AN EXPERIMENT TO DETERMINE THE RELATIVE EFFECTIVENESS 0F NEWSPAPER ADVERTISING SCHEDULES AS AFFECTED BY ADVERTISEMENT SIZE AND FREQUENCY OF INSERTIDN Thesis for theDegree of M. A _ MICHIGAN STATE UNIVERSITY ARTHUR WILLIAM NOVAK 1969 ABSTRACT AN EXPERIMENT TO DETERMINE THE RELATIVE EFFECTIVENESS OF NEWSPAPER ADVERTISING SCHEDULES AS AFFECTED BY ADVERTISEMENT SIZE AND FREQUENCY OF INSERTION BY Arthur William Novak The purpose of the study was to determine the relative effectiveness of running a single advertisement of relatively large size vs. a sequence of three smaller advertisements in a newspaper advertising schedule. An attempt was made to design a project that would be free from many of the extraneous variables which had plagued previous studies of space size and repetition. 911 telephone interviews were completed with stu- dents and non-students who held season coupon books to Michigan State University's Performing Arts Company pro- ductions. These reSpondents also were readers of newspapers carrying advertising for these theatrical productions, and thus potentially were exposed to either three small adver- tisements appearing at weekly intervals or to one larger advertisement appearing concurrently with the initial small Arthur William Novak advertisement. All advertisements were for one of two Performing Arts Company productions and appeared in one of two newspapers. Respondents were telephoned during two-day calling periods before and after the appearance of each advertisement. During each calling period, they were asked whether they were aware of the play being advertised, whether they planned to attend the play, and whether they recalled seeing any advertising for it. Responses were tabulated and subjected to Chi-square analyses and corre- lation analyses. Although the large advertisements were found to generate higher initial recall, the high rate of forgetting in both the large and the small-advertisement treatment groups suggested that, when advertising recall must be sustained over time, a smaller advertisement (12 column inches) appearing at weekly intervals for three weeks is more effective than a single, larger insertion (35-40 column inches). Accepted by the faculty of the Department of Advertising, College of Communication Arts, Michigan State University, in partial fulfillment of the re- quirements for the Master of Arts Degree. fi/fiéw Director of Thesis % 21,17; 7 AN EXPERIMENT TO DETERMINE THE RELATIVE EFFECTIVENESS OF NEWSPAPER ADVERTISING SCHEDULES AS AFFECTED BY ADVERTISEMENT SIZE AND FREQUENCY OF INSERTION BY Arthur William Novak A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS College of Communication Arts Department of Advertising 1969 ACKNOWLEDGMENTS Appreciation is extended to the following persons for helping to make this project possible:‘ Dr. Gordon E. Miracle of Michigan State University's Department of Advertising, for his guidance in all phases of the project; The students of Dr. Miracle's Advertising Research course (spring, 1969), for conducting the interviews; Dr. E. C. Reynolds and Dr. John Baldwin of Michigan State University‘s Department of Theatre, for allowing me to regulate advertising for two Performing Arts Company productions in accordance with the requirements of this project. ii TABLE OF CONTENTS LIST OF TABLES O O O O O O O O O O O O O O O O O 0 LIST OF FI Chapter GURES O O O O O O O O O O O O O C O O O I I 0 INTRODUCTION 0 O O O I O O O O O O O O O 0 II. RE The Problem . . . . . . . . . . . . . . . Purposes of the Study . . . . . . . . . . Hypotheses . . . . . . . . . . . . . . . Methodology . . . . . . . . . . . . . . . VIEW OF LITERATURE . . . . . . . . . . . Studies of the Effects of Repetition on Advertising Effectiveness . . . . . . . . Studies of the Effects of Increased Space Size on Advertising Effectiveness . . . . "Forgetting" as a Factor Affecting the Relative Effectiveness of Larger Space vs. Repetition . . . . . . . . . . . . . Present Status of the Problem . . . . . . III. PRESENTATION, ANALYSIS, AND INTERPRETATION OF DATA 0 O O I O C O O O O O O O C O O I 0 Awareness of Next PAC Production . . . . Advertising Recall . . . . . . . . . . . Plans to Attend . . . . . . . . . . . . . IV. CONCLUSIONS 0 O O O O O O O O I O O O O O O BIBLIOGRAPHY O O O O O O O O O O O C O O O O I O O APPENDICES iii Page iv H nbbJNH 21 21 29 33 37 42 48 53 82 88 94 97 LIST OF TABLES Table Page 1. Respondent Awareness of Next Production . . . . 43 2. Respondent Recall of Newspaper Advertising . . 44 3. Respondent Plans to Attend Next Production . . 45 4. Positive Responses as a Percentage of Total Responses During Each Calling Period . . . . 47 Appendix B: Chi Square Computations . . . . . . . . . . . . lll Correlation Analyses . . . . . . . . . . . . . ll6 Donnahoe's Square Root Formula . . . . . . . . 118 iv LIST OF FIGURES Figure Page 1. Small Advertisement - Marat/Sade . . . . . . . 8 2. Small Advertisement - Little Mary Sunshine . . 8 3. Large Advertisement - Marat/Sade . . . . . . . 9 4. Large Advertisement - Little Mary Sunshine . . 10 5. Calling Periods During Advertising Schedules in Two Newspapers . . . . . . . . . . . . . 12 6. Respondent Awareness of Next Production . . . 49 7. Respondent Recall of Newspaper Advertising . . 54 8. Correlation Analysis: Recall of Newspaper Advertising--L vs. S groups . . . . . . . . 61 9. Correlation Analysis: Small-ad Treatment groups-Awareness of Newspaper Advertising vs. Recall of Next Production . . . . . . . 61 10. Schedule of Non-Experimental Publicity . . . . 72 ll. Respondent Plans to Attend Next Production . . 83 Appendix A: Questionnaire Number 1: Pretest for Preliminary Survey Questionnaire . . . . . . 97 Interviewer Patter and Questionnaire #1 Preliminary Survey . . . . . . . . . . . . . 99 Questionnaire "A" . . . . . . . . . . . . . . 101 Questionnaire "B" . . . . . . . . . . . . . . 103 Questionnaire "C" . . . . . . . . . . . . . . 105 Questionnaire "D" . . . . . . . . . . . . . . 108 CHAPTER I INTRODUCTION The Problem As advertising has grown into a billion-dollar institution, a number of recurring problems have remained unsolved. One problem of importance involves the choice between large-size advertisements and continuity. Rel- atively little attention has been given to measuring the effectiveness of infrequently-appearing, larger advertise- ments as compared to the effectiveness of smaller adver- tisements appearing more frequently. Shortcomings of Past Studies A basic shortcoming of past studies is that, while there have been many efforts to examine the effects of repeat advertisements and of increasing their size, no recent study has attempted to look at these factors simul- taneously. A simultaneous comparison is desirable because the media planner, working with a limited budget, is forced to make conclusions about these variables simultaneously. He knows that an increase in advertisement size necessitates a sacrifice somewhere in frequency and continuity. While existing studies may tell him optimum size or optimum number of repetitions for a given set of advertisements, they do not give adequate indication of comparative effec- tiveness values for an advertisement of size A repeated x times vs. an advertisement of size B repeated y times. A second shortcoming has been the difficulty of isolating the experimental variables from extraneous var- iables. Within individual studies, the advertisements used have appeared in more than one media vehicle, and, often, a number of different types of items have been advertised. Under these circumstances, it is difficult to draw valid conclusions on the effects of varying either space size or number of repetitions. Finally, different studies have employed different measurement criteria of advertising effectiveness. Whereas one study might measure interest in a product or awareness of that product, another might use recall or recognition of the advertisements as a measuring stick. Purposes of the Study The present study examines the effects of space size and repetition simultaneously. As a result, media planners may be able to apply some of the findings to their own problems. In focusing upon the variables of size and repeti- tion, other variables which would limit the usefulness of the data were eliminated. The advertisements are for the same items and appear at regular, predetermined intervals. The design of the study includes analysis of data on size and repetition with respect to three criteria of advertising effectiveness--awareness of the advertised item, success of the advertising in "selling" that item, and recall of the advertisements themselves. Since each of these criteria will be applied to the same advertising schedules, various relationships between criteria also can be analyzed. Since the experiment took place over a three-week period, forgetting of the advertising, also will be measured and analyzed. The problem of optimizing an advertising schedule cannot be solved until more is known about for— getting and advertising recall. Hypotheses The study gathered evidence to use as a basis for testing the following null hypotheses: 1. There is no difference in awareness of the name of an advertised play between persons recently exposed to a small advertisement and those recently exposed to a large advertisement. 2. There is no difference in the level of recall of an advertisement one to two days after exposure and the level of recall four to six days after exposure. 3. There is no difference in the amount of for- getting of the advertising that occurs between persons exposed to a large advertisement and those exposed to a small advertisement. 4. There is no difference in the amount of for- getting of the advertising that occurs between persons exposed to only one advertisement and those exposed to more than one advertisement. 5. There is no difference in advertising recall between persons exposed to one large advertisement and those exposed to three small advertisements. 6. There is no difference in advertising recall between persons recently exposed to a large advertisement and those recently exposed to a small advertisement. 7. There is no difference over time between changes that occur in awareness of the name of an advertised play and changes that occur in recall of the advertising. 8. There are no differences in stated plans to attend a play resulting from either exposure to or forget- ting of the advertising. Methodology Review of Literature Previous studies on the effectiveness of various advertisement space sizes and numbers of repetitions have been summarized and appear in the following chapter. While these studies provide some insights into the solution of the research problem with which this thesis deals, and also provide direction in formulating the design of the present experiment, the review of the literature serves primarily to point up the need for additional research. The Experiment An experiment was carried out as follows: two advertising schedules were set up to run in two newspapers-- a three-insertion schedule of relatively small advertise- ments and a one-insertion schedule consisting of a rel- atively large advertisement, about three times the size of each of the smaller advertisements. Both the large and the small-advertisement schedules appeared twice, once in each of two newspapers for one of two plays presented by Michigan State University's Performing Arts Company. The treatment groups consisted of respondents who were exposed to the large-advertisement schedule and re- spondents exposed to the small-advertisement schedule. The two large advertisements were almost identical in content to the six smaller advertisements, so that differences resulting from variations in the item being advertised or in the composition and pictorial elements of the advertise- ment were minimized. Also, the two treatment groups had roughly the same demographic composition, so respondent bias was negligible. Background Information on Performing Arts Company LPAEL.