Violence on Canadian Television and Some of Its Cognitive Effects

André Gosselin (Université Laval)

Jacques DeGuise (Université Laval)

Guy Pacquette (Université Laval)

Laplante Benoit (Université du Québec)

Abstract: Inspired by George Gerbner's method of cultural indicators, content analysis of fictional programs broadcast on Canada's eight largest television networks has indicated that the level of violence on Canadian television, although overall not as high as American television's violence content, reached comparable levels on occasion, particularly on the private networks. A survey conducted on a sample of university students provides evidence that television viewing affects beliefs concerning the level of violence in one's surroundings, even in a population of media-literate adults, although their level of fear does not seem to be influenced in the same way.

Résumé: A l'instar de l'approche des indicateurs culturels de George Gerbner, une analyse de contenu des émissions dramatiques diffusées sur les huit plus grands réseaux de télévision au Canada démontre que le niveau de violence à la télévision canadienne, sans être globalement aussi élevé que celui observé aux États-Unis, atteint occasionnellement des niveaux comparables, surtout aux réseaux privés. Un sondage auprès d'un échantillon d'étudiants universitaires montre que l'écoute de la télévision a un effet sur les impressions qu'ont même des adultes formés aux médias à l'égard du niveau de violence dans la société, quoique la télévision ne semble pas influer sur leur niveau de peur de la même manière.

Violence on television has, for many years, been a subject of great concern for Canadians. This concern was clearly expressed in the early 1990s by a petition signed by several hundred thousand supporters and presented to the Prime Minister of Canada by a young girl, Virginie Larivière, whose sister had been brutally murdered. Pressure from both public opinion and the Canadian government urged broadcasters to adopt, in 1994, a code of ethics regarding violence on television.1

Along with other researchers (Bandura, 1986; Comstock, 1989), we remain sceptical as to the importance of television as an indicator of violence in our society, insomuch as very few studies exist in Canada on the subject of violence portrayed on Canadian television and how it is perceived. The goals of this study were twofold: (1) to conduct a content analysis of violence on Canadian television according to George Gerbner's methodology; (2) to verify, in a Quebec student population, if the cultivation analysis method was conclusive, notably with regard to the hypothesis that "heavy" viewers (as compared to "light" viewers) tend to perceive their surroundings in a certain way and hold certain beliefs, mirroring an image of reality portrayed on television.

Violence on Canadian television

Various methods of analyzing violence on television have been applied in previous studies (Signorielli, 1984). In Canada, one of the most significant studies on this topic was conducted in 1976 (Williams, Zabrack, & Joy, 1982). It dealt with varying degrees of violence by distinguishing "conflicts," physical attacks, and verbal attacks. This study did not treat Canadian television specifically, but rather examined 109 programs which had the highest BBM ratings nationally. As Canadians are great consumers of American television, only 77% of the programs in the airtime studied was Canadian content.2

Our collective goal was not to study the kind of television Canadians watch but what Canadian networks broadcast. The most appropriate method, in these conditions, seemed to be the "violence index" perfected by the Annenberg group, led by Gerbner and colleagues (Gerbner, Gross, Jackson-Beeck, Jeffries-Fox, & Signorielli, 1978; Signorielli, 1984). Although the subject of controversy (Blank, 1977), this method has been used in many other countries, notably in the Netherlands (Bouwman & Stappers, 1984), in Australia (McCann & Sheehan, 1985), and in Japan (Iwao, De Sola Pool, & Hagiwora, 1981).

We may recall that Gerbner's violence index was composed of five elements: (a) the percentage of programs3 containing violence (% P), (b) the number of violent acts per program (R / P), (c) the number of violent acts per hour (R / H), (d) the percentage of main characters involved in acts of violence (% V), and (e) the percentage of main characters involved in homicides (% K). The index was calculated as follows: VI = (% P) + 2(R / P) + 2(R / H) + (% V) + (% K).

In operational terms, for the purpose of this study, violence was defined as any explicit act of force destined to injure or kill, or the expression of any serious threat to injure or kill a character, whether human or humanlike, regardless of the context in which the act occurred. Also included as violent acts were accidents and natural phenomena resulting in injury or death.4 The programs studied were of a fictional nature and were broadcast during the week of March 21-28, 1993. This week was chosen as it appeared to present no particularities and was therefore considered as being representative of what was normally shown on Canadian television.

The fictional programs retained for analysis originated from the major networks, namely, the public (Société Radio-Canada and the Canadian Broadcasting Corporation), private (TVA, Télévision Quatre Saisons, CTV, and Global Television), and educational networks (Radio-Québec in Quebec and TVO in Ontario). Only those programs aired each weeknight during prime time and on weekends between 6 a.m. and 6 p.m. were retained for analysis. The hours examined during the week were 7 to 11 p.m. for the anglophone stations, and 6 to 10 p.m. for the francophone stations. The corpus was comprised of programs totaling 6,936 minutes (115.6 hours). American statistics used for comparison covered programming during one week in 1992, namely, 94 programs presented during 54.8 hours by CBS, ABC, and NBC, in the evening between 7 and 11 p.m. and on the weekend between 6 a.m. and 6 p.m. (Gerbner, 1993).

