More than three decades of cultivation research support small,
though significant, effects of television (TV) exposure on perceptions
of social reality (e.g., Morgan & Shanahan, 1996). Theoretically,
the most commonly cited explanation for such effects focuses on the
cognitive processes underlying them (e.g., Shrum, 1995), specifically
the accessibility of constructs from memory as a function of the
heuristic processing of media fare. Although much cultivation literature
focuses on determining the key variables that might influence the
relationship between TV exposure and perception of the social world, the
potential for personality traits to influence the elicitation of
cultivation effects has been ignored. This is surprising, given that
personality traits, like TV exposure, likely influence construct
accessibility. Given the centrality of accessibility in current
cultivation theorizing, this research examines how certain personality
traits might impede (or facilitate) the cultivation process. As such,
they have the potential to illuminate the conditions under which
stronger cultivation effects might be expected or, conversely,
conditions under which cultivation effects might be unlikely to be
found.
Cultivation Theory
Cultivation theory asserts that common conceptions of reality are
cultivated by overall patterns of TV programming to which communities
are regularly exposed over long periods of time (Gerbner, 1969; Gerbner,
Gross, Morgan, Signorielli, & Shanahan, 2002). Gerbner and his
colleagues propose that compared to light TV viewers, heavy viewers are
more likely to perceive the world in ways that mirror reality as
presented on TV rather than more objective measures of social reality.
Researchers have tested and found support for the cultivation hypothesis
in a range of contexts (e.g., Morgan & Shanahan, 1996); however, the
bulk of cultivation research focuses on TV violence and its effects on
perceptions of real-world incidences of crime and victimization (e.g.,
Potter, 1993, for a review). Numerous content analyses have documented
that the number of violent acts on American network TV greatly exceeds
the amount of real-world violence (e.g., Diefenbach & West, 2001).
In turn, heavy TV viewers: (a) overestimate the incidence of serious
crime in society (first-order effects), and (b) are more likely to
believe that the world is a mean place where people cannot be trusted
and are just looking out for themselves (second-order effects; Gerbner
et al., 2002; Gerbner, Gross, Morgan, & Signorielli, 1980).
Although originally a more sociologically based theory, cultivation
theorizing has taken a decidedly psychological approach in recent years.
Shrum's research program in particular offers evidence that
overestimates of crime prevalence are likely the result of heuristic
processing of TV programming, allowing TV-based constructs to enjoy
higher accessibility in the minds of heavy viewers (e.g., Shrum, 2001).
That is, because heavy viewers are recently and frequently exposed to
certain common images and themes on TV, those themes become more
accessible in memory and thus more influential in making judgments, like
violence prevalence estimates.
Several methodological and conceptual critiques of cultivation
(e.g., Doob & Macdonald, 1979; Potter, 1993) have motivated
consideration of potential moderators of the cultivation effect. Most
notable is personal experience (Doob & Macdonald, 1979; Shrum &
Bischak, 2001; Weaver & Wakshlag, 1986), which has been proposed to
have two potential effects. Resonance suggests cultivation effects may
be amplified in situations where viewers have more real-world experience
whereas mainstreaming suggests that TV exposure might override
differences in perspectives that might ordinarily result from personal
experiences (e.g., Gerbner et al., 2002).
In addition to personal experience, the cultivation literature has
revealed several other variables that moderate the cultivation effect,
including viewing motivations (e.g., Carveth & Alexander, 1985);
attention level; need for cognition (Shrum, Burroughs, &
Rindfleisch, 2005); nation of origin (Woo & Dominick, 2003); and
elaboration styles (Shrum, 2001). Despite these advances, however,
scholars argue for the need to investigate other potential moderators of
the relationship between mass media exposure and crime perceptions
(e.g., Shrum & Bischak, 2001). In considering viable directions for
such investigations, it seems reasonable to step back and ask:
On what premises are cultivation effects based, and what about the
receiver is likely to shape those outcomes? As the cultivation paradigm
is predicated on the assumption that TV socializes individuals from
infancy and thus its effects predate most experiences, individual
differences that predate TV exposure would be of considerable interest.
Chief among these are personality traits, which may not only impact
processing of TV messages but may also influence construct
accessibility, the prime mechanism posited to underlie cultivation
effects.
