni kan dao duoshao she?
2sg see RC how many snake
How many snakes did you see?
NNS: *- zhang ,Eo
Yi zhang de she.
One CL NOM snake
One snake.
NS: FI,-- z ?
Ni kan dao yi tiao she?
2sg see RC one CL snake
You saw one snake?
NNS: --
YI tiao.
One CL.
One (snake).
Recall data
The measure word again. I have learned snake before. Second year, but I,
it was the first semester of advanced, and it was really difficult, I kind of
remember, it was a story, like a fairy tale, but we never learned it. And I
learned "tiao", but never for snake. I know it is for long object, but didn't
know that was for long object, but we always used it for "he", or I mean
really long and thin object. You never put it with animals. I thought it is for
inanimate objects.


O: The learner didn't use any measure word in the original sentence as shown in

example 4-13.


142









scored lower than his CHL peers, and could not activate what he acquired by age ten

within a two-day treatment, in spite of his relatively older age of arrival and much longer

formal schooling in the home country compared with other CHL learners in the current

study,

In summary, the CHL learners developed a rich implicit knowledge of classifiers

due to their early exposure to the language, which may not always be visible in their oral

production. However, with some assistance (such as reoccurrence in the pre- and post-

test), some of this knowledge could be activated, as we saw among the CHL learners in

the control group. Being facilitated by feedback, the CHL learners made even greater

increases on the oral post-test. However, since their initial levels of implicit knowledge

have already approximated ceiling levels, although meta-linguistic feedback still

outperformed recasts, its superiority was not significant. Those who have scored the

maximum even showed slight regression on the post-test.

Table 6-3. The profile of learners TW and SBY
TW SBY
Generation First First
Age of arrival Ten Ten
Schooling in home country Four years Four years
Cultural identity Less stronger Stronger
Step father: native English Father: Mandarin/English
Home language speaker
Home language speaker Mother: Mandarin/ English
Mother: English/ Mandarin
Home literacy environment Not so many Many

Written cloze test. In contrast to findings from the oral test, the results of the

written tests showed that feedback type significantly affected learners' increase in their

performance from the pre-test to the post-test. That is, meta-linguistic feedback led to a

greater increase than recasts. However, this superiority was found only among the non-

CHL learners, but not among the CHL learners. That is, the scores of the non-CHL


202









classifier, which they used on the pre-test and in the treatments. This resulted in their

decreased score on the post-test. Taking learner LL as an example, she used the

accurate classifier shuang for chopsticks on her pre-test, as well as during her two

treatments. However, she replaced it with an incorrect classifier tiao on her post-test. A

similar case also occurred with learner SBY, who used the target classifier pi for horse

on his pre-test and the two treatments, but replaced it with the incorrect classifier tiao on

the post-test. It might not be a coincidence that both shuang and pi were the classifiers

appeared in the original recorded sentences on the post-test, in which learners were

expected to modify the original sentence if necessary. A plausible explanation for this

type of "regression" is that, as discussed earlier in Chapter 6, CHL learners rely heavily

on their past listening experience when dealing with classifiers, particularly when they

have rich exposure to the HL at home. However, these experiences are not always

accurate. Sometimes learners assume an incorrect classifier as an alternative one (see

Example 6-6). Therefore for LL and SBY, it is possible that in the past they had heard

both the target classifier and the one used in the original recorded sentences on the

post-test, and assumed that they were interchangeable.

The above explanation is further evidenced by the fact that the same type of

"regression" was not seen among the two non-CHL learners, as well as learner ZZD,

who scored the lowest in the CHL learner group. The two non-CHL learners' "lack of

progress" was mostly caused by their reduced usage of the general classifier ge on their

post-test. Although this resulted in the decrease of their score, it showed the

improvement in their awareness of specific classifiers. On the other hand, the CHL

learner ZZD's "no progress" on the post-test was mostly due to his incorrect choice of


200









The former was used to measure learners' implicit knowledge, while the latter was

employed to assess their explicit knowledge.


Day 1


CHL learner


Interview


Non-CHL learner


Exit


Figure 4-1. The research design.

The experiment was conducted on three consecutive days. The short testing

period was chosen due to the characteristic of the targeted linguistic forms: the

association between a classifier and available head nouns is arbitrary and conventional,


108


Questionnaire




Pre-test




Treatment 1




Treatment 2




Post-test




Simulated Recall


Day 2








Day 3









tasks generate more negotiations of meaning among learners (Doughty & Pica, 1986;

Gass & Varonis, 1985; Long, 1988; Pica, 1994), but others have questioned this

argument (Duff, 1986; Nakahama et al., 2001), In order to minimize the task effects, the

current study employed both a one-way task and a two-way task.

Lastly, the context in which feedback was provided was designed to enable recasts

to serve multiple discourse functions, in order to make it as implicit as possible.

Previous research showed that recasts could be more or less implicit, depending on

various recast features (e.g., length, number of changes, prosodic emphasis, etc.) and

the discourse context (Loewen & Philp, 2006). For instance, in the classroom, learners

are sometimes not sure whether a recast provided by the teacher was a corrective

feedback or an approval of their response, since teachers often use both. In the tasks

designed for the current study, although recasts were only provided on errors made by

learners, they were done in as natural a way as possible. The learners could therefore

interpret them simply as a discourse move made by the researcher to carry on the

interaction. I will be looking at precisely this question, specifically seeking to determine

whether CHL and non-CHL learners perceived this feedback differently.

With the above careful considerations, the current study employed two task-based

NS-NNS conversational interactions in the treatment sessions. In each interaction

session, learners completed one communicative task. On Day 1, they engaged in a

one-way interaction task ("a story telling"); on Day 2, they completed a two-way task

("spot the differences"). During the interaction, participants received feedback on not

only the problematic classifiers, but also any errors that they made. Both tasks used

Power Point slides to elicit the target form. On each slide, several pictures were


123









the classroom. Thus his knowledge of Chinese was almost entirely developed in the

classroom. Unlike YHX, he did not have rich past experiences of this language that he

could activate. Therefore feedback played an important role in helping him to figure out

his untargetlike language.

Lastly, it would be interesting to look closely into those who did not make any

increase or even regressed from the pre-test to the post-test. A post hoc analysis of the

pre- and post-test scores of these learners is presented in table 6-2.

Table 6-2. Learners who increased zero or decreased in the experimental groups.
Learner Group Test type Pre-test Post-test Increase
TW RHL Oral 46 42 -4
SN RHL Oral 90 88 -2
SBY RHL Oral 92 88 -4
LL MHL Oral 90 88 -2
ZZD MHL Oral 23 23 0
LJD RNHL Oral 13 13 0
LAW RNHL Oral 25 21 -4

The results in table 6-2 reveal the following interesting facts: the majority of these

learners (5/7) are CHL learners. Three of them (SN, SBY, and LL) were those who

achieved the maximum scores on the pre-test, whereas the other two (TW and ZZD)

were those who scored the lowest in the CHL learner group. In contrast, among the two

non-CHL learners in the recast group, LJD scored relatively lower than the mean score,

but was not among those who scored the lowest. The lowest score in the group was

only six out of a possible one hundred. The final learner, LAW, scored slightly higher

than the mean score (M: 19.63).

For the three CHL learners who have already achieved maximum scores on the

pre-test, it would not have been not possible for them to further increase their score on

the post-test. However, it is surprising that they sometimes "forgot" the accurate


199









takes place in a three-dimensional framework with intersecting planes of time, space,

and identity, a CHL learner will form a new cultural identity which inherits "some of the

'Chineseness' from his family and his neighborhood but will enable him to become a

very different kind of Chinese-American from his family and his neighbors". (He, 2008b,

p. 110). In addition, she notes that, "the degree to which a learner's CHL develop is

dependent upon the degree to which s/he is able to find continuity and coherence in

multiple communicative and social worlds in time and space and to develop hybrid,

situated identities, and stances" (p. 116). In her identity-based model for CHL

development, she hypothesized that the degree of success in CHL development

positively correlates learners' desire to be connected with their heritage culture and CHL

community members for the long run, the desire to communicate in CHL in a moment-

by-moment manner, the home language of the learner, the availability and diversity of

the CHL input, as well as the extent to which the learner has created a niche in the

English-speaking linguistic, social, and cultural community.

Motivation. Chinese was ranked as one of the most difficult languages for native

English speakers to learn (Hadley, 2001). Therefore regardless of CHL or non-CHL

learners, a strong motivation must be present, since it is one of the key factors

influencing the rate and success of second/foreign language (L2) learning (Dornyei,

1994; Ely, 1986; Gardner, 1985). Gardner defined motivation as "the extent to which the

individual works or strives to learn the language because of a desire to do so and the

satisfaction experienced in this activity" (Gardner, 1985, p. 10). Gardner's well-known

socio-educational model distinguished two types of motivations: instrumental and

integrative. The former one refers to a pragmatic reason that drives learners to learn a










the recast group (M=58.87) and those in the meta-linguistic group (M=57.22) was very

close. They scored higher than learners in the control group (M=41.66). On the other

hand, for non-CHL learners, learners in the recast group (M=19.63) scored slightly

higher than those in the meta-linguistic group (M=15.8). The mean scores of learners in

the control group (M=16.44) were between the scores of learners in the two

experimental groups.

Table 5-10. Summary of oral pre-test scores for CHL learner group
Group N Minimum Maximum M SD
RHL 12 38 92 58.87 18.28
MHL 11 23 90 57.22 19.15
CHL 8 10 85 41.66 27.75


Table 5-11.


Summary of oral pre-test scores for non-CHL learner group


Group N Minimum Maximum M SD
NRHL 12 6 56 19.63 13.32
NMHL 11 4 48 15.80 13.13
NCHL 9 10 21 16.44 3.67


60-

50-

40-

U) 30-

20-

10-


Group

Figure 5-4. Mean scores of oral pre-test

Tables 5-12 and 5-13 below summarize the descriptive statistics by CHL learner

groups and non-CHL learner groups during the post-test. The means scores of learners

in all the groups are also graphically presented in figure 5-5. For CHL learners, those in


155


v









required to indicate whether each sentence was grammatically correct or not, report

their degree of certainty, and indicate whether they applied any rules in their judging

process. The meta-linguistic knowledge test was formed by five sentences. During the

test, learners were told that these sentences were ungrammatical, and were requested

to correct the error, and provide the reason for the errors. The results showed significant

group differences in the oral imitation test, but not in the untimed grammaticality

judgment test in terms of ungrammatical sentences. No significant group differences

were found in both tests in terms of the grammatical sentences.

Loewen and Nabei (2007) used three different testing instruments in their

experiment that investigated the effects of feedback on English question formation: a

timed grammaticality judgment test that limited the completion time of each item to 1.8

to 5 seconds, an un-timed grammaticality judgment test, and an oral production task.

The untimed grammaticality judgment test was designed to measure learners' L2

explicit knowledge, while the oral production task assessed their L2 implicit knowledge.

The same set of 40 items was used in both the timed and untimed grammaticality

judgment tests. Thus any differences in learners' performances could be interpreted as

differences in their two types of knowledge: when the time was controlled, learners were

pushed to use their "feel," and there was "little need or opportunity to access meta-

linguistic knowledge" (R. Ellis, 2005, p.157). On the other hand, the untimed task

allowed learners to involve their meta-linguistic knowledge. The oral production task

consisted of two-spot-the-differences tasks. The results showed an increase in the

timed grammaticality judgment test, but not in the untimed grammaticality judgment test

on the oral production test.


126










introspective comments made by learners in the recast group were classifier-related,

whereas the percentage made by learners in the meta-linguistic group was 89.19%. On

the other hand, for the non-CHL learners, 57.26% of introspective comments made by

learners in the recast group were classifier-related, whereas the percentage made by

learners in the meta-linguistic group was 81.99%. Figure 5-3 graphically compares the

percentage of classifier-related comments provided by learners across the four

experimental groups.

Table 5-5. Classifier-related comments provided by CHL learner group
Group N Minimum Maximum M SD Percentage
RHL 155 24 91 69.35 22.5 71.43
MHL 165 71 100 91.31 9.58 89.19


Classifier-related comments provided


by non-CHL


learner group


Group N Minimum Maximum M SD Percentage
RNHL 217 29 93 59.61 20.26 57.26
MNHL 264 62 94 81.91 11.29 81.99


100-

80-

S60-
E
0
40-

20-


0. I I I
RHL MHL RNHL MNHL
Group

Figure 5-3. Percentage of classifier-related comments across experimental groups

In summary, the descriptive statistics showed that learners in the meta-linguistic

feedback group made more classifier-related comments than those in the recast group ,


151


Table 5-6.









CHAPTER 1
INTRODUCTION

The present study examines the effects of implicit and explicit feedback on the

acquisition of Mandarin classifiers by learners from heritage language (HL) and non-

heritage language (non-HL) backgrounds. The study is rooted within an interactionist

framework, drawing from previous research in second language acquisition (SLA) on

the topic. Implicit feedback is operationalized in the form of recasts, which are defined

as a native speakers' (NS) reformulation of learners' problematic utterances; explicit

feedback takes the form of meta-linguistic feedback, which refers to a provision of meta-

linguistic information to indicate an error made by the learner through highlighting the

nature and the characteristic of the target language (TL) form.

Previous research on the effects of implicit and explicit feedback has revealed

controversial results: some researchers found that implicit and explicit feedback worked

equally effectively (e.g., Carroll, 2001; DeKeyser, 1993; Kang, 2009; Kim & Mathes,

2001; Leeman, 2003; Loewen & Nabei, 2007), while others reported the superiority of

explicit feedback over implicit feedback (e.g., Ammar & Spada, 2006; Lyster, 2004;

Muranoi, 2000; R. Ellis et al., 2006). In previous studies, learners usually studied the TL

as a foreign language (FL) or a second language (SL) through formal classroom

instruction; therefore they had developed their explicit knowledge and literacy skills, but

were weak in their implicit knowledge. However, little empirical research has been done

to investigate learners who have developed implicit knowledge but lack explicit

knowledge and literacy skills due to their exposure during childhood, such as HL

learners. In addition, in the United States, HL instruction and research has been

attracting more and more interest from researchers and language educators, along with









addition, the relationship between the two types of knowledge has also been much

debated. These controversies resulted in very different pedagogical recommendations.

There are three major positions: the non-interface position, the weak-interface position,

and the strong interface position.

Krashen' (1981) non-interface model completely denies any associations between

implicit and explicit knowledge. He believes that these two types of knowledge have

their independent mental processes and storage. Explicit knowledge can never be

converted into implicit knowledge, and vice-versa. He claims that implicit and explicit

knowledge are caused by different acquisitional mechanisms. Implicit knowledge is the

result of acquisition, whereas explicit knowledge is caused by learning. Learned

competence does not become acquired competence. The results of learning can never

lead to implicit knowledge. Therefore the role of L2 instruction should really be to

provide large quantities of comprehensible input to foster implicit learning, not to provide

explicit rules and systematic practice of these rules. The acquired implicit knowledge is

used by learners in producing utterances; whereas the learned explicit knowledge only

functions as a monitor to edit the utterances produced by acquired knowledge, which is

known as the Monitor Hypothesis (Krashen, 1981; 1985). Krashen's position was

criticized for not operationable as it failed to provide precise definitions of the two critical

constructs in this model: acquisition and learning (e.g., DeKeyser, 1997; Gregg, 1984).

In contrast, DeKeyser (1998) in the strong interface position claims that explicit

knowledge can not only be derived from implicit knowledge, but also be converted into

implicit knowledge through intentional practice. He claims that L2 knowledge is learned

by learners as declarative knowledge first, and then changed into procedural knowledge









I bought four pens.
NS: BAM f j4t ff ?
Na si zhi bi shi shenme
That four CL pen is what
What color are those four pens?


NNS: --^ AF--,1
Yi zhi shi
One CL is
One is green,
-A-11 go


Ivse
green


yanse de
color NOM


de
NOM


Yi zhi shi hongse de
One CL is red NOM
One is red.
Recall data
This is like zhi bi, but I was like ge bi.
what it was.


I didn't, or I forgot, I didn't remember


Example 6-17 Learner XS in treatment 2
NNS: *- --4"o


Wo you san ge
Isg have three CL
I have three brushes.
NS: _-- o


Wo ye you san zhi
Isg also have three CL
I also have three brushes.


maobi
brush


maobi
brush


Ni you duoshao gangbi
2sg have how many pen
How many pens do you have ?
NNS: iS~W o
Liang zhi gangbi
Two CL pen
Two pens.
Recall data
I didn't know what gangbi was. And then the measurement word for maobi
and gangbi, I didn't know.

Zhi is a frequently used classifier for long-rigid objects, which was found to emerge

in L1 Mandarin Children's vocabulary as early as age three (Hu, 1993). Therefore there


188









computer-delivered acceptability judgment test to measure the effects of the feedback.

The results showed no significant differences between two feedback types for both

immediate and delayed post-tests.

In sum, the 18 studies detailed revealed mixed results in terms of the effects of

implicit and explicit feedback on learners' L2 development. Some studies showed that

explicit feedback was more effective than implicit feedback, especially when the study

was conducted in a classroom. On the other hand, results from most of the laboratory

and CALL settings found no significant differences between implicit feedback and

explicit feedback. Regardless of the setting in which the research was conducted, the

treatment in most of these studies only lasted for a very short period of time.

DeKeyser's (1993) study, which lasted for a full school year, was the only longitudinal

study of the 18 reviewed here. No statistical differences were found between implicit

and explicit feedback groups. This study reminded researchers that the time span of

feedback may also affect the degree to which feedback is effective. Besides the settings

of the studies, the effects of feedback were also correlated with other factors. For

instance, the effects of feedback were affected by linguistic forms. Nagata found at

relatively more implicit feedback significantly outperformed more explicit feedback on

particles, but not on verbal predicates, which may suggest that a linguistic form may

favor one type of feedback over the other. Learners' language proficiency was also

found to be associated with the effectiveness of certain types of feedback in Ammar and

Spada's study. Prompts worked more effectively for learners with lower proficiency. For

learners with higher proficiency, however, both recasts and prompts seemed to work

equally efficiently. This result provides some insights to the current study: since learners









4.2.5.2 Treatment sessions

In designing the treatment tasks, the following issues were carefully taken into

consideration: task effects and contextual factors.

Previous studies suggested that learner factors (e.g., ethnicity, gender,

proficiency, etc.) affected their learning in task-based instructional settings (Bitchener,

1999; Gass & Varonis, 1986; Slimani-Rolls, 2005). Given the heterogeneous nature of

the participants in the current study, each task was carefully designed to make sure it

not only elicited the targeted classifiers, but also enabled each learner to complete it at

their most comfortable level. In particular, HL learners usually have greatly advanced

speaking skills compared with their non-HL counterparts (Valdes, 1995; 2001) even

though they are enrolled in the same class, and present similar or even weaker

grammatical knowledge on standard tests. Therefore the pictures used for stimuli were

carefully chosen (such as the color and shape of items, the actions of animals, etc.) to

make sure that learners had a lot to say when their language abilities allowed. Thus

more interactions could occur.

According to the flow of the information, tasks that are commonly employed in

SLA research can be divided into two categories: one-way tasks and two-way tasks

(Doughty & Pica, 1986; Long, 1988; Pica, 1987; Varonis & Gass, 1985). In a one-way

task, only one party of the dyad holds the information that is needed to complete the

task. Thus during the interaction, the information flows from the information holder to the

other party of the dyad. In contrast, in a two-way task, both parties of the dyad have the

information, and they are expected to complete the task through exchanging the

information. Previous studies have provided controversial results in terms of the task

effects of one-way and two-way tasks: many researchers have argued that two-way


122









4 RESEARCH QUESTIONS AND METHODOLOGY ................................... 101

4.1 Research Questions.............................. ............... 101
4 .2 M ethodo logy ........................................... ...... .............. 103
4.2.1 Operationalizations .............. ......... ........ .. ............... 103
4 .2 .2 D e s ig n ................................................................... 10 6
4.2.3 Participants ........... ......... .......................... ......... ...... 109
4.2.4 Target Linguistic Items ...... ..................... ............... 116
4.2.5 M materials .......................................................... .................. 120
4.2.5.1 Background information questionnaire .............. ............... 120
4.2.5.2 Treatm ent sessions .................................... ...................... 122
4.2.5.3 Testing m material .................................. ... ... ............. 124
4.2.5.4 The oral im itation test ........... .................. ... ............. 127
4.2.5.5 The untimed written cloze test...... ....... .............. .......... 129
4.2.5.6 Advantages and disadvantages of the testing design in the
current study .............. .......... .. .... .... .... .......... 130
4.2.5.7 Measurement of learners' perception of the feedback......... 131
4.2.5.8 Interview ............ .......... ....................... 135
4.2.6 Procedure .......... .......... .......................... ............... 136
4.2.7 Transcriptions ................. ...... ............... ........... 139
4.2.8 Scoring ...... ...................... .......... ........ 139
4.2.9 Coding ............ ............................. ............... 139
4.2.9.1 Classifier-related com ments ............................ ... .............. 140
4.2.9.2 Non classifier-related comments.......................... ................. 143
4.2.10 Inter-Rater Reliability ................................................ 146

5 RESULTS ................... ............................. 148

5.1 Research Question 1 .............. ........... ........................ 148
5.1.1 Descriptive Statistics ......................... ..... ............ ......... 148
5.1.1.1 C lassifier-related com m ents ........... ............... ............... 150
5.1.1.2 Non classifier-related comments........................ ............... 152
5.1.2 Inferential Statistics ............................................................ 153
5.2 Research Question 2........................................................ 154
5.2.1 Descriptive Statistics .................................... ....... .. ............... 154
5.2.1.1 Oral test ...... ............................... ..... 154
5 .2 .1.2 W ritte n test ............. .......... .... ................................... 156
5.2.1.3 Learners' previous knowledge reflected in their verbal
com m ents ........................................................ .............. 159
5.2.2 Inferential Statistics .................................................... ........ ............... 161
5.3 Research Question 3......................................... ............... 164
5.3.1 Descriptive Statistics .................................... ....... .. ............... 164
5.3.1.1 O ral im station test............................. ...... ............. 164
5.3.1.2 Written cloze test ...... .................... ............... 165
5.3.2 Inferential Statistics .................................................... ........ ............... 167




6









applied with a probability of less than 1" (p.119). The results also showed that heritage

speakers with literacy skills consistently outperformed those without literacy skills.

Romanova suggested that literacy in L1 may be a major factor influencing the

development of input-based rules and probability mechanisms in morphological

processing (Rothman, 2007).

A small-scale study conducted by Isurin and Ivanova-Sullivan (2008) investigated

seven heritage speakers, 11 L2 learners, and five native speakers of Russian. All the

participants were requested to tell a story based on a children's picture book, A boy, a

dog, a frog and a friend. The elicited speech samples from the three groups of learners

were analyzed with a focus on selected morphosyntactic categories (e.g., aspect, case),

and word order. They found that heritage speakers outperformed L2 learners in all three

morphosyntactic categories.

Lynch (2008) conducted a qualitative study on five HL and four L2 learners of

Spanish. The five HL learners selected in this study were either born in the U.S. or

migrated here before age two, and had two to five years of formal study of Spanish. In

other words, they were all typical lower-proficiency HL learners. On the other hand, the

L2 learners in the study all had more than five years of formal study of Spanish,

meaning they were generally at a more advanced level of Spanish L2 learning. Data

were collected from individual interviews in Spanish by the researcher. Selected

grammatical features were quantitatively analyzed, such as noun-adjective gender

agreement, aspectual and mood distinction, subject-verb word order, etc. The overall

results showed more similarities than differences between the two groups. The two









Explicit feedback in an implicit condition was operationalized by Rosa and Leow

(2004) as a simple indication of whether the answers made by learners are correct or

not. Negative evidence, which was employed in Leeman's study, was operationalized

as feedback containing evidence "on ungrammatical noun-adjective agreement" (2003,

p. 49). With repetition, teachers "replicate the student's error verbatim, usually with

rising intonation and stress to highlight the error" (Lyster, 2004, p. 405). Elicitation, in

contrast, can include three techniques: first, using direct questions such as 'how do we

say that in French?'; second, pausing to allow the student to complete the utterance

initiated by the teacher; third, asking the learner to reformulate the utterance (Lyster &

Ranta, 1997).

In Carroll and Swain's study (1993, 2001), learners in an explicit hypothesis

rejection group were told that they made an error, followed by an explicit explanation of

that error. Meta-linguistic feedback has been frequently compared with recasts

regarding the impact of each on L2 development (e.g., Ammar & Spada, 2006; Carroll &

Swain, 1993; R. Ellis et al., 2006; Kang, 2009; Kim & Mathes, 2001; Sauro, 2009;

Sheen, 2007). The former usually contains "comments, information, or questions related

to the well-formedness of the student's utterance, without explicitly providing the correct

form" (Lyster & Ranta, 1997, p. 47). As the correct linguistic form is not provided in the

meta-linguistic feedback, learners are pushed to produce the correct form by

themselves based on the provided meta-linguistic information.

The above dichotomy of implicit and explicit feedback has been questioned by

researchers (e.g., R. Ellis et al., 2006; Loewen and Nabei, 2007), who argued that the

line between these two types of feedback is not always clear: to what extent should









social, linguistic, and demographic backgrounds: they can be foreign-born immigrants

who moved to the U.S. at a young age, native-born children with foreign-born immigrant

parents, or native-born children of native-born parents with immigrant background

(Valdes, 2005). Because of the heterogeneous nature of the population in this group,

defining the HL learner is not an easy task. In the literature, three criteria have been

employed to define HL learners: their membership in the HL community, their personal

connection through family background, and their greater or lesser linguistic proficiency

in the HL (Carreira, 2004). Based on the above criteria, numerous definitions have been

proposed. These definitions can be roughly divided into two groups: a broadly defined

group, and a narrowly defined group (Kondo-Brown, 2005).

Definitions belonging to the broadly defined group tend to focus on learners' ethnic

background connection to their "own ancestral language (i.e., indigenous, colonial, and

immigrant languages)" (Kondo-Brown, 2005, p.564), or their "heritage motivation" (Van

Deusen-Scholl, 2003, p.222). The most influential and representative definition in this

group proposed by Fishman, defined HL learners based on two criteria: learners who

speak LOTEs (languages other than English), and who have a particular family

relevance to the language (2001).

Since definitions in the broadly defined group only emphasize the ancestral, ethnic

background connection to the language, without considering proficiency level, they often

include within the group many members who actually cannot speak a word in the HL

(Van Deusen-Scholl, 2003). Consequently, in university-level FL classes, students who

seek to "reconnect with their family's heritage through language" are also considered as

HL learners, "even though the linguistic evidence of that connection may have been lost









(only one case), which indicated their confidence about their pronunciation. In contrast,

the non-CHL learners made most of their comments about phonology, particularly

regarding the tone. Examples 6-12 and 6-13 illustrated these results.

Example 6-12 A CHL learner' phonological comment
NNS: *-A A .o
You liu duo hua
Have six CL flower
There are six flowers.
NS: T,,--7 ?
Ni de hua dai bu dai zhizi?
You NOM flower have no have stem
Does your flower have a stem?
NNS: o
You
Have
(it) has.
NS : J[ T,-fo
Na ni you liu zhi hua
Then 2sg have six CL flower
Then you have six flowers.
Recall data
I couldn't pronounce duo.

Example 6-13 A non-CHL learner's phonological comment
NNS: *E-to
Wu ge hua
Five CL flower
(There are) five flowers.
NS: ------ I A A-4A to
Yi er san si wu liu, you liu duo hua
One two three four five six You six CL flower
(Count) one, two, three, four, five, six, there are six flowers.
NNS: Sorryo
Recall data
I got confused with huar (tone 1) and huar(tone 4).

On the other hand, the CHL learners in the recast group made most

of their comments about morphosyntac issues. Interestingly, like the non-


183









In Example 6-3, the CHL learner did not perceive the provided recast as a

corrective feedback, which was reflected in his response toward it: instead of producing

an immediate uptake to modify the non-targetlike classifier, he simply responded to

continue the interaction. A post hoc analysis was conducted to investigate this learner's

immediate uptake pattern after feedback during the interaction. Uptake here refers to

"the learners' modification of their original utterance following the NS's provision of

feedback through recasts or meta-linguistic feedback" (Mackey et al., p. 492). As

discussed in Chapter 3, uptake may not be a reliable measure of noticing of the

feedback, particularly in the case of recasts. However, it may be related to learners'

perceptions about feedback at the time of the feedback. The result showed that learner

XSB tended to provide uptake after receiving feedback. In addition, for those recasts for

which he did not provide uptake, none was accurately perceived. These results are

shown in tables 6-1.

Table 6-1. Frequency of uptake and perception of learner XSB
+Uptake (n=18) 81.82% -Uptake (n=4) 18.18%
Perceived Not perceived Perceived Not perceived
10 8 0 4
(55.56%) (44.44%) (0%) (100.00%)

In addition, a second post hoc analysis showed that learner XSB used the same

problematic classifier zhang in both his oral and written pre- and post-tests. All this

evidence suggested that this learner assumed zhang was the correct classifier for mirror.

The target classifier mian provided by the interlocutor at the moment of the interaction

was apparently assumed to be just an alternative one by this learner, and never

assimilated into his interlanguage system.


174









the rapid growth of the HL population in postsecondary FL/ SL classrooms since the

1970s (Valdes, 2005). In less than two decades, the field of CHL instruction has made

astounding progress. More and more language teachers in postsecondary Chinese

programs have to face CHL learners and non-CHL learners in the same classroom. Due

to their different language backgrounds in Chinese, these two groups of learners have

different language behaviors and language needs. However, little empirical research

has been done in the past to compare the linguistic behaviors of these two groups of

learners.

Motivated by gaps in the literature, the present study empirically investigates the

effects of recasts and meta-linguistic feedback on learners with CHL background, in

comparison with learners without CHL background. The following three issues are

particularly explored: learners' perceptions of recasts and meta-linguistic feedback; the

developmental benefits of recasts and meta-linguistic feedback; the relationship

between feedback type and the increase in learners' performance on the oral and

written tests.

Chapter 2 highlights some of the fundamental issues associated with HL learners,

with a particular emphasis on CHL learners. In particular, it asks and answers the

following questions: who are HL learners? How do they differ from non-HL learners? In

the first part, the chapter will examine HL learners in general. It will provide definitions of

HL learners based on previous studies. Then it will discuss issues related to HL

learners' language proficiency, including various variables that could be employed to

predict their proficiency levels. The chapter will then present a review of previous

research on the comparison between HL learners and non-HL learners. The second










the recast group (M=73.45) scored slightly lower than those in the meta-linguistic group

(M=78.05). Learners in both experimental groups scored much higher than learners in

the control group (M=50.54). On the other hand, for non-CHL learners, learners in the

recast group (M=35.07) also scored slightly lower than learners in the meta-linguistic

group (M=36.64). Learners in the control group (M=15.52) scored much lower than

learners in the two experimental groups.

Table 5-12. Summary of oral post-test scores of CHL learner group
Group N Minimum Maximum M SD
RHL 12 42 92 73.45 16.53
MHL 11 23 94 78.05 21.18
CHL 8 19 83 50.54 22.59


Table 5-13.


Summary of oral


post-test scores of non-CHL learner group


Group N Minimum Maximum M SD
RNHL 12 10 75 35.07 23.24
MNHL 12 6 40 36.64 11.52
CNHL 9 6 90 15.52 21.54


80-


60-


CO 40-


20-


RHL MHL CHL RNHL MNHL CNHL
Group
Figure 5-5. Mean scores of oral post-test


5.2.1.2 Written test

Tables 5-14 and 5-15 below summarize the descriptive statistics by CHL learner

groups and non-CHL learner groups during the pre-test. For CHL learners, learners in


156









much longer length of formal Chinese study than the other non-HL participants. Four

non-HL learners were excluded because they were heritage speakers of Spanish (n= 4).

Although their HL is not Chinese, their HL experience might provide them some

advantages over the English monolinguals. Seventeen HL learners were excluded

because their home dialect was a dialect other than Mandarin (Cantonese=12;

Fujian=33; Wenzhou=1; Shanghai=1). Four HL learners withdrew from the experiment in

the middle due to a scheduling conflict. Thus a total of 64 (male=35; female=29)

learners were left in the final pool of participants. Among them, there were 23 HL

learners and 24 non-HL learners in the experimental groups; eight HL learners, and

nine non-HL learners were included in the control group.

When the data collection was conducted, 15 participants were enrolled in the

Heritage Class, which is equivalent to a second-year level traditional Chinese class.

Twenty-one participants were enrolled in the third year class and 28 were enrolled in the

second year course. Their ages ranged from 18 to 23, with an average age of 19.9

(SD=1.20). The average length of formal Chinese study was 10.9 months for CHL

learners (SD=6.8); 19.4 months for non-CHL learners (SD=7.9). The average length of

stay in China was 1.3 months (SD=8.0). All participants completed an informed consent

form (see Appendix B), and agreed to receive 0.3-0.5% extra points for their final class

grade from their instructors upon the completion of the study.

Why were only HL learners with Mandarin L1 and non-HL learners with English L1

involved in this study? Why were learners at high-intermediate and advanced level

chosen for the experiment? These two questions will be answered below.


110









months for the CHL learners, but 19.4 months for the non-CHL learners. Table 6-4

summarizes learners' use of Chinese characters on the written pre- and post-test.

The results in Table 6-4 indicate that with the exception of the CHL learners in the

recast group, learners generally show a similar level of knowledge of Chinese

characters. The non-CHL learners used even more Chinese characters than the CHL

learners. Although the non-CHL learners in the two experimental groups showed a

similar level of knowledge of Chinese characters, those in the meta-linguistic feedback

group outperformed those in the recast group by approximately eight percent. A

possible explanation for this is that the meta-linguistic information in the feedback

helped learners connect the target classifier and the associated Chinese character

which they had acquired in other contexts in the past, which is illustrated in Example 6-

25.

Table 6-4. Learners' knowledge of Chinese characters on the pre- and post-test
CHL learners Non-CHL learners
Recast Meta Recast Meta
Pre-test 136 (51.5%) 74 (30.5%) 159(60.7%) 148(56.0%)
Post-test 78 (29.5%) 40 (16.5%) 94 ((35.6%) 77 (29.2%)
Pre-test (without ge4) 77 (29.2%) 34 (14.0%) 51 (19.3%) 48 (18.2%)
Post-test (without ge) 72 (27.3%) 37 (15.3%) 54 (20.5%) 74 (28.0%)

Example 6-25 A non-CHL learner in the recast group
NNS: *-#' o
Yi jian maozi
One CL hat
One hat.
NS: ---1i iJ on top.o IJR41I --'. 7 414It ?
You wu duo huar
Have five CL flower
(There are) five flowers.
NNS: ...
NS: Top of the head il'Mt ? Top of the mountain ?
NNS : -INII-To


206









In summary, the descriptive statistics of the oral and written pre-tests showed that

learners in the recast group scored slightly higher than learners in the meta-linguistic

group. That indicated that learners in the recast group had a slight advantage over

those in the meta-linguistic group before they received any feedback during the

treatment. However, during the post-test, learners in the recast group were

outperformed by those in the meta-linguistic group. The only exception occurred for

CHL learners in the written test: learners in the recast group outperformed those in the

meta-linguistic feedback group. These comparisons are summarized in table 5-18

below. Inferential statistics were employed to test whether any group differences existed

in both pre- and post-tests. The results will be presented in 5.2.2.

Table 5-18. Comparison of pre-test and post-test scores across the groups
Test Pre Post Pre Post
Oral RHL>MHL>CHL MHL>RHL>CHL RNHL>CNHL>MNHL MNHL>RNHL>CNHL
Written RHL>MHL>CHL RHL>MHL>CHL RNHL>MNHL>CNHL MNHL>RNHL>CNHL


5.2.1.3 Learners' previous knowledge reflected in their verbal comments

Tables 5-19 and 5-20 summarize learners' previous knowledge of the target

classifiers based on their verbal reports collected in the stimulated recall. Figures 5-8

and 5-9 graphically present the results. The percentage of (re)-acquisition is

approximately two times higher in the CHL learner group (M=14.0%) than in the non-

CHL learner group (M=6.8%), indicating that CHL learners had knowledge of these

classifiers before. CHL learners also present a higher percentage in the category of A2

(M=48.8%) than non-CHL learners (M=35.2%), suggesting that they knew a wider range

of classifiers. A slightly higher percentage for the category of Al in non-CHL groups

(M=31.3%) indicated that they were more likely to use the general classifier ge instead


159









one hand, and ignore the competence on the other. Since HL learners are more likely to

have a much wider range and higher level of competence compared to their

performance, their proficiency levels tend to be underestimated under the current

performance-oriented proficiency assessment system.

In order to access the real picture of a HL learner's existing linguistic knowledge

system, it would be best to conduct longitudinal research, observing each linguistic

development stage of the learner. However, it is not realistic to do so for every learner.

Nevertheless, the more we know about the background of a HL learner, the better we

can infer his/her language skills. In fact, a number of social variables have been

identified as possible predictors of HL learners' proficiency.

Learner generation has been found as an important predictive variable of HL

maintenance and attrition. Generally speaking, HL learners of an older generation

usually have higher proficiency in their HL than those of a younger generation (e.g.,

first-generation compared to the second-generation, second-generation compared with

the third-generation, etc). For first- generation immigrants, age of arrival is an important

predictive variable: usually older arrivals with exposure to English after 12 years old

show higher proficiency than younger arrivals in their HL, especially where

pronunciation (Flege et al., 1995; Yeni-Komshian et al., 2000), and lexical retrieval are

concerned (McElree et al., 2000). However, some individuals have lower proficiency in

English. Excluding the first-generation immigrants, birth order plays an important role in

predicting HL learners' proficiency: first-born children tend to have higher levels of HL

proficiency than their younger siblings (Lambert & Taylor, 1996; Zentella, 1997).









CHL and non-CHL learners' profiles. The following three learner profiles

represent two types of CHL learners and the non-CHL learners in the current study. For

the two CHL learners, learner WKW represents learners who were born in the U.S.,

whereas learner YHX represents learners who were foreign-born, but emigrated to the

United States at young age. These two profiles briefly illustrate how these two types of

CHL learners developed their HL before they took formal college-level Chinese courses.

WKW. WKW was born in the U.S. His parents spoke only Chinese with him at

home until he attended kindergarten at age five. Besides his parents, he also practiced

his Chinese with his grandfather when the latter visited the United States. He learned

many Chinese classical poems from his grandfather. There was a clear change before

and after he started kindergarten: before attending kindergarten, his mother read

Chinese stories to him. He also watched Chinese TV programs; after attending

kindergarten, his dominant language rapidly shifted from Chinese to English. He still

watched some Chinese TV programs, such as the TV series The Journey to the West,

The One Hundred and Eight Heroes, but he could only understand the plots, not the

details. Worrying that he might totally forget his Chinese, his parents sent him to a

community Chinese language school when he was seven years old. He went there

every Sunday for a three-hour class until he was sixteen. At that time he felt learning

Chinese was useless since it was not commonly used in this country. Later when he

took postsecondary Chinese classes, he started to realize that he was re-acquiring

many words he learned when he was young. When the data was collected, he had been

taking Chinese course at the University of Florida for eight months.


114









linguistic feedback made a greater increase in the written post-test than those who

received recasts during the treatment. The effect size showed that this factor accounted

for 72.8% of the variance in the data, which is a relatively large effect. In contrast, the

main effect of the CHL background variable, is not statistically significant (F1,43 = .032, p

= .859, partial eta-squared= .01 power= .054). However, the interaction between

feedback type and CHL background was found to be statistically significant (F1,43 =

4.663, p = .036, partial eta-squared= .098 power= .560). The effect size showed that

the interaction factor accounted for 56.0% of the variance in the data, which is a

relatively small effect. Thus we really want to know whether the combinations of

feedback and learners' language background are statistically different from each other.

A post hoc test was employed with an adjusted p-value of .0121. The result showed that

for the CHL learners, the feedback type was not very important (p=.772). In contrast, for

the non-CHL learners, the feedback type was very important (p=.002): those in the

meta-linguistic feedback group performed significantly better than those in the recast

group.


1 .012 is the cut-off value of False Detection Rate (FDR) caculated by R statistical
program.


168









was instructed to try again, and make a correct selection before moving to the next item.

For the implicit condition, learners only received feedback that indicated whether the

answer was right or wrong, which was similar to the implicit feedback employed in

Nagata's study. Two tests were employed in this study: multiple-choice recognition tests

and controlled-production tests. In the multiple-choice recognition test, learners needed

to choose the correct answer from four options. In the controlled-production test,

learners were asked to complete sentences by providing an appropriate verb form. The

results showed that learners who received explicit feedback reported higher levels of

awareness of the target structures than those who received implicit feedback. In

addition, learners who reported being aware of the target structure did significantly

better than those who did not report being aware of it.

Sanz (2003) employed computer delivered input processing instruction, in which

learners received a response from the computer of "ok" if they gave a correct answer,

and "sorry, try again" if their response was incorrect. Two interpretation tests were

employed in this study: a sentence completion test and a written video retelling test.

Although both groups did significantly better on the post-test; no significant differences

were found between the two groups.

Sauro (2009) compared recasts and meta-linguistic feedback on the English zero

article among 23 high intermediate/ advanced EFL adult learners. The treatment

consisted of two computer-mediated collaborative writing activities, in which the learner

was paired with a native English-speaking chat partner in individual chat rooms using

the Virtual Classroom chat tool from Blackboard. Feedback was provided by the chat

partner whenever the learner made an error with the target form. The study employed a









Swain, M. (2005). Verbal protocols: What does it mean for research to use speaking as
a data collection tool? In M. Chaloub-Deville, M. Chapelle & P. Duff (Eds.),
Inference and generalizability in applied linguistics: Multiple research
perspectives (pp. 97-113). Amsterdam: Benjamins.
Swain, M. & Lapkin, S. (2002). Talking it through: Two French immersion learners'
response to reformulation. International Journal of Educational Research, 37,
285-304.
Szeto, K.-S. K. (1998). The acquisition of Cantonese classifiers. Unpublished doctoral
dissertation, University of Hong Kong.
Tse, L. (2000). The effects of ethnic identity formation on bilingual maintenance and
development: An analysis of Asian American narratives. International Journal of
Bilingual Education and Bilingualism, 3, 185-200.
Valdes, G. (1995). The teaching of minority languages as academic subjects:
Pedagogical and theoretical challenges. The Modern Language Journal, 79, 299-
328.
Valdes, G. (2001). Heritage language students: Profiles and possibilities.. In J. K.
Peyton, D. A. Ranard & S. McGinnis (Eds.), Heritage languages in America:
Preserving a national resource (pp. 37-77). McHentry: the Center for Applied
Linguistics/ Delta Systems.
Valdes, G. (2005). Bilingualism, heritage language learners, and SLA research:
opportunities lost or seized? The Modern Language Journal, 89, 410 426.
Van Deusen-Scholl, N. (2003). Toward a definition of heritage language: sociopolitical
and pedagogical considerations. Journal of Language, Identity, and Education.,
2,211-230.
Varonis, E. & Gass, S. M. (1985). Non-native/non-native conversations: a model for
negotiation of meaning. Applied Linguistics, 6, 71-90.
Warden, M.,Lapkin, S.,Swain, M. & Hart, D. (1995). Adolescent language learners on a
three-month exchange: Insights from their diaries. Foreign Language Annals, 28,
537-550.
Weger-Guntharp, H. (2006). Voices from the Margin: Developing a profile of Chinese
heritage language learners in the FL classroom. Heritage Language Journal, 4,
29-46. retrieved November 11, 2009, from
http://www.international.ucla.edu/languages/heritagelanguages/journal/article.asp
?parentid=53016.
Wiley, T. G.,Klerk, G. d.,Li, M.,Liu, N.,Teng, Y. & Yang, P. (2008). Attitudes toward
Mandarin, heritage languages, and dialect diversity among Chinese immigrants
and international students in the United States. In H. A. Weiyun & X. Yun (Eds.),
Chinese as a heritage language: Fostering rooted world citizenry (pp. 67-87).
Honolulu: University of Hawai'i Press.
Wong Fillmore, L. (1991). When learning a second language means losing the first.
Early Childhood Research Quarterly, 6, 323-346.
Wu, S.-M. (2008). Robust learning for Chinese heritage learners: Motivation, linguistics
and technology. In K. Kondo-Brown & J. D. Brown (Eds.), Teaching Chinese,
Japanese, and Korean heritage language students (pp. 271-297). New York:
Lawrence Erlbaum Associates.


237










Classroom studies of effects
Target
participants Trgt
structure
35 Dutch high Variety of
school seniors features,
learning L2 predominant
French ly
morphosynt
actic


Table 3-3.
Study
DeKeyser
(1993)








Muranoi
(2000)







Havranek
& Cesnik
(2001)


Indefinite
article to
denote new
information





Varity of
English
phonological
, lexical, and
grammatical
features


of implicit and explicit feedback
Design


Groups: (A) extensive
explicit corrective feedback
during normal class
activities, (B) limited explicit
corrective feedback.
Treatment: 10 class periods




Groups: (A) interaction
enhancement (IE) by means
of requests for repetition
and recasts in
communicative task +
formal debriefing (explicit
grammar explanation), (B)
IE + meaning-focused
debriefing, (C) control
Data on 1700 corrective
feedback episodes from
normal English lessons.


Tests Results


Three oral communication
tasks (interview, picture
description and story-
telling). Fill-in-the-blank
test. Tests administered
twice.




Design: pre/immediate
post/delayed post-test
Tasks: 1. oral story
description task, 2. oral
picture description task, 3.
a written picture
description task, 4.
grammaticality judgment
test.
Class-specific tests
(translation, correction,
reading aloud, and written
and spoken completion
tasks). Directed at
corrected items.


114 Japanese
college EFL
students (first
year)





207 university
students
specializing in
English


No statistically significant
differences evident
between groups A and B.
Learners with high
previous achievement,
high language aptitude,
high extrinsic motivation,
and low anxiety benefited
the most from error
correction.
Both experimental groups
outperformed the control
group on both post-tests.
Group A outperformed B
on immediate post-test but
not on delayed post-test.



Elicited self-correction was
the most efficient
combinations for all
learners.









least one native-Japanese-speaking parent presented significantly higher proficiency

than the rest of the learners. Her findings indicate that HL learners with lower

proficiency are more like FL learners in terms of their linguistic skills. Even within the HL

learner group, the between-group difference is quite remarkable. Therefore simply

separating HL learners from FL learners without considering their proficiency

differences will risk putting learners with completely different language abilities in the

same classroom. Lynch (2008) compared HL learners with lower-proficiency and L2

learners at advanced level. He found more similarities than differences between the two

groups. I will provide more details about Lynch's study in section 2.1.4.

2.1.3 Predictive Variables of HL Learners' Proficiency In Their HL

HL learners come to the classroom with a wide range of linguistic knowledge and

abilities, which set them apart from L2/ FL learners who had no previous exposure to

the target language. Even with extremely limited language input, the general sense and

intuition about the language and the culture that they developed still gives HL learners

advantages over their non-HL counterparts. Thus unlike non-HL learners who need to

learn everything from scratch, HL learners already know something about the language.

The question is: what do they know? And to what extent do they know it? Without

knowing the answers to these questions, it is impossible to provide appropriate and

efficient instruction to HL learners. However, finding an answer to these questions is an

extremely difficult and complex task. As Hendrys (2008) argued, HL learners'

knowledge system is similar to an iceberg, which is formed by a smaller visible

performance on the surface, but a much bigger invisible competence underneath.

Therefore HL learners' real proficiency is hard to assess with the current proficiency

assessment methods, which are usually overwhelmingly based on performance on the









6.4.2 Pedagogical Implications

The participants in the current study included a combination of traditional FL

learners of Chinese and learners who had early exposure to Chinese as a HL. They

represent the diverse learner population in postsecondary Chinese programs in the

United States. Therefore the findings in the current study provide invaluable implications

for Chinese language education, particularly in language curriculum development and

classroom instruction.

Currently, postsecondary Chinese language classes generally implement either a

single-track or a dual-track language program. The former puts all the learners into the

same curriculum, ignoring their language background; the latter separates learners with

CHL background from those without CHL background. A dual-track language program

usually includes two types of classes: the regular class, in which CHL learners and non-

CHL learners attend the same class; and the HL class, in which CHL learners with

strong oral speaking proficiency and some literacy skills are enrolled. In either track, the

curriculum is designed based on an assumption that the traditional speaking, listening,

reading and writing skills should be equally emphasized. The participants of the current

study were recruited from these two types of classes. However, the results showed that

the CHL learners had a much higher level of oral proficiency than literacy skills. In

contrast, the non-CHL learners demonstrated more evenly distributed oral and literacy

skills. Thus the current curriculum is apparently more able to accommodate the

language needs of non-CHL learners than CHL learners. The results of the current study

showed a great gap between the oral and written skills presented by the CHL learners.

In addition, the CHL learners scored very close to the non-CHL learners on their written

test. Therefore although CHL learners still need to improve their oral productive skills in


210









1993; Lyster & Izquierdo, 2009) to measure learners' L2 development. In these studies,

a pre-test was given to the participants in order to ensure group comparability prior to

the treatment (Mackey & Gass, 2005). The short-term effects of treatment were

measured by a immediate post-test, whereas the long-term effects were measured by a

delayed post-test. A wide range of tasks have been employed for tests, including

grammaticality judgment tests, sentence completion, picture prompt tests, translation

tests, oral imitation tests, picture description tests, sport the differences tasks, jigsaw

tasks, consensus tasks, ordering tasks, and consciousness-raising tasks (for a review

see Mackey & Gass, 2006 ). The target linguistic structures also vary. The most

frequently examined ones have been syntactical, morphosyntactical, and morphological,

such as dative verbs (Carroll & Swain, 1993; Kim & Mathes, 2001), verbal predicates

and particles (Nagata, 1993), indefinite articles (Muranoi, 2000), derivations of nouns

from verbs (Carroll, 2001), noun-adjective agreement (Leeman, 2003), pronouns (Sanz,

2003), gender (Lyster, 2004), verb past tense (R. Ellis et al., 2006; R. Ellis, 2007), and

question forms (Loewen & Nabei, 2007). The target structures in these studies have

different degrees of complexities in terms of learners' processing and operational

burdens. Some structures might only require simple explicit knowledge (Leeman, 2003),

while others, such as question forms (Loewen & Nabei, 2007), may require learners to

acquire more complex rules.

The designing of tests is highly related to validity and reliability of the final results.

The tests employed in previous studies often more likely involved grammaticality

judgment tests, sentence completions, picture prompt tests, and translation tests, etc.

All of these tests favor explicit knowledge rather than implicit knowledge. Therefore









experiences; however, the following two language-specific factors also play important

roles.

The first is the home dialect factor. As Valdes (2001) pointed out, in many cases

immigrant students are speakers of non-prestige varieties of their HL. In terms of CHL

learners, as we have seen above, they or their ancestors were originally from different

hometowns where various Chinese dialects are spoken. Therefore their home dialects

also vary accordingly. For instance, HL learners who are originally from Hong Kong

mostly speak Cantonese at home, while those from mainland China more likely speak

Mandarin, or a mix of Mandarin and other dialect(s) (Wiley et al., 2008).

In the U.S., Mandarin is the most widely taught variety of Chinese in Chinese

language classrooms. According to a survey conducted by the Modern Language

Association (MLA), in 2006, 51,582 students enrolled in Mandarin class at the

postsecondary-level, which ranked as the 7th most commonly taught language other

than English in the U.S. On the other hand, other dialects of Chinese are hardly even

introduced into mainstream educational institutions. According to the same survey, the

enrollments in Cantonese classes were only 178, while only 21 students took

Taiwanese, which is one of the varieties of Min spoken in Taiwan. No records were

available for other dialects. This indicates that those dialects are either not taught in the

classroom at all, or the enrollments are too small to be considered in the survey

(Modern Language Association, 2006).

Thus for many HL learners in the U.S., learning their HL in a postsecondary-level

classroom means learning a second dialect (Valdes, 2005) when their home dialect is









The first type is known as a qualifying classifier, while the second type is called a

quantifying classifier.

There are overwhelmingly more quantifying classifiers than qualifying classifiers.

The estimated number of classifiers runs from as many as 900 in Hanyu Liangci Cidian

aif~ ,"A dictionary of Chinese classifiers" (Bitchener, 1999), to as few as

several dozens (Chao, 1968; Li & Thompson, 1981). However, qualifying classifiers

were estimated at only about 70 (Hu, 1993).

Although the number of qualifying classifiers is much smaller than that of

quantifying classifiers, the acquisition of qualifying classifiers challenges L1 Chinese

children more.

Qualifying classifiers include a general classifier and dozens of specific classifiers.

Specific qualifying classifiers categorize nouns into semantic classes based on their

different physical properties, such as shape, size, animacy, etc. (Harred et al., 1972;

Hu, 1993), making the classifier selection in Chinese an extremely complex task for

young children. Children have to choose appropriate classifiers by relying solely on the

semantic co-occurrence constraints on the classifier- head noun structure, with no

morphological cues.

The most frequently used qualifying classifier in Mandarin is 4 ge, also known as

a general classifier. It is usually the first classifier acquired by Chinese children

(Erbaugh, 1986; Hu, 1993; Liu, 2008; Loke, 1991; Loke & Harrison, 1986; Lu & Li,

2008)4. Specific classifiers emerge after the general classifier, following a certain order

determined by the perceptual salience of the object properties. For instance, children


118









meta-linguistic knowledge. On the other hand, tests designed to measure implicit

knowledge expect learners to produce the TL in a spontaneous manner within a limited

time period.

However, solely relying on one type of tests harmed the construct validity of the

research results. For instance, R. Ellis (2007) criticized some research on the effects of

feedback for relying too much on testing instruments that better measure explicit L2

knowledge. As a result, explicit feedback might be found to be more effective than

implicit feedback in these studies. In order to avoid the above problems, it was decided

to employ multiple testing instruments that could measure both implicit and explicit

knowledge, as a solution to fully portray the degree of improvement caused by the

feedback.

