‘n - LIBRA D v Michigan 5:: .e University This is to certify that the thesis entitled PROGRAMMIM} STRATEGIE AND TELEVISION °ROGRAM POPULARITY FOR CHIIDREN presented by Jacob J. Wakshlag has been accepted toWards fulfillment of the requirements for Ph. D. degree in comunication «04le L 1 Major profegr DateNO’U. 17,1191? 0-7639 PROGRAMMING STRATEGIES AND TELEVISION PROGRAM POPULARITY FOR CHILDREN BY Jacob Joseph Wakshlag A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Communication 1977 ABSTRACT PROGRAMMING STRATEGIES AND TELEVISION PROGRAM POPULARITY FOR CHILDREN BY Jacob J. Wakshlag This research addressed children's television viewing by focusing on stability and instability in television program popularity. 0f central concern was the relationship between programming strategies and the popularity of Prime Time and Saturday morning programs for children. TV viewing data were collected from approximately 300 fourth, sixth and eighth graders in the Fall, Winter and Spring of the '75—'76 television season. These data were converted to provide popularity scores for programs. Program popularity was the dependent variable in this study. Estimates of reliability and stability for program popular- ity were generated. Several independent variables were assessed regarding their ability to affect program popularity. These variables fell into three categories: audience characteristics, programming strategies and program attributes. Multiple regression and Analysis of Variance techniques were used to assess the relationship between program popularity and these vari— ables. Results of the analyses as well as post hgg_ana1yses indicated a problem with the typology employed to describe programs. Some variables, however, were found to be significant predictors of program popularity. These variables were: sex of lead character (matching sex Jacob J. Wakshlag of respondent), time of broadcast (Prime Time only), and in the Fall, whether a program was returning or new. Counterprogramming did not emerge as a viable programming strategy. Block programming received partial support. Inheritance effects did not emerge. Saturday morning program popularity was negatively related to age of respondent. Only Sports emerged as a program type which was significantly more popular among one sex (boys) than another. Reliability of the program popularity measure was quite high. Stability in program popularity was found to increase over time, more so for younger respondents. This suggested that program popularity is more stable for these viewers and that it takes longer for them to acclimate to the television season. A discussion of program types as they relate to program preferences and actual viewing suggested that other factors must be considered before types emerge which have predictive value for viewing. These other factors include viewer characteristics, programming strategies, program attributes and their interaction. ACKNOWLEDGEMENTS The author gratefully acknowledges the assistance of Dr. Bradley S. Greenberg, Chairman of the Advisory Committee for his advice, encouragement, understanding and patience during the past few years. Thanks are also due to the remaining members of the Advisory Committee, Dr. Charles Atkin, Dr. Thomas Baldwin and Dr. R. Vincent Farace for their invaluable insights and encouragement. The author is also grateful to his wife, Paula and son, Joey for bearing with him during difficult times. This project was funded in part by grant 90-C-635 to Michigan State University from the Office of Child Development, Department of Health, Education and Welfare. Project coders were: Cathy Muckler, Gary McGee and Mike Zgraggen. Their assistance is gratefully acknowledged. ii TABLE OF CONTENTS LIST OF TABLES . . . . . . . . . . . . . . CHAPTER I. II. III. RATIONALE AND REVIEW OF LITERATURE Rationale . . . . . . . . . . . . Review of Literature . . . . . . . The Search for Program Types Children's Program Preferences Reasons for Program Selection Predicting Program Popularity Summary . . . . . . . . . . . Hypotheses . . . . . . . . . . . . METHODOLOGY . . . . . . . . . . . Variables . . . . . . . . . . . . Procedures . . . . . . . . . Selection of Programs . Respondents . . . . . . . . . Data Collection . . . . Validity and Reliability of Program Popularity Measure . . . . . . . . . Statistical Analysi . . RESULTS . . Program Popularity . . . . . . . . Prime Time . . . . . . . . . Saturday Morning . . . . . . Nielsen Ratings . . . . . . . Stability and Reliability . . Programming Strategies . . . . . Counterprogramming . . . . . Inheritance Effect . . . . . Block Programming . . . . . . Sex of Lead Character . . . . Start Time . . . . . . . . Spinoffs . . . . . . . . . Returning Programs . . . . . Programming Strategies Summary Popularity of Program Types by Sex and Respondent . . . . . . . . . iii 43 43 49 49 50 50 52 . 53 . 58 58 58 59 61 62 64 74 75 79 83 85 87 88 9O . 90 Program Type Popularity by Sex of Respondent . . Program Type Popularity by Age of Respondent . . Changes Over Time . . . . . . . . . . . . . . . . . . Post Hgg_Ana1yses . . . . . . . . . . . . . . . . . . Stepwise Regression . . . . . . . . . . . . . . Factor Analyses . . . . . . . . . . . . . . . . IV. DISCUSSION, CONCLUSIONS, LIMITATIONS AND IMPLICATIONS . Discussion . . . . . . . . . . . . . . . . . . . . . The Dependent Variable--Program Popularity . . . Programming Strategies . . . . . . . . . . . . . Audience Descriptors . . . . . . . . . . . . . Predicting Program Popularity . . . . . . . . . General Conclusions . . . . . . . . . . . . . . . . . Limitations . . . . . . . . . . . . . . . . . . . . . Implications . . . . . . . . . . . . . . . . . . . . REFERENCES 0 O O O O O O I O O O I O O O O O I O O O O O O O 0 APPENDIX A: NIELSEN PROGRAM TYPES . . . . . . . . . . . . . APPENDIX B: EXPOSURE MEASURE EXAMPLE . . . . . . . . . . . . iv . 91 . 92 . 93 . 94 . 94 . 97 . 99 . 99 104 111 114 119 120 123 128 132 133 Table 10. 11. 12. 13. 14. 15. LIST OF TABLES Summary of Research on Program Types . . . . . . . . . Prime Time Shows by Type . . . . . . . . . . . . . . . Saturday Morning Programs by Type . . . . . . . . . . Mean Popularity of Prime Time Shows by Time of Survey, Grade and Sex of Respondent . . . . . . . . . . . . . Proportion of Respondents Reporting Having Viewed Non broadcast Prime Time Shows . . . . . . . . . . . . Mean Popularity of Saturday Morning Shows by Time of Survey, Grade and Sex of Respondent . . . . . . . . . Proportion of Respondents Reporting Having Viewed Nonbroadcast Saturday Mbrning Shows . . . . . . . . . Correlation Matrix for National Nielsen Ratings by Age and Program Popularity by Grade . . . . . . . Estimated Reliability and Temporal Stability of Program Popularity by Grade . . . . . . . . . . . . Correlations for Predictor Variables with Prime Time Program Popularity by Season, and Grade and Sex of Respondent . . . . . . . . . . . . . Multiple Regression Analyses for Predicting Prime Time Program Popularity . . . . . . . . . . . . . Multiple Regression Analyses for Predicting Prime Time Program Popularity by Season and Grade . . . . . Multiple Regression Analyses for Predicting Prime Time Program Popularity by Season and Sex of Respondent Correlations for Predictor Variables with Saturday Morning Program Popularity by Season, and Grade and Sex of Respondent . . . . . . . . . . . . . . . Multiple Regression Analyses for Predicting Saturday Morning Program Popularity . . . . . . . 21 44 47 S9 60 . 6O 61 62 63 . 66 67 68 69 70 71 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. Multiple Regression Analysis for Predicting Saturday Mbrning Program Popularity by Season and Grade . . . . . . . . . . . . . . . . . . Multiple Regression Analysis for Predicting Saturday Morning Program Popularity by Season and Sex of Respondent . . . . . . . . . . . . . . . . . . Programming Strategies Summary . . . . . . . . . . Program Type Popularity by Sex . . . . . . . . . . . Program Type Popularity by Grade . . . . . . . . . . Differences in the Popularity of Returning vs. New Programs Over Time . . . . . . . . . . . . . . Stepwise Regression Analysis for Predicting Prime Time Program Popularity . . . . . . . . . . . . . . . Stepwise Regression Analyses for Predicting Saturday Morning Program Popularity . . . . . . . . . . . . . Prime Time Program Groupings Obtained via Factor Analysis . . . . . . . . . . . . . . . . . . . Saturday Morning Program Groupings Obtained via Factor Analysis . . . . . . . . . . . . . . . . . . vi 72 73 90 91 92 94 95 96 98 98 CHAPTER“; RATIONALE AND REVIEW OF LITERATURE Rationale This research addresses children's viewing patterns by focusing on stability and instability in the popularity of television programs. Of central concern is the relationship between network programming strat- egies and the popularity of programs for children. Anecdotal information concerning the time and effort expended by network executives indicated a great deal of concern regarding the renewal of programs and other scheduling strategies. For example: A series that has been high on the popularity scale for many years may be showing clear signs of attrition, indicating it may flop if renewed one more season. Conversely, careful study of rating histories may reveal that certain program series which performed indifferently during the season had the potential of becoming hits if placed on a different even- ing, or at a different hour (Brown, 1971, p. 51). Several media personnel have written descriptive accounts of the development and utilization of programming strategies. According to Shanks (1976): With increasing sophistication they realized that individual shows, though fully sponsored, could pull down the shows on either side of them or be incompatible with these shows, not only in gross numbers but in audience differences. Thus, were born program flow, block booking and counterprogramming (p. 100). Discussing why shows get cancelled Doan (1970) claims: After a show gets on the air, there is one major factor other than its inherent appeal which influences its fate. The programs night of the week and hour, and the competition it faces in that time period, help make it or break it (p. 124). Similarly, broadcasting textbooks emphasize the importance of strategic 2 program placement. For example, Kingson, Cowgill and Levy (1955) con- clude that "A given time period cannot be programmed without reference to its relationship to others on the schedule (p. 177)." Research has yet to address these issues especially for children. No evidence exists to date which suggests whether these strategies influence program popularity for these viewers. Research is needed which examines the effects of these strategies on differing populations so as to determine if they are effective across age groups or are differentially effective for differing age groups. An analysis of program popularity would be incomplete were it only to investigate changes in popularity. Equally important is the analysis of stability, the lack of change in program popularity. A stable pattern would indicate regular exposure to programming, characters and behaviors. Researchers concerned with the effects of mediated content usually assess viewing patterns at a single point in time. In order to be valid such data must reflect general enduring exposure patterns among those surveyed. If it does not, valid estimates of the effects of televised content upon viewers will not be obtained. Recent research reported by Greenberg and Atkin (1976) followed a more cautious approach. The popularity of programs for a sample of children was assessed at three points in time in order to identify programs to which children were highly and consistently exposed. Given the absence of information concerning the stability of program popularity such procedures are probably necessary. Information on stability would allow the research to determine whether such cautious procedures are justified. Research has yet to assess the stability of program popularity over time. Theories of cognitive development (Piaget, 1952) would 3 justify differential stability by age (or stage of development). Given the flurry of new programs and the rescheduling of old programs in the Fall, program popularity estimates are likely to be less stable than at other times of the television season. Consequently, research which uses early season exposure data as the index of media exposure would most likely underestimate the relationship between televised content and a respondent's cognitions, attitudes and behaviors. The industry itself employs continuous monitoring of program popularity. Professional rating services such as Nielsen and Arbitron are paid substantial sums to provide continuous monitoring of the American television diet. An analysis of the stability of television program popularity would pro— vide information which could be used to justify (or fail to justify) these large expenditures by the broadcasting industry. Review‘gf_Literature The major programming strategies referred to briefly in the pre- vious section have been labeled "block programming" and "counterpro— gramming." Block programming stresses the advantage of having a suc- cession of similar type shows on a single evening. This strategy assumes that the placement of a program within a succession of similar type programs will attract a larger and more homogeneous audience than would be the case were the program in a succession of shows different from itself. Hence, a show does best when it is adjacent to shows of a similar type. Such large homogeneous audiences are attractive to the advertiser because the audience type is more easily defined. Counterprogramming, on the other hand, stresses the advantage of scheduling shows that are different from those of competitors. Rather than compete for an audience with particular tastes, networks program 4 to appeal to differing segments of the mass audience. Both block programming and counterprogramming have been discussed in reference to program type (see Doan, 1970; Brown, 1971; Shanks, 1976). Steiner (1952) and Owen, Beebe and Manning (1974) have used the counter- programming concept for constructing models of media economic systems and predicting viewer satisfaction. These models assume that the mass audience can be segmented into groups with different program type pre- ferences. Concerning the broadcasting industry Thayer (1963) noted: It is evident to most people in the broadcasting business that different kinds of persons are attracted to different kinds of programs. Knowing the precise differences that exist, then, becomes extremely important to the broadcaster and the advertiser (p. 218). While the above suggests attraction to particular program types, Goodhardt, Ehrenberg and Collins (1975) prepose a different process. They suggest the possibility of a . . . negative or inhibitory effect, e.g. that being addicted to one situation comedy or gng_western series is enough -- there is no need to watch other series of the same type (p. 46)." Regardless of which hypothesis is true, an acceptable typol- ogy for programs is needed. The Search for Program Types. A_priori category systems. Probably, the most widely recognized typology for programs is that used by A. C. Nielsen Co. (see Appendix A). This category typology is employed by both the networks and I! EEIQE} The principle advantage of this typology is its comprehensiveness. However, its comprehensiveness also makes it very complex. Recognizing this problem, Nielsen reduces its 36 categories into the following types for audience analysis: General Drama, Suspense and Mystery Drama (Sus— pense/Mystery, Police and Private Detective programs), Situation Comedy, 5 Variety and Feature Films. Notably absent are Action shows. Analysis of the Action category for February, 1976 indicates that only two shows (Bionic Woman and Six Million Dollar Man) constituted this category. The Suspense and Mystery Drama category was entirely composed of police and detective shows such as Baretta and Police Woman. There appeared to be little difference between these two categories. Combining the two should not significantly alter this category schema. Viewing preference or appreciation. This section describes research related to the establishment of program types by assessing the viewing preferences, or preferred programs, of the audience.' Wells (1969) ana— lyzed data obtained from TvQ, a syndicated rating service. Each respon- dent rated all network programs on a five point scale ranging from "One of my favorites" to "Poor." An additional category was provided for those who have not seen the program in question. Returns from about 2,000 respondents were analyzed for each of the years 1962, 1964, 1966, and 1968. For each year, the ratings of all programs between six and eleven P.M. were intercorrelated and analyzed via principal components factor analysis with varimax rotation. No criteria for the selection of factors were stated. The amount of variance explained by each factor or the set of factors is not reported either. Four types of programs, "Westerns," "News," "CBS Adults Situation ' and "Variety" shows appeared in all four years. Westerns were Comedy,‘ preferred by those in blue collar households, from small towns and rural areas and from families living in the south central section of the coun- try. Males provided these programs with higher ratings than did females in 1962 but these differences disappeared in later years. Demograph- ically, the only pattern to appear for news programs was that older 6 persons preferred these programs more than younger persons. The CBS Adult Situation Comedies were preferred by females over males and six through eleven year olds. There were no coherent demographic patterns for the Variety shows. The Variety factor was judged to be rather weak because some programs with a variety format failed to load upon this factor. Also, some programs which did load on this factor one year failed to do so in another. According to the researcher, these results indicate that programs normally classified as variety shows do not appeal to the same audience. Three of the four years produced factors defined as Action and Kids- Animals. For both factors, however, programs that loaded in one year failed to load in other years. Teenage Situation Comedies, Panel Shows and Supernatural Shows emerged in 1962 but disappeared by 1968. Teenage Situation Comedies were presented by ABC while Adult Situation Comedies were aired by CBS. Inter- estingly, this parallels the present situation in television programming. Demographic analysis of the preference data indicated that Teenage Situation Comedies were preferred by females more than males and, as might be expected, by the 12-17 year olds. The sex difference for these programs was higher than that for Adult Situation Comedies. Panel shows were also preferred by females, but, the age profile indicated that these shows were preferred by older people. Supernatural Shows were preferred by 12-34 year olds and members of larger families. Three program types appeared in 1962 but never appeared again. They were: Cartoons, Handsome Heroes and Drama. Cartoons were highly popular with.the youngest age group but popularity dimished for each subsequent age group. These programs were preferred more by males than by females 7 and by members of blue collar families. The Handsome Heroes category was composed of shows such as "Surfside 6" and "77 Sunset Strip." They were preferred by teenagers and blue collar families. No differences were found by sex. The Drama category drew high preference ratings from the following groups: females, older viewers, educated viewers, and residents of metropolitan areas. Three groups of shows emerged after 1962 and were still present in 1968. They were: Movies, Newlywed-Dating Game and Science Fiction. Movies were preferred by middle groups in terms of age and education. Earlier studies revealed that these shows were preferred by females. However, by 1968 sex differences had disappeared. In format the Newly- wed—Dating Game shows resembled defunct panel shows but their demo- graphic profiles were more like those of defunct Teenage Situation Comedy. Strongest preferences for these quiz shows existed for teen- agers and females. Science Fiction shows were most strongly preferred by younsters and teenagers. Doctor Shows emerged as a factor in 1964 but were absent by 1968. While they were on, these shows were preferred by women, people with less than a college education and blue collar families. The Wells (1969) study was the first comprehensive attempt to derive program types based on viewer preferences. The use of viewer pre- ferences or satisfaction might seem less important than other dependent measures such as actual viewing. However, it must be indicated that the achievement of high viewer satisfaction is one of the major goals of broadcasting policy (Emmett, 1967). Hence, the search for program types related to program preferences has direct policy implications. Rothman and Rauta (1969) attempted to follow the above goals by addressing the following question; "Can viewers be described and possi- bly categorized in terms of the programmes they like (p. 45)?" Such information, they suggest, would help to establish a rationale for the allocation of television time to different program types. Based upon preliminary research with programmers and interviews with viewers they derived the following gross categories for adult pro— grams: Religious, Sports, News, Drama, Comedy, Situation Comedy, Music- Variety, Music-Dance, Quizzes, and Serious Discussion. Subcategories within each major category were also specified; A set of 200 respon- dents were presented with a list of programs which they evaluated on an appreciation scale. The program appreciation scale was a Guttman scale developed earlier in the research. The results were factor analyzed and yielded nine factors. Criteria for selection of factors were not pre- sented. The amount of variance explained by each factor or the set as a whole was also absent. Using a procedure similar to canonical analysis, these nine factors were utilized as criterion variables which were maxi- mally related to a set of predictor variables. This was done in order to identify groups of shows which are liked by particular groups of people. Among the predictor variables were standard demographic descriptors, an appreciation scale for program types and a series of scales designed to measure the respondents attitudes toward specific aspects of television and television as a whole. A description of these last variables was not provided. The above procedures produced eight groups of programs. The first, labeled "Light viewing" was liked by people who preferred the following types of programs: Comedy, Variety, Spy, Crime, Adventure, or Westerns. In other words, preferences for these program types appear to run together. Sports was the second group of programs identified. They were generally preferred by men. Some specific sports shows such as "Professional Wrestling" were also preferred by those who generally like variety shows. In other words, two groups of people appeared to have different reasons for watching these shows. The third group of programs were labeled "Women's programs." These programs included romantic serials as well as cooking shows. The remaining program preference groups are self-explanatory and were listed as follows: Current Affairs, Religious, Culture, Quiz, and Pop Music. Comparisons to a_priori categorization schema indicate the fact that the a_priori schema are more complex than is necessary for a description of program groups, at least for prime time. Frost (1969) had respondents evaluate all programs offered by the BBC and ITV, on 58 pretested semantic differential scales. The results were subjected to Principal Components Analysis. Criteria for the inclusion of factors were not presented. Information concerning the amount of variance explained was also not accounted for. The results present nine factors. Having identified the nine factors used by view~ ers to discriminate among programs it was possible to compute program attitude ratings across all nine factors. Each program received a score on each of the nine factors. These program profiles were submitted to cluster analysis in order to find programs which covaried across the nine factors. The research identified six main clusters. The first was labeled "Family Entertainment." Programs in this cluster were concerned with relaxation and entertainment, had little or no crime content, were humorous and suitable for children. Cluster two was labeled "Pop." These programs were particularly uninformative, highly 10 informal, noisy, and suitable for children. The third cluster was labeled "Adult Crime/Violence." This cluster had considerable violence and was perceived as being more suitable for adults than the first two clusters. This appears to be like the traditional "action/adventure" category. Apart from containing "quiet" programs, Cluster Four had considerable feminine appeal and was labeled "Women's Romance." Cluster Five contained programs that were seen as being highly informative, unro- mantic, quiet, and for adult viewing. This cluster was labeled "Intel- lectual" but it appears similar to the traditional discussion and informational shows. The final cluster was dominated by sports programs and was labeled "Sports." Again an extensive categorization schema does not appear necessary in order to categorize programs. Goodhardt £3 31. (1975) report having identified six main program clusters people say they really like to watch during prime time. They identified these clusters by " . . . noting the cases where people who say they 'really like to watch' one particular program include an espe- cially high proportion of people who say they 'really like to watch' another (p. 112)." These groupings seem to correspond to prevalent a priori category schema. Types reported were: Sports, Current Affairs, Light Entertainment (dramatic serials, general, situation comedy), Adventure, and Children's. The sixth type did not seem to fit into common sense categories. Shows in this group were labeled as "Cult" programs (Monty Python, Star Trek and Pink Panther). These researchers do not attempt to partition viewers among these types as did Wells (1969). Rather, these categories merely reflect groupings of likings. Consistent with Wells (1969) these researchers found that some programs (e.g. "Professional Wrestling") seemed to fit into more than one 11 category. For example, preferences for "Professional Wrestling" were highly related to some sport shows but not others. In addition, pre- ference for this program was highly related to preferences for some light entertainment programs. They account for this by resorting to a " . . . 