--The Performing Arts Company at Michigan State Uni- versity is an organization of graduate students in theatre, operating within the University's Department of Theatre. During the 1968-69 academic year, the PAC presented eight plays. The pattern of advertising for the last two plays of the season, Marat/Sade and Little Mary Sunshine, was the object of analysis in the present experiment. Advertising Schedules.--Advertising schedules for the two plays were determined in advance so that advertise- ments would appear systematically for experimental purposes. For Marat/Sade, advertisements appeared in the following newspaper editions: State Journal (Lansing area daily newspaper)--April 6, April 13, April 20; State News (MSU daily newspaper)--April 8. For Little Mary Sunshine, the schedule patterns were reversed: State Journal--May 4; State News--May 2, May 9, May 16. Advertisements, then, appeared in two types of schedules: a one-insertion and a three-insertion schedule. A one-insertion schedule of a "large" advertisement for one of the plays in one of the newspapers ran parallel to a three-insertion schedule of "small" advertisements for the other play in the other paper. During the second half of the study, the plays and newspapers were switched in both treatment groups. Size and Composition of Advertisements.--Advertise- ments were designed so that the large advertisement in a schedule was approximately three times the size of each of the three smaller advertisements. The sizes of the Marat/ Sade advertisements were 5 columns x 7 inches in the News. and 3 columns x 4 inches in the Journal. The Little Mary Sunshine advertisements were 4 columns x 10 inches in the Journal and 2 columns x 6 inches in the News. (One column is approximately 1-3/4 inches wide.) Thus, on an inch-per- inch basis, the three small advertisements combined would approximate the advertising space and cost of the single larger advertisement. The three small advertisements for a given play were identical, except for minor changes in copy necessitated by the opening of mail orders, openings and closings of the box office, and the actual openings of the plays. The only major difference in the large advertisement was the size. Attention was also given to graphic elements of the adver- tisements across the two plays. For both plays, a photo- graph was the focal point of the advertising. Titles were set in large type of equal size for the two plays. (See following pages for copies of advertisements.) The Sample.--The sample was drawn from the PAC coupons of 2560 students and 1173 non-students on file with the MSU Theatre Department. Those with coupons on file had purchased PAC coupon books at the start of the school year. coaaou Lo CON» .ooEO xom 229:0»; *0 N. .32 mar—Lo.» coocgoto .3370? :20 0—0..." :0 £3.02. .5... o 3-8 .62 mzfmzzm >M> . . . . . . . . . . . 2.30 05.623 05 mo; 93: 05 van Cams» no 9325 no no; Em 05 9.0;; 23335 035 3380.5”. H H .( om >m2||m3mz mumumv mcflsmCSm Nsz mapuflqllucmfimmfiuum>©¢ Hamsmll.m musmflm wow Hflumdnlam:H50h mpmumv momm\umumzuupqmsmmHuum>ca Hamsmuu.a musmflm 25.8% 8&2. .5 assoc: FE 8:8 5n 35...“... a sum .5 32.2... 225.52..— »zE—go at: cages—cu.— < id m 5.56:». 04.105.3— h~.~N .=E< mn.<130iol310£51‘3 Pastas—>— 1 53.953 1.. 8:38.: o..— .123 if. at... 185. 3.3.! 2% on. 3..- two «It... $.55; Eb. 8. £2 E >$x ha a: fiesta; «In 2 I 9.2. E339; 2.: . . .. lunusblvn-\Tbvt|ll\ \I- I: u~n-\I!u v.6! \.i.\ n\s\~ Am HHHQ¢IIm3mz mumumv momm\umumzllucmfimmflunm>©< mmumqll.m musmam cozoucomota .EaanU 3.3 95:623. 4 zomooo mo 8.3 moasso xom odmost .3. 3 1:5: ozfimxfim >> Wmm>m m4>OZ mmmomo 1:52 ._2.a m mm._.m< a... (.0: {tit ‘52....) 8.10;»; of 80- VIC T1... at}... 23:30; it: 2 53.... :33 r it» a.“ Q; 1.0.3.10) 2:1 0; a. on“ *0 mmUDEc_ 0:“ >n VOELO%LmQ mu 32.x: 7...: Zvatc 8‘» co co_.uc_»u»n< can cozaummrod of. 10 Remember those musicals where the girl was as sweet as sugar, and the hero was the salt of the earth? . . . . . . Well, we’re bringing back that old-lime flavor. THE.PERFORM|NG ARTS COMPANY PRESENTS LITTLE MARY SUNSHINE MAY 20-25 FAIRCHILD mum 8.P.M. MAIL ORDERS TO COUPON HOLDERS START TOMORROW MAY 5 TICKETS ON SALE EVERY WEEKDAY AFTERNOON STARTING MAY 12 AT FAIRCHILD BOX OFFICE. $2.00-OR COUPON. Figure 4.~-Larcm_ Advertisement—-Little Mary Sunshine (State Journal--Ma§' 4) 11 A coupon book entitled the owner to admission to each PAC event at a discounted price. Cost of the book was $7.00. Consequently, the coupon-holders, as a group, were expected to have an above-average interest in theatre in general and in the activities of the PAC in particular. A total of 911 interviews was completed over the course of the study. Determination of appropriate sample sizes was com- plicated by the nature of the universe. Out of some 4300 purchasers of PAC coupon books, 3733 had identifying coupons on file with the Theatre Department. This universe of 3733, however, was not divided equally between students and non- students. 2560, or nearly 69% of the coupon book-holders, were students. Only ll73--about 31%--were non-students. Consequently, with equal-size samples of students and non- students desired, maximum sample size was limited by the number of non-students. Readership habits of non-students further compounded the problem. Only students who read the Ngw§_(but not the Journal) and non-students who read the Journal (but not the News) could qualify as subjects. Although about 90% of the students were readers of the New§_only, only about 53% of the non-students were readers of the Journal only. Universe size (coupon—holders with the stipulated readership habits) was then reduced to 2560 x .90 = 2304 students; 1173 x .53 = 622 non-students; or 2304 + 622 = 2926, total universe size. 12 A final complication in arriving at a reasonable sample size stemmed from the fact that a different sub- sample was to be interviewed before and after each adver- tisement appeared, as follows: Marat/Sade Calling Periods** Journal group (non-students) Slb * Sla 82b * 82a 53b * 83a News group (students) le * Lla L2b L2a L3b L3a Little Mary Sunshine Calling Periods** Journal group (non-students) le * Lla L2b L2a L3b L3a News group (students) Slb * Sla 52b * 82a 83b * 83a **"8" and "L" stand for the small and large-adver- tisement treatment groups, respectively. The numbers 1, 2, and 3 represent the week during which an advertisement appeared. The six columns across refer to the six calling periods. The letters "b" and "a" refer to calling periods before and after probable exposure to a particular adver- tising insertion. These insertions are represented by asterisks. In the L group, asterisks appear only between calling periods lb and la, because only one large adver- tisement appeared, during the first week of interviewing for each play. Subsequently, L-group respondents were phoned at the same times as respondents in the correspond- ing S-group subsamples. Figure 5.--Calling Periods During Advertising Schedules in Two Newspapers In order to determine how large a subsample had to be to achieve a given level of statistical reliability, it was necessary first to determine how many subsamples would 13 provide data in each individual statistical analysis. A typical statistical analysis in this project is a Chi-square computation involving comparison of positive and negative responses for Sla (for both plays) with those for Lla. If 50 students and 50 non-students were interviewed (as was planned) after the appearance of the large and initial small advertisement for each play, the subsamples Lla + Sla would equal 200. The next step was to determine whether this sample size of 200 would allow acceptable statistical reliability. The minimum sample size necessary was computed by means of the following formula:1 2 n = k (lOO-P) (10,000) E3'( p ) where k = the number of standard deviations from the true universe percentage cor- responding to an area under the normal curve (des— ignating a specified level of reliability). r = percentage within which universe percentage is to be estimated. p = assigned probability of getting a positive response. Assuming that positive and negative responses to questions would be equal in number, P will have a value of .50. Taking a l-in-S chance that the acquired data will lHarper W. Boyd, Jr. and Ralph Westfall, Marketing Research: Text and Cases (Homewood, 111.: Richard D. IrWfi, Inc. ' 196D ' p. 380. 14 not be within 5 percentage points of the true universe percentage, the relative percentage r can then be computed as 35 or 10%. The k-value (for 80% confidence) is 1.28. In other words, if we are willing to take a one- in-five chance that a given result will be more than 10% away from the true universe percentage, necessary sample size will be, 1.64 X 100-50 100 50 X 10,000 = 164 Since this sample size is greater than 5% of the universe of 2926, an adjustment must be made in the sample size, according to the formula below:2 n . . n' = where n = unadjusted sample Size 1 + . . and N universe Size ZII'J Substituting into the formula, , _ 164 _ n ‘W '155 In actuality, as a result of shortages of inter- viewers and interviewer time, the ideal of 50 student and 50 non-student completed interviews during each calling period for each play generally was not attained. Conse- quently, the average total subsamples in the L vs. S or 21bid., p. 379. 15 before vs. after computations fell short of 200. Actual subsample totals, however, for 21 out of the 30 Chi-square computations in the study, exceed 155. Sampling Technique.--Sixty lists of thirty names each (thirty student lists and thirty non-student lists) were typed out and distributed randomly to each interviewer. During a calling period, an interviewer was expected to complete 10-12 interviews with qualified respondents from his list of thirty names. The lists were typed in alpha- betical order of respondent's last name, and interviewers were instructed to phone respondents in the order their names appeared on the lists. Since an anticipated 300 non-students were to be interviewed on MaratZSade advertising, and since there were only approximately 622 qualified non-student respondents, it was felt that it would have been difficult or impossible to obtain a completely different non-student sample for Little Mary Sunshine. Consequently, both the non—student and the student samples for Little Mary Sunshine were drawn from the same respondent lists as were the Marat/Sade sam- ples. For the Little Mary Sunshine phase of the project, respondent lists were redistributed among the interviewers, with respondents that had completed interviews for Maratz §a§e_designated by red dots placed next to their names. Interviewers were then instructed to phone respondents in the order their names appeared on the lists, as before, but 16 with the proviso that 60% of completed interviews be ob— tained from "red dot" respondents. In this way, the per- centage of respondents who were interviewed a second time was roughly equalized for the large and small-advertisement treatment groups, and the factor of previous experience with the project should not have had a significant biasing effect on comparative data between the two groups. In the second half of the interviewing, then, sampling was not random. Those who had been interviewed previously had a greater chance of being interviewed again. It is felt, however, that this fact does little harm to the results of the project, since it is not comparisons between the plays that is the main object of analysis, but compar- isons between the large and small-advertisement treatment groups, each of which contain an approximately equal number of respondents from each of the plays. Telephone Interviewing.