DIFFERENCES IN KNOWEEDGE ACQUISITION AMONG READERS OF THE PAPER AND ONEINE VERSIONS OF A NATIONAE NEWSPAPER By David Tezvksbury and Scott L. Althaus Relative to traditional neivspapers, Internet-based papers provide fewer cues about nexvs story importance and give readers more control over storif selection. As a resnit, rciuiers of an online paper niaif acqnire less information abont national, international, and political events than zvoutd print paper readers. This article reports the results of a multi-day experiment which compared the differentia! effects of exposure to print and online versions of the New York Times. Online readers of the Times appear to have read fewer national, international, and political news stories and were less likely to recognize and recall events that occurred during the exposure period. The news media serve several functions in American society. Although communication scholars rarely agree on what constitutes a definitive list of the social functions of the press, the role of informing audiences about important events is often considered central.' According to this view, the news media serve a surveillance function, monitoring the activities of government and other social institutions and providing citizens with useful information about the state of the nation and the world. Af the heart of modern democratic theory, in fact, is the principle that a free society is best assured by an unhindered flow of information.- In the process of packaging and presenting news, the major media communicate to audiences what it is they should know about their world. While normative appraisals of this function include concerns about whether the news media adequately present the diversity of available information and opinion, there is genera] agreement in principle that it is important to effective, responsive democracy.•* Part of the educational function of the press is rooted in standardized news presentation formats. News selection practices and procedures provide audiences with recognizable cues about the importance and meaning of issues, thereby helping to guide the news consumption process for many people. Led by national outlets like the Nexv York Times, Washington Post, and major television network news programs, the press applies relatively standard rules to decisions regarding story selection and layout prominence, rules that incorporate the values and beliefs of news editors about what is important to report."* In the main, recent research in masscommunicationhas ewkshtiry is mi assistant professor of Speech Coiuiuutiicatioii and Scott L. Althaus ;s tin iissistant professor of Speech Coiiinniiiicntion niui Political Scifiict'. University of Vol.77. Mo. 3 Illinois al Urbiinii-Clianipai^u. Fuintin^^ far this stiuti/ u>ns provkicd by the iltUC CunipnsAiitiimii 2000 Research Bonn!at tlieUniversitifofllliiJoisiU Urhaim-Chaiupiiign. Tlieanthors thankDimui 4^7-479 mOOOAEJMC Basham and George Wickei/for their assistance in called iiii^ till' data nsciiiii this project. The mithors tilso thank Diivid Swanson, Wayne Wanta, and the anonymous reviaversfor their coininents on earlier versions of this article. DlhtLRtNCES IN KNUWLLIX^ ACQWS 457 been concerned with this process in terms of the agenda-setting role of the press.'^ However, story importance cues also play an influential role in citizen learning, the focus of the present study. Recent developments in news presentation over the Internet suggest that the traditional gatekeeper role of the press may be changing. During the piist few years, most of the major newspapers in America have launched online versions of their products." Based in the World Wide Web portion of the Internet, these online news services often largely mimic the content of the original versions. However, there are potentially important differences in presentation. In their transition to the Internet, news organisations tend to give relatively equal prominence to a larger variety of stories, and they usually present audiences with a substantiiil ability to shape the news flow7 So, in place of the newspaper front page, with its distinctive editorial function ofhighlightingimportantstories, online newsoften presents consumers with a menu-based format and a much wider variety of information options. Not bound by space constraints in the way most trad i tional newspapers are, these online news services are able to provide a larger number of stories on any given topic. Inasmuch as the traditional media represent a linear mode of news presentation, the newer online services engender a parallel structure that gives consumers substantially more choice and control. Online services provide easy access to information that is often buried in the depths of traditional newspapers, and they frequently incorporate links between news items and related stories and Web sites. These interactive elements of online newspapers mean that their editors may exercise less influence over what and how people learn about ptiblic affairs. Graber suggests that the two most important criteria readers of traditional newspapers apply to article selection are story importance cues supplied by editors and the match between story topics and their own interests.^ Participants in Graber's influential study of news reading indicated the stories that caught their attention in the newspaper, and analysis of these reports revealed that people use article location (the closer to the front of the paper, the more likely selected), headline or visual size (the larger of either, the more likely selected), and story length and repetition to help guide their decisions. When asked about articles they missed, respondents in the study noted the stories' relatively low prominence in the newspaper. Newspapersupplied cues are not the entire picture, however. Graber reports that interest in a topic can easily override the prominence criterion. For example, crime stories were more popular than international stories, regardless of the relative prominence of the two topics in the paper. Overall, however, Graber reports that 72 percent of reading comes from the first section of the paper, the section typically heavy in national, international, atid political news.'^ The focus of the present study is whether the movement of traditional news practices to computer-oriented presentation structures may change the nature of the relationship between editor and reader. The key question is whether online formats provide readers with more flexibility in story selection and whether readers will use that flexibility to personalize their news consumption. If so, then it is likely that fewer news readers will acqttire the knowledge that editors think they should have, and more readers will acquire the knowledge that personally interests them. Traditional conceptions of the role of the media in democratic societies would tend to evaluate the possibility of an increasingly "personalized" news flow in largely negative terms. Because the standardized format of a traditional newspapers creates a hierarchical arrangement of stories, readers 4do JouKSAUSM & MASS COMMUNICATIOH QuAKn:Ri\ give more time and attention to those articles that editors consider important.