QOOl

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the
thesis entitled
Violence on American Television:
Do the Perpetrators' Prosocial Qualities Matter?

presented by

Jennie M. Hwang

has been accepted towards fulfillment
of the requirements for

M. A. degree in Mitigation

 

Major professor

 

Date 5/9/01

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

PLACE IN RETURN Box to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE [ DATE DUE DATE DUE

 

OCT 9 6 (i002

 

90237 11:2 8le

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/C|RC/DatODue.l£5-p.15

i’l. K’IIIII-l l'.’

VIOLENCE ON AMERICAN TELEVISION:
DO THE PERPETRATORS’ PROSOCIAL QUALITIES MATTER?

By

Jennie M. Hwang

AN ABSTRACT OF A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF ARTS
Department of Communication

2001

Professor Stacy L. Smith

ABSTRACT

VIOLENCE ON AMERICAN TELEVISION:
DO THE PERPETRATORS’ PROSOCIAL QUALITIES MATTER?

By
Jennie M. Hwang

The present study content analyzed prosocial perpetrators on American television.
By conducting a secondary analysis of the data collected from the 3rd year of the National
Television Violence Study (Smith et al., 1998), the prevalence and context of prosocial
versus nonprosocial violence was assessed. Differences in prosocial and nonprosocial
violence were also measured in prime time and children’s shows. Overall, the findings
revealed that prosocial perpetrators engaged in less violence than did nonprosocial
perpetrators in terms of sheer amount. In regards to context, prosocial perpetrators were
more likely to be white humans, engage in violence to protect lives of others, and the
violence was usually shown as justified. Similar trends emerged in prime time.
However, prosocial violence in prime time resulted in less short- and long-tenn harm
than did nonprosocial violence. Children’s programs were more likely to feature
prosocial perpetrators in their childhood or teenaged years and commit justified violent
behaviors. These and all other findings were discussed in terms of risk exposure to

prosocial perpetrators may pose to the audience.

TABLE OF CONTENTS

INTRODUCTION ................................................................................... 1
Content Analyses ................................................................................ 4
NTVS and the Present Study .................................................................. 7

METHOD ............. i ............................................................................... 9
Sample ............................................................................................ 9
Definition of Violence ......................................................................... 9
Units of Analysis ............................................................................... 10
Contextual Measures .......................................................................... 11

PAT Level Variables ............................................................ 11
Scene Related Variables ........................................................ 13
Program Level Variables ....................................................... 15
Coder Training and Reliability .............................................................. 16

RESULTS ........................................................................................... 17
Analysis Plan .................................................................................. 17
Research Question 1 .......................................................................... 18
Research Question 2 .......................................................................... 24
Research Question 3 .......................................................................... 27

DISCUSSION ...................................................................................... 32
Footnotes ....................................................................................... 41

APPENDICES ..................................................................................... 45

REFERENCES ..................................................................................... 49

Violence on American Television:
Do the Perpetrators’ Prosocial Qualities Matter?

Recently, the issue of television violence in this country has provoked
considerable concern (Friedrich-Cofer & Huston, 1986; Paik & Comstock, 1994). The
reason people are worried about TV violence is that exposure to such content may be
teaching our children to act aggressively. In fact, one public opinion poll shows that 40%
of Americans believe that exposure to media violence contributes to violent behavior in
society and an additional 28% see it as a “considerable” factor (Plagans et al., 1991). A
recent survey by Annenberg Public Policy Center also indicates that 81% of parents are
concerned about the violence their kids see in movies or on TV (Oldenburg & Healy,
1999). Consistent with these polls, over forty years of social science research reveals that
viewing TV violence may increase the risk of learning aggressive thoughts, attitudes, and
behaviors (American Medical Association, 1996; American Psychological Association,
1993; Centers for Disease Control, 1991; US. Surgeon General, 1972).

Although exposure to TV violence may contribute to learning, it is also the case
that not all violent portrayals pose the same degree of risk to viewers. Some portrayals
may increase the risk of learning or enacting aggression whereas other portrayals may
decrease such risks. Indeed, research reveals that the context or way in which violence is
presented influences how viewers respond to televised acts of aggression (Gunter, 1985;
Wilson et al., 1997). For instance, the blockbuster film The Terminator glamorizes
aggression as a means of social problem solving whereas the movie Schindler ’s List

illustrates the moral, social, and psychological devastation caused by violence in society.

Surely, both films illustrate that the context or way in which violence is shown can
influence individuals’ reactions to the content differently.

It has been argued that one of the most important contextual factors may be the
nature of the perpetrator (Hoffner & Cantor, 1991; Reeves, 1979; Reeves & Grecnberg,
1977; Reeves & Miller, 1978). According to Social Cognitive Theory, “attractive”
perpetrators are more likely to be attended to and identified with than are unattractive
ones (Bandura, 1986). Thus, the influence of attractive perpetrators may pose
considerable risk for viewers’ learning and imitation of aggression.

But what qualities render a perpetrator attractive? One attribute that may be
important is the character’s prosocial orientation. Prosocial perpetrators are those who
act benevolently, help others, and are motivated to consider the needs of others before
themselves (Smith et al., 1998, p. 14). For example, Mel Gibson in the Lethal Weapon
series or Bruce Willis in Die Hard films are two prosocial perpetrators who use violence
to fight crime and/or drug infiltration in society. Prosocial perpetrators are not limited to
only human characters, however. Children’s programs such as the Powerpujf Girls and
Teenage Mutant Ninja Turtles often feature anthropomorphized characters that fight evil
forces attempting to destroy the earth. These examples underscore the idea that television
and film may be saturated with violent perpetrators who are portrayed in a prosocial light.

There is a great deal of evidence that children are attracted to shows featuring
prosocial perpetrators of violence. Nielsen data reveals that Power Rangers, a group of
prosocial karate chopping teens that fight alien creatures, is one of the most popular
programs in the early aftemoon among 2- to 11-year-olds (Stipp, 1993). Not only are

children watching this program, but they are also purchasing product-related toys and

games. By 1995, Bandai toys had sold roughly 6 million action figures from the show
and forced 14 production plants into overdrive to keep up with orders (Benezra, 1995).

Besides the triumph in the toy market, empirical research reveals more directly
that prosocial characters influence viewer liking. Studies show that characters or
individuals who possess positive attributes (e.g., altruistic, kind) are better liked than
those who possess negative attributes (e. g., selfish, cruel) (Berscheid, 1985; Hoffner &
Cantor, 1985; Liefer & Roberts, 1972; Wilson, Cantor, Gordon, & Zillmann, 1986;
Zillmann & Bryant, 1975). For instance, Zillmann and Cantor (1977) exposed children to
a short film with either a malevolent, benevolent, or neutral main character. Immediately
after exposure, children rated how much they liked the protagonist. The results showed
that the benevolent protagonists were liked significantly more than the malevolent ones.
Additionally, Cohen (1999) found that prosocial traits were the most prevalent reason
(34%) teens cited for choosing their favorite character on television.

Research also shows that prosocial characters can affect wishful identification.
Wishful identification refers to the process whereby a viewer desires to possess a
character’s attributes or lifestyle (von Feilitzen & Linne, 1975). Assessing the predictors
of WIShfiJl identification among 7- to l2-year-olds, Hoffner (1996) found a positive
relationship between prosocial attributes and wishful identification. That is, the more
prosocial traits (i.e., kind, helpful, and caring) the character possessed the higher the
wishful identification for that character. A negative relationship between antisocial
characteristics (i.e., mean, selfish) and wishful identification also was observed.

Interestingly, a character’s use of violence did m3 influence wishful identification. Thus,

at least this one study shows that prosocial characters are potent role models that younger
viewers identify with and desire to be like.

There is also evidence that children may imitate the behavior of prosocial
characters. In terms of research in the television violence arena, studies show that
prosocial perpetrators are more likely to be imitated than those perpetrators who lack
benevolent qualities (Boyatzis, Matillo, & Nesbitt, 1995; Liefer & Roberts, 1972). For
example, children exposed to a cartoon featuring violent superheroes were more likely to
act aggressively immediately after exposure than were those who watched violent
cartoons featuring perpetrators who were not superheroes (Liss, Reinhardt, & F redriksen,
1983). Other studies show that exposure to prosocial models can also increase children’s
task persistence, self-control (Stein & Friedrich, 1972), generosity (Bryan & Walbek,
1970), and even helping behavior (Sprafltin, Liebert, & Poulos, 1975).

In sum, the above research suggests that prosocial perpetrators may significantly
impact young viewers’ learning of aggression. That is, viewers may be more likely to be
attracted to and imitate the actions of prosocial perpetrators. Consequently, it becomes
important to assess the prevalence and context surrounding prosocial perpetrators on
American television. In the next section, I review all of those content analyses that have
quantified the relative presence of benevolent perpetrators across the television
landscape.

Content Analyses

To date, only a handfial of studies have content analyzed prosocial perpetrators

involved with violence on television. Perhaps the most notable study was conducted by

George Gerbner and his research team (Signorielli, 1990). These scholars sampled an

intact week of network prime time and weekend morning shows from 1967-1985.
Violence was defined as “the overt expression of physical force against self or other,
compelling action against one’s will on pain of being hurt or killed, or actually hurting or
killing” (Gerbner & Gross, 1976, p. 184). The findings revealed that weekend morning
programs were more likely to feature violence than were those prime-time programs
(90% vs. 72% respectively). In terms of characters, the results showed that 36% of all
men and 24% of all women involved with violence were good (e.g., positive or hero).

Greenberg, Edison, Korzenny, Fernandez-Collado, and Atkin (1980) also assessed
the amount of anti-social behavior on television. To this end, the researchers sampled a
single episode from all fictional prime time and Saturday morning television shows from
1975-78. Antisocial behavior was defined as “that which is psychologically or physically
injurious to another person or persons whether intended or not, and whether successful or
not” (1980, p. 102). The results showed that Saturday morning programs feature more
physically aggressive acts per hour than did prime time shows (23 vs. 12 acts per hour,
respectively). The findings also revealed that villains are four times more likely to
engage in physically aggressive acts per hour than are heroes (2.08 vs. .50 acts per hour,
respectively).

Looking more specifically at characters involved with violence, Potter and Ware
(1987) assessed the context surrounding violent heroes and villains on television. Two
intact weeks of prime-time programming across the three major broadcast networks (e.g.,
ABC, NBC, CBS) were sampled. The researchers defined an antisocial act as “any
attempt by one character to harm another character” (p. 672). Overall, the results

demonstrated that 8.1 physical acts of aggression occurred per prime-time hour. Further,

almost all of the violent actions of heroes are likely to be shown as rewarded (92%) and
justified (97%).

The previous content analyses suggest that there is a great deal of violence on
television, especially weekend mornings. Yet the past research reveals very little about
the nature of the characters involved with violence. At least two of the past studies
examined the prevalence of heroic instigators of violence on TV (Greenberg et al., 1980;
Potter & Ware, 1987). However, heroes may be depicted far less frequently on television
than prosocial perpetrators (Wilson et al., 1997; Wilson et al., 1998). Thus, we still do
not know how frequently prosocial perpetrators are portrayed on American television.

