Signs of Learning: Child Nonnative Speakers in Tutoring Sessions With a Child Native Speaker

27
Signs of Learning: Child Nonnative Speakers in Tutoring Sessions With a Child Native Speaker Sabrina Peck Los Angeles City College Learning acts, which included conversation-maintaining acts and language- learning acts, were coded and compared in a group of native Spanish-speaking children (n = 9) during individual tutoring sessions with a native English speaker. In each session, an 8 year-old NS of English tutored one kindergartener for about twenty-five minutes at an elementary school. Three kindergarteners had been rated high, three average and three low In overall achievement by their teachers. For each child, the total number of words and number of unique words were computed. In the nine tape recordings, the high group ranked ahead of the low group in the quantity and sophistication of their English, and the number of learning acts pro- duced. The medium group resembled the high group more than the low group on most features. Each group had a characteristic pattern of learning act use. This study is an investigation of one setting for child second language acquisition: a tutoring session at school in which an older L1 child helps a younger L2 child.* The work reported here is part of a broader study, a description of both teaching by the child native speaker and learning by the nine nonnatives (Peck 1985). This paper focuses on discourse acts through which the L2 child may be learning English. The study is concerned with two broad questions. The first is, how do child nonnative speakers talk to a child native speaker in a peer-tutoring session at school? What is the range of their behaviors and of their apparent knowledge of English? The answer to this question is a descrip- tion. The second question is more speculative: how might children be learning language through conversations with an older child native speaker? More specifically, given a thorough description of the range of behaviors, at what points might the nonnatives be learning some aspect of English? The two questions are answered here in the context of a Los *I would like to thank Evelyn Hatch, Roger Andersen, Thomas Weisner, Barbara Hawkins and Vanessa Flashner for their help with this project, as well as Michael Long and Richard Day for their constructive criticism of the manuscript. 545

Transcript of Signs of Learning: Child Nonnative Speakers in Tutoring Sessions With a Child Native Speaker

Signs of Learning: Child Nonnative Speakers in Tutoring Sessions With a Child Native Speaker

Sabrina Peck Los Angeles City College

Learning acts, which included conversation-maintaining acts and language- learning acts, were coded and compared in a group of native Spanish-speaking children (n = 9) during individual tutoring sessions with a native English speaker. In each session, an 8 year-old NS of English tutored one kindergartener for about twenty-five minutes at an elementary school. Three kindergarteners had been rated high, three average and three low In overall achievement by their teachers. For each child, the total number of words and number of unique words were computed. In the nine tape recordings, the high group ranked ahead of the low group in the quantity and sophistication of their English, and the number of learning acts pro- duced. The medium group resembled the high group more than the low group on most features. Each group had a characteristic pattern of learning act use.

This study is an investigation of one setting for child second language acquisition: a tutoring session at school in which an older L1 child helps a younger L2 child.* The work reported here is part of a broader study, a description of both teaching by the child native speaker and learning by the nine nonnatives (Peck 1985). This paper focuses on discourse acts through which the L2 child may be learning English.

The study is concerned with two broad questions. The first is, how do child nonnative speakers talk to a child native speaker in a peer-tutoring session at school? What is the range of their behaviors and of their apparent knowledge of English? The answer to this question is a descrip- tion. The second question is more speculative: how might children be learning language through conversations with an older child native speaker? More specifically, given a thorough description of the range of behaviors, at what points might the nonnatives be learning some aspect of English? The two questions are answered here in the context of a Los

*I would like to thank Evelyn Hatch, Roger Andersen, Thomas Weisner, Barbara Hawkins and Vanessa Flashner for their help with this project, as well as Michael Long and Richard Day for their constructive criticism of the manuscript.

545

546 Language Learning VOI. 37, No. 4

Angeles-area elementary school, and ten subjects (an English NS second grader and nine Spanish-speaking kindergarten children). The nine learn- ers included three successful children, three average, and three unsuc- cessful, according to holistic ratings by their teachers.

Chosen by their teachers, these nine children represent the range of academic and social achievement for Hispanic kindergarteners at one school. An analysis of their speech helps to define a range of language proficiency and language use of such children. The analysis yields a quantitative range for their English proficiency and for their language- learning acts.

The research questions are: 1) How does the output (total words, words per minute, and unique words) differ for the three groups? 2) How do the three groups compare in the number and range of learning acts? a) as measured by conversation-maintaining acts, b) as measured by language-learning acts.

BACKGROUND

For children acquiring English, or any second language, discourse with other children who speak the new language is an important part of their exposure, and thus one part of the process of language learning. Researchers of first-language acquisition have described child-child dis- course and shown that the older child can adjust his or her speech for younger children (Shatz & Gelman 1973; Sachs & Devin 1975). Some of the adjustments convey affection, and others simplify the input and probably make it more comprehensible for the young child. Similarly, in a second language study, child native speakers adjusted and simplified their speech for child nonnatives (Wong Fillmore 1975). Many studies of child-child discourse by a native and nonnative speaker have focused on how the native speaker-learner (L 1-L2) conversations might contribute to language learning.

Most researchers have concentrated on the contributions of the child native speaker to the other child’s learning, but some have described learner strategies and achievements as well. Gilmore (1977) described a Swahili-based private pidgin language that was constructed by two pre- school children, neither a native speaker of Swahili. Several studies have reported language play and suggested that the play could provide a useful and fun way for the nonnative speaker to learn about the new language (Berman, undated; Peck 1978; Nemoianu 1980.)

Peck 547

Although counterexamples exist (Saville-Troike, McClure & Fritz 1982), conversational experience appears very important for many chil- dren’s second-language acquisition. In a study of five Spanish-speaking learners of English, Wong-Fillmore (1 979) outlined strategies that her subjects used while acquiring English over a school year. In conversa- tions with their L1 friends, learners used a variety of techniques to main- tain conversations with peers. They took advantage of contextual clues, they pretended that they understood, and they built their grammars around formulaic expressions. In another study set in an elementary school, (McClure, Saville-Troike & Fritz 1982), limited English-speaking children maintained conversations with other children by using nonver- bal cues and routines, by repeating a previous utterance, and by extend- ing the functions of their English vocabulary and grammar.

