1 The Kom Experimental Mother Tongue Education Pilot Project Report for 2012 Stephen L. Walter and Kain Godfrey Chuo Executive Summary For five years, an experimental program in multilingual education has been functioning in Boyo Division in the North West Region of Cameroon. In this program, 12 experimental schools use the local language—Kom—as a medium of instruction in Classes 1-3. After Class 3, children in the experimental schools return to the standard practice of English-only instruction. The 12 experimental schools are matched with 12 comparison schools which continue to use English as the language of instruction. At the end of year 5, standardized testing was once again completed for children in both comparison and experimental schools in Class 5. The findings are presented in the various sections of this report. The first part of the report summarizes the findings from the first four years of the experimental project to provide the background needed to interpret the results seen in Class 5. Then the report takes a closer look at the results in Class 3—the final year of the experimental intervention, at Class 4, the first year after the end of the experimental intervention, and finally at Class 5 which is the focus of the assessment work done this year. The Class 5 assessment reveals the following: 1. The performance of students who had been in the experimental program has dropped quite dramatically though it is still statistically better than that of students coming from comparison schools. 2. Performance on all subjects tested was low with math being especially low—just 16- 17 percent. 3. English and reading comprehension are the two areas where the KEPP students show the greatest advantage over the comparison students—a relative gain or advantage of a little over 30 percent. Performance in math was almost the same for both groups. Taken as a whole, the Class 5 assessment suggests the following: 1. The three year period of the intervention is not long enough to adequately prepare students for an effective transition to L2 instruction. (Note: Those who have been in English-only schools for all 5 years are even less prepared for the demands of Class 5 than are the children from experimental schools.) 2. The students coming from the experimental schools still show in Class 5 some of the educational benefits derived from having been in the experimental program. 3. The teaching of math in the area is woefully inadequate especially when done in English. 4. The overall level of proficiency in English reading proficiency and comprehension is probably at about Class 2 to 2.5 compared to a native-speaking population. The final section of the report presents a series of suggestions for improving the quality of educational delivery in Boyo Division based on the evidence of the research done over the 5 year history of the project. Pages 2-12 give detail on the history of the project and findings for previous years. Results for Class 5 for this year may be found beginning on page 13.
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
The Kom Experimental Mother Tongue Education Pilot Project
Report for 2012
Stephen L. Walter and Kain Godfrey Chuo
Executive Summary
For five years, an experimental program in multilingual education has been functioning in
Boyo Division in the North West Region of Cameroon. In this program, 12 experimental
schools use the local language—Kom—as a medium of instruction in Classes 1-3. After
Class 3, children in the experimental schools return to the standard practice of English-only
instruction. The 12 experimental schools are matched with 12 comparison schools which
continue to use English as the language of instruction. At the end of year 5, standardized
testing was once again completed for children in both comparison and experimental schools
in Class 5. The findings are presented in the various sections of this report.
The first part of the report summarizes the findings from the first four years of the
experimental project to provide the background needed to interpret the results seen in Class 5.
Then the report takes a closer look at the results in Class 3—the final year of the
experimental intervention, at Class 4, the first year after the end of the experimental
intervention, and finally at Class 5 which is the focus of the assessment work done this year.
The Class 5 assessment reveals the following:
1. The performance of students who had been in the experimental program has dropped
quite dramatically though it is still statistically better than that of students coming
from comparison schools.
2. Performance on all subjects tested was low with math being especially low—just 16-
17 percent.
3. English and reading comprehension are the two areas where the KEPP students show
the greatest advantage over the comparison students—a relative gain or advantage of
a little over 30 percent. Performance in math was almost the same for both groups.
Taken as a whole, the Class 5 assessment suggests the following:
1. The three year period of the intervention is not long enough to adequately prepare
students for an effective transition to L2 instruction. (Note: Those who have been in
English-only schools for all 5 years are even less prepared for the demands of Class 5
than are the children from experimental schools.)
2. The students coming from the experimental schools still show in Class 5 some of the
educational benefits derived from having been in the experimental program.
3. The teaching of math in the area is woefully inadequate especially when done in
English.
4. The overall level of proficiency in English reading proficiency and comprehension is
probably at about Class 2 to 2.5 compared to a native-speaking population.
The final section of the report presents a series of suggestions for improving the quality of
educational delivery in Boyo Division based on the evidence of the research done over the 5
year history of the project.
Pages 2-12 give detail on the history of the project and findings for previous years. Results
for Class 5 for this year may be found beginning on page 13.
2
Introduction
The year 2012 marks the fifth year of the operation of the experiment in mother tongue based
multilingual education in the Boyo Division of North West Region, Cameroon. In the design,
12 experimental schools have been selected as sites for the implementation of a curriculum
using the mother tongue (Kom) as the primary language of instruction. Twelve (12)
matching schools were selected as control schools. These 12 schools implement the standard
model of instruction in which English is used as the medium of instruction for all subjects
and classes1. The schools participating in the experimental program were matched by type
(private or public), location (remote village, small town), and size (smaller or larger) with
schools in the standard program. There was a certain amount of selection bias in the
identification of schools to be included in the experimental program in that poorer performing
schools were somewhat more likely to be included to reduce or eliminate any concerns about
hand-picking better schools to create a more favorable outcome.
The intervention
The primary element of the intervention is that of language of instruction. We consider this
variable foundational in light of (a) the research findings of Thomas and Collier (1997, 2002),
and (b) the fact that all developed countries consider it a given that the language of
instruction in their own schools should be the primary language of the country whether
English, Spanish, German, French, Danish, Russian, or whatever. The issue of language of
instruction becomes controversial in such countries only when dealing with minority
populations—be they indigenous or the product of immigration.
The primary intervention was accompanied by several supporting elements worthy of
investigation in their own right, though we have lacked the resources (and local permission)
to carefully manipulate and measure the impact of these variables. Nevertheless, they are
reported here for the sake of completeness. (1) Teachers received three weeks of training in
how to teach reading and writing in their own language since they had neither previous
experience nor training in using their own language for literacy. This variable is probably
offset by the fact that teachers in the standard program receive regular in-services as well. (2)
The teachers and students were provided with textbooks. Textbooks exist as well for children
in the standard program but buying them is at the discretion of parents and many fail to do so.
(3) Since teaching literacy in Kom was an additional subject in the curriculum for those in the
experimental program, some adjustments had to be made in the weekly timetable to make
time for this additional subject. This was accomplished primarily by bundling some of the
‘more peripheral subjects’ into a Kom-language reader to be used both for reading practice
and to teach the content of the more peripheral subjects. (4) Because teachers were
implementing instructional methods new to them (and the community) at least two additional
supervisory visits were arranged to provide support and to do on-the-spot problem solving.
At this point, we are treating the intervention as being unitary in nature though we know it is
very possible that some of the sub-elements may have had a measurable impact on outcomes.
In fact, some education officials in the area have raised the question of whether having
textbooks in the English-medium schools might impact outcomes in those schools as well.
1 In this part of Cameroon, the term “class” is preferred to the term “grade” in reference to the yearly levels of
students in basic primary education.
3
Past findings
The results during the first four years of the Kom Education Pilot Project (KEPP) showed a
marked—even strong advantage in educational achievement for the children in the
experimental program. Measured improvements in all learning outcomes except oral English
were dramatic—as much as 600 percent in some cases with 100 percent gains being typical.
In the case of oral English, the children in the experimental program also outperformed the
children in the standard English-medium program but by much smaller amounts ranging from
5 to 35 percent depending on the class. Even this small advantage was contrary to prevailing
public expectation that the best way for children to learn and improve their English was to be
immersed in an all-English instructional environment.
Previous reports for the years 2008, 2009, 2010 and 2011 contain the details of the findings
for those years. This year—year 5 of the experimental project—means that the first cohort of
children have now had two years back in the standard model in which English is the sole
language of instruction. The consequences of this transition will be a major theme of this
report. We are aware of no research which predicts or documents with any kind of specificity
the likely educational consequences of this kind of transition in a developing country after
having spent three years in a mother tongue model. Thus, we expect the details of this year’s
testing to be of great interest especially among those committed to the MLE model and, more
generally, to the larger international education community.
The best prevailing model of what to expect in this circumstance is that of Thomas and
Collier (1997, 2002) which makes a number of relevant statements. First, their model states
quite categorically that so-called early-exit mother tongue instructional programs—of which
the KEPP program is an instance—will produce weaker educational outcomes than late-exit
programs, those providing instructional support in the mother tongue for up to six years.
Secondly, the model indicates that early-exit programs will still produce a certain amount of
enduring improvement in educational outcomes and is thus still an improvement over
submersion programs (such as the standard model in Cameroon) in which children are taught
in English or French from day one regardless of whether they speak either of these languages.
Thirdly, the model suggests that after children leave early-exit programs and move into a
standard language program, they will tend to stagnate and lose ground to mainstream children
(children being educated entirely in their first language).2
Since the model put forth by Thomas and Collier is based entirely on research done in the
US, they made no claims about its aptness for describing educational experience in
developing countries. It is for this reason that we believe readers will be very interested in
the findings of this report especially those pertaining to Class 4.
The analysis of Class 5 data will present summary results of the testing of learning outcomes
for the year and will also include some preliminary thinking and analysis as to the reasons for
these results. This reflection will include a review of results in the light of the theoretical
predictions coming from the best models on multilingual education. In this process, attention
will be given to distinctive features of basic education in the project area as these contrast
significantly from the educational practice upon which the Thomas and Collier model is
based.
2 It is important to note that this model was developed based on research and educational practice in the US. In
countries like Cameroon, few or relatively few children receive their entire education in their first language. In
Cameroon’s case, that is because few children speak standard English or standard French as their first language
so there is no large population of children being educated in their first language that can be referred to as “the
mainstream.”
4
Some notes on the educational assessments which were carried out
To assist first time readers, some background information is provided here on the nature of
the assessments and assessment process used in the project area. As indicated above, the
experimental project consists of 12 experimental schools in which the local language (Kom)
is used as the language of instruction in Classes 1-3 for all subjects but English (as a subject).