Personality Traits and Cultivation Effects
According to H. J. Eysenck (1990), personality is a hierarchical
structure comprised of single and habitual cognitions or acts,
dimensions or patterns of thoughts and feelings (i.e., traits), and
intercorrelations between traits. Because personality traits are
relatively enduring over time, they are generally presumed to be
governed by genetic factors (H. J. Eysenck, 1992; for a review, see
Matthews, Deary, & Whiteman, 2003). However, there is some evidence
that environment may play a role in shaping personality over the life
span (e.g., Roberts, Walton, & Viechtbauer, 2006). H. J. Eysenck
(1990) argues for three broad personality dimensions: neuroticism,
psychoticism, and extraversion. According to Eysenck, neuroticism
represents emotional instability, and its component traits include
anxiety, moodiness, tension, shyness, and low self-esteem. Psychoticism
represents emotional independence and is marked by a hostile
disposition, lack of empathy, impulsivity, creativity, and
tough-mindedness. In contrast, extraversion is characterized by
sociability, activity, and sensation seeking.
According to Wober (1986), personality traits have been remarkably
understudied in media research up until the mid 1980s. Since then,
researchers have focused largely on how personality traits impact media
selection and preference (e.g., Bruggemann & Barry, 2002; Weaver,
1991; Weaver, Brosius, & Mundorf, 1993), with much less attention
paid to the ways in which personality variables shape media effects
processes and outcomes. Still, some evidence exists in this domain. For
example, Gunter found that neuroticism (Gunter & Furnham, 1983) and
psychoticism (Gunter, 1983) enhance perceptions of violent TV content.
Relatedly, Zillmann and Weaver (1997) found that after exposure to
gratuitous violence, high psychotic males were more accepting of
violence as a means of conflict resolution. These findings suggest that
personality traits impact the ways in which individuals respond to mass
media messages. However, such traits fail to be integrated into broader
theoretical frameworks. The goal of this research is to incorporate
notions of personality trait into the cultivation paradigm.
Although cultivation research rarely includes personality variables
(though Bryant, Carveth, & Brown, 1981, offer a notable exception),
recent theoretical advancements lay the foundation for this sort of
integration. Recall that Shrum's (1995) model posits that TV
exposure leads to first-order cultivation effects because it makes TV
images and themes readily accessible in memory. Social cognition
research, however, has demonstrated that both external stimuli and
internal factors have the ability to affect construct accessibility
(Bargh, Lombardi, & Higgins, 1988). Further, certain constructs may
become chronically accessible, even in the absence of a prime, when a
person has frequent and consistent related experiences given their
unique life history and social encounters (e.g., Higgins & King,
1981; Higgins, King, & Mavin, 1982).
Here, it is argued that personality traits represent internal
factors that affect one's unique life history and social encounters
and, therefore, promote chronic construct accessibility. This view is
supported by the evidence that personality traits are related to
cognitions (Calvo & Castillo, 2001 b), opinions and attitudes
(Zillmann & Weaver, 1997), behavioral tendencies (McCroskey, Heisel,
& Richmond, 2001), and direct experiences (McCrae, 1996). If, as
asserted, certain personality traits predispose individuals to have
related constructs already chronically accessible, it would limit
TV's opportunity to impact accessibility. Thus, it is expected that
cultivation effects are somewhat dependent on personality traits. Below
key traits related to H. J. Eysenck's (1990) personality dimensions
are elaborated and predictions are offered as to how these traits might
interact with TV viewing to affect cultivation outcomes.
Neuroticism/Trait Anxiety
Generally speaking, those high in neuroticism exhibit cognitive
biases towards negative, dangerous thoughts; experience more stress; and
tend to have lower life satisfaction than low neurotics (for a review,
see Matthews et al., 2003). Neuroticism's subcomponent of trait
anxiety is of particular interest as it has been a focus in the media
effects literature (Bryant et al., 1981). By definition, trait anxiety
is a person's proneness to perceive potential harms in his or her
environment (H. J. Eysenck, 1992). Research on the relationship between
anxiety and cognitive processes has shown that trait-anxious individuals
exhibit an attentional bias towards threatening stimuli (MacLeod &
Matthews, 1988), are more likely to jump to negative conclusions when
presented with ambivalent information (Calvo & Castillo, 2001 a),
and judge ambiguous situations as threatening (Calvo & Castillo,
2001b). Further, they view their personal experiences as more
threatening as they scan their environments for potential dangers (H. J.