When learners act to maintain the conversation, whatever learning goes on or could go on is also maintained. In a study of an adult learner, Chesterfield (1980) found three such acts: the repair of communication breakdown, appeals to the children for information, and the repetition of a correction. Scarcella and Higa (1982) found a number of ways in which L2 adolescents worked more actively than L2 children to sustain conversations with an NS. The adolescent speech contained more ques- tions, more self- and other-repetition, more initiations of cooperative dialogues, more topic initiation and more conversational fillers.

Thew (1979) developed a taxonomy of language-learning acts for her comparison of talk by child-older sibling and child-mother pairs. Thew found that the young children’s speech was richer in language-learning acts with their mothers than with their older siblings. She also found evidence of more competition in the child-child pairs.

METHOD

SUBJECTS

The 10 subjects were all students at the same Los Angeles-area elemen- tary school. Nine of the children were Mexican or Mexican-American kindergarten students and native Spanish speakers. The tenth was a sec- ond grader and native speaker of English.

The nine children were chosen from a pool of 20 children at the school who fulfilled all the criteria for the study. They were all in their first year of schooling and ranged in age from approximately 4 to 5 years. They

548 Language Learning Vol. 37, No. 4

came from three different kindergarten classrooms. According to tests administered by the bilingual coordinator’s office, they all had at least average intelligence. They all spoke Spanish as a first language. In English they ranged from complete beginners to bilinguals who could probably succeed in an all-English kindergarten.

In December, when the children had been in school for about 10 weeks, each kindergarten teacher made a holistic rating of her own stu- dents who were eligible for the study. Teachers rated each child as high, medium or low. The instructions to the teachers read: “DO you think of him or her as high-achieving, compared to other children from Spanish- speaking families that you have taught, medium achieving, or low achieving? [Olverall achievement [is to be considered,] not necessarily achievement in learning or using English.” The researcher discussed the questionnaire briefly with each teacher, pointing out that an overall rat- ing would take into account such factors as the child’s social maturity, small muscle coordination (using scissors, drawing), listening skills, and arithmetic achievement as well as English or Spanish oral language and pre-reading skills.

Of the 20 children who appeared eligible for the study, five were rated high, or unusually successful overall. Two of these high-rated children could not be included in the study: because one was a kindergarten repeater, and the other a frequent absentee who was never tape recorded. There were three low-rated, or unusually unsuccessful children, all of whom were included in the study. The remaining 12 children, 10 of whom had already been tape recorded, were rated medium. Tapes from three of these medium children were chosen for analysis: two girls, so that the number of boys and girls in the study would be more closely balanced, and one boy, because the tape quality for his session was higher than that of the other available sessions. Thus, tapes from three children from each group (high, medium and low) were chosen for the study, for a total of nine learners.

Several important variables were not controlled for, such as level of English, socioeconomic status, parents’ education, and sex. The goal of the study was not to isolate variables that most or least contribute to child second language acquisition, but to describe a range of behaviors by a group of children from the same school.

The child native speaker, who is called Ed here, was a Black American, second-grade student at the same school as the learners. He was bright, and ahead of his classmates in all subjects. Ed was chosen from several

Peck 549

second graders, all recommended by their teachers. All were willing to help “teach games to kindergarten kids who speak Spanish and some English,” as the researcher expressed it. As training, Ed played the games twice, then practiced teaching them to the researcher, who role-played a Spanish-speaking kindergarten child. Ed was chosen over the other sec- ond graders because he was more outgoing and talkative in the role play, and also because his teacher was willing to let him leave the classroom at more periods of the day. Ed knew a few Spanish words for numbers, a few colors, and classroom commands such as “iSientate!” (Sit down!). Many of his classmates were Spanish speakers, though all of them spoke English as well.

PROCEDURE

Each kindergarten child was met at his classroom, and asked if he or she wanted to go and play some games in English with the researcher and Ed. Several children chose not to participate, but all except one decided to participate at a subsequent time, after seeing their classmates leave and safely return with Ed and the researcher.

Each learner met for about 25 minutes with Ed. The researcher sat at the table with the children. She took notes, tape recorded the session, and sometimes talked with the children. Ed directed three activities, and at times helped the learners with unfamiliar words or directions.

The three activities (a game of fish, a story retelling task, and a picture sheet task) allowed Ed to elicit and teach familiar color, number and animal vocabulary, as well as a short narrative. The cards for the first task, the fish game, were pairs of squares, triangles and circles in five colors. Next, in the story task, Ed told a story about a preschool-age boy who saw a bear sleeping in his bed. The boy immediately told his family, who initially ignored him, and finally listened to him. While Ed told the story, he showed photocopies of pictures from the book version of the story (Guilfoile 1957). He then asked the learner to retell the story while using the same pictures. For the last activity, the picture sheet task, Ed and the learner used sheets that had the same nine, small pictures mounted on them, but ordered and numbered differently on the two sheets. First, Ed described each picture on his sheet, asking the learner for that picture’s number on the other sheet. Then, using another set of sheets, the learner described the pictures to Ed, who supplied their num- bers on his sheet.

550 Language Learning Vol. 37, No. 4

Over a period of one month, Ed taught the games to 17 different kindergarten children. Most weeks, he conducted one or two sessions per day, two days a week. As described earlier, the tapes for 9 of the 17 children were chosen for transcription and analysis: the three high chil- dren, the three low children, and three of the remaining children, all rated average.

Transcription of the nine audiotapes included notation of latches, overlaps, pause length, as well as phonemic transcription of unintelligi- ble syllables. The letter x was used to indicate a completely unintelligible syllable. Each child was assigned a three-letter code. The codes end in -h, -m, or -1, for the high, medium, and low children, respectively.

In order to count each child’s words, two edited subtranscripts were produced from each transcript. One subtranscript was made up of only Ed’s words, and one of only the learner’s words. The researcher’s talk and the children’s talk to the researcher were removed, along with any commentary. Differing phonemic forms of the same word were replaced with normal spelling so that they would all be counted as instances of the same word. Other phonemic notations were replaced with x‘s. Next, counts of the total number of words and the total number of unique words were obtained by running the WordPlus program (Micropro Corp.) on these child-words-only versions of the transcripts. WordPlus defines words as sequences of letters bounded by spaces. Thus, x‘s or sequences of x‘s, which stand for unintelligible syllables here, were counted as words. This procedure ensured that unintelligible syllables contributed to the word count. Since self-repetitions were also counted, the word count reflects how much the children spoke, including dysfluencies.