These 12 experimental schools have been paired with 12 control schools in which all
instruction is in English, the standard educational model of the area (and the province).
Each year, standard tests (developed by the research team) are administered to children in
both control and experimental schools in the language of instruction. This means,
obviously, that children in the control or comparison schools are tested with instruments
rendered in English and those in the experimental schools are tested with instruments
rendered in Kom. Testing of proficiency in oral English was done in English.
Content for all assessment instruments came directly from the curriculum materials used for
instructional purposes. In the case of math, the instruments were identical with the exception
of the language of presentation. In the case of reading and language arts, the instruments can
best be described as ‘highly comparable.’ That is, if a sample text for reading
comprehension was taken from Lesson 3 in the standard program, a sample text for reading
comprehension in Kom was also taken from Lesson 3 of the instructional textbook developed
for that program. If the English instrument contained 5 words to test word recognition, the
Kom-medium instrument also contained 5 words drawn from parallel lessons and of similar
length and perceived difficulty. The oral English assessment was identical for all students at
a given level.
The assessment designs for Classes 1 and 2 were identical apart from the adjustments made
in the language of the assessment because children were being taught in two different
languages. The assessment design for Class 3 included an innovation in that ALL children in
both programs took the same reading test in English. Those in the experimental program also
took an additional reading/language arts test in Kom since that is their language of instruction
for reading. This modification in the assessment design was motivated by a desire to better
measure the nature of the reading skills being developed in the experimental schools with
respect both to English AND to Kom. Consequently, no assessment of oral English was
given to Class 3 students.
The assessments for Classes 4 and 5 were quite straightforward; identical tests in English
given to all students coming from both experimental and control schools. No test of oral
English was given in Class 4 though the results for that year suggest there may be a need for
a more thorough assessment of proficiency in English especially at the basic level of
vocabulary development. Such a component was added to the Class 5 assessment this year
and will receive due attention in the report which follows.
Table 1 (next page) summarizes the assessment design for all classes in the experimental
program.
5
Table 1. Description of assessment design being used in the KEPP project.
Standard Program Experimental Program
Language of
instruction
Language of
testing
Language of
instruction
Language of
testing
Class 1
Language arts/reading
Math
English
English
English
English
English
English
English
Kom
Kom
English
Kom
Kom
English
Class 2
Language arts/reading
Math
English
English
English
English
English
English
English
Kom
Kom
English
Kom
Kom
English
Class 3
Language arts/reading
Math
English
English
English
English
Kom
Kom
Kom and English
Kom
Classes 4 and 5
Language arts/reading
Math
English
English
English
English
English
English
English
English
All testing was done by non-school personnel to increase the level of reliability of assessment
results.
Some comments on the presentation of results
The report for 2012 will first give summary results for Classes 2-4 for previous years.3 Then,
more detailed information is given for Classes 3 and 4 for previous years since performance
in Class 3 represents the high water mark of achievement for those in the experimental
program, and the results for Class 4 bring into focus the results of the shift from Kom-
medium to English-medium instruction.
Summary of test results from 2011
Table 2 summarizes the major findings for all three classes included in the 2011 assessment.
Table 2. Performance on standardized tests administered to both Standard and Experimental (KEPP)
schools in 2011.
Class 2 Class 3 Class 4
Test Component Standard KEPP Gain
4
(%) Standard KEPP
Gain
(%) Standard KEPP
Gain
(%)
Language Arts 22.0 61.0 177.3 24.2 41.1 69.8 28.1 38.7 37.7
Math 21.5 54.3 152.6 21.0 41.7 98.6 27.3 40.0 46.5
Oral English 52.9 61.3 15.0 * * * * *
Overall Test 34.4 59.2 72.1 23.0 41.3 79.6 27.6 39.5 43.1
Several preliminary interpretive notes are in order. First, apart from the very basic oral
English assessment, children in the standard program consistently score in the 20-30 percent
range. Since the assessment instruments used primarily a multiple choice format, this range
3 The interested reader may access previous reports from the primary author ([email protected]), from SIL
Cameroon in Yaounde, or from the Regional Delegate for Basic Education in Up Station, Bamenda, Cameroon.
4 Many of the tables included in this report will contain a column called ‘Gain (%).’ The numbers in this
column express the extent of the improvement or gain in efficiency for those in the KEPP program relative to
those in the standard program which, in this case, is treated as the ‘normal’ or default program.
6
can be considered equivalent to random guessing. While an examination of individual test
scores indicates that some children in the standard program are clearly learning, the group
mean scores persist in the range one associates with random guessing.
Second, we note what appears to be a significant drop in performance in the KEPP program
between Class 2 and the next two classes in Table 2. The most likely reason for this apparent
drop is the fact that the group mean scores reported in Table 2 for Classes 3 and 4 reflect
performance on an English-medium assessment instrument rather than a Kom-medium
instrument which is the language of instruction in Class 3. As will be demonstrated in a
subsequent section of the paper, when children in the experimental program in Class 3 were
tested using Kom-content reading materials, they scored in the 65 percent range.
Probably the most intriguing feature of Table 2 is the general similarity between mean scores
in Class 3 and Class 4 (when Class 3 results are based on English-language assessments).
Children in/from the KEPP program are consistently scoring around 40 percent. Children in
the standard program averaged 23 percent in Class 3 with an up-tick to 27.6 percent in Class
4.
Finally, note that no test of proficiency in oral English was given to children after Class 2.
The reason for this change was that the entire test for children in Classes 3 and 4 was in
English so a separate oral test was not deemed necessary. This decision may well be
reconsidered in future assessments. (In addition, a Kom reading test was given to children in
the experimental schools.)
Figure 1. Summary comparison of overall results between the two models in all four classes for 2011.
Interpretation of the results shown in Figure 1 must be done with caution. First, we note what
appears to be strong performance of the KEPP children in Classes 1 and 2. We believe this to
reflect what we will term “instructional adequacy.” Both children and teachers5 speak Kom
as a first language so there is a high level of effective communication in the classroom. The
classroom atmosphere is dynamic and electric as children compete to answer questions and
demonstrate their abilities. Conversely, the performance of the children in the English-
5 A few of the teachers are not actually native speakers of Kom but are highly fluent having lived in the area for
many years.
0
10
20
30
40
50
60
70
Class 1 (2010) Class 2 Class 3 Class 4
Class
Gro
up
Mean
Sc
ore
English KEP
7
medium schools is effectively lower than pictured having been elevated by scores on the oral
English assessment which was probably not rigorous enough. Children in these classrooms
are quiet, hesitant, and unwilling to participate actively. The fact that many of the teachers
speak English poorly further detracts from instructional adequacy.
In Classes 3 and 4 we observe a narrowing of the differential between the two groups. The
children in or from the experimental program lose—from a testing point of view—the
advantage of being tested in their language of instruction. Children in the English program
reflect performance levels similar to those observed in Classes 1 and 2 though obviously the
tests for Classes 3 and 4 are more difficult than the earlier tests. The significance of these
comparisons will be investigated more fully when we take a closer look at the results from
Classes 3 and 4.
Results for Class 3
Overall comparisons
The testing design employed for Class 3 allows us to address some interesting questions
about the relative effectiveness of the mother tongue component of the experimental
program. This will be explained below.
In the research design, Class 3 children in all 24 schools took the same test of knowledge and
skills in language arts and math. This test measured knowledge of English grammar, word
recognition skills, and reading comprehension. The test in math covered a range of topics
reflecting the curricular content of the Class 3 math textbook. Children in the experimental
KEPP program also took a reading test in Kom having a structure similar to the English
reading test. The texts used for reading comprehension (in both languages) used local names
and events to avoid potential threats to comprehension posed by using unfamiliar story
content.
A more detailed examination of performance by Class 3 children in 2011
In Table 3 we break down results of the testing by the sub-skills included in the assessment
both for language arts and for math.
Table 3. More detailed comparison of results by program for Class 3 (2011).
English KEPP Gain (%) Reading
in Kom Gain (%)
Language Arts 24.2 41.1 69.2 66.4 174
Grammar 27.4 42.1 53.6 65.8 140
Word recognition 21.0 41.4 97.1 * *
Reading comprehension 23.2 39.6 70.7 67.9 193
Math 21.0 41.7 96.7 * *
Number system 27.6 46.8 69.6 * *
Sets 14.1 23.8 68.8 * *
Simple calculations 18.8 45.2 140.4 * *
Story problems 14.8 35.7 141.2 * *
Overall Test Result 23.0 41.3 79.6 * *
The first assessment (in English) allows us to gain insight into performance when children
from both programs are placed on an equal footing with respect to language of assessment.
Any observed difference must, therefore, be largely attributable to effects stemming from the
experimental intervention which has motivated this study. At the same time, the inclusion of
8
the assessment of reading skill in Kom gives us a “true” read on the development of reading
skills in the experimental program.
A quick scan of scores for the children in the standard (English) program again (or still)
reflects group performance at or below the level of random guessing. About the only
plausible conclusion to be drawn is that children in the standard program, as a group, are still
not reading at all. Of course, this characterization as non-readers has to be recognized as
being intertwined with weak or very weak skills in English. However, the fact that the KEPP
children scored in the range of 40 percent on all sections of the English test pushes us to one
of two conclusions: either (a) their level of English is substantially higher or (b) they have
learned to read reasonably well in Kom and can leverage this ability to perform at a higher
level despite weak skills in spoken English.
The evidence from the math section of the test lends additional support to the argument that
the primary effect influencing performance is language of instruction rather than proficiency
in English. Note, for example, the disparity in the performance when doing basic
calculations. This is not a language-based task yet the KEPP children scored 140 percent
higher on this sub-skill. The primary implications would seem to be that the KEPP children
have made more progress in mastering math facts and processes BECAUSE of their
instruction in a language they understand.
