19

CHAPTER 2
A Review of Related Literature and Studies
2.1 Prologue:

Review is an activity which assumes a lot of importance in

research. It is so because it provides the researcher with the information
about the researches done in the past in the related area. This
information enriches the knowledge of the researcher and provides
direction for identifying the problem for his/her research. Since the review
of related studies provides information about the research work already
done and ultimately one gets the information about the research work
which has not yet been done. Thus it gives insight into the problem and
also the direction for dealing with the problem. The review of related
literature provides knowledge about the literature available, the selection
of appropriate tools that can be used and also the methodology of
research. The following section deals with the literature and studies
relevant to the present study.

2.2 Review of related literature

In order to have a better understanding of the problem under study,

its methodological concepts and to select the appropriate statistical
method and tools for analysing the data a thorough study of the related
literature was undertaken.

2.2.1 About the Battery of Selection Tests:

The multiple choice objective-type test battery used for selection of

employees in banks in the clerical cadre consisted of the following tests.
 20

Test I: Reasoning Ability - This test consisted of 50 objective type

questions. Each question had five alternatives out of which one was the
right answer. Out of the 50 questions 25 questions were on verbal
reasoning in which a problem was described and the candidates were
supposed to solve it and 25 questions were of non-verbal type i.e. figures
(diagrams) were given and the candidates had to find the answer by
establishing the relationship between the given figures (diagrams). This
test aimed at assessing the Analytical Ability of the candidate.

Test 11: English Language - This test consisted of 50 objective type

questions, each question had five alternatives out of which one was the
right answer. Questions were based on a passage to assess the reading
comprehension, functional grammar and lexical skills. The test did not
require any knowledge about English literature. This test was aimed at
assessing the facility with language one has. However, since it was a
language test and had all the chances of being biased in favour of
candidates from urban areas this test was only a qualifying test and
scores on this test were not reckoned for merit-ranking.

Test III: Numerical Ability - Consisted of 50 objective type questions

with five alternatives, one of the alternatives being the right answer. In
this test 25 questions were on only computation while 25 questions were
problems on arithmetic reasoning. This test assessed the proficiency in
Maths. This test was aimed at assessing the facility with numbers and
basic computation skill and did not require in-depth knowledge of
mathematical concepts and fomriula. Besides computation, the 25
questions on arithmetic reasoning also assess the ability to interpret the
problem described and solving it.
 21

Test IV: Clerical Aptitude - Consisted of 50 objective type questions with

five alternatives one being correct answer. 35 questions were on
matching names and numbers while 15 questions were on classification
and simple table reading. This test was to assess how speedily and
accurately one can work with numbers and letters.

2.2.2 Standardization of Tests

Testing is a process used for testing the knowledge of an individual.

It is used in schools and colleges to test the subject knowledge learnt as a
part of curriculum. Besides the school/college education, there are tests
which are used for testing various abilities that an individual has. The
tests used are basically of two types (i) Objective type and (ii) Essay type.
In an essay type test, the examinees are asked to write their own answers
to the given questions, whereas in on objective type test the examinees
are asked to choose the correct answer from amongst the given
alternatives. However, the objective type items sometimes are of match
the column or completion type items. These can be one word item or
short one sentence answer item. The one sentence type items can come
under essay type short answer items as well.

The assessment of essay type questions is bound to be influenced

by the personal views, knowledge and style of the examiner whereas, the
assessment of objective type questions is not affected by change in
assessor, once the right answer key is prepared. Usually the teacher
made classroom tests for the use of school/college examinations to be
given to the school/college students are informal and not standardized. In
the school/ college examinations the tests used are teacher made tests
based on the syllabus and when the scoring scheme is given assessment
 22

can be done by anybody. The score given by any assessor will always
remain the same.

The tests can be classified as informal or standardized.

Technically, a standardized test is a systematic process for

assigning numerical values to sample of behaviour, such that the same or
equivalent items are administered to all test-takers using uniform
directions and scoring methods. Standardization can be extended to
include uniform methods of test interpretation derived through the
development of norms for the group for which the test was designed,
information on the consistency of test performance (reliability) and
evidence that the test measures what it was designed to measure
(validity).

2.2.2.1 Performance Criteria for Standardized Tests:

Criteria for the perfomnance of standardized test can be classified

into two categories: (i) requirements for obtaining accurate and
appropriate scores; and (ii) requirements for making accurate and useful
interpretations of the scores. Requirements for obtaining accurate and
appropriate scores include matched level of tests to the students, clarity
and comprehensiveness in the administration of instructions and logistics.
Requirements associated with test interpretations include nonns, reliability
and validity.
 23

2.2.2.2 Norms:

Norms permit an interpretation of a test takers score to be made

relative to the scores made by a large number of similar individuals.
These interpretations do not address how much content was mastered but
how well the test-taker did in comparison to his/her peers.

2.2.2.3 Administration of the tests under standard conditions:

The conditions under which the tests are administered, play an

important role in deciding the performance of the test-takers. It is
therefore, extremely important and necessary to conduct the tests under
specific conditions to have the valid interpretation of the test scores.
Therefore for the standardized tests, not only the question paper and the
answer paper are important but also the manual of instructions for test
administration are equally important. The manual of instruction includes
the following information.

(a) Description about testing conditions.

(b) Procedure of test administration step-by-step in detail.
(c) Directions to be given to the examinees by the examiner. The
instructions which are to be read aloud are highlighted.
(d) Information required for answering general queries of the
examinees.
(e) Time limits and the method to be followed for observing these time
limits.
(f) Instructions for handling, packing and despatch of the test material.
 24

It may be noted that the meaningfulness of the test results depends on

strict adherence to the procedures described In the Instruction manual
(Wardrop, 1976, pp.2-3).

2.2.2.4 Objective scoring of the standardized tests:

Objective scoring means any two equally trained scorers would get
exactly identical scores for the same test-paper. This is possible only for
the objective type tests for which the scoring key can be fixed and the
scorer has to just match the right answer key with the examinee's
responses and count the responses which are correctly answered i.e.
matching with the right answer key. Therefore the essay type of
subjective tests are not used in standardized tests but only the objective
type tests can be standardized. The scoring of a standardized objective
type tests is not influenced by the subjective reactions of the scorers. The
scoring of objective tests can now be done by using the automated
scoring machine known as "Optical Mari< Reader" (Wardrop, 1976, p.3.).

2.2.2.5 Construction of a Standardized Test:

The development of a test involves the following steps and

asepects for consideration as described in Wardrop (1976, pp.72, 73).

Test Content: Content analysis is an extremely important aspect in

development of a standardized test. Since the standardized test is
expected to be used for many years, it attains greater importance to
analyse the contents carefully. For content analysis the author is required
to analyse the curriculum. Therefore not only the content analysis but also
the curriculum analysis is required to be done. The curriculum should be
 25

decided in such a way that it remains useful for the years to come. In the
content analysis one has to prepare a blue print, which gives the topics to
be covered, the number of questions from each topic and the distribution
of difficulty level of questions under each topic and the weightage to be
given for each topic. In short, the blue print of a test gives the detailed
picture of the test.

Item writing: The next step to content analysis is item writing. To write
good items is a difficult task. The items written must go through the
stages of review, modification and editing before they are put to actual
use. Even with the careful process of review, a number of items are likely
to be discarded after actual try out. So at least two to four times the
number of items should be written than that of the actual requirement.

Item Analysis:
Item Analysis is a technique to know the qualities of items, on the
basis of responses given by the candidates who have taken the test. It is
necessary to ensure the effectiveness of the test as a whole, in terms of
each of its units i.e. items.

