There is no doubt that testing is an essential part of language teaching and learning. A
language test in general can be a “ sample language behavior and infer general ability in
the language learnt.”(Brown D.H, 1994:252). In other words, from the results of the test,
depending on different kinds of tests with different purposes as well, the teacher infers a
certain level of language competence of his students in such different areas as grammar,
vocabulary, pronunciation, or speaking, listening, writing and reading.
It is obvious that the teacher plays a very important role in the process of assessment and
measurement which is conducted through testing. It is said that “ language testing is a form
of measurement. It is so closely related to teaching that we cannot work in testing without
being constantly concerned with teaching.”(Heaton, 1988:5).
There are various types of test which serve different purposes in foreign language teaching
and learning. Among the kinds of tests and testing, writing tests are said to be less reliable
from the point of both scorer and testee. This situation can be seen clearly at Nghe An
Junior Teacher Training College. For many years, English writing has been considered the
most difficult skill to be tested among teachers. Teachers have found it difficult to mark the
achievement writing tests accurately, in particular mark compositions, as they blame that
there is no rating scale for scoring compositions, or the provided rating scale is too general.
Apart from this, many students are still worried about the results of the writing
achievement tests, especially the task of writing a composition as they wonder if their
writings are accurately evaluated by raters.
That is the reason for choosing the topic of the research: A study on the reliability of the
achievement writing test for the second year English major students at N.A.JTTC. It is
hoped that the study will be helpful to the author, the teachers at the English department of
N.A JTTC and to those who are concerned with language testing in general and the study
of the reliability of writing achievement tests in particular.
2. AIMS OF THE STUDY
The major aims of this study are:
to explore the relevant notions of language testing
to analyze the achievement writing test for the second year English major
students on the basis of the syllabus, purposes of teaching and testing; and
available data such as test scores and scores of sample compositions for
evidences on its validity and reliability with a focus on reliability.
to provide some suggestions for test- designers as well as raters.
3. SCOPE OF THE STUDY
Evaluating an achievement writing test consists of complex procedures and needs a
number of criteria to be set up. However, due to the availability of data and limitation of
time, this study focuses mainly on the reliability of the achievement writing test for the
second year English major students at N.A JTTC. The results can be seen as the basis for
providing some suggestions for test designers as well as raters.
4. METHODS OF THE STUDY
On the basis of analyzing the teaching aims, and syllabus for the second-year English
major students as well as the content of the writing test (term 1) as the practical base for
the study, the quantitative method, which focuses on analyzing the test scores of 156
second year students and the scores of 15 sample compositions collected randomly, is used
to measure the reliability of the test.
5. DESIGN OF THE STUDY
The study is comprised of three parts:
Part I: Introduction
provides information on the rationale for choosing the topic, the aims, the scope, and the
methods of the study.
Part II: Development ( divided into 3 chapters)
*Chapter one: reviews the literature related to language testing (definitions, approaches,
roles, purposes and relationships between teaching-learning- testing), testing writing
(types of writing, criteria of testing writing and problems in testing writing) and the criteria
of a good test with a focus on reliability in which factors affecting language test scores,
methods to determine the reliability of a language test and measures for improving test
reliability are mentioned.
* Chapter two: presents an overview of the teaching, learning and testing situations at
Nghe An Teacher Training college including a description of the second year English
teaching aims, the writing syllabus /course book as well as the content of the writing
* Chapter three: presents the methodology of the analysis of the format of the writing
achievement test, of the data collected from the test scores of 156 second year students and
the scores of 15 random sample compositions in order to find out “to what extent is the
marking of the writing achievement test reliable?
Part III: Conclusion
presents a summary and some recommendations/suggestions for test- designers and raters.
CHAPTER 1: LITERATURE REVIEW
In this chapter, the theoretical background for the study is established. Firstly, the term’
language testing’ including the approaches, roles, purposes as well as the relationships
between language testing, teaching and learning will be explored. Then testing of writing
will be discussed and followed by an examination of the criteria of a good language test
with a focus on the reliability.