Eysenck, 1992).
Given their cognitive biases towards negative thoughts and
threatening situations, high trait-anxious people are likely to have
danger-related constructs chronically accessible. Consequently, exposure
to threatening, violent media messages is unlikely to make such
constructs any more accessible than they already are. Thus, for high
trait--anxious individuals, there should be little relationship between
TV viewing and violence-related cultivation outcomes. Individuals low in
trait anxiety, on the other hand, do not see the world as dangerous and
threatening, and are not likely to have related constructs accessible.
Therefore, exposure to the violent world of TV should increase the
accessibility of danger and crime-related constructs, leading to the
traditional cultivation effect. Thus, the following interaction is
expected:
[H.sub.1]: Trait anxiety will interact with TV viewing such that
heavy TV viewing will result in a stronger cultivation effect for those
low, versus high, in trait anxiety.
This prediction is somewhat different from what might be expected
from a resonance standpoint. Shrum and Bischak (2001) have argued that
the media can exert an additive impact on construct accessibility, above
and beyond that of direct experience. From this perspective, heavy
exposure to violent images would resonate with the threatening,
dangerous thoughts and perceptions of high trait-anxious individuals,
leading to an even stronger cultivation effect for them compared to
those low in trait anxiety. Thus, the resonance hypothesis would predict
an interaction between trait anxiety and TV viewing, but in the opposite
direction as that proposed above, resulting in the following alternative
hypothesis:
[H.sub.1alt]: Trait anxiety will interact with TV viewing such that
heavy TV viewing will result in a stronger cultivation effect for those
high, versus low, in trait anxiety.
Psychoticism
High psychotic individuals tend to be neither sympathetic nor
empathetic towards others (Richendoller & Weaver, 1994). They are,
however, curious about morbid events (Zuckerman & Litle, 1986),
argumentative (McCroskey et al., 2001), and likely to embrace violent
conflict resolutions (Zillmann & Weaver, 1997). Further, high
psychotics seem to select media heavy in violence (Weaver, 1991), and
tend to see violent media as more amusing, interesting, and exciting
than low psychotics (Bruggemann & Barry, 2003). In fact,
considerable research links high psychoticism to antisocial behavior and
criminality (H. J. Eysenck, 1996).
Combined, this research suggests that high psychotics are likely to
have aggressive, violent, and hostile constructs chronically accessible
(e.g., Zillmann & Weaver, 1997). If so, an accessibility ceiling for
high psychotics that limits the potential for cultivation effects is
expected. For low psychotics, however, TV exposure should increase the
accessibility of aggression-related constructs and thus enhance
cultivation effects. Thus, it is proposed:
[H.sub.2]: Psychoticism will interact with TV viewing such that
heavy TV viewing will result in a stronger cultivation effect for those
low, versus high, in psychoticism.
An alternative prediction can also be offered based on the
resonance hypothesis. That is, the violent imagery on TV might resonate
with the violent world perceptions of high psychotics, amplifying the
cultivation effect by even further increasing the accessibility of
violence-related constructs. Thus,
[H.sub.2alt]: Psychoticism will interact with TV viewing such that
heavy TV viewing will result in a stronger cultivation effect for those
high, versus low, in psychoticism.
Extraversion/Sensation Seeking
Within the broader personality dimension of extraversion, media
researchers have emphasized the trait of sensation seeking, particularly
in studies of media violence. According to Zuckerman (1979), sensation
seeking is a biologically based personality trait that represents
"... the need for varied, novel, and complex sensations and
experiences and the willingness to take physical and social risks for
the sake of such experiences" (p. 10). There is conflicting
research, however, about the chronic accessibility of constructs
pertaining to crime and violence-related topics for high sensation
seekers. On one hand, high sensation seekers are more aggressive
(Zuckerman, Kuhlman, Joireman, Teta, & Kraft, 1993), more curious
about morbid events (Zuckerman & Litle, 1986), more attracted to
dangerous behaviors (Roberti, 2004), and more likely to have committed a
nonviolent crime than low sensation seekers (Horvath & Zuckerman,
1993). Thus, high sensation seekers might have aggression and
violence-related constructs chronically accessible. However, it is
likely that sensation seekers seek out dangerous situations because they
fail to recognize the potential harmful consequences of their actions
(Joireman, Anderson, & Strathman, 2003), which suggests they do not
have danger-related constructs accessible. Indeed, high sensation
seekers are more likely than low sensation seekers to view the world as
nonthreatening and to fail to perceive dangerous stimuli (Roberti,
2004).
