Findings

(i) First cohort second administration

Exploratory analysis by means of examination of the plots did not provide any evidence to suggest that there was any difference in teaching mode effect in the first cohort, second administration for the secondary language assessment data. That is there were no apparent differences in oral, listening or writing scores gathered from secondary students taught by NETs, Locals or Both.

Analytical investigation revealed that teaching mode effect was not interacting with time effects for either oral scores, listening scores, coordinated oral and listening scores or writing scores. In other words, irrespective of school and form level, in the first cohort secondary longitudinal analysis there was no evidence of a NET effect on language assessment scores.

The interaction effect of oral and listening within-subjects factors and school level, form level, and teaching mode between-subject factor was significant However, since the Eta squared value was small (0.028) and the teaching mode effect was not significant in lower level interactions, there is not sufficient evidence in these results to suggest that a significant NET effect exists in the sampled data.

(ii) Second cohort second administration

Exploratory analysis revealed a possible teaching mode effect in the form of an apparently marked difference between the oral scores produced by students taught by a NET and those taught by a local teacher. The plot is reproduced below.

One year teaching effect For a« students in 2nd cohort 2nd administration

26r

Longitudinal effect: November 1999 to May 2000

Analytical investigation revealed that the difference in oral scores was significant The teaching mode effect was interacting with the time effect for oral scores. Analytical analysis also revealed that the teaching mode was interacting with time for coordinated oral and listening scores, but not for either writing scores or listening scores.

In other words in the second cohort, teaching mode effect was not interacting with the time effect for either listening scores or writing scores. However teaching mode effect was interacting with the time effect for oral scores and coordinated oral and listening scores. The interaction of teaching mode and time was significant (for oral p= O.001 Eta Squared=0.054;

for oral & listening p = 0.007 Eta Squared=0.020) but the magnitude of the differences was small. This provides evidence that exposure to a NET was influential in increasing the oral assessment and coordinated oral and listening scores of secondary school students at all levels of school and all ability levels, and that this influence was significantly greater than that of local teachers teaching similar students.

There was a significant mtetaetion effect of oral within-subject factor, form level, school level and teaching mode (p-value < 0-001), however the Eta Squared value was less than that for the Intearaction effect of the oral within-subjects factor and teaching mode, therefore no further analysis is needed. On the other hand, the interaction effect of the listening within-subjects factor, school level and teaching mode was significant (p=0.007 Eta Squared=0.028). But the interaction of teaching mode and listening within-subjects factor was not significant Therefore graphical analysis is needed to further investigate &efindm

(B) Primary

(i) First cohort second administration

Examination of the first cohort plots revealed one that contained possible evidence of a teaching mode effect The plot for the first cohort, second administration writing scores seemed to suggest a sizeable difference between scores obtained fiom students in the second administration of the writing tests.

66

Two year teaching effect

For all pupils in 1st cohort 2nd administration

1 -40'

g> .38 1 .36

1 -34 1 -32 I .30 1 •V) O£M28'

UJ .4<LO

s

/

s*

+

y

-.40

•.38 ..36 -.34 .32 -.30 .28

Teaching mode

mt ^m

•NET to NET

— "

* Local to Local

Longitudinal effect March 1999 to May 2000

Analytical investigation, however, revealed that the difference in writing scores was not significant.

On the other hand, in the analytical investigation significant differences were revealed in the listening scores. This provided evidence to suggest that there was a possible teaching mode effect on the listening scores of pupils who were taught by a NET in their first year and a local teacher in their second year, when compared to listening scores produced by pupils taught by local teachers. The difference was statistically significant (p = 0.018 Eta Squared = 0.027), however the magnitude of the difference was very small.

In other words, in the first cohort, the teaching mode effect was not interacting with the time effect for either oral scores or writing scores. However the teaching mode effect was interacting with the time effect for listening scores suggesting that pupils who were taught by a sequence of NET to Local over two years scored significantly higher in the listening assessment than pupils taught by a NET for two years and than pupils taught by a local teacher for two years. In interpreting this finding it needs to be borne in mind that the reliability of the primary listening test on the second administration was marginally unacceptable (please refer to Section 6.1.3.1 C above). For this reason the finding needs to be treated with caution.

A similar result was found for coordinated oral and listening scores (= < 0.001 Eta Squared = 0.072) however the magnitude of the difference was small.

The interaction effects for the oral within-subjects factor, school level and teaching mode, and the listening within-subjects factors, school level and teaching mode, were significant, but with Eta Squared values of less than that of the interaction effect of the listening within-subject factor and teaching mode. Therefore, farther analysis was not comi^^^

The interaction effect for both the oral within-subject fector, form level and teaching mode was significant (p=< 0.001 Eta Squared=0.122). Therefore, graphical analysis is needed to farther investigate fee findings.

(ii) Second cohort second administration

Exploratory analysis revealed a possible interaction of teaching mode and time for the writing scores of pupils in the second cohort. The plot is reproduced below.

One year teaching effect

For afl pupils in 2nd cohort 2nd administration

0.0 "BOTH

Longitudinal effect: November 1999 to May 2000

Analytical investigation of all the language assessment scores, however, revealed that there was no interaction of teaching mode and time for the writing scores, the listening or the coordinated oral and listening scores. However there was a significant finding for the oral scores.

In the second cohort, the teaching mode effect was interacting with the time effect for oral scores. The effect was significant (p - 0.003 Eta Squared = 0.063), but the magnitude was very small. This finding suggests that the influence of the NET was significant in producing higher oral assessment scores than those of local teachers or a combination of local teachers and NETs.

In other words, pupils taught by a NET scored significantly higher in the oral assessments than pupils taught by other teachers*