Program analysis was conducted by a five-member team of coders who underwent a one-week training session. Each program was analyzed by two coders working on an individual basis. However, in the event of a disagreement, a consensus was required to obtain satisfactory results. To produce results that were comparable to those from the United States, rigorous attention was given to the coding standards laid down by Gerbner and his colleagues.

The violence index was measured using a series of variables considered to be relevant to the Canadian context. The following variables were established: the networks, type of network (private, public, and educational), broadcast time slots, audience (adults and children), language used (English and French), production location, program tone (comical, serious), subject matter (comedy, drama, etc.), and format (cartoon, film, soap opera, series, children's program5 [-.25m]).6

The overall violence score for the Canadian programs registered at 106.1 (see Table 1), which was only significant when compared to a similar situation. Being that the stations used for comparison were American commercial networks, the "educational" stations were not considered, leaving public and private Canadian stations for valid comparison. A score of 129.9 was therefore achieved, which compared favourably with the American score of 169.6 (see Table 2), indicating 23.4% less violence in Canadian programs than in American programs. Each item of the Canadian score was lower than those of the American score, with the exception of percentage of characters involved in murders (5.9 in Canada versus 4.8 in the United States). This led us to deduce that if the Canadian characters are less often violent, their violence was overall much greater than that of their American counterparts.

Table 1
Violence Index for Canadian Television Networks
No. of all programs 21 27 18 53 28 41 28 12 228
% of programs with
violence (% P) 66.7 51.9 83.3 7.5 50 75.6 536 25 48.2
No. of violent episodes 110 226 81 16 67 230 49 9 788
Violent episodes
per program (R / P) 5.2 8.4 4.5 0.3 2.4 5.6 1.8 0.8 3.5
Violent episodes
per hours (R / H) 10.1 11.6 4.5 1.6 3.6 11 4.3 1.5 6.8
No. of leading characters 104 135 90 108 142 181 93 38 891
% of leading characters
involved in violence
(% V) 36.5 32.6 51.1 11.1 23.9 43.6 30.1 10.5 32
% of leading characters
involved in killing
(% K) 1.9 9.6 6.7 2.8 2.1 9.4 3.2 0 5.3
Program score (PS)
(% P) + 2(V/ E)
+ 2(V/ H) 97.3 91.9 101.4 11.3 62 108.9 65.6 29.5 68.8
Character score (CS)
(% V) + (% K) 38.5 42.2 57.8 13.9 26.1 53 33.3 10.5 37.3
Violence index PS+CS 135.8 134.1 159.1 25.1 88 161.9 99 40 106.1

In both countries, children's programs were more violent than adult programs (an index of 192.5 and 114.4 respectively -- a difference of 68.1%; see Table 2). They contained four times more violent scenes per hour and 76.9% of these programs contained violence, compared to only 58.9% for adult programs. Approximately the same differences were observed in the United States, but with higher percentages.

CTV (161.9) and TQS (159.1, see Table 1) registered the highest violence ratings; however, TVA presented the highest number of violent acts per hour. As was expected, educational television (RQ and TVO) registered the least amount of violence (25.1 and 40.0). It is interesting to note that SRC scored higher than TVA mainly because of children's programming which was mostly made up of violent cartoons. When the latter were not considered in the compilation, the SRC score dropped to 64.4,7 almost the same level as the educational networks, while remaining practically unchanged on the other networks.

Table 2
Violence Index for Adult and Children's Programs
on Canadian and American Television
(Educational television excluded)
Canadian television American television
Adult Children's Total Adult Children's Total
Violent programs 58.9 76.9 63.2 74.1 82.5 77.7
Violent episodes
per program 3.7 7.8 4.7 3.4 7.8 5.2
Violent episodes
per hour 5.4 21.5 7.7 4.0 32.0 9.0
Leading characters
in violence 31.0 52.9 36.1 47.2 78.9 58.6
Leading characters
in killing 6.5 4.0 5.9 5.7 3.3 4.8
Index 114.5 192.5 129.9 141.7 244.2 169.6

Private networks had more violent programming than did public television (scores of 140.2 and 107.6 -- a difference of 30.3%; see Table 3), yet this score remained lower than that of the American networks (140.2 and 169.5). The English- and French-language networks showed almost the same amount of violence (see Table 3). English-language networks rated slightly higher, all stations considered. However, French-language networks scored higher when educational television was excluded. Nevertheless, French-language stations reported 54.9% of all violent scenes compiled (433/788; see Table 3).

Table 3
Violence Index for Canadian Television by Status and Language
(Educational television excluded)
Language Status
French English Private Public
Violent programs 65.2 61.9 65.8 57.1
Violent episodes
by program 6.3 3.6 5.1 3.6
Violent episodes
per hour 8.6 6.8 8.4 6.0
Leading characters
in violence 38.9 33.9 39.5 29.3
Leading characters
in killing (% K) 6.4 5.5 7.8 2.0
Index 140.3 122.0 140.2 107.6

Analysis of programs broadcast by non-educational networks according to production location revealed that programs originating in the United States were the most violent, achieving a rating of 153.0 (see Table 4) compared to only 85.2 for programs originating in Canada. Among the latter, those produced in Quebec reported a score of 49.6, compared to 111.9 for programs produced in the other Canadian provinces.