The previous research suffers from a few other limitations as well. First, the
definitions of violence or physical aggression used are very broad or liberal in nature
including acts such as natural disasters and invasions of privacy.1 Thus, the previous
estimates of violence on television are probably artificially inflated. Second, the samples
may not be representative of today’s viewing environment. The previous research was
conducted at least 15 to 30 years ago when there were only three broadcast networks on
television. Today, there are at least five broadcast networks and 70% of the nation
subscribes to cable programming (Nielsen Media Research, 2000). Consequently, we do
not know how often prosocial perpetrators are depicted in the current multi-channel
environment. Third, most of the previous research primarily focused on the amount of
violence and not its context. As noted above, the context or way in which violence is
shown on television is more important than sheer amount. Thus, a content analysis
assessing the context of violence surrounding prosocial violence on American TV has yet

to be conducted.

NTVS a_n_d the Present Study

Given the risk that prosocial perpetrators may pose, the goal of my thesis is to
content analyze their prevalence on television using the database from the third year of
the National Television Violence Study (Smith et al., 1998). This database was chosen to
overcome the limitations associated with previous research. First, the NTVS scholars
crafted a highly conservative definition of violence grounded in social science research.
That is, the definition counted only those intentional acts that were designed to physically
harm or injure a human or human like target. This conceptualization is quite consistent
with the stimuli used in most experimental research examining the impact of exposure to
media violence.

Second, the researchers sampled a composite week of television content from
October 1996 to June of 1997 across 23 of the most frequently watched broadcast and
cable channels from 6:00 am. to 11:00 pm. (PST). Clearly, this is the largest and most
representative sample of television content in the history of social science research.

Third, NTVS developed a coding scheme uniquely sensitive to assessing the
context of violence on television. In particular, the scheme measured eight major
contextual variables documented by research to either increase or decrease the risk of
learning. For instance, studies show that violence involving attractive perpetrators
(Bandura, 1986, 1994), justification (Berkowitz & Geen, 1966; Berkowitz & Rawlings,
1963), guns (Berkowitz, 1973; Berkowitz & LePage, 1967; Page & O’Neal, 1977),
extensiveness and graphicness (Leflcowitz et al., 1977; Swart & Berkowitz, 1976), little
or no consequences (Baron, 1971a, 1971b), rewards or no punishments (Bandura, Ross,

& Ross, 1963b; Lando & Donnerstein, 1978), humor (Baron, 1978; Berkowitz, 1970),

and realism (Atkin, 1983; Thomas & Tell, 1974), all increase the risk of learning or
enacting aggression.

Using this framework, the NTVS researchers found that 61% of all programs
contained violence. Further, a great deal of the violence on television was glamorized,
sanitized, trivialized, and not chastised. In terms of characters, a full 28% of all violent
interactions (N = 16,000) involved “good” perpetrators. Unfortunately, the NTVS
scholars considered good characters as those that were either “good” or “good and bad.”
As such, these findings do not reveal how often purely prosocial characters (i.e., good
only) engage in violence nor the context surrounding such aggression. Therefore, the
first research question asks:

RQl: Does the prevalence and context surrounding prosocial perpetrators differ

from those who lack prosocial qualities?

Given the fact that young children may be the most susceptible to learning
aggression (Paik & Comstock, 1994), it becomes important to also assess the amount and
nature of prosocial violence in shows that are popular with youngsters. Data consistently
reveals that children are drawn to specific time slots and program genres of TV content
(N ielson Media Research, 2000). That is, the most popular time of day with 2- to 11-
year-olds is prime time and the most popular genre is children’s shows (Stipp, 1993).
Thus, the second and third research questions ask:

RQ2: In children’s programs, does the prevalence and context surrounding

violence of prosocial perpetrators differ fi'om those who lack prosocial qualities?

RQ3: In prime-time programming, does the prevalence and context surrounding

violence of prosocial perpetrators differ from those who lack prosocial qualities?

Method

To answer the research questions outlined above, the database from the third year
of the National Television Violence Study (Smith et al., 1998) was used. As a result, the
sample, definition of violence, units of analysis, contextual features, coder training and
reliability are explicated briefly below. For the full report, see Smith et al. (1998).
Simple

Programs were randomly sampled from October of 1996 to June of 1997 to build
a composite week of TV programming from 6:00 am. to 11:00 pm. (PST). Programs
were sampled across 23 channels that include the broadcast networks (ABC, CBS, Fox,
NBC), independent broadcasters (KCAL, KCOP, KTLA), the public broadcasting
network (KCET), basic cable (A&E, AMC, BET, Cartoon Network, Disney, Family
Channel, Lifetime, MTV, Nickelodeon, TNT, USA, VH-l), and premium cable
(Cinemax, HBO, Showtime). All programs were aired and taped in the Los Angeles
market. In sum, a total of 3,212 programs were sampled in the composite week of
television content.2
Definition of Violence

The NTVS definition placed emphasis on three key features: intention to harm,
the physical nature of harm, and the involvement of animate beings. More specifically,
violence was defined as “any overt depiction of a credible threat of physical force or the
actual use of such force intended to physically harm an animate being or group of beings.
Violence also includes certain depictions of physically harmful consequences against an

animate being or group that occur as a result of unseen violent means” (Smith et al.,

1998, p. 30). In total, there were three types of violent depictions: credible threats,
behavioral acts, and harmful consequences.3
Units of Analysis

To fully assess the context of violence, the NTVS researchers measured
aggression at three distinct units or levels of analysis. The first and most microscopic
was the violent interaction. A violent interaction was defined as an aggressive exchange
involving a specific perpetrator (P) performing a particular type of act (A) against a
specific target (T). Any time the perpetrator, the target or type of act changed, a new
violent interaction was created. The PAT was created so that the rapidly changing
context of violence between characters (e.g., means used, extent of means used, imminent
consequences of violence) could be assessed accurately.

The second unit of analysis was the violent scene. A violent scene was defined as
a related series of violent behaviors that occurred without a significant break in the flow
of actual or imminent violence (Smith et al., 1998, p. 31). One or more PATs might
occur within a given violent scene among the same characters or types of characters.
Several contextual variables were measured at the scene level (e.g., rewards, punishment,
explicitness, humor, graphicness).
The most macro unit of analysis was the entire violent program. After watching the
violent show, coders assessed variables (e.g., realism of violence, patterns of punishment,
anti-violence) to capture the overall theme or message represented across a violent

program (Smith et al., 1998, p. 31).

10

ContextuaL Measures

To assess the 8 major contextual variables (i.e., nature of perpetrator, justification,
means or weapon used, extent/graphicness, consequences, reinforcements, humor,
realism), 27 measures were developed. Each contextual variable will be defined at the
level of analysis with which it was measured.

PAT Level Vargbl—es. Two types of contextual measures were examined at the
PAT level: character related and violence related. In terms of character variables,
demographic characteristics of the perpetrators and victims of violence were measured.
The first demographic variable was type of character. Characters were categorized as
either human, animal, supernatural creature, anthropomorphized animal, or
anthropomorphized supernatural being" This variable was later collapsed into four
categories: humans, anthropomorphized creatures, supernatural creatures, and others.
Each violent character’s sex also was assessed. Sex was coded as either male or female.
The next demographic variable was the age of the violent character. Each violent
character was coded as a child (between infancy and 12 years of age), teen (13-20 years
of age), adult (21-64 years of age), or senior citizen (65 years of age or older). Finally,
each character’s apparent ethnicity was assessed. Each character was categorized as
White, Hispanic, Black, Native American, Asian/Pacific Islander, or Middle Eastern.
Due to the low frequency observed in some of the ethnic groups, this variable was
collapsed into four categories: White, Black, Hispanic, and others.

In addition to demographic data, characters were assessed in terms of specific
attributive qualities. The first is orientation towards others. Characters were coded as

good, bad, good and bad, or neutral. The second quality is centrality to the plot. A

11

character was considered central to the plot if his/her principal role was essential to the
story. Each character was coded as either a primary character or not a primary character.
The third is law status. Characters who worked for the state/federal government as
officials, police officers, or military personnel were coded as having law status.
Characters who did not work for the state/federal government were coded not having law
status.

In addition to character variables, several violence-related variables were assessed
at the PAT level. The first contextual factor was the reason for violence. There were six
values for this measure: protection of life, retaliation, anger, personal gain, mental
instability, and other. These values were collapsed into four major reasons: protect life,
anger, personal gain, and other. The second variable was justification. Justified violence
was defined as those aggressive acts and/or threats that were portrayed as “morally
correct” or “just” given the circumstances of the plot (Smith et al., 1998, p. 32). Each
interaction was coded as justified or unjustified. The third was the means used in each
violent interaction. Means were defined as any object, weapon, or device that a
perpetrator used to threaten and/or harm a target. There were six values for this variable:
use of body only (e.g., arm, leg), unconventional weapons, conventional weapons non-
firearms, handheld firearms, heavy weaponry, and bombs. Due to the low fi'equency of
occurrence of some values, the means variable was collapsed into four values: use of
body, unconventional weapons, conventional weapons, and others. The fourth variable
assessed the extent of means used. For behavioral acts only, the extent variable was

coded as one (single example of act), some (between 2-9 examples of the act), many

12

(between 10-20 examples of the act), or extreme (over 20 examples of the act). These

values were later collapsed into two categories: one act versus repeated acts.

Several variables assessed the immediate consequences of violence at the PAT
level. The amount of depicted pain a target experienced as a consequence of violence
was measured. Pain was defined as the audible (e.g., screams, moans, yells, gasps) or
visible (e.g., facial expressions, physical reactions such as clutching of a wound/injury)
expression of physical suffering that occurred as a result of violence. Each interaction
was coded as featuring no pain, mild pain, moderate pain, or extreme pain. Next, the
amount of depicted harm to the target was recorded. Depicted harm referred to the actual
portrayal of physical injury or incapacitation that was caused by violence. Depicted harm
was coded as none, mild, moderate, or extreme. The amount of likely harm also was
assessed. Likely harm referred to the amount of physical injury and/or incapacitation that
would have been experienced if the same violence occurred against an average sized
human in real life. Likely harm was measured as none, mild, moderate, or extreme. At
the analysis level, the amount of likely harm was subtracted from the amount of depicted
harm to yield an estimate of the extent of unrealistic harm on television. Finally, the
amount of serious violence was assessed. By using the likely harm variable, all
interactions featuring moderate or extreme likely harm were coded as “serious” or lethal
in nature. All interactions featuring no or mild levels of likely harm were coded as “not”

serious.

Scene Related Vaflbles. Several contextual variables were assessed at the scene
level.’ The first contextual factor was rewards. Rewards were defined as any verbal or

nonverbal positive reinforcement that was given to or taken by a violent perpetrator.

l3

(Smith et al., 1998, p. 34). Three types of rewards were assessed at this level: self praise
(e. g., perpetrator says to the self, “good job”), praise from others (e.g., another character
says to perpetrator, “good job”), or material praise (e. g., perpetrator is given money for
acting violently). Each of these variables are coded as present or absent. Ifany of these
types of rewards were featured in a violent scene, then at the analysis the scene level was
coded as rewards “present.”

