Inconsistency, Loose Construing and Schizophrenic Thought Disorder

Brit.J. Psjchiat.(1973), i@, 209-17
Inconsistency,
Loose Construing and Schizophrenic
Thought Disorder
By ELSIE T. HAYNES and J. P. N. PHILLIPS
Bannister
INTRODUCTION
et ci. (1971),
Spelman
ci ci. (i@7i),
normals, neurotics and (in the 1962 study only)
depressives. This finding has been confirmed,
Williams (‘97'), Frith and Lithe (1972) and
McFayden and Foulds (1972), all found that,
on the whole, consistency scores tended to
discriminate thought-disordered
schizophrenics
from other groups about as well as intensity
scores. Secondly, in all those studies which
reported consistency scores (i.e. all except
sometimes
in slightly
number
of subsequent
McPherson
and Buckley, and
scores tended to be positively
Bannister
(1960,
1962)
found
that
thought
disordered schizophrenics hadsignificantly lower
scores on a repertory grid test of the intensity of
the relationships between their constructs than
did
non-thought
disordered
schizophrenics,
different
ways, by a
studies (vide infra). On
the basis of his results, Bannister put forward
the theory that schizophrenic thought disorder
is characterized
can
account
by ‘¿loosened
construing',
for all its clinical signs, and
stems from repeated
invalidations
it would
introduce
which
which
of construing.
an effectively
random
these
with
intensity scores, usually significantly so.t Thus
it might simply be that thought-disordered
schizophrenics differ from other groups only in
consistency, or lack of it, and that this variable
is related to both consistency scores and
This theory has been followed up in a number of
subsequent investigations (Bannister, 1963, 1965;
Bannister and Salmon, 1966).
However, an alternative hypothesis which will
account for the finding without the need to
introduce the concept of loosened construing
is that thought-disordered schizophrenics are
simply more inconsistent than other subjects
in their performance on the repertory grid.
Such inconsistency might result from a private,
idiosyncratic and possibly delusional conception
of what the experimenter has asked them to do,
or from difficulties of discrimination (cf. Frith
and Lillie,
1972),orjustfrom a chaoticdisorga
nizationof theirthinking;but, whatever its
origin,
Williams)
correlated
intensity
scores on grid tests.
However, the evidence of these studies (with
the exception of another and more important
part of that of Frith and Lillie) is not conclusive,
for the consistency
scores used in them are
contaminated
by the intensity of the relation
shipsbetween constructs,
being indeed com
puted from the very same measures of intensity
as are used to compute the intensity scores, so
that it is also possible that low intensity in
thought-disordered schizophrenics was the cause
of the low consistency scores, instead of the
otherway round. Specifically,
the consistency
scoreswere computed by running correlations
* We
have
found
a number
of inconsistencies
(which
it
element into the patients' responding, and thus
lower the absolute magnitude of the correlations
wouldbetedious
toenumerate)
inthetables
andfigures
of
these studies. In most cases it has not been possible to
between
it has been possible the differences from the correct
results have not been great, and the conclusions do not
seem to be materially affected.
their constructs.
There is some indirect support for this
alternative hypothesis in Bannister's own studies
and the confirmatory studies mentioned above.
Firstly, Bannister (1960*, 1962*), Bannister
and Fransella (1966*), Foulds et ci. (1967),
Poole
(5968),
McPherson
and
Buckley
(‘970),
determinewhatthecorrectresultsshouldbe, butwherever
t Bannister and Mair (i968) quote also an unpublished
BA. dissertation
by D. V. Bridges.
Unfortunately,
the
most extensive searches failed to obtain us a copy, and it
nowseemslikelythatall onceextantcopiesof thisdisserta
tion have been lost.
209
210
INCONSISTKNCY,
LOOSE
CONSTRUING
AND
between measures of intensity obtained from
one grid and the corresponding measures of
intensity
obtained
from a second
grid using the
same constructs.*
Now, in the studies by
Bannister (5960, 1962), where the second grid
used
different
elements
from
the
first,
the
measures of intensity were essentially phi
correlation
coefficients, and their standard
errors of estimate would be inversely propor
tional to their magnitudes.
