Content-Type: text/html This paper was presented at the Association for Education in Journalism and Mass Communication in San Antonio, Texas August 2005. If you have questions about this paper, please contact the author directly. If you have questions about the archives, email rakyat [ at ] eparker.org. For an explanation of the subject line, send email to [log in to unmask] with just the four words, "get help info aejmc," in the body (drop the ""). (Feb 2006) Thank you. Elliott Parker ==================================================================== Missing the Market: Character Salience in Television Program Websites James R. Angelini Debbie P. C. Goh Jason A. Rosow Tyler Dodge Wenchang Deng Na Zhou Susan Tyler Eastman Indiana University, Bloomington Correspondence: James R. Angelini Doctoral Student Dept. of Telecommunications 1229 E. Seventh Street Indiana University Bloomington, IN 47405-5501 Tel: (812) 323-2615 Fax: (812) 855-7955 E-mail: [log in to unmask] Manuscript submitted to Mass Communication and Society Division Association for Education in Journalism and Mass Communication April 1, 2005 ABSTRACT The television media promote primetime programs in ways that reflect different target markets. Analysis of the gender, ethnic, and age demographics of 1306 images of characters on the six broadcast networks' 124 program websites showed close similarities between old and new media in gender and ethnicity stereotyping but not in age stereotyping. In addition, an index of character and program salience measured the relative prominence of characters by demographic type. Missing the Market: Character Salience in Television Program Websites Over the five decades of television's history, scores of studies have documented the changing patterns of media coverage in terms of gender, ethnicity, and age . These studies generally arise from the belief that distorted or misleading portrayals of social groups may lead to negative social consequences for those groups . One of the primary purposes of such research has been to determine the magnitude of television network representations of racial minorities, women, and individuals of varying ages because relative proportions for excluded groups have been interpreted as contributing to the reinforcement of existing stereotypes . Television portrayals have been shown to have an impact on lasting beliefs about an individual's own social groups, as well as other groups' status in the larger, general societal context . This study extends investigation of television character portrayals tp the new media of network program websites to determine whether they have the potential for mitigating or exacerbating television's social stereotyping problem. Cultivation theorists argue that television has long-term effects that are small, gradual, and indirect but cumulative and significant . They emphasize the effects of television viewing on the attitudes rather than the behaviors of viewers. Heavy watching of television is seen as "cultivating" attitudes that are more consistent with the world depicted in television programs than with the everyday world . This research agenda has produced a large body of work that highlights the importance of analyzing the kinds of people appearing in programs on the grounds that the content of television affects some people's attitudes and values, and it leads to subjugation and the reproduction of historical inequalities. This rationale has also been applied in studies about promoting programs to viewers, contending that on-air promos have special importance because people who do not watch many of the promoted programs nonetheless are influenced by the promos they see in great quantities . George Gerbner and his associates relate the formation of attitudes and values about various social groups to amounts and kinds of television consumption. Although traditional cultivation theorists primarily limit the scope of their theory to "heavy" viewers, recent work by Zillmann (1999, 2002) has demonstrated the utility of expanding the horizon of media influence to include less compulsive viewing. His exemplification theory posits that extended distribution of events or images defines them as exemplars, irrespective of any distortion, and exemplars are bound to influence related concepts and behaviors . In particular, many scholars have shown that continued exposure to the televised portrayals of various social groups, particularly those groups of which the viewer is not a member, corresponds to stereotypical attitudes about gender , race/ethnicity , and age . It is well understood that television programs deal in visual and auditory stereotypes for quick and efficient communication with a large and varied audience. Analyses of primetime programming has repeatedly shown that white male characters overwhelmingly dominate television portrayals, especially of figures of authority, and that television under-represents the young, the old, and all minority groups except, in recent years, African Americans, and that most characters are visualized with easily-recognized stereotypical attributes. Furthermore, researchers have found that promotional messages were a microcosm of primetime programs that also powerfully underrepresented younger and older people, as well as women and ethnic minorities, in stereotyped roles, and thus potentially contributed to viewers' skewed understandings of groups' social positions . The focus of these previous content analyses of television portrayals, however, has been limited to programs themselves and on-air