Tải bản đầy đủ


(đánh giá độ tin cậy của bài thi trắc nghiệm THứ NHấT TRÊN MáY
TíNH cuối kỳ 4 dành cho sinh viên năm thứ hai không chuyên
ngành tiếng anh trờng đại học kinh doanh và công nghệ hà nội)
Minor Programme Thesis
Field: Methodology
Code: 601410
HANOI, 2008

(đánh giá độ tin cậy của bài thi trắc nghiệm THứ NHấT TrÊN MáY
TíNH cuối kỳ 4 dành cho sinh viên năm thứ hai không chuyên ngành
tiếng anh trờng đại học kinh doanh và công nghệ hà nội)
Minor Programme Thesis
Field: Methodology
Code: 601410
Supervisor: Nguyễn Thu Hiền. M.A
HANOI, 2008
I hereby state that I: Nguyen Thi Viet Ha, Class 14A, being a candidate for the degree of
Master of Arts (TEFL) accept the requirements of the College relating to the retention and
use of Master of Arts Thesis deposited in the library.
In terms of these conditions, I agree that the origin of my thesis deposited in the library
should be accessible for the purposes of study and research, in accordance with the normal
conditions established by the librarian for the care, loan or reproduction of the thesis.
In the completion of this thesis, I have received a great deal of backup. Of primary
importance has been the role of my supervisor, Ms. Nguyen Thu Hien, M.A, Teacher of
Department of English and American Languages & Cultures, College of Foreign
Language, Vietnam National University, Hanoi. I am deeply grateful to her for her
precious guidance, enthusiastic encouragement and invaluable critical feedback. Without
her dedicated support and correction, this thesis could not have been completed.
I am deeply indebted to my dear teacher, Mr. Vu Van Phuc, M.A, Head of Testing Center,
College of Foreign Languages, VNU, who provided me with a lot of useful suggestion and
assistance towards my study.
I would also like to express my sincere thanks to all teachers and colleagues in English
Department, HUBT, for their help in conducting the survey, sharing opinions and making
suggestions to the study. Especially, my thanks go to Ms. Le Thi Kieu Oanh, Assistant of

English Department, HUBT for her willingness to offer test score data.
I wish to show my special thanks to the students of K11 at Hanoi University of Business
and Technology who have actively participated in the survey..
Finally, it is my great pleasure to acknowledge my gratitude to beloved members of my
family, especially my husband who constantly encouraged and helped me with my thesis.
The main aim of this minor thesis is to evaluate the reliability of the final Achievement
Computer-based MCQs Test 1 for the 4th semester non-English majors at Hanoi
University of Business and Technology.
In order to achieve this aim, a combination of both qualitative and quantitative research
methods were adopted. The findings indicate that there is a certain degree of unreliability
in the final achievement computer-based MCQs test1 and there are two main factors that
cause the unreliability including test item quality and test- takers performance. ’
Having carefully considered a thorough analysis of the collected data, the author made
some suggestions in order to improve the quality of the final achievement test and the
MCQs test 1 for the non-majors of English in the 4
semester in Hanoi University of
Business and Technology. Firstly, the test objectives, sections and skill weight should be
adjusted to be more compatible with the course objectives and the syllabus. Secondly, a
testing committee should be set up for the construction and development of a multi choice
item bank including test items which are of good p-value and discrimination value.