Thus ifthe intensities
in the two grids were low they would be less
stably estimated than if they were high, so that
the measures of intensity would agree less well
together from the first grid to the second. This
factor alone could have caused a positive
correlation between intensity and consistency
in these two studies. The subsequent studies
use the Bannister-Fransella test (Bannister and
Fransella, 1966, 5967), in which the second
grid is just an immediate
readmiistration
first one, and where the measures
are Spearman
rhos. According
of the
of intensity
to Bannister
and
Fransella, intensity scores and consistency scores
SCHIZOPHRENIC
THOUGHT
DISORDER
appear in principle feasible to devise a test of
intensity that is unaffected by inconsistency; for
inconsistency, as envisaged in the alternative
hypothesis, is analogous to a random noise
component in the thought-disordered schizo
phrenics' responses, and there seems to be no
way of eliminating it directly. However, it is
possible to devise a test of consistency which is
unaffected
by intensity
: one such was used by
Frith and Lillie, and anothe@ is described below.
By the use of such a pure test of consistency, the
inconsistency component of intensity scores can
be partialled
out statistically,
making possible a
test of the alternative hypothesis, which is the
purpose of the present study.
METhOD
Subjects. These were:
Ten diagnosed schizophrenics
at Dc la Pole
Hospital (eight male and two female), judged to be
thought-disordered
by the responsible Consultant
Psychiatrist and, where he was a different
also by the Physician Superintendent;t
person,
Ten schizophrenics (nine male and one female
on this test are ‘¿mathematically independent';
nevertheless,
it is clear that if the intensities are
similarly judged not to be thought-disordered;
low in absolute value the measures of them will
cluster more closely about zero than if they are
high: thus, any changes in these measures
resulting from changes in the rankings of the
elements will cause a greater perturbation of
their rank order from the first administration to
the second, and hence there will be a lower
correlation between the measures of intensity
for the two administrations if the intensities
are low in absolute value than if they are high.
Ten normal controls (five male and five female),
artisan, clerical and nursing staff of the hospital.
This factor also could have resulted in a positive
correlation
between
intensity
and consistency
in all the studies.
Since, therefore, consistency can affect in
tensity scores (as mentioned in the second
paragraph),
and intensity can affect consistency
scores (as just shown), it would be desirable to
administer
pure tests of both intensity and
consistency to thought-disordered schizophrenics
and other clinical groups, in order to determine
whether it is consistency, or intensity, or both of
them, which is, or are, relevant to schizophrenic
thought disorder. Unfortunately, it does not
* For
(1972).
a
fuller
discussion
of
this
approach
see
and
The threegroupswerematchedon age,but not
on intelligence or sex: previous work by Bannister
(i96o,
1962)
and
Bannister
and
Fransella
(1967)
found no evidence that these three variables were
related to intensity and consistency scores on repertory
grid tests; Frith
and Lillie found
intelligence,
but not
age or sex, to be highly significantly related to a pure
measure
of
consistency;
when
this
measure
was
partialled out statistically, the only highly significant
relationship
was between
age and Bannister's
measure of consistency. It did not appear desirable to
attempttomatchthegroupson measured
intelligence
since this could result in the groups being less
differentiated on thought disorder (which is likely to
be related to measured intelligence).
An attempt was
made to match the groups on pre-morbid intelligence,
asestimated
by pre-morbid
occupational
status,
but
paucity of records rendered this impossible.
Test material. This consisted of photographs of
eight men and eight women unknown to the subjects4
t We are grateful to Dr. J. A. It. Bickford, Physician
Superintendent,
and to Dr. I. C. Church for these
assessments.
We are grateful to Professor Bannister for making
Slater
these
photographs
available.
BY
ELSIE
T.
HAYNES
AND
Procedure.The subject was presented with four of the
photographs,
and
asked
to arrange
them
P.
N.
Scoring. This
211
PHILLIPS
comprised
three
stages:
Stage I : Since the twenty subsets of four photo
graphs each constituted
a balanced
incomplete
in order,
from the most Likeable to the least. When he had
@
3.
done this, another four were presented with the same block design, every possible pair of photographs
request.
Thiswas continued
untiltwentydifferentoccurred just once in some subset. It was therefore
subsets of four had been presented and ranked, the possible, following Durbin (1951), Schucker
subsets
constituting
a balanced
incomplete
design,
as illustrated
by the data
in Table
Gulliksen
block
and Tucker
(1961) and Slater (5965), to
derive from the rankings an implied table of pair
I, which
showstherankings
producedby thefirst
non-thought comparisons, as may be illustrated by the example
in Table I.
disordered schizophrenic.