^" This attention-directing function of the newspaper therefore tends to leave mass audiences with similar conceptions of the importance of various issues or events, as the agenda-setting literature has demonstrated.'' Online services have tended to provide far less direction to their readers regarding what deserves attention, and so people gathering news from online sources might be learning about a wider range of topics than their print-reading counterparts. With this increased opportunity to personalize the flow of news, fewer people may be exposed to politically important stories. As a consequence, online news providers may inadvertently develop a readership that is more poorly informed than readers of traditional newspapers about the core events that shape public life.'^ This traditional assessment might nonetheless be given a morepositive gloss by those who are dissatisfied with the way that traditional newspapers have exercised their attention-directing potential. By placing more control over the flow of news in the hands of individual readers, online news audiences may become collectively more knowledgeable about a broad range of events and issues even though individually the knowledge base of this audience might be rather idiosyncratic. In this way, online news distribution might contribute to the development of "issue publics," small segments of dispersed individuals who specialize in particular kinds of public affairs information.'-' In terms of their standing in democratic theory, such a populace might compare favorably to an audience of individual "generalists" who are aware of important events but have little depth of understanding. To be sure, communication research has not been ignoring the potentially important implications of developments in computer mediated communication. Investigating the differences between media in terms of audience learning and recall is a traditional theme in mass communication research.^•' Unfortunately, relatively few studies have specifically examined the effects of differences between computer-based news presentations and other media. Onenotable exception compared audience recall of information from news presented by newspaper, television, radio, and computer.^^ The results indicate that exposure to newspaper and computer-based news results in roughly equivalent learning and that both are superior to radio and television. Edwardson, Kent, and McConnell compared information gain from television news and videotex stories."" They report that subjects exposed to videotex delivered via television recalled more information from a story than did subjects exposed to a television news version of the same story. Fico, Heeter, Soffin, and Stanley compared the differences between reading an indexed version of a newspaper (similar to a menu-based layout) and reading one with a traditional layout.''' They found that readers of an indexed paper are less likely to read stories about local, national, and international current events. Subjects in their study considered the indexed paper to be most useful for goal-directed reading, whereas the traditional layout was better for headline scanning and for its ease of use. While the results of these and related studies shed important light on an understudied topic, a limitation they share is the frequent use of relatively artificial versions of online news presentations. In summary, the way news sources based on the World Wide Web use visuals, graphics, and inter-connectivity features may encourage people to process online stories differently from how they process traditional newspaper articles.^** Additionally, the lessening of editorial control over story prominence suggests that online news consumers will expose themselves to a broader range uf topics. Given current audience demand for crime, health, and sports news,'" the greater freedom afforded by online news suggests that audiences for online services may acquire less information about public affairs and events than might audiences for traditional news outlets. Focus: Tlie New York Times 460 To better understand how the mode of online information delivery affects learning about public affairs, we devised an experiment that compared the experience of reading the print version of the New York Times with that of reading its World Wide Web incarnation. Given its status as a leader in the market, the T/i(ics is particularly appropriate for use in this study. Most of the major daily newspapers in the country take story treatment and priority cues from this and other elite Eastern papers.^" Moreover, the Times has a large nationwide readership in both its print and online forms. The paper's overall weekday circulation was approximately 1.1 million and the daiiycirculationofthenational edition was 268,000 in 1997.-' Meanwhile, the online edition boasted 5 million registered users and S00,00() unique visitors during August 1998.-- Thus, by virtue of both its relatively wide circulation and its prestige in the industry, the Tinier provided a useful focus for our research. As daily newspapers add World Wide Web versions of their products, they make a number of important editorial decisions regarding layout, use of graphics, and story selection and placement. The paper and electronic versions of the Times are relatively similar to each other in terms of informational content. However, there are differences between them which can be expected to affect tht? news consumption process. Figure 1 reproduces the home page of the Tiinf:^ from Thursday, 26 February 1998, one of the days on which our study was conducted. A small number of headlines are visible on this page; the majority of the space is occupied by links to focused sectionsand topical indices. What is particularly important is the difference between the headlines given here and the articles that appeared in the paper version of the Times. The most prominent story in Figure 1 is the story about the Israeli Mossad. In contrast, readers of the print version of the Times found stories on this topic on page 6 of the Thursday edition and on page 9 of Friday's paper. The second most prominent story in Figure 1 deals with campaign finance reform. This topic was the lead story in the next day's paper version. The electronic planner article in the top lefthand corner of Figure 1 appeared on thefrontsection of a special technology section in Thursday's print version. The CIA story appeared on page 11 of the Thursday print version front section, and the Michael Milken story appeared below the fold of the front page of Friday's print version. Of particular interest for this study is the use of indices ("news by category" in Figure 1) to guide news readers. The international index for Thursday, 26 February 1998, reproduced in Figure 2, was the page presented to a reader who clicked on the "international" category in Figure 1. The list of topics shown there is a relatively close match to the international stories in that day's national edition of the paper. All but one of the articles in the list appeared in the paper version. However, with only a few exceptions, the relative rank of stories on the list bears very little relationship with their apparent importance in the print version. For example, the journal article about Indonesia appears on the third page of the paper version, carries a 4column headline, features a photograph and a map, and is 29 column-inches FIGURE 1 The New York Times Home Page Gtaroalto • tacwit • * .S isracH Mossad Agents Cmighl Planting FliofW 'Bugm' in S^vit2e^land N<w Hofic tor Kru^ed CJ.A. Drvilopa Coxn Tk O*<nbn><* I • r»LITICS UPMkil Onipaigti Pinaitu Oti: Supporters Unable to 3reak. Senate Filibinter UWTTMM-IUMW • * • ri...tfi«it CoDtuU vmc »ydncy4e' KeOKnt liavMl II) & fMiwoid !.