To contribute to make the test valid in terms of effectiveness of the

test to tap the desired abilities, aptitudes, etc., there are certain features
within the test. For a test to be valid, there is a pre-requisite that the test
should consist of items which are effective and are free from any defect.
Each test item should be homogeneous to the entire test, thus contributing
to the purpose for which it is developed. A thorough item analysis also
includes a number of quantitative procedures. Specifically, the numerical
indicator are often derived during an item analysis: Item difficulty, Item
discrimination and distractor power statistics.
 26

Item Difficulty Index (p):

Item difficulty statistics is an appropriate choice for aptitude tests

when the items are scored dichotomously i.e. correct vs. incorrect. The
facility Index (p-Value) is the percentage of students who have given
response as that particular answer choice. The p-Value of answer choice
which is the 'key' for the item, is the percentage of students who have
attempted the item correctly. Higher the p-Value easier is the item e.g. if
the p-Value of an Item is .68 means 68% examinees have answered it
correctly, while if the p-Value of another item is .15 i.e. 15% of the
examinees have answered it correctly. Thus the item having p-Value of
.68 is easier than the item with p-Value of .15. It is expected for obvious
reasons that more number of high scoring examinees answered the item
correctly compared to the low scoring examinees. When p is the
percentage of passing the item and q is the percentage of failing, pq is the
variance. For p = .5, q = .5, variance = .25 which is the maximum possible
value of variance. Therefore items with p-value .5 are most preferred
items. Since .5 is the ideal value, practically the items with p-values
ranging between moderate values of .4 to .6 are preferred to the much
easier or much harder questions.

Item Discrimination Index (D):

The item discrimination analysis addresses the validity of the items

on a test, that is the extent to which the items trap the attribute they were
intended to assess. Similar to the Item Difficulty Index, the Item
Discrimination Index can be derived whenever the items can be scored
dichotomously as correct or incorrect.
 27

Index of Discrimination is the item-test "point biserial correlation

coefficient". The 'test score' of a candidate for a 50 items test is a
continuum ranging from 0-50, whereas the 'item score' is a dichotomous
score viz '0' if the response is not the correct answer and ' 1 ' if the
response given is the correct answer. This dichotomous score of '0' or ' 1 '
is correlated with continuum score on the test. The statistical method
used for this purpose is known as "point Bi-Serial Correlation" according to
Garrett (1981. p.97).

When the items are scored dichotomously, the assumption of

normality is unwarranted and therefore point biserial r is more appropriate
than the biserial r which assumes normality of distribution. Moreover,
biserial 'r' estimates the correlation on higher side as compared to point
biserial 'r'. The biserial correlation is used when the dichotomous variable
is a "forced" one like pass or fail and the point biserial correlation is used
when the dichotomous variable is a "true" one like score of 1 or 0. In case
of Objective test items the dichotomous variable is a "true" one which is
score 1 or 0 and therefore point biserial correlation is calculated.

The index of discrimination can also be defined as the difference

between p-values of high performing group (upper group U) and the low
performing group (lower group L), and is designated as 'D'. Thus if the
percentage of candidates in the upper group responding correctly to an
item is 'U' and the percentage of candidates in the lower group responding
correctly to the same item is 'L' then for that particular item D = U - L.
Generally, items with .20 index of discrimination are acceptable while
items with moderate index of discrimination above .30 are preferred.
 28
2.2.3. Reliability

Reliability refers to the consistency of measurement. We would

always like to use the tools that will give us a consistent measure of one's
ability or aptitude; meaning thereby, the scores on a test taken at one
time should not significantly differ from the scores on the same or similar
test taken at some other time. But since the scores on psychological tests
are derived from human responses, and one characteristic of human
being is that their behaviour tends to fluctuate from time to time and from
situation to situation.

True Score - Refers to the score that an individual would have obtained on
a test under perfect conditions and it could be measured without error.
However, this is a difficult, rather only a hypothetical situation so far as the
psychological tests are concerned. Therefore, the scores obtained on a
test can be represented by the following equation -

X = X +E
t T
where X = obtained score
t

X = True Score
T

E = Error which can be positive or negative.

It may be noted here that certain concepts viz. true score, error,
true variance, error variance, etc. cannot be directly measured but they
can only be estimated.

A score obtained on a particular test on a particular day should not

be considered as an exact point but rather it should be thought of as
 29

representing a zone in which the student's true ability lies. That is, if a
particular student scores at the 60'^ percentile, it is wrong to conclude that
this is his exact standing. It is more appropriate to say that his true score
lies somewhat near the 60**' percentile. Scores obtained on any particular
test on any particular day are somewhat biased. High scores are too high
and low scores are too low. it's not only the ability of the student that
determines the score, but it is also the luck factor that contributes. The
students who scored high on test, are not only high on ability but also they
had a better luck at that particular point of time and the students who
scored low are not only low on ability but also they had a bad luck at that
particular point of time. If an alternate form of test is administered on
some other day, the students who scored high on the first occasion, will
as a group, score little lower while the group of students who scored low
will as a group, score on little higher side. This is called the tendency of
regression towards mean. However, this tendency does not necessarily
apply to each individual. Few students who score high on the first
occasion may score still higher on the second occasion and few students
who scored low on first occasion may score still lower on the second
occasion. All scores above mean tend to be somewhat biased upward i.e.
they are probably higher than they should be.

2.2.3.1 Test-Retest Reliability

In this method of determining the reliability of test scores, an

identical test is administered on a second occasion. The correlation
coefficient between the test scores obtained on the first and the
second occasion determines the reliability of the test. The error variance
 30

corresponds to the random fluctuations from one test session to another.

This could be due to any one or more reasons like change in testing
conditions such as change in weather, light, venue, test administration
personnel, distractions like noise, etc. or even a broken pencil point are
some of the factors that may affect an individual's score. "Retest
reliability shows the extent to which scores on a test can be generalised
over different occasions. The higher the reliability, the less susceptible the
scores are to the random daily changes in the condition of the test takers
or of the testing environments.

While reporting the Retest reliability, the length of interval on which

the test was readministered should also be mentioned. It is desired that
the interval should not be very long since a remarkable development may
take place over a period of time which may rise or drop an individual's
status. For example, additional qualification or job-experience acquired
over a period of time may lead to improvement in an individual's
performance, on the other hand in case of old persons or housewives the
performance may go down over a period of time. Therefore, the interval is
desired to be short particularly in case of young children, in whose case
the progressive changes are fast and can take place in a very short period
of time.

Although apparently simple and straight forward, the test - retest

technique presents certain difficulties. Though the interval is kept short,
persistent practice may bring in some improvement in an individual's
performance. Also, if the interval is short an individual may be able to
recall his responses on the earlier occasion which may affect his
performance. Retesting with the identical test does not have control over
these factors.
 31

2.2.3.2 Alternate Form Reliability

This method of determining reliability takes care of the difficulties

found in the test - retest reliability. In this method an individual is tested
on two different but equivalent forms on two different occasions. The
correlation between the scores on two forms represents the reliability
coefficient, gives us the measure of two types of reliabilities viz. temporal
stability or the reliability of test scores obtained on different occasions and
consistency of response to different item samples. Thus alternate form
reliability provides more useful measures for evaluating a test.

Alternate form reliability should also be reported alongwith the

interval of time between the administration of two tests. If the two forms
are administered one after the other in immediate succession, the
correlation coefficient represents the reliability across two forms and not
across two occasions, and the error variance represents fluctuation in
performance on two sets of items and not on two different occasions.

In this type of reliability, while developing the alternate form it is

necessary to ensure that the two forms i.e. the two sets of items are truly
parallel. The two sets of tests may be constructed independently with
same specifications. The number of items under each area should be the
same, moreover the items should be of same difficulty level. Time allotted
for both the tests should be same, instructions and the sample items if
any, should also be equivalent.

In this method the constraint of practice effect gets reduced but not
completely eliminated. The forms of test being parallel, contain the
problems based on the same principles. Once the principle is understood
 32

by an individual, it may not be difficult applying the same principle and

solving a similar problem. Moreover, developing truly equivalent forms
may not be always possible.

2.2.3.3 Split-Half Reliability

In this method the scores are obtained for each student in two parts
viz. for two halves of a test. The test is split into two equal halves, no. of
questions in each part being equal. This can be done by taking first half
and next half questions, or by taking alternate questions in the two parts or
by any other method so as to ensure that the two parts of the test are
equivalent. The method of splitting would depend upon the structure of
the test. Usually, in most of the tests, first half and second half would not
be equivalent, owing to differences in nature and difficulty level of items as
well as cumulative effect of warming up, practice, fatigue, boredom and
any other factors varying progressively from the beginning to the end of
the test. A procedure that is most suitable for most of the purposes is to
find the scores on odd and even items of the test. Such a method of
division yields most nearly equivalent half-scores. However, while dividing
the test into two halves, if a group of items is based on common
information, all those items in the group should be put in the same part.