--Telephone interviews were conducted by 17 students in a senior-level advertising research course at Michigan State University. The students were assigned calling periods for interviewing PAC coupon book-holders. One group of interviewers phoned only stu- dents, the othergroup only non-students. A given respondent was to be phoned within a 2-day period before or after the insertion date of an advertisement. Student respondents were phoned before or after the appearance of an advertise- ment in the State News; non-students, before or after an 17 advertisement in the State Journal. During weeks when only a small advertisement appeared in one of the papers, those in the large-advertisement treatment group were phoned concurrently with those in the small-advertisement group. Thus, for each play, there were three "before" and three "after" phoning periods for student respondents and three such "before" and "after" periods for non-students. Interviewing Procedure and Questionnaire Forms.-- Interviewers were divided into four groups corresponding to the following calling periods: (1) non-student respondents before an insertion; (2) student respondents before an insertion; (3) non-student respondents after an insertion; and (4) student respondents after an insertion. The questionnaire was adapted for each of these four groups. Thus, four questionnaires were used, iden— tified as A, B, C, and D (see Appendix A). Questionnaire forms A and B were administered to non-students and stu- dents, respectively. Since the project was specifically concerned with the effectiveness of particular advertising schedules in particular newspapers, it was essential that respondents be relatively high in exposure potential to an advertisement in one of the papers and relatively low in exposure potential to an advertisement in the other paper. Consequently, in order to qualify as a respondent, a non- student had to be a regular reader (at least three times per month) of the Sunday State Journal (all Journal 18 advertisements appeared in Sunday editions); a student respondent had to be a regular reader (at least four times per week) of the State News. Dividing the sample into student and non-students with these readership habits eliminated the problem of biased subsamples. For purposes of analysis, the students' responses in the large-adver- tisement treatment group for one play were combined with the responses of non-students in the large-advertisement treatment group for the other play, and this total was compared with total responses in the small-advertisement treatment group. Consequently, there were approximately equal numbers of students and non-students in each treatment group. After being questioned as to their readership habits, those respondents who met readership requirements were asked three additional questions: (1) Can you tell me the name of the next Performing Arts Company production? (If they could not, they were told the name.) (2) As of now, do you think you will attend this play when the Performing Arts Company presents it this spring? (3) Do you recall seeing any newspaper advertising for the play? Questionnaire form C was used for non-students interviewed during calling periods after the appearance of an advertisement in the Journal; similarly, questionnaire form D was used for students during calling periods 19 following the appearance of an advertisement in the News. These forms differed from the A and B questionnaires in one respect: in form C, respondents were asked whether they had read the Theatre section of the preceding Sunday's Journal; in form D, they were asked if they had read the News on the day of the last previous advertisement inser- tion. The purpose of the C and D forms was to increase the validity of results in the "after-advertisement insertion" calling periods by including only those respondents with the highest exposure potential to the advertising. Possible effects of the use of the C and D forms will be discussed in a later section. Preliminary Survey.--Six weeks prior to the start of the main series of interviews, a preliminary set of telephone calls was made. The questionnaire used for this survey had been pretested with calls to five students and five non-students. (See Appendix A for questionnaire forms.) In the preliminary survey, thirty students and thirty non-students--all PAC coupon book-holders--were interviewed. The primary purpose of the preliminary survey was to get some indication of whether there were any differences in knowledge of and interest in the two plays. Respondents who correctly identified Marat/Sade as "serious drama" and Little Mary Sunshine as "light comedy" were rated "very familiar" with the plays. Those who could not 20 correctly classify the plays but claimed to have heard of them before were rated "fairly familiar." Those who had not heard of the plays were rated "unfamiliar." Responses of students with respect to Marat/Sade and non-students with respect to Little Mary Sunshine were added together and compared to the total of responses of students with respect to Little Mary Sunshine and non- students with respect to Marat/Sade. These were the two key groups for comparison because they corresponded to those who would be exposed to the large-advertisement and small-advertisement treatments, respectively. With respect to the "very familiar" criterion, the data revealed a higher level of positive response in the large-advertisement treatment group. However, when the criterion was "fairly familiar" there was little indication of any difference between the treatment groups.» Since a reader who is "fairly familiar" with a play might have as much interest in an advertisement for that play as would someone "very familiar" with it, the preliminary survey does not indicate that knowledge of the plays is different enough between treatment groups to bias subsequent findings. CHAPTER II REVIEW OF LITERATURE In reviewing previous work on the problems at hand, it is necessary to summarize two distinct bodies of data: (1) experimentation on the effectiveness of repeat adver- tisements; (2) studies on the effectiveness of increasing size of advertisements. In both instances we will be con- cerned not only with levels of advertising effectiveness with various sizes of advertisements or different numbers of repetitions, but also with the rate of forgetting of the advertising over time. Studies of the Effects of Repetition on Advertising Effectiveness Researchers disagree on the question of how large an increment of recall a repetition of an advertisement will produce. At one extreme, some studies indicate that, regardless of frequency, a repeated advertisement will be seen by at least as many new readers (or readers who did not recall seeing it before) as saw it the first time.3 3Roger Barton, "Repetition-Repetition-Repetition," Media/scope, Vol. 12 (July, 1968), p. 62. 21 22 This result was obtained in a study of the Steering Commit- tee of Survey of the Effectiveness of Repeated Advertise- ments, a group of more than 300 advertisers, agencies, and publishers,4 and in thirteen studies made by McGraw-Hill Research on various trade publications.5 According to Daniel Starch, advertisements can be repeated at least six times in monthly issues of a magazine without losing readership.6 In another test of trade publications, groups of advertisements were repeated in Steel and Product Engineer- ing, a weekly and a monthly, respectively. One of the major conclusions of the study was that, "when an ad is repeated, it is seen and read by at least an equal number of new readers or readers who do not recall seeing it before."7 Studies in the consumer magazine field support the notion of equal readership for repeated advertisements. From 1955 to 1958, 80 advertisements were inserted a total 41bid. 5James Playsted Wood, Advertising and the Soul's Belly (Athens, Ga.: University of Georgia Press, 1961), p. 87. 61bid., p. 88. 7Albert Wesley Frey, Advertising (Third Edition; New York: The Ronald Press Company, 1961), p. 365. From "The Effectiveness of Repeated Advertisements," Industrial Marketing, Vol. 32 (Sept.-Nov., 1947). 23 of 234 times in different issues of a consumer magazine. Readership was almost constant for each advertisement up to eight insertions (the maximum level of insertions in the study). The intervals between insertions ranged from several weeks to several months.8 A more recent study by McGraw-Hill used 183 advertisements repeated twice and 40 advertisements repeated three times.9 Advertisements repeated twice had "read" indices of 104.1 and 99.3 (com- pared to 100 for the original); those repeated three times received scores of 92.7, 98.7, and 95.4. In other words, little falling off of readership was observed. The evidence on "constant readership" with repeti- tion is widespread, but far from conclusive. A major, criticism of the studies is that a person being interviewed to determine whether he has seen a repeated advertisement, may respond positively because he saw the original. That is, even if a respondent is asked about a particular issue of a magazine or newspaper, it will be easy for him to assume, if he remembers an advertisement, that he saw it in that issue. According to Howard Sawyer, when an advertise- ment scores as high after the second insertion as it did after the first, this is an indication that the second group of respondents "contains peOple who saw the ad the 8Barton, Media/scope, Vol. 12, p. 62. 91bid. 24 first time . . . and have no way of recalling in which issue they saw the ad."10 Furthermore, this second group of respondents may have seen the advertisement the second time, but only remembered seeing it because they had been exposed to it once before. Studies of memory have shown that "a relevant 11 In other words, association link or tag" improves memory. it is entirely possible that a respondent will have no recall of an advertisement, then see a repeat of it and recall both the repeat and the original. This phenomenon is the result of an "interactive effect" of repetition, which will be discussed further. Herbert A. Zielske partially solved some of the shortcomings of other studies of repetition in his analysis of "The Remembering and Forgetting of Advertising."12 Respondents were mailed a copy of the same advertisement for thirteen consecutive weeks or every four weeks for a year. After each mailing, a different subsample of re- spondents was questioned on their recall of the advertising. Unlike the McGraw-Hill and similar studies of repetition, 10Howard G. Sawyer, "Are Repeat Ads as Good as They Seem?," Medialscope (July, 1968), p. 60. 11W. R., K. C., F. R. S. Bausfield, The Basis of Memor (London: Kegan, Paul, Trench, Trubner & Co., Ltd., 9 p p. 106. ‘ 12Hubert A. Zielske, "The Remembering and Forgetting of Advertising," Journal of Marketing, Vol. 23 (January, 1959), pp. 239-243. 