Based on this evidence, high sensation seekers should be less
likely than low sensation seekers to have danger-related constructs
chronically accessible. Therefore, exposure to the often extreme dangers
portrayed on TV likely make violence-related constructs more accessible
for high sensation seekers than they might have been otherwise.
Consequently, high, but not low, sensation seekers, should be more prone
to violence-related cultivation effects. Thus, the following interaction
is proposed:
[H.sub.3]: Sensation seeking will interact with TV viewing such
that that heavy TV viewing will result in a stronger cultivation effect
for high, versus low, sensation seekers.
Again, an alternative prediction derived from the resonance
hypothesis can be offered. That is, for low sensation seekers, TV's
mean and dangerous world might resonate with their perception of the
world as dangerous, leading to an enhanced cultivation effect relative
to that evidenced for high sensation seekers. Thus, it is proposed:
[H.sub.3alt]: Sensation seeking will interact with TV viewing such
that heavy TV viewing will result in a stronger cultivation effect for
low, versus high, sensation seekers.
Personality Trait and Selective Exposure
The above hypotheses are predicated on the assumption that
personality traits influence the degree of cultivation possible due to
their effects on the chronic accessibility of violence-related
constructs. However, it could be argued that any cultivation differences
found might be the result of the traits' impact on the selection of
media content. Indeed, some evidence suggests that personality traits
influence media preferences, though the literature presents somewhat
inconsistent findings. For example, high sensation seekers seem to
prefer media messages with high sensation value (i.e., fast-paced,
novel; Stephenson & Palmgreen, 2001), and they seek out and enjoy
horror and X-rated movies (Zuckerman & Litle, 1986),
action-adventure (Nolan & Patterson, 1990; Perse, 1996), music
programs (Perse, 1996), news (Zillmann, 1991), sports (Krcmar &
Greene, 1999; Nolan & Patterson, 1990), and even comedies and game
shows (Nolan & Patterson, 1990). Yet, Rowland, Fouts, and Heatherton
(1989) found that after controlling for age, sensation seeking did not
relate to program content preference. Further, Krcmar and Greene (1999)
found contradictory evidence with some dimensions of sensation seeking
positively associated with watching sports but negatively related to all
violence.
Evidence linking trait anxiety with media preferences is also
inconclusive. Some studies suggest that anxious individuals are more
likely to select and enjoy crime dramas in which justice triumphs
(Wakshlag, Vial, & Tamborini, 1983), and research on the
"parent" trait of neuroticism shows that high neurotics watch
more tragic movies, listen to downbeat music, and watch news and
information programs (Weaver, 1991). On the other hand, Weaver et al.
(1993) found no links between neuroticism and movie genre preferences.
A clearer usage pattern emerges for psychoticism. Weaver (1991)
found that high psychotics were less interested in comedy but strongly
preferred violent horror movies (see also Bruggeman & Barry, 2002,
and Weaver et al., 1993), and they found the violence to be more
enjoyable, interesting, and humorous than low psychotics (Bruggemann
& Barry, 2002). Thus, high psychotics seem to seek out and enjoy
violent media while avoiding lighthearted options.
Although this line of research is far from complete, it gives
reason to suspect that personality traits might drive media exposure
patterns, which could potentially explain any differences in cultivation
patterns identified. To explore this possibility, it is asked:
[RQ.sub.1]: Do differences in genre consumption explain any
cultivation differences associated with personality traits?