1. CBT: Computer-based testing
2. HUBT: Hanoi University of Business and Technology
3. MC: Multi choice
4. MCQs: Multi choice questions
5. ML Pre- : Market Leader Pre-intermediate
6. KD: Kuder- Richardson
7. SD: Standard deviation
1. Table 1 Types of tests
2. Table 2 Scoring format for each semester
3. Table 3 The syllabus for 4
semester (for non English majors– )
4. Table 4 Time allocation for language skills and sections
5. Table 5 Specification grid for the final computer-based MCQs test 1
6. Table 6 Main points in the grammar section
7. Table 7 Main points in the vocabulary section
8. Table 8 Topics in reading section
9. Table 9 Items in the functional language sections
10. Table 10: Test reliability coefficient
10. Table 11: p-value of items in 4 sections
11. Table 12: Discrimination value of items in 4 sections
12. Table 13: Number of test items with acceptable p-value and discrimination
value in 4 sections
13. Table 14: Suggested scoring format
14. Table 15: Proposed test specifications
12. Chart 1 Students response on test content’
13. Chart 2 Students response on item discrimination value’
14. Chart 3 Students response on time length ’
15. Chart 4 Students response arbitrariness ’
16. Chart 5 Students response on relation between test score and their ’
1.1. Rationale for the study
1.2. Aims and research questions
1.3. Theoretical and practical significance of the study
1.4. Scope of the study
1.5. Method of the study
1.6. Organization of the paper
2.1. Language testing
2.1.1. What is a language test?
2.1.2. The purposes of language tests
2.1.3. Types of language tests
2.1.4. Criteria of a good language test
2.2. Achievement test
2.2.1. Definition
2.2.2. Types of achievement test
2.2.3. Considerations in final achievement test construction
2.3. MCQs test
2.3.1. Definition
2.3.2. Benefits of MCQs test
2.3.3. Limitations of MCQs test
2.3.4. Principles on designing a good MCQs test
2.4. Reliability of a test
2.4.1. Definition
2.4.2. Methods for test reliability estimate
2.4.3. Measures to improve test reliability
2.5. Summary
Chapter 3: The Context of the Study
3.1. The current English learning, teaching and testing situation at HUBT
3.2. The course objectives, syllabus and materials used for the second non-
majors of English in Semester 4.
3.2.1. The course objectives
3.2.2. Business English syllabus
3.2.3. The course book
3.2.4. Specification grid for the final achievement Computer-based MCQs test
in Semester 4.
Chapter 4: Methodology
4.1. Participants
4.2. Data collection instruments
4.3. Data collection procedure
4.4. Data analysis procedure
5.1. The compatibility of the objectives, content and skill weight format of
the final achievement computer-based MCQ test 1 for 4
semester with
the course objectives and the syllabus
5.1.1 The test objectives and the course objectives
5.1.2. The test item content in four sections and the syllabus content
5.1.3. The skill weight format in the test and the syllabus
5.2. The reliability of the final achievement test
5.2.1. Reliability coefficient
5.2.2. Item difficulty and discrimination value
5.3. The attitude of students towards the MCQs test 1
5.4. Pedagogical implications and suggestions on improvements of the
existing final achievement computer-based MCQs test 1 for the non-
English majors at HUBT.
5.5. Summary
6.1. Summary of the findings
6.2. Limitations of the study
6.3. Suggestions for further study
Grammar, Reading, Vocabulary and Functional language check list
Survey questionnaire (for students at HUBT)
Students test scores’
Item analysis of the final achievement computer-based MCQs test 1- 150
items, 349 examinees
Item indices of the final achievement computer-based MCQs test 1
Chapter 1: Introduction
1.1. Rationale of the study
Testing plays a very important role in teaching and learning process. Testing is one
form of measurement which is used to point out strengths and weaknesses in the learned
abilities of the students. Through testing, especially tests scores we may discover the
performance of given students and of teachers. As far as students are concerned, test scores
reveal what they have achieved after a learning period. As for teachers, test scores indicate
what they have taught to their students. Based on test results, we may make improvement in
teaching, learning and testing for better instructional effectiveness.
Another reason for the selection of testing a matter of study lies in the fact that the
current language testing at Hanoi University of Business and Technology (HUBT) has
been under a lot of controversy among students and teachers. Testing is mainly carried out
in the form of two objective tests on computers (named test 1 and test 2) which are
administered at the end of each semester. The scores that a student gets on these tests are
the main indicators of his or her performance during the whole semester. There are
different comments on the results of these tests, especially the test 1 for the second-year
non-English majors. Some subject teachers claim that these tests do not truly reflect the
students’ language competence. Others say that these tests are appropriate to what students
have learnt in class and compatible with the course objectives and therefore reliable. Also,
among the students, do opposite ideas exist. Many think that these tests are more difficult
than what they have learnt and studied for the exam, others say that these test items are
easy and relevant to what they have been taught. Therefore finding out whether the tests are
closely related with what the students have been learnt and what the teachers have taught,
also, whether these tests are of reliability is indispensable.