R. Ellis and his colleagues (R. Ellis, 2007; R. Ellis et al., 2006) used three different

types of tests to measure the effects of implicit and explicit feedback on English past

tense -ed: an oral imitation test, an untimed grammaticality judgment test, and a meta-

linguistic knowledge test. Among the three tests, the oral imitation test aimed at

measuring learners' implicit knowledge, while the un-timed grammaticality judgment test

and the meta-linguistic knowledge test were used to assess learners' explicit

knowledge. The oral imitation test consisted of 36 audio-recorded belief statements.

During the test, participants first indicated on an answer sheet whether they agreed,

disagreed, or were not sure about the statement. Next they were requested to repeat

each statement in correct English. The untimed grammaticality judgment test consisted

of 45 written sentences, including seven grammatically correct sentences and eight

grammatically incorrect sentences, as well as 30 distractor sentences. Learners were


125









The differences and similarities between HL and non-HL learners provide

researchers in field of SLA with a good opportunity to re-conceptualize and expand the

field by "examining possible interactions between SLA and the area of language

instruction currently referred to as the teaching of heritage languages" (Valdes, 2005, p.

410). Lynch (2003) argued that "the framing of a coherent agenda for research and

theory building in the field of heritage language acquisition (HLA) depends in part upon

the research and theory already existent in second language acquisition

(SLA)... Clearly, the general sorts of questions asked in SLA are questions that HLA

researchers must be asking, and the research methodologies used to respond to those

questions in SLA are methodologies that would lend themselves fruitfully to HLA

endeavors" (p.1). The following two studies approached HL acquisition with typical SLA

research designs.

Kim (2008) investigated sentence processing differences between 88 HL and 40

non-HL learners of Korean. Based on their self-reported L1, these learners were

subcategorized into five groups. HL learners were divided into Korean L1 speakers,

Korean and English L1 speakers, and English L1 speakers. Non-HL learners were

divided into two groups: Japanese L1 speakers, whose language is also a SOV

structure like Korean, and Korean L2 learners whose L1 is a SVO structure (e.g.,

English, Chinese, etc). All the participants were given a picture selection task, which

contained 84 test items, aiming to assess learners' comprehension of Korean relative

clauses. The results showed that HL learners with Korean as their L1 and L2 learners

with Japanese as their L1 earned the highest scores among all the participants.

Moreover, the analysis of error types indicated that HL learners found case markers









as a productive, hypothesis-testing device whereby learners think aloud as they work to

solve L2 problems'" (p.53).

Ohta's study is important for the following reasons. First, in contrast to studies

conducted within an interactionist framework, this study employed a qualitative

discourse-analytic approach in its data analysis, which enabled the researcher to

examine the efficacy of recasts from an emic rather than etic perspective. The

researcher used Conversation Analysis with specific attention to overlap, pauses, as

well as the levels of volume of learners' utterances. The collected data were viewed in

an holistic fashion. Second, besides the learner who was addressed by the teacher and

received the recasts, the study also looked at the effects of recasts on learners who

were not being addressed, but participated in the class as auditors. In fact, they were

the ones who produced private speech more often than the learners who received the

recasts did. The private speech produced by the auditors also provided counter-

evidence to the notion that recasts are ineffective, since they elicited fewer responses

from the learners being addressed. In this study, the auditors were found to actively

participate in the class through private speech. As argued by Ohta, "private speech

provides powerful evidence of the mental activity triggered by the noticing of contrasts

between ill-formed and correct utterances. The efficacy of recasts should not be

doubted based on the presence or absence of an overt oral response" (p.66).

A problem of employing private speech to examine the effectiveness of recasts is

that learners may not be equally talkative. In fact, among the seven learners who

participated in the study, only four of them produced private speech with moderate to

high frequency. Ohta's study suffered less from this drawback as it was a longitudinal









feedback be overt enough to be considered explicit? How covert must feedback be to

be categorized as implicit? For instance, although recasts are generally considered

implicit, not all of them are as implicit as is generally assumed. In fact some of them are

quite explicit, especially when saliency cues (e.g., phonological emphasis) are involved

(R. Ellis et al., 2006). These cues refer to "additional intonational or verbal cues or

signals to highlight the error and/ or push the learner further to respond to the feedback"

(Nassaji, 2007, p. 526). For instance, in Doughty and Varela's study (1998),

phonological stress was employed to indicate the source of the problem.

Researchers have proposed that instead of adopting the dichotomic distinction of

implicit and explicit feedback, the explicitness of negative feedback should be

considered as constituting a continuum, as shown in Figure 3.1 (based on Loewen &

Nabei, 2007, p. 362).

Loewen and Nabei proposed a feedback classification scheme that classifies

feedback on two criteria: (a) provision of L2 models and pushed output, and (b)

explicitness of feedback. Based on the two criteria, they divided six commonly

employed feedback types in the classroom (Lyster & Ranta, 1997) into two subgroups:

prompts (e.g., Ammar & Spada, 2006; Lyster, 2004; Lyster & Mori, 2006) and provide

(Loewen & Nabei, 2007). The prompt group includes meta-linguistic feedback,

elicitation, repetition, and clarification requests. These four types of feedback share one

characteristic in common: they all prompt learners to produce output without providing

L2 models. In contrast, the provide group consists of two types of feedback: explicit

correction and recasts. As the name provide suggests, both types of feedback provide









or input enhancement (Norris & Ortega, 2000). However, they think the role of practice

is only limited to item learning and fluency improvement. Some researchers believe that

explicit learning works effectively in L2 acquisition through systematic practice (e.g.,

DeKeyser, 2003; Schmidt, 1990, 1994, 1995; Schmidt & Frota, 1986; Swain, 1985),

because practice could "gradually bridge the gap between explicit knowledge and

language use" (DeKeyser, 2003, p. 328).

3.1.2 Implicit vs. Explicit Knowledge

There are two types of linguistic knowledge internalized in learners' mind: implicit

knowledge and explicit knowledge. Implicit knowledge refers to "knowledge that learners

are only intuitively aware of and that is easily accessible through automatic processing",

whereas explicit knowledge consists of "knowledge that learners are consciously aware of

and that is typically only available through controlled processing" (R. Ellis et al., 2006, p.

340). Implicit knowledge is knowledge of language, which can be further distinguished into

two types: formulaic knowledge, which is internalized in learners' mind as unanalyzed

units. For instance, NS usually know a large number of fixed or semi-fixed expressions.

Another type is rule-based knowledge, which facilitates learners to produce novel

sentences. On the other hand, explicit knowledge is knowledge about the language, which

is analyzed, abstract, and explanatory (R. Ellis, 1994). Learners usually are able to

verbalize explicit knowledge if called upon. However, this knowledge is not dependent on

meta-linguistic knowledge (Bialystok, 1994, p.567).

Implicit and explicit knowledge distinguished from each other mainly in the following

areas: awareness, type of knowledge, systematicity and certainty of L2 knowledge,

accessibility of knowledge, use of L2 knowledge, self-report, and learnability (R. Ellis,

2005), which are summarized in Table 3-1. Implicit knowledge is intuitive, in the sense









Example 4-13
Feedback episode
NNS: *RJ Mo
Wo mai si bi.
1sg buy four pen
I bought four pens.
NS: &- 7 IV ?
Ni mai le si zhi bi.
2sg buy PFV four CL pen
You bought four pens.
NNS: MV0o
Si zhi bi.
Four CL Pen
Four pens.
Recall data
Again with the measure word, it's always the measure words.

4.2.9.2 Non classifier-related comments

SE: the semantic category was operationalized as general comments about

communicating meaning, creating understanding, or being unable to express an

intended meaning.

Example 4-14
Feedback episode
NNS: IEajL^S o )IUlik o o
Wo zhu zai zher foluolida. Nar hen da, hen
1sg live in here Florida. There very big, very.
I live in Florida, There is very big.
NS: Ro
Hen re.
Very hot.
Very hot.
NNS: fM1J 7 --i7\ 7 o
Suoyi wo mai le yi ge maozi.
So 1sg buy PFV one CL hat.
So I bought one hat.
N S: wT N $11-T t /AP- 0 ?
Ni na ding maozi shi shenme yanse.
2sg that CL hat is what color
What color is your hat?


143









3.4 CHL Learners, Implicit and Explicit Feedback

CHL learners usually use Chinese when communicating with family members,

such as grandparents or parents, who normally do not speak English (Dai & Zhang,

2008). When speaking with siblings and Chinese friends, CHL speakers tend to switch

to English. Some of them even feel embarrassed to speak their HL with people other

than their family members, because they consider Chinese a language only spoken at

home. The shyness toward their HL suggests that their acquisition and maintenance of

their HL is in a "vertical and reciprocal intimate relation" (Dai & Zhang, 2008, p.41), and

is restricted to an informal setting with limited topics of daily life. Moreover, many CHL

learners speak a non-prestige dialect, which in many cases is unintelligible to Mandarin.

Finally, due to their different levels of exposure to the HL, the language abilities of CHL

learners vary from individual to individual, which leads to tremendous variations in their

linguistic knowledge and abilities when beginning Chinese classes at college or

university (Hendryx, 2008). However, as a group, CHL learners show some consistent

patterns: although having developed a considerable level of fluency in their oral speech

skills, they tend to have a narrow range of lexical and syntactic alternatives (Valdes,

2001). In addition, the majority of CHL learners tend to have high-level receptive skills

(especially listening skills) but low-level productive skills (e.g., speaking), which often

resemble other HL learners (Valdes, 2001).

The imbalance between the receptive and productive skills observed in CHL

learners reminds us of learners in Swain's well-known study on Canadian French

immersion programs (1985; 2005). Both learners in the two CHL learner groups and

French immersion program learner groups speak a dominant language, as well as a

second language which they started to learn from a young age. Learners in both groups









Chinese outside the classroom. He also tried to watch some Chinese movies, cartoons,

or TV programs, but he admitted that he could only pick up some words occasionally.

Interlocutor. The interlocutor of the task-based interaction was served by the

researcher, who is a female native Mandarin speaker with 13 years experience teaching

Chinese as a second language. She started to teach undergraduate Mandarin courses

at UF in 2007. In order to avoid sampling bias, all the students whom she was teaching

when the data was collected were excluded from the experiment at the subject

recruitment stage. The same interlocutor elicited the stimulated recall data and

conducted the interview. She was in her late 30s when the data was collected.

4.2.4 Target Linguistic Items

Chinese classifiers. Previous studies have investigated feedback effects in

various linguistic domains, such as syntactical, morphosyntactical, and morphological

forms, in the forms of dative verbs (Carroll & Swain, 1993; Kim & Mathes, 2001), verbal

predicates and particles (Nagata, 1993), indefinite article (Muranoi, 2000), derivations of

nouns from verbs (Carroll, 2001), noun-adjective agreement (Leeman, 2003), pronouns

(Sanz et al., 2009), gender agreement (Lyster, 2004), verb past tense formation (N. Ellis

& Larsen-Freeman, 2006; R. Ellis, 2007), and question formation (Loewen & Nabei,

2007). The target linguistic items in these studies differed in degree of complexity in

terms of learners' processing operation burden. Some structures required simple explicit

knowledge, while others required more complex rules. Thus the choice of linguistic

structures relates to the results of the research, and needs to be made with careful

consideration.

The current study chose Chinese numeral classifiers as the linguistic item,

motivated by theoretical, methodological, and pedagogical grounds. To my knowledge,


116









3.3 Previous Empirical Research Comparing the Effects of Implicit and Explicit
Feedback

Numerous studies have been done in the past comparing of the effects of different

types of corrective feedback on L2 acquisition. Generally, the results of these studies

showed that both implicit and explicit feedback facilitate L2 acquisition. However, in

terms of which type of feedback has more advantages in L2 development, the results

were mixed. Since these studies were conducted in a wide range of settings (e.g.,

laboratory, classroom, or computer-based interactions), and using a variety of measures

(e.g., grammaticality judgment tests, sentence completion, translation tests, etc), and

treatment tasks (e.g., communicative tasks, and mechanical exercises), it is not wise to

rush to a conclusion about which type of feedback has a greater impact on L2

acquisition, without carefully considering the substantial differences among these

studies. For instance, researchers argued that the settings of the interaction affect the

effects of feedback, particularly where recasts are considered. In the classroom context,

recasts are provided in an interactional context (Lyster, 1998b), therefore it may be

more challenging for learners to perceive the corrective intentions of recasts as they

would in a laboratory setting. Lyster (1998b; 2007) found that recasts might not function

as effectively as other types of feedback in classroom settings that are more meaning-

oriented than form-oriented. In laboratory settings, variables can be easily controlled,

allowing for feedback to be delivered in an intensive manner focusing on the target

linguistic forms.

This section will focus on 18 previous studies that compared the effects of implicit

and explicit feedback conducted in three different settings: laboratory (Carroll & Swain,

1993; Kang, 2009; Kim & Mathes, 2001; Leeman, 2003, McDonough, 2005,









certain areas, such as vocabulary, register, etc., they should identify writing skills

development as a priority. A new curriculum that can better satisfy the different learners'

language needs is strongly needed in the field of Chinese language instruction.

Besides the language curriculum, Chinese language instructors also need to

adjust their classroom instruction to maximize the learning effects. The results of the

current study showed that the CHL learners tended to rely on what they have heard in

the past when assessing the accuracy of their language, instead of using rule-based

grammatical knowledge. In addition, they had a natural tendency to focus their attention

more on content rather than linguistic forms when engaging in a conversational

interaction. As claimed in the Noticing Hypothesis, learners' noticing of the gap between

their internal language system and the target language system is the precondition for

converting input to intake (Schmidt, 1990, 1993, 1994, 1995), which can further lead to

a change in their interlanguage. Therefore, one of the most critical goals in Chinese

classroom instruction should be facilitating CHL learners to notice the discrepancies

between their problematic output and the targetlike input provided by the instructor. As

CHL learners normally acquire their HL implicitly (at least partially), they tend to lack

explicit meta- and rule-based knowledge that is needed in order to identify the gap.

Therefore feedback that explicitly indicates learners' errors, such as meta-linguistic

feedback employed in the current study, can be an efficient tool when the instructor tries

to draw CHL learners' attention from meaning to certain linguistic forms, particularly for

CHL learners' stabilized non-target like language.

On the other hand, non-CHL learners develop their Chinese knowledge mainly

through explicit learning in a classroom setting. Therefore they develop explicit


211









4.2.7 Transcriptions

The researcher transcribed all the tests, treatments, stimulated recall, and the

interview data. For the pre- and post-test data (128 tests, approximately 23 hours), only

the classifier used by the learners was transcribed. For the treatment data (94 tasks,

7.45 hours for CHL learners, 11.9 hours for non-CHL learners;), the stimulated recall

data (6.35 hours for CHL, 14.2 hours for non-CHL learners), and the interview data (13

hours), the entire recording was transcribed.

4.2.8 Scoring

On oral tests, two points were awarded for target language (TL) production, and

zero points for non target language (NTL) production. One point was given when a

learner used ge to replace the appropriate classifier. TL production was operationalized

as TL suppliance of the targeted classifier (e.g., jian for chenshan "shirt"). NTL

production was operationalized as NTL suppliance of the classifier (e.g., tiao for

chenshan "shirt").

On written tests, three points were awarded for TL production in Chinese

characters (e.g., T(jian) for chenshan "shirt"), two points for TL production in Hanyu

Pinyin7 (e.g., jian for chenshan "shirt"). One point was awarded for the general classifier

ge, regardless of whether in Chinese characters or in Hanyu Pinyin8. Zero points were

given for NTL production (e.g., tiao for chenshan "shirt").

4.2.9 Coding

The stimulated recall protocols were coded according to perception categories that

were generated based on previous research on learners' perception about the target of

the feedback. Learners' perception of feedback has been coded into six categories in

previous research-lexical, semantic, phonological, morphosyntactic, no content, and


139









but irregular, and not rule-governed. Frequent input consisting of classifiers that are

presented with associated head nouns makes the acquisition of these classifiers easier.

In addition, all the classifiers employed in the current study were relevant to objects or

animals usually seen in daily life. Therefore HL learners have far more opportunities to

have exposure to these classifiers compared with non-HL learners. In other words, the

longer the testing period lasted, the more likelihood HL learners could hear these

classifiers at home when conversing with their family members. In contrast, since non-

HL learners only had access to Mandarin in the classroom, it was unlikely for them to

have equal opportunities to have exposure to the classifiers. In order to control for

possible external input, the current study employed a two-day testing period.

The treatments were both video- and audio-recorded. The video-recording of the

treatments were edited into short clips, each of which contained one feedback episode.

The video clips were used as stimuli during the stimulated recall session. The

stimulated recall was used to elicit the noticing data on the provided feedback from the

47 learners in the experimental groups. During the stimulated recall session, learners

were asked to verbally report their thoughts after watching each video clip which

contained a feedback episode taken from the two treatments. They could ask the

interviewer to pause the video anytime in between when they wished to share any

thoughts at any particular moment during the interaction. (See section 4.2.5.4 for more

details.)

4.2.3 Participants

CHL and non-CHL learners. A total of 93 volunteers initially participated in the

experiment, however 29 of them were excluded from the study and their data will not be

reported here. Among them, four non-HL learners were excluded because they had a


109









Long, M. (2007). Recasts in SLA: The story so far. In M. Long (Ed.), Problems in SLA
(pp. Mahwah, NJ: Erlbaum.
Lu, X. & Li, G. (2008). Motivation and achievement in Chinese language learning: A
comparative analysis. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage
language: Fostering rooted world citizenry (pp. 89-108). Honolulu: University of
Hawai'i Press.
Lynch, A. (2008). The linguistic similarities of Spanish heritage and second language
learners. Foreign Language Annals, 41, 252-281.
Lyster, R. (1998a). Negotiation of form, recasts, and explicit correction in relation to
error types and learner repair in immersion classrooms. Language Learning, 48,
183-218.
Lyster, R. (1998b). Recasts, repetition and ambiguity in L2 classroom discourse.
Studies in Second Language Acquisition, 20, 51-81.
Lyster, R. (2004). Research on form-focused instruction in immersion classrooms:
implications for theory and practice.
Lyster, R. (2007). Learning and teaching languages through content: A counterbalanced
approach. Amsterdam: Benjamins.
Lyster, R. & Izquierdo, J. (2009). Prompts Versus Recasts in Dyadic Interaction.
Lyster, R. & Mori, H. (2006). Interactional Feedback and Instructional Counterbalance.
Studies in Second Language Acquisition, 28, 269-300.
Lyster, R. & Ranta, L. (1997). Corrective feedback and learner uptake: Negotiation of
form in communicative classrooms. Studies in Second Language Acquisition, 19,
37-66.
Mackey, A. (1999). Input, interaction, and second language development: An empirical
study of question formation in ESL. Studies in Second Language Acquisition, 21,
557-587.
Mackey, A. (2006). Feedback, noticing and instructed second language learning.
Applied Linguistics, 27, 405-430.
Mackey, A. & Gass, S. M. (2005). Second language research: Methodology and design.
Mahwah: NJ: Lawrence Erlbaum Associates.
Mackey, A.,Gass, S. M. & McDonough, K. (2000). How do learners perceived
interactional feedback? Studies in Second Language Acquisition, 22, 471-497.
Mackey, A. & Philp, J. (1998). Conversational interaction and second language
development: Recasts, responses, and red herrings. The Modern Language
Journal, 82, 338-356.
Mak, D. (1991). The acquisition of classifiers in Cantonese. Unpublished doctoral
dissertation, University of Reading.
McDonough, C. (2005). The role of prosody and complexity in infants' detection of
repeated speech patterns. Dissertation Abstracts International, 65, 4873-B.
McElree, B.,Jia, G. & Litvak, A. (2000). The time course of conceptual processing in
three bilingual populations. Journal of Memory and Language, 42, 229-254.
McLaughlin, B. (1978). The monitor model: Some methodological considerations.
Language Learning, 28, 309-332.
Modern Language Association. (2006). Enrollments in languages other than English in
United States institutions of higher education. Retrieved November 15, 2009,
from http://www.mla.org/pdf/06enrollmentsurvey_final.pdf.


234









these uncertainties may make recasts more implicit for CHL learners. Second, recasts

are less likely to be perceived as corrective feedback compared with other types of

corrective feedback because of their multiple discourse functions. For CHL learners, as

they have more experience communicating with their family members in HL, they tend

to regard language exercises in the classroom as interactional conversation (Gass &

Lewis, p.99), especially when the correction is embedded in a complete rather than in a

partial reformulation. In addition, their broader implicit knowledge gives them a privilege

over non-CHL learners. That is, they do not have to know every detail in the utterance

before they can comprehend. Therefore they tend to be less sensitive to the alternative

forms provided by NS. Lastly they tend to treat a NS's reformulation as a yes/no

question or as a continuation of the conversation, since they are not pressed to produce

output.

In sum, due to their unique language exposure to Chinese at home, CHL learners

develop an internal knowledge system that differs from non-CHL learners. CHL learners

have developed strong implicit knowledge, but much weaker explicit knowledge.

Consequently, their reaction toward implicit and explicit feedback may differ from non-

CHL learners who have the opposite distribution of implicit and explicit knowledge. Thus

the findings from previous studies on the effects of implicit and explicit feedback, which

mainly focused on FL/SL learners, may not apply to CHL learners. However, to the best

of my knowledge, little empirical research has been done in this area so far. This gap in

the literature motivated the research questions of the current study, which I will present

in Chapter 4. Assuming that HL learners had indeed developed a certain sensibility









targeted in the current study was classifiers, which select their nouns based on different

semantic domains, such as shape, function, animacy, etc. Thus the semantic

relationship between a classifier and the associated noun is intricate and complex (Hu,

1993). In addition, although the target classifiers had already been encountered by non-

CHL learners in the classroom, the extent to which they had been acquired varied.

Some were only briefly introduced as supplementary vocabulary words; some were

thoroughly taught in the text and appeared in the homework and exams. Therefore,

learners might have forgotten those classifiers that were not used, as shown in Example

6-17.

Example 6-17
NNS: *R --, -- -B o
Wo you yi yi ge maozi
Isg have one one CL hat
I have one hat.
NS: Ni you yi ding maozi shi ma
2sg have one CL hat is QP
You have one hat, don't you?
NNS: *--JJ -5T o
yi ding maozi
one CL hat
One hat.
Recall data
Yeah, I just cannot remember the measure word. But I was trying to make
it more conversation. I was trying to respond or ask questions back...
I wish I were more fluent. That is what I was thinking. I felt better the
second day compared with the first day. Because a lot of things we
discussed on the first day were things we had learned, I had studied, but I
don't ever use them. I can never talk about flowers or moon pies, or
anything like that, I never talked about them. So I forget how to say them.
So if you remind me, usually I can remember the majority of them.

Besides the classifiers, non-CHL learners also had to worry about their

pronunciation, particularly the tones. The descriptive statistics results of learners'


190









Xiao, Y. (2006). Heritage learners in the Chinese language classroom: Home
background. Heritage Language Journal, 4, 47-56. Retrieved November 10,
2009, from http://www. international. ucla.edu/media/files/YX. pdf.
Xiao, Y. (2008). Home literacy environment in CHL development. In A. W. He & Y. Xiao
(Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp.
151-166). Honolulu: University of Hawai'i Press.
Yeni-Komshian, G. H.,Flege, J. E. & Liu, S. (2000). Pronunciation proficiency in the first
and second languages of Korean-English bilinguals. Bilingualism: Language and
Cognition, 3, 131-149.
Zentella, A. C. (1997). Growing up bilingual. Maiden, MA: Blackwell Publishers.
Zhang, H. (2007). Numeral classifiers in Mandarin Chinese. Journal of East Asian
Linguistics, 6, 43-59.


238









difficult, which affected their processing of relative clauses. Kim argued that although HL

learners may not have fully acquired the syntactic structures of Korean, they can use

their semantic and contextual knowledge to comprehend sentences, even when they

contain complex structures. In addition, since they acquired their HL implicitly, and lack

explicit meta-linguistic, or rule-based knowledge, grammar-focused instruction could

play an important role in making them notice some features of the target language.

Hence, "in formal language instruction, it may be crucial for teachers to provide HL

learners with consistent interaction with negative feedback that draws attention to the

form, rather than meaning" (p.124).

The study conducted by Gass and Lewis (2007) was the first attempt to approach

the linguistic differences between HL and non-HL learners within an interactionist

framework. Thirteen FL learners and six HL learners of Italian participated in the study.

All the participants took part in a task-based interaction (a spot-the-difference task) with

an interlocutor who was a near-native speaker of Italian. Feedback was given on non-

targetlike utterances during the interaction. Learners' perceptions about the feedback

were measured using stimulated recall methodology. The results showed a between-

group difference of HL and non-HL learners in terms of their perception of provided

corrective feedback. Non-HL learners outperformed HL learners in perceiving

phonological and morphosyntactic feedback; while HL learners perceived semantic and

lexical feedback more correctly than non-HL learners. In addition, HL learners also tend

to incorrectly interpret feedback on other linguistic items as semantic feedback as well.

Based on these results, Gass and Lewis argued that the exposure to HL during

childhood may make HL learners "regard language as a form of real communication"









knocks on the table. The recall prompt signal made "learners more alert to the details of

that speech than they may have been ordinarily" (p.109). Egi also used immediate recall

in her study, but with a different operationalization (2004). Instead of asking learners to

repeat previous content from the recasts, she asked them to verbalize their thoughts

about the language episodes after the recasts. Obviously, higher levels of noticing are

required for learners when being asked to verbalize their thoughts, compared to simply

repeating the limited items that appear in immediately preceding recasts. Although the

immediate recall operationalized in Egi's study might better capture "qualitative aspects

of noticing" (2004, p.248), because of the longer time learners spent on commenting

about the recasts, it might place a heavier burden on their working memories. In

addition, as the knocking recall signals occurred frequently during the interaction, the

flow of normal conversation was consistently interrupted. In fact, the results showed that

the average length of immediate recall protocols was significantly shorter than

stimulated recall. As Egi noted, it may suggest that learners wanted to keep the

interruption to the communicative tasks as small as possible by trying to keep their

comments concise.

In stimulated recall, participants are prompted to recall thoughts they had while

performing a task or participating in an event. It is assumed that some tangible (perhaps

visual or aural) reminder of an event will stimulate recall of the mental processes in

operation during the event itself. Through the use of stimulated recall, "a subject may be

enabled to relive an original situation with great vividness and accuracy if he is

presented with a large number of the cues or stimuli which occurred during the original

situation" (Bloom, cited in Mackey & Gass, 2005, p.17).









During the test, the audio-recording of the test items was played one item at a

time. Each item was played only once. The learners were required to first indicate orally

whether they agreed with, disagreed with, or were unsure about the statement in each

item based on what they heard. Learners were told that they could disagree when they

spotted anything wrong in the sentence (e.g., grammar, choice of vocabulary, etc).

However, regardless of which indication they gave, learners were required to repeat

each sentence. If they disagreed with the sentence, they were asked to repeat the

sentence in correct Chinese, as shown in Example 4-6.

Example 4-6

Wo zuotian qu shangdian mai le san tiao kuaizi.
1sg yesterday go shop buy PFV three CL chopsticks
Yesterday I went to shop and bought three chopsticks.

The learner was expected to disagree with the sentence, since the classifier used in

the sentence was not correct. Upon the completion of the judgment, the learner was

requested to repeat the sentence orally in "correct" Chinese. "The basic assumption

underlying elicited imitations is that if a given sentence is part of one's grammar, it will be

relatively easy to repeat" (Mackey & Gass, 2005, p.55). Classifiers generally have

relatively low communicative value in a Chinese conversation, since they can almost

always be replaced with the general classifier 'ge'; learners would have not focused their

attention on classifiers if they were not already in their lexicon, especially when the

sentence is reasonably long. Other elements in the sentence, such as verbs, nouns, or

adjectives might more likely draw their attention compared with classifiers.

Although no minimal time was set to complete the oral imitation test, participants are

still under time pressure as each of the 33 audio-recorded items was only played once.


128









very limited literacy skills even though they are fluent in the oral language (Li, 2006). As

a result, for CHL learners, the goals for their HL learning are more likely to improve their

literacy skills in order to read and write well. In contrast, for non-CHL learners, they rank

speaking and listening as the most important immediate goals in their Chinese learning

(Wu, 2008).

Cultural identity. CHL learners bring to the classroom not only linguistic skills, but

also their cultural identity, which has been influenced by both Chinese and American

cultures. This is also not shared by their non-CHL counterparts.

Similar to their linguistic skills, their cultural identities also vary from individual to

individual, depending on the home and environment they grew up in. Dai and Zhang

(2008) conducted a survey among 80 college students with CHL background, and their

participants perceived their own cultural identities differently. The majority of the

participants viewed themselves as a combination of both Chinese and American

cultures, and they drew on different cultural knowledge when interacting with different

people, at different times, and in different social contexts. The rest of the participants

either considered themselves "less Chinese than the Chinese, and less American than

the Americans" (p.44), and thus indicated that they belonged to neither Chinese or

American culture, or accepted one culture but rejected the other. G. Jia (2008) found

that HL speakers who identified a stronger connection and preference to their HL

culture tended to not only use more HL, but also self-rate their reading and writing skills

higher.

In a related study, He considers learners' identity as a prime dynamic force rather

than just the background in CHL learning. She argued that through learning CHL, which









of recasts, which set them apart from other types of negative feedback: first, the

corrections in recasts are done in an implicit and incidental manner, with no clear

indicator of the source of errors. Therefore, learners have to infer by themselves that

their utterances are problematic (Carroll & Swain, 1993). Second, the juxtaposition of

the learner's ungrammatical utterance and the teacher's reformulation may enhance the

salience of the target form and promote learner noticing, which provides the learner with

an ideal opportunity to make a cognitive comparison and to notice the gap between the

target-like and non-targetlike forms (Doughty & Pica, 1986; Long et al., 1998). Third,

since corrections are embedded in a context, the flow of communication will not be

interrupted (Long, 1996, 2007). As a result, recasts can serve two functions: a

pragmatic function that keeps the communication channel open, and a corrective

function that indicates that an error was made by the learner. Thus the function of

recasts can often be ambiguous, as it can be used both for providing and seeking

confirmation and for additional information (Lyster, 1998b). In fact, classroom studies

showed that corrective recasts were more likely to be interpreted by learners as non-

corrective repetitions during the discourse in SLA classrooms, as both of these two

types of feedback share the same pragmatic functions (Lyster, 1998a).

Clarification requests can be used for problems in either comprehensibility or

accuracy, and include phrases such as 'pardon me', 'what do you mean by X'? (Lyster

& Ranta, 1997). Similar to recasts, the corrective intention of clarification requests is

also not so easily noticed because of the dual purpose it serves: clarifying meaning as

well as prompting learners to self-repair their problematic utterances. Therefore, it is

also considered more implicit than explicit (Loewen & Nabei, 2007).









test was employed by many researchers to examine the effects of implicit and explicit

feedback in the past (e.g., Ammar & Spada, 2006; Carroll & Swain, Kang, 2009;

Leeman, 2003; R. Ellis et al., 2006; Sheen, 2007). R. Ellis and his colleagues did not

find that significant differences existed between implicit and explicit feedback groups in

the immediate post-test; however, they found them in the delayed post-test (R. Ellis et

al., 2006). Sheen (2007) found that the meta-linguistic feedback group significantly

outperformed the recast group on both immediate and delayed post-tests. Previous

research found that it takes time for learners to incorporate the target structure into their

interlanguage systems, thus the effects of feedback become more evident on in the

delayed test, particularly when the effects were measured by tests that favored implicit

knowledge (Mackey, 1999, also see R. Ellis et al., 2006). Due to the lack of a delayed

post-test, learners' measurements of improvements of target forms in the current study

were limited to only short-term effects, as shown on the immediate post-test, thus

potentially weakening the findings of research questions 2 and 3.

A final shortcoming of the test design was the inconsistency between the oral and

written test in terms of identifying non-targetlike classifiers. On the written test, learners

were required to write in Chinese characters as much as they could, so that when a

learner used a homonym of the target classifier it could be identified. However, the

same type of error could not be identified on the oral test. This limitation particularly

favored the CHL learners on their oral test, as they acquired their HL mainly through

listening to their parents. In many cases, they only acquired the oral form rather than the

meaning of a classifier. For instance, the CHL learners' written test results showed that

they often misused R zhi, which is the classifier for animals, to classify flowers, which


214









Table 5-32. Summary of increases on the written


N Minimum Maximum


M


post-test for recast group
SD


RHL 12 3 33 15.33 9.30
RNHL 12 -3 18 9.17 5.62


Table 5-33. Summary of increases on the written
group


post-test scores for meta-linguistic


Group N Minimum Maximum M SD
MHL 11 4 34 16.45 9.92
MNHL 12 2 36 21.92 11.18

Table 5-34 presents the descriptive statistics of learners in the control groups. The

results show that the increase made by CHL learners in the control group was 4.13

points, which indicated that these learners showed a slight improvement of the target

form by simply taking the tests. In contrast, the increase of non-CHL learners was only

.56 point, which indicated that these learners did not benefit from taking the tests.


RHL MHL


Figure 5-11.

Table 5-34.


RNHL MNHL


Group
Increases on the written post-test across the experimental groups

Summary of increases on the written post-test for control group


Group N Minimum Maximum M SD
CHL 8 -7 12 4.13 6.47
NCHL 9 -6 7 .56 4.16


In summary, the descriptive statistics results show that learners in the meta-

linguistic feedback group generally made higher increase in both oral and written tests

regardless of their language background. Interestingly, whereas CHL learners in the


166


Group


0__









most advanced L2 learners consistently outperformed or performed equally well

compared to the most advanced HL learner.

To summarize the above studies: low-proficiency HL learners, who are either

simultaneous or near-simultaneous bilinguals in their HL and the societal language,

show more similarities with L2 learners (Montrul, 2002). Exposure to the HL usually

begins at birth or at a very early age for these learners, and comes from their family or

from community members through daily life. This exposure is totally input-based, and

thus these HL learners develop an implicit rather than explicit knowledge of the HL.

Although their meta-linguistic knowledge may also emerge, as it does for monolingual

children, there are usually no opportunities to make it fully develop through reading and

writing practice, since they mostly attend English-dominant mainstream schools. With a

few exceptions, most of them only develop literacy skills in English, and remain illiterate

in their HL. Without formal study and literacy skills, HL learners' linguistic knowledge of

their HL is more implicit and intuitive, thus differing from traditional FL and L2 learners,

who routinely acquire and store explicit linguistic knowledge of the target language

through formal classroom instruction (Lynch, 2008).

Although there are so many similarities between low-proficiency HL learners and

L2 learners, the exposure to the HL at an early age has lasting benefits on HL learners,

even though their language experience is limited or incomplete, is interrupted by

attending English-dominant mainstream schools, and undergoes attrition. These lasting

benefits enable them to have advantages over L2 learners in some areas, such as

phonology but not in others, such as morphosyntax (Au et al., 2002).









contrast, ding, the classifier for hat, is less frequently used. Therefore it could be

possible that WKW associated the classifier mian with the words he knew (such as

mianbao) to remember it. In the current study, both the CHL learners and the non-CHL

learners reported that they tend to use this analogous strategy when acquiring an

unknown classifier.

In contrast, the non-HL learners usually relied on the explicit knowledge they

learned in the classroom and feedback to figure out their errors. Example 6-22

illustrates this.

Example 6-22 A non-CHL learner ZK
NNS:* t_ ,A--o
Wo ye mai yi zhang da jingzi
1sg also buy one CL big mirror
I also buy a big mirror.
NS: -M-RR1 -4----"a ?
Yi mian hen da de jingzi ma
One CL very big NOM mirror QP
A very big mirror?
NNS: 3o
Dui
Correct
Correct.
Recall data
At first I thought the measure word would be zhang forjingzi, cause it is flat
surface. But it wasn't, I was kind of surprised for a second, I heard you corrected
me.

In Example 6-23, learner ZK use the rule (zhang is a classifier that modifies an

object with a flat surface) that he learned from the class to make a reasonable

assumption for mirror. When a recast was provided by the interlocutor, he realized that

he had been corrected. His profile (see Chapter 4) showed that he had been taking

Chinese courses at the university for 15 months when the data was collected. He had

never been to China, and had very limited opportunities to practice his Chinese outside


198









5.3 Research Question 3

Research question 3 asked whether feedback type or language background affects

the increase in CHL and non-CHL learners' performance on a) an oral imitation test, b)

an untimed written cloze test.

5.3.1 Descriptive Statistics

5.3.1.1 Oral imitation test

Tables 5-29 and 5-30 present the number of participants, minimum, and maximum

scores, as well as the mean scores of the increase made by learners in all the

experimental groups. The increase was calculated by the score of the post-test minus

that of the pre-test. Figure 5-10 shows a graphical representation of the increase by

learners in the experimental groups. All the learners made an increase from pre-test to

post-test. In the recast group, CHL learners and non-CHL learners made exactly the

same increase (M=15.33). In the meta-linguistic group, there was also little difference

between the two learner groups. In terms of feedback condition, learners in the meta-

linguistic feedback group made a greater increase than those in the recast group by

approximately five points.

Table 5-29. Summary of increase on the oral post-test for recast group
Group N Minimum Maximum M SD
RHL 12 -4 32 15.33 12.38
RNHL 12 -4 35 15.33 12.01

Table 5-30. Summary of increase on the oral post-test for meta-linguistic group
Group N Minimum Maximum M SD
MHL 11 -2 42 20.91 15.31
MNHL 12 2 44 21.00 13.14


164









perceptions showed that non-CHL learners made more phonology-related comments

than CHL learners when they failed to perceive the feedback accurately. Example 6-18

presented such a case of a non-CHL learner.

Example 6-18
NNS: *R A)JLo
Wo you liu huar
1sg have six flower
I have six flowers.
NS: AIt)L ?
Ni You liu zhi huar
2sg have six CL flower
You have six flowers?
NNS: AWt JLo
Liu zhi huar
six CL flower
Six flowers.
Recall data
I just wanted to make sure that I made the correct tone for that one.

As a final explanation of lack of differences between the two feedback conditions

in the current study, the length of treatment in the current study was relatively short: the

total length of the two treatment tasks was approximately 20 minutes for CHL learners,

and 40 minutes for the non-CHL learners. The different length of treatment for the two

learner groups was due to their different fluency: the CHL learners engaged in the

interaction with the NS in a smooth manner. They always responded to the NS very

quickly. In contrast, the non-CHL learners tended to take some time before

comprehending the NS's question. It took them an even longer time to respond to the

NS.

Previous studies that revealed significant differences between different feedback

groups usually spread their treatments over a longer period of time, ranging from one

hour (R. Ellis et al., 2006) to several weeks (Ammar & Spada, 2006; R. Lyster, 2004). In


191










learners in the recast group (M=53.14) scored relatively higher than those in the meta-

linguistic group (M=48.79). Learners in the control group (M=33.89) scored much lower

than those in both two experimental groups. On the other hand, for non-CHL learners,

learners in the recast group (M=36.76) scored relatively lower than learners in the meta-

linguistic group (M=46.72). Learners in the control group (M=17.06) scored much lower

than learners in the two experimental groups. Figure 5-7 graphically present the mean

scores of all the groups.

Table 5-16. Summary of written post-test scores for CHL learner group


Group


N Minimum Maximum


RHL 12 16 88 53.14 18.84
MHL 11 7 72 48.79 19.02
CHL 8 4 67 33.89 24.76


Table 5-17.


Summary of written post-test


scores for


non-CHL learner group


Group N Minimum Maximum M SD
RNHL 12 13 70 36.76 16.76
MNHL 12 11 80 46.72 21.56
CNHL 9 2 29 17.06 8.89


60-

50-

40-
0
CO 30-

20-

10-

0-
RHL MHL CHL RNHL MNHL CNHL
Group

Figure 5-7. Mean scores of written post-test


158











Target
Study participants Design Tests Results
structure


Classroom +
laboratory
study 25
French L2
undergraduate
learners


34 private
language
school ESL
students (low
intermediate).


66 EFL
students at a
Japanese
university
(Average
years of
studying
English: 7)


French
feminine
endings


English past
tense -ed





English
question
formation


Lyster &
Lzquierdo
(2009)


Pre-post-delayed post test
1. object-identification task,
2. picture-description task,
3. binary-choice and
reaction-time measures.


Groups: (A) recast, (B)
prompt (clarification
requests followed by a
repetition).
Classroom treatment: form-
focused instructional unit (3
hr/ day over 2 weeks),
provided for all participants.
Laboratory treatment: three
oral tasks, one-to-one with
an interlocutor.
Groups: (A) recasts, (B)
meta-linguistic information,
(C) control.
Treatment: two story
narrative tasks. Participants
were assigned in a triads
with the researcher.
Groups: (A) Recasts, (B)
Clarification request, (C)
Meta-linguistic feedback,
(D) No feedback, (E)
control.
Treatment: 1. a spot-the-
difference, and 2. a guess-
the-storyline task, one-to-
one with the researcher.


No significant differences
found between groups A
and B.


Group B significantly
outperformed the other two
groups for both the
delayed imitation and
grammaticality judgment
post-tests.

Groups A, B, and C
significantly outperformed
groups D and E only on
timed grammaticality
judgment test, but not on
untimed grammaticality
judgment and oral
production tests.


Design: pre-/ immediate
post-/ delayed post-test.
Tasks:1. oral imitation, 2.
untimed grammaticality
judgment test, 3. meta-
linguistic knowledge test.

Design: pre/ post-test
Tasks: 1. timed
grammaticality judgment
test,2. untimed
grammaticality judgment
test, 3. oral production
task.


Ellis et al.
(2006)





Loewen
& Nabei
(2007)


Table 3-3.


Continued









stimuli in short-term memory" (Mackey et al., 2000, 474). Awareness, in contrast, was

assessed as the ability "to verbalize the rules of the sequence in answer to a question

about a pattern underlying the stimulus presentations" (Robinson, 1995, p.300).

In the SLA literature, various retrospective methodologies have been employed in

collecting noticing data, including diary studies (Schmidt, 2001), questionnaires ), and

uptake sheets (Warden et al., 1995). However, these methods tend to take long spans

of time. As cognitive processing is claimed to occur in a short period of time, more time-

efficient verbal reports have been employed, including think-aloud protocols (e.g., Leow,

2000; Rosa & O'Neill, 1999), immediate recall (Egi, 2004), and stimulated recall

(Adams, 1991; Mackey et al., 2000; Slimani-Rolls, 2005).

The verbal reports can roughly be categorized into two types based on their

currency of verbalization: concurrent or retrospective (Adams, 2003; Egi, 2004). In

previous studies, two types of verbal reporting have been used by many researchers in

collecting noticing data related to recasts: immediate recall (e.g., Egi, 2004; Philp,

2003), and stimulated recall (Adams, 1991; Egi, 2004; 2007; Nabei & Swain, 2002).

Immediate recall is a technique that can be employed to "elicit data immediately

after the completion of the event to be recalled" (Mackey & Gass, 2005, p.85). This

technique has been commonly used within cognitive psychology to access detection

and rehearsal in short-term auditory memory. Philp (2003) employed this technique in

her study to investigate various constraints preventing learners from noticing the gap

between their ill-formed utterances and the target forms, after receiving recasts. In her

study, learners were asked to repeat the last part of the recast they heard during a

conversational turn immediately after a recall prompt signal, which consisted of two









of the targeted classifier. They also tended to skip classifiers (M=26.7%) more often

than the CHL learners (M=10.3%), which is reflected in the category of O.

Table 5-19. CHL learners' previous knowledge of target classifiers
Group R (%) A1 (%) A2 (%) O (%)
RHL(n=155) 9 (5.8) 63 (40.6) 70 (45.2) 13 (8.4)
MHL(n=166) 36(21.7) 24(14.5) 86(51.8) 20(12.0)
Total(n=320) 45 (14.0) 87 (27.2) 156 (48.8) 33 (10.3)
Note: R=(re)acquisition, Al=substitution of ge, A2=wide range of (non-target/incorrect)
classifiers, O=skip a classifier

Table 5-20. Non-CHL learners' previous knowledge of target classifiers
Group R (%) A1 (%) A2 (%) O (%)
RNHL(n=217) 16(7.4) 96 (44.2) 69 (31.8) 36(16.6)
MNHL(n=266) 17(6.4) 55 (20.7) 101 (38.0) 93 (35.00)
Total(n=483) 33 (6.8) 151 (31.3) 170 (35.2) 129 (26.7)


Category
DR
MA1
OA2
Ho


Figure 5-8. CHL learners' previous knowledge of target classifiers


Group
OR
]A1
DA2
Mo









Figure 5-9. Non-CHL learners' previous knowledge of target classifiers


160









R: reacquisition was operationalized as learners' comments that clearly indicate

they knew which measure word should be used before participating in the study, as

shown in Example 4-10.

Example 4-10
Feedback episode
NNS: ] ,J 7--;j o
Qu dongwuyuan wo kan dao le yi ge long.
Go zoo 1sg see RC PFV one CL dragon
I went to the zoo and saw a dragon.
NS: 'i^EAM ?
Na tiao long you duo chang?
That CL dragon have how long
How long was that dragon?
NNS: IR-Ko
Hen chang.
Very long.
Very long.
Recall data

(I knew it should be 'yi tiao long', but I forgot.)

Al: Acquisition 1 was operationalized as learners' comments indicating that they

did not know which measure word should be used before participating in the study, and

used the general measure word 'ge' to replace it, as shown in Example 4-11.

Example 4-11
Feedback episode
NNS: *R 7 o
Wo mai le san ge yaodai.
1sg buy PFV three CL belt.
I bought three belts.
NS: 4tT, 0, ?
Ni de san tiao yaodai shi shenme yanse de?
2sg NOM three CL belt is what color NOM
What color are your three belts?
NNS: --o -o-" dao MEI P iI'M" o
San tiao. Yi ge dao shi hongse de. Hai you heise de.
Three CL. One CL UK is red NOM Also have black NOM


141









effective in the L2 classroom also apply to HL learners? As pointed out by Montrul,

"without proper understanding of how similar or different these two types of learners are,

it is difficult to tell at this point whether the exact same methods applied to L2 learners in

the classroom should also be applied to heritage language learners" (2008, p. 500).

Numerous researchers investigated similarities and differences between HL and

non-HL learners within the framework of UG (Isurin & Ivanova-Sullivan, 2008; Montrul,

2004; Montrul et al., 2008). They consider HL learners as incomplete learners, since the

HL they "grow up speaking (or only hearing)" as an L1 was "replaced by another

language as dominant and primary" (Polinsky, 2008, p.1) at some point in their life,

consequently they usually "fail to develop full linguistic ability in the heritage language"

(Montrul et al., 2008). In addition, the earlier they start to become bilingual in their HL

and the societal language, and "the more intense the exposure to the sociolinguistically

dominant language, the more incomplete the adult grammar may turn out to be"

(Montrul, 2002, p.61). Due to the incomplete or interrupted acquisition of the HL, HL

learners share some linguistic behaviors with L2 learners, such as transferring errors

from the majority language. On the other hand, HL learners generally lack literacy and

formal instruction in the HL, which differs from L2 learners. These UG linguists try to find

out whether learners with HL background have advantages over those without HL

background.

Montrul and her colleagues (2008) investigated the knowledge of Spanish gender

agreement among 69 HL learners and 72 L2 learners, compared with the baseline data

of 22 native speakers of Spanish. The proficiency of the HL learners was carefully

controlled. Only learners who met the following three criteria were chosen: they were









Northern dialects as its base dialects" (Chen, 1999, p. 24). It is called putonghua HMfiB

"common speech (language)" in mainland China and Hong Kong, guoyu Hiff"national

language" in Taiwan, and huayu fif "Chinese" in Singapore. It is also the widely taught

variety in Chinese language classrooms (He, 2008a). Henceforth, the term Mandarin

only refers to the standard dialect of Chinese in this dissertation.

In mainland China, the government has played an active role in promoting

Mandarin, or putonghua _ii, since the 1950s. According to survey results released

by the National Language Committee (2005) 53% of the Chinese population in mainland

China speaks Mandarin. The percentage of the Mandarin-speaking population varies

dramatically with age and education. Generally speaking, the younger the age, and the

higher the education, the higher the percentage of Mandarin speakers within the

population. For instance, while 40.59% of people from 45-59 years old speak Mandarin,

the percentage jumps to 70.12 % when the age of the population decreases to between

15 and 29 years old. In terms of education, while 25.49 % of people with elementary-

school education speak Mandarin, 86.77% of people with more than two-years of

college education are Mandarin speakers.

Compared with mainland China, Taiwan was more strict in promoting Mandarin, or

guoyu Hif in the 1980s, at the expense of other Chinese dialects and aboriginal

languages. All dialects other than guoyu Hi were banned in schools and mass media.

As a result, Taiwan, which regarded guoyu Hi, as "the High language and the local

Chinese dialects and aboriginal languages as the low languages", was more

"characterized by bilingualism and diglossia" than mainland China (Chen, 1999, p. 60).

In 1991, about 90% of the population of Taiwan spoke guoyu Hi which was much









Harred, J. F.,Knight, A. R. & Mclntyre, J. S. (1972) Dow Chemical Company, USA.
Havranek, G. & Cesnik, H. (2001). Factors affecting the success of corrective feedback.
In S. Foster-Cohen & A. Nizegorodzew (Eds.), EUROSLA Yearbook (pp.
Amsterdam: Benjamins.
He, A. W. (2004). Identity construction in Chinese heritage language classes.
Pragmatics, 14, 199-216.
He, A. W. (2006). Toward an identity theory of the development of Chinese as a
heritage language. Heritage Language Journal, 4, 1-28. Retrieved October 20,
2009, from http://www.international.ucla.edu/media/files/AH.pdf.
He, A. W. (2008a). Chinese as a heritage language: An introduction. In A. W. He & Y.
Xiao (Eds.), Chinese as a heritage language: Fostering rooted world citizenry
(pp. 1-12). Honolulu: University of Hawai'i Press.
He, A. W. (2008b). An identity-based model for the development of Chinese as a
heritage language. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage
language: Fostering rooted world citizenry (pp. 1-12). Honolulu: University of
Hawai'i Press.
Heigham, J. & Sakui, K. (2009). Ethnography. In J. Heigham & R. Croker (Eds.),
Qualitative research in applied linguistics: A practical introduction (pp. 91-111).
Hampshire: Palgrave Macmillan.
Hendrys, J. D. (2008). The Chinese heritage language learners' existing linguistic
knowledge and abilities. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage
language: Fostering rooted world citizenry (pp. 53-66). Honolulu: University of
Hawai'i Press.
Hendryx, J. D. (2008). The Chinese heritage language learners' existing linguistic
knowledge and abilities. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage
language: Fostering rooted world citizenry (pp. 53-66). Honolulu: University of
Hawai'i Press.
Hinton, L. (1999). Involuntary language loss among immigrants: Asian-American
linguistic autobiographies. ERIC Digest, Retrieved March 18, 2010, from
http://www.cal.org/resources/digest/digest_pdfs/9910-hinton-langloss.pdf.
Hu, Q. (1993). The acquisition of Chinese classifiers by young Mandarin-speaking
children. Unpublished doctoral dissertation, Boston University.
Huang, s. (1993). Yuyan, shehui yu zuqun yishi. [Language, society, and ethnicity].
Taipei: Crane.
Isurin, L. & Ivanova-Sullivan, T. (2008). Lost in between: The case of Russian heritage
speakers. Heritage Language Journal, 6, 72-103. Retrieved November 4, 2009,
from http://www.international.ucla.edu/media/files/l+l-S_HLJ_6_1.pdf.
Iwashita, N. & Liem, I. (2005). Factors Affecting Second Language Achievement in
Primary School. Australian Review of Applied Linguistics, 25, 1-21.
Jia, G. (2008). Heritage language development, maintenance, and attrition among
recent Chinese immigrants in New York City. In A. W. He & Y. Xiao (Eds.),
Chinese as a heritage language: Fostering rooted world citizenry (pp. 189-203).
Honolulu: University of Hawai'i Press.
Jia, L. (2009). Contrasting models in literacy practice among heritage language learners
of Mandarin. Journal of Asian Pacific Communication, 19, 56-75.


231









LIST OF FIGURES

Figure page

2-1 A continuum of heritage language learners' proficiency ......................... 22

3-1 The explicitness of the feedback .............................................. 61

4-1 The research design .............................. ........................ .............. 108

4-2 Experimental procedure and approximate time ......................................... 138

4-3 Coding scheme of the recall data ................................................ 140

5-1 Feedback episodes provided for Chinese heritage language group................. 149

5-2 Feedback episodes provided for non-Chinese heritage language group......... 149

5-3 Percentage of classifier-related comments across experimental groups.......... 151

5-4 Mean scores of oral pre-test .................. .............. ... ............... 155

5-5 Mean scores of oral post-test .......................... ......... ................ 156

5-6 Mean scores of written pre-test .............................................. 157

5-7 Mean scores of written post-test.............................. ............... 158

5-8 Chinese heritage language learners' previous knowledge of target classifiers. 160

5-9 Non-Chinese heritage language learners' previous knowledge of target
classifiers ............... .... .......... ......... .................. ........... 160

5-10 Increase on the oral post-test across the experimental groups ..................... 165

5-11 Increases on the written post-test across the experimental groups.................. 166









BIOGRAPHICAL SKETCH

Ye Han was born in Beijing, China. She received her B.A. in Teaching Chinese as

a Second Language from Beijing Language and Culture University in China. She

received her M.A. in Journalism from Sophia University in Tokyo Japan. She received

her Ph. D in Linguistics from the University of Florida in the summer of 2010.


239









required a smooth communication flow with as little interruption as possible. Particularly

during the story-telling task, if a learner is frequently interrupted by prompt signals, he/

she may lose interest in carrying on the story. Second, in previous studies, the average

length of immediate recall protocols was significantly shorter than stimulated recalls

since the immediate recall was carried out under the pressure of continuing the

interaction (Egi, 2004). As learners' perception of feedback is one of the dependent

variables investigated in the current study, the more detailed information provided by

the learners, the more their perception of this feedback could be explored.

The procedure of the stimulated recall in the current study followed Gass and

Mackey (2000). A typical example of stimulated recall is shown below:

Example 4-9
N: Ok, Let's look at the video we took from yesterday.
(Played the video.)
L: 3 7 -- -o
Wo mai le yi jian maozi.
1sg buy PFV one CL hat.
I bought one hat.
N: ,, 3 7 -I- i--7-?
Ni mai le yi ding maozi.
2sg buy PFV one CL hat.
You bought a hat?
L: 3 7 7 -.
Wo mai le yi ding maozi
1sg buy PFV one CL hat.
I bought one hat.
N: (Paused the video.)
What were you thinking at that time?
L: I didn't really know how to say one hat. I was not sure which measure
word I should use for "hat". And she (the NS) corrected me.

