Matrix Tests

By JOHN C. RAVEN, [Link].

Research Department, Royal Eastern Counties Institution, Colchester

Methods of Measuring Mental Ability

The simplest method of assessing mental ability is, of course, to ask questions,
award marks and to compare the marks gained by one person with the marks gained
by others. Its simplicity, range of application and possible refinements are the
chief assets of this method and its chief defect is that the marking, however
conscientiously it is carried out, remains arbitrary. The persons examined are really
classified according to their ability to please the examiner and though this involves
mental ability, training is also essential.
A more reliable method of assessing mental ability is to use a standard group of
problems which persons of average ability, without previous training, are just able

j-
 MENTAL HEALTH 11

to solve at a
given age and of such a nature that persons of more than average ^
are able to solve them at an earlier age; while those of less than
average ability
if they solve them at
all, succeed only at a later age. Such a series of problems, ea
of which can be marked
simply as solved or unsolved, providesa means o
a person's mental development at any given age and thence the rate at w ic asse^*
ev
ment is taking
"
place. This method leads naturally to the conception of
"
ages and intelligence quotients". It has the merit of showing t e ex en
which a bright young child and an older duller child are similar
mentally and, at the
same time,
the way in which they are fundamentally different. rovi e a
is neither Pf
deliberately coached nor abnormally prevented from acquiring su C1?
general information, mental
age increases steadily with chronologica age w le
intelligence quotient remains relatively constant throughout life. e me o
therefore considered a means of measuring innate ability. It has two serious 1
advantages. The first is that one can give no satisfactory definition o * e a 11
measured. The second is that one can never be sure that testees have had
equa
^
opportunities for acquiring the necessary general information. The t
ings o
everyday life vary from place to place and so do language and educationa aci i ies.
To overcome the
difficulty the problems used have been made numerous an
varied, but whilst this decreases the effect of any one unsuitable test it increases
the possibility of unsuitable
tests occurring and obscures their existence. For" t is
reason other methods of
measuring mental ability are used. Some tests em o y
a
period of practice before the marked test
begins, other tests are designed to e
concerned with things with which
every person, in the author's opinion, should
be familiar,
or, on the contrary, with material which no one is
likely to have
encountered. To ensure that the
ability measured can be clearly defined some tests
are made
entirely verbal, others non-verbal; some are designed to measure the rate
of mental
work, others measure only the accuracy of thinking.
Each method has its
advantages and limitations. The superintendent of an
institution finds mental
ages convenient when grouping defectives for supervision.
A teacher finds tests
of school attainments useful when
classes. When individuals are placing scholars in suitable
to be selected for special
tests of innate training or occupation,
ability and specific aptitude are needed. But a test is useful only
in so far as its
reliability is known and the ability it measures is definable.
Tests of Eduction
The work of Spearman (1927) may appear an abstract statistical controversy
" " "
concerning g and s but his great contribution to practical mental
is his clear enunciation testing
of the laws of "
eduction or
" "
He has
noegenesis .

shown that the whole qualitative

development of creative mental activity is a con-
tinuous process of?
1. Apprehending the characters of
experience,
2. Educing relations between the
characters apprehended, and
3. The creation of correlates
bearing specific relationships to experienced
characters.
 12 MENTAL HEALTH

The more mental is studied, the more clearly these three phases of
activity
eduction, interacting the
one uponother, are seen to be the characteristic qualities
" "

experienced in all mental processes which are in any way rational original
"adjustive", or, in ordinary language, intelligent, as distinct from those which are
purely "instinctive", "habitual", "reproductive"; in other words, repetitive.
It follows that what is needed is a good test of eduction andgood tests of repetitive
ability.
To determine the best tests of eduction, testees have been trained to educe
relations of different complexity (Spearman, 1927). The causes of failure and
conditions of success have been studied. It has been shown that verbal (Stephenson,
1931) and performance tests (Alexander, 1935) are influenced by processes other than
eduction while visually presented meaningless figures provide the most satisfactory
means estimating a person's innate eductive ability. Perceptual tests, such as
of
those shown in Figs. 1 to 4, may appear useless artistic stunts or obscure mathematical
problems but, upon investigation, success in solving them is found to depend upon
the ability for logical thought which is the essential factor in all intelligent conduct.
According to the type of problem employed and the method of testing adopted such
"
tests can be used to measure either the rate or the accuracy of eduction. Progressive
Matrices "* is a series of such tests designed to measure the accuracy of eduction.