For the two reasons mentioned above, the author would like to undertake this study
entitled A study on the reliability of the final achievement Computer-based MCQs Test“
1 for the 4th semester non-English majors at Hanoi University of Business and
Technology” with the intention to examine rumors about this test. In addition, the author
hopes that the study results help to raise awareness among teachers as well as those who
are interested in this field. At the same time, study results, in some extent, can be applied to
improve the current testing situation in HUBT.
1.2. Aims and research questions
The main aim of the study is to investigate the reliability of the existing final
achievement MCQs test 1 (4
semester) for non-English majors at HUBT through
analyzing the test objectives, test content and test skill weight format, students’ scores, test
items, perception and comments from students on the test and then to make suggestions
towards the test’s improvement.
To achieve this aim, the following research questions are set for exploration:
1. Are the objectives, content and skill weight format of the final achievement
computer-based MCQs test 1 compatible with the course objectives, the
syllabus content and skill weight format ?
2. To what extend is the test 1 reliable?
3. What is the student’s attitude towards the final achievement Computer-based
MCQs test 1?
1.3. Scope of the study
The existing final achievement Computer-based MCQs test 1 in the 4
semester for
the second-year non-English majors at HUBT
1.4. Theoretical and practical significance of the study
Theoretically, the study proves that testing is crucial in order to measure and
evaluate the quality of learning and teaching. Also, test reliability is one of the most
important criteria for the evaluation of a test.
Practically, the study presents how reliable the final achievement MCQs test 1
administered at HUBT is and how to improve its quality.
1.5. Method of the study :
Both qualitative and quantitative methods are used.
Regarding literature review on language testing, course objectives, syllabus, the
objectives, content and format of the achievement test 1 for 4
term, results of the
questionnaires for students, qualitative method is applied.
With reference to test scores and test items analysis, quantitative method is used.
1.6. Organization of the paper
The study is composed of 6 chapters.
Chapter 1- Introduction briefly states the rationale, aims and research questions,
scope of the study, theoretical and practical significance of the study, method of the study
and organization of the paper.
Chapter 2- Literature review discusses relevant theories of language testing, final
achievement test, Computer-based MCQ tests and test reliability.
Chapter 3- The context of the study deals with English learning, teaching and testing
situation at HUBT, course book, syllabus and check list for the test.
Chapter 4- Methodology presents participants, data collection instruments, data
collection and data analysis procedure.
Chapter 5– Results and Discussions presents and discusses the results of the study.
Suggestions for the improvement of the achievement test 1 are also proposed in this
Chapter 6- Conclusion summarizes the findings, mentions the limitations and
provides suggestions for further study.
Chapter 2: Literature review
2.1. Language testing
2.1.1. What is a language test?
There are a wide variety of definitions of a language test which have one point of
similarity. That is to say, a language test is considered as a device for measuring
individuals’ language ability.
According to Henning (1987, p.1), “Testing, including all form of language test, is
one form of measurement”. In his opinion, tests such as listening or reading
comprehension are delivered in order to find out the extent to what the abilities of these
skills are present in the learners. Similarly, Bachman (1990, p.20) stated: “A test is a
measurement instrument designed to elicit a specific sample of an individual’s
behavior”. He also considered obtaining the elicited sample of behavior as the
distinction of a test from other types of measurement.
Brown H.D (1995, p.384) presented the notion in a simpler way: “A test, in plain
words, is a method of measuring a person’s ability or knowledge in a given domain”.
He explained that a test first and foremost is a method which includes items and
techniques requiring the performance of testees. Via this performance, a person’s
ability or language competence is measured.
These viewpoints show that a language test is an effective tool of measuring and
assessing students’ language knowledge and skills and providing precious information
for better future teaching and learning.