The tasks in each audiotape were timed. In each session, some time was taken up with the researcher’s talk to the children and the children’s talk to her. This talk lasted from an estimated 30 seconds in one session to 6 minutes and 40 seconds in another. Adjusted times were obtained by subtracting the estimated duration of talk by and to the researcher in each session from the raw time for that session. These ranged from 27.6 to 15.0 minutes. The average adjusted time for all nine sessions was 23.06 minutes.

Four output measures were computed for each transcript. The mea- sures were minutes per session, total number of words, number of unique words, and number of words per adjusted time. Adjusted counts were obtained for all four measures. To do this, the average adjusted time for

Peck 551

all sessions, 23.06 minutes, was divided by the adjusted time for the session in question. The resulting number was a normalizing factor for that session. For example, for a high-rated child called Veh, the normal- izing factor was 23.06 divided by 26.88 (her adjusted time), or .86. Adjusted counts were then calculated by multiplying the raw counts by the normalizing factor for that child. For instance, the adjusted number of words for Veh was obtained by multiplying Veh’s total raw number of words produced (421) by Veh’s normalizing factor (.86), yielding 362 adjusted words.

The transcripts were coded with two kinds of features: conversation- maintaining acts and language-learning acts. The researcher coded all nine transcripts, and another experienced rater learned the coding system and coded three of the transcripts. The two raters agreed substantially, though no formal measure of interrater reliability was made.

The conversation-maintaining acts were largely adapted from Scar- cella and Higa (1982). Counts of conversation-maintaining acts are intended to show how actively and often the nonnative speakers worked to maintain the conversation. The features are:

1 . A ratio of the total raw number of the learner’s words to the total raw number of words in that session;

2. The learner shows a lack of understanding;

3. The learner repeats Ed’s utterance;

4. The learner starts a new topic;

5 . The learner uses fillers; and

6. The learner uses self-repetitions.

Most of the features are expressed as frequencies (the number of occurrences) and this generally means the number of turns in which the feature appears. However, 1) is a ratio, 5 ) refers to the number of fillers, not to the number of turns containing fillers, and 6) refers to the number of self-repetitions.

In light of Scarcella and Higa’s study (1982), it was expected that the high group would have the highest mean number of learning acts, fol- lowed by the medium and then the low groups.

Six features in this study define language-learning acts by the kinder- garten children:

552 Language Learning Vol. 37, No. 4

1. Ostensive definitions. The learner points to something, perhaps a color or a story character and names it. The effect is that the learner shares the name with Ed.

2. The learner repeats Ed’s correction of truth or expression. By repeating the correction, the learners acknowledge, at least, that they heard the correction.

3. The learner makes a sincere request for information.

4. The learner responds to Ed’s insincere request for information. For example, Ed asks “What color?” and the learner responds, correctly or incorrectly.

5 . Contingent queries. The learner asks a question about some part of Ed’s immediately preceding utterance.

6 . The learner responds to Ed’s contingent queries.

Two of the features, (4) and (6), describe responses to Ed‘s teaching acts. These two features are expressed as ratios, but the other language- learning features are expressed as frequencies. Raw frequencies were determined, and adjusted frequencies were then computed, using the same normalizing factors that were described earlier.

In light of Thew’s work, Ed could conduct himself more like a parent or more like a sibling. If Ed acted like a parent with all the learners, the three groups might show mean numbers of learning acts according to the level of their English. By this reasoning, the high group would rank first in learning acts, followed by the medium, and then the low group. If Ed acted more like a sibling, one would expect a great deal of competition, rivalry, and negative feedback. One would expect the lowest mean of learning acts for the high group because the high-rated children might have the most competition with Ed. In this scenario, the ranking for the mean number of learning acts would be:

Low > Medium > High.

Of the language-learning features, three were expressed as ratios of some utterance by Ed to some response by the learner. The other features were frequencies. Raw counts were computed for each feature. Adjusted counts were computed as described above.

Chi-square tests were used to describe the significance of the apparent

Peck 553

difference among the adjusted group means. In several cases, where expected cell means were lower than 5 , chi square could not be inter- preted. For some of these cases, Kruskal-Wallis tests were used.

RESULTS

OUTPUT RESULTS

Question 1: How does the output differ for the three groups?

Tables 1 and 2 detail the output data for each group.

1. Minutes per session. The actual times for the nine sessions ranged from 29.6 to 17.8 minutes. Adjusted times ranged from 27.6 to 15.0 minutes. The average times for the three groups were insignificantly different.

2. Total number of words. The range for the total number of words per learner was from 760 to 0 (raw) and 669 to 0 (adjusted). The high group produced the greatest adjusted number of words, while the low children produced the lowest number.

3. Unique words. The range for the number of unique words per learner was from 147 to 0 words (adjusted). The high children as a group pro- duced the highest mean number of unique words, the medium children ranked second, and the low group ranked third.

4. Number of words per adjusted time. The range was 28.9 to 0 words per adjusted minute. As a group, the high children produced the largest number of words per minute, the medium children ranked second and the low children ranked third. However, two of the medium children actually produced about the same number of words as two of the high children. Thus the medium group was not significantly different than the high group.

The high children ranked above the low children in both the quantity and sophistication of their English. The high group was significantly higher than the low group on all output measures, except for minutes per session. On the other hand, the medium and high groups were not signifi-

Tabl

e I

Out

put D

ata

for

the

Nin

e Le

arne

rs

Hig

h G

roup

M

ediu

m G

roup

Lo

w G

roup

Alh

V

eh

Gah

Xh

SDh

Mam

M

om

JSm

X

m

SDm

K

al

Jel

Jql

Xl

SDI

la.

Min

utes

per

se

ssio

n (R

) 29

.0

28.3

17

.9

25.0

6.

2 22

.9

29.6

25

.1

25.9

3.