Note also the low performance by children from both programs on problems having to do
with sets (union and intersection). Both programs performed at or below the level of random
guessing. Even though sets are introduced into the curriculum in Cameroon in Class 1, the
concept of sets is clearly not being learned. The consistently poor performance on problems
involving sets suggests that the concept of sets is probably too abstract for young children
(and maybe for their teachers as well). We would recommend removing this topic from the
curriculum for the early grades and giving the corresponding time to basic math facts which
are more useful to children and which clearly need more instructional time to develop
mastery.
Performance profile for Class 3
The performance profile for Class 3 (2011) in Figure 2 (next page) was constructed by using
the assessment scores for the English-medium program straight from the data. However, the
data for the KEPP children is a composite of their performance on the common math test
(given in English) and their performance on the Kom reading assessment rather than the
English reading assessment. We felt this would produce a profile more consistent with the
instructional experience of these children.
9
Figure 2. Performance profile for Class 3 based on overall scores (2011).
The distributions for both groups more closely approximate a normal distribution than we
saw in the profile for Class 2. The primary difference is that the distributions have radically
different shapes. The English profile is centered on 20-30 percent while the KEPP profile is
centered on 50-60 percent. Those in the KEPP program are clearly exhibiting substantial
learning while those in the English program are still characterized by performance which is
not very different from that generated by random guessing. In fact, there are still more
scores in the two lowest categories than one would predict from a random guessing model,
but fewer than seen for Class 2. A slightly cynical interpretation would be that the children in
Class 3 have become better at random guessing than those in Class 2. A more positive
interpretation is that we are seeing the faint outlines of actually being able to figure out the
content of the test but little evidence of actually being able to read.
It is reasonable to ask of the KEPP program why there are still a substantial number which
lag behind in the range of 30-50 percent. Our best guess at this point is that we have a case of
a potentially effective instructional model which is running into the various systemic
limitations of rural education in Cameroon—inadequate time on task, weak pedagogy, under-
age children in class, lack of supervision, etc. Improvement in educational outcomes
dependent directly on the intervention are evident, but there are constraints on the overall
reach of this intervention due to the broader systemic limitations of the educational system in
place.
Results for Class 4
The results for Class 4 are of intense interest as this group includes a cohort of children who
completed 3 full years of Kom-medium instruction and are now fully immersed in an all
English environment along with those who have been in an English-medium environment
from Class 1.
As in the past, the assessment instrument for Class 4 was based on content of the curriculum.
The evidence is that some teachers/schools covered most or all of the curriculum during the
year while others did not. We noticed, in processing the data that children in some schools
consistently did well on certain tasks or questions while others often did poorly.
0
20
40
60
80
100
120
140
160
0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100
Performance Categories (percentage)
Nu
mb
er
of
Stu
de
nts English KEP
10
The testing format used this year for Class 4 was a little more challenging than in past years,
especially for the reading section of the assessment. Following assessment strategies
currently being employed in the US, we selected longer reading passages and then asked
questions about these passages which probed not merely basic comprehension, but also more
general comprehension, ability to draw inferences from the passage, ability to generalize
about the passage, as well as knowledge of specific lexical items encountered in the various
passages.
In addition, the multiple choice format was made more challenging by making use of more
subtle distinctions than in past years. (For example, a question about a particular outcome of
a story might include response options which were culturally plausible, but not correct in
terms of the content of the passage.) We expected the assessment to be quite challenging for
Class 4 students and a good test of their strengths and weaknesses as readers.
The math section of the assessment included 31 items while the language arts section had 23
items. The details of what these items covered will be reflected in the subsequent tables
which present the results from the assessment.
Overall results
The assessment for Class 4 included just two major components—language arts (reading) and
math. The results are presented in Table 4. In the table, we will continue to use the labels
standard or English to designate children coming from this program and KEPP to refer to
children who came through the experimental program (though they are no longer in a mother
tongue program). In Class 4 ALL children are now in English-medium classrooms receiving
the same instruction in the same language. An examination of the data revealed very little
movement or mixing of students in schools so the cohorts by classroom continue largely as
they were in the earlier classes.
Table 4. Summary results from Class 4, 2011.
Standard KEPP
N Mean SD N Mean SD Gain
(%) Statistics
Language
Arts/Reading 361 28.1 11.4 204 38.7 13.5 37.7 T = 9.48; p = 0.000
Math 361 27.3 15.9 204 40.1 19.2 46.9 T = 8.09; p = 0.000
Overall 361 27.6 11.5 204 39.5 14.7 43.1 T = 9.93; p = 0.000
Several observations can be made about these data. First, the children coming from the
standard program continue to score at the level of random guessing or just slightly above that
level. Second, the scores of the children coming from the KEPP program no longer show the
larger spreads or differentials we saw in the earlier classes. Third, even receiving identical
instruction and in a second language, the children who participated in the KEPP program still
show substantial carryover benefits from having been in that program. Overall, these
children scored 43 percent higher than did the children from the control group. Fourth,
neither of the groups scored well. We will explore possible explanations in the interpretive
section which comes later.
A more detailed look at the data
Table 5 examines in more detail the content of the assessment and the relative performance of
the two groups of Class 4 students.
11
Table 5. Detailed comparison of Class 4 students on the sub-skills of the assessment (2011).
Standard Program KEPP Program
N Mean SD N Mean SD
Gain
(%) Statistics
Language
Arts/Reading 361 28.1 11.4 204 38.7 13.5 37.7 T = 9.48; p = 0.000
Comp. factual 361 32.5 25.3 204 46.4 26.8 42.7 T = 6.07; p = 0.000
Comp. holistic 361 28.5 31.9 204 42.4 37.5 48.8 T = 4.45; p = 0.000
Conceptualization 361 24.1 24.7 204 30.6 26.8 27.0 T = 2.82; p = 0.005
Word knowledge 361 25.2 19.4 204 24.5 18.6 -.03 T = 0.39; p = 0.698
Punctuation 361 22.6 23.3 204 46.1 32.1 104 T = 9.15; p = 0.000
Grammar 1 361 35.6 33.5 204 49.8 35.3 40.0 T = 4.66; p = 0.000
Grammar 2 361 33.4 26.5 204 40.8 27.4 22.2 T = 3.13; p = o.002
Math 361 27.3 15.9 204 40.1 19.2 46.9 T = 8.09; p = 0.000
Calculations 361 28.9 20.2 204 42.0 23.9 45.3 T = 6.57; p = 0.000
Fractions 361 18.6 20.8 204 25.7 25.4 38.2 T = 3.40; p = 0.001
Geometry 361 28.3 24.7 204 45.1 27.6 59.4 T = 7.19; p = 0.000
Story problems 361 27.5 23.6 204 39.4 25.3 43.3 T = 5.52; p = 0.000
Overall 361 27.6 11.5 204 39.5 14.7 43.1 T = 9.93; p = 0.000
First, we note that the performance pattern for those coming out of the standard programs
continues to be in the range of 20-30 percent, what we have so far characterized as the level
of random guessing. At the same time, we see some glimmers of upward performance gains
with scoring on three sub-skills above 30 percent—comprehension of factual information in a
text, grammar 1 (the proper use of the English comparative), and grammar 2 (use of
tense/aspect in English). There is also an increase in the number of mean scores which are in
the high 20s approaching 30 percent.
On the KEPP side the major finding is that scoring has dropped from the 60 percent range
which characterized Classes 1 and 2 and Class 3 when assessment involved reading
comprehension in the Kom language. In Class 4 we find performance hovering in the low 40
percent range with a few sub-skills dipping below the 40 percent range.
Several of the sub-skills stand out as being atypical of the overall pattern. Knowledge of
English vocabulary was very low for both groups as was ability to handle fractions. Students
from both programs also had difficulty thinking broadly about a text (main point, possible
title, summary statement). The very low performance level on fractions is actually rather
common in developing countries. Obviously, fractions are hard to teach and to learn. An
examination of individual test papers indicated that students definitely wrestled with the
problems on fractions but simply had not mastered the key conceptual notion of a fraction
and its relationship to a whole nor the processes for dealing with operations involving
fractions. This is an area which clearly needs more attention in instruction and more practice
on the part of students.
Looking at the math section, it is clear that even basic math facts have not been mastered by
the end of Class 4. Students commonly did addition when the indicated symbol (and written
instruction) was to multiply. Similarly, students did multiplication (usually incorrectly) when
division was the indicated task. In our judgment, in addition to the limitations imposed by
language of instruction, students do not receive enough instruction or enough practice to
consistently master even basic math tasks.
A few comments about English and English reading are also worth noting at this point. First,
the KEPP children have clearly learned more about the formalisms of language such as
punctuation and grammar. It is also evident that their previous experience with Kom reading
12
has imparted some basic skills in terms of how to attack a text for comprehension purposes.
The KEPP children were more adept at finding information in a text whether literally given in
the text or expressed in an indirect manner. They were also somewhat better at thinking
globally or holistically about a text.
One of the striking findings of the assessment is that both groups have a poor grasp of
English vocabulary beyond the basics. Neither group was able to use text and context to
determine the sense of a word which they might recognize but which was being used in a
more nuanced way. This lack of knowledge, in turn, significantly reduced reading
comprehension (Droop and Verhoeven, 2003).
A performance profile for Class 4
Figure 3 presents a performance profile for all students tested in Class 4. We draw two
primary conclusions from the profile. First, there has been a significant reduction in the
distinctiveness of the profiles seen for the earlier classes. The two profiles are now closer
together (in their mean) and more similar in their distribution. Secondly, the profile of the
students originating from standard schools is still centered in the 20-30 percent category
indicative of random guessing.
Figure 3. Performance profile for children in Class 4 (2011).
It is not difficult to conclude from this profile and the other data presented above that children
in Class 4 whose previous educational experience was in the standard program have either
still not learned to read or have not learned enough English to understand what is being
taught in the classroom (or both). Their performance on the math section of the Class 4
assessment (which requires almost no language knowledge) certainly supports a conclusion
that minimal learning is taking place whether because of the language of instruction or
because of other factors in the educational environment.