The first subset of four contains photographs 3, ‘¿5,
This procedure was then repeated with the further
8 and 5, which the patient has ranked in that order.
constructs
Mean, Good, Unusual,Narrow-minded,
Sincere,
&lfish,
Bannister,
Unreliable
and
1962>, a different
Kind
(taken
This implies that he judges 3 to be more Likeable
than 15, 8 and 5, so ones are entered in columns 15,
from
randomization
of the
basic balanced incomplete block design (given by
Cochran and Cox, 5957), being used for each
8 and 5 of row 3 (and noughts in rows 15, 8 and 5 of
construct.
more
column
T@nii
3):
it also implies
Likeable than 8 and
that
he judges
15 to be
5, so ones are entered in
I
Table
Total
12345678910111213141516I00000000000000I131410i62I
315
6Rankings
48 75
IPairComparison
0000000050010I9122II3II
111111I1591374I
110110
10110110I5410II5I
10
0100110010I32,i6611111
101
I
414i67
8I
II•5
9
8
0
0
0
0
0
1100101
0
I
I
0
0
I
6
8
0
0550010I
‘¿3
111510I6353II•10I
10000I9810III100000000015II14I12I
10000000
o o
I65i6131100101101103
I
0
I
I
0
I
I
0
I
I
@
I 0
I
I
I
I
6557
4
3
6•@
8.5
15
157214911111011
128
1512
412 13
15
13
001
111111111I6
@
Rank
I
8
1010
2
15
254
i61000000001000
000000000000000
12
53
5
I
9
8
3
7
4
ii
14
i6
I
I
0
I
I
0
I
I
I
I
I
I
I
2
14
10
3 ‘¿5
9
4 8 5
4 ‘¿35
II 14 10
‘¿0
3
4
11.5
3
13@5
II
0
I
9
0
I
3
‘¿3
0
Circular
Triads
d=5
I
0
I
5
I3•5
3
i6
212
INCONSISTENCY,
LOOSE
CONSTRUING
AND
SCHIZOPHRENIC
THOUGHT
DISORDER
totals. In this example, d = 620—615 = 5, and the
columns 8 and 5 of row 15 (and noughts corre
circular
triads
areshownatthefootofthetable.
spondingly): it finally implies that he judges 8 to be five
The second measure of inconsistency which may be
moreLikeable
than5,soa oneisentered
incolumn5
ofrow 8 (anda noughtcorrespondingly).
Treating used is Slater's (I96o, 1961) statistic i, the minimum
number of (corresponding pairs of) cells it is necessary
each subset of four in the same way, the pair com
to change in the pair comparison table in order to
parison table is filled in completely.
Stage @:From such a table there can be derived
both a ranking of the photographs and also measures
make it completely consistent (i.e. without
circular triads whatsoever),
or, equivalently,
of consistency (or rather, inconsistency). There are
in fact two measures of inconsistency which may be
used. The first is Kendall and Babington Smith's
minimum number of errors
one must supposethe
subject to have made. In the present sample the table
can be made completely consistent, and all circular
(1939;
triads eliminated, by changing the entries in row 4
column @,in row 10 column 14, and in any one of
Kendall,
circular
where
@
14,
I@8)
statistic
d,
the
number
triads such as, in this example,
the patient has judged
14 more
Likeable
than
io
of
2, 14, 10,
2 more
Likeable than
and
more
io
thethreecells
row 3 column9,row 9,column15or
row i column 3 (and also their correspondingmirror
Likeable
than 2 d is readily computed by the formula
d =@:I2@@
where
E
image
T@aii II
Permutated Pair Comparison Table
6 3 15 9 12 4 13 8 5 7 10 2 II 14 I
3 ‘¿5 85
7'
64
6
13
14
10
16
9
12
2
II
7
54
10
II
32
i
6
12
12
I
0
0
12
0
0
8 14
4 i6
13
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
‘¿â€˜â€˜â€˜III
0
I
0
0
0
II
7000000000
10
I
II
14
I
3
12
5
7
i6
10
4
‘¿3 82
6
15
10
2
13
5
I
0
0
0
0
8ooooooo
13
4
II
I
I
10
0
I
101
63
98
3
7
I
I
400000
500000100
9
8
I
11111
14
12
I
9010
2
69
I
101
57
15
I
30
15
13
15
4, etc.), so that i =
0
0
I
I
I
I
I
I
I
I
I
I
I
I
II
I
I
I
I
I
I
1
7
I
I
I
I
I
I
I
I
1
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
0
0.