4 17

.8

27.6

29

.6

25.0

6.

3 Ib

. (A

) 26

.3

26.9

17

.4

23.5

5.

3 21

.6

27.6

24

.0

24.4

3.

0 15

.0

21.5

27

.5

21.3

6.

3

2a.

Tota

l num

ber

of w

ords

(R

) 76

0 42

1 30

3 49

5 23

7 46

4 47

2 30

0 41

2 97

.1

0 67

17

3 80

87

.2

2b.

(A)

669

362

403

478

167

496

396

288

393

104

0 72

14

5 72

72

.5

3a.

Num

ber

of

uniq

ue w

ords

(R

) 16

7 10

9 10

8 12

8 33

.8

130

138

79

116

32.0

0

34

75

36

37.5

3b

. (A

) 14

7 94

14

4 12

8 29

.8

139

116

76

110

31.9

0

36

63

33

31.6

4.

Raw

wor

ds p

er

adju

sted

min

utes

28

.9

15.7

17

.4

20.7

7.

2 21

.5

17.1

12

.5

17.0

4.

5 0

3.1

6.3

3.1

3.1

Tabl

e 2

Chi

-squ

are

Test

s on

Gro

up M

eans

for

Out

put D

ata

Hig

h/M

ed./L

ow

Hig

h/M

ed.

Hig

h/L

ow

Med

./Low

d

f=2

d

f=l

df=

l d

f=l

lb.

Min

utes

per

ses

sion

(A)

.22

(NC

) (N

C)

(NC

)

2b.

Tota

l # w

ords

(A

) 29

2. **

* 8.

10**

29

8.**

* 22

0. * *

* 3a

. #

uniq

ue w

ords

(R)

53.6

***

.5

50.4

9**

41.0

6***

4 W

ords

(R) pe

r m

ins.

(A)

12.7

**

.39

13.1

***

8.28

**

Key

: (A

) =

adj

uste

d da

ta

(R) =

raw

dat

a (NC) =

test

not

con

duct

ed;

*p <

.05,

*

p < .O

1 *I*,

< .0

01.

556 Language Learning Vol. 37, No. 4

cantly different from each other on any of the measures except total number of words. As for individuals in the medium and high groups, two medium-learners and two high-learners had similar scores on the output features. In summary, the children formed two groups. One was made up of the low children. The other was made up of the medium children and all of the high children except Alh.

As individuals, the nine children varied widely on all three word mea- sures. Two children, Kal and Alh, were at each end of the range. Kal did not speak at all during her session. At the other extreme, Alh talked the most (669 total words, adjusted), used the largest number of unique words (147, adjusted) and talked the most per minute (28.9 raw words per adjusted minute). The other children were somewhere between Kal and Alh in the output measures, and as discussed above, they fell into two groups, not three.

RESULTS FOR LEARNING ACTS: CONVERSATION-MAINTAINING ACTS

Question 2a: How do the three groups of learners compare in the number and range of learning acts as measured by conversation-maintaining acts?

Tables 3 and 4 detail the conversation-maintaining acts for each group. In light of Scarcella and Higa’s study (1982), it was expected that the

high group would have the highest mean number of learning acts, fol- lowed by the medium and then the low groups. As we will see, the hypothesis was supported for three of the five features.

I . NNS words per total words by the NNS and NS combined. The children in the high group produced a higher proportion of the total words in each session than did the low group. One of the high children, Alh, produced a much higher proportion (.44) of the total words than did any of the other children. The rest of the high and medium children grouped together. The low children also grouped together, producing 0, .06, and .12 of the total words.

Several children produced about the same proportion of the total words but they used different word-producing techniques such as, self- repetition, fillers, and repetition. In the story task for example, Gah frequently used two conversation-maintaining acts; self-repetition and fillers:

Tabl

e 3

Conv

ersa

tion-

Mai

ntai

ning

Act

s by

Each

Lea

rner

Hig

h G

roup

M

ediu

m G

roup

Lo

w G

roup

Alh

V

eh

Gah

X

h SD

h M

am

Mom

Js

m

Xm

SD

m

Kal

Je

l Jq

l X

l SD

l 1.

N

NS

wor

ds/

NS

+ N

NS

wor

ds (R

)

2a.

Show

lack

of

unde

rsta

ndin

g (R)

2b.

(A)

3a.

Rep

eat

NS

3b.

(A)

utte

ranc

e (R

)

4.

Star

t a

new

top

ic (R

)

5a.

Use

5b.

(A)

fille

rs (

R)

6a.

Self-

re

petit

ions

(R)

.44

7 6.2

22

19.4

1 84

74

45

.27

11

9.5

10

8.6

0 13

11.2

3

.27

.33

.1

.26

4 7.

33

3.51

0

5.3

7.00

2.

21

0

19

17.0

6.

24

57

25.3

17

.8

8.47

61

0 .3

3 .5

8 0

31

42.7

36

.9

17

41.2

42

.1

31.4

18

.2

22

23.3

21

.0

14

.27

1 .8

50

42

0 24

20.2

0

.I9

1 1 20

19.2

0 61

58.6

0

.24

.I5

0 .0

6

.67

.58

0 1

.60

.53

0 1

42.3

19

.6

0 6

40.7

20

.9

0 6.

4

0 0

0 0 4

34

23.6

0

32.3

22

.8

0 4.

3

4.7

8.1

0 2.

0

.I2

.06

.06

0 .3

3 .5

8 0

.33

.58

27

11

14.2

22

.7

9.1

11.7

0 0

0

17

I 8.

89

14.3

6.

2 7.

3

0 .6

7 1.

15

2.1

0 .7

0 1.

07

6b.

(A)

39.6

2.

6 29

.3

23.8

19

.1

15

0 0

4.7

8.7

0

Tabl

e 4

Chi

-squ

are

and

Kru

skal

- Wal

lis T

ests

on

Gro

up M

eans

for

Con

vers

atio

n-M

aint

aini

ng a

cts

Hig

h/M

ed ./

Low

H

igh/

Med

. H

igh/

Low

M

ed./L

ow

df=

2

df=

1

df=

l d

f=l

1.

NN

S w

ords

/NS

+ N

NS

wor

ds (R

) 6.