'Wrestling' therefore, appears to appeal functional explanation to somewhat different groups of people but for two different reasons -- to one it is sport, to the other light entertainment (p. 115)." Frank, Becknell and Clokey (1971) criticize earlier research con- cerning program types because they fail to control for confounding variables. Specifically citing the Wells (1969) study the researchers suggest that the factor structure for program types might be influenced by two sets of variables in addition to content; the personal character- istics of the respondent and the scheduling characteristics of the pro- gram. In other words, Wells (1969) factors may be due to other things besides program type. Frank et_§l, (1971) were the first to incorporate explicit statistical procedures in the experimental design to adjust for scheduling factors before factor analysis. Preference scores were adjusted by removing the effects of demographic, socioeconomic and tele- viewing variables and factor analyzing the residuals. Three factor analyses were run. In the first, neither individual nor scheduling factors were controlled. In the second, individual demographic (sex, age, family size, & market size), socioeconomic (education, occupation & income) and television (number of sets owned, color T.V., total viewe ing per week) characteristics were controlled. In the third, scheduling variables (day of the week, starting time, network, and the individuals preference for the program's lead-in) were controlled. The first two factor analyses yielded highly consistent results. The average number 12 of factors over a period of three years was ten. The factors accounted for 92% of the possible variance. However, the factor analysis of residuals following the removal of scheduling effects was markedly dif- ferent. The average number of factors increased to 26 and the variance accounted for reduced to 70%. The authors interpret these findings to mean that program scheduling has a major influence on the results of raw factor analyses such as that of Wells (1969). The authors were unable to identify any content related factors among the 26 generated in their last factor analysis which did not also emerge in the earlier two. Eleven factors were identified based on the results of the first two factor analyses. They were: Movies, Peyton Place, Solid Citizen, Conflict, Situation Comedy, Unreal Escapist, Westerns, Doctors, Comic Strip, Mating Game, and Teenage Music. Again, all factors represent clusters of "liking." The first two factors are self—explanatory. The third, labeled "Solid Citizen" suggests a single cluster for two a priori content categories. According to the researchers, there appears to be a positive relationship between preferences for musical variety programs (Andy Williams, Dean Martin, etc.) and panel shows (What's My Line, I've Got a Secret). The remaining eight factors did not hold up across all three analyses and their labeling is quite tenuous. For example, the factor labeled "Situation Comedy" consisted of 12 programs all of which were broadcast on C.B.S. Obviously, this factor should disappear when the effect of scheduling variables (such as network) were controlled. This, in fact, was the case. Similar problems exist for the "Unreal Escapist," "Western," "Conflict," and "Mating Game" factors. The remaining factors ("Doctors," "Comic Strip," and "Teenage Music") may have some content clustering since they continued to exist after 13 scheduling variables were controlled. The Frank g£_al, (1971) study was a distinct departure from earlier studies because it removed the effects of confounding variables prior to conducting factor analysis for program types. However, the authors proceeded to identify content factors which disappeared when the effects of scheduling variables were controlled (e.g. the "CBS-Situation Comedy" factor). Such a procedure is misleading. Such factors cannot be attributed to content elements alone. Labeling these factors according to content is obviously inappropriate. The decision to control for individual differences by Frank etual. (1971) was questioned by Gensch and Ranganathan (1974). These research— ers question what the resultant abstract factors represent. Since the factors arrived at were based upon data in which individual or subgroup differences were removed, the factors cannot be related to any subgroups. Hence, the abstract factors would appear to be of no utility for identi— fying program type clusters which differentially attract types of people. In other words, the derived factors would be of questionable use to market segmentation or programming. The preceding discussion was concerned with the search for program types which have been observed in studies of viewer preference. A summary of this research as it relates to the present study will appear following the discussion of program types as determined from viewing behavior . Notably absent from research on preference are studies designed to assess program types for children. While a considerable number of studies have used typologies and assessed preferences and behavior, the program types had not been established on the basis of research (see 14 Rubinstein, Comstock & Murray, 1972). These studies were not designed to discover program types. Rather, they attempted to categorize viewing behavior and preferences within "intuitive" classification schemes. This research is covered in the section labeled "Children's Program Preferences." Viewing behavior. This section reviews the literature concerning the establishment of program types by the analysis of viewing data. In a landmark study, Kirsch and Banks (1962) attempted to validate commonly used program types for their utility as predictors of individual viewing preferences. This study was the first to utilize factor analysis of audience responses as a procedure for establishing a program typology. The typology evaluated was not presented in the report. However, the number of categories (36) is identical to that utilized by Nielsen. 220 A.R.B. diaries from the Chicago area were used as the data base. Each program was categorized by type. For each type the researchers divided the total number of viewings by the number of opportunities to view. The correlations between program types were factor analyzed and graphi- cally rotated to simple structure. Five factors were extracted and accounted for 33% of the variance. No criteria for factor extraction were reported. Only three factors were discussed in the report. The first was described as a factor with adult programs at one end and children's programs at the other end. Factor two contained programs with easily identified protagonists. The third factor was described as containing more "subtle" programs. Perhaps realizing that the research provided only one clear dimen- sion the researchers conducted a second study. In this study they decided not to employ any §_priori schema. Rather, they let actual 15 viewing behavior dictate program types. The data consisted of 596 A.R.B. diaries from the eastern time zone for February, 1961. Only the viewing patterns of males were analyzed. Analysis of the viewing pat- terns indicated that correlations were high for programs on the same network on the same evening and highest between adjacent programs. The correlation matrix was subjected to factor analysis with varimax rota- tion. The criterion for factor inclusion was whether the factor accounted for at least two percent of the variance. Six factors were extracted and accounted for 24% of the variance. The factors were labeled: "ABC Westerns," "CBS Situation Comedies," "Music/Variety," "ABC Action" (adventure, mystery, drama), "NBC Westerns," and "CBS Action." This last factor was extremely weak (among the shows on this factor was "Leave it to Beaver"). According to the researchers, there were considerable differences among the reasons for viewing groups of shows. While men preferred to view Westerns, they often viewed programs which were selected by others in their households. Both situation comedies (sitcoms) and music/ variety programs were identified as program types which were viewed because they were selected by women in the household. They suggest that a good deal of research would be necessary before program type could reliably be used as a predictor of viewing patterns. The research of Kirsch and Banks (1962) is important for several reasons. First, it was the initial attempt to use audience behavior as the criterion for establishing program types. Second, it was the first to provide clear empirical evidence for the existence of lead-in or inheritance effects. Finally, it was the first study to suggest that pre—established typologies did not accurately reflect viewing patterns. 16 Swanson (1967) conducted a similar factor analytic study to search for program types related to exposure patterns. Data were obtained by personal interviews of 2371 female heads of household. Among other things, respondents were asked to recall their exposure to nighttime network television programming during the 1964—65 season. Correlations between exposure to each show were submitted to principal component factor analysis. The criteria for factor inclusion were: a) the factor must account for at least two percent of the variance and b) the factor must have at least two programs with loadings no less than .30. This procedure resulted in twelve factors which were rotated to an "objec- tive" solution. The factors accounted for 45% of the total variance. The researcher's labeling of factors consisted of the programs which loaded heavily on them. Factor one was labeled "Donna Reed-Patty Duke-Ozzie and Harriet" and accounted for twelve percent of the vari- ance. They appear to be situation comedies. The second factor contained mostly quiz shows plus "Perry Mason" and "Andy Griffith" (7% of the variance). Factor three was a series entitled "90 Bristol Court" (5%). The fourth factor contained mostly drama programs (3%), the fifth.most1y westerns (3%) and the sixth contained sports (3%). While this last factor was rather small, its loadings were quite high. Appar- ently, women who watch one sports event are quite likely to watch others. Six more factors, each accounting for 2% of the variance were presented in the report but are not worth mention. The researcher suggested that the study's results provide a parsi- monious and efficient analysis of viewing patterns. This author disagrees. The entire set of factors accounted for less than half of the variance in viewing patterns. Furthermore, each of the last half 17 of the set explained only two percent of the variance. The study as a whole did little to substantiate the significance of program type effects on viewing. While further criticism might focus upon the use of recall rather than actual viewing or diaries, Goodhardt et_§l, (1975) suggest that recall methods may overestimate actual viewing but do not distort patterns of viewing. Ehrenberg (1968), in reviewing Kirsch and Banks (1962) and Swanson (1967), pointed out quite clearly that the two works contained major conceptual errors. Summarizing Ehrenberg's (1968) critique, and entend— ing it a bit, Gensch and Ranganathan (1974) stated: He shows that the correlations among programs are due not only to program content (show types) but also to other influences such as time of day, day of the week, and network. It is obviously wrong to try to interpret configurations in terms of only one dimension (show types) when the form of the configuration is significantly influenced by other dimensions. Ehrenberg reexamined Swanson's data and pointed out that the two major factors which Swanson identified in terms of show types actually represent the ABC and NBC networks (p. 390). Hence, the major conceptual problem of factor analysis, factor interpretation, has clouded up an area which appeared to be proceeding satisfactorily. Ehrenberg (1968), using ARB diaries as a data base, found fairly consistent patterns of viewing. He reported a small but positive correlation (.2) between the viewing of programs on the same day on the same channel. Interestingly, the identical correlation was found for programs on the same channel but one week apart. This sug- gests that the inheritance effect which appeared first in the Kirsch and Banks (1962) study may in fact be an artifact of channel loyalty, the propensity for viewers to prefer a particular channel. While cast— ing some doubt upon the strength of the inheritance effect Ehrenberg's (1968) analysis does support that of Kirsch and Banks (1962) because 18 the highest correlations in both studies were observed for adjacent pro- grams. This latter result supports the existence of an inheritance effect which is limited to adjacent programs. Continuing with the findings of the Ehrenberg (1968) study, the re- searcher reported a small negative correlation (-.1) for the viewing of programs on the same day but on different channels. An average correla— tion of -.4 was reported for programs on simultaneously. The evidence again suggests channel loyalty. However, one would expect a larger neg- ative correlation between viewing programs which are on simultaneously. This result suggests that there is some simultaneous viewing or channel switching going on among programs. The overall pattern, however, is negative, as should be expected. Usually, people watch one program and do not engage in channel switching or simultaneous viewing. Analysis of the relationship between viewing programs on the same channel but on different nights of the week yielded a slight but posi- tive correlation (.l). A similar analysis for programs of different channels on different nights of the week yielded an average correlation of zero. Ehrenberg (1968) stressed that these results could not have been obtained by factor analysis since there are no established procedures for quantitatively utilizing prior knowledge concerning structural factors such as channel loyalty and the inheritance effect. Gensch and Ranganathan (1974) disagreed. They stated that the problem was not inherent in factor analytic procedure but rather in the fact that it had been misused and misinterpreted by Swanson (1967) and Wells (1969). In a study similar to that of Frank EEHEA° (1971) Gensch and Ranganathan (1974) factor analyzed their data after controlling for the 19 effects of scheduling variables. Hence, the factors which emerged are not confounded by channel loyalty and other structural factors. Fifteen factors were extracted which had an eigenvalue of at least 2.0. They explained 65% of the residual variance. A program was included on a di- mension if it had a loading of .38 or better. The researchers replicated this procedure with a second sampling of subjects from the same data set (respondents of the Brand Rating Index) in order to determine which fac- tors were stable. Fifteen factors emerged which explained 60% of the residual variance. Eight factors emerged which were common to both sets. These common factors were labeled: "Movies," "Action," "Light Enter- tainment," "Traditional Values" (Lawrence Welk, Andy Griffith, Green ' and Acres), "Peyton Place," "Dating and Newlywed Games," "Westerns,' "Variety." These factors accounted for 45% of the residual variance. Darmon (1976) attempted to utilize a_priori program types and channel loyalty simultaneously to predict program ratings. Using regression techniques he found that 72% of the variance in TV ratings was explained by the use of these two variables. Both variables were found to have a significant and unique effect on program ratings. Summary. Early research on program types, whether concerned with behavior or preference, suffered because of the confounding effects of structural variables like time and the inheritance effect. When this was pointed out (Ehrenberg, 1968) other researchers employed procedures which controlled for the variability attributable to these structural elements. These researchers were still able to find program types which influenced adult viewing patterns and preferences. This confirma- tion of earlier research, however, is tempered by the fact that the typologies which emerged were not entirely consistent with each other. 20 None of these studies were done among child viewers. An analysis of the program types which emerged in each study is reported in Table 1. Excluded from the table are any types which emerged in only one study since there are a sufficient number of studies to indicate stable and replicated types. The following program types have been found to emerge repeatedly: "Westerns," "News and Current Affairs," "Sitcom," "Variety," "Action/Adventure," "Children's," "Panel or Quiz," "Drama," "Films," and "Sports." Each type has emerged in both preference and viewing studies with the exception of "News and Current Affairs." This type emerged repeatedly in preference studies but failed to emerge in any of the viewing studies. It must be recalled, however, that very few such programs exist during prime time, which was the central concern of these studies. The earlier research on program viewing and preference as well as the more recent have confirmed the existence of program types which influence patterns of viewing. Only one study, however, (Darmon, 1976), has employed both structural and program type variables to predict pro- gram popularity or ratings. Cluster and factor analytic studies have only indirectly addressed this issue. They were an important first step upon which to build predictive models for program popularity. Unfortunately, Darmon (1976) ignored this body of research and used an §_priori program typology. Models which employ the knowledge generated from these studies of program type have yet to be applied. Children's Program Preferences. Notably absent from the literature concerning the derivation of program typologies are studies concerning children. This author has not encountered a single piece of research concerned with the 21 Table 1 Summary of Research on Program Types Study Gensch Rothman Good- Kirsch & & hardt Frank & Ranga- Wells Rauta Frost g£_al, 25 El: Banks Swanson nathan Type (1969) (1969) (1969) (1975) (1971) (1962) (1967) (1974 Action— Adven- X X X X X X ture Child- X X X ren s Drama X X X X X X X Films X X X News & Current X X X X Affairs Panel & Quiz X X X X Situa- tion X X X X X X Comedy Sports X X X X Variety X X X X X X X was" x x x x x erns 22 establishment of a program typology for this category of viewers. Usually, within the larger category schema, in a single category, are lumped all animated or non-animated programs which appear, to the re- searcher, to be made for children. No direct empirical evidence exists which justifies the lumping of these programs together. Children watch a great deal more than children's programs. The research to be covered in this section will indicate the broad variety of programs that they do watch. According to Lyle (1972) children begin viewing at a very early age and soon develop program preferences and habits. Lyle and Hoffman (1972b) indicated that preschool children are able to name their favor- ite programs. Over 80% of the three year olds interviewed were able to name a favorite program. Even the youngest children(three year olds) indicated that they watch television on a daily basis. Upon catego- rizing their expressed program preferences by type, cartoons accounted for almost two thirds of their choices. In a study aimed at analyzing the program preferences of schoolchildren from first through tenth grade Lyle and Hoffman (1972a) found that first graders preferred sit- uation comedies and cartoons. Sixth graders had dropped the cartoon programs and showed a heavy preference for situation comedies and an increased preference for adventure shows. Tenth graders preferred ad- venture as well as music-variety and drama. Sex differences were also apparent. Boys appeared to develop a preference for adventure shows before girls. Girls, on the other hand, appeared to develop prefer- ences for sitcoms earlier than did boys. Further analysis of the data indicated that program.preferences are more diversified as children grow older. However, it is not known whether this is due to a general 23 desire to diversify or due to the greater variety of programs available to the older children. In a national survey LoSciuto (1972) asked mothers to specify their child's favorite programs. The children's ages ranged from three to twelve. Reports were obtained for about 140 youngsters. The research- er described respondents' preferred programs as cartoons, other child- ren's programs and family comedies. These clusters were based on the researcher's intuition. Few of the programs were prime time. Most were daytime programs. The LoSciuto (1972) study suffers in several respects. Firstly, grouping the viewing habits of such a diverse age group provides relatively little information which could be used to determine the viewing behaviors of any of these children. It has already been seen (Lyle and Hoffman, 1972a, b) that the viewing pre— ferences of children change as they mature. Furthermore, research by Greenberg, Ericson and Vlahos (1972) and others indicates that mothers are not good estimators of their childs television exposure. Never- theless, the results are consistent with other studies. Murray (1972) analyzed the viewing of six year old inner city boys with regard to program preferences and found situation comedies to be the most popular. This was followed by cartoons and action programs. General drama, sports and news programs were at the low end of the popularity continuum. Program type preferences for these youngsters were highly correlated (.76) over two points in time suggesting good temporal stability. Situation comedies and cartoons accounted for about 70% of all preferred programs. Action and adventure shows accounted for another 20%. The preference pattern was closely related to actual viewing as recorded by diaries. 24 Streicher and Bonney (1974) conducted interviews with children aged six through twelve concerning their program preferences. Children were asked to name the programs they liked and disliked. Programs which appeared to be of the same type to the researchers were clustered toget- her and given a label. Empirical criteria for the clustering would have been preferable. The most frequently mentioned favorites were cartoons, fantasies, sitcoms, and mystery-suspense dramas. No examples of each category were reported. Cartoons were favored more by boys than by girls. However, the popularity of cartoons for both groups decreased with age. While dramas were also more popular for boys than girls, their popularity increased with age for both groups. Apparently, the preference for cartoons is replaced with a preference for mystery/sus- pense programs. The category of fantasy programs were also preferred by boys over girls. No age trend was apparent for this program type. Girls overwhelmingly preferred situation comedies. This category was hardly mentioned at all by boys. The program types disliked most by this sample of youngsters were news and talk shows. The research thus far indicates that certain program types are preferred over others and that preferences appear to change as a child grows older. Preschoolers prefer cartoons while first graders prefer situation comedies as well. As children approach their teens, prefer- ences for cartoons are replaced with preferences for action/adventure shows. These preferences for action/adventure shows develop more rapidly for boys than for girls. Girls seem to remain loyal to situation comedies. No reasons are offered in these studies which can explain why these sex differences appear. 25 Reasons for Program Selection Most of the research relevant to program selection is more concern- ed with reasons for exposure to television rather than exposure to a particular program or program type. At the initial stage, the potential viewer must decide to watch television or to engage in some other activ- ity. Next, having decided to watch television the viewer must select among the programs offered. While these decisions are somewhat related, considerably more research has addressed the general issue of television exposure than the issue of program selection. It should be obvious that reasons which affect the choice between viewing and other activities are directly related to program popularity. One area of research concerned with exposure to television has examined the relationship between exposure and background variables of the viewer such as age, sex, socio-economic status, education, etc. Several studies document these relationships (for recent examples see Lyle and Hoffman, 1974a, b; Greenberg, 1976). However, these research efforts do not provide a satisfactory explanation for television view- ing behavior. For example, why is it that children watch more televi- sion until the age of twelve and then slowly begin to watch less? This type of question concerns itself more with the use of television by the audience than it does the effect of television on the audience. Inves- tigations concerned with the individual's use of mass media have generally been labeled "uses and gratifications" research. Von Feilitzen (1976) summarized this approach as follows: Investigations of this kind . . . are based on the assumptions that the individual, by his use of the mass media, obtains a reward in the form of needs gratification. The individual is selective and chooses (more or less consciously) mass medium and mass medium content on the basis of the functions, or the meaning, which the medium and the content have for him, 26 and the availability of functional alternatives. The functions are steered by the individual's needs, which are dependent in turn upon both psychological factors existing in the individual . . . and the more social factors. Thus, mass media effects do not occur unless the individual himself chooses to use the mass media in a certain way . . . (p. 95). Several researchers have attempted to outline the functions or needs of television for children. Von Feilitzen (1976) presented four major functions: entertainment, informative or cognitive, social and escapist, based upon interviews with children. Her research indicated that compared to the other mass media television satisfies the most needs. Secondly, children felt that it satisfied these needs best. However, as children grow older, radio appears to satisfy the need for escape more than does television. Perhaps this is due to the fact that television is more often viewed by the family together than is radio where listening is more of an individual or peer activity. Further information concerning the ability of television to satisfy a multiplicity of needs was provided by Brown (1976). He had 800 children (aged 7 - 15) respond to a series of functional statements by indicating which medium serves the functions best. Examples of these functional statements are: "Which of these do you do when you feel lonely?" and "Which of these do you find most exciting?" For nine of thirteen statements television received more choices than any other. On only two occasions did fewer than 25% of the children select television. These cases were times when the child was sad or wanted to forget. In both cases, recordings showed a marked increase. Res— ponses to these statements also varied by age. Reported incidence of the use of television for need satisfaction increased between seven and nine years of age and remained at about the same level until they dropped for teenagers. This pattern is highly consistent with that of 27 actual television exposure. One interesting finding indicated that the use of television for social utility purposes was high for the youngest groups but was being displaced by records and peers as the child grew older. By the age of 15 television, records and peers had an equal chance of being a topic of conversation. These studies are examples of research aimed at the uses and gratifications of television. They suggest that television does have some specific uses for children. They also indicate that these func- tions differ in importance as the child grows. Hence, viewing levels by age may be attributable to the declining dependence of the child on television for need fulfillment. The gratifications obtained from television by British children were studied by Greenberg (1974). Application of factor analysis to the responses of these children on a set of scales yielded eight factors. No criteria for the extraction of factors were presented. The entire set of factors accounted for 56% of the variance. The factors and the common variance explained by each were: "Learning" (20%), "Habit" (14%), "Arousal" (13%), "Companionship" (11%), "To Relax" (14%), "To Forget" (13%), and two "Pass Time" factors each accounting for five percent. The research provided no reason for the splitting of the "Pass Time" items. However, the distinction appears to be that one set provided reasons for watching TV while the other provided situations where TV may be viewed. It also may be that the questions aligned themselves into two factors because of their phra- seology. The first of the two factors began with the word "when" while the other began with the word "because". This difference could be enough to result in unique factors. 