In the above example, the researcher first let the learners watch a video clip taken

from the treatment in which the learner received the feedback. Then she paused the

video, and asked the learner to verbally report what he had in mind when the feedback


134









learners' language experience and home environment in order to help the researcher to

prepare the interview guide (see Section 4.2.5.5). The questionnaire consisted of 14

items, 10 of which were closed-item questions that related to the most fundamental

information about the learners' language experience, such as learners' birthplace, L1,

home language, etc.

Table 4-1. The targeted classifiers.
Semantic Stimuli
Domain Classifier Meaning Stimuli Nouns nouns
Domain nouns

[] zhi ,i
zh inhuman chicken, cat
Animacy E pi inhuman horse
inhuman dragon, snake
-s tiao j 6
.X shuang paired-item chopsticks, shoe,
single-item gloves, shoe
Arrangement R zhi hi


tB ba hand-tool )J,
upper-body garments toothbrush,
t jian hat scissors
unn ding supported objects shirt, sweater
Function vehicles hat
A.jia jia
Sjia piano
$ liang car

tiao long-flexible tie, belt
drooping flowers flower
duo long, hard, stick-like I pen, flower
,1 zhi flat-thin picture, desk
flat-thin mirror
zhang round-small ii, -- pearl, grape
9 mian block-like beef, mooncake
Shape flat and thin leaf, bread
S ke needle needle
A- kuai A, 0t
Span "j -,
& gen -+

General Ge General general


121










models of the target form, therefore learners will not be pushed to produce output as

they do with prompt.



Prompt




o



Metalinguistic Elicitation Repetition Clarification
S Feedback request
o


Explicit Implicit

Explicit
Correction Recast






Provide

Figure 3-1. The explicitness of the feedback (Loewen & Nabei, 2007, p.362. Figure
15.1)

The subtypes of feedback under prompts and provide are ordered according to

explicitness, as shown with horizontal double sided arrows in Figure 3-1. Here,

explicitness is defined in terms of the identification of the error that triggered feedback.

Meta-linguistic feedback is considered the most explicit form of prompt-type feedback

because it explicitly identifies the source of the problem that led the NS to provide

feedback. In contrast, clarification requests are on the implicit end of the continuum

because learners have to infer that "the form of their utterance is responsible for the

interlocutor's comprehension problems" (Carroll & Swain, 1993, p.361). Under the term









linguistic form (Chinese classifiers) in the current study could be both semantic and

lexical (Zhang, 2007), it was hypothesized that CHL learners would perceive both

explicit and implicit feedback on Chinese classifiers more accurately than non-CHL

learners.

RQ 2: Will explicit feedback work more effectively than implicit feedback for CHL

and non-CHL learners' language development?

H 2: Learners' language proficiency level may impact the effects of different types

of feedback. Ammar and Spada (2006) found that prompts-which included meta-

linguistic information, repetition, and elicitation--worked more effectively than recasts for

learners with lower proficiency, but not for learners with higher proficiency. As CHL

learners usually have higher proficiency than their non-CHL counterparts, particularly in

speaking and listening skills, it was hypothesized that explicit feedback would prove

more effective than implicit feedback in non-CHL learners' L2 development, but not in

CHL learners' HL development.

RQ 3: Does feedback type or language background affect the increase in CHL and

non-CHL learners' performance on a) an oral imitation test, b) an untimed written cloze

test?

H 3: The two tests employed in the current study were designed to measure

different types of knowledge. On the oral imitation test, learners are more likely need to

tap their implicit knowledge to perform their oral production under time pressure. On the

other hand, on the written cloze test, learners have time to access their explicit

knowledge since the test is untimed. In addition, their performance is determined by


102









As pointed out by Gass and Mackey, with regard to reliability, previous studies

showed that verbalization affects cognitive processes "only if the instructions require

verbalization of information that would not otherwise be attended to" (2000, p.109).

Therefore verbal reports that are "elicited with care and interpreted with full

understanding of the circumstances under which they were obtained, are a valuable and

thoroughly reliable sort of information about cognitive processes" (2000, p.109).

In addition to the validity and reliability of the instrument, several things could be

done to improve the validity of these studies. First of all, external validity could be

increased by recruiting a larger sample size. Second, internal validity could be

increased by controlling the proficiency level of the subjects through a proficiency test,

such as an Oral Proficiency Interview (OPI) before the data collection. As in Mackey et

al.'s study, the number of years of prior study of English varied from as short as three

years, to as long as 14 years; it is therefore hard to tell whether the results of the study

would be valid for learners with various proficiency levels, or only for learners with

relatively high proficiency levels given that half of the participants in the ESL group had

been learning English for over ten years. In Gass and Lewis's study, the heritage

learners' proficiency level was based solely on their course enrollment, which is also

problematic as I argue in Chapter 2.

Both stimulated recall and immediate recall are important methods in eliciting

noticing data on recasts and other corrective feedback. The advantages and

disadvantages of the two methods were closely examined by Egi in her experimental

study (2004). She argued that immediate recall has an advantage over stimulated recall

in being "free from a memory decay problem" as it elicits noticing data when information









episode occurred. Upon the completion of the learner's recall, the researcher started to

play the next video clip.

4.2.5.8 Interview

As discussed in Chapter 2, HL learners' linguistic proficiency should not only be

assessed by their visible language performance, but also by their underlying language

competence, which developed over time through their language exposure (Hendryx,

2008). Thus, we should not rely solely on traditional tests to assess a HL learner's

proficiency level. In fact, the more we know about a learner's past language experience

and home language environment, the closer we come to describing his/her real

linguistic knowledge and abilities (Jiang, 2008). The ideal approach to fully portray a HL

learner's existing linguistic system is to record every step of his/her language

development through longitudinal research. Obviously, it was not realistic to do so when

many participants were investigated, as in the current study. Thus the current study

employed a semi-structured interview. There are several advantages to using this data

collection method: first, the researcher could collect needed information by asking pre-

prepared, well-organized questions. Secondly, the researcher could ask in-depth or

unstructured questions when necessary, with sufficient flexibility. Thus, the researcher

could collect all the intended information and get a big picture of all the HL participants,

while simultaneously collecting specific information concerning each individual learner

(Heigham & Sakui, 2009).

Previous studies have shown that HL learners' proficiency can be predicted by a

variety of social variables, such as age of arrival, cultural identity, socioeconomic status,

language environment, etc. (Jia, 2008). Moreover, motivation was also found to play an


135









born and schooled in the U.S.; had no schooling experience in their home country; and

became bilingual in Spanish and English before age five. All the participants completed

one oral picture description task and two written tests. The results showed that HL

learners performed more like native speakers than the L2 learners on the oral test,

which required more implicit knowledge; however, they were less accurate on the two

written tests, which required more explicit knowledge. Moreover, like the L2 learners,

the errors made by the HL learners were also systematic. Based on these results,

Montrul and her colleagues argued that HL learners have "fast, implicit, and

automatically processed knowledge (typically acquired early in childhood)," which

developed during a critical period; but lack "meta-linguistic knowledge," as they usually

had no or very limited formal education in their HL. Therefore their knowledge of gender

in Spanish "might be stored, represented, and deployed differently" from L2 learners (p.

541).

Romanova (2008) studied the morphological processing of 50 real verbs and 30

nonce verbs by 14 heritage speakers, 17 beginning L2 learners, and 28 native speakers

of Russian. Of note, seven heritage speakers in her study had a few years of

elementary schooling in Russia before moving to the U.S. The results showed that the

heritage speakers performed closer to native speakers, and greatly outperformed L2

learners in their recognition of real verbs, especially real verbs of high-frequency.

However, in terms of nonce verbs, it was the L2 learners who performed more like

native speakers. Romanova attributed the differences to the limited linguistic input

received by the heritage speakers in her study: "it seems that because of interrupted or

incomplete first language acquisition, the weights of rules are not set and rules are









APPENDIX B
CONSENT FORM

I.R.B No.: 2009-U-0172

Informed Consent
Protocol Title: The effects of implicit and explicit feedback
Please read this consent document carefully before you decide to
participate in this study.

Purpose of the research study: The purpose of this study is to examine the effects of
feedback techniques on heritage Chinese language learners.

What you will be asked to do in the study: You will participate in three-days of
activities: Day 1: You will be interviewed by the researcher about your Chinese heritage
language learning experience at home (approximately 30-40 minutes). The interview will
be audio-taped. You will also complete a set of written and oral questions (five minutes
for each) in Chinese, the purpose of which is to measure your current proficiency level
in Chinese; Day 2: You will be asked to tell a story based on provided pictures. In
between, the researcher may ask you some questions related to your story
(approximately 40 minutes). The process will be both video and audio taped. You will
also complete a set of written and oral questions in Chinese in order to confirm your
proficiency level in Chinese. Day 3: You will be shown some video clips chosen from the
interview of Day 1, and will be asked to give some comments about them
(approximately one hour). Your participation in the study has no impact on your school
grade, and your results of written and oral questions will not be communicated to your
instructors. The process will be audio taped. All the video and audio recordings will only
be used for data analysis purposes.

Time required: 3-4 hours

Risks and Benefits: You will be practicing Chinese with a native speaker by
participating in the study. No risks will be involved in the study.

Compensation: No compensation will be given for participating in this research.

Confidentiality: Your identity will be kept confidential to the extent provided by law.
Your information will be assigned a code number. The list connecting your name to this
number will be kept in a locked file in my faculty supervisor's office. When the study is
completed and the data have been analyzed, the list will be destroyed. Your name will
not be used in any report.

Voluntary participation: Your participation in this study is completely voluntary. There
is no penalty for not participating.

Right to withdraw from the study: You have the right to withdraw from the study at
anytime without consequence.


218









THE EFFECT OF IMPLICIT AND EXPLICIT FEEDBACK: A STUDY ON THE
ACQUISITION OF MANDARIN CLASSIFIERS BY CHINESE HERITAGE AND NON-
HERITAGE LANGUAGE LEARNERS



















By

YE HAN















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA

2010









Jiang, S. (2008). Developing a Chinese placement instrument in response to the
diversity in a student population at the university level. In T. Hudson & C. Martyn
(Eds.), Case studies in foreign language placement: Practices and possibilities
(pp. 119-131). Honolulu: University of Hawai'i Press.
Kang, H.-S. (2009). The relative efficacy of explicit and implicit feedback in the learning
of a less-commonly-taught foreign language. IRAL, 47, 304-324.
Ke, C. (1998). Effects of language background on the learning of Chinese Characters
among foreign language students. Foreign Language Annals, 31, 91-100.
Killingley, S.-y. (1981). The Semantic grouping of mensural classifiers in Cantonese.
Anthropological Linguistics, 23, 383-435.
Kim, H.-S. (2008). Heritage and nonheritage learners of Korean: Sentence processing
differences and their pedagogical implications. In K.-B. Kimi & B. James Dean
(Eds.), Teaching Chinese, Japanese, and Korean heritage language students
(pp. 99-134). New York: Lawrence Erlbaum Associates.
Kim, H.-S. H. (2006). Processing strategies and transfer of heritage and non-heritage
learners of Korean. Unpublished doctoral dissertation, The University of Hawai'i.
Kim, H. & Mathes, G. (2001). Explicit vs. implicit corrective feedback. The Korea TESOL
Journal, 4, 1-51.
Koda, K.,Lv, C. & Zhang, Y. (2008a). Effects of print input on morphological awareness
among Chinese. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language:
Fostering rooted world citizenry (pp. 125-135). Honolulu: University of Hawai'i
Press.
Koda, K.,Zhang, Y. & Yang, C.-L. (2008b). Literacy development in Chinese as a
heritage language. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage
language: Fostering rooted world citizenry (pp. 125-135). Honolulu: University of
Hawai'i Press.
Kondo-Brown, K. (2005). Differences in language skills: Heritage language learner
subgroups and foreign language learners The Modern Language Journal, 89,
563-581.
Kondo, K. (1997). Social-psychological factors affecting language maintenance:
Interviews with Shin Nisei University students. Linguistics and Education, 9, 369-
408.
Krashen, S. (1981). Second language acquisition and second language learning.
London: Pergamon.
Krashen, S. (1982). Principles and practice in second language acquisition. New York:
Pergamon.
Krashen, S. (1985). The input hypothesis: Issues and implications. Londaon: Longman.
Lambert, W. E. & Taylor, D. (1996). Language in the lives of Ethnic minorities: Cuban
American families in Miami. Applied Linguistics, 17, 477-500.
Lee, J. S. (2002). The Korean language in America: The role of cultural identity in
heritage language learning. Language, Culture, and Curriculum, 15, 117-133.
Leeman, J. (2003). Recasts and second language development: Beyond negative
evidence. Studies in Second Language Acquisition, 25, 37-63.
Leeman, J. & Martinez, G. (2007). From identity to commodity: Ideologies of Spanish in
heritage language textbooks. Critical Inquiry in Language Studies, 4, 35-65.


232









during the stimulated recall, and their acquisition of classifiers were assessed by their

test scores. The data were both quantitatively and qualitatively analyzed. The results

showed that both feedback types were effective in facilitating learners' acquisition. In

addition, learners' language background was also found to affect their perceptions of

feedback, as well as their acquisition. The findings of this study expand our knowledge

about implicit and explicit feedback. In addition, they also provide invaluable information

particularly for educators and administrators who are involved in HL instruction.









toward recasts, would they form a new perception toward recasts in the classroom after

observing the teacher for a certain period of time?


100









were found to acquire shape classifiers earlier than function classifiers (Loke & Harrison,

1986).

After the above brief sketch of Chinese classifiers, I will now discuss the

theoretical consideration for choosing Mandarin qualifying classifiers as the target

linguistic structure in the current study. First of all, although a conventional association

exists between a qualifying classifier and a fixed set of head nouns, its communicative

value is quite limited, as the general classifier ge could almost always be used to

replace the specific classifier within a context. Doing so usually will not jeopardize the

comprehension of the utterance. Secondly, unlike gender agreements in the Romance

languages, no morphophonolocial inflection occurs between the classifiers and the head

noun. The classifier is a single, discrete item to be learned, with no further nomifications

on the sentence. Thus the feedback given in the treatments could provide learners with

a context in which a classifier should be applied, and enable them to (re)acquire it within

a relatively short period of time.

From a methodological aspect, classifiers are relatively easy to elicit in meaning-

focused interaction. As a classifier always associates with a head noun, in the two

communication tasks in the current study, head nouns were used as stimuli to elicit the

target classifiers.

Lastly, on pedagogical grounds, previous studies showed that Chinese L1

children usually do not complete their acquisition of classifiers until preschool or early

school years (Chang, 1983; Erbaugh, 1982; Liu, 2008; Lu & Li, 2008; Mak, 1991).

Classifiers start to emerge in the children's utterances as early as 1; 07 (Szeto, 1998).

By three years old, they acquire the general classifier ge 4, but tend to use it to replace


119









sent by his parents when he was young. Both examples indicated that he acquired

classifiers and their associated nouns as a chunk. In other words, his learning

experience of classifiers was overwhelmingly input-based. As a result, his knowledge

was unbalanced. For instance, he knew the target classifier for horse, which was

considered relatively difficult by the non-CHL learners in the current study. However, he

did not know the less-difficult classifier for belt, as he did not wear belts often in daily life,

and therefore did not have the opportunity to acquire the classifier for it.

The descriptive statistics results in tables 5-19 and 5-20 provided further evidence

on this issue: the results showed that the CHL learners were more likely to use a non-

target classifier that was supposed to associate with other nouns than the non-CHL

learners (48.8% vs. 35.2%), which suggested that the CHL learners had heard a wider

range of classifiers than the non-CHL learners. During the interaction, they mainly relied

on these personal experiences to figure out whether a classifier was correct or not.

Sometimes their experience was not accurate. Example 6-6 illustrates this.

Example 6-6
NNS: ,-i -o
Haiyou liang zhang jingzi
Also two CL mirror
(there are) also two mirrors.
NS: "R#"1,fr-]--EUT iM -lT JMA-l-o
Zhang women yiban shi shuo pingmian de dongxi
CL 1pl usually is say flat surface NOM thing
Zhang, we usually (use it to) call something with a flat surface

Danshi shuo jingzi women bu shuo zhang
But say mirror 1pl no say CL
zhang, we usually use it to describe an object with a flat surface,
But we don't use it for a mirror.
NNS: -TY?
Jingzi


177









simply expand to studies conducted in classroom settings, where learners are exposed

to mixed types of feedback on various linguistic forms (Lyster & Ranta, 1997; Sheen,

2006). Another limitation that might weaken the generalizability of the current findings

concerns the participants of the study. Due to the difficulties of recruiting adequate

numbers of participants, the current study only included CHL learners who grew up in a

Mandarin-speaking family. Learners who grew up in a home where a Chinese dialect

other than Mandarin was spoken (e.g., Cantonese, Shanghainese, etc), were excluded

from the current study. It is not clear whether CHL learners with different dialect

backgrounds would benefit from implicit and explicit feedback differently. In addition, the

sample size of the current study (N = 64) was relatively small, due to the difficulty of

recruiting a large number of participants at the intermediate level and above for a less

commonly taught language. Lastly, due to the small participant population, an

experimental group that engages only in an interaction without receiving any type of

feedback was not included in the current study. Without a comparison with this

interaction-only group, it was unclear whether the improvements learners made from the

pre-test to the post-test was entirely due to the feedback, or partially due to interaction

only. The CHL learners are particularly in question concerning this issue given the fact

that scores of the CHL learners in the control group increased considerably from the

pre-test to the post-test, particularly on their oral test. Further studies should include an

additional control group to allow for further comparison.

Secondly, in terms of the design shortcoming, the current study did not employ a

delayed post-test due to the difficulty of recruiting adequate subjects who were willing to

commit one more day several weeks apart from the three-day activity. A delayed post-


213









control group made improvements in their increase in both oral and written tests, non-

CHL learners made little or even regressed improvement.

5.3.2 Inferential Statistics

In order to answer the question of whether different feedback types led to an

increase in different types of knowledge, a 2 X 2 full-factorial analysis of variance

(ANOVA) was conducted to determine the effects of feedback type and learners'

language background on their increase. The use of a factorial ANOVA design in the

analysis of between-group differences was motivated by the following fact: in the current

study, there are two independent variables that split the participants into separate

groups: the feedback type and the CHL background of the learners. Therefore we can

also call it a two-way between-groups design.

I will first report the results for the oral test. The main effect of feedback type,

ignoring the CHL background variable, is not statistical (F1,43 = 2.123, p = .152, partial

eta-squared= .047 power= .297). It indicates that learners who received meta-linguistic

feedback did not increase a significantly higher score on the oral post-test than those

who received recasts during the treatment. Similarly, it was found that learners' CHL

background did not produce a statistically significant effect for the main effect of CHL

background (F1,43 = .000, p = .991, partial eta-squared= .000 power= .050), indicating

that overall, learners made similar increase regardless of their language background.

The interaction between feedback type and CHL background was not found to be

statistically significant either.

Next I will report the results of the written test. The main effect of feedback type,

ignoring CHL background variable, is statistically significant (F1,43 = 6.888, p = .012,

partial eta-squared= .138 power= .728). It indicates that learners who received meta-


167









5-31 Summary of increase on the oral post-test for control group.......................... 165

5-32 Summary of increases on the written post-test for recast group..................... 166

5-33 Summary of increases on the written post-test scores for meta-linguistic
g ro u p ...................... .. .. ......... .. .. ......... ...................................... 1 6 6

5-34 Summary of increases on the written post-test for control group.................... 166

6-1 Frequency of uptake and perception of learner XSB............................. 174

6-2 Learners who increased zero or decreased in the experimental groups. ......... 199

6-3 The profile of learners TW and SBY............................................... 202

6-4 Learners' knowledge of Chinese characters on the pre- and post-test............. 206









Regardless of the language background, when learners failed to make classifier-

related comments, 85% of their comments were either unrelated to the feedback

episode at all or related to lexis (see tables 5-7 and 5-8).

The descriptive statistics showed that the number of non classifier-related

comments given by the CHL learners in the recast group (n=62) was approximately

three times more than by CHL learners in the meta-linguistic group (n=20). A similar

pattern was also seen with the non-CHL learners (recast: 162; meta-linguistic: 58). The

gap between the two feedback groups indicated that learners who received recasts

were undoubtedly less aware of the classifier-related feedback. In addition, the non-

CHL learners were more likely to make no content recall comments compared with the

CHL learners. Learners' no content comments also showed that the CHL learners were

more focused on telling a story, whereas the attention of the non-CHL learners was

more easily captured by things that seemed unusual to them in the interaction.

Examples 6-8 and 6-9 illustrate this.

Example 6-8 A CHL learners' no content comment

NNS: *R3--i -To
Wo mai yi ge maozi
1sg buy one CL hat
I bought one hat.
NS: ON11t)' Im t /AP ?
Na ding maozi shi shenmeyanse de
That CL hat is what color NOM
What color is that hat?
NNS: R-leo
Baise de
White NOM.
Recall data
Yeah, I was just describing the picture.

Example 6-9 A non-CHL learners' no content comment


180









feedback condition. This study employed three tests: an object-identification task, a

picture-description task, an oral imitation test, a computerized binary-choice and

reaction-time test. The results showed no significant difference between recast and

prompt groups for any of the measures.

The following four empirical studies attempted to address the effectiveness of

computer-assisted-language-learning (CALL), particularly where the question of

providing feedback is concerned. Researchers argued that feedback provided through

CALL is more effective since it can be provided immediately, thus it can draw learners'

attention to problems in their utterance (Nagata, 1993).

Nagata (1993) compared two versions of Japanese computer-assisted-language

instruction (CALl) exercises: traditional CALl (T-CALI), and intelligent CALl (I-CALl). T-

CALl only provided learners feedback about what was wrong, such as "GA is not

expected to be used here". On the other hand, I-CALI not only pointed out what was

wrong, but also provided detailed meta-linguistic explanations on the reason for the

errors. Nagata found significant main differences between the two experimental groups

in their achievement and retention for the particle errors, but not for verbal errors.

Rosa and Leow (2004) examined the role of awareness in L2 development. In their

study, explicit feedback was operationalized as a meta-linguistic explanation on the

linguistic form selected by learners. When the selection was correct, in addition to the

provision of the reason why the choice was correct, learners also received an

explanation of how the targeted structure works in target language. On the other hand,

when the learner made an incorrect choice, the reason for the error was provided,

followed by an explanation of how the targeted structure should work. Then the learner









are able to access an abundance of comprehensible input either in the immersion

classroom or at home. Like the CHL learners, the immersion learners were also found

to be weak in their productive skills (speaking and writing), although they were deemed

native-like in other language areas. Swain concluded that what the immersion learners

lacked was being pushed to produce modified output. She argued that modified output

is necessary for learners' acquisition because:

"while attempting to produce the target language vocally or silently (sub-
vocally), learners may notice that they do not know how to say (or write)
precisely the meaning they wish to convey... Under some circumstances,
the activity of producing the target language may prompt second language
learners to recognize consciously some of their linguistic problems: It may
bring their attention to something they need to discover about their second
language" (Swain, 2005, p. 474).

Similar to learners in Swain's study, CHL learners also may not have enough

opportunities to be pushed to produce modified output. Although many of them grow up

in a Chinese-speaking family, the conversation between parents and children is often a

mixing of English and Chinese. Usually the parents speak in Chinese, but the children

respond in English. For these learners, comprehending input is easier as they can

understand its meaning without precise knowledge of morphology and syntax. On the

other hand, in order to produce output, they need to focus their attention on the forms in

their inter-language and tap their language competence. When providing corrective

feedback to CHL learners, the following questions need to be considered: which type of

feedback is most likely to help learners notice the corrective intention? Which type of

feedback can best push CHL learners to make modifications and produce modified

output?




Full Text

PAGE 1

1 THE EFFECT OF IMPLICIT AND EXPLICIT FEEDBACK: A STUDY O N THE ACQUISITION OF MANDARIN CLASSIFIERS BY CHINESE HERITAGE AND NON HERITAGE LANGUAGE LEARNERS By YE HAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNI VERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Ye Han

PAGE 3

3 To Jiuru Han and Yulan Zhou

PAGE 4

4 ACKNOWLEDGMENTS First of all, I own my deepest gratitude to my diss ertation committee advisor Theresa Antes for her encouragement, supervision and support throughout the time it took me to complete the project. I thank t he members of my dissertation committee, Diana Boxer, Brent Henderson, and Ester De Jong, who have gen erously contributed their time and expertise to improve my work. I would also like to thank Takako Egi. It would not be possible to complete the project without her inspiration and guidance at the preliminary stage of my dissertation project. I am gratef ul for the faculty and teaching assistants in the Chinese program, Cynthia Chennault, Cynthia Shen, Elinore Fresh, Han Xu, I Chun Peir, Richard Wang, Shizhong Deng, Binmei Liu, Jimmy Huang, Rui Cao, Ya Gang Fan, and Yeng Jung Lee. I also thank all the stud ents who participated in my research. I would also like to acknowledge Janet Rose and Judy Shoaf for their help in arranging the data collection location and equipment. I would also like to thank all my fellow graduate school friends, for their encourageme nt and support. My thanks must go also to my research assistant, Jingya Zhong, who spent many hours assisting in data coding. I particularly want to thank my parents, who have supported me in every respect during my five years in graduate school. I thank my father, who inspired me to develop a life long passion toward books and knowledge by consistent perusing higher levels of achievement in his own field through out his entire career. I thank my mother, who inspired me to become a brave and strong person who is not afraid of challenges and obstacles in life.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 11 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 14 2 H ERITAGE L ANGUAGE LEARNERS ................................ ................................ .... 18 2.1 Who Are H eritage L anguage Learners? ................................ ......................... 18 2.1.1 Defining H eritage Language learners ................................ .................. 18 2.1.2 Heritage Language L ........................... 21 2.1.3 Predictive Variables of Heritage Language Proficiency In Their Heritage Language ................................ ................................ ..... 24 2.1.4 Research on Similarities and Differences between Heritage Language and Non Heritage Language Learners .............................. 27 2.2 C hinese Heritage Language Learners ................................ ............................ 34 2.2.1 Chinese Dialects ................................ ................................ .................. 34 2.2.2 Chinese People in the U.S. ................................ ................................ .. 36 2.2.3 Defining Chinese Heritage Language Learners ................................ ... 38 2.2.4 The Differences between Chinese Heritage Language and Non Chinese Heritage Language Learners ................................ ................. 42 3 IMPLICIT AND EXPLICIT FEEDBACK ................................ ................................ ... 50 3.1 Theoretical Constructs ................................ ................................ ................... 50 3.1.1 Implicit vs. Explicit Learning ................................ ................................ 50 3.1.2 Implicit vs. Explicit Knowledge ................................ ............................. 53 3.1.3 Implicit vs. Explicit Feedback ................................ ............................... 56 3.2 Measuring Effectiveness of Corrective Feedback ................................ .......... 62 3.2.1 Uptake and Learner Repair ................................ ................................ 62 3.2.2 Private Speech ................................ ................................ .................... 64 3.2.3 ................................ ............................. 66 3.2.4 Test Perf ormance ................................ ................................ ................ 72 3.3 Previous Empirical Research Comparing the Effects of Implicit and Explicit Feedback ................................ ................................ ................................ ....... 75 3.4 Chinese Heritage Langu age Learners, Implicit and Explicit F eedback .......... 93

PAGE 6

6 4 RESEARCH QUESTIONS AND METHODOLOGY ................................ .............. 101 4.1 Research Questions ................................ ................................ ..................... 101 4.2 Methodology ................................ ................................ ................................ 103 4.2.1 Operationalizations ................................ ................................ ............ 103 4.2.2 Design ................................ ................................ .............................. 106 4.2.3 Participants ................................ ................................ ......................... 109 4.2.4 Target Linguistic Items ................................ ................................ ...... 116 4.2.5 Material s ................................ ................................ ............................ 120 4.2.5.1 Background information questionnaire ................................ 120 4.2.5.2 Treatment sessions ................................ .............................. 122 4.2.5.3 Testing material ................................ ................................ .... 124 4.2.5.4 The oral imitation test ................................ ........................... 127 4.2.5.5 The untimed written cloze test ................................ .............. 129 4.2.5.6 Advantages and disadvantages of the testing design in the current study ................................ ................................ ......... 130 4.2.5.7 n of the feedback .......... 131 4.2.5.8 Interview ................................ ................................ ............... 135 4.2.6 Procedure ................................ ................................ .......................... 136 4.2.7 Transcriptions ................................ ................................ .................... 139 4.2.8 Scoring ................................ ................................ .............................. 139 4.2.9 Coding ................................ ................................ .............................. 139 4. 2.9.1 Classifier related comments ................................ ................. 140 4.2.9.2 Non classifier related comments ................................ ........... 143 4.2.10 Inter Rater Reliability ................................ ................................ ......... 146 5 RESULTS ................................ ................................ ................................ ............. 148 5.1 Research Question 1 ................................ ................................ .................... 148 5.1.1 Descriptive Statistics ................................ ................................ ......... 148 5.1.1.1 Classifier related comments ................................ ................. 150 5.1.1.2 Non classifier related comments ................................ ........... 152 5.1.2 Inferential Statistics ................................ ................................ ........... 153 5.2 Research Question 2 ................................ ................................ .................... 154 5.2.1 Descriptive Statistics ................................ ................................ ......... 154 5.2.1.1 Oral test ................................ ................................ ................ 154 5.2.1.2 Written test ................................ ................................ ........... 156 5.2.1.3 ledge reflected in their verbal comments ................................ ................................ ............. 159 5.2.2 Inferential Statistics ................................ ................................ ........... 161 5.3 Research Question 3 ................................ ................................ .................... 164 5.3.1 Descriptive Statistics ................................ ................................ ......... 164 5.3.1.1 Oral imitation test ................................ ................................ .. 164 5.3.1.2 Written cloze test ................................ ................................ .. 165 5.3.2 Inferential Statistics ................................ ................................ ........... 167

PAGE 7

7 6 DISCUSSION AND CONCLUSION ................................ ................................ ...... 169 6.1 Research Question 1 ................................ ................................ .................... 169 6.2 Research Question 2 ................................ ................................ .................... 185 6.3 Research Question 3 ................................ ................................ .................... 193 6.4 Implications ................................ ................................ ................................ .. 207 6.4.1 Theoretical Implications ................................ ................................ ..... 207 6.4.2 Pedagogical Implications ................................ ................................ ... 210 6.5 Limitations and Future Research ................................ ................................ 212 6.6 Conclusion ................................ ................................ ................................ ... 215 APPENDIX A TRANSCRIPTION CONVENTIONS ................................ ................................ ..... 217 B CONSENT FORM ................................ ................................ ................................ 218 C QUESTIONNAIRE ................................ ................................ ................................ 220 D WRITTEN PRE TEST ................................ ................................ ........................... 222 E WRITTEN POST TEST ................................ ................................ ........................ 224 F TREATMENT 1 ................................ ................................ ................................ ..... 226 G TREATMENT 2 ................................ ................................ ................................ ..... 227 LIST OF REFERENCES ................................ ................................ ............................. 228 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 239

PAGE 8

8 LIST OF TABLES Table p age 3 1 Key characteristics of implicit and explicit knowledge ................................ ......... 54 3 2 Laboratory studies of effects of implicit and expl icit feedback ............................ 87 3 2 Continued ................................ ................................ ................................ ........... 88 3 3 Classroom studies of effects of implicit and explicit feedback ............................ 89 3 3 Continued ................................ ................................ ................................ ........... 90 3 3 Continued ................................ ................................ ................................ ........... 91 3 4 C omputer assisted language learning studies of eff ects of implicit and explicit feedback ................................ ................................ ................................ ............. 92 4 1 The targeted classifiers. ................................ ................................ ................... 121 4 2 The experimental and control groups. ................................ .............................. 137 5 1 Feedback episodes provided for C hinese heritage language learner group ..... 148 5 2 Feedback episodes provided for non Chinese heritage language learner group ................................ ................................ ................................ ................ 149 5 3 Commented feedback episodes by Chinese heritage language learner group 150 5 4 Commented feedbac k episodes by non Chinese heritage language learner group ................................ ................................ ................................ ................ 150 5 5 Classifier related comments provided by Chinese heritage language learner group ................................ ................................ ................................ ................ 151 5 6 Classifier related comments provided by non Chinese heritage language learner group ................................ ................................ ................................ .... 151 5 7 Chinese heritage language ed as classifier related ................................ ................................ ........................... 153 5 8 Non Chinese heritage language identified as classifier related ................................ ................................ ........... 153 5 9 Summary of results of LSD post hoc test ................................ ......................... 154 5 10 Summary of oral pre test scores for Chinese heritage language learner group 155

PAGE 9

9 5 11 Summary of oral pre test scores f or non Chinese heritage language learner group ................................ ................................ ................................ ................ 155 5 12 Summary of oral post test scores of Chinese heritage language learner group ................................ ................................ ................................ ................ 156 5 13 Summary of oral post test scores of non Chinese heritage language learner group ................................ ................................ ................................ ................ 156 5 14 Summary of written pre test scores f or Chinese heritage language learner group ................................ ................................ ................................ ................ 157 5 15 Summary of written pre test scores f or non Chinese heritage language learner group. ................................ ................................ ................................ ... 157 5 16 Summary of written post test scores f or Chinese heritage language learner group ................................ ................................ ................................ ................ 158 5 17 Summary of written post test scores f or non Chinese heritage language learner grou p ................................ ................................ ................................ .... 158 5 18 Comparison of pre test and post test scores across the groups ....................... 159 5 19 Chinese heritage language nowledge of target classifiers 160 5 20 Non Chinese heritage language classifiers ................................ ................................ ................................ .......... 160 5 21 One way ANOVA for the oral pre test ................................ .............................. 161 5 22 One way ANOVA for the written pre test ................................ .......................... 161 5 23 One way ANOVA for the oral post test ................................ ............................. 162 5 24 One way ANOVA for the written post test ................................ ........................ 162 5 25 LSD p ost hoc test for Chinese heritage language learner s on the o ral test ...... 162 5 26 LSD p ost hoc test for non Chinese heritage language learner s on the oral test ................................ ................................ ................................ .................... 162 5 27 LSD p ost hoc test for Chinese heritage language learner s on the written test 163 5 28 LSD p ost hoc test for non Chinese heritage language learner s on the written test ................................ ................................ ................................ .................... 163 5 29 Summary of increase on the oral post test f or recast group ............................. 164 5 30 Summary of increase on the oral post test for meta linguistic group ................ 164

PAGE 10

10 5 31 Summary of increase o n the oral post test for control group ............................ 165 5 32 Summary of increases o n the written post test for recast group ....................... 166 5 33 Summary of increases o n the written post test scores for meta linguistic group ................................ ................................ ................................ ................ 166 5 34 Summary of increases o n the written post test f or control group ...................... 166 6 1 Frequency of uptake and perception of learner XSB ................................ ........ 174 6 2 Learners who increased zero or decreased in t he experimental groups. ......... 199 6 3 The profile of learners TW and SBY ................................ ................................ 202 6 4 knowledge of Chinese characters on the pre and post test ............. 206

PAGE 11

11 LIST OF FIGURES Figure page 2 1 A c ontinuum of h eritage l anguage ................................ ..... 22 3 1 The explicitness of the feedback ................................ ................................ ........ 61 4 1 The r esearch design. ................................ ................................ ........................ 108 4 2 Experimental procedure and appr oximate time ................................ ................ 138 4 3 Coding scheme of the recall data ................................ ................................ ..... 140 5 1 Feedback episodes provided for C hinese h eritage l anguage group. ................ 149 5 2 Feedback episodes provided for non C hinese h eritage l anguage group. ......... 149 5 3 Percentage of classifier related comments acro ss experimental groups .......... 151 5 4 Mean scores of oral pre test ................................ ................................ ............. 155 5 5 Mean scores of oral post test ................................ ................................ ........... 156 5 6 Mean scores of written pre test ................................ ................................ ........ 157 5 7 Mean scores of written post test ................................ ................................ ....... 158 5 8 C hine se h eritage l anguage 160 5 9 Non C hinese h eritage l anguage classifiers ................................ ................................ ................................ .......... 160 5 10 Increase on the oral post test across the experimental groups ........................ 165 5 11 Increases o n the written post test across the experimental groups .................. 166

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE EFFECT OF IMPLICIT AND EXPLICIT FEEDBACK: A STUDY ON THE ACQUISITION OF MANDARIN CLASSIFIERS BY CHINESE HERITAGE AND NON HERITAGE LANGUAGE LEARNERS By Ye Han August 2010 Chair : Theresa Antes Major: Linguistics Previous studies revealed mixed results in terms of the relative effects of implici t and explicit feedback: some found that explicit feedback worked more efficiently than implicit feedback; others found no difference between the two feedback types. These contrasting results called for further investigations into this issue, particularly examining those underexplored learner groups who have developed their target language (TL) in a different path from traditional foreign/ second language learners investigated in previous studies, such as heritage language (HL) learners. Therefore the prese nt study aims at contributing to the literature by investigating the relative effects of implicit feedback in the form of recasts and explicit feedback in the form of meta linguistic feedback on the acquisition of Mandarin classifiers by Chinese heritage l anguage (CHL) and non heritage language (non CHL) learners. This study employed a pre test treatment post test research design, in which 64 participants (CHL=35, non CHL=29) were randomly divided into four experimental groups and two control groups. Feedb ack was provided

PAGE 13

13 during the stimulated recall, and their acquisition of classifiers we re assessed by their test scores. The data were both quantitatively and qualitatively analyzed. The results ffect their perceptions of feedback, as well as their acquisition. The findings of this study expand our knowledge about implicit and explicit feedback. In addition, they also provide invaluable information particularly for educators and administrators who are involved in HL instruction.

PAGE 14

14 CHAPTER 1 INTRODUCTION The present study examines the effects of implicit and explicit feedback on the acquisition of Mandarin classifiers by learners from heritage language (HL) and non heritage language (non HL) backgro unds. The study is rooted within an interactionist framework, drawing from previous research in second language acquisition (SLA) on the topic. Implicit feedback is operationalized in the form of recasts, which are defined feedback takes the form of meta linguistic feedback, which refers to a provision of meta linguistic information to indicate an error made by the learner through highlighting the nature and the character istic of the target language (TL) form. Previous research on the effects of implicit and explicit feedback has revealed controversial results: some researchers found that implicit and explicit feedback worked equally effectively (e.g., Carroll, 2001; DeK eyser, 1993; Kang, 2009; Kim & Mathes, 2001; Leeman, 2003; Loewen & Nabei, 2007), while others reported the superiority of explicit feedback over implicit feedback (e.g., Ammar & Spada, 2006; Lyster, 2004; Muranoi, 2000; R. Ellis et al., 2006). In previous studies, learners usually studied the TL as a foreign language (FL) or a second language (SL) through formal classroom instruction; therefore they had developed their explicit knowledge and literacy skills but were weak in their implicit knowledge. Howev er, little empirical research has been done to investigate learners who have developed implicit knowledge but lack explicit knowledge and literacy skills due to their exposure during childhood, such as HL learners. In addition, in the United States, HL ins truction and research has been attracting more and more interest from researchers and language educators, along with

PAGE 15

15 the rapid growth of the HL population in postsecondary FL/ SL classrooms since the 1970s ( Valds, 2005) In less than two decades, the fiel d of CHL instruction has made astounding progress. More and more language teachers in postsecondary Chinese programs have to face CHL learners and non CHL learners in the same classroom. Due to their different language backgrounds in Chinese, these two gro ups of learners have different language behaviors and language needs. However, little empirical research has been done in the past to compare the linguistic behaviors of these two groups of learners Motivated by gaps in the literature, the present study e mpirically investigates the effects of recasts and meta linguistic feedback on learners with CHL background, in comparison with learners without CHL background The following three issues are meta linguistic feedback; the developmental benefits of recasts and meta linguistic feedback; the relationship between feedback type and written tests Chapter 2 highlights some of the fundamental issues associated with HL learners, with a particular emphasis on CHL learners. In particular, it asks and answers the following questions: who are HL learners? How do they differ from non HL learners? In the first part, the chapter will examine HL learners in g eneral. It will provide definitions of HL learners based on previous studies. Then it will discuss issues related to HL predict their proficiency levels. The chapter will then present a review of previous research on the comparison between HL learners and non HL learners. The second

PAGE 16

16 part of the chapter will focus on CHL learners in particular. It starts with an overview of Chinese language and Chinese immigrants in the U.S followed by a review of definitions of CHL learners provided in the literature. Finally, the chapter will discuss the differences between CHL learners and non CHL learners in linguistic, cultural and social aspects based on the review of previous researc h. Chapter 3 will discuss effects of implicit and explicit feedback within an interactionist framework. The chapter will first introduce some theoretical constructs which relate to the research goal of the present study, including implicit and explicit learning; implicit and explicit knowledge; and implicit and explicit feedback. Then the chapter will introduce four measures that have been frequently employed in previous studies to assess the effectiveness of corrective feedback, such as uptake and lear ner chapter will then review previous empirical research on the comparison of the effects of implicit and explicit feedback. Finally, the chapter will provide a working defin ition of implicit feedback and explicit feedback for the present study, based on a discussion of Based on the theoretical basis built from the literature review in C hapters 2 and 3, Chapter 4 will pr esent research questions that were motivated by gaps in the previous research, as well as the methodology employed to answer the research questions. It will describe the materials used to collect the data in the study, as well as the analytical tests emplo yed to evaluate the data. The results of the experiment will be presented in Chapter 5. The chapter will include three sections, each of which will report one

PAGE 17

17 research question. For e ach research question the results of descriptive statistics will be repo rted first, followed by the results of inferential statistics. In C hapter 6, the findings will be discussed with regard to the research questions. I about meta linguisti c feedback and recasts ( R esearch Q uestion 1), and then discuss the relative effects of the two types of feedback ( R esearch Q uestion 2). Finally, I will discuss the findings of the relationship between feedback type and their performan ce on the post test s will be considered as the two most important independent variables through out the whole discussion Following the discussion of the findings, t he theoretical and pedagogical implication s and some limitations of the current study will be discussed. Possibilities of future research will be raised. The chapter concludes with a summary of how the findings of this study can extend our knowledge of effectiveness of feedback on learners with di fferent language backgrounds.

PAGE 18

18 CHAPTER 2 HL LEARNERS 2.1 Who Are HL L earners ? 2.1.1 Defining HL l earners In the U.S., before the term HL learner was introduced in the FL teaching profession in the mid 1990s, HL learners w ere labeled as native speakers, qua si native speakers, or bilingual students. However, none of the se terms is able to effectively distinguish this group of learners from traditionally acknowledged learner groups in the literature such as an L1 learner group, or an L2 learner group. In fact HL learners share both similarities and differences with L1 and L2 learners : l ike L1 learners, they experience early exposure to the HL (Montrul 2008, p.500) However, these exposures are usually either restricted or in sufficient Therefore their knowledge and ability of the HL are usually incomplete, which is similar to L2 learners (Carreira, 2004; Egi, 2007; Gregg, 1984; Han, 2000; Koda et al., 2008a; Loewen, 2004; McLaughlin, 1978; Montrul, 2008; R usso et al., 1989; Sanz et al., 2009; Sauro, 2009; Slimani Rolls, 2005; Valds, 1995; 2001; Wu, 2008 ) On the other hand their historical, family and ethnic connection s to their HL are not shared by L2 learners and differ from L1 learners. The te r m HL l earner high lighted this unique connection. Since HL learners have unique language experiences and linguistic abilities which set them apart from the traditional L1, L2 and FL learners, i n the language education field, it is important to clarify who are the HL learners particularly for those who work on curriculum design, materials selection, student placement and assessment, and teacher training (Carreira, 2004 ). However, this group consists of people with various hi storical,

PAGE 19

19 social, linguistic, and demographic background s : t hey can be foreign born immigrants who moved to the U.S. at a young age, native born children with foreign born immigrant parents, or native born children of native born parents with immigrant bac kground (Valds, 2005 ). Because of the heterogeneous nature of the population in this group, defining the HL learner is not an easy task. In the literature, t hree criteria have been employed to define HL learners: their membership in the HL commu nity, their personal connection through family background, and their greater or lesser linguistic proficiency in the HL (Carreira, 2004 ) Based on the above criteria, numerous definitions have been proposed. These defin itions can be roughly divided into two groups: a broadly defined group and a narrowly defined group (Kondo Brown, 2005 ) Definitions belong ing to the broadly defined group immigrant languages) (Kondo Brown, 2005 (Van Deusen Scholl, 2003 p.222). The most influential and representative definition in this group proposed by Fishman, defined HL learners based on two criteria: learners who speak LOT E s (languages other than English), and who have a particular family relevance to the language (2001 ). Since definitions in the broadly defined group only emphasize the ancestral, ethnic background connection to the language, with out considering proficiency level they often include withi n the gro u p many members who actually cannot speak a word in the HL (Van Deusen Scholl, 2003 ) Consequently i n university level FL classes, students who also considered as HL learners,

PAGE 20

20 (Van Deusen Scholl, 2003 p. 222 ). Thus definitions in the broadly defined group provide language educators with little help in distinguishing HL learners from traditional FL /L 2 learners in the classroom On the other hand, definitions in the narr owly defined group adopt language proficiency and language affiliation as the criteria to restrict the members within the HL commitment or connection to the home, community, or ( Li & Duff, 2008, p.16 ) The most widely accepted definition has been pro posed by Valds. In her definition, a HL learner refers to someone who wa at leas t understands the language and In this definition, Valds sets three main criteria for identifying HL learners: the home language, minimal proficiency in the HL, and the societal language. T his definition restricts HL learners to those who have achieved some degree of bilingual proficiency (Kondo Brown, 2005 ), exclud ing individuals with strong family or personal connections to the HL but who cannot speak the language. Nevertheless, the proficiency of learners i s defined in a relatively broad and vague nition proficiency level was given in the National Standards for Foreign Language Education Project (1999 ) : nts may come to class able to converse in the language in home and community situations but may lack the abilities to interact comfortably in more formal settings. Further, they may be quite comfortable with oral language but possess

PAGE 21

21 limited skills in read This de scription highlighted at least they tend to have stronger oral proficiency than reading and writing abilities; in terms of register, they are more likely to control inf ormal than formal registers However, fully assessing the language proficiency of a HL learner is a far more complex issue, which I will address in the next section. 2.1.2 HL L L anguage P roficiency Although HL learners grow up in a bilingual ( or even multilingual ) environment, they hardly have domain of interaction, and rare ly use (Valds, 2005 p.414). In fact, HL learners mostly use HL only with their family members (e.g. parents, grandparents, etc) and people from the com munities where the HL is spoken (e.g. church language school ) They usually receive their formal educati on at mainstream schools where the entire curriculum is conducted in English. Unless being taught at home or sent to after school community HL schools, they usually develop their literacy only in English (Valds, 1995; 2001 ). Therefore HL learners are not full bilinguals or balanced bilinguals who have absolutely equivalent proficiencies in two languages. Instead, they are (Valds, 2005 p.414). Their proficiency can be considered within a linguistic abilities and communicative strategies that an individual may access in one or the other of his or her two languages a t a specific moment, for a particular purpose, in a (Valds, 1995 p.316), a s illustra ted in Figure 2 1.

PAGE 22

22 Figure 2 1 A c (Valds, 2005, p. 414) In Figure 2 1, the size of the fonts indicates language strengths in languages A and B respectively. The two letters --at the left and right ext reme s indicate monolinguals who speak only language A or B. The two same middle indicate absolute equivalent language proficiency in languages A and B, which is only theoretically possible. As illustrated in Figure 2 1, due to l imited linguistic input, HL learners usually never acquire all the domains in their HL completely such as grammatical, textual, illocutionary, and sociolinguistic competence (Bachman, 1990 ). Therefore HL learners usually have unbalanced abilities in their H L and societal language, and this u nbalance usually will be further expanded when the quantity of linguistic input and linguistic interaction necessary to maintain the full lexical, phonological, morphological, and syntactic distinctions that are made by fluent competent speakers (Anderson, 1982 p.91) become unavailable or less available. Previous studies found that immigrant children rapidly lose their HL skills after starting their schooling in mainstream schools and their dominant language will quickly shift from HL to English (Fishman, 1991; Wong Fillmore, 1991 ). Some of them may lose their HL completely; others may try to reconnect with their HL through formal education such as taking language courses at universities or colleges. However, even

PAGE 23

23 after re acqu iring the once lost HL, whether or to what extent a HL learner can maintain the HL depends on many factors, such as personal determination, motivation, language environment, language needs, etc. Therefore, at the individual level, proficiency is consistently undergoing change in their lifetime (Valds, 2005 ). Since the home and community environment where HL learners grow up are quite different, t he distribution of strengths in the ir HL also vary from individual to individual (Valds, 2005 ). For instance, some HL learners are strong in receptive grammars, but limited in productive ones. They can comprehend what they are told in their HL, but are only able to respond in English. Some HL learners have no problems when engaging in an informal conversation, but cannot handle styles and registers requested in academic or professional settings. If we consider HL proficiency as a continuum, some HL learners fall in to the lower extreme, while some fall in to the higher extreme and ot hers fall in between In o ther words, s ub groups with various proficiencies exist in the HL learner group. In the U.S., many postsecondary level language programs have been trying to answer the needs of rapidly growing HL learner populations by providing a dual t rack for FL and HL learners. However, a lot of the time, HL learners are simply put in a separate track without considering the heterogeneous nature of the population T his simple dichotom y h as been questioned by some researchers (Kondo Brown, 2005; Lynch, 2008 ). Kondo Brown (2005) investigated the grammar, reading, and listening skills o f 185 HL learners and FL learners of Japanese. She found great similarities between FL learners and American born HL learners who have neither Japanese speaking parents nor experience living in Japan. On the other hand, HL learners with at

PAGE 24

24 least one nativ e Japanese speaking parent presented significantly higher proficiency than the rest of the learners. Her findings indicate that HL learners with lower proficiency are more like FL learners in terms of their linguistic skills. Even within the HL learner gro up, the between group difference is quite remarkable. Therefore simply separating HL learners from FL learners without considering their proficiency differences will risk putting learners with completely different language abilities in the same cl assroom. Lynch (2008) compared HL learners with lower proficiency and L2 learners at advanced level. He found more similarities than differences between the two ct ion 2.1.4. 2.1.3 Predictive V ariables of HL L P roficiency I n T heir HL HL learners come to the classroom with a wide range of linguistic knowledge and abilities which set them apart from L2/ F L learners who had no previous exposure to the target language. Even with extremely limited lang uage input, the general sense and intuition about the language and the culture that they developed still give s HL learners advantages over their non HL counterparts. Thus unlike non HL learners who need to learn everything from scratch, HL learners already know something about the language. The question is: what do they know? And to what extent do they know it ? Without knowing the answers to these questions it is impossible to provide appropriate and efficient instruction to HL learner s However, finding a n answer to the se question s is an extremely difficult and complex task. As Hendrys (2008 ) argued, knowledge system is similar to an iceberg which is formed by a smaller visible performance on the surface, but a much bigger invisible compete nce underneath. Therefore is hard to assess with the current proficiency assessment methods which are usually overwhelmingly base d on performance on the

PAGE 25

25 one hand, and ignore the competence on the other Since HL learners are more likely to have a much wider range and higher level of competence compared to their performance, t heir proficiency levels tend to be underestimated under the current performance oriented proficiency assessment system. In order to access the real pictu system, it would be best to conduct longitudinal research, observing each linguistic development stage of the lea r ner. However, it is not realistic to do so for every learner Nevertheless, the more we kno w about the background of a HL learner, the better we can infer his/her language skills In fact, a number of social variables have been identified as possible Learner generation has been found as an important pred ictive variable of HL maintenance and attrition Generally speaking, HL learners of an older generation usually have higher proficiency in their HL than those of a younger generation (e.g. first generation compared to the second generation, second generat ion compared with the third generation, etc). For first generation immigrants, age of arrival is an important predictive variable: usually older arrivals w it h exposure to English after 12 years old show higher proficiency than younger arrivals in their HL especially whe re pronunciation (Flege et al., 1995; Yeni Komshian et al., 2000 ), and lexical retrieval are concerned (McElree et al., 2000 ) However, some individuals have lower proficiency in English. Excluding the first generation immigrants, birth order plays an important role in predict ing first born children tend to have higher levels of HL proficiency than their younger siblings (Lambert & Taylor, 1996; Zentella, 1997 ).

PAGE 26

26 In real life, however, we sometimes encounter exception s F or instance, a third generation child speaks better than a second generation child ; a younger arrival is superior to a n older arrival ; or a third born child speaks more fluently than the first born child in their HL Focusing on Spa nish HL learners, Lynch tries to find explanations for t hese exceptions from s peaker social networks T hat is social networks including active speakers of Spanish would fare better in terms of language acquisition and maintenan ce than those with less dense social networks in (Lynch, 2003 p. 9). In a related study, Zentella (1997 ), shows that the Spanish proficiency of five working class Puerto Rican girls in New York are close ly social networks Thus supporting Based on her studies on Chinese HL learners, He argued that C ultural identity may be the most critical factor to HL development. She hypothesized that HL (2008b p. 116). Previous studies found HL cultural identity and HL proficiency are closely associated: Lee (2002 ) examined the identity and language choices of 40 second generation Korean HL learners. The results show the stronger the identification with the Korean orie ntation items, the higher the Korean language proficiency. Another study on 114 adult second generation and first generation (early arrival) Korean HL learners also found that those w it h higher Korean HL proficiency reported a stronger Korean cultur al iden tity (Cho, 200 0 ). Similar results were also revealed in Kondo (1997 ) and Tse (2000 ).