25 responses here were meant to be cumulative. That is, re- spondents were not asked whether they recalled seeing the advertisement during a particular week, but simply whether they could recall the advertising, regardless of when, or how often, they had seen it. Under these circumstances, the first repetition (in the weekly-interval treatment), increased recall by about 13%, the second repetition by about 9%, and the third by about 6%. These percentages of incremental recall are low enough to suggest a law of diminishing returns with regard to repeat advertisements and the validity of Sawyer's contention that, when studies show constant levels of readership for repeat advertise- ments, many of the respondents are not recalling the most recent insertion but simply an insertion or combination of insertions. In any case, it should be clear from the above that researchers have attacked the problem of repetition from two different angles: (1) studying recall of a series of advertisements individually; (2) studying cumulative recall of the series of advertisements. The second method, that employed by Zielske, would appear to have a distinct advan- tage. Studies of individual advertisement recall--even if they do measure recall of the most recent advertisement—- have no way of estimating the part played in recall by earlier advertisements in the series. By using a measure of cumulative recall, however, the effect of earlier 26 advertisements can be observed with some accuracy (although a major factor preventing complete accuracy is that not all respondents are likely to have been exposed to each adver- tisement in the series). Others have tried to show that, because incremental recall declines with each succeeding repetition, advertisers would be wise to keep repeat advertisements to a minimum. A leading exponent of this viewpoint is Julian Simon.13 He cites extensive data, including a study by Politz for ngk magazine which found the first advertisement in a series reaching 3.7 million readers, the second an additional 3.2 million, and the third another 2.6 million. Simon, however, has not considered a possible benefit of repetition that does not make itself obvious in the data: the interactive effect of the repeated adver- tisement with the originally-seen advertisement. Research on recognition of rapidly-presented words has shown that speed of recognition increases as familiarity or previous exposures increase.14 In other words, less clarity is needed for correct identification of a familiar stimulus. From these findings, Walter Weiss has concluded that 13Julian L. Simon, "New Evidence for No Effect of Scale in Advertising," Journal of Advertising Research, Vol. 9 (Mar., 1969), p. 39. 14Walter Weiss, "Repetition in Advertising," Psy- chology in Media Strategy, ed. Leo Bogart (Chicago: American Marketing Association, 1966), p. 63. 27 "repeated exposures to an ad will increase the speed with which a quick glance or fleeting contact suffices to produce recognition and, as a consequence, recall of the gross elements of the ad, or at least those which are part of the 15 This character- retained content of previous exposures." istic of the repeated advertisement can be invaluable to advertisers, since readers often may not have the mental energy to give an advertisement more than a “quiCk glance" or the time for more than "a fleeting contact." Simon himself cites a study suggesting the existence of the interactive effect of repeated advertisements. He notes that a 1965 Politz study showed that for "Familiarity with Product," "Rating of Product Quality," and "Suscepti- bility to Prescribe Product," four insertions for ethical drugs in Modern Medicine produced less incremental effect 16 than did three insertions. The important point here, however, is not (as Simon suggests) that incremental effect decreased, but, rather, that there is any incremental effect at all. For if attitudes toward a product improve with repeat insertions, then repetition has a measurable quali- tative value. Even though the repetition may not seem worthwhile in a purely quantitative sense (i.e., the 15Ibid. 16John Stewart, Repetitive Advertising in News- papers:_tA.Study of Two New Products (Boston, Mass.: Harvard Business SchooI) 1964). 28 increments are decreasing), the greater possibility that a repeated advertisement will register positively on the consciousness may make the difference between a consumer's vague interest in a product and his intention to buy, or between mere intention and the actual purchase. Roger Barton makes this same point when he cites research indicating that generated attitudes are more resistant to satiation effects than the recall of learned information.17 In other words, though recall of a given advertisement may reach a maximum point, attitudes toward the item being advertised are likely to change more slowly and reach a maximum or "satiation" only after many repetitions. Another researcher, John Stewart, suggested that repetition might be necessary to carry the consumer through the hierarchy of responses leading to the purchase decision. In his study, zero, four, eight, and twenty consecutive newspaper advertisements were tested for two products. He concluded, "Perhaps the most important finding from the study was that it did require substantial repetition to achieve efficient purchase results. In contrast to the rapid rise in brand awareness, there was a time lag before the awareness transformed into purchase behavior."18 17Roger Barton, "Advertising Overkill, or How Much is Enough?," Media/scope, Vol. 12 (December, 1968), p. 66. 18 Simon, Journal of Advertising Research, Vol. 9, p. 39. 29 Studies of the Effects of Increased Space Size on Advertising Effectiveness Conclusive findings on the effectiveness of larger- space advertisements are hampered by inadequate isolation of the effects of size increases. However, those studying Space size do have an advantage in ease of measurement, since it is more feasible to average the readership scores on different size advertisements than to average out scores at various stages in repetition schedules. In the latter case, we cannot determine which advertisement exposure in the repeat schedule was most important in increasing recall. In the case of space size, we are not faced with this ambi- guity. Thus, if readership of an advertisement is twice as great when the size is tripled, we know that the increased readership is due to the larger advertisement and not the result of the combined effect of two advertisements. The main research findings on space size suggest that increments of space will increase readership, but not proportionately. Allan S. Donnahoe analyzed Starch reader- ship scores of 16,000 advertisements and found that average "noting" scores did not increase in direct proportion to 19 size. Instead, he discovered a "square root law," ac- cording to which the "per cent noting" scores tended to 19Everett M. Nelson, "Large or Small Space? What's the Objective?," Editor & Publisher, Vol. 100 (June 17, 1967), p. 20. 30 fall near the square root of the number of agate lines in an advertisement. In other words, a lOO-line advertisement would yield--on the average--a 10% "noting" score, a 200- line advertisement only 14%. In line with these findings, J. Heads found that doubling the size of newspaper adver- tisements in two newspapers in the United Kingdom, the Daily Mirror and the Daily Express, increased proven recall 20 by only 3%. Daniel Starch has offered a widely-quoted explana- tion of such findings which is both misleading and illog- ical. In the book, Advertising, Albert Wesley Frey cites Starch as his source for maintaining that, for print adver- tisements, increase in attention generally is not pr0por- 21 tionate to increase in size. Starch's reasoning is as follows: If a half-page advertisement attracts 20 out of 100 readers of a publication, a full page advertisement (assuming the additional space is used with equal efficiency) should attract only 20% of the remaining 80 readers, or 16. The total readership would then be 20 + 16 or 36, less than twice the readership with exactly twice the space. The flaw in this reasoning should be apparent. When the larger advertisement appears, the new readership 20J. Heads, "Ad Sizes and One Ad Recallers," Journal of Advertising Research, Vol. 8 (December, 1968), p. 27. 21Frey, p. 314. 22Ibid. 31 scores should be computed from scratch, not by adding a predetermined percentage increment onto the smaller-adver- tisement readership. Starch assumes that incremental readership can be computed by taking 20% of the remaining readers. This is an invalid approach because the larger advertisement is not perceived as two halves, with a certain percentage of persons responding to a half of the advertise- ment and the same percentage of the remaining Subjects re- sponding to the whole. When the larger advertisement appears, everyone's response is a response to the larger advertise- ment in its entirety. How many readers the advertisement attracts will depend upon the response thresholds of the respondents, and there is no reason why positive responses cannot be doubled or more than doubled by doubling space. In other words, the effectiveness of a large advertisement cannot be estimated simply as the effectiveness of the sum of its parts, as Starch has done. Indeed, several space size studies show that incre- ments of size can yield equally large or larger increments of advertising readership. An average of studies reported by Media/sc0pe show the following levels of effectiveness, on an index basis, for advertisements of various page . 23 Sizes: 23"Effect of Space Size in Business Papers," Media/ scope, Vol. 10 (August, 1966), p. 53. 32 2 pages - 132—150 1 page - 100 1/2 page - 44 1/3 page - 33 1/4 page - 29 While relatively little difference was found between read- ership levels for the l and 2-page advertisements, doubling the space from l/2 to 1 page more than doubles the reader- ship. Tripling the size of a 1/3 page advertisement triples the readership, and going from a 1/4 to a l-page advertise- ment almost quadruples readership. In the main, then, these figures contradict Starch's assumption of inevitable diminishing returns. In 1957, McGraw-Hill obtained medians of "reader interest" for various advertisement sizes when it asked readers of 189 issues of 19 publications to indicate every advertisement that "was of interest to you when you first went through the issue."24 Results showed reader interest in the advertisements increasing in almost direct proportion to size. Using sales as a barometer, McGraw-Hill found in- crements of space size yielding even larger increments of sales. For advertising of a six-book set called "Library of Power Plant Practice," McGraw-Hill found that a 50% increase in space produced an 83% sales gain.25 241bid., p. 56. 251bid., p. 57. 33 As is the case with repetition and its "interactive effect," larger space size also offers a qualitative feature not fully revealed by the data. The intangible asset of a large-space advertisement is that it can create an "aura of importance" and "feeling of dominance" about a product,26 and, in so doing, "impress the trade" and "steam up the sales troops,"27 not to mention the consumer. According to a report issued by McGraw-Hill's Laboratory of Advertising Effectiveness, not only do larger advertisements draw more readers, but "reader comments also indicate that larger space permits the advertisers to get across a more complete message and a better impression of the company."28 :Forgetting" as a Factor Affecting the Relative Effectiveness of Larger Space vs.