28 The remaining results of the Greenberg (1974) study are unique in terms of the depth of information provided. Firstly, the researcher presented factor strengths which reflect the mean score for each factor. This information indicated the importance of each of these reasons for watching TV for the sampled children. All the factors averaged about three (out of a maximum of four) with the exception of relaxation and forgetting. These two factors were lower than the others in terms of strength, especially the latter. Secondly, the researcher reports correlates of each of these functions. Of particular relevance to this study are the relationships between these functions and exposure to different types of programs. First of all, general exposure to television was positively correlated with the gratifications labeled "relaxation" and "habit." Concerning specific types of programs, exposure to violent content was positively related to reports of the use of television for arousal. The particular types of content most strongly related to the arousal gratification were action-adventure and science fiction shows. Watching out of habit was positively correlated with the viewing of nonviolent programming but also with science fiction violence. Watching to forget was positively related to viewing nonviolent content. The above data indicate that exposure to certain program types are related to the gratification obtained or sought from television. Greenberg (1976) pointed out that the parallelism between gratification sought vs. those obtained was still a problem for gratification research. Furthermore, be indicated that research to date had been unable to separate the gratifications obtained from the medium of television vs. those obtained from its content. However, this piece 29 of research does signify that there is some relationship between program content and television gratifications. This is not to suggest that particular program types are monofunctional, serve only one function, but that some program types appear to serve particular functions better than others. Further indications of the relationship between program type and gratifications were suggested by Lyle (1969). He stated: "Individuals do differ . . . in their selection of content within each medium." Con- tinuing: " . . . the person who is troubled or physically fatigued may seek 'escape' content from the media . . . (p. 210)." The author also cited research which indicated that people with higher intelligence levels more frequently select media content that offer deferred grati- fications than do people with moderate or low intelligence levels. These propositions and citations support the contention that different program types attract differing audiences. In a more conceptual theoretic overview of an individual's mass communication needs Atkin (1973) suggested two needs which are fulfilled by the mass media. First, there is the need for information. This is a function of two factors labeled extrinsic uncertainty and intrinsic uncertainty. Extrinsic uncertainty was defined as being produced by a perceived discrepancy between one's current level of certainty and a criterion level. The criterion level is determined by the importance of environmental objects to adaptation requirements. Intrinsic uncertainty, on the other hand, is generated by a perceived discrepancy between one's present state and a goal determined by personal interest. The second need specified was the need for entertainment. The magnitude of this need is a function of the discrepancy between one's 30 present state of enjoyment and some criterion level. He specified that intrinsic uncertainty and enjoyment needs are satisfied during exposure. Extrinsic uncertainty, on the other hand, is satisfied following exposure. Messages that reduce extrinsic uncertainty have instrumental utility for the receiver. They help one make decisions. Paralleling the "fractions of selection" discussed by Schramm (1972), Atkin (1973) indicated that the benefits of exposure to a message must be weighed against expenditures and liabilities. Expendi— tures are a function of the available resources and perceived amount of energy required to expose oneself and decode the message. Liabili- ties are perceived harmful consequents of exposure to a message. Liabilities include social sanctions for attending to particular stimuli, foregone benefits which might have accrued had the individual engaged in other activities, increases in uncertainty caused by expo- sure to the message, aversive emotional arousal caused by exposure to the message, etc. These factors of benefit, expenditure and liability are quite important. For if they operate as suggested, they should predict program selection and popularity. The program that maximizes benefits relative to expenditures and liabilities is the one which will obtain the highest popularity in a given time slot. Predicting_Prqgram Populariyy. The use of "block programming" by type suggests, as was noted earlier, that it is advantageous to schedule shows which are similar in type close together. The corollary,of course, is that programs will do worse if they are programmed adjacent to programs which are different in type. If networks follow this line of reasoning one should expect 31 to find programs of a similar type following each other on actual net- work schedules. Furthermore, other things being equal, programs scheduled adjacent to same type show should have larger audiences than programs which are adjacent to programs of a different type. In other words, the benefits to the viewer are greater, relative to expenditures and liabilities, when programs which are adjacent are of the same type. Empirical research concerning these predictions is virtually non- existent. Related research, however, has been conducted concerning the existence and extent of inheritance effects for programming regardless of type. Goodhardt e£_§l, (1975) described the inheritance effect thusly: Program planners and television commentators have often sub- scribed to the belief that once viewers have switched on their sets, they tend to continue watching the same channel throughout the evening. Such a belief puts a high premium on attracting a large audience early in the evening. Once attracted, it is regarded as relatively easy for subsequent programmes on that channel to 'inherit' part of this audience (p. 43). This notion appeared to derive support from the Kirsch and Banks (1962) study where high correlations were noted between audiences of different programs on the same evening on the same channel, and the highest figures were noted for immediately adjacent programs. However, in the absence of norms for the usual amount of duplication neither the size nor the limitations of this effect could be determined. Goodhardt g£_§l, (1975) suggested that audience duplication for pro— grams on the same channel on the same evening should be compared to audience duplication figures for shows on the same channel but on different evenings. Upon doing so they discovered that the inheri— tance effect is far more limited than was originally supposed. Channel loyalty, the propensity for people to view a particular channel, 32 accounted for most of the higher correlations among shows on the same channel. For the most part, whether two shows were on the same or dif- ferent nights had no effect on the size of the audience that watched both shows. The only consistent exception to this rule was when shows were directly adjacent to each other. Their research in England indi- cated that audience duplication between adjacent programs was 15% higher than for nonadjacent programs. Parallel research in the U.S. (Goodhardt g£_§l,, 1975) indicated that audience duplication between adjacent programs is 24% higher than for nonadjacent programs. The U.S. study also revealed some inheritance effects for programs that were one show apart (called adjacent plus one). Audience duplication between adjacent plus one programs was found to be 7% higher than shows which were neither adjacent or adjacent plus one. Had channel loyalty not been taken into account, the inheritance effect would have been overstated. They summarized their findings as follows: These then are the general findings. It is now clear why many interested parties have overestimated the significance of catching a large audience early in the evening. It is because the basic nature of channel loyalty (as reflected in the dupli- cation of viewing law) was not widely understood. One could say that there is a general inheritance effect for programmes on the same channel shown on different days (i.e. channel loyalty). The apparent pattern of inheritance on any single evening is then only a facet of this general effect. The special inheri- tance effect for programmes close together is additional to this and can be due to three possible causes. Either people stay tuned to the next programmes out of inertia, or because the programme has ended part-way through the programmes on the alternative channels, or because they tuned in to the previous programme to wait for the programme scheduled to appear next (a "lead out" rather than a "lead in" effect) (pp. 45-46). The existence of an inheritance effect, however, does not provide the media programmer with much useful information for programming decisions. All it says is that the ratings of adjacent programs are related and that it is advantageous to have a program adjacent to a 33 highly rated one and that programs which occur earlier but are not adjacent have minimal, if any, effects on a program's popularity. The block programming concept, however, suggests an explicit strategy. It suggests that the inheritance effect will be maximized if adjacent pro- grams are the same type. No research has addressed this issue. As described earlier, counterprogramming involves designing a sche- dule of programs so that they are different from those offered by the other networks in the same time period. Rather than compete for the same audience networks appeal to differing segments of the larger mass audience. Employing this logic, Steiner (1952) and Owen, Beebe and Manning (1974) have constructed elaborate models designed to predict viewer satisfaction given different media configurations and economic systems. However, their research is lacking in two ways. First, the basic assumption that an audience can be divided into segments with unique program preferences (Steiner, 1952) or groups with ranked pro- gram type preferences (Owen g£_al,, 1974) is not supported by research. Secondly, their models have yet to be empirically verified. While models are a method for unraveling the impacts of complex phenomena they often oversimplify and fail to predict accurately. No information is available to suggest the extent to which the models proposed make accurate predictions. Research by Bruno (1973) indicated significant correlations between viewing programs of the same type. The correlations ranged from -.17 to 40. The existence of any negative correlations was quite surprising since the reported correlations were for programs of the same type. One problem in this research was the failure to control for confounding factors such as channel loyalty. Negative correlations may reflect 34 the fact that two programs of the same type were on different channels and that the channel loyalty factor was stronger than the effect of program type. It might also be the case that some shows of the same type were on different channels at the same time. No controls for such factors were reported by the researcher. However, some of the research reviewed earlier concerning viewing behavior and program type (Gensch and Ranganathan, 1974; Darmon, 1976) did control for these factors and verified the existence of a program type effect on viewing patterns. Hence, counterprogramming by program type appears to be a viable strat- egy for increasing program ratings. Were program type not a deter- mining factor, the ratings of programs would not be affected by the types of programs which were on other channels at the same time. The available research justifies consideration of counterprogramming by type as a viable programming strategy. The preceding discussions of counterprogramming and block program— ming centered on the relationship between a programs own type and the types of programs which are adjacent or competing. However, type is only one possible description of a program. Also of concern is whether a program is new or returning from the prior season. When a program is known to the viewer, benefits and expenditures can be estimated with greater ease than when a program is new. Since returning programs are those which were popular, it stands to reason that these shows should do well compared to shows whose benefits are not known. Con- cerning block programming, it should be better to have a program adjacent to returning rather than new programs. The same logic applied to counterprogramming suggests that programs which are opposite new programs should do better than those opposing returning programs. 35 These effects, however, are contingent upon the viewers familiarity with the programs. As familiarity with the newer programs increases, the impact of returning programs should fade. Hence, by the end of the season, programs which were new are no longer new. Thus, returning programs should have no greater impact upon the ratings of programs which are adjacent or opposite than do new programs by the end of the season. In addition to the major discussions of such strategies of counter- programming and block programming, there are several others which should be examined. These include the use of "spinoffs," the time a program is broadcast and sex of the leading character. One programming strategy which is increasing in popularity is the "spinoff." These programs contain characters who first appeared in other programs. Viewers are familiar with the characters since they appeared on successful shows. The benefits and liabilities of such programs could be based on those of the earlier programs providing spinoffs with a considerable advantage over other programs. Another directly manipulable variable is the time a program is broadcast. Research on adults by Goodhardt g£_ a}, (1975) indicated that the potential audience for later prime time programs is smaller than for those broadcast earlier. Casual inspection of Nielsen reports also suggests this trend. Such a factor should be even more pronounced for children whose bedtimes come earlier than adults'. The time a program starts should be an important determinant of the popularity of programs for children. Sprafkin (1975), in a laboratory setting, found that children viewed programs that stressed characters which were of the same sex 36 as the viewer. Hence, programs with male leading characters should be attractive to boys while programs with female leading characters should be more attractive to girls. Other predictors of program popularity may exist but are not sub- ject to the manipulation of programmers. For example, the network programmer cannot take advantage of the fact that programs on a parti- cular channel do better than programs on other channels. This may be due to reception problems at the viewers'set or the fact that one provides superior programming. Summary The preceding section has attempted to review the literature in order to identify major elements which impact upon the popularity of a program, with special attention to children. The elements appear to be organized in three areas: Audience characteristics, scheduling strategies and program characteristics. Audience Characteristics Scheduling Strategies Prpgram Attributes Age Counterprogramming by Type Type Sex Block programming by Spinoff Type Counterprogramming by Sex of lead New vs. Old character Block programming by New vs. Old New vs. Old Time of program Popularity of Adjacent Program Included are elements subject to manipulation by programmers as well as others which appear to be beyond their control. The first section attempted to establish a relationship between program.popularity and 37 program type for the purpose of counterprogramming and block program— ming. The research suggests that program types do emerge when struc- tural factors are controlled. The emergence of these types suggests the viability of audience segmentation. Resting upon the assumption that each brand of a product sells effectively to only a segment of the total market, segmentation is used in order to market or advertise the product to the appropriate segment of the market. It appears that the research on program type has followed these notions by attempting to segment the television audience by program type. While most of the research has addressed the adult viewing audience, research on children suggests a similar strategy may be effective. Broadcasters appear to agree that such a segmentation is possible. According to the broad- casters interviewed by Cantor (1974), if a program is aimed at boys it will be adventure, western or space fantasy; girls prefer comedy and rock and roll groups. Certain program types have emerged with regularity over several studies. The ones which apply to prime time broadcasting appear to be as follows: Action-Adventure News & Current Affairs Sports Drama Panel & Quiz Variety Feature Films Situation Comedy Westerns "Westerns," "Panel & Quiz" and "News and Current Affairs" have prac- tically disappeared from the network's prime time schedules. An addi- tional category labeled "Children's programs" has emerged but is of no utility for the purpose of segmenting this market or predicting prime time program popularity for children. These categories have emerged in studies which used adults rather 38 than children for their data. Some researchers might argue that children and adults are so different that it would be useless to apply a typology based on an adult sample for an analysis of children. On the other hand, this researcher would argue that there are good reasons for utilizing the existing typologies in spite of the fact that they were not generated by children. First, several types have emerged regularly in several studies. These studies were conducted with samples of different sexes as well as geographic locations. The fact that these types emerged regularly suggests that they have some gener- alizeability and stability across different samples. Second, is the issue of parsimony. It is the author's contention that one should attempt to generalize extant knowledge to new samples rather than generate new knowledge for different samples. The end result of the latter procedure would be to have several disconnected and unrelated areas of knowledge concerning television and human behavior. When one has a finding that occurs with some regularity for one group of people, they should attempt to discover whether the same phenomenon occurs with other groups of people. Ignoring prior research on program types would not allow one to assess the generalizeability of these program types for children. The generation of program types for children should follow attempts to generalize existing knowledge concerning program types. Attempts at discovering a new typology should begin after research indicates that the old typology does not work for younger populations. The generation of program types for children at this stage seems premature. What is needed is an attempt to gener- alize existing program types to a younger population. The use of factor analysis would not allow one to take prior knowledge and 39 research into account. Ehrenberg (1968) stated the issue succinctly: The factor analytic problem ... is that there are no techniques for taking such prior knowledge into account. Factor analysis is in fact generally described as an explor- .a£gry technique, to be used when one happens to know nothing much about the subject matter. Every time one wants to use factor analysis one has in effect to pretend ignorance of one's subject-matter. This may be more difficult for some of us than others (p. 58). While no explicit typologies have been generated for Saturday morning programs, research does provide a basis upon which to create a typology. The research of Greenberg and Atkin (1976) suggests a dichotomy between animated and nonanimated programs. The research of Cantor (1974) suggests that a useful dichotomy would be comedy vs. adventure. Such findings should be used and generalized before at- tempts are made to create new typologies by exploratory methods. The discovery of consistency across several studies dealing with the viewing preferences and behaviors of adults and children suggests temporal stability in viewing patterns and program popularity. How- ever, few studies have addressed this problem directly, especially for children. The issue is especially important for studies attempting to establish the magnitude of the relationship between exposure to tele- vision content and viewer attitudes and behaviors. Given that the predominant measure of exposure is the self-report, the issue becomes one of the stability of these viewers reports over time. Do children View (or report that they view) the same programs throughout the season or do they view (or report that they view) different programs? Mea- sures of exposure which attempt to assess enduring exposure patterns or program popularity at only one point in time will suffer if insta- bility is high. The issues addressed by this research were divided into four sets. 40 The first set concerned the validity, reliability and stability of pro- gram popularity based on respondent self-reports. It was a qualitative assessment of the measurement of the dependent variable, program popu- larity. The second set assessed the effects of programming strategies which were set forth by the hypotheses. The next set examined program popularity for each program type by respondent characteristics such as age and sex. The final issues addressed concerned changes in the ef— fects of programming over time. Hypotheses A comment about the forthcoming hypotheses is necessary at this point. The hypotheses are not derived from any formal theory. Each hypothesis is an attempt to formally state the principles of program- ming suggested by various writers as well as those principles suggested by empirical research as they apply to program popularity for children and adolescents. Sufficient information was generated to Specify di- rection in all but one hypothesis. Hypothesis 2 is nondirectional and more exploratory than the others. The following hypotheses concern the stability of program popu— larity over time. 1. The stability of program popularity is higher later in the season than earlier. 2. The stability of program popularity differs for children of different ages. The following hypotheses concern counterprogramming. 3. Program popularity is positively related to the number of different type programs opposite a program. The 5a. 5b. The The viewing 9. 10. 11. 41 Program popularity is negatively related to the number of returning programs opposite a program. 4a) The effect of returning programs which are opposite a program diminishes over time. following hypotheses concern the inheritance effect. Program popularity is positively related to the popularity of a programs' lead-in. Program popularity is positively related to the popularity of the program which follows a program. The correlation between the popularity of adjacent programs is contingent upon whether they are of the same type. The correlation is higher when adjacent programs are of the same type. following hypotheses concern block programming. Programs which are adjacent to programs of the same type are more popular than programs which are adjacent to programs of a different type. Programs which are adjacent to returning programs are more popular than programs which are adjacent to new programs. 