Table 4
Violence Index of Canadian Television by Origin of Programs
(English Canada, Quebec, United States) (Educational television excluded)
English Total for United
Canada Quebec Canada States Others
Violent programs 61.5 26.3 47.6 72.2 62.5
Violent episodes
per program 2.6 0.9 1.9 6.4 4.2
Violent episodes
per hour 5.5 2.0 4.0 8.8 8.9
Leading characters
in violence 32.5 15.2 24.9 42.4 32.9
Leading characters
in killing 1.7 2.2 1.9 8.0 0.0
Index 111.9 49.5 85.2 153.0 121.9

It was astonishing to discover that the programs produced in Quebec scored so low, considering that the francophone networks achieved a score almost equal to that of the anglophone networks. This paradox was explained by the fact that during the hours retained for analysis, the francophone networks showed more films recording a higher violence index than other types of programs.

In analyzing the various formats, the cartoons were rated separately from the other children's programs, scoring the highest violence index (206.8; see Table 5), followed by films (178.4) and series (98.9). Soap operas achieved an even lower score (55.6), with only 35.3% of them containing any violence. With the exception of cartoons, programs for children displayed almost no violence (1.9%) and less than 2% of the main characters presented to children were involved in violence (% V = 1.7). Therefore, the violence shown to children originated almost exclusively from cartoons. Of these, 79.5% contained violence, averaging 24.8 violent scenes per hour. In addition, 57.7% of the main cartoon characters were violent.

Table 5
Violence Index for Canadian Television by Program Format
Soap Children's
Cartoons Movies operas Series programs Total
Violent programs 79.5 79.3 35.3 53.9 1.9 48.2
Violent episodes
per program 7.9 9.6 0.9 2.1 0.0 3.5
Violent episodes
per hour 24.8 6.3 1.5 4.3 0.2 6.8
Leading characters
in violence 57.7 52.1 13.4 27.4 1.7 32.0
Leading characters
in killing 4.3 14.1 2.1 4.6 0.0 5.2
Index 206.8 177.4 55.6 98.8 4.0 106.0

Faced with the challenge of defining violence and rendering this definition operational, measuring violence on television was not an easy task. Opting for Gerbner's definition was of help in making comparisons with the American situation. This rigorous method left little or no room for the coder's subjectivity, and therefore appeared reliable. This approach, however, was not exempt from criticism precisely because it did not consider the circumstances in which the violence was committed. This discrepancy was particularly visible in the cartoons which were objectively very violent, but in which the violence was often totally unrealistic or presented in a comical fashion, lessening if not dissipating the generated anxiety and also, perhaps, the consequent negative effects on children. It is in fact true that this kind of violence, offered as entertainment, may appear to be sanctioned by society, thus clouding what opinion children may have on the subject.

Gerbner's index is also a theoretical construct that may seem arbitrary, notably in its heterogeneous considerations and the weight accorded to them. Its main attribute resides in the fact that it has long been used in the United States and has become, of sorts, a standard of measurement. Made up mostly of percentages, it is more qualitative than quantitative in nature: It does not indicate how much violence is shown, only that the programs are very or not very violent. To an extent, if a station presented but one very violent fictional program, its index would nevertheless score very highly. Similar findings were obtained from TQS's programming that, despite showing only 81 violent scenes, recorded a higher violence index than TVA which showed three times as much violence.

Programs broadcast in Canada were significantly less violent than those broadcast in the United States. This difference would have been significantly higher if Canadians did not import so many programs from their American neighbour. American programming not only contained more violence, but the present study revealed that 65 of the 96 violent programs broadcast in Canada (67.7%), as well as 579 of the 733 violent scenes (79%), were produced in the United States. Although Canadians were not forced to choose these programs, the fact remains that the violent nature of their programming is influenced by American culture.

It was also observed that the programs from private stations were significantly more violent that those of public stations. In analyzing format, it was even more troubling to notice that apart from the cartoons, films showed the most violence, followed by weekly series, and far behind, by soap operas. In fact, the same order was obtained when rating programs according to cost, suggesting that the higher the production costs, the more violent a program was likely to be. This, perhaps, explains the violent nature of American productions that benefit from more generous budgets than Canadian productions.

The effects of watching television

Content analysis showed that violence was prevalent in both Canadian and American television, thus supporting the common view that since the advent of television in Canada, both French and English Canadians have been exposed to levels of television violence quite similar to those observed in the United States. Canadian stations have always broadcast many American programs and films, with the consequence that this medium has given to Canadians an exaggeratedly violent image of the world. In the Canadian context, it should be quite easy to verify Gerbner's thesis that the more people watch television and are exposed to its violent content, the more likely they will be to perceive their society or the world around them as being more violent than it really is.