(.£ In » ABtiiUcr Ncs VHTK TiBci Hunii; Ucbiiri News by Category ^ ^ 'nlKIOMlilUUtl Q National O Mctrn O Politic* ^^ ^ O '^'• "^' "• ' 9 Biuiiuu a QbiliuuiM © 19.9^ The Niiv York Tinier; Compumj. Reprinted By Permission. long. Yet, this article appears last on the index in Figure 2. Prior research has demonstrated that an index-based structure may prompt readers to widen their story selection, exposing themselves toa different set of stories than they would have if using the paper version. While the index format provides readers with a clear set of choices about news content, it also severely mutes the effect of editorial presentation decisions. These indices include few of the traditional cues regarding story importance (e.g., headline size, article length, or visuals) that may guide readers' choices. For example, the second story in Figure 2, a report of how United Nations weapons inspectors do their work, began above the fold on the front page of the paper version, carried a 3-column headline, featured a color photograph and 4 black-and-white photographs, a map, and was 104 column inches long. Thus, while a reader of the print version would encounter a number of very salient cues about this story's significance, an online reader would have only the item's content and relative placement on the index to indicate importance. Most of the articles on any one day appear in both incarnations of the Times. However, there are a few stories that appear only in the paper version and vice-versa. In particular, the online version is able to present substans IN KNOWLEDGE AcQuis 461 FIGURE 2 The New York Times International News Index International Steer your jny. Fvlirwar) 26, I99« U> C.I.A. tJ.N. Si Piu.-l-sJ4Hiitf Turks Aim ui Fprm New Panv ArreMcd in Swit^crand, In Iiy^tfl, Ortminals Take li> the Cafrpaign Tmil V.S. Expecied lo Waive r>rmf *it^rM.^i<|{^<j |^ynin« f olomhia High-Tech Rxecjlives Ask for I^ccu-av nn f nreifln Worittrs j Q l l: Etrntv E Bb I ill ll Id Wh K Bowbi Indonesia With Kp«ir International We © 1998 The Neiv York Times Company. Reprinted By Permission. tially more content than is the paper version. There are links to related stories from prior days, and there are constant updates as new stories come in from the wire services. A core question for the present study is whether these differences in the presentational attributes of the two versions mean that audiences may be less likely to attend to public affairs news, content that may benefit from the presentational emphasis it receives in the print version of the paper. Setting the attributes of online papers against the story selection criteria Graber describes for print papers suggests some important differences between them.^^ The structure of online newspapers rearranges the relationship between medium-supplied cues and readers' pre-existing interests. On the one hand, the cues newspapers typically give regard ing a story's importance are both reduced and re-organized when the paper goes online. Article length and headline size cues are eliminated or muted in most online newspapers. It is often the case that only a uniformly-si/ed headline and perhaps a single paragraph are visible to readers. Only when they click on the story do additional cues become visible. On the other hand, online newspapers appear to facilitate the matching of readers with topics that interest them. The menu structure and links to related stories mean that readers can quickly identify topics that match their interests, and it means they can often delve more deeply into a topic than they can with a traditional newspaper. These elements of online news and the findings of related prior research suggest that readers of online newspapers may be less likely to read the public affairs stories that typically appear in the "front section" of most traditional daily newspapers. This suggests the following hypothesis: 462 laURS'AUSM & M.«.* COMMUNICAJIUN Qll.'MiTLRU HI: Online news readers will be exposed to fewer international, national, and political news stories than will print news readers. It may be that these groups will also differ in the amount of detail they recall about stories they have read. That is, not only may story selection differ between these groups but so may the conditions of exposure. It may be that onegroup will be more cognitively engaged with the news and so may recall more details about tbe stories they have read. Therefore, our first research question asks whether there will be observed differences in the amount of story detail readers may be able to recall from short-term memory. To the extent that online news readers are exposed to fewer public affairs stories, these readers may be less informed about newsworthy events of the day. This means they will be less likely than readers of the print version to recognize public affairs topics that appeared in the news during the time of exposure. They will also be less likely to recall information about those topics. H2: Online news readers will recognize fewer international, national, and political news topics than will print news readers. H3: Online news readers will recall fewer facts about international, national, and political news topics than will print news readers. What is unclear from prior research is whether the online readers will be less inclined to retain information about the stories tbey read. It may be that one medium encourages more depth in story reading. Therefore, it may be tbat subjects in one of the conditions will be less likely to recall details of the information they read. This leads to our second research question: Will there be group-based differences in the depth of recall about stories subjects report reading? The study consisted of a multi-exposure between-subjecfs experiment designed to assess whether there are appreciable differences in how people consume nearly identical news content over different media channels. Sample. A convenience sample of student volunteers at a large Midwestern university completed the study. The use of a student sample in this research has some advantages. Chief among them is the students' relative familiarity with the Internet. All but one of the subjects in the online group reported at least occasional use of the World Wide Web prior to the study. This level of familiarity meant that differences between exposure groups were unlikely to be due to the novelty of Internet use. An advertisement was placed in the campus newspaper and posted to bulletin boards throughout the university. The notice offered seven dollars for each hour of participation in a study on the mass media. Most of those completing the study were undergraduates (89%), and almost two-thirds of them (63%) were female. Design and Procedures. On the first day of the study (a Sunday afternoon), subjects completed a common pretest questionnaire. Following the pretest, one randomly assigned group of 47 control subjects was asked to retum after six days {42 of these subjects completed the study). They were not DIFFERENCES nv KNOWLEDGE ^cQureiTKW 463 told anything of the actual purpose of the research, and they were not asked to modify their news consumption hahits. A second group was asked to report to a computer-equipped classroom for one hour on each of the following five days {43 of 45 subjects completed the study in this group). Gathered in groups of 10 to 20, subjects in this condition used individual personal computers to access the New York Times online site (www.nytimes.com) for a minimum of 30 minutes and a maximum of 60 minutes during the sessions (most students in both exposure groups read for about 40 minutes).'