8a) The effect of new vs. returning adjacent programs diminishes over time. following hypotheses concern audience characteristics and the of program types. Action-adventure programs are more popular among boys than girls. Situation comedies are more popular among girls than boys. Variety programs are more popular among girls than boys. 12. 13‘. 14. 15. 16. 17. 42 Sports programs are more popular among boys than girls. Boys prefer programs with male leading characters. Girls prefer programs with female leading characters. The preference for action-adventure programs is positively related to age level of subject. The preference for feature films is positively related to age level of subject. The preference for Saturday morning programming is negatively related to age level of subject. The following hypotheses concern other manipulable programming strategies. 18. 19. 20. The time a show starts is negatively related to program popularity for prime time shows. Programs which are spinoffs are more popular than programs which use only novel characters. Returning programs are more popular than new programs. 203) The advantage of being a returning program diminishes over time. CHAPTER II METHODOLOGY This chapter describes the manner in which the study was conducted. The first section presents the variables and their operational defini— tions. The second section describes the procedures employed in the execution of the study. It includes a description of the programs selected, the respondents, data collection and procedures used to assess validity and reliability of measurement. This section concludes with the procedures employed for statistical analysis of hypotheses as well procedures for pggp Egg analysis. Variables Program'Type. Program type was used as an independent variable as well as the input for constructing variables concerned with programming strategy (counterprogramming & block programming). A total of 132 programs were analyzed. The prime time categories and the shows assigned to each are presented in Table 2. There were 103 prime time shows. Out of these shows six were deemed uncodeableauuiassigned missing values for the type variable. These shows were: "Almost Anything Goes," "American Music Awards," "Bugs Bunny/Road Runner," "Circus of the Lions," "Sixty Minutes," and "Walt Disney." There was disagreement between the two coders who assigned types to programs in only two instances. These were discussed between them and a type was mutually agreed to. No criteria for coding a program into a particular type were 43 44 Table 2 Prime'Time Shows by'Type Action Adventure (n=29) Situation Comedy (n=30) Drama (n=16) Barbary Coast (F) Baretta Barnaby Jones Bionic Woman (W,S) Blue Knight (w,s) Bronk Cannon City of Angels (S) Columbo (W,S) Emergency Harry-O Hawaii Five—O Hawk (S) Invisible Man (F) Jigsaw John (W) Joe Forrester Kojak Matt Helm (F) McCloud (F) Moblile One (F) Police Story Police Woman Rockford Files Rookies Six Million Dollar Man Starsky & Hutch Streets of San Francisco S.W.A.T. Switch Variety (n=7) Carol Burnett Cher (Sonny & Cher; Donny & Marie (v.5 Mac Davis (S) Rich Little (W,S) Tony Orlando & Dawn All in the Family Barney Miller Big Eddie (F) Bob Newhart Chico & the Man Cop & the Kid (W) Doc Dumplings (F) Fay (F) Good.Times Grady (W) Happy Days Honeymooners (W) Jeffersons Joe & Sons (F) Laverne & Shirley (W,S) Mary Tyler Moore MASH Montefuscos (F) On the Rocks (F,S) Beacon Hill (F) Doctor's Hospital (F) Ellery Queen Family Holvak (F) Kate McShane (F) Little House on the Prairie Marcus Welby Medical Center Medical Story (F) Movin' On (F,W) Petrocelli (F,W) Rich Man, Poor Man (w) Sara (S) Swiss Family Robinson (F) Three for the Road (F) Waltons One Day at a'Time (w,s) Feature Films (n=12)* Phyllis Popi (W) Rhoda Sanford & Son That's My Mama (F) The Practice Welcome Back Kotter When'Things Were Rotten (F) Sports (n=3) Basketball Playoffs (3) Boxing (S) NFL Football (F) Saturday Night Live (F) ABC Monday Night (3) ABC Friday Night (F,W ABC Saturday Night (S ABC Sunday Night (F,s CBS Thursday Night (F CBS Friday Night (w,s NBC Monday Night (F,s NBC'Thursday Night (W,S) NBC Saturday Night NBC Sunday Night (S) Note--Letters which follow program title indicate when true popularity data were collected; F=Fall, W=Winter, S=Spring. lasted the entire season. *There were two ABC Saturday Night and Sunday Night Movies in the Spring. Programs without letters Ac\1 ”a..- 45 established a priori. Given the ease with which the task was done, no such criteria appear necessary. However, the coders did state that the differences between categories were not always clear. Some programs were difficult to code. The elements used to determine whether a program was action-adventure were discussed by the coders after the task was completed. They stated that such programs stressed physical action often with violence. Also, they seemed to deemphasize inter- personal relationships. Drama programs, on the other hand, had more emphasis on interpersonal relationships and less stress on action and violence. While no formal typologies had been established by previous research, information was available which provided input for a Saturday morning program typology. Cantor (1974) suggested a dichotomy related to con- tent such as adventure vs. comedy. Other research suggested a dichot- omy related to mode of presentation e.g., animated vs. nonanimated. Inspection of the set of Saturday morning programs indicated that while most cartoons were comedy some were adventure programs. Simi- larly, while most nonanimated programs were adventure programs there were some comedy programs. 'Therefore, programs were categorized into the following four groups: Animated Comedy, Animated Adventure, Nonanimated Comedy and Nonanimated Adventure. Saturday morning programs are arranged by type in Table 3. 'There were a total of 29 Saturday morning programs. Programs were assigned to categories by one coder and the author since the other coder was not familiar with the pro- grams. Agreement was reached on all but one program. 'The difference was discussed and a category assigned to the program. "Go-USA" did not fit into this category scheme. #6 Table 3 Saturday Morning Programs by Type Animated Adventure (n=4) Animated Comedy (n=14) Emergency Plus Four Adventures of Gilligan Return of the Planet of the Apes Bugs Bunny/Road Runner Super Friends (S) Fat Albert Valley of the Dinosaurs Groovy Goolies (W) Hong Kong Phooey Jetsons (W,S) Josie and the Pussycats Oddball Couple Pebbles & Bamm-Bamm Pink Panther Scooby Doo Secret Lives of Waldo Kitty Speed Buggy (W.S) Tom & Jerry/Grape Ape Nonanimated Adventure (n=6) Nonanimated Comedy (n=4) Isis Far Out Space Nuts Land of the Lost Ghost Busters Lost Saucer Sigmund & the Sea Monsters (F) Run Joe, Run Uncle Croc's Bloc (F,W) Shazam Westwind Note--Letters which follow program title indicate when true popularity data were collected; F=Fall, W=Winter and S=Spring. Programs without letters lasted the entire season. Program Popularity The dependent variable for this study was program popularity. Respondents were provided with a checklist of programs offered by the major networks arranged by day of the week. 'They were instructed to check off only programs which they watched every week or almost every week. Program popularity was operationalized as the percentage of respondents who checked off the program. Shows which were on only 47 once such as specials and sporting events were checked off if they had been watched “that one time. Progam 2m Counterprogramming This variable was operationally defined as the number of programs opposite a program which were different in type from itself. The values of this variable ranged from zero to two. A program's counterprogramming.status was only evaluated once, at the beginning of the program. Changes in opposing programs during a program's duration were not evaluated. There were several reasons for this decision. First, Goodhardt gt 3;. (1975) stated that such changes could have only minimal impacts on the individual viewer. People usually watch a whole program. In the case of half- hour programmes, about 93% of those who watch the first quarter-hour also watch the second. With much longer pro- grammes more substantial erosion of the audience occurs -- up to about 20% of initial viewers may be lost by the end, and even more late in the evening (p. 19). In their extensive research, the authors have consistently used the first quarter-hour estimate as their estimate of the size of the audience. The results of the second quarter-hour were hardly different and certainly would not affect the conclusions (Goodhardt §§_g;,, 1975). A second reason was this author's inspection of a set of national Nielsen Ratings for February, 1976 (this concides with the time in which data were collected for the present study). The inspection indicated minimal (less than one percent) variation for the duration of the vast majority of programs, including feature films. Hence, it was surmised that changes in opposing programs could have only minimal effects on a program's popularity. Program'Pype Block Programming This was operationally defined as the number of adjacent shows which were of the same type as the show of concern. Lead-in and follow- in shows were analyzed separately. Hence, this variable was broken down into two. The first was whether the lead-in (preceding) program was the same type. The second was whether the following program was the same type as the show itself. Both variables were treated dichotomously. Ngwuyg. Returning Counterprogramming A program was coded as new if it became part of a network's regular schedule during the season analyzed or if it made its initial appearance during that season. The counterprogramming status of a program in regard to new vs. returning programs was operationally defined as the number of returning programs opposite a program. Its range was zero to two. .Nggwx§, Returning Block Programming This variable concerned the number of returning shows adjacent to a program. Lead-in and following shows were analyzed separately. Two dichotomous variables were generated, one for the lead-in and one for the following program. Tim; Time was defined as the time a program was scheduled to start. A 24 hour clock was used in order to distinguish between A.M. and P.M. An 8 A.M. program was assigned a value of 8.00 while an 8 P.M. program was assigned a value of 20.00. 49 Spinoffs Programs categorized as spinoffs were those whose leading character(s) originally appeared in a different program. Grass Respondents were given questionnaires which were color coded as to their school grade. Sex The respondent's sex was obtained via self report (Boys=1, Girls=0). Network Information regarding the network which presented a program was obtained from.TXfl§uid§. Effect coding was used for each Network. A variable was created for two of the three Networks. The last Network (NBC) was assigned a minus one (-1) on each of the other two variables. This yielded a treatment effect attributeable to the network a program was on. 'This treatment effect is defined as the deviation of the group's (network's) meanflrml the grand mean. The grand mean in this case is the mean popularity of all programs. Sga‘gf Leading Character This variable was independently assessed for each program by two coders. Programs with a male leading character were assigned a value of 1. Programs with female leading characters were assigned a value of 0. Programs with both male and female leading characters were assigned a value of .5. Procedures Selection‘gf’Programs The programs selected for study were those broadcast during the week prior to each administration of the questionnaire. Prime time and 50 Saturday morning offerings of the three major networks were selected according to T! figidg. Prime time was defined as 8-11 P.M. Monday through Saturday and 7-11 P.M. on Sunday. Saturday morning shows were those offered between 8 A.M. and 1 P.M. Because it was not felt to be a "Saturday morning" type show, popularity data were not collected on American Bandstand (air time was 12:30 P.M.). Respondents Respondents were students in one suburban elementary and middle school in Mid-Michigan during the '75-'76 academic year. 'The elementary school was among others that fed students to the middle school. Completed questionnaires were obtained from 300 in the fall, 286 in the winter and 281 in the spring. Representation by sex was approxi- mately 50-50 at all administrations. There were 101 fourth graders, 106 sixth graders and 93 eighth graders in the fall. In the winter there were 100 fourth graders, 97 sixth graders and 89 eighth graders. Data collection in the spring yielded 99 fourth graders, 101 sixth graders and 79 eighth graders. Data Collectigg Data were collected in October, February and May of the '75-'76 school year. All grades were queried on the same day. Questionnaires were administered to the same intact classes. Questionnaires were ad- ministered to each class by the researcher and two colleagues. After the questionnaires were distributed instructions were read by the administrators. The instructions stated that the researchers wanted to know what television shows the child had watched over the last few weeks. They were instructed to "... put an X next to the shows you watch every week or almost evegy week." Following the instruction the 51 programs were presented by day of the week as follows: Monday Night Tuesday Night ____HONEYMOONERS ____HAPPY DAYS ‘____RICH MAN, POOR MAN ____LAVERNE AND SHIRLEY ____RHODA ____THE ROOKIES The administrators stressed that the children were to check off programs they had watched and not just ones that they liked. A sample of the exposure measure is in Appendix B. Responses were coded onto optical scanning sheets and automatically punched onto cards. The popularity of each program was determined by calculating the percentage of respondents who placed a mark next to the show. This information was placed on a set of schedule sheets which were structured by day, season (Fall, Winter and Spring), time and Network. Each program was placed in its appropriate location on these sheets. In this way coders were able to see which programs preceded, followed or competed with each other. After assigning each program a type and entering its popularity score, coders were able to determine 'the appropriate values for the counterprogramming and block programming variables. The appropriate values for each program were transcribed onto coding sheets by each of two coders. ‘The coders exchanged sheets eund checked each other's work. Errors were observed to be minimal. lkata from the coding sheets were tranferred onto computer cards 'Via.manual keypunching. This resulted in a data set where each case :represented a program. Since the central concern of this project ‘was program popularity, this was the appropriate unit of analysis. 52 Validity and Reliability pf Program Popularity Measure Quality of measurement was assessed in several ways. The winter questionnaire contained several shows for which it was absolutely impossible to have been exposed for the few weeks prior to adminis- tration. This set of shows contained a program which had not been on yet (City of Angels), two that had been early cancellations ( Big Eddie & Sigmund and the Sea Monsters) and two that hadn't been on for several years (Gunsmoke & Winky Dink). The extent to which children reported viewing these shows could be interpreted as some indication of validity of the checklist exposure measure. A further assessment of the quality of the data was obtained by comparing program popularity for the sample of respondents in the winter to comparable figures for a national sample. The Nielsen National TV Ratings for February 9-22, 1976 were used for this purpose. Nielsen program popularity for age groups is based on the diary method. Ratings for fourth graders were compared to Nielsen's 6-11 year olds. Eighth graders were compared to teens (12-17). Sixth graders were compared to both of Nielsen's groups. The appropriate Nielsen Station Index (local Nielsen data) was solicited by this author from a local television station but the data were not made available. These data would have been preferred over the national figures for assessing concurrent validity. On the other hand, the comparison of this study's popularity estimates to ratings from a national sample provides information concerning the generalizeability of this study's popularity estimates to a national sample. While some discrepancies should occur due to differences in the samples and reporting methods, it was felt that a comparison was useful. 53 The final assessment of the reliability of program popularity is associated with testing the first two hypotheses. The procedures em- ployed for testing these hypotheses also provided estimates of reliabi- lity for each grade level. The coefficient of reliability is defined as the ratio of true score variance to total variance. The latter is composed of true score variance and error variance. This definition is equivalent to the classical definition of reliability, the square of the correlation between observed and true (corrected for measurement error) scores. Estimates of true score variance and error variance are calculated based on the three waves of observed scores. The formulas employed for this estimation are presented in Wiley and Wiley (1970). Statistical Analysis Stability. Stability is defined as "... the correlation between the true scores at times i and j (Wiley & Wiley, 1970, p. 115).' As such it is a test-retest correlation between measured variables whose measurement error has been removed. These stabilities can be estimated given three waves of observed scores. Separate analyses were conducted for each grade in order to determine if age related differences emerged. Programming Strategies. Simple correlations were employed as the initial test for the programming strategies outlined in the hypotheses. The variables were correlated with program popularity in order to determine whether the hypotheses had any justification. Subsequently, multiple regression was employed to determine the significance of the unique contribution of each of the predictor variables for predicting 54 program popularity. The predictor variables and their associated hypotheses were: counterprogramming by type (3), counterprogramming by new vs. returning (4), popularity of lead-in (5a), popularity of following program (5b), block programming by type (7), and block programming by new vs. returning (8). The latter two variables were analyzed by separately considering the programs which preceded and followed a program. In addition, the time a program started (18), whet- her it was new or returning (20) and whether it was a Spinoff (19) were included as predictor variables. Length of program (in hours) and effect coded variables for network were included in the equations as controls. The three sets of data (Fall, Winter, Spring) were analyzed inde- pendently. For each season, separate analyses were conducted for prime time and Saturday morning programs. Within prime time and Saturday morning, analyses were conducted over all respondents. Separate analyses for each grade and sex were also conducted. Regression analyses for each sex group included one additional variable, sex of leading charac- ter, in order to test Hypotheses 13 and 14. Audience differences. Hypotheses about program types preferred by one sex over the other (9-12) were assessed via one tailedut test. Differences by grade as specified by the hypotheses (15-17) were tested via one way Analysis of Variance (AOV) and subsequent t tests. Changes over time. Hypothesis 20 was concerned with whether a returning program would have an advantage over a new program. Hypothesis 20a indicated that the advantage would diminish over time. The first of these hypotheses was tested in the multiple regression equations (described earlier. The second was tested by comparing the size of the 55 difference between the ratings of new and old programs at each season. The best estimate of the difference between new and returning programs was the regression coefficient (unstandardized). It was deemed superior to other estimators because it adjusts for the effects of the other variables in the equation. Hence, the differences reported for each season were uniquely attributable to whether a program was new or returning. 'The differences among the regression coefficients for Fall, Winter and Spring were tested via one way AOV and subsequent t tests. Hypothesis 4a was concerned with the diminishing impact or return- ing competing programs (those on another channel at the same time). Again, the best estimate of the unique impact . . . . ecafisecfios 2. NH 3 .3 an. an. won. «a. S. S. 2. 3. an. . . . . . . . ...- . . ...... . . . . 88 3 3 S .283 ton. tun.Iac~.Itcn.I.-M.~ Ieonjehn It: Inc... IL.~.Is~m.Itm.—.I enmitoafu .Itwnftmnfnhnf 259 Panama“ in; I mo I Hm Ituw Itcmftmwf tmnf 49.: an: mH.I.O~.ItnN.I «A: ON. OH: 8. MO., 8. Loaocsogo anon O O I O O O O- O- O I O O O 0 x0 . m." I mo I ma I OH I an I a." r so .3 .no I an I .3 I 8 I mo I on: no: 2.: «a: co: uucdcmswmuag . . . . . . .. . 2:329: .c 8 no I no I 00 I 00. no.4 mo I Ow I HA I 2": 8.: 8.: 8: an: «H: OH... on}. mm: ”5552 o o o o o o o o o o . . . cal 6 Ac tan emu #2” LR. tau tan to“ tun :5 can. can. can. to“. #9... tom. to“. tea. tea. enema V .2... on m... .8. a. on. ma. 2. 3.. mo. .8... ..o. :3. mo... 2. 8. S. 8. wfiwcwfimflmW _ ”new . . . . .. 1.3.38...— 3003 to.— am to... tun. 3.3. eon. ion .3. 4.8. .2. Om. cm. .3. no. sew. 0A. «a. mu. we.“ £5833 Me A v 0 O O O o o o . o o . o . . QdLGH—a Oh D can tom tom :3 ..mn can. «A .3 son .3 9... row. am. no. an. «a. .3. an) 54:3.— «0 ha «.— omsmoa any Amvuoote 09.3.32...” on: .8: 00$ 00: 3": n1: an: 3. mo? 3. an: AG? .3: mo. 3. mo. 8. 8. so: .n> we." . ..ESoe sec: .2. .3. .2. .ma. .2. .mn. 2. fl. 2. .2. 3. 3. 8. no... 8. 8. .8. 8. use some . cases.» . Ignace-30 h x c o 4 Han h z . c e .- Hue ..— z a o .— Han Rom. 0.5.5 J35 non 0.3.8 Iago. new 395 £95 . have? ......“de L35: Hana common . . .30chon we no» can 0098 and ace-com bu . 80330.28 kiddos .52.. It. .3...— fis. :23...» .333: .8 OH OAR!" 66 ability of each variable to predict program popularity. In multiple regression, any changes in the dependent variable (program popularity) which are attributable to more than one variable are partialled out of both. This joint effect, however, remains and contributes positively to the amount of variance in the dependent variable which is explained by the set of predictor variables (R2). The results of the multiple regression analyses for prime time shows are reported in Tables 11 through 13. Table 11 presents the results by season across all res- pondents. Table 12 presents results by season and grade of respondent. Table 13 presents results by season and sex of respondent. The figures reported in the tables are unstandardized regression coefficients. These coefficients represent the expected change in the criterion variable attributeable to a change of one unit in a predictor variable, controlling for the other predictors in the equation.. Hence, the figures in the tables are the expected change in program.popularity, in. percentage points, for a change of one unit in the corresponding pre- dictor variable, controlling for the effects of all the other predictor variables in the equation. The analysis of the correlations between the hypothesized predict- ors and the popularity of Saturday morning shows is reported in Table 14. The multiple regression analyses for predicting the popularity of Saturday morning shows are in Tables 15 through 17. Table 15 presents results by season across all respondents. Table 16 presents the results by season and grade of respondent. Table 17 provides results by season and respondent's sex. The results of the correlational and regression analyses are presented in the text of this report as they pertain to the research hypotheses presented in Chapter 1. 67 Table 11 Multiple Pagression Analyses for Predicting Prime Time Program Popularity Season Wt ' ' Fill ‘Hinter Sgrigg Counterprograming: . by W -e78 '- 5059 by Returning vs. New .53 -.92 -2.0h Block Programme Lead-in Same as Program 41.51 -l6.6l «- Pbllowing Same as Program -h.27 -- 8.116 Lead-in Returning -17.91 -9.56 5.13 Following Returning 9.11; 2.25 -2.h9 Popularity of Lead-in .22 .30 .23 Ponularity of following -- .29 .20 Sbimffa 12 e 35 2 e 33 -he 31‘ Returning vs. New Program 25.27. 7.91 5.96 Starting Till. .60m ‘5063 -50h6 Length of Program 4.9.81 43.56 40.75 Chanel: L30 «2.13 3.30 13.77 as ‘lefl ... 4.36 NBC 3.63 «— 1.91 Constant 1039' 51.85” 18.80 R2 .657 .522 .h92 Adjusted R2. em .03}: '033h P (d!) 1.91 . 1.19 0.60 (11.15) (11.12) (13.3) late—me absence of a re ssion coefficient indicates that the 988 default value (1‘ (.01 excluded the variable from the analysis. The clones of these variables were essentially sero. ‘ Adjusted 32' 3.2- “3%) (1-32) . where k- the number of independent . variables and N- the number of cases. 13605 ' 68 Table 12 Multiple Regression Analyses for Predicting Prime Time Program Popularity by Season and Grade . Season Predictor —‘_ p.11 Winter m Grade th 6th 8th th 6th 8th Counter-cro- gaming: by ”fleece .lelo '9- “W 1.51; 9e96 «1.09 3e29 5e“ 3e” by Return- 108‘s“ 1.95 e92 .55 '3019 "" '3072 -10; 4.340 ’1015 Block Pro- gunning: I‘d-in 3 5.31 ..- ‘h. 39 £1.17 -13 .87 .17 06h ... 3 on -.. Following 38"....uol -3.05 ... -6.00 low 503‘} 3087 8.115 8078 ... Lead-in Rtming. all-12 .83 .15.93 -19 cm ‘7e63 .12 em '13 .57 10.“ ‘ 6092 he}? Following Returning. 01 heZL heZh Sew Sal-‘2 3.“: heh6 .- 40$ ‘11.” Popularity of LQ‘d-mJ .10 .29 .27 e26 .28 .1314 .21} 031‘ .28 Popularity of follow- 1118‘. e e e e e e e o.- .013 e13 e32 e17 029 .21 e10 .20 Sinatra... .. 8.61; 7.10 16.59 2.15 ... 2.06 -3.53 43.25 «6.35 hturning '3 e new progru. 29.50" 23.7h* 23 .hT" 9.78 3.87 7.32 6.59 5.11: 7.26 at.” Till. .9e5h. ‘Seao -3.” $.70 -5095 “'- ‘80119 'lbesll «2.71 3h“ “mu L10.61 '15e2h ‘19.23 .211.“ 42.86 '19e96 .10082 .10.” -1205? Channel: I ABC -6.71 .2 .13 .97 1.08 5.02 5.68 3.01 5.32 7 .81 CE CO. ‘2 .92 -3 .18 ... ... .2 .17 -1061} .2 .115 -3015 NBC ~- 5.05 2.21 "I- ... “3051 .1e37 ‘Zea‘, 41o“ Constant 36.73 32 .68. 37 .01: 55.59 50.72 113 .22 17 .96 111.97 213.01! :2 .am .612 .639 .589 .529 .mm .552 .510 .m Adz e R2 .11.? e32? .279 .0511 0097 -0270 .0011; -01168 ‘e3b6 ? at) 2.15 2.15 1.77 1.10 1.22 0.62 .93 .52 .55 1 10 ll 2 I3 10 (12 9 1h 7 12 9 Note-Jrhe absence of a regression coefficient indicates that the SPSS default value (P(.01) excluded the variable from the analysis. p<.05. (59 Table 13 Multiple Regression Analysis for Predicting Prime Time Popularity by Season and Sex of Respondent Program Season Predictor Fall Winter SEing Sex: Boys Girls Boys Girls Boys Girls Counterprogramming: by w .