0
0
0
0
0
0
0
0
0
0
0
0
11011
I
200000000000
II
14
0
0
0
0
0
0
0
0
0
0
0
0
0
0
I
0
0
0
100000000000000
‘¿4
i6
3.
the three nearest adjoining orders which differ least
from the obtained data: inconsistencies stand out as
noughts in the upper right-hand half of the table
of the squares of the row
Rankings
9
5 column
rows and columns have been permuted into one of
I@2
15
cells row
This may be seen more clearly in Table II, where the
—¿
a@ is the sum
any
the
i6
0 o o 0 0 0 0 0 0 0 0 0 0 0 0
i3
-
3159
3
15
910
0
1593
I 0
15
I
9
310
0
I 0
9
I
30
15
9 3 ‘¿5
10
I
I
0
BY
ELSIE
T.
HAYNES
and ones in the lower left-hand half. It will be noted
from the sub-tables that photographs 3, 15 and 9
may be put in that order or in the order 15, 9, 3, or in
the order 9, ‘¿3,
‘¿5
(depending upon which of the
three cells mentioned above is changed) without
altering the number of inconsistencies, so that there
are here three nearest adjoining orders. The determi
nation of Slater's i is not straightforward, and was
carried out by computer (Phillips, 1969, 1971).
The ranking of the photographs taken from each
pair comparison table was that given by the row
totals
(as in Table
I) rather
than the nearest
adjoining
order, because, as has just been seen, there were
sometimes more than one of the latter.*
(It can now be seen that the procedure of ranking
subsets of four photographs at a time was a compro
mise between obtaining
possible
pair
all one hundred
comparison
judgements
and twenty
of the
photo
graphs, which would have provided maximal infor
mation on internal consistency but have been im
practicably
lengthy,
and
ranking
the whole
set,
which although perfectly practicable provides no
information on internal consistency.)
Stage
@:Two measures of inconsistency
obtained
for each subject,
by summing
dover
were
all nine
constructs (Zd) and by summing i over all nine
constructs
(Zi).
To obtain measures of intensity for each subject,
Spearman
rhos were first computed
for all thirty-six
pairs of constructs. There were then two alternatives.
Bannister
(1960,
1962) used a score which,
when
the
subject has to dichotomize the elements on each
construct is equal to the absolute value of the phi
correlation coefficient multiplied by half the number
of elements: therefore, analogously, the sum of the
absolute values of all thirty-six rhos (Z@ @I)may be
taken as a measure of intensity. On the other hand
Bannister and Fransella (1966, 1967), and the studies
* For a further method of estimating a ranking from
an inconsistentpair comparison table,see Kendall 1955.
AND
3.
P.
N.
213
PHILLIPS
subsequent to them, used the sum of the squares of
Spearman rhos (Zp2), which also may be taken as a
measure of intensity.
RESULTS
There were significant differences between the
groups
on intensity
scores,
Specifically,
on both
Z@iI and Zp2 the
thought-disordered
schizophrenics
scored signi
ficantly
less than the non-thought-disordered
schizophrenics
(p
controls
.01),
(p
<
<
.05)
the
method:
197
Ryan,
1959;
O'Neill
the normal
two
groups
not
and
Wetherill,
I).
These results confirm the previous findings
cited in the introduction, but are not crucial
for the alternative hypothesis. As regards the
latter,
(i)
there
that
related
are
three
schizophrenic
only
to low
possibilities,
thought
intensity,
namely
disorder
(ii)
that
is
it is
related both to low intensity and to low con
sistency, and (iii) that it is related only to low
consistency. Presumably Bannister's theory im
plies that one or other ofthe first two possibilities
is the case, whereas the alternative hypothesis
is that it is the third which is true.
According to possibility (i), the thought
disordered schizophrenics should not differ
from the other two groups on the pure measure
of inconsistency. In fact, the differences between
the gi oups on this variable were more significant
than
those
in intensity,
as may
be seen from
Table III: thought-disordered schizophrenics
scored (on both Zd and L'i) significantly more
inconsistently than either of the other two
[email protected]
[email protected]@057.145<0I£
Group
.£@j709.1242@[email protected]()<.()fJ@ZjI59@967@837@I20<.001
and
latter
differing significantly from one another (Tukey's
D!fferences between groups on intensity and consistencyscores
p26@47I4@205•99@<.@Consistency
as may be seen from
Table III.