The profile of the KEPP group is much less positive than in previous classes. Two or three
conclusions are possible. (1) As many have suggested, the early exit from a more supportive
mother tongue instructional environment has negative educational consequences as reflected
in the relative similarity of the two profiles. (2) None of the children are truly ready for
English-medium instruction. The smaller observed advantage for the KEPP children is
merely a carry-over from the more favorable instructional experience in Classes 1-3. (3) The
0
20
40
60
80
100
120
140
160
0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100
Performance Categories (percentage)
Nu
mb
er
of
Stu
de
nts
in
Ra
ng
e
English KEP
13
lower performance in Class 4 by the KEPP students is temporary. Once they begin to get a
handle on English, the educational foundation laid earlier in Classes 1-3 will enable them to
make stronger progress in school in Class 5 and beyond.
The next section presents findings from Class 5 which begin to shed some light on which of
these three possible conclusions is most apt.
Results from 2012 for Class 5
The assessments
The assessment instruments and methodology used in 2012 for Class 5 is very similar to that
of previous years especially for Grades 3 and 4. Students in both comparison and
experimental schools were tested in English/reading and in math. In the case of math, there
were 30 test items taken from the Class 5 math textbook.
The assessment of English/reading was somewhat more exploratory in nature. Half of the
assessment was devoted to reading comprehension and half to an assessment of knowledge of
English vocabulary. The intent of this latter component was to develop a better
understanding of the relative impact of the two models—comparison and experimental—to
the developing of mastery and skill in English. This seemed especially relevant for several
reasons. First, we wanted to test the widespread assumption that children will learn more
English by being placed entirely in an English-speaking environment in school. Second, we
want to investigate in greater depth the impact of the initial L1 instruction on mastery of a
second language. Third, we wanted to attempt to distinguish somewhat between the process
of “learning to read English” and that of “learning English.” All too commonly, when these
are treated as identical processes, one or both fail to develop in a healthy and robust manner
especially in developing countries. Fourth, we wanted to attempt to investigate the impact on
educational outcomes when teachers speak the language of instruction only as a second
language with varying degrees of proficiency. The following table gives a more detailed
breakdown of the content of the two assessments.
Table 6. Content of the two assessments administered to Class 5 in 2012.
Number of
Test Items
Comments
Reading comprehension
of English texts
All test items in this section were 4-option multiple choice.
‘The Fire’ story 9 Text taken as is from the Class 5 reader used as a textbook in
Cameroon. The items tested comprehension of the story at several
levels and three items of English grammar.
‘Tenali’ 10 Text taken from a Class 3 standardized test used in the US. All items
were comprehension questions
English vocabulary All words in this vocabulary assessment were taken from the Class 5
reader used as a textbook in Cameroon
Word meaning 7 Find the best definition of the target word
Synonyms 3 Find the word with a meaning closest to that of the target word.
Antonyms 3 Find the word with a meaning most opposite to that of the target word.
Words in context 7 Choose the word that most naturally completes the target sentence.
Mathematics 10 free response items and 20 multiple choice items. All items adapted
from the Class 5 math textbook.
Basic operations 6 Multiplication and division
The number system 6 Sq. root, conversion between fractions and decimals
Fractions 7 Addition, subtraction, multiplication and division
Geometric operations 3 Area and perimeter
Word problems 4 Using basic operations
Reading a graph 4 Interpreting a bar graph about traffic in Bamenda
14
A note on the schools and the tracking of students
Before presenting the findings from the assessments done in 2012, it is necessary to give
some background about the research population. The discussion here will focus on changes
in that population between 2010, the year the current Class 5 students were in Class 3, and
the present.
The Boyo Division is quite densely populated. One consequence is that parents often have a
number of choices as to where they send their children. Typically, younger children are sent
to the closest school available. However, as children get older, other factors influence such
decisions including cost (if one wishes to send a child to a private school), distance, one’s
job, the ability of the child, and public perception about the quality of the school (or teacher)
involved.
The research results for Class 3 in 2010 were based on assessment data coming from 24
schools—twelve experimental schools and twelve comparison or match schools. There were
247 students in the experimental schools and 314 students in the comparison schools.
However, a comparison of enrollment data between the 2010 and 2012 school years indicates
the presence of very significant changes in the research population during this period of time.
The key data are set forth in the following table. (Warning: This table will be somewhat
confusing until the reader has read the explanation which follows.)
Table 7. Data on the identity of students in the 2012 assessment of Class 5.
12 Experimental Schools 12 Comparison Schools
KEPP English KEPP English
2010 (Class 3) 247 - - 314
2012 (In the Class 5 data set) 163 100 5 2966
2012 (From 2010) 118 - - 79
A comparison of the enrollment data between 2010 and 2012 for the same set of schools
reveals that there have been rather dramatic changes in the research population between these
two years. The original population in 2010 included 247 students in the 12 experimental
schools which had been present in these schools for 3 years. The Class 5 data set contains
only 163 students coming from experimental schools. Of these, only 118 or 47.8 percent are
still in the same schools they were in during the 2010 school year. Obviously, it would be
relevant to know what happened to the 84 KEPP children who are not present in the Class 5
data set. Reports from educational personnel in the area indicate there is a lot of movement
of families into and out of the area for agricultural purposes. Also, there is movement within
the area as children (or parents) opt to go to different schools.
The changes in the comparison schools are equally dramatic. The Class 3 population in the
comparison schools was 314 in 2010. The population of children coming from comparison
schools was still 296 in Class 5 suggesting much continuity and little loss. However, a
comparison of actual enrollment details indicates that only 79 of the original cohort of 314
students in the comparison schools in 2010 are still present in the same schools in 2012. This
means a turnover of almost 75 percent in the span of just two years. Furthermore, we find
that there are now 100 students from English medium schools (but usually not from the
original 12 schools picked for comparison) now enrolled in the experimental schools. An
examination of just which experimental schools had the biggest gains of English-medium
children indicates that those experimental schools which had achieved the highest mean
scores in 2010 are now the schools attracting the greatest number of English-medium
6 The total number of students in this category was originally 328. Thirty-two of these were removed because of
apparent or admitted irregularities in test-taking.
15
children. Apparently, parents notice which schools are doing well and hope that sending their
children to these schools, even though the medium of instruction has reverted to English, will
improve their chance for success.
These changes have been taken into consideration in the analysis which follows. We begin
with an overall comparison of performance on the two assessments.
Table 8. Comparison of experimental and comparison children 2 years post intervention (Class 5).
Standard KEPP
Mean SD Mean SD Gain
(%)
Effect Size
(D)7
Statistics
Language
Arts/Reading 26.5 9.88 34.9 10.7 31.7 .777 T = 9.01; p = 0.000
Math 16.3 7.53 17.3 8.22 6.1 .129 T = 1.43; p = 0.153
Overall 22.0 6.99 27.2 7.47 23.6 .692 T = 7.89; p = 0.000
These overall results are not encouraging. Both groups performed at a low level though the
experimental group scored higher on both assessments than the comparison group. The mean
score of 26.5 on the English/reading assessment is barely above the level of random guessing
given that all items were 4-choice multiple choice items. The advantage of the experimental
group is 31.7 percent though the performance of this group is still low with a mean 34.9
percent. Note, however, that this difference is considered to be highly significant statistically
with an effect size of .777.
The math result is especially striking by how low it is. Because of the structure of the test, a
theoretically minimum possible (or likely) score is 16.67 if one knows nothing at all and
simply makes random guesses as to the answer to each question. The actual group mean
scores were 16.3 for the English group and 17.3 for the KEPP group. This very low result
does not mean that none of the children in the assessment had any math knowledge as will be
shown later. Rather it is an indication that math knowledge and skill is limited AND that
there were a good number of children who did not even try to answer many questions. More
will be said about the issue of math in a later section.
English and Reading –the big picture
Apart from experimental programs like this one, virtually all educational instruction and
materials in Anglophone Cameroon is in English. Even in the case of this experimental
project, all education beyond Class 3 is in English. This makes a reading mastery of English
critical to being successfully educated. The English assessment included two subsections:
reading comprehension and English vocabulary.
Reading comprehension
Reading comprehension was tested by means of two stories (a total of 19 test items). One of
the two stories—“The Fire”—came from the Class 5 English textbook used in the area. The
theme and the content are very evocative of rural Cameroon. The second story was a simple
narrative about a village farmer in India written at a Class 3 (in the US) level.
7 Effect size is a statistic used to give an idea of how large a difference is in a “unit” which is independent of
that used in measurement. Values of D usually fall between 0 and 1 though it is possible for D to go higher than
1. A general rule of thumb for interpreting this statistic is as follows: 0 - .3 = weak, .4 – .7 = moderate, .8 – 1 =
strong.
16
Each story was followed by a series of comprehension questions designed to measure
understanding of the story. In the case of the first story (The Fire), several of the questions
also probed for knowledge of English grammar.
The overall results for these two stories are given in Table 9.
Table 9. Reading comprehension for Class 5 students on two English texts.
Standard KEPP
Mean SD Mean SD Gain % Effect Size
(D)
Statistics
The Fire 35.8 18.8 47.2 20.0 31.8 .595 T = 6.43; p = 0.000
Tenali 21.3 14.3 27.1 13.6 27.2 .411 T = 4.47; p = 0.000
Combined 28.2 12.5 36.6 13.7 29.8 .654 T = 7.13; p = 0.000
Several facts are immediately apparent. First, comprehension of the story, “The Fire”, was
much higher than for the story, “Tenali.” None of the measures of comprehension was
especially good being below 50 percent in all cases—a level indicating that these students, on
average, are still struggling to read basic English. Third, the children with the KEPP
foundation consistently outscored the children from the English-only background by a sizable
margin, at least statistically. A later section will examine more closely specific strengths and
weaknesses in the development of reading skills in the research population.