1.1 Language testing
1.1.1 Definitions of language testing
Testing is an important part of every teaching and learning experience and becomes one of
the main aspects of methodology. Many researchers have given out definitions of testing
with different points of view.
Allen (1974:313) emphasizes testing as an instrument to ensure that students have a sense
of competition rather than to know how good their performance is and in which condition a
test can take place. He says ‘test is a measuring device which we use when we want to
compare an individual with other individuals who belong to the same group.’.
According to Carroll (1968:46), a psychological or educational test is a procedure designed
to elicit certain behavior from which one can make inferences about certain characteristics
of an individual. In other words, a test is a measurement instrument designed to elicit a
particular behavior of each individual.
According to Bachman (1990:20), what distinguishes a test from other types of
measurement is that it is designed to obtain specific sample of behavior. This distinction is
believed to be of great importance because it reflects the primary justification for the use
of language tests and has implications for how we design, develop and use them to their
best use. Thus, language tests can provide the means for more focus on the specific assure
Besides, Ibe (1981:1) points out that “a sample of behavior under the control of specified
conditions aims towards providing a basis for performing judgment.”. The term a sample
used here is rather broad and it means something else rather than the
traditional types of paper and pencils. Read (1983) shares the same idea with Ibe in the
sense that a sample of behavior suggests language testing certainly includes listening and
speaking skills as well as reading and writing ones.
However, Heaton (1988:5) looks at testing in a different way. In his opinion, tests are
considered as a means of assessing the students’ performance and to motivate the students.
He looks at tests with positive eyes as many students are eager to take tests at the end of
the semester to know how much knowledge they have. One important thing is that he
points out the relationship between testing and teaching.
In short, from the above descriptions, testing is an effective means of measuring and
assessing students’ language knowledge and skills. It is of great use to both language
teaching and learning. In order to understand more about language testing, we should have
a look at the different approaches to language testing in the following part.
1.1.2. Approaches to language testing
According to Heaton (1988:15), there are four main approaches to testing: (i) the essay
translation approach, (ii) the structuralist approach, (iii) the integrative approach, and (iv)
the communicative approach.
(i) The essay approach is considered as an old method in which tests often focus on essay
writing, translation and grammatical analysis. This approach requires no special skills or
expertise in testing, but the subjective judgment of the teacher is considered to be of
(ii) The structuralist approach places the main focus on testing concrete skills without
relating them to the context. The skills of listening, speaking, reading and writing are
separated from one another as it is considered essential to test each one at a time. The
learners’ mastery of the separate elements of the target language (phonology, vocabulary
and grammar) are also tested using words and sentences completely out of contexts so that
a large number of samples of language forms can be covered in the test in a comparatively
short time. Thus, the results of the students’ tests completely depend on the accurate forms
of separate language aspects or skills tested rather than the total meaning of the discourse
or ability to use the language appropriately and effectively.
However, this approach is still valid for some types of tests for specific purposes as it is
considered to be objective, precise, reliable and scientific. That is the reason why the
typical type of test following this approach, multiple choice, is still widely used nowadays
even though there is still a limited use for multiple choice items in many communicative
(iii) The integrative approach, in contrast, involves the testing of language in contexts and
is thus concerned primarily with meaning and the total communicative effect of discourse
(Heaton, 1988). Integrative tests, instead of separating the language into different aspects,
are designed to test two or more skills at the same time (especially focusing on reading and
listening or language components in integration as grammar and vocabulary. In other
words, this type of tests is concerned with students’ global proficiency, not their mastery of
separate elements or skills. The typical types of tests following this approach are cloze
tests, dictation, oral interviews, translation and essay writing. However, according to
Heaton (1988:16), integrative testing involves ‘ functional language’ but not the use of
functional language, and thus they are weak in communication.