214
INCONSISTENCY,
LOOSE
CONSTRUING
AND
groups (p < 00I in both cases), which did not
differ significantly
from one another.
Indeed,
there was (just barely) no overlap between the
thought-disordered schizophrenics and all the
other subjects on either Ed or Zi. Thus possi
bility (i) above must be rejected : schizophrenic
thought disorder clearly is related to inconsis
tency,
and
from
the
relative
magnitude
of
the
effect (a pure measure of inconsistency differen
tiating better than a measure of intensity
confounded
judged
by
inconsistency)
it
might
be
to be the more relevant of the two
variables.
The view that intensity scores differentiate
the groups only to the extent that they are
contaminated
by inconsistency receives some
support
from
the
fact
that
these
scores
were
SCHIZOPHRENIC
THOUGHT
DISORDER
the test-retest reliability ofrankings ofitems, the
intensities
of the
relationships
between
concepts
were as high as, or higher than, those of normals.
However, unfortunately for the purpose of the
present study, they do not report any assessment
of their schizophrenics for thought disorder,
and indeed the procedure of experimentally—as
opposed to statistically—matching schizophre
nics
with
normals
on
consistency
is likely
to
have selected schizophrenics who were not
thought-disordered.
Thus nothing about the
intensity scores of thought-disordered schizo
phrenics can be concluded from this study.
Koh and Shears (1970), on the other hand,
reported that schizophrenics
showed poor
internal consistency in psychological scaling tasks
( judgement@
of
line
lengths
and
preferences
for
tude estimation)
as compared
@
significantly correlated with the pure measure
ofinconsistency (p < .001) Whichever measure
of intensity, E@ or Ep2, and whichever
Ag am, the authors
do not report
measure
Ed or Zi, was used, the
oftheir schizophrenic.s for thought disorder; but
@
overall
in the
here, unlike with the Miller and Chapman
study, it seems likely that a good many of the
of consistency,
correlation
was
whilst the within-groups
no case varied
another,
region
of —¿o 7,
correlation,
which in
significantly
from one group
to
was in the region of —¿o
‘¿5.
However,
the crucial test between
possibilities
musical
excepts
by
subjects must
disordered.
Most crucially
category
actually
rating
have
relevant,
and
magni
with normals.
any assessment
been
however,
thought
is the study
of Frith and Lillie (1972), who retrospectively
(ii) and (iii) must be made by analysis of co
variance of the intensity scores, with the pure
measureofinconsistency
ascovariate.
Ifthought
disorderis relatedto both inconsistency
and
intensity, then the intensity scores, even with
inconsistency partialled out, should still differen
tiate between the groups; whilst if it is related
variables
only to inconsistency
they should not. In fact,
the latter turned out to be the case: for all four
tion between
photographs,
combinations of intensity and consistency score,
the F-ratios were less than unity (indeed less
immediate
than 0.2),
one used here was of internal consistency, their
and not significant
even at the 75 per
examined
the relationship
between
psychiatric
ally assessed thought disorder (and a number of
which
do not concern
us here)
scores on the B-F test, including
calculated
pure measure
and
a specially
of consistency,
namely
element consistency, which is the squared correla
the first and second rankings of the
summed over all six constructs
in
the test. Although
test-retest
their pure measure was of
consistency,
whereas
the
cent level: thus there is no evidence in the results were in the relevant respects identical
data that schizophrenic
thought disorder is with those of the present study. Intensity scores
thought-disordered
relatedto intensity,
once the contaminationof significantly differentiated
intensity
scores by inconsistency
is removed.
patients
from others
pure (element)
DISCUSSION
thought-disordered
(x2
=
5'I5,
p <
.05):
consistency scores differentiated
patients
from
others
still
more significantly (x = I4'43, p < .ooi),
There are two studiesof possible
relevance, and were significantly correlated with both
and one of certain relevance, which must be intensity scores (r = o'68, p < .ooi) and
Bannister's
consistency
scores (r = o 6o, p <
mentioned here. Miller and Chapman (1968)
.oor); intensity scores, corrected for their
found that when schizophrenics were matched
with normals on consistency, as measured by relationship with pure (element) consistency,
Related
findings
BY
ELSIE
T.
HAYNES
showed no relationship with psychiatric
ment of thought disorder.
AND
assess
Inconsistency, loose construing and schizophrenic
thought disorder
From
these results,
and
those
of the present
study it would appear to follow, firstly, that there
is no evidence in the studies cited in the intro
duction that schizophrenic thought disorder
is characterized
constructs
by weak
relationships
; and secondly,
that
3.