The second variable was punishments. A punishment was defined as any verbal
or nonverbal sign of disapproval or disappointment that was expressed towards a
perpetrator for acting violently. Punishments were classified as involving self-
condemnation that a perpetrator expresses for acting violently (e. g., perpetrator says to
self for acting violently, “I shouldn’t have done that”), condemnation fi'om others (e.g.,
another character tells perpetrator “you shouldn’t have done that”), nonviolent action to
stop or penalize violence (e. g., arresting a perpetrator for acting violently), and violent
action by a third party to terminate further violence (e.g., someone other than the victim
using violence to stop perpetrator). Each of these variables are coded as present or absent
at the scene level. Again, if any punishment was featured in the violent scene, then
punishment was coded at the scene level as “present.”

The third factor was graphicness. Graphicness was defined as the quantity or
amount of blood, gore (e.g., innards, viscera), and/or dismemberrnent depicted within the
violent scene. Each scene was coded as featuring no graphicness, mild graphicness,
moderate graphicness, or extreme graphicness. Very few scenes used the upper limits of
this scale. Thus, these levels were collapsed into two categories: graphicness vs. some

graphicness. The last scene level measure was humor. Humor was defined as the use of

14

speech, actions and/or behaviors that a character engaged in that were intended to amuse
the self, another character or characters, and/or the viewer. Humor was coded as present
or absent at the end of each violent scene.

Program Level Variables. Several contextual variables were assessed at the end
of the entire violent program. The first was whether the program featured an anti-
violence theme or not. An anti-violence theme is one that emphasizes that violence is
morally and/or socially wrong. Programs were coded as featuring an anti-violence theme
if any one of the four following criteria were met: (1) alternatives to violent actions were
presented and/or discussed; (2) main characters discussed repeatedly the negative
consequences of violence; (3) the physical pain and emotional suffering that resulted
from violence was emphasized; or (4) the punishments for violence outweighed the
rewards clearly and consistently throughout the plot (Smith et al., 1998, p. 35).

The second variable was program level harm/pain. This variable assessed the
extent to which negative consequences that resulted fi'om violence were presented across
the program. Negative consequences could include any physical harm/pain, emotional
cost, financial loss, or psychological suffering that occurred at a result of violence. There
were three values for harm/pain: (1) none, (2) short-terrn only, or (3) long-term or
extended suffering. The next program level contextual measure was realism. Realism
referred to the actuality of the characters, settings, and events that were presented in a
program. Two variables were created to assess the realism of violence: authenticity and
style of presentation. Authenticity referred to the programs level of realism when
presenting violence. Programs were coded as actual reality, re-creation of reality, fiction,

and fantasy. Later, this variable was collapsed into two categories: reality (actual reality,

15

recreation of reality, and fiction) versus fantasy (fantasy). Thus, a given violent program
could show footage of actual events enacted by the actual people involved in real time,
the past, and/ or creative contexts. Fantastic violent shows, however, featured characters
and/or settings that cannot possibly happen in the real-world as we know it today. In
terms of the presentational style, each program was coded as animation, live action, or a
mix of both. Again, this variable was collapsed into two categories: live action only
versus animated action.

Coder Training and Reliability

A total of 56 undergraduate research assistants at the University of California,
Santa Barbara were trained as coders to evaluate the sample of program content. Prior to
coding, the RA’s received approximately 40 hours of classroom instruction to learn the
conceptual and operational definitions in the codebook. In addition, the coders spent
roughly 20 hours in the laboratory unitizing and quantifying the contextual variables.
Once coding began, all research assistants independently watched and coded randomly
sampled program content for violence.

Two levels of reliability were assessed for NTVS: unitizing and the degree of
consistency in coders’ judgments on the contextual measures. In terms of unitizing, the
modal number of interactions and scenes agreed upon in each reliability test was
assessed. The number of coders that came within 20% of the mode on interactions and
scenes in each reliability test was figured. Across all of the programs assessed for
reliability (n = 20), most coders agreed upon the number of interactions (68% median

agreement) and scenes (78% median agreement) within a 20% interval around the mode.

16

After coders agreed on the same level of analysis, their agreement on choosing a
value on each of the variables in the scheme was assessed. The NTVS researchers
computed a level of confidence for each of 540 reliability coefficients (27 variables on
each of the 20 programs in the reliability tests). Across all 20 tests, the median reliability
coefficients for each of the variables used in the present study were as follows:
orientation towards others (.74), perpetrator type (.95), sex (.91), apparent ethnicity (.84),
age (.80), primary character (.85), law status (.70), type of act (1.0), reason (.81),
justification (.91), means used (.93), extent (.89), depicted harm (.76), likely harm (.77),
pain (.70), self praise (.92), praise from other (.98), material praise (.97), self
condemnation (1.0), condemnation from others (.93), graphicness (.91), humor (.90),
anti-violence theme (1.0), harm/pain (.75), realism (.93), style of presentation (1 .0).

Results
Analysis Plan

Three aspects of the analysis plan need to be explicated. The first aspect concerns
how “prosocial” perpetrators are defined. As noted above, the NTVS researchers coded
orientation towards others as good, bad, both good and bad (blended), or neutral (neither
good nor bad). Using these categories, prosocial characters are defined in this study as
only those “good” perpetrators who act benevolently, helped others, and/or are motivated
to consider the needs of others (N = 4,731). Characters are considered to lack prosocial
qualities if they are coded as “bad,” that those who act primarily in their own self-interest
and have very little regard for others (N = 7,200). All other categories are excluded for

analysis purposes.

17

The second concerns how to frame and/or interpret all of the findings. In
explanation, prosocial status is measured at the PAT level of analysis. Yet several
contextual variables are measured at the scene and program level. As such, the study
“brings down” the scene and program variables to the PAT level interaction.
Consequently, the correct interpretation of findings at the scene level would be “the
percentage of prosocial perpetrators shown engaging in violence in a humorous context.”
At the program level, the results should be framed as “the percentage of prosocial
perpetrators featured in contexts with little or no extended consequences from violence.”

The third aspect concerns the significance of any particular finding. A chi-square
statistic is executed to assess whether the distribution of prosocial violence varied by
each contextual factor. Only those analyses that were significant at p < .05 are reported
below. Because the sample size used in this study is large (N > 11,000), even small
differences in the depiction of violence will yield significant chi-square values. Thus,
only differences of 5% or greater between percentages are deemed to be of practical
significance.

Research Question 1

Amount. Research Question 1 asks, “Does the frequency and context surrounding
prosocial perpetrators differ from those who lack prosocial qualities?” To address the
first part of this question, two variables are examined. The first is rate per hour. Rate per
hour is calculated by dividing the total number of violent interactions featuring prosocial
(N = 4,731) or nonprosocial (N = 7,200) perpetrators by total program hours (2,586

hours) in the sample. The results show that prosocial perpetrators occur less fi'equently

18

per hour than nonprosocial perpetrators (1.83 vs. 2.78, respectively) across the entire
composite week of programming.

In addition to rate per hour, the type of violence characters engaged in is assessed.
Four types of acts are measured at the PAT level: credible threats, behavioral acts,
harmfirl consequences, and accidents. A significant difference in act type is observed, x2
(3, N = 11,931) = 148.55, p < .05, 31"‘6 = .11 Prosocial perpetrators are more likely to
engage in behavioral acts (72.1%,) than are nonprosocial perpetrators (64.5%.). No other
differences are found in this analysis. A total of 28.3% of acts involve credible threats,
2.2% harmful consequences, and only 1.9% of acts are accidental.

Contextual Variables. To answer the second half of Research Question 1, both
character and context variables are examined. In terms of character-related variables,
both perpetrator and victim demographics are assessed. The first measure is perpetrator
type. A chi-square analysis revealed a significant difference in perpetrator type by
prosocial status, x2 (3, N = 11,931) = 119.48, p < .05, y* = .10. As seen in Table 1,
74.5%., of prosocial perpetrators of violence are significantly more likely to be humans
than are nonprosocial perpetrators (66.2%.). Two other categories account for nearly all
the remaining perpetrators: anthropomorphized creatures (48.8%) and supernatural
creatures (10.6%). However, these two types do not differ by prosocial status.

The second variable is perpetrator sex. A significant difference in perpetrator sex
by prosocial status is observed, x2 (1, N = 10,205) = 174.89, p < .05, ¢ = .13. The
findings show that males are significantly less likely to be featured as prosocial

perpetrators (83.4%.) than as nonprosocial perpetrators (91 .9%b). In contrast, women are

19

significantly more likely to be featured as prosocial perpetrators (16.6%.) than as
nonprosocial perpetrators (8.1%.) (see Table l).

The third measure is age of perpetrator, which also varies significantly by
prosocial status, 12 (3, N = 9,895) = 192.62, p < .05, 2* = .14. The results show
prosocial perpetrators (8.7%.) are significantly more likely to be teens than are
nonprosocial perpetrators (3.6%.). Yet prosocial perpetrators (84.6%.) are significantly
less likely to be adults than are their nonprosocial counterparts (93.1%.) No differences
are found in the other subcategories (see Table 1).

The ethnicity of perpetrators is the fourth variable. For humans only, the analyses
indicate a significant difference in perpetrator ethnicity, x2 (3, N = 8,254) = 104.80, p <
.05, \_l* = .11. As seen in Table l, prosocial perpetrators (77.0%.) are significantly more
likely to be white than are nonprosocial perpetrators (71.0%.). No other differences are
observed. Only 5.5% of perpetrators are Black, and just 2.9% are Hispanic.

Primary character is the fifth demographic variable assessed. A significant
difference is observed, x2 (1, N = 11,538) = 554.93, p < .05, 4; = .22. The results
demonstrate that prosocial perpetrators (88.0%.) are significantly more likely to be
primary characters than are nonprosocial perpetrators on television (69.0%.) (see Table
1).

Finally, a significant difference emerges for law status, 12 (1, N = 11,931) =
281.07, p < .05, ¢ = .15. The findings indicate that law enforcers are significantly more

likely to be portrayed as prosocial perpetrators (20.0%,) than as nonprosocial perpetrators

(9.2%.) (see Table 1).

20

Together, the results reveal that prosocial perpetrators of violence on television
are most likely to be humans, adult males, and white. Although less frequent in nature
overall, prosocial perpetrators are more likely to be female and in their teenage years than
are nonprosocial perpetrators. Also, prosocial perpetrators are more likely to be primary
to the plot and possess an occupation as a law enforcer than are nonprosocial
perpetrators.

In addition to perpetrator variables, the demographics of the victim are analyzed.
As seen in Table 1, no differences are found in terms of victim type, ethnicity, and law
status. However, a significant chi-square is observed in victim gender by prosocial
status, x2 (l, N = 9,591) = 212.73, p < .05, (b = .15. The results show that prosocial
perpetrators are more likely to target males (92.8%b) than are nonprosocial perpetrators
(82.6%.). A difference also is found for female victims. Prosocial perpetrators (7.2%.)
act less violently towards female victims than do nonprosocial perpetrators (17.4%.) (see
Table l).