Method
Participants and Procedures
Surveys were completed by 433 undergraduates during class periods
in exchange for course extra credit. Of these, 6 reflected outlying
responses on the key dependent measure of first-order beliefs and were
thus excluded from analysis, leaving 427 valid cases. The sample was 57%
female. The participants' average age was 21.15 (SD = 2.83), and
81% were Caucasian.
Although college student samples are not necessarily ideal, they
are still suitable for cultivation investigations focused on
accessibility as there is little about college students to suggest they
differ from more general samples of adults in their cognitive processes
(see Shrum, 2001, for a more detailed discussion on these matters). As
to the nature of this investigation, which focuses on issues of violence
in particular, 19% of the sample reported having personally been a
victim of a violent crime and 50% reported that a family member or close
friend had been a victim of violent crime. As these figures do not
include other forms of violence the students might also have experienced
(e.g., physical abuse in family or dating relationships), it appears
many of the students have the sorts of experiences that would allow for
a fair test of the alternative, resonance-based hypotheses.
Measures
Measures are presented in the order in which they appeared in the
survey. First- and second-order beliefs were assessed first. The former
represents prevalence estimates of events in the real world whereas the
latter represents value judgments or general attitudes about the state
of the world at large (Shrum, 1995). Of note, scholars have found more
consistent evidence for first-order, rather than second-order, effects
(Hawkins & Pingree, 1982; Potter, 1991a, 1991b). However, in the
interest of comprehensiveness, both are examined.
Violence Prevalence Estimates. Fourteen belief questions used in
past research (e.g., Nabi & Sullivan, 2001) measured
respondents' estimations of crime frequency. Of these, 10 focused
on crime generally and 4 on personal likelihood of victimization.
Responses were translated into z scores and combined to create a
"relative estimation" scale of each participant's
tendency to over- or underestimate crime prevalence relative to the
sample as a whole. The 14 items formed a reliable scale of overall crime
prevalence estimations ([alpha] = .84). Separate scales based on the 10
societal estimate items ([alpha] = .80) and the 4 personal victimization
likelihood estimate items ([alpha] = .79) were also created as these
have been shown to have different effects in past research (e.g., Shrum
& Bischak, 2001; Tyler, 1980).
Mean world attitude was assessed with the three classic mean world
items, but in anticipation of their poor reliability ([alpha] = .61)
given the scale's history, eight 5-point Likert items were included
based on the original mean world scale. Of these, seven formed a
single-factor reliable measure ([alpha] = .77), and thus served as the
measure of second-order beliefs. The seven items included: Most people
will try to take advantage of you if given the chance; Generally
speaking, the world is a dangerous place; Most people would not bother
to go out of their way to help someone; In general, you cannot be too
careful in dealing with people; Most people are basically honest; Most
of the time people try to be helpful; and Generally speaking, most
people can be trusted (the latter three items were reverse coded).
Although not often assessed in cultivation research, behavioral
intentions and behaviors are of interest as the concern over cultivation
effects often stems from the effects that altered perceptions of the
real world as a function of TV viewing might have on viewers'
actions. Thus, for exploratory purposes, both were assessed consistent
with Nabi and Sullivan (2001). Protective behavioral intentions were
measured by asking respondents how likely they would be to take more or
less risky action in response to eight different scenarios that could
present some personal danger ([alpha] = .75). Protective behavior was
measured by asking, on 5-point scales, how often the respondents
generally take each of 16 precautions against crime victimization
([alpha] = .77).
Three personality traits were then assessed. Trait anxiety was
measured with the 20-item trait anxiety inventory (Spielberger, Gorsuch,
& Lushene, 1970; [alpha] = .89). Sensation seeking (SS) was assessed
with Ferguson, Valenti, and Melwani's (1991) 10-item measure of
impulsivity and adventurous risk taking ([alpha] = .87). These items are
more general than Zuckerman's (1979) measure and are more
consistent with past media research in which thrill/adventure-seeking
and disinhibition have been predictive of media selection (Krcmar &
Greene, 1999). Finally, psychoticism was measured with Eysenck, Eysenck,
and Barrett's (1985) 12-item revised psychoticism scale ([alpha] =
.66). Of note, correlations among the traits were generally small (rs =
.07 -.33; see Table 1), and thus they can be considered separate
constructs.