The split-half reliability provides a measure of consistency with

regard to content sampling. It obviously does not provide the temporal
stability of the test since the test is given in a single session. This type of
reliability coefficient is sometimes called a coefficient of internal
consistency, since only a single administration of a single form is required.
Once the scores on the two halves are obtained, correlation between the
two sets of scores is obtained. This correlation coefficient represents the
 33

reliability coefficient. Actually, this type of reliability coefficient gives the

reliability of half the test, since the number of items in each part is half of
the items in the original test. Other things being equal, the reliability of a
test increases with its length. A longer test will be more reliable. With a
longer sample of behaviour, we can arrive at a more adequate and
consistent measure. The effect that lengthening or shortening a test will
have on its coefficient can be estimated by means of the Spearman-Brown
formula given in Anastasi (1997, pp.95-96).

2.2.3.4 Internal Consistency Reliability

It involves single administration of a single form. It is based upon

the consistency of responses to all the items in the test. The inter-item
consistency depends upon the (i) content sampling and (ii) heterogeneity
of the sample. The inter-item consistency increases with the homogeneity
of the test. For instance, if a test contains questions on Maths, Science
and English as against another test containing question only on English
will have lower inter-item consistency. Further a test of English covering
areas of grammar, vocabulary and comprehension as against a test of
English only covering grammar will have lower inter-item consistency.
Content sampling will depend upon the criterion that the test is meant for
predicting. A highly heterogeneous criteria cannot be tested by a highly
homogeneous test. In such case it is desirable to construct several
homogeneous tests to measure different aspects of heterogeneous
criteria.

The most common procedure for finding inter-item consistency is

developed by Kuder and Richardson. It involves the calculation of Test-
Reliability coefficients based upon the method of rational equivalence.
 34

The method of rational equivalence is an attempt to get an estimate of the

reliability of a test, free from the objections raised in other methods. Two
forms of a test are said to be parallel when the corresponding items in the
two tests are interchangeable and the inter-item correlation is same for
both the forms. The method of rational equivalence stresses the inter-item
correlations and also the correlation of Items with the whole test. Out of
the various formulas derived the most widely used one is known as
"Kuder-Richardson formula 20" (KR-20). A simple approximation of above
formula KR-20 is KR-21 as per Guilford (1978, pp.427-429).

2.2.3.5 Significance of reliability coefficient

There is no hard and fast rule as to how high should be the

reliability coefficient of a test to be reliable. It depends upon the type of
the test and the purpose for which it is used. As stated in Guilford (1984,
pp.388-389), for research purpose, lower reliabilities are acceptable as
compared to the one for practical purposes of diagnosis and prediction.
Sometimes, we can make best use of the test even with a reliability as low
as 0.50. For some purposes, even a test of low reliability adds enough to
prediction to justify its use, particularly when used in a battery along with
other tests. According to Kaplan (1977, p. 121), reliability estimates in the
range of 0.70 and 0.80 are good enough for most purposes in basic
research. Also, since KR21, correlation coefficient is an approximation of
Pearson's product moment correlation, and referring to table Q in Guilford
(1978, pp.531-532), a correlation coefficient as low as 0.081 for df = 1000
is highly significant i.e. significant at .01 level, it can be concluded that any
reliability coefficient more than 0.5 will be highly significant if the sample
size is more than 1000.
 35

An index of homogeniety given us the estimate about internal

consistency of the test, however, it does not give direct indication of the
amount of variability or error that can be expected in an individual test score.
The difference between the obtained score and the true score is the error
score or Xt = X j +E where Xt is the obtained score and XT the true score. E
is the error which can be positive or negative. Standard error of
measurement is the standard deviation of the distribution of error scores and
is computed by Sm = Sx 7 1 - TK

Where Sm is the standard error of measurement, Sx is the

standard deviation of obtained scores and rtt is the reliability coefficient.

95% confidence limits can be obtained by Xt +1.96 Sm that means

the true score can be more or less by 1.96 Sm than the obtained score.
99% confidence limits can be obtained by Xt + 2.58 Sm. Confidence limits
give us the range by which the obtained score can differ from the true score.

2.2.4. Validity:

Any test is always designed with some purpose and the test scores
have some meaning only when they are related to some other variable,
e.g. an achievement test is meant to know the academic performance of
students in the class-room, an aptitude test is meant for knowing certain
abilities of an individual which in turn will predict the behaviour of an
individual in certain domain. A test will be useful when its scores can be
correlated with some other variable and can be meaningfully interpreted. If
 36

a test has perfect consistency but its scores cannot be correlated with any
other variable, such a test will be of no use and hence will have no validity.
Validity is a generic term and can be defined at various levels in various
ways. The validity of a test can be determined from the answers to the
questions like - How well does the test measure what it is supposed to
measure?, What traits does it measure?, Does it really measure what it is
supposed to measure? Does it supply the infomnation that can be used in
decision making? What interpretation can be given to the test scores?
What percent of the variance in the test score is attributable to the variable
that the test measures?

Validity can be determined by the proportion of the true variance

that is relevant to the purposes of the testing i.e. the extent to which the
scores are attributable to the variable that test measures. Thus, as
explained in Brown (1970, pp.97-98) the validity of a test is defined either
by (1) the extent to which the test measures the hypothesized underlying
traits, construct or factor, or (2) the relationship between test scores
and some extra-test criterion measure.

2.2.4.1 Content Validity.

The class room tests are used for assessing the knowledge
acquired by the students through the class room teaching. These tests are
syllabus based and the items included in the test are just a sample as it is
not feasible to cover entire syllabus in the test. The teacher thus has to
test the entire knowledge through these test items and it is therefore
necessary that the tests are representative of the behavioural domain.
Scores on the tests are used not as ends in themselves, but rather to
make inferences about performance in the wider domain. The purpose of
 37

class room examination is to provide an objective basis for making an

inference about the students' knowledge of the material covered in the unit
over which they are being examined. Because it is not possible to ask
every question of the unit, a sample of possible item is selected and is
given in the form of a test. On the basis of the student's performance on
this sample of items, his degree of knowledge in the entire unit of subject
matter can be inferred. To the extent that the items are a good sample of
the universe our inferences will be valid; to the extent that any bias is
introduced in the selection of items, the inferences will be in error, and the
test will be invalidated.

Although the concept of content validity is usually associated with

achievement testing, it can also be applied to other areas of psychological
testing. The definition of the content validity as per Brown (1970, pp.135-
138) states that the items on the test must be representative sample of the
universe of possible content or behaviours. Representative sample means
taking a sample that includes in due proportion or frequency every
relevant or required characteristic of the whole.

The principal method of determining content validity involves

establishing one-to-one correspondence between the test items and the
universe of behaviour with the help of an expert judge. However, this
method does not have any quantitative index or even a set of agreed upon
qualitative categories, secondly owing to individual differences, different
judges will arrive at different conclusions. To overcome this problem, a
well specified definition of the content universe is called for.
 38

2.2.4.2 Face Validity:

It is not the validity in technical sense. It does not refer to what it

actually nneasures, but it refers to what it appears to be measuring. Face
validity refer to whether the test looks like valid to the examinees, the
administrative users and other untrained personnel. Thus face validity is
determined by a superficial examination of the test by the test-taker and
considers only obvious relevances. Content validity, on the other hand, is
established by the thorough and systematic evaluation by a
psychometrically sophisticated judge and considers both subtle and
obvious aspects of relevance.

Face validity is important to the extent that the appearance of the

test has an influence over the motivation of test takers. For example,
if in a particular situation if the test does not have face validity i.e. it does
not appear to be valid to the test-taker, the test taker may feel the test has
no relevance with the decision to be taken and hence may not be
motivated to write the test, thereby affecting his performance. Thus, face
validity, may not guarantee accurate measurement, may be an important
influence on test-taking motivation and thus on the validity of the obtained
scores.