A significant difference emerges for victim age as well, 12(3, N = 9,472) =
134.86, p < .05, 31* = .12. The findings showed that prosocial perpetrators (5.1%.) are
significantly less likely to harm teen victims than are nonprosocial perpetrators (10.5%.).
Adults, however, are more likely to be the targets of prosocial violence (908%.) than
nonprosocial violence (82.4%.). No other differences were found in the analyses (see
Table 1).

A final difference is found in the distribution of primary characters who are
victims of prosocial or nonprosocial violence. A chi-square analysis reveals a significant

difference in primary character victims, x2 (1, N = 11,425) = 157.87, p < .05, (b = .12. As

21

seen in Table 1, prosocial perpetrators (63.7%.) are less likely to act violently against
primary characters that are victims than are nonprosocial perpetrators (74.7%b).

Besides character-related variables, violent contextual variables are measured.
Among all variables, no differences are observed in extent, graphicness, reinforcement,
humor, and anti-violence theme (see Table 2). However, the reasons for violence differ
significantly by prosocial status, x2 (3, N = 11,917) = 4,129.77, 2 < .05, \_l* = .59. The
findings indicate that protection of life is significantly more likely to be the reason for
violence by prosocial perpetrators (56.9%.) than by nonprosocial perpetrators (6.0%.). In
contrast, anger (29.5%.), personal gain (44.8%,), and other reasons (19.6%,) are
significantly more likely to be the motives for violence by nonprosocial perpetrators (see
Table 2).

The second variable is justification, which varies significantly by prosocial status,
xz (1, N = 11,881) = 4,830.95, 9 < .05, d) = .64. The results show that justified violence is
significantly more likely to be portrayed by prosocial perpetrators (64.1%.) than by
nonprosocial perpetrators (5.2%.) (see Table 2).

The third variable is the means or weapon used. A significant difference in
weapon use by prosocial status is observed, x2 (3, N = 11,931) = 69.99, p < .05, X” = .08.
As seen in Table 2, prosocial perpetrators are significantly more likely to use their own
bodies to enact violence (409%,) than are nonprosocial perpetrators (35.7%.). No other
differences are found in this analysis. A total of 18.4% of perpetrators used
unconventional weapons, 35.9% used conventional weapons, and only 7.9% of
perpetrators used other means for violence (i.e. heavy weaponry, bombs, means

unknown).

22

The fourth variable is consequences. Although several types of consequences are
assessed (i.e. pain, depicted harm, unrealistic harm), only likely harm reveals a
significant difference, x2 (3, N = 6,996) = 34.83, p < .05, y* = .07. The results show that
prosocial violence (28.0%.) is less likely to involve in extreme likely harm to the victims
than is nonprosocial violence (34.6%.) (see Table 2).

The authenticity of the violence is the last variable assessed. Two types of
authenticity are measured: realism and presentation style. A chi-square analyses shows a
significant difference in realism, 12 (1, N = 11,929) = 35.88, p < .05, 41 = .06.
Specifically, prosocial perpetrators (57.6%.) are significantly more likely to be featured
in contexts with realistic violence than are nonprosocial perpetrators (52.1%.). In
contrast, unrealistic violence is more likely to be shown with nonprosocial (47 9%.) than
with prosocial characters (42.4%.) (see Table 2).

In regards to presentation style, a significant difference by prosocial status
appears as well, x2 (1, N = 11,930) = 66.11, p < .05, d) = .07. The results show that
prosocial perpetrators (67.7%.) are significantly more likely to be portrayed in live action
contexts than are nonprosocial perpetrators (60.4%.). Prosocial violence (32.3%.),
however, is less likely than nonprosocial violence (39.6% b) to be shown in animated
action (see Table 2).

In brief, prosocial perpetrators in overall programming tend to use their own
bodies to commit violence, their victims are adult males, and the most likely reason for
aggression is to protect life. Most often, the violence of prosocial perpetrators is justified

and depicts aggression in realistic contexts.

23

Research Question 2

Research question 2 asks, “In children’s programs, does the prevalence and
context of violence surrounding prosocial perpetrators differ from those who lack
prosocial qualities?” To answer this question, only children’s shows were selected in the
NTVS database. The NTVS researchers defined children’s show as those that were
designed for audiences 16 years of age or younger. A total of 479.50 hours of children’s
shows were in the NTVS database.

Amou_nt. In terms of the first half of Research Question 2, rate per hour is
examined. To this end, the frequency of prosocial (N = 1,585) and nonprosocial
perpetrators (N = 2,811) involved in violent interactions is divided by the total program
hours in children’s programming (479.50 hours). The results show that prosocial
perpetrators occur less frequently per hour than do nonprosocial perpetrators (3 .31 vs.
5.86, respectively) in children’s shows.

The second measure assesses type of act. A significant difference is observed, x2
(3, N = 4,396) = 85.75, p < .05, \_l* = .14. Prosocial perpetrators (80.2%.) are more likely
to engage in behavioral acts than are nonprosocial perpetrators (67.9%.). Prosocial
perpetrators (16.7%.), however, are lese likely to enact credible threats than are
nonprosocial perpetrators (28.8%.). No other differences are found in this analysis. A
total of 0.5% of acts involve harmfirl consequences, and just 2.7% of acts are accidental.

Contextual Variables. To address the second half of Research Question 2,
character and contextual variables are examined. In terms of character-related variables,
only perpetrator type, sex, age, and primary character vary by prosocial status. A chi-

square analysis reveals a significant difference in type of perpetrator by prosocial status,

24

x2 (3, N = 4,396) = 40.11, p < .05, 2* = .10. The results show that prosocial perpetrators
(6.9%.) are less likely to be supernatural creatures than are nonprosocial perpetrators
(12.8%.). No other differences are found (see Table 3).

For perpetrator sex, a significant difference is observed as well, 780, N = 3,388)
= 35.43, p < .05, d) = .10. The findings show that prosocial perpetrators (85.9%.) are
significantly less likely to be males than are nonprosocial perpetrators (92.3%.). In
contrast, females are less likely to be featured as prosocial perpetrators (14.1%.) than as
nonprosocial perpetrators (7.7%.) (see Table 3).

Age of perpetrator is the next variable that varies significantly by prosocial status
in children’s shows, x2 (3, N = 2,901) = 318.29, p < .05, 31* = .33. The results reveal that
prosocial characters are significantly more likely to be children (13.6%.) and teens
(17.9%.) than are nonprosocial characters (3.8%. and 2.8%., respectively). Also, a
difference is observed for adult perpetrators. Prosocial perpetrators (67.1%.) are less
likely to be adults than are their nonprosocial counterparts (91.9%.) (see Table 3).

Finally, a significant difference is found for primary characters, x2 (1, N = 4,288)
= 122.29, p < .05, d) = .17. As seen in Table 3, the findings demonstrate that prosocial
perpetrators (89.2%.) are significantly more likely to be primary characters in children’s
programs than are nonprosocial perpetrators (75.3%.).

In terms of the victim profile, only age and primary character are found to
significantly differ (see Table 3). Specifically, a significant difference in victim age
emerges in the analysis, x2 (3, N = 2,671) = 78.53, p < .05, \_/* = .17. The results show
that teens are significantly less likely to be victims of prosocial perpetrators (6.5%.) than

of nonprosocial perpetrators (16.8%.). On the other hand, adults are more likely to be the

25

targets of prosocial violence (84.9%.) than nonprosocial violence (70.1%.). No
difference is observed in other subcategories. Children only account for 10.2% of total
victims in children’s shows (see Table 3).

The final significant difference is observed in primary character victim, x2 (1, N =
4,241) = 33.35, p < .05, 41 = .09. The findings illustrate that prosocial characters (72.2%.)
are less likely to act violently towards a primary character who is a victim than are
nonprosocial perpetrators (80.0%.) (see Table 3).

In sum, prosocial perpetrators in children’s programs are most likely to be
anthropomorphized creatures, males, and white adults. Prosocial perpetrators also are
more likely to be females in their childhood and teenaged years, when compared with
nonprosocial perpetrators. Prosocial perpetrators also are more likely to be shown as
primary characters and committing violence against adult victims.

Among all violence measures, extent, graphicness, reinforcement, humor, realism,
and anti-violence theme do not vary by prosocial status. However, the reasons for
violence do reveal a significant difference, x2 (3, N = 4,396) = 1,831.91, 9 < .05, 2* =
.65. The results show that protection of life is significantly more likely to be the reason
for prosocial violence (60.9%.) than for nonprosocial aggression (4.4%.). Yet anger
(27.1%.), personal gain (49.1%.), and other reasons (19.4%.) are significantly more
likely to be the reasons for violence by nonprosocial perpetrators than prosocial
perpetrators (see Table 4).

Justification also differs by prosocial status, 780, N = 4,371) = 1,924.29, 9 < .05,

d) = .66. The results show that justified violence is significantly more likely to be

26

portrayed by prosocial aggressors (64.4%.) than by nonprosocial aggressors (4.0%.) (see
Table 4).

A significant difference is observed in the types of means used as well, x2 (3, N =
4,334) = 17.40, p < .05, 31* = .06. As seen in Table 4, prosocial perpetrators use their
own bodies to enact violence (45.8%.) significantly more than do nonprosocial
perpetrators (40.8%.). No other differences are found in this analysis.

Depicted harm varies significantly by prosocial status, 980, N = 2,624) = 29.70,
p < .05, \_l* = .11. The findings reveal that prosocial violence is less likely to yield no
depicted harm (48.5%.) than is nonprosocial violence (55.6%.). Prosocial perpetrators
(11.0%.), in contrast, are significantly more likely to inflict extreme depicted harm on
their victims than are nonprosocial perpetrators (5.7%.) (see Table 4).

Also, a significant difference in unrealistic harm by prosocial status is found, x2
(1, N = 2,350) = 8.20, p < .05, d) = .06. When compared to nonprosocial perpetrators
(56.5%.), the results show that violence is less likely to result in unrealistically low levels
of harm for the targets of prosocial perpetrators (50.6%.) (see Table 4).

Together, children’s programs are more likely to show prosocial perpetrators
engaging in justified violence using natural means to protect lives. They also are shown
most fi'equently in contexts with more extreme depicted harm but less unrealistic harm,
when compared with nonprosocial perpetrators.

Regarch Question 3

Research Question 3 asks, “In prime-time programming, does the prevalence and

context surrounding violence of prosocial perpetrators differ fiom those who lack

prosocial qualities?” The NTVS researchers defined a prime-time program as any that

27

was aired between 8:00 and 11:00 pm. (PST) Monday through Saturday or 7:00 to
11:00 pm. (PST) on Sunday. A total of 466.50 prime time hours of programming were
captured in third year NTVS sample.