2.1.2. The purposes of language tests
Language tests regarding their purposes are perceived from different perspectives
by different scholars. Typically, Henton (1990) mentioned 7 points which can be
represented as follows:
• Finding out about progress
• Encouraging students
• Finding out about learning difficulties
• Finding out about achievement
• Placing students
• Selecting student
• Finding out about proficiency
In general, a language test is used to evaluate both teaches and students’
performance, to make judgment and adjustment to teaching materials and methods, and
to strengthen students’ motivation for their further study.
2.1.3. Types of language tests
Language tests can be classified into different types according to their purposes.
Henton (1990), Brown (1995), Harrison (1983) and Hughes (1989) pointed out that
language tests include four main types: proficiency tests, diagnostic tests, placement
tests and achievement tests with characteristics illustrated in the following table:
Type of test Characteristics
Proficiency test Measure people’s abilities in a language regardless of any
training they may have had in that language
Diagnostic test Check students’ progress for their strengths and weaknesses
and what further teaching is necessary
Achievement test Assess what students have learnt as known syllabus
Placement test Classify students into groups at different level at the beginning
of a course
Table 1: Types of tests
Another researcher, Henning (1987) divided tests into objective and subjective ones
on the basic of the manner in which they are scored. Subjective tests obtain scoring by
opinionated-judgment on the part of the scorer while objective tests are scored by
comparing examinee responses with an established set of acceptable responses or
scoring key.
2.1.4. Criteria of a good language test
Just like any measuring device, a language test presents potential error
measurement. For the purpose of investigating and evaluating and “testing” a test,
researchers such as Brown (1995), Henning (1987), Bachman (1990) and Harrison
(1983) identified criteria to determine if a test is good or not. A good language test
must feature four most important qualities: reliability, validity, practicality and
The reliability of a test is its consistency (Brown, 1995; Harrison, 1983). A test is
reliable only when it yields the same results whether it is administrated under any
circumstances or scored by any markers. The validity of a test refers to “the degree to
which the test actually measures what it is intended to measure” (Brown, 1995, p.387).
A test is considered to be valid if it possesses content validity, face validity and
construct validity. The practicality of a test is administrative. A test is practical when it
is time and money- saving. Also, it is easy to administer, mark and interpret. The
discrimination of a test is the extent to which a test separates the students from each
other (Harrison, 1983). In other words, it is the capacity of the test to discriminate
among different students and to reflect individuals’ performance of the same group.
2.2. Achievement test
2.2.1. Definition
Achievement tests are of extensive use at different levels of education due to their
distinguished characteristics. Researchers define the notion of achievement tests in
various ways.
Henning (1987, p.6) held that:
Achievement tests are used to measure the extent of learning in a
prescribed content domain, often in accordance with explicitly stated
objectives of a learning program. .
From this definition, it followed that an achievement test was a measurement tool
designed to examine language competence of learners over a period of instruction
learning and to evaluate instruction program. In the same token, Hughes (1989) put that
achievement tests were intended to assess how successful individual students, groups of
students or the courses themselves have been in achieving objectives. Achievement
tests play an important role in the education programs, especially in evaluating
students’ acquired language knowledge and skills during a given course.
2.2.2. Types of achievement test
Achievement tests can be subdivided into the final achievement and progress
achievement according to the time of administration and the desired objectives
(Henton, 1990).
Final achievement tests are usually given at the end of the school year or at the end
of the course to measure how far students have achieved the teaching goals. The
contents of these tests must be related to the teaching content and objectives concerned.
Progress achievement tests are usually administrated during the course to measure
the progress that students are making. The results from the test enables teachers to
identify the weaknesses of the learners, diagnose the areas not properly obtained by
students during the course in order to have remedial action.
Henton (1990) also stated the two types of test differ in the sense that final
achievement tests are designed to cover a longer period of learning and it should
attempt to cover as much of syllabus as possible.