P.
N.
PHILLIPS
subsume
the elements
(behaviour
of the
acquaintance)
under consideration
and may
itself be, in some way, modified (sic). This
modification could take the form of loosening
the construct, that is to say, weakening its
relationship with other constructs, so that it
leads to varying predictions (the definition of
a loose
construct)
. Thus
loosely
to
construe
a
person as “¿loving―
would not be automatically
to anticipate
between
and
any appearance
215
from
his “¿sincere―
and
so on behaviour,
since
the
“¿likeable―
relationships
which is characteristic of schizophrenic thought
between the constructs normally constellated
with “¿loving―
have been weakened.' Bannister
thus appears to regard the ‘¿continual
shifting of
disorder.
elements
Before considering how this conclusion relates
to the ‘¿loosened
construing' theory of schizo
which according to Kelly (bc. cit.) characterizes
phrenic
loose construing, but as being the precursor and
to the contrary probably
tamination
of intensity
thought
resulted from con
scores by inconsistency,
disorder,
it is necessary
to be
a loose
from
one pole
construct,
to the contrasting
as not
in
itself
pole',
constituting
clear as to how the phrase is to be taken, since,
cause of loosening,
as will
fication of the construct, namely weakening its
relationships
with other constructs. Further
be shown,
it has
been
given
two
quite
which
is a subsequent
modi
distinct meanings.
more,
Bannister
treats
loosened
construing
as
Kelly (1955) lays down the definition, which
must presumably be taken as authoritative, ‘¿Asynonymous with weak structure (e.g. ‘¿thought
loose construct is one which leads to varying
disordered schizophrenics had very weak con
(loosened
construing)'
Ban
predictions but retains its identity.'
(Vol. I, ceptual structure
p. 533). He gives numerous
illustrations
of this
definition, e.g. ‘¿If
breakfast is loosely construed,
then the meal one ate at i , A.M. might at once
be construed
both as breakfast and as something
other than breakfast, or as breakfast at one time and
as somethingcontrastingwith breakfasta momentlater.
This is loose construction.'
(Vol. II, p. 1058).
Thus Kelly's concept of loose construction seems
to mean
both
internal
inconsistency,
as opera
tionally defined in the present study, and test
retest
unreliability,
as operationally
defined
by
element consistency in the study of Frith and
Lilile.
Bannister's concept of loose construing seems
quite
different.
The
best
statement
of it occurs
nister,
states
1965, p. 377),
‘¿From
our point
whereas
ofview,
Kelly
loose
explicitly
construction
is not to be considered as the same as lack of
structure.'
(Vol. II, p. 1030).
Therefore, whilst the present results, and
those of Frith and Lillie, suggest that there is
no evidence
that schizophrenic
thought
disorder
is characterized by ‘¿loosened
construing' in
Bannister's sense of weak relationships between
constructs, they are perfectly consistent with,
and indeed strongly support the proposition
that it is characterized by ‘¿loose
constructs' in
Kelly's sense of inconsistently used constructs.
However,
it should
be pointed
out that it does
not require
the elaborate
conceptual
apparatus
of personal
construct
theory
to tell us that
in his ig6o study (p. 1246): ‘¿For
example, if the
construing
ofa personas “¿loving―
isinvalidated thought-disordered
schizophrenics
are incon
then the tendency
may be to construe
the person
sistent. Pluralitas non est ponenda sine necessitate.
in the
contrast
pole
as “¿hating―, but the
acquaintance
may then exhibit
behaviour
more
appropriate
to the “¿loving―
pole and we may
Sui,mi&py
i. It is suggested
that
the repeatedly
con
duly shuffle
him back to be construedunder
that pole. However, if invalidation occurs in
firmed finding of low intensity
this manner
schizophrenics, which Bannister has taken as the
basisfor his ‘¿loosened
construction'
theoryof
to regard
repeatedly,
the
construct
the tendency
as itself
may
inadequate
be
to
between
constructs)
(of relationships
scores in thought-disordered
216
INCONSISTENCY, LOOSE CONSTRUING AND SCHIZOPHRENIC THOUGHT DISORDER
schizophrenic
thought
disorder,
is in
fact
an
artefact of the contamination
of intensity scores
by inconsistency.