Development of English vocabulary
Researchers such as Nation have long argued that text or reading comprehension is very
heavily dependent on the mastery of the vocabulary which make up the text. Hu and Nation
(2000) provide research suggesting that encountering as few as 1 in 20 unfamiliar words can
limit one’s ability to fully comprehend a text and to use context to grasp the meaning of the
unknown word.
Other researchers have studied the relationship between mastery of vocabulary and academic
success in general (Cunningham and Stanovich, 1997). The evidence is compelling that there
is a close relationship between vocabulary mastery and ability to do well in school (Beck and
McKeown, 2007).
Accordingly, the assessment of English sought to provide some measure of English
vocabulary as a means of understanding growth in the mastery of this aspect of the English
language. The words chosen for the assessment were all taken from the Class 5 Textbook for
English. Obviously, this does not guarantee that the words are known but it does suggest that
all students should have been exposed to this vocabulary during the year.
Three different strategies were used for testing knowledge of vocabulary. The first strategy,
labeled “meaning” in the following table, asks the student to find the best one-word definition
of the target word from those words in the list. The second strategy asks students to pick
words with similar meaning (synonyms) or opposite meaning (antonyms). The third section
asks students to choose the word which most naturally completes a sample sentence. The
following example illustrates this strategy.
1. When my uncle got up to go, he put his ________ on his head.
a. hat b. cane c. hair d. book
Every one of the choices could, in principle, be correct. One of the four choices, however, is
the most natural and logical word to complete the sentence in most contexts and that is the
17
word, ‘hat’. While it is certainly conceivable that a person could carry a cane on his head,
this would be rather strange. If a person has a cane with them, it is because they need it to
help them walk. The last thing such a person would do is carry it on their head.
It is also conceivable that a person could put his ‘hair’ on his head but this is something of a
collocational clash in that, if one has hair, it is always on his/her head. Furthermore, if
someone has a toupee or hairpiece, that person is unlikely to take it off in a public setting.
Finally, it is common to see people carrying books on their head but these are school
children, not older established adults. Furthermore, in this area, men are increasingly likely
to wear a hat as they get older.
Such items require that the respondent tap a range of knowledge about the language as well
as about their knowledge of the world—often gained by exposure to print—to make the best
choice. Table 10 presents the results of the vocabulary assessment.
Table 10. Assessment of knowledge of basic English vocabulary.
Standard KEPP
Mean SD Mean SD Gain % Effect Size
(D)
Statistics
Meaning 22.1 18.2 29.3 21.0 32.6 .377 T = 4.08; p = 0.000
Syn. and
antonyms 24.5 18.7 31.2 21.5 27.3 .343 T = 3.70; p = 0.000
Usage 27.5 18.1 38.9 18.4 41.5 .627 T = 6.79; p = 0.000
Overall 24.7 11.7 33.2 13.1 34.4 .701 T = 7.61; p = 0.000
Performance on the vocabulary assessment parallels what we saw earlier with the assessment
of reading comprehension. While overall performance is low, the children with the KEPP
foundation consistently scored 35-40 percent higher (relative gain) than children coming
from an all-English background. All differences are statistically significant and the effect
sizes are in the moderate to strong category.
If vocabulary knowledge is, in fact, essential to reading comprehension, we would expect to
find a correlation between vocabulary knowledge and text comprehension, especially of the
story “The Fire” which comes from the Class 5 textbook. A test for such a correlation found
an r of .358 which is statistically significant at the 0.000 level. While being statistically
significant, this correlation is probably lower than some would expect. This is very likely
due to the fact that the vocabulary items tested were randomly selected from the text and may
or may not actually occur in the story, “The Fire.”
Since the vocabulary included in the assessment all came from the Class 5 English textbook,
it seems safe to conclude either that the level of the language in the textbook is far too
advanced OR that the children in Class 5 are not progressing as they should be in developing
English language skills (or both).
Discussion of the findings on English and Reading
It is clear from the Class 5 data that the 3 years of Kom-medium instruction has had a
positive and ongoing impact on these students even after having switched to an English-
medium classroom. The three years spent learning to read in Kom and learning other
subjects via Kom language instruction has not compromised their ability to develop English
proficiency. In fact, the data suggest quite strongly that this early foundation has given these
students a significant advantage in making the transition to English as reflected by the 31.7
percent advantage over their peers in reading comprehension and a 34.4 percent advantage in
learning English vocabulary.
18
At the same time, we have to recognize that neither the KEPP nor the English-only students
have become proficient readers of English even after 5 years of schooling. The earlier
assessments noted that the KEPP children were reading fairly well by the end of Class 2 and
were quite proficient readers of Kom by the end of Class 3 (68 percent comprehension from
Table 3). This proficiency seems to have conveyed an advantage in learning to read in
English. At the same time, the data would seem to support a claim that these children are not
ready for full-fledged study in an all-English environment as demonstrated both by the low
levels of reading proficiency in English AND the low levels of mastery of English
vocabulary.
The literature on Multilingual Education suggests that so-called early exit models are
inherently weak in that they (a) provide little or no support for the L1 after the exit from L1
classrooms at the end of the designated time in L1—typically 3 years; and (b) do not lay an
adequate academic and cognitive foundation for making the transition to L2 instruction.
Much of this criticism of early exit models is based on the research of Cummins and others
that it appears to take 5-7 years for children to learn a second language well enough to be
effectively educated in that language. The core argument advanced by these theorists is that
language development, academic development, and cognitive development are
interdependent features of human intellectual development. When students are in an L1
classroom, these are all developing together. At the same time, these children are beginning
to learn the L2 so as to be ready for instruction in the second language. But, is it possible for
children to learn enough L2 in a span of three years to be ready for the more demanding
academic work of Class 4 and beyond? The best current research says that three years is not
enough to prepare children for this linguistic, cognitive, and academic jump to a second
language (Thomas and Colllier, 1997). The Class 5 data lends support to this critique of early
exit models. Look at Figure 4.
Figure 4. History of performance of a cohort of MLE students in an early exit model of L1 instruction.
Figure 4 traces the performance of KEPP children on assessments of English/reading and
math for the three year period beginning with Class 3—the last year of L1 instruction—
through Class 5 (results for the current year). As is evident, performance has dropped each
year—a trajectory entirely consistent with the predictions of the model put forth by Thomas
and Collier.
To be entirely fair, we cannot discount other possible explanations for this performance
history. For example, the tests have become more difficult each year in keeping with the
specifications of the national curriculum and the content of the textbooks. Maybe the tests
were too difficult. It is also possible that, at the more advanced academic levels of Classes 4
19
and 5, the teachers are less able to teach the more demanding content of this curriculum in
English. (For more on this point, see the parallel report on the assessment done in the
Teacher Training College.)
Nonetheless, we cannot escape the reality that the Class 5 children including the KEPP cohort
have not come close to mastering the vocabulary used in the Class textbook and reader on
English. If this is true of the English textbook, it is almost certainly true as well of the Math
textbook. However we want to frame a possible explanation of the findings, it seems clear
that while the foundation laid by the 3 years of L1 instruction has given these children a clear
educational advantage in English and reading, these 3 years of L1 instruction have not been
sufficient to make a productive jump to L2 instruction in Class 4 and beyond.
Results on the Assessment of Mathematics
Even a casual comparison of the math textbooks for Classes 4 and 5 indicates that a large
number of new math concepts are introduced in Class 5. Furthermore, the level of conceptual
complexity and abstractness increases as well. Together, these factors make math in Class 5
a challenging subject.
The assessment administered in Class 5 drew sample problems from the full breadth of the
Class 5 math textbook used in the region. In addition some basic foundational skills taught in
previous years were included to establish an overall indication of math skill development
during the earlier years. See Table 12 (two pages ahead) for a breakdown of the content of
the math assessment.
Ten of the items were free response items (do the necessary calculations and write in the
resultant answer) and 20 were 4-item multiple choice questions. The free response
mechanism was used for the more basic operations (addition, subtraction) assumed to have
been taught and mastered in earlier years. It was also used for factual items (What is the
square root of 16?) and for the easiest of the new material (What is the decimal equivalent of
5/8?). The overall performance of the comparison and experimental groups is given in Table
11—16.3 percent for the comparison group and 17.3 percent for the experimental group.
The results from the math assessment can be divided and compared in a number of ways.
The following tables are included to provide additional insight into the nature of the
performance on this assessment.
Table 11. Item-based comparison of performance on the Class 5 math assessment.
English KEPP
Category of Items Mean SD Mean SD Gain % Effect Size Statistics
Free response 8.9 12.7 9.4 11.2 5.6 .041 T = 0.43; p = .664
Multiple choice 19.9 9.25 21.2 10.3 6.5 .136 T = 1.45; p = 0.147
Combined 16.3 7.53 17.3 8.22 6.1 .129 T = 1.43; p = 0.153
By any measure, performance on the math assessment was extremely low. On the face of it,
flipping a coin would have produced results just as high (or low). However, when we
examine the data by individual performance, we find that some appeared to do pretty well
including on the free response items which were not possible to answer by guessing. The
following pair of histograms gives an additional picture of how children performed on this
assessment.
20
Figure 5. Performance on the Free Response section of the math assessment by group.
Out of 564 Class 5 students tested, 6 were able to correctly answer at least five of ten free
response items correctly, items judged to be the easiest or most basic in the assessment.
Seventy-five percent of both groups correctly answered one or fewer items correctly with the
majority of these failing to answer any items correctly.
Figure 6. Performance on the set of multiple choice items included in the assessment for Class 5.
The histogram of performance on the part of the assessment which was presented using a
multiple choice format (20 or 30 items with 4 choices to choose from) presents an entirely
different and instructive profile. The light and dark solid bars capture the distribution of
performance of the two groups of students. The set of patterned bars reflects likely or
predicted performance based on pure random guessing. The overall performance profile for
both groups closely mimics that of the random guessing model. In fact, it would appear that
performance would actually have been better if students had chosen to use a random guessing
strategy to answer the multiple choice questions since their actual performance was below
that predicted by a random guessing model.