(iv) The communicative approach is considered to be interactive, purposive, authentic,
contextualized and should be assessed in terms of behavioral outcomes. Although both
integrative and communicative approaches emphasize the importance of the meaning of
utterances rather than their form and structure, communicative tests are concerned
primarily with how language is used in communication ( Heaton, 1988:19). The
communicative approach emphasizes the evaluation of language use rather than usage
(‘use’ is concerned with how people actually use language for different purposes while
‘usage’ concerns the formal patterns of language). However, the communicative approach
is claimed to be less reliable because of various real- life situations in different areas/
countries and according to Heaton (1988), in order to increase the reliability, especially in
scoring, a very carefully drawn -up and well- established criteria must be designed.
In brief, each approach to language testing has its weak points as well as strong points.
Therefore, in Heaton’s point of view, a useful test will generally incorporate features of
several of these approaches (Heaton, 1988:15).
1.1.3. The roles of language testing
According to Mc Namara (2003:4), language tests play a powerful role in many people’s
lives, acting as gateways at important transitional moments in education, in employment,
and in moving from one country to another. Language testing is used for the assessment,
employment, selection and considered by raters as a means of placing students on
particular courses. Moreover, language tests are also considered as the criterion to evaluate
the language proficiency of the researchers who want to carry out their research in
F. Bachman (1990:2) has also mentioned to the importance of language testing. Language
tests can be valuable sources of information about the effectiveness of learning and
(language teachers regularly use tests to help diagnose students’ strengths and
weaknesses, to assess students’ progress, and to assist in evaluating students’
achievements); language tests are also frequently used as sources of information in
evaluating the effectiveness of different approaches to language teaching, as sources of
feedback on learning and teaching; and thus language tests can provide useful input into
the process of language teaching. Conversely, insights gained from language learning and
teaching can provide valuable information for designing and developing more useful tests.
1.1.4. The purposes of language testing
Language tests usually differ according to their purposes. Hughes (1989:7) points out the
different purposes of testing based on the different kinds of tests such as:
- to measure language proficiency regardless of any language courses that
candidates may have followed.
- to discover how far students have achieved the objectives of a course of study.
- to diagnose students’ strengths and weaknesses, to identify what they know and
what they do not know.
- to assist placement of students by identifying the stage or part of a teaching
program most appropriate to their ability.
According to Henning (1987:1-2-3), there are six major purposes of language testing
which can be represented as follows:
(i) Diagnosis and Feedback ( in order to find out students’ strengths and weaknesses)
(ii) Screening and Selection ( in order to decide who should be allowed to participate
in a particular program of study)
(iii) Placement ( in order to identify a particular level of the student and place him or
her in a particular program of study)
(iv) Program Evaluation ( in order to provide the information about the effectiveness
of the programs)
Providing Research Criteria ( in order to provide a standard of judgment in a
variety of other research contexts)
(vi) Assessment of Attitudes and Sociopsychological Differences
1.1.5. Relationship between Language Testing and Language Teaching & Learning
Though a large number of examinations and tests in the past tended to separate testing
from teaching, Heaton (1988:5) emphasizes that teaching and testing in some ways are so
interwoven and independent that it is very difficult to tease apart “Both testing and
teaching are so closely interrelated that it is virtually impossible to work in either field
without being constantly concerned with the others”. Heaton (1988:5) also notes: “Tests
may be constructed primarily as devices to reinforce learning and motivate the students or
as a mean of assessing the students’ performance in the language”. In the former case,
testing is geared to the teaching, whereas in the latter case, teaching is often geared largely
to the testing.
According to Hughes (1989:1), the effect of testing on teaching and learning is known as
backwash which can be harmful or beneficial. He puts more focus on the harmful test. If
the test content does not go with the objectives of the course, then the backwash can be
really harmful and it leads to problems of teaching in one way but testing in another way.
In his view, some language tests have harmful effects on teaching and often fail to measure
accurately whatever they are intended to measure. That is proved in the case of a writing
test with only multiple choice items in which learners only concentrate on practicing such
items rather than practicing the skill of writing itself.