To assess TV viewing hours, respondents indicated how many hours of
TV they watched during each of four time periods (6 a.m.-noon, noon-6
p.m., 6 p.m.-midnight, midnight-6 a.m.) during the average weekday, the
average Saturday, and the average Sunday. These data were combined
(weighting the "average weekday" questions by a factor of 5
compared to the "Saturday" or "Sunday" questions) to
create an "average TV viewing hours/day" measure. The
sample's mean daily TV viewing hours was 4.36 (SD = 2.66), which is
comparable to Nielsen data on the average adult's daily TV viewing
of 4.45 hours (Nielsen, 2000).
Rubin's (1981) perceived TV realism scale was also included.
Of the five items, four formed a single factor scale with notably higher
reliability than the full scale ([alpha] = .79 vs. [alpha] = .64). For
this reason, the four-item index was used in the analyses.
Respondents then rank-ordered their preference for 10 genres of TV
programming (comedy, drama, news, soap operas, sports, daytime talk
shows, late-night talk shows, documentaries, action adventure, and game
shows) and indicated their estimated number of weekly viewing hours for
each programming type. Finally, respondents were asked to list up to six
TV programs that they make efforts to watch every week. These programs
were then categorized into 16 genres (including the 10 noted above, with
the exception of action-adventure, and including competitive
reality-based TV, noncompetitive reality-based TV, news magazine
programs, entertainment news, cartoons, music TV, and "other")
by two coders blind to the purposes of the study. Coder reliability was
high (Kappa = .95).
Finally, age and race were assessed along with personal experience
as a victim of violent crime (no = 0; yes = 1). Enjoyment of violent
programming was then assessed with two items (I enjoy TV shows more if
they have lots of excitement, like fights and explosions; As far as I am
concerned, the more violent a TV show is, the better; r = .63, p <
.001).
Results
For hypothesis testing, three regression analyses were conducted
with comparable structures. Demographic and past experience variables
typically included as controls in cultivation studies were entered in
Block 1 : age, race (0 = non-White; 1 = White), gender (0 = male; 1 =
female), past victimization experience (0 = not victimized; 1 =
victimized), and perceived TV realism. Enjoyment of violent media was
also included given its association with the predictor and dependent
measures (e.g., SS r = .11; psychoticism r = .28; daily TV viewing r =
.23; behavior r = -.19). The key cultivation-related TV viewing measure
(daily TV viewing hours) and the targeted personality trait (as a
continuous measure) were then entered in Block 2, and the interaction
between the latter variables were entered in Block 3. TV viewing hours
and personality trait interactions were created by centering each
variable based on its mean and multiplying the transformed variables.
All partial correlations among the traits and dependent measures are
reported in Table 1.
Cultivation effects have consistently been recognized as small in
magnitude, both generally (average r = .09) and for violence
specifically (average r = .103, Morgan & Shanahan, 1996). This study
has power of approximately .55 to detect a small effect of r = .10, and
.98 to detect an effect of r = .20 at p = .05.
Television Viewing and Prevalence Estimates of Violence
A hierarchical regression with demographics, perceived TV realism,
and liking violent programming entered in Block 1 ([R.sup.2] = .12, p
< .001), and daily TV viewing hours entered in Block 2 indicated a
positive association between daily TV viewing and overall prevalence
estimates of violence-related occurrences ([beta] = .12, [R.sup.2] =
.012, p = .02). This suggests a general cultivation effect consistent
with past literature (e.g., Morgan & Shanahan, 1996). When
considering societal versus personal estimations separately, TV viewing
maintained a comparable association with societal estimates as reported
above ([beta] = .12, [R.sup.2] = .013, p = .02), but the association
between TV viewing and personal victimization estimates decreased
([beta]) = .07, [R.sup.2] = .005, ns). These data are consistent with
past research indicating that the cultivation effect holds more for
societal than personal estimates of victimization likelihood. Further,
there was a small association between daily TV viewing hours and mean
world beliefs, though it only approached standard conventions of
statistical significance ([beta] = .09, [R.sup.2] = .007, p = .08).