— 4.60 ... 2.89 3.61 15.00 by “tn” '8 M :e78 ... 1.20 .1051 “" “2079 Block Programming: lead-in Same as Program -9.65 2.66 - -11.07 -- 5.711 Following Same as Progra- -8.3h 1.39 6.65 -1.83 6.58 13.35 Lead-in Returning -l8.36 -1h.28 -8.76 ~7.b2 5.20 13.37 Follalling Ratumng 7.38 8.60 107,4 .‘ .. 4.10 PWHW Of u‘d'in e18 011‘ e18 ..- e30 e27 Popwuty 0f POI-loving 0.. .10 ’e01‘ 030 .09 032 Sex of Lead Character 10.65 -8.68 o3.78 -17.86 12.126 47.“: Spinoff 10.311 16.130 -2e9s 11.35 ‘8092 .3063 Returning vs. New Program 212.57n 28.58" 7.81 13.85 6.83 9.73 Mm T1” $.18 .6096 -5085 .18032 -5088 -7.§ “mu or W” 43032 '12095 ‘15.“ 41o” -1h.71 4.216 Channels ABC 1e& -Yels 6.37 hob} 6076 3eb2 cas 4.65 2.11. -5.3o 4.71. -- -3.0h NBC 2.03 5.01 -1007 -1e69 .- -e Constant 1.1.01. 39.51. 50.57 66.57’ 18.1.7 12.29 3’ .632 .721. .153 .562 .m .6? Adjusted 3’ .261. .hoz -.1hh ..008 ..235 -.201 P (df) 1.72 2.25 0.76 0.99 0.67 0.77 (13.13) (115.12) (12.11) (13.10) (12.9) (15.6) Note-orhe absence of a regression coefficient indicates that the SPSS default value (P<.01) excluded that varable from the analysis. The slopes of these Vii-ables were essentially sero. * p<.05. 70 . mo.va a. .e e e» no ago 2? swabkoaemooee-H eumomwoa o no on none coho; cH Proms: ...wwoz 8. ...... .6 . H . r... on. R. a. .R. .24. . . . .. . .S. .3. .2. .... 3.....3883 R: «932.- 2: «03%... R: enffiaf R... an? S... 2? E...- 8: on: 3: i.- 8.65883 8: mm. 2.. 2. 2. a. d- a. ma. n . . 8. .- N. S. a. a .- . .3825 .83 . a H we we a e 8 no now can: 8. .8. 8.- «a. 3. «a. 2. mo. .2. 8. 2. mo. .82 .3.- 2: 8.18. 9...- 2.538 9.26023 NH. o~.- .8.- no: mo: 8.- 2. .8: 8f 3. 2. 8. um. mu. 9... em. on. 2. ”5.558 . fi-oaflsv S. :7 ma... 8: my. .3.- S. 5.1:? 8: 3. fit 8. 8f 8? 8.. S. 8: 2.6 . . . . . 8.3%.; 8. a. 9... S. 3.. ms. fl. 2 a S 8 8 .3. .3. .3. .3. .2. .2. T... ...-...“... . . ulohmonm :83 ...... ma. 8. an. .3. ma. an. 2...? S. 3. 3. 8. 2. 2. S. R. S. E. 1.508 no b25583 m“. 8. 8. .8. 2. i. 3...... R. on. 60.2. R. R. .52.”. on. on. ...-...... no $2.98.... 1338:. coneauuoscH Neel “Nel 0°62. mHel Dflel OHeI NHeI JNel ONel sol 8e! Fuel ONe OHe 8e ”Ce JNe OH. in on} we... 0 O O O O O O 'Esva EAJV mm? 3 - 3... .....f 9...- on.- ma - so - 8 ma - mu - 2.. - an... mu... fl? 3... on? 9...- RFH. .EAC Ends—Esau 6.5.3958 nxeoad.~=eoa Hfi.uze,oafi. Now ...-.5 .905. now 23.5 1.35. New one-5 nacho . no 0 Ava-ovflfom Cut-:35: _ . Aaévnek a «can cocoon aconconeem uo Rom on: scene ace £8.33 P b.2336. e88.— ucho... 025... 5c. 823...» .3369... .8 8635959 4H 0.5..“ 71 Table 15 Kultiplo Regression Analyses for Predicting ‘ Saturday Morning Program Popularity p . ¢ Season redicoor ---- Fall Winter Spring Counterprogreming: by W .6007 100& .106" by Returning vs. New -- -12.h3 -- ‘ A Block Programing: ' Lead-in Same as Program 4.25 «0.0.98 1.91: Following Same as Program 4.2.1.5 - -9.73 Lead-in Returning 46.23 1.68 -7.80 Following Returning -1h.86 -7.30 -h.99 Popularity of Lead-in .32 -.72 -— Popularity of Following .39 «56 6.27, Spinoff: -- ’ -26.30 416.31 Returning vs. New Program 25.1114” 7.77 7.72 Starting Tim ~- . .10038 ‘105? Length of Program 18.140 25.30 25.71 Channel: ABC ... .2207? -8001 CBS 11.02 211.65 9.117 rm .- -1088 dohé Re .771 0838 0565 Adjusted R2 .1181- .1-15 -.306 (10.8) (13.5) (12.6) NoteuThe absence of a re ression coefficient indicates that the SPSS default value (P901 excluded the variable from the analysis. The slopes of these variables were esqentially zero. * p905. 72 Table 16 Multiple Regression Analyses for Predicting Saturday Morning Program Popularity by Season and Grade ‘ Season Predictor 23.1.} ' ___W1 "tar £231.23 Grade: 11th 61.1- 8111 h'th. 61:11 81.1- hth 6th 8th Counterpro- gramming: - by W000. no. -" ”.- ..- 6.” 11.19 C“ .6.” 2.72 by Return- ingvs.nou. -6-22 -5.08 -—- -- -6.21 9.68 -S.65 -§.88 8.39 Block Pro-. cramming: Lead-in sam 7.06 -b.95 7.97 7.07 7.63 32.27 ~3.S3 -15.13 26.87 Following 38M....... -8007 -111012 ‘11036 ... 5.5; 12032 ‘9072 4110117 2.91 “turning. qr .1090 ‘9 .00 ... ‘9 01° -7011]. “16018 -7014? 2002 “11073 Following Beturning.l -- -16.h3 -- -- 2.35 -3.h8 -—- -8.59 -1b.37 Popularity Of Lead-in. 037 019 032 02h 031 ... 02h -009 ‘ezh Popularity of follow- ~ Wee... coo --"' 0’39 ..- .C. -019 -026 -033 011‘ .23 SMtISOOOOGt u. on... 6.07 .6055 ’6027 .10038 -17096 .12056 .1607? nae-mung vs.‘ 39.66" 21.56" 15.33" 111.8h* 13.29 10.03 8.89 1.81 12.57 now progrmnl at.“ Till. 0". 1". - m m 5058 .6051} 41.02 6.7]. Show Length ... 18.89 —- 23.52 55.57 111.22 22.53 32.13 21.13 Channel: ' ABC 12.00 "- hell? —. 41.10 -180111 .11036 -5057 '6076 CBS bozo 3.911 m 6012 60120 10096 11036 9058 1.85 no -16020 ... -.. ... .2.” .7th ‘70m .6100]. 16091 Constant 28.51 - -oh9 5.56 25.23 -8.0h -36.09 66.93 62.07 -18.03 22 .831 .663 .531 .599 .629 .696 .617 .603 .717 Ad? . 32 0696 .2111 e297 03M '03}? -0096 .150 -0788 ’0273 df) 6.15” 1. 57 2.27 2. 35 0.65 0.38 O. 80 0. L3 0.72 (a 101 (10 a) (5 121 (7-11) (11,9) (11 s) (12.6) (11.11) (ll-,1.) Note--The absence of a regression coefficient indicates that the SPSS default value (F(.01) excluded the variable from the analysis. 13605. 73 Table‘l7 Multiple-Regression.Analysis for Predicting.Seturday Morning Prograa Popularity by Season and Sex of Respondent Season Predictor ' Fall Winter SEEing Sex: Boys Girls Boys Girls Boys Girls CounterprogranMng: by Type 4.52 4.0.67 -- -- -- 6.82 by Returning vs New - --- —6.05’ --33 -3.25 -- Block Programming: lead-in Same as Program -- 1.85 -- 1h.hh ‘ -- -12.76 Following Same as Program 45.69" 42.39 -—- 10.92 45.18 48.52 Lead-in Returning -- -- -9.27 -- -3.S9 5.h5 Following Returning —- ... --- 11.26 o8." 4.1.29 Popularity of Lead-in -— .31 . .12“ .52 ... ..- Popularity of Following -- -—- -- -.21 -—- .hl Sex of 103d Character 26098. “O 18.90. 2033 ‘17037. 16.09 31311101.! can 9 e21 .- 5.2 3 ‘17 02° -16 .02 Returning vs. New Program 23.18‘ 211.36" 16.05. 8.26 12.89" 3.59 Starting Tm ..- .— 2027 5032 -2e8h 'llelb Length of Program 19.13 -- 617.56” 31~69 bl.h8 -- Channel: ABC -"- 50m ... -3 01.11 -5 02° .6009 CE 9.79' 5.33 11.28 7.113 6.89 10.89 N80 .- -1007“ ... 4.1.02 -1069 .hem a? .818 .713 .751 .688 .771 .758 AdJuSth R2 0727 061814 0552 -0122 .1311 0273 P (d!) 8.98" 3.1.1 " 3.7? 0.85 2.111 1.56 (6.12) (8.10) (8.10) (13.5) (11.7) (12.6) Hoteu‘rhe absence of a regression coefficient indicates that the SPSS default value (F<.Ol) excluded that varable from the analysis. The slopes of these variables were essentially aero. * p<.05. 74 Counterprogramming Hypothesis 2, Hypothesis 3 stated that program pepularity is positively related to the number of different type programs opposite a program. An examination of Tables 10 and 14 presents the relevant correlations. The zero order correlations were significant and in the hypothesized direction over all respondents as well as for all age and sex analyses for prime time programs in the Spring. However, the multiple regression analyses failed to produce even a single instance of support for this hypothesis. In an effort to discover why counter- programming by type emerged as a significant correlate in the Spring, it was found that this was the only instance where this variable was significantly correlated with the variable start time (r=-.27). In other words, in the Spring, later programs appeared to be counterpro— grammed to a lesser degree than earlier programs. This fact coupled with the observation that starting time was consistently negatively related to program popularity (line #18 in Table 10) provides some explanation for why counterprogramming by type emerged as a significant correlate of program popularity in the Spring. Counterprogramming was more prevalent earlier in the evening when more children watch regard- less of programming strategy. The correlation between these two vari- ables also provides the reason for the absence of a counterprogramming effect in the multiple regression analyses. The effect of counterpro- gramming by type and the decline in audience size as the evening grows later were confounded. Hence, neither variable was credited with this effect. In the judgment of this researcher, Hypothesis 3 was not supported. 75 Hypothesis 3, Hypothesis 4 stated that program popularity is negatively related to the number of returning programs opposite a pro- gram. Neither the zero order correlation analysis nor the regression analysis yielded a single instance of support for this hypothesis. Correlations in prime time ranged from -.18 to .09 (Table 10). Corre- lations on Saturday morning were more divergent. They ranged from -.28 to .24 (Table 14). The lesser fluctuation for prime time shows is most likely attributeable to the larger number of shows upon which the corre- lations were based. Hypothesis 4 was not supported. Counterprogramming summary. The hypotheses testing the effects of counterprogramming on program popularity for school children failed to be supported. While partial support existed for prime time shows in the Spring regarding counterprogramming by type, it appeared that the relationship was spurious. The time a program started appeared to be the cause for the observed relationship between counterprogramming by type and prime time popularity in the Spring. Inheritance Effect. Hypotheses Sa,b. The general hypothesis stated that program pop- ularity is positively related to the popularity of adjacent programs. The effects of lead-in (5a) and following (5b) program popularity were assessed independently. The popularity of a program's lead in was a significant correlate of prime time program popularity across all respondents as well as all age and sex groups in the Spring (Table 10). The correlations ranged from .34 to .43. Analysis of the correlations between the popularity of the lead-in and a program's own popularity in the Winter also yielded some significant correlations. The correlation was significant for the 76 popularity of programs across all subjects (.28) and for eighth graders (.30). Again, however, the regression analyses failed to corroborate the zero order correlations. This suggests that the popularity of a program's lead-in is related to other variables in the equation. Anal- ysis of the intercorrelations among predictor variables yielded two significant correlates of a lead-in program's popularity, they were sex of a program's lead character (r-—.28) and the time a program started (-.26). Both of these variables were negatively related to program popularity for these school children (a negative correlation for sex of lead character indicates higher program popularity for programs with female lead characters). The first correlation suggests that shows with female lead characters had stronger lead-ins. The second suggests that later programs had weaker lead-ins than earlier ones. This latter result is of course due to the decline in audience size for later programs. Hence, the popularity of a program's lead-in is necessarily confounded with the time a program starts. The regression analysis suggests that when the effects of a program's starting time is con- trolled, the popularity of a lead-in has minimal if any effect upon the p0pularity of prime time programs. Analysis of the correlations between the popularity of Saturday morning programs and their lead-ins is reported in Table 14. There was a significant correlation between these two variables for sixth grade viewers (.38) in the Fall. In the Winter significant corre- lations were obtained over all subjects (.43), for fourth graders (.39), for boys (.38), and for girls (.41). Multiple regression analyses yielded one case where the variable contributed uniquely and signifi- cantly to program popularity. The popularity of a program's lead-in 77 was a significant predictor of the program's own popularity (B=.42) for boys in the Winter (Table 17). An increase of one point in the popu- larity of a Saturday morning program's lead-in was estimated to result in an increase of .42 points in a program's own popularity for these respondents. Based on the analysis of lead—ins, there appears to be minimal support for the inheritance effect. The correlations between prime time program popularity and the popularity of following shows are reported in Table 10. A significant correlation (.26) between these two variables was found for eighth graders in the Fall. In the Winter, the variables were related signi- ficantly for girls (.30). In the Spring the correlation was significant over all subjects (.38). The pattern held across all grade levels with the correlation for fourth graders being .41, for sixth graders .32, and eighth graders .28. The correlation for boys was not significant while it was for girls (.46). Multiple regression analysis failed to uphold these findings. The effects of a following program's popularity did not remain when other variables were controlled. Especially problematic was program length. This variable was significantly (—.36) correlated with rating of the following program. Longer programs had following programs which were less popular than shorter programs. Another way of presenting the interpretation of the correlation is that programs which followed longer programs were less popular than programs which followed shorter programs. This makes senSe since programs which. followed longer programs appeared later in the evening than did programs which followed shorter programs. Again, the time a program started negatively affected program popularity. It appears that when a variable is significantly related to prime time program popularity it is also 78 related to the time a program starts. The correlations for the popularity of following programs and Saturday morning program popularity are in Table 14. Only one signifi- cant correlation emerged. The popularity of a program was significantly related to the popularity of the following program for girls in the Spring (r=.44). Again, multiple regressions failed to bear this out (Table 17). An analysis of the correlates of following program popu- larity indicated that it was significantly related to the control variable which was established to assess the impact of network. Fol- lowing program popularity was significantly correlated with ABC (-.52) and CBS (.53). This is due to the greater popularity of programs on CBS and the lesser popularity of ABC Saturday morning programs. This trend appeared to be stronger for girls than for boys. Program popu- larity for boys was correlated -.21 with the ABC variable and .30 with the CBS variable. Program popularity for girls was correlated -.43 with the ABC variable and .58 with the CBS variable. Obviously the differ- ence was much larger for girls. Including these network related variables as controls appears to have reduced the relationship between program popularity and the popularity of following programs. Therefore, the effect of following programs which emerged in the correlational analyses were confounded by the large differences between CBS and ABC in terms of the relative popularity of their Saturday morning programs, especially for girls. The only possible source of support for Hypothesis 5a, was the emergence of lead-in program popularity as a significant predictor of Saturday morning program pOpularity for boys in the Winter (Table 17). The set of analyses do not support Hypothesis 5b. 79 Hypothesis Q, The hypothesis stated that the correlation between the popularity of adjacent programs is contingent upon whether they are of the same type. The relationship was posited to be higher when adjacent programs were of the same type. All programs were divided into two groups. One group contained all programs that were preceded by programs which were identical to themselves. The correlation between the popularity of these programs and their lead—ins was .31. The other group contained all programs which were preceded by programs which were different in type from themselves. The correlation between these pro- grams and their lead-ins was .19. The difference between these corre— lations was not significant. Hypothesis 6 was not supported. Inheritance effects summary. Hypotheses 5 and 6 concerned with inheritance effects were not supported. Failure to support Hypothesis 5 appears to be due to the relationship between program popularity and starting time. When significant correlations between the popularity of adjacent programs emerged inclusion of a program's starting time almost always attenuated the relationship. The effects of adjacent program popularity and starting time were confounded in the analysis of prime time programs. Where starting time was not related to program popularity (Saturday morning) other factors emerged to confound the relationship between the popularity of adjacent programs and a program's popularity for schoolchildren. Block Prggramming Hypothesis 2, Hypothesis 7 stated that programs which are adjacent to programs of the same type are more popular than programs which are adjacent to programs of a different type. Similarity between a pro- gram and its lead-in were assessed as were the effects of similarity 80 of a program and that which followed it. Whether a program's lead-in was of the same type as it was found to be a significant correlate of prime time program popularity in only two instances. In the Spring the correlation was significant for sixth graders (r=.26) and girls (r=.28) (Table 10). Regression analyses failed to support either the correlation for grade (Table 12) or girls (Table 13). The variable was significantly correlated with program length (-.55). Programs which had identical lead-ins were shorter than programs which had lead-ins which differed from themselves. Entering both into the regressions would attenuate their unique effects on pro- gram popularity. Given earlier results, the programs which were shorter (which were more likely to have lead-ins which were identical to them- selves) should also have following programs with higher ratings. This is because programs which follow shorter programs are on earlier in the evening and benefit from the larger audience. This indeed was the case (r=.43). Again, the effects of time are confounded with the effects of other variables when assessed together in the regression equations for predicting prime time program popularity among schoolchildren. Saturday morning program popularity was significantly correlated with whether a program's lead-in was the same as it, but only in the Fall (Table 14). The correlation based on all respondents was .48. Correlations by grade were .49 for fourth graders, .41 for sixth graders, and .42 for eighth graders. Correlations by sex were .45 for boys and .45 for girls. Regression analyses failed to bear out these results. None of the regressions yielded a significant regression coefficient for this variable in the Fall, or at any other time (Tables 15, 16, 17). While having a lead-in which was identical to the show itself was 81 positively related to the control variable length (.42) this did not seem to cause attenuation since length was not significantly correlated with program popularity. Having an identical lead—in was negatively correlated with whether the program itself was opposite programs dif- ferent from itself (-.39). However, this variable was not significantly related to program popularity either. Some higher order combination of variables attenuated the relationship between having an identical lead- in and program popularity. Pp§£_hgg_analysis via stepwise multiple regression revealed that this variable did enter the equation early but that its effects diminished gradually as other variables entered the equation. The stepwise analysis (Table 23), therefore, supports the contention that some set of variables attenuated the unique contribution of having a lead-in which is identical to the program itself. However, due to the gradual decline in the effect of this variable no small set of predictors can be identified as producing the attenuation. Given that no identifiable variable or set of variables can account for the relationship between Saturday morning program popularity and having an identical lead-in, the data lends partial support to the hypothesis that having an identical lead-in positively effects a program's popularity among schoolchildren. Whether the program which followed a program was of the same type as the program itself was found to be a significant correlate of prime time program popularity in every case (Table 10). The correlations ranged from .39 to .19. Again, however, the regression analyses failed to support the importance of this variable as a predictor of program popularity. Having a following program be the same type as the program itself did not produce a significant regression coefficient for the 82 prediction of program popularity across all subjects (Table 11), for any age group (Table 12), or for any sex group (Table 13). Analysis of the intercorrelations among the independent variables indicated that at all three seasons programs which were followed by same type programs tended to be shorter and were more likely to be spinoffs than others. Follow- ing programs also tended to be returning rather than new (In the Fall being followed by same type programs was correlated -.35 with same pro- gram length, .25 with spinoff and -.48 with whether the following pro- gram was old. In the Winter these correlations were -.37, .26, and -42. In the Spring they were -.29, .27, and -.37). Spinoff was a correlate of program popularity and could contribute to the attenuation of the effect of having a same type following program on program popularity (see Table 10). Whether the following program was returning or not did not emerge as a significant correlate of program popularity (Table 10). The control variable of program length, however, was consistently a correlate of program popularity. In the Fall it was correlated -.28 with program popularity, in the Winter the correlation was -.36 and in the Spring the correlation rose to -.41. Apparently, the inclusion of this control variable attenuated the relationship between program popu- larity and whether the following program was the same type as the program itself. Saturday morning program popularity was not significantly corre- lated with whether the program was followed by a same type program (Table 14). Multiple regression analysis did yield one significant regression coefficient for the predictor variable but it was in the direction opposite to that hypothesized (Table 17). The analysis indi— cated that programs which were followed by like type programs were less 83 popular than programs followed by different type programs. Hyppthesis Z_summary. Having a lead-in of the same type as a pro— gram itself appeared to enhance Saturday morning program popularity among schoolchildren in the Fall. This was the only example of support for block programming by matching a program's type to its lead-in. Matching a program's type to that of its following program was corre- lated with prime time program popularity in every case. However, regres- sion analysis failed to yield significant regression coefficients for this variable's prediction of prime time program popularity. This sug- gested that the variable was related to other variables in the equation. Analysis of the zero order correlation among the predictor variables in the regression equations indicated that these other variables were whether the program itself was a spinoff as well as the program's length. Hypothesis 8, This hypothesis stated that programs which are ad- jacent to returning programs are more popular than programs which are adjacent to new programs. The effect of lead-in returning and following returning programs were analyzed separately. Neither variable emerged as a significant correlate of program popularity in prime time (Table 10) or Saturday morning (Table 14). Neither did they ever emerge as significant predictors of program popularity in the multiple regression analyses. Hypothesis 8 was not supported. Sex gf_Lead Character Hypothesis 13, The hypothesis predicted that boys prefer programs with male leading characters. The sex of character variable was coded with male characters assigned a value of one and females a zero. Sex of lead character was not a significant correlate of prime time popularity for boys in this sample (Table 10). Multiple regression analysis for 84 program popularity among these boys also failed to yield any significant relationship (Table 13). Sex of lead character was a significant corre- late of Saturday morning program popularity for boys in the Spring (r=.35) (Table 14). The correlations observed in the Fall (.28) and Winter (.29) were not significant. Multiple regression analysis found that sex of leading character was a significant predictor of Saturday morning program popularity for boys in every season (Table 17). Fall programs with male leading characters were expected to have a popularity score 26.98 points higher than Fall programs with female leading charac- ters, for boy respondents. In the Winter the expected difference was 18.9 points. The expected difference in the Spring was 17.37. Hypoth- esis 13 was partially supported. Hypothesis 14, The hypothesis stated that girls prefer programs with female leading characters. The sex of character variable was coded with male characters assigned a value of one and females a zero. The hypothesis predicts a significant negative correlation. This signifi- cant negative correlation was observed in the Winter (-.35) and in the Spring (-.43) for prime time shows. The correlation between sex of lead character and prime time program popularity in the Fall was -.17 but not significant (Table 10). However, regression analysis for the same popularity data failed to yield a significant regression coefficient for sex of leading character and the other variables in the equation indicated that programs which had female lead characters were more likely to be spinoffs (r=-.18) and were shorter (r=.18). Both of these variables were significantly related to prime time program popularity for girls. The correlation between popularity and Spinoff was .30 in the Winter and .27 in the Spring (Table 10). The correlation between 85 program length and popularity for these respondents was —.40 in the Winter and —.39 in the Spring. All three of these predictor variables were intercorrelated and significantly related to program popularity for girls. Including all three in the regression equations attenuated the unique predictive ability of each. Since including other variables reduced the relationship pp§£_hgg_stepwise analysis was conducted to determine which of the variables entered the regression equation. The results indicated that both sex of lead character and program length entered in the Winter. Only sex of lead character entered the equation in the Spring. Hence, the most powerful single predictor among the three variables, controlling for the effects of those others in the equation, was sex of lead character. Sex of lead character was not a correlate of Saturday morning program popularity for girls (Table 14). Regression analyses for the same data did not yield any significant regression coefficients for prediction of program popularity by sex of lead character for girl respondents (Table 17). Based upon the preceding discussions, Hypothesis 14 is partially supported. Starg Time Hypothesis 18. The hypothesis stated that program popularity is negatively related to the time a prime time show starts. Start time was a significant and negative correlate of prime time program popularity in every case (Table 10). The correlation ranged from «.52 to —.28. As ‘ would be expected, the size of the correlation was negatively related to grade of respondent. Regression analysis indicated that starting time was a significant predictor of program popularity for fourth 86 graders in the Fall (Table 12). This was the only case in which starting time yielded a significant regression coefficient. In the Fall starting time was significantly correlated with program length (.23); later programs were longer and tended to have lower ratings. The inclusion of the length variable appears to be principally responsible for attenuation of the effect of starting time as a predictor of program popularity. In the Winter another significant correlate of program popularity, whether the following program was the same type was also significantly correlated to starting time (-.22). Programs which appear— ed later in the evening were more likely to be followed by different type programs than were programs which were on earlier in the evening. This relationship may have been sufficient to attenuate the predictive ability of starting time in the Winter regression equations. Starting time was correlated to two significant correlates of program popularity in the Spring. They were counterprogramming by type (-.27) and the popularity of the program's lead—in (-.26). Insertion of the three variables into a single regression equation attenuated the unique impacts of each, and none of them emerged as a significant unique predictor of program popularity. Nevertheless, the size and consistency of the correlations coupled with the unique effect of starting time for fourth graders supports the hypothesis. The relationship between audience size and start time suggests that the popularity of a program's lead-in is attributable to the time the lead—in program starts. The later it starts the smaller the audience. Hence, the logical choice between the two variables is start time rather than lead-in rating. The decrease in a program's popularity cannot cause it to appear later. On the other hand, it is well known that later programs have smaller audiences. 