In order to test this hypothesis empirically and in an effort to prepare future international comparisons with the United States (Carlson, 1983; Gerbner & Gross, 1976; Gerbner, Gross, Eleey, Jackson-Beeck, Jeffries-Fox, & Signorielli, 1977a, 1977b, & 1977c; Gerbner et al., 1978; Gerbner, Gross, Morgan, Signorielli, & Jackson-Beeck, 1979; Hirsch, 1980, 1981; Hughes, 1980), Great Britain (Gunter & Wober, 1983, 1988; Wober, 1978; Wober & Gunter, 1982), Australia (Pingree & Hawkins, 1981), Holland (Bouwman & Stappers, 1984), Sweden (Hedinsson & Windahl, 1984), and Ontario (Doob & Macdonald, 1979), we designed a 200-item questionnaire using variables, indices, and questionnaire items similar to those used in studies from these countries and adapting them to the Canadian context. Methodological critiques of the cultivation analysis theory of Gerbner and associates were considered in the adaptation of the questions (Doob & Macdonald, 1979; Hirsch, 1980, 1981; Hughes, 1980; Newcomb, 1978); we also used a somewhat larger set of attitudes most likely to be influenced by media depiction of violence (Barrile, 1984; Carlson, 1985; Doob & Macdonald, 1979; Gunter, 1987; Gunter & Wober, 1988).

In addition to the standard sociodemographic and socioeconomic variables, and to the questions on the uses of media and on other cultural activities, the questionnaire included over 150 items related to a wide range of representations, beliefs, and attitudes related to real or fictional violence. These items covered: (a) pessimism and social demoralization, including Gerbner's Anomie Index (Signorielli, 1990); (b) distrust, including Gerbner's Mean World Index (Bryant, Carveth, & Brown, 1981; Morgan & Signorielli, 1990; Wober, 1978; Zillmann & Wakshlag, 1985); (c) evaluative perception of the actual violence, including Gerbner's Images of Violence Index (Carveth & Alexander, 1985; Doob & Macdonald, 1979; Gerbner et al., 1977a, 1977b, 1977c; Hawkins & Pingree, 1980; Ogles & Sparks, 1989, 1994; Potter, 1991); (d) perception and fear of the danger, risks, or vulnerability when encountering actual violence, including Gerbner's Perception of Danger Index (Gerbner & Gross, 1976; Gunter & Wober, 1983; Hawkins & Pingree, 1980; Ogles & Sparks, 1989, 1994; Wober, 1978; Zillmann & Wakshlag, 1985); (e) tendency towards violence; (f ) tolerance of violence and inclination towards punitive justice; (g) approval of violence; (h) tendency towards criminal activity (hidden delinquency); (i) perception of the context of television violence; ( j) perception of police work; (k) perception of judges and the justice system; and (l) attitude towards the rights of the accused. We also included in the questionnaire some indicators that are not usually found in cultivation analysis research, but that appeared to us to be of interest while preparing future research on the effects of television viewing: (m) causal attributions of the violence and criminal activity of individuals (variables inspired by the cognitive theory on causal attributions); (n) public opinion on the role of information media with regard to the way real violence is reported; (o) perception of the effects of television violence, whether real or fictional; and (p) opinion in the public debate on violence in the media.

For the purpose of the present study, the questionnaire was answered by 360 first-year students enrolled in the Département d'information et de communication at Quebec City's Université Laval during the first week of September 1993. This sample is obviously not drawn from the general population and was chosen for two reasons related to the objectives of this particular study. One of our goals was to assess the reliability and validity of a series of composite indices used in a number of studies on the effects of television viewing. Surprisingly, while most studies in this area report empirical evidence of a relationship between television viewing and the answers given to a set of questions, little is known of the metric properties of the instruments used to measure the attitudes and beliefs that are supposed to be affected. Using a somewhat captive sample made it possible to investigate the reliability and validity of a larger selection of composite indices than would have been possible on a sample of the general population. The use of such samples for this purpose is common practice in psychological studies. This part of our work showed that most of the item-based indices used in the research on the effect of television viewing, including those most commonly used in cultivation analysis studies, have poor or very poor metric properties (Laplante & Gosselin, 1996).

Collecting data in this sample also gave us an opportunity to examine the relationship between television viewing and attitudes towards the world in a manner different from what is usually done. First-year students, especially in a department of communications, are more educated, younger and, hopefully, more sophisticated about television and the media than the general population. They make up a special population that should be more aware of the conditions of television programming as well as of the mechanisms of its reception by viewers and, for this reason, more immune to it. Any evidence of an effect of television viewing detected in such a sample could be considered as very robust empirical evidence. The ways in which television viewing would affect these people could be thought as the minimum mechanism by which television viewing can affect people.