2.2.4.3 Construct Validity:

Construct validity of a test is the extent to which the test may be

said to measure a theoretical construct or trait. Any data throwing light on
the nature of the trait under consideration and the conditions affecting its
development and manifestations represent appropriate evidence for this
validation. Construct validity is said to be appropriate when the test user
wishes to infer the degree to which the individual possesses some
 39

hypothetical trait or quality (construct) presumed to be reflected in the test

performance. Thus, whenever a test is to be interpreted as a measure of
some attribute or construct that people are expected to be possessing;
construct validity becomes important.

The construct validity is classified into two types viz. convergent

validity and discriminant validity. When a test or some other measure of
the proposed trait correlates strongly with other instruments designed to
measure the same trait, it is said to have convergent validity. According to
Guilford (1978, p.436), when a test correlates very little or not at all with
measures of other traits; it is said to be having discriminant validity. Factor
Analysis is the most commonly used technique to establish construct
validity.

2.2.4.4 Criterion Related Validty:

The most common use of tests, other than the class room tests is
to predict the performance in the domain of behaviour. For example,
achievement tests are used to place the students in various class
sections, while aptitude tests are used for predicting performance on job.
The variable that is predicted by the test is called criterion, and the validity
thus obtained is called criterion-related validity. Since the test is used to
predict the criterion, the criterion-related validity is sometimes referred to
as predictive validity; and since it involves collection of empirical data
on the relationship between test scores and the criterion measure, this
type of validity is also referred to as empirical validity. The proper
measure of a test's criterion related validity, and thus its usefulness, is an
index of its relative contribution, over and above that of other measures
and sources of information, to increased decision making accuracy. We
 40

are not interested in the test scores per se, but are interested in the test
because it predicts some important criterion behaviour.

As described in Anastasi (1997, p.119), for certain uses of

psychological tests, concurrent prediction is the most appropriate type and
can be justified in its own right. The logical distinction between predictive
and concurrent validation is based not on time but on objectives of testing.
Concurrent validation is relevant to tests employed for diagnosis of
existing status, rather than prediction of future outcomes.

2.2.4.4.1 Validity Coefficient method of determining criterion-related

validity:

One of the most commonly used methods of determining criterion-

related validity is by way of Validity Coefficient' i.e. by correlating the test
scores with criterion scores. The procedure involves following steps as
explained in Brown (1970, pp. 109-111).

(1) Selecting an appropriate group to serve as subjects in the study

(2) Administering the test to the designated group
(3) Applying the relevant treatment
(4) Collecting the criterion data and
(5) Correlating the test scores with criterion scores.

The correlation coefficient thus obtained is called validity

coefficient. Usually, the type of correlation coefficient is Pearson's Product
Moment Correlation. The pre-condition for Pearson's Product Moment
correlation is to have the linear relationship between the two variables to
 41

be correlated. Therefore, it is necessary that the test-scores and the

criterion scores are linearly related to each other.

One way of interpreting validity coefficient is comparative, i.e. to

choose a test with highest validity coefficient. Being more valid, it is
expected to be more useful. Validity coefficient can also be interpreted in
terms of percent variance. The percent variance accounted for is obtained
by squaring the obtained correlation coefficient. Thus, if, validity
coefficient r = 0.5, variance = 0.25 = 25%. We can say that 25% of the
variance is shared by the two measures, or that 25 percent of the variance
in the criterion measure is attributable to variations in predictor scores.
The validity coefficient can also be interpreted as a measure of predictive
efficiency i.e. as the ratio of the average criterion score made by the
persons selected by a 'test', to the average criterion score made by same
number of persons selected on the basis of criterion scores. The major
advantage of correlational validity coefficient is the ability to predict the
criterion scores. Given the test scores, the criterion score can be
predicted by using the regression equation y' = a+bx; where y' is the
predicted criterion score, x is the test-score; a is a constant to correct for
the difference between x and y and b is the regression constant.

2.2.4.4.2 Decision Making Accuracy:

Since decision making on the basis of test performance is the

ultimate aim, accuracy of decision is of utmost importance to the decision-
maker. Decision maker's performance can be determined by calculating
the proportion of accurate decisions out of the total decisions made.
When the psychological tests are used for decision-making, the proportion
of correct decisions made gives the index of the effectiveness of the test.
 42

Higher the proportion of correct decisions made, higher is its effectiveness

and validity. To calculate the index of decision making accuracy the
subjects can be classified into groups on the basis of their performance in
test and also on the basis of criterion performance. They can be classified
into two groups acceptable or above average performance and
unacceptable or below average performance. Thus we get the subjects
divided into four groups as follows:

Criterion Unacceptable Acceptable

performance (- ve, below average) (+ ve, above average)

Test Decision
Acceptable A B
(+ve, above average)
Miss Hit
Unacceptable C D
(-ve, below average) Miss
Hit

The index of decision making accuracy is the ratio of correct

decisions made to the total decisions made:

PcTot = B-t-C = B+ C = Total Hits

A+B+C+D N N

Pctot, the index of decision making accuracy is referred as an index

of validity when the validity is defined as accuracy of decision making.
 43

In some situations when the success or failure of those who were

not selected is unimportant, the ratio of successful selected persons to the
total selected becomes more appropriate.
PcPos = B
A+B

A and B are the selected candidates who are unsuccessful and

successful respectively. Pcpos is the index of validity in terms of ratio of
selected persons who are successful to the total persons selected as
described in Brown (1970, pp. 118-120).

2.2.4.4.3 Interpreting Criterion Related Validity or Decision

Making Accuracy Data:
Following factor influence the validity indices calculated in terms of
Decision making accuracy.

A. Sample:
It is important to select the sample which will well represent the
population. Size of the sample also plays an important role. As the size
of the sample increases, errors of measurement tend to counterbalance
each other and the obtained results could be expected to be more
balanced. Therefore, with a larger sample, the results can be more
expected to be statistically significant.

B. Base Rates:
As described in Brown (1970, p. 127) the base rate may be defined
as the rate of occurence of a phenomenon in an unselected population.
Thus, the base rate would be the proportion of people who would be
successful on a job or in an academic program if there were no selection.
In practical situations the selection is always made using some or the
 44

other method applying some criterion, whatsoever it may be. Thus, in

validity studies, the base rates more appropriately refer to the rate of
occurrence of the phenomenon in relatively unsystematically selected
group.

B. Selection Ratio: According to Kaplan (1977, p. 180), selection ratio is

defined as the percentage of applicants selected or admitted. If the
predictor is positively related to the criterion, by being more and more
selective, we can increase the probability that any person selected will
be successful. The effect of selection ratio on selection efficiency has
been presented in Taylor and Russell's table. As the selection ratio
increases, selection efficiency decreases for any validity index.

2.2.5 Job Performance:

Job performance is an important factor, which deals with the quality

and the quantity of work done by an employee posted for a certain job. It's
not only the quality of the work or the quantity of the work done by an
employee but it is the linkage between the two i.e. how quickly as well
accurately one can perform the assigned job. To determine the job
performance of any role incumbent it is important to identify the key
performance areas (KPAs) and various dimensions of the job assigned.
Key performance areas are the critical functions to be performed by the
role incumbent over a given period of time. It is important to define these
categories meaningfully. These functions should specify what the
employee should be doing rather than what results are expected from him.
The key performance areas i.e. important functions to be performed by
incumbent would obviously differ for different organizations. KPAs can be
obtained by an elaborate and extensive job description. However,
 45

broad job-descriptions, may not be helpful in identifying the KPAs as they

do not give the exact and clear picture of the functions involved in a job.

Identification of Performance Areas may involve following steps:

1. A small group consisting of 4 or 5 persons may be formed involving
those who are presently performing the role and those who have
performed this role in the past and are presently supervising this
role. The group may list out all the tasks associated with the role
and then classify them into meaningful categories.

2. Out of the entire list of all the tasks, the group has to identify the
main activities being performed, what is the exact role in performing
these activities, how much time is spent on each of these activities
and rank them in the order of importance to determine the
weightage for each activity. The group may also spell out the
expectations from a role incumbent if his performance is to be rated
as excellent.