87 Hypothesis 18 is supported. Spinoffs Hypothesis 12, The hypothesis stated that programs which are spinoffs are more popular than programs which use only novel characters. Being a spinoff was positively correlated to prime time program popu- larity across all subjects in the Fall (.21) and Spring (.24) but not in the Winter (.19). It was a significant correlate for fourth graders (.28) and sixth graders (.25) in the Spring only. Spinoffs were more popular to eighth graders in the Fall (r=.26). Spinoffs never emerged as a significant correlate of prime time program popularity for boys. However, it was a significant correlate of program popularity for girls in the Fall (.31), Winter (.30) and Spring (.27). Regression analyses of the same data failed to yield a single significant regression coef- ficient for the spinoff variable. Its effect was confounded with others. In the Fall it was found to be related to other correlates of prime time program popularity. These were program length (-.29) and whether a pro- gram was followed by a same type program (.25). In the Winter, the following significant correlates of prime time program popularity were also significantly related to spinoff: length (-.29), sex of lead character (-.18) (spinoffs were more likely to have female lead charac- ters than other programs), and whether a program was followed by a same type program (.26). The same three variables were significant corre- lates of program popularity and were also correlated with spinoff in the Spring. The correlations for each were: length (.29), sex of lead character (-.18) and whether a program was followed by a same type pro- gram (.28). Inclusion of these variables in the same regression equa- tion reduced the contribution for the variable designed to measure the 88 impact of spinoffs. Discussions accompanying the analysis of the impact of sex of lead character indicated that sex of lead character and pro— gram length were stronger predictors of program popularity than was the spinoff variable. The latter variable appears to be the least useful of the set. Hypothesis 19 was not supported. Returning Programs Hypothesis 29. The hypothesis stated that returning programs are more popular than new programs. Whether a show was returning or not was a moderate but consistent correlate of prime time program popularity among schoolchildren in the Fall. The correlations were significant across all analyses and ranged from .51 to .55 (Table 10). Significant correlations also emerged in the Winter but they were considerably lower and did not emerge in all analyses. The correlation across all respon- dents was .23. For girls and fourth graders it was .28. No significant correlations emerged in the Spring. These correlations ranged between .05 and .11 (Table 10). Multiple regression analyses of the same data corroborated the correlational analysis. Returning programs were more popular than new programs in the Fall. Significant regression coefficients emerged in every regression analysis of Fall prime time program popularity. Re- turning programs were predicted to have an advantage of 25.3% over new programs (Table 11). The advantage for returning programs was about the same for sixth (23.7) and eighth graders (23.4). The advantage for fourth graders was slightly higher (29.5) (Table 12). The difference for boys was 24.6% while the difference for girls was 28.6% (Table 13). Analysis of the popularity of Saturday morning programs yielded similar results. Correlations in the Fall were all significant. They 89 ranged from .51 to .67. In the Winter the correlations were somewhat lower and fewer of them were significant. Significance was obtained for the relationship beween returning vs. new programs and Saturday morning program popularity for all subjects (.46), for fourth graders (.37) and for girls (.36). In the Spring, returning programs were more popular among fourth graders only. The correlation was .36 (Table 14). Returning Saturday morning programs were estimated to have a 25.4% larger audience than new programs in the Fall (Table 15). Table 16 indicates that this was far higher for fourth graders (39.66). The advantage of these programs for sixth graders was 24.6%. For eighth graders the advantage of returning programs was smaller (15.3) but still significant. The popularity of returning programs for fourth graders was also higher than that for new programs in the Winter (14.8) (Table 16). Multiple regression analyses for the popularity of Saturday morning programs for boys indicated that returning programs advantage decreased as the season wore on but was still significant in the Spring. The advantage of returning programs for this group was 23.2% in the Fall, 16.1% in the Winter and 12.9% in the Spring. The advantage of returning programs for females only appeared in the Fall. The advantage was 24.4%. Hypothesis 20 was supported. 90 Programming Strategies Summary Table 18 summarizes the results of the analyses regarding the hypotheses concerning programming strategies. Table 18 Programming Strategies Summary Hypothesis Support Yes Part No 3--Counterprogramming by Type X 4--Counterprogramming by New vs. Returning X 5--Inheritance Effect for Program Popularity X 6--Inheritance Effect is contingent upon Adjacent Programs being the Same Type X 7--Block Programming by Type enhances Program Popularity X 8--Block Programming (being adjacent to returning pro- grams) enhances Program Popularity X 13--Boys Prefer Programs with Male Leads X 14--Girls Prefer Programs with Female Leads X 18--Later Start Time reduces Prime Time Program Popularity X 19--Spinoffs are more popular than Other Programs X 20--Returning programs are more popular than New Programs but the advantage dissipates over time. X Popularipyppf_Program Types py_ Sex and Age pf_Respondent Hypotheses 9-12 predicted that particular types of programs would be more popular for one sex group than another. The results of the analyses of these hypotheses are reported in Table 19. Hypotheses 15- 17 predicted age related popularity differences by program type. Tests of these hypotheses are reported in Table 20. 91 Program Type Popularity py_Sex of_Respondent Table 19 Program Type Popularity by Sex Program Type Prime time: Boys Girls Action Adventure (n=29) 32.66 27.44 Drama (n=16) 21.92 29.48 Feature Film (n=12) 17.12 14.06 Situation Comdedy (n=30) 39.82 43.30 Sports (n=3) 50.75a 9.75a Variety (n=7) 31.62 41.88 Saturday morning: Animated Adventure (n=4) 40.97 33.97 Animated Comedy (n=14) 41.55 41.30 Nonanimated Adventure (n=6) 39.34 41.60 Nonanimated Comedy (n=4) 45.17 35.68 Note--Means with same subscript in'a row are significantly different from one another (p(.05). Table 18 reports the mean popularity of each program type by sex of respondent. Hypothesis 9, which stated that action-adventure shows are more popular among boys than girls was not supported. Hypothesis 10, which stated that situation comedies are more popular among girls than boys was not supported. Hypothesis 11, which predicted that variety programs were more popular among girls than boys was not sup- ported despite the fact that the difference was over 10% and in the hypothesized direction. Only Hypothesis 12, which predicted that sports programs were more popular among boys than girls, was supported. The observed difference in popularity was 41% (t=7.39, df=4). 92 Program Type Popularity_py_Age.of_Respondent Table 20 Program Type Popularity by Grade Program Type Grade 4th 6th 8th Prime time: Action Adventure (n=29) 31.58 28.82 27.72 Drama (n=16) 31.22 24.00 24.15 Feature Film (n=12) 14.60 14.87 18.11 Situation Comedy (n=30) 40.93 40.80 39.41 Sports (n=3) 26.00 33.50 33.75 Variety (n=7) 37.84 36.79 34.43 Saturday Morning: Animated Adventure (n=4) 49.23a b 37.48 23.63a Animated Comedy (n=14) 54.463: 39.548 28.19b Nonanimated Adventure (n=6) 47.88a 40.99 30.548 Nonanimated Comedy (n=4) 53.678 38.778 26.90a Note--Means with same subscript in a row are significantly different from one another (p<.05). Table 20 reports the mean popularity of each program type by grade of respondent. Hypothesis 15, which stated that older respondents prefer action adventure programs more than younger subjects was not sup- ported. Hypothesis 16, which predicted the same pattern for feature films was, likewise, not supported. Hypothesis 17 predicted that Saturday morning program popularity was negatively related to age of viewer. This hypothesis was supported. Saturday morning adventure shows, whether animated or nonanimated, were significantly more popular among fourth graders than eighth graders. Animated comedy shows were more popular among fourth graders than sixth or eighth graders. Non- animated comedy shows were more popular among fourth graders than sixth graders and more popular among sixth graders than among eighth graders. 93 Changes Over Time This section presents the results concerning the effects of new vs. returning programs over time. Two of the three related hypotheses were not testable since the main hypotheses were not supported. Hypoth- esis 4a was not tested since Hypothesis 4 was not supported. Hypothesis 4 predicted that program popularity is negatively related to the number of returning programs opposite a program. Hypothesis 4a predicted that this effect would decrease over time. Hypothesis 8a was not tested since Hypothesis 8 was not supported. Hypothesis 8 stated that programs which are adjacent to returning programs are more popular than programs adjacent to new programs. Hypothesis 8a predicted that this effect would diminish over time. Hypothesis 20a, however, was testable since its main hypothesis was supported. Hypothesis 203 stated that the advantage of being a returning rather than a new program would diminish over time. Table 21 reports the results of the analysis of Hypothesis 20a. The results indicated that, for both prime time and Saturday morning shows, the advantage held by a returning program was signifi- cantly greater in the Fall than in the Winter or Spring. The advantage of returning programs is fairly well dissipated by Winter. Hypothesis 20a was supported. 94 Table 21 Differences in the Popularity of Returning vs. New Programs Over Time Daypart Time Fall Winter Spring Prime Time 25.273"b 7.91a 5.96b Saturday Morning 25.448’b 7.778 7.72b Note--Cell entries reflect the advantage of returning over new programs based upon the unstandardized regression coefficients. Entries with the same subscript in a row are significantly different (p<;05). Post Hoc Analyses Stepwise Regression Stepwise multiple regression analyses were conducted in order to identify the most parsimonious set of predictors of program popularity. Separate analyses were conducted for prime time and Saturday morning. Prime time. Stepwise analyses across all respondents were conducted for Fall, Winter, and Spring program popularity. Results of these anal- yses are reported in Table 22. In the Fall three variables were observed to be significant predictors of program popularity. They were: whether a program was returning, its starting time and its length. These three variables accounted for an adjusted R2 of .47. This figure was significant (F=8.59, ii?3,23). The use of adjusted R2 rather than R2 as the estimate of variance explained is justified because the latter is a more biased estimator of variance explained in the population, especially when the number of variables is large relative to the number of programs in the analysis. 95 Stepwise Regression Analysis for Predicting Prime Time Program Popularity Predictor Sgggop Fall Winter Spring Returning vs. New Program 24.03 --- --- Starting Time3 —7.81 —11.46 -8.48 Length -12.59 -16.57 Constant 39.97 44.05 53.65 R2 .528 .207 .309 Adjusted R2 b .468 .171 .236 F (df) 8.59 5.73 4.24 (3,23) (1,22) (2,19) Note—-Since the criterion for entry into the stepwise regressions was significance at the p .05 level all entries in the table are significant. a Start time for prime time shows was measured as.a deviation from 8 P.M. in hours. Hence, 8:30 = .5, 9:00 = 1.0, etc. k—l b Adjusted R2=R2- N:k'(l-R2), where k = the number of predictors in the equation and N = the smallest number of cases (programs) for any correlation upon which the regression is based. This N equals the total degrees of freedom plus one. In the Winter stepwiSe regression starting time emerged as the only significant predictor of prime time program popularity. The ad- justed R2 was .17 and significant (F=5.73, df?1,22). In the Spring both starting time and length emerged as significant predictors. Adjust- ed R2 was .24 and significant (F=4.24, df§2,19). 96 Table 23 Stepwise Regression Analyses for Predicting Saturday Morning Program Popularity Predictor Season Fall Winter Spring Lead-in same 13.42 Returning vs. New Program 17.77 13.84 Constant 26.41 30.55 R2 .523 .210 Adjusted R2 a .463 .164 F (df) 8.77 4.52 (2,16) (1,17) Note—-Since the criterion for entry into the stepwise regressions was significance at the p .05 level all entries in the table are significant. 12:1 a Adjusted R2=R2- N-k (l-RZ), where k = the number of predictors in the equation and N = the smallest number of cases (programs) for any correlation upon which the regression is based. This N equals the total degrees of freedom plus one. Saturday morning. Stepwise analyses over all subjects were con- ducted in the Fall, Winter, and Spring for Saturday morning program popularity. Results of these analyses are reported in Table 23. In the Fall two variables entered into the equation. They were whether the lead-in program was the same type as the program itself and whether the program itself was returning from last season. The adjusted R2 for the equation with these two variables was .463 and significant (F=8.77, dfé2,16). In the Winter, the only variable to enter was whether the program was returning. Adjusted R2 was .164 and significant (F=4.52, 97 dfyl,l7). No variables entered the stepwise analysis in the Spring. Stepwise regression analysis summary. The stepwise analyses tended to yield a greater number of significant predictors of program popularity than did the earlier standard multiple regression analyses. This is attributable to intercorrelations among the predictor variables used in this study. No single set of variables emerged from the stepwise anal- yses of prime time and Saturday morning program popularity. However, consistency was found regarding the effect of returning vs. new program- ming. Once again, the variable was found to have a significant impact on program popularity in the Fall. Start time appeared to be an impor- tant predictor of prime time program popularity, as did program length. In general, both analyses indicated that the variables employed in this study were better predictors of program popularity for younger rather than older viewers. Factor Analyses Prime time. A post hoc factor analysis of prime time viewing was conducted in order to identify potential program types for schoolchildren. Five factors were extracted which accounted for 37% of the total vari- ance. The program groupings and the percentage of common variance explained by each are presented in Table 24. Dramatic programs appeared to cluster regardless of whether they were detective, medical or police drama. The second group were shows that, for one reason or another, were more popular among girls than boys. The third factor indicated that the viewing of some shows on ABC were negatively related to viewing family dramas. Factor four consisted of sitcoms with the exception of Cannon. The final group contained programs with characters of power or strength. 98 Table 24 Prime Time Program Groupings Obtained via Factor Analysis CBS/Popular ABC vs. Among Family Drama (42%) Girls (23%) Drama (17%) Comedy (10%) Power (9%) Barnaby Jones Carol Burnett Baretta All in the Bionic Woman Emergency Mary Tyler Moore Happy Days Family Happy Days Joe Forrester Bob Newhart Welcome Back, Cannon Six Million Harry-O Phyllis Kotter Chico & $ Man Hawaii S-O Rhoda Little House the Man SWAT Klute on the Doc Kojak Prairie (-) Jeffersons Medical Center Barney Miller MASH Police Story Waltons (-) Sanford & Police Woman Son The Rookies Switch Marcus Welby Note--All loadings were positive except as indicated. Saturday morning. A factor analysis of Saturday morning viewing yielded four factors which explained 42% of the total variance. Program groupings and the percentage of common variance explained by each are presented in Table 25. Table 25 Saturday Morning Program Groupings Obtained via Factor Analysis Early ABC + Non- Animated Comedyp(62%) Adventures of Gilligan Far Out Space Nuts Ghostbusters Hong Kong Phooey Lost Saucer Oddball Couple Tom & Jerry/ Grape Ape _NBC (19%) Josie & the Pussycats Land of the Lost Return to the Plan— et of the Apes Westwind Animated (11%) Shazam-Isis Bugs Bunny/ (9‘) Road Runner Isis Emergency Plus 4 Shazam Pebbles & Bamm- Bamm Scooby-Doc Speed Buggy CHAPTER IV DISCUSSION, CONCLUSIONS, LIMITATIONS AND IMPLICATIONS This chapter concludes the study. The first section reviews and interprets the findings of the study. The second section summarizes the findings into a set of general conclusions. Part three discusses the limitations of the study. The final section discusses the study's implications for programming and further research. Discussion The Dependent Variable--Program Popularity The checklist based measure of program popularity appeared to relate to other measures of viewing and behaved as would be expected from earlier research. The data reflected the well documented trend for viewing levels to decline in the Spring. A longitudinal interpre- tation of the crossectional data reflects a trend for viewing levels to drop as children become teenagers. This pattern was marginally reflected in prime time but was strongly indicated for Saturday morning programming. This last finding does not correspond precisely with data presented by Schramm, Lyle and Parker (1961). These authors reported that viewing increases until about sixth grade and then declines. The present study indicates that the age at which peak viewing occurs may have changed since fourth graders reported the greatest number of regularly viewed prime time shows (as reflected by the fact that their mean program 99 100 program popularity score was higher than that of the two other groups). However, this result may be attributable to this study's finding that program popularity is more stable for these younger viewers. While they may actually watch less TV than older viewers, they may have a larger set of regularly viewed programs. This would result in higher estimates for the present study (as compared to Schramm 35.213: 1961) since the present study's questionnaire asked the respondents to check off only those shows which were viewed with regularity. The analysis of nonviewable shows provided mixed results concernw ing validity. The program which had not been on the air (City of Angels) had an extremely low reported incidence of viewing. Similar results emerged for "Winky Dink," a program that went off the air several years ago. However, other nonviewable shows, which were more familiar to respondents, had higher reported viewing levels. For example, the reported popularity of the program "Big Eddie" was almost identical for all three age groups in the Fall, before it was cancelled. Despite this equality in the Fall, considerable inequality emerged in the Winter, after the show had been off the air for months. Analysis by grade indicated that fourth graders reported viewing these shows far more than did the other grades. There may be two reasons for this problem. First, the younger respondents might have been more likely to forget or ignore the instructions and check off familiar or preferred programs, rather than ones they actually viewed. While the questionnaire clearly stated, and the administrators stressed 101 the desire for information on actual viewing, the instructions might not have been clear, were forgotten or ignored by the time respondents moved part way through the checklist. Another explanation considers the fact that the time frame provided to respondents was ambiguous. Respondents were asked to check off those programs viewed every week or almost every week. Explicit information concerning the time frame was not provided. While the ' administrators told the respondents to think back over the "past few weeks," this frame may not have been adequate or clear to the respondent. Hence, programs which had been regularly viewed earlier in the season might be checked off on a later questionnaire. This problem may have been strongest for fourth graders because their conceptions of time may not be fully developed. They are less likely to have entered what Piaget (1952) has labeled the "formal operations" stage of development. It is at this stage where the capability for dealing with abstraction emerges. For these children, the phrase "last few weeks" may include a time span of months. Alternatively, the children may recall viewing a program but not how long ago it was viewed. To play safe, they would check it off. The mean popularity score across all programs for fourth graders was almost three times as high as the mean Nielsen Rating for the same set of programs for 6-11 year olds. The ratio was closer to 2:1 for sixth.graders with the same Nielsen group. A comparison of the reported viewing of sixth and eighth graders with.the Nielsen 12—17 year old sample yielded a 3:1 ratio. These differences were expected since 102 Nielsen's diaries report the viewing of a single episode while the present data are cumulative over a period of weeks. Goodhardt 25 El- (1974) indicate that on the average, 55% of those who view an episode of a program one week will watch it any other week. Hence, the present study's popularity estimates should be higher than that of Nielsen. It would be quite disturbing had the difference not emerged. The most crucial test of the two sets of data are their intercorre- lations. The correlations between the checklist and Nielsen data were all positive and patterned as expected. Fourth graders correlated most highly with Nielsen's 6-11 year olds (r=.36), eighth graders with Nielsen's 12-17 year olds (r=.60), and sixth graders about equally with both of Nielsen's groups (r=.48, .41). The correlation for the youngest age group (.36) was not high, suggesting some discrepancies. However, on the whole, the correlations were about as high as could be expected given the differences in measurement and the samples employed. The last bit of information relevant to quality of measurement is the reliability of the study's measure of program popularity. Reliabil— ity, in this case, is the correlation between the observed popularity of a program and the popularity of the program corrected for measurement error. Reliability estimates were .90 or higher. The evidence indi- cated high reliability for program popularity across all grades. In summary, reliability for the measurement of program popularity was strong. Minimal problems were suggested by the reported viewing of nonviewable shows by fourth graders. This problem appeared to be highest for programs which respondents could have previously viewed. On the other hand, the popularity measure reflected well known patterns both seasonally and longitudinally, and corresponded as well as expected with 103 national Nielsen audience estimates. Stability. The stability estimates are essentially correlations between two sets of data collected