promotion of them (or advertising images) and has ignored the increasing relevance and importance of the new media, especially Internet web pages. This shortcoming is surprising given the importance of the networks' official websites dedicated to television programs. For such shows as Survivor (II, Vanuatu, Australian Outback), it has been argued that the official websites may have been as important to audiences as the programs themselves . Not only the audiences but also the networks have a considerable investment in program websites, which have cost networks millions of dollars to develop . Indeed, although program websites possess a high entertainment component, they serve as more than mere novelties: They can increase program audiences and generate significant revenue . Online promotion offers a unique opportunity for examining some fundamental differences between old and new promotional media. First, while broadcast promos for programs are primarily persuasive media, Internet web pages are largely informational media . That is, on-air promos, radio spots, billboards, and periodical advertisements are all media that aim to persuade potential viewers to watch, listen, or subscribe. In contrast, online sites must be sought out by Internet users on computers or mobile phones, and those users are typically searching for some kind of information or experiential enhancement of their television experience. Furthermore, on-air and most print media messages tend to approach the audience member, whether that person is interested or not, while online media have to be purposively accessed by the user to serve that person's needs. Still, this is not to minimize the role of media as entertainment: Both persuasive and informational promotional messages must be visually and auditorily engaging to command the audience's attention. Second, on-air and online promotional messages have differing contextual and structural attributes.[1] A broadcast promo is necessarily presented next to or within another program; it may be situated within a popular or unpopular program, and it may be surrounded by many other commercial messages or be relatively isolated, but in any case, the promo is secondary to the program and an interruption to it. In contrast, an online promotion normally consists of numerous web pages and links, constituting an extended experience, thus being the principal content rather than an interlude. Program web sites can be likened to on-air programs, while banner ads and pop-ups are the functional equivalent of on-air promos. Further, a broadcast promo appears at a network-determined time and with a fixed duration, and viewers cannot normally extend or repeat the spot (except with TiVo-like DVR controls). In contrast, web-based program sites are persistent and available anytime and are intended for active manipulation. Moreover, web sites consist of both general introductory pages (the home page for a network, the usual entry path) and program specific pages (beginning with splash pages of varying lengths that are connected to other pages and sites by click-on links), a structure not at all parallel to on-air promotion. Accessing of program websites varies by genre (greater for reality and some dramas), while on-air promos tout programs of all entertainment genres proportionate to the individual network's economic need. Still, despite the widespread use of online program sites by television audiences, little is known about their composition in terms of character portrayals and their relationship to broadcast promos. Foundational Questions In an attempt to fill this lacuna, we begin by asking general research questions to establish some baselines for studying online program promotion. RQ1: What is the proportionate distribution of the gender, ethnicity, and ages of program characters portrayed on program splash pages and network home pages? How do the distributions of characters' gender, ethnicity, and age on web pages compare with the proportions for characters in on-air promos and programs reported in other studies? At the same time, we must ask whether systematic variations between the six networks or between program genres confound interpretation of any results. Beyond Quantity Although the frequency and proportion of portrayals has been shown to impact viewer beliefs, a new theoretical lens has developed that illuminates the role of portrayals as a key factor in distinguishing viewer impact. Salience Theory contends that effectiveness in program promotion arises from maximal prominence of certain structural and content variables that collectively create an advantage that positively affects a promoted program's ability to attract an audience . Although Salience Theory has traditionally been applied to on-air promotion, by extension, it also suggests that maximally salient character portrayals in online program promotion may influence website use and, as a function of repetition and selective prominence, viewers' beliefs about demographic groups. This study extends research about the salience of promotional messages in three ways. First, it adapts the attributes of salience that have been previously demonstrated for broadcast promos to the programs and characters appearing on network promotional web sites. Second, it suggests new attributes of salience that are specific to web sites, namely selective sound and motion. Third, it introduces the use of salience indexes for characters and web pages as a method of summarizing relative prominence so that producers' treatment of characters and programs