2.2.3. Considerations in final achievement test construction
On the basis of its characteristics, Heaton (1990) put that covering much the
contents of a syllabus or a course book is a requirement for designing a final
achievement test. Testers should avoid basing the test on their own teaching rather than
on the syllabus or course book in order to establish and maintain a certain standard. In
addition, Mc Namara (2000) stated that test writers should draw out a test specification
before writing a test. Test specification is resulted from the process of designing test
content and test method. Test specification has to include information on the length, the
structure of each part of the test, the type of materials with which the candidates will
have to engage, the source of materials, the extent to which authentic materials may be
altered, the response format and how responses are scored. They are usually written
before the tests and then the test is written on the basis of the specifications. After the
test is written, the specification should be consulted again to see whether the test
matches the objectives set in the specifications.
2.3. MCQs test
2.3.1. Definition
Multi-choice questions tests (MCQs tests) are objective tests which require no
particular knowledge or training in the examined content area on the part of the scorer
(Henning, 1990). They are different from subjective tests in terms of scoring methods.
That means no matter which examiners mark the test, a testee will get the same score
on the test (Heaton, 1988).
MCQs tests use multi-choice questions which is also called multi-choice items as a
testing technique. An MC item is a test item where the test taker is required to choose
the only correct answer from a number of given options (McNamara, 2000; Weir,
1990). .
In the view of Heaton (1988), MC items take many forms but their basic structure
includes two parts. The initial part is known as stem. The primary purpose of the stem is
to present the problem clearly and concisely. The stem needs to provide the testees a
very general idea of the problem and the answer required. The stem may be in the form
of an incomplete statement, a complete statement or a question. The other part is the
choices from which the students select their answers and is referred as options/
responses or alternatives. In an MC item there may be three, four or five options of
which one is the correct options or key while the others are distractors of which the task
is to distract the majority of poor students from the correct option. The optimum
number of options in most public test for each multi choice item is five. And it is
desirable to use four options for grammar items and five for vocabulary and reading.
2.3.2. Benefits of MCQs test
MC items are undoubtedly one of the most widely used types of items in objective
test (Heaton, 1988). The popularity of this testing technique results from its efficiency.
Researchers such as Weir (1990), Heaton (1988) and Hughes (1989) pointed out a
number of benefits which are presented as detailed below.
Firstly, the scoring of MCQs test is perfectly reliable, rapid and economical. There
is only one correct in the format of an MC item so that the scorers’ interference into the
test is minimized. The scorers are not permitted to impose their personal expertise,
experience, attitudes and judgment when giving marks to testees’ responses. The
testees, thus, always get a consistent result whoever the scorers are and whenever their
tests are given marks. In addition, MCQs tests can be marked mechanically with
minimal human intervention. As a result, the marking is not only reliable and simple
but also more rapid and often more cost effective than other forms of written test (Weir,
Secondly, an MCQs test can cover a much wider sample of knowledge than a
subjective test. When taking an MCQs test, a candidate has only to make a mark on the
paper and therefore it is possible for testers to add more items in a given period of time
(Hughes, 1988). With a large number of items in the test, the coverage of knowledge in
MC items is so broad and is very useful for identifying students’ strengths and
weaknesses and distinguishing their ability.
Thirdly, MCQs tests increase test reliability. According to Heaton (1988) and Weir
(1990), it will not be difficult to obtain reliability for MCQs tests because of perfectly
objective scoring. Besides, due to the fact that the testees do not have to deploy the skill
of writing as in open-ended one and MC items have clear and unequivocal format, the
extent to which measurement errors exert on the trait being assessed is narrowed.
Another benefit is that MC items can be trialed beforehand fairly easily. From these
trials, the difficulty level of each item and that of the test as a whole are usually
possible to be estimated in advance (Weir, 1990). The results from item difficulty
estimate make a great contribution to the success of designing a more appropriate test
to candidates’ level of language.
In addition, Heaton (1988, p.27) claimed that “multi choice items can provide a
useful means of teaching and testing in various learning situation (particularly at the
lower levels) provided that it is always recognized such items test knowledge of
grammar, vocabulary, etc. rather than the ability to use language”. MC items can be
very useful in measuring students’ ability to recognize correct grammatical forms, etc
and therefore can help both teacher and students to identify areas of difficulty.