3. These activities may be classified into meaningful categories of

functions, getting thereby the KPAs.

4. Assign weightages to the activities as per their importance.

Actual work behaviour i.e. job-performance is likely to result from a

combination of motivational forces, moderated by skills and
abilities. Even, a well executed plan of job-performance is doomed
to failure if workers do not have skill and potential for the job. Both,
performance outcomes and job satisfaction may be affected by
aforesaid factors and may be affecting each other. The model
 46

presented in fig.2.1 depicts a few internal and external motivating

factors which could be determinants of job performance in industrial
organizations as explained in Rao (1984, pp.32-35).

Determinants of Job-Performance

PERSONALITY
VARIABLES K>B
(VALUES, ASPIRATION. ATlSFACnON
LOCUS OF
CONTROL)
fEUCEPTlON OF
JOB SITUATION
JOB (ORGANISATIONAL
SITUATION CtlMATE)

BACKGROUND
VARIABLES
lACC. SEX. MARITAL
STATUS. INCO\fE.
SESi

fig. 2.1

2.2.5.1 Measurement of Job-Performance:

Perfonnance of an employee is usually measured by way of

performance rating scale filled up by the supervisor (superior) or
the reporting authority under whom the employee is directly
working. This being an Important tasks, it Is very important and
necessary that the ratings being given by the rater are unbiased
and are made objectively. To achieve this it is necessary that the
rating scales are designed with utmost care giving minimal scope
for any subjective element. Reporting authority on the basis of
his/her observations about the employees performance fills up
 47

the performance rating proforma. Judgement about an individual's

performance is made on tfie basis of these ratings. The
performance rating scale is usually a five-point or seven-point
scale. The assessment may be made using categories like
"excellent" or "outstanding", "good", "above average", "average",
"below average", "poor" etc. If categories are used they need to be
defined and points may be assigned to them e.g. outstanding = 7,
Very good = 6, Above Average = 5, Average = 4, Below average =
3, Poor = 2 and Very poor = 1.

In one type of appraisal system, the appraisee is asked to write the

self-appraisal report giving details of the activities performed during
the year of appraisal, and submits the same to the reporting
authority. The appraiser (reporting authority) has to study the
report and give his ratings based on his own observations as well
as on the basis of appraisee's self appraisal report. This is followed
by a one-to-one discussion between the appraiser and appraisee in
which the appraiser can share additional information with the
appraiser. In other type of appraisal system, there is no provision
for self appraisal but the reporting authority has to make his
observations and give his ratings on the basis of his own
observations.

Smith and Kendall (1963) have evolved a procedure for developing

evaluative rating scales anchored by examples of expected
behaviours. The format proposed for these rating scales is a series
of continuous graphic rating scales, arranged vertically. Behavioural
descriptions, exemplifying various degrees of each dimension, are
printed beside the line at different heights according to their scale
 48

positions as determined by judgements of those who are expected

to use scales. The behavioural descriptions are intended as
anchors to define the levels of the characteristic, and as operational
definitions of the dimension being rated.

The anchors are defined in terms of behaviours expected to be

shown by the ratee. Ratings are to be made by checking at any
position along the line. This type of rating scales are also called
descriptive rating scales. Depending upon the requirement the
proforma is descriptive or othenwise. In the descriptive rating scale
the levels of performance are described and therefore it becomes
easier for the rater to give his/her observations. The traits on which
the employees are to be observed is decided on the basis of job-
requirement situations. This differs from organization-to-
organization depending upon the activities of the organization. The
performance rating proforma designed for the use in present study
is a combination of descriptive and objective rating scale. Traits
decided and the levels described are decided taking into
consideration the job requirements in banks at clerical level.

2.2.6 Job-Satisfaction:

According to Vroom (1978, p.99), the terms job-satisfaction

and job-attitude are typically used interchangeably. Both refer to
effective orientation on the part of individuals toward work roles,
which they are presently occupying. Positive attitude towards the
job is conceptually equivalent to job-satisfaction and negative
attitudes towards the job are equivalent to job dissatisfaction. Job
satisfaction plays a central role in the study of behaviour at work.
 49

Knowledge of the determinants, the consequences, and other

correlates of job satisfaction are vital. Once an individual joins an
organization, a vector of scores on a well-constructed, validated set
of job satisfaction scales becomes the most informative data. Job
satisfaction is an emotional reaction to a job that results from the
incumbent's comparison of actual outcomes with those that are
desired. General job satisfaction involves components not caused
by the immediate job-situation. One is temperamental or
happiness, another is trust in management. Both can act as
causes, effects or quasi moderators, and each is likely to be related
to co-operative and adaptive behaviour. Since neither can be
changed easily, both should be measured and the extent of their
influences estimated. The management of an organization may
considered changing certain aspects of a job-situation, viz. change
in job-design to improve satisfaction with task characteristics, or
introducing training to improve supervisor's interpersonal skills.
Changes are generally introduced In the hope that improvements in
facet satisfaction will in turn affect broader areas of employee
satisfaction and eventually improve behaviour and reduce costs.
The managements in general are interested in improving job
satisfaction. Greater the job-satisfaction, better is the quality of life,
health and stability. Following points have an important place in
determining the job satisfaction as explained In Cranny et.al (1992,
pp.45-47).

1. Job-satisfaction must be measured reliably.

2. Different results may be expected according to the scales
used.
 50

3 Satisfaction cannot be judged in absolute terms, but involves

comparison.
4. The relationship between person and environment is
interactive.
5. Job satisfaction can be broken down into facets or
components.
6. Each of these facets can be tied to one or more aspects of
work environment and the job.
7. Satisfactions with these aspects are inter-correlated defining
a general overall factor.
8. Some of this general factor can be attributed to relatively
permanent characteristics of the individual.
9. Job level is typically correlated with satisfaction with all
aspects of the job.
10. Community characteristics can account for a large proportion
of the inter-correlation of satisfactions when the sample has
been drawn from a number of communities.

Job satisfaction, as observed by Herzberg consists of two

distinct dimensions - job-satisfaction and job-dissatisfaction.
These are two different states and depend upon factors - hygiene
factors and motivational factors. The hygiene factors viz. working
conditions, salary, inter-personal relationship etc. when present do
not lead to job-satisfaction but when not present they bring in job-
dissatisfaction. Hence they are the dissatisfiers. The motivational
factors viz. achievement, responsibility etc. when present lead to
job-satisfaction. Therefore, satisfaction depends upon motivational
factors while dissatisfaction results from the absence of hygiene
factors.
 51

The job-satisfaction is to be treated as a complex set of

variables, because the workers who are highly satisfied with one
diniension of their job e.g. work-itself, may be dissatisfied with
another dimension say "wages". It is therefore theoretically, as well
as practically useful to consider it as a set of different dimensions.

2.2.6.1 Measurement of Job-Satisfaction:

Job-satisfaction of employees is usually measured by way of

questionnaire. The questionnaire consists of questions aiming to
know about the employee's feeling about the work assigned and
being performed by him/her. The questionnaire is given to the
employee for filling up and indicating his/her feelings about the job.
Another way of measuring job-satisfaction is interviewing the
employee and asking hinrVher the feelings about his/her job. Job-
Description-Index (JDI) is one such questionnaire designed by
Smith, Kendall and Hulin. It gathers information about employees
feelings about the work itself, supervision, people (Co-workers),
pay and promotion. The researcher has used JDI as tool for the
present study.

2.2.7. Multiple Regression Analysis:

Multiple regression is a technique used to predict scores on
a single outcome variable Y on the basis of scores on several
predictor variables, the Xjs. Multiple correlation is the estimate of
the correlation between one dependent variable and two or more
independent variables. When the number of independent variables
becomes more than two, the calculations become too tedious to be
carried out manually. Therefore, the calculations are required to be
 52

carried out mechanically using software packages. It gives the

estiniate of proportion of variance in the dependent variable that is
accounted for by the independent variables.

