SHORTER ALTERNATIVES TO THE CALIFORNIA PSYCHOLOGICAL INVENTORY (CPI)
Monica Albu
(articol apărut în Volumul de lucrări ale
Conferinţei Internaţionale de Psihologie Aplicată, 17-19 mai
2001, Sinaia, Editura Augusta, Timişoara, 2002).
Abstract.
A California Psychological Inventory deficiency is the great number of its items
(over 400, in each of its versions). In this paper are presented two reduced
versions of CPI, in which each scale contains less items than the original
however has a better internal consistency.
The
California Psychological Inventory (CPI) was built by H. G. Gough, in order to
provide psychologists with an instrument that could be used for diagnosis and
prediction of interhuman behaviour. The items in the inventory refere to social
behaviour, attitudes and personal relations. The intentions stipulated for the
scales of the inventory are: (a) to predict what people will say and do in
specified contexts, and (b) to identify individuals who will be evaluated and
described in diferentiating and interpersonally significant ways.
There
are more versions of CPI, that differ from each other from the point of view of
items number (480, 462 and, recently, 434). The research presented below used
the version with 462 items.
This
version has 20 standard scales, designed in such way as to highlight folk
concepts – that is, concepts that arise from and are linked to the processes
of interpersonal life, and that are to be found everywhere that humans
congregate into groups and establish societal functions. The standard scales
were grouped in the following categories (Gough, 1987):
·
Measures of Poise, Self-Assurance, and
Interpersonal Proclivities: Do (Dominance),
Cs (Capacity for Status), Sy
(Sociability), Sp (Social Presence), Sa
(Self-acceptance), In (Independence),
Em (Empathy).
·
Measures of Normative Orientation and
Values: Re (Responsibility), So
(Socialization), Sc (Self-control), Gi (Good Impression), Cm
(Communality), Wb (Well-being), To
(Tolerance).
·
Measures of Cognitive and Intellectual
Functioning: Ac (Achievement via
Conformance), Ai (Achievement via
Independence), Ie (Intellectual
Efficiency).
·
Measures of Role and Personal Style: Py
(Psychological-mindedness), Fx (Flexibility),
FM (Feminity/Masculinity).
On
analysing the version with 462 items, translated and adapted for the Romanian
language by H. Pitariu, based on answers given by 1,000 people, two important
flaws have been noticed:
·
completing the questionnaire takes about one hour;
·
the internal consistency of the scales is low.
In
order to eliminate these problems we tried to build a questionnaire based on
items from CPI, to which the following conditions were imposed:
·
the total number of items should be smaller than 462;
·
the questionnaire should consist of 20 scales, corresponding to the 20 standard
CPI scales;
·
the scales should have better internal consistency than the CPI scales.
The
building of the new questionnaire was accomplished in 4 steps:
1.
Six psychologists identified those items in CPI which were formulated
incorrectly (they could be interpreted in more ways, they addressed only a part
of the subjects or they contained negatives). 130 items were found. It was
decided not to include those items in the questionnaire.
2.
For each standard CPI scale, groups of 3 psychologists indicated items which
could be included in it, considering Gough’s definition for the concept
measured by the scale.
3.
For each standard CPI scale, a group of items was formed, consisting of:
· items from the original scale which were not
eliminated during step 1;
· items formulated by psychologists in step 2 and
which originally didn’t belong to that particular scale.
Two
psychologists analysed the items and highlighted those items which, in their
opinion, were the most representative for what scale measured.
4.
On the computer and using an original algorithm, more alternatives were built
for each standard scale. For each of them, we started from a representative item
(marked during step 1), and we took out items from the group formed for that
particular scale, pursuing the best internal consistency for that scale. To
achieve this, for each step of the algorithm, we chose the item that, by being
added to the scale in the form it had at that moment, produced the highest
increase of the Kuder-Richardson 20 (KR20) internal consistency coefficient. The
algorithm stopped when the value of the coefficient started to go down.
During
this step, for each standard scale we used two methods:
a.
We didn’t limit the number of items in the scales we formed. Of the scales
that were built, the one with the best internal consistency was chosen. This way,
we obtained the CL questionnaire, with 249 items.
b.