As far as computer-based MCQs tests are concerned, according to McNamara
(2000) many important national and international language tests, including TOEFL, are
moving to computer-based testing (CBT) since there have been rapid developments in
computer technology. The main feature of CBT is that stimulus texts and prompts are
presented not in examination booklets but on the screen, with candidates being required
to key in their responses. The advent of CBT has not necessarily involved any change in
the test content but often simply represents a change in test method. McNamara (2000)
noted that the proponents of computer-based testing can point to a number of
advantages. First, just as paper-done MCQs tests, scoring of fixed response items can be
done automatically and the candidate can be given a score immediately. Second, the
computer can deliver tests that are tailored to the particular abilities of the candidate.
This type of test, as also called computer-adaptive test, can provide far more information
about the testees’ ability.
2.3.3. Limitations of MCQs tests:
Despite the fact that MCQs tests bring lots of benefits, especially, to test
administrators, there are several problems associated with the use of MC items. These
problems were identified by a number of researchers such as Weir (1990), Hughes
(1989), Heaton (1988), McCOUBRIE. P (2004) and McNamara (2000).
First of all, Hughes (1989) criticized that MCQ technique tests only recognition
knowledge. To do a given task, a testee just needs to look at the stem and four or five
options and then picks out the key. His or her performance is not much more than the
recognition of the right form of language. It shows no evidence that this person can
produce the language. Obviously, this type of test presents a lack between at least some
candidates’ productive and receptive skill and therefore the performance on an MCQs
test may give an inaccurate picture of these candidates’ ability (Hughes, 1989). Heaton
(1988) also pointed out that an MC item does not lend itself to the testing language as
communication and the process involved in the actual selection of one out of four or five
options does not bear much relation to the language used in most real life situation.
Normally, in everyday situation we are required to produce and receive language while
MC items are merely aimed to test receptive skills.
Another problem arises when using MCQs tests is that “the multi choice item is one
of the most difficult and time consuming types of items to construct” (Heaton, 1988,
p.27). In order to write a good item, test designers have to strictly follow certain
principles. For example, they have to write many more items than they actually need for
a test. After that they have to pre-test and analyze students’ performance on the item
evaluate items and recognize the usable ones or even to rewrite the items for a
satisfactory final version. These procedures take a lot of test constructors’ time and need
far more careful preparation than subjective tests.
Furthermore, objective tests of MCQs type encourage guessing (Weir, 1990;
Heaton, 1988; Hughes, 1989). Hughes estimated the chance of guessing the correct
answer in a three option multi choice item is roughly 33%; in four or five option item it
is 25% or 20% respectively. The format of MC items makes it possible for testees to
complete some items without reference to the texts they are set on. As a result, the score
gained in MCQs maybe suspect and the score range may become narrow.
Some other limitations in the use of MC items involve backwash and cheating.
Backwash may be harmful because MQ items require students to memorize as many
structures and forms as possible and do not stimulate them to produce language. Thus
practicing MC items is not a good way to improve learners’ command of language.
Cheating may be facilitated as MC items make students easy to communicate with each
other and exchange selected response nonverbally.
Referring to computer-based tests, according to McNamara (2000), this type of test
requires the prior creation of item bank which have been thoroughly trialed. The
preparation for a standardized item bank to estimate difficulty for candidates at given
levels of ability as precisely as possible is not at ease. In addition, delivering CBT raises
the question of validity and reliability. For example, different levels of familiarity with
computers or of reading texts on computer screens will affect students’ performance.
These differences might lead to difficult conclusion about a candidate’s ability.
2.3.4. Principles to construct MC items
In order to construct a good MC item, there are a large number of principles which
can be summarized as follows (Heaton, 1988):
• Each MC item should have only one answer
• Only one feature at a time should be tested
• Each option should be grammatical correct when placed in the stem, except for the
case of specific grammar test items.
• All multi-choice items should be at a level appropriate to the proficiency level of
the testees.
• Multi choice items should be as brief and as clear as possible
• Multi choice items should be arranged in rough order of increasing difficulty and
there should be one or two simple items to “lead in” the testees.