Amount. In terms of rate per hour, the total number of violent interactions
featuring prosocial perpetrators is 843. Nonprosocial perpetrators were featured in 1,370
violent interactions. Using these figures and the total amount of prime time hours in the
sample, prosocial perpetrators are featured less frequently per hour than are nonprosocial
perpetrators (1.81 vs. 2.94, respectively) in prime-time programs. The type of violence
characters engaged in also is measured. No significant differences in act type by
prosocial status are found. A total of 29.6% of acts are credible threats, 66.0% behavioral
acts, only 3.3% are harmful consequences, and 1.0% of acts are accidents.

Contextual Variables. Both character and context variables are examined to
answer the second part of Research Question 3. In terms of character-related variables,
the only perpetrator variables that vary by prosocial status are type, sex, ethnicity,
primary character, and law status (seen Table 3). To illustrate, a significant difference in
perpetrator type is observed, x2 (3, N = 2,213) = 60.27, p < .05, y* = .17 . The findings
show that prosocial perpetrators (94.1%.) are significantly more likely to be humans than
are nonprosocial perpetrators (84.2%.). No other differences are found (see Table 3).

Second, a significant difference in perpetrator sex by prosocial status is obtained,
x2 (1, N = 1,947) = 63.49, p < .05, d) = .18. The results indicate that prosocial perpetrators
are significantly less likely to be males (79.5%.) but more likely to be females (20.5%.)

than are nonprosocial perpetrators (91.9%. vs. 8.1%., respectively) (see Table 3).

28

The ethnicity of the perpetrator reveals significant differences by prosocial status
as well, x2 (3, 1:1 = 1,944) = 55.55, p < .05, y* = .17. The findings show that prosocial
perpetrators (80.8%.) are significantly more likely to be white than are nonprosocial
perpetrators (66.8%.). No other differences are observed in this analysis (see Table 3).

A chi-square significance in primary character is observed, x2 (l , N = 2,100) =
55.49, p < .05, ¢ = .16. As seen in Table 3, the results demonstrate that prosocial
perpetrators (83.0%.) are significantly more likely to be primary characters than are
nonprosocial perpetrators (68.4%.).

A final significant difference is found for law status, x2 (1, N = 2,213) = 237.90, p
< .05, (b = .33. The findings reveal that law enforcers are more likely to be prosocial
perpetrators (39.3%.) than nonprosocial perpetrators (11.3%.) (see Table 3).

On the other hand, almost identical results are observed in victim demographics
by prosocial status. That is, the significant differences are found in victim type, sex,
ethnicity, primary character, and law status. The results show a significant difference in
victim type by prosocial status, x2 (3, N = 2,213) = 22.48, p < .05, \_l* = .10. The findings
reveal that humans are less likely to be the target of prosocial violence (85.8%.) than of
nonprosocial violence (91.0%.). No other differences are obtained in the analyses (see
Table 3).

The victim sex also varies significantly by prosocial status, x2 (l, N = 1,862) =
62.47, p < .05, ¢ = .18. The results show that males are significantly more likely to be
the victim of prosocial perpetrators (91.6%.) than are nonprosocial perpetrators (77.5%.).
In contrast, females are less likely to be the target of prosocial violence (8.4%.) than of

nonprosocial violence (22.5%.) (see Table 3).

29

A significant difference is obtained in the ethnicity of the victim, 780, N = 1,963)
= 18.38, p < .05, \_I* = .10. The findings indicate that victims of prosocial perpetrators
(70.2%.) are significantly less likely to be white than are the victims of nonprosocial
perpetrators (77.8%.). No other differences are observed in this category (see Table 3).

Primary character is the next victim demographic variable that differs by
prosocial status, x2 (1, N = 2,115) = 7.65, p < .05, a = .06. The results show that
prosocial perpetrators (62.2%.) act violently significantly less against primary character
victims than do nonprosocial perpetrators (68.1%.) (see Table 3).

Finally, the law status of victims varies significantly by prosocial status, x2 (1, N
= 2,213) = 25.40, p < .05, d) = .11. As seen in Table 5, the findings show that law
enforcers are significantly less likely to be the victim of prosocial violence (12.0%.) than
of nonprosocial violence (20.3%.).

All together, prosocial perpetrators are most likely to be portrayed in prime-time
shows as humans, males, and white adults. They are more likely to be primary to the plot
and possess an occupation as a law enforcer.

In addition to character-related variables, violent contextual variables are
measured in the PAT, scene, and program levels. Only differences in reason,
justification, consequences are significant (see Table 4). In terms of reasons, the results
reveal a significant difference by prosocial status, x2 (3, N = 2,213) = 883.89, p < .05, _V*
= .63. The findings show that protection of life is significantly more likely to be the
reason for violence by prosocial perpetrators (64.5%.) than by nonprosocial perpetrators

(6.9%.). However, anger (27.2%.), personal gain (45.5%.), and other reasons (20.4%.)

30

are significantly more likely to be the reasons for violence by nonprosocial perpetrators
(see Table 4).

A significant chi-square is observed in justification, x2 (1, N = 2,205) = 1,013.47,
9 < .05, d) = .68. The results indicate that prosocial perpetrators (68.4%.) are significantly
more likely to be shown engaging in justified violence than are nonprosocial perpetrators
(5.0%.) (see Table 4).

Also, a significant difference in depicted harm by prosocial status is found, 12 (3,
N = 1,298) = 34.83, p < .05, y* = .07. The findings reveal prosocial violence (39.1%.) is
more likely to yield no depicted harm to the target than is nonprosocial violence
(32.1%.). In contrast, extreme depicted harm is less likely to be shown against the targets
of violence of prosocial perpetrators (24.1%.) than of nonprosocial perpetrators (29.2%.)
(see Table 4).

As seen in Table 4, although extreme depicted harm shows a 5% difference
between prosocial and nonprosocial perpetrators, the chi-square is not significant at the p
< .05 level.

Finally, a significant difference in the extended consequences of violence is
observed, it” (2, N = 2,213) = 13.29, p < .05, y* = .08. The findings indicate that
prosocial perpetrators (40.8%.) are significantly less likely to be shown in settings with
long-term harm than are nonprosocial perpetrators (48.5%.) (see Table 4).

Together, prime-time programming depicts prosocial perpetrators engaging in
justified violence to protect self and others. Further, the violence of prosocial

perpetrators is likely to yield little depicted harm to the target.

31

Discussion

Generally, the findings of this study reveal that prosocial status influences the
nature and amount of violence a perpetrator engages in. Overall, prosocial perpetrators
engage in less violence than their nonprosocial counterparts, yet genre and time of day
seem to exert an influence on prosocial perpetrators violent acts. Further, the
prototypical violent interaction involving prosocial perpetrators features a human adult
that is white engaging in justified violence that is depicted in realistic contexts. These
findings are interpreted below in light of the risk that exposure to prosocial violence may
pose to viewers.

In terms of Research Question 1, the prosocial status of a perpetrator affects the
amount and type of violence shown on television. Across the entire composite week of
programming, prosocial perpetrators engage in less violence per hour than do
nonprosocial perpetrators. Despite this positive trend, the average child watches roughly
three hours of television per day (N ielson Media Research, 2000). Thus, the typical
youngster is being exposed to approximately 5.49 violent interactions involving prosocial
perpetrators per day or 2,004 depictions per year. Repeated exposure to prosocial
violence may teach or reinforce aggressive thoughts, attitudes, and behaviors (Huesmann
& Miller, 1994).

Additionally, prosocial status exerts an influence on the type of violence a
perpetrator enacts. When compared to their nonprosocial counterparts, prosocial or good
characters are more likely to engage in violent behavioral acts. This trend might be
particularly problematic for younger children who rely heavily on visual depictions for

learning (Huston-Stein et al., 1981; Valkenburg & Cantor, 2000). By actually seeing the

32

behavior enacted by prosocial perpetrators, younger children may be more likely to
attend to and learn such overtly depicted violent actions (Huesmann, Lagerspetz, & Bron,
1984; Liss, Reinhardt, & Fredriksen, 1983).

The second half of Research Question 1 addresses the context of violence
surrounding prosocial vs. nonprosocial perpetrators. First, notable differences were
found in the types of demographic characteristics assigned to prosocial and nonprosocial
aggressors. Prosocial perpetrators are more likely than nonprosocial perpetrators to be
human, white, and primary to the plot. Based on the notion of perceived similarity
(Hoffner & Cantor, 1991), these findings suggest that white viewers who strongly
identify with prosocial perpetrators may be more at risk for learning aggression.
Although less frequent in nature, prosocial perpetrators also are more likely than non
prosocial perpetrators to be law agents or government authorities, females, and in their
teenaged years. When considering victim profiles, prosocial status seems to exert less
influence. Victims of prosocial violence tend to be adult males. Because viewers tend to
identify and empathize with similar others (F eshbach & Roe, 1968; Jose & Brewer, 1984;
von Feilitzen & Linne, 1974), white males may experience more fear when watching the
victims of prosocial violence than other subgroups of viewers.

The last aspect of Research Question 1 concerns the context of prosocial violence.
A few attributes of prosocial violence may pose risk for learning. For example, a
majority of the violent interactions involving prosocial perpetrators are depicted as
protecting the self or other from harm and are portrayed as justified. Such socially
sanctioned violence may communicate to the viewer that violence is acceptable in certain

instances. Indeed, studies show that exposure to justified violence increases the

33

probability of aggression among children and adults (Berkowitz & Geen, 1966;
Berkowitz & Rawlings, 1963; Meyer, 1972).

Another trend that was found is that prosocial violence may be more imitable than
nonprosocial violence. This trend was witnessed across three separate variables. First, a
full 41% of prosocial violence involves the perpetrator’s own body. Violence involving a
perpetrator’s own body (e.g., punching, karate chops) may pose an increased risk than
some other types because the weapon is readily accessible to all viewers, especially
children. In fact, at least one study showed that children were significantly more
aggressive after viewing episodes of the Power Rangers and their emulation of violence
was the precise imitation of the character’s physical acts, such as flying karate kicks or
chops (Boyatzis, Matillo, & Nesbitt, 1995). The second and third variables suggest that
prosocial violence is presented in more realistic contexts. That is, prosocial perpetrators
are more likely to be shown in contexts that are real in nature and are presented in live
action. Numerous studies indicate that realistic portrayals of violence can increase the
probability of learning among both child and adult viewers (Atkin, 1983; Berkowitz &
Alioto, 1973; Geen, 1975).

It must be noted, however, that there are several similarities between prosocial
and nonprosocial violence that may influence risk. First, the findings reveal that violence
is sanitized on television. Independent of prosocial status, violence on television shows
little or no harm (e.g., injury or blood) and/or pain to the victim. Studies show that
presenting pain and suffering can reduce aggressive responding (Baron, 1971a, 1971b).
Thus, failing to portray such consequences may increase the learning and enactment of

aggression. Second, violence is often not chastised on television. Violent interactions

34

are most likely to be featured in scenes with no rewards or punishments. Research
reveals that these types of depictions can increase the probability of learning (Bandura,
Ross, & Ross, 1961, 1963b; Lando & Donnerstein, 1978; Paik & Comstock, 1994). This
effect is presumably due to the fact that unpunished violence communicates to the viewer
that violence does not have negative repercussions. Third, violence is often trivialized
on television. Regardless of prosocial orientation, a little over 40% of violent
interactions are featured in contexts that contain humor. Juxtaposing humor and violence
has been found to increase aggression (Baron, 1978; Berkowitz, 1970).

