The Measure of Paranormal Beliefs
After recoding the items that were negatively worded, we combined the scores
across the 20 items of paranormal beliefs to form an additive index. Cronbach’s
alpha on this index was .88. Evidence for the measure’s validity is indicated by the
fact that it was significantly correlated with the tendency for respondents to report
that they had experienced a paranormal event in their own life [r
ϭ .46, n ϭ 190; p Ͻ
.001]. Sparks, et al. (1997) found the same correlation (r
ϭ .47) in their study. In
order to determine if the structure of this measure was multi-dimensional, the 20
items were submitted to a maximum likelihood factor analysis with varimax
rotation. As observed by Sparks, et al. (1997), five factors emerged initially from this
analysis with eigenvalues greater than 1.0. However, in this case, the first factor
accounted for 30% of the variance and none of the remaining factors accounted for
at least 10% of the variance. Moreover, the items loading on this first factor were not
easily interpretable and the reliability of an additive index consisting of these items
(.60) failed to meet the conventional minimum established for alpha (.70). Conse-
quently, we decided to employ the entire 20-item measure as our main measure of
paranormal beliefs. It should be noted that the results of the factor analysis do
diverge somewhat from those obtained by Sparks, et al. (1997), who found an
interpretable, two-factor solution for this measure, both of which formed reliable
sub-scales.
The Measures of Television Viewing
Two measures of viewing were constructed from the responses. The first measure
was a “total viewing” measure in hours-per-week and is described above. Only six
respondents in the sample (3%) reported viewing no television at all during a typical
week. At the other extreme, one viewer reported viewing 74-hours of TV during a
typical week (10
ϩ hours per day). The median number of hours viewed per week
was 18, or about 2–3 hours-per-day. A second viewing measure was designed to
assess viewing of programs that were known to feature paranormal phenomena
regularly. The number of times that respondents indicated seeing each of the
programs was summed together to form a total measure for paranormal program-
ming.
Testing the Hypothesis and Research Questions
In order to test the first hypothesis, we initially computed correlations between the
measures of TV viewing and the measure of paranormal beliefs. The measure of
total TV viewing was significantly correlated with paranormal beliefs [r(190)
ϭ .19,
p
Ͻ .01]. Consistent with H1’s prediction that this relationship should be more likely
for the viewing of paranormal TV programs, we found that the measure of
paranormal TV viewing was significantly correlated with paranormal beliefs [r(191)
ϭ
.31, p
Ͻ .001]. Since the second research question sought information about the
impact of demographic variables on the relationship between TV exposure and
paranormal beliefs, we ran two regression equations to explore this issue— one for
each of the two viewing measures. The independent variables for each equation
were identical. Age, sex, income, education, weekly attendance at a religious service,
and intensity of religious belief were all entered into the equation in a single block.
This was followed by entering, in respective equations, either the total TV viewing
measure, or the measure of paranormal viewing. This permitted us to examine how
106
COMMUNICATION MONOGRAPHS
much variance in paranormal beliefs could be accounted for by TV viewing after
controlling for these demographic factors. Table 2 displays the results of the
equation using the total TV viewing measure as a predictor variable; Table 3
displays the results using the measure of paranormal viewing.
As Table 2 reveals, the demographic variables accounted for 16% of the variance
in paranormal beliefs [F (6,175)
ϭ 5.56, p Ͻ .001]. Total TV viewing accounted for
an additional 2% of the variance, but this was not enough to meet the conventional
level of significance [p
Ͻ .07]. The entire regression model accounted for 18% of the
variance in paranormal beliefs [F (7,174)
ϭ 5.31, p Ͻ .001]. The similar equation for
paranormal TV viewing (Table 3) shows that viewing of paranormal TV accounted
for an additional 4% of the variance in paranormal beliefs [p
Ͻ .003]. The entire
regression model accounted for 20% of the variance in paranormal beliefs [F (7,175)
ϭ 6.16, p Ͻ .001].
The equations in Tables 2 and 3 reveal information pertinent to RQ2. Age,
income, weekly attendance at a religious service, and intensity of religious belief
proved to be unrelated to paranormal beliefs. However, sex and education did
predict belief in the paranormal. The signs of the beta coefficients indicate that
females were more likely to express belief in the paranormal than were males and
people with lower levels of education were more likely to express belief than were
people with higher levels. The effect associated with education is much stronger than
the effect for sex.
TABLE 2
R
EGRESSION
R
ESULTS FOR
P
REDICTING
B
ELIEF IN THE
P
ARANORMAL FROM
T
OTAL
T
ELEVISION
V
IEWING
Variables Entered
Multiple R
R
2
Beta
Step 1:
Age
Ϫ.06
Sex
.17* (
p
Ͻ .05)
Income
.07
Education
Ϫ.32* (p Ͻ .001)
Weekly Religious Service
Ϫ.13
Intensity of Religious Belief
.07
.40
.16
Step 2
Total TV Viewing
.42
.18
.13
Note. The entire regression model was significant [
F (7,174)
ϭ 5.31; p Ͻ .001].
TABLE 3
R
EGRESSION
R
ESULTS FOR
P
REDICTING
B
ELIEF IN THE
P
ARANORMAL FROM
P
ARANORMAL
V
IEWING
Variables Entered
Multiple R
R
2
Beta
Step 1:
Age
Ϫ.07
Sex
.17* (
p
Ͻ .05)
Income
.07
Education
Ϫ.31* (p Ͻ .001)
Weekly Religious Service
Ϫ.12
Intensity of Religious Belief
.07
.39
.15
Step 2
Viewing Paranormal
.45
.20
.22* (
p
Ͻ .003)
Note. The entire regression model was significant [
F (7,175)
ϭ 6.16; p Ͻ .001]. The betas for Step 1 are slightly
different than the ones listed in Table 1 due to the fact that one respondent had missing data for the analysis in
Table 1 and was not included in the analysis.
107
TELEVISION AND PARANORMAL BELIEFS