at two points in time. The difference between the stability estimates and common test-retest procedures is that the stability figures are correlations between program popularity scores which have been corrected for measurement error (reliability). Differences between the common procedures and those employed in the present study were minimal since reliability was very good. The stability estimates were quite substantial; .71 between Fall and Winter, .88 between Winter and Spring. However, there appears to be some need to account for the lower correlation between Fall and Winter. This is probably attributeable to the flurry of new programs in the Fall as well as the quick replacement of shows that are at the bottom of the ratings heap. By Winter most of this shuffling has occurred and the programming is probably more stable between the Winter and Spring. Squaring the stability estimates indicates that a considerable amount of variability; 50% between Fall and Winter, 23% between Winter and Spring is attributeable to systematic changes between seasons. While some may be attributeable to programming changes, research has yet to consider the specific elements which result in fluctuations in televi— sion program popularity during a season. In order to account for this variability or instability one must consider seasonal viewing patterns, increasing familiarity with new programs and characters, the number and variety of programs offered at each season and changing tastes as a function of the child viewers maturation during the year, in addition to changes in program schedules. The present study merely described the degree of stability (or instability) in program popularity. Future 104 research should address the factors which result in stability (or instability). Analysis of stability across the three age groups in this study indicated that the difference in stability between Fall and Winter vs. Winter and Spring was greater for younger subjects. While all three groups were about the same between Fall and Winter, the stability coef- ficients rose more for younger groups. Apparently, it takes longer for these younger respondents to become satisfactorily acclimated to the new television season, but once acclimated they are more loyal viewers. While this study's findings support this interpretation of the data, they do not eliminate possible alternative explanations. One such explanation rests on the assumption that the time frame "last few weeks" is more problematic for younger respondents. If so younger respondents may have checked off programs which they no longer viewed but viewed earlier in the season. This would affect later stability estimates more than earlier ones, resulting in some inflation of the youngest respon- dents' Winter-Spring stability estimates. The extent to which this factor affected differences attributeable to age of respondent cannot be determined from the present study. Probably, age related differences in acclimation and loyalty do exist. However, the size of the differences may be overestimated in this study. Programming78trategies By_£ypg. Two program type strategies were tested in this study, they were: counterprogramming by type and block programming by type. Neither strategy was found to be effective. One reason for this study's failure to support the efficacy of these strategies could lie in the typologies utilized. The typology 105 was generated by a review of literature rather than from the respondents themselves. Had a respondent generated classificatory schema been used these strategies might have emerged as important predictors of program popularity. The problem is compounded by the fact that the typology which emerged was primarily based on research with adults. The popu- larity of programs for a younger population of viewers was of concern here. A respondent generated typology might have alleviated these problems but at a cost. The use of existing program typologies is justified when generalizability is unknown or has not been assessed. The present research attempted to apply the prevalent typologies in order to examine their generalizability. Had the commonly used typolo- gies been confirmed as having predictive value for this sample of child- ren, the results would have been far more useful than results derived from a typology generated by and for a relatively small sample. Factor analyses of viewing data were conducted poo; hog. The clusters do not appear to correspond strongly to those employed in the study. Action shows do not seem to have been differentiated from drama shows by these young viewers. In fact, the first dimension is dominated by later prime time shows. Thus, it might be a factor attributeable to ' differing bedtimes. The other factors likewise seem to compound sched- uling variables and types. Similar problems emerged for the Saturday morning analysis. Quite frankly, these results were expected based upon earlier research reported in chapter one of this study. There are too many other factors beside program type and program type preference which influence viewing patterns. Unless these other factors are accounted for, the relationship between progam types and viewing will continue to be unclear. 106 The relationship between program content and viewing requires study. When is preference for a program type strong enough to cause a person to watch television? Some scholars (Bogart, 1972), argue that this is never the case. Rather, they argue that program type preferences only affect program selection after the individual has already decided to view. Others (Steiner, 1952; Besen and Mitchell, 1976), argue that program types may affect the decision to watch television at all. Certainly, when preference for a particular program is strong enough, one may consciously set aside time to view it, as the popularity of special programs such as "Super Bowl," or "World Series" suggest. However, such programs appear to be the exception rather than the rule. No evidence to date has been able to support the contention that preference for a program_£ypg is strong enough to cause viewers to set aside time to view that particular program type, whenever a program of that type appears. Preference for a particular program type is not likely to be a strong predictor of television viewing. It is more likely to affect selection of programs after the viewer has decided to switch on the television. Even in this case, however, several other factors mediate between individual preference for a particular program type and actual viewing. Such factors may include the amount of promotion for the program. The influence of program promotion has yet to be empiri- cally examined. Certainly, promotions influence one's decision to view a program. The influence may be cognitive (existence of the program) and/or affective (viewer may judge the extent to which a program appeals to him or her on the basis of the promotions). Preferences of others who share the television as well as the individual's "mood" at that particular time should also be considered. A useful conceptualization 107 of these mood states might be the gratifications sought from television at the time. Generalized long term preferences may reflect continuing and enduring gratifications sought from media. In order to become more conversant in national affairs an individual may view television news. In order to function socially with others an individual may choose to watch programs which significant others watch. If the kids at school discuss the programs which are on television, a child wishing to become part of the group must watch shows which the others watch. If a certain type of show is preferred by peers, a viewer may select those types of shows over others. Short term gratifications may also be important. The desire for excitement has been shown to be related to the viewing of particular program types, specifically those which have violent content (Greenberg, 1974). These short term gratifications probably interfere with viewing patterns which would be dictated by long term viewing preferences. In other words, a person may forego his or her usually preferred program type if they crave stimulation and the normally preferred show does not usually result in this effect. Other types of descriptors of programming may also influence selection of a program. For example, sex of leading character appeared to influence the popularity of programs for boys and girls. However, the pattern was not consistent. Females preferred programs with female leading characters in prime time but not on Saturday mornings. Males preferred programs with male leading characters on Saturday morning but not in prime time. Again, some other factors must be influencing same sex preferences such that these patterns emerged. The interaction be- tween the audience, descriptors of programs such as type and sex of lead 108 character, as well as scheduling and programming factors must be consid- ered in order to make sense of their individual contributions to viewing patterns and program popularity. Watt and Krull (1974) suggested another alternative to the use of program type categories as program descriptors. They proposed that the structural or form.characteristics of a program have an effect on view- ing. In their study both form and program type were employed for pre- dicting nonrandom viewing patterns. Both were successful at predicting deviations from randomness. The study is important because it suggests that noncontent factors influence viewing choice. Further research is needed to determine the extent and manner in which form and content (or program type) interact, and operate with scheduling factors for predict— ing viewing and program popularity. Contrary to Watt and Krull's (1974) finding concerning the impor— tance of program types on viewing patterns is that of Goodhardt g£_al, (1974). They suggest that the principal value of program types related to content is in the prediction of appreciation or liking of programs. While appreciation may affect viewing, the research reported by Goodhardt EE El: (1974) indicates that if such a relationship exists, it is very weak. Program types have emerged for appreciation. People do say they like programs of a similar type. However, their research indicated that these preferences had virtually no effect on actual viewing behavior. A resolution of the differences between Goodhardt SE al. (1974) and Watt and Krull (1974) is necessary. Differences do appear between the two sets of research in terms of respondent samples and data collection procedures. The Watt and Krull (1974) findings are based upon adolescents who reported the frequency 109 of viewing particular programs via recall. The Goodhardt §£_§l,.(1974) research is based on adult samples who report viewing in diaries. Attributing differences to data collection procedures is possible. In this case one would hypothesize that for some reason, program types related to content are more likely to affect recall of viewing than on- the-spot diary entries of viewing. Such reasons might include the greater possibility for program type preference to interfere with recall processes. Selective recall might emerge with respondents indicating greater viewing for programs of preferred types. Differences might also be attributeable to the different respondent samples used. Perhaps, adolescent's viewing patterns are more easily dictated by program type than adult viewing patterns. One might argue that adults are more like- ly to be influenced by quality of programming regardless of type while youngsters are attracted to specific types of programming like cartoons. Combining the two explanations is also possible. Perhaps there is an interaction between respondent age and method of data collection. Future research must assess the reason for the differences between the Goodhardt ggugl. (1974) findings and those of Watt and Krull (1974) concerning individual viewing patterns. The present study suggests that program types minimally influence program popularity. §y_new vs. returning programs. Counterprogramming and block programming by new vs. returning programs were also considered as programming strategies. It was hypothesized that programs adjacent to returning programs and programs opposite new programs would do better than other programs. Neither strategy was found to be effective in this study. Being opposite one or more returning programs (counterprogramming) 110 did not hurt a program, even in the Fall, when returning programs had a considerable popularity advantage over new programs. This appeared to be a rather puzzling situation. Klein (1971) argued that audience size is amazingly constant at any one time from day to day. Bogart (1972) and Owen 22 al. (1974) have presented data supporting Klein's notion of stability in audience size. If audience size is stable, how can returning shows have an advantage over new ones, and yet not affect the popularity of opposing programs? One reason might be that the constancy phenomenon fluctuates seasonally. While the researchers indicate that constancy is a general trend, less constancy, or stability, may be exhibited in the Fall when acclimation to TV programming occurs. Since no new evidence emerged from this study concerning the cOnstancy of audience size during any time slot, the above explanation must be viewed as largely speculative. Another explanation involves previously discussed measurement problems. Returning programs might have been viewed during their summer reruns making them likely to checked off in the Fall. The nonspecifi— city of time span and problems involving familiarity could only help the popularity of returning programs. It should be noted that these advantages would be observed without regard to a programs' current competition. A returning program would receive inflated popularity scores without detracting from the popularity of it's current competi- tion. Being adjacent to returning programs was likewise ineffective as a unique predictor of program popularity. Even in the Fall when the effect was hypothesized to be strongest, no evidence was found which could support the existence of a relationship between program popularity 111 and whether adjacent programs were new or not. Even the zero order correlations were small suggesting that multicollinearity was not a problem. The advantages thought to accrue to programs which were adjacent to returning programs simply did not emerge in this study. SummaryeCounterprogramming and Block Programming. The results of this study did not support the efficacy of either strategy. Counter- programming or block programming by new vs. returning was not successful despite the fact that returning programs were more popular than new ones in the Fall. Counterprogramming by type was likewise ineffective in influencing program popularity. Block programming by type did appear to receive partial support, but the only clear instance of support was in the Fall, for Saturday morning programs. This emerged despite apparent differences between the typology employed in this study and that derived from the actual viewing patterns of respondents. The pre- ceding discussion suggested that a great deal of work is necessary before a clear typology of television programs emerges which would effectively predict program popularity and viewing patterns, at least for the age group presently studied. It was suggested that other descriptors of programming such as sex of lead character be considered when analyzing the content of television. In addition to program type, form and other program descriptors it was suggested that research be conducted which considers audience descriptors and scheduling factors. Audience Descrippors The audience was analyzed regarding differing preferences by age and sex. The results of these analyses indicated that only one type of program, "Sports,' was differentially preferred by sex. Boys preferred these types of programs far more than did girls. It should be stated 112 however, that the number of programs employed in these analyses resulted in extremely conservative tests of the hypotheses. For example, "Vari- ety" programs were 10% more popular among girls than boys, yet, because there were only seven such shows, this difference was not large enough for statistical significance. In part, this is also due to the large amounts of variability around the respective means. Inspection of the mean popularity of each program type yielded large standard deviations. Usually the standard deviations were at least half as large as their corresponding means. Similar ratios of popularity to standard deviation were observed without distinquishing among program types. Stratifi- cation of programs by type only slightly reduced the amount of error variance in program popularity. The existence of such large standard deviations indicates that large differences in program popularity within each type existed. Consequently, the typology employed in this study was not successful in accounting for variance in program popularity among schoolchildren. It may be more successful, however, once other factors which influence program popularity are isolated. One factor whiCh might contribute to variability within program type is the previously discussed program structure or form (Watt and Krull, 1974). Variation within each program type may be attributeable to differences in form. These researchers have isolated two dimensions of form labeled "Dynamics" and "Unfamiliarity." The dimension labeled "Dynamics" is associated with action. The number and randomness of scenes, the number of characters who talk and differences in the lengths of their verbalizations cause variability on this dimension. The "Unfamiliarity" dimension is affected by frequent cutting from indoor to outdoor scences and familiarity with 113 the situations portrayed in programs. Any variability along these dimensions within content category would cause the content categories to be poorer predictors of viewing. Watt and Krull (1974) found that this in fact was the case. Viewing patterns were affected by both content and form. Not controlling the form factors in the present study contributed to variability within each content category. Hence, two paths for research are indicated. First, better identification of program types and second, the inclusion of form variables for predicting variability in program popularity and viewing patterns. Another analysis of the content of programs might stress the characters. Their sex, attractiveness and power might also influence who, and how many, watch a television program. Age related differences in the popularity of Saturday morning pro— grams were found. Popularity declines with age. The sharpest declines were observed for comedy programs regardless of whether they were car- toon or live. This suggests that Saturday morning comedy shows are the first to lose popularity as the child matures. Research discussed earlier in this study suggested that action- adventure and drama programs replace comedy programs as children mature. The results of the present study do not support this contention. The notion of replacement implies that the popularity of action or drama shows should increase as the popularity of comedy or cartoon programs decreases. Reduction in the popularity of all Saturday morning program- ming was observed. No concomitant increase in the popularity of action or drama shows emerged. Therefore, substitution could not have occurred. The data suggest that age related reductions in viewing do not occur uniformly across all program types. Reductions in total viewing 114 appear to be primarily a function of decreased Saturday morning viewing, especially comedy viewing. The popularity of prime time program types did not appear to be influenced by age among these respondent groups. Schramm 35 al. (1961) present exposure data which, for the most part, corroborate these findings. They found that cartoons (most of which are comedy) declined sharply in popularity between fourth and sixth grade. Such programs were even less popular among their eighth graders. They also presented American Research Bureau data for 1954 which indicated that the mean rating of "Kid shows" declined more sharply than any other program type between the ages of nine and 14 (the age range of respondents in the present study). Neither Schramm SE El: (1961) nor ARB distinguished among types of children's shows as did the present study. The present study indicated that while cartoons did drop in popularity, animated comedy dropped more rapidly than animated adventure. Similarly, nonanimated comedy dropped off in popu- larity more quickly than did nonanimated adventure. Hence, the rates of dropoff do not appear to be related to mode of presentation (cartoon vs. noncartoon). The rate at which popularity declined was related to program content with Saturday morning comedy programs declining more rapidly than Saturday morning adventure programs. Predicting Program Popularity. Strong predictors of program popularity for schoolchildren did emerge.- The strongest of these was whether §_program itself was new op returning. New programs suffer a strong disadvantage in the Fall. However, by Winter the new programs which are on the air are as popular as those which returned from the prior season. Sex of lead character also emerged as a predictor of program popularity. Boys preferred 115 Saturday morning programs with male lead characters over program with female lead characters. Girls preferred prime time programs with female lead characters over programs with male lead characters. The existence of strong sags EEE effects for young boys has been observed in several other studies. Such studies have dealt with pre- ferred models (Miller & Reeves, 1975; Greenberg, Heald & Wakshlag, 1976) as well as with viewing preferences (Sprafkin, 1975). Hence, the sur- prising finding is the absence of such effects for prime time programming. Most of the research has dealt with children who were younger than those in the present study. Since the Saturday morning programs are viewed by a greater proportion of fourth graders, one would expect results based on Saturday morning shows to most closely parallel the earlier studies. Others (Grusec and Brinker, 1972) would argue that this trend should get stronger as the child matures. They posit that sex-typing increases with age. Other things being equal, this suggests that the effect of sex of lead character should be greater in prime time, when a greater proportion of a program's audience consisted of eighth graders. The evidence used to support their contention was that seven year old boys learned more from male than female models than did five year old boys. Again, these subjects are considerably younger than those used for the present study. Generalizing differences between the two very young age groups to the complete range of ages is not warranted at this stage. No evidence exists to support such a linear extrapolation. In fact, the present study suggests that the opposite is true. The effect of sex of lead character on the popularity of programs for boys appeared only in those shows viewed by proportionally more fourth graders than eighth graders. Program preferences appeared to be less likely to be 116 affected by sex of lead character as age of male respondent increased. Another possibility is that the relationship is curvilinear. As boys develop sex role identification they may prefer to observe programs with male models. However, as they approach the teenage years, this prefer- ence may disappear. Concern with members of the opposite sex may draw attention to programs with female leads. Other explanations for the observed differences between Saturday morning and prime time viewing might center on differences in program- ming rather than audience composition. Perhaps sex of lead character counterprogramming is extensively employed on Saturday morning and not so in prime time. The creation of Saturday morning shows designed to attract a principally boy or girl audience was noted by Cantor (1974). No such concerted pattern of purposive male-female audience segmentation has been documented for prime time programming. The single exception to this may be for the only regularly scheduled prime time sports program "Monday Night Football." This was the only program type observed to result in audience segmentation by sex. The question which needs to be asked is "Why was this so?" It is that it is devoid of female charac— ters, participants or commentators while its competition (Maude and All in the Family) contain strong female characterizations? What is being suggested is that audience segmentation by program type is too simple a formulation for predicting viewing behavior with any accuracy. The appeal of a program to differing segments of an audience is based on several factors which may include type. One suggestion for such a factor might be sex of lead character. Another might be amount of violence, since Gerbner (1972) indicated that male viewers prefer vio- lent programming over female viewers. Other types of segmentation of 117 course, should be investigated. Segmentation by age and intelligence of characters and program structure (Watt and Krull, 1974), are likely candidates for inclusion in future analyses. Another predictor of program popularity for prime time programs was startipg time. The variable was a significant and negative corre- late of program popularity in every case. As expected, the size of the correlation was negatively related to grade of respondent. This var- iable was highly related to several other variables which were hypothe- sized to be predictors of program popularity. Inclusion of these interrelated independent variables resulted in multicollinearity problems which attenuated each variable's power to predict program popularity. Thus, none of these variables emerged as significant unique predictors of program popularity. Of the set of variables, however, the logical choice among them was start time. Start time was the reason that the popularity of a program's lead-in was related to program popularity (later programs are less popular and since their lead-ins are also later, the lead-ins are likewise less popular). The reverse causal sequence is illogical. Of the several variables investigated in this study, the three variables discussed in this section appear to hold the most promise. Whether a program is new or old is a significant predictor of program popularity for schoolchildren. The time a program starts is a schedul- ing factor which significantly alters the popularity of a television program. Sex of lead character when matched with sex of viewer is positively related to program popularity for schoolchildren of either sex. One must consider variation attributeable to the variables employed 118 in this study in the context of other factors which were not evaluated. Exogenous variables unrelated to programming or programming strategy do influence program popularity. Such variables include factors in the child's social environment such as parental influence over the child's television viewing. This includes restrictions over total time and individual shows as well as encouragement to view other shows. McLeod, Atkin and Chaffee (1972) indicated that such control is more prevalent among younger than older adolescents. Other variables which may affect program selection include sibling viewing preferences as well as the viewing preferences p£_parents. These effects may operate through the other family members greater control over the television or through a child's modeling of other's program preferences. Variation in program popularity for children is not going to be sufficiently explained by investigating programs or programming strate- gies in isolation. Program characteristics, programming strategies and audience characteristics must all be considered in order to more fully account for variability in program popularity. The emergence of three consistently good predictors is the first step toward a fuller explanation of the determinants of program popu- larity, and change in the popularity of programs over time. More variables need to be identified and evaluated. The preceding discus- sions have presented variables which appear to be easily quantifiable. Qualitative factors, on the other hand, pose more severe operational difficulty. Elements such as durability of plot, quality of script, acting, directing, etc., need to be examined as well. No matter how ingeniously a program is scheduled, a poor quality program should be the "most objectionable" and unpopular program in a time slot. 119 General Conclusions This section presents general conclusions supported by the present study. 