^'' A second group of 42 students was asked to report to classrooms each day of the following week (38 subjects completed the study in this condition).-^ This group was asked to read a traditional print version of the national edition of the Times. Subjects in both exposure conditions were repeatedly told to restrict their news exposure to what they received in the visits, an exposure control procedure used successfully by other researchers.^'' At the end of the reading sessions on the final day of exposure (a Friday), the two experimental groups completed a short questionnaire to assess their exposure to news on that particular day. On the final day of the experiment (a Saturday) all three of the groups completed a posttest questionnaire, received a debriefing about the nature of the study, and collected payment for their participation.-' The overall pretest-posttest retention rate wa's 92%.=" There are several important advantages to this design. The use of an experimental approach provided the study with a great deal of precision in being able to identify causal relationships. The use of random assignment to condition and the inclusion of the control group provided substantial leverage in assessing the proper causal direction and ruling out alternative explanations. Because the control group subjects were not asked to behave in an unusual fashion, they acted as a "normal" group against which the others could be compared. Thus, differences observed in the experimental groups can be interpreted as patterns that resulted from exposure, not from some naturally occurring event. Conducting the experiment over a weeklong period allowed us to observe the process and effects of news consumption as they actually happen, in a cumulative fashion over succeeding days.-" In contrast, shorter experiments restrict researchers to studying the atypical circumstance in which consumers are exposed to a single news story or a single day's news, The present study avoids another common element of artificiality in studies of this sort by using the naturally occurring flow of news. Rather than having read a highly stylized version of the news, subjects in this study read the actual news of the day in precisely the format that hundreds of thousands of people across the country read it. Finally, the design limits the effects of format novelty. Measurement of subjects' reading habits was done on the final day of exposure. Four days of using the paper or online versions of the Times should have allowed subjects to develop their own preferred approach to the paper.^" Exposure Control Check. Embedded in the pretest and posttest were identical sets of measures of subjects' exposure to television and radio news. Both sets of questions asked subjects to estimate the number of days in the previous week on which they had used the respective medium and the number of hours they typically used it on those days. For each medium the answers to the two questions were multiplied. Final Day Reading Patterns. At the end of the Friday exposure sessions, subjects in the paper and online groups completed a short question464 joiiK^i.viM & MASH CoMMUf^icAi'iON QIIAKIT.KLY naire (called a mid-test here). First Seelian Raui. Early in this instrument was a question asking subjects to identify the "section" they selected first when they began reading that day. The item had response options of "international, national, and political news," "business news," "sports news," "arts and leisure news," and "other." Thirteen percent of respondents completing the question indicated "other." Following this option was a blank space in which subjects wrote in "weather," "front page," "late breaking news," and similar descriptions. All responses to the first-section-read question were collapsed to one of two categories, international, national, and political news, and all others. This renders a dichotomous variable that tests whether subjects started with Times "front section" public affairs stories or began elsewhere. Time Spent with FjKliStxtioii. Using the list of options provided for the previous question, subjects then reported the approximate percentage of time they spent with each section that day. Finn! Day Story E.xposure. Early in the Friday mid-test was a question that asked subjects to list the stories they could recall having read. This was accompanied by a set of 12 blank lines. Subjects were asked to use 1 line for each story topic. Responses to this item were analyzed by a trained coder. The coder matched the brief story descriptions with a complete list of news pieces that had appeared in the paper and online versions of the Times that day. These story items were then collapsed into the major news categories used by the Times and aggregated to categories relevant to testing the hypotheses: public affairs news (national, international, politics) and other news (New York metro; business and technology; sports, arts and leisure; and science, opinion columns, editorials, and obituaries). In a final category of "uncodable" were placed topics that could not be matched with news pieces from that day. Ten percent of responses were analyzed by a second coder. The reliability of the coding was very high (Cohen's Kappa = .94). Also identified was the number of words subjects used to describe each of the topics (Cohen's Kappa - .96). Filial Day Story Detail. Subjects were asked to give detailed descriptions of the first two stories they had listed in response to the previous question. A trained coder separated these comments into discrete thoughts. A second coder analyzed lO'^ of the cases. The reliability of this measure was acceptable (Guetzkow's U - .05). The mean number of descriptions was 5.09 for the first story (SD - 2.07} and 4.26 for the second (SD = 2.26). Memory for Events of Week. Toward the end of the Saturday posttest was a battery of 15 questions (3 for each of the preceding 5 days of exposure) designed to assess subjects' ability to recognize and describe news events from the week. The events appeared in both versions of the Times and were chosen to represent a range of topics and levels of prominence in the newspaper {the complete list of question prompts can be found in Table Al in the appendix). Recogiiitiaii of Events. Each item began with the phrase, "Have you seen or heard any news stories this week about..." and then gave a brief description of the event. The response choices were "yes" and "no." The second half of the question