PAGE 27

27 Socioeconomic class is another important predictive variable. Two studies o n Spanish HL speakers based on Mexican Americans in Texas revealed contradictory results. Snchez (1983 ) found that those from lower socioeconomic class es tend ed to use more Spanish compared with those from the middle class. However, the opposite was found by Amastae (1982 ) Jia (2008 ) hypothesized that the differen t results found in the two studies may have been due to a generational effect, which was partially confirmed in her research on Chinese immigrants in New York She found that l ower family income sig nificantly predicted higher self assessed HL ability in reading, speaking, and writing She concluded that first generation immigrants with higher income, higher educational levels, and higher English proficiency have a stronger desire to assimilate into t he mainstream society. Despite the different results of the above research, socioeconomic class was proven to correlate with HL proficiency of HL learners. 2.1.4 Research on S imilarities and D ifferences b etween HL and N on HL Learners In the U.S., the pr esence of HL learners in postsecondary level language classrooms has been a challenge in the FL teaching field since the mid 1970s The tea ching professionals in this field who were trained to teach FL learners, found themselves hav ing to face a group of students who the language than they were, but who could not talk about the language using the terminology used in the teaching of traditional grammar who time learning grammar rule taught to (Valds, 1995 p. 304). Through classroom observation s language teachers know that these learners differ from FL learners. However, one must ask : how different are they ? Besides the differences, are there any similarities between learners of these two groups? Do methods proven to be

PAGE 28

28 effective in the L2 classroom also apply to HL learners? As pointed out by M o ntrul, w ithout proper understanding of how similar or different these two types of learners are, it is difficult to tell at this point whether the exact same methods applied to L2 learners in 00). Numerous researchers investigated similarities and differences between HL and non HL learners within the framework of UG (Isurin & Ivanova Sullivan, 2008; Montrul, 2004; Montrul et al., 2008 ). They consider HL learners as incomplete learner s since the an (Polinsky, 2008 p. 1 ) at some point in their life consequently (Montrul et al. 2008) In addition, the earlier they start to become bilingual in their HL and the societal language dominant language, the more incomplete the adult grammar may turn out t (Montrul, 2002 p.61). Due to the incomplete or interrupted acquisition of the HL, HL learners share some linguistic behaviors with L2 learners, such as transferring errors from the majority language. On the other hand, HL learners generally lack l it eracy and formal instruction i n the HL, which differ s from L2 learners. These UG linguists t ry to find out whether learners with HL background have advantages over those without HL background. Montrul and her colleagues (2008 ) investigated the knowledge of Spanish gender agreement among 69 HL learners and 72 L2 learners compar ed with the baseline data of 22 native speakers of Spanish. The proficiency of the HL learners wa s carefully controlled Only learners who met the following three criteria were chosen: they were

PAGE 29

29 born and schooled in the U.S. ; had no schooling e xperience in their home country; and became bilingual in Spanish and English before age five All the partici pants completed one oral picture description task and two written tests. The results showed that H L learners performed mo re like native speakers than the L2 learners on the oral test which require d more implicit knowledge ; however, they were less accurate o n the two written tests which required more explicit knowledge Moreover, like the L2 learners, the errors made by the HL learners were also systematic. Based on these results, Montrul and her colleagues argued that HL learners have automatically processed knowledge (typically acquired early in chil dhood) developed during a critical period; meta linguistic knowledge as they usually had no or very limited formal education in their HL. Therefore their knowledge of g ender represented (p. 541). Romanova (2008 ) studied the morphological processing of 50 real verbs and 30 nonce verbs by 14 heritage s peakers, 17 beginning L2 learners, and 28 native speakers of Russian. Of note seven h eritage speakers in her study had a few years of elementary schooling in Russia before moving to the U.S. The results showed that the heritage speakers performed closer t o native speakers and greatly outperformed L2 learners in their recognition of real verbs, especially real verbs of high frequency. H owever, in terms of nonce verbs, it was the L2 learners who performed more like native speakers. Romanova attributed the d ifferences to the limited linguistic input received by the heritage speakers in her study : incomplete first language acquisition, the weights of rules are not set and rules are

PAGE 30

30 applied with a probability of less tha n 1 (p.119) The results also showed that heritage speakers with literacy skills consistently outperformed those without literacy skills. Romanova suggested that literacy in L1 may be a major factor influencing the development of input based rules and pro bability mechanisms in morphological processing (Rothman, 2007) A small scale study conducted by Isurin and Iva n ova Sullivan (2008 ) i nvestigated seven h eritage speakers, 11 L2 learners, and five native speakers of Russian. All the A boy, a dog, a frog and a friend The elicited speech samples from the three groups of learners were analyzed with a focus on selected morphosyntactic categories ( e.g., aspect, case), and word order. They found that heritage speakers outperformed L2 learners in all three morphosyntactic categories. Lynch (2008 ) conducted a qualitative study on five HL and four L2 learners of Spanish. The five HL learners selected in this study were either born in the U.S. or migrated here before age two and had tw o to five years of formal study of Spanish. In other words, they were all typical lower proficiency HL learners On the other hand, the L2 learners in the study all had more than five years of formal study of Spanish meaning they were generally at a more advanced level of Spanish L2 learning Data were collected from individual interview s in Spanish by the researcher Selected grammatical features were quantitatively analyzed, such as noun adjective gender agreement, aspectual and mood distinction, subject verb word order, etc. The overall results showed more similarities than differences between the two groups. The two

PAGE 31

31 most advanced L2 learners consistently outperformed or performed equally well compared to the most advance d HL learner. To summarize the a bove studies: low proficiency HL learners who are either simultaneous or near simultaneous bilinguals in their HL and the societal language, show m ore similar ities with L2 learners (Montrul, 2002 ) Exposure to the HL usually begins at birth or at a very early age for these learners, and comes from their family or from communi ty members through daily life T his exposure i s tota lly input based, and thus these HL learners develop an implicit rather than explicit knowledge of the HL. Although their meta linguistic knowledge may also emerge, as it does for monolingual children there are usually no opportunities to make it ful ly develop through reading and writing practice, since they mostly attend English dominant mainstream schools. With a few exceptions, most of them only develop literacy skills in English, and remain illiterate in their HL. Without formal study and literacy their HL is more implicit and intuitive, thus differ ing from traditional FL and L2 learners, who routinely acquire and store explicit linguistic knowledge of the target language through formal classroom instruc tion (Lynch, 2008 ). Although there are so many similarities between low proficiency HL learners and L2 learners, the exposure to the HL at an early age has lasting benefits on HL learner s, even though their language experience is limited or incomplete, is interrupted by attending English dominant mainstream schools, and undergoes attrition. These lasting benefits enable them to have advantages over L2 learners in some areas, such as phono log y but not in other s such as morphosyntax (Au et al., 2002 ).

PAGE 32

32 The differences and similarities between HL and non HL learners provide rese archers in fie l d of SLA with a good opportunity to re conceptualize and expand the field by instruction currently referred to as the teaching of heritage languages (Valds, 200 5 p. 410). Lynch (2003 theory building in the field of heritage language acquisition (HLA) depends in part upon the research and theory already existent in second language acquisition Clearly, the general sorts of questions asked in SLA are questions that HLA researchers must be asking, and the research methodologies used to respond to those questions in SLA are methodologies that would lend themselves fruitfully to HLA The following two studies approached HL acquisition with typical SLA research design s Kim (2008 ) investigated sentence processing difference s between 88 HL and 40 non HL learners of Kor e an Based on their self reported L1 these learners were subcategorized into five groups. HL learners were divided into Korean L1 speakers, Korean and English L1 speakers, and English L1 speakers. Non HL learners were divided into two groups: Japanese L1 speakers whose language is also a SOV stru cture like Korean and Korean L2 learners whose L1 is a SVO structure ( e.g., English, Chinese, etc). All the participants were given a picture selection task which contain ed clause s. The results showed that HL learners with Korean as their L1 and L2 learners with Japanese as their L1 earn ed the highest scores among all the participants. Moreover, the analysis of error types indicated that HL learners found case markers

PAGE 33

33 difficult, wh ich affect ed their processing of relative clauses. Kim argued that although HL learners may not ha ve fully acquired the syntactic structures of Korean they can use their semantic and cont e xtual knowledge to comprehend sentences, even when they contain com plex structures In addition, since they acquired their HL implicitly, and lack explicit meta linguistic, or rule based knowledge, grammar focuse d instruction could play an important role in making them notice some features of the target language. Hence, in formal languag e instruction, it may be crucial for teachers to provide HL learners wit h consistent interaction with ne gative feedback that draws attention to the form, rather than meaning (p.124). The study conducted by Gass and Lewis (2007 ) w as the first attempt to approach the linguistic differences between HL and non HL learners within an interactionist framework. Thirteen FL learners and six HL learners of Italian participated in the study. All the participants took part in a task based int eraction (a spot the difference task) with an interlocutor who was a near native speaker of Italian. Feedback w as given on non were measured using stimulated recall meth odology. The results showed a between group difference of HL and non HL learners in terms of their perception of provided corrective feedback. Non HL learners outperformed HL learners in perceiving phonological and morphosyntactic feedback; while HL learne rs perceived semantic and lexical feedback more correctly than non HL learners. In addition, HL learners also tend to incorrectly interpret feedback on other linguistic items as semantic feedback as well. Based on these results, Gass and Lewis argued that the exposure to HL during regard language as a form of real communication

PAGE 34

34 whereas L2 learners who acquire language only in a classroom setting may treat the This study hol ds extremely important implication s for language instruction. That is, the feedback technique s that have proven to be effective for FL/L2 learner s may also work for HL learners, but in a different way. 2.2 CHL Learners 2.2.1 Chinese D ialects 1 Chinese is an umbrella term for seven major dialects in the Chinese language family: Mandarin 2 Wu Xiang Gan Min Cantonese and Hakka (He, 2006 ) T hese seven dialects can be subdivided into two groups based on geography : Northern Chinese and Southern Chinese Mandarin is a label applied to dialects spoken in Northern China. More precisely, i t refers t o dialects spoken along the Yellow Plain and the Loess Pl ateau, including Northwestern, Northern proper, River, and Southwestern varieties Despite having these varieties, Mandarin is highly unified D ifferent dialects of Mandarin are mutually intelligible On the other hand, Southern Chinese consists of six mutually unintelligible dialects 3 : Wu Xiang Gan Min Cantonese and Hakka The biggest obstacle for Northern Chinese and Southern Chinese speakers to understand each other is the phonol ogical difference s among the dialects : generally speaking, Southern Chinese dialects tend to have more tones than Northern Chinese dialects. In addition, Southern Chinese dialects also ha ve sounds which do not exist in Northern Chinese, such as stop ending s (Sun, 2006 ) In the SLA literature, Mandarin refers to the standard dialect of Chinese officially recognized by the governments in mainland China, Taiwan, Hong Kong, and Singapore It is based on the and

PAGE 35

35 Northern dialects as its base (Chen, 1999 p. 24). It is cal led putonghua and Hong Kong guoyu huayu It is also the widely taught variety in Chinese language classrooms (He, 2008a ). Henceforth, the term Mandarin only refers to the standard dialect of Chinese in th is dissertation. In mainland China, the government has played an active role in promot ing Mandarin, or putonghua since the 1950s. According to survey results released by t he National L anguage Committee (2005 ) 53% of the Chi n ese population in mainland China speaks Mandarin. The percentage of the Mandarin speaking population varies dramatically with age and education. Ge neral ly speaking, t he younger the age, and the higher the education, the higher the percentage of Mandarin sp eak ers within the population For instance, while 40.59% of people from 45 59 years old speak Mandarin th e percentage jumps to 70 .12 % when the age of the population decreases to be tween 15 and 29 years old In terms of education, while 25 49 % of people with elementary school education speak Mandarin, 86.77% of people with more than two year s of college education are Mandarin speakers Compared with mainland China, Taiwan was more strict in promoting Mandarin or guoyu in t he 1980s, at the expense of other Chinese dialects and aboriginal languages. All dialects other than guoyu w ere banned in schools and mass media. As a result, Taiwan which regarded guoyu Chinese dialects and aboriginal languages as the low langu was more (Chen, 1999 p. 60). In 1991, about 90% of the population of Taiwan spoke guoyu which wa s much

PAGE 36

36 higher than the percentage of the Mandarin speaking population in mainland China at the same era. At the same time, 82.5% of native speakers of Southern Min the most widely spoken dialect other than Mandarin in Taiwan -are bilingual in Southern Min and guoyu ; 88% of native speakers of other Chinese dialects and aboriginal languages are bilingual in guoyu and thei r native tongue However, since the late 1980s, there has been an increased awareness and enhance ment of regional identity and local dialects. In 1987, the Ministry of Education formally allowed school children to speak dialects at school. Consequently, t h e use of Southern Min ha s dramatically increased in the society (Huang, 1993 ). Although 97% of Hong Kong people speak Cantonese as their native language, English instead of Cantonese or Mandarin had been regarded as the High language until July 1997, the year when Hong Kong became a Special A dministrative Region of China. However, after 1997, the use of Cantonese in public functions of the government has been dramatically increased. P utonghua is also becoming more and more important in the society (Chen, 1999 ). 2.2.2 Chinese P eople in the U.S. According to the U.S. Census Bur eau, in 2000, among the over 2.2 million people of Chinese ethnicity in the United States 83% of them spoke Chinese, while the rest of them were either English monolinguals or speakers of other languages. Moreover, the majority of the Chinese speakers wer e foreign born, and ranked as the fourth largest immigrant group in the U.S. These people or their ancestors, moved to the U.S. mainly as the result of three waves of large scale immigration from China to the U.S. in the past 1 6 0 years Th e first

PAGE 37

37 wave ca n be traced back to 1849, when the California gold rush occurred. At that time Chinese immigrants were poor farmers with no literacy skills, who were mainly from Guangdong Province and spoke solely Cantonese. T he second wave of Chinese immigration to the U.S. took place between the 1950s a nd 1970s. The Chinese who immigrated during this period usually came from more educated and wealthy background s than their Cantonese speaking predecessors Among them, those from mainland China and Taiwan spoke mostly Man darin, while those from Hong Kong spoke Cantonese. The largest and still on gong Chinese immigration started in the 1980s, and has continued to date. The majority of these immigrants are scholars from mainland China Regardless of which province in China t hey are origina te from, in addition to the local dialects (if there are any), these people all speak Mandarin (Chang, 2003 ). The above brief historical sketch shows that the language diversity that currently exist s among the Chinese in the U.S. correlated with the historical fact that these people or their ancestors original ly came from differen t parts of China where various dialects were spoken. Based on th is observation, people who migrated from mainland China and Taiwan speak better M andarin than those who were from Hong Kong. This observation was partially proven by an on line survey conducted i n 2006, by Wiley and his colleagues They investigated Mandarin and other dialects spoken among Chinese immigrants and international students in the U.S. Overall, more than 50% of the respondents claimed that they spoke Mandarin wel l or very well regardless of the regions they were originally from. However, people from mainland China we re most confident about their fluency in Mandarin. In fact, 91.3% of them reported that they

PAGE 38

38 spoke Mandarin very well, which wa s higher than the perce ntage reported by people from Taiwan ( 79.8% ). O nly 18.4% of people from Hong K ong thought they spoke Mandarin very well (Wiley et al., 2008 ) Although international student s and HL speakers were not differentiated in the survey the results still show that the region s of origin of the Chinese people in the U.S. can be used as a useful vari able to predict their proficiency in Mandarin. 2.2.3 Defining CHL L earners As with HL learners in general, CHL learner s are defined either broadly based on their ethnic, historical, and family connection to the language, or narrowly according to their a ctual linguistic proficiency in Chinese. Following Fishman (2001 ), CHL learner s can be defined as individuals who have a personal interest or involvement in Chinese, a language spoken by their ancest ors (Li & Duff, 2008 ). This definition includes a wide range of learners with vari ous language backgrounds, such as learners who are ethnically Chinese but with no or limited exposure to Chinese language and culture, learners who speak Chinese fluently but with no literacy skills, learners who speak a dialect other than Mandarin, etc. Based on actual linguistic competence and familial affiliation, following Valds (2001 Chinese is spoken and who speaks or at least understands the language and is to some degree b ilingual in Chinese (He, 2006 p.1). grasp of the HL f alling along a continuum of having very little HL knowledge to being (Li & Duff 2008, p. 17). Like HL learners of other languages, the cause

PAGE 39

39 experiences; however, the following two language specific factors also play important roles. The f irst is the home dialect factor. As Valds (2001 ) pointed out, in many cases immigrant students are speakers of non prestige varieties of their HL. In terms of CHL learners, as we have see n above they or their ancestors were originally from diffe r ent hometowns where various Chinese dia lects are spoken. Therefore their home dialects also vary accordingly. For instance, HL learners who are originally from Hong Kong mostly speak Cantonese at home, while those from mainland China more likely speak Mandarin, or a mix of Mandarin and other di alect(s) (Wiley et al., 2008 ). In the U.S. Mandarin is the most widely taught vari et y of Chinese in Chinese language classrooms. A ccording to a survey conducted by the Modern Language Association ( MLA ) in 2006, 5 1, 582 students enrolled in Mandari n class at the postsecondary level, which ranked as the 7 th most commonly taught language o ther than English in the U.S. On the other hand, other dialects of Chinese are hardly even introduced int o mainstream educational institutions. According to the same survey, the enrollments in Cantonese classes were only 178, while only 21 students took Ta iwanese w hich is one of the varieties of Min spoken in Taiwan. No records we re available for other dialects. This indicates that those dialects are either not taught in the classroom at all, or the enrollments are too small to be considered in the survey (Modern Language Association, 2006 ) Thus for many HL learners in the U.S., learning their HL in a postsecon dary level classroom means learning a second dialect (Valds, 2005 ) when their home dialect is

PAGE 40

40 not Mandarin. This is more challenging for learners whose home dialect is not mutually intelligible wi t h Mandarin, such as Cantonese, Min, etc. The se cond factor is the orthography system. Chinese people all use one unifying writing system Unlike English and other phonographic writing systems, the Chinese simultaneously encoding various concepts or ideas totally divorce d (Sun, 2006 p.102). Therefore, regardless of which dialect is spoken, a literate Chinese person can read any character with his/her own local pronunciation and write sentences that can be understood by people from any dialect without be ing able to speak tha t dialect Thus, u nlike English, l iteracy in Chinese is measured by the number of characters a learner has mastered rather than the number of words. In mainland China, a student i s requ ired to master minimally 2 500 characters when gr aduating from elementary school and 3 500 characters when graduating from college (Chen, 1999 ) It was estimated that there are over 56 000 Chinese characters, with 3 000 of them commonly used in modern Chinese As l earning characters is s o difficult and time consumi ng for Chinese children, the simplification of Chinese character s started in the early 1900s. In mainland China, b etween the 1950s and the mid 1960s, over 2 000 of the most commonly us ed characters were officially simplified, an d used in mainland China. This is known as jianhuazi Today simplified Chinese is also officially used in Singapore (He, 2008a ) In Taiwan and Hong Kong, however, the older script, which is known as fantizi is still used Therefore i n the U.S., HL learners whose families are originally from Taiwan and Hong

PAGE 41

41 Kong te nd to support instruction in traditional script while those from mainland China prefer simplified script (Chen, 1999; Sun, 200 6 ) Thus there is a high likelihood that HL learners even with some literacy skills have to re learn the writing systems when their home script is different from the school script. In addition, regardless of the simplified or traditional script, ther e exists a much gre a t er distance between spoken Mandarin and its orthography than that in a language with a phonographic writing system Therefore, the unbalance and disparity in the four language skills tends to be even more pronounced in CHL learners tha n HL learners of languages wi th a phonographic writing system. CHL learners are by definition, bilingual, but are not very often bi literate when they arrive in our classrooms (Dai & Zhang, 2008; Hendryx, 2008; Koda et al., 2008b ). To summarize, when considering the proficiency of C HL learners in the U.S., besides the factors which ar e observed in HL learners of other languages, we also need to look at l e home dialect s and writing scripts. Thus a learner from a Mandarin speaking family may still have to face the situation that the classroom script differ s from his/ her home scri pt A learner whose home dialect is unintelligible to Mandarin may have difficulty with aspects of the spoken language, but may not have equal trouble in writing if the classroom script is the same as his/ her home script (He, 2008a ) Many Cantonese speaking learners belong to this category. Based on classro om observation s at the postsecondary level in Chinese program s Hendryx identified five commonly seen subgroups among CHL learners (Hendryx, 2008).

PAGE 42

42 Learners in the first group usually have little command of the HL language, but do not lack intuitions abo ut it In addition, they also store a considerable amount of cultural knowledge. Learners in the second group have slightly better abilities than those in the first group, especially in terms of their speaking and listening skills. However, these skills ar e usually more receptive rather than productive. They usually have no or quite limited abilities in reading and writing. To summarize, CHL learners in the first and second group s fall into the group of lower proficient HL learners (Lynch, 2008 ), who share more similarities with L2 learners in terms of their language skills and linguistic behavior (Kondo Brown, 2005; Li, 2008 ; Lynch, 2003 ). Nevertheless, their limited exposure to the HL still gives them some advantages over those without any HL experience which enables them to acquire the target language faster than L2 learners. Learners in the third group have excellent speaking and listening skills, but very limited or even no reading and writing skills. On the contrary, learners in the fourth group have fairly developed reading and writing skills, but little abilities in Spoken Mandarin. In the U.S., most of these latter learner s are from Cantonese, and Taiwanese speaking families. For them, HL learning means learning the second dialect. Learners in the fifth group have near native proficiency in all four language skills. These learners all received formal or semi formal educat ion in the target language, either through after school program s community based schools, or by stud ying at schools in their home country. 2.2.4 T he D ifferences between CHL and Non CHL L earners In the U.S., CHL instruction and studies is a newly emerg ing field with approxi m at e ly 10 years of history. S i nce the mid 1990s, it has att r ac t ed a rap idly growing body of scholars from various disciplines, such as SLA, bilingualism, reading

PAGE 43

43 research, discourse analysis, orthography analysis, and/or language pedagog y (He, 2008a ) studies, CHL learners pre sented unique characteristics whic h differ from their non CHL counterparts in terms of linguistic knowledge system (Dai & Zhang, 2008; Hendryx, 2008 ), literacy skills (Jia, 2009; Koda et al., 2008b; Xiao, 2008 ), morphological awareness (Koda et al., 2008a ), Chinese character learning (Ke, 1998; Xiao, 2006 ), and their motivation (Lu & Li, 2008; Weger Guntharp, 2006 ) and identity (He, 2004; He, 2008b ) L inguistic proficiency CHL learners tend t o have uneven skills in speaking, listening, reading, and writing (Hendryx, 2008; Weger Guntharp, 2006; Wu, 2008 ) which result from their heterogeneous past exposure to and learning experience in Chinese Their C HL acquisition usually starts at home relying mainly on a vertical and reciprocal intimate relation with their parents or grandparents. With very few exceptions, the ir acquisition heavily emphasize s speaking and listening rather than reading and writing skills (Dai & Zhang, 2008) Consequent ly most of them are bilingual in English and Chinese, but not bi litera te particularly when little or limited print materials are available at home (Koda et al., 2008b ). In addition, b ecause of the dominance of English language and limited expos ure to the HL, HL learners usually experience great decline of their HL skills right after they start schooling in mainstream schools which is known as language attrition (Fishman, 1991; Wong Fillmore, 1991 ). R. Jia (2008) investigated 85 Chinese HL speakers in New York City, age d from 16 to 3 0 years old. More than 90% of the m were foreign born and e migrated to the U.S. before 20 years old. Data used in this stud y were collected through a tailor made language background questionnaire which a sked learners to self rate their proficiency in both languages for

PAGE 44

44 each two year segment of the entire time span from their systematic exposure to English The results showed HL skills all experienced g reat attrition after five years of residence in the U.S. on the one hand, and a great improvement in English on the other hand Moreover, a mong the four skills, reading and writing abilities showed larger s cales of attrition compared with speaking and listening skills. In contrast, non CHL learners usually have relatively even distribution of the four language skills, which almost entirely developed through formal classroom learning. Similar results were fo und by Xiao (2006) who conducted two studies comparing the language skills of the two learner groups The first one compared 18 CHL and 18 non CHL learners from a high beginning level intensive Chinese course in terms of their listening, grammar, reading, and character writing skills. The results showed that CHL learners significantly outperformed non CHL learners in listening and grammar, but not in reading comprehension and character writing T he second study was conducted among 94 HL and 54 non HL learn ers of Chinese from three different instructional levels: beginning, intermediate, and advanced. A 25 item grammaticality judgment test and a 6 item English to Chinese translation task were distributed during a 20 min u te class time. The results showed that HL learners did significantly better o n the grammaticality judgment test, but not o n the translation t a s k Ke (1998 ) also found that CHL learners had no significant advantages over non CHL learners in terms of Chinese character learning. He conducted a s tudy among 60 FL and 85 HL learners of Chinese, who just finished their first year of Chinese instruction at seven institutions. The demographic background information of the HL participants in this study is quite limited, and only briefly mentioned that t hese

PAGE 45

45 literacy experience ( e.g., home schooling, attending community schools, etc ). All th e participants were given one character recognition task, and one character production task. Each test consists of 30 characters chosen from the first year Chinese textbook used by these learners. The results showed no significant difference between the tw o groups of learners in terms of their recognition and production of Chinese characters. In sum, CHL learners are superior i n listening, speaking, and grammar which could be attributed to the large amount of HL input they are expose d to at home through d aily oral communication with their family members even though their skills are more rec eptive rather than productive. This advantage is not shared by their non CHL counterparts. In contrast, literacy skill development requires more systematic learning an d rich HL print resources at home (Xiao, 2008). Compared with English, which has inflectional morphemes in structurally transparent alphabet based words, the Chinese writing system is a logographic system, which do es not phonologically associate with the c haracter. Therefore, character learning is considered the hardest part in Chinese learning (Everson, 1998 ). Ma ny Ch inese parents try to teach their children at home. However, whether the learning could proceed in a systematic manner depends on give up trying (Hinton, 1999 ) and send their children to community weeken d Chinese schools for their CHL literacy development. However children usually lose their C HL literacy skills after formal schooling since their home literacy does not relate to their school literacy (Wong Fillmore, 1991 ). Thus many CHL learners are illiterate or with

PAGE 46

46 very limited literacy skills even though they are fluent in the oral language (Li, 2006 ). As a result, f o r CHL learners, the goals for their HL learning are mo re likely to improv e their literacy skills in order to read and write well. In contrast, for non CHL learners, they rank speaking and listening as the most important immediate goals in their Chinese lea rning (Wu, 2008). C ultural identity CHL learners bring to the classroom not only linguistic skills, but also their cultural identity which ha s been influenced by both Chinese and American cultures. This is also not shared by their non CHL counterparts. Similar to their linguistic skills, their cultur al identities also vary from individual to individual, depending on the home and environment they gr e w up in Dai and Zhang (2008) conducted a survey among 80 college students with CHL background, and their participants perceive d their own cultural identities differently. The majority of the participants view ed themselves as a combination of both Chinese and American cultures, and they drew on different cultural knowledge when interactin g with different peopl e, at different time s and in different social context s The rest of the participants either considered themselves and thus indicated that they belonged to neither Chinese or Ame rican culture, or accept ed one culture but reject ed the other. G. Jia (2008) found that HL speakers who identified a stronger connection and preference to their HL culture tend ed to not only use more HL, but also self rate their reading and writing skills higher In a related study, than just the background in CHL learning. She argued that through learning CHL, which

PAGE 47

47 takes place in a three dimensional framework with intersecting planes of time space, and identity, a very different kind of Chinese (He, 2008b p. 110). In addi tion, she notes that, dependent upon the degree to which s/he is able to find continuity and coherence in multiple communicative and social worlds in time and space and to develop hybrid, situated identities, based model for CHL development, she hypothesized that the degree of success in CHL development community members for the lon g run, the desire to communicate in CHL in a moment by moment manner the home language of the learner, the availability and diversity of the CHL input, as well as the extent to which the learner has created a niche in the English speaking linguistic, soci al, and cultural community. Motivation Chinese was ranked as one of the most difficult languages for native English speakers to learn (Hadley, 2001 ). Ther efore regardless of CHL or non CHL learners, a strong motivation must be present, since it is one of the key factors influencing the rate and success of second/foreign language (L2) learning (Drnyei, 1994; Ely 1986; Gardner, 1985 ) Gardner defined motivation as "the extent to which the individual works or strives to learn the language because of a desire to do so and the satisfaction experienced in this activity" (Gardner, 1985, p. 10). known s ocio educational model distinguished two types of motivations: instrumental and integrative. The former one refers to a pragmatic reason that drives learners to learn a

PAGE 48

48 FL or L2, such as career development, travelling, passing exams, moving to a new countr y, etc. On the other hand, integrative motivation closely relates to a desire of getting closer or even becom ing the member of a language community. Lu and Li (2008) conducted a comparative analysis of the effect of different motivational factors on 59 CHL and 61 non CHL learners from ni n e Chinese college classes. All the participants were required to complete a three part questionnaire, consisting of 63 questions, followed by an interview. The results showed that although both types of motivation played an important role in the Chinese learning by participants from both learner groups, CHL learners were significantly more influenced by instrumental motivation than their non CHL counterparts. In addition, their integrative motivation was more highly correlat ed to their perceptions on their listening and speaking abilities, but not their reading or writing skills. For the non CHL learners, the correlation was found between their integrative motivation and listening, speaking, as well as writing, but not readin g skills In sum, the previous research revealed differences between CHL and non CHL learners in linguistic, cultural and social aspects. Although research in this field h a s made great progress in recent years, t he majority of these studies were non expe rimental the data in which were mainly collected through questionnaires, interviews, classroom observations, case stud ies To further explore the cause and effect questions on the learning process and developmental path of CHL learners, experimental studi es are certainly necessary. To my knowledge, no experimental research has been conducted in comparison of the two learner groups under an

PAGE 49

49 interactionist framework such as th e benefits o f different types of corrective feedback, which motivated the current study. 1 In the current study, the term dialec t refers to all the local varieties spoken in mainland China, Taiwan, Hongkong which share a high degree of uniformity in terms of grammar and writing system. 2 All the Chinese characters used in this paper take the form of simplified Chinese. 3 Uninte lligible here refers mainly to the phonological differences between Southern and Northern dialects. Both the two dialects share the uniformed writing system. In the current study, speakers of both Southern and Northern dialects were also considered as nati ve speakers of Chinese.

PAGE 50

50 CHAPTER 3 IMPLICIT AND EXPLICI T FEEDBACK This chapter will focus on implicit and explicit feedback. First, two pairs of related theoretical constructs -implicit vs. explicit learning and implicit vs. explicit knowledge -will be introduced, foll owed by the definition of implicit vs. explicit feedback The n ext se ction will review the main measures that have been employed in the literature in examining the effects of corrective feedback. Then the discussion will move to the effects of feedback on 8 previous studies. Last ly implicit and explicit feedback will be discussed in relation to HL learners who are the target learner population in the current study and have been less studied in the pa st. 3.1 T heoretical Constructs 3.1.1 Implicit vs. E xplicit L earning Learning a language can be either implicit or explicit. Young children learn their L1 in a manner as effortless as they grow. That is t hey not only do not have any goal s or motivation s f or learning, but also do not even realize they are learning. It seems that as long as you immerse children in the target language environment, they can automatically become a competent speaker in that language. No explicit instruction or correction is ne eded (Rober, 1989 ) is unconscious rather than conscious. In terms of complex grammar, children simply rely on their exposure to whatever linguis tic data is available around them in a natural setting, such as communication with adults or older children TV program s etc. Linguists working from an innate position claime d that children are born with a set of abstract principles and parameters that f orm the underlying grammars of all natural

PAGE 51

51 languages Therefore they can completely master the language with incomplete input in the language data regardless of whether it is well formed or not (Chomsky, 2001 ) In contrast, l inguists working from an interactionist position disagre e that UG is the only d ; though they admit that children are born with some s ort of innately specific knowledge they argue that it interacts with the environment and input (Gass, 1997 ) Despite these discrepanc ies acquisi W hat works perfectly for children may not work equally well for adults. Compar ed to oriented, and motivated by various reasons. T hey usually need to acquire rules in or der to master grammar ( at least certain features of grammar), instead of simply rely ing on unconscious exposure to natural linguistic data (R. Ellis, 1994 ). Thus learning a L2 can occur through either implicit learning or explicit learning. Implicit learning refers to owledge about the underlying structure of a complex stimulus environment by a process that takes place naturally, simply and without conscious individual makes and tests hypoth (R. Ellis, 1994 p. 1). The fundamental difference that sets these two types of learning apart is whether a learner is conscious of what is being learned (Dekeyser, 1998 ). Conscious here includes at least two meanings : the expressibility issue and the intuitive judgment issue. Expressibili ty is how well a learner can verbally express the goes to went

PAGE 52

52 knowledge but not declarative knowled ge, both of which are stored in their long term memory (Chamot & O'Malley, 1994 ). Thus t heir linguistic knowledge is more implicit than explicit. Intuitive judgment refers to the notion that the acquired information is for making grammaticality judgments in t (Rober, 1989 p. 230). In other words, when making judgments, learners do not bring the knowledge resulting from implicit learning to the awareness level, instead the judgments are intuitive. Krashen (1981; 1982; 1985 ) views implicit and explicit learning as two types of learning with contrastive features. He clearly distinguished language acquisition from language learning: acquisition is incidental a nd implicit, with no selective attention to features of input that feed into the learning process, which results in knowing the language; learning on the other hand, is intentional and explicit, with selective attention, which results in knowing about the language. Krashen argued that only implicitly acquired knowledge can be the basis of fluent production, which is particularly effective for complex structures. He claimed that the traditional explicit grammar instruction and error correction in second lan language development. Many researchers did not agree with Krashen, particularly about the role of explicit learning. They argued that explicit learning is effective in L2 acquisition, which was supp orted by empirical evidence from laboratory studies ( e.g., DeKeyser, 1995; Leow, 1998 ). Some of them believe that explicit learning plays a significant role in L2 acquisition, because it could help learners notice the gap between input and th eir own output (R. Ellis, 1994 ) through the explicit teaching of grammar, explicit error correction,

PAGE 53

53 or input enhancement (Norris & Ortega, 2000 ). However, they think the ro le of practice is only limited to item learning and fluency improvement. Some researchers believe that explicit learning works effectively in L2 acquisition through systematic practice (e.g., DeKeyser, 2003; Schmidt, 1990, 1994, 1995; Schmidt & Frota, 1986 ; Swain, 1985), language use (DeKeyser, 2003, p. 328). 3.1. 2 I mplicit vs. E xplicit K nowledge There are two types of linguistic knowledge internalized in learners mind : impl icit ciously aware of (R. Ellis et al., 2006, p. 340). Implicit knowledge is knowledge of language which can be further distinguished into two types: formulaic knowledge which is internalized units. For instance NS usually know a large number of fixed or semi fixed expressions. Another type is rule based knowledge, which facilitate s learners to produce novel sentences. On the other hand, explicit knowledge is knowledge about the language which is analyzed abstract, and explanatory (R. Ellis, 1994) Learners usually are able to verbalize explicit knowledge if called upon. However, this knowledge is not dependent on meta linguistic knowledge (Bialystok, 1994 p.567). Implicit and explicit knowledge distinguished from each other mainly in the fol lowing areas: awareness, type of knowledge, systematicity and certainty of L2 knowledge, a ccessibility of knowledge, use of L2 knowledge, self report, and learnability (R. Ellis, 2005) which are summarized in Table 3 1. Implicit knowledge is intuitive, in the sense

PAGE 54

54 (R. Ellis, 1994, p. 85) On the other hand, explicit knowledge is tances of its use), abstract (in the sense that it takes the form of some underlying generalization of actual linguistic behavior) and explanatory (in that the logical basis of the knowledge is (R. Ellis, 1994, p. 85) Implicit knowledge involves automatic processing. It is usually not verbali zable, and some argue that certain aspects of it can never be learned after a certain age. On the other hand, explicit to achieve control in demanding and that involves controlled processing. It is potentially verbalizable, and can be learned at any age (R. Ellis, 2005, p.150). Table 3 1. Key characteristics of implicit and explicit knowledge (R. Ellis, 200 5, p.151) Characteristics Implicit knowledge Explicit knowledge Awareness Intuitive awareness of linguistic norms Conscious awareness of linguistic norms Type of knowledge Procedural knowledge of rules and fragments Declarative knowledge of grammatical r ules and fragments Systematicity Variable but systematic knowledge Anomalous and inconsistent knowledge Accessibility Access to knowledge by means of automatic processing Access to knowledge by means of controlled processing Use of L2 knowledge Access t o knowledge during fluent performance Access to knowledge during planning difficulty Self report Nonverbalizable Verbalizable Learnability Potentially only within critical period Any age Although l inguists generally recognize the important role played by implicit knowledge in L2 acquisition, they dispute about the role of explicit knowledge In

PAGE 55

55 addition, the relationship between the two types of knowledge has also been much debated. These controversies resulted in very different pedagogical recommendati ons T here are three major positions: the non interface position, the weak interface position, and the strong interface position. non interface model c ompletely denies any association s between implicit and explicit knowledge He believe s that these two types of knowledge have their independent mental processes and storage E xplicit knowledge can never be convert ed into implicit knowledg e and vice versa. He claims that implicit and explicit knowledge are caused by different acquisitional mechanisms Implicit knowledge is the result of acquisition whereas explicit knowledge is caused by learning Learned competence does not become acquired competence The results of learning can never lead to implicit knowledge T herefore the role of L2 i nstruction should really be to provid e large quantities of comprehensible input to foster implicit learning, not to provid e e xplicit rules and systematic practice of these rules. The acquired implicit knowledge is used by learners in producing utterances; whereas the learned explicit knowledge only functions as a m onitor to edit the utterances produced by acquired knowledge which is known as the Monitor Hypothesis (Krashen, 1981; 1985 ) position was criticized for not operationable as it failed to provide precise definitions of the two critical constructs in this model: acquisition and learning (e.g., DeKeyser, 1997; Gregg, 1984 ). In contrast, DeKeyser (1998) in the strong interface position claims that explicit knowledg e can not only be derived from implicit knowledge, but also be co nverted into imp l icit knowledge through intentional practice He claims that L2 knowledge is learned by learners as declarative knowledge first, and then changed into procedural knowledge

PAGE 56

56 thr ough communicative practice. Declarative knowledge is defined by DeKeyser as repeated target behaviors On the other hand, procedural knowledge is defined as behavior inv action pair s that state what is to be done under certain The strong interface position recommended that grammar rule should knowledge such as let learners practice in communicative tasks In the weak interface model, which was developed by R. Ellis (1993; 1994 ) explicit knowledge is considered to play an important role in L2 acquisition, particularly facilities learners to notice some specific linguistic features of t he input through comparing their own output and the input. This position allows two way conversions between implicit and explicit knowledge However, these conversions are only allowed under certain stringent conditions: explicit knowledge can be converte d into implicit knowledge when learners are developmentally ready to acquire a new form; implicit knowledge can also be converted into explicit knowledge when explicit knowledge can be pr oposes that explicit and implicit knowledge can be taught separate ly: e xplicit knowledge can be taught through consciousness raising tasks whereas implicit knowledge can be taught in task based activities. 3.1 3 Implicit vs. E xplicit F eedback In SLA, n egative feedback is a mechanism that responds to errors made by learners during their practice. It utterance s (Leeman & Martinez, 2007 ). Negative feedback can be either implicit or explicit, depending on how an error was indicated: when the error was c overtly

PAGE 57

57 indicated, the feedback is considered implicit; when the error was overtly indicated, it is considered explicit (Ellis et al., 2006 ). In the previous studies the most prevalent form that has been taken for implicit feedback is recasts (e.g., Ammar & Spada, 2006; Carroll, 2001; Carroll & Swain, 1993; R. Ellis, 2007; Iwashita & Liem, 2005; Kang, 2009; Leeman, 2003; Long, 1996; Lyster, 2004; Muranoi, 2000; Nassaji, 2007; Sheen, 2007 ). Besides recas ts, implicit feedback was also operationalized in the forms of clarification request s (e.g., Loewen & Nabei, 2007 ; McDonough, 2005 ) On the other hand, explicit feedback has been operat ionalized in a variety of forms. Some of the studies involve only minimal explicit feedback, such as expl icit feedback in a n implicit condition (Rosa & Leow, 2004 ) negative evidence (Leeman, 2003), repetition (e.g., Kang, 2009 ; Lyster, 2004 ) or elicitation of the correct form ( e.g., Lyster, 2004 ; Ammar & Spada 2009 ) Others involve more detailed explanation such as explicit rejection (e.g., Carroll & Swain, 1993; Kim, 2008 ), meta lingui s tic feedback (e.g., Ammar & Spada, 2006; R. Ellis et al., 2006; Kim & Mathes, 2001; Loewen & Nabei, 2007 ; Lyster, 2004 ), and explicit correction (DeKeyser, 1993 ) Reca st s generally refer to a a minus the error (Lyster & Ranta, 1997 ). Long provided a more detailed definition, in which a recast is defined as preceding utterance in which one or more nontarget like (lexical, grammatical, etc.) item(s) is/are replaced by the corresponding target language form(s), and where, throughout the exchange, the focus of the interlocutors is on meaning, not language as (Long, 2007 p.77). high light ed several i mportant characteristics

PAGE 58

58 of recasts which set them apart from other types of negative feedback : first, the correction s in recasts are done in a n implicit and incidental manner, with no clear indicator of the source of errors Therefore learners have to i nfer by themselves that their utterances are problematic (Carroll & Swain, 1993). Second the juxtaposition of may enhance the salience of the target form and promote learner noticing, w hich provides the learner with an ideal opportunity to make a cognitive comparison and to notice the gap between the ta r get like and non targetlike forms (Dought y & Pica, 1986 ; Long et al. 1998 ). Third, s ince corrections are embedded in a context, the flow of communication will not be interrupted (Long, 1996, 2007). As a result, recast s can serve two functions: a pragmatic function that keeps the communication c hannel open, and a corrective function that indicates that an error was made by the learner. Thus the function of recasts can often be ambiguous as it can be used both for providing and seeking confirmation and for additional information (Lyster, 1998b ). In fact, classroom studies showed that corrective recast s were more likely to be interpreted by learners as non corrective repetition s during the discourse in SLA classrooms, as both of these two types of feedback share the same pragmatic functions (Lyster 1998a). Clarification request s can be used for problems in either comprehensibility or accuracy, and include phrases such as pardon me what do you mean by X ? (Lyster & Ranta, 1997). Similar to recasts, the corrective inten t ion of clarification request s is also not so easily noticed because of the dual purpose it serve s : clarifying meaning as well as prompting learners to self repair their problematic utt erances. Therefore, it is also considered more implicit than explicit (Loewen & Nabei, 2007).

PAGE 59

59 Explicit feedback in a n implicit condition was operationalized by Rosa and Leow (2004) as a simple indication of whether the answers made by learners are correct or not N egative evidence, which was was operationalized p. 49). With r epetition teachers Elicitation in contrast, can include three techniques: first, using ; second, pausing to allow the student to complete th e utterance initiated by the teacher; third, ask ing the learner to reformulate the utterance (Lyster & Ranta, 1997 ). (1993, 2001) learners in an explicit hypothesis rejection group were told that they made an error, followed by an explicit explanation of th at error Meta linguistic feedback has been frequently compared with recast s regarding the impact of each on L2 development (e.g., Ammar & Spada, 2006; Carroll & Swain, 1993; R. Ellis et al., 2006; Kang, 2009; Kim & Mathes, 2001; Sauro, 2009; Sheen, 2007 ). The former to the well (Lyster & Ranta, 1997 p. 47). As the correct linguistic form is not provided in the meta lingui s tic feedbac k, learners are pushed to produce the correct form by themselves based on the provided meta linguistic information. The above dichotomy of implicit and explicit feedback has been questioned by researchers (e.g., R. Ellis et al. 2006; Loewen and Nabei, 2 007), who argued that the line between these two types o f feedback is not always clear: to what extent should

PAGE 60

60 feedback be overt enough to be considered explicit? H ow covert must feedback be to be categorized as implicit? For instance, a lthough recasts are generally considered implicit, not all of them are as implicit as is generally assumed I n fact some of them are quite explicit especially when saliency cues (e.g., phonological emphasis) are involved ( R. Ellis et al. 2006) These cues al intonational or verbal cues or (Nassaji, 2007 p. 526). For instance, i phonological stress was employed to indicate the source of the problem. R esearchers have proposed that instead of adopting the dichotomic distinction of implicit and explicit feedback, the explicitness of negative feedback should be considered as constituting a continuum as shown in Figure 3. 1 (based on Loewen & Nabei 2007, p. 362). Loewen and Nabei proposed a feedback classification scheme that classifies feedback on two cr i t e ria: ( a ) provision of L2 models and pushed output and ( b ) explicitness of feedback. Based on the two criteria they divided s ix commonly employed f eedback types in the classroom (Lyster & Ranta, 1997) into two sub groups: prompts ( e.g., Ammar & Spada, 2006; Lyster, 2004; Lyster & Mori, 2006) and provide (Loewen & Nabei, 2007). The prompt group include s meta lingui s tic feedback elicitation, repetition, and clarification request s These four types of feedback share one characteristic in common: they all prompt learners to produce output without providing L2 models In contrast, the provide group consist s of two types of feedback: explicit correction and recast s As the name provide suggests, both types of feedback provide

PAGE 61

61 models of the target form, therefore learners will not be pushed to produce output as they do with prompt Figure 3 1 The explicitness of the feedback (Loewen & Nabei, 2007, p.362. Figure 15.1) T he subtypes of feedback under prompts and provide are ordered according to explicitness, as shown with horizontal double sided arrows in Figure 3 1. Here, explicitness is defined in terms of the identification of the error that triggered feedback. Meta linguistic feedback is consider e d the most explicit form of prompt type feedback because it explicitly identifies the source of the problem that led the NS to provide feedback. In contrast, clarification requests are on the implicit end of the continuum nder the term Prompt Explicit Implicit Explicit Correction Recast Metalinguistic Elicitation Repetition Clarification Feedback request More Mordified output Less Provide

PAGE 62

62 provide recast is closer to the implicit extrem e because it does not explicitly suggest that an error was made. In contrast, explicit correction is on the explicit end of the continuum because it clearly indicate s that what the student has s aid is incorrect (Lyster & Ranta, 1997). 3.2 Measuring E ffectiveness of C orrective F eedback The effectiveness of corrective feedback on L2 acquisition has been assessed using a range of measures in the SLA literature. These measures can be roughly divided into four types: 1. uptake and learner repair (e.g., R. Ellis et al., 2001; Lyster & Ranta, 1997; Mackey & Philp, 1998; Panova & Lyster, 2002 ); 2. p rivate speech (Oh ta, 2000 ); 3. lea r ner noticing of feedback (Gass & Lewis, 2007; Mackey et al., 2000; Philp, 2003 ); 4 test performance (e.g.,Carroll & Swain, 1993; Leeman, 2003; Sheen, 2007 ). 3.2.1 Uptake and L earner R epair Uptake refers to response immediately follow ing a s corrective feedback (Loewen, 2004 ; Lyster & Ranta, 1998) This can be further subcategorized into two types according to the results of the correction: repair or needs repair (Lyster and Ranta, 1997). Repair n a Needs repai r indicates that the utterance still contains an error and includ e s lingui s tic feedback), same error, different error, hesitation, partial repair, and off target. R. Ellis et al. (2001) employed two different terms to name the subgroups with similar definitions: successful uptake and unsuccessful uptake Thus Uptake relates to the construct of modified output. It has been employed to examine the effects of feedback in general, and recasts in particular.

PAGE 63

63 Under this approach, feedback followed by a high rate of learner uptake is considered effective, whereas th at with a low rate of learner uptake is viewed as being less effective or ineffective Lyster and Ranta (1997) in their influential study investigated the relationship between corrective feedback and learner uptake in four French immersion classrooms. They found that although recast is the most frequently employed feedback technique in the classroom, it leads to the least uptake compared with other types of feedback, such as meta linguistic feedback, elicitation, and repetition. Therefore they concluded that recast is less effective than the other types of feedback in more explicit forms In a n ESL classroom based descriptive study Lo e wen (2004) found a higher rate of learner uptake compared with the study d one by Lyster and Ranta ( 1997 ) He attributed the different results to the different learning contexts employed in the two studies. He argued that in ESL classroom s L2 learners may be more ready to produce uptake as they tend to focus on linguistic forms even in meaning focused activities, which contrast s with in immersion or content based classes, wh ere students tend to attend to con tent and meaning. Therefore he argued that characteristics of the learning context, such as age and the previous L2 language exposure of learners, as well as classroom uptake Sheen (2006) found that learner uptake f ollowing recasts also significantly relate s to various characteristics of recasts, including the length of recasts, the linguistic focus (e.g., pronunciation vs. grammar), the type of change (e.g., substitution vs. addition) the mode of utterance (e.g., d eclarative vs. interrogative), and the number of changes (e.g., one vs. multiple).