21
Taken together, Figures 5 and 6 pretty much imply that math knowledge and skill—at least as
presented in this particular assessment—are extremely low, and for many students, almost
non-existent.
Math knowledge and skill by specific content areas
In this section we present another perspective on math performance based on specific
domains of knowledge and skill. These are: basic operations (6 items); the number system (6
items); applying the basic operations to fractions (7 items); geometric figures (3 items); word
problems involving basic operations (4 items); and reading a graph (4 items).
Table 12. Performance by domain of math skill and knowledge among Class 5 students.
English KEPP
Mean SD Mean SD % Gain Effect Size Statistics
Basic operations 10.8 18.4 11.0 16.0 1.85 .011 T = 0.12; p = 0.905
The number system 16.1 14.5 18.8 15.1 16.8 .184 T = 2.05; p = 0.041
Operations with fractions 12.7 13.8 12.2 12.1 -3.9 -.038 T = 0.44; p = 0.661
Geometric figures 23.4 23.5 24.6 22.0 5.1 .052 T = 0.57; p = 0.572
Basic operations in word
prob.
19.4 19.7 19.6 20.1 1.03 .010 T = 0.11; p =0.914
Reading a graph 22.4 20.6 25.4 24.1 13.4 .138 T = 1.52; p = 0.129
Overall 16.3 7.53 17.3 8.22 6.1 .129 T = 1.43; p = 0.153
Scanning Table 12, it is clear that math knowledge and skill is low in ALL domains of math
tested in Class 5. While the KEPP group shows a slight advantage in all but one domain of
knowledge or skill, the differential is, in most cases, negligible. In only one case, ‘The
number system,’ is there a statistically significant difference and even in this case
performance by both groups is exceedingly low.
It is suggestive that the two math tasks in which both groups scored the highest entail a visual
presentation of information or data—working with geometric figures and reading a graph.
Conversely, the weakest two domains were basic computation (multiplication and division)
and computations involving fractions. It is our informed hunch that a major reason for the
poor performance in these areas is simply the lack of actual practice. Typically, a teacher
will demonstrate an example or two on the blackboard, ask students to copy these into their
notebooks, and thereby assume that the operation in question has been learned.
What about the overall poor performance in math? Having worked in the area for 6 years
observing classes and talking to teachers and educational officials at all levels, we conclude
the following:
1. Time on task. Under the best of conditions, the school week contains only
approximately 20 instructional hours. The curriculum consists of 13-15 subjects to be
taught each of which must be given due time. The time designated for math in the
timetable is 5.5 hours. Classroom visits and informal conversations with education
personnel suggest that the actual amount of instruction time given to math is
frequently less than this.
2. Practice and drill. Very few children have textbooks and there is no provision for
workbooks or seatwork to practice the material taught in the classroom. Homework is
not often given.
3. Automatic promotion. The current system of automatic promotion means there is
little concern or accountability for subject mastery. This applies especially to math
which is a much feared and dreaded subject among both students and teachers.
22
4. Local mythology about math. A history of low performance in math has produced a
local phobia about math. Many children enter school assuming that math is difficult
if not impossible so learn to fear and dread math classes.
5. Subject level mastery by teachers. From the onset of this research project, we have
heard comments that Cameroonian students struggle with math. The data from Class
5 provide vigorous empirical support for this somewhat vague but apparently accurate
conclusion. Various observers have questioned whether the teachers know the math
themselves. Testing done in one of the teacher training colleges along with
discussions with education personnel with experience in the area point to weak
teacher skills as a significant part of the problem. For example, we have learned that
new teachers can navigate their way through middle school, high school, and teacher
training without being required to take a single math course. This means such
teachers are teaching math almost entirely based on what they themselves learned in
primary school. The Class 5 data reflect the consequences. We also learned that
while there is a qualifying exam students must take to enter the teacher training
college, administrators are forced to depress the qualifying marks for math in order to
meet their government specified quotas for new teachers. Finally, we’ve actually
tested the math skills of students in the teachers training college which has
corroborated the suspicion that weak teacher knowledge and skill is a major
contributing factor to the current situation.
An obvious question is, “If the Class 5 children have done so poorly in even basic math, how
do they manage to achieve passing marks on the First School Leavers’ Exam (FSLE) which
is required for entry into a secondary school?” Two local explanations have been proffered.
(1) The division (school district) provides intensive tutoring for those preparing to take the
FSLE. (2) The FSLE is taken in a very “supportive environment.”
Interpreting the Class 5 results with respect to the Kom multilingual education project.
One of the initial motivations for the KEPP research project was that of investigating the long
term impact of L1 or mother tongue instruction in a school context. Research in various
Western nations had established expectations and even models of how L1 instruction worked
in a Western context in terms of long term academic impact. There are, however, many very
significant differences between educational practice in the West and that of many low-income
countries.
The KEPP project is now in its fifth year of operation. Two cohorts of children have been
through the complete three-year cycle of L1 instruction and have returned to the traditional
L2 instructional model (via English) which has been in place in the region for many years.
We now have assessment data for all five years of program operation including for the initial
cohort which just completed Class 5. We are now in a position to study and model the
interaction between language of instruction and educational outcomes in a rural Cameroonian
context. What can we learn from this data? Table 13 contains the essential data.
23
Table 13. Summary of results in the Kom research project for 5 years.
Standard KEPP
Mean SD Mean SD Gain
(%)
Effect Size
(D) Statistics
Class 1 - 2008 Language Arts 8.4 11.4 62.8 25.3 648 2.79 T = 35.54; p = 0.000
Math 9.9 17.0 44.9 32.8 354 1.24 T = 17.16; p = 0.000
English (oral) 35.4 21.9 52.5 26.3 48 .708 T = 9.12; p = 0.000
Overall 17.6 12.5 53.4 22.2 203 2.00 T = 25.67; p = 0.000
Class 2 - 2009 Language
Arts/Reading 26.2 17.5 51.5 24.3 98.6 1.203 T = 14.77; p = 0.000
Math 23.6 21.5 51.3 26.6 116 1.150 T = 14.32; p = 0.000
English (Oral) 51.0 21.3 59.5 20.1 16.7 .410 T = 5.09; p = 0.000
Overall 35.6 15.4 54.7 17.8 53.6 1.150 T = 14.34; p = 0.000
Class 3 - 2010 Language Arts &
Reading (Eng.) 19.9 13.0 40.6 16.8 104 1.400 T = 16.46; p = 0.000
Langage Arts &
Reading (Kom) - - 65.6 23.2 - -
Math 21.8 17.5 49.5 24.3 127 1.333 T = 15.67; p = 0.000
English (Comp.) 17.4 17.8 38.9 23.7 124 1.043 T = 12.29; p = 0.000
Kom (Comp.) - 66.4 30.0 282 2.070 T = 23.75; p = 0.000
Overall 20.1 12.7 53.9 20.1 133 1.690 T = 18.09; p = 0.000
Class 4 - 2011 Language
Arts/Reading 28.1 11.5 38.9 13.5 38.4 .888 T = 10.18; p = 0.000
Math 27.3 16.0 40.0 19.2 46.5 .738 T = 8.44; p = 0.000
Overall 27.6 11.5 39.5 14.7 43.1 .932 T = 10.67; p = 0.000
Class 5 - 2012 Language
Arts/Reading 26.5 9.88 34.9 10.7 31.7 .777 T = 9.01; p = 0.000
Math 16.3 7.53 17.3 8.22 6.1 .129 T = 1.43; p = 0.153
Overall 22.0 6.99 27.2 7.47 23.6 .692 T = 7.89; p = 0.000
The data in Table 13 track two ‘cohorts8’ of children through five years of education in Boyo
Division. Assessment results have been broken out by Class, model (KEPP or English), and
subject. In addition to the mean scores, the table includes several other columns of interest to
researchers and analysts. The ‘% Gain’ column shows the improvement or percent of gain of
children in the experimental program when compared to those in standard or English-medium
schools. A ‘gain’ of 100 percent means that the experimental group has scored at a level
twice as high as that of the comparison group. The column label “Effect Size” gives a value
favored by many researchers to indicate the size of the impact of the experimental variable.
See footnote 9 for an explanation of how to interpret the values in this column. The column
labeled “Statistics” contains the test statistics performed on the data.
Several aspects of the table stand out. (1) Through Class 3, gain scores of 100 to 300 percent
were common. (2) After Class 3, gain scores are generally below 40 percent though most of
these are still statistically significant. (3) Directly related to points 1 and 2, the mean scores
for the experimental group declined in Classes 4 and 5 though they still exceed those of the
8 On the face of it, these two cohorts—one in the 12 experimental schools and one in the 12 comparison schools
are the exact same children being tracked over a 5 year period of time. In actuality, as we have already pointed
out, there is a substantial yearly turnover in the schools of the area meaning that the groups tested in Class 5 are
not the exact same children assessed in Class 1 in 2008.
24
comparison group. (4) Overall mean scores for the comparison group exceeded 30 percent
only once—Class 2— during the 5 year research period.9 (5) Math scores dropped quite
dramatically in Class 5.
To provide a visual picture of the results, the following two graphs have been prepared. The
curves have been plotted on the exact same Y-scale so that more direct comparisons were
possible. We chose not to put all of the curves on a single graph because of the visual clutter
this produced.
Figure 7. Five year history of results for all comparison students (English only) for the period 2008 to
2012.
From the graph it is clear that performance has consistently hovered in the 20-25 percent
range on both subjects tested. There does appear to be a slight upward trend in performance
on the language arts assessments of about 4 percent per year. Since the assessments have
been based upon the specified learning outcomes of the curriculum, the data indicate that this
cohort has reached the point at the end of Class 5 where it is doing better than merely
guessing at the answer. This coincides fairly well with the baseline data which suggested that
students actually begin making progress in learning to read English during Class 5 with this
progress accelerating in Class 6.