-Repetition The data on both repetition and space size vary considerably, but in a somewhat similar fashion. A repeated advertisement (total space doubled) and an advertisement twice the size of another each yield increments of reader— ship, but readership is not likely to be increased in pro- portion to the space increase (i.e., readership probably will not be doubled). However, positive response to an 26Nelson, Editor & Publisher, Vol. 100, p. 20. 27"Effect of Space Size in Business Papers," Media! scope, Vol. 10, p. 83. 28 Ibid., p. 55. 34 advertisement is more likely to be doubled, or more than doubled, if (1) in the case of an advertisement twice the size of the original, the response criterion is reader interest in the advertisement, or (2) in the case of a repeated advertisement, we consider a second impression on an individual of equivalent value to reaching a different individual. Up to this point, then, no clear-cut advantage has been found for either repetition or larger space over the other, when total advertising space is the same in each instance. However, we have not yet considered one addi- tional factor--"forgetting." Advertising schedules are subject to the effects of the passage of time, and a lower rate of forgetting can be expected for a regularly-running advertisement than for a single insertion. Lucas and Britt cited two benefits of repeating an advertisement: (1) it combines "the influence of the new with the residue of the old;" and (2) "impressions established through spaced repe- titions . . . fade more slowly in each successive stage."29 Clearly, the first benefit--the "interactive effect“ of repetition--is responsible for the second, the slowing down of the forgetting process. 29Darrell Blaine Lucas & Steuart Henderson Britt, Advertisin Psychology and Research (New York: McGraw- HiII Book Co., 1950), p. 81. 35 Most studies of repetition do not even account for forgetting. In such studies, readership is measured only after repeat exposures, not before, so the rate of forgetting is not determined. Zielske's study, cited earlier, is an exception. Measurements of advertising recall were made every week, and respondents exposed to advertisements every four weeks were found to have extremely high rates of for- getting between advertisements. Three weeks after the appearance of the first advertisement, recall dropped from 14% to 3%. With succeeding repetitions, the percentage of respondents forgetting the advertising decreased sharply. Zielske-maintained that some forgetting also occurred in the weekly advertisement treatment, "but one week was the smallest unit of time measured in the study."30 The key finding of Zielske's study was that the presence of a high rate of forgetting makes continuous repetition of an advertisement more desirable in many cases than a more highly-concentrated schedule which stops en— tirely after a relatively short period of time. Although one-third more different subjects at least temporarily remembered the advertising in the once-a-week treatment, the average weekly number of housewives remembering the advertising was higher in the once-every-four-weeks treat- ment. In addition, there was little remembering of the 3OZielske, Journal of Marketing (January, 1959), p. 240. 36 advertising over a large portion of the year among respond- ents subjected to the concentrated schedule. A single insertion of a large advertisement can be considered an extreme case of Zielske's concentrated schedule, with an even lower average weekly recall over an extended period of time. As high as the rate of forgetting was in Zielske's experiment, it might be expected to be even higher in the case of actual newspaper advertising. In Zielske's study, the advertisements were mailed to the subjects without any other materials. If the women had been exposed to the advertisements in a newspaper, which have to compete with dozens of other commercial and news messages,.the impact of the advertising would very likely have been dimmed and "forgetting" accelerated. The relative effects of repetition vs. discontinuity over the long term are demonstrated by a McGraw-Hill study on advertising for Kaiser Engineers.31 Early in 1958, recognition of Kaiser Engineers was equal (15.4%) among readers of general business publications and readers of publications of the mining industry, although it had been increasing more rapidly in the former case. At that point, advertising was discontinued in the general business mag- azine, while it was continued, at the previous level, in 3J'Malcolm R. Ochs, "The Fight for Awareness," Media/sc0pe, Vol. 10 (July, 1966), p. 59. 37 the mining publications. By late 1962, recognition had fallen to 12.2% among readers of the general business pub- lications; it had climbed to over 32% among readers of the mining publications. Studies of how rates of forgetting vary with changes in space size are noticeably lacking in the literature. From what is known about increased reader interest with exposure to larger advertisements, we can infer that for- getting of a larger advertisement would occur at a lower rate than forgetting of a smaller advertisement. Present Status of the Problem The foregoing summary points to the fact that ad- vertising researchers have far to go before arriving at definitive answers to questions on the relative effective- ness of repetition and size of advertisements. The lack of conclusive data stems largely from two shortcomings of past studies. One of these shortcomings has been the inability of researchers to isolate the effects of size and repetition of advertisements from other factors affecting readership potential. For instance, Dik Twedt, in a factor analysis of 34 advertising variables, found Space size to account for 12% of the total variance in readership of the adver- tisements studied. Yet, "pictorial and color aspects of the ad," "typographic size and variety," "informative 38 nature of the ad," and "influence of surrounding matter" together accounted for 55% of total variance.32 The ques- tion is, how can any of these factors be separated from the variable of space size, since each of them are contingent upon Space size? Similarly, although "influence of previous schedule" accounted for only 1% of total variance in Twedt's study, the foregoing analysis of repetition has suggested that the previous advertising schedule can affect perception of all the other variables mentioned by Twedt. As one author states, "repetition substitutes recognition for recall . . . and recognition, actually re-cognition, a knowing again, is always easier than recall."33 Conse- quently, a factor like "pictorial composition" would improve an advertisement's readership potential more when the ad- vertisement is repeated. The second shortcoming of past studies involves experimenters' failure to handle the two variables simul- taneously, studying the effects of forgetting upon each in a study which could control for most intervening variables. Possibly the only major simultaneous studies of space size and repetition were conducted by Edward K. Strong, in the early part of the century. In 1914, Strong 32"Effect of Space Size in Business Papers," Media/ scope, Vol. 10, p. 53. 33Wood, p. 104. 39 compared the levels of recognition of monthly, l/4-page advertisements running for 4 consecutive months, l/2-page ads running every other month, and l-page advertisements running once. He found that the efficiency of space in- creases approximately as the square root of the area, while the efficiency of repeated advertisements increases approx- imately as the cube root of the area.34 From this, Strong concluded that a single, one-page advertisement would bring about more recognition than would two half pages appearing at monthly intervals. However, since his intervals between advertising exposures were always at least one month apart, the incidence of forgetting apparently played an important part in even the repeat-advertisement schedules, thereby preventing the interactive effect of a weekly repeat sched— ule from showing up in the results. Strong himself sug- gested that "there is an Optimum length of interval between successive presentations," and "more recent work of mine shows clearly that intervals of . . . a week are superior to that of a month."35 However, two years later Strong presented data showing that the same total amount of space is more effec- tive used in large amounts, less often, than when used in 34Edward K. Strong, Jr., "The Effect of Size of Advertisements and Frequency of their Presentation," Psy- chological Review, Vol. 21 (March, 1914), pp. 136-152. 35 Ibid., p. 152. 40 smaller amounts at-weekly intervals.36 This conclusion seems surprising in view of the fact that he also found the attention value of large advertisements less than propor- tional to their size. His conclusion that larger space, less often, is preferable to smaller space at weekly in- tervals is the result of the poor performance of repeat advertisements in this study. This poor performance, how- ever, can be blamed on the artificial nature of Strong's experiment. Subjects were instructed to look through "dummy" magazines which contained only the experimental advertisements. Under such conditions, levels of adver- tising exposure could be expected to be very high. With exposure high initially, increments of recognition from exposure to repeat advertisements would tend to be small-- small enough to erase the initially greater proportional effectiveness of the smaller advertisements and the greater potential for forgetting over time among those exposed to fewer large advertisements. When considering the results of studies of advertising recall carried out by McGraw-Hill under more realistic conditions, it appears likely that Strong's results underestimated the ability of a repeat advertisement to increase advertising recall under the more natural conditions prevailing in an actual newspaper or magazine. 36Edward K. Strong, Jr., "The Factors Affecting a Permanent Impression Developed Through Repetition," Journal of Experimental Psychology, Vol. 1 (February, 1916), pp. 319-338. 41 The only conclusion, then, that can be gleaned from existing data, is that, when the objective is to maximize recall of the advertising, no clear-cut advantage exists between giving over a certain amount of space to a smaller number of large advertisements or choosing instead a larger number of small advertisements. In both cases, there is likely to be an increase in recall that is less-than- prOportional to the increase in space size. However, when it becomes important to sustain advertising recall over a protracted period of time, the strong negative effect of forgetting suggests the superiority of a smaller-advertise- ment, greater-repetition type schedule. Undoubtedly, there is some Optimum point in any situation beyond which it would be unwise to further out advertisement size and increase frequency of insertions, but researchers have barely begun the work of establishing guidelines that will help determine that point. CHAPTER III PRESENTATION, ANALYSIS, AND INTERPRETATION OF DATA Results of interviews with PAC coupon book-holders during six calling periods prior to the opening of each play (as explained earlier) are presented on the three pages following. The first page of data relates to coupon- holders' awareness of the next PAC production; the second, to their recall of the advertising for that production; the third, to their interest in attending the play. On each page, responses to a question are grouped into columns of "positive," "negative," and "total" responses for each calling period. For the "awareness" question, "Can you tell me the name of the next Performing Arts Company production?," positive responses are composed of the total of those who "gave correct response and seemed sure of it," "gave correct response but did not seem sure of it," and "mentioned cor- rect response among several possibilities." (See question- naire forms, Appendix A.) All three of these responses were considered positive because in each instance mention of the correct play may have been the result of exposure to the experimental advertising. 