1. Reliability of the checklist measure of program popularity was good. Some minor validity problems emerged for fourth graders which detracted from the strength of some of the findings in the study. Corre- lations between the popularity checklist and Nielsen national estimates were as strong as could be expected and improved with age of respondent. 2. Concerning stability, it was found to increase over time. Size of the increase was negatively related to respondent age. This indicates that acclimation to the new season takes longer for younger respondents. Stability was highest for fourth grade program popularity suggesting that once acclimated, young viewers are more loyal. 3. Concerning counterprogramming, no evidence appeared which sup- ported the efficacy of counterprogramming by program type or returning vs. new programs. Returning programs do not significantly detract from the p0pularity of competing programs despite the fact that returning programs themselves are more popular than new ones, at least in the Fall. 4. Concerning inheritance effects, no variation in program popu- larity is attributeable to the popularity of adjacent programs. The popularity of adjacent programs does not have a unique impact on program popularity for schoolchildren. The popularity of adjacent programs is attributeable to the time the programs start. Inclusion of the latter variable attenuated the relationship between adjacent program popularity. 5. Block programming did not emerge as a significant determinant of program popularity in this study. The results may be due to the program typology utilized and/or the absence of variables designed to 120 control for variability within program type. 6. Sports programs are the only program type which appeals more to one sex than the other, with greater popularity among boys. The absence of sex related differences in other program categories was probably due to large variability in popularity within each type. Again, control for this variability is needed. 7. Among the other program descriptors suggested was sex of lead character. Saturday morning programs with male leading characters are more popular among boys than other Saturday morning shows. Girls appear to prefer programs with female leads in prime time. 8. Concerning popularity differences by age, the popularity of all Saturday morning program types decreases with age of respondent. Comedy shows, regardless of whether they are animated or nonanimated, exhibit the sharpest decline. 9. Starting time is negatively related to prime time program popularity. The relationship is strongest for younger viewers. 10. Returning programs are more popular than new programs but this difference disappears by Winter. The disappearance is probably due to the network's quick replacement of unpopular new shows and an increase in respondent familiarity with the new programs. Limitations Among the problems encountered in the study were those involving the typologies. The absence of any type related counterprogramming or block programming effects may be symptomatic of the typologies used in this study.. Other typologies may support the hypotheses of other writers concerning the efficacy of counterprogramming and block pro- gramming. Until future research studies replicate the findings of 121 this study concerning counterprogramming and block programming with other typologies, the generalizeability of the present study's findings are limited. This is especially true for the Saturday morning typology since it was developed in the present study. Future research is needed to examine its utility for predicting audience viewing patterns. Multidimensional scaling may be useful for discovering the dimen— sions people use to classify programs. The output of multidimensional scaling is a Euclidean space locating each program on a dimension. These dimensions may be empirically related to unidimensional concepts which provides the researcher with an empirical basis for naming concepts used in the perception of television shows. Such procedures have been successfully employed to determine perceptions of television characters (Reeves & Greenberg, 1977) and mass media (Lometti, Reeves & Bybee, 1977). Another limitation of the present study is attributeable to meas- urement of the dependent variable "program popularity." Generalization to other measurement systems such as diaries or telephone coincidental would be tenuous. Further research is needed to establish the relation— ships among the several methods currently used to measure exposure and program popularity. Further limitations are imposed by the respondent sample used. Generalization of the program popularity scores for the same programs but with different respondents may be tenuous. The correspondence between Nielsen national estimates and the present study's checklist based popularity scores was not very high. The highest correlation was for this study's eighth graders and Nielsen's 12-17 year olds. Even then, however, the common variance was only 36%. While the smallness 122 of this figure is in part attributeable to differing measurement proce- dures, it is undoubtedly also due to differences in the samples employed. Generalization of these results to other samples, national or local, would be premature. The dayparts examined also limit the generalizeability of the study. Assuming that the results of the present study are valid for dayparts other than prime time or Saturday morning would be tenuous. Also of concern is the fact that the study analyzed viewing and programs during a single season. If that season alone were of concern, statistical tests would be superfluous. The best linear combination of the indepen- dent variables for predicting prime time and Saturday morning program popularity for the '75-'76 season are reported by the regression equa- tions. This study contains close to the universe of all shows broadcast during that season. However, the researcher was concerned with more than one season. The use of statistics allows some generalizeability to a larger universe of programs and situations. However, until further research is done, generalizing the present results to other seasons or day parts may not be warranted. One final limitation of the present study involves problems encountered concerning missing values. The standard missing value procedure employed in multiple regression analysis is labeled listwise deletion. In listwise deletion a case is excluded from the computation of gpy_regression statistics if it contains even a single missing value. Thus, a program with only one missing value would be cmmpletely elimi— nated from any regression related analyses. This procedure often results in large losses of information. The alternative procedure is labeled pairwise deletion. In pairwise deletion, a missing value for a 123 particular variable causes that case to be eliminated from calculations involving that variable only. Hence, the computed statistics for each variable or pair of variables could have a different number of cases. However, each correlation may be based on a different segment of the sample. This could result in highly unusual results, such as negative sums of squares and §_ratios, or multiple correlations greater than one. Listwise deletion is usually preferred but was not employed in this study. Attempted use of listwise deletion reduced the number of cases (shows) by two thirds and deleted them systematically. For example, all shows which had no data concerning lead-in program popularity were deleted. Since no data were collected on the popularity of shows before 8 P.M., all of these shows would have been dropped from the study. No popularity data were obtained for programs which began on or after 11 P.M. Hence, all programs which ended at 11 P.M. or later would have been dropped from the study. The loss of all of these shows would have been disastrous. Hence, pairwise deletion was used. A comparison of both procedures did yield some discrepancies. However, whether they are due to problems inherent in pairwise deletion or small N's in list- wise deletion cannot be determined from the study. Hence, presentation of the discrepancies here would be fruitless. Further research which includes complete information on early and late evening programs must be used to resolve this problem. Implications The experimenter would suggest from the findings of this study that the checklist measure of program popularity was highly reliable. However, while reliability was high, validity needs further evaluation. The consistency suggested by the reliabilities may not be caused by 124 viewing patterns and viewing patterns only. Future research must address this issue more fully. The correspondence between recall measuresof exposure and actual viewing requires documentation. Studies have been conducted to evaluate Nielsen logs and audimeters which indi- cate that they have problems as well. At least researchers can adjust for these problems if necessary. Such is not the case for checklist systems and other recall measures. The fact that such procedures have face validity is not sufficient to support their use. The persistence of research which finds marginal support for the relationship between television exposure to specific content and actual behavior is at least partly due to measurement error. Improved measurement should increase the strength of findings which relate actual television exposure to "real world" behavior. Further assessments of the veracity of self-reports by schoolchildren are necessary. Those concerned with the process of acclimation to television programs and the establishment of program loyalty would be interested in the findings of the present study. The data suggest that a viewer's age is positively related to the rate of acclimation but negatively related to program loyalty. Such findings can be interpreted within a "tolerance for uncertainty" paradigm. This would suggest that the younger viewer's ability to cope with uncertainty is less than that of older viewers resulting in longer acclimation periods. The apparent greater loyalty of younger viewers may be a function of uncertaintly as well. Rather than rationally cope with the new programs, the younger viewer sticks to the old preferred programs. This interpretation also corresponds to developmental perspectives. Alternative explanations emerge from the literature on perceived value and effort. Younger 125 viewers work harder to find their viewing pattern and preferred shows. This in turn causes them to value the shows they finally view more than the older viewer. This increased valuation results in greater loyalty. Of course, more research is needed in order to confirm the findings of this study concerning acclimation and program loyalty as well as the explanations advanced for viewer related differences. The specific components of the process of acclimation to new programs needs development. What are the procedures or stages indivi- duals go through when acclimating to the new television season? How closely do they parallel those in the traditional diffusion of innovation paradigm? Such research is necessary for both adult and nonadult popu- lations. Much of the research reviewed earlier suggested that there is a large difference between actual viewing of, and expressed likings for, program types. The research also suggested that program types may be adequate descriptors of program preference. Such conjecture, if valid, could be the beginning of research designed to assess the preferences of adult and nonadult viewers. Some authors (see Rothman & Rauta, 1969) argue that program preferences be used to allocate television time to different program types. Indeed, it may be argued that preference rather than actual viewing should be the major determinant of TV time allocation since the latter is often confounded by factors not relevant to program- ming in the public interest. In addition, the former is much easier to measure with some precision while the latter has posed problems that have yet to be overcome. Other policy issues are raised by the findings of this research. Specifically, this study provides some data relevant to the impact of 126 programming variety on the audience. Speculation concerning the promise of cable television centers on that medium's ability to deliver a large variety of programming, and to do so profitably. Steiner (1952) and Besen and Mitchell (1976) argue that greater variety in programming should increase total audience viewing without detracting significantly from the popularity of already existing programming. The addition of a new program which differs from those already broadcast should attract new viewers. Other scholars such as Bogart (1972) argue that variety will only cause greater segmentation of the present viewers in any time slot. They argue that the television audience for any time slot is determined by exogenous factors. New programming could not increase the total number of viewers. These two positions have been labeled the "active" and "passive" models of the television viewer. Both are in fact, likely to be true to some extent. The data in this study, however, conflict with the "passive" model, and are consistent with the "active" model. In this study, a program's popularity was not affected by the absence or presence of variety during its broadcast. Program popularity was not affected by the number of competing programs which were of the same type as the program itself. This result does not support the passive models' contention that changes in variety would alter each program's audience without affecting total audience size. The result is consistent with the view that each program's popularity would remain the same and that the total audience should increase. The latter is, of course, the view of the "active" model. These last findings support those who suggest that more variety in television programming is needed. Since variety is not likely to emerge within the present broadcast structure, the future of cable and 127 multipoint distribution systems appears to promise the increased variety. Research in the future should investigate with greater rigor the effects of these communication innovations. The way viewers respond to the increased variety in the content and uses of communication media will strongly influence the success or failure of future communication innovations. LIST OF REFERENCES 128 LIST OF REFERENCES Atkin, C. Instrumental utilities and information seeking. In P. Clark (Ed.), New models for mass communication research. Beverly Hills: Sage, 1973. Besen, S.M., & Mitchell, B.M. Watergate and television: An economic analysis. Communication Research, 1976, 3, 243-260. Bogart, L. The age of_television. New York: Ungar, 1972. Brown, J.R. Children's uses of television. In J.R. Brown (Ed.), Child- ren and television. Beverly Hills: Sage, 1976. Brown, L. Television: The business behind the box. New York: Harcourt, brace, 1971. Bruno, A.V. The network factor in TV viewing. Journal o£_Advertising Research, 1973, 13(5), 33-39. Cantor, M. Producing television for children. In G. Tuchman (Ed.), The T!_establishment. Englewood Cliffs: Prentice Hall, 1974. Darmon, R.Y. Determinants of TV viewing. Journal of_Advertising Research, 1976, 16(6), 17-20. Doan, R.K. Why shows are cancelled. In B. Cole (Ed.), Television. New York: Free Press, 1970. Ehrenberg, A.S.C. The factor analytic search for program types. Journal of_Advertising Research, 1968, 8(1), 55—63. Emmett, B.P. A new role for broadcasting research? Paper presented at the Western Association for Public Opinion Research, 1967. von Feilitzen, C. The functions served by the media. In J.R. Brown (Ed.), Children and television. Beverly Hills: Sage, 1976. Frank, R.E., Becknell, J.C., & Clokey, J.D. Television program types. Journal of_Marketing Research, 1974, 8, 204-211. Frost, W.A.K. The development of a technique for TV programme assessment. Journal of_the Market Research Sociepy, 1969, 11, 25—44. Gensch, D., & Ranganathan, B. Evaluation of television program content for the purpose of promotional segmentation. Journal of_Marketing Research, 1974, 11, 390-398. 129 Gerbner, G. Violence in television drama: Trends and symbolic func- tions. In G.A. Comstock & E.A. Rubinstein (Eds.), Television and social behavior. Vol. 1: Media content and control. Washington, D.C.: U.S. Government Printing Office, 1972. Goodhardt, G.J., Ehrenberg, A.S.C., & Collins, M.A. The television audience: Patterns of_viewing. Lexington, Mass.: Lexington, 1975. Greenberg, B.S. Gradifications of television viewing and their corre- lates for British children. In J. Blumler and E. Katz (Eds.), The uses of_mass communication. Beverly Hills: Sage, 1974. Greenberg, B.S. Viewing and listening parameters among British children. In J.R. Brown (Ed.), Children and television. Beverly Hills: Sage, 1976. Greenberg, B.S., & Atkin, C. Parental mediation of children's social learning from television (continuation request). A proposal sub— mitted to the office of Child Development, Department of Health, Education and Welfare. Department of Communication, Michigan State University, 1976. . Greenberg, B.S., Heald, G., & Wakshlag, J. TV character attributes, identification and children's modeling tendencies. Paper presented at the annual meeting of the International Communication Association, Portland, 1976. Grusec, J.E. & Brinker, D.B., Jr. Reinforcement for initation as a social learning determinant with implications for sex role develop— ment. Journal of Personality and Social Psychology, 1972, 21, 149- 158. Joreskog, K.G., & Van Thillo, M. LISREL: A general computer program for estimating a linear structural equation system involving multiple indicators of unmeasured variables. Princeton, N.J.: Educational Testing Service, 1972. Kingson, W.K., Cowgill, R., & Levy, R. Broadcasting television and radio. New York: Prentice Hall, 1955. Kirsch, A.D., & Banks, S. Program types defined by factor analysis. Journal of_Advertising Research, 1962, 2, 29-31. Klein, P. The men who run TV aren't that stupid . . . They know us better than you think. New York, 1971, 20-29. Lometti, G.E., Reeves, B. & Bybee, C.R. Investigating the assumptions of uses and gratifications research. Communication Research, 1977, 4, 321-338. ' 130 LoScuito, L.A. A national inventory of television viewing behavior. In E.A. Rubinstein, G.A. Comstock, & J.P. Murray (Eds.), Television and social behavior. Vol. 4: Television ip_day-to-day life: Patterns of_p§g, Washington, D.C.: U. S. Government Printing Office, 1972. Lyle, J. Contemporary functions of the mass media. In R.K. Baker & S.J. Ball (Eds.), Violence and the media. Washington, D.C.: U.S. Government Printing Office, 1969. Lyle, J. Television in daily life: Patterns of use (overview). In E.A. Rubinstein, G.A. Comstock, & J.P. Murray (Eds.), Television and social behavior. Vol. 4: Television ip_day-to-day life: Patterns of_p§g, Washington, D.C.: U.S. Government Printing Office, 1972. Lyle, J., & Hoffman, H.R. Children's use of television and other media. In E.A. Rubinstein, G.A. Comstock, & J.P. Murray (Eds.), Television and social behavior. Vol. 4: Television ip_day-to-day life: Patterns 2£.2§E3 Washington, D.C.: U.S. Government Printing Office, 1972. (a) Lyle, J., & Hoffman, H.R. Explorations in patterns of television view- ing by preschool-age children. In E.A. Rubinstein, G.A. Comstock, & J.P. Murray (Eds.), Television and social behavior. Vol. 4: Television ip_day-to-day life: Patterns of_p§g, Washington, D.C.: U.S. Government Printing Office, 1972. (b) McLeod, J.M., Atkin, C.K. & Chaffee, S.H. Adolescents, parents and television use: self-report and other-report measures from.the Wisconsin sample. In G.A. Comstock & E.A. Rubinstein (Eds.), Television and social behavior, Vol. 3: Television and adolescent aggressiveness. Washington, D.C.: U.S. Government Printing Office, 1972. McNemar, Q. Psychological statistics. (4th ed.) New York: Wiley, 1969. Miller, M.M. Television and sex-typing in children: A review of theory and research. Unpublished manuscript. Department of Communication, Michigan State University, 1976. Miller, M.M. & Reeves, B.B. Children's occupational sex role stereotypes: The linkage between television content and perception. Paper present- ed at the annual meeting of the International Communication Association, Chicago, 1975. Nielsen, A.C., Co. Nielsen national TV ratings. Second February Report. Northbrook, Ill.: A.C. Nielsen, 1976. Owen, B.M., Beebe, J.M., & Manning, W.G., Jr. Television economics. Lexington, Mass.: Lexington, 1974. Reeves, B. & Greenberg, B.S. Children's perceptions of television charac- ters. Human Communication Research, 1977, 3, 113-127. 131 Rothman, J., & Rauta, I. Toward a typology of the television audience. Journal pf_the Market Research Society, 1969, 11, 45-70. Rubinstein, E.A., Comstock, G.A., & Murray, J.P. (Eds.), Television and social behavior. Vol 4: Television ip_day—to-day life: Patterns of_use. Washington, D.C.: U.S. Government Printing Office, 1972. Schramm, W. The nature of communication between humans. In W. Schramm & D.F. Roberts (Eds.), The process and effects of_mass communication. Urbana: University of Illinois, 1972. Schramm, W., Lyle, J. & Parker, E.B. Television in the lives of our children. Stanford: Stanford University Press, 1961. Shanks, B. The cool fire. New York: Norton, 1976. Sprafkin, J.N. Sex and sex role as determinants of children's television program selection and attention. Unpublished doctoral dissertation, State University of New York, 1975. Steiner, P.O. Program pattern preferences, and the workability of com- petition in radio broadcasting Qparterly Journal of Economics, 1952, 66. - Streicher, L.H. & Bonney, N.L. Children talk about television. Journal .of_Communication, 1974, 24, 54-61. Swanson, C.E. The frequency structure of television and magazines. Journal pf_Advertising Research, 1967, 7, 8-14. Thayer, J.R. The relationship of various audience composition factors to television program types. Journal of_Broadcastiog, 1963, 7, 215-225. Watt, J.H., Jr. & Krull, R. An information theory measure for television programming. Communication Research, 1974, 1, 44-68. Wells, W.D. The rise and fall of television program types. Journal of Advertising Research, 1969, 9, 21—27. Wiley, D.E. & Wiley, J.A. The estimation of measurement error in panel data. American Sociological Review, 1970, 35, 112-117. APPENDICES 132 APPENDIX A Nielsen Program Types Adventure Award Ceremonies and Pageants Audience Participation Child Multi-Weekly Child Day-Animation Conversations, Colioquies Child Evening Child Day-Live Concert Music Situation Comedy Comedy Variety Devotional Daytime Drama Documentary, News Documentary, General Evening Animation Western Drama Feature Film Format Varies General Drama General Variety Instructions, Advice Musical Drama News Official Police Political Popular Music-Contemporary .Private Detective Popular Music-Standard Quiz-Give Away Quiz-Panel Sports Anthology Sports Commentary Sports Event Science Fiction Suspense/Mystery other than OP, PD Unclassified 133 APPENDIX B Exposure Measure Sample WE ARE INTERESTED IN FINDING OUT WHICH TELEVISION SHOWS YOU HAVE BEEN WATCHING THIS FALL. PLEASE PUT AN X NEXT TO THE SHOWS YOU WATCH EVERY WEEK OR ALMOST EVERY WEEK. Monday night Tuesday night _BARBARY COAST (01) _HAPPY DAYS (10) ____NFL FOOTBALL (02) ___WELCOME BACK, KOTTER (11) ____RHODA (03) ___THE ROOKIES (12) ___PHYLLIS (04) _MARCUS WELBY, M.D. (13) _ALL IN THE FAMILY (05) ____GOOD TIMES (14) _MAUDE (06) ____JOE AND SONS (15) ____MEDICAL CENTER (07) ____SWITCH (16) ___THE INVISIBLE MAN (08) ___BEACON HILL (17) MOVIN' ON (18) POLICE STORY (19) JOE FORRESTER (20) Wednesday night Thursday pighg _WHEN THINGS WERE ROTTEN (21) _BARNEY MILLER (31) _THAT'S MY MAMA (22) ____ON THE ROCKS (32) _BARETTA (23) _THE STREETS OF SAN FRANCISCO (33) ___STARSKY AND HUTCH (24) ___HARRY O (34) _____TONY ORLANDO AND DAWN (25) _THE WALTONS (35) _______CANNON (26) _____THE MONTEFUSCOS (37) ____KATE McSHANE (27) _____FAY (38) _LITTLE HOUSE ON THE PRAIRIE ____ELLERY QUEEN (39) (28) _MEDICAL STORY (40) DOCTORS HOSPITAL (29) PETROCELLI (30) '. - - . 1 .1 '4 I _ ~71 «1‘ ~ ' 1" I ' k c | _ '7, I . - —, ‘1 1‘ L . 1 1 , . w' - _, \ ' . 1 . 1 ‘ 1 1 1 . ' . 1 . . x O 1 I'. . . I .1 . ‘ 1 1' , 1 A 1 . 1 i ‘ O A -1 . 1 ‘ . ‘1 ‘ V (1‘ ~ |, ~~ I . , . . d‘~ v * f. ’ 7~" IV 1 1 o __ A l 7‘. ". 0 ~ — "l . 11 ' ' .1' ' _ :1 ' ‘ “Vi. ‘ ER ‘ ‘ 7 ' zfi .1 p r, '1. ' . . 1. ~- 1 : u . 1. ~11 ‘ ‘1‘1‘ 1" ’ "" . > J . ' , :J. .‘ 1 ‘ ' J v 7 l,‘ ' .1 .._ .‘ -' ‘ 1293 03177 9253 11111111111mumm?“Ilmulmnum11111111111111