can be evaluated. Because applying salience theory to program web pages requires adapting it to the unique affordances of the online context, two research questions address the relationships between the portrayals of social groups. RQ2: As measured by an index of image salience elements, on average, how much salience is allotted to women and minority characters, in relation to the averages for the other gender and ethnic groups portrayed on program splash and network home pages? It would be valuable to know whether any particular elements contribute significantly to the amount of image salience for particular gender, ethnic, or age groups. In other words, does prominence for white male characters or young adults come from particular web page elements? Another aspect of the relationship is whether a program's salience index is closely related to each show's popularity. RQ3 Is there a significant correlation between the salience indexes for specific programs on their web splash pages and the program's current rating? Proportion, Salience, and Portrayals The large body of research about character portrayals in broadcast programs and promotion, when combined with recent literature regarding Internet websites, points to some predictions about what might be expected regarding the proportion and salience of portrayals on Internet web pages. Both a study of television programs and a study of television promos concluded that television continues to overwhelm viewers with images of white men while placing women and minority characters in supporting roles. Clearly, in program and promos analyses, men are overrepresented in the population of television characters while women are underrepresented . Although African-Americans are now generally represented on television in proportion to their U.S. Census numbers, Hispanics, Asians, and Native Americans remain markedly unrepresented . Thus, we safely predict two outcomes: H1: On program web pages, the proportions of male and white characters will be higher than female and other ethnic group characters. Although it has been argued that characters in the age group 18 to 49 are overrepresented in programs and promos, while all other age demographics are underrepresented, the Internet may be different . Analyses of contemporary Internet users consistently conclude that teens and young adults are more likely to use the Internet than other age groups , despite recent rises in use by older adults . Moreover, most television networks strive to attract younger not older adult viewers to please advertisers, and Internet use is greatest among the younger age group. Thus, given that a chief goal of promotion is to generate connections and identification with the characters by users , we predict that H2: The proportion of teens and young adult characters promoted in web pages will be significantly larger than the proportion of any other age group. Looking at broadcasters' websites, researchers have sought to identify the changes and continuities in notions of the audience from on-air to online, concluding that the television networks have merely expanded from on-air to online but not really changed their conceptions of audiences as target groups . Accordingly, in comparisons with Eastman and Walker (2005), we should not expect big differences between new and old media. H3: The proportions for gender and ethnic in web pages will not differ significantly from their overall proportions in on-air promos, while the proportions of teens and young adults will be larger on web pages than in on-air promos. Nonetheless, television programming has witnessed an emergence of network specific variation in target audiences even at the broadcast level. ABC, CBS, and NBC generally target the largest possible audiences, and FOX targets teens and younger adults, while UPN and the WB predominantly seek either male or female viewers, and UPN has proportionately more programs with African-American roles. Thus, we can confidently expect that H4: The gender, ethnicity, and age of characters depicted on web pages will differ significantly by network, reflecting programming for different target audiences. Furthermore, contemporary television programming includes only limited number of television genres within primetime, consisting of comedies, dramas, reality, sports, and news. Because notable differences by genre have emerged in previous studies , particular genres may be associated with particular types of character portrayals. Specifically, such reality programs as Survivor, Fear Factor, and Amazing Race, may be associated with different age groups than situation comedies or dramas. Given the diversity found in programs of specific genres, we predict that H5: The gender, ethnicity, and age of characters depicted on web pages will differ significantly by genre. Going beyond demographic quantities, we can look at the prominence of characters and programs. Although the salience index is a new and untested tool for measuring differences in web page promotion, it draws on a sizable body of measurement of salience within on-air promotion. Given the prevalence of white males both as television characters and Internet users , it seems logical to presume that not only white males�especially young ones�will be promoted more on websites but that the kind of promotion they receive will be more salient than that given to other groups. Therefore, we predict that H6: The average salience index for young white males will be higher than the average salience index for any other gender or ethnic group. Method