42 43 Table l.--Respondent awareness of next production. 4 A Tabulation of Responses Tabulation of Responses Large Advertisement Treatment Group SmalliAdvertisement Treatment Group ‘-_-T'7;_7' >1 1 a) H m m 0 LI m m o «0 >. m:» p :0 >4 «a» a 2:0 m 010 a 2:0 m 020 Calling a -5 2.1 :32: a .5 a 2:23 Per-1001* I; as .= 88... s as .c 88.... H us p Hmu H as u umu a was 0 ‘ h - w .. m m "g. 8 is 8. .. 9;. Calling g g a) I? a) :0 2 cu m a) :0 Period* m c: H 4J (nu—I s: -—I 4J (OI-l \\ 044 a. as:a. -044 a. sszm u FLC 0:) P lot: 010 m +Jm .c 0 am: a +Jm .a O£L£i u +Jc .9 many u I +1: u $4w4u m was a o m was 0 o 2 run m mmm 2 Am m mmm RecalIed ‘ ' Advertising, 5 12 17 24% 13 18 31 41% Did Not lb Recall Adv. 34 19 53 76% 27 18 45 59% Total ‘89 31 70 100% 40 36 76 100% Recalled 5.2. _Advertising 21. 26 47 54% 25 22 47 55% Did Not la Recall Adv. 24 16 . 40 46% 13 25 38 45% Total 45 42 87 1008 38 47 85 100% :Recalled ' i 7 . ii Advertising 11 13 24 33% 19 10 29 45% Did NOT; ‘ I h, [I 2b Recall Adv. 19 ‘ 30 3 49 77%: 22 13 5~85 55% Total 30 43 73 100% 41 23 64 100% Recalled .Advertising 21 24 45 54% 23 28 51 61% . id Not 2a Recall Adv. 22 ‘16 , 38 46% 16 «17 33 39% Total 43 40 83 100% 39 45 84 100% .———+ Recalled Advertising 34 15 49 61% 16 12 28 55%; Did Not I 3b Recall AdN- 16 15 . 31 39% 14 9 23 45%~ Total 50 30 80 100% 30 21 51 100% - Recall?d. 30 31 61 73% 22 20 42 57% Advertising ' _ id Net . 3a Recall Adv. l4 9 . 23 27% 19 13 32 43% Total 44 4o 84 100% 41 33 74 100% “D _V *Calling periods are represented by the abbreVlations at left. The numbers 1, 2, and 3 refer to the week of the experiment during which an advertising insertion appeared. The letters "b” and "a" tell whether the calling period was before or after the advertising insertion date. 45 Table 3.--Respondent plans to attend next production. TabuIation of Tabulation of Responses _yy Responses Large Advertisement Small Advertisement Treatment Group Treatment Group, 8 m 1 . r: m . 8.141... "3 w >~ mg) 8 s. s 88 Period* 8 22 ,3 4; $3 (‘3 c: .—| .p mu—I \ (1)-H m cam \ (1)-H m can. u FMS Q)O +: F4: 010 a urn r: OILS m +1m J3 Otis .. 1:8 ‘5 888 :4 13.5 s 888’ g .401 00 0.0400 2 am 00 0.0400 Would ‘ Attend 34 ‘ 23 ’ 57 81% 29 20 49 64% Would Not lb Attend 5 8 13 19% ll 16 27 36% Total 39 31 70 100% 40 36 76 100% Would ' ' Attend 32 34 66 76% 26 25 51 60% would Not ' la Attend l3 . 8 ' 21 24% 12 22 34 40% Total 45 ‘ 42 , 87 100% 38 47 85 100% WouId ‘—' , ’ ' Attend 17 36 53 73% 25 8 33 52% WOuld Not ' 2b Attend l3 7 20 27% 16 15 31 48% Total 30 p 43 . 73 100% _ 41 23 .‘64 100% . Would Attend 28 37 65 78% 28 24 52 62% Would Not ‘—“‘ 2a Attend 15 ' 3 ‘ 18 22% ll 21 32 38% Total 43 40 83 100% 39 45 84 100% Would ' ' Attend 34 24 58 72% 22 ll, 33 65% Would Not ‘ 3b Attend 16 . 6 22 28% 8 10 18 35% Total 50 30 80 100% 30 21 51 100% Would Attend 24 ‘ 29 ‘ 53 63% 27 19 46 62% ould Not 3a Attend 20 ' 11 ’ 31 37% l4 14 28 38% Total 44 40 84 100% 41 33 74 100% .___H. *Calling pefio s are represented by the abbreviations at left. The numbers 1, 2, and 3 refer to the week of the experiment during which an advertising insertion appeared. The letters "b" and "a" tell whether the calling period was before or after the advertising insertion date. 46 For the "interest," or "plans to attend," question, those responding with "yes" or "probably" were grouped together as positive, and on the "recall" question, "yes" and "thinks so" responses were classified together as pos- itive. Grouping responses in this manner, rather than analyzing them individually, was believed to be advantageous because of (l) the relatively small size of the subsamples and (2) the possibly nebulous distinction between a "yes" and a "probably" or a "yes" and a "thinks so" during the course of a telephone interview. For example, whereas respondent x might say "yes" and mean "probably," respondent y might say "probably" and mean "yes." Note that column totals refer to responses for both of the large—advertisement (L) treatments and both of the small-advertisement (S) treatments. As we have already emphasized, by combining responses in this manner, several possible biases have been eliminated: bias caused by student-non-student differences, differences between the two newspapers, and differences in advertising perception between the two plays. Each of the variables in each pair of variables was divided almost equally between comparable subsamples. The only factor that could not be controlled for was the fact that different subsamples contained dif- ferent play-respondent combinations (e.g., students exposed to Marat/Sade advertisements made up a part of the L treat- ment group, but no students in the S group were exposed to MaratZSade advertisements). 47 The responses during each calling period in both the L and S groups have been converted into percentages. These percentage figures appear at the bottom of each column. Since limitations were imposed on interviewer time, and because some interviewers experienced difficulty in finding adequate numbers of qualified respondents, the number of respondents varies somewhat across calling periods and be- tween treatment groups. For this reason, the percentage figures frequently will be referred to for purposes of analysis rather than the raw data used in Chi-square analyses. Generally, data will be analyzed separately for each question, with emphasis on the statistics for advertising recall. The following table presents positive reSponses as a percentage of total responses for each of the three meas- urement criteria and during each of the six calling periods. This table will serve as a convenient reference for much of the material that follows. Table 4.--Positive Responses as a Percentage of Total Responses During Each Calling Period Awareness of Upcoming Plans to Advertising Production Attend Recall Calling Period L-group S-group L-group S-group L-group S-group 1b 51% 53% 81% 64% 24% 41% 1a 60 61 76 60 54 55 2b 68 52 73 52 33 45 2a 69 71 78 62 54 61 3b 78 75 72 65 61 55 3a 76 74 63 62 73 57 48 Awareness of Next PAC Production A line graph representing fluctuations in the percentages of positive responses over the six calling periods for both the L and S treatment groups on the "awareness" question, appears on the following page. In calling period lb, the lack of significant differences between treatment groups suggests that, prior to adver- tising, the interest-arousal capabilities of advertising would be similar in the two groups. With the appearance of the large and the first small advertisements, awareness of the plays increased almost equally in each treatment group, so that the two groups remained roughly equal in levels of positive re- sponse during la (x2 = .03).37 37All significant probabilities (less than .05 under the null hypothesis) and their corresponding X2 values are presented in the main body of the manuscript. Probabilities greater than .05 and their X2 values also are presented in the main body of the manuscript if they relate specifically to one of the hypotheses being tested. All other X2 values can be found in Appendix B. 49 80%> I 75%. 70%4 65%m ON C T U1 U! T U1 0 W .1} U1 i 40%” 35%“ 30%l Percentage Positive Responses zssn 20%- 15%~ L-group 1'0 %- — —————— S-group 5%., T l l I 1 T lb la 2b 2a 3b 3a Calling Period Figure 6.--Respondent awareness of next production (percentage positive responses). 50 Hypothesis 1 The validity of null hypothesis 1 now can be eval- uated. The hypothesis states that: There is no difference in awareness of the name of an advertised play between persons recently exposed to a small advertisement and those recently exposed to a large advertisement. The null hypothesis is not rejected (probability greater than .80 under the null hypothesis). Since the proportion of positive responses on the awareness "cri- terion" was almost identical between the two treatment groups immediately following, as well as immediately pre— ceding, the appearance of the initial advertisements, it may be concluded that the large advertisement was not significantly more effective than the initial small adver- tisement in increasing the level of awareness. Chi-square analysis was employed for comparisons of positive and negative responses for le vs. Lla, Slb vs. Sla, and for the totals (L + S) lb vs. 1a. In all three cases, X2 was not significant. The fact that differences between the lb and la calling periods were so small may be attributed to the initially high levels of awareness in 1b, which may have restricted the potential for increase as a result of the initial advertisements. High levels of awareness prior to any advertising should not seem surpris- ing, since respondents were members of a "pre-sold" group, PAC coupon book-holders. 51 Going from period la to 2b, positive responses increase in the L group, while they fall off in the S group, resulting in a significant difference between L2b and 82b (x2 = 4.70, probability less than .05). Whereas a decrease in awareness was evident in the S group, heightened awareness seemed to result in the L group. Data suggest that perhaps the larger advertisement did provide greater impact, although this impact was not discernible in the la period. A possible explanation of the increase in L2b is that the greater impact of the larger advertisement gen- erated word-of-mouth advertising passed on by those who had seen it, whereas the smaller advertisement did not produce a similar effect. At best this interpretation must be regarded cautiously, since changes in response levels were not significant in either group. (For additional interpre- tation, see discussion of non—experimental publicity in the following section of this chapter.) A significant difference was noted between L2b and S2b (x2 = 4.70, probability less than .05). Although nei- ther treatment group had changed significantly from their positive response levels during 1a, response levels changed in opposite directions, thus accounting for the difference in 2b. After the appearance of the second small advertise- ment, however, S responses rose significantly over their 2b level (x2 = 6.09, probability less than .02). possibly, 52 because of its interaction with the memory of the first small advertisement, the second advertisement was more effective than the first in establishing conscious aware- ness of the plays. During the fifth and sixth calling periods, the percentages increase moderately, then level off from this plateau. Since other sources of publicity were abundant during these final calling periods, the data probably pro- vide a poor reflection of the effectiveness of the adver- tising per se. Thus, the small decrease in positive re- sponse after the appearance of the third small advertisement can be attributed to the fact that publicity for the play already may have reached a saturation point. Those who were susceptible to communications regarding the plays, had already received enough communications so that they could no longer forget about the plays. Finally, an analysis was made of the total of all responses received before the appearance of an advertisement (le, Slb, 82b, S3b) vs. the total of all responses made after an advertisement's appearance (Lla, Sla, 82a, 83a). X2 was 4.31, probability less than .05. This value suggests both the influence of the advertising and the importance of the "forgetting" factor. 