There is some support for this
alternative hypothesis in the literature, since it
has been found in eleven studies that consistency
scores tend to be positively, and usually signifi
candy, correlated with intensity scores, and to
discriminate thought-disordered
schizophrenics
from other groups about
evidence
is not conclusive,
possibility
as well. However,
this
since there is also the
that consistency
overcome
grid
test
this
was
difficulty,
a
administered
form
of
to ten
thought-disordered
schizophrenics,
ten non
thought-disordered
schizophrenics
and ten
normal controls, in which the subjects ranked
various
overlapping
subsets
of elements,
rather
than, as is conventional, the whole set, thus
making possible a measure of pure internal
consistency. Although intensity scores signifi
candy discriminated
the thought-disordered
schizophrenics
from
the
others,
pure
internal
consistency
scores
(which
were
significantly
correlated
with them)
did so better and when
pure internal
consistency
was partialled
out of
the intensity scores, the latter no longer dis
criminated the groups significantly. Effectively
identical results have been obtained by Frith
and Lillie.
3.
It is shown
construction'
senses,
that
the
phrase
—¿
(Kelly)
and
‘¿loosened
intensity
(Ban
nister). Thus the results give no confirmation
Bannister's
theory,
that schizophrenic
to
but strongly
support
the idea
thought
disorder
is charac
terized by loosened
construing,
—¿
in Kelly's sense.
(‘962).
‘¿The
nature
and
(1963).
‘¿Thegenesis
of
—¿
(1965).
‘¿The genesis
schizophrenic
of
of
and
FIWI5EUA,
F.
(1966).
thought
dis
schizophrenic
thought
‘¿Agrid
test
of
schizo
phrenic thought disorder.' Brit. 3. soc.din. Psychol.,@,
95—102.
Test
of
Aoienw,
J.
(‘97').
‘¿@haractthstics
and
Psycliol.,
10,
144—51.
and
SALMON,
P.
(,g66).
‘¿Schizophrenic
thought
215—9.
COCHRAN,W. G., and Cox, G. (i@@7). &perimental Design.
Duamu, J. (‘95'). ‘¿Incompleteblock designs in ranking
experiments.' Brit. 3. Psycliol., Statist. Section, 4, 85—go.
Fouus, G. A., HOPE, K., McPnzasoN, F. M., and M@so,
P. R.
(i@6@').
‘¿Cognitive disorder
among
the
schizo
phrenias. I. The validity of some tests of thought
process disorder.'
Brit. 3. Psyc/ziat., ii@, 1361—8.
Farm, C. D., and Lni.re, F. J. (1972). ‘¿Why
does the
repertory grid test indicate thought disorder?' Brit. 3.
soc. din. Psyc/zol.,ii, 73—8.
Guuixszx, H., and Tucicza, L. (z@6,). ‘¿A
general pro
cedure for obtaining paired comparisons from
multiple rank orders.' Psychometrika,26, 173-83.
KELLY,
G.
A.
(i9@@).
The
Psychology
ofPersonal
Coartructs.
New York: Norton.
KENDALL,M. G. (@8). Ram@CorrelationMethods. London:
Grilfin.
—¿
(‘955).
‘¿Further
contributions
to
the
theory
of paired
comparisons.' Biometrics,ii, 43-62.
—¿
and
Ssnm,
B.
BABINOTON
paired comparisons.'
KON,
S.
D.,
scaling
and
(ig@).
‘¿On the
method
of
Biornetrilca, 35, 324-45.
Sisa@as,
G.
(i@@o).
by schizophrenics
and
‘¿Psychophysical
normals.'
Arch. gen.
Psjchiat., 23, 249—59.
MOFAYDEN, M., and Fouus,
G. A. (1972). ‘¿Comparison
of provided and elicited grid content in the grid test
of schizophrenic
thought
disorder.'
121, 53—7.
MCPHERSON, F., and Bucia.zv,
F.
Brit.
3.
(1970).
Psyc/iiat.,
‘¿Thought
process disorder and personal construct subsystems.'
Brit. 3. soc. din. Psycliol.,9, 380-I.
MILLER, G. A., and Cii4@p@N, L.J. (,g68). ‘¿Response
bias
and schizophrenic beliefs.' 3. aM. Psycliol., 7@, 252—5.
R., and WETHERILL,
G. B. (@g7i).
‘¿The
present
PmLUI's,J. P. N. (i96g). ‘¿A
further procedure for deter
mining
Slater's
i and all nearest
adjoining
orders.