The multiple regression equation Is used for prediction of a

dependent variable from the knowledge of values of independent
variables. The prediction equation for two independent variables Xi
& X2 and one dependent variable Y is Y = bo + bi Xi + b2 X2 as
mentioned in Harris, (1985, pp.50-51) where bi and ba are weights
and bo is the constant. The equation gets expanded with increase
in number of variables.

2.2.8 Hem Characteristic Curve (ICC)

A valuable way to learn about item is to graph their

characteristics, which can be done with the 'Item Characteristic Curve'
(ICC). It is a curve plotted by taking test scores along X-axis and the
p-value or Itenvfacility index along the Y-axis. It is expected of a
good test-item that more number of high performing candidates
answer the item correctly as compared to the low performing
candidates. Therefore, It is expected that, as the test-score increases
the p-value also increases and the ICC for a good test item should be
rising up as we move from left to right on the X-axis. An ideal ICC is
shovwi in Figure 2.2.
ITEM CHARACTERISTIC CURVE

90-

1 1 70- y^
2 1
^ %
60-
50-
y "/
0 g
0) ra
40-
^,1^
";;^^
S5 30-
c •^ -;;^^^
4) o) 20 -
2 £ —^r-^^^^"^"*^"^
SIS 10-
°' 0.
0-10 11-20 21-30 31-40 41-50
Score Range

Fig. 2.2
 53

2.3 Review of related studies:

A large number of studies were reviewed with a view of

getting insight into the problem. Of these, the following studies
were found relevant to the present study.

2.3.1 Studies related to Job-Performance and Job-

Satisfaction:
(A) Studies conducted by NIBM/IBPS
Deshpande, et.al (1974 pp.417-444) studied - Relationship
Between Job Performance of Clerical Employees and Type of
Selection Test: Objective and Descriptive.

Objectives of the study was "To compare the performance of two

groups recruited at about the same time through two different -
objective and descriptive - examinations". The two groups of
candidates (i) recruited through NIBM objective type tests and (ii)
recruited through traditional descriptive type test were selected as
sample. Data as regards their performance was obtained from their
supervisors through a mixed rating scale - performance rating
proforma. The data were analysed for comparison on different
abilities which are desired of a bank employee.

Findings of the study were - (1) The candidates selected on the

basis of job-related objective tests proved to be not only effective
job performers but they were also equally good or better performers
than those recruited through the traditional type of descriptive
examination. (2) The results of the pilot study also indicated that
candidates recruited on the basis of objective tests possess higher
 54

potential for promotability as compared to those recruited on the

basis of a traditional examination.

The results of the study did indicate the strength of objective tests
vis-a-vis the descriptive papers in clerical selection. However, it
must be borne in mind that the tests have to be related to the
important components of the job and also have to be carefully
constructed. Any objective test poorly constructed or improperly
administered may not prove effective.

Mankidy (1977, pp.64-79) carried out a study - How appropriate

was the clerical selection?: A validation of NIBM clerical selection
Tests.

Objectives of the study were (i) to study the validity of selection

tests for clerical recruitment in banks or how far the selection
had achieved the contemplated and desired objectives, (ii) To find
out whether the selection tests had given to the banks "appropriate"
type of clerical recruits.

The validity of the tests was tested against the criteria "on the job
performance". Findings of the study were (1) A considerably large
proportion of employees recruited through NIBM were evaluated by
their supervisors as very good performers. Even in regard to the
specific abilities, the evaluation was on the same lines. (2) it
confirmed that the selection strategy evolved at NIBM had been
more effective in selecting clerks with higher potentials for
supervisory and management jobs in comparison with the
traditional system of selection.
 55

(B) Other Studies:

Amarsingh (1985, pp. 1069-1070) studied, Correlates of Job

Satisfaction Among Different Professionals.

The important findings of the study were (i) The job-intrinsic

variables such as job-concrete and job-abstract correlated
positively and significantly with job-satisfaction of professionals, (ii)
The job-extrinsic variable including psycho-social, enconomic and
community growth factors was found to be positively related to job-
satisfaction of professionals, (iii) age was found to be positive
correlate of job-satisfaction, (iv) Experience correlated positively
and significantly in case of advocates and doctors with job-
satisfaction while in case of teachers and engineers this correlation
was not significant, (v) Self-esteem was found to be positively
related with job-satisfaction, (vi) high scores on extraversion
affected the job-satisfaction of teachers, engineers, advocates and
doctors negatively.

Chaudhari and Lahiri (1968, pp.41-62) - Carried out a study on

Perceived Job Characteristic as Satisfier and Dissatisfier by Manual
Workers.

The study was based on the work attitudes of 100 skilled blue-collar
workers. Questionnaires were employed to determine worker's
satisfaction with 8 job-context factors and 5 job-content factors.
Satisfaction index for each factor was derived by subtracting the
importance score (need-strength) from the extent to which the
 56

need was met (input). Satisfaction was predicted when the

difference was equal to or greater than zero and dissatisfaction was
predicted when the difference was less than zero. The results
showed that hygienes as well as motivators may contribute to
sources of worker's dissatisfaction. It was concluded that the two
sets of work attitude determinants known as job-context and job-
content factors were not independent of each other as sources of
employees satisfaction and dissatisfaction at least for blue-collar
workers.

Hulin (1968, pp. 122-126) studied - Effects of changes in Job

Satisfaction levels on employee turnover.

It was observed that the turnover rate of female employees of the

company was as high as 30%,' which was higher than other
companies in the area. To find out the reason, the researcher
administered JDI to the 345 female clerical workers. The results of
the scores on JDI indicated that dissatisfaction of clerical workers
was occurring in all areas. Analysis of their responses
substantiated this finding. After the survey study and implementing
changes in the company's policies, this rate came down. Moreover,
there was a significant increase in satisfaction with four out of five
of the job areas.

These findings indicated that the difference in the multiple

correlation was non-significant. Even with a turnover rate of 12% a
significant amount of the variance in individual termination
decisions was attributable to differences in satisfaction.
 57

Kolte & Supe (1972, pp.405-413) studied - Determinants of Job-

satisfaction of village level workers: a test of herzberg's dual-factor
theory.

The study was conducted in Vidarbha region of Maharashtra. The

V.L.Ws were given the schedule in the training class. They were
assured of the anonymity of their responses and rapport was built-
up. The schedule was connected with their job and were asked to
describe. The factors responsible for inducing job-satisfaction and
job-dissatisfaction were presented. It was revealed by the critical
incidents that satisfies, in order of the frequency of incidents in
which they were cited, included achievement, recognition, work
itself, good superior, advancement, money and responsibility
whereas dissatisfiers included, again in the order of frequency,
policy and administration, work itself, interpersonal relations,
working conditions, advancement, salary, lack of recognition and
poor technical supervision.

Thus the study explored the factors associated with both job-
satisfaction and dissatisfaction and found that the factors were not
necessarily distinct and separate. The factors were found to
interact among themselves; the feelings of either satisfaction or
dissatisfaction being determined by the achievement of the aim of
promotion and salary. Thus, Herzberg's dual factor theory was not
supported by the empirical evidence in this study.

Koustelios and Baoiatis (1997, pp.469-475) University of Thessalv

studied - The employee Satisfaction Inventory (ESI): Development
of a scale to measure satisfaction of Greek employees.
 58

A pool of 130 items was collected by interviewing Greek

employees. It was reduced to 83 items. These items were
distributed over the six subscales viz. working conditions (12
itenns): Supervisor (24 items) pay (12 items): Job itself (17 items):
Organisation as a whole (12 items) and Promotion (9 items).

The selection of items for inclusion in the final subscales was

carried out and finally it got reduced to 24 items from 83 items
representing the six refined subscales.

The subjects were requested to indicate their agreement with each

item on a 5-point scale ranging from strongly agree (1) to strongly
disagree (5).

The main purpose of this study was to develop an instrument to

measure employee's job satisfaction based on Greek samples. In
this respect, the factor analytic results were encouraging: a six
factor solution emerged. Further, results from confirmatory factor
analysis confirmed that the six factor model was fairly good.

Kulsum U. (1985, pp. 1090-1097) studied the Influence of School

and Teacher Variables on Job Satisfaction and Job Involvement of
Secondary School Teachers in the City of Bangalore.