This study analyzed all network homepages and splash pages of primetime-program websites on the six major English-language television networks--ABC, CBS, FOX, NBC, UPN and WB. The pages were captured on a single day to avoid weekly site changes by webmasters. A single day (November 12, 2004) was purposively selected from the first half of November because (1) most new as well as continuing programs would have already premiered, (2) that period avoids holidays, and (3) during a November sweeps period, the networks would be exhibiting their best efforts at promoting their fall schedules. Despite weekly changes in plot summaries and plot-related photographs on websites, it is likely that the site content on the sample day typifies a much longer period of time for most programs. The primary unit of analysis was the primetime character, including human and human-like characters that were portrayed as capable of speech or social interaction. Instrument For each splash page, coders noted the (1) program's title, (2) network (ABC, CBS, FOX, NBC, UPN, WB) and (3) genre (comedy, drama, reality, news and other).[2] To measure image characteristics, coders recorded the following for each character: (4) name, (5) gender (male, female, unknown for ambiguous gender in cartoon characters and aliens), (6) ethnicity (White, Black, Hispanic, Asian, Native American, Other for aliens and cartoons, and Unknown), and (7) age (child 0-12, teen 13-19, college/young adult 20-34, settled adult 35-64, grandparent/elderly 65+, and unknown for ambiguous aliens and cartoon characters).[3] Studies have shown that the frequency of repetition of a message usually contributes to the impact of promotional messages . It follows that repetition of images of specific characters on web pages would strengthen their impact relative to unrepeated character images. Thus, (8) every picture of a single character was enumerated for images larger than a thumbprint (about 100 x 150 pixels), both in the main parts of the page and any promotional ads (such as DVD sales) for the program providing there were recognizable character images related to the main picture. Because photographs on websites were often small and difficult to judge, coders made use of actor bios and cast descriptions to verify names and intended ages of the characters. Teens, for example, are often played by older actors but makeup, script, acting style, and allotted position in family hierarchies define their supposed ages, giving coding of the actor as "teen" greater ecological or external validity because it would likely resonate with viewers' experience of the program and thus the website.) Because the pilot coding showed that advertising messages for non-program-related products had no recognizable faces in them, such ads were excluded. Each image on a home page was later added to the totals for the appropriate character. Salience of a character image online was measured using an eight-factor composite index drawn from Eastman's index of factors impacting the effectiveness of on-air promotion . The eight factors relevant to online salience were (9) picture size (large, medium, small), (10) framing of image (portrait, a mid-to-full body shot, or a long shot), (11) shot composition (single, pair, or multiple characters), (12) picture position on the web page (top, middle, bottom), (13) picture motion (video, slide show, special effects), (14) character movement (high, medium, low), (15) automatic voice audio (if recognizable), and (16) other visual elements contributing to dominance (extreme close-ups, foreground placement, larger relative size, dramatic color/value contrast). In subsequent data entry, factors 9 through 14 were assigned three points for the most salient condition (large, portrait, single, top, video, high, voice sound), two for the mid condition, and one point for the least salient condition (small, long shot, multiple, bottom, special effects, low). One point was awarded for any visual dominance elements; no points were given for the absence of picture motion, character movement, or audio.[4] Totaling these points for each individual character provided an index of salience above and beyond the number of pictures and distinct from the character's gender, ethnicity, and age. Totaling the points by program for all characters provided an index of the salience the program had been given by the web page producers. Totaling all indexes for a network or a genre provided comparisons by network or program genre. Coding Process For a pilot test, each of the seven researchers coded three program splash pages from one of the six networks. The researchers then discussed the results as a group and fine-tuned definitions and instructions to prepare a code book. On the actual coding day, six researchers coded all identifiable primetime character images appearing on the splash pages and homepage of one network. A seventh researcher randomly selected 24 program splash pages from the six networks, stratified as 4 programs per network, for intercoder reliability tests. In addition, all 6 network home pages were recoded by a single researcher to assess reliability. Altogether, reanalysis was conducted of 24% (29 of 122) of the page database and 20% (161 of 826) of the total characters. Reliability was assessed using simple agreement and Cohen's kappa among paired coders.