asked, "If yes, what exactly happened?" and provided 3 lines for subjects to explain the event. This format is adapted from the National Election Studies 1989 pilot study."" Because testing the second and third hypotheses requires an analysis of learning about public affairs news, all of the stories on which these questions were based were selected from the front section of the Times' paper 465 version. Seven of the stories were international in focus, although the United States and its officials figured prominently in 4 of them (e.g., 1 of the stories focused on President Clinton's reaction to a new U.N. weapons inspection accord with Iraq). The stories were selected to fit into one of three relative prominence categories, high, middle, and low. The page on which the story appeared in the paper version, its head line width, and its total length were all used to identify the appropriate category for the story (the assignment of category is given in the table in the appendix). As an example, the high prominence story for the second day of exposure (24 Fehruary 1998) was President Clinton's reaction toa new accord with Iraq. It appeared on the first page, had a 4-coIunin headline, and was 36.75 column inches long. The middle story for the day, the reduction in charges against two men suspected of purchasing anthrax, appeared on page 12 (3 pages into the National Report section), had a 3-column headline, and was 11.25 inches long. The low prominence event, an explosion near Algiers, appeared on page 4 (2 pages into the International section), had a 2-coluinn headline, and was 4.5 inches long.'An approximately equal number of domestic and international events appeared in each of the three levels of prominence. The prominence categorization for one event, revelations ahout Israeli spy activity in Switzerland, proved to be problematic. This story had an unusually high prominence in the T/mfs Web page but carried only low prominence in the print version. Therefore, the data for this event were removed from the all of the grouped analyses and were examined separately. A set of three indices of event recognition—high, middle, and low story prominence—was created from the remaining items. These measures represent the number of stories in each category that subjects indicated they recognized (Ms - 1.51, 1.33, and .46, respectively; SDs - 1.09,1.31, and .67, respectively).^^ Recall ofEi'cnts. The open-ended recall of event details was analyzed for accuracy. A trained coder first divided each response into discrete thoughts, and a second coder analyzed 10% t>f cases. The reliability of the unitization was very high (Guetzkow's U = .008). The primary coder then examined each statement for accuracy hy comparing it to the content of the relevant news article.-** Comments were categorized as correct, incorrect, or off-topic (e.g., "I think 1 read something about this"). A second coder analyzed lO'M. of cases. Thereliability of the categorization was fair (Cohen's Kappa - .75). These data were then collapsed to a dichotomy for each story (0 = no correct recall statements, 1 = at least one correct recall statement). A set of three indices of event recall accuracy—high, middle, and low story prominence—was created by summing these dichotomous variables (Ms = 1.25, .91, and .14, respectively; SDs = 1.06,1.17, and .39, respectively). These measures assessed subjects' knowledge of details regarding events that fell within the 3 categories. Also created was a count of the number of responses each subject gave to the open-ended prompt. This was computed by adding the number of responses, accurate and inaccurate, that subjects gave for each topic to which they responded. This procedure was used to create thought counts for high, middle, and low prominence stories.-^'^ Results 466 Preliminary Analysis: Exposure Control Check. Self-reports of television and radio news exposure were examined to determine whether subjects in the twoexperimental groups were exposed to current events news outside of the experimental setting. The three groups did not differ on these variables jOUKNMJfM & MASS CoMMUIVICAnON QllAfltKLY TABLE 1 Portion of Total Time Spent with Each Type of Article Paper Online International, National, and Political Articles 59.74 (32.55) 44.53 (30.99) 2.15* Business Articles 7.84 (13.57) 13.62 (22.12) -1.38 Sports Articles 7.68 (14.52) 13.48 (19.65) -1.49 Arts and Leisure Articles 18.53 (23.61) 17.86 (22.02) .13 Other types of articles 6.42 (20.17) 10.46 (14.65) -1.02 100% 100% Note: Degrees of freedom for t-tests are 77-79, depending on the test. Ceil entries are percentage of stories read that fall within each category. Entries in parentheses are standard deviations, t /' < ,10, * p < .05, ** p < .01, ***;; < .001 in the pretest (see Table A2 in the appendix for a report of the group means), but, as expected, their responses differed substantially in the posttest. While the media use of control subjects did not differ over time, the exposure reported by the experimental groups fell dramatically.''^ Findings: Final Exposure Patterns. HI suggests that subjects in the online group will be less likely than their paper group counterparts to have read national, international, and political news, all typical Times "front section" topics. Three variables were gathered in the mid-test to test this hypothesis: self-reports of the first section read, self-estimates of time spent with each section, and recall of stories read that day. These data were gathered immediately after the final exposure setting and measured reading patterns on that final day only. First Section Read. A significantly higher percentage of subjects in the paper group reported reading international, national, or political news first (84% versus 60% in the online group), X' (1- f^O) ^ 5.94, p = .02.-^^ This suggests that readers of the online version of the Times were willing to take advantage of the large number of story options available in that format as they chose which article to read first. Time Spent with Each Section. A more direct test of total exposure for the day comes from the measure of the portion of time subjects report spending with various sections of the paper. Group means for this variable are displayed in Table 1. As predicted by HI, subjects in the paper group spent a larger portion of their total reading time with international, national, and political articles than was the case with those in the online group.-^ The group means for the other newspaper sections are not significantly different, but they show a predictable reciprocal pattern of exposure. Thus, the online DiTrLRENCESw KNOWLEDGE AcQUKn ION 467 TABLE 2 Recall ofStoric:> Read on FimI Day Number of Stories Recalled Paper Online 1.95i 63% (30%) 41% (25%) 3.60' .98 (1.06) 1.22 20% (17%) 137, (13%) 1.90+ 1.66 (1.62) 1.44 (1.40) .89 26% (23%) 20'^. (21%) 1.31 1.03 (.96) .56 (.83) 2.50* 177, (14%) 87. (13%) 3.09" -3.65* 34% (27%) 507. (22%) -2.9r Paper Online 3.97 (2.^3) 2.98 (2.06) National 1.29 (1.06) International Politics Total Public Affairs Nezus Percentage of Stories Recalled 2.13 165 a.88) a.86) Metro .32 (.53) .53 (1.14) -1.09 57, (97.,) 77n (157o) -.62 Business & Technology .18 (-46) 1.00 (1.09) -4.29* 3% (77.) 