In sum, the profile of prosocial violence involves several elements of risk.
Prosocial perpetrators are likely to be depicted engaging in justified violence to protect
life in realistic contexts. Further, the typical prosocial perpetrator is a human who is
white, adult, and male.

Research Question 2 asked if prosocial status influenced the amount and nature of
violence in children’s programs. Notable differences emerged on both amount variables
in kid’s shows. First and similar to the overall landscape, prosocial perpetrators are more
likely than nonprosocial perpetrators to commit behavioral acts of violence. However,
prosocial perpetrators are less likely to threaten their targets. Second, prosocial violence
occurs at a lower rate per hour than does nonprosocial violence (3.31 vs. 5.86,
respectively). Although this trend is similar to the pattern of rate per hour across the
entire composite week, children’s shows actually feature more prosocial and
nonprosocial aggression. The rate per hour in this specific genre is roughly twice as high

as rate per hour in the overall landscape. Thus, a young child who generally consumes

35

two to three hours of children’s shows daily will see roughly 9.93 violent incidents per
day or 3,624 violent incidents per year.

The second half of Research Question 2 addressed the context of violence by
prosocial status. First, substantial differences were found in the demographic
characteristics of prosocial and nonprosocial aggressors/victims. When compared with
nonprosocial perpetrators, prosocial perpetrators are more likely to be portrayed as
primary characters that are females in their childhood and teenaged years. Although
these were significant differences by prosocial status, it is still the case that most
prosocial aggressors are males and adults. Given the increase in child and teen prosocial
perpetrators, younger children have more role models similar in age in children’s
programming. As noted earlier, the research on perceived similarity suggests that
children attend to and identify with models that are similar to the self (see Hoffner &
Cantor, 1991). Thus, this pattern is especially risky for younger viewers who repeatedly
consume children’s programming. In terms of victims, only two differences were
observed. Prosocial perpetrators are more likely to be adults than are nonprosocial
perpetrators and they are less likely to attack primary characters.

One other difference in the demographics of characters involved with violence in
children’s shows must be addressed. Children’s shows are more likely to feature
anthropomorphized characters. Although anthropomorphized characters may pose little
risk to older children, younger viewers who have not mastered the fantasy/reality
distinction may be potentially susceptible to such portrayals. Thus, anthropomorphized

characters may be potent role models to younger preschool age children who have not yet

36

mastered the differences between fantasy and reality (Bandura, Ross, & Ross, 1963a;
Howard, 1998; Jaglom & Gardner, 1981).

The context of aggression involving prosocial perpetrators involves elements that
may both increase and decrease risk. In terms of increasing risk, the findings show that
prosocial perpetrators are more likely to use their own physical bodies (e.g., hands, legs,
etc.) to protect others violently and their aggression is usually justified. As noted above,
these trends illustrate that violence in children’s programs — just like the overall
landscape — is glamorized. In terms of decreasing risk, prosocial violence in children’s
shows is more likely to show the consequences of aggression. This was evidenced across
two separate measures. First, prosocial violence results in significantly more extreme
depicted harm than does nonprosocial violence. Second, the violence involving prosocial
perpetrators is less likely to result in unrealistically low levels of harm. These results
suggest that prosocial violence in children’s shows may actually teach that aggression has
negative repercussions, and thus may reduce viewers’ likelihood of learning.

Two other points about children’s shows are worth noting. First, violence
involving prosocial and nonprosocial characters in this type of programming is more
likely to feature humor than is the overall landscape. Specifically, violent interactions in
children’s shows were featured in contexts with humor substantially more often than
were violent interactions in the overall landscape. As documented above, the admixture
of humor and violence can increase the probability of aggression (Baron, 1978;
Berkowitz, 1970). Second, children’s shows are more likely to be depicted in fantastic

and animated contexts. Although these formats pose less risk for older children, younger

37

viewers may be especially vulnerable to violence in such contexts (Dorr, 1980; Thomas
& Tell, 1974).

Together, the findings from Research Question 2 suggest that prosocial violence
in children’s shows may be especially risky for younger viewers. Similar to the overall
landscape, prosocial violence is justified and intended to protect life. However,
children’s shows are a unique genre that features prosocial violence involving more
younger and anthropomorphized perpetrators in fantastic contexts. The only positive
trend is that prosocial violence is more likely to depict the consequences of aggressive
behaviors.

The last research question asked about the frequency and context of violence in
prime time by prosocial status. Prime time differs very little from the overall landscape
in terms of the amount of violence. The only difference observed was for rate per hour.
Prosocial perpetrators engage in less violence per hour than do nonprosocial perpetrators
(1.81 vs. 2.94, respectively). Interestingly, prime time features less aggression per hour
than does children’s shows or the overall landscape.

Prosocial status also influences the types of perpetrators and victims shown in
prime time. When compared to nonprosocial perpetrators, prosocial perpetrators are
more likely to be female, human, white, primary to the plot, and possessing law status.
Although most of the prosocial perpetrators are male, it is interesting to note that prime
time features more females than children’s shows or the overall landscape. Thus, young
females may have more aggressive role models in prime time now than ever before.
Prosocial status also influences the types of victims shown in prime time. Prosocial

aggressors attack individuals that are different from them demographically. That is,

38

prosocial perpetrators are less likely than nonprosocial perpetrators to attack humans,
females, whites, primary characters, and law agents.

In terms of context, only three notable differences were observed. First, prosocial
perpetrators are more likely to be shown protecting the self or others than are
nonprosocial perpetrators. Again, this pattern may heighten the probability of learning or
enacting aggression. Second, the violence of prosocial characters is often depicted as
justified which has been shown to increase violent behaviors in adults (Berkowitz &
Geen, 1966; Berkowitz & Rawlings, 1963). Third, the violence prosocial perpetrators
engage in is more likely to be sanitized. That is, prosocial violence results in more
victims showing no depicted harm and in contexts with less extended suffering than does
nonprosocial violence. Such sanitized trends, as noted earlier, increase the risk of
aggression.

Overall, the aim of this study was to assess the distribution of violence by
prosocial status. Every study is not without its limitations, however. First, this analysis
only examined prosocial aggression in the overall landscape, children’s shows, and prime
time. As such, we still know little about the nature and context of prosocial violence in
other types of shows, such as dramatic series, reality television, or movies. Future
research should examine how often and in what context prosocial perpetrators are shown
in such genres.

Second, the present study was a secondary analysis of a large data set. As a
result, all aspects of the study were already executed. One potential limitation with the

NTVS study is that the authors only captured physical aggression or the threat thereof in

39

their definition of violence. Future research should also monitor verbal violence and how
prosocial status may influence such the context surrounding such aggressive acts.

In spite of these limitations, the results from my thesis suggest that at least two
follow up experiments need to be conducted. The first experiment could examine the
impact of prosocial violence involving females. In explanation, the findings from this
study show that prosocial perpetrators are more likely to be females than are nonprosocial
perpetrators. Although less frequent in nature than male perpetrators, this pattern
suggests that female youngsters may perceive same sex prosocial perpetrators as role
models that are similar to them, and hence inflate the risk of learning. Yet several studies
have found that male characters were chosen as role models by girls more often than
female characters were chosen by boys, due to the fact that male characters were more
plentiful and appeared in more exciting and interesting roles than did female characters
(Howitt & Cumberbatch, 1976; Miller & Reeves, 1976). This research was conducted
over 25 years ago and needs to be updated. It may be the case that girls now identify
with exciting female prosocial perpetrators such as Zena the Warrior Princess or The
Powerpufi‘ Girls. Or, it may be the case that young girls watching television receive a
“double dose” when viewing by identifying with both males and females who are
aggressive for prosocial purposes. Clearly, more research is needed on the influence of
female perpetrators on television that engage in violence for prosocial motives.

The second experiment could assess the impact of several contextual features in a
violent scene on aggressive behavior. That is, most of the previous research in the
violence arena have taken a univariate approach and only manipulated one or two

variables at a time. Yet, the common prototype for violence involving prosocial

4O

 

perpetrators also features justified violence in realistic settings. Future research needs to
assess how several factors in a violence scene interactively impact the learning of
aggression. That is, does adding another factor to a violent scene increase exponentially
the risk of aggressive behavior in the audience? Which contextual feature (prosocial
status, justification, realism) poses the most risk relative to all others? Answering such
questions experimentally will help us to more fully understand the risk that television
violence is posing to youth.
Footnotes

1 The focus of Gerbner and his colleagues’ research was to determine the
relationship between amount of exposure to televised violence and viewers’ beliefs about
violence in the real world. However, their definition did not distinguish different
subtypes of violence within the broad category. That is, the conceptualization of violence
included intentional and accidental acts; acts of nature were counted as well. All acts of
violence were summed and weighted equally in the analysis. Yet not all of these types of
violence my have similar effects in terms of learning aggression (see Wilson et al.,
1997). As Greenberg’s study was designed to examine a fuller range of negative social
behaviors on TV, the categorization of physically aggressive acts accounted for physical
control or restraint of others (e. g., grabbing, shoving, pushing, holding), physical invasion
of privacy, and elaborated fighting. Although these definitions are not well-suited for the
purposes of this research (i.e., a concern about learning physical aggression only), it is
important to note that the concept of aggression used by these researchers was
appropriate for their areas of inquiry. For example, Gerbner was interested in how

exposure to all types of violence (intended or not) contributes to the mean world

41

syndrome. Greenberg, on the other hand, was interested in how a wide range of behavior
may contribute to antisocial behaviors in society.

2 Per the NTVS contract with the National Cable Television Association, several
types of programs were sampled but excluded from the content analysis. These program
types include: religious programs, game shows, “infomercials,” instructional programs,
sports, and hard-breaking news. These program types comprise 14% (n=462) of the
shows in the sample. Thus, the total number of programs sampled and coded for violence
was 2,750.

3 In the analysis, four types of violence were examined. First, credible threat
refers to the perpetrator demonstrates a clear intent to harm the target physically and has
the means ready to do so, but for some reason does not follow through immediately.
Second, a behavioral act is any overt behavior or string of behaviors that is intended to
harm a target, whether successful or not. Third, harmfirl consequences are instances in
which an injured victim is depicted but the violence itself is not shown on screen.
Finally, accidental violence features situations where a character experienced
unintentional harm in the context of ongoing violence.

4 There are time when violent characters aggress individually (e.g., bank robber
shoots gun) or in multiple units (e. g., war scene). Typically, demographic and attributive
qualities of single characters are easy to assess. However, when a series of characters act
violently as a collective group, demographic and attributive characteristics may be
difficult to code. For example, it is possible that a group of aggressors may feature both
males and females as well as blacks and whites. When a group was not homogeneous in

a particular trait or quality, the group or multiple unit was coded as “mixed.” Anytime

42

there was not enough information given in the plot to assess a particular character-related
variable, it was coded as “can’t tell”.