In general, testing and teaching -learning process have a very close relationship. Tests can
be effective for both teaching and learning process and vice versa. Therefore, good tests
are useful, desirable and can be used as a valuable teaching device. For educational
purposes, educators should improve both language tests and language teaching methods to
have beneficial backwash.
1.2. Testing writing
In this section, to understand what testing writing is, it is crucial to discuss the writing
types, criteria to test writing accurately and problems in testing writing.
1.2.1. Types of writing
Writing is defined as the productive skill in the written mode. It is complex and difficult to
teach and test as well since it requires the mastery not only of grammatical and rhetorical
devices but also of conceptual and judgmental elements in which the use of judgment is
considered to be the most important in teaching and testing of writing. Normally, after
clarifying the purpose of writing, the particular audience and register, an essay or a
composition may be written in accordance with the four main types: narrative, descriptive,
expressive, or argumentative. However, Heason (1988:136) divides the types of writing
according to learners’ levels. In his point of view, there are three main levels: basic,
intermediate and advanced. At the basic level, learners learn how to write letters,
postcards, diaries, and forms. With the already -learnt types in the basic level, learners
should be taught the ways to guide and set instructions at the intermediate level. And at the
advanced level, learners have to learn more how to write newspaper reports and notes.
1.2.2. Criteria of testing writing
There is a wide variety of writing tests which is needed to test many kinds of writing tasks
that we engage in. According to Madsen (1983:101), there are three main types of writing
tasks and tests as well: controlled, guided and free writing which correspond with their
testing criteria. He points out the advantages and limitations of each type in which scoring
in guided and free writings is considered to be subjective. Thus, McNamara (2000:38)
says in the past, writing skills were assessed indirectly through examinations of control
over the grammatical system and knowledge of vocabulary because of the problem of
subjectivity. However, nowadays educators place more emphasis on teaching and testing
learners’ communicative language abilities. That is the reason why there is a current trend
in shifting from testing isolated items to testing writing compositions in which the
managing of the rating process became an urgent necessity. John Boker indicates the
disadvantages of testing writing compositions such as limited amount of content sampled,
time- consuming of score, and low reliability in scoring. Therefore, in order to evaluate
learners’ ability in the writing skill accurately, raters have to be carefully trained in rating
learners’ writings, and pay much attention to some micro skills involving in the writing
process that learners need to acquire. Heason (1988:135) describes those micro skills as
* language use: the ability to write correct and appropriate sentences;
* mechanical skills: the ability to use correctly those conventions peculiar to the
written language (e.g. punctuation, spelling....);
* treatment of content: the ability to think creatively and develop thoughts, excluding all
* stylistic skills: the ability to manipulate sentences and paragraphs, and use language
* judgment skills: the ability to write in an appropriate manner for a particular purpose
with a particular audience in mind, together with an ability to select, organize and
order relevant information.
He also represents the minimum criteria of testing writing which learners need to obtain
based on the different levels. This can be shown as belows:
* Basic level: No confusing errors of grammar or vocabulary; a piece of writing legible
and readily intelligible; able to produce simple unsophisticated sentences.
* Intermediate level: Accurate grammar, vocabulary and spelling, though possibly
with some mistakes which do not destroy communication; handwriting generally
legible; expression clear and appropriate, using a fair range of language; able to
link themes and points coherently.
* Advanced level: Extremely high standards of grammar, vocabulary and spelling;
easily legible handwriting; no obvious limitations on range of language candidate is
able to use accurately and appropriately; ability to produce organized, coherent
writing, displaying considerable sophistication.