15% (197o) -3.59" Sports, Arts, & Leisure 1.42 (1.32) 1.79 (1.10) -.99 227.. (23%) 237c. (21%) -.25 Editorials, Obits., Etc. .21 (.41) .33 (.57) -1.03 4% (77,,) 5% (9%) -.77 .27 (.58) .70 (.77) -3. IS" 3% (9%) 9% (9%) 6.32 (2.16) 7.33 (2.39) 100% 100% Tohil Other Nm's UiKodable Total All News -2.85** Note: Degrees of freedom for all t-tests are 79. Cell entries are number of stories recalled from each category. Entries in parentheses aro standard deviations. + f < .10," p < 05, ** p < .01, *** p < .{im readers in this study were less likely to attend to stories that traditionally are grouped in the front section of most newspapers. Final Day Story Exposure. Categories of stories that subjects recalled having read on the final day of exposure are presented in Table 2. Subjects in the online group recalled a significantly larger number of stories. Therefore, the data are presented as both the mean number of stories recalled from each category and as the percentage of all recalled stories that fell within each 46S ]OUKNA\liiM fr MflSS CW QlMHTEHLC category. HI suggests that subjects in the print group will be more likely to recall having read public affairs news (national, international, and political news). The aggregate measure of this news category on the left side of the table shows a marginally significant difference in the predicted direction.^'* Within the public affairs category, all of the individual items are in the predicted direction, but only one difference, recall of political news, is statistically significant The raw scores for the aggregate "other news" category show a significant reciprocal difference in item recall. People in the online group recalled having read a significantly larger number of non-public affairs stories. Again, all of the mean differences in the specific categories were in a uniform direction, but only one was significantly so. The percentagedata on the right side of the table paint a clearer picture. All of the public affairs reading differences become much sharper. In particular, the combined public affairs category accounted for almost twothirds of stories recalled by the print group, whereas it accounted for fewer than half in the online group. Additionally, the gap in national news reading approaches statistical significance (/' = .06). Not surprisingly, the pattern of "other news" reading is correspondingly muted in the percentage data. Still, this category accounted for half of all stories online readers recalled but only abouta third of those that print readers reported. In sum, the final day recall data offer relatively strong support for HI. Recall that the two versions of the Times on this day were not exact duplicates. A substantial difference between them was a larger number of articles in the online version, with 112 stories in the newspaper and 177 available at the primary Tinier Web site. Since public affairs articles comprised a larger share of the total in the print version {34"'(,) than in the Web version (28%), it would be useful to determine whether this difference alone could account for the tendency of print readers to recall more public affairs news. Toexaminethispossibility,wecalculatedavariable that adjusts for the relative availability of public affairs stories. This was done for each subject by multiplying the percentage of public affairs news recalled by the percentage of public affairs news iii'nf/nWt'in his or her exposure condition. The mean print group score on this variable was .21 (the 63'/^. public affairs story recall rate multiplied by the 34'^ public affairs story availability = .21), and the mean online group score was .12 (4r>;. x 28'^ = .12). In other words, print subjects recalled a higher percentage of public affairs stories even after controlling for differences in news availability. Consistent with our expectations, this difference was significant (t(79) - 5.18, ;> < .01). Thus, the differences in public affairs recall cannot be explained only by the relative availability of those stories. Final Day Story Detail. The first research question asks whether there will be group differences in the depth of recall about stories read on the final day of exposure. Subjects in the paper group provided significantly more detailed explanations of the two stories we asked them to describe. Whereas subjects in the online group used a mean of 4.23 thoughts (SD - 1.59} to describe each of the stories, those in the print group used a mean of 5.17 (SD = 2.04, t(79) - 2.32, p - .02).^" Thus, there is some evidence that subjects in the print group may have absorbed a bit more information about the stories they read. Unfortunately, it does not seem possihle to discern more than this from these data. Memory for Events of the ViJeek. Subjects' self-reports of stories they read on Friday, discussed above, indicated that print readers were more DtiTLHSNCEs in KNOWLEDGE AcQuismoN 469 TABLE 3 Recognition of News Stories (Closed-Ended) High Profile Stones Middle Profile Stories Low Profile Stories Paper Online Control 2.18, (1.14) 1.60, (.90) {.77) 2.16, (1.42) 1.19, (1.22) 74, (.86) .76, (.79) .47, {.67) (.38) .81, 17 • c 21. 14.90* 8.70* Note: Coll entries are number of stories recalled from each category. Entries in parentheses are standard deviations. Means with different subscripts in a piirticular row differ in planned comparisons with p < .05. Degrees of freedom for all F ratios are 2 and 119 or 120, depending on the equation. t ^ < .10, ' [' < .05, ** p < .01, '** p < .001 likely to read public affairs news. The generalizability of these data is limited somewhat by the fact that subjects were reporting their reading for a single day only, and there were some stories that appeared in one version of the paper but not the other. TTiat was most often the case when a breaking news story appeared in the online version late in the afternoon one day and in the print version the next day. The lack of story overlap on Friday afternoon would be relevant for our results only to the extent that the presence or absence of some specific topic influenced readers' likelihood of reading public affairs news that day. One way to eliminate the effects of this temporal issue is to look at subjects' exposure to news over an extended period of time. We conducted such an analysis using subjects' recognition and recall of public affairs events that occurred over the five days of the experiment. H2 suggested that subjects exposed to the paper version of the Times would be most likely to recognize brief descriptions of the public affairs news events that had appeared in both versions of the paper that week. In addition, all of these stories appeared on the same day in both versions of the Times. The only important difference between them was the style of presentation. Table 3 reports the results of analyses comparing subjects' recognition of stories. For all three of the story prominence levels, event recognition was relatively low for the control group. Both of the exposure groups recognized more events. More importantly for H2, planned comparisons between groups indicated that the print group recognized significantly more of the week's events than did the online group. Thus, H2 is supported.'" The news channel appears to have affected audience members' knowledge of public affairs events that occurred during the week.