5 Whenever accessing scene level variables, coders were instructed to watch the
entire scene as well as the immediately adjacent scene. In some cases, information about
specific scene level variables does not occur until the imminent threat of violence is
actually over. For example, a scene may depict a heroic character saving a town fi'om a
villain or supernatural creature. In the scene after the hero kills the villain, the town
presents him with an award for his bravery and courage. In this case, the hero is
rewarded but not until the immediate threat of violence is over. To accurately capture the
related frequency of reward and punishment, coders were instructed to watch both the
violent scene as well as the scene immediate after.

6 In a chi-square analysis, Cramer’s V is a statistic that measures the strength of
the statistical dependence between two and/or more variables. Variables with strong
relationships will result in higher Cramer’s V values. Cramer’s V is used when one of
the variables in the analysis features more than two categories.

7 In all chi-square analyses, variables (e. g., type, ethnicity) with an “other”
subcategory are not presented in any of the tables. If any of the “other” subcategories are

significant, they are discussed in the results section of my thesis.

43

APPENDICES

44

Table 1

Pmtrator and Victim Profiles by Prosocial Status

 

 

 

Not Prosocial Prosocial
Perpetrators

Type7

Humans 66.2%. 74.5%.

Anthropomorphized creatures 22.8 19.0

Supernatural creatures 6.7 3.2
Gender

Males 91.9. 83.4.

Females 8.1. 16.6.
Age

Children 2.3 5.2

Teens 3.6. 8.71,

Adults 93.1. 84.6.
Ethnicity

Whites 71.0. 77.0.

Blacks 5.1 6.1

Hispanics 4.3 1.0
Primary Characters 69.0. 88.0.
Law Status 9.2. 20.0.

Victims

Type

Humans 69.5 69.9

Anthropomorphized creatures 20. 3 19.6

Supernatural creatures 3.6 5.8
Gender

Males 82.6. 92.8.

Females 17.4. 7.2.
Age

Children 5.5 3.1

Teens 10.5. 5.1.

Adults 82.4. 90.8.
Ethnicity

Whites 74.7 71.4

Blacks 6.7 5.8

Hispanics 2.0 4.0
Primary Characters 74.7. 63.7.
Law Status 11.3 9.9

 

Note. Percentages reported in the table have to meet two criteria First, the overall chi-square for the variable has
to be significant at p < .05. Second, the percentages have to differ by 5% or greater. Ifthese two conditions were
met, then the percetages do not share a common subscript

45

Table 2
Violent Contextual Variables m Prosocial Status

 

Not Prosocial Prosocial

 

Reasons
% of PATs w/protect life 6.0%. 56.9%.
% of PATs w/anger 29.5. 23.4.
% of PATs w/personal gain 44.8. 10.9.
% of PATs w/other reasons 19.6. 8.8.
Justification 5.2. 64. 1.
Means
% of PATs w/use of body 35.7. 40.9.
% of PATs w/unconventional weapons 8.3 8.7
% of PATs w/conventional weapons 36.8 34.6
Extent & Graphicness
% of PATs w/one act 37.5 40.4
% of PATs w/repeated acts 62.5 59.6
% of PATs w/lethal violence 34.5 33.3
% of PATs in scenes w/blood & gore 19.0 18.3
Consequences
% of PATs w/no pain 48.4 48.9
% of PATs w/no depicted harm 40.5 41.9
% of PATs w/extreme depicted harm 20.2 17.2
% of PATs w/no likely harm 16.3 17.9
% of PATs w/extreme likely harm 34.6. 28-0.
% of PATs w/unrealistic harm 33.5 32.4
% of PATs in programs w/long-term harm 29.8 28.8
Reinforcement
% of PATs in scenes w/rewards 35.3 36.0
% of PATs in scenes w/punishments 38.9 41.4
% of PATs in scenes w/no rew. or puns. 57.8 61.2
Humor
% of PATs in scenes w/humor 41.4 43.4
Realism of Violence
% of PATs in programs w/realistic violence 52.1. 57.6.
% of PATs in programs w/fantastic violence 47.9. 42.4.
% of PATs in programs w/live action 60.4. 67.7.
% of PATs in programs w/some animated aciton 39.6. 32.3.

Anti-Violence Theme
% of PATs in programs w/anti-violence 2.6 2.4

 

Note. Percentages reported in the table have to meet two criteria. First, the overall chi-square for the variable
has to be significant at p < .05. Second, the percentages have to differ by 5% or greater. If these two conditions
were met, then the percetages do not share a common subscript.

46

 

 

 

 

 

Chrldrens Programs Prime Time
Not Prosocial Prosocral Not Prosocial Prosocial
Perpetrators

Type

Humans 32.8% 34.1% 84.2%. 94.1%.

Anthropomorphized creatures 47.8 50.5 6.6 4.3

Supernatural creatures 12.8. 6.9. 5.1 0.4
Gender

Males 92.3. 85.9. 91.9. 79.5.

Females 7.7. 14.1. 8.1. 20.5.
Age

Children 3.8. 13.6. 1.6 1.4

Teens 2.8. 17.9. 3.9 6.4

Adults 91.9. 67.1. 94.3 90.9
Ethnicity

Whites 76.7 74.3 66.8. 80.8.

Blacks 1.0 3.4 6.7 5.2

Hispanics 2.5 0.6 5.6 1.3
Primary Characters 75.3. 89.2. 68.4. 83.0.
Law Status 3.8 7.1 11.3. 39.3.

Victims

Type

Humans 33.5 31.2 91.0. 85.8.

Anthropomorphized creatures 47.0 49.3 5.0 7.0

Supernatural creatures 7.3 11.2 1.2 3.8
Gender

Males 90.3 93.9 77.5. 91.6.

Females 9.7 6.1 22.5. 8.4.
Age

Children 11.7 7.5 2.8 1.3

Teens 16.8. 6.5. 8.7 5.5

Adults 70.1. 84.9. 87.7 92.5
Ethnicity

Whites 73.7 76.5 77.8. 70.2.

Blacks 2.7 2.8 6.1 6.1

Hispanics 0.7 3.0 3.5 6.1
Primary Characters 80.0. 72.2. 68.1. 62.2.
Law Status 6.1 4.2 20.3. 12.0.

 

 

Note. Percentages reported in the table have to meet two criteria. First, the overall chi-square for the variable
has to be significant at p < .05. Second, the percentages have to differ by 5% or greater. Ifthese two conditions
were met, then the percetages do not share a common subscript

Table 4

 

 

 

Violence Variables in Children's and Prime-Time m m Prosocial Status
Children's Programs Prime Time
Not Prosocial Prosocial Not Prosocial Prosocial

Reasons

% Of FATS w/protect life 4.4%). 60.9%. 6.996. 64.9%.

% of PATs w/anger 27.1. 18.9. 27.2. 18.6.

% of PATs w/personal gain 49.1. 9.7. 45.5. 10.1.

% of PATs w/other reasons 19.4. 10.5. 20.4. 6.8.
Justification 4.0. 64.4. 5.0. 68.4.
Means

% of PATs w/use of body 40.8. 45.8. 35.7 34.9

% of PATs w/unconventional weapons 25.8 26.6 13.4 16.6

% of PATs w/conventional weapons 24.1 20.3 45.7 46.0
Extent & Graphicness

% of PATs w/one act 39.5 41.5 36.6 36.8

% of PATs w/repeated acts 60.5 58.5 63.4 63.2

% of PATs wllethal violence 36.3 38.8 33.2 34.2

% of PATs in scenes w/blood & gore 2.1 1.5 37.3 35.6
Consequences

% of PATs w/no pain 58.8 56.9 39.3 43.5

% of PATs w/no depicted harm 55.6. 48.5. 32.1. 39.1.

% of PATs w/extreme depicted harm 5.7. 11.0. 29.2. 24.1.

% of PATs w/no likely harm 16.7 15.6 15.3 17.2

% of PATs w/extreme likely harm 31.2 28.9 40.3 34.3

% of PATs w/unrealistic harm 56.5. 50.6. 21.5 25.4

% ofPATs in programs wllong-term harm 10.8 7.8 48.5. 40.8.
Reinforcement

% of PATs in scenes w/rewards 42.8 45.3 33.0 30.8

% of PATs in scenes w/punishments 35.8 37.4 41.8 44.4

% of PATs in scenes w/no rew. or puns. 61.3 65.5 57.5 60.9
Humor

% of PATs in scenes w/humor 62.9 63.2 29.3 33.5
Realism of Violence

% of PATs in programs w/realistic violence 9.7 9.1 67.5 70.2

% of PATs in programs w/fantastic violence 90.3 90.9 32.5 29.8

% of PATs in programs w/live action 6.3 8.3 87.8 89.1

% of PATs in programs w/animated aciton 93.7 91.7 12.2 10.9
Anti-Violence Theme

% of PATs in programs w/anti-violence 2.1 1.6 3.0 3.4

 

 

Note. Percentages reported in the table have to meet two criteria. First, the overall chi-square for the variable
has to be significant at p < .05. Second, the percentages have to differ by 5% or greater. Ifthese two conditions
were met, then the percetages do not share a common subscript.

48

REFERENCES

American Medical Association. (1996). Physician gaide to media violence. Chicago, IL:
Author.

American Psychological Association. (1993). Violence fld youth: Psychology}
response. Washington, DC: Author.

Atkin, C. (1983). Effects of realistic TV violence vs. fictional violence on aggression.
Journalism Quarterly. 60. 615-621.

Bandura, A. (1986). Social founfiationsgof thought and action: A social cognitive theory
Englewood Cliffs, NJ: Prentice-Hall.

Bandura, A. (1994). Social cognitive theory of mass communication. In J. Bryant & D.
Zillmann (Eds), MediLeffects (pp. 61-90). Hillsdale, NJ: Erlbaaum.

Bandura, A., Ross, D., & Ross, S. A. (1961). Transmission of aggression through
imitation of aggressive models. Journal of Abnormal and Social Psychology, 63,
575-582.

Bandura, A., Ross, D., & Ross, S. A. (1963a). Imitation of film-mediated aggressive
models. Journal of Abnormmnd Social Psychology, 66, 3-11.

Bandura, A., Ross, D., & Ross, S. A. (1963b). Vicarious reinforcement and imitative
learning. Journal of Abnormal and Social Psychology, 67, 601-607.

Baron, R. A. (1971a). Aggression as a function of magnitude of victim’s pain cues, level
of prior anger arousal, and aggressor-victim similarity. Joummf Personality and

Social Psychology, 18, 48-54.

Baron, R. A. (1971b). Magnitude of victim’s pain cues and level of prior anger arousal
as determinants of adult aggressive behavior. Journal of Personality and Social

Psychology, 17, 236-243.