Our study of the metric properties of the indices widely used in cultivation analysis research had shown that these had weak metric properties and that researchers should be cautious when using them to assess the effect of television viewing on viewers (Laplante & Gosselin, 1996). However, we found that one of these indicators -- perception of danger -- was indeed made of two different dimensions and that both could be measured with some reliability using a different subset of questions that made up the original index. One of these dimensions clearly refers to the beliefs an individual has about the level of danger in the surrounding world, while the other has more to do with emotion and can be interpreted as a measure of how much an individual fears the surrounding world.8 We therefore decided to label these two dimensions, respectively, cognition and fear. Since they were the only constructs that turned out to have some metric robustness out of the various cultural indicators we had examined, we became interested in investigating their dependence upon television viewing.

Cultivation analysis is mainly focused on providing empirical evidence of the effect that television viewing has on perception. However, our work on the metric properties of the cultural indicators had shown that it seemed difficult to measure perception as a whole and that what cultivation analysis theory commonly presents as a single concept really breaks down into two different components, a belief and an emotion. With such a result, it became quite impossible to carry further any reflection on the effect of television viewing within the conventional framework of cultivation analysis. At this stage, we felt in need of enlarging our theoretical framework. Because we had some evidence that television viewing affected viewers through belief and emotion rather than sheer perception, we drew our main hypothesis from a conventional cognitivist framework. Basically, we hypothesized that if television viewing had an effect on viewers' beliefs and emotions, and if this effect appeared as a set of relationships between television viewing, cognition, and fear, it should be made of at least two relations: an influence of viewing on cognition and an influence of cognition on fear. Such an hypothesis is not unheard of in the area of cultivation analysis: Hawkins & Pingree (1990) have already proposed a reformulation of the original thesis which relies on a similar mechanism.

We proceeded to test our hypothesis through the use of regression analysis.9 The results are reported in Table 6.

Table 6
Regressions of Viewing, Cognition, and Fear
Equation number 1 2 3 4 5 6 7 8
variable Viewing Cognition Cognition Fear Fear Fear Fear Fear
Womena 1.985 (.831)c .286 (.295) .186 (.295) 1.523 (.153)e 1.695 (.192)e 1.713 (.192)e 1.716 (.188)e 1.140 (.328)e
Less than 20 .280 (.922) 1.009 (.326)d 1.014 (.323)d .307 (.170) .267 (.353) .258 (.352) .165 (.345) .232 (.345)
25 through 34 .613 (1.284) .101 (.464) .053 (.460) .255 (.240) .502 (.353) .493 (.352) .446 (.345) .481 (.343)
35 or more 2.050 (1.684) 2.334 (.587)e 2.436 (.583)e .412 (.304) .723 (.488) .676 (.489) .847 (.480) .744 (.480)
Viewing .050 (.019)d .012 (.010) .006 (.010) .006 (.010)
Cognition .112 (.027)e .034 (.045)
Less than 20 .011 (.402) .026 (.402) .007 (.393) .102 (.394)
25 through 34 .435 (.478) .438 (.478) .362 (.467) .388 (.465)
35 or more 1.830 (.620)d 1.792 (.620)d 1.627 (.607)d 1.398 (.614)c
Women and
cognition .120 (.056)
Origin 10.917 (.735)e 4.938 (.260)e 4.398 (.328)e 2.316 (.135)e 2.027 (.152)e 1.902 (.184)e 1.418 (.215)e 1.794 (.277)e
R2 .022 .081 .099 .251 .271 .274 .309 .564
The reference category is Men.
The reference category is aged 20 through 24.
Significant at the .05 level.
Significant at the .01 level.
Significant at the .001 level.

In our population, women watch television about two hours less than men in a typical week; there are no significant differences between age groups (equation 1). The strength of the belief in the dangerous nature of the surrounding world varies according to age: younger people are more prone to sharing such a belief than older people (equation 2). This relationship holds once television viewing is controlled. Viewing itself, as hypothesized, has an effect on cognition: each hour of television viewing adds .05 to the score on the cognition scale (equation 3).

Women fear the surrounding world more than men; the level of fear does not seem to vary across age groups (equation 4). However, when testing for the interaction between age and gender (equation 5), we see that the pattern is somewhat more complex. The level of fear does not vary significantly across age groups among men (their scores on the fear scale range from 2.027 to 2.750), but it does among women: under 35, women have a greater fear of the surrounding world than men of the same age group (their scores range from 3.722 to 4.000) while women aged 35 or more, with a score of 2.615, are less prone to fear than younger women and have a level of fear similar to men.10

Viewing does not seem to have any direct effect on fear (equation 6). Cognition, on the other hand, has a positive influence on the level of fear: people who believe the world is dangerous appear to fear it the most (equation 7). However, this interpretation is misleading: testing for interaction between gender and belief shows that this relationship holds only for women (equation 8).

These results are summarized in Figure 1. Two relationships are at the centre of this system: the influence of television viewing on cognition and the effect of cognition on fear. These two relationships are linear and both are in accordance with our hypothesis: people who watch more television tend to believe the world is more dangerous than those who watch less television, and people who believe the world is dangerous fear it the most. However, these results do not support clearly the existence of influence of television viewing on fear. Even though the first relationship is true for our entire population, the second one holds only for women: men who believe the world is dangerous do not fear it more than men who believe otherwise. Furthermore, there is no raw influence of television viewing on fear, which means that even for women, the existence of the effects of viewing on belief and of belief on fear cannot be interpreted as evidence of an indirect path of influence of viewing on fear: the covariance of viewing and cognition and the covariance of cognition and fear do not intersect.