53 Advertising Recall The second graph presents the percentage positive responses to the question on advertising recall. These responses will be analyzed somewhat more intensively than the data on "awareness" and "plans to attend," because re- sponses to this question should give the most reliable indication of the actual patterns of advertising effective- ness in the experimental schedules. A myriad of factors might influence awareness of the next play or plans to attend, but only the advertising itself should determine the nature of response to a question on advertising recall. Vertical Analysis (Between Calling Periods) That the initial levels of positive response are as high as they are (in lb) can be attributed to previous exposures to advertising of the season coupon book (in which Marat/Sade and Little Mary Sunshine were both referred to) and to other productions of Marat/Sade (the film version of the play recently had appeared both on and off-campus, and East Lansing High School had presented its version of the play several weeks before calling period lb). A significant rise in the percentage of positive responses was found between periods lb and la, for the two treatment groups combined, and for L alone. X2 for L and S combined was 15.15, probability less than .001. For L alone, before vs. after the large-advertisement insertion, 54 80%1 75%w 70%“I 65%- 60%- 7* 55%: a. / ‘0’ ’ A p Ln 0 m <3 w w c» 1 L n \ OJ UI 09 g Percentage Positive Responses (» to w A 25%- 20%- 15%— L-treatment —————— S-treatment 10%- 5%- _l A t I T l r i lb la 2b 2a 3b ' 3a Calling Period Figure 7.—-Respondent recall of newspaper advertising (per- centage positive responses), 55 X2 was 14.22, again significant at the .001 level. Al- though the X2 value for the S group was not significant, its probability under the null hypothesis was less than .10. The significant increases in advertising recall going from lb to la constitute reasonably conclusive evi- dence that the first insertions were read and digested. The percentage increase here was more than three times as great for the L group as for the S group. Clearly, the large advertisement was more effective insofar as respond- ent's recall of the advertising was concerned. However, it will be remembered that the large advertisement did not seem significantly more effective than the small advertise— ment in fostering awareness of the plays during the la period. From this, we may logically conclude that many persons who are made aware of a product or service by a small advertisement will not remember that advertisement as being the source of their awareness. The dubious advantage of the large advertisement, in this case, is that although L-group respondents did not remember the information commu- nicated by the advertisement any better than did S-group respondents, there was a stronger tendency for L-group respondents to recall the advertisement itself after being reminded of the play. The drop in positive response going from la to 2b was very similar to the increase that took place between 56 the preceding calling periods. X2 for the combined L-S group was 7.78 (probability less than .01); for the L group, X2 was 4.65 (probability less than .05); and for S, X2 was 1.46, not significant. Hypothesis 2 The validity of null hypothesis 2 now can be eval- uated. The hypothesis states that: There is'no‘difference_in the level of recall of an advertisement one to two days after exposure and the level of recall four to six days after exposure. The null hypothesis is rejected for the L and S groups combined (probability less than .01 under the null hypothesis). For the L group, the null hypothesis again is rejected (probability less than .05 under the null hypo- thesis). For the S group, the null hypothesis is not rejected (probability greater than .10 under the null hypothesis). The significance of the total drop in percentage positive response between periods la and 2b provides strong backing for the proposition that forgetting does occur, even in a period as short as several days after an adver- tising exposure. In the "awareness" analysis, it will be remembered that no conclusive evidence of forgetting was found between the la and 2b calling periods. Logically, any incidence of forgetting would be more obvious with regard to advertising recall than to awareness of the item 57 advertised. Conceivably, a respondent could remember the name of a play after he had forgotten where he came upon this information. The reverse situation, however, would be a logical impossibility (knowing where he had come upon information that he has forgotten). Also, it should be recalled that extraneous variables, such as the word-of- mouth that may have been generated by the large advertise- ment, are likely to increase awareness but should not influence advertising recall. The fact that the drop in positive response level (between la and 2b) is not significant may be related to the smaller increase in recall in that group (relative to the L group) between the preceding calling periods. This relationship will be discussed further in the correlation analysis that follows. Between periods 2b and 2a, significant increases occurred: for L and S combined, 10.29 (probability less than .01); and for L, 8.79 (probability less than .01). No significant difference was found for S, although the X2 probability value was less than .10. The above increases are misleading, since other communication variables probably were already working to a significant degree (see section on non-experimental pub- licity near the end of this chapter). There could be no other explanation for the increase of 21 percentage points between L2b and L2a. 58 No further significant changes in positive response levels were found between successive calling periods. Response percentages rose slightly from period L2a to L3b and fell off slightly for the corresponding periods 52a to S3b. There were insignificant gains in positive response in both treatment groups between calling periods 3b and 3a. The fact that S group responses remained at fairly constant levels during the final calling periods, despite pervasive non-experimental publicity, suggests that adver- tising had already reached a saturation point in the S group by calling period 2a. However in the L group, pos- itive response increased enough during the 3b and 3a calling periods so that L3a positive response was significantly higher than positive response for 83a (X2 = 4.36, proba- bility less than .05). A possible explanation for the widened gap between the two treatment groups is that a considerable number of students (S group) were immune to any communications involving Little Mary Sunshine, while many non-students (also S group) rejected all communications relating to Marat/Sade. Thus, the S group peaked during the 2a period, while the L group, as suggested by the higher levels of positive response on the "plans to attend" question (see section on "plans to attend," this chapter), may have had a greater potential for positive response. X2 analysis compared total before-insertion re- sponses (le, Slb, 52b, S3b) with the total after-insertion 59 responses. X2 was 15.81 (probability less than .001). As would be expected, positive response is significantly higher in the "after-insertion" periods, and the difference between "before“ and "after" periods is more pronounced in the case of "recall" than it was with respect to "awareness." A major influence on the above data is, of course, the experimental advertising. Unfortunately, combining the data in this manner cannot conclusively prove the value of repetition in increasing recall and retarding forgetting. This shortcoming must be blamed, once again, on the effects of publicity extraneous to the study. James Playsted Wood maintained that "memory must be roused from a latent to an active state by repetition of the original stimulus or something which brings the original stimulus into play again."38 During the later calling periods, many extraneous stimuli were bringing the experimental advertising into play again (or being confused with it). Consequently, fluctuations in the data during these periods do not accu- rately reflect the effects of repetition and forgetting of the experimental advertising. Nevertheless, the more marked incidence of forget- ting throughout the advertising recall analysis (as compared to the awareness analysis), bears out the contention, sug- gested earlier in the chapter, that an advertised item 38wood, p. 104. 60 tends to be remembered after the actual source of informa- tion about that item (advertising recall) has been forgot- ten. The nature of the relationship between advertising exposure and advertising recall and forgetting is explored further in the following section. Correlation Analysis (First Three Calling Periods) Additional vertical analysis involved a correlation of percentages of positive responses for the first three calling periods of the L group with those of the S group. A cursory glance at the data, and the fact that graphed coordinates of the data fell in an almost straight line, suggested a high positive correlation. Computing the coefficient of correlation (see Appendix B), an almost perfect correlation of +1 was obtained. (Both graphs of correlations in this study appear on the following page.) Perhaps the most significant pattern in the data is that the rate of changeover to positive response going from lb to 1a was the same as the rate of changeover to negative response going from la to 2b, and this relationship held true for both the L and the S groupy. In the L group, 39% of those respondents answering negatively during period lb, answered positively during la; and 39% of respondents answering positively in la, answered negatively during 2b. In the S group, the corresponding rise and fall was 18%. 61 55- la 50. 45- 40. 35- 30- 25J 1b 20- 154 10- r = +1.00 2b L Group U1 1 0 I I j i I I I I I I I 5 10 15 2025 30 3540 45 5055 S Group Percentage Positive Response Percentage Positive Response Figure 8.--Correlation Analysis: Recall of newspaper advertising—-L vs. S groups (percentage positive responses). 65- 60‘ la 55‘ 1b a 50* 45— 40- 35 .I. 30- 25" 204 15‘ r = +.93 10— 57 0 02b l i I I I | I i’1 i 5101520253035 455055 (Reca11-—S Group) Percentage Positive Response Percentage Positive Response (Awareness-S Group) Figure 9.--Correlation Analysis: Smallead treatment groups-- awareness of next production vs. recall of news- paper advertising (percentage positive responses). 