1.2.3. Problems in testing writing
Many problems may occur in testing writing; however, Hughes (1989:75) represents the
main considerations in testing writing which can be also seen as the main problems in
testing writing. They involve the setting of the writing / testing tasks that are properly
representative of the population of tasks, truly represent the students’ abilities, and can be
As being mentioned above about the subjectivity in scoring writings which possibly results
in the unreliability of the test - the major problem in testing writing, in this section, we will
discuss about the causes that lead to this problem. According to Mc Namara (2000:37), the
rating given to a candidate in testing writing is a reflection , not only of the quality of the
candidate’s performance, but of the qualities as a rater of the person who has judged it. In
fact, different raters may assign different scores on the same tests and even a grade that one
rater gives on the same test may vary from day to day. Thus, the subjectivity of the writing
test is mostly caused by different raters’ interpreting and applying the same rating scale in
different ways. Besides, raters may be affected by some random factors such as health,
climate, and so on. Or as Sako (1983:219) states about personal bias which tend to affect
“ The personal bias may stem from a variety of factors, such as the appearance of the
essay, poor spelling, paucity of ideas, or what has been called the ‘halo effect’, that is, a
higher or lower grade resulting from previous pleasant or unpleasant experiences with the
writer.” Moreover, the unreliability is possibly the result of the vague criteria set for the
evaluation of writing tests. Therefore, test designers find it difficult to write a good and
well- designed rating scale and this also cause the difficulties in the scoring process.
In short, in order to obtain the reliability in scoring writing tests, test designers and raters
should be careful in designing, interpreting and applying the rating scale.
1.3. Criteria of a good language test
Before making a test, test designers often ask themselves many questions such as: How do
we design a test that is a true indicator of students’ communicative ability? How do we
know that it is a good one? and so on. To write a good test, Harrison (1983:10) claims that
it is essential for test designers to considerate the four basic characteristics of all good tests
namely validity, reliability, practicality and discrimination. Among those, two of the
qualities- validity and reliability- are considered to be critical for tests and sometimes
referred to as essential measurement quality. This is because these are the qualities that
provide the major justification for using test scores ( numbers) as a basis for making
inferences or decisions (Bachman, 1996:19). In this section, we will discuss briefly what
those qualities are and how they must be considered.
Validity is the quality that most affects the value of a test and the most critical factor to be
judged in the total program of language testing. According to Harrison (1983:10), the
validity of a test is the extent to which the test measures what it is intended to measure.
Two questions must be considered when determining test validity of a language test: What
aspects of the language are designed to measure, and how well does it, in fact, measure
the global skills or the discrete elements of the language? (Sako, 1983:24). The validity of
a language test is determined by comparing the results it gives with some outside or
independent criterion. Sako also presents five primary validity concepts:
(i) content validity ( in order to ascertain how thoroughly a language test samples the
instructional objectives or’ the universe of criterion behavior’)
(ii) concurrent validity ( by correlating the language test scores with the criterion behavior
or language performance established )
(iii) predictive validity ( in order to determine how well the test scores predict language
(iv) construct validity ( by examining correspondence between the content of the test and
the various components of the skill being measured)
(v) face validity (a judgment about a test based on the way the test looks to educators,
students and the general public)
In short, validity is the “must” for testers to take into considerations when they construct a
It would be unwise if test designers are to consider tests’ validity and reliability apart from
its practicality. According to Harrison (1991:13), a valid and reliable test is of little use if it
does not prove to be a practical one. Test practicality refers to financial limitations, time
constrains, ease of administration, scoring and interpretation. A test is impractical in case it
is prohibitively expensive and it takes much time to construct.
In conclusion, a test has practicality if it does not involve much time or money in
constructing, implementing and scoring it.
It is said to be incomplete without considering the discrimination of a test. According to
Heaton (1988:165), an important feature of a test is its capacity to discriminate among the
different candidates and to reflect the differences in the performances of the individuals in
the group. It is true for both teacher- made tests and standardized tests. In order to have
this feature, a test must have a scale ranging from extremely easy items to extremely
difficult items. However, the extent of the need to discriminate varies depending on the
purpose of the test. The more efficiently a test discriminates among students, the easier it is
to divide them into suitable groups and the more clearly it shows the level of each
individual in the group.
