6.2 ANALYSIS OF LANGUAGE ASSESSMENT DATA . 43
6.2.7 SecondStage Analysis of'Primary and Secondary Oral Assessment
6.2.7.7 Findings . ;
(A) Secondary Schools
(i) 1ST Administration - Inter-teaching Mode Group Comparison
The sample size for NET and NNET is 33 and 39, respectively. Significant difference was indicated for three variables on the first administration: Max Length of Turn (NET > Local), % of Clauses that are Statements (Local > NET) and % of Clauses that are Questions (NET>Local) using non-parametric testing - Mann-Whitney Test These are all indicators for productivity.
However, teaching mode effect is not consistent.
(ii) 2nd Administration - Inter-teaching Mode Group Comparison
Significant difference was found on both Parametric (ANOVA) and Non-Parametric testing (Kruskal Wallis Test) for at least one of the teacher groups for the variable % Grammatical Accuracy of Attempted Statements - Gp 3 > Gp 4 according to one-way ANOVA and Kraskall Wallis; and Gp2 > Gp 4 according to Kruskall Wallis, Here there is some indication that those students exposed to both NET and Local teachers had a higfter percentage of grammatical accuracy for attempted statements than those exposed to Local teachers alone.
(iii) 2nd Administration - Inter-teaching Mode Group Comparison with 1st
administration values controlled
In ANCQVA, significant difference is found between groups on the variable % Grammatical Accuracy of Statements (ga_as2), using LSD methodin paired comparison (since only 3 groups exist)* Students taught by teachers with a transition from NET to Local elicited more correct clauses than those with a transition from NET to NET.
Results of (i) to (iii) are summarised in the following table.
Secondary; 1st & 2nd admin.; Parametric and Non-parametric tests
T 7 * 'L.I
Variable
Number of Utterance Mean length of
Utterance Median Max Length of
Utterance Max Length of
Turn No. of Attempted
Clauses
% of Utterances that are clauses
Number of Attempted Statements
% of Clauses that are Statements
%Grammatical Accuracy of
Statements Number of Attempted Questions
% of Clauses that are Questions
<3™wmtical Accuracy of Questions
1** Administration - Comparison between NET and Local Independent
T-Test
Mann-Whitney Sample: NET=33; NNET=39
tevene
0.06
0.69
0.83
0.73
0.60
0.05
0.08
0.03*
0.16
027
0.40
0.16
0.83
p-value
0.69
0.58
0.73
0.13 ,
0.10
0.89
0.81
0.67
0.08
0.97
0.10
0.08
025
p-value
0.45
0.52
0.48
0.08
<0.01*
(NET>Local) 0.76
0.83
0.9 0.01*
(LocaONET) 0.83
0.12 0.01*
(NET>Local) 0.10
2nd Administration - Comparison between 4 Groups
ANOVA
Kruskal
Wallis ANCOVA
Sample: 1=22; 2=7; 3=20; 4=17
levene
0.36
0.015*
<0.01*
0.195
<0.01*
0.247
*0.049
0.262
0315
0.825
0.899
0315
0.849
p-value
0.069
0.657
0.447
0.639
0.037*
0.664
0.587
0.702
0.37 0.003*
2>4 3>4 0.744
0.37
0.832
p-value
0.40
0.24
0.88
0.54
0.70
0.61
0.558
0.73
0.42 0.01*
2>3>1>4 N/A
0.42
0.54
Levene
0.45
0.15
0.15
0.07
0.02*
0.93
0.06
0.85
0.67 0.91
0.84
0.55
0.63
p-valu e
0.70
0.44
0.03*
0.88
0.33
0.24
0.77
0.20
0.84 0.046*
3>4 0.36
0.97
0.71
ANCOVA
2-N->L; 3-L->N;
a=0,05/6 (6 paired comparison from 4 groups)
92
(iv) Variables with Significant correlation between the 2 administrations
Correlation of the thirteen variables between 1st and 2nd Administration shows the following 7 variables to be significantly correlated in some teaching mode class groups,
1. Number of Utterance 2. Mean Length of Utterance 3. Median
4. Max Length of Utterance 6. Number of Attempted Clauses 8. Number of Attempted Statements 11. Number of Attempted Questions
These results are summarised in the table below.
Secondary; 2nd admin.; correlation tests
Variable
1. Number of Utterance 2. Mean length of Utterance 3. Median length of Utterance 4. Max Length of Utterance 5. Max Length of Turn
6. Number of Attempted Clauses 7. % of Utterances that are clauses 8. Number of Attempted
Statements
9. % of Clauses that are Statements 10. %Grammatical Accuracy of
Statements
11. Number of Attempted Questions 12. % of Clauses that are Questions 13. % Grammatical Accuracy of
Questions
Group 1 Sample = 22 Coef.
0.55 0.61 0.07 0.58 0.15 0.59 024 0.61 0.22 -0.07 0.16 0.22 0.32
P-val.
