Chapter 5
Reliability
Reliability of a test concerns the consistency of test scores. Suppose that the same test is administered twice without a time interval. If the conditions of test administration are identical, the mark each student receives in both sittings should be almost the same. Does this happen always?
This chapter discusses the issue of consistency of test scores.
What is Reliability?
If the same test is administered to a same student, the marks he receives may vary due to many reasons.
There are differences in the way the test is administered.
It is possible that learning takes place between the two tests.
It is also possible that the learner forgets between the two tests.
If marking is done by two examiners and if the questions require judgment of the examiner, it is likely that marks vary.
It is quite natural that human beings perform the same task slightly differently in two occasions.
However, when a student gets a mark like 72 in a test, we have some idea of his performances. If this varies from time to time without a valid reason, we may not believe the test results. We say that the test is NOT reliable. Reliability of a test is the consistency of test scores. The more similar the test scores in two sittings, the better the reliability will be. In fact, reliability is an important aspect of testing. Even though some conditions are inevitable, a good tester must see that his test is reliable.
Reliability Co-efficient
It is possible to quantify reliability. For this purpose two sets of scores are needed5. Consider the following two sets of scores.
Student Number
|
Test 1
|
Test 2 (Standard test)
|
1
|
82
|
82
|
2
|
56
|
56
|
3
|
37
|
37
|
4
|
79
|
79
|
5
|
64
|
64
|
6
|
24
|
24
|
The two sets of scores are identical. This demonstrates perfect reliability. The reliability co-efficient is 100% or 1. But in natural test situations this is very unlikely. Consider the following testing situation.
Student Number
|
Test 1
|
Test 2 (Standard test)
|
1
|
82
|
34
|
2
|
56
|
28
|
3
|
37
|
74
|
4
|
79
|
17
|
5
|
64
|
91
|
6
|
24
|
48
|
The two sets of scores are not related at all. The reliability co-efficient is 0. In most practical test situations, this is also not very much likely.
In real testing situations, this value ranges between 0 and 1. If it is closer to 1, there is higher degree of reliability. (This value can be easily calculated once the two sets of data are fed into Excel programme.)
Language testing involves testing of skills. In measuring skills, objective as well as subjective test items are helpful. For instance, a skill like reading can be tested through objective test items only. These items require one unique answer and do not involve judgment of the examiner. Therefore, high reliability can be obtained. But in the case of productive skills such as writing, where we include direct test items, subjectivity comes into play. The degree of reliability will be relatively less in such instances. The reliability co-efficient value of above 0.90 is expected in reading tests, while in listening it could be around 0.80. However, in writing, it could be 0.70 and in speech even a little lower than that.
Do'stlaringiz bilan baham: |