Progressive Matrices

Progressive Matrices consists of

sixty perceptually presented tests. Each test
" "
consists of a design or matrix from which part has been removed. The testee
has to examine the matrix and decide which of the pieces given below is the right one
to complete the matrix. Twelve tests complete a Set and there are five Sets, lettered
A to E. Each Set develops a different theme. The initial tests in each Set are easy
so as to be self evident and these are followed by tests of increasing difficulty; the
order in which they are presented provides the necessary training.
Three forms of Progressive Matrices are in use. In the published form the tests
are printed in black on white paper in a booklet. This is intended for use with
ordinary people over six years old. The testee indicates his choice by pointing to
the piece so that the psychologist can record his choice, or by writing down the number
of the piece he selects on a scoring form.
For young children the Board Form of the test is used.j This consists of
Sets A and B only. Each Set is in an attractive box and each test, drawn in bright
colours, is in a separate folder. The matrix is on the upper part of the folder and
" "
has a real gap to be filled, the bits are in the lower part of the folder and the

child picks them out and puts them into the gap. Seeing the result of his choice
trains the child much better than verbal instructions. To ensure that colour
discrimination does not affect success, dark patterns are used on a light ground.

*
H. K. Lewis & Co. Ltd., London. (Sets A, B, C, D and E.)
t Board Form of Sets A and B, obtainable from the Research Dept., Royal Eastern Counties
Institution, Colchester.
 MENTAL HEALTH 13

<?> <?>
i> ffl
?D
Example 1 Example 2

O-o Q> @
0 Im
Aa A

A Za) A A

Example 3 Example 4

From article published in Brit. Journal Med. Psychol., 1936

14 MENTAL HEALTH

To prevent manipulation, the gap and the insets are asymmetrical. The advantage
of the Board Form of the test is that it is almost irresistible to young children. The
disadvantage is that it is difficult to make and therefore expensive.
For special cases a portfolio form of the test is used. In this form the black
and white prints are cut out and mounted on grey paper. The testee indicates
his choice by any gesture of which he is capable, the psychologist recording
his choice.
The procedure is simple. All that is required is a quiet room, a table at which
the testee can sit comfortably, the most suitable form of the test, and a skilled and
experienced psychologist. The psychologist watches the testee build up a system of
thought under standard conditions, notes the degree to which he is successful and
the nature of his errors, guides his attention with as little comment as possible, and
ensures that errors, when they occur, are genuine failures of eductive ability. An
average child of three is well able to solve the initial problems of Set A and the
complete series of sixty problems presents difficulty to quite able adults. The whole
range of eductive ability is covered and from the results the psychologist can class
any testee according to where his score falls on a percentile scale. The reliability
of the testee's total score is indicated by the scores on the component Sets; the time
taken and the nature of the testee's errors are indicative of temperamental and
emotional traits.
Under the guidance of an experienced psychologist one of these forms of the
test can be used with almost any testee. It has been used with normal children from
three to fourteen years old, with mentally defective children and physically defective
children including those with partial sight. Being independent of language,
satisfactory results have been obtained from deaf children and foreign children
(Spanish refugees). It is equally applicable to adults.* An embossed form of
Matrix Test for use with blind subjects could be prepared.
For clinical purposes the test has distinct advantages. So little need be said
or done that children's tears and antagonisms are circumvented. The testee finds
himself able to succeed and the co-operation of even psychotic patients is secured.
"
*
Writing in Mental Welfare (April 1939) Professor Burt says: Of those (intelligence tests
for adults) at present available none is wholly satisfactory."
"
Matrix Tests, which are, of course, "judgment tests, have been given to only 150 adults.
" "
The results, however, have been uniformly satisfactory. Adults have not been resentful or
" "
embarrassed by any resemblance between the test and school examinations. Their comments
were: "I'm doing these all right" (Feeble-minded adult); "They make you think, don't they?"
" "

(Average adult); "You've got me beat this " time (Sergeant-major); I'm staking on number
" " "
six (Officer); What is the answer to E 8 ? (Student); What I like is the way your test grips
" "
me (Psycho-analyst); I'm sorry. I can't say how long 1 took. I was interrupted and did the last
" "
few in my bath (a lady). A solicitor's opinion" was that they were problems of pure logic ".
An artist considered them mainly questions of good design ". Both obtained "
high scores but
while the solicitor apparently based his arguments on what Burt describes as the explicit step by
step inference of the logician the artist apprehended the solution by what Burt describes as a
" '

complex synthetic activity, comparable to what is popularly described as intuition ', whereby
we implicitly comprehend the intelligible character of a whole, without explicitly analysing it into
its component parts or distinctly formulating their relations." Testees frequently perceive a matrix
of relations as a whole without clearly perceiving that explicit relations exist between the individual
"

figures. It was for this reason that the writer liked the name Matrix Test ".
 PROGRESSIVE MATRICES
boards for little children
In the form of hoards
 MENTAL HEALTH 15