2.4. Reliability of a test
2.4.1. Definition
In research, the term reliability means ‘repeatability’ or ‘consistency’. A test is
considered reliable if it would give us the same result over and over again assuming that
what we are measuring isn't changing. Lynch (2003, p.83) stated that reliability refers to
“the consistency of our measurement”. In the same vein, Harrison (1983) explained that to
be reliable, tests should not be elastic in their measurement. Whatever the version of the
test a testee take, whatever the occasion the test is administrated, and whatever raters who
score the test, it still yields the same results.
2.4.2. Methods of test reliability estimate
Reliability may be estimated through a variety of methods which is presented below:
* Test-retest method is a classic way to calculate the reliability coefficient of a test. The
test is given to a group of students and then given again to these students immediately
afterward (the interval between two test administration is no more than two weeks). The
test is assumed to be perfectly reliable if the students get the same score on the first and the
second administration (Alderson, J.S. et al., 1995)
* Parallel-form methods involve correlating the scores from two or more similar (parallel)
tests which are administrated to the same sample of persons. A formula for this method
may be expressed as follows:
Rtt = rA,B (Henning, 1987)
Rtt: the reliability coefficient
rA,B: the correlation of form A with form B of the test when administered to the same
people at the same time.
* Inter-rater method is applied when scores on the test are independent estimates by two
or more raters. It involves the correlation of the ratings of one rater with those of another.
The following formula is used in calculating reliability:
Rtt = (Henning, 1987)
1 + (n-1)r A,B
Rtt: inter-rater reliability
n: the number of rater who combines estimates from the final mark for the examiner
rA,B: the correlation between the raters, or the average correlation among all rater if
there are more than two
* Internal consistency method judges the reliability of the test by estimating how
consistent test-takers’ performances on different parts of the tests with each other
(Bachman, 1990). The following are internal consistency measures that can be used:
Split-half reliability involves dividing a test into two, and correlating these two
halves. The more strongly the two halves correlate, the higher the reliability will be. This
method uses the following formula:
Rtt = (Henning, 1987)
1 + r A,B
Rtt: Reliability estimated by the split half method
rA.B: The correlation of the score from one half of the test with those from the
other half
Kuder-Richardson Formula 20 (KD20) is based on item level data and is used
when the tester has the results for each test item. The KD-20 is as follows:
n s
- ∑s
Rtt = ( ) (Henning, 1987)
(n-1) s
Rtt: The KR 20 reliability estimate
n: The number of items in the test
: The variance of test scores
: The sum of the variances of all items (or ∑pq)
Kuder- Richardson Formula 21 (KD-21) is based on total test scores and assumes that all
items of an equal level of difficulty. The KD-21 is as follows:
n x – x
Rtt = ( 1 - ) (Henning, 1987)
( n-1) s
Rtt : The KR 20 reliability estimate
n: The number of items in the test
x: The mean of scores on the test
: The variances of test scores
Alderson, J.S. et al (1995) stated that for the internal consistency reliability, the
perfect reliability index is +1.0. In the same view, Hughes (1989, p.31-32) noted that “ the
ideal reliability coefficient is 1- a test with a reliability coefficient of 1 is one which would
give precisely the same results for a particular sets candidates regardless of when it
happened to be administrated”. Reliability coefficient for a good vocabulary, structure and
reading test is usually in the 0.90 to 0.99 range, for an auditory comprehension test is more
often in the 0.80 to 0.89 range and for an oral production test it may be in the 0.70 to 0.79
range while an MCQs test typically has the reliability coefficient of more than 0.80
(Hughes, 1989).
Among the above ways of estimating reliability, test-retest and parallel methods
require at least two test administrations while the inter-rater and internal consistency
methods need only a single administration. For the reason of convenience and satisfaction,
KD20 and KD 21 are often chosen more than the others and are considered the two most
common formulae (Alderson. J.S. et. al., 1995).
Concerning MCQs tests, besides estimating test reliability coefficient, item analysis
including item difficulty and item discrimination provides more concise insight into the test
reliability (Henning, 1997).