Brit. 3. math. statist. Psycliol.,22, 97-101.
schizo
disorder: re-test of the serial invalidation hypothesis.'
Brit. 3. Psyc/ziat.,III, 377—82.
—¿
Grid
disorder: specific or diffuse?' Brit. 3. med.Psychol.,39,
ment.ScL, io6, 1230-49.
measurement
order: a serial invalidation hypothesis.' Brit. 3.
Piat.,
109, 68o-86.
—¿
(ig67).
Soc. (B), 33, 218—41.
in thought
phrenicthoughtdisorder.'3. ment.Sci., io8, 825-p.
—¿
F.
state of multiple comparison methods.' 3. Ro,. statist.
D. (196o). ‘¿Conceptualstructure
disordered schizophrenics.'J.
and
Soc. din.
REFERENCEs
—¿
[email protected],
—¿andM@ua,J. M. M. (@68). The Evaluationof Personal
Constructs.London: Academic Press.
O'NEILL,
BAIsvaR,
and
validity of the grid test of thought disorder.' Brit. 3.
has been used in two different
consistency
D.,
Sthizop/trenic Thought Disorder. Manual. Barnstaple:
Psychological Test Publications.
Second Edition. New York: Wiley.
scores are contamina
ted by intensity, as well as the other way round.
2. To
repertory
[email protected],
—¿
(ig7i).
‘¿Slater's
1 and
nearest
adjoining
orders
from
pair comparisons
or balanced
incomplete
block
design rankings (new edition).' (Computer program
abstract.) Brit. 3. math. statist. Psychol.,24, 125-6.
Poo@i, D. A. (i968). ‘¿A
retrospective investigation of the
clinical usefulness of the Bannister-Fransella
Grid
Test of schizophrenic thought disorder.' Unpublished
M.Sc. Thesis, University of London.
Rv@, A. T. (i959).
‘¿Multiple
comparisons
inpsycho
logical research.' Psychol.Bull., 56, 26-47.
BY ELSIE T. HAYNES
Scauczza,
paired
R. E. (i@@g). ‘¿A
note on the use of triads for
comparisons.'
Psychometrika, 24, 26-47.
AND J. P. N. PHILLIPS
217
Si@ma, P. (1972).‘¿The
measurement of consistencyin
repertory grids.' Brit. 3. Psychiat., 121, 45-51.
Si@vaa,
P.(ig6o).
‘¿The
analysis
ofpersonal
preferences.SPELMAN,M. S., HARRISON,A. W., and Mzwop, G. W.
Brit. 3. statist. Pol.,
13, 119-35.
(‘97,).‘¿Grid
test for schizophrenicthough disorder in
—¿ (@6i).
‘¿Inconsistencies
in a schedule
of paired
com
acute and chronicschizophrenia.'
Psychological
parisons.'
—¿
(i@6@).
Biometrika, 48, 303—12.
‘¿The
test-retest
reliability
of
some
methods
of
multiple comparison.' Brit. 3. math. statist. Psycho!.,
i8, 227—42.
Medicine, I@234—8.
Wn.u@uss,E. (i@7i). ‘¿The
effect of varying the elements in
the Bannister-FransellaGrid Test of thought disorder.'
Brit. 3. Psychiat., 119, 207—12.
Elsie T. Haynes, B.A., A.B.PS.S.,Psychology Department, Dc La Pole Hospital, Willerby, E. Yorks. Present
address: Psychology Department Brandesburton Hospital, Brandesburton, E. r@ks.
J. P. N. Phillips,M.A.,Ph.D.,Dip.Psych.,
Psychology
Department,
TheUniversity,
Hull,E. Yorks.
(Received 8 August 1972)
Inconsistency, Loose Construing and Schizophrenic Thought
Disorder
ELSIE T. HAYNES and J. P. N. PHILLIPS
BJP 1973, 123:209-217.
Access the most recent version at DOI: 10.1192/bjp.123.2.209
References
Reprints/
permissions
You can respond
to this article at
Downloaded
from
This article cites 0 articles, 0 of which you can access for free at:
http://bjp.rcpsych.org/content/123/573/209#BIBL
To obtain reprints or permission to reproduce material from this paper, please write
to [email protected]
/letters/submit/bjprcpsych;123/573/209
http://bjp.rcpsych.org/ on June 18, 2017
Published by The Royal College of Psychiatrists
To subscribe to The British Journal of Psychiatry go to:
http://bjp.rcpsych.org/site/subscriptions/