Basic HTML Version
340
F.
J.
EVANS AND D. SCHMEIDLER
Implications for Dimensionality of Hypnotic Phenomena. Theoreti-
cally, the two scales are measuring a common factor
or
dimension.
Recent evidence by Hammer et al.
(1963)
and Evans
(1966)
indicates
that items representative
of
hypnotic phenomena may be multidimen-
sional: the usually assumed single factor in such scales cannot ade-
quately account for the common variance indicated by the item corre-
lations. Furneaux
(1964)
has presented evidence which questions the
application of factor analytic techniques to dichotomized items of the
present type.? Without conducting such an analysis, there is consider-
able evidence from the present results that the two scales are measur-
ing more than a single dimension.
If
the scales measure common dimensions, then the correlation be-
tween HGSHS:A and SHSS:C should be equal, statistically,
t o
the
maximum limit to such a correlation set by the respective reliabilities
of the scales (.74 and .75). The maximum possible correlation between
HGSHS:A and SHSS:C is .745: the obtained value of
.59
is signifi-
cantly less than this maximum
( p
<
.0125,
one-tailed).
At
least one
of the scales is measuring attributes not measured by the other, al-
though they do measure common attributes to a large extent.8
Descriptively, about
55%
of
the total variance in performance on
each scale is reliable variance. About 35% of the reliable variance is
accounted for by the correlation between the tests,
or
the attributes
measured in common, and
20%
of the reliable variance is accounted
for by attributes not common to the scales.
There is strong evidence in the present study that passive motor
(primary) suggestibility should be distinguished from challenge sug-
‘Furneaux (1964) has shown that the correlations among items similar to the
items in the two scales depend upon item difficulty. When an easy item is
di-
chotomized
so
that fewer
Ss
pass
the item, then the correlation between the item
and other measures of hypnosis is higher when the item is made harder. The
regressions of the continuous variables underlying the dichotomized items are
not linear.
A
similar result would probably occur
if
items
in
those scales were
dichotomized
using
different criteria, for there is a marked negative correlation
between proportion of
Ss
passing
each item and item validity coefficients (biserial
correlations of item with adjusted total scores). Typical
rank
order values are:
SHSS:A, -.75 (calculated from Weitrenhoffer
&
Hilgard, 19591, HGSHS:A,
-.17 (calculated from Shor
&
E.
Ome, 1963)
and
-60
(present sample); SHSS:C,
-.45 (calculated from Weitsenhoffer
&
Hilgard, 1962) and 3 9 (present sample).
’
It
is
noted that
a
Similar conclumon can
be
derived from resulta reported by
Weitsenhoffer and Hilgard (1962). The reliabilities of SHSS:A and
SHSS:C
were reported to be
.83
and
85
respectively. If the two scales
are
measuring the
same dimension they would correlate
.84.
The obtained correlation between
SHSS:A
and SHSS:C of
.72
is
significantly less than
this
maximum value
(N
=
203,
p
<
,001).
Downloaded by [University of Macedonia] at 04:28 30 March 2012