71##

(4

(e

l (e

l 2b

. Sh

ow la

ck o

f un

ders

tand

ing

(A)

5.72

##

(el

(4

(e)

3b.

Rep

eat

NS

utte

ranc

e (A

) 22

.7**

* 8.

2**

1.83

17

.9* *

* 4.

St

art a

new

top

ic (

R)

(NC

) (N

C)

WC

) (N

C)

5b.

Use

fill

ers

(A)

25.6

***

1.04

25

.2**

* 16

.4* *

* 6b

. Se

lf-re

petit

ions

(A

) 31

.3**

* 11

.5**

* 19

.9**

* 1.

7 K

EY: (

R)

= r

aw d

ata,

(A

) =

adj

uste

d da

ta,

(e)

= e

xpec

ted

cell

mea

ns o

r sa

mpl

e siz

e to

o sm

all,

(NC

)= t

est

not

cond

ucte

d.

*p <

.05;

**p

< .0

1; *

**p

< ,0

01 - f

or C

hi-s

quar

e tes

ts.

#p c .

05; #

#p <

.01;

###

p <

,001

~ fo

r K

rusk

al-W

allis

test

s.

Peck 559

T17Ga: 4 Ed: He’s going down where?

Ed: No, down STAIRS Gah: With, with his mother.

Gah: (sigh) He say, Mother! There’s, is, in, in, in my, in my room, is, is a bear. (p.4:) Father, is a, in the in my room, is, is a, is er, is a bear. (. . .)

While retelling the same part of the story, Mam repeated Ed’s utterances, line by line:

TSMA: 7 Ed: Andrew came running downstairs, saying there’s a bear in my room

Mam: /n/ tair. Ed: (prompting:) Andrew,

Mam: Andrew ca- Ed:

Mam: Andrew came, Ed: run //ning//

Mam: //run// ning down, Ed: the stairs,

Mam: /I/ downstairs

-ca. Andrew came running dow:n,

It is evident that although Mam and Gah each produced about .27 of the total words in their sessions, a larger proportion of Gah’s words was fillers or part of self-repetitions, and a larger proportion of Mam’s was part of other-repetitions.

2. Show lack of understanding. The children in the high group showed Ed their lack of understanding more than did the other children. Veh (from the high group) ranked first on this feature, and showed her lack of understanding 11 times. She asked Ed for clarification and for defini- tions; in this example she wanted a confirmation:

T8V: (10) (Ed is asking Veh to name colors before the fish game) Veh: Green Ed: No:! This is.

Veh: This over here? Ed: This right, THERE . . .

Veh: Blue Ed: This,

Veh: This? Ed: Mm hm.

Veh: Red.

560 Language Learning Vol. 37, No. 4

The infrequency with which the medium and low learners showed their lack of understanding is striking. In the following example from the low group, Jql did not understand the rules for the fish game, and he was also unfamiliar with the English names for colors and shapes. Even so, he lasted out the fish game without asking for explanations.

T14JQ: 4 Ed: Do you have an red triangle? . . . Square. Say RED triangle. I mean, red

Jql: (shakes his head) Ed: No? . . . Now you ask me. (6s) Ask me. (6s) Ask me do I have a, o:range

circle or-orange square (. . .) Now ASK me again. Do you have a brown circle or orange, circle?

square.

Jql: (nods) Ed: (whining:) Which one? . . . No . . . Ask me again. (points) Ask me! Do I

have a:, a umm, square, whatever (. . .)

Jql, along with the other low children often appeared not to have the needed English vocabulary or enough confidence to show his lack of understanding to Ed.

3. Repeat NS utterance. The groups are significantly different. The medium children repeated Ed’s utterances more than the high children did, and more than the low children, who ranked last on this feature. Much of the repetition of Ed’s utterances occurred during the story task. While the high children generally retold the story themselves and the low children spoke little during this task, the medium children retold much of the story by repeating after Ed, turn by turn.

These two examples, both from the story task, contrast Mom, who repeated Ed’s utterances more than any other child, with Jel, who rarely repeated. Here Mom recounted how the bear was captured:

T13MO: 9 Ed: The zoocatcher,

Mom: Zoo, catcher, Ed: Got the net.

Mom: Got the net. Ed: And put it over his hea:d.

Mom: put it over head. Ed: He couldn‘t fit the cage.

Mom: He couldn’t /ke/. Ed: So he just put it, /ps/, so he just put his HEAD. In there.

Mom: And he put his head, in there,

Peck 561

Jel, a low child, did not repeat after Ed in the story task even after Ed urged him to:

TlOJe: 6 Ed: (. . .) Andrew came, (10s) Andrew came . . . Say it. (15s) Andrew came

running downstairs, say it (8s) Andrew came running downstairs . . . . . . You have the book . . . Look at the book. Andrew came, (10s) (to researcher:) Doesn’t say it.

4. Start a new topic. Only one child, Alh, started a new topic. After the fish game, Ed was starting to introduce the next task, the story about Andrew:

T2AL: 5 Ed: Could you read?

Ed: Could you read?

Ed: (points to book) What does it say?

Ed: Tape recorder

Alh: What?

Alh: I can’t READ.

Alh: Uh . . . . . . (points to tape recorder) For what this?

The other children stayed within the bounds of the task at hand, of the instructions to the task, or of being drilled on the names of colors or objects. Alh seemed to be the most fluent, as well as the most assertive and competitive of the,learners. He stood out from the other high chil- dren on several other measures that will be discussed later.

5. Use fillers. All of the learners (except Kal, who did not speak at all during her session) used some fillers (4-84, raw data). The high and medium groups each used significantly more than the low group. For each group, the standard deviation is high.

6. Self-repetition. The high group used self-repetition significantly more than the other two groups did. However, note that four children (two medium and two low) did not produce any self-repetitions. The follow- ing example illustrates Alh’s frequent use of self-repetition. He is explaining that Andrew told the next door neighbor about the bear.