The formula for calculating item difficulty is:
p = (Henning, 1987)
p: proportion correct
∑Cr : the sum of correct responses
N: the number of students
Henning (1987) pointed out that p value for each item should be between 0.33 and
0.67 and thus the level of difficulty of the item is acceptable. If p value is below 0.33, the
item is considered as too difficult. If it is above 0.67, the item is too easy.
The formula for computation of item discrimination is:
D = (Henning, 1987)
Hc + Lc
D: discriminability
Hc: the number of correct response in the high group
Lc: the number of correct response in the low group
The optimal size of each group is 28% of the total sample. For very large samples of
examinees, the number of examinees in the high and low groups are reduced to 20% for
computational convenience. The acceptable discrimination value by sample separation
method is >= 0.67 (Henning, 1987)
2.4.3. Measures to improve test reliability
Reliability may be improved by eliminating its sources of error. Hughes (1989)
makes a list of recommendation to improve test reliability as follows:
• Take enough sample of behavior
• Do not allow candidates too much freedom
• Write unambiguous items
• Provide clear and explicit instructions
• Ensure that the test are well laid out and perfectly legible
• Candidate should be familiar with format and testing techniques
• Provide uniform and non-distracting conditions of administration
Furthermore, item difficulty and item discriminability show that the reliability of an
MCQs test. is low or high (Henning, 1987). Therefore the most straight forward ways to
improve test reliability is to design MCQs items with good level of difficulty and
discrimination value.
2.5. Summary
This chapter presents the theoretical framework for the study. In Section 2.1, the
notion of a language test as a measuring device of people’s ability is reviewed.
Additionally, the purposes of language testing, types of language tests and criteria of a
good test are also discussed. Section 2.2 classifies achievement tests into two types and
mentions consideration in designing final achievement tests. The definition, benefits and
limitations of MCQs tests and principles of this type of test construction are dealt with in
section 2.3. The final Section - 2.4 is concerned with test reliability, methods for estimating
test reliability, and ways to make language tests more reliable.
Chapter 3: The Context of the Study
3.1. The current English learning, teaching and testing situation at HUBT
There are over 1500 second-year non-majors of English at HUBT. English is their
required subject for foreign language. Their levels of proficiency vary because of their
different backgrounds, knowledge of language, exposure to English, characteristics,
learning attitudes, motivations and so on. These students have to cover a comparatively
large amount of knowledge of English as English hold the highest credits among all
subjects. In the English Department, HUBT there are totally 62 teachers who work with the
non-English majors enthusiastically to help them with the foreign language. They are all
dedicated and qualified with an average of five years’ teaching experience.
With the aim to equip students with business English and communication skills
necessary for their future career, learning and teaching activity for the second-year non-
English majors mainly focus on developing speaking and listening skills. However, testing
process is quite complicated and can be described as follows.
In semester 4 the students have to experience daily assessment and go through four
tests all together. Daily assessment includes checking vocabulary, speaking skill, and doing
tasks in the course book and practice files. The four tests comprise of two paper tests and
two computer-based MCQs tests. These tests are designed by teachers of English
Department, HUBT and . The paper tests, given in the middle of the term (week 9) and at
the end of the term (week 17) focus on listening, writing, grammar and vocabulary. The
computer-based MCQs tests are administered on computers in the week 19. Each test lasts
2 hours and includes 150 multi choice items emphasizing on vocabulary, grammar,
reading and functional language. The construction of the first test (hereafter achievement
test 1) is based on the three units of the course book (Unit 7, 8, 9) that the students have
already learnt. The second one (achievement test 2) is designed on the basis of the last
three units of the course book (Unit 10, 11, 12). Items of MCQs tests are selected by one
person in charged of teaching English in the 3
and 4
semester for the second year
The Computer-based MCQs test administered in HUBT is similar to a paper-done
one. The main different is that the test is delivered on computers and students simply click
mouse for their chosen response among A, B, C, D. This kind of test is different from
computer adaptive tests which are tailored to the particular abilities of the candidate. In

Tài liệu bạn tìm kiếm đã sẵn sàng tải về

Tải bản đầy đủ ngay