What to look for in a psychological test

Tests are designed for a purpose and the use of

a particular test will vary according to the
objectives of assessment.

Some broad distinctions between different

categories of tests
can be made
 are used to assess knowledge and skills acquired
through education and instruction.
 mathematics knowledge,
 foreign language proficiency
 mastery in a craft
Such tests tend to be narrowly defined in content and
targeted at the achievement of specific standards.
Like ability tests, such tests are generally designed so that
there is only one correct answer to each test question. The
test score is usually the total number of questions
answered correctly.
 assess broader areas of what a person can do. While scores on
such tests are influenced by education and training, they are not
designed to assess specific areas of knowledge or skill.
 measures of verbal reasoning
 (the ability to comprehend, interpret and draw conclusions from oral
or written language),
 numerical reasoning
 (the ability to comprehend, interpret and draw conclusions from
numerical information),
 spatial reasoning
 (the ability to understand and interpret spatial relations between
objects)
 Mechanical reasoning
 (understanding of everyday physical laws such as force and leverage
involved in the use of tools and equipment).
 used to assess how well an individual is likely
to perform in a training programme or in a
job.

 Attainment tests, ability tests and personality

tests are all used to predict future
performance, and so the term aptitude has
more to do with prediction than with a
specific category of test.
 which are used to assess how a person is likely to
react emotionally to situations and other people,
the types of relationship they prefer to establish
with others, and the motivational factors that
influence a person’s actions.
 In contrast to attainment and ability tests, tests
of disposition do not generally contain questions
to which there is only one correct answer.
Rather, the answers given to questions reflect a
tendency to interpret situations or respond to
other people in particular ways.
 Typical qualities assessed by such tests are
anxiety, sociability, perseverance,
dominance, fear of failure and resistance to
stress.

 Personality tests are the most widely known

form of this type of test.
 are similar in their design to tests of disposition, but
focus on the activities, hobbies and types of work that
a person might enjoy or might be best suited for.

 They are frequently used in

 careers counselling to gauge priorities in life (career,
family, life style)
 orientation towards work (service to others, invention,
entrepreneurship) in order to help focus a person’s
thinking
 to suggest possibilities the person may not have
considered previously.
 are among the most complex form of psychological
test in dealing with areas that are both sensitive and
difficult to diagnose.
 They are also among the most diverse group of tests
in covering a number of conditions and symptoms,
and their use requires both general clinical expertise
as well as specific knowledge of a particular test.
 They include assessments of neuropsychological
damage resulting from physical trauma or from
pathological
 conditions.
What does it measure and how does it measure it?
 Rationale behind a test
 a description of why the test was developed
and why it was constructed in the way it
was.
 Contains such information in the form of a brief history of
the test.

 This history should include any relevant theory supporting

the test,
 the steps taken to construct the test,
 details of research and
 summaries of the results of such research.
 The manual should also state
 whether the test was designed for a broad,
general range of uses, or
 whether it was designed for use with
specific groups of individuals (e.g. ages,
occupations, types of condition, as an aid to
specific diagnoses or decisions).
 With a statement of what the test is
supposed to measure, we can then look for
numerical evidence of how successful the test
construction process has been.
Some of the general quality checks
that should be reported in a test
manual.
 When a test is administered, the outcome
is an observed score on the quality
measured by the test.

 However, the observed score may be

misleading without information on how
successful the test has been in measuring
that quality accurately.
 Two types of score need to be considered: the
person’s true score on the quality measured,
and how closely the observed score matches
the person’s true score.
 Statistical indices known as reliability coefficients
provide the evidence required to judge how
accurately or precisely a person’s true score has
been estimated by a test.

 Reliability coefficients take the form of a proportion,

and a value of +0.7 is generally held as a minimum
requirement for the use of a test. Such a value
states that 70 per cent of the differences between
people as measured by a test are due to true
measurement, while 30 per cent of such differences
are due to measurement error.
 Of course, designers should seek reliability
coefficients greater than +0.7 in constructing a
psychological test.
 The test manual should state
which type of reliability coefficient
is reported and why.
 Internal Consistency
 Evaluates the extent to which the questions in a
test are consistent in their contribution to the
accuracy of the overall test score.

 Low internal consistency would suggest that the

test contains poorly constructed questions or
questions that measure qualities other than that
intended in the design of the test.
 This type of reliability coefficient is appropriate when
the test is made up of discrete questions which are
each answered independently.

 It is not appropriate when the items in a test are

ranked one against the other as in some personality
and interest inventories, or when the score is based
purely on the speed of responses or continuous
performance on a task as in tests of clerical speed or
hand-eye coordination.
 Alternate Forms
 Evaluates the extent to which the ordering of
people from higher to lower test scores is
consistent across two or more versions of a test.
 If a person is in the top 10 per cent of people
tested on Version A of a test, then one
would reasonably expect that person to
score in the top 10 per cent on Versions B or
C of the same test.
▪ This coefficient is inappropriate unless there is an
explicit statement that different tests are
alternate forms or versions of the same test.
 Test-retest or Stability
 Evaluates the extent to which the ordering of
people from higher to lower scores is consistent
across time. If it is expected that the quality
measured by a test is unaffected by factors
associated with time, then a person‘s standing at
Time A should correspond to their standing on the
same test at Time B.
 The stability coefficient is not appropriate if
change over time is an important feature of
the quality measured by a test.

 An example would be a measure of mood

bwhich could vary day-today or hour-to-
hour.
 While reliability is concerned with how
accurate or precise a test score is, validity is
concerned with what the test score actually
measures.
 The observed score is rarely used to interpret the
information from a psychological test.

 Test score is usually transformed into

 norm scores (in the case of tests of ability or
disposition)
 criterion scores (in the case of tests of attainment or
dysfunction). To be able to interpret these
transformed scores, a test user must understand the
process by which these scores are arrived at and what
they represent.
Many tests of disposition and interest
generate several scores rather than
one single score. Accurate
interpretation of these scores depends
on understanding the pattern of
relationships between them.
 The test manual should clearly state the
procedures for creating transformed scores,
why these procedures were chosen, and how
these transformed scores are to be
interpreted.
 It is possible that factors such as sex,
ethnicity or social class may act to obscure,
mask or bias a person’s true score on a test. If
this is the case, the observed test score may
not be an accurate or valid reflection of the
quality assessed through the test.
 This has been a concern of test designers for a
considerable time, and an entire body of
psychometric research has been devoted to
developing methods for evaluating whether a
test score is biased against different population
subgroups.

 Test manuals should state whether the test has

been evaluated for potential bias, what methods
have been used to carry out such an evaluation
and the results obtained.