TZAL: 10 Alh: He nex-, nex-nex-next door neighbor, shouting, “Next door neighbor! Uh-

one bear is in my, my room, uh, and he-

562 Language Learning Vol. 37, No. 4

Of all the children, Alh produced the most self-repetitions. In summary, the high group appeared to be the best at maintaining the

conversation. They talked more. They used significantly more fillers than did the low children, and more self-repetition than either of the other groups. They showed their lack of understanding more often to the native speaker.

The medium children seemed second best at maintaining the conversa- tion. They repeated Ed’s utterances significantly more than either of the other groups. Typically, they repeated while retelling the story, where a high child might have formed original utterances or a low child might have used a series of ostensive definitions.

The low children appeared the least able to maintain the conversation. This is not surprising, considering how little they spoke compared to the children in the other groups. They also spoke little, relative to Ed, and were unlikely to repeat Ed’s utterances. They rarely showed Ed their lack of understanding, and they used fillers and self-repetition infrequently. Like all of the other children except Alh, the low children initiated no new topics.

RESULTS FOR LEARNING ACTS: LANGUAGE LEARNING ACTS

Question 2b: How do the three groups of learners compare in the number and range of learning acts as measured by language-learning acts?

Tables 5 and 6 detail the language learning acts for each group.

I . Ostensive definitions. The three groups were not significantly differ- ent. The mean for the low group was the highest, followed by the high and then the medium group. Three individuals, each from a different group, produced the most ostensive definitions. In their transcripts, most of the ostensive definitions are found during the story task. Jql, for example, got through the story task mainly by naming characters in the pictures:

Tl4JQ: 8 Ed: Say there’s a bear in my room! Jql: Bear in room, Ed: (turns to the next page, a picture of Andrew, his father, his mother, and the

cleaning lady) Jql: Girl . . . Mother. Mother. Ed: (turns to the next picture, which shows the brother and the sister)

Tabl

e 5

Lang

uage

-Lea

rnin

g A

cts

by A

N N

ine

Lear

ners

Low

Gro

up

Hig

h G

roup

M

ediu

m G

roup

Alh

V

eh

Gah

X

h SD

h M

am

Mom

Js

m

Xm

SD

m

Kal

Je

l Jq

l XI

SDI

la.

Ost

ensi

ve

37

13

20.8

lb

. (A

) 1.

8 14

.6

9.3

8.6

6.4

0 4.

2 16

.3

6.8

8.5

0 2.

1 31

11

17

.3

defi

nitio

ns (R

) 2

17

7 8.

7 7.

6 0

5 17

7.

3 8.

7 0

2

2a.

Rep

eat

a

2b.

(A)

.9

1.7

4 2.

2 1.

6 0

3.4

1 1.

5 1.

7 0

1 1.

7 2.

1 3.

4 1.

5 1.

7 co

rrec

tion

(R)

12

3

21

0

4 1

1.7

2.1

0 1

4

3a.

Sinc

ere

requ

est

for

info

rmat

ion

(R)

6 0

5 3.

7 3.

2 0

1 1

.7

.6

01

0

.33

.6

3b.

(A)

5.3

0 6.

6 4.

0 3.

5 0

.8

1 .6

.5

0

1 0

.3

3 .6

4a.

Res

pond

to a

n in

- si

ncer

e re

ques

t fo

r in

form

atio

n (R

) 7/

8 18

/19

10/1

3 7/

10

17/2

9 13

/17

0/4

4/19

8/

23

Dec

imal

for

m:

.88

.95

.77

.86

.09

.7

.59

.76

.68

.09

0 .2

1 .3

5 .1

9 .1

8

5a.

Con

tinge

nt

quer

ies

(R)

66

2

4.7

2.3

1 0

1 .7

.6

0

1 0

.3

3 .6

5b

. (A

) 5.3

5.2

2.7

4.4

1.5

1 0

1 .66

.6

01

0

.33

.6

6a.

Res

pond

to

cont

inge

nt

quer

ies

(R)

1/1

3/3

2/3

112

10/1

3 3/

3 0/

0 9/

10

414

Dec

imal

for

m:

1 1

.67

39

.1

9 .5

.7

7 1

.76

.25

0 .9

1

.63

.55

LAN

GLE

AR

N/3

7:4

87

Ob3

Tab

le 6

. P1

77 M

B

r2

Tabl

e 6

Chi

-squ

are

and

Kru

skal

-Wal

lis T

ests

on

Gro

up M

eans

for

Lan

guag

e Le

arni

ng A

cts

- H

igh/

Med

./Low

H

igh/

Med

. H

igh/

Low

M

ed./L

ow

df=

2

df=

l d

f=l

- d

f=l

lb.

Ost

ensi

ve d

efin

ition

s (A

) 1.

01

(e)

(el

(el

2b.

Rep

eat a

cor

rect

ion

(A)

.62

(K-W

) (e

l (e

l (e

)

3b.

Sinc

ere

requ

est

for

info

rmat

ion

(A)

1.64

(K-W

) (e

l (e

) (e

l

4a.

Res

pond

to in

sinc

ere

requ

est

for

info

rmat

ion

(R)

I. 20

##X

(el

(el

(4

5b.

Con

tinge

nt q

uerie

s (A

) 6.

00#

(el

(e)

(e)

6a.

Res

pond

to

cont

inge

nt q

uerie

s (R

) .6

5 (K

-W)

(el

(el

(el

KEY

: (R

) =

raw

dat

a (A

) =

adj

uste

d da

ta

(e)

= e

xpec

ted

cell

mea

ns o

r sa

mpl

e siz

e to

o sm

all.

(K-W

) =

Kru

skal

-Wal

lis te

st

# p

< .0

5; #

* p <

.01;

#*#

p <

.001

for

Kru

skal

-Wal

lis te

sts.

Peck 565

Jql: Girl. Boy. Mother. Ed: (turns to a picture of the nextdoor neighbor and his dog) Jql: Man, boy. Perro.

Vey (who produced 17 ostensive definitions) used the same strategy:

T8V: 7 Ed: (. . .) From now on, everybody listen to Andrew, now YOU say the story.

Veh: . . . (naming pictures:) /stIp/. Bear. . . . This is the: the frie:nd, this, boy? and the father, the dog, and the mother.

