Construct Validity
Construct validity is related to the teaching process. Second language teaching is based on a learning theory. Behaviourists’ believe that language is learnt bit by bit. The skills are isolated and language is decontextualized and presented to the learners. In such context, testing should reflect the learning principles used in the teaching process. Language items are decontextualized in testing too. In contrast, according to communicative approach to language teaching, skills are integrated and language items are contextualized. Test items will reflect this feature. In essence, the theories used in teaching should be reflected in test item construction.
In language teaching, most of the skills are mental constructs, which require to be converted into measureable entities. For instance, the construct ‘reading’ is abstract and requires to be measured in some form which can be assessed. In this case, we decide the underlying abilities of reading – one’s ability to read and extract information from a text, get the main idea of a text etc.
Construct validity too ensures the fact that test paper measures what it claims to measure and nothing else.
Concurrent Validity
This is an aspect of validity through which tester can validate the test results. Suppose that a candidate scores 80% in a speech test. What does this mean? The syllabus includes, say for example, fifty language functions1, which the candidate should master in the study programme. Does 80% mean that he can use eighty percent of these functions? The test has not obviously tested all these functions in the test. How can he be sure of the learner performances? These functions are different from each other. One’s ability to perform in one function such as ‘agreeing’ does not mean that he can perform in other functions such as ‘requesting’ or ‘apologizing’. How does the teacher ensure that he has selected ‘a representative sample’ of language items for testing? In such case, can the teacher validate the results?
For such a speech test to be valid, the teacher should test all language functions given in the syllabus! This is not practical at all.
To validate the test results, the tester should compare the results of some standard measurement. This means that there must be two tests. One test is the general test administered to all the students where some selected language functions are tested.
The second test is a standard form of measurement. The tester randomly selects a few students, say six students, and administers another test, where all the language functions given in the syllabus are tested. It is practical because it involves only a limited number of students. Both tests are administered almost at the same time.
At this stage of test validation, there are two sets of scores, the second set comprises marks of only six students. These test scores of the selected six students can be something like the following:
Student Number
|
Test 1
|
Test 2 (Standard test)
|
1
|
72
|
72
|
2
|
46
|
46
|
3
|
57
|
57
|
4
|
81
|
81
|
5
|
64
|
64
|
6
|
39
|
39
|
What does this mean? In both test situations, the candidates have performed equally. Since in the standard test all language functions have been tested it is certain that candidates demonstrate the skills as specified in the syllabus. The fact that there is no difference in the two tests means that the other test – the general test administered to all candidates – also gives us the same results. It is possible to conclude that the general test is equal to standard form of measurement. It is a valid test. Both tests are 100% equal. The general test is validated in this way. The validity co-efficient is 1 (100 out of 100).
But such results are not always realistic. What about the following two sets of scores?
Student Number
|
Test 1
|
Test 2 (Standard test)
|
1
|
72
|
34
|
2
|
46
|
81
|
3
|
57
|
28
|
4
|
81
|
42
|
5
|
64
|
31
|
6
|
39
|
84
|
These results show that we cannot believe one set of scores: since the standard for of measurement must be accurate, Test 1 results should not be true. In other words, Test 1 results are not valid. This shows that the validity co-efficient2 is 0 (0 out of 100).
This may also not be the case always. The validity co-efficient will range between 0 and 1 in actual test situations. If it is closer to 1, the results can be interpreted as valid. Suppose we get a value of 0.67 as validity co-efficient. Is this a sufficient value for us to say the test is valid? The answer is that it always depends on the purpose of the test. If it is for general assessment in a final achievement test, it may be sufficient, but in a selection test where we intend to select the most suitable 10 candidates for a particular diplomatic post, it is a lower value.
As explained here, the standard assessment is the criterion against which the test is validated. Since the test and the criterion test are administered almost at the same time, this is called concurrent validity. In fact, it is criterion-reference validity.
Do'stlaringiz bilan baham: |