Score Comparisons
For experimental work sixty-five problems were used. In Table I the mean
scores for interesting groups of adults and children are shown. The scores of the
university students showed little correlation with their individual scholastic attain-
ments, but the difference between the mean score for the group of students and the
mean score for the
group of soldiers is significant. The Headquarter Staff of a
regular Infantry Battalion was tested. The scores are approximately wnat might
be expected from a
group of average adults and there appears to be little growth
of eductive
ability after the age of thirteen. There is, however, a characteristic
difference between the average score for a child of thirteen years and that of a child
?f eight years.
Usually during its ninth year a child begins to solve problems similar
to those shown in
Figs. 3 and 4 in which it is necessary to reason by analogy.
Apparently higher thought processes begin to mature, and the fact, that children over
nine years need
problems which are unsuitable for children under eight years of age
made the construction of a continuous series of tests
extremely difficult.
Table I.

No. tested. Mean score. Standard Deviation.

UniversityStudents 24 54-3 4-7

Soldiers 44 44-2 9-6
Children 13-14 years 178 42-5 10-3
Children 8-9 years 53 20-6 7-0
Feeble-minded adults 25 19-8 5-2

The average score for a

group of high-grade feeble-minded adults is approxi-
mately equal to the score of the average child of eight years. If the simple ament
is given
time, and especially if he is allowed to work by trial and error, he may, like
the average child of
eight years or less, acquire skill in solving problems similar to
those shown in Figs. 1 and 2, but in solving
problems of the type shown in Figs. 3
and 4, where it is
necessary to reason by analogy, the ament remains, throughout
life, characteristically incompetent. The thought processes which normally begin to
appear during the ninth year fail to mature and it is probably just the inability to
reason by
analogy which renders the majority of mentally defective adults incapable
of "

'
managing themselves and their own affairs and mentally retarded children
incapable of receiving proper benefit from instruction in ordinary schools .

Boys attending London schools for the mentally defective were tested. The
mean standard score for those
making no progress in school work was found to be
?2'5a-*, while the mean standard score for those who were making progress was
?1 5cr.
?

Some boys over thirteen years of age, although seriously retarded in school
attainments, obtained relatively normal Matrix Test scores and were clearly able to
reason by
analogy. It was interesting to find that the percentage of boys able to
reason by
analogy before leaving the school agreed with the percentage of boys
"who had retained
regular employment after leaving the school.
*<r is
simply a statistical unit of measurement; + if above normal, if below normal.
16 MENTAL HEALTH

Physically defective children were tested. They were first classified according
to the nature of their ailment and then sub-classified according to whether their
teachers considered them normal in school work, backward due to loss of schooling
or retarded due to mental dullness. The results are shown in Table II. The
teacher's ratings and the test scores both show that the association of mental dullness
with congenital abnormalities is greater than its association with acquired diseases
even when neurological abnormalities of whatever origin are considered as a separate
class. The result is interesting because it shows the agreement between the test
scores, the teacher's ratings, and the findings of other investigators (Dawson, 1931).
The teacher's ratings show a high incidence of backward and dull children, but the
mean standard score for the whole group is, as it should be, just normal. The mean
score for the children classified as mentally dull is distinctly below normal but the

mean for those considered backward but not dull suggests that, as a group,
score

they are evenslightly brighter than those considered normal in school work. These
findings are probably correct; in general a backward child has to be brighter than a
child of average attainments if it is to impress the teacher that it is backward but
not dull, and the No child classified as backward obtained a test
test showed this.
score significantly On
below normal. the other hand one child classified as dull
obtained a score of+2o*and enquiries showed that the child was genuinely intelligent,
but extremely backward.
Table II.

Physically Defective Children

Normal. Backward. Dull. Total. Mean <x score.

Intercurrent Disease .. 32 43 + 0.35

Neurological Conditions 15 23 + 0-1
Congenital Abnormalities 15 25 + 0-05

Total 62 15 14 91
Mean Standard score + 0-4 + 0-5 1-0 -0-2

A revised and standardized series of sixty matrices and the Terman Merrill
Revision of the Binet Scale was given to 131 children referred to a child guidance
clinic; 57 were sent for examination before emigration and 74 were referred to the
clinic on account of psychological difficulties. The children examined before
emigration from all parts of the British Isles and were chiefly orphans of good
came

average mental ability. The children referred on account of psychological difficulties

contained eighteen cases of school failure, twenty-four cases of anti-social conduct,
twenty-four cases of unsatisfactory habits such as enuresis, and eight cases of emotional
abnormalities, fears, etc.
For comparative purposes Terman I.Q.s and Matrix Test scores were each
converted into percentile ratings. According to both scales the group contained a
rather low percentage of children of average ability, and rather high percentages of
exceptionally dull and exceptionally bright children.
 MENTAL HEALTH 17