The performance in math is less optimistic. The overall slope for math is just 1.6 percent, an
amount which could either indicate learning or it could be a testing artifact having to do with
the difficulty level of the assessments used or the increasing difficulty of the subject matter
(of math). The substantial drop in Class 5 appears to coincide with a rather steep increase in
the level of difficulty of the content of the Class 5 math curriculum.
Now, we turn to an examination of the five year trajectory of the Kom medium students set
forth in Figure 8.
9 The mean score for the comparison group in Class 2 was somewhat inflated by high scores on an Oral English
task included in the Class 2 assessment which was probably too easy. Of course, the mean score for the
experimental group would have been inflated by the same amount.
25
Figure 8. Five year history of assessment results for all experimental (KEPP) students for the period 2008
to 2012.
Figure 8 has a vertical black bar separating the two distinct education phases of the KEPP
program—the Kom-medium instruction in Classes 1-3 and the English-medium instruction in
Classes 4 and 5. The performance contrast between these two phases is dramatic. During the
first three years of the mother tongue instruction, the KEPP students demonstrated overall
mean scores of around 53% for each year. Their performance on the assessments of language
arts and reading comprehension were generally higher than that. However, once these
children shifted to English-medium schools in Class 4, their performance began to decline.
In Class 4, the mean performance had dropped to 40 and by Class 5 it had dropped further to
27 percent. Even so, the performance of the Kom-medium students has exceeded that of the
English-only students. See the following graph for a cross-group comparison of overall
performance.
Figure 9. Comparison of overall scores for experimental and comparison groups for 5 years.
Apart from the somewhat high (and explainable) performance of the comparison students in
Class 2, the comparison group shows a generally consistent level of performance between 20
and 30 percent. The KEPP program performed consistently at the 53-54 percent level
26
through Class 3 and then began dropping when instruction returned to English in Class 4. In
Class 5 the advantage has shrunk to 6 percentage points in absolute terms and to a relative
advantage of about 30 percent. This relative advantage is still statistically significant but not
very compelling in the overall picture.
Visualizing the impact of the KEPP innovation
In Tables 8-12, we presented the mean performance of each group of students in Class 5 in
statistical terms. In this section, we present some additional methods for understanding the
nature of the persisting impact of the KEPP program even though the students in Class 5 are
two years removed from that program. Look at Figure 10.
Figure 10. Distribution of Class 5 students into four equal quadrants based on performance and program
history.
In Figure 10, each bar represents the percentage of each group performing at each level. Q1
is the lowest performing group (1/4 of the entire group), Q2 is the second lowest performing
group and so on. Looking at the graph, it is easy to see a pattern. There is a much higher
presence of English-only students in the lowest performing 25 percent of all students.
Conversely, in the highest 25 percent of all students, the percentage of KEPP background
students is twice as great as that of English only students. The continuing impact of the
KEPP background manifests itself clearly in this ranking of all Class 5 students (who
participated in the assessment).
Another (and somewhat abstract) method of portraying the continuing educational impact of
the KEPP experience is to make use of the distributional characteristics of the performance of
the students in combination with the fact that intelligence (IQ) is normally distributed in most
natural populations. Using the mathematical characteristics of the normal distribution, the
overall score on the Class 5 assessment can be converted into IQ scores for all students which
can then be graphed against performance on the actual assessment administered to the Class 5
students. (In looking at the graph, it is VERY IMPORTANT to note that there is never a
linear relationship between a given test and actual real IQ. The data in the graph ignores this
fact BUT assumes that variation between the test taken and real IQ is probably equally
randomly distributed in both groups.) The horizontal spacing between the two plots can be
interpreted as educational benefit (expressed in IQ units) which has accrued to students in the
KEPP program due to the 3 year experience in mother tongue classrooms.
27
Figure 11. Impact of the KEPP innovation expressed in terms of an estimated IQ.
If one draws a horizontal line from any point on the vertical axis (the assessment scores) to
the point where this horizontal line intersects the two lines of markers and then drops a
vertical line to the X axis (the estimated IQ scores), the difference between the two points of
intersection on the X axis is a measure of the “IQ impact” of the KEPP program. The
following table gives some examples to make this point easier to follow.
Table 14. Estimated IQ needed to achieve a given score on the Class 5 Assessment.
Selected scores on the
Class 5 assessment
English
only KEPP Difference
10 74 65 9
20 96 86 10
30 117 105 12
40 139 126 13
Table 14 gives specific results of applying the technique described in the preceding
paragraph. We will pick the score of 30 on the Class 5 assessment to illustrate how this
works. In the row where the score of 30 is found we find the number 117 for the English
only group and 105 for the KEPP group. These numbers suggest that a student coming from
the English only program would likely need an estimated IQ of 117 to score 30 on the Class 5
assessment, while a student coming from the KEPP program would only need an IQ of 105 to
achieve the same score. Plainly stated, the implication is that children coming from English-
only schools need a higher IQ to score at the given level compared to their peers coming from
an MLE school.
The Class assessment results suggest that the educational benefit of the KEPP experience
translates into a performance advantage in the classroom equal to approximately 12 points on
the standard IQ scale. This is quite a large difference. The following example from the US
will help the reader grasp the significance of such a difference. An IQ of 105 is just slightly
above average. A student with an IQ of 105 finishing secondary might or might not be
accepted at a low level state university for further study and, if accepted, would have to work
hard to finish a first degree in a relatively non-challenging academic area such as social work,
English, or general business. On the other hand, a person with an IQ of 117 could easily go
to almost any major university, take almost any course of study, and likely be accepted to do
a master’s degree if he or she so chose.
28
Again, we want to emphasize that the data in Table 14 do NOT state or imply that KEPP
students have higher IQs than English students. Rather, the point of the table is that the
educational start the KEPP children got in Classes 1-3 in the mother tongue program has
given them an educational advantage equivalent to having a higher IQ.
The issue of mixed classrooms
There is one final feature of the data set we need to investigate before presenting our final
summary of the data. In Classes 1-3, strict separation was maintained between models.
Children were generally not permitted to move between the experimental and control schools
though there were a small number of transfers into the KEPP schools in Class 3. The impact
of this late entry was investigated and quantified in the 2010 report.
However, all instruction after Class 3 is now in English. Little or no effort was made to
maintain strict separation between the programs in terms of the background of students
entering a school which had been designated as an experimental school. In fact, upon
examining the data, it appears that the experimental schools which had the strongest Class 3
students tended to attract a lot of enrollment of children coming from English-only schools
(though not always comparison schools). The result was that the schools assessed in 2012
can be divided into three categories:
1. Comparison schools (N = 12) whose students are entirely or almost entirely from English-
only schools. We found 5 students out of a population of 296 enrolled in these schools.
2. Mixed schools (N = 8) whose Class 5 populations were approximately a half and half
mixtures of KEPP and English-only students. Solely for the sake of reference, we will refer
to these schools as “mixed” schools.
3. KEPP schools (N=4) whose Class 5 populations were primarily (at least 75%) made up of
KEPP background children. Not insignificantly, these four schools tended to be among the
weaker of the KEPP schools. (Solely for the sake of reference, we will refer to these schools
as “pure” schools.
This distributional effect gives one an additional opportunity to look for possible educational
effects deriving from the KEPP MLE model. If, for example, the children in the “pure”
schools performed differently (either higher or lower) when compared to those in the mixed
schools, we learn something further about the continuing impact of the KEPP innovation.
Table 15 presents the results of this additional analysis.
Table 15. Comparison of results from 3 levels of ‘mixing’ of experimental and comparison students in
Class 5 in 2012.
Comparison Schools Mixed Schools ‘Pure’ KEPP Schools
English (296) KEPP (5) English (96) KEPP (108) P-value English (4) KEPP (55)
The Fire 35.3 42.2 37.6 49.3 0.000 30.6 43.4
Tenali 21.2 22.0 21.6 27.6 0.002 25.0 26.7
Reading 27.9 31.6 29.2 37.9 0.000 27.6 34.6
Vocabulary 24.5 24.0 24.8 34.7 0.000 33.8 31.1
English 26.3 27.7 26.7 36.2 0.000 30.8 32.8
Math 16.1 24.7 17.2 18.0 0.448 8.3 15.2
Overall 21.9 26.4 22.5 28.3 0.000 21.0 25.2
Table 15 does not reveal anything dramatic though there are some tendencies. The KEPP
children in the mixed schools tended to outperform the KEPP children in other schools by
roughly 10 percent. The KEPP children in the comparison schools fared the worst but since
there were only five of them, we are not able to draw reliable conclusions about this
educational situation. The English-only children in the mixed schools tended to outperform
their peers in the comparison schools but not by a large amount. This tendency probably
supports the earlier observation that the mixed schools are probably a little stronger
29
academically than the “pure” KEPP schools which is why they attracted many more English-
only students in the first place.
Gender
There is a small skew in the distribution of students by gender into the two programs in Class
5. Males are slightly over-represented in the English-only cohort and females are slightly
over-represented in the KEPP cohort. Statistically, the difference is significant (p = 0.009).
Among English-only students, males outscore females on most measures but none of these
differences are statistically significant. Among the KEPP background children, female
children outscore male children on all measures but math but the differences are not
statistically significant. Accordingly, gender does not appear to be an explanatory value in
explaining group differences.
Age
Children in Class 5 range in age from 7 to 16. About 5 percent of children are under the age
of 10 at the end of the school year. There is a slight negative correlation between age and
performance but the correlation is not strong enough to be statistically significant. The 7 year
old actually scored well above average so age did not seem to be an issue in that person’s
participation at least academically.
Socioeconomic status (SES)
The assessments this year included a series of questions for estimating a student’s
socioeconomic status. (This item was added because so many in the international community
have inquired as to whether we have monitored or controlled for this variable (which is
known to be significant in western countries).) We found virtually no relationship at all
between the proxy measures of SES we used and overall student performance.
Education of mother
Much educational research in developing countries has found or suggested a link between the
level of a mother’s education and the potential for success in school on the part of her
children. We did not find such an effect in the Kom data. In fact, we found a slightly
negative correlation between mother’s education and educational performance.