Reliability is a necessary characteristic of any good test. According to Sako (1983:28), the
reliability of a language test is concerned with the degree to which it can be trusted to
produce the same result upon repeated administration to the same individual, or to give
consistent information about the value of a learning variable being measured. Sako has
represented several external factors which may affect the reliability of a test such as:
variation in testing conditions (lighting, temperature, distraction, or noise), differences in
administrative instructions, test compromise, inaccurate in scoring, inadequate sampling of
test items, lack of motivation (fatigue or illness). On the other hand, he has given some
factors that can improve the reliability of language tests: standardizing and optimizing the
testing conditions, using a uniform procedure in administering the test, increasing the
number of test items, providing an adequate sampling of test items, and reducing
subjective scoring of the test.
In short, according to Hughes (1989), there are two main sources of reliability: the
consistency of performance from candidates and scoring. The former is affected by several
factors such as the number of questions, test administration and test instructions. Possibly
the most important factor affecting the test reliability is the scoring method. In” A Guide to
Language Teaching”, Henning (1987:76) notes:’ Subjectivity in scoring or mechanical
errors in the scoring process may introduce the inconsistencies in scores and produce
unreliable measurement”. The more reliable the test result will be, the less subjective the
In general, those are the criteria for a good test that test- designers should consider when
designing tests. However, in this study we have placed more emphasis on the reliability
which is defined as a measure to evaluate the consistency of language test scores through
analyzing students’ test scores and compare
the results of students’ compositions.
Therefore, in this section, three issues about the reliability of a language test will be
discussed in more details: (i) factors affecting language test scores, (ii) methods to
determine the reliability of a language test, and (iii) measures for improving test reliability.
188.8.131.52. Factors affecting language test scores
As mentioned above, the language test scores are affected by not only
language knowledge and skills, but by other factors which has been concerned much and
will be discussed here. William Wiersma (1989:290) divides these factors into two major
categories: personal factors associated with individuals being tested
and the factors
associated with the test setting called test administration factors. According to Bachman
(1990:163), the factors which can affect test scores, external to the language knowledge
and skills of the students, can be grouped into the broad categories: (i) test method
facet,(ii) personal attributes, and(iii) random factors. They are represented as follows:
(i) Test method facets can be seen as the characteristics of the test method which affect the
test performance. Bachman (1990:118) represents five categories of test method facets: the
test environment, the test rubric, the nature of input the test taker receives, the nature of the
expected response to that input, and the relationship between input and response.
(ii)Personal attributes which are not related to language ability include individual
characteristics such as cognitive style and knowledge of particular content areas, and group
characteristics such as sex, race and ethnic background. These attributes are considered to
be systematic in the sense that they are likely to affect a given individual’s test
(iii) Random factors include unpredictive and largely temporary conditions such as his
mental alertness or emotional state, and uncontrolled differences in test method facets
(changes in the test environment from one day to the next, or idiosyncratic differences in
the way different test administrators carry out their responsibilities).
In a nutshell, there are numerous factors that can affect test scores, and those factors are
rather complicated. Therefore, it is necessary to come to the conclusion that the
administers should find the ways to control and minimize the effects of those factors to
obtain the high reliability of the test.
184.108.40.206. Methods to determine the reliability of a language test,
According to Hughes (1989:31), it is possible to quantify the reliability of tests in the form
of reliability coefficients that allow us to compare the reliability of different tests. He
points out 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 set of candidates
regardless of when it happened to be administered. According to Hughes (1989), Bachman
(1990), Heaton (1989), and Weir (2005:24), there are four methods (or types of scoring
validity) to evaluate the reliability of a language test: test- retest , alternative or
parallel/equivalent form, split half (internal consistency), rater consistency ( intra- rater,
inter- rater). They are described as belows:
(i) Test- retest method: according to Hughes (1989:32), in order to obtain test- retest
reliability, it is necessary to get a group of subjects to take the same test twice and then
make a comparison of two sets of scores. However, he points out the drawbacks of this
method:” If the second administration of the test is too soon after the first, the subjects are
likely to recall the items and responses. If there is too long a gap between administrations,
learning or forgetting will take place and the subjects’ motivations may be reduced.”