<0.01*
<0.01*
0.77
"<0.01*
0.50 O.01*
0.27 O.01*
0.31 0.77 0.46 0.31 0.14
Group 2 Sample = 7 Coef.
0.62 0.46 0.61 0.56 -0.18 0.67 0.51 0.61 0.05 -0.48
0.82 0.05 0.68
P-val.
0.14 0.29 0.14 0.19 0.69 0.10 0.24 0.15 0.91 0.27 0.02*
0.91 0.09
Group 3 Sample =20 Coef.
0.55 0.72 0.20 0.47 0.01 0.67 0.43 0.58 0.33 0.03 0.61 0.33 0.17
P-val.
0.01*
<0.01*
0.39 0.03 0.96
<0.01*
0.06
<0.01*
0.16 0.90 0.01*
0.16 0.47
Group 4 Sample = 17 Coef.
0.46 0.78 -0.01 0.52 -0.15
0.25 0.13 024 0.14 0.36 024 0.14 -0.08
P-val.
0.06
<0.01*
0.96 0.03*
0.56 0.32 0.62 0.35 0.59 0.15 0.36 0.59 0.75
Notably the variable Mean Length of Utterance had significant correlation among three out of the four groups, (Group 1,3, & 4) meaning that students' performance was most stable as measured by this variable between the first and second administration. Those who scored highly in the first administration also did so in the second administration. Their performance did not fluctuate wildly ftom the first to the second administration. Similarly, with the lower scores, the same positive correlation was shown for these seven variables. Except for one
variable #11, which has good correlation within one teaching mode class group, all other variables have good correlation in two teaching mode class groups. There were no variables that had any significant correlation within Group 2 which was also too small in sample size for the results to be reckoned with.
(v) Variables with significant difference between the 2 administrations - per each of 4 groups
Paired Sample t-test results show significant differences between the 1st and 2nd administration in the following 7 variables among the four groups of teaching class modes:
2. Mean Length of Utterance 4. Max Length of Utterance
8. Number of Attempted Statements 9. % of Clauses that are Statements 10. % Grammatical Accuracy of Statements 11. Number of Attempted Questions
12. % Grammatical Accuracy of Questions
The table on the next page summarises the results.
94
Secondary; 2nd admin.; Paired sample t-tests
Variable 1. Number of Utterance 2. Mean length of Utterance 3. Median
4. Max Length of Utterance 5. Max Length of Turn
6. Number of Attempted Clauses 7. % of Utterances that are clauses 8. Number of Attempted Statements 9. % of Clauses that are Statements
10.%Grammatical Accuracy of Statements
1 1 . Number of Attempted Questions 12. % of Clauses that are Questions 13. % Grammatical Accuracy of Questions
Group 1 Sample size= 22 Mean
-0.45 1.26 12.0 1.50 -20.9
4.27
62.0 9.77 83.5 30.4-1.41
15.5 3.79
Std.er
4.51 0.59 3.61 1.66 4.95
•2.29 5.10 2.06 3.24 3.10
0.95
3.24 0.94p-val.
0.92
0.047*<0.01*
0.37
<0.01*
0.07
<0.01*
<0.01*
<0.01*
<0.01*
0.15
<0.01*
<0.01*
Group 2 Sample size = 7
Mean -2.57
1.49 6.71, ,2.57 -17.7 -2.00
68.4 -5.28 91.8 40.6 -4.287.10
3.42
Std.er
6.84 0.57 3.39 2.86 13.3 5.38
3.47 4.60 3.65 5.68 1.523.65 2.29
p-val.
0.72
0.04*0.09 0.40 0.23 0.72
<0.01*
0.29
<0.01*
<0.01*
<0.01*
0.10 0.18
Group 3 Sample size = 20 Mean
-10.0 1.46 6.15
2.2517.6 -3.30
61.9 1.00 88.9 36.7 -1.35 10.0 4.12Std.er
5.86 0.61 1.57
2.7714.21
2.74
5.59 2.81 2.43 2.720.84
2.43 1.22p-val.
0.10
0.03*
<0.01*
0.43 0.23 0.24
<0.01*
0.73
<0.01*
<0.01*
0.13
<0.01*
<0.01*
Group 4 Sample size = 17 Mean
2.12 1.14 8.20 3.94 28.0 -0.88 55.4 4.35
85.1 21.1 -2.0513.8 2.61
Std.er
4.23 0.43 1.68 1.72
18.4 3.94 4.84 3.52 3.36 3.65 1.10 3.36 1.24p-val.
0.62 0.01*
<0.01*
0.04
0.14 0.82<0.01*
0.23
<0.01*
<0.01*
0.08
<0.01*
0.052
There is gain in the top four listed variables and loss in the lower 3 listed. All the variables showing significant gain are from productivity variables: Mean Length of Utterance, Max Length of Utterance, Number of Attempted Statements and Percentage of Clause that are Statements, The variable Mean Length of Utterance is the strongest indicator of gain with this indicated for three out of four groups (NET •> NET, Local ->NET, and Local -> Local). This also points to an indication that students* output has increased over time* However, this phenomenon is true of all groups of students - those with NET, those without NET exposure as well as those with both types of exposure.