*^ A similar pattern can be seen in the analysis of open-ended questions asking subjects to list what they could recall about each of the listed events. These data are presented in Table 4. H3 predicted that subjects exposed to the paper version of the Times would be more likely to recall accurate details about the events of the week. For the most part, this expectation is supported by the data. As before, subjects in the exposure groups were able to recall 470 ]0 TABLE 4 Recall of News Story Detail (Open-Ended) Paper Online Control High Profile Stories 1.77a (1.20) 1.30, (.86) .69, (.81) 13.90* Middle Profile Stories 1.56^ (1.42) 7% • .40, (.86) 12.88*" (1.22) .28, (.51) .48, (.74) .07, (.26) Z39 Low Profile Stories f: Cell entries are number of stories recilled from each category. Entries in parentheses are standard deviations. Means with different subscripts in a particular row differ in planned comparisons with p < .05. Degrees of freedom for ail F ratios are 2 and 121. t ;7 < .10, * p < .05, **;) < .01, "* ;> < .001 more detail than were control subjects. In addition, contrast analyses indicated that those exposed to the paper version were significantly more likely to recall details about top and middle-level stories. The difference between recall of low prominence stories was not significant {p = .10).'''' The overall pattern of the data supports H3.''^ People exposed to the online version of the Tinit's were less likely to recall details about public affairs events .**-'' These analyses of H2 and H3 had omitted one of the week's events, a story about Israeli spy activity in Switzerland. While removing the story from the event recognition and recall indices improved the validity of those measures, it also provided an opportunity to focus on the effects of differences in how the paper and online versions of the Timefi are constructed. The Thursday paper article about this event appeared on the fourth page of the international section, ran under a 4-column headline and was 22.25 column inches long. Similarly, the article in the Friday paper ran on the seventh page of the international section, had a 1-column headline and was 18.25 column inches long. This placed the issue in the low prominence category for both days. In the online version of the Times, the stories about the situation were considered "late-breaking" at the time subjects were exposed to them. Thus, they were given special prominence on the online version of the paper. In particular, one story appeared on the upper right hand corner of the Times home page, the most prominent location on the site. Subjects' awareness of the situation appears to have been affected by this pattern. Members of the online group were more likely to recognize the topic than were those in the print or control conditions (Ms - .37, .13, and .02, respectively; SDs - .49, .34, and .15, respectively), F(2,120) - 10.54,;' <.()1). The same pattern was observed for recall of details about this specific story (Ms - .35, .05, and .02, respectively; SDs = .48, .23, and .15, respectively), F(2, 120) = 12.95,;' <.O1 ** ).• ' This situation points to an important attribute of online news. The spy story was a relatively recent event at the time the subjects in theonlinegroupsatdowntoreadther/mcs. Thus, it appeared in a prominent location on the Web site. When editors of the print version had weighed the story against other news of the day, however, the story was apparently judged to be substantially less important on some set of criteria. This case highlights the potential for the online editors of a newspaper to give perhaps special weight to a story because it is new."*^ It also demonstrates that, as Graber notes, salience cues supplied by the media do matter."'^ The second research question asked whether there would be group differences in the amount of detail in subjects' recall of event information. None of the comparisons between the print and online groups revealed significant differences. The only significant overall effect was observed for the low prominence issues. When offering event information about low prominence events, subjects in the paper group tended to offer more detail than did those in the control group, (Ms - 1.61, 1.31, and 1, for the print, online, and control groups; SDs = .67, .63, and 0, respectively}, F{2,43) = 3.25, ;' < .05, Thus, there is no evidence of greater event recall by one experimental group relative to the other. Discussion The data presented give a relatively clear picture of the potential differences in the interaction between reader interests and news presentation features. Readers of the online version of the New York Times were less likely tobegin their reading with international, national, and political news. They were also less likely to spend a large portion of their reading time with those sorts of stories. As an apparent result of differences in exposure, online news readers were less likely to recall having read national and political news topics that appeared in the Times and more likely to recall business and other news topics. These differences in story exposure selections resulted in significant and substantial differences in readers' knowledge about public affairs events that occurred during the exposure week. The analyses of recognition and recall data demonstrate that by reducing and reorganizing story salience cues online news formats can alter the consumption and retention of intemationaL national, and political news. This pattern was particularly pronounced for stories that editors of the paper version for the Times had considered to be of top and mid-level importance. Thus, our data indicate that the mode of delivery had its greatest impact on the most proiiiiiiciil stories of the day. That thechannel had less impact on relatively trivial events should come as no comfort to those worried about the impact of Internet news reading. We also found evidence that salience cues associated with online news formats can substantially redirect reader attention. As the story about Israeli spies in Switzerland illustrates, when online editors give special prominence to the most up-to-the-minute news, readers are willing and able to follow their lead. A similar process occurs when traditional article importance cues are removed in online news. As the online version presents fewer cues about the importance of events, it appears that people are more willing to use their own interests as the guiding criterion. That is, in the balance between newspaper-supplied cues and reader interests, online news appears to give the latter an advantage by organizing the stories topically and by supplementing articles with additional information."''^ Naturally, there are some limitations to the present study. There are always problems when one attempts to reproduce "real world" behaviors and processes in a laboratory setting. The subjects in the two exposure 472 lni!KNAU conditions may have applied difterent news reading behaviors than they would outside of this setting. However, it seems reasonable to expect that the groups would differ from the norm in a similar fashion. That is, if an effect of the artificial setting is to increase attention to news, the groups should experience an approximately equal bot)st in attention.