Although Henning (1987) makes a suggestion for the period of time between two
administrations” The second administration should be organized after no more than 2
weeks.” this type of reliability, according to Weir ( 2005:24), is seldom carried out;
therefore the parallel form reliability is the preferred choice.
(ii) Parallel form method: according to Bachman (1990:168), parallel tests are two tests of
the same ability that have the same means and variances and are equally correlated with
other tests of that ability. It is defined as the extent to which 2 tests taken from the same
domain measure the same things. Henning (1987) also points out that in order to carry out
this method, the tests must show equivalent difficulty, variance and co-variance, and that
no learning practice should be done between two occasions. However, in practice, it is very
difficult to obtain high reliability based on the parallel form method because students may
get unequal results for the same level’s tasks of parallel forms ( they may find some tasks
familiar with them or sometimes their results may be affected by random factors such as
health, mood, climate, etc.). Therefore, alternative forms are often simply not available.
Because of the problems associated with parallel form and test-retest methods, the other
method, which is known as split half and focuses on the consistency with each other of a
test’s internal elements, is represented as bellows.
(iii) Split half method, according to Bachman (1990:172), is the method used to examine
the internal consistency (concerned with how consistent test takers’ performances on the
different parts of the test are with each other) of a test in which we divide the test into two
equivalent halves and determine the extent to which scores on these two halves are
consistent with each other. According to Bachman, there are two ways of splitting a test
into two halves (dividing into the first and second halves, the odd and even halves) in
which the odd- even method is particularly applicable to tests in which items are ordered in
terms of difficulty, designed to measure the same ability and to be independent of each
other. However, Bachman recommends that in situations where it is not possible to retest
individuals, or to administer equivalent forms, and some estimate of internal consistency is
the only means of examining the reliability of the test, we must make every attempt to split
the test into halves in such a way as to maximize their equivalence and their independence.
The reliability coefficient is calculated using what is known as the Spearman- Brown
Prophecy formula as follows:
2* reliability for 1/2 test
Reliability of scores on total test =
1+ reliability for 1/2 test
However, the most frequency employed method for determining internal consistency is the
Kuder- Richchardson approach, particularly formulas KR20 and KR21.
Shohamy (1985), on the other hand, also presents Kuder- Richardson Formula 21 in
another form which is illustrated as follows:
x (K- x)
Rxx = 1-
x: the mean
SD: standard deviation
K : number of items
This formula can be used as a means to calculate the reliability of a test. This will be
explored in the next chapter.
(iv) Rater consistency reliability can be discussed in cases the tests scores are evaluated
subjectively such as test scores of compositions or oral tests. There are two types of raterreliability: intra- rater reliability and inter- rater reliability which will be represented as
* Intra- rater reliability is defined as the extent to which the same rater is consistent in his
rating from one occasion to another (Shohamy, 1985). Actually, if the rater applied the
same set of criteria consistently in rating the language performance of different individuals,
this will yield a reliable set of ratings. However, in some cases, inconsistency may occur in
the rating criteria themselves or the way in which they are applied. For instance, in scoring
compositions, some raters pay more attentions to grammatical errors and that affects their
ratings or in oral tests, raters may, sometimes, unconsciously relax their ratings. According
to Bachman (1990: 179), in order to examine the reliability of ratings of a single rater, we
need to obtain at least two independent ratings from this rater for each individual’s
language sample. He also recommends two approaches to calculate the reliability of the
ratings: computing the appropriate correlation coefficient between the two sets of ratings,
or computing a coefficient alpha.
* Inter-rater reliability is the extent to which different raters are consistent in their ratings.
This is a major concern in tests of writing and speaking proficiency which are subjectively
scored. According to Bachman (1990:180), ratings given by different raters can also vary
as a function of inconsistency in the criteria used to rate and in the way in which these
criteria are applied. He proves that with the same rating scale given, different raters yield
different results for the same tests since different raters may form their own focuses (on
grammar, vocabulary or content...) in their rating process. Moreover, although they have
the same rating scale, they may interpret it differently.