The final sample of the study had 586 secondary school teachers
selected on a proportionate stratified random sampling technique.
Tools used to measure the job-satisfaction and job-involvement
inventory were Indiresan's job satisfaction inventory and a job
involvement scale respectively. Winer's Leadership Behaviour
 59

Description Questionnaire (LBDQ) and Lawler and Porter's Job

Performance Scale were used to measure the leadership behaviour
and job-performance respectively. To measure the attitude towards
teaching profession and a scale on teacher effectiveness were
constructed by the researcher. Sharma's SOCDQ was used to
quantify organizational climate.

Findings of the study were - (i) Permanent teachers had a higher

level of job-involvement as compared to temporary teachers, (ii)
Teacher's job satisfaction, teacher effectiveness, teacher's attitude
towards the teaching profession, students size and teacher's
performance turned out to be the significant predictors of teacher's
job involvement, (iii) Teacher's attitude towards the teaching
profession and teacher effectiveness turned out to be the common
predictors of both job-satisfaction and job-involvement.

Roy A (1982, p.560) carried out a study of the General Aptitude

Test Battery (GATB) in respect of its Parts and Aptitudes, and Job
Performance of Clerical and Supervisory Technical Personnel in
Textile Industry.

The investigation was carried out in three textile mills of

Ahmedabad, on 230 clerical and 170 supervisory technical
personnel. Two short performance standard inventories, each
having ten items were used for evaluating the performance of
subjects in the sample. Seven out of 12 GATB were selected and
administered to the same sample. Means and SDs of all the scores
were computed separately for two groups across the different
 60

demographic variables. ANOVA and t-test were applied to test the

significance of differences in means.

The main findings were - In clerical group (i) there were no

significant differences among the mean scores across departments,
(ii) personnel with length of service upto 25 years showed
significantly higher mean scores than those having length of service
from 26 to 35 years, (iii) on some parts of GATB, graduates scored
significantly higher than matriculates while matriculates scored
significantly higher than non-matriculates.

Findings for supervisory technical group were also similar.

2.3.2 Studies Related to Validity and Reliability:

AndrewAS J et.al (1994, pp. 179-184) carried out a validity study of

Bigg's three factor model of learning approaches : a confirmatory
factor analysis employing a Canadian sample.

Sample consisted of 205 students from six high schools; with males
and females in almost equal proportion. In order to test the
validity of the three factor model of learning approaches, structural
equation modelling techniques were employed. The present
problem required the use of confirmatory factor analysis. The
overall fit of the data to the model was very good resulting in a
comparative fit index of 0.97 based on 11 degrees of freedom. As
required for the test of the model, all three latent variables were
maintained in an orthogonal configuration. The lack of correlation
between the three latent variables and the resulting good fit of the
 61

data to the model, support the theoretical orthogonality of the

endogeneous factors.

The results of the study strongly supported Biggs three factors

model underlying the Learning Process Questionnaire. The major
results indicated that (1) the overall fit of the model to the data was
very good. (2) The three basic factors were clearly identified and
were orthogonal and (3) particular parts of the model did not have a
good fit.

Gokhale (1981, pp.149-153) studied Validity of tests used in the

selection of spearhead teams.

The sample consisted of 28 persons who were finally selected for

the job. They were in the age-group of 21-28 years and were from
rural background. At the end of six months training in rural
background they were rated on a Three-point scale on Six
dimensions. The test-retest reliability of the scale was 0.8.
Correlation coefficients between these six dimensions and the
dimensions underlying tests were calculated.

It was found that the tests used were appropriate for predicting only
some measures of job performance. In fact traits viz. hard-work,
convincing people and planning had very poor values. The test
battery, however, needed to consist of tests which could be
sequentially given so that the desirable traits could be held
common and other cognitive tests then administered to assess the
planning and other abilities of the applicants.
 62

Puhan (1978, pp.95-100) studied, Psychometric Invariance in

Reliability and Validity Contexts Psychometric invariance as a
concept and as a requirement of psychological tests, was relatively
new. Its implications, usability, and overall suitability as a scientific
concept in psychometrics, would largely depend on its meaningful
relations with other existing requirements of psychological tests,
viz. reliability and validity. Therefore, the basic purpose of the
study was to facilitate logical connections between the concept of
psychometric invariance and the concept of reliability and validity. It
could be concluded that psychometric invariance of a test were far
more useful than those of traditional reliability and validity
assessment of a test. A psychometric invariance assessment
across time would examine the stability of factor loading patterns of
that test across two different occasions. This would represent a
case of test-retest reliability. Such an invariance assessment could
also be used to test the validity. Psychometric invariance
assessment simultaneously evaluated reliability and validity of a
test in terms of its factor loading components.

2.3.3. Studies Related to Construction and Stadardization of

Tests:
Bhatt. G.C. (1981, pp.477-478) studied Construction and
standardization of verbal Reasoning Test for the students studying
in Grades VIII and IX of secondary schools in Saurashtra Area.

The sample consisted of 5,449 students selected from ninety-six

different schools of sixty-two different places of Saurashtra region
by using the technique of stratified random sampling. Initially the
 63

test consisted of 200 items for pre-tryout. After the item-analysis,

134 items were retained for the pilot test and divided into two forms.
The final form of the test consisted of sixty items.

Descriptive statistics like central tendencies, SD and Skewness

were worked out. Percentile scores, standard scores, T-Scores
and Stanines were developed. Reliability was established by test-
retest, split-half and Kuder-Richardson formulas 20 and 21. The
reliability coefficients obtained by these four methods were found to
be 0.82, 0.93, 0.91 and 0.82 respectively. Validity of the test was
established by correlation with intelligence tests, aptitude tests like
abstract reasoning, numerical ability and verbal reasoning test.

Conclusion: (i) The means of boys and girls of Grade IX were

higher than those of Grade VIII. (ii) The means of boys were higher
than those of girls in Grades VIII and IX in the total sample and (iii)
Urban and rural area differences were observed only in the case of
the Grade IX sample.

Deb Maya (1966, pp.73-76) studied, Standardisation of a Group

Intelligence Test Earlier days Binet Scale or its adaptations were
commonly in use as intelligence tests. However, as these tests
being of individual type could not be administered to a group at a
time. Shri S K Bose and Shri S C Datta from Calcutta University,
constructed a test comprising questions in the areas (1)
Arithmetical Reasoning, (2) Verbal Reasoning, (3) Vocabulary, (4)
Direction, (5) Number Completion and (6) Analogy. It was a speed
test. Subsequently, the author of this paper undertook the
standardisation of the test.
 64

The test had an uniform procedure of administration and scoring.

For each item there was only one predetermined correct answer.
No personal equation of the administrator could alter the test-score.
Thus the test could be considered as an objective test.

The scores were found to be normally distributed with the Mean

score of 35 and SD of 8.7 having range of 10 to 59.

Reliability was tested by the split-half method. Correlation

Coefficient with Spearman-Brown formula was found to be 0.92,
hence highly reliable.

The test was earlier validated against academic achievement. The

correlation was found to be 0.38. (significant at .01 level), which
was statistically significant. Later on the test was validated against
a standard intelligence test, Raven's Standard Progressive
Matrices. The correlation was found to be 0.67 (significant at .01
level). Thus the test could be considered as valid.

Mishra,(1985, p.354) studied the Construction and Standardization

of a Verbal Group Test of Intelligence in Oriya for the Age Group
12+to 15+.

The item areas of the test were verbal analogy, verbal reasoning,
vocabulary general information and numerical relations. The test
was standardized on a sample of 2000 boys and girls chosen on
a stratified random basis. Split-half, test-retest and other
reliability coefficients were calculated. The inter-item correlation
 65

and factor analysis with varimax rotation were used for a study of
validity of the test.

The study resulted in developing a verbal group test of intelligence

in two parallel forms. The test had 5 subtest areas of 50 items and
required 30 minutes for administration in the classroom situation
using answersheets.

The reliability indices were high in the range of 0.73 to 0.92; and
the validity coefficients in the range of 0.52 to 0.73. The factors
identified through factor analysis were general reasoning and
verbal comprehension.