We tried to obtain, in each scale, about half the number of items in the
standard CPI scale and, besides, we wanted the KR20 coefficient to be bigger
than the KR20 in the standard scale. This way, we obtained the CS questionnaire,
with 213 items.
In
order to check if the two questionnaires measured the same concepts as the CPI
standard scales, we calculated, for each standard scale, the linear correlation
coefficient between the scores of standard scales and the scores of
corresponding scales in CL and CS. To accomplish this, we used answers provided
to CPI by a number of 1141 persons (668 males and 473 women). In table 1 it can
be noticed that all the linear correlation coefficients are significant on the
threshold p=0.001. We can conclude
from here that the new scales measure the same concepts as the standard CPI
scales. So, to interpret the scores of the subjects we can use the descriptions
given by Gough to “high” scores and, respectively, “low” scores (Pitariu,
Albu, 1993).
Using
the same group of subjects we compared the internal consistency of standard CPI
scales with that of CL and CS scales (Table 2). With very few exceptions, the
KR20 coefficients of new scales are higher than those of standard scales.
The
conclusion is that CL and CS questionnaires measure the same concepts as the
standard scales in the questionnaire with 462 items of CPI, but contain much
fewer items than CPI (CL has 249 items and CS, 213) and their scales have a
better internal consistency than standard CPI scales.
Table
1. Linear correlation coefficients for questionnaires CPI, CL and CS
(N=1141)
Scale
CPI
CPI CL
CL
CS CS
Do
.934
.921 .988
Cs
.667 .669
.971
Sy
.868 .826
.962
Sp
.769 .642
.928
Sa
.629 .540
.971
In
.873 .854
.972
Em
.761 .716
.919
Re
.719 .711
.939
So
.791 .660
921
Sc
.958 .940
.983
Gi
.907 .907
1.000
Cm
.668 .639
.931
Wb
.884 .877
.989
To
.802 .751
.934
Ac
.617 .587
.979
Ai
.595 .449
.907
Ie
.888 .877
.960
Py
.709 .688
.873
Fx
.846 .846
1.000
FM
.816 .816
1.000
Table 2. KR20 internal consistency coefficients for scales in questionnaires
CPI, CL and CS.
(668 males and 473 females)
Scale
CPI
CL
CS
Length
KR20
Length
KR20
Length KR20
M
F
M
F
M
F
Do
36
.766 .791
20
.790 .829
18 .778
.815
Cs
28
.654 .690
18
.750 .792
14 .731
.776
Sy
32
.674 .725
18
.758 .814
14
.725 .784
Sp
38
.655 .700
21
.752 .785
17 .750
.775
Sa
28
.446 .584
20
.808 .830
14 .789
.810
In
30
.616 .619
19
.678 .737
15 .692
.742
Em
38
.534 .648
18
.559 .635
12 .582
.617
Re
36
.657 .600
24
.674 .633
18 .658
.645
So
46
.675 .682
37
.796 .783
23 .763
.762
Sc
38
.810 .824
33
.824 .832
28 .810
.817
Gi
40
.831 .828
30
.826 .833
30 .826
.833
Cm
38
.497 .565
24
.637 .632
19 .616
.584
Wb
38
.767 .812
24
.789 .816
22 .775
.805
To
32
.566 .564
23
.618 .624
16 .593
.569
Ac
36
.687 .694
22
.742 .725
19 .730
.726
Ai
36
.577 .602
26
.672 .716
16 .628
.652
Ie
42
.632 .681
28
.714 .763
21 .695
.757
Py
28
.513 .582
24
.693 .751
14 .671
.728
Fx
28
.573 .594
14
.596 .619
14 .596
.619
FM
32
.484 .430
15
.546 .490
15 .546
.490
References
Gough, H.G. (1987), California
Psychological Inventory Administrator's Guide, Consulting Psychologist Press,
Palo Alto.
Pitariu, H.D., Albu, M. (1993),
Inventarul Psihologic California: prezentare şi rezultate experimentale, Revista
de Psihologie, nr. 3, 249-263.