Therefore, Bachman recommends Ebel’s approach to calculate the reliability in case there
are more than two raters, in which we sum the ratings of different raters and then estimate
the reliability of these summed ratings by computing a coefficient alpha in the following
k: number of raters
si2: variance of ratings for a given rater
sx2: variance of the summed ratings
si2: sum of the variances of different raters’
220.127.116.11. Measures for improving test reliability.
As mentioned above about the factors to improve the reliability of a language test given by
Sako (1983:28), Weir (2005:117) restates Hughes’ guidelines for making the test task itself
more likely to produce reliable scores as follows:
(i) take enough sample of behavior;
(ii) do not allow candidates too much freedom;
(iii) write unambiguous items;
(iv) provide clear and explicit instructions;
(v) ensure that tests are well- laid out and perfectly legible;
(vi) make candidates familiar with format and testing techniques;
(vii) provide uniform and non- distracting conditions of administration;
(viii) use items that permit scoring which is as objective as possible;
(ix) make comparisons between candidates as direct as possible.
And in relation to scoring of test performance itself:
(i) provide a detailed scoring key;
(ii) train scorers;
(iii) agree acceptable responses and appropriate scores at the outset of scoring;
(iv) exclude items which do not discriminate well between a weaker and stronger
(v) identify candidates by number, not name;
(vi) employ multiple, independent scoring.
In a nutshell, in order to achieve the above factors, we need the efforts and attempts of the
test designers, test administers and test raters.
In this chapter, we have attempted to establish the theoretical framework for the thesis,
consisting of three main issues: language testing (definitions, approaches, roles, purposes
and relationships between testing and teaching- learning); testing writing (types of writing,
criteria of testing writing and problems in testing writing); and criteria of a good test with a
focus on reliability (factors affecting language test scores, methods to determine test
reliability, and measures for improving test reliability).
From those issues, we draw out the main principles for our research in the next chapter.
They are represented as follows:
(i) Reliability is considered as one of the criteria to make a good test, especially
unreliability is often found in scoring compositions as it is scored subjectively.
(ii) One of the useful methods which is used to measure the reliability of a writing test,
especially a composition, is inter- rater reliability.
(iii) The reliability coefficient is calculated by the following formula presented by
x (K- x)
Rxx = 1-
x: the mean
SD: standard deviation
K : number of items
(iv)The good range of the reliability coefficient for tests of essay types may be in the 0.7
to 0.79 range.
(v)The measures for improving test reliability in relation to scoring of test performance
are pointed out clearly in 18.104.22.168. (p.18).
AN OVERVIEW OF THE TEACHING, LEARNING AND TESTING
This chapter provides the background information about the teaching at Nghe An Junior
Teacher Training College. The information presented here includes the reality of English
language teaching, learning and testing; the learners and teachers’ profile; and a
description of the current writing syllabus as well as the teaching aims.
2.1. The students, teaching staff and the English teaching & learning situation at N.A
Junior Teacher Training College.
2.1.1. The students and their backgrounds
Nghe An Junior Teacher Training College has been training English teachers for junior
high schools in the province since 1999. In general, most of the students who enter
N.A.JTTC are aged 18-20 and are female students (90%). Each course consists of about 3
or 4 classes with 30-40 students in each. A large proportion of them comes from the rural
and remote areas of Nghe An province, so their levels of English proficiency are low and
especially, they are not used to learning four skills (speaking, writing, reading and
listening) in which writing is considered to be one of the most difficult skills. They are
only good at the controlled writing, not familiar with guided and free writings, especially
teachers are worried about testing and evaluating their writings.
2.1.2. The English teaching staff
The English Department is staffed with 40 teachers aged between 26-55, 6 of them are
“reserved” Russian teachers. The proportion of teachers who have M.A Degrees is 4/40.
Now other six are studying for M.A Degrees at VNU, one at a university in America and
five are studying for M.A Degrees at Dalana University of Switchland. In general, they are
well- trained and rather professionally experienced with at least 2 years in teaching