62 This virtually perfect correlation across treatment groups and calling periods (see Appendix B for calculations) has several implications, the likelihood of which are en— hanced by the fact that non-experimental publicity was minimal during the first three calling periods. First, this correlation may suggest a possible explanation of the nature of attitude change. One author has stated, "most studies of the immediate and long-range effects of a single exposure to a communication show that there is a tendency for attitudes to return to the positions held before ex- posure to the communication."39 Indeed, the data suggest that attitudes tend to return to their initial levels as forgetting of a first exposure occurs. Another finding suggested by these data is that forgetting occurs rapidly after only one advertising inser- tion. Backing up this assertion is the fact that the rate of decrease several days after the first insertions was equal to the earlier rate of increase immediately following these insertions. In other words, it appears likely that many of the respondents who went from negative to positive response between times 1b and 1a, also went from positive to negative between 1a and 2b. This phenomenon is almost equally evident in both treatment groups. 39Martin Weinberger, "Does the 'Sleeper Effect' Apply to Advertising?," Journal of Marketing, Vol. 25 (October, 1961), p. 65. 63 Indeed, it can be demonstrated that the ratio of "percentage of respondents forgetting the advertising be- tween periods la and 2b" to the "percentage of respondents changing from negative to positive response between periods lb and la," is almost equal in both treatment groups: In the L group, 39% = percentage respondents shifting from positive to negative responses between periods 1a and 2b. 56% = percentage of total respondents answering positively in period 1a who had answered negatively in period lb 70% = percentage of those who forgot the first advertisement out of total who had seen no advertising prior to the first experimental insertion In the S group, the corresponding figures were 18%, 25%, and a ratio of 72%. The fact that the L and S ratios are so similar suggests that a predictable proportion of those who claimed no familiarity with the advertising prior to the first insertion, but did see the first insertion, soon forgot seeing it. Note that the larger advertisement appears only slightly more successful than the smaller advertisement in preventing forgetting (70% as compared to 72%). Hypothesis 3 The validity of null hypothesis 3 now can be eval- uated. The hypothesis states that: 64 There is no difference in the amount of forgetting that occurs between those exposed to a large advertisement and those exposed to a small advertisement. The null hypothesis is not rejected. The percen- tages of those who forgot the initial advertisements (out of the total who said they had seen no advertising prior to the initial experimental exposures) are almost identical (70% and 72%) in the two treatment groups. Since subsample sizes varied between calling periods, comaprison can only be made on a percentage basis. Consequently, the null hypothesis only can be accepted tentatively, without sta- tistical proof of its validity. The similarity of the percentages, however, leads to the conclusion that, while a larger advertisement is likely to be seen by a significantly greater number of persons than a smaller advertisement, those who do see it will forget seeing it just as quickly as those seeing a smaller advertisement. For purposes of more clearly demonstrating certain implications of this analysis, assume for the moment that initial levels of positive response (calling period lb) had been equal in the two treatment groups, say 24%. Converting 39% of the negative responses in the L group before the first insertion to positive responses, and converting 18% of those in the S group (see preceding page for meaning of these percentages), the positive response levels at time la would be 54% and 42%, respectively. For the L group, it is 65 already known that a "forgetting ratio" of 70% brings the positive response level in period 2b to 33%. A “forgetting ratio" of 72% would bring the S-group positive responses for this same period down to 29% (rate of forgetting = 31% (72% x 43%); 31% x 42% = 13%; 42% - 13% = 29%). We can see, then, that when both treatment groups start out with equal percentages of positive responses in period lb, the advantage of the larger advertisement, reflected in the la period, will almost have been lost at point 2b (where L would equal 33%, just four percentage points above S). The reason for this phenomenon, as we have demonstrated, is simply that, with only one advertising exposure, the percentage of forgetting is extremely high, regardless of the size of the advertisement. Now that we have determined rates of increased recall and forgetting, we can draw conclusions.as to the relative effectiveness of the l-large vs. the 3-small advertisement schedules. We have seen that larger size will influence a reader's seeing an advertisement in the first place but will not retard the rate of forgetting. Rather than advertisement size, the important variable in recall or forgetting of the advertising may be the number of times an advertisement is seen by an individual. The nearly identical "forgetting ratios“ for the two treatment groups suggest that those respondents who replied negatively in period lb played a more important part in the forgetting 66 that occurred between 1a and 2b than did those who gave positive responses during lb. In other words, repetition of an advertisement may have an important effect in retard- ing the rate of forgetting among those who have seen the advertisement (or an advertising-type stimulus for the advertised item) previously. Data from Zielske's experiment supports this notion; he found the rate of forgetting be- tween advertisement exposures to decrease with successive exposures. Consequently, we could expect the small-adver- tisement schedule of repetitions in the present study, which, by period 2b, was already registering almost as many positive responses as the L group, to gain a steadily widening advantage over the L group during the final three calling periods. Unfortunately, the data in the present study during the final three calling periods do not bear out the above reasoning. As stated earlier, the effects of non-exper- imental publicity during these later calling periods pro- duced invalid data. Extrapolation of the pattern of response during the first three periods would be an unfeas- ible alternative, since not enough is known about the effectiveness of repeat advertisements in increasing recall and retarding forgetting, or about the rate of forgetting without repetition, to quantify predictions. 67 Hypotheses 4 and 5 As a result of the effects of non-experimental publicity, no conclusive evaluation can be made regarding the validity of null hypotheses 4 and 5. However, the data do provide enough information to make analysis of these hypotheses profitable. The hypotheses state that: 4. There is no difference in the amount of forgetting of the advertising that occurs between those exposed to only one advertisement and those exposed to more than one advertisement. 5. There is no difference in advertising recall between those exposed to one large advertisement and those exposed to three small advertisements. The correlation analysis suggests that most of those who forgot seeing the first experimental advertise- ments were those who had not received any related communi- cation prior to exposure to these advertisements. If this was the case, the factor of additional previous exposure may have retarded forgetting (in the group answering pos- itively during calling period lb) and the null hypothesis would be rejected. If these findings, in combination with what has been learned about increases in recall and decreases brought about by forgetting in the two treatment groups, were extrapolated over the entire three-week period, the superiority of the S group would become more pronounced, and null hypothesis 6 also would be rejected. 68 Horizontal Analysis (Between Treatment Groups) During initial calling period lb, the proportion of positive responses was significantly higher in the S group than it was in the L group (X2 = 4.50, probability less than .05). However, no significant differences were present between groups Lla and Sla (X2 = .04). Hypothesis 6 The validity of null hypothesis 6 now can be eval— uated. The hypothesis states that: There is no difference in advertising recall between persons recently exposed to a large advertisement and those recently exposed to a small advertisement. The null hypothesis is not rejected (probability greater than .80 under the null hypothesis). However,_ since the proportion of positive responses had been sig- nificantly greater in the S group than in the L group prior to the first insertions, it may be concluded that the large advertisement generated a greater increase in recall than did the initial small advertisement. The relative increases in recall correspond closely to values predicted by Allan S. Donnahoe's square root formula (mentioned earlier) for estimating "per cent noting" scores of advertisements. The advertisement sizes used in the present experiment averaged out to 168 agate lines for the smaller advertisements and 525 agate lines for the 69 larger advertisements. Taking the square roots of these numbers, as Donnahoe suggests, per cent noting scores of 13% and 23%, respectively, are obtained. These values are considerably lower than the actual obtained values of 24% and 39% (percentage of respondents answering positively in period la who had answered negatively in lb). The higher values can be explained to be the result of the above- average interest level of respondents in the item being advertised--a PAC production,-—as demonstrated by their purchase of PAC coupon books. In spite of the fact that per cent noting scores are higher than the square root formula would predict, the ratio of these scores is very close to that predicted by the formula. According to the formula, the large adver- tisement should have generated 1.8 times as many positive responses as the small advertisement. In the present study, the large advertisement was found to be 1.6 times as effec- tive (see Appendix B for calculations). Although the S group registered a moderately higher percentage of positive responses in calling period 2b, the difference was not significant. The narrowing margin be— tween the two treatment groups can be attributed to the equalizing effect of the high rates of forgetting in both groups between la and 2b. In calling periods 2a and 3b, there were no sig- nificant differences in response levels between treatment 70 groups. As cited previously, L3a positive response was significantly greater than S3a. The effects of extraneous publicity on the data during these later calling periods have been mentioned previously and will be discussed in more detail later in the chapter. Inter-Variable Correlation (Awareness and Recall) A second correlation analysis was carried out. Percentage data suggested a similar pattern of response to the "awareness" and "recall" questions during the first three calling periods. Correlation analysis (see Appendix B and Figure 9) yielded correlation coefficient r = +.93 between positive response percentages for the questions on awareness of the next production and recall of the adver- tising over calling periods lb, la, and 2b in the S treat- ment group. This correlation indicates that the small advertisements did exert an important influence upon aware- ness of the plays, and reinforces the likelihood of the finding that forgetting of the advertising message, re- flected in a drop in awareness of the next production, occurs quite rapidly. The absence of a similarly-high correlation in the 1L group (r = +.32) may be attributed to the greater impact cxf the large advertisement, which could sustain awareness