PAGE 64

64 Moreover, uptake is an optional discourse move ; it is up to a learner to produce uptake or not to produce any uptake at all ( R. Ellis et al. 2001). Therefore when learne rs do not produce any uptake it does not necessarily mean that the feedback does not have any impact on them. Sometime s learners are just not given enough opportunities for uptake after recasts in ESL classrooms (Braidi, 2002; Oliver, 1995; 2000 ). Some researchers questioned whether uptake is a reliable measure of noticing, partic ularly for recasts. For instance, L ong (2007) argued that recasts and feedback of a more explicit nature were designed with different instructional foci: recasts can be either interpreted as a corrective feedback, or a continuous topic, therefore they are more ambiguous. In contrast, explicit feedback ( e.g., meta linguistic feedback, elicitation, repetition) aims at pushing learners to produce more output. Thus explicit feedback tends to elicit more uptake than recasts, which does not necessarily support th e argument that learners notice fewer recasts than explicit feedback. 3.2.2 Private S peech (2000 ) longitudinal study examined the effectiveness of recasts through private speech which was recorded with small tape r ecorders and individual microphones clipped to seven learners of Japanese over an academic year. In her study, private speech was defined as an adaptation to an interlocutor, and lack of response from an interlocutor P rivate speech development and cognitive development are mutually constitutive and proceed. For L2 learners, private speech serves as a cognitive tool for the internal ization of the L2, and

PAGE 65

65 as a productive, hypothesis testing device whereby learners think aloud as they work to solve L2 problems in contrast to studies conducted within an interactioni st framework, this study employed a qualitative discourse analytic approach in its data analysis, which enable d the research er to examine the efficacy of recast s from a n emic rather than etic perspective. The researcher used Conversation Analysis with spec ific attention to overlap, pauses, as well as the leve l s of volume of learners utterance s The collected data were viewed in a n holistic fashion. Second, besides the learner who was addressed by the teacher and received the recasts, the study also looked at the effect s of recasts on learners who were not being addressed, but participated in the class as auditor s In fact, they were the one s who produced private speech mo re often than the learners who received the recasts did. The private speech produced by the auditors also provided counter evidence to the notion that recasts are ineffective since t hey elicit ed fewer responses from the learners being addressed In this study, the auditors were found to actively participate in the class through private spe ech As argued by Ohta provides powerful evidence of the mental activity triggered by the noticing of contrasts between ill formed and correct utterances. The efficacy of recasts should not be doubted based on the presence or absence of an (p.66). A problem of employing private speech to examine the effectiveness of recasts is tha t learners may not be equally talkative. In fact, among the seven learners who participated in the study, only four of them produced private speech with moderate to

PAGE 66

66 study, in which each learner could be observed over long spans of time. On the other hand, if private speech was used in a cross sectional study a theoretical triangulation (using multiple perspectives to analyze the same set of data), or a stimulated recall could be helpful in capturing the impact of recasts on learners who do not produce private speech frequently, but participate in the class eq ually actively in their mental process ing (Mackey & Gass, 2005). 3.2.3 Learner N oticing of F eedback Many researchers argue that c orrective feedback is facilitative in L2 acquisition because it connects several mechanisms in the process, including input, internal learner capacities, output, and selective attention of learners (Long, 1996). Among these mechanisms, attention is a limited capacity system, which is required for all learning (Schmidt, 1995, Tomlin & Villa, 1994) ne way in which the input becomes more manageable is by the learner focusing attention on a limited and hence .8). Tomlin and Villa (1994) distinguished attention into three functions: alertness, orientation, and detection. They claimed that detection processing of a stimulus at higher levels, such as storage and rehearsal in short term 296), which does not necessarily involve awareness. In contrast, Schmidt (1990, 1993, 1995) argued that awareness at the level of noticing is crucial for learning. In his well known Noticing Hypothesis (Schmidt, 2001 ), Schmidt claimed that learners must consciously notice input in order for it to become intake. Hence noticing is considered requisite for learning to happen (Schmidt, 1995). Noticing here refers to a

PAGE 67

67 stimuli in short et a l. 2000 474 ). Awareness in contrast, was In the SLA literature, various retrospecti ve methodologies have been employed in collecting noticing data, including diary studies (Schmidt 2001 ), questionnaires ), and uptake sheets (Warden et al., 1995 ). However, these methods ten d to take long spans of time. As cognitive processing is claimed to occur in a short period of time, more time efficient verbal reports have been employed, including think aloud protocols (e.g., Leow, 2000; Rosa & O'Neill, 1999 ), immediate recall (Egi, 2004 ), and stimulated recall (Adams, 1991; Mackey et al., 2000; Slimani Rolls, 2005 ). The verbal reports can roughly be categorized into two types based on their currency of verbalization: concur rent or retrospective (Adams, 2003; Egi, 2004 ). In previous studies, two types of verbal reporting have been used by many researchers in collecting noticing data related to recasts: immediate recall (e.g., Egi, 2004; Philp, 2003 ), and stimulated recall (Adams, 1991; Egi, 2004; 2007 ; Nabei & Swain, 2002 ). Immediate recall is a technique that can be immediately after the completion of t technique has been commonly used within cognitive psychology to access detection and rehearsal in short term auditory memory. Philp (2003) employed this technique in her study to investigate v ariou s constraints preventing learners from noticing the gap between their ill formed utterance s and the target forms after receiving recasts. In her study, learners wer e asked to repeat the last part of the recast they heard during a conversational turn immed iately after a recall prompt signal, which consisted of two

PAGE 68

68 immediate recall in her study, but with a different op e rationalization (2004). Instead of asking learners to repeat previous content from the recasts, she asked them to verbalize their thoughts about the language episodes after the recasts. Obviously higher level s of noticing are required for learners when being asked to verbalize their thoughts compared to simply repeating the limited items that appear in immediately preced ing recasts. Although the immediate recall might o because of the longer time learners spent on commenting about the recasts it might place a heavier burden on their working memories. In addition, as the knocking recall signals occurred frequently during the interaction, the flo w of normal conversation was consistently interrupted. In fact, the results showed that the average length of immediate recall protocols was significantly shorter than stimulated recall. As Egi noted, it may suggest that learners wanted to keep the interru ption to the communicative tasks as small as possible by trying to keep their comments concise. In s timulated recall participants are prompted to recall thoughts they had while performing a task or participating in an event. It is assumed that some tangible (perhaps visual or aural) reminder of an event will stimulate recall of the mental processes in operati o enabled to relive an original situation with great vividness and accuracy if he is presented with a large number of the cues or stimuli which occurred during the original

PAGE 69

69 In two descriptive studies (Gass & Lewis, 2007 ; Mackey et al. 2000), stimulated recall was employed to examine the perception s of L2 learners about feedback given on various linguistic domains including phonology, morphology, lexis, and semantics. In the two studies, a communicative task between a learner and a NS interviewer was video tape d, and played by a second interviewer to the same learner immediately upon the completion of the activity. When watching the video, the learner was asked to pause it anytime when she/ he wished to recall any thoughts she/ he had at any particular moment in the interaction. After each interactional feedback episode, the researcher also paused the video and asked the learner to recall his/ her thoughts at the time. The goals of the studies were to examine whether learners could correctly perceive feedback wit h various linguistic foc i and the differences between learners with different language background s (e.g., heritage vs. non heritage learners) in terms of perception s about feedback. There are several benefits in employing stimulated recall immediately after the activity to examine the perception s of learners c on cerning recasts and other feedback technique s : first, it increases the likelihood that learners trace their thoughts from short term memor y As the memories are still fresh, they should be more close to the real thoughts that learners had at the moment of interaction. If the stimulated recall was conducted with a longer delay following the activity learners may tend to say what they think the researchers want to hear because the event is not sha rply focused in their memories anymore. Second, the use of the video recording serves as a strong stimulus

PAGE 70

70 On the other hand, concerns have been raised questioning the validity and reliability of the method (for a review, see Gass & Mackey, 2000). Validity includes reflection of cognitive processes; protocols are non veridical if they fail to cap S timulated recall is considered valid under the assumption that the cognitive processing of learners is directly accessible and available through their verbal reporting. However, do the verbal comments made by participant s truly reflect the thought processes of participants? Researchers such as Nisbett and Wilson argued that since conscious awareness can only relate to the products of mental proces ses, and the processes themselves cannot be reached through introspection, argument agains t Nisbett and Wilson is that the thought processes should be accessible through stimulated recall (Gass & Mackey, 2000 p.107). A second concern is reactivity which refers to V erbal results of previous studies with regard to the reactivity of stimulated recall are conflicting. For inst ance, verbalizations were found either to decrease or improve (Leow & Morgan Short, 2004 ; Rosso et al., 1989 )

PAGE 71

71 As pointed out by Gass and Mackey, w ith regard to reliability, previous studies verbalization of informa (2000, p.109) Therefore verbal reports that are understanding of the circumstances under which they were obtained, are a valuable and thoroughly reliable sort of i 2000, p.109 ). In addition to the validity and reliability of the instrument, several things could be done to improve the validity of the se stud ies First of all, external validity could be increased by recr uiting a larger sample size. Second, internal validity could be increased by controlling the proficiency level of the subjects through a proficiency test, such as an Oral Proficiency Interview (OPI) before the data collection. As in Mackey et al. the number of years of prior study of English varied from as short as three years, to as long as 14 years ; it is therefore hard to tell whether the results of the study would be valid for learners with various proficiency level s or only for learners with relatively high proficiency level s given that half of the participants in the ESL group ha d solely on their course enrollment, which is also problematic as I argue in C hapter 2 Both stimulated recall and immediate recall are important methods in eliciting noticing data on recasts and other corrective feedback. The advantages and disadvantages of the two methods were closely examined by Egi in her experiment al study (2004). She argued that immediate recall has an advantage over stimulated recall

PAGE 72

72 is still fresh in a terms of practicality, immediate recall requires a stimulated recall is super ior t o immediate recall in the following areas. First, the interlocutor in the interactional activity and the researcher in the stimulated recall session can be served by different researchers, thereby reduc ing the pressure on the subjects to have to avoid reporting any negative comments about the interaction to the same researcher who conducted the activity. Moreover, the researcher can clarify any uncertainties and doubts about stimulated recall is not intrusive since it is carried on after the completion of the oral interaction. Lastly, I also find that stimulated recall requires less training prior t o data collection compared with immediate recall for both the researchers and the participants. 3.2.4 Test P erformance In c lassroom based observational resear ch, as various variables are likely to be present within the same context, it is hard to conclude whether a cause and effect correlation exists between the variable under study and the outcome. In contrast, in experimental settings, variables can be tightl y controlled and manipulated. In addition subjects can also be randomly assigned to different experimental groups. Therefore the experimental approach has an advantage over classroom based observation in determi ni ng the correlation between the tested vari able and the outcome (Mackey & Gass, 2005). L effectiveness of recasts and other feedback type s in experimental studies employing an immediate post test (e.g., Loewen & Nabei, 2007; McDonough, 2005; Russo et al., 1989 ) or both immediate and delayed post test s (e.g., Ammar, 2008; Carroll & Swain,

PAGE 73

73 1993; Lyster & Izquierdo, 2009 ) In these studies, a pre test wa s given to the participants in order to ensur e group comparability prior to the treatme nt (Mackey & Gass, 2005) The short term effects of treatment were measured by a immediate post test whereas the long term effects were measured by a delayed post test A wide range of tasks have been employed for tests, including grammaticality judgment tests, sentence completion, picture prompt tests, translation tests, oral imitation tests, picture description tests, sport the differences tasks, jigsaw tasks, consensus tasks, ordering tasks, and consciousness ra ising tasks (for a review see Mackey & Gass, 2006 ). The target linguistic structures also vary. The most frequently examined ones ha ve been syntactical, morphosyntactical, and morphological, such as dative verbs (Carroll & Swain, 1993; Kim & Mathes, 2001) verbal predicates and particles (Nagata, 1993), indefinite article s (Muranoi, 2000), derivations of nouns from verbs (Carroll, 2001), noun adjective agreement (Leeman, 2003), pronouns (Sanz, 2003), gender (Lyster, 2004), verb past tense ( R. Ellis et al., 2006; R. Ellis, 2007), and question forms (Loewen & Nabei, 2007) The target structures in these studies have different degree s of complexities in terms of and operation al burden s Some structures might only require simple explicit kn owledge (Leeman, 2003), while others, such as question forms (Loewen & Nabei, 2007), may require learners to acquire more complex rules T he designing of tests is highly related to validity and reliability of the final results. The t ests employed in previ ous studies often more likely involv ed grammaticality judgment tests, sentence completion s picture prompt test s and translation tests, etc. All of these tests favor explicit knowledge rather than implicit knowledge. Therefore

PAGE 74

74 studies which employed these types of tests ha d a higher likelihood of finding explicit feedback to be more effective than implicit feedback ( R. Ellis, 2007). I n order to examine the effect of feedback, employing tests that t ap the implicit knowledge of learners is thus also critical For instance, an elicited imitation test can be used to designed to manipulate ce the sentences accurately is a (Mackey & Gass, 2005). Caution should been used in designing appropriate instruments to test what the researc h intends to test (Mackey & Gass, 2005). For instance, Han (2000 ) suggested that grammaticality judgment tests suffer a reliability dge of grammatical rules. The results found in his study indicated that learners used different types of knowledge under different conditions. For instance, they tend to use their implicit knowledge under time pressure, but explicit knowledge when adequate time is given. For linguistic forms with highly complex and difficult structure s they tend to cope with intuitions as well as various strategies. However in this small scale study only one type of grammaticality judgment test was employed Han urged res earchers to be careful before making any conclusion based on grammaticality judgment tests. Thus employing multiple tests that examine different types of knowledge is crucial in test design Section 3.3 will introduce several studies that employed multipl e tests.

PAGE 75

75 3.3 Previous Empirical Research Comparing the Effects of Implicit and Explicit Feedback Numerous studies have been done in the past comparin g of the effects of different types of corrective feedback on L2 acquisition. Generally the result s of these studies showed that both implicit and explicit feedback facilitate L2 acquisition. However, in terms of which type of feedback has more advantage s i n L2 development, the results were mix ed Since t hese studies were conducted in a wide range of s ettings (e.g., laboratory, classroom, or computer based interaction s ), and using a variety of measures (e.g., grammaticality judgment tests, sentence completion, translation tests, etc), and treatment tasks (e.g., communicative tasks, and mechanical exerci ses) it is not wise to rush to a conclusion ab o ut which type of feedback has a greater impact on L2 acquisition without carefully considering the substantial differences among these studies. For instance, researchers argued that the settings of the inter action affect the effects of feedback, particularly where recast s are considered In the classroom context, recasts are provided in an interactional context (Lyster, 1998b), therefore it may be more challenging for learners to perceive the corrective inten tions of recasts as they would in a laboratory setting. Lyster (1998b; 2007 ) found that recasts might not function as effectively as other types of feedback in classroom settings that are mor e meaning oriented than form oriented. In laboratory settings, variables can be easily controlled, allow ing for feedback to be delivered in an intensive manner focusing on the target linguistic forms. T his se ct ion will focus on 1 8 previous studies th at co mpar ed the effects of implicit and explicit feedback conducted in three different settings: laboratory (Carroll & Swain, 1993; Kang, 2009; Kim & Mathes, 2001; Leeman, 2003 McDonough, 2005,

PAGE 76

76 s ummarized in Table 3 2), classroom (Ammar & Spada, 2006; DeKeyser, 1993; R. Ellis et al., 2006; Havranek & Cesnik, 2001; Loewen & Nabei, 2007; Lyster, 2004; Lyster & Izquierdo, 2009; Muranoi, 2000; Sheen, 2007 summarized in Table 3 3), and computer based inte raction (Nagata, 1993; Rosa & Leow, 2004; Sanz et al., 2009 summarized in Table 3 4). The review will particularly addr ess the test design and results in these studies. The five laboratory studies all incorporated recasts as the form for implicit feedback. Leeman (2003) contrasted these with explicit negative evidence McDonough (2007) was the only researcher who also use d request s for clarification as an implicit feedback type. The explicit feedback also took several different forms in the studies, including meta linguistic feedback (Carroll & Swain, 1993; 2007; Kang, 2009), explicit rejection (Carroll & Swain, 1993) and enhanced silence (Leeman, 2003). Treatments in these studies generally involved communicative tasks, which were conducted one on one between the interlocutor and the learner. All the studies employed pre and post test designs, three studies also used dela yed post test s (Carroll & Swain, 1993; Leeman, 2003; Kang, 2009). The tasks of the tests varied, but only one study employed multiple tasks during the test session (Kang, 2009) the other three only used one task. The results of these studies were mixed. Carroll and Swain (1993) compared the effects of four types of feedback: explicit correction, direct meta linguistic feedback, indirect meta linguistic feedback, and recasts. The first two are more explicit, whereas the last two are more implicit. A recall task was employed in measuring the effects of the feedback, in which learners were asked to recall the linguistic items they saw during the treatment sessions. There were

PAGE 77

77 two recall sessions: the first recall session was conducted immediately after the fe edback session, whereas the second one happened one week later. All the experimental groups did significantly better than the control group on both recall sessions, which indicated a learning effect of all the feedback types tested in this study. However, only the direct meta linguistic group, in which participants were told that an error was made and were provided with an explicit semantic or phonological explanation, significantly outperformed the other experimental groups. There were no significant diffe rences among the other three experimental groups. This study was replicated by Kim and Mathes (2001). However, they only compared two types of negative feedback: the explicit negative feedback group with explanation, and the modeling plus implicit negative feedback group. The results showed no significant differences between the two feedback groups. Leeman (2003) did not intend to compare the effects of implicit and explicit feedback, rather she was interested in the role played by negative evidence in re casts. Therefore she manipulated positive and negative evidence in the feedback employed in her study, and employed four types of feedback: recasts, negative evidence, enhanced salience of positive evidence, and unenhanced positive evidence. Among the four types of feedback, recasts were relatively more implicit, whereas the other three were more explicit. The test used in this study was a picture description task, in which learners were given two digitally altered photographs and were asked to work individ ually to find differences between the two photographs. The findings showed no significant differences between the four experimental groups. In addition, only the recast and enhanced salience groups significantly outperformed the control group. McDonough

PAGE 78

78 (2 007) used a narrative task to measure the effects of two types of implicit feedback given on the emergence of English simple past activity verbs: recasts and requests for clarification. The results showed no significant differences between the two feedback groups, al though both experimental groups did significantly better than the control group. However, the post hoc analysis showed that the clarification requests group used significantly more progressive activity verbs than the recast group. The findings s stages simultaneously, whereas recasts may have a more concentrated impact on a Kang (2009) employed two tests in assessing the effec ts of implicit and explicit feedback on 34 English speaking learners of Korean as a less commonly taught foreign language in the U.S: an untimed grammaticality judgment test and a picture description test. Implicit feedback took the form of repetitions and recasts, whereas the explicit feedback took the form of meta linguistic comments, plus an alternative form or solution to the error. The results showed no significant differences between the two experimental groups, although both of them significantly out performed the control group. In classroom settings, recasts have been frequently used in the comparison of the effectiveness of prompt type (which includ e s clarification requests, repetition of learner error, meta linguistic cues, and elicitation). Among these feedback types, clarification requests are more implicit, whereas the other three types of prompts feedback are more explicit. The following nine classroom studies were conducted across a spectrum of classroom settings, including immersion classrooms (Lyster, 2004), high schools (DeKeyser, 1993), postsecondary EFL classrooms (Havranek & Cesnik, 2001; Kim &

PAGE 79

79 Mathes, 2001; Muranoi, 2000; Sheen, 2007), intensive elementary EFL classrooms (Ammar & Spada, 2006), and postsecondary French L2 classrooms (Lyste r & Izquierdo, 2009). Among them, six studies found that implicit feedback is less effective than more explicit feedback types (Ammar & Spada, 2006; R. Ellis et al. 2006; Havranek & Cesnik, 2001; Lyster, 2004; Muranoi, 2000; Sheen, 2007). The other four f ound no differences between the different feedback groups (De K eyser, 1993; Loewen & Nabei, 2007; Kim & Mathes, 2001; Lyster & Izquierdo, 2009). Havranek & Cesnik (2001) collected, in total, 1, 700 instances of corrective feedback data from L2 learners with a wide range of age and proficiency levels, out of which 12 types were identified as the most frequent combinations. The four types of correction analyzed in this study were: recasts with repetition, recasts without repetition, successfully elicited s elf correction, and explicit rejection before recasts without learner repetition. It is the only study among the 18 studies reviewed in this se ct ion that examined the effects of feedback not only on learners who received it, but also on their peers who obs erved the feedback happening. The study employed a variety of task types, including written and spoken completion tasks, translation, correction, and read aloud tasks. Elicited self correction turned out to be the most efficient combinations for all learne rs, whereas recast without repetition was the least effective combination. R. Ellis et al. (2006) compared the effects of two types of feedback recasts versus meta linguistic information -among three classes of lower intermediate EFL learners. Three differ ent tasks were employed, including an oral elicited imitation test, an untimed grammaticality judgment test, and a meta linguistic knowledge test. For the first two tests, grammatical and ungrammatical items were examined separately in order to

PAGE 80

80 measure dif ferent types of knowledge: ungrammatical sentences provided a stronger measure of explicit knowledge, whereas grammatical sentences provided a measure of implicit knowledge. The meta lingui s tic feedback group did significantly better than the recast group on the oral imitation and grammaticality judgment tasks during delayed post test s, which were conducted two weeks after the treatments. Sheen (2007) also compared the effects of recasts and meta linguistic feedback. The tests employed in this study include d a speeded dictation test, a writing test, and en error correction test. All these tests favored explicit knowledge over implicit knowledge. The results showed that the meta linguistic group significantly outperformed the recast and control groups on bo th the immediate post test and the delayed post test However, the recast group did not perform significantly better than the control group. Both of the following studies compared the effects between prompts and recasts on 118 participants who received fi ve weeks of form focused instruction (FFI) in a French elicitation, repetition, meta linguistic feedback, and clarification requests. In this study, all of these four type s of feedback were put under the prompts group, in comparison with a recast group. The study employed two oral tests and two written tests. The oral test included an object identification test and a picture description test, whereas the written test consis ted of binary choice and text completion t ask s. The results showed that the prompts group significantly outperformed the control group on all measures. The r ecast group significantly outperformed the control group in some of the measures Prompts in Ammar elicitation, repetition, and meta linguistic feedback. In this study, the three types of

PAGE 81

81 feedback were put under the prompts group, in comparison with the recast group. The study was conducted in ESL classrooms over a four week period. Two tasks types were employed in the study, both of which required simultaneous focus on form and meaning: one was a written meaning oriented contextualized grammaticality judgment task, in which learners were asked to identify all errors in a story and provide the correct alternatives. The other one was an oral picture description task, in which learners had to describe a set of pictures. Both experimental groups significantly outperformed the control group on all t he measures. The prompts group significantly outperformed the recast group on all the tests. In addition, Ammar and Spada also found that prompts were especially effective for learners who had lower scores on the pre test, whereas learners with higher scor es on the pre test benefited similarly from both recasts and prompts. Muranoi (2000) examined the impact of interaction enhancement (IE). This refers to a communicative instructional technique, in which the interaction is enhanced by means of feedback (e .g., requests for repetition and recasts), followed by two types of debriefings, during which the instructor reviews student performance in class in the debriefing group, with explicit grammar explanation; (B) IE + meaning focused of accuracy in communicating the message, not in terms of accuracy of the target forms. The study employed four tasks: an oral story description task, an oral picture description task, a written picture description task, and a grammaticality judgment test.

PAGE 82

82 On the delayed post test Muranoi found that group A who received a formal debriefing, significantly outperfor med Group B, who received a meaning focused debriefing. Loewen and Nabei (2007) employed two tests in measuring the effects of three types of feedback: recasts, request for clarification, and meta linguistic feedback. The results showed a learning effec t of all the feedback types, but only on a timed grammaticality judgment test, not on untimed and oral production tests. And no significant differences were found among the three feedback groups. DeKeyser (1993) studied 35 Dutch high school seniors who we re learning L2 French in two classes taught by two different teachers. One teacher was asked to correct mistakes made by learners as frequently and explicitly as possible, whereas the other was asked to avoid error correction as much as possible. The treat ments lasted a full school year with no particular forms targeted. DeKeyser conducted three oral tests and one written test twice, once at the beginning, and once at the end of the school year. No statistical differences were found between the two groups o f students. Lyster & Izquierdo (2009) investigated the effects of prompts and recasts, in a classroom + laboratory setting, on the acquisition of grammatical gender by 25 intermediate level French L2 learners. Prompts in this study were operationalized as clarification requests followed by a repetition. The classroom treatment was held during a two week period for all participants by the same instructor in a form focused instructional unit, in which a workbook was designed to target a set of French gender endings. It was followed by a laboratory feedback treatment, in which participants completed three different oral tasks. Either recasts or prompts were given to the participants on their gender related errors by the researcher, depending on the

PAGE 83

83 feedback co ndition. This study employed three tests: an object identification task, a picture description task, an oral imitation test, a computerized binary choice and reaction time test. The results showed no significant difference between recast and prompt groups for any of the measures. The following four empirical studies attempted to address the effectiveness of computer assisted language learning (CALL), particularly where the question of providing feedback is concerned. Researchers argued that feedback provi ded through attention to problems in their utterance (Nagata, 1993). Nagata (1993) compared two versions of Japanese computer assisted language instruction (CALI) exe rcises: traditional CALI (T CALI), and intelligent CALI (I CALI). T CALI not only pointed out what was wrong, but also provided de tailed meta linguistic explanations on the reason for the errors. Nagata found significant main differences between the two experimental groups in their achievement and retention for the particle errors, but not for verbal errors. Rosa and Leow (2004) ex amined the role of awareness in L2 development. In their study, explicit feedback was operationalized as a meta linguistic explanation on the linguistic form selected by learners. When the selection was correct, in addition to the provision of the reason w hy the choice was correct, learners also received an explanation of how the targeted structure works in target language. On the other hand, when the learner made an incorrect choice, the reason for the error was provided, followed by an explanation of how the targeted structure should work. Then the learner

PAGE 84

84 was instructed to try again and make a correct selection before moving to the next item. For the implicit condition, learners only received feedback that indicated whether the answer was right or wrong, which was similar to the implicit feedback employed in choice recognition tests and controlled production tests. In the multiple choice recognition test, learners needed to choose the correct answer from four options. In the controlled production test, learners were asked to complete sentences by providing an appropriate verb form. The results showed that learners who received explicit feedback reported higher levels of awareness of the target structures than those who received implicit feedback. In addition, learners who reported being aware of the target structure did significantly better than those who did not report being aware of it. Sanz (2003) employed computer delivered input processing instruction, in which gain employed in this study: a sentence completion test and a writ ten video retelling test. Although both groups did significantly better on the post test ; no significant differences were found between the two groups. Sauro (2009) compared recasts and meta linguistic feedback on the English zero article among 23 high intermediate/ advanced EFL adult learners. The treatment consisted of two computer mediated collaborative writing activities, in which the learner was paired with a native English speaking chat partner in individual chat rooms using the Virtual Classroom c hat tool from Blackboard. Feedback was provided by the chat partner whenever the learner made an error with the target form. The study employed a

PAGE 85

85 computer delivered acceptability judgment test to measure the effects of the feedback. The results showed no s ignificant differences between two feedback types for both immediate and delayed post test s. In sum, the 18 studies detailed revealed mixed results in terms of the effects of hat explicit feedback was more effective than implicit feedback, especially when the study was conducted in a classroom. On the other hand, results from most of the laboratory and CALL settings found no significant differences between implicit feedback and explicit feedback. Regardless of the setting in which the research was conducted, the treatment in most of these studies only lasted for a very short period of time. s tudy of the 18 reviewed here. No statistical differences were found between implicit and explicit feedback groups. This study reminded researchers that the time span of feedback may also affect the degree to which feedback is effective. Besides the setting s of the studies, the effects of feedback were also correlated with other factors. For instance, the effects of feedback were affected by linguistic forms. Nagata found at relatively more implicit feedback significantly outperformed more explicit feedback on particles, but not on verbal predicates, which may suggest that a linguistic form may found to be associated with the effectiveness of certain types of feedback in Ammar and learners with higher proficiency, however, both recasts and prompts seemed to work equally efficiently. This result provides some insights to the current study: sin ce learners

PAGE 86

86 with HL background usually tend to have higher proficiency compared with their non HL counterparts, would non HL learners also favor prompt s over recast s ? In terms of test design, some of the studies only employed tests that can be considered results showed an advantage of explicit feedback over implicit feedback. However, since the tests employed in this study only tested the explicit knowledge of learners, the validity of the results was questionable. In contrast, some of the studies employed multiple tests that can assess both implicit knowledge and explicit knowledge. The most representative example is Muranoi, who used two oral tests which measured implicit knowledge, a s well as two written tests which tested the explicit knowledge of learners. Unfortunately, he did not take advantage of these tests to really compare the effects of implicit and explicit feedback.

PAGE 87

87 Table 3 2 Laboratory studies of effects of implicit a nd explicit feedback Study participants Target structure Design Tests Results Carroll & Swain (1993) 100 Spanish adult ESL learners (low intermediate) Dative verbs Groups: (A) direct meta linguistic feedback, (B) explicit rejection, (C) recasts, (D) indirect meta linguistic feedback, (E) control. Treatment: two feedback sessions, each followed by recall Treatment as pre test Two recalls served as post and delayed post test All of the treatment groups significantly outperformed the control group on both recall tasks. Group A significantly outperformed other experimental groups. Leeman (2003) 74 university learners of L2 Spanish (first year) Spanish noun adjective agreement Groups: (A) recast, (B) negative evidence (source or p roblem indicated but not corrected), (C) enhanced salience with no feedback, (D) control. Treatment: Two communicative tasks: 1. Object placement task, 2. catalog shopping task. One on one with researcher. Design: Pre /Po st / delayed post test Task: picture description task. Only groups A and C outperformed the control group on all post treatment measures. No difference between A and C. McDonough (2007) 74 EFL university students (first year) Simple past ac tivity verbs Groups: (A) recasts, (B) clarification requests (e.g., huh? Pardon? What? A gain ?, ( C) control. Treatment: 1. two way information exchange, 2. One way information gap task, one to one with the researcher. Design: Pr e post test Task: a dream narration task Both groups A and B significantly outperformed group C, but no significant differences shown between these two groups.

PAGE 88

88 Table 3 2. Continued Study participants Target structure Design Tests Results Kang (2009) 34 learners of Korean (with Korean speaking parents) Korean past tense Groups: (A) explicit (meta linguistic comments, alternative form ), (B) Implicit (partial or full repetition or reformulation), ( C) control Treatment: 1. story sequencing, 2. spot the difference tasks, one to one with the researcher. Design: Pre post delayed test. Tasks: 1. grammaticality judgment, 2.picture description tests. No significant difference bet ween experimental groups on post test. Kim & Mathes (2001) 20 Korean adult ESL learners (high beginners and intermediate) Dative verbs Groups: (A) received explicit meta linguistic feedback, (B) received recasts. Treatment: Feedback was presented in two sessions 1 week apart each followed by production with no feedback. Controlled production tasks (as in the treatment) without feedback. Differences between performance on first and second production tasks were not significant. Differences between groups for increases in production were not significant. Learners expressed preference for explicit feedback.

PAGE 89

89 Table 3 3 Classroom studies of effects of implicit and explicit feedback Study participants Target structure Design Tests Results DeK eyser (1993) 35 Dutch high school seniors learning L2 French Variety of features, predominant ly morphosynt actic Groups: (A) extensive explicit corrective feedback during normal class activities, (B) limited explicit corrective feedback. Treatment: 10 class periods Three oral communication tasks (interview, picture description and story telling). Fill in the blank test. Tests administered twice. No statistically significant differences evident between groups A and B. Learners with h igh previous achievement, high language apti t u d e, high extrinsic motivation, and low anxiety benefited the most from error correction. Muranoi (2000) 114 Japanese college EFL students (first year) Indefinite article to denote new information Groups: (A) interaction enhancement (IE) by means of requests for repetition and recasts in communicative task + formal debriefing (explicit grammar explanation), (B) IE + meaning focused debriefing, (C) control Design: pre/immediate post/delayed post test Tasks: 1. oral story description task, 2. oral picture description task, 3. a written picture description task, 4. grammaticality judgment test. Both experimental groups outperformed the control group on both post test s. Group A outperformed B on imme diate post test but not on delayed post test Havranek & Cesnik (2001) 207 university students specializing in English Varity of English phonological lexical, and grammatical features Data on 1700 corrective feedback episodes from normal English lessons. Class specific tests (translation, correction, reading aloud, and written and spoken completion tasks). Directed at corrected items. Elicited self correction was the most efficient combinations for all learners.

PAGE 90

90 Table 3 3. Continued Study participants Target structure Design Tests Results Lyster (2004) 148 (grade 5) 10 11 year olds in a French immersion program French grammatical gender (articles +nouns) Groups: (A) received form focused instru ction (FFI) + recasts, (B) FFI + prompts (including explicit feedback), (C) FFI only, (D) control group. Treatment: workbook, providing a communicative context. Design: pre/post delayed post test Tasks: 1. object identification test, 2. picture description test, 3. written binary choice a, 4. text completion test. Group B was only group to outperform control group on all measures (post tests 1 and 2). Group A outperforme d control group on 4 and 8 measures. Statistically significant differences were found between groups B and C not between A and B. Ammar & Spada (2006) 64 ESL student (Grade 6 intensive English ESL class) Third person possessive determiners in English Groups: (A) recasts, (B) prompts (elicitation, repetition, and meta linguistic feedback), (C) control. Treatment: 330 495 min of communicative activities Design: Pre immediate post delayed post test. Task: 1. mea ning oriented contextualized grammaticality judgment task, 2. picture description task. Groups B significantly outperformed group A. Sheen (2007) 80 EFL community college students (intermediate) English articles Groups: (A) recasts, (B) me ta linguistic correction, (C) control. Treatment: retell two narratives to the class. Design: aptitude test, pre/post/delayed post test. Tasks: 1. speeded dictation test, 2. writing test, 3. error correction test Meta linguistic group significantly outperformed Groups A and C. Recast group did not perform significantly better than C.

PAGE 91

91 Table 3 3. Continued Study participants Target structure Design Tests Results Lyster & Lzquierdo (2009) Classroom + laborat ory study 25 French L2 undergraduate learners French feminine endings Groups: (A) recast, (B) prompt (clarification requests followed by a repetition). Classroom treatment: form focused instructional unit (3 hr/ day over 2 weeks), p rovided for all participants. Laboratory treatment: three oral tasks, one to one with an interlocutor. Pre post delayed post test 1. object identification task, 2. picture description task, 3. binary choice and reaction time measures. No significant differences found between groups A and B. Ellis et al. (2006) 34 private language school ESL students (low intermediate). English past tense ed Groups: (A) recasts, (B) meta linguistic information, (C) control. T reatment: two story narrative tasks. Participants were assigned in a triads with the researcher. Design: pre / immediate post / delayed post test. Tasks:1. oral imitation, 2. untimed grammaticality judgment test, 3. meta linguisti c knowledge test. Group B significantly outperformed the other two groups for both the delayed imitation and grammaticality judgment post tests. Loewen & Nabei (2007) 66 EFL students at a Japanese university (Average years of studying Engli sh: 7) English question formation Groups: (A) Recasts, (B) Clarification request, (C) Meta linguistic feedback, (D) No feedback, (E) control. Treatment: 1. a spot the difference, and 2. a guess the storyline task, one to one with th e researcher. Design: pre/ post test Tasks: 1. timed grammaticality judgment test,2. untimed grammaticality judgment test, 3. oral production task. Groups A, B, and C significantly outperformed groups D and E only on timed grammaticality judgment test, b ut not on untimed grammaticality judgment and oral production tests.

PAGE 92

92 Table 3 4 CALL studies of effects of implicit and explicit feedback Study participants Target structure Design Tests Results Nagata (1993) 32 second year univ ersity learner of L2 Japanese Japanese passive structures; verbal predicates and particles Groups: (A) feedback indicating what was missing or not expected, (B) received same feedback + meta lingui s tic explanations. Written test using same format as treat ment task. Group B significantly outperformed group (A) on particles but not verbal predicates. Learners expressed preference for meta linguistic explanation. Rosa & Leow (2004) 100 adult university learners of L2 Spanish enrolled in advan ced courses Contrary to the fact conditional sentences in the past Groups: (A) meta linguistic explanation + opportunity to try a gain if incorrect, (B) implicit feedback indicating whether the answer was right or wrong, (C) control group. Treatment: input based jigsaw task characterized by task essentialness Design: pre/immediate post/ delayed post test Tasks:1. multiple choice recognition tests, 2. written controlled production tests. Learners who received explicit feedback reported higher levels of awareness than those who received implicit feedback. In addition, learners who reported being aware of the target structure did significantly better than those who did not report being aware of it. Sanz (2003) 28 first year university learners of Spanish Position of clitic pronouns between object and verb Groups: (A) explicit meta linguistic feedback, (B) implicit feedback. Treatment: Computer delivered input processing instruction without p rior explicit instruction. Design: pre/ post test Interpretation tests. Task: 1. sentence completion. 2. written video retelling. Both groups significantly increased their abilities to interpret the O clitic V sentences, but no significant differenc es showed between the two groups. Sauro (2009) 23 high intermediate advanced adult ESL learners. English zero article Groups: (A) Recasts, (B) meta linguistic feedback, (C) control. Treatment: computer me diated collaborative writing activities, paired with a native English speaking interlocutor. Groups: (A) r ecasts, (B) meta linguistic feedback, (C) control. Treatment: computer mediated collaborative writing activities paired with a native English speaking interlocutor. No significant advantage for either feedback type on immediate or sustained increase s, the meta linguistic group showed significant immediate increase s relative to the control condition.

PAGE 93

93 3. 4 C HL L ea rners, I mplicit and E xplicit F eedback CHL learners usually use Chinese when communicating with family members, such as grandparents or parents, who normally do not speak E nglish (Dai & Zhang, 2008) When speaking with siblings and Chinese friends, CHL spe akers tend to switch to English S ome of them e ven feel embarrass ed to speak their HL with people other than their family members because they consider Chinese a language only spoken at home. Th e shyness toward their HL suggests that their acquisition and maintenance of is restricted to an informal setting with limited topics of daily life. Moreover, many CHL learners speak a non prestige dialect, which in many case s is unintelligible to Mandarin. Finally, due to their different level s of exposure to the HL, the language abilities of CHL learners vary from individual to individual, which leads to tremendous vari ations in their linguistic knowledge and abilities when beginning Chinese classes at colleg e or university (Hendryx, 2008). H owever, as a group, CHL learners show some consistent patterns: although h av ing developed a considerable level of fluency in their oral speech skills, they tend to have a narrow range of lexical and syntactic alternatives (Valds, 2001). In addition, the majority of CHL learners tend to have high level receptive skills (especially listening skills) but low level productive skills ( e.g., speaking), which often re semble other HL learners (Va lds, 2001). The imbalance between the receptive and productive skills observed in CHL learners remind s known study on Canadian French immersion program s (1985; 2005 ). Both learners in the two CHL learner group s and French immersion program learner groups speak a dominant langu age, as well as a second language which they started to learn from a young age. Learners in both groups

PAGE 94

94 are able to access an abundance of comprehensible input either in the immersion classroom or at home. Like the CHL learners, the immersion learners were also found to be weak in their productive skills (speaking and writing), although they were deemed native like in other language areas. Swain concluded that what the immersion learners lack ed was being pushed to produce modified output. She argued that mo dified output : vocally), learners may notice that they do not know how to say (or write) precisely the meaning they wish to convey... Under some circumstances, the activity of producing the target language may prompt second language learners to recognize consciously some of their linguistic problems: It may bring their attention to something they need to discover about their second langu Similar to CHL learners also may not have enough opportunities to be pushed to produce modified output. Although many of them grow up in a Chinese speaking family, the conversation between parents an d children is often a mixing of English and Chinese. Usually the parents speak in Chinese, but the children respond in English. For the se learners, comprehending input is e asier as they can understand its meaning without precise knowledge of morphology and syntax. On the other hand, in order to produce output, they need to focus their attention on the form s in their inter langu a ge and tap their language competence. When providing corrective feedback to CHL learners, the following questions need to be consid ered: which type of feedback is most likely to help learners notice the corrective inten t ion? Which type of feedback can best push CHL learners to make modifications and produce modified output?

PAGE 95

95 T he fundamental difference between CHL learners and non C HL learners lies in their internalized language system: CHL learners have already developed an implicit linguistic knowledge system of Chinese even before starting their formal Chinese study at schools, which is attribute d to their earlier exposure to this language from family members during their childhood (Valds, 2005). In contrast, non CHL learners are not equip ped with such a knowledge system beforehand, and are more likely expected to construct it solely through formal classroom instruction. In terms of explicit knowledge, however, HL learners are not sup erior to their counterpart s For instance, although they seem to know the language and are able to use a set of internalized grammatical rules, they do not necessarily have the metalanguage to talk ab out the grammatical system itself, and may also have difficulties learning grammar rules that are commonly taught in foreign language or second language classrooms (Valds, 1995). Non HL learners usually initiate their target language study by obtaining e xplicit knowledge through explicit instruction, and later only select portion s of the explicit knowledge that they feel ready to incorporate into their interlanguage system This could possibly be convert ed into implicit knowledge (Han, 2003). As CHL learn ers have already developed an implicit knowledge system, the y would need to convert the knowledge system in an opposite process to that of the non CHL learners. Having discussed the profile and internal knowledge system of CHL learners, I propose the foll owing working definition of explicit feedback in my study : information or comments which a. clearly indicate that an error has been made; b. use meta linguistic language to point out the source of the error. On the other hand, implicit

PAGE 96

96 feedback refers to i nformation or comments which a. do not overtly indicate that there was an error; b. do not interrupt the flow of communication. E xplicit feedback in my study is provided in the form of a meta linguistic technique, wh ereas implicit feedback takes the form o f recasts as shown in examples 3 1 and 3 2 below. Example 3 1 Explicit feedback 010 NNS : Wo mai si bi 1sg buy four pen I bought four pens. 011 NS : In Example 3 1 the learner produce s a problematic sentence in 010, which trigger s the feedback by the NS in 011. This feedback contains two parts: the first part that the learner m ade an error. The second part provides a with a meta language hint with regard to the correct form that should be used here Meta linguistic feedback as defined in my study may be perceptually mo re salient to CHL learners for the following reasons. First i t has been estimated that HL learners have the potential ability to acquire 80% 90% of grammatical rules that govern words, phrases, and sentences (Ming & Tao, 2008). However, certain problems were also found to persist in their grammatical competency d espite years of learning the language (Kim, 2008). These problems will not automatically disappear solely from the exposure t o positive evidence (Han, 2003, p.134). Moreover, compared with English speaking non HL learners, although HL learners share similar sentence processing problems in the language, they display greater comprehension ability in listening. This superiority in comprehension by

PAGE 97

97 HL learners is due to their strong er semantic k nowledge and contextual skill s but not to the in syntactic structure processing skills (Kim, 2008). E nvironmental support in the form of comprehensible input is necessary for language learning, but insufficient for learning certain specifiable aspects of an L2. Paradoxically, comprehensible input may actually inhibit learning on occasion, because it is often possible to understand a message without understanding all the structures and lexical items in the language encoding it, without being aware of not ( Long, 1996, p. 425). Thus the meta linguistic hint provided in meta linguistic feedback may push learner s to correct their errors through searching their implicit linguistic system Second, aning to the form. Previous studies found that HL learners tend to focus on semantic issues rather than other linguistic features. Therefore during the interaction HL learners are more likely to attend to meaning rather than form (Gass & Lewis, 2007, Kim, 2008). On the other hand, although HL learners have developed an implicit knowledge system of their HL from the exposure at home during their childhood, they tend to lack explicit meta linguistic and rule based knowledge which would possibly help them not ice the gaps between their interlanguage and the target form s (Kim, 2008). Therefore they need consistent corrective feedback from the teachers to draw their attention to the form s of the language, rather than meaning when engaging in interaction (Kim, 200 8). Based on all of these factors, meta linguistic feedback stands out compared to other feedback technique s since t he meta linguistic information provide d in the feedback not only help s learners figure out the correct linguistic form, but also contribute s system. Currently in North America, many post secondary Chinese programs are designed for either foreign language learner, or native Chinese learners who need little explicit instruction of grammar and p ronunciation (He, 2008). However, neither of t hese

PAGE 98

98 t racks may be truly effective for CHL learners who may need less explicit instruction than foreign language learners but more explicit instruction than the native speakers Example 3 2 Implicit feedback 020 NNS : Wo mai si bi 1sg buy four pen I bought four pens. 021 NS : 2sg mai le si zhi bi You buy PF V four CL pen You bought four pens? 022 NNS : Si zhi bi Four PFV pen Four pens. In the current study, implicit feedback is opera t ionalized in the form of recasts for the following reasons. First, although general agreement has been made that recasts can add material to a during L1 acquisition disagreement remain s on whether it c ould also provide negative evidence for children on the ir ungrammatical utterances (Nicholas et al. 2001). Despite this in SLA classrooms, recast is the most frequently used corrective feedback form by L2 teachers (Oliver, 1995; Long, 2007; Lyster and Ranta, 1997; Sheen, 2004). Beside corrective recasts, tea response s The dual function s of recasts usually cause ambiguity and make recasts more implicit for second language or foreign language learner s To further complicate the is sue, it remains unclear whether HL learners are relatively more sensitive toward recasts because of their previous language exposure. That is, they have noticed the dual functions served by recasts. However, from time to time, they may n ot process adequate knowledge to accurately interpret which function a particular recast serves All

PAGE 99

99 these uncertainties may make recasts more implicit for CHL learners. Second, recasts are less likely to be perceived as corrective feedback compared with other types of corre ctive feedback because of their multiple discourse functions. For CHL learners, as they have more experience communicat ing with their family members in HL, they tend to regard language exercise s in the classroom as interactional conversation (Gass & Lewis, p.99), especially when the correction is embedded in a complete rather than in a partial reformulation. In addition, their broader implicit knowledge gives them a privilege over non CHL learners T hat is they do not have to know every detail in the utter ance before they can comprehend. Therefore they tend to be less sensitive to the alternative forms provided by NS. Lastly they tend to treat a question or as a continuation of the conversation, since they are not pressed to p roduce output. In sum, due to their unique language exposure to Chinese at home, CHL learners develop an internal knowledge system that differs from non CHL learners. CHL learners have developed strong implicit knowledge, but much weaker explicit knowled ge. Consequently, their reaction toward implicit and explicit feedback may differ from non CHL learners who have the opposite distribution of implicit and explicit knowledge. Thus the findings from previous studies on the effects of implicit and explicit f eedback, which mainly focused on FL/SL learners, may not apply to CHL learners. However, to the best of my knowledge, little empirical research has been done in this area so far. This gap in the literature motivated the research questions of the current st udy, which I will present in Chapter 4. Assuming that HL learners had indeed developed a certain sensibility

PAGE 100

100 toward recasts, would they form a new perception toward recasts in the classroom after observing the teacher for a certain period of time?

PAGE 101

101 CHAPTER 4 RESEARCH QUESTIONS AND METHODOLOGY Drawing from previous research on interaction, implicit and explicit feedback, and noticing, as well as HL and non aimed at investigating the effects of feedback in relation to feedback type as well as learner language background This chapter will present the research questions and hypothes e s as well as the details of the methodology employed to answer the research questions in this study 4 .1 Research Questio ns There are three research questions in the current study. RQ 1: Do CHL learners perceive both explicit and implicit feedback on Chinese classifiers more accurately than non CHL learners? H 1: Learners with different language backgrounds would have diff erent perceptions about the feedback provided during the interaction. As HL learners develop language sensitivity and intuitions through exposure to the HL in a natural language setting at home, they can comprehend the meanings without having to know every detail in the utterances. Consequently, they tend to focus on meaning rather than on linguistic form when engaging in a conversational interaction. The findings in previous : learners with HL background showed higher accuracy in perceiving feedback on semantic and lexical items compared with their non HL counterparts. In contrast, non HL learners were more likely to perceive feedback on semantic and lexical items as no conten t, a term which refers to something related to the content of the conversation between them and the interlocutor (Gass & Lewis, 2007; Kim, 2008). Since the target

PAGE 102

102 linguistic form (Chinese classifiers) in the current study could be both semantic and lexical (Zhang, 2007 ), it was hypothesized that CHL learners would perceive both explicit and implicit feedback on Chinese classifiers more accurately than non CHL learners. RQ 2 : Will explicit feedback work more effectively than implicit feedback for CHL and non H 2 of feedback. Ammar and Spada (2006) found that prompts which included met a linguistic information, repetition, and elicitation -worked more effectively than recasts for learners with lower proficiency, but not for learners with higher proficiency. As CHL learners usually have higher proficiency than their non CHL counterparts, particularly in speaking and listening skills, it was hypothesized that explicit feedback would prove more effective than implicit feedback in non RQ 3 : Does feedback type or language background affect the increase in CHL and non an untimed written cloze test? H 3 : The two tests employed in the current study were designed to measure different types of knowledge. On the oral imi tation test, learners are more likely need to tap their implicit knowledge to perform their oral production under time pressure. On the other hand, on the written cloze test, learners have time to access their explicit knowledge since the test is untimed. In addition, their performance is determined by

PAGE 103

103 their knowledge of Chinese characters as a full score for each target linguistic form is given only when it is written in Chinese characters. HL learners have a unique internal language knowledge system that sets them apart from non HL learners in the process and development of language acquisition (Gass, 1997 ). Before receiving formal language instruction at school, HL learners have usually already dev eloped their implicit knowledge, but lack explicit knowledge as well as literacy skills In the case of CHL learners, they have no a dvantage over non CHL learners in their learning of Chinese characters In addition, p revious research also found that HL learners performed more like native speakers than the non HL learners on their oral produ ctions, but more similarly with non HL learners on their written productions (Montrul et al. 2008). Therefore the following results are expected: a) CHL learners and non CHL learners benefit from feedback in a different way ; b) CHL learners outperform no n CHL learners on the oral imitation test, but not on the written cloze test. Thus i t was increase in performance on the oral imitation test, but not on the written cloze test. 4 .2 Methodology 4 .2.1 Operationalizations Feedback Following R. Ellis et al. (2006), explicit feedback was in the form of meta linguist i c information, while implicit feedback was i n the form of recasts. In order to minimize the perceptual salience of the target ed linguistic items, f eedback was provided on any errors made by learner s and not limite d to classifier related errors only Following Lyster and Ranta (1997), meta linguist i c feedback wa s operationalized ated to the well

PAGE 104

104 A n example of meta linguistic feedback taken from the data collected for the current stud y is given in 4 1 Transcription conventions are shown i n Appendix A. Example 4 1 Meta lingui s tic feedback NNS : Wo mai le yi chenshan. 1sg buy PFV one shirt. I bought one shirt. N S : You need a measure word for clothing linguistic feedback you wear on the top As shown i n this example meta linguistic feedback includes a provision of meta linguistic information to indicate the error made by the learner by highlight ing the nature and the characteristic of the TL form. In this exa mple the feedback pointed out which type of words needed correcting in the In addition, it also indicated the specific semantic association between the target classifier and the related head noun Recast w as operational problematic utterances produced by learners. In order to make recasts as implicit as possible, the corrective part in recasts was not stressed or realized with higher pitch, in order to avoid any pa rticular salience (Leeman, 2003). Moreover, the recasts o f were always in the form of a complete sentence (Sheen, 2006). Example 4 2 Recast NNS : Wo mai le yi tiao chenshan. 1sg buy PFV one CL shirt I bought one shirt. N S :

PAGE 105

105 2sg buy PFV one CL shirt You bought a shirt. In Example 4 2, th e same error as in Example 4 1 triggered the recast provide d by the NS. When providing the recast, the NS reformulated the learner s original sentence with a complete sentence Ni mai le yi jian chenshan in either a declarative or a n interrogative form 1 (see Example 4 3). N o additional features (e.g., stress, intonation, and segmentation ) that could signal the error were added to the recast. Perception of feedb ack. Following Mackey et al. (2000), as well as Gass and was operationalized as any verbal comments that indicate d that the learner had noticed, or paid attention to feedback on the cla ssifier related error at the time. A typical example of accurate perception of feedback i s shown below: Example 4 3 NNS : Wo mai le yi jian maozi. 1sg buy PFV one CL hat. I bought one hat. N S : ? Ni mai le yi ding maozi. 2sg buy PFV one CL hat. You bought a hat? NNS : Wo mai le yi ding maozi 1sg buy PFV one CL hat. I bought one hat. Recall data know how to say one hat. I was not sure which measure word 2 corrected me.

PAGE 106

106 In example 4 3, the learner clearly verbalized that the NS corrected his error on t he classifier. 4.2.2 Design In the literature, the effects of corrective feedback have been investigated from both classroom based observation (Panova & Lyster, 2002 ; Sheen, 2004 ) and laboratory base d research (Ayoun, 2004; Han, 2002; Leeman, 2003 ). Each approach has its ad vantages: classroom specific theori (Mackey & Gass, 2005 p.186). However, in classroom based research, it is hard to determine a cause and effect relationship between the investigated independent variable(s) and the dependent variable(s) since all the var iables are present in the same context simultaneously without any control. This disadvantage could be overcome in an experimental or quasi experimental study in which independent variables are tightly controlled and manipulated. Thus the results are more straightforward and reliable. Since the goal of the current study was to investigate the association between feedback types and the learning outcome by learners with different language background s an experimental approach in which different variables (e.g ., feedback type, learner group) could be controlled was desirable. The current study employed a pre test post test randomized control group design (see Figure 4 1). By randomly assigning 64 participants to experimental and control groups, it was assur ed that there were no biases in the assignment of subjects. Forty seven learners in experimental groups participated in two treatments, during each of

PAGE 107

107 which they engaged in a conversational interaction with a NS of Mandarin. Feedback was provided by the NS on any errors made by the learners during the interaction. Learners in each experimental group received feedback in only one form, either recast or meta linguistic, depending on the group to which they were assigned Their short term language development was measured by pre and post test s. Further details about feedback was measured by stimulated recall data, which was collected on the last day of the experiment. experience and home language background. Non HL learners did not participate in the interview. Seventeen learners in a control group took only the pre and post test without participating in the treatments. Their scores o n the two tests were compared with those of the experimental groups, to ensure that any improvements made by learners in the experimental groups were due to the treatments, rather than t o test effect The pre test was us post test was employed to measure the treatment effect. In designing the pre and post test the following issues were carefully conside red. First, as the current study intended to look at the effects of feedback on each classifier, the same set of 33 sentences w as used in both tests in order to ensure comparability between the two tests. In order to minimize practice effects, however the sentences were randomly arranged in each test. Second, in order to draw different types of language knowledge from the learners, both the pre and the post test used two different testing instruments: an oral imitation test and a written test.

PAGE 108

108 The former was used to measure learners employed to assess their explicit knowledge. CHL learner Non CHL learner Figure 4 1 The r esearch design. The experiment was conducted o n three consecutive days. The short testing period was chosen due to the characteristic o f the target ed linguistic forms: the association between a classifier and available head nouns is arbitrary and conventional, Day 1 Treatment 2 Questionnaire Day 2 Pre test Treatment 1 Post test Simulated Recall Int erview Day 3 Exit

PAGE 109

109 but irregular, and not rule governed F requent input consisting of classifiers that are presented with associated head nouns makes the acquisition of these classifiers easier. In addition, all the classifiers employed in the current study were relevant to objects or animals usually seen in daily life. Therefore HL learners have far more opportunities to have exposure to these classi fiers compared with non HL learners. In other words, the longer the testing period lasted the more likelihood HL learners could hear these classifiers at home when conversing with their family members. In contrast, since non HL learners only had access to Mandarin in the classroom, it was unlikely for them to have equal opportunities t o have exposure to the classifiers. In order to control for possible external input, the current study employed a two day testing period. The treatments were both video an d audio recorded. The video recording of the treatments were edited into short clips, each of which contained one feedback episode. The video clips were used as stimuli during the stimulated recall session. The stimulated recall was used to elicit the noti cing data on the provided feedback from the 47 learners in the experimental groups. During the stimulated recall session, learners were ask ed to verbally report their thoughts after watching each video clip which contained a feedback episode taken from the two treatments. They could ask the interviewer to pause the video anytime in between when they wished to share any thoughts at any particular moment during the interaction ( See section 4.2.5.4 for more details. ) 4.2.3 Participants CHL and non CHL learners. A total of 9 3 volunteers initially participated in the experiment, however 2 9 of them were excluded from the study and their data will not be reported here. Among them, four non HL learners were e xcluded because they had a

PAGE 110

110 much longer length of formal Chinese study than the other non HL participants. F our non HL learners were excluded because they were heritage speakers of Spanish ( n = 4). Although their HL is not Chinese, their HL experience might provide them some advantages over the English monolinguals. Seventeen HL learners were excluded because their home dialect was a dialect other than Mandarin (Cantonese=12; Fujian=3 3 ; Wenzhou=1; Shanghai=1). F our HL learners withdrew from the experiment in the middle due to a scheduling conflict. Thus a total of 64 (male=35; female=29) learners were left in the final pool of participants. Among them, there were 23 HL learners and 24 non HL learners in the experimental groups; eight HL learners, and nine non HL learners were included in the control group. When the data collection was conducted, 15 participants were enrolled in the Heritage Class, which is equivalent to a second year level traditional Chinese class. Twenty one participants were enrolled in th e third year class and 2 8 were enrolled in the second year course. Their ages ranged from 18 to 23, with an average age of 19.9 ( SD =1.20). The average length of formal Chinese study was 1 0.9 months for CHL learners ( SD = 6.8 ); 19.4 months for non CHL learner s ( SD =7.9). The average length of stay in China was 1.3 months ( SD =8.0). All participants completed an informed consent form ( s ee Appendix B), and agreed to receive 0.3 0.5% extra points for their final class grade from their instructors upon the completio n of the study. Why were only HL learners with Mandarin L1 and non HL learners with English L1 involved in this study? Why were learners at high intermediate and advanced level chosen for the experiment? These two questions will be answered below.