Pillai K K (1978, p.500) studied The construction and

standardization of a Verbal Test of Intelligence in Tamil (for the age
group 10 +to 15+)

For pre-tryout a sample of 100 students was used. For further try-
out a sample of 750 students selected from three schools in
Chidambaram was used. For final administration, 5000 students
from 34 schools in one of the districts in Tamilnadu were selected
by the method of stratified proportionate sampling. The test
used included seven sub-tests: Synonym, antonym, analogy,
classification, mixed words, reasoning (Verbal) and reasoning
(numerical). A total of 110 items were there in the test.

The coefficient of reliability computed by test retest method was

found to be 0.84 and that by the split half method was found to be
0.88. The content validity was considered on the basis of various
 66

types of behaviours assessed by the sub-tests. Norms were determined

in respect of the total sample, grades and age groups.

2.3.4 Study based on Factor Analysis:

Nalr, K.S. (1972, p.493) carried out An Analytical Study of the Factor
Pattern of Verbal and Non-Verbal Tests of Intelligence.

The sample for item analysis consisted of 370 students and the sample for
final analysis consisted of 420 students chosen from the secondary
schools of Trivandrum educational districts.

Correlations among the sixteen variables were computed using the

Pearson's Product moment method of correlation.

(a) Verbal and Non-Verbal tests were formed mainly on the basis of
content.

(b) A third factor, identified as numerical ability, was the same as the one
identified by others.

(c) A fourth factor which showed possibilities for tests to be grouped on its
basis, would have emerged if more tests had been used.

(d) Factor I which had high loadings on the tests of analogies, series,
spatial relations, classifications, water-reflection and arithmetic
reasoning could be identified as Non-Verbal factor.
 67

(e) Factor II which had high loadings on vocabulary tests as well as water
reflection which was classified as a Non-Verbal item could be termed
as a Verbal factor.

(f) Factor III which had high loadings on arithmetic reasoning tests,
number series and number classification, could be termed a numerical
reasoning factor.

2.4. Implications of the Review:

A thorough review of the related studies and literature was of great help as
it gave information about the studies already carried out by various
researchers in this area.

2.4.1. Implications of Review of Literature:

While reviewing the literature on validity and reliability, details of
various methods and their relevance in different situations was studied. It
was observed that the product moment correlation for calculating validity
coefficient was the most widely used method. The concept of 'Criterion
Related Validity' as 'Decision Making Accuracy' was found to be most
relevant for present study since most of important part of the selection
process in banking industry is the decision making accuracy in selection of
personnel based on the objective type tests. Therefore, it was decided to
use this method for calculation of validity coefficient, though it was not
used by the researchers of the studies under review. KR-21 formula was
used for establishing the reliability (internal consistency) of the tests.
Multiple Regression Analysis was used for estimating the Job-
Performance and Job-Satisfaction on the basis of performance on
objective type tests. Literature on Job-Performance was useful for
designing the performance rating proforma for the present study, as it
 68

gave information about various traits on which the performance should be

rated in different situations and also about the types of scales that could
be used in different situations. Literature on Job-Satisfaction gave
information about how to measure job-satisfaction, its importance in job-
situation, the elements considered while designing a tool for measurement
of Job-Satisfaction and also the various tools which were readily available.
Job Description Index i.e. JDI, ^ tool designed by Smith et.al. was
choosen for the present study since found relevant for the situation.

2.4.2 Implications of Review of Studies:

2.4.2.1 implications of Studies Related to Job Performance and Job-

Satisfaction

(A) Implications of Studies conducted by NIBM/IBPS

The study of Deshpande et.al (1974) and Mankidy (1977) had given
the information about the types of tests being used at the inception of the
system, work done by NIBM in relation to testing as selection strategy for
bank recruitment. It answered the question like why more importance was
given to the objective type tests as Compared to the 'traditional' descriptive
type tests and how the selection system was evolved. Since the present
study basically dealt with the present selection strategy for bank
recruitment and was an effort to Establish the validity of objective type
tests, the above information was very much required and useful.
Mankidy's study also threw light on how the validity of tests was
established against the criteria of job-performance, tools and traits
selected for measuring job performance. Present study was an extension
 69

of these studies In the changed scenario. Review of these studies

confirmed the fact that though there were changes in patterns of tests
used by NIBM/IBPS, no such study was undertaken in the past more than
20 years.

(B) Implications of other Studies:

Charles Hulin, had used JDI for measurement of job satisfaction

while studying reason for high turnover rate among female employees.
This study told us how the job satisfaction or dissatisfaction affects an
employee and areas of satisfaction covered by the JDI. Job satisfaction
was one of the criteria used for establishing validity in the present study
and the JDI was the tool used for measuring job satisfaction, Hulin's study
was useful in obtaining the information particularly about the effect of job
satisfaction or dissatisfaction on an employee. Studies by Kolte & Supe
(1972), Koustelios & Baoiatis (1997) and Amar Singh (1985) provided
information about the factors which determine the job-satisfaction of an
individual and the types of tools that could be deviced and used for
measuring the job-satisfaction of professionals in different areas at
different levels. This helped in deciding the factors relevant for the
present study and selecting the tool for measurement of job-satisfaction.
The study by Kulsum (1985) dealt with the relationship between various
factors and their effect on job-satisfaction and job-involvement of teachers
from different strata. It involved measurement of job-satisfaction, job-
performance and job-involvement and establishing inter-relationship
among them. It helped by giving insight into how to measure various
factors using different tools and how to establish the relationship; what
were the factors affecting job-satisfaction and job-performance.
 70
Study of Roy A (1982) dealt with finding out the factors that affect the job-
performance. This threw light on what were the important traits on which
the performance should be assessed and about the tools for measuring it.
Study of Chaudhari & Lahiri (1968), dealt with the analysis of job elements
that act as satisfiers and dissatisfiers for blue collared workers. Thus, it
gave us information about the factors determining job-satisfaction of the
employees.

Studies on job-satisfaction and job-performance provided information

about the factors working as satisfiers and dissatisfiers methods
and tools for measurement of job satisfaction and job-performance and
development of scales for the purpose which was very much useful for the
present study.

2.4.2.2 Implications of Studies Related to Validity and Reliability:

Gokhale (1981) studied the validity of tests with a purpose of

predicting occupational success and finding out the utility of tests in
predicting various traits of job performance. This study along with other
studies on validity and reliability threw light on different methods of
establishing reliability and validity. Since the purpose of the present study
was to establish the validity and reliability of the tests and ultimate
purpose of the tests was predicting occupational success, these studies
were helpful.

2.4.2.3 implications of Studies Related to Construction and

Standardization of Tests:
Studies by Bhatt (1981), Pillai (1978), Mishra (1985) and Maya Deb
(1966) deal with the construction and standardization of objective type
tests in different areas as such Test of Reasoning (verbal and non-verbal).
 71

Arithmetical Reasoning, Vocabulary, Test of intelligence, etc. They

explained various methods of establishing reliability viz. split-half, test-
retest, etc. Inter-item correlation, factor analysis for study of validity,
methods of calculating correlation coefficient viz. Speamrian-brown
correlation coefficient, product moment correlation, etc. It also provided
the ranges of coefficients of reliability and validity throwing light on
expected or usually acceptable ranges of these coefficients.
Standardization forms an important part of test-construction and involves
ensuring objective administration and scoring of tests, establishing
reliability and validity and working out norms. The present study dealt
mainly with the objective type tests on Reasoning (verbal and nonverbal),
English, Numerical Ability and Clerical Aptitude (speed test) it undoubtedly
involved these procedures and therefore the above studies were very
much relevant and useful.

In the various research studies reviewed, not a single study was

found in which the effect of change in item position on the performance
was studied. The present study had considered this additional aspect of
the test administration and studied the effect of change in item position.

2.5 Epilogue:

A thorough review of the related literature and research studies was

undertaken with a view to obtain information about the methods available
and used in different situations and their suitability for the present study.
The review of studies provided information about the research work
already done and the research work which was not done so far but was
necessary to be done.