2. The learner repeats a correction of truth or expression. The group means were all about the same. The range for individuals was 0-4 repeti- tions of a correction. The nine children mainly repeated Ed’s corrections during the story task and the picture tasks. Examples of errors are “two apples” instead of “cherries,” “brother” instead of “mother,” and “cat” instead of “cherry.” Three children, each from a different group, were the highest ranked on repeating a correction.

3. The learner makes a sincere request for information. The high group had the highest mean number of sincere requests for information (4.0, adjusted data) and the means for the other two groups were much lower (0.6, 0.33, adjusted data). A three-way Kruskal-Wallis test shows that these differences across groups are not statistically significant.

Two high children, Gah and Alh, made 6 and 5 requests for informa- tion, respectively. They asked Ed to supply names of objects and charac- ters. Alh asked a number of other questions, perhaps to distract from the tasks. He inquired about noise outside the building, the name and func- tion of the tape recorder, and about what Ed was doing. The number of questions that Alh asked, as well as his assertiveness and curiosity make him stand out from the other children. Except for Alh and Gah, every other child made only one request for information.

4. The learner responds to Ed‘s insincere requests for information. Ed mainly used insincere requests to drill colors, shapes, and names of people in the story. As a group, the high children had the highest rate of response to Ed’s requests, followed by the medium group and then the low group. The three means were significantly different. As for individ- ual variation, Ed addressed fewer requests to Kal (the child who did not

566 Language Learning Vol. 37, No. 4

speak at all) than to any other child. Alh, the high child who was more talkative and more assertive than any other child, received the next low- est number of insincere requests for information.

The low children received fewer insincere requests for information per child than the medium children did, and they were also less successful in responding to the requests. For example, Kal and Jel responded to none of Ed‘s insincere requests for information in their sessions at the begin- ning of the fish game:

T3KA: 1 Ed: Know what this is? (42s) Do you got two oranges?

Kal: (shows two of her cards, both circles)

TlOJE: 1 Ed: Do you know your colors? (8s) Roja? (6s) (Showing cards:) Circle? . . . Do

Jel: (Stares straight ahead and looks frightened) you? (5s) You do?

In contrast, Jsm from the medium group responded to Ed’s insincere requests 13/17 times. In the following example, Jsm responded to both of Ed‘s requests:

Tl5JS: 7 Ed: (. . .) What is this?

Ed: He said, there’s a bear in my room. Okay. (5s) What is he saying? Uh, what Jsm: The bear.

was he saying on the next page? Jsm: Umm, where’s my brother? Ed: Good! What is he saying to his father?

5. Contingent queries. The rank for the mean number of contingent que- ries is High > Medium > Low. The three means are significantly differ- ent. Two high children (Ah and Gah) each produced 5 contingent queries while the medium and low children each produced 0 or 1 contingent query. Alh’s contingent queries were in all three tasks. He used them to get Ed to repeat or clarify something in the previous turn. For example:

TZAL: 16 Alh: (. . .) You have two:, two, apple?

Alh: Cherries?

Alh: //Yuh?//

Ed: That’s cherries.

Ed: Uh //huh// Uh huh. (. . .)

Peck 567

6. Respond to contingent queries. Nonverbal responses were not counted. The high group verbally responded to Ed’s contingent queries the highest proportion of the time, followed by the medium group and then the low group. Note that the high group actually received fewer contingent queries (1, 3, 3 , raw) while the other two groups received more (medium: 2, 13, 3, raw; low: 0, 10, 4, raw). Two children, Alh and Kal (both of whom, as noted above, received the fewest insincere requests for information), also received the fewest contingent queries.

In summary, the high group produced the most language-learning acts. For five features, the mean for the high children is the highest, followed by the means for the medium and low groups. The high children appear more able to ask and answer questions, and to repeat a correction than the low children. In contrast, the low children rank first on ostensive definitions. Note, however, that most of the expected cell means are too small for a chi-square or Kruskal-Wallis test.

DISCUSSION

This analysis has revealed some ways in which the three groups of children are different. Some of the features were more useful than others in differentiating learning acts and language-learning acts by the three groups.

GROUPS

The high and low groupings corresponded to the children’s ranking in the learning acts. The high children appeared to know more English than the low children. They produced more words, more unique words, and more words per minute. However, the high children as a group did not appear very different from the medium children. The teacher ratings for the high and low groups did result in two distinct groups: one talkative, and one much less talkative, and seemingly less competent in English.

The high group performed the most learning acts, and the other groups each ranked first on one learning act. The high group’s use of self-repetition and fillers suggests that these are techniques or skills that fluent children are more likely to exercise than less fluent children. Like the adolescents studied by Scarcella and Higa (1982), the high group produced more conversation-maintaining acts than did the other groups. The high group also produced more language-learning acts (adapted

568 Language Learning Vol. 37, No. 4

from Thew 1979) than did the low group. The medium children repeated Ed’s utterances more than did the other children. For instance, they repeated lines of the story at points where the high children might have constructed an original utterance, or the low children might have used ostensive definitions: “Man. Mother. Boy.” More ostensive definitions were used by the low group than by any other group. Perhaps their high rank on this feature was by default. The low children may not have had enough vocabulary or fluency to use other techniques such as self- repetition or fillers. It appears that the high and medium children, who used all the other language-learning acts, did not need to use ostensive definitions in the tutoring session.

THE NATIVE SPEAKER

There are signs that Ed, the native speaker, adjusted his questioning to the level of the learner. He addressed contingent queries and insincere requests for information the least frequently to the high group. Perhaps this happened because Ed did not need clarification or confirmation from the high children, and did not need to quiz them on color terms or other English vocabulary. Ed asked these questions the most frequently with the medium children. Perhaps their speech was at a suitable level for contingent queries: less clear, less correct and less developed than the high children’s speech. Perhaps the medium children were also at a suit- able level for insincere requests for information: they were less familiar with the English terms for colors, numbers and animals than were the high children.

CONCLUSION

The first, and most obvious conclusion, is that the kindergarten chil- dren differed from each other. Their output varied, and the frequency of their conversation-maintaining acts and language-learning acts varied as well. This finding is not surprising, since the original goal of the study was to analyze transcripts from a range of children.