The correlation between the two test ratings is shown in Table III. There is
considerable agreement between Terman and Matrix Test classifications, but in
eight cases the ratings differ by more than one class. The case notes of these eight
children are striking.
Three children obtained Terman I.Q.s of over 130 but only average Matrix
Test scores. All three had attended secondary schools, had proved failures,
and had reacted by anti-social behaviour. All three showed exceptional verbal
fluency.
One child, examined before emigration, obtained a Terman I.Q. of 96, but
from his Matrix Test score he appeared intellectually defective. Unfortunately
no case notes are available.
The remaining four cases were all children under 10| years of age who were
about to emigrate. Their Matrix Test scores indicated that they were of superior
mental ability but their Terman I.Q.s ranged from 90 to 109. One child had
recently come from Cornwall, another from Scotland while a third was Irish.
All four were slow in
following instructions and concerning one child the writer
received these notes. His mother hates him and told me that she would do
"

anything get
to rid of him and hoped to have him classed as subnormal
. . .

mentally so that she might have him put away permanently?she tried to produce
this condition in him him shut up in a bedroom. He was not allowed
by keeping
to go to school .
any normal child subjected to the treatment this boy
. .

has received would show the effects of it."

Table III.

Terman Percentile Class. Matrix Percentile Class.*

Totals.
D C B

[Link].
128 and over
112 to 127
89 to 111
73 to 88
72 or less

Totals 14 16 52 30 19 131

A testee's
ability is estimated by comparinghis score on the Matrix Tests with the scores
obtained by other testees of the same age. Thus he can be classed as :
A?"
Intellectually superior " if his score exceeds that of 95 per cent, of the testees of his own
age group.
B?"
Definitely above average " if his score exceeds that of 75 per cent.
C?" Average "
ability if his score lies between that obtained by 25 to 75 per cent.
D?"
Definitely below average " if his score is exceeded by 75 per cent, of testees o is age
group.
t" "
Intellectually defective if his score is exceeded by 95 per cent.

Commentary
To meet the difficulties and
shortcomings experienced when using other mental
tests a series of Progressive Matrices has been
prepared and standardized. It is
18 MENTAL HEALTH

"
independent of language and training but is in no sense a performance test
It measures eductive ability which is definable and the essential creative factor
in conduct.
intelligent
The of testees have been shown to differentiate children and adults into
scores

five classes according to whether a person is intellectually defective, dull, normal,

bright or of superior intellectual ability. The standard series fails to differentiate
between individuals within these groups but the efficient range of each matrix is known
(Raven, 1939) and it is a simple matter to design further series of matrices which

can be used to differentiate between persons of approximately

equal ability and to
measure either the rate or the clearness of eduction.
The test provides a reliable means of differentiating between backwardness due
to disorganized schooling and backwardness due to mental impairment.
An interesting finding during the experimental work merits further investigation.
The scores of feeble-minded adults resemble those of a child of eight years or less,
but remain characteristically unlike those of an average person of more than nine
"

years. Psychologically they may be described as intellectually defective ".

" "
Intellectual defect can be diagnosed directly from test results and may be defined

as the permanent inability to form comparisons and reason by

analogy. It does not
necessarily occur in all persons certified as mentally defective, but it is probably the
" "
chief cause of social failure. Intellectual defect may exist in persons who, from
good repetitive ability, make stable social adjustments. In such cases the certification
of mental defect would be unwarranted, but the diagnosis of intellectual defect would
be justified.
A somewhat similar state of affairs is found to exist at the opposite extreme of
ability. A certain superficial brightness and verbal fluency appears to account for
a high Terman Merrill I.Q. and for early school success, but superior intellectual as

well as repetitive ability appears essential if scholastic success and social stability are
to be maintained.
Matrix Tests have been designed in accordance with psychological principles.
Even if these principles are disputed, the results show that in practice the tests work
as they should; and this,in the end, is what matters.
The writer is indebted to Dr. L. S. Penrose and Dr. L. G. Fildes for much helpful
assistance and criticism. Thanks are also due to the Child Guidance Council.
The work was carried out under the auspices of the Medical Research Council and
the Darwin Trust.

REFERENCES
Alexander, W. P., 1935. "Intelligence, Concrete and Abstract." Brit. J. Psychol. Monog.
Suppl., No. 19.
Dawson, S., 1931. "Intelligence and Disease." Spe. Rep. Ser., Med. Res. Coun., London,
No. 182. H.M. Stationery Office.
Raven, J. C., 1939. "The R.E.C.I. Research Series of Perceptual Tests; An experimental
survey." Brit. J. Med. Psychol., XVIII, 1.
Spearman, C., 1927. The Nature of Intelligence and Principles of Cognition. Macmillan & Co.,
London.
Stephenson, W., 1931. "Tetrad differences for verbal and non-verbal subtests." Amer. J-
Educ. Psychol. 22.