PAGE 111

111 In Fall 2008, I conducted a preliminary survey among approximately 300 learners who enrolled in ten classes in the Chinese program at UF. Out of 267 returned responses, 79 learners were identified as CHL learners. Among them, 40 % of learners reported that th ey grew up in a Mandarin s peaking family, followed by 32.5.% of learners who reported Cantonese as their family language. The rest of learners reported various dialects spoken at home, such as Shanghai, Fujian, Wenzhou, etc. Based on the survey results, I initially proposed to include both Mandarin and Cantonese speakers in the experimental group. The final decision of excluding HL learners with dialects other than Mandarin from the experiment was made based on the following two reasons First, only 12 Cant onese speaking HL learners volunteered themselves to participate in the research, which was much lower than the originally proposed number of 30. The second reason was due to a theoretical consideration. Although the majority of Mandarin classifiers are a lso available in other dialects, there are variations among different dialects. Since Cantonese was the second most commonly spoken dialect by the HL volunteers in the current study, I will use Cantonese as an example to make a comparison with Mandarin. Ma ndarin and Cantonese share most of the same classifiers. However, for a same head noun, Mandarin and Cantonese may apply different classifiers. For instance, zhen the classifier used in Mandarin is gen while ngaan is used in Cantonese. In addition, even though a classifier is used in both dialects, the semantic meaning may differ. For instance, zhi is the classifier used in Mandarin to refer to small anima ls such as birds, mice, and cats. It also classifies one of a pair of small objects, such as shoes, gloves,

PAGE 112

112 and socks. On the other hand, jek in Cantonese, sorts not only the small animals listed above, but also large animals, such as horses and oxen. It also refers to round objects, such as eggs and teeth. Even pans, windows, and phonograph records also use it as the classifier, although this would be considered incorrect in Mandarin. Besides the differences in the use of classifiers, the most important difference that sets Mandarin and Cantonese apart is the phonology, given that the two languages are not mutually intelligible phonologically in the first place. The distance between them is as much as French differs from Spanish, or Swedish from German. The two languages have a different consonant and vowel system. Some consonants and vowels can only be found in Mandarin but not in Cantonese, and vice versa. For instance, the dental stops [z], [c], and retroflex stops [zh], and [ c h] in Mandarin are parti cularly difficult for Cantonese speakers to learn. In addition, their tonal system is also different. Mandarin has only four tones, whereas Cantonese has at last six distinctive tones. For instance, the pitch value in Mandarin rising tone is 35; however, C antonese has 35 or 25 for high rising, and 23 or 13 for low rising (Argus & Matthews, 1991 ). T hus in the current experiment, HL learners with dialects other than Mandarin were excluded in order to reduce the extra variables that may have influenced the effects of the treatments. Non HL learners with English L1 were chosen for the experiment due to the following concern. As all the HL learners participating in the current study received their formal education in English dominant mainstream schools, their dominant language was English. In fact, all 23 HL learners interviewed by the researcher consider ed English to be their native language. Previous research showed that the L1 variable was crucial in

PAGE 113

113 determining the variability of both HL and non HL learners. HL learners showed similar processing strategies as L2 learners when their dominant language wa s the same as (Kim, 2006 ). Thus it would be interesting to compare the linguistic behavior of members in these two learner groups. Learners at high intermediate and advanced levels were chosen due to the following concerns: first of all, as discussed in C hapter 2 HL learners usually tend to have strong speaking and listening skills even with very little formal classroom instruction. In order to ensure the communicative activities adopted in the treat ment appropriate for learners in both groups not too easy for CHL learners on the one hand, but also not too challenging for non CHL learners on the other hand non CHL learners need to reach a proficiency level that allows them to carry on communication wi th the NS. Secondly, since the majority of the target classifiers had already been introduced in the textbook at the beginning and intermediate level, all the non CHL learners should have already acquired, or at least encountered all or the majority of the target classifiers from their textbook through formal classroom learning. For CHL learners, although they may not have acquired all the classifiers in the classroom as some of them skipped the beginning level, they were more likely to have heard all of th ese classifiers that are frequently used in daily life. Third, for HL learners, after at least one year of formal instruction at the university level, their explicit knowledge of Mandarin, which started to grow at earlier years of their life through langua ge exposure at home, was re acquired. Thus, the difference in the performance between them and the non HL learners w as more likely due to their HL experience, which the current study was mostly eager to explore.

PAGE 114

114 CHL and non The fo llowing three learner profiles represent two types of CHL learners and the non CHL learners in the current study. For the two CHL learners learner WKW represents learners who were born in the U.S., whereas learner YHX represents learners who were foreign born, but e migrated to the U nited S tates at young age. These two profiles briefly illustrate how these two types of CHL learners developed their HL before they took formal college level Chinese courses. WKW. WKW was born in the U.S. His parents spoke onl y Chinese with him at home until he attended k indergarten at age five. Besides his parents, he also practiced his Chinese with his grandfather when the latter visited the United States He learned many Chinese classical poems from his grandfather. There wa s a clear change before and after he started k indergarten: before attending k indergarten, his mother read Chinese stories to him. He also watched Chinese TV programs; after attending k indergarten, his dominant language rapidly shifted from Chinese to Engli sh. He still watched some Chinese TV programs, such as the TV series T he J ourney to the W est T he O ne H undred and E ight H eroes, but he could only understand the plots, not the details. Worrying that he might totally forget his Chinese, his parents sent him to a community Chinese language school when he was seven years old. He went there every Sunday for a three hour class until he was sixteen. At that time he felt learning Chinese was useless since it was not commonly used in this country. Later when he too k postsecondary Chinese classes, he started to realize that he was re acquiring many words he learned when he was young. When the data was collected, he had been taking Chinese course at the University of Florida for eight months.

PAGE 115

115 YHX. She was born in Ta iwan, and came to the U.S. when she was seven. Before moving to the U.S., she spoke Chinese with her parents and grandparents at home. She learned approximately two hundred Chinese characters in kindergarten, as well as in elementary school, where she stud ied for half a year. When she had just c o me to the U.S., her parents asked her to copy Chinese characters from a dictionary every day, but she quickly gave up. Between age seven and age nine, she attended a Chinese community language school every Sunday. A lthough her English overtook her Chinese and became dominant at around age ten, Chinese has always been part of her life: she speaks Chinese with her parents at home; she enjoys going to Chinese Karaok e with her Chinese friends; she also watches Chinese TV program s with her parents, and she estimates that she can understand 80% of them. On the other hand, although there are many Chinese books and magazines at home, she almost never bothers to read them due to her limited literacy skills. When the data was c ollected, she had been taking Chinese course at the University of Florida for eight months. Learner ZK represents the non CHL learners in the current study, who have been learn ing their Chinese mainly through formal classroom education at a postsecond ary level. ZK Learner ZK is a native English speaker, who was born in the United States. He had been taking Chinese course at the University of Florida for 1 5 months when the data was collected. He never visited China, and his entire skills in Chinese ca me from classroom learning. He had Chinese language partners, with whom he sometimes practiced his Chinese. But basically he did not have many opportunities to speak

PAGE 116

116 Chinese outside the classroom. He also tried to watch some Chinese mo vies cartoons, or TV programs, but he admitted that he could only pick up some words occasionally. Interlocutor. The interlocutor of the task based interaction was served by the researcher, who is a female native Mandarin speaker with 13 years experience teaching Chinese as a second language. She started to teach undergraduate Mandarin courses at UF in 2007. In order to avoid sampling bias, all the students whom she was teaching when the data was collected were excluded from the experiment at the subject recruitment stage. The same interlocutor elicited the stimulated recall data and conducted the interview. She was in her late 30s when the data was collected. 4.2.4 Target Linguistic Items Chinese classifiers. Previous studies have investigated feedback effects in various linguistic domains, such as syntactical, morphosyntactical, and morphological forms, in the forms of dative verbs (Carroll & Swain, 1993; Kim & Mathes, 2001 ), verbal predicates and particles (Nagata, 1993 ), indefinite article (Muranoi, 2000 ), derivations of nouns from verbs (Carroll, 2001 ), noun adjective agreement (Leeman, 2003 ), pronouns (Sanz et al., 2009 ), gender agreement (Lyster, 2004 ), verb past tense formation (N. Ellis & Larsen Freeman, 2006 ; R. Ellis, 2007 ), and question formation (Loewen & Nabei, 2007 ). The target linguistic items in these studies differed in degree of complexity in te knowledge, while others required more complex rules. Thus the choice of linguistic structures relates to the results of the research, and needs to be made with careful c onsideration. The current study chose Chinese numeral classifiers as the linguistic item, motivated by theoretical, methodological, and pedagogical grounds. To my knowledge,

PAGE 117

117 no studies on feedback effect have been done in the past that chose classifiers as the target structure. Before discussing the theoretical grounds of the target linguistic form, I will first provide a brief introduction t o Chinese classifiers. Chinese is a numeral classifier language. Chinese classifiers can be subcategorized into three types: nominal, verbal, and compound classifiers. The current study chose nominal classifiers, because they are the most frequently used category, possessing the largest paradigm. ve, or certain (Killingley, 1981 Example 4 4 yi zhi mao one CL(animate inhuman) cat one cat Example 4 5 yi dai mi one CL (measuring unit) rice one bag of rice In fact, examples 4 4 and 4 5 represent two types of nominal classifiers: nominal occur with certain cla zhi in Example 4 4; those (Hu, 1993 p.9), such as the classifier dai in Example 4 5.

PAGE 118

118 The first type is known as a qualifying classifier, while the second type is called a quantifying classifier. There are overwhelmingly more quantifying classifiers than qualifying classifiers. The estimated number of classifiers runs from as many as 900 in Hanyu Liangci Cidian (Bitchener, 1999 ), to as few as several dozens (Chao, 1968 ; Li & Thompson, 1981). However, qualifying classifiers were estimated a t only about 70 (Hu, 1993). Although the number of qualifying classifiers is much smaller than that of quantifyin g classifiers, the acquisition of qualifying classifiers challenges L1 Chinese children more. Qualifying classifiers include a general classifier and dozens of specific classifiers. Specific qualifying classifiers categorize nouns into semantic classes b ased on their different physical properties, such as shape, size, animacy, etc (Harred et al., 1972; Hu, 1993 ) making the classifier selection in Chinese an extremely complex task for young children. Children have to choose appropriate classifiers by relying solely on the s emantic co occurrence constraints on the classifier head noun structure, with no morphological cues. The most frequently used qualifying classifier in Mandarin is ge, also known as a general classifier. It is usually the first classifier acquired by Ch inese children (E rbaugh, 1986; Hu, 1993; Liu, 2008; Loke, 1991; Loke & Harrison, 1986; Lu & Li, 2008 ) 4 Specific classifiers emerge after the general classifier, following a certain order determined by the perceptual salience of the object properties. For instance, childr en

PAGE 119

119 were found to acquire shape classifiers earlier than function classifiers (Loke & Harrison, 1986 ) After the above brief sketch of Chinese classifiers, I will now discuss the theoretical consideration for choosing Mandarin qualifying classifiers as the target linguistic structure in the current study First of all, although a conventional association exists between a qualifying classifier and a fixed set of head nouns, its communicative value is quite limited, as the general classifier ge could almost always be used to replace the specific classifier within a context. Doing so usually will not jeopardize the comprehension of the utterance. Secondly, unlike gender agreements in the Romance languages, no morphophonolocial inflection occurs between the classifiers and the head noun. The classifier is a si ngle, discrete item to be learned, with no further nomifications on the sentence. Thus the feedback given in the treatments could provide learners with a context in which a classifier should be applied, and enable them to (re)acquire it within a relatively short period of time. From a methodological aspect, classifiers are relatively easy to elicit in meaning focused interaction. As a classifier always associates with a head noun, in the two communication tasks in the current study, head nouns were used as stimuli to elicit the target classifiers. Lastly, on pedagogical grounds, previous studies showed that Chinese L1 children usually do not complete their acquisition of classifiers until preschool or early school years (Chang, 1983; Erbaugh, 1982; Liu, 2008; Lu & Li, 2008; Mak, 1991 ) Classifier (Szeto, 1998 ). By three years old, they acquire the general classifier ge but tend to use it to replace

PAGE 120

120 any specific classifiers (Erbaugh, 1982; Hu, 1993 ) This overgeneralization of ge continues until they are four to five years old (Liu, 2008; Lu & Li, 2008 ). By six years old, children rapidly improve their ability to use appropriate specific classifiers (Erbaugh, 1982; Liu, 2008 ). However, they do not fully master the classifier system until the age of nine (Mak, 1991 ). Acquiring classifiers is also a challenge for learners of Chinese as a FL, especially when their L1 is a non classifier language, such as English. The Mandarin classifiers in the current study. In the current study, 20 Mandarin qualifying classifiers (see Table 4 1) were chosen from Integrated Chinese I and II 5 ; these classifiers were introduced as either required or suppl ementary vocabulary in the classroom These classifiers belong to five categories : animacy, shape, arrangement, function, and the general classifier ge In the treatments, 29 nouns were given as stimuli to elicit the target classifiers. Am ong them, 1 1 classifiers used two stimuli Since shuang shared the same two stimuli with zhi, one more stimul us was added. Se ven other specific classifiers have a more limited choice of head nouns and t herefore only one stimulus was used. The general classifier could be used to replace any of the spec ific classifiers. The only exception is zhi, when used to modify a flower with one or several blossoms on the stem. 4.2.5 Materials 4.2.5.1 Background information questionnaire The background information questionnaire (see Appendix C) was designed to achieve the following two goals: first, identifying the two types of targeted population in the current study: CHL learners who grew up in a Mandarin speaking family, and non HL learners with English as their L1. Second, collecting information regarding H L

PAGE 121

121 in order to help the researcher to prepare the interview guide (see Section 4.2.5.5). The questionnaire consisted of 14 items, 10 of which were closed item questions that related to the most fundamental home language, etc. Table 4 1 The targeted classifiers Semantic Domain Classifier Meaning Stimuli Nouns Stimuli nouns Animacy zhi pi tiao inhuman inhuman inhuman c hicken, cat horse dragon, sna ke Arrangement shuang zhi paired item single item c hopstick s shoe, gloves, shoe Function ba jian ding jia liang hand tool upper body garments hat supported objects vehicles t oothbrush, s cissors s hirt, sweater hat piano car Shape tiao duo zhi zhang mian ke kuai pian gen long flexible drooping flowers long, hard, stick like flat thin flat thin round small block like flat and thin needle , t ie, belt flower pen, flower picture, desk mirror pearl, grape beef, mooncake leaf, bread needle General Ge General g eneral

PAGE 122

122 4.2.5.2 Treatment sessions In designing the treatment task s, the following issues were carefully taken into consideration: task effects and contextual factors. Previous studies suggested that learner factor s (e.g., ethnicity, gender, proficiency, etc.) affected their learning in task based instructional setting s (Bitchener, 1999; Gass & Varonis, 1986; Slimani Rolls, 2005 ). Given the heterogeneous nature of the participants in the current study, each task was carefully designed to make sure it not only elicited the targeted classifier s, but also enabled each learner to complete it at their most comfortable level. In particular, HL learners usually have greatly advanced speaking skills compared with their non HL counterparts (Valds, 1995; 2001 ) even though they are enrolled in the same class, and present similar or even weaker grammatical knowledge on standard tests. Therefore the pictures used for stimuli were carefully chosen (such as the color and s hape of items, the actions of animals, etc ) to make sure that learners had a lot to say when their language abilities allowed. Thus more interactions could occur According to the flow of the information, tasks that are commonly employed in SLA research c an be divided into two categories: one way task s and two way task s (Doughty & Pica, 1986; Long, 1988; Pica, 1987; Varonis & Gass, 1985 ) In a one way task, only one party of the dyad holds the information that is needed to complete the task T hus during the interaction, the information flows from the information holder to the other party of the dyad In contrast, in a two way task, both parties of the dyad have the information and they are expected to complete the task through excha ng ing the information Previous studies have provided controversial results in terms of the task effects of one way and two way task s : many researchers have argued that two way

PAGE 123

123 task s generate more negotiations o f meaning among learners (Doughty & Pica, 198 6; Gass & Varonis, 1985; Long, 1988; Pica, 1994), but others have questioned thi s argument (Duff, 1986; Nakahama et al., 2001 ), In order to minimize the task effects, the current study employed both a one way task and a two way task Lastly, the con text in which feedback was provided was designed to enable recast s to serve multiple discourse functions in order to make it as implicit as possible Previous research showed that recast s could be more or less implicit, depending on various recast feature s (e.g., length, number of changes, prosodic emphasis, etc ) and the discourse context (Loewen & Philp, 2006 ). For instan ce, in the classroom, learners are sometimes not sure whether a recast provided by the teacher was a corrective feedback or an approval o f their response, since teachers often use both. In the tasks designed for the current study, although recast s w ere onl y provided on errors made by learners they were done in as natural a way as possible. The learner s could therefore interpret them simply as a discourse move made by the researcher to carry on the interaction. I will be looking at precisely this question, specifically see k ing to determine whether CHL and non CHL learners perceive d this feedback differently. With the above careful considerations, the current study employed two task based NS NNS conversational interaction s in the treatment sessions In each interaction session, learners complete d one communicative task. On Day 1 they engaged in a one way interaction task ( a story telling ); on Day 2 they completed a two way task ( spot the differences ). During the interaction, participants received feedback on not only the problematic classifiers, but also any errors that they made Both tasks used Power Point slides to elicit the target form. O n each slide, several pictures were

PAGE 124

124 displayed. Each picture portrayed the items that reflected the target h ead nouns with one picture usually portray ing one target item (see Appendix F ). In the one way interaction task, learners were gi ven a set of Power Point slides, o n each of which four to ni n e target i tems ( e.g., furniture, food, an animal, etc) were dis played Learners were required to tell a story using all of the provided items In their story, they needed to provide the name and the quantity of each item. The interlocutor asked questions in Chinese (e.g. what did you see at the zoo? What did you buy for your mother for to help learners develop their story. In the two w ay interaction task, each member of the dyad was shown a set of Power Point s lide s on two separate computer screens Care was taken to make sure that neither member of the dyad could see the screen of the other party. The goal of the task was to find the differences between the learner Point slides, relying solely on the conversational interaction. Each slide contained 3 6 items. T he same it ems were already used to elicit classifiers i n the one way interaction task but using different pictures in the 4.2.5.3 Testing m ateri al In previous experimental studies, various tests were employed to measure the e ffects of feedback. Some of them favored explicit knowledge, such as grammaticality judgment (Muranoi, 2000; Nagata, 1993 ), sentence completion (e.g., Sanz, 2003 ), and translation tests (e.g.,Havranek & Cesnik, 2001 ) knowledge, such as picture description (DeKeyser, 1993; Leeman, 2003), and story telling ta s ks (DeKeyser, 1993) T ests designed to measure explicit knowledge provide learners with unlimited time, which enables them to have enough time t o apply their

PAGE 125

125 meta linguistic knowledge. On the other hand, tests designed to measure implicit knowledge expect learners to produce the TL in a spontaneous manner within a limited time period However, s olely relying on one type of tests harmed the construct validity of the research results For instance, R. Ellis (2007) criticize d some research on the effects of feedback for relying too much on testing instruments that better measur e explicit L2 knowledge As a result, explicit feedback might be found to be more effective than implicit feedback in these studies. In order to avoid the above problems, it was decided to employ multiple testing instru ments that could measure both implici t and explicit knowledge as a s olution to fully port r a y the degree of improvement caused by the feedback R. Ellis and his colleagues (R. Ellis, 2007; R. Ellis et al. 2006 ) used three different types of tests to measure the effects of implic it an d explicit feedback on English past tense ed : an oral imitation test an untimed grammaticality judgment test, and a meta linguistic knowledge test. Among the three tests, the oral imitation test aimed at the un timed grammaticality judgment test and the meta linguistic knowledge test were knowledge. The oral imitation test consisted of 36 audio recorded belief statements. During the test, p articipants first indicated on a n answer sheet whether they agreed disagreed, or were not sure about the statement Next they were requested to repeat each statement in correct English The untimed grammaticality judgment test consisted of 45 written sentences, including seven grammatic ally correct sentences and eight grammatically incorrect sentences, as well as 30 distracto r sentences. Learners were

PAGE 126

126 required to indicate whether each sentence was grammatically correct or not, report their degree of certainty, and indicate whether they a pplied any rules in their judging process. The meta linguistic knowledge test was formed by five sentences. During the test, learners were told that these sentences were ungrammatical, and were requested to correct the error, and provide the reason f or the errors. T he results showed significant group differences in the oral imitation test, but not in the untimed grammaticality judgment test in terms of ungrammatical sentences. No significant group differences were found in both tests in terms of the grammat ical sentences. Loewen and Nabei (2007) used three different testing instruments in their experiment that investigated the effects of feedback on English question formation : a timed grammaticality judgment test that limited the completion time of each item to 1.8 to 5 seconds an un timed grammaticality judgment test, and an oral production task. The untimed grammaticality judgment explicit knowledge, while the oral production task assess ed their L2 implicit kn owledge. The same set of 40 items was used in both the timed and untimed grammaticality judgment tests. Thus any differences in differences in their two types of knowledge: when the time was controlled, learne rs were linguistic k (R. Ellis, 2005 p .157 ). On the other hand, the untimed task allowed learners to involve their meta linguistic knowledge. The oral production t a s k consisted of two spot the differences tasks. The results showed an increase in the timed grammaticality judgment test, but not in the untimed grammaticality judgment test on the oral production test

PAGE 127

127 In sum, the above studies showed that employing tests designed to tap different types of linguistic knowledge generated mix ed results which provided deeper insights for researchers to investigate the effects of implicit and explicit feedback. Based on the review of these previous studies, t he current study employed two types of instruments: a n oral imitation test and a written cloze test. On the oral imitation test, learners would be focused primarily on meaning under pressure to perform in real time 6 intuitive and implicit. On the other hand, learners on the untimed written test were predicted to be more focused on form, therefore it was expected that this task would based and explicit. Their literacy skills were particularly important for completing this task. 4.2.5.4 The oral imitation test T he oral imit ati on test consisted of 33 audio recorded i te ms. Among them, 22 items related to the targeted classifiers. Half of them used correct classifiers, and the other half used incorrect ones. M oreover, 11 items served as distract o rs. All 33 items were in the for m of declarative sentence s In designing the testing items, care was taken to make sure of the following issues: first, all the vocabulary used in the testing items was chosen from Integrated Chinese I and II In other words, all the participants in the current study should have already acquired, or at least encountered these vocabulary words at the time the experiment was conducted. Second, each sentence was made long enough (containing approximately 10 TM, so that they needed to tap their implicit knowledge rather than simply memorizing it in order to complete the task (Mackey & Gass, 2005 ).

PAGE 128

128 D uring the test, the audio recording of the test items w as played one item at a time. E ach item was played only once The learners were required to first indicate orally whether they agree d with, disagree d with or were unsure about the statement in each item based on what they heard. Learners were told that they could disagree when t he y spot ted anything wrong in the se ntence (e.g., grammar, choice of vocabulary etc ) However, regardless of which indication they gave, learners were required to repeat each sentence. If they disagr eed with the sentence, they were asked to repeat the sentence in correct Chinese as shown i n Example 4 6 Example 4 6 Wo zuotian qu shangdian mai le san tiao kuaizi 1sg yesterday go shop buy PFV three CL chopsticks Yesterday I went to shop and bought three chopsticks. The learner was expected to disagree with the sentence, since the classifier used in the s entence was not correct. Upon the completion of the judgment, the learner was underly (Mackey & Gass 2005, p.55). Classifiers generally have relatively low communicative value in a Chinese conversation, since t hey can almost ; learners would have not focused their attention o n classifiers if they were not already in their lexicon, especially when the sentence is reasonably long. Other elements in the sentence, such as verbs, nouns, or adjectives might more likely draw their attention compared with classifiers. Although no minimal time was set to complete the oral imitation test, participants are still under time pressure as each of the 33 audio recorded items was only played once.

PAGE 129

129 As soon as the participant produces his/ her utterance, the next item will be played. There is not time for reviewing or checking back. 4.2.5.5 The untimed written cloze test The written test consisted of 22 it e ms, each of which was in the form o f a declarative sentence. As with the oral test, all the vocabulary in the written test appeared in Integrated Chinese I and II Therefore participants should have already acquired or at least encountered these i tem s in the classroom. During the test, lear ners were requested to fill in the blanks in Chinese characters by referring to the matched English sentence as shown in example 4 7 They were told to write in Hanyu Pinyin when they could not remember the Chinese characters. Example 4 7 Wo you san zhi mao he liang pi ma 1sg have three CL cat and two CL horse I have three cats and two horses. In example 4 7 the learner was expected to provide the two missing sho rt phrases and h There was no time limitation set for completion of the test. Besides the 21 targeted clas sifiers, 12 other words were also taken out of same sentences to serve as dis tract o rs as shown in example 4 8 In example 4 8, besides the phrase which contained the targeted classifier learners were also requested to provide the Chinese equ ivalent of the word

PAGE 130

130 Example 4 8 Mingtian shi wo de shengri Tomorrow is my NOM birthday Tomorrow is my birthday Mama gei wo mai le yi shuang xi e Mom Pre 1sg buy PFV one CL shoe M y mom bought a pair of shoes for me. 4.2.5.6 A dvantages and disadva ntages of the testing design in the current study The pre test post test design enabled me to directly examine the effects of f eedback in the current study In the SLA literature, the effects of feedback w ere also measured by uptake, which was defined as corrective feedback (e.g., Loewen, 2004; Lyster, 1998b; Panova & Lyster, 2002 ) However its validity was questioned by a number of researchers (Long, 2007; Lyster, 1998a; 2004; Ohta, 2000 ) Long argued that up tak e episodes produced toward different types of feedback should not be compa red with each other since different types of feedback were designed to achieve different instructional goals. For instance, explicit feedback (e.g., meta linguistic, elicitation, repetition) usually elicit s more uptake than implicit feedback (e.g., recast) does as it was designed to push learners to produce more output in the first place In addition, the occurrence of lear ner uptake significantly relate s to various characteristics of recasts, such as the length of recasts, the linguistic focus, the type of change, the number of changes, etc (Sheen, 2006 ) Moreover e ven when learners d o produce uptake following recasts it does not necessarily lead to subsequent L2 development. In fact, learners with lower proficiency may not be able to identify the corrective inten t ion embedded in recasts, especially when their

PAGE 131

131 developmental level is not ready for the target forms Their responses toward recasts could be just red h errings (Mackey & Philp, 1998 ) By employing a pre test and post test design feedback under investigation could be manipulated in a context in which the independent variables were carefully controlled. knowledge of the t arget forms could be measured any improvement they gained during the post test could be attributed to the effects of the treatments Thus, the correlation found between the effect s of feedback was more reliable On t he other hand, the pre test post test design also has disadvantages. First of all, as the pre test was administrated before the treatment session, it could cue learners to the target linguistic forms. Particularly, since the HL learners investigated in the current study had already stored a conside rable amount of implicit knowledge of the target forms from their childhood HL exposure, the pre test that knowledge. In contrast, the non HL learners have had exposure to these items on ly through formal classroom instruction. Therefore it may not be equally likely for them to that they once learned but forgot later since these forms were never fully mastered In other words, the pre test post test design could favor HL learners more than it does non HL learner s Second, the pre test post test design does not allow the researcher to measure the effects of any feedback that arises incidentally during the interaction (Nassaji, 2009 ) 4.2.5.7 Measure ment the feedback usually directed toward particular aspects of the language. However, those aspect s may not always be the aspects that

PAGE 132

132 to with the corrective feedback Previous studies showed a correlation between intention of the feedback and L2 development (Adams, 1991; Egi, 2007; Mackey, 2006; Rosa & Leow, 2004; Swain & Lapkin, 2002 ) In the literature, l s been exa mined mainly using introspective methods in which learners (Mackey & Gass, 2005 p. 201) either in a written form, such as learning journals, diaries, questionnai res, etc or in verb al report s such as think aloud protocols, prompted repetition of feedback, meta talk, stimulated recall, immediate recall, etc. These methods can be either conducted on line (in real time), or off line (no t in real time) (see Egi, 2004 for a review of these methods ). Among the above mentioned methods, verbal reports have been employed as a a wareness. Verbal report is generally defined as consists of gathering protocols, or reports, by asking individuals to say what is going (Mackey & G ass, 2005, p.77). In order to be taken as evidence of noticing (Schmidt, 2001 p. 77). v erbal rep ort s have to be collected either concurrently or immediately following the expe rience. The two most commonly used types of verbal report s are stimulated recall (Adams, 1991; Mackey et al., 2000; Nabei & Swain, 2002 ), and immediate recall (Egi, 2004; Leow, 1997; 2000; Philp, 2003; Rosa & O'Neill, 1999 ).

PAGE 133

133 In a stimulated re call, learners are prompted to recall thoughts they had while performing a task or participating in an even t Through the use of stimulated recall, a if he /she is pre sented with a large number of cues or stimuli which occurred during the Immediate recall elicit s data immediately after the completion of the event to be re called 5) During the interaction, learners a re asked to verbalize their thoughts or repeat the last thing of the recast they heard during a conversational turn immediately after a recall prompt signa l (Egi, 2004; Philp, 2003 p.85) Both stimulated recall and immediate recall are important methods in eliciting noticing data of recasts and other corrective feedback Both have their advantages. The biggest advantage of immediate recall over stimulated recall l ie s in the fact that it is the event when the information is still fresh (Egi, 2004 p.259). On the other hand, stimulated recall is sup erior to immediate recall in the following aspects: first, it doesn he communication flow since it is carried o ut after the completion of the oral interaction. Second, the researcher can clarify an y spot. In addition, stimulated recall requires less training prior to data collection compared with immediate recall for both the researchers and the participants The current study chose stimulated recall over immediate recall due to the following considerations: first, the two communication tasks employed in the experime nt

PAGE 134

134 required a smooth communication flow with as little interruption as possible. Particularly during the story telling task, if a learner is frequently interrupted by prompt signals, he/ she may los e interest in carrying on the story. Second, in previous s tudies the average length of immediate recall protocols was significantly shorter than stimulated recall s since the immediate recall was carried out under the pressure of continuing the interaction (Egi, 2004) of feedback is one o f the dependent variables investigated in the current study, the more detailed information provide d by the learners, the more their perception of this feedback could be explored. The procedure of the stimulated recall in the current study followed Gass a nd Mackey (2000). A typical example of stimulated recall i s shown below: Example 4 9 (Played the video.) L: Wo mai le yi jian maozi. 1sg buy PFV one CL hat. I bought one hat. N: ? Ni mai le yi ding maozi. 2sg buy PFV one CL hat. You bought a hat? L: Wo mai le yi ding maozi 1sg buy PFV one CL hat. I bought one hat. N: (Paused the video.) What were you thinking at that time? word I should u In the above example, the researcher first let the learners watch a video clip taken from the treatment in which the learner received the feedback. Then s he pau s ed the video, and asked the learner to verbally r eport what he ha d in mind when the feedback

PAGE 135

135 episode occurred. Upon the complet ion play the next video clip. 4.2.5.8 Interview only be assessed by their visible language performance but also by their underlying language competence which developed ove r time through their language exposure (Hendryx, 2008 ) Thus we should not rely solely on traditional tests to assess proficiency lev el In fact, t and home language environment, the closer we come to describing his/her real linguistic knowledge and abilities (Jiang, 2008 ) The ideal approach to fully portra y a HL development through longitudinal research. Obviously i t wa s not realistic to do so when many participants were investigated as in the current study Thus the current stu dy employed a semi structured interview There are several advantage s to using this data collection method : first, the researcher could collect needed information by asking pre prepared, well organized questions. Secondly, the researcher could ask in depth or unstructured questions when necessary, with sufficient flexibility. Thus, the researcher could collect all the intended information and get a big picture of all the HL participants, while simultaneously collect ing specific information concerning each i ndividual learner (Heigham & Sakui, 2009 ) Previous studies ha ve show n a variety of social variables, such as age of arrival, cultural identity, socioeconomic status, language environment, etc (Jia, 2008 ) Moreover, motivation was also found to play an

PAGE 136

136 important role i language skills (Lu & Li, 2008 ) In addition, home environment was found to have a (Koda et al., 2008a; Koda et al., 2008b; Xiao, 2008 ) Based on the findings of these previous studies, t he interview in the current study consiste d of four main parts: a. language background information about the learner (e.g., What was the first language or languages they spoke? At what age did English become their dominant language? Did they receive any systematic literacy education either at home or in Chinese community schools? b. th eir motivation f or learning Chinese; c background information about the ir parents (e.g., What do their parents do for a living? What languages do they speak? How are thei r English skills in the traditional four domai ns?); d. the home environment (e.g. What are the available Chinese print How often do parents and children engage in parent child HL literacy related home activities? ) Before each interview, the researcher prepared an interv iew guide based on careful reviews of the background questionnaire of each interviewee. The interview guide contained questions that both were general and individual specific. During the interview session, the researcher mostly followed the interview guide However, she also listened to the interviewee very carefully in order to ask flexible questions that were not pre prepared and thus complete d s unique language profile. 4.2.6 Procedure The experiment was conducted o n three consecutive d ays. At the beginning of the experiment, all participants were requested to sign the consent form and fill out the background information questionnaire. Based on the self reported information in the

PAGE 137

137 questionnaire, the participants were identified as either HL learners or non HL learners and were randomly assigned to a group (see Table 4 2) Table 4 2 The experimental and control groups. Recasts Meta linguistic Control CHL Non CHL CHL Non CHL CHL Non CHL 12 12 11 12 8 9 The 47 l earners in the exper iment al groups then took the pre test They first took the oral test, then the written test. Conducting t he oral test before the written test was due to the following considerations: since written test was untimed, learners would have adequate time to tap their explicit knowledge. Therefore if they took the oral test after taking the written test they could easily draw the explicit knowledge that was still fresh in their memory instead of tapping their implicit knowledge which would defeat the purpose of the test design By employing oral test before the written test, learners could be pushed to use their explicit knowledge to a greater extent in the written test if they encountered problems that could not be resolved with their implicit knowledge in the o ral test. Following the pre test s they completed the picture description task in Treatment 1. On D ay 2 The 47 learners in experimental groups first participated in the spot the differences task in Treatment 2, then took the post test A gain they first to ok the oral test, then the written test. On Day 3 all the learners in experimental groups participated in the stimulated recall session, in which they met the researcher individually. After the stimulated recall, the non HL learners exited the experiment. The 23 HL learners were interviewed by the researcher individually The procedure and approximate time spent for each task are summarized in Figure 4 2.

PAGE 138

138 CHL learner Non CHL learner Figure 4 2 Experimental procedure and approximate time The 17 Learners in the control group took the pre test on Da y 1. They also first took the oral test, then the written test. On day 2, they completed the oral and written post tests in the same order as they did in pre test. Day 1 Treatment 2 (10 20 min) Consent form, background questionnaire (5 min) minminm Day 2 Pre tests: Oral test (5 15 min) Written test (5 15 min) Treatment 1 (10 20 min) Post tests: Oral test (5 15 min) Written test (5 15 min) Simulated Recall (30 40 min) Interview (15 20 min) Day 3 Exit

PAGE 139

139 4.2.7 Transcriptions The researcher transcribed all the tests, treatments, stimulated reca ll, and the interview data. For the pre and post test data (128 tests, approximately 23 hours), only the classifier used by the learners was transcribed. For the treatment data (94 tasks, 7.45 hours for CHL learners, 11.9 hours for non CHL learners;), the stimulated recall data (6.35 hours for CHL, 14.2 hours for non CHL learners), and the interview data (13 hours), the entire recording was transcribed. 4. 2.8 Scoring On oral tests, two points were awarded for target language (TL) production, and zero p oints for non target language ( NTL ) production. One point was given when a learner used ge to replace the appropriate classifier. TL production was operationalized as TL suppliance of the targeted classifier (e.g., jian for chenshan productio n was operationalized as NTL suppliance of the classifier (e.g., tiao for chenshan On written tests, three points were awarded for TL production in Chinese characters (e.g., ( jian ) for chenshan ) two points for TL production in Hanyu Pi nyin 7 (e.g., jian for chenshan One point was awarded for the general classifier ge regardless of whether in Chinese characters or in Hanyu Pinyin 8 Zero points were given for NTL production ( e.g., tiao for chenshan 4.2.9 Coding The stimulated recall protocols were coded according to perception categories that into six categories in previo us research lexical, semantic, phonological, morphosyntactic, no content, and

PAGE 140

140 unclassifiable (Mackey et al. 2000; Gass & Lewis, 2007). Since the current study focuse d on classifier s s were first divided into two categories bas ed on whether the learner perceive d the feedback as classifier related. related comments no n classifier related comments oded into the following four subcategories: n classifier into the following five subcategories: Semantics (SE), phonology (PH), morphosyntax (MO), lexis (LE), and no content (NC). The coding scheme is summarized in figure 4 3. Figure 4 3 Coding scheme of the recall data 4.2.9.1 C lassifier related comments Th e f our subcategories under classifier related comments were tailor made for classifiers. Classifier related R A1 A2 Not classifier related PH LE Recall comments O NC SE MO

PAGE 141

141 R: reacquisition was operationalized as comments that clearly indicate the y knew which measure word should be used before participating in the study, as shown in Example 4 10. Example 4 10 Feedback episode NNS: Qu dongwuyuan wo kan dao le yi ge long. Go zoo 1sg see RC PFV one CL dragon I went to the zoo and saw a dragon. NS: Na tiao long you duo chang? That CL dragon have how long How long was that dragon? NNS: Hen chang. Very long. Very long. Recall data ( I kn A1: Acquisition 1 was operationalized as comments indicat ing that the y did not know which measure word should be used before participating in the study and as shown in Example 4 11 Example 4 11 Feedback episode NNS: Wo mai le san ge yaodai. 1sg buy PFV three CL belt. I bought three belts. NS: Ni de san tiao yaodai shi shenme yanse de? 2sg NOM three CL belt is what color NOM What color are your th r ee belt s ? NNS: dao San tiao. Yi ge dao shi hongse de. Hai you heise de. Three CL. One CL UK is red NOM Also hav e black NOM

PAGE 142

142 Three. One is red, Also have (one) black. Recall data A gain when I was doing that, I know I didn't know the measure words. It makes, kind of, my personality, I like to know what I talk about. I knew I just kept saying "ge" because I didn't know what the actual measure word was. A2: Acquisition 2 was operationalized as comments made by l earner s to indicate that they thought a wrong measure word as the correct one as shown in example 4 12 Example 4 12 Feedback episode N S: ? Ni kan dao duoshao she? 2sg see RC how many snake How many snakes did you see? NNS: zhang Yi zhang de she. One CL NOM snake One sna ke. NS: Ni kan dao yi tiao she? 2sg see RC one CL snake You saw one snake? NNS: YI tiao. One CL. One (snake). Recall data The measure word a gain I have learned snake before. Second year, but I, it was the first semester of advanced, and it was really difficult, I kind of remember, it was a story, like a fa i ry tale, but we never learned it. And I learned "tiao", but never for snake. I know it is for long object, but didn't know that was for long object, but we always used it for "he", or I mean really long and thin object. You never put it with animals. I thought it is for inanimate objects. O : The learner didn t us e any measure word in the original sentence as shown in example 4 13.

PAGE 143

143 Example 4 13 Feedback episode NNS: Wo mai si bi. 1sg buy four pen I bought four pens. NS: Ni mai le si zhi bi. 2sg buy PFV four CL pen You bought four pens. NNS: Si zhi bi. Four CL Pen Four pens. Recall data A gain with the measure word, it's always the measure words. 4.2.9.2 No n classifier related comments S E: t he semantic category was operationalized as general comments about communicating meaning, creating understanding, or being unable t o express an intended meaning. Example 4 14 Feedback episode NNS: Wo zhu zai zher foluolida. Nar hen da, hen 1sg live in here Florida. There very big, very. I live in Florida, There is very big. NS: Hen re. Very hot. Very hot. NNS: Suoyi wo mai le yi ge maozi. So 1sg buy PFV one CL hat. So I bought one hat. NS: Ni na ding maoz i shi shenme yanse. 2sg that CL hat is what color What color is your hat?

PAGE 144

144 NNS: Beise de. White NOM, White. Recall data I was actually trying to think very bright, like light. PH: t he phonological category was o p erationalized as specific comments about pronunciation. Example 4 15 Feedback episode NNS: Wo mai le yi ge huar. 1sg buy PFV one CL flower I bou ght one flower. NS: Zhi you yi zhi. Only have one CL. Only one (flower)? NNS: Yi zhi. One CL. One (flower). Recall data I didn't realize the tone difference between painting a nd flowers, huar (T1) and huar (T2). MO: t he morphosyntax category was operationalized as comments about sentence formation and structure or word order, as well as comments on specific aspects of grammar. Example 4 1 6 Feedback episode NNS: Yi ge jingzi. One CL mirror. One mirror.

PAGE 145

145 NS: Ni na mian jingzi hua le duoshao qian 2sg that CL mirror spend PFV how much money How much mone y did you spend for that mirror? NNS: Mian CL Mian Recall data I always forget the pattern you state the object that is something next to you first, like English is next to that table, but in Chinese, you say table next to is something. LE: t he lexical category was operationalized as containing specific comments about a known or unknown word, including the provision of a synonym and comments about a synonym, or the word itself. Example 4 1 7 Feedback episode NNS: yuegao Liang ge yuegao Two CL yuegao Two moon cakes. NS: Liang kuai yuebing Two CL moon cakes Two moon cakes. NNS: Yuebing. Mooncake. Recall data I just can't remem ber "bing", I just know "gao". NC : no content was operationalized as instances in which the subject participated verbally in the recall, yet said nothing about the content. Example 14 1 8 Feedback episode NS: Hai you shenme?

PAGE 146

146 Also h ave what? What else (do you) have? NNS: Hai you yi gangqin. Also have one piano Also have one piano. NS: A, ni hai mai le yi jia gangqin. A, 2s g also buy PFV one CL piano. You also bought one piano. ? Ni na jia gangqin duoshao qian 2sg that CL piano how much money How much money did you spend on that piano? NNS: Hen duo, hen duo qian Very many, very many money A lot of money. Recall data "Gangqin henduo qian". They are really expensive. 4.2 .10 Inter R ater R eli ability The researcher scored 100% of the test and coded 100% of the stimulated recall data. A female NS Chinese graduate student assistant scored 10% of randomly selected test data and 10% of randomly selected stimulated recall data. The student assistant received a training session from the researcher, in which the scoring and coding protocols were explained. In addition, a set of test and stimulated recall data were also provided for the student assistant to practice on before beginning the formal scorin g and coding, the results of which were not included in the analysis. A simple agreement percentage was employed to assess the consistency in scoring and coding. This measure indicated high rates of agreement between the raters for both data sets: inter rater reliability was 99% for the test, and 95% for the stimulated recall data.

PAGE 147

147 1 Sheen (2006) found that declarative recasts elicited a higher amount of uptake than interrogative recasts. 2 easure word used interchangeably with classifier in the current study. 3 Fujian belongs to Min dia lect group, Wenzhou and Shanghai belong to Wu dialect group. 4 Szeto (1998) argued that ge / should only be treated as a candidate for the general classifier, but has not treated as the statues of it yet. 5 Integrated Chinese I and II are required te xtbooks in the first and second year Chinese program at UF. 6 Although no minimal time was set for the oral imitation test, learners were still under pressure to perform within limited time given audio recorded sentence was only played once to the learner. 7 Hanyu Pinyin refers to the Romanization system for Mandarin. 8 the appropriate specific classifier, no extra point has given when written in Chinese character.

PAGE 148

148 CHAPTER 5 RESULTS This chapter presents descriptive and inferential statistics results for each casts and meta lingui s tic feedback. R esearch question 2 looked into the effectiveness of recasts and meta lingu i stic feedback on CHL and non Research question 3 addressed the relationship between feedback type language background, and the degree of improvement on the post test. All statistical analyses were conducted using SPSS Version 17.0 for Windows. 5.1 Research Question 1 Research question 1 asked whether CHL learners perceive both implicit and explicit feedback more accurately than non CHL learners. 5.1.1 Descriptive S tatistics The database included 1,192 feedback episodes: 428 were provided for CHL learners, whereas 764 were for non CHL learners. Of the 428 feedback episodes that were provided for CH L learners, 224 were recasts, whereas 204 were meta lingui s tic feedback. On the other hand, of the 764 feedback episodes that were provided for non CHL learners, 427 were recasts, 337 were for meta linguistic feedback. Tables 5 1 and 5 2 and figures 5 1 an d 5 2 present these results. Table 5 1 Feedback episodes provided for CHL learner group Group N M SD Percentage RHL 224 18.67 7.79 52 34 MHL 204 18.55 11.01 4 7. 66 Note N = raw frequency; M = mean frequency of feedback episodes provided per learner in each feedback group ; SD = standard deviation (frequency). MHL= CHL meta linguistic feedback group; RHL= CHL recast group.

PAGE 149

149 Figure 5 1 Feedback episodes provided for CHL group Table 5 2 Feedback episodes provided for non C HL learner group Group N M SD Percentage RNHL 427 35.58 12.81 5 5.8 9 MNHL 337 28.08 5.78 44 11 Note N = raw frequency; M = mean frequency; SD = standard deviation (frequency). MNHL= non CHL meta linguistic group; RNHL= non CHL recast group. Figure 5 2 Feedback episodes provided for non CHL group Of the 428 feedback episodes that were provided for CHL learners, 9 3 9 3 % ( n =4 02 ) of them were provided with introspective comments by learners. Of the 224 recast episodes, 96. 88 % ( n =21 7 ) w ere commented on by learners. Similarly, of the 204 meta linguistic feedback episodes, 9 0 69 % ( n = 1 8 5 ) were commented on by learners. Table 5 3 summarizes these results. All these introspective comment data were collected during the stimulated recall.

PAGE 150

150 Tabl e 5 3 Commented feedback episodes by CHL learner group Group N M SD Percentage RHL 21 7 18.00 7.79 96. 88 MHL 18 5 18.00 10.86 9 0 69 Note N = raw frequency; M = mean frequency; SD = standard deviation (frequency). On the other hand, of the 764 episode s that were provided for non CHL learners, only 9 1 .7 5 % ( n =7 0 1) of them were commented on by learners. This breaks down as follows: of the 427 recast episodes, 8 8 76 % ( n =3 7 9 ) were provided with introspective comments; whereas of the 337 meta linguistic feed back episodes, 9 5 55 % ( n =32 2 ) were commented on by learners. Table 5 4 presents these results. Table 5 4 Commented feedback episodes by non CHL learner group Group N M SD Percentage RNHL 3 7 9 33.25 12.26 8 8 76 MNHL 32 2 26.83 6.45 9 5 55 Note N = raw f requency; M = mean frequency; SD = standard deviation (frequency). In summary, learners across all the experimental groups provided introspective comments for more than eighty eight percent of the feedback episodes In the recast group, t he percentage of the commented feedback episodes provided by the CHL learners (96. 88 %) was higher than those provided by the non CHL learners ( 88 76 %) ; in the meta linguistic feedback group, t he percentage of the commented feedback episodes provided by the CHL learners (9 0 69 %) was slightly lower than those provided by the non CHL learners (9 5 55 %). 5.1.1.1 Classifier related comments Next I will report the results of classifier related comments made by all the learners. Tables 5 5 and 5 6 summarizes the number, minimum a nd maximum, mean, standard deviations, and percentage of classifier related comments provided by learners across the four experimental groups. For the CHL learner group, 71 .4 3 % of

PAGE 151

151 introspective comments made by learners in the recast group were classifier related, whereas the percentage made by learners in the meta linguistic group was 8 9 19 %. On the other hand, for the non CHL learners, 5 7 2 6% of introspective comments made by learners in the recast group were classifier related, whereas the percentage mad e by learners in the meta linguistic group was 8 1 99 %. Figure 5 3 graphically compares the percentage of classifier related comments provided by learners across the four experimental groups. Table 5 5 Classifier related comments provided by CHL learner group Group N Minimum Maximum M SD Percentage RHL 155 24 91 69.35 22.5 71 .4 3 MHL 16 5 71 100 91.31 9.58 8 9 19 Table 5 6 Classifier related comments provided by non CHL learner group Group N Minimum Maximum M SD Percentage RNHL 217 29 93 59.61 20.26 5 7 2 6 MNHL 26 4 62 94 81.91 11.29 8 1 99 Figure 5 3 Percentage of classifier related comments across experimental groups In summary, the descriptive statistics showed that learners in the meta linguistic feedback group made more classifier relate d comments than those in the recast group

PAGE 152

152 regardless of their language background. C HL learners made more classifier related comments in both recast and meta linguistic feedback groups compared with non CHL learners in the recast group. However, CHL lear ners in the recast group did not outperform non CHL learners in the meta linguistic feedback group. 5.1.1.2 No n classifier related comments Tables 5 7 and 5 to make classifier related comme nts. Among the five sub categories, the percentage of no content was over 50% for all the learners regardless of their experimental groups. This indicate d that all the learners tended to give irrelevant comments on feedback when they did not make comments on the targeted Chinese classifiers. Following no content lexis was the second most common category of comments made by all the learners. However, there were also minor differences between the CHL learners and the non CHL learners. In the recast condit ion, the third most common category was morphosyntax for the CHL learners, but phonology for the non CHL learners. The least common category was phonology for the CHL learners, but semantics for the non CHL learners. In the meta linguistic feedback condit ion, the CHL learners did not comment on any phonological and morphosyntactic issues, but 5% of their comments were about semantics In contrast, semantics was the least common category of comments for the non CHL learners, at 0%. The non CHL learners also made limited comments on phonology (3.4%) and morphosyntax (1.7%).