11.5P 22P 1

In brief, our analyses show that television viewing affects the beliefs people have about the level of danger in society, even when controlled for the effects of age and gender and even in a population made of people who are prone to know the media and how they work much more than the general population. However, television viewing does not seem to influence the emotion people feel about the surrounding world, neither directly or indirectly. The relationship between cognition and fear holds for women but does not involve television viewing: since it links a belief to an emotion, it may well be interpreted in a cognitivist perspective, and since it involves differences between men and women, it may well be seen as a case for gender studies. But, at least in our population, the influence of television viewing seems restricted to beliefs.

Our empirical study of the effect of television viewing has shown first that most of the cultural indicators commonly used in cultivation analysis research have poor metric properties. Then it provided us with evidence that, when using two somewhat valid and reliable indices, television viewing does not have any influence on the emotion people feel about the surrounding world. Oddly enough, these results give some support to the cultivation thesis. Even though the original indicators were dismissed as poor measurements of the components of perception which could be influenced by television viewing, they became the stuff we used to craft better instruments. Although we found no evidence of influence of television viewing on an emotion often associated with it, we discovered that television viewing affects the beliefs about the level of violence in society even in media literate people. This can be seen as robust evidence of the role television plays in the shaping of how we see the world.


This study is by no means conclusive. Research on the cognitive aspects of the effects of television is relatively recent and many studies are still to come. Content analysis of programs has thus far failed to consider the context in which the instances of violence appear, addressing only physical violence and completely ignoring psychological violence. Our study was limited to examining only one aspect of the perception of the world among those aspects possibly affected by watching television. To obtain more conclusive results, future research must focus on broader populations, using upgraded instruments to measure those dimensions of the perception of reality affected by watching television.

Generally speaking, it has become in fact more and more apparent that if watching television affects the beliefs, opinions, and attitudes of viewers, it is not solely responsible, and the mechanisms by which it works are certainly not simple. Any researcher interested in this question must then choose between two approaches, namely, either to seek to reveal the existence of a definite influence of watching television on the representation of reality or to better understand the role of television viewing in the development of the representation of reality.

Gerbner et al. adopted the first approach in their empirical research, using data tables and bivariate or trivariate testing to determine the plain effect of television viewing on attitudes, beliefs, or opinions, exactly the same way variance analysis does in testing the effect of an experimental treatment. Empirical analysis does not consider the mechanisms by which watching television acts on perception, no more than experimenting with a new drug on two volunteer test groups shows how the drug works. Although Gerbner proposed an interpretation of his theory, this interpretation did not constitute a mechanism nor the subject of empirical analysis.

On the other hand, the second approach attempts to decipher just how watching television acts on -- or within -- the perception, and proceeds with an empirical analysis of the mechanism (or mechanisms) by which it works. The Hawkins & Pingree (1990) reformulation, simple though it may be, provides a starting block for this type of research, which is obviously much more complex and as much theoretically as methodologically demanding: there is a definite need for more detailed mechanisms dealing with the representation of certain aspects of social reality, mechanisms that will integrate the elements of both the social and psychological processes, at least one cultural practice (television viewing), and, of course, measures of the system of beliefs, attitudes, and opinions believed to be affected by the process.

The two research approaches differ greatly and, in a sense, represent opposite methodological strategies. Studies on the effects of television viewing, and particularly those on the cultivation effect, have until now been conducted along the lines of the first approach. We suggest the time has come for research opting for the second approach. The present study has attempted to do just that.