Baron, R. A. (1978). The influence of hostile and nonhostile humor upon physical
aggression. Personality and Social Psychology Bulletin. 4, 77-80.

Benezra, K. (1995, Superbrands). No stopping the Rangers. Adweek Eastern ed, 135-
138.

 

Berkowitz, L. (1970). Aggressive humor as a stimulus to aggressive responses. Journal
of Personality and Social Psychology, 16, 710-717.

Berkowitz, L. (1973). Words and symbols as stimuli to aggressive responses. In J.

Knutson (Ed), Control of aggression: Implications from basic reseaih (pp. 113-
143). Chicago, IL: Aldine-Atherton.

49

Berkowitz, L., & Alioto, J. T. (1973). The meaning of an observed event as as
determinant of its aggressive consequences. Journal of Personality and Social

Psychology, 28, 206-217.

Berkowitz, L., & Geen, R. G. (1966). Film violence and the cue properties of available
targets. Joumflf Personality and Social Paychology, 3, 525-530.

Berkowitz, L., & LePage, A. (1967). Weapons as aggression-eliciting stimuli. Joumal of
Personality @119 SochL Psychology, 7, 202-207.

Berkowitz, L., & Rawlings, E. (1963). Effects of film violence on inhibitions against
subsequent aggression. Jourfl of Abnormal and Social Psychology, 66, 405-412.

Berscheid, E. (1985). Interpersonal attraction. In G. Lindzey & E. Aronson (Eds), I_h_§
Hand Book of Social Psychology (Vol. 2, pp. 413-484). New York: Random
House.

 

Boyatzis, C. J ., Matillo, G. M., & Nesbitt, K. M. (1995). Effects of “The Mighty
Morphin Power Rangers” on children’s aggression with peers. Child Study
Journal 25, 45-55.

 

Bryan, J. H., & Walbek, N. H. (1970). Preaching and practicing generosity: Children’s
actions and reactions. Child Development, 41. 329-353.

Centers for Disease Control. (1991). Position papers from the Third National Injury

Conference: Setting the National _A_genda for Injury Control in the 19905.
Washington, DC: Department of Health and Human Services.

Cohen, J. (1999). Favorite characters of teenage viewers of Israeli serials. Journal of
Broadcasting & Electronic Media 43. 327-345.

Dorr, A. (1980). When I was a child I thought as a child. In S. B. Withey & R. P. Abeles
(Eds), Televisioa_ and socia behavior: Beyond violence aad children (pp. 191-
230). Hillsdale, NJ: Erlbaum.

von Feilitzen, C., & Linne, O. (1975). Identifying with television characters. Journal of
Communicatioa, 25, 51-55.

Feshbach, N. D., & Roe, K. (1968). Empathy in six- and seven-year-olds. Child
Development, 39, 133-145.

Friedrich-Cofer, L., & Huston, A. C. (1986). Television violence and aggression: The
debate continues. Psychological bulletin. 100. 364-371.

50

Geen, R. G. (1975). The meaning of observed violence: Real vs. fictional violence and
consequent effects on aggression and emotional arousal. Journal of Research in
Personality, 9, 270-281.

Gerbner, G., & Gross, L. (1976). Living with television: The violence profile. Journal of
Communication 26, 173-199.

 

Gerbner, G., Gross, L., Morgan, M., & Signorielli, N. (1980). The “mainstreaming” of
America: Violence profile no. 11. Journal of Communicatiop, 30, 10-29.

Greenberg, B. 8., Edison, N., Korzenny, F., F emandez-Collado, C., & Atkin, C. K.
(1980). Antisocial and prosocial behaviors on television. In B. S. Greenberg
(Ed), Life on Television (pp. 99-128). Norwood, NJ: Ablex.

Gunter, B. (1985). Dimensions of television violence. Aldershots, England: Gower.
Hoffner, C. (1996). Children’s wishful identification and parasocial interaction with

favorite television characters. Journal of Broadcasting & Electronic Media, 40.
389-402.

 

Hoffner, C., & Cantor, J. (1985). Developmental differences in responses to a television
character’s appearance and behavior. Developmental Psychology, 21, 1065-1074.

Hoffner, C., & Cantor, J. (1991). Perceiving and responding to mass media characters.
In J. Bryant & D. Zillmann (Eds), Responding to the screen (pp. 63-101).
Hillsdale, NJ: Lawrence Erlbaum.

Howard, S. (1998). Unbalanced minds? Children thinking about television. In S.
Howard (Ed), Wired-up: Young people and the electronic media. London: UCL
Press.

Howitt, D., & Cumberbatch, G. (1976). The parameters of attraction to mass media
figures. In R. Brown (Ed), Children aad television (pp. 167-183). Beverly Hills,
CA: Sage.

Huesmann, L. R, Lagerspetz, K., & lEron, L. D. (1984). Intervening variables in the TV-
violence-aggression relation: Evidence fi'om two countries. Developmental

Psychology, 20, 746-775.

Huesmann, L.R., & Miller, L. S. (1994). Long-term effects of repeated exposure to
media violence in childhood. In L. R. Huesmann (Ed), Aggressive Behaaior:
Current perspectives (pp. 153-186). New York: Plenum Press.

Huston-Stein, A., Fox, 8., Greer. D., Watkins, B. A., & Whitaker, J. (1981). The effects
of TV action and violence on children’s social behavior. Journal of Genetic

Psychology, 138, 183-191.

51

Jaglonr, L. M., & Gardner, H. (1981). The preschool viewer as anthropologist. New
Directions in Child Development, 13, 9-29.

Jose, P. E., & Brewer, W. F. (1984). Development of story liking: Character
identification, suspense, and outcome resolution. Developmental Psychology, 20,
91 1-924.

Lando, H. A., & Donnerstein, E. I. (1978). The effects of a model’s success or failure on
subsequent aggressive behavior. Journal of Resarrch In Personalig, 12, 225-234.

Leflrowitz, M. M., Bron, L. D., Walder, L. Q., & Huesmann, L. R. (1977). Growing up
to be violent: A longitudinal study of the development of aggression. New York,
NY: Pergamon Press.

Liefer, A. D., & Roberts, D. F. (1972). Children’s responses to television violence. In J.
Murray, E. Rubinstein, G. Comstock (Eds), Television and social behavior. (Vol.
2, pp. 43-180). Washington, D. C.: United States Government Printing Office.

Liss, M. B., Reinhardt, L. C., & Fredriksen, S. (1983). TV heroes: The impact of rhetoric
and deeds. Journal of Applied Developmeraal Psychology, 4, 175-187.

Meyer, T. P. (1972). Effects of viewing justified and unjustified real film violence on
aggressive behavior. Jouraal of Personality and Social Psycholo 23, 21-29.

Miller, M., & Reeves, B. (1976). Dramatic TV content and children’s sex-role
stereotypes. Jourml of Broadcasting, 20, 35-50.

National Institute of Mental Health. (1982). Televisiorflnd behavior: Ten years of
scientific progress and implications for the eighties (Vol. 1). Summary report.

Washington, DC: US. Government Printing Office.
Nielsen Media Research. (2000). 2000 Report on television. New York, NY: Author.

Oldenburg, A., & Healy, M. (1999, May 6). Parents see violence, not sex, as biggest
concern with media. USA Today, p. D1.

Page, D., & O’Neal, E. (1977). “Weapons effect” without demand characteristics.
Psychological Reports, 41, 29-30.

Paik, H, & Comstock, G. (1994). The effects of television violence on antisocial
behavior: A meta-analyses. Communication Research. 2_l, 516-546.

Plagens, P., Miller, M., Foote, D., & Yoffe, E. (1991, April 1). Violence in our culture.
Newsweek, 117, 46-52.

52

Potter, W. J ., & Ware, W. (1987). An analysis of the contexts of antisocial acts on prime-
time television. Comflnication Regarch. 14. 664-686.

Reeves, B. (1979). Children’s understanding of television people. In E. Wartella (Ed),

Children communicating: Media and development of thought, speech,
understanding (pp. 115-155). Beverly Hills, CA: Sage.

 

Reeves, B., & Greenberg, B. S. (1977). Children’s perceptions of television characters.
Human Communication Research 3. 113-127.

Reeves, B., & Miller, M. M. (1978). A multidimensional measure of children’s
identification with television characters. Journal of Broadcasting, 22, 71-86.

Signorielli, N. (1990). Television’s mean and dangerous world: A continuation of the
cultural indicators perspective. In N. Signorielli & M. Morgan (Eds), Cultivation
analysis: New directions in mediaaeffects research (pp. 85-106). Newbury Park:
Sage.

Smith, S. L., Wilson, B. J ., Kunkel, D., Linz, D., Potter, J ., Colvin, C. M., & Donnerstein,
E. (1998). Violence in television programming overall: University of California,
Santa Barbara study. National Television Violence Study. (Vol. 3, pp. 5-220).
Thousand Oaks, CA: Sage.

 

Sprafkin, J. N., Liebert, R M., & Poulos, R. W. (1975). Effects of a prosocial televised
example on children’s helping. Journal of Experimeptal Child Psychology, 20,
119-126.

Stein, A. H, & Friedrich, L. K. (1972). Television content and young children’s
behavior. In J. P. Murray, E. A. Rubinstein, & G. A. Comstock (Eds), Television
m sociflbalryior: Reports and papers: Vol. 2. Television a_rrd social learning
(pp. 202-317). Washington, DC: Government Printing Office.

Stipp, H. (1993). The challenge to improve television for children: A new perspective.
In G. L. Beny & J. K. Asamen (Eds), Children & Televis_i<_)r_1 (pp. 296-302).
Newbury Park, CA: Sage.

Surgeon General’s Scientific Advisory Committee on Television and Social Behavior.
(1972). Television and growing up: The impact of televised violence.
Washington, DC: US. Government Printing Office.

Swart, C., & Berkowitz, L. (1976). Effect of a stimulus associated with a victim’s pain
on later aggression. Journal of Personality and Soci_al Psycholo 33, 623-631.

Thomas, M. H, & Tell, P. M. (1974). Effects of viewing real versus fantasy violence
upon interpersonal aggression. Journal of Research In Personality, 8, 153-160.

53

Wilson, B. J ., Kunkel, D., Linz, D., Potter, J., Donnerstein, E., Smith, S. L., Blumenthal,
E., & Gray, T. (1997). Violence in television programming overall: University of
California, Santa Barbara study. National Televiaion Violence Study. (Vol. 1, pp.
3-268). Thousand Oaks, CA: Sage.

Wilson, B. J ., Kunkel, D., Linz, D., Potter, J ., Donnerstein, E., Smith, S. L., Blumenthal,
E., & Gray, T. (1998). Violence in television programming overall: University of
California, Santa Barbara study. National Television Violence Study. (Vol. 2, pp.
2-204). Thousand Oaks, CA: Sage.

 

Wilson, B. J ., Cantor, J ., Gordon, L., & Zillmann, D. (1986). Affective responses of
nonretarded and retarded children to the emotions of a protagonist. Child Study
Journal 16, 77-93.

 

54

 

 

_ _ “ms..—