[5] Overall, intercoder reliability was 98.8%, and agreement was 100% for gender, 100% for ethnicity, and 98% for age. Cohen's kappa for number of pictures was 100% and for character salience was 96%. Results Every program carried in primetime in mid-November 2004 or soon-to-premiere had a web page. Analysis of the 122 program pages on the six networks' websites generated 1319 images (pictures of actors) for 826 characters. ABC's 23 program web pages (22 splash and 1 home) contained pictures promoting 145 characters; CBS's 23 pages promoted 136 characters; FOX's 23 pages promoted 185 characters; NBC's 23 pages promoted 169 characters; UPN's 13 pages promoted 67 characters; and the WB's 15 pages promoted 124 characters.[6] Thus, only UPN, and to a lesser extent, the WB, are underrepresented in this sample. Demographic Differences Hypothesis 1 predicted that men and white characters would be the largest gender and ethnic groups. As expected, of the 826 portrayed characters, 56% were male and 44% were female (which compares to 49% men and 51% women in the U.S. population). Ethnic distribution was overwhelmingly white (81%), followed by 14% African American, 2% Hispanic, 2% Asian, and 1% uncodable cartoons or aliens (see Table 1). Thus, both outcomes predicted in the first hypothesis were supported. Table 1 about here Hypothesis 2 dealt separately with age and predicted that teen and young adults combined would be the dominant age group. In contrast to our expectations, the ages of portrayed characters skewed toward the older age groups. Table 1 shows that about 5% of characters were children, 11% teens, 27% young adults, 53% settled adults, and 5% elderly (fewer than 1% were uncodable). The combined 37% for teens and young adults is far outweighed by 53% for settled adults. More than half the characters on the web pages were beyond youth by any perspective (as is true of the U.S. population), and thus, the second hypothesis--claiming that websites would skew toward youth because users do--was not supported. Hypothesis 3 presumed that the portrayals of characters in the new media and the old media would match on such attributes as gender and ethnicity but not on age, where teen and young adult age groups would be larger on web pages. In addition to the online findings in the first column of Table 1 already discussed, and the U.S. Census equivalents in the third column, the second column lists the findings for on-air promos from Eastman and Walker's recent study (2004). The table shows that while the proportions by gender in this study matched on-air promo closely, the proportions by ethnicity actually match the census more closely than they do the findings for on-air promos. As Table 1 shows, although the most recent U.S. Census shows a 49:51 gender split, Eastman and Walker (2005) found a 60:40 ratio in on-air promos for primetime programs, comparable to the 56:44 ratio in this study. For ethnicity, Table 1 shows that apart from UPN, which has a 40:60 white and other ethnic group ratio, the overall proportions for online (81:19) are not significantly different from the proportion in on-air promos (88:12). Both are also clearly different from U.S. Census figures. Moreover, the percentages of young adults and settled adults in online promotion differ significantly from their proportions in on-air promotion in Table 1, although the skew is, as predicted, toward younger people. Only the proportion of young adults matches the census, the rest differing markedly, and the proportions in all groups (except elderly) differ significantly from findings for on-air promos. Thus, characters promoted on websites differ both from the census and from those promoted in on-air promos, except that the bias in promos toward male characters appears also in website portrayals. Although some support appears for the second half of the third hypothesis�regarding favoritism for youth on web sites, when the three groups (teen plus young and settled adults) are collapsed, the finding is no longer significant, suggesting that the differences may be artifacts of measurement. In this web study, adults from 20 to 34 years were considered "young adults," while the on-air study broke the categories apart at age 25. This means that a huge proportion of television characters appear to be in their late 20s or early 30s, although only about a quarter of the general U.S. population falls in this age range. In sum, only in gender proportions and emphasis on white characters�but not in minority or age categorization--was the third hypothesis supported. Hypothesis 4 raised the issue of differences by network and predicted significant differences between the networks because of their different programming targets. Table 2 shows this study's findings for gender, ethnicity, and age by network. In gender, ABC exhibited a 50/50 split while FOX, NBC, and UPN circled closely around the 60:40 overall ratio. Only UPN appeared a distinct outlier with its reversed proportion of more women than men (42:58). In ethnicity, both CBS and WB were somewhat higher in proportion of white characters (92% and 89%) and correspondingly lower in minority classifications, while UPN was significantly lower in proportion of whites (40%) and dramatically higher in ethnic groups (60%), again reflecting differences in target audiences. However, because UPN represents only a small portion of the character database (8%), it seemed reasonable to treat the networks as a single unit in some subsequent analyses. Significant age differences for characters among the networks also appeared. As Table 2 shows, WB had a higher proportion of teen characters (27%) than the overall average of 11%. ABC had a significantly lower proportion of young adult characters (1%) by comparison