To determine if these relationships might in part be explained by
personality traits, the regressions were repeated with the three traits
entered in a separate block prior to TV viewing hours. Although the
traits combined to explain some variance in the overall prevalence
estimate ([R.sup.2] = .015, p = .07), societal estimate ([R.sup.2] =
.013, p = .10), and mean world perception ([R.sup.2] = .08, p < .001)
measures, the magnitudes of the TV viewing-prevalence relationships did
not meaningfully change (overall [beta] = .11, p = .03; societal [beta]
= .11, p = .03; personal [beta] = .07, ns). The TV viewing-mean world
relationship, however, did decrease ([beta] = .03, ns).
Personality Traits and the Cultivation Effect
Trait Anxiety. [H.sub.1] and [H.sub.1alt] offered competing
interaction predictions regarding the relationship between trait anxiety
and heavy TV viewing. Compared to those low on trait anxiety, [H.sub.1]
suggested those high on trait anxiety would experience weakened
cultivation effects whereas [H.sub.1alt] suggested those high on trait
anxiety would experience enhanced cultivation effects. Regression
results indicated that the interaction between trait anxiety and daily
TV viewing hours was not significant for the overall prevalence
estimates (p = .19), societal prevalence (p = .47), or mean world
perceptions (p = .26). However, the interaction for the personal
victimization likelihood estimate was significant ([beta] = -.09, p =
.05; see Table 2). Consistent with [H.sub.1], the nature of the
interaction indicated that among heavier TV viewers, those higher in
trait anxiety evidenced less of a cultivation effect regarding personal
estimates of victimization likelihood than those lower in trait anxiety.
This is most clearly depicted with partial correlations between TV
viewing and personal victimization likelihood estimates for those high
in trait anxiety, [r.sub.p](208) = -.02, ns, and those low in trait
anxiety, [r.sub.p](199) = .1 7, p = .02, based on a median split on the
trait variable (see also Figure 1).
Psychoticism. [H.sub.2] and [H.sub.2alt] offered competing
interaction predictions regarding psychoticism and cultivation effects.
[H.sub.2] predicted those low in psychoticism would experience a
stronger cultivation effect whereas [H.sub.2alt] predicted that those
high in psychoticism would experience the stronger effect. The
interaction between psychoticism and daily TV viewing hours approached
significance for the overall prevalence estimates ([beta] = -.09, p =
.07) and societal estimates ([beta] = -.09, p = .07) but not personal
estimates (p = .28) or mean world perceptions (p = .32; see Table 2).
Consistent with [H.sub.2], the nature of the interactions suggested that
those high on psychoticism experienced weaker cultivation effects than
those low on psychoticism. This relationship is well depicted with
partial correlations between TV viewing and both overall and societal
estimates for those high in psychoticism, [r.sub.p](198) = .09, ns, and
[r.sub.p](198) = .06, ns, versus those low in psychoticism,
[r.sub.p](209) = .20, p = .003, and [r.sub.p](209) = .24, p = .001,
based on median splits on that variable. The discrepancy between those
high and low in psychoticism on the societal prevalence estimate measure
appears particularly meaningful (see Figure 2).
Sensation Seeking. [H.sub.3] and [H.sub.3alt] offered competing
interaction predictions regarding sensation seeking and cultivation
effects. [H.sub.3] predicted high sensation seekers would experience
stronger cultivation effects whereas [H.sub.3alt] predicted low
sensation seekers would evidence stronger effects. The regression
analyses did not reveal significant interactions for overall prevalence
estimates (p = .84), societal estimates (p = .58), or mean world
perceptions (p = .61; see Table 2). The interaction with personal
prevalence estimates was weak ([beta] = .07, p = .12), though suggestive
that high sensation seekers may experience greater cultivation effects,
[r.sub.p](211) = .13, p = .055, than low sensation seekers,
[r.sub.p](196) = .00, ns, based on the variable's median split. To
the extent this difference might be meaningful, it supports [H.sub.3].
[FIGURE 1 OMITTED]
Personality Traits and Television Exposure
The results thus far support the idea that violence-related
constructs that are chronically accessible due to personality traits
minimize cultivation effects rather than enhance them, as resonance
might predict. An alternative explanation for the cultivation
differences identified thus far, however, might be variations in
patterns of media exposure for those high and low on each personality
trait. Indeed, controlling for gender, age, race, and past
victimization, enjoyment of violent programming correlated with
sensation seeking ([r.sub.p] = .13, p = .006), and psychoticism
([r.sub.p] = .14, p = .005), though not trait anxiety ([r.sub.p] = .04,
p = .37). However, enjoyment is a poor substitute for actual exposure.