As of January 1, 1994, private broadcasters in Canada have voluntarily adhered to a code of ethics in television, recommending they refrain from showing gratuitous scenes of violence that condone, glorify, or encourage violence. Programs containing violent scenes are broadcast only after 9 p.m. The Canadian Radio-television and Telecommunications Commission (CRTC) has approved this code, which was adopted by the Canadian Association of Broadcasters (CAB), and expects all private stations and networks to comply. The CRTC, which has often threatened to impose such rulings on the industry, will closely follow the application of the code and will demand changes should it not see results. Both the CRTC and CAB agree on the need to establish a national program rating system to be integrated within this code.
According to this study, Canadian viewers witnessed 18.5 "conflicts" per hour, which included 9.0 physical attacks and 7.8 verbal attacks. Surprisingly, situation comedies were the most heavily laden with acts of physical violence (40% of identified segments), followed by cartoons (27.8%) and "crime" shows (27%).
Gerbner considered a "program" to be any production constituting an entity in itself, such as a film, an episode of a soap opera, or even a cartoon. This definition differed from that which is normally used. For example, a children's program may contain several cartoons and extend over several episodes; this last definition is used in the present study.
Acts of violence were not counted per se, but rather violent scenes. A violent scene began with initiation of violent behaviour and ended with its interruption or with the arrival on the scene of a new main character. If a new main character became involved in the scene, it then constituted another violent scene. Only the main characters were considered, namely, those who played a major part in the action taking place and who generally had a speaking part. Extras were excluded from the analysis. The main characters were usually humans but could also have been androids (robots, extraterrestrials) or animals (when playing a humanlike character, as is often seen in cartoons).
Only in program format was any distinction made regarding cartoons and children's programs.
This study was also made up of other variables not yet examined, such as data on psychological and verbal violence, characterization of violent individuals and their victims, acceptance of the act and its moral implications, physical consequences of the violent acts, and so forth.
Of the 21 fictional programs studied, SRC accounted for 11 animated programs.
An important step of the procedure used to assess the internal consistency of the indices used in television viewing research was to test the unidimensionality of each index through the use of factor analysis. The factor analysis of the original set of items used to build the perception of danger index showed that it was not a unidimensional index and was really made up of two different dimensions. Further analysis showed that these dimensions could reasonably be interpreted as cognition and fear, and that it was possible, with the available data, to build unidimensional indices that could measure them with some reliability. For the purpose of the regression analyses, we computed the two variables as factor scores of two different factor analyses: this procedure ensured that the variables would have the highest possible reliability, that they would retain their natural correlation even though they were computed as factor scores, and that their correlation would be disattenuated. The Cronbach's alphas are .65 for cognition and .64 for fear; their correlation is .2296 and significant at the .001 level (Laplante & Gosselin, 1996).
The rules for the empirical verification of causal relationships with non-experimental data have first been formulated some 40 years ago by Paul F. Lazarsfeld. A good introduction to the fundamentals can be found in Rosenberg (1968). Extensions to more sophisticated statistical techniques are quite well presented in Long (1983a, 1983b).
In a multiple regression equation, where some of the independent variables are categorical variables represented by a set of so-called "dummy" variables, the regression coefficients of the dummy variables are deviations. The mean scores of the subgroups are thus computed by adding the origin and the deviations from origin of each subgroups. Since the reference category for gender is "Men" and the reference category for age is "From 20 to 24," the score for men aged 20 to 24 is given by the origin. The scores for other men's groups are computed by adding the regression coefficient of the corresponding age group to the origin. The scores for women are computed from the scores for men by adding the "main" difference between men and women and the difference between men and women specific to each age group.


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

Barrile, L. (1984). Television and attitudes about crime. In R. Surette (Ed.), Justice and the media. Springfield: C. Thomas.

Blank, D. M. (1977). The Gerbner violence profile. Journal of Broadcasting, 21(3), 273-279.

Bouwman, H., & Stappers, J. (1984). The Dutch violence profile: A replication of Gerbner's message system analysis. In G. Melischek, K. E. Rosengren, & J. Stappers (Eds.), Cultural indicators: An international symposium (pp. 113-128). Vienna: Verlag der Osterreichischen Akademie der Wissenschafter.

Bryant, J., Carveth, R., & Brown, D. (1981). Television viewing and anxiety: An experimental examination. Journal of Communication, 31(1), 106-119.

Carlson, J. M. (1983). Crime show viewing by pre-adults: The impact on attitudes toward civil liberties. Communication Research, 10(4), 529-552.

Carlson, J. M. (1985). Prime time law enforcement. New York: Praeger.

Carveth, R., & Alexander, A. (1985). Soap opera viewing motivations and the cultivation process. Journal of Broadcasting and Electronic Media, 29(3), 259-273.

Comstock, G. (1989). The evolution of American television. Newbury Park, CA: Sage Publications.

Doob, A. N., & Macdonald, G. E. (1979). Television viewing and fear of victimization: Is the relationship causal? Journal of Personality and Social Psychology, 37(2), 170-179.

Gerbner, G. (1993). Violence in cable-originated television programs: A report to the national cable television association. Philadelphia: The Annenberg School for Communication, University of Pennsylvania.

Gerbner, G., & Gross, L. (1976). Living with television: The violence profile. Journal of Communication, 26(2), 173-199.

Gerbner, G., Gross, L., Eleey, M., Jackson-Beeck, M., Jeffries-Fox, S., & Signorielli, N. (1977a). TV violence profile no. 8: The highlights. Journal of Communication, 27(2), 171-180.

Gerbner, G., Gross, L., Eleey, M., Jackson-Beeck, M., Jeffries-Fox, S., & Signorielli, N. (1977b). The Gerbner violence profile: An analysis of the CBS report. Journal of Broadcasting, 21(3), 280-286.

Gerbner, G., Gross, L., Eleey, M., Jackson-Beeck, M., Jeffries-Fox, S., & Signorielli, N. (1977c). One more time: An analysis of the CBS "Final comments on the violence profile." Journal of Broadcasting, 21(3), 297-303.

Gerbner, G., Gross, L., Jackson-Beeck, M., Jeffries-Fox, S., & Signorielli, N. (1978). Cultural indicators: Violence profile no. 9. Journal of Communication, 28(3), 176-207.