Classroom competence, as judged globally by the teacher ratings, cor- responded to the children’s communicative and linguistic competence as measured in this study. The high children appeared more proficient in English than the low children. If one considers only the high and low groups, then proficiency in English seems to go hand-in-hand with fre-

Peck 569

quency of learning acts. In every category except one, the high group produced more learning acts than the low group. It is not necessarily predictable that high-achievers would be more active than low-achievers in conversations with another child. The results show that high- achievers, as globally rated by the teachers, produced more signs of learning in the child-child tutoring sessions than did low-achievers. The discrete, quantifiable measures reported here appear to corroborate the teachers’ holistic ratings.

Thus, in many cases the features used here in analyzing the transcripts divided the high group from the low group. This division was usually neat, and often marked a significant difference between the counts for the two groups on a feature. However, one feature, “learner nominates a new topic,” did not help to discriminate among groups, because only one topic was nominated in the whole corpus. The medium group, in general, was lower than the high group on many of the output measures and learning acts. The difference between the high and medium groups was usually not significant.

Reification is one limitation of this study. The features in this study were used to define an action or skill (maintaining the conversation) and mental activity (language learning). In this study, the process of “learn- ing” was investigated by means of counting linguistic products. We can- not be sure, however, that a string of words labeled “repeat native speaker utterance” actually represents a “learning act,” a type of “main- taining the conversation.” The limitation here is that the products are not the process, and we cannot surely say that one child is learning more than another or better than another. We can count conversation-maintaining and language-learning acts, yet only speculate about what and how the children are learning language. This limitation is not unusual in social science research, but deserves mention.

The study is also limited by the small amount of data for each pair. In addition, for each pair the session was a first meeting, so that the shyer kindergarten students probably spoke less with the native speaker than they might have if they had worked with him before. The data here, in other words, might not be representative of future tutoring sessions for each pair, if more sessions were arranged. So it seems that tutoring sessions analyzed here are like testing sessions, too.

Another limitation of the study is that the learning acts that were counted and analyzed are all verbal. Nonverbal interaction, such as eye contact and smiles, was not taken into account. Thus, the consistent

5 70 Language Learning Vol. 37, No. 4

lowest ranking counts for the low group on the learning acts correspond with their small output. It is interesting to note that Kal, who never spoke, interacted a great deal with Ed (especially during the card game) with flirtatious looks and a sudden, staccato grab at each card as Ed named it. In fact, an analysis of Ed’s talk with Kal reveals more adjust- ments to her level, as well as more efforts to get her attention, and to maintain the conversation, than in Ed’s talk with Jel or Jql. If interaction between a beginning level NNS and an NS is at issue, nonverbal acts need to be examined more than they were here. Future projects could avoid the last two limitations by gathering data in school sites that already have child-child tutoring programs, and by using videotape as well as audiotape.

This study sheds light on the range of the nonnatives’ output and the ways in which they may be learning language through the tutoring ses- sion. The high group spoke more than the low group, and in the process, produced more learning and language-learning acts. The high group was best at maintaining the conversation: the high children used more fillers, more self-repetition, and asked more questions than the other groups. For the medium group, a typical way to maintain the conversation was to repeat the native speaker’s utterances. The low group, who spoke little, relied on ostensive definitions. Thus, the results suggest that ostensive definitions are easier to produce than other-repetition, which is easier than asking the NS a question. Such information will be useful in the design of interactive research on NNS child-NS child tutoring and learn- ing. It can also contribute, in the long run, to the development of curric- ulum materials and teachers’ guides for groupwork or peer work by child nonnative speakers and other children, native or nonnative.

REFERENCES

Berman, R . (Undated). Shelli once again: the re-emergence of a bilingual. Unpublished paper, Tel Aviv Unversity.

Chesterfield, K. (1980). The child as a teacher: an investigation into child native speaker- adult learner discourse. Unpublished M.A. thesis, UCLA.

Fillmore, L. W. (1975). The second time around: Cognitive and social structures in second language acquisition. Unpublished. Ph.D. dissertation, Stanford University.

Fillmore, L. W. (1979). Individual differences in second language acquisition. In C. J . Fillmore, D. Kempler and W. Wang (Eds.), Individual differences in linguistic ability and linguistic behavior. (pp. 203-228) New York: Academic Press.

Guilfoile, E. (1957). Nobody lisrens to Andrew. Chicago: Follett Publishing Co.

Peck 5 71

Gilmore, P. (1977). A children’spidgin: The case of aspontaneouspidgin for two. Unpub- lished paper, University of Pennsylvania.

McClure, E., Saville-Troike, M. & Fritz M. (1982). Children’s communicative tactics across language boundaries. Paper presented at the Chicago Linguistics Society meeting.

Nemoianu, A. M. (1980). The boat’s gonna leave: A study of children learning a second language from conversations with other children. Amsterdam: John Benjamins B.V.

Peck, S. (1978). Child-child discourse in second language acquisition. In E. Hatch, (Ed.), Second language acquisition: A book of readings. (pp. 383-400) Rowley, Massachusetts: Newbury House.

Peck, S. (1985). Signs of teaching and learning: A descriptive study of children’s language in native speaker-nonnative speaker tutoring sessions. Unpublished Ph.D. dissertation, UCLA.

Sachs, J., & Devin J. (1975). Young children’s use of age-appropriate speech styles in social interaction and role-playing. Journal of Child Language, 3, 81-98.

Saville-Troike, M., McClure E. & Fritz M. (1982). Communicative tactics in children’s second language acquisition. Paper presented in Milwaukee at the University of Wiscon- sin Symposium on Universals of Second Language Acquisition.

Scarcella, R., & Higa C. (1982). Input and age differences in second language acquisition. In S. Krashen, R. Scarcella, & M. Long (Eds.), Child-adult differences in second lan- guage acquisition. (pp. 175-201). Rowley, Massachusetts: Newbury House.

Shatz, M., & Gelman R. (1973). The development of communication skills: Modifications in the speech of young children as a function of the listener. Monographs of the Society for Research in Child Development. Serial no. 152, 38, 5.

Thew, C. (1979). The role of mothers and firstborn female siblings in teaching and encour- aging language skills. Unpublished. Ph.D. dissertation, University of British Columbia.