to 27% overall, while FOX and UPN had significantly higher proportions (55% and 42%) than the 27% overall. Also, ABC, CBS, and NBC all had significantly higher proportions of settled adults (72%, 63%, and 66%) by comparison to 53% overall, while FOX and UPN had significantly lower proportions (28% and 40%). These findings support the claim for network differences in Hypothesis 4, but the differences are most dramatic for age and far less so for gender and ethnicity. Table 2 about here Hypothesis 5 looked at the role of genre and predicted that differences between genres for gender, ethnicity, and age would be significant and thus could potentially confound interpretation. Table 3 shows that the distribution of gender, ethnicity, and age for four genres. In gender, all the genres except reality shows hovered around the 60% male to 40% female ratio. Reality shows were more nearly equal in their male-female distribution (48% male, 52% female). In ethnicity, all genres except comedy had about 82% white characters in their web promotion, with comedy somewhat lower (75%) in proportion of white characters. Subsequent selective analysis by the younger age groups showed that children and teens associated most closely with the comedy genre (26%), while young adults associate mostly with the reality genre (48%). Settled adults figured most prominently in news (86%), and then in dramas (60%). Thus, the fifth hypothesis predicting differences by genre was clearly supported for age groups, but for gender and ethnicity, relatively small differences by genre emerged. Table 3 about here Salience for Characters The final hypothesis turns to the index of salience, as previously described--an index collapsing internal and external image variables to quantify the degree of prominence each character and each program received. Hypothesis 6 predicted that males who are young and white would be given more prominence on pages by webmasters than any other combination of gender, ethnicity, or ages. However, overall analysis defeated that hypothesis. It appeared that the huge quantity of male characters obliterated any distinction between those in the foreground and those in the background. In consequence, the researchers turned to a tertiary split of the index data created low, mid, and high groups of characters according to their salience to look for patterns of character receiving salient treatment by site producers. Table 4 shows the results of that tertiary analysis, revealing that only age contributed significantly to differences in salience (p=.001). Children and elderly were more likely to have low salience. Young adults were more likely to have mid salience, while teens were more likely to have high salience. Examination of Table 4 also shows that Hispanics were more likely to have been placed in the background and less frequently in the foreground than other ethnic groups. Thus, while the results do not support our hypothesis that more prominence will be given to young males specifically, it does add strength to our earlier argument that attention will be paid to attracting younger viewers. Table 4 about here Individual and program salience were also investigated. Notably, just three characters had very large salience scores (well over 80 points): Dr. Gregory House in FOX's drama House (110 points); Lea Masters in CBS's Survivor: Vanuatu (111 points), and Sydney Bristow in ABC's Alias (185 points). In two cases, House and Alias, the shows had yet premiered at the time of coding. The exceptionally high salience for the lead characters on the program splash pages may reflect the networks' attempts to establish them firmly in viewers' minds. High salience for the Survivor character Lea is an artifact of his being voted off the series in the previous week's episode (the last person to depart is always highly promoted on the program's splash pages). A comparison of the mean differences in salience scores of the 122[7] characters who appeared on both home and splash pages showed that they scored significantly higher on splash pages (_=17.02, t=4.65, p<.001) than on home pages (_=11.91). Conclusions This study's finding that salience for most characters--except teens�did not differ significantly from on-air promos indicates that the network web producers have failed to harness the features of the Internet to the fullest, despite their belief that it will serve as an invaluable promotion tool. Network program websites have essentially "missed the market." This can be likened to the early days of introducing new information and communication technologies, when development experts extolled the merits of each new technology and sought to implement it without due consideration on how to best engage it appropriately. The Canadian Development International Agency today reminds those considering harnessing new communication technologies that while these technologies are "one important driver of economic growth, they should not be considered a panacea." One needs to assess the characteristics of a technology and use it effectively. In the business arena, researchers have argued that content providers prosper only when they pay heed to what information and service users seek in the online environment . Given the importance of the networks' official websites in increasing program audiences and generating significant revenue, it is thus critical for networks web managers to reconsider the blueprints for their sites. On a broader front, this study asked about the general distribution of gender, ethnicity, and age groups on web