PAGE 153

153 Table 5 7 related Group PH (%) SE (%) MO (%) LE (%) NC (%) RHL(n=62) 1 (1.6) 3 (4.8) 5 (8.1) 18 (29.0) 35 (56. 5) MHL(n=20) 0 (0.0) 1 (5.0) 0 (0.0) 2 (10.0) 17 (85.0) Note: PH=phonology, SE=semantics, MO=morphosyntax, LE=lexical, NC=No content Table 5 8 Non related Group PH (%) SE (%) MO (%) LE (%) NC (%) RNHL (n=162) 8 (4.9) 5 (3.1) 6 (3.7) 58 (35.8) 85 (52.5) MNHL (n=58) 2 (3.4) 0 (0) 1 (1.7) 13 (22.4) 42 (72.4) In summary, when learners failed to comment on classifiers, they were most likely to comment on things unrelated to the fee dback episode, or to comment on lexical issues. There were minor differences between the CHL learners and the non CHL learners in their comments on semantic, morphosyntactic, and phonological issues. 5.1.2 Inferential S tatistics In order to test whether C HL learners perceive recast and meta linguistic feedback on Chinese classifiers more accurately than non CHL learners, a one way analysis of variance (ANOVA) was used to determine whether the means of classifier related comments provided by learners signif icantly differed across all the experimental groups. For the one way ANOVA, the alpha level was set at .05. The results revealed a significant difference between the groups, F (3, 43) = 7.77, p = .00. Since a significant between group difference was found classifier related comments, the null hypothesis of no differences between the groups was rejected. An LSD post hoc test was employed to locate the source of statistical significance. The results are summarized in t able 5 9. The LSD post hoc test revealed that CHL learners receiving meta linguistic feedback provided significantly more classifier related comments than CHL learners in the recast group ( p = .003). They also significantly outperformed non CHL learners in the

PAGE 154

154 recast gr oup ( p = .000), but not those in the meta linguistic groups ( p = .192). One the other hand, CHL learners in the recast group did not significantly outperform non CHL learners in either experimental group. These results suggested a superiority of meta linguis addition, it also indicated an advantage of CHL learners in the meta linguistic feedback group over non CHL learners in the recast group in terms of their accuracy of perceivi ng classifiers. Table 5 9 Summary of results of LSD post hoc test Commented feedback Group comparison P MHL>RHL .0 03 ** RHL= R N HL 167 RHL=MNHL 077 MHL>RNHL .00 0 ** MNHL > R NHL .002 ** MHL= MNHL .192 P < .05 ** P < .01 Note MHL= CHL meta ling ui s tic feedback group; RHL = CHL recast feedback group; RNHL = non CHL recast feedback group; MNHL = non CHL meta linguistic feedback group 5. 2 Research Question 2 R esearch question 2 asked whether explicit feedback w as more effe ctive than implicit feedback for CHL and non CHL lea 5. 2 .1 Descriptive Statistics 5. 2 .1.1 Oral test Tables 5 1 0 and 5 1 1 below summarize the number of participants, minimum and maximum, mean scores, and standard deviations for CHL learner groups and non CHL learner groups during the pre test F igure 5 4 graphically present s the mean scores of the two learners groups respectively For CHL learners, the means scores of learners in

PAGE 155

155 the recast group ( M =58.87) and those in the meta l inguistic group ( M =5 7 22 ) was very close They scored higher than l earners in the control group ( M = 41.66 ). On the other hand, for non CHL learners, learners in the recast group ( M =19.63) scored slightly higher than those in the meta linguistic group ( M =15. 8). The mean scores of learners in the control group ( M =16.44) were between the scores of learners in the two experimental groups. Table 5 1 0 Summary of oral pre test scores for CHL learner group Group N Minimum Maximum M SD RHL 12 38 92 58.87 18.28 MHL 11 23 90 57.22 19.15 CHL 8 10 85 41.66 27.75 Table 5 1 1 Summary of oral pre test scores f or non CHL learner group Group N Minimum Maximum M SD NRHL 12 6 56 19.63 13.32 NMHL 11 4 48 15.80 13.13 NCHL 9 10 21 16.44 3.67 Figure 5 4 Mean scor es of oral pre test Tables 5 1 2 and 5 1 3 below summarize the descriptive statistics by CHL learner groups and non CHL learner groups during the post test The means scores of learners in all the groups are also graphically presented in f igure 5 5 For CH L learners, those in

PAGE 156

156 the recast group ( M =73.45) scored slightly lower than those in the meta linguistic group ( M =78.05) Learners in both experimental groups scored much higher than learners in the control group ( M =50.54). On the other hand, for non CHL le arners, learners in the recast group ( M =3 5 07 ) also scored slightly lower than learners in the meta linguistic group ( M = 36 64 ). L earners in the control group ( M =15.52) scored much lower than learners in the two experimental groups. Table 5 1 2 Summary of oral post test scores of CHL learner group Group N Minimum Maximum M SD RHL 12 42 92 73.45 16.53 MHL 11 23 94 78.05 21.18 CHL 8 19 83 50.54 22.59 Table 5 1 3 Summary of oral post test scores of non CHL learner group Group N Minimum Maximum M SD RNHL 12 10 75 35 07 23.24 MNHL 12 6 40 36.64 11.52 CNHL 9 6 90 15 52 21.54 Figure 5 5 Mean scores of oral post test 5. 2 .1.2 Written test Tables 5 1 4 and 5 1 5 below summarize the descriptive statistics by CHL learner groups and non CHL learner g roups during the pre test For CHL learners, l earners in

PAGE 157

157 the recast group ( M =36.12) scored relatively higher than those in the meta linguistic group ( M =30.50) T he mean scores of l earners in the control group ( M =29.60) were close to the mean scores of lear ners in the meta linguistic group On the other hand, for non CHL learners, learners in recast group ( M =26.56) scored slightly higher than those in the meta linguistic group ( M =21.54). L earners in the control group ( M =16.89) scored relatively lower than le arners in the two experimental groups. Figure 5 6 graphically present the mean scores of all the groups. Table 5 1 4 Summary of written pre test scores f or CHL learner group Group N Minimum Maximum M SD RHL 12 7 76 36.12 21.66 MHL 11 0 62 30.50 20.08 CHL 8 2 59 29.60 21.33 Table 5 1 5 Summary of written pre test scores f or non CHL learner group Group N Minimum Maximum M SD RNHL 12 4 61 26.56 15.88 MNHL 12 3 50 21.54 13.76 CNHL 9 1 30 16.89 9.67 Figure 5 6 Mean score s of written pre tes t Tables 5 1 6 and 5 1 7 below summarize the descriptive statistics by CHL learner groups and non CHL learner groups during the written post test For CHL learners,

PAGE 158

158 learners in the recast group ( M =53.14) scored relatively higher than those in the meta ling uistic group ( M =48.79) L earners in the control group ( M =33.89) scored much lower than those in both two experimental groups On the other hand, for non CHL learners, learners in the recast group ( M =36.76) scored relatively lower than learners in the meta linguistic group ( M =46.72). L earners in the control group ( M =17.06) scored much lower than learners in the two experimental groups. Figure 5 7 graphically present the mean scores of all the groups. Table 5 1 6 Summary of written post test scores f or CHL l earner group Group N Minimum Maximum M SD RHL 12 16 88 53.14 18.84 MHL 11 7 72 48.79 19.02 CHL 8 4 67 33.89 24.76 Table 5 1 7 Summary of written post test scores f or non CHL learner group Group N Minimum Maximum M SD RNHL 12 13 70 36.76 16.76 MN HL 12 11 80 46.72 21.56 CNHL 9 2 29 17.06 8.89 Figure 5 7 Mean score s of written post test

PAGE 159

159 In summary, the descriptive statistics of the oral and written pre test s showed that learners in t he recast group s cored slightly higher than learners in th e meta linguistic group That indicated that learners in the recast group ha d a slight advantage over those in the meta lingui s tic group before they received any feedback during the treatment. However, during the post test learners in the recast group wer e outperformed by those in the meta linguistic group. The only exception occurred for CHL learners in the written test: learners in the recast group outperformed those in the meta linguistic feedback group. These comparisons are summarized in table 5 18 be low. I nferential statistics were employed to test whether any group differences existed in both pre and post test s. The results will be presented in 5.2.2 Table 5 1 8 Comparison of pre test and post test scores across the groups Test Pre Post Pre Po st Oral RHL>MHL>CHL MHL>RHL>CHL RNHL>CNHL>MNHL MNHL>RNHL>CNHL Written RHL>MHL>CHL RHL>MHL>CHL RNHL> MNHL> CNHL MNHL>RNHL>CNHL 5.2.1.3 Tables 5 1 9 and 5 2 0 dge of the target classifiers based on their verbal reports collected in the s t imulated recall Figures 5 8 and 5 9 graphically present the results. The percentage of ( re ) acquisition is approximately two times higher in the CHL learner group ( M =1 4 0 %) tha n in the non CHL learner group ( M = 6 8 %), indicating that CHL learners had knowledge of these classifiers before. CHL learners also present a higher percentage in the category of A2 ( M =4 8 8 %) than non CHL learner s ( M = 35 2 %), suggesting that they knew a wide r range of classifiers A slightly h igher percentage for the category of A1 in non CHL groups ( M =31.3%) indicated that they were more likely to use the general classifier ge instead

PAGE 160

160 of the targeted classifier They also tend ed to skip classifiers ( M = 26 7 %) more often than the CHL learners ( M =10. 3 %) which is reflected in the category of O. Table 5 19 Group R (%) A1 (%) A2 (%) O (%) RHL(n=155) 9 (5.8) 63 (40.6) 70 (45.2) 13 (8.4) MHL(n=16 6 ) 36 (2 1 7 ) 24 (14. 5 ) 8 6 (51. 8 ) 20 (12. 0 ) Total(n=320) 4 5 (1 4 0 ) 87 (2 7 2 ) 15 6 ( 48 8 ) 33 (10. 3 ) Note : R = ( re ) acquisition A1 = substitution of ge A2= wide range of (non target/ incorrect ) classifiers O = skip a classifier Table 5 2 0 Non ledge of target classifiers Group R (%) A1 (%) A2 (%) O (%) RNHL(n=21 7 ) 16 (7.4) 96 ( 44 2 ) 6 9 (3 1 8 ) 36(16.6) MNHL( n=266 ) 17 (6.4) 55 (20.7) 101 (3 8 0 ) 93 ( 35.00 ) Total (n=483) 33 ( 6 8 ) 151 (3 1 .3) 1 7 0 ( 3 5 2 ) 129 ( 26 7 ) Figure 5 8 Figure 5 9 Non

PAGE 161

161 5. 2 .2 Inferential Statistics pre test scores were submitted to the one way analysis of variance ( ANOVA ) measure to test for group equivalence. No s ignificant group differences were found in pre test scores across all e xperimental and control groups, in dicating group equivalence at the onset of the study. The results are summarized in table s 5 2 1 and 5 2 2 below. Table 5 2 1 One way ANOVA for the oral pre test Group D f MS F p CHL 2 807.54 1.78 .19 Non CHL 2 49.38 .38 .69 P <.05 Note CHL = include s CHL learners in all the experimental and control groups Non CHL = include s non CHL learners in all the experimental and control groups Table 5 2 2 One way ANOVA for the written pre test Group D f MS F p CHL 2 49.38 .38 .69 Non CHL 2 134. 07 .30 .74 P <.05 In order to answer the research question whether explicit feedback works more effectively than implicit feedback for CHL and non development, a one way ANOVA measure was employed to examine group differences in the post test The independent variable was the feedback type received by learners during the treatment. The dependent variable was their scores during the pre and post tests. The scores of each learner group were separately submitted to the one way ANOVA. The alpha level was set at .05. A post hoc LSD test was conducted to locate the source of statistical significance. scores o n the post test were submitted to the one way ANOVA measure to examine whether there wa s a significant difference between

PAGE 162

162 group s S ignificant between group differences were found in both the oral and written post test for the non CHL learner group, as well as in the oral post test for the CHL learner group. No significant between group differences were found for the CHL learn er group in the written post test The results are summarized in table s 5 2 3 and 5 2 4 below. Table 5 2 3 One way ANOVA for the oral post test Group D f MS F p CHL 2 1931.85 4.89 .015 Non CHL 2 1360.35 3.37 .048 P <.05 Table 5 2 4 One way ANOVA for the written post test Group D f MS F p CHL 2 2221.94 7.70 .13 Non CHL 2 929.58 2.20 .002 P <.05 Tables 5 2 5 and 5 2 6 summarize the results of t he LSD post hoc test for the oral post test CHL learners in the two experimental groups significantly outperformed those in the control group. However, statistical significance was not found between the two experimental groups ( p =.584). Similar results were found for non CHL learners. Table 5 2 5 LSD p ost hoc test for CHL learner s o n the oral test P airs P RHL MHL .584 RHL CHL .018 MHL CHL .006 P <.05 Table 5 2 6 LSD p ost hoc test for non CHL learner s o n the oral test Pairs p RNHL MNHL .849 RNHL CNHL .035 MNHL CNHL .024 P <.05

PAGE 163

1 63 Tables 5 2 7 and 5 2 8 summarize the results of the LSD post hoc test for the written post test Table 5 2 7 showed that CHL learners in the control group were significantly outperformed by those in the recast group (p=.049), but not by those in the meta linguistic group. On the other hand, non CHL learners in the tw o experimental groups did significantly better than those in the control group. However, no significant results were found between the two experimental groups for either CHL or non CHL learners. Table 5 2 7 LSD p ost hoc test for CHL learner o n the writte n test Pairs p RHL MHL .616 RHL CHL .049 MHL CHL .130 P <.05 Table 5 2 8 LSD p ost hoc test for non CHL learner o n the written test Pairs p RNHL MNHL .183 RNHL CNHL .013 MNHL CNHL .001 P <.05 In summary, the inferential statistics revealed significant between group differences across the experimental and control groups regardless of their CHL background. Overall, learners in the experimental groups did significantly better than learners in the control groups, but no significant between grou p differences were found between the recast group and the meta linguistic feedback group. However, i n the written post test CHL learners in the recast group significantly outperformed the control group, whereas those in the meta linguistic feedback group did not do significantly better than the control group.

PAGE 164

164 5. 3 Research Question 3 R esearch question 3 asked whether feedback type or language background affects the increase in CHL and non an unt imed written cloze test 5. 3 .1 Descriptive Statistics 5. 3 .1.1 Oral imitation test Tables 5 29 and 5 3 0 present the number of participants, minimum, and maximum scores as well as the mean scores of the increase made by learners in all the experimental gr oups. The increase was calculated by the score of the post test minus that of the pre test Figure 5 1 0 shows a graphical representation of the increase by learners in the experimental groups All the learners made an increase from pre test to post test. I n the recast group, CHL learners and non CHL learners made exactly the same increase ( M =15.33). In the meta linguistic grouop, there was also little difference between the two learner groups. In terms of feedback condition, learners in the meta linguistic feedback group made a greater increase than those in the recast group by approximately five points. Table 5 29 Summary of increase on the oral post test f or recast group Group N Minimum Maximum M SD RHL 12 4 32 15.33 12.38 RNHL 12 4 35 15.33 12 .01 Table 5 3 0 Summary of increase o n the oral post test for meta linguistic group Group N Minimum Maximum M SD MHL 11 2 42 20.91 15.31 MNHL 12 2 44 21.00 13.14

PAGE 165

165 Figure 5 1 0 Increase on the oral post test across the experimental groups Table 5 31 presents the descriptive statistics of learners in the control groups. The results show that the increase made by CHL learners in the control group was 9 points, which indicated that these learners showed an improvement of the target form by simply ta king the tests. In contrast, the increase of non CHL learners was minus 1.11 point, which indicated that these learners did not make any progress by taking the tests. Table 5 3 1 Summary of increase o n the oral post test for control group Group N Mini mum Maximum M SD CHL 8 2 21 9.00 8.00 CNHL 9 13 19 1.11 10.24 5. 3 .1.2 Written cloze test Tables 5 3 2 and 5 3 3 present the descriptive statistic s of data o n the post test of learners in the experimental groups Figure 5 1 1 shows a graphical represent ation of increase s made by learners in all the experimental groups. The results show that in the recast group the CHL learners made greater increase s ( M = 15.33) than the non CHL learners ( M = 9.17) In contrast, in the meta linguistic group the non CHL le arners ( M = 21.92) outperformed the CHL learners ( M = 16.45)

PAGE 166

166 Table 5 3 2 Summary of increase s o n the written post test for recast group Group N Minimum Maximum M SD RHL 12 3 33 15.33 9.30 RNHL 12 3 18 9.17 5.62 Table 5 3 3 Summary of increase s on the written post test scores for meta linguistic group Group N Minimum Maximum M SD MHL 11 4 34 16.45 9.92 MNHL 12 2 36 21.92 11.18 Table 5 34 presents the descriptive statistics of learners in the control groups. The results show that the increas e made by CHL learners in the control group was 4.13 points, which indicated that these learners showed a slight improvement of the target form by simply taking the tests. In contrast, the increase of non CHL learners was only .56 point, which indicated th at these learners did not benefit from taking the tests. Figure 5 1 1 Increase s o n the written post test across the experimental groups Table 5 3 4 Summary of increase s o n the written post test for control group Group N Minimum Maximum M SD CHL 8 7 12 4.13 6.47 NCHL 9 6 7 .56 4.16 In summary, the descriptive statistics results show that learners in the meta linguistic feedback group generally made higher increase in both oral and written tests regardless of their language background. Interesti ngly, whereas CHL learners in the

PAGE 167

167 control group made improvements in their increase in both oral and written test s non CHL learners made little or even regressed improvement. 5. 3 .2 Inferential Statistics In order to answer the question of whether differ ent feedback type s led to an increase in different types of knowledge, a 2 X 2 full factorial analysis of variance (ANOVA) was conducted to determine the effects of feedback type and lea r n e language background on the ir increase The use of a factorial A NOVA design in the analysis of between group differences was motivated by the following fact: in the current study, there are two independent variables that split the participants into separate groups: the feedback type and the CHL background of the learne rs. Therefore we can also call it a two way between groups design. I will first report the results for the oral test. The main effect of feedback type, ignoring the CHL background variable, is not statistical (F 1,43 = 2 1 23 p = 15 2 partial eta squared= .0 4 7 power= 2 97 ). It indicat es that learners who received meta linguistic feedback did not increase a significantly higher score o n the oral post test than those who received recasts during the treatment. Similarly it was found that back ground did not produce a statistical ly significant effect for the main effect of CHL background (F 1,43 = 000 p = 991 partial eta squared= 000 power= 050 ) i ndicating that overall, learners made similar increase regardless of their language backgrou nd The interaction between feedback type and CHL background was not found to be statistical ly significant either Next I will report the results of the written test. The main effect of feedback type, ignoring CHL background variable, is statistical ly sig nificant (F 1,43 = 6 888 p = 0 12 partial eta squared= 1 38 power= 728 ). It indicate s that learners who received meta

PAGE 168

168 linguistic feedback made a great er increase in the written post test than those who receive d recasts during the treatment The e ffect size showed that this factor accounted for 72.8 % of the variance in the data, which is a relatively large effect. In contrast, t he main effect of the CHL background variable, is not statistical ly significant (F 1,43 = .0 32 p = 8 59 partial eta squared= .0 1 power= 05 4 ). However, t he interaction between feedback type and CHL background was found to be statistical ly significant (F 1,43 = 4.663, p = .036, partial eta squared= .098 power= .560) The effect size showed that the interaction factor accounted f or 56.0% of the variance in the data, which is a relatively small effect. Thus we really want to know whether the combinations of A post hoc test was employed with an a djusted p value of .0 12 1 The result showed that for the CHL learners, the feedback type was not very important ( p =.772). In contrast, for the non CHL learners, the feedback type was very important ( p =.002): those in the meta linguistic feedback group perf ormed significantly better than those in the recast group. 1 .0 12 is the cut off value of False Detection Rate (FDR) caculated by R statistical program.

PAGE 169

169 CHAPTER 6 DISCUSSION AND CONCL USION In this chapter, I will discuss the findings of the study with regards to the research questions, followed by a consideration of the theoretical and pedagogi cal implications. Then I will address some limitations of the study, and suggest areas for future research. Finally, I will conclude the chapter by addressing how the current study extends our knowledge about the effects of implicit and explicit feedback. 6.1 Research Question 1 Research question 1 asked whether CHL learners perceive both implicit and explicit feedback on Chinese classifiers more accurately than non CHL learners. In this study ational interaction, learners with CHL background did not show any statistically significant advantage over those without CHL background in the same feedback condition. The only significant difference between the two learner groups was seen between the CHL learners in the meta linguistic feedback group and the non CHL learners in the recast group. On the other hand, learners in the meta linguistic feedback group all significantly outperformed those in the recast group, regardless of their language backgroun d. Comparing the findings of current study with those of previous research (e.g., Lyster & Ranta, 1997; Lyster 1998a, 1998b; Panova & Lyster, 2002), i t is noticeable that the percentage of accurate perception of feedback was high er, particularly in the case of recasts ( 57 % ~88% ). The discrepancy of results between the current study and previous studies could be due to the following factors: the learning context; the

PAGE 170

170 First of all, p revious res earch found that learning context could play an important noticed in a FL classroom rather than in an immersion classroom. This is because in a FL classroom, linguistic f orms are usually emphasized more often; whereas in a n immersion classroom, instructors focus more on communication and meaning (R. Ellis & Sheen, 2006; Lyster & Mori, 2006; Nicholas et al., 2001; Sheen, 2004 ). The current study was conducted in a laboratory setti ng, in which recasts were delivered intensively on the target forms in a dyad between a learner and the NS interlocutor. As a result, recasts in the current study might have become less implicit. Therefore, they were more easily noticed by the learners re gardless of their language background. T he stimulated recall data in example 6 1 illustrate that, despite the distracto r s used in the study and the corrections provided for all errors, the repeated appearances of target classifiers ss of recasts for these structures. Example 6 1 A non s NNS: Wo mai le si ge bi 1sg buy PFV four CL pen I b ought four pens. NS: Ni mai le si zhi bi 2sg buy PFV four CL pen You bought four pens? NNS: Si zhi bi Four CL pen Four pens. Recall d ata Cause when you asked, and I said, and you said it after. I tried to trouble listen for the measure words. So I would know. Cause I don't want to say a wrong measure word. People may be like: what you are talking about?

PAGE 171

171 In E xample 6 1 the l earner rep orted that he noticed he was frequently corrected by the NS interlocutor by her r eformulation of his original words. In addition, he focused his attention to trouble listen particularly for the target classifiers. In summary, the laboratory setting employ ed in the current study might have made recasts less implicit than was expected. As a result, they were more easily noticed by learners, regardless of their language background. Although the CHL learners ( 71 .4 3 %) still showed an advantage over the non CHL learners (5 7 2 6%) in their perceptions the difference was not great enough to be statistically significant. Secondly, l earners orientation may have also affect ed their perceptions of feedback Generally, HL learners share a similar pattern in their per ceptions of feedback due to their exposure to the HL in the early years. E ven though the length and the extent of the exposure are limited, the effect of the experience still lasts for a very long time (Au et al., 2002). W hen engaging in a conversational i nteraction with a NS interlocutor learners with HL background are more likely to focus on lexical and semantic issues compared with those without HL background (Gass & Lewis, 2007). In addition, they tend to trea t the interaction as a real life communicat ion that they have been experiencing with family and community members. In contrast learners without HL experience are more likely to consider the interaction as a language exercise that they always experience in a classroom (Kim, 2008) Therefore, HL lea rners could be considered more meaning oriented, whereas non HL learners are m ore form oriented. Due to this different orientation, HL learners perceive d feedback on semantic and lexical items more accurately than non HL learners in Gass and Lewis ( 2007). This finding was supported by the descriptive statistics of current study (see tables 5 5 and

PAGE 172

172 5 6), g iven that classifiers were also semantic and lexical in nature However, Gass and Lewis neither examined the relationship between feedback type and learners perception, nor employed inferential statistics to assess the significance of their results. These two gaps were filled by the current study. In the current study, t he CHL learners in the meta linguistic feedback condition were significantly more accurate in their perceptions of feedback than the non CHL learners in the recast condition ( p =.000) In addition, within the CHL learner group, those in the meta linguistic feedback group also performed significantly better than those in the recast group ( p feedback can be equated with attention, then this result could be interpreted as providing evidence that meta linguistic feedback is superior to recasts in helping CHL learners temporarily shi ft their attention from meaning oriented communication to the 2 provides support for this interpretation. Example 6 2 s NNS: Yi tiao hua One CL flower One flower? NS: Zheme yi ge zhengge de hua you jing lian qilai T his kind one CL whole NOM f lower by stem connect CP This kind of one whole flower (that was) connected by a stem Jiao shenme Call what What (do we) call (it)? NNS: Tiao CL

PAGE 173

173 Ti ao Recall data 1 ( I forgot th e word that I did not use much. In daily life I don t talk about flowers much. When I speak Chinese with my parents, I can say them ( su ch as fl owers). I do not think whether I (should) call it duo zhi or mian You made me think what they (should be) called. I forgot many of (the measure words). In Example 6 2, the CHL learner s verbal comments implied that he usually does not pay much attention to his choice of word s when speaking with his parents in Chinese. However, the meta linguistic feedback given by the interlocutor forced him to attend to the appropriate linguistic form which he had acquired in the past but subsequently forgot. In contrast, recasts could sometimes be interpreted as an alternative way fo r saying the same thing with the purpose of keeping the communicative channel open. E xample 6 3 illustated this. Example 6 3 CHL learner verbal comment NNS: F angjian fang zhe yi zhang jingzi Room put CP One CL mirror There is a mirror in the room. NS: Na mian jingzi gao bu gao T hat CL mirror tall not tall Is that mirror tall or not? NNS: Bu gao Not tall Not tall. Recall data 2 That one is OK. I forgot for a second the word mirror for C hinese.

PAGE 174

174 In Example 6 3, the CHL learner did not perceive the provided recast as a corrective feedback, which was reflected in his response toward it: instead of producing an immediate uptake to modify the non targetlike classifier, h e simply responded to continue the interaction. A post hoc analysis was conducted to investigate this learner s immediate up take pattern after feedback during the interaction Uptake here refers to the learners modification of their original utterance fo llowing the NS s provision of feedback through recasts or meta linguistic feedback ( Mackey et al., p. 492). As discussed in Chapter 3, uptake may not be a reliable measure of noticing of the feedback, particularly in the case of recasts. However, it may b perceptions about feedback at the time of the feedback. The result showed that learner XSB tended to provide uptake after receiving feedback. In addition, for those recasts for which he did not provide uptake, none was accurately per ceived. The se results are shown in tables 6 1 Table 6 1 Frequency of uptake and perception of learner XSB +Uptake ( n =1 8 ) 81 82 % Uptake ( n = 4 ) 18 18 % Perceived Not perceived Perceived Not perceived 1 0 ( 55.56% ) 8 ( 44 44 %) 0 ( 0 %) 4 ( 100 00 %) In addition, a second post hoc analysis showed that learner XSB used the same problematic classifier zhang in both his oral and written pre and post tests. All this evidence suggested that this learner assumed zhan g was the correct classifier for mirror The target classifier mian provided by the interlocutor at the moment of the interaction was apparently assumed to be just an alternative one by this learner, and never assimilated into his interlanguage system.

PAGE 175

175 In summary, m eta linguistic feedback overtly indicates the error with meta linguistic information, thus it helps CHL learners shift their attention from meaning oriented communication to the target form. Compared with meta linguistic feedback, recasts are le based or content based lessons, in which meaning and communication, rather than linguistic forms, are in focus (Lyster, 1998a, 1998b; Lyster & Ranta, 1997). Thirdly, l previous learning experience might also have had impact on their perceptions of feedback, particularly recasts. When Chinese L1 children acquire classifiers at an early age, they tend to treat the classifier and noun construction as an inseparable linguist ic unit and learn them by simple memorization (Fang, 1985 ) This stimulated recall in the current stu dy. Examples 6 4 and 6 5 below are two of the representative verbal reports. Example 6 4 At home NNS: Ye mai le san zhi pidai Also buy PFV three CL belt (I) also bought three belts. NS : Na san tiao pidai shi shenme yanse de That three CL bel t is what color NOM What color are those three belts? NNS : San tiao pidai Three CL belt Three belts, Hongse heise de Red black NOM Red and black,

PAGE 176

176 Hai you yi ge huise de Also have one CL gray NOM (There is) also a gray one. Recall data either. I have never worn belts before. For things I often use, such as one pen, a piece of paper, I would know the classifiers for them. But I have n ot Example 6 5 At community Chinese language school NNS: Yi ge she One CL snake NS: ? Na tiao she sh i shenme yanse de? That CL snake is what color NOM What color is that snake? NNS: Na tiao She. Baise, huangse de That CL snake white yellow That snake (is) white, yellow. Rec all data poem, (My mom sent me to a Chinese school when I was young, where many parents taught their children how to speak Chinese. I remembered there was this lesson, in which the teacher taught us a poem. I forgot what the poem was about. But I remember it was about a zoo. So I remember how The verbal comments in examples 6 4 and 6 5 were reported by the same CHL learner: Example 6 4 was about his learning experience at home, whereas Example 6 5 was drawn from his learning experience at a Chinese community school where he was

PAGE 177

177 sent by his parents when he was young. Both examp les indicated that he acquired classifiers and their associated nouns as a chunk. In other words, his learning experience of classifiers was overwhelmingly input based. As a result, his knowledge was unbalanced. For instance, he knew the target classifier for horse which was considered relatively difficult by the non CHL learners in the current study. However, he did not know the less difficult classifier for belt as he did not wear belts often in daily life, and therefore did not have the opportunity to acquire the classifier for it. The descriptive statistics results in tables 5 19 and 5 20 provided further evidence on this issue: the results showed that the CHL learners were more likely to use a non target classifier that was supposed to associate with other nouns than the non CHL learners (48.8% vs. 35.2 %), which suggested that the CHL learners had heard a wider range of classifiers than the non CHL learners During the interaction, they mainly relied on these personal experiences to figure out whether a classifier was correct or not. Sometimes their experience was not accurate. Example 6 6 illustrats this. Example 6 6 NNS: Haiyou liang zhang jingzi Also two CL mirror (t here are) also two mirrors. Zhang women yiban shi shuo pingmian de dongxi CL 1pl usually is say flat surface NOM thing Zhang, we usually (use it to) call somet hing with a flat surface Danshi shuo jingzi women bu shuo zhang But say mirror 1pl no say CL zhang, we usually use it to describe an object with a flat surface, But we NNS: ? Jingzi

PAGE 178

178 Mirror. Recall data ( yizhang jingzi I definitely have heard it, it sounded very familiar. Therefore when I was thinking (dur ing the interaction), I could not see anything wrong there. ) In Example 6 6 the learner assumed zhang was the correct classifier because she had heard it in the past, even though the interlocutor clearly pointed out that her assumption was inco rrect. A post hoc analysis showed that this learner still used zhang on her post test. In contrast, the acquisition of classifiers by non CHL learners comes mainly from formal classroom instruction. Their knowledge of classifiers was found to relate po sitively to their proficiency levels: the higher their proficiency level in Chinese, the greater their knowledge of classifiers (Liang, 2008 ). Unlike CHL learners, they acquire classifiers through explicit learning rather than simply memorizing classifiers and their associated nouns as a unit. Example 6. 7 illustrates a typical thinking process by a non CHL learner toward the provided recast, which was reported during the stimulated recall. Example 6 7 A non NNS: Wo ye mai yi zhan g 1sg also buy one CL I also buy one, Yi zhang da jingzi One CL big mirror One big mirror. NS: Yi mian hen da de jingzi ma One CL very big NOM mirror QP

PAGE 179

179 One big mirror? NNS: Dui. Yes. Recall data At first I thought the measure word would be zhang for jingzi cause it is flat surface. But it wasn't, I was kind of su rprised for a second, I heard you corrected me. In Example 6 7 surprised about the target classifier provided by the interlocutor. Unlike CHL learners, he relied on the meta linguistic kno wledge that he had acquired through classroom learning rather than memorization of the classifier and an associated noun unit. In summary, the previous learning experience helped both CHL learners and non CHL learners to accurately perceive the recasts pr ovided during the interaction. However, it worked in a different manner. producti 452). In order for feedback to play a facilitating role in promoting learning during interaction, learners need to focus their selective attention on linguistic forms and notice the gap between their problematic interlanguage and the ta rget like forms (Schmidt & Frota, 1996). If learners verbal reports of their perceptions can be quated with attention, it is interesting to know to which particular aspect of language learners attended when they failed to perceive feedback accurately. For the current study, it is particularly interesting to investigate this issue by comparing the two learner groups

PAGE 180

180 Regardless of the language background, when learners failed to make classifier related comments, 85% of their comments were either unrelated to the feedback episode at all or related to lexis (see tables 5 7 and 5 8). The descriptive statistics showed that the number of non classifier related comments given by the CHL learners in the recast group ( n =62) was approximately three times more than by CHL learners in the meta linguistic group ( n =20). A similar pattern was also seen with the non CHL learners (recast: 162; meta linguistic: 58). The gap between the two feedback groups indicated that learners who received recasts were undoubtedly less aw are of the classifier related feedback. In addition, the non CHL learners were more likely to make no content recall comments compared with the no content comments also showed that the CHL learners were more focused on telling a sto ry, whereas the attention of the non CHL learners was more easily captured by things that seemed unusual to them in the interaction. Examples 6 8 and 6 9 illustrate this. Example 6 8 no content comment NNS: Wo mai yi ge maozi 1sg buy one CL hat I bought one hat. NS: Na ding maozi shi shenme yanse de That CL hat is what color NOM What col or is that hat ? NNS: Baise de White NOM Recall data Yeah, I was just describing the picture. Example 6 9 A non no content comment

PAGE 181

181 NNS: Y i ge long O ne CL dragon One dragon NS: Zhiyou yi tiao long dui ma Only have one CL dragon right QP (there is) only one dragon, right? NNS: CL One (dragon). Recall data Like when you asked me can I see the dragon, I thought it was funny cause dragon is not real. Although participants in the current study have already acquired or encountered all the stimuli nouns, as well as the majority of the vocabulary that emerged during the interaction (e.g., colors) 3 through classroom instruction, many of them were still struggling due to a lack of adequate lexicon to describe the pictures. Although HL learners generally have a much wider range of vocabulary than non HL learners, they may st ill have difficulties when describing things that are less frequently encountered in daily life. Example 6 10 lexical comment NNS: Wo mai le san ge yaodai 1sg buy PFV three CL belt I bought three belts NS: Na san tiao yaodai shi shenme yanse That three CL belt is what color What color are (those) three belts ? NNS: Hongse de, heise de, haiyou Red NOM black NOM also have Recall data

PAGE 182

182 I didn't know how to describe the color In Example 6 10, the CHL learner did not know how to describe the color of the belt, which was a light brown in the picture. In this particular case, the word for light brown was more important for maintaining the flow of communication compared to the attenti on was focused on how to describe the color rather than the classifier. Compared with the CHL learners, the non CHL learners have more difficulties in their vocabulary, which is illustrated in Example 6 11. Example 6 11 A non ment NNS: Liang ge Two CL Two. NS: Zhe liang zhi xiaoji zai zuo shenme? This two CL little chicken Pre do what What are the two little chicke ns doing? NNS: Zai, zuo, bu zhidao Pre do no know Recall data I didn't know the word for standing at all. And I didn't know the word for chicken. I know I heard it before in Ch ina, but I didn't put it together. In Example 6 11, the non CHL learner did not know the two most fundamental vocabulary items that were needed to carry on the conversation. Therefore her full attention was focused on the lexical issue rather than the c lassifier. Besides the categories of no content and lexis the two learner groups showed some differences in the distribution of categories of phonology, semantics, and the morphosyntax. The CHL learners were least likely to make a comment on phonology

PAGE 183

183 (o nly one case), which indicated their confidence about their pronunciation. In contrast, the non CHL learners made most of their comments about phonology, particularly regarding the tone. Examples 6 12 and 6 13 illustrated these results. Example 6 12 A CHL learner NNS: You liu duo hua Have six CL flower There are six flowers. NS: Ni de hua dai bu dai zhizi? You NOM flower have no have stem Does your flower have a stem? NNS: You Have (it) has. NS Na ni you liu zhi hua Then 2sg have six CL flower Then you have six flowers. Recall data I couldn't pronounce duo Example 6 13 A non phonological comment NNS: Wu ge hua Fiv e CL flower ( There are ) five flowers. NS: Yi er san si wu liu, you liu duo hua One two three four five six You six CL flower (Count) one, two, three, four, five, six, there are six flo wer s. NNS : Sorry Recall data I got confused with huar (tone 1) and huar(tone 4). On the other hand, the CHL learners in the recast group made most o f their comments about morphosyntac issues. Interestingly, like the non

PAGE 184

184 CHL learners, they were also distracted by English word order. Examples 6 14 and 6 15 show two representative cases. Example 6 NNS: Wu ge hua Five CL flower (There are) five flowers. NS: Yi er san si wu liu, you liu duo hua One two three four five six You s ix CL flower (Count) one, two, three, four, five, six, there are six flowers. NNS: Recall data adjective to describe (I was thinking what wo rd I should put in front of the word picture. As something like adjective should be put to describe the shape of an object in Chinese. It should be put after the word in English. It is both very difficult for me). Example 6 tactic comment NNS: Yi ge jingzi One CL mirror One mirror. NS: Ni na mian jingzi hua le duoshao qian 2sg that CL mirror spend PFV how much money How much money did you spend for that mirror? NNS: Pangbian de gangqin Next NOM piano The pinao next (to the mirror). Recall data I alway s forget the pattern you state the object that is something next to you first, like English is next to that table, but in Chinese, you say table next to is something. I always have to stop and think.

PAGE 185

185 In summary, the CHL and the non CHL learners tended t o focus their attention on different language elements in the feedback when they did not perceive feedback as classifier related. Regardless of their language background, they were more easily drawn to the areas in which they were less confident. 6 2 Research Question 2 Research question 2 asked whether explicit feedback work s more effectively than implicit feedback for CHL and non The results of the current study indicated that the answer to this research questio n which differs from findings of some previous research (e.g., Ammar & Spada, 2006; R. Ellis et al. 2006; Lyster, 2004). Therefore I will focus the discussion by asking why no significant differences emerge d between feedback types among either the CHL learners or the non CHL learner s The one way ANOVA on the oral post test scores revealed identical results for the recast and the meta linguistic feedback group s among both CHL and non CHL learner s. T hat is, regardless of the feedback type learner s received during the treatment they all significantly outperformed those in the control groups Similar results were also found for non CHL learners in the written test. Since group equivalence was not found at the onset, any increase s by learners made i n the post test could be attributed to the effectiveness of the feedback they received during the treatment These results suggested that both implicit and explicit feedback provided in communicative sition of the target linguistic forms. However, a n unexpected finding in the current study was that no significant

PAGE 186

186 between group differences were found for CHL learners in the written test. The LSD post hoc test revealed that learners in the control group w ere significantly outperformed by learners in the recast group but not by those in the meta linguistic group ( p =. 13 ). I will discuss this issue at the end of this section. Despite the significant differences existing between the experimental groups and the control group, n o significant difference between the recast group and the meta linguistic feedback group were found. These results echoed findings made in some previous studies ( e.g., Carroll & Swain, 1993; De K eyser, 2003; K ang, 2009; Lo e wen & Nabei, 2 007; Nagata, 1993; Sanz, 200 3 ; Sauro, 2009 ), but contrast ed with results of other researchers (e.g., Ammar & Spada, 2006; Lyster, 2004; R. Ellis et al. 2006). Next I will explain why significant differences did not emerge between the two feedback conditio ns in the current study by comparing my results with some of the previous research. The first explanation relates to the explicitness of the feedback provided during the interactions. As explained in Chapter 4, implicit feedback in the current study wa s operati onalized in the form of recasts, which were always provided as implicit ly as possible: they were always delivered in a complete sentence without any illocutionary cues (e.g., stress or higher pitch) to increase the salience of the reformulation. I n contrast, explicit feedback was operationalized in the form of meta linguistic feedback, which overtly indicated error s made by learners with specific meta linguistic clues. Previous research found that the explicitness of feedback closely related to lea noticing of feedback. For instance, Rosa and Leow found that learners who received explicit information had higher levels of awareness than those who received more

PAGE 187

187 implicit information on their target structures (2004). However, as discussed for res earch question 1, learners in the current study noticed more than half of the feedback, regardless of the type of feedback they received during the interactions. In other words, it seemed that recasts and meta linguistic feedback w ork ed equally effectively As discussed in section 6.1, t he laboratory setting might have played an important role in increasing the likelihood of noticing which supported the findings of previous studies (Ha n, 2002; Nicholas et al., 2001). Thus recasts in the present study might have become more explicit compared with those in previous classroom studies that found significant differences between explicit and implicit feedback (e.g., Ammar & Spada, 2006; R. L yster, 2004; R. Ellis et al. 2006). con tribute to the lack of difference between the two feedback types. The descriptive statistics showed that CHL learners scored close to 60 % i n the oral pre test which suggested that they could have already been quite close to the ceiling levels of the target classifiers. With such a high initial level, it might have be en hard for them to improve significantly with in a short period of time reg ardless of which type of feedback they received during the treatment This result was supported by the findings made of Ammar and Spada (1996). Examples 6 16 and 6 17 illustrated a typical case of a CHL learner who failed to use the target classifier even after he noticed the recasts a couple of times during the interaction Example 6 16 Learner XS in treatment 1 NNS: W o mai le si, si, si ge bi. 1sg buy PFV four four four CL pen

PAGE 188

188 I bought four pens. NS: Na si zhi bi shi shenme yanse de That four CL pen is what color NOM What color are those four pens ? NNS: Yi zhi shi lvse de One CL is gre e n NOM One is green, Yi zhi shi hongse de One CL is red NOM One is red. Recall data This is like zhi bi but I was like ge bi I didn't, or I forgot, I didn't remember what it was. Example 6 17 Learner XS in treatment 2 NNS: Wo you sa n ge maobi Isg have three CL brush I have three brushes. NS: Wo ye you san zhi maobi Isg also have three CL brush I also have three brushes. Ni you duoshao gangbi 2sg have how many pen How many pens do you have NNS: Liang zhi gangbi Two CL pen Two pens. Recall data I didn't know what gangbi was. And then the measurement word for maobi and gangbi I didn't know. Zhi is a frequently used classifier for long rigid object s which was found to emerge in L1 Mandarin Children s vocabulary as early as age three (Hu, 1993). Therefore t here

PAGE 189

189 was a high likelihood t hat the learner XS had heard it in the past This assumption was supported by the verbal reports in Example 6 16 The verbal reports also indicated that he noticed the target classifier provided by the NS in the recast. However, on the following day, when a different noun that required the same target classifier was presented to the learner, he f ailed to provide it a gain As discussed in section 6.1, this could be explained as follows : since CHL learners tend to memorize a classifier and its associated noun as an inseparable linguistic unit they may only be able to apply the classifier in the same unit, but not in a new context It was noticeable that in both treatments, he not only noticed the target classifier, but also provided successful uptake. However the post hoc analysis of his post test showed that he used the general classifier ge instead of zhi in new contexts this suggests that CHL learners need time to incorporate what they have noticed in the feedback into their internal language system In addition, it may take some time and practice for them to produce it. In contrast, non CHL learners in the current study had a much lower initial level of the target classifiers ( M <20) compared with CHL learners Thus, i t seemed reasonable to assume tha t meta linguistic feedback would be more effective than recasts based on the findings made by Ammar and Spada (1996). However, although the descriptive statistics showed a slight advantage in the mean scores achieved by learners in the meta linguistic feed back group over those in the recast group, no statistically significant difference was found in the current study. The difference between the two studies could be due to the different target structure s : the target form employed in Ammar and w as the English third person singular possessive determiners his and he r which agree with the gender of the possessor. In contrast, the linguistic form

PAGE 190

190 targeted in the current study was classifiers, which select their nouns based on different semantic dom ains, such as shape, function, animacy, etc. Thus the semantic relationship between a classifier and the associated noun is intricate and complex (Hu, 1993). In addition, a lthough the target classifiers had already been encountered by non CHL learners in t he classroom the extent to which they had been acquired varied. Some were only briefly introduced as sup plementary vocabular y words ; some were thoroughly taught in the text and appeared in the homework and exams. Therefore, learners might have forgotten t hose classifiers that were not used, as shown in E xample 6 17 Example 6 17 NNS: Wo you yi yi ge maozi Isg have one one CL hat I have one hat. NS: Ni you yi ding maozi shi ma 2sg have one CL hat is QP You h ave one hat, do NNS: yi ding maozi one CL hat One hat. Recall data Yeah, I just cannot remember the measure word. But I was trying to make it more conversation I was trying to respond or ask questions back I wish I were more fluent. That is what I was thinking. I felt better the second day compared with the first day. Because a lot of things we discussed on the first day were things we had learned, I had studi ed, but I don't ever use them. I can never talk about flowers or moon pies, or anything like that, I never talked about them. So I forget how to say them. So if you remind me, usually I can remember the majority of them. Besides the classifiers, non CHL l earners also had to worry about the ir pronunciation particularly the tones The descriptive statistics results of learners

PAGE 191

191 perceptions showed that non CHL learners made more phonology related comments than CHL learners when they failed to perceive the fe edback accurately. Example 6 1 8 presented such a case of a non CHL learner. Example 6 1 8 NNS: W o you liu huar 1sg have six flower I have six flowers. NS: Ni You liu zhi hua r 2sg have six CL flower You have six flowers? N N S: Liu z hi huar six CL flower Six flowers. Recall data I just wanted to make sure that I made the correct tone for that one. As a final explanation of lack of differences between the two feedback conditions in the current study, the le ngth of treatment in the current study wa s relatively short: the total length of the two treatment tasks was approximately 20 minutes for CHL learners, and 40 m inutes for the non CHL learners. The different length of treatment for the two learner groups wa s due to their different fluency: the CHL learners engaged in the interaction with the NS in a smooth manner. They always responded to the NS very quickly. In contrast, the non CHL learners tended to take some time before t took them an even longer time to respond to the NS. Previous studies that revealed significant differences between different feedback groups usually spread their treatment s over a longer period of time, rang ing from one hour (R. Ellis et al. 2006) to several w eeks (Ammar & Spada, 2006 ; R. Lyster, 2004). In

PAGE 192

192 fact, the descriptive statistics in the current study showed that learners in the meta linguistic feedback groups generally scored higher than those in the recast groups o n the post test despite th eir lower initial scor es o n the pre test This advantage could possibly become statistical ly significant if learners were given a longer time to internalize what they have increased from the feedback. However, it is not clear to what extent the length of t reatment would influence the effects of the feedback (Lowwen & Nabei, 2007). A l onger treatment time may not necessarily lead to statistical differences between different feedback groups. For instance, the academic year longitudinal study conducted by DeKe yser (1993) found no statistical differences between implicit and explicit feedback groups. Therefore the length of time that w ould be needed to optimize the effects of different feedback needs further empirical research Before end ing this section, I wil l discuss why CHL learners in the meta linguistic group did not outperform those in the control group o n the written post test to the extent expected It is also noteworthy that, although the difference between the recast group and the control group was st atistically significant, the p value was .0 49 which was only marginally significant. difference between the feedback group and the control group. The written tes t in the present study scored three points for each target classifier written using a correct Chinese character, whereas two points w ere given for each target classifier written in correct Hanyu Pinyin. Thus, to achieve high scores o n the written test, a l earner not only needed to know the target classifier, but also the Chinese character. As explained in Chapter 3, most HL learners only develop their literacy skills in English, but not in their

PAGE 193

193 HL. The descriptive statistics showed that CHL learners scored much lower o n their wri tten test than o n the oral test indicating that even though they know a target classifier, they might not have known the Chinese character, or even the correct Hanyu Pinyin. Since feedback was provided through oral interaction, lea rners could not get help from the treatment in writing Chinese characters or Hanyu Pinyin. 6. 3 Research Question 3 Research question 3 asked whether feedback type or language background affects on test, b) an untimed written cloze test. Oral imitation test. The results of the current study showed that neither feedback oral imitation test. In addition, the descrip tive statistics showed that the CHL learners and the non CHL learners increased by virtually the same amount when they received the same type of feedback. These results seemed to suggest that the two groups of learner benefited equally from the same type o f feedback. However, the descriptive statistics in table 5 31 showed that the CHL learners in the control group increased by nine points from the pre test to the post test, which was more than half of what was achieved by the CHL learners in the recast gr oup ( M =15.33), and only slightly less than half of what was scored by the CHL learners in the meta lingu i stic feedback group ( M = 20 91 ). Based on this result, it is reasonable to expect that the CHL learners in the experimental groups would have increased t heir score even without any feedback. This could not be interpreted as practice effect, given the fact that the non CHL learners in the control group not only did not make any

PAGE 194

194 increase, but also decreased their score by 1.11 points. Apparently a more plaus ible interpretation is needed here. As reviewed in Chapters 2 and 3, HL learners usually start their exposure to the HL at birth. In most of the cases they develop their HL through daily oral communication with family members. As a result, they tend to ha ve stronger oral proficiency than FL/ SL learners. In fact, HL learners were found to perform more like native speakers than the FL/ SL learners on oral production tasks. The descriptive statistics in the current study, which were presented in tables 5 10 and 5 11, also showed that the CHL learners ( M > 57%) had a higher level in their initial knowledge of the target classifiers than the non CHL learners ( M outstanding performance on oral production tasks to their implicit and automatically processed knowledge, which is typically acquired during childhood (e.g., Montrul et al., in their verbal comments collected durin g the stimulated recall, as illustrated in Example 6 19. Example 6 19 A CHL learner YHX NNS: Ranhou yi zhi she Then one CL snake Then (there is) one snake. NS: She shi changchang de, ruanruan de, xixi de Snake is long NOM soft NOM thin Snake is long, soft, and thin. NNS: Yi tiao she One CL snake One snake. Recall data

PAGE 195

195 (I knew it was wrong right after I said it. Because I first thought of yitiaoshe whether it is correct or not until I say it. I would only know whether it is correct or not after listening to it. According to the past experience, and according to whether I have heard it or not). In Example 6 19, the learner immediately realized that the classifier she used was incorrect based on what she heard in the past. In other words, she could have figu red out the error even without the feedback provided by the interlocutor. Therefore this case could probably be used to explain the increase obtained by the CHL learners in the through simple (repeated) exposure to the target items. For this particular learner, It is impossible to know how much of this increase was due to her implicit and intuitive knowledge developed from her past experience, and how much was due to the effect of the feedback. A post hoc analysis of pre and post test showed that she increa sed her score from 56.3 to 89.6, which was much higher than her group average level ( M =20.91). It noteworthy that her pre test score was slightly lower than the group av erage ( M = 57.22). Her outstanding performance could be related to her higher degree of exposure to the language compared with many of her peers in the group (see her profile in Chapter 4). Unlike many young arrivals who quickly totally shifted their domina ntly language to English, YHX has kept her HL as part of her life: she speaks Mandarin with her parents, goes to Chinese Karaoki with her Chinese friends, and also watches Chinese TV programs with her parents at home. However, the feedback should have at l east helped her confirm her intuitions. If her increase was mostly due to the

PAGE 196

196 activation of her implicit knowledge, which type of feedback she received during the interaction should not have affected her performance. Now I will present another learner WKW, who received recasts during the treatment. A post hoc analysis of his pre and post test results showed that he scored 50% on the pre test, which was lower than the group average ( M =58.87); and 77.1% on the post test, which was slightly higher than th e group average ( M =73.45). His increase ( M =27.1) was also higher than the group average ( M =15.33). The interview data (see his profile in Chapter 4) showed that before taking a postsecondary Chinese class, his HL learning experience consisted mainly of ora l communication with his parents at home. After his dominant language shifted from Mandarin to English right after age five when he started to attend mainstream schools, communication with his parents also became a mixture of Mandarin and English. Compared with YHX, his exposure to the language was more limited. Consequently, his HL knowledge was closely related to what he frequently heard from his parents during everyday life. For those classifiers that he did not hear from his parents, he usually did not have opportunities to obtain the knowledge, even those objects that were common in daily life, such as belt (see Example 6 4), mirror (Example 6 20), and hat (Example 6 21). Example 6 2 0 A CHL l earner WKW NNS: Ye mai le liangge jingzi Also buy PFV two CL mirror Also bought two mirrors. N S: Ni mai le liang mian jingzi 2sg buy PFV two CL mirror. You bought two mirror. NNS :

PAGE 197

197 Liang mian mirror Two CL mirror Two mirrors. Recall data Jingzi jingzi de liangci (the classifier fo r mirror). Usually if it is flat, I use yizhi or yizhang yizhang So whenever I am thinking something flat, I automatically assume it is yizhang I got the impression from my mother. Normally when I do my home said: both of these are yizhang Example 6 2 1 WKW NNS: Wo qu shangdian mai le yi ge maozi I go shop buy PFV one CL hat I went to shop and bought a hat. N S: Ni na ding maozi shi shenme yanse de 2nd that CL hat is what color NOM What is the color of your (that) hat? NNS : Baise de White NOM White. Recall data (At that moment I was thinking that I went shopping for clothing with my In both example s 6 20 and 6 21, the learner did not know the appropriate classifier for either mirror or hat However, a post hoc analysis of his post test showed that he used the correct classifier for mirror but not for hat He did not recall that he learned either of the two classifiers from his mother. This difference could then be explained as follows: Besides being a classifier, mian (for mirror ) also appears in many words which are frequently used in everyday life, such as mianbao (bread), miantiao (noodle). In

PAGE 198

198 co ntrast, ding the classifier for hat is less frequently used. Therefore it could be possible that WKW associated the classifier mian with the words he knew (such as mianbao ) to remember it. In the current study, both the CHL learners and the non CHL learn ers reported that they tend to use this analogous strategy when acquiring an unknown classifier. In contrast, the non HL learners usually relied on the explicit knowledge they learned in the classroom and feedback to figure out their errors. Example 6 22 illustrates this. Example 6 2 2 A non CHL learner ZK NNS: Wo ye mai yi zhang da jingzi 1sg also buy one CL big mirror I also buy a big mirror NS: Yi mian hen da de jingzi ma One CL very big NOM mirror QP A very big mirror? NNS: Dui Correct Correct. Recall data At first I thought the measure word would be zhang for jingzi cause it is flat surface. But it wasn't, I was kind of surprised for a second, I heard you corrected me. In Example 6 23, learner ZK use the rule ( zhang is a classifier that modifies an object with a flat surface) that he learned from the class to make a rea sonable assumption for mirror When a recast was provided by the interlocutor, he realized that he had been corrected. His profile (see Chapter 4) showed that he had been taking Chinese courses at the university for 15 months when the data was collected. H e had never been to China, and had very limited opportunities to practice his Chinese outside

PAGE 199

199 the classroom. Thus his knowledge of Chinese was almost entirely developed in the classroom. Unlike YHX, he did not have rich past experiences of this language th at he could activate. Therefore feedback played an important role in helping him to figure out his untargetlike language. Lastly, it would be interesting to look closely into those who did not make any increase or even regressed from the pre test to the po st test. A post ho c analysis of the pre and post test scores of these learners is presented in table 6 2. Table 6 2 Learner s who increased zero or decreased in the experimental groups Learner Group Test type Pre test Post test Increase TW RHL Oral 46 42 4 SN RHL Oral 90 88 2 SBY RHL Oral 92 88 4 LL MHL Oral 90 88 2 ZZD MHL Oral 23 23 0 LJD RNHL Oral 13 13 0 LAW RNHL Oral 25 21 4 The results in table 6 2 reveal the following interesting facts: the majority of these learners (5/7) are C HL learners. Three of them (SN, SBY, and LL) were those who achieved the maximum scores on the pre test, whereas the other two (TW and ZZD) were those who scored the lowest in the CHL learner group. In contrast, among the two non CHL learners in the recast group, LJD scored relatively lower than the mean score, but was not among those who scored the lowest. The lowest score in the group was only six out of a possible one hundred. The final learner, LAW, scored slightly higher than the mean score ( M : 19.63). For the three CHL learners who have already achieved maximum scores on the pre test, it would not have been not possible for them to further increase their score on the post test

PAGE 200

200 clas sifier, which they used on the pre test and in the treatments. This resulted in their decreased score on the post test. Taking learner LL as an example, she used the accurate classifier shuang for chopsticks on her pre test, as well as during her two treat ments. However, she replaced it with an incorrect classifier tiao on her post test. A similar case also occurred with learner SBY, who used the target classifier pi for horse on his pre test and the two treatments, but replaced it with the incorrect classi fier tiao on the post test It might not be a coincidence that both shuang and pi were the classifiers appeared in the original recorded sentences on the post test in which learners were expected to modify the original sentence if necessary. A plausible explanation for this on their past listening experience when dealing with classifiers, particularly when they have rich exposure to the HL at home. However, these ex periences are not always accurate. Sometimes learners assume an incorrect classifier as an alternative one (see Example 6 6). Therefore for LL and SBY, it is possible that in the past they had heard both the target classifier and the one used in the origin al recorded sentences on the post test, and assumed that they were interchangeable. The above explanation is further evidenced by the fact that the same type of seen among the two non CHL learners as well as learner ZZD who scored t he lowest in the CHL learner group. The two non CHL learners lack of progress was mostly caused by the ir reduced usage of the general classifier ge on their post test. Although this resulted in the decrease of their score, it showed the improvement in t heir awareness of specific classifiers. On the other hand, the CHL test was mostly due to his incorrect choice of

PAGE 201

201 specific classifiers. For instance, he used zhang for dragon, tiao for gloves on the post test. T he in terview data reveals that ZZD had much more limited exposure to Mandarin compared with SN, SBY, and LL: as the third born of the family, and with parents who speak fluent English, ZZD communicates with his family members mostly in English. Although he atte nded Chinese language school once a week between age eight and ten, his formal education in Mandarin at the postsecondary level ( n =1 month) was much shorter compared with the non CHL learners in the current study. In other words, he did not have as wide a range of classifiers as learners SN, SBY, and LL had on the one hand; he also did not have some explicit knowledge that the non CHL learners possessed. Most of the non CHL learners in the current study knew that zhang should be used for an object that is f lat or has a flat surface. None of them used zhang t o classify a dragon. that of SBY and LL: he used the accurate classifier shuang for chopsticks on his pre test, as well as during his two treatments, but replaced it with an incorrect classifier tiao on his post test. Table 6 3 summarizes the profile of TW and SBY, who both decreased the most from the pre test to the post test among the CHL learners. Despite the similariti es in their generation, age of arrival, and schooling in the home country, which learners scored very differently on their tests: one scored the highest in the group the other was one who scored the lowest. The most distinctive difference that set these two

PAGE 202

202 sco red lower than his CHL peers, and could not activate what he acquired by age ten within a two day treatment, in spite of his relatively older age of arrival and much longer formal schooling in the home country compared with other CHL learners in the curren t study, In summary, the CHL learners developed a rich implicit knowledge of classifiers due to their early exposure to the language, which may not always be visible in their oral production. However, with some assistance (such as reoccurrence in the pre and post test), some of this knowledge could be activated, as we saw among the CHL learners in the control group. Being facilitated by feedback, the CHL learners made even greater increases on the oral post test. However, since their initial levels of imp licit knowledge have already approximated ceiling levels, although meta linguistic feedback still outperformed recasts, its superiority was not significant. Those who have scored the maximum even showed slight regression on the post test. Table 6 3. Th e profile of learners TW and SBY TW SBY Generation First First Age of arrival Ten Ten Schooling in home country Four years Four years Cultural identity Less stronger Stronger Home language Step father: native English speaker Mother: English/ Manda rin Father: Mandarin/ English Mother: Mandarin/ English Home literacy environment Not so many Many Written cloze test. In contrast to findings from the oral test, the results of the written tests showed that feedback type significa performance from the pre test to the post test. That is, meta linguistic feedback led to a greater increase than recasts. However, this superiority was found only among the non CHL learners, but not among the CHL l earners. That is, the scores of the non CHL

PAGE 203

203 learners in the meta linguistic feedback group ( M =21.92) increased significantly more than those of the non CHL learners in the recast group ( M =9.17). In contrast, the CHL learners in the two experimental groups increased similarly (15.33 vs. 16.45). As discussed earlier in this chapter, recasts were more explicit in the current study than in some previous studies due to the laboratory setting Consequently both recast and meta linguistic feedback were generally perceived as overtly corrective by learners. However, meta linguistic feedback indicated the source of learners error with meta linguistic information which can help learners obtain rule based knowledge. In contrast, recasts simply reformulated learners problematic utterance without any explanation. Unlike the CHL learners who had intuitive and implicit knowledge, the non CHL learners mostly relied on the feedback to figure out the gap between their utterance and the target linguistic forms. Thus the fee dback type had a greater impact on the non CHL linguistic form was relatively complex, as shown in examples 6 23 and 6 24. The two classifiers for flowers were found to be a ch allenge among the participants of the current study, particularly the non CHL learners. In Example 6 23, the interlocutor clearly indicated that flowers that grow on the same stem should not be classified by duo she understood the 24, the interlocutor only embedded the target classifier in her recast. The learner reported that he was still confused about these two classifiers even after the two day treatment. A post hoc analysis also showed that the learner in Example 6 23 used the correct classifier on the written post test, whereas the learner in Example 6 24 did not.