With the exception of the number of questions for Group 2, all the other variables showing significant loss relate to grammatical accuracy of statements or questions. This may be related to the fact that grammatical accuracy decreases as productivity increases since students tend to speak more in time but may not have developed a parallel control for accuracy needed for language with a higher productivity. The loss shown in the number of attempted questions may also be related to the fact that asking questions is a comparatively more demanding task than making statements and students have shown improvement via more statements rather than in more questions.
Summary
In Summary, there are indicators of gains and losses among various different variables and among groups of different teaching mode between the first and second administration, as well as differences in baseline performance among groups within each administration. However, the results are rather scattered and show no clear cut significant difference that can be linked to all variables of productivity or grammatical accuracy or to a particular mode of teaching, whether NET only, Local only or both.
This lack of clear evidence of the comparative benefits of the NET exposure may actually reflect the fact that there is a host of multiple factors influencing language learning in ESL in the Hong Kong context The mere exposure to a NET may not be a strong or sufficiently isolated independent variable to effect changes in language gain or loss when there are multiple changing factors related to at least three main categories of factors: NET characteristics in personality and teaching style, the curriculum with various degrees of flexibility and emphasis on oral language development as well as student characteristics in interest, motivation, degree of participation with the NET, previous language ability as well features of home and school emraronments that may or may not be conducive to ESL learning.
96
(B) Primary Schools
(i) Which variable is significantly correlated between the 2 administrations?
Correlation Analysis of Variables Primary; Td admin.; Correlation tests
Variable
Number of Utterances Mean length of Utterance Median
Max Length of Utterance Max Length of Turn
Number of Attempted Clauses
% of Utterances that axe clauses Number of Attempted Statements
% of Clauses that are Statements
% Grammatical Statements
Group 1
'•
Group 2
^
Group 3
-Group 4
v'
The correlation of the ten variables between 1st and 2nd Administration shows the following two variables to be significantly correlated by teaching mode groups: Mean Length of Utterance for Group 2 and % of Utterances that are Clauses for Group 4. These indicate that student performance is fairly stable and does not fluctuate wildly from the first to the second administration for these two variables.
(ii) Which variable shows a significant Difference between the 2 administrations - per each of 4 groups?
Paired Sample t-test
Variable
Number of Utterances Mean length of Utterance Median
Max Length of Utterance Max Length of Turn
Number of Attempted Clauses
% of Utterances that are clauses Number of Attempted Statements
% of Clauses that are Statements
% Grammatical Statements
Group 1
•five Gain
-five Gain
Group 2 -iveLoss
•five Gain -iveLoss -five Gain
•five Gain
-fiveGain -five Gain
Group 3 Group 4
-five Gain
-five Gain
-rtveGain
Paired S ample t-test results for the ten variables b etween 1st and 2nd Administration s how significant gains or losses made among some teaching mode groups in the following variables.
Five variables show gain in Group 2 (NET->Local):
3. Median
4. Max Length of Utterance
7, % of Utterances that are Clauses 9, % of Clauses that are Statements 10. % Grammatical Accuracy of Statements Two variables show a loss in Group 2:
1, Number of Utterance 5. Max Length of Turn
The results are quite scattered in that gain is shown for botiuproductivity and grammatical accuracy variables but not for all of the concerned variables in each category. Loss is registered pnly in Group 2 where students sre exposed to both NET and Local teachers. The strongest indicators would be the two variables - % of Utterances that are Clauses and % Grammatical Accuracy of Attempted Statements that show gain in all three groups (Group 1,2, & 4).
(iii) 1ST Administration - Inter-teaching Mode Group Comparison
No significant difference was found between NET and Local groups in all 10 variables in parametric or non-parametric testing.
(iv) 2Bd Administration - Inter-teaching Mode Group Comparison
Significant differences via non-parametric testing (Kruskal Wallis) are indicated for two variables namely:
There is some evidence that students with exposure to only local teachers perform better in these two variables than students that are exposed to the NET only or to both in the Number of Utterances. Also in the variable % of Clauses that are Statements, there is evidence that students exposed to locals only perform better than students exposed to both. However, the NET only group also showed evidence that their students also performed better on this variable than the group exposed to both.
Summary
In Summary, findings from the primary results are similar to those for the secondary* There are indicators of gains and losses among various different variables and among groups of different teaching mode between the first and second administration, as well as differences in baseline performance among groups within the second administration for isolated variables. However, the r^ulte arc rather scattered and show no clean cut significant differenced^
all variables of productivity or grammatical accuracy or to a particular mode of teaching, whether NET only, Local only or both.