pages, differences between online and on-air promotion of programs, and about differences between types of web pages, network targets, and program genres as potentially confounding variables. The results of this analysis lead to four main conclusions. First, web sites numerically favor male characters, white characters, and settled-adult characters. It is clear that web pages give somewhat less attention to women characters, considerably less attention to minority characters, and practically no attention to the children and the elderly, little attention to teens and somewhat more to young adults, but favoring the settled-adult group over all others. In each case, the favored group was given more prominence in foreground placement on web pages as well as middle and background placement. By comparison to Second, web promotion of programs closely resembles over-the-air promotion of network programs. Surprisingly few differences between online and on-air promotion emerged, the most salient being the age distinction between young and settled adults, possibly a measurement artifact. While the Eastman and Walker study showed that on-air promos favored younger adults, the web pages analyzed here favored the settled-adult group. It is likely, however, that our age classification of characters more accurately reflects how network producers intend audience to see these characters (and perhaps how viewers see them), because this study draws verification from the cast and bio descriptions. Thus, this study corrects rather than challenges previous work in age proportions by Eastman and Walker (2005). Altogether, this study's findings support our prediction that the gender and ethnic proportions of web page characters match promo characters, they clearly challenge our prediction that web pages ought to emphasize younger characters in greater quantity. Third, possible confounding variables tended to support rather than contradict the main findings. Subset comparison of splash pages to home pages showed significant differences in these variables overall, as characters received lower mean salience scores on home pages than splash pages. This is expected as space on network homepages is distributed across all programs, as opposed to splash pages, which are dedicated solely to the respective programs and their characters. Network differences in web pages emerged but predictably, following the trend of the four major networks seeking broad mid-aged audiences and the smaller networks seeing younger subset audiences and ethnically oriented viewers. Genre differences appeared but tended to follow the general patterns with a few exceptions: Reality program had more gender balance, while comedy had somewhat more ethnic and younger representation. But overall, the analyses by the six networks and four main genres produced patterns consistent with the male, white, and mid-aged web pages. Fourth, the salience measurements produced strong differentiations that paralleled most of the previous findings�with one big difference. Males received by far the most prominence in size of pictures, framing of images, shot composition, picture position, character actions, video motion, and recognizable voice audio. Children and the elderly and Hispanics were allotted the least salience by comparison to other characters. The one notable difference from the other findings in this analysis and the Eastman and Walker analysis of on-air promos was that, of all age groups, teen were clearly given the most prominence. This is the sole finding that supports our expectation that web sites would try to attract and hold the interest of the age group that sees computers as essential tools for living, the group that has never lived without them and are thus most attune to using them for information and entertainment. The fundamental assumption behind studies of the demographic composition of characters in television programs is that, for the social good, programs ought to provide equitable, representative portrayals over the long term for both women and men, for people of varied ages and ethnic backgrounds. Over time and across many programs and networks, television ought to reflect the composition of the United States�or at least the best of America. By avoiding distortions from under-representation as well as over-representation of men, whites, and young adults, television could potentially lessen the negative economic, employment, and social consequences of narrow stereotyping of members of minority groups. Web page media include a large amount of pre-existing content, including standard portraits of the cast, official graphics relating to show logos and typography, and still images taken from the shows themselves. Accordingly, the correlations between structural features suggest that the stock content being put online is designed and utilized consistently. That is, guidelines like network style sheets and production techniques inform the development of stock content. Still, in traditional media, content is leveraged to promote character salience reflecting conscious or unconscious framing biases, but this is not the case with television program web pages: the lack of correlation between structural features and character demographics suggests that content is not being leveraged for character presentation. Given the nascent state of online promotion in comparison to such well-understood media as magazine and televised promotion, we may reason that the program web sites are being built by designers who are implementing extant content online