PAGE 204

204 Example 6 23 A non CHL learner in the meta linguistic feedback group NS: ? Ni mai le duoshao huar 2sg buy PFV how many flower How many flowers did you buy? NNS: Wu duo huar. Five CL flower Five flowers. NS: duo ) NS: Yi zhi hua One CL flower One flower. NNS: Yi zhi hua One CL flower O ne flower. Recall data I was processing, putting into my head, zhi means stick like thing. After you told me, I was like oh, ok, so I should remember zhi f or things that are like sticks. Example 6 2 4 A non CHL learner in the recast group NS: Ni na zhi baise de huar shang you duoshao huar 2sg that CL white NOM flower on have how many flower How many flowers are on that white bunch of flowers? N NS: You wu duo huar Have five CL flower (There are) five flowers. NS: You wu duo huar Have five CL flower (There are) five flowers. NNS: Yo u wu duo huar Have five CL flower (There are) five flowers.

PAGE 205

205 Recall data Did you tell me different measure words, one of them was for flowers, one of them was for the individual flowers, or something? I w as confused right there, c ause I was trying to figure out. I think there was one flower, or one plant with a bunch of flowers. I was trying to say how many flowers were on that plant, but it sounded like I had a bunch of flowers. So I guess I was thinking how am I goanna say this, I am goanna to say the differences between those two things. It is noteworthy that the learner in Example 6 23 did not use the accurate classifier on the oral post test. A possible explanation for this result is that the learner did not have time to draw on his newly acquired explicit knowledge under the time pressure. In fact, the non CHL learners generally scored higher on the written test than on the oral test. I f explicit knowledge is most useful when learners have time to dra w on it, then it is not surprising that the non CHL learners performed better on the untimed written test than the oral imitation test. In contrast, the CHL learners scored lower on their written test than oral test, which also went along with the discussi on presented earlier. That is, these learners had limited explicit knowledge compared with their rich implicit knowledge. related to their knowledge of Chinese characters, as a full mark was given to each target classifier only when it was written in Chinese characters. As reviewed in Chapter 2, HL learners usually develop their HL literacy skills much later than their oral proficiency. Many of them even remain illiterate f or the rest of their life. Previous research did not find that CHL learners had any advantages over non CHL learners in their Chinese character learning (Ke, 1998; Xiao, 2006). In the current study, the average length of postsecondary level formal classroo m instruction of Chinese was 10.9

PAGE 206

20 6 months for the CHL learners, but 19.4 months for the non CHL learners. Table 6 4 and post test. The results in Table 6 4 indicate that with the exception of the CHL learners in the recast group, learners generally show a similar level of knowledge of Chinese characters. The non CHL learners used even more Chinese characters than the CHL learners. Although the non CHL learners in the two experimental groups showed a similar level of knowledge of Chinese characters, those in the meta linguistic feedback group outperformed those in the recast group by approximately eight percent. A possible explanation for this is that the meta linguistic information in the fe edback helped learners connect the target classifier and the associated Chinese character which th ey had acquired in other context s in the past which is illustrated in Example 6 25. Table 6 4 knowledge of Chinese characters on the pre and p ost test CHL learners Recast Meta Non CHL learners Recast Meta Pre test 136 (51.5%) 74 (30.5%) 159 (60.7%) 148 (56.0%) Post test 78 (29.5%) 40 (16.5%) 94 ((35.6%) 77 (29.2%) Pr e test (without ge 4 ) 77 (29.2%) 34 (14.0%) 51 (19.3%) 48 (18.2%) Post test (without ge ) 72 (27.3%) 37 (15.3%) 54 (20.5%) 74 (28.0%) Example 6 2 5 A non CHL learner in the recast group N NS: Yi jian maozi One CL hat One hat N S: on top You wu duo huar Have five CL flower (There are) five flowers. N NS: NS: Top of the head Top of the mountain NNS

PAGE 207

207 Recall data That helps. I will never forget that after you said touding Like the mountain, the top of the mountain. I remembered from that ding would be the top and peak, so. That was good. 6.4 Implications 6.4.1 Theoretical I mplications The main purpose of the current study was to compare the relative effects of explicit and implicit feedback by empirically investigating learners with heterogeneous language background s and learning experie nce s Therefore the findings of the study particularly extended our knowledge of pr e vious findings that dr ew largely from EFL/ ESL learners who had homogeneous language background s and learning experience s Previous research generally found that both exp licit and implicit feedback facilitate d language acquisition. However, in terms of the relative effects of implicit and explicit feedback the results were mix ed Generally, studies conducted in laboratory and CALL settings found no significant differences between implicit and explicit feedback (e.g., Carroll & Swain, 1993; K ang, 2009; Kim & M athes, 2001; Leeman, 2003; Nagata, 1993; Sauro, 2009). On the other hand, studies conducted in classroom settings revealed more complicated results: s ome found no sign ificant differences between implicit feedback and explicit feedback (e.g., De K eyser, 1993; Loewen & Nabei, 2007; Kim & Mathes, 2001; Lyster & Izquierdo, 2009); while others showed that feedback in a more explicit form (e.g., prompts) worked more efficientl y than feedback in a more implicit form (e.g., recasts) (e.g., Ammar & Spada, 2006; Havranek & Cesnik, 2001; Lyster, 2004; Muranoi, 2000; Sheen, 2007 ; R. Ellis et al., 2006 ). The contrasting results found in the previous studies suggest ed that further inve stigation into the relative effects of implicit and explicit feedback wa s necessary.

PAGE 208

208 In the investigation of the relative effects of implicit and explicit feedback, the current study, overall, did not find a significant difference between explicit feedbac k in the form of meta linguistic feedback and implicit feedback in the form of recasts. Although this result was typical in laboratory studies, the current study expanded the generalizability of previous findings particularly by including the following two areas that ha d Learners recruited in previous studies were usually FL or SL learners who o nly received formal la n guage instruction in classrooms. As a result, they had higher level s of explicit knowledge than implicit knowledge. Although extensive research has been conducted to investigat e the effects of implicit feedback in the form of recasts and explicit feedback in the form of meta linguistic feedback few of these studies has recruited learners who had a higher level of implicit knowledge than explicit knowledge due to their exposure to the HL during the childhood. Kang was one of the few researchers who recruit ed HL learners ; background information and include it in the discussion of his results (2009) Ammar s and the effects of implicit and explicit feedback, and found that high proficiency learners benefit ed equally from both types of feedback. In contrast, l ow proficiency learners benefited more from explicit feedback in the form of prompts (including meta linguistic feedback, repetition, and elicitation) than from recasts. However, their learner population had no HL background. The current study filled the gap in the literature by investigating HL learners in comparison with learners without HL background. In

PAGE 209

209 addition, t o the best of my kn owledge, it was the first research that targeted CHL learners feedback, the current study also found that CHL learners perceived feedback more accurately than non CHL learners (Gas s & Lewis, 2007). In addition, the current study took a further step to investigated in the previous research The results showed that CHL learners significantly outperformed non CHL learners i n accurately perceiving meta linguistic feedback This result suggested that meta linguistic feedback may be more effective in orarily from meaning to form during task based interaction, particularly when learners are more m eaning oriented, and possess a high level of implicit knowledge, but lack explicit knowledge to help them notice the gap between their interlanguage and the target language, as was the case with the CHL learners in the current study. Attention that focuse s on forms, as well as noticing of the gap (Schmidt & Frota, However, for the actual change to eventually occur and show up performance a certain amount of time is need ed for learners to incorporate the newly acquired knowledge into their language system (Mackey et al. 2000 ) This argument may explain why meta linguistic feedback did not bring a significant advantage over recasts in the increase of knowledge made by CHL learners in the post test although it helped them to notice more discrepancies between their non targetlike forms and the target like forms

PAGE 210

210 6.4.2 Pedagogical Implications The participants in the current study included a combination of traditional FL lea rners of Chinese and learners who had early exposure to Chinese as a HL. They represent the diverse learner population in postsecondary Chinese programs in the United States. Therefore the findings in the current study provide invaluable implications for C hinese language education, particularly in language curriculum development and classroom instruction. Currently, postsecondary Chinese language classes generally implement either a single track or a dual track language program. The former puts all the le arners into the same curriculum, ignoring their language background; the latter separates learners with CHL background from those without CHL background. A dual track language program usually includes two types of classes: the regular class, in which CHL l earners and non CHL learners attend the same class; and the HL class, in which CHL learners with strong oral speaking proficiency and some literacy skills are enrolled. In either track, the curriculum is designed based on an assumption that the traditional speaking, listening, reading and writing skills should be equally emphasized. The participants of the current study were recruited from these two types of classes. However, the results showed that the CHL learners had a much higher level of oral proficien cy than literacy skills. In contrast, the non CHL learners demonstrated more evenly distributed oral and literacy skills. Thus the current curriculum is apparently more able to accommodate the language needs of non CHL learners than CHL learners. The reul ts of the current study showed a great gap between the oral and written skills presented by the CHL learners. In addition, the CHL learners scored very close to the non CHL learners on their written test. Therefore although CHL learners still need to impro ve their oral productive skills in

PAGE 211

211 certain areas, such as vocabulary, register, etc., they should identify writing skills development as a priority. language needs is strongly needed in the f ield of Chinese language instruction Besides the language curriculum, Chinese language instructors also need to adjust their classroom instruction to maximize the learning effects. The results of the current study showed that the CHL learners tend ed to rely on what they have heard in the past when assessing the accuracy of their language, instead of using rule based grammatical knowledge. In addition, they had a natural tendency to focus their attention more on content rather than linguistic forms when engaging in a conversational their internal language system and the target language system is the precondition for converting input to intake (Schmidt, 1990, 199 3, 1994, 1995), which can further lead to a change in their interlanguage. Therefore, one of the most critical goals in Chinese classroom instruction should be facilitating CHL learners to notice the discrepancies between their problematic output and the t argetlike input provided by the instructor. As CHL learners normally acquire their HL implicitly (at least partially), they tend to lack explicit meta and rule based knowledge that is needed in order to identify the gap. Therefore feedback that explicitly linguistic feedback employed in the current study, can be an efficient tool when the instructor tries zed non target like language. On the other hand, non CHL learners develop their Chinese knowledge mainly through explicit learning in a classroom setting. Therefore they develop explicit

PAGE 212

212 knowledge and literacy skills before implicit knowledge. These l earners need to convert their explicit knowledge into implicit knowledge in order to become more fluent in their oral production. The explicit feedback in the context of a communicative task can facilitate the conversion from explicit knowledge into implic it knowledge. For this reason, implicit feedback such as recasts could provide non HL learners more intentional practice. Lastly, the results of the current study also suggested that both CHL learners and non CHL learners need to improve their literacy s kills, particularly where Chinese characters are concerned. While the oral corrective feedback, particularly explicit feedback, was found to help the non CHL learners to relate the Chinese characters that they acquired in the past to new linguistic context s, it did not work equally efficiently for the CHL learners. Based on these findings, Chinese instructors may find it helpful to provide their feedback in a written form in the instruction of Chinese characters. 6.5 Limitations and Future Researc h Having discussed the potential contributions of the findings to the field of SLA and HL instruction theoretically and pedagogically I will point ou t some li mitations of the current study from the following three aspects: the generalizability of the curr ent findings, the design, and the methodology. First of all, the current study was conducted in a tightly controlled laboratory setting, in which the explicitness of corrective feedback and the linguistic forms were tightly controlled. In addition, the fe edback was delivered intensively on the target linguistic forms in dyadic interactions between an individual participant and the NS interlocutor o n a one to one manner. All these factors might have weakened the ecological validity of the current findings Thus t he implications of the study may not

PAGE 213

213 simply expand to studies conducted in classroom settings, where learners are exposed to mixed types of feedback on various linguistic forms (Lyster & Ranta, 1997; Sheen, 2006). Another limitation that might weaken the generalizability of the current findings concerns the participants of the study. Due to the difficulties of recruiting adequate numbers of participants, the current study only included CHL learners who grew up in a Mandarin speaking family. Learners w ho grew up in a home where a Chinese dialect other than Mandarin was spoken (e.g., Cantonese, Shanghainese, etc), were excluded from the current study. It is not clear whether CHL learners with different dialect background s would benefit from implicit and explicit feedback differently. In addition, the sample size of the current study ( N = 64) was relatively small, due to the difficulty of recruiting a large number of participants at the intermediate level and above for a less commonly taught language. Last ly, due to the small participant population, an experimental group that engages only in an interaction without receiving any type of feedback was not included in the current study. Without a comparison with this interaction only group, it was unclear wheth er the improvements learners made from the pre test to the post test was entirely due to the feedback, or partially due to interaction only. The CHL learners are particularly in question concerning this issue given the fact that scores of the CHL learners in the control group increased considerably from the pre test to the post test, particularly on their oral test. Further studies should include an additional control group to allow for further comparison. Secondly, in terms of th e design shortcoming the current study did not employ a delayed post test due to the difficulty of recruiting adequate subjects who were willing to commit one more day several weeks apart from the three day activity. A d elayed post

PAGE 214

214 test was employed by many researchers to ex amin e the effects of implicit and explicit feedback in the past (e.g., Ammar & Spada, 2006; Carroll & Swain, Kang, 2009; Leeman, 2003; R. Ellis et al. 2006 ; Sheen, 2007 ). R. Ellis and his colleagues did not find that significant differences existed betwee n implicit and explicit feedback groups in the immediate post test ; however, they found them in the delayed post test (R. Ellis et al. 2006). Sheen (2007) found that the meta linguistic feedback group significantly outperformed the recast group on both im mediate and delayed post test s. Previous research found that it takes time for learners to incorporate the target structure into their interlanguage systems, thus the effects of feedback beco me more evident on in the delayed test, p articularly when the eff ects were measured by tests that favored implicit knowledge (Mackey, 1999 also see R. Ellis et al. 2006). Due to the lack of a delayed post test measurements of improvements of target forms in the current study were limited to only short term effects as show n o n the immediate post test thus potentially weaken ing the findings of research questions 2 a nd 3 A final shortcoming of the test design was the inconsistenc y between the oral and written test in terms of identifying non targetlike classifiers. On the written test, learners were required to write in Chinese characters as much as they could, so t hat when a learner used a homonym of the target classifier it could be identified. However, the same type of error could not be idenfified on the oral test. This limitation parcicularly favored the CHL learners on their oral test, as they acquired their HL mainly through listening to their parents. In many cases, they only acquired the oral form rather than the meaning of a classifier. For instance, the CHL learners written test results showed that they often misused zhi which is the classifier for anim al s to classify flowers, which

PAGE 215

215 is supposed to be zhi A possible solution to solve the inconsistency between the oral and the written test in future studies would be to require the learner to orally describe to which zhi he/ she is referring, although t his would detract from the naturalness of the interaction.. Finally, I will discuss a feedback. The current study employed stimulated recall, which ha s been widely used in previous studies to elicit introspective verbal comments about feedback (e.g., Adams, 2003; Egi, 2007; Gass & Lewis, 2007; Mackey et al., 2000; Swain & Lapkin, 2002). However, the method has some limitations: first of all, verbal comments collected through stimulated recall p resent only may not provide verbal comments even though they have noticed feedback. They also may only report one type of perception about feedback when they actually perceive it in several ways (Egi, 2004) Lastly, there were two shortcomings in processing the stimulated recall in the current study. The first one was that the stimulated recall was not conducted immediatel y after the treatment. The second one was that the stimulated recall and the treatment were both conducted by the researcher. As always, further research is needed. 6. 6 Conclusion The present study empirically investigated the effects of implicit feedback in the form of recasts and explicit feedback in the form of meta linguistic fee dback on Chinese The results showed that both feedback t ion of feedback. In addition, both feedback groups showed significant in crease relative to the

PAGE 216

216 control group on the post test indicating the overall effectiveness of meta linguistic feedback and recasts regardless of their language background On the other hand, the study also found that learners have some impacts on the effects of meta linguistic feedback and recasts. The study examined the relative effectiveness of meta linguistic feedback and recasts on an unexplored learner group: CHL learners who acquired Chinese main ly through implicit learning at home and in the Chinese community, with only a short time of formal education at the postsecondary level; in comparison with non CHL learners who acquired Chinese mainly through explicit learning in the classroom. The result s shed some light on how the relative language background, particularly with the following findings: a) the CHL learners in the explicit feedback condition perceive d feedback more accurately than the n on CHL learners in the implicit feedback condition ; b) explicit feedback proved superior to recasts in prom o ting the non CHL learners writing skills ( p =.002) However, the same superiority was not seen for the CHL learners. However, due to the limitatio ns in design and methodology, further research is needed to increase our understanding of how C HL learners acquire or require their C HL. 1 The transcription remained the original comments made by learners without modifying their grammatical errors. 2 The transcription remained the original comments made by learners without modifying their grammatical errors. 3 This may not apply to all the CHL learners as some of them skip the first year of the Chinese course. 4 The general classifier ge was the most fr equently used Chinese character on the written post test.

PAGE 217

217 APPENDIX A TRANSCRIPTION CONVEN TIONS CL Classifier CP Compliment CRS Currently relevant state ( le ) NOM nominalizer PFV perfective aspect ( le ) Pre preposition QP question particle 1sg First person singular 2sg Second person singular 3sg Third person singular 1pl First person pl ural

PAGE 218

218 APPENDIX B CONSENT FORM I.R.B No.: 2009 U 0172 Informed Consent Protocol Title: The effects of implicit and explicit feedback Please read this consent document carefully before you decide to participate in this study. Purpose of the research study: The purpose of this study is to examine the effects of feedback techniques on heritage Chinese language learners. What you will be asked to do in the study: You will participate in three days of activities: Day 1: You will be inter viewed by the researcher about your Chinese heritage language learning experience at home (approximately 30 40 minutes). The interview will be audio taped. You will also complete a set of written and oral questions (five minutes for each) in Chinese, the p urpose of which is to measure your current proficiency level in Chinese; Day 2: You will be asked to tell a story based on provided pictures. In between, the researcher may ask you some questions related to your story (approximately 40 minutes). The proces s will be both video and audio taped. You will also complete a set of written and oral questions in Chinese in order to confirm your proficiency level in Chinese. Day 3: You will be shown some video clips chosen from the interview of Day 1, and will be ask ed to give some comments about them (approximately one hour). Your participation in the study has no impact on your school grade, and your results of written and oral questions will not be communicated to your instructors. The process will be audio taped. All the video and audio recordings will only be used for data analysis purposes. Time required: 3 4 hours Risks and Benefits: You will be practicing Chinese with a native speaker by participating in the study. No risks will be involved in the study. Com pensation: No compensation will be given for participating in this research. Confidentiality: Your identity will be kept confidential to the extent provided by law. Your information will be assigned a code number. The list connecting your name to this num ber will be kept in a locked file in my faculty supervisor's office. When the study is completed and the data have been analyzed, the list will be destroyed. Your name will not be used in any report. Voluntary participation: Your participation in this stu dy is completely voluntary. There is no penalty for not participating. Right to withdraw from the study: You have the right to withdraw from the study at anytime without consequence.

PAGE 219

219 Whom to contact if you have questions about the study: Han Ye, Graduat e Student, Linguistics Program, phone: 352 262 8841. Email: hanye@ufl.edu 4131 Turlington Hall, PO Box 115454 Gaine sville FL 32611 5454. Antes, Theresa. Professor, Romance Languages and Literatures, phone: (352) 392 2101 x 236, Email: antes@rll.ufl.edu Dauer Hall PO BOX 115565, Gain esville Florida Whom to contact about your rights as a research participant in the study: IRB02 Office, Box 112250, University of Florida, Gainesville FL 32611 2250; phone 392 0433. Agreement: I have read the procedure described above. I voluntarily ag ree to participate in the procedure and I have received a copy of this description. Participant: _____________________________ Date: _________________ Principal Investigator: _____________________ Date: _________________

PAGE 220

220 APPENDIX C QUESTIONNAIRE 1. Name ________________________________________________ (both your Chinese and English name) 2. Gender: F M 3. Class Standing: Freshman/Sophomore/ Junior / Senior/ Graduate _______________________ 6. Were you born in the U.S.? Yes. No. 7. If your answer is No in No.4, your age of immigration _________________ Did you migrate from the following regions? Please circle one: Mainland China Taiwan Hong K ong 8. What was the first language you spoke? English Mandarin Cantonese Others ____________ 9. Which language do you speak with your parents NOW (Choose all the languages you speak with them)? English Mandarin Cantonese Others _____________ 10. Which language do you speak with your siblings NOW (Choose all the languages you speak with them)? English Mandarin Cantonese Others _____________ 11.Whi ch language do your parents speak with each other (Choose all the languages they speak)? English Mandarin Cantonese Others _____________ 12. Which language do you speak with your friends who can speak Chinese NOW? Man darin Cantonese Others _____________ English _____________ 13. Which language do you speak with your grand parents? English Mandarin Cantonese Others _____________

PAGE 221

221 14. Which language did your p arents teach you how to write at home? Mandarin Cantonese English _____________ Others _____________

PAGE 222

222 APPENDIX D WRITTEN PRE TEST Please complete the following sentences. You can write either in Chinese characters or in Pin Yin 1. _____________ _______________ I have three cats and two horses. 2. __________ _____________ I ate a chicken for dinner. 3. _________________________ My little sister drew a dragon. 4. _______ _____________ s pet is a snake. 5. _____________ ____________ There are a pair of chopsticks and two pairs of scissors on the table. 6. ________ __________ Tomorrow is my birthday, my mom bought a pair of shoes for me. 7. ___________ ________________ I lost one of my gloves yesterday when I went to watch a movie. 8. ________ _________ He is really weird, he only wears one shoe on his foot. 9. _____________ ___________ There is one toothbrush and one pen in her bag. 10. _________ ________ ________________ ______________ Yesterday I went to a shopping center and bought a lot of stuff: two shirts, one sweater, and one belt. 11. ____________________ _________________ 12. ______ ______ ______ I spent $15000 on buying a car. 13. _________ ___________________ __________________ 14. __________ __________________ _____________________ 15. ___________________ 16. ______________ There are eleven pearls in my bracelet. 17. _____________ ___ ______________________________

PAGE 223

223 18. ________ ________________ At the Mid Autumn festival, I ate two mooncakes. 19. ________________________ There are twenty three tables in the classroom. 20. __________ __ ________________________ There is beef on the plate on the table in the kitchen 21. ________________ _________________ There are five leaves and five flowers in the box. 22. ______________________ _________________ My sister wants to learn cross stitch, so she went to the shop and bought three sewing needles.

PAGE 224

224 APPENDIX E WRITTEN POST TEST Please complete the following sentences. You can write either in Chinese characters or in Pin Yin. 1. _______ _____________ ake. 2. _____________ _______________ I have three cats and two horses. 3. __________ _____________ I ate a chicken for dinner. 4. ______________________ _________________ My sister wants to learn cross stitch, so she went to the shop and bought three sewing needles. 5. ________________________ There are twenty three tables in the classroom. 6. ___________ _________________ I lost one of my gloves yesterday when I went to watch a movie. 7. __________ ______ ____________ _____________________ 8. _____________ _________________________________ 9. _____________ ___________ There i s one toothbrush and one pen in her bag. 10. __________ __________________________ There is beef on the plate on the table in the kitchen 11. _____________ ____________ There are a pair of chopsticks and two pairs of scissors on the table. 12. ____ _____ _________ ________________ ______________ Yesterday I went to a shopping center and bought a lot of stuff: two shirts, one sweater, and one belt. 13. ________________ _________________ There are five leaves and five flowers in the box. 14. __ __________________ _________________ 15. _________________________ My little sister drew a dragon.

PAGE 225

225 16. _________ ___________________ __________________ nt to give him a tie and a hat. 17. ___________________ 18. ______________ There are eleven pearls in my bracelet. 19. ________ __________ Tomorrow is my birthday, my mo m bought a pair of shoes for me. 20. ________ ________________ At the Mid Autumn festival, I ate two mooncakes. 21. ______ ____________ I spent $15000 on buying a car. 22. ________ _________ He is really weird, he only wears one shoe on his foot.

PAGE 226

226 APPENDIX F TREATMENT 1

PAGE 227

227 APPENDIX G TREATMENT 2

PAGE 228

228 LI ST OF REFERENCES Adams, K. L. (1991). The Influence of Non Austroasiatic Languages on Numeral Classification in Austroasiatic. Amastae, J. (1982). Language shift and maintenance in the lower Rio Grande Valley of Southern Texas. In F. Ba rkin & J. Ornstein Galicia (Eds.), Bilingualism and language contact: Spanish, English and native American languages (pp. 261 277) New York: Teachers College Press. Ammar, A. (2008). Prompts and recasts: Differential effects on second language morphosynta x. Language Teaching Research, 12, 183 210. Ammar, A. & Spada, N. (2006). Once size fits all? Recasts, prompts, and L2 learning. Studies in Second Language Acquisition, 28, 543 574. Anderson, R. W. (1982). Determining the linguistic attributes of language attrition. In R.D. Lambert & B. F. Freed (Eds.), The loss of language skills (pp. 83 118) Rowley, MA: Newbury House. Argus, M. V. & Matthews, E. K. (1991). Stratigraphic excavation techniques for paleontologists. Journal of Paleontology, 17, 119 127. Au, T.,LM, K.,SA, J. & JS, O. (2002). Overhearing a language during childhood. Psychological Science, 13, 238 243. Ayoun, D. (2004). The Effectiveness of Written Recasts in the Second Language Acquisition of Aspectual Distinctions in French: A Follow Up Study. Bachman, L. F. (1990). Fundamental considerations in language testing. Oxford: Oxford University Press. Bialystok, E. (1994). Representation and ways of knowing: Three issues in second language acquisition. In N. Ellis (Ed.), Implicit and explicit learnin g of languages (pp. 549 569) London: Academic Press Limited. Bitchener, J. (1999). The negotiation of meaning by advanced ESL learners: the effects of individual learner factors and task type. Unpublished doctoral dissertati o n, The University of Auckland. Braidi, S. M. (2002). Reexaming the role of recasts in native speaker/ nonnative speaker interactions. Language Learning, 42, 1 42. Carreira, M. (2004). Seeking explanatory adequacy: A dual approach to understanding Heritage Language Journal 2, 1 28. Retrieved October 20, 2009, from www.international.ucla.edu/article.asp?parentid=14647 Carroll, S. (2001). Input and evidence: The raw materi al of second language acquisition. Amsterdam: John Benjamins. Carroll, S. & Swain, M. (1993). Explicit and implicit negative feedback: An empirical study of the learning of linguistic generalizations. Studies in Second Language Acquisition, 15, 357 366. Ch amot, A. U. & O'Malley, J. M. (1994). Language learner and learning strategies. In N. Ellis (Ed.), Implicit and explicit learning of languages (pp. 371 392) London: Academic Press Limited. Chang, H. W. (1983). Pre schooler's use of classifiers in Mandarin Chinese. National Taiwan University Press. Chang, I. (2003). The Chinese in America. New York: The Penguin Group. Chao, Y. R. (1968). A grammar of spoken Chinese. University of California Press.

PAGE 229

229 Chen, P. (1999). Modern Chinese: History and sociolinguistic s. Cambridge: Cambridge University Press. Cho, G. (2000). The role of heritage language in social interactions and relationships: Reflections from a language minority group. Bilingual Research Journal, 24, 369 384. Chomsky, W. (2001). David Kimhi's Hebrew grammar (Mikhlol). New York: Bloch. Dai, j. h. E. & Zhang, L. (2008). What are the CHL learners inheriting? Habitus of the CHL learners. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 37 51) Honolulu: U niversity of Hawai'i Press. DeKeyser, R. M. (1993). The effect of error correction on L2 grammar knowledge and oral proficiency. The Modern Language Journal, 77, 501 514. DeKeyser, R. M. (1995). Learning second language grammar rules: An experiment with a miniature linguistic system. Studies in Second Language Acquisition, 17, 379 410. DeKeyser, R. M. (1997). Beyond explicit rule learning: Automatizing second language morphosyntax. Studies in Second Language Acquisition, 19, 195 221. Dekeyser, R. M. (1998). Beyond focus on form: cognitive perspectives on learning and practicing second language grammar. In C. J. Doughty & J. Williams (Eds.), Focus on form in second language acquisition (pp. 42 63) New York: Cambridge University Press. Drnyei, Z. (1994). Mot ivation and motivating in the foreign language classroom. Modern Language Journal, 78, 273 284. language acquisition? TESOL Quarterly, 20, 305 325. Duff, P. (1986). Another lo ok at interlanguage talk: Taking task to task. In R. Day (Ed.), Talking to learn: Conversation in second language acquisition (pp. 147 181) Rowley, MA: Newbury House. Egi, T. (2004). Verbal reports, noticing, and SLA research Language Awareness, 13, 243 2 64. Egi, T. (2007). Interpreting recasts as linguistic evidence: The roles of linguistic target, length, and degree of change Studies in Second Language Acquisition (pp. 511 537) Ellis, N. C. & Larsen Freeman, D. (2006). Language emergence: Implications for applied linguistics introduction to the special issue. Applied Linguistics 2006, 27, 558 589. Ellis, R. (1993). Second language acquisition and the structural syllabus. TESOL Quarterly, 27, 91 113. Ellis, R. (1994). The study of second language acquis ition. Oxford: Oxford University Press. Ellis, R. (2005). Measuring implicit and explicit knowledge of a second language: A psychometric study. Studies in Second Language Acquisition 141 172. Ellis, R. (2007). The differential effects of corrective feedba ck on two grammatical structures. In A. Mackey (Ed.), Conversational interaction in second language acquisition (pp. 339 360) Oxford: Oxford University Press.

PAGE 230

230 Ellis, R.,Basturkmen, H. & Loewen, S. (2001). Learner uptake in communicative ESL lessons. Langu age Learning, 51, 281 318. Ellis, R.,Loewen, S. & Erlam, R. (2006). Implicit and explicit corrective feedback and the acquisition of L2 grammar. Studies in Second Language Acquisition, 28, 339 368. Ellis, R. & Sheen, Y. (2006). Re examing the role of recas ts in second language acquisition. Studies in Second Language Acquisition, 28, 576 600. Ely, C. (1986). An analysis of discomfort, risktaking, sociability, and motivation in the L2 classroom. Language Learning, 36, 238 244. Erbaugh, M. (1982). Coming to or der: Natural selection and the origin of syntax in the Mandarin speaking child. Unpublished doctoral dissertation University of California, Berkeley. Erbaugh, M. (1986). Taking stock: The development of Chinese noun classifiers historically and in young children. In C. Craig (Ed.), Noun classes and categorization (pp. 399 435) Philadelphia: John Benjamins. Everson, M. E. (1998). Word recognition among learners of Chinese as a foreign language: Investigating the relationship between naming and knowing. Mo dern Language Journal, 82, 194 204. Fang, F. (1985). An experiment on the use of classifiers by 4 to 6 year colds. Acta psychologyca sinica, 17, 384 392. Fishman, J. A. (1991). Reversing language shift: Theoretical and empirical foundations of assistance to threatened languages. Clevedon, England: Multilingual Matters. Fishman, J. A. (2001). 300 plus years of heritage language education in the United States. Washington DC: Center for Applied Linguistics/ Delta Systems. Flege, J. E.,Munro, M. J. & Mackay, I. R. A. (1995). Factors affecting strength of perceived foreign accent in a second language. Journal of the Acoustical Society of America, 97, 3125 3134. Gardner, R. C. (1985). Social psychology and second language learning: The role of attitudes and moti vation. London: Edward Arnold. Gass, S. M. (1997). Input, interaction, and the second language learner. Mahwas, N.J.: Lawrence Erlbaum Associates. Gass, S. M. & Lewis, K. (2007). Perceptions about interactional feedback: differences between heritage langua ge learners and non heritage language learners. In A. Mackey (Ed.), Conversational interaction in second language acquisition (pp. 79 99) Oxford: Oxford University Press. Gass, S. M. & Mackey, A. (2000). Stimulated recall methodology in second language re search Mahwah, NJ: Erlbaum. Gass, S. M. & Varonis, E. (1986). Sex differences in nonnative speaker nonnative speaker interactions. In R. Day (Ed.), Talking to learner: Conversation in second language acquisition (pp. 327 351) Rowley, MA: Newbury House. Gr Applied Linguistics, 5, 79 100. Hadley, A. O. (2001). Teaching language in context. Boston, MA: Heinle & Heinle. Han, Y. (2000). Grammaticality judgment tests: How reliable and valid are they. Applied Li nguistics, 11, 177 204. Han, Z. (2002). A study of the impact of recasts on tense consistency in L2 output. TESOL Quarterly, 36, 543 572.

PAGE 231

231 Harred, J. F.,Knight, A. R. & McIntyre, J. S. (1972) Dow Chemical Company, USA. Havranek, G. & Cesnik, H. (2001). Fact ors affecting the success of corrective feedback. In S. Foster Cohen & A. Nizegorodzew (Eds.), EUROSLA Yearbook (pp. Amsterdam: Benjamins. He, A. W. (2004). Identity construction in Chinese heritage language classes. Pragmatics, 14, 199 216. He, A. W. (200 6). Toward an identity theory of the development of Chinese as a heritage language. Heritage Language Journal 4, 1 28. Retrieved October 20, 2009, from http://www.international.ucla.edu/ media/files/AH.pdf He, A. W. (2008a). Chinese as a heritage language: An introduction. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 1 12) Honolulu: University of Hawai'i Press. He, A. W. (2008b). An identity based model for the development of Chinese as a heritage language. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 1 12) Honolulu: University of Hawai'i Press. Heigham, J. & Sakui, K. (2009). E thnography. In J. Heigham & R. Croker (Eds.), Qualitative research in applied linguistics: A practical introduction (pp. 91 111) Hampshire: Palgrave Macmillan. Hendrys, J. D. (2008). The Chinese heritage language learners' existing linguistic knowledge an d abilities. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 53 66) Honolulu: University of Hawai'i Press. Hendryx, J. D. (2008). The Chinese heritage language learners' existing linguistic knowledge and abilities. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 53 66) Honolulu: University of Hawai'i Press. Hinton, L. (1999). Involuntary language loss among immigrants: Asian American linguistic autobiog raphies. ERIC Digest Retrieved March 18, 2010, from http://www.cal.org/resources/digest/digest_pdfs/9910 hinton langloss.pdf Hu, Q. (1993). The acquisition of Chine se classifiers by young Mandarin speaking children. Unpublished doctoral dissertation Boston University. Huang, s. (1993). Yuyan, shehui yu zuqun yishi. [Language, society, and ethnicity]. Taipei: Crane. Isurin, L. & Ivanova Sullivan, T. (2008). Lost in between: The case of Russian heritage speakers. Heritage Language Journal 6, 72 103. Retrieved November 4, 2009, from http://www.international.ucla.edu/media/files/I+I S_HLJ_6 _1.pdf Iwashita, N. & Liem, I. (2005). Factors Affecting Second Language Achievement in Primary School. Australian Review of Applied Linguistics, 25, 1 21. Jia, G. (2008). Heritage language development, maintenance, and attrition among recent Chinese imm igrants in New York City. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 189 203) Honolulu: University of Hawai'i Press. Jia, L. (2009). Contrasting models in literacy practice among heritage language l earners of Mandarin. Journal of Asian Pacific Communication, 19, 56 75.

PAGE 232

232 Jiang, S. (2008). Developing a Chinese placement instrument in response to the diversity in a student population at the university level. In T. Hudson & C. Martyn (Eds.), Case studies in foreign language placement: Practices and possibilities (pp. 119 131) Honolulu: University of Hawai'i Press. Kang, H. S. (2009). The relative efficacy of explicit and implicit feedback in the learning of a less commonly taught foreign language. IRAL, 4 7, 304 324. Ke, C. (1998). Effects of language background on the learning of Chinese Characters among foreign language students. Foreign Language Annals, 31, 91 100. Killingley, S. y. (1981). The Semantic grouping of mensural classifiers in Cantonese. Anth ropological Linguistics, 23, 383 435. Kim, H. S. (2008). Heritage and nonheritage learners of Korean: Sentence processing differences and their pedagogical implications. In K. B. Kimi & B. James Dean (Eds.), Teaching Chinese, Japanese, and Korean heritage language students (pp. 99 134) New York: Lawrence Erlbaum Associates. Kim, H. S. H. (2006). Processing strategies and transfer of heritage and non heritage learners of Korean. Unpublished doctoral dissertation The University of Hawai'i. Kim, H. & Mathes G. (2001). Explicit vs. implicit corrective feedback. The Korea TESOL Journal, 4, 1 51. Koda, K.,Lv, C. & Zhang, Y. (2008a). Effects of print input on morphological awareness among Chinese. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fo stering rooted world citizenry (pp. 125 135) Honolulu: University of Hawai'i Press. Koda, K.,Zhang, Y. & Yang, C. L. (2008b). Literacy development in Chinese as a heritage language In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fostering r ooted world citizenry (pp. 125 135) Honolulu: University of Hawai'i Press. Kondo Brown, K. (2005). Differences in language skills: Heritage language learner subgroups and foreign language learners The Modern Language Journal, 89, 563 581. Kondo, K. (1997) Social psychological factors affecting language maintenance: Interviews with Shin Nisei University students. Linguistics and Education, 9, 369 408. Krashen, S. (1981). Second language acquisition and second language learning. London: Pergamon. Krashen, S (1982). Principles and practice in second language acquisition. New York: Pergamon. Krashen, S. (1985). The input hypothesis: Issues and implications. Londaon: Longman. Lambert, W. E. & Taylor, D. (1996). Language in the lives of Ethnic minorities: Cuban American families in Miami. Applied Linguistics, 17, 477 500. Lee, J. S. (2002). The Korean language in America: The role of cultural identity in heritage language learning. Language, Culture, and Curriculum, 15, 117 133. Leeman, J. (2003). Recasts and se cond language development: Beyond negative evidence. Studies in Second Language Acquisition, 25, 37 63. Leeman, J. & Martinez, G. (2007). From identity to commodity: Ideologies of Spanish in heritage language textbooks. Critical Inquiry in Language Studies 4, 35 65.

PAGE 233

233 Leow, R. (1998). Toward operationalizing the process of attention in SLA: Evidence for Tomlin and Villa's (1994) fine grained analysis of attention. Applied Psycholinguistics, 19, 133 159. Leow, R. & Morgan Short, K. (2004). To think aloud or n ot to think aloud: The issue of reactivity in SLA research methodology. Studies in Second Language Acquisition, 26, 35 57. Leow, R. P. (1997). Attention, awareness, and foreign language behavior. Language Learning, 47, 467 505. Leow, R. P. (2000). A study of the role of awareness in foreign language behavior: Aware versus unaware learners. Studies in Second Language Acquisition, 22, 557 584. Li, D. & Duff, P. A. (2008). Issues in Chinese heritage language education and research at the postsecondary level. I n H. A. W & X. Y (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 13 36) Honolulu: University of Li, G. (2006). Culturally contested pedagogy : Battles of literacy and schooling between mainstream teachers and As ian immigrant parents. Albany: State University of New York Press. Li, R. (2008). Using a web delivered questionnaire to improve placement in Chinese. In T. Hudson & M. Clark (Eds.), Case studies in foreign language placement: Practices and possibilities ( pp. 111 117) Honolulu: University of Hawai'i Press. Liang, N. S. y. (2008).The acquisition of Chinese shape classifiers by L2 adult learners. In M. K. M. Chan & H. Kang (Eds.), the 20 the North American Conference on Chinese Linguistics : 309 326) Columbus Ohio: The Ohio State University. Liu, Y. (2008). The effects of error feedback in second language writing. Arizona Working Papers in Second Language Acquisition and Teaching, 15, 65 79. Loewen, S. (2004). Uptake in incidental focus on form in meaning foc used ESL lessons. Language Learning, 54, 153 188 Loewen, S. & Nabei, T. (2007). Measuring the effects of oral corrective feedback on L2 knowledge. In A. Mackey (Ed.), Conversational interaction in second language acquisition (pp. 339 361) Oxford: Oxford University Press. Loewen, S. & Philp, J. (2006). Recasts in the adult English L2 classroom: Characteristics, explicitness, and effectiveness. The Modern Language Journal, 90, 536 556. Loke, K. K. (1991). A semantic analysis of young children's use of Manda rin shape classifiers. In A. Kwan Terry (Ed.), Child language development in Singapore and Malaysia (pp. 98 116) Singapore University Press. Loke, K. k. & Harrison, G. (1986). Young Children's use of Chinese (Cantonese and Mandarin) aortal classifiers In H. S. R. Kao & r. Hoosain (Eds.), Linguistics, Psychology, and the Chinese Language (pp. 125 146) University of Hong Kong. Long, M. (1988). Instructed interlanguage development. In L. Beebe (Ed.), Issues in second language acquisition: Multiple perspecti ves (pp. 115 141) Rowley, Mass: Newbury House. Long, M. (1996). The role of the linguistic environment in second language acquisition In W. Ritchie & T. Bhatia (Eds.), Handbook of second language acquisition (pp. 413 468) San Diego, CA: Academic Press.

PAGE 234

234 Long, M. (2007). Recasts in SLA: The story so far. In M. Long (Ed.), Problems in SLA (pp. Mahwah, NJ: Erlbaum. Lu, X. & Li, G. (2008). Motivation and achievement in Chinese language learning: A comparative analysis. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 89 108) Honolulu: University of Hawai'i Press. Lynch, A. (2008). The linguistic similarities of Spanish heritage and second language learners. Foreign Language Annals, 41, 252 281. Lyster, R. (19 98a). Negotiation of form, recasts, and explicit correction in relation to error types and learner repair in immersion classrooms. Language Learning, 48, 183 218. Lyster, R. (1998b). Recasts, repetition and ambiguity in L2 classroom discourse. Studies in S econd Language Acquisition, 20, 51 81. Lyster, R. (2004). Research on form focused instruction in immersion classrooms: implications for theory and practice. Lyster, R. (2007). Learning and teaching languages through content: A counterbalanced approach. Am sterdam: Benjamins. Lyster, R. & Izquierdo, J. (2009). Prompts Versus Recasts in Dyadic Interaction. Lyster, R. & Mori, H. (2006). Interactional Feedback and Instructional Counterbalance. Studies in Second Language Acquisition, 28, 269 300. Lyster, R. & Ra nta, L. (1997). Corrective feedback and learner uptake: Negotiation of form in communicative classrooms. Studies in Second Language Acquisition, 19, 37 66. Mackey, A. (1999). Input, interaction, and second language development: An empirical study of questi on formation in ESL. Studies in Second Language Acquisition, 21, 557 587. Mackey, A. (2006). Feedback, noticing and instructed second language learning. Applied Linguistics, 27, 405 430. Mackey, A. & Gass, S. M. (2005). Second language research: Methodolog y and design. Mahwah: NJ: Lawrence Erlbaum Associates. Mackey, A.,Gass, S. M. & McDonough, K. (2000). How do learners perceived interactional feedback? Studies in Second Language Acquisition, 22, 471 497. Mackey, A. & Philp, J. (1998). Conversational inter action and second language development: Recasts, responses, and red herrings. The Modern Language Journal, 82, 338 356. Mak, D. (1991). The acquisition of classifiers in Cantonese. Unpublished doctoral dissertation University of Reading. McDonough, C. (2 005). The role of prosody and complexity in infants' detection of repeated speech patterns. Dissertation Abstracts International, 65, 4873 B. McElree, B.,Jia, G. & Litvak, A. (2000). The time course of conceptual processing in three bilingual populations. Journal of Memory and Language, 42, 229 254. McLaughlin, B. (1978). The monitor model: Some methodological considerations. Language Learning, 28, 309 332. Modern Language Association. (2006). Enrollments in languages other than English in United States ins titutions of higher education. Retrieved November 15, 2009, from http://www.mla.org/pdf/06enrollmentsurvey_final.pdf

PAGE 235

235 Montrul, S. (2002). Incomplete acquisition and attrition of Spanish t ense/aspect distinctions in adult bilinguals. Bilingualism: Language and Cognition, 5, 39 68. Montrul, S. (2004). Subject and object expression in Spanish heritage speakers: A case of morphosyntactic convergence. Bilingualism: Language and Cognition, 7, 12 5 142. Montrul, S. (2008). Second language acquisition welcomes the heritage language learner: Opportunities of a new field. Second Language Research, 24, 487 506. Montrul, S.,Foote, R. & Perpin, S. (2008). Gender agreement in adult second language learn ers and Spanish heritage speakers: The effects of age and context of acquisition. Language Learning, 58, 503 553. Muranoi, H. (2000). Focus on form through interaction enhancement: integrating formal instruction into a communicative task in EFL classroom. Language Learning, (57), 511 548. Nabei, T. & Swain, M. (2002). Learner awareness of recasts in classroom interaction: A case study of an adult EFL student's second language learning. Language Awareness, 11, 43 63. Nagata, N. (1993). Intelligent computer f eedback: For second language instruction. The Modern Language Journal, 77, 330 339. Nakahama, Y.,Tyler, A. & Van Lier, L. (2001). Negotiation of meaning in conversational and information gap activities: A comparative discourse analysis. TESOL Quarterly, 35 377 405. Nassaji, H. (2007). Elicitation and reformulation and their relationship with learner repair in dyadic interaction. Language Learning, 57, 511 548. Nassaji, H. (2009). Effects of recasts and elicitations in dyadic inteaction and the role of feed back explicitness. Language Learning, 59, 411 452. National Language Committee. (2005). Figures of Chinese language usage survey. Retrieved October 22, 2009, from http://www.moe.gov.cn/edoas/website18/level3.jsp?tablename=1262&infoid=144 93 National Standards in Foreign Language Education Project (1999). Standards for foreign language learning in the 21st century. (pp. KS: Allen Press, Inc. Nicholas, H.,Lightbown, P. & Spada, N. (2001). Recasts as feedback to language learners. Language Learning, 51, 719 758. Norris, J. M. & Ortega, L. (2000). Effectiveness of L2 instruction: a research synthesis and quantitative meta analysis. Language Learning, 50, 417 528. Ohta, A. S. (2000). Rethinking recasts: A learner centered examination of corrective feedback in the Japanese language classroom. In J. K. Hall & L. S. Verplaeste (Eds.), The construction of second and foreign language learning through classroom instruction (pp 47 71) Mahwah, NJ: Lawrence Erlbaum Associates. Oliver, R. (1995). Negative feedback in child NS NNS conversation. Studies in Second Language Acquisition, 18, 459 481. Oliver, R. (2000). Age differences in negotiation and feedback in classroom and pairw ork. Language Learning, 50, 119 151. Panova, I. & Lyster, R. ( 2002 ). Patterns of corrective feedback and uptake in an adult ESL classroom. TESOL Quarterly, 36, 573 595.

PAGE 236

236 Philp, J. (2003). Constraints on "noticing the gap": Nonnative speakers' noticing of recasts in NS NNS interaction. Studies in Second Language Acquisition, 25, 99 126. Pica, T. (1987). Second language acquisition, social interaction and the classroom. Applied Linguistics, 81, 3 21. Polinsky, M. (2008). Gender under incomplete acquisition: knowledge of noun categorization. Heritage Language Journal 6, 40 71. Retrieved November 4, 2009, from http://www.international.ucla.edu/media/files/M_P_HLJ_6 _1.pdf Rober, A. (1989). Implicit learning and tacit knowledge. Journal of experimental psychology, 118, 219 235. Romanova, N. (2008). Mechanisms of verbal morphology processing in heritage speakers of Russian. Heritage Language Journal 6 105 126. Retri eved October 20, 2009, from http://www.heritagelanguages.org Rosa, E. & Leow, R. (2004). Awareness, different learning conditions, and second language development. Applied Psycholinguistics, 25, 269 292. R osa, E. & O'Neill, M. D. (1999). Explicitness, intake, and the issue of awareness. Another piece of puzzle. Studies in Second Language Acquisition, 21, 511 556. Russo, J. E.,Johnson, E. J. & Stephens, D. L. (1989). The validity of verbal protocols. Memory and Cognition, 17, 759 769. Snchez, R. (1983). Chicano discourse: Socio historic perspectives. Rowley, MA: Newbury House. Sanz, C.,Lin, H. J.,Lado, B.,Bowden, H. W. & Stafford, C. A. (2009). Concurrent Verbalizations, Pedagogical Conditions, and Reactivit y: Two CALL Studies. Sauro, S. (2009). Computer mediated corrective feedback and the development of L2 grammar. Language Learning & Technology, 13, 96 120. Schmidt, R. (2001). Attention. In P. Robinson (Ed.), Cognition and second language instruction (pp. 3 32) New York: Cambridge University Press. Sheen, Y. (2004). Corrective Feedback and Learner Uptake in Communicative Classrooms across Instructional Settings. Sheen, Y. (2006). Exploring the relationship between characteristics of recasts and learner upt ake. Language Teaching Research, 10, 361 392. Sheen, Y. (2007). The effects of corrective feedback, language aptitude, and learner attitudes on the acquisition of English articles. In A. Mackey (Ed.), Conversational interaction in second language acquisiti on : A series of empirical studies (pp. 301 322) Oxford: Oxford University Press. Slimani Rolls, A. (2005). Rethinking task based language learning: What we can learn from the learners. Language Teaching Research, 9, 195 218. Sun, C. (2006). Chinese: A lin guistic introduction. Cambridge: Cambridge University Press. Swain, M. (1985). Communicative competence: Some roles of comprehensible input and comprehensible output in its development. In S. Gass & C. Madden (Eds.), Input in Second Language Acquisition (p p. 235 256) New York: Newbury House.

PAGE 237

237 Swain, M. (2005). Verbal protocols: What does it mean for research to use speaking as a data collection tool? In M. Chaloub Deville, M. Chapelle & P. Duff (Eds.), Inference and generalizability in applied linguistics: Multiple research perspectives (pp. 97 113) Amsterdam: Benjamins. Swain, M. & Lapkin, S. (2002). Talking it through: Two French immersion learners' response to reformulation. International Journal of Educational Research, 37, 285 304. Szeto, K. S. K. (199 8). The acquisition of Cantonese classifiers. Unpublished doctoral dissertation University of Hong Kong. Tse, L. (2000). The effects of ethnic identity formation on bilingual maintenance and development: An analysis of Asian American narratives. Internat ional Journal of Bilingual Education and Bilingualism, 3, 185 200. Valds, G. (1995). The teaching of minority languages as academic subjects: Pedagogical and theoretical challenges. The Modern Language Journal, 79, 299 328. Valds, G. (2001). Heritage lan guage students: Profiles and possibilities. In J. K. Peyton, D. A. Ranard & S. McGinnis (Eds.), Heritage languages in America: Preserving a national resource (pp. 37 77) McHentry: the Center for Applied Linguistics/ Delta Systems. Valds, G. (2005). Bil ingualism, heritage language learners, and SLA research: opportunities lost or seized? The Modern Language Journal, 89, 410 426. Van Deusen Scholl, N. (2003). Toward a definition of heritage language: sociopolitical and pedagogical considerations. Journa l of Language, Identity, and Education., 2, 211 230. Varonis, E. & Gass, S. M. (1985). Non native/non native conversations: a model for negotiation of meaning. Applied Linguistics, 6, 71 90. Warden, M.,Lapkin, S.,Swain, M. & Hart, D. (1995). Adolescent lan guage learners on a three month exchange: Insights from their diaries. Foreign Language Annals, 28, 537 550. Weger Guntharp, H. (2006). Voices from the Margin: Developing a profile of Chinese heritage language learners in the FL classroom. Heritage Languag e Jou r nal 4, 29 46. retrieved November 11, 2009, from http://www.international.ucla.edu/languages/heritagelanguages/journal/article.asp ?parent id=53016 Wiley, T. G.,Klerk, G. d.,Li, M.,Liu, N.,Teng, Y. & Yang, P. (2008). Attitudes toward Mandarin, heritage languages, and dialect diversity among Chinese immigrants and international students in the United States. In H. A. Weiyun & X. Yun (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 67 87) Honolulu: University of Hawai'i Press. Wong Fillmore, L. (1991). When learning a second language means losing the first. Early Childhood Research Quarterly, 6, 323 346. Wu, S. M. (2008). Robust learning for Chinese heritage learners: Motivation, linguistics and technology. In K. Kondo Brown & J. D. Brown (Eds.), Teaching Chinese, Japanese, and Korean heritage language students (pp. 271 297) New York: Lawrence Erlbaum Associates.

PAGE 238

238 Xiao, Y. (2006). Heritage learners in the Chinese language classroom: Home background. Heritage Language Journal 4, 47 56. Retrieved November 10, 2009, from http://www.international.ucla .edu/media/files/YX.pdf Xiao, Y. (2008). Home literacy environment in CHL development. In A. W. He & Y. Xiao (Eds.), Chinese as a heritage language: Fostering rooted world citizenry (pp. 151 166) Honolulu: University of Hawai'i Press. Yeni Komshian, G. H.,Flege, J. E. & Liu, S. (2000). Pronunciation proficiency in the first and second languages of Korean English bilinguals. Bilingualism: Language and Cognition, 3, 131 149. Zentella, A. C. (1997). Growing up bilingual. Malden, MA: Blackwell Publishers. Zh ang, H. (2007). Numeral classifiers in Mandarin Chinese. Journal of East Asian Linguistics, 6, 43 59.

PAGE 239

239 BIOGRAPHICAL SKETCH Ye Han was born in Beijing, China. She received her B.A. in Teaching Chinese as a Second Language from Beijing Language and Cultu re University in China She received her M.A. in Journalism from Sophia University in Tokyo Japan. She received her Ph. D in L inguistics from the University of Florida in the summer of 2010.