School quality as a variable
It is normal to find variation in the quality of schools in any educational setting and this is
true in the Kom research project as well. Figure 12 plots the school means for all schools
participating in the study.
30
Figure 12. Comparison of the performance of all schools in the research project.
The range between high and low-performing schools varies from 7.8 for the experimental
schools to 9.5 for the comparison schools. Otherwise, they appear to exhibit generally
parallel variation in performance. The three comparison schools which seemed to perform at
a slightly lower level than would be predicted do not seem to have much in common. Two
are in more remote parts of the Kom area and one is in the largest market town which is the
government center for the area.
A statistical analysis reveals that for the entire research population, the two variables of
school quality and model or program (KEPP – Standard) accounts for 27 percent of the
variation in students scores. Of this amount, school quality accounts for about 75 percent and
program history accounts for about 25 percent. This is imminently plausible since students
from the two programs are now somewhat mixed between the two sets of schools.
However, if we examine just the experimental schools, we find that the proportion of
variance accounted for by the program variable is higher—around 45 percent. This is
probably because more KEPP-background students are found in these schools.
Conclusions and recommendations
The summary reports from earlier years of this research project have simply reported the
results of the KEPP experimental program using the assessment data as the primary evidence
of impact of the innovation. We are now in the fifth year of this longitudinal study. These
five years of research have begun to establish a clearer picture of the nature of education in
Boyo Division and some of the primary issues affecting the quality of educational outcomes
in the area (and by extension in much of the North West Region). Accordingly, this year’s
report will take a step beyond simply reporting what has been found. Included will be a
section which offers some steps which could be taken to very significantly enhance the
quality of basic education at least in the rural reaches of the Region. We recognize, of
course, that we have no authority to make or even propose changes. Rather, we offer these
suggestions in the spirit of attempting to translate careful research work into useful and
effective change BASED on the findings of the research being reported.
31
Summary of findings to date
1. The experimental innovation—use of the first language (mother tongue) as a medium
of instruction—did bring about very significant improvements in learning outcomes in
the first three years of the innovation. Reading comprehension reached 66 percent in
Class 3 and students were consistently scoring around 50 percent on the math
assessments. The experimental students also demonstrated stronger knowledge and
proficiency in English.
2. The data also show that the performance of the children in the experimental program
began dropping rather seriously once they left that program and returned to English-
medium classrooms—a decline that extended into Class 5 and appears to have
accelerated at least in the case of math.
3. Despite the sharp decline in performance on the part of the experimental children,
considerable evidence of the stronger foundation developed in Classes 1-3 persists in
the assessment data gathered in Class 5 especially with respect to reading and
knowledge of English.
4. The educational advantages developed in the experimental program with respect to
math knowledge and skill declined somewhat in Class 4 and has essentially
disappeared in Class 5.
In their research, carried out in the US, Thomas and Collier(1997) found that early exit
models raised long term educational performance from the 11th
percentile to the 34th
percentile relative to the mainstream (native English speaking population). This is certainly
an improvement, but still leaves children in the bottom third of all children in the US. In the
US setting, the objective of multilingual education is to bring children to full educational
parity (the 50th
percentile or above) with the native English speaking population. The
educational model which best achieves parity is that of late exit multilingual education. In
this model, children receive some instruction in both L1 and L2 through all 6 years of basic
education. This gives children six years to develop academically and cognitively in a
familiar language while also developing the proficiency in the second language needed to
achieve parity.
In the Cameroonian case, there is no such reference population—such as a large population
of native speakers of English— against which we can compare the achievement of children in
a multilingual educational model. What educational standard do we use to measure the
achievement of an educational innovation such as mother tongue education? It would appear
that we have two choices: (1) we can use the performance of children in the normal English-
only schools as a frame of reference, or (2) we can use the learning standards or outcomes of
the national curriculum as a frame of reference. The problem with the first option is that, in
current theory about multilingual education, the English-only model is the weakest of all
possible educational models for a sociolinguistic setting like that of Cameroon so any
comparisons almost always show favorable results for an innovation. The problem with the
second option is that it appears to be such a high standard given current educational practice,
especially in this part of Cameroon, that it seems almost out of reach. Nonetheless, option 2
seems our best choice at this point.
Using the national curriculum as our frame of reference and the Class 5 assessment as
administered this year, the data suggest that the English-only Class 5 children have mastered
approximately 22.0 percent of the specified learning outcomes while the Class 5 students who
came from the KEPP MLE program have mastered 27.2 percent of the curriculum. The
KEPP MLE model produced an improvement of 31.7 percent relative to the standard model.
This is progress but is it adequate progress?
32
These findings seem to confirm the best current theorizing in multilingual education that
early exit programs of which the Kom project is an example provide good initial instruction
but do not provide the foundation needed to make a successful transition to instruction in a
second language. This is why there was a steep drop in performance by KEPP students in
Classes 4 and 5. However, the Cameroonian educational environment is not similar to that
(the US) from which their theory has been derived—hence a part of the motivation for the
current research work in Kom.
Two major issues are raised at this point. First, is it really the goal of basic education in
Cameroon to achieve the objectives explicitly or implicitly incorporated in the curriculum?
And, if so, what are the steps that could or should be taken to do so?
Let us assume that the answer to the first question is, “Yes.” The strong achievement
exhibited in reading and math in Classes 1-3 suggests that, under the right educational
conditions, students are capable of achieving at reasonably higher levels than is typical. We
will build on that logic plus the findings of the Class 5 assessment to make a series of
suggestions as to steps we believe could be taken to come closer to achieving the learning
outcomes stated in the curriculum on a much broader scale.
1. Extend the L1 (KEPP) model to additional Classes 4 and/or 5 and/or 6
It has been widely assumed that the use of L1 instruction decreases the likelihood that
students will learn L2 (English in this case). The data from the KEPP program has
consistently shown that this is an invalid assumption. The KEPP students have
consistently done better in English than the English only students. It appears that the
ability to learn to read facilitated by doing so in L1 leverages and enhances one’s ability
to grapple with and learn a second language.
2. Upgrade teachers’ English (see the report on GTTC in Fundong)
3. Upgrade teachers’ math knowledge and skill (see the report on GTTC in Fundong)
4. Upgrade math instruction by:
a. Upgrading teachers (see point 2 above)
b. Assuring adequate time on task (lengthen the instructional day, eliminate peripheral
subjects in the first 4 years)
c. Providing supporting practice materials for teachers
d. Establishing some intermediate levels of accountability by doing better year-end
testing of student achievement and using the results to improve instruction
At present mastery of basic math is reduced by: (1) lack of instructional time; (3) teaching
of new and abstract concepts via a second and largely unknown language; (3) lack of
practice time to develop individual mastery; and (4) limitations in the knowledge and skill
of teachers to teach many of the basic constructs of math especially via English. The lack
of any kind of intermediate assessment means that all teachers can avoid accountability
for the poor performance of students.
By far, the most compelling lesson from the Class 5 assessment is confirmation of the
assertion by MLE theorists that early exit models of multilingual education are inherently
weak. Many writers about MLE use the term “weak” to refer to the sociolinguistic status of a
language vis-à-vis another more dominant language AND to the level of support and
endorsement given to the L1 as a vehicle for education. The data gathered during this fifth
year of the KEPP MLE project adds further nuance to the term “weak” providing evidence
that early exit programs are also educationally weak. Though students do not lose all of the
educational benefit gained during the early years of L1 instruction, they lose much of it.
We are holding open the possibility that math instruction in the research site is a special or
extreme case. Only further experience and research will answer this question.
33
References
Beck, I.L., & McKeown, M.G. (2007). Increasing young low-income children’s oral
vocabulary repertoires through rich and focused instruction. The Elementary School
Journal, 107(3), 251–271.
Bialystok, E. 1988. Levels of bilingualism and levels of linguistic awareness. Developmental
Psychology 24(4), 560-567.
Cummins, J. 1976. The influence of bilingualism on cognitive growth: a synthesis of
research findings and explanatory hypotheses. Working Papers on Bilingualism 9, 1-43.
Cunningham, A.E., & Stanovich, K.E. (1997). Early reading acquisition and its relation to
reading experience and ability 10 years later. Developmental Psychology, 33(6), 934–945.
doi:10.1037/0012-1649.33.6.934
Dale, E., & O’Rourke, J. (1981). The living word vocabulary: A national vocabulary
inventory. Chicago: World Book/Childcraft International.
Droop, M. and Verhoeven, L. 2003. Language proficiency and reading ability in first- and
second-language learners. Reading Research Quarterly 38(1), 78-103.
Galambos, J. and Hakuta, K. 1988. Subject specific and task-specific characteristics of
metalinguistic awareness in bilingual children. Applied Psycholinguistics 9, 141-162.
Hu, M and I.S.P. Nation 2000. Unknown vocabulary density and reading comprehension.
Reading in a Foreign Language 13:1.403-30.
Thomas, W. and Collier, V. 1997. School effectiveness for language minority children.
http://www.ncela.gwu.edu/files/rcd/BE020890/NCBE_School_Effectiveness.pdf Last
accessed 14 July 2010
Thomas, W. and Collier, V. 2002. A National Study of School Effectiveness for Language
Minority Students’ Long-Term Academic Achievement. Center for Research on Education,
Diversity and Excellence, Santa Cruz CA.
http://crede.berkeley.edu/research/crede/research/llaa/1.1_final.html Last accessed 5
August 2010
Walter, S. and Trammell, K. 2010. The Kom experimental mother-tongue education project
report for 2010. Unpublished research report. Available from the authors or SIL
Cameroon.
Walter, S and Trammell, K. 2009. ‘The Kom MLE report for 2009’, Unpublished research
report. Available from the authors or SIL Cameroon.
Walter, S. and Trammell, K. 2008. ‘Results of the First Year of the Kom Experimental
Education Project’, Unpublished research report. Available from the authors or SIL
Cameroon.
Related Documents