but without making the effective marketing decisions associated with designers of traditional promotion. The causes may be several: The designers may not understand the theoretical and practical foundations of promotion, audience response, and perceived salience; or they may be following the dictates of style specifications toward standardizing pages that need frequent updating, thus curtailing their creativity. In any case, the result is a network website that publishes high-quality media related to television programs but nonetheless fails to leverage the media in ways fundamental to promotion, including establishing character salience in the minds of viewers. In short, the networks are missing the market. References Table 1. Demography of Webpage Characters in Percents Online Characters On-air Characters (Eastman & Walker, 2005) U.S Census** (n=826) (n=5241)* Gender Character Male 55.7 59.5 49.1 Female 44.3 40.5 50.9 (?2, N = 826) = .00, p < .001 Ethnicity Character White 80.6 88.0 75.1 Black 14.0 8.6 12.3 Hispanic 2.3 -- 12.5 Asian 1.7 -- 3.6 Alien/Toon/Other 1.7 -- 0.9 (? 2, N = 827) = .00, p < .001 Age Character Child 4.8 2.2 Teen 10.9 5.8 25.7 = <18 Young Adult 26.5 9.1 23.8 Settled Adult 53.0 76.4 38.0 Elderly 4.5 6.7 12.4 Others 0.2 (? 2, N = 827) = .00, p < .001 *Primetime only; some categories collapsed. *Derived from U.S. Census population estimates as of 2000. Table 2. Demographic Percentages for Online Pages by Network ABC CBS FOX NBC UPN WB (n=145) (n=136) (n=185) (n=169) (n=67) (n=124) Gender of Character Male 50.3 58.1 60.0 58.6 41.8 55.7 Female 49.7 41.9 40.0 41.4 58.2 44.3 (x2, N=826)=.100, Cramer's V=.100 Ethnicity of Character White 83.4 91.9 79.5 78.7 40.3 88.7 Others 16.6 8.1 20.5 21.3 59.7 11.3 (x2, N=826)=.001, Cramer's V=.001 Age Character Child 9.0 2.9 6.5 1.8 1.5 5.6 Teen 14.5 5.9 7.0 2.4 14.9 27.4 Young Adult 1.4 16.9 54.6 28.4 41.8 13.7 Settled 72.4 62.5 28.1 66.3 40.3 46.0 Elderly 2.8 11.8 3.8 1.2 1.5 6.8 (x2, N=826=0.001, Cramer' V=.001 Table 3. Demographic Percentages for Online Promotions by Genre Comedy Drama Reality News (n=245) (n=296) (n=250) (n=35) Gender Character Male 57.6 59.8 48.4 60.0 Female 42.4 40.2 51.6 40.0 (x2, N=826)=.046, Cramer's V=.046 Ethnicity Character White 75.5 82.4 82.0 82.9 Others 24.5 17.6 18.0 17.1 (x2, N=826)=.172, Cramer's V=.172 Age Character Child 9.4 2.7 3.6 0 Teen 16.3 13.9 3.6 0 Young Adult 15.9 19.6 48.4 2.9 Settled 50.6 60.1 42.4 85.7 Elderly 6.9 3.7 2.0 11.4 (x2, N=826)=.001, Cramer's V=.001 Missing the Market Missing the Market 4 4 Table 4. Salience by Gender, Race, and Age for Online Characters Low Salience* Mid Salience High Salience Percentage of total (n=302) (n=250) (n=274) images Gender Character Male 37.4 29.3 33.3 55.7 Female 35.5 31.4 33.2 44.3 Average Salience (?2, N = 826) = .784, Cramer's V = .784 (?2, N = 826) = .652, Cramer's V = .652 Most Salient (?2, N = 824) = .00, p < .001 Ethnicity Character White 36.8 29.0 34.2 80.3 Black 32.8 37.9 29.3 14.0 Hispanic 47.4 31.6 21.1 2.3 Asian 28.6 35.7 35.7 1.7 Alien/Toon 50.0 16.7 33.3 1.7 (?2, N = 826) = .623, Cramer's V = .623 (?2, N = 826) = 1.00, Cramer's V = 1.00 Age Character Child 52.5 30.0 17.5 4.8 Teen 29.4 21.2 49.4 10.9 Young Adult 29.0 42.4 28.6 26.5 Settled 38.6 28.7 32.6 53.0 Elderly 56.8 24.3 18.9 4.5 (?2, N = 826) = .001, Cramer's V = .001 (?2, N = 826) = .302, Cramer's V = .302 * Based on a tertiary split of the data, Low = 0 to 9, Mid = 10 to 16, High (Most Salient) = 17-186 Missing the Market Missing the Market 4 4 Notes [1] Print advertising and billboards for programs more closely resemble on-air promotion than they do online program sites; this is often called the push/pull distinction: Print, billboards, and on-air promos are pushed out at viewers, whereas online sites pull in viewers. [2] "Sport" was originally included as a genre, but dropped when only UPN's WWE Smackdown had a network website among the primetime sites. WWE Smackdown was classified as reality programming for the purposes of this study. Clicking on Monday Night Football referred the user to ESPN's NFL site. [3] The age category of 35-64 is called "Settled" rather than merely "middle-aged" to capture not only chronological age but social lifestyle (as previously used by Signorelli, 2004). [4] . Interactive elements of pages (navigational devices, mouse-over changes, click-on spots) were not coded because they require further action on the part of site users and are thus readily avoidable and do not contribute to the immediate salience of a character portrayal. [5] Following Neuendorf (2002), tests accounting for chance were judged inappropriate for two-coder measurements of such obvious physical characteristics as gender and race/ethnicity, and for age where many cases were revised based on group discussion, but they were deemed appropriate for enumerations and for salience characteristics because multiple options had about equal chance of being selected. [6] ABC, CBS, and NBC supply programs for 22 hours of primetime during the seven days of each week, while UPN and the WB program only14 hours, accounting for the far fewer programs and thus characters on the UPN and WB web sites. Although FOX supplies programs for just 15 hours a week, they are almost exclusively half-hour shows (rather than hour-long programs or movies), thus substantially increasing the number of programs for which FOX has web sites and the concomitant number of characters. [7] Dr. Gregory House's was excluded from this analysis to prevent his single high salience score in this dataset from skewing the data.