Thus, the relationships between personality traits and programming
exposure were explored in multiple ways.
[FIGURE 2 OMITTED]
First, each respondent's favorite genre was categorized as
either high violence/ conflict (drama, soap operas, daytime talk, news,
action adventure, sports) or low violence/conflict (situation comedies,
late-night talk, game shows, documentaries). Second, the number of hours
respondents reported viewing each of the 10 genres were collapsed into
high and low violence/conflict viewing hours as described above. Third,
the favorite programs listed were collapsed into high versus low
violent/conflict programming as follows: high violence/conflict (drama,
soap operas, daytime talk, news, news magazines, sports,
reality-conflict, cartoons) and low violence/conflict (sit-coms,
late-night talk, game shows, documentaries, reality-non conflict, and
entertainment news). Given their rarity, the music and "other"
categories were dropped.
Focusing on preferred genre, partial correlations indicated no
significant association between personality traits and preference for
more violent TV genres: anxiety [r.sub.p](390) = .02, ns; SS
[r.sub.p](390) = -.01, ns; psychoticism [r.sub.p](390) = -.02, ns. In
addition, partial correlations suggested that those high in anxiety
watched significantly more hours of violent programming, [r.sub.p](363)
= .16, p = .002, and those high in psychoticism watched more hours of
both violent, [r.sub.p](363) = .13, p = .01, and nonviolent programming,
[r.sub.p](363) = .09, p = .07. When controlling for daily TV viewing
hours, however, these associations were eliminated for psychoticism
([r.sub.p] = .00 and [r.sub.p] = .02, ns) and reduced for anxiety,
[r.sub.p] = .08, p = .12. Finally, partial correlations involving the
number of low and high violent/conflict programs watched regularly
indicated no significant relationships between trait anxiety, sensation
seeking, or psychoticism and preference for low conflict, [r.sub.p](421)
= -.04 to .04, ns, or high conflict programming, [r.sub.p](421) = -.03
to .05, ns. Of note, the proportion of favorite programming likely to
contain conflict and violence was roughly equal for those high and low
on each personality trait (anxiety p = .73; SS p = .84; psychoticism p =
.49). Thus, in answer to [RQ.sub.1], these data offer little support for
the notion that the personality trait-cultivation effect relationships
could be explained by selective programming exposure.
Discussion
This research sought to understand the role that personality traits
might play in the cultivation process, and specifically to identify
conditions under which stronger (or weaker) effects than ordinarily
expected within the cultivation framework might occur. These data
indicated an overall cultivation effect for violence prevalence
estimates consistent with past research. They further revealed that,
consistent with predictions, personality traits may interact with TV
exposure under some circumstances, and in ways consistent with the
personality trait/chronic accessibility perspective, but not the
cultivation-based resonance viewpoint. In particular, evidence from both
the regression and correlation analyses suggested trait-anxious
individuals appeared less, not more, likely than their low
trait--anxious counterparts to evidence cultivation regarding
estimations of personal victimization likelihood. Although the
interactions did not reach standard levels of statistical significance,
the correlational evidence for psychoticism and sensation seeking
suggested differences in magnitude of cultivation effects for those high
and low on these traits consistent with the main hypotheses. Those lower
in psychoticism evidenced a significant cultivation effect in terms of
societal estimates of crime and violence whereas those higher in
psychoticism did not, and high sensation seekers evidenced a borderline
significant cultivation effect regarding personal victimization
likelihood estimates whereas low sensation seekers did not. Further, the
magnitudes of these relationships were consistent with or greater than
those found in previous cultivation research. However, one must be
cautious in interpreting the correlational evidence given the
interaction results fell outside the conventional significance level of
p < .05.
To the extent the identified differences are real, the data offered
no evidence to suggest that they might stem from differences in the
consumption of particular TV genres. Although those high in trait
anxiety seemed to watch more violent/conflictual programming, this would
suggest high trait-anxious individu