Gerbner, G., Gross, L., Morgan, M., Signorielli, N., & Jackson-Beeck, M. (1979). The demonstration of power: Violence profile no. 10. Journal of Communication, 29(3), 177-196.

Gunter, B. (1987). Television and the fear of crime. London: Libbey.

Gunter, B., & Wober, J. M. (1983). Television viewing and public trust. British Journal of Social Psychology, 22, 174-176.

Gunter, B., & Wober, J. M. (1988). Violence on television: What the viewers think. London: Libbey.

Hawkins, R. P., & Pingree, S. (1980). Some processes in the cultivation effect. Communication Research, 7(2), 193-226.

Hawkins, R. P., & Pingree, S. (1990). Divergent psychological processes in constructing social reality from mass media content. In M. Morgan & N. Signorielli (Eds.), Cultivation analysis: New directions in media effects research (pp. 35-50). London: Sage Publications.

Hedinsson, E., & Windahl, S. (1984). Cultivation analysis: A Swedish illustration. In G. Melischek, K. E. Rosengren, & J. Stappers (Eds.), Cultural indicators: An international symposium (pp. 389-406). Vienna: Verlag der Osterreichischen Akademie de Wissenschafter.

Hirsch, P. M. (1980). The "scary world" of the non viewer and other anomalies: A reanalysis of Gerbner et al.'s findings of cultivation analysis. Part I. Communication Research, 7(4), 403-456.

Hirsch, P. M. (1981). On not learning from one's own mistakes: A reanalysis of Gerbner et al.'s findings on cultivation analysis. Part II. Communication Research, 8(1), 3-37.

Hughes, M. (1980). The fruits of cultivation analysis: A re-examination of the effects of television watching on fear of victimization, alienation, and the approval of violence. Public Opinion Quarterly, 44(3), 287-302.

Iwao, S., De Sola Pool, I., & Hagiwora, S. (1981). Japanese and U.S. media: Some cross-cultural insights into television violence. Journal of Communication, 31(2), 28-36.

Laplante, B., & Gosselin, A. (1996). L'écoute de la télévision et la formation de la perception de la réalité sociale. Sainte-Foy, QC: INRS-Culture et Société.

Long, J. S. (1983a). Confirmatory factor analysis. Beverly Hills, CA: Sage Publications.

Long, J. S. (1983b). Covariance structure models. Beverly Hills, CA: Sage Publications.

McCann, T. E., & Sheehan, P. W. (1985). Violent content in Australian television. Australian Psychologist, 20(1), 33-42.

Morgan, M., & Signorielli, N. (1990). Cultivation analysis: Conceptualisation and methodology. In M. Morgan & N. Signorielli (Eds.), Cultivation analysis: New directions in media effects research (pp. 13-34). London: Sage Publications.

Newcomb, H. (1978). Assessing the violence profile studies of Gerbner and Gross: A humanistic critique and suggestion. In G. C. Wilhoit (Ed.), Mass communication review yearbook (Vol. 1, pp. 451-469). Beverly Hills, CA: Sage Publications.

Ogles, R. M., & Sparks, G. G. (1989). Television violence and viewers' perceptions of criminal victimization. Mass Communication Review, 16(1), 2-11.

Ogles, R. M., & Sparks, G. G. (1994). Question specificity and perceived probability of criminal victimization. Mass Communications Review, 20(1-2), 51-61.

Pingree, S., & Hawkins, R. (1981). U.S. programs on Australian television: The cultivation effect. Journal of Communication, 31(1), 97-105.

Potter, W. J. (1991). The relationships between first and second order measures of cultivation. Human Communication Research, 18, 92-113.

Rosenberg, M. (1968). Logic of survey analysis. New York: Basic Books.

Signorielli, N. (1984). The measurement of violence in television programming: Violence indices. In J. R. Dominic & J. E. Fletcher (Eds.), Broadcasting research methods. Boston: Allyn and Bacon.

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

Williams, T. M., Zabrack, M. L., & Joy, L. A. (1982). The portrayal of aggression on North American television. Journal of Applied Social Psychology, 12(5), 360-380.

Wober, J. M. (1978). Televised violence and paranoid perception: The view from Great Britain. Public Opinion Quarterly, 42, 315-321.

Wober, J. M., & Gunter, B. (1982). Television and personal threat: Fact or artefact? A British survey. British Journal of Social Psychology, 21, 239-247.

Zillmann, D., & Wakshlag, J. (1985). Fear of victimization and the appeal of crime drama. In Dolf Zillmann & Jennings Bryant (Eds.), Selective exposure to communication (pp. 141-156). Hillsdale, NJ: Erlbaum.

  •  Announcements
    Atom logo
    RSS2 logo
    RSS1 logo
  •  Current Issue
    Atom logo
    RSS2 logo
    RSS1 logo
  •  Thesis Abstracts
    Atom logo
    RSS2 logo
    RSS1 logo

We wish to acknowledge the financial support of the Social Sciences and Humanities Research Council for their financial support through theAid to Scholarly Journals Program.