^" Therefore, it seems reasonable to attribute a good portion of the effects observed here to differences in the news channels. Another potential concern in the interpretation of the data here is with the possibility that the observed differences between exposure groups are primarily due to lack of overlap in the content of the newspaper versions. That is, it may be that observed differences in the exposure to and recall of public affairs news topics is merely due to an overrepresentation of other sorts of information on the Web version of the Times. The data presented here suggest that differences in the recognition and recall are not due solely to differences in content between the two versions. To be sure, the extra content in the online version may serve to divert attention away from public affairs information. However, that diversion is only part of the process here. The results of this study suggest that differences in layout—most visibly, a reduction in importance cues and the use of indices—may have independent effects on the kinds of stories readers notice and read. In general, the useof college students as subjects in experiments is often a cause for concern.''' In this case, it does not appear to be particularly worrisome. As it happens, educated young adults are just the people that online service providers tend to desire most. Their familiarity with computers as well as their relatively high levels of disposable income make them a prime audience for these services. In fact, part of the impetus behind the movement of newspapers into online publishing has been a dramatic aging of the newspaper reading cohort in America. So, while a probability sample is difficult to obtain for an experimental design of this sort, the use of college students as subjects turns out to be somewhat appropriate. As use of the Web becomes more widespread in the general populace, future research in this area can take advantage of more representative samples. While there has been quite a bit of scholarly discussion of the potential impact of new communication technologies on the democratic process, little of it has featured empirical research.^^ Until very recently, the creation and use of new communication media have been uneven and invariably less dramatic than the field has expected. This tendency appears to be changing. The development and adoption of the World Wide Web represents the first sustained foray into world of computer-based information dissemination for a mass audience. Clearly, there is a need for more research that empirically examines the effects of developments in new communication technologies. Appendix mid Notf^ follow. 473 APPENDIX Table AI Event Rt'call Prompts Event Prompt Prominence 1) ...how Iraqi citizens have been thinidng about the standoff with the U.S.? Middle 2) .. .the fate of two men charged with possessing anthrax? Middle 3) .. .the price of human donor eggs? Top 4) ...new evidence regarding a series of bombings in Georgia and Alabama? Top 5) ...the latest developments in Paula Jones's lawsuit against President Clinton? Low 6) ...youth gangs in Los Angeles? Top 7) ...about President Clinton's reaction to the new accord with Iraq? Top 8) .. .questions regarding elements of Sgt. Major Gene McKinney's defense? Low 9) ...the Canadian government's new budget? Low 10) ...Senator I.ott's reaction to the new accord with Iraq? Top 11) ...Mexico's chief drug trafficker? Middle 12).. .Israeli Mossad agents in Switzerland? Low 13) .. .Library of Congress plans to post a new exhibit online? Low 14) .. .national fruit and vegetable prices? Middle 15) .. .an explosion near Algiers in Africa? Low Table Al Exposure to Television and Radio News Paper Television News (Pretest) Television News (Posttest) Radio News (Pretest) 1.78 (2.W) Control 1.84 (2.63) 2.03 (3.29) .51 .29 (1.83) 1.99 (4.77) 3.15+ (2.84) .62 1.05 (1.82) .72 .60 (2.35) (1.33) Radio News (Posttest) Online .04 .19 .69 (.17) (-96) (1-64) 3.76* Nolc: Cell entries are mean hours of exposure in the preceding week. Entries in parentheses are standard deviations. Degrees of freedom for all F ratios are 119 or 120, depending on the test. .10,* 474 .05, .01, ***;'<.001 'fM & M/\ss CoMMijMQ!77fiN QIMKHKLV NOTES 1. Charles K. Atkin, "Communication and Political Socialization," in HandbookofPoliticalCoriinnmication,cd.DanD.NimmoMMiKc'iih. R.Sanders (Newbury Park, CA: Sage, 1981), 299-328; Doris Graber, M(7ss Maiia and Aim-iicaii Politics, 5th ed. (Washington, DC: Congressional Quarterly Press, 1997); Michael Curevitch and jay C. Blumler, "Political Communication Systems and Democratic Values," in DcKwcraci/fljiiif/;^Mass MfiV/ff^ed. Judith Lichtenberg {New York, NY: Cambridge University Press, 1990), 269-89; Harold D. Lasswell, "The Structure and Function of Communication in Society," in Mass Comimmications, ed. Wilbur Schramm (Urbana, IL: University of Illinois Press, 1960), 117-30. 2. See, for example, John S. Mill, "Of the Liberty of Thought and Discussion" in Utilitaiianism. On Liberty, and Considerations on Representative Government, ed. Harry Acton (London: J.M. Dent, 1859/1972). 3. For normative appraisals, see Lance W. Bennett, News: The Politics of Illusion, 3d ed. (White Plains, NY: Longman, 1996); Craber, MaiiS Media and American Politics; Benjamin Page, Who Deliberates? Mass Media in Modern Democrac\f (Chicago, IL: University of Chicago Press 1996); fora discussion of the role of news media in a democracy, see Commission on Freedom of the Press, A Free and Responsible Press (Chicago, IL: University of Chicago Press, 1947). 4. Herbert Gans, Deciding;; What's News (New York, NY: Pantheon Books, 1979); Graber, Mass Media and American Politics; David M. White, "The Gatekeeper: A Case Study in the Selection of News," in People, Societi/, and Mass Conimnnications, ed, Lewis Anthony Dexter and David M. White (New York, NY: Free Press, 1964), 160-71. 5. Fora recentcompilation on agenda setting and its conceptual cousins, see Maxwell McCombs, Donald L. Shaw, and David Weaver, Communication and Democracy: Exploring the Intellectual Frontiers in Agenda-Setting Theory (Mahwah, NJ: Erlbaum, 1997). 6. ErikK. Meyer, "An Unexpected WiderWebforthe World's Newspapers," American Journalism Review News Link, http://ajr.newslink.org/ emcoliahtml.org/, July 1998. 7. Carrie Heeter, Natalie Brown, Stan Soffin, Cynthia Stanley, and Michael Salwen, "Agenda-Setting by Electronic Text News," journalism Quarterly 66 (spring 1989): 101-106. 8. DorisGraber, Processing the News: How People Tame the Information Tide, 2d ed. (New York, NY: Longman, 1988). 9. Similarly, another major study of newspaper reading habits found that the modal reading style (45 percent of respondents report it) is to start at the front of the paper and to "go right though" the paper, Leo Bogart, Press ami Public: Who Reads What, When, Where, and Why in American Nezvspaper:^ (Hillsdale, NY: Erlbaum, 1989), 154. 10. Graber, Processing the Ne-ws. 11. McCombs, Shaw, and Weaver, Communication and Democracy. 12. For a discussion of these concerns, see Richard Davis and Diana Owen, New Media and American Politics (NY: Oxford University Press, 1998); Russell W.Neuman, T/«'f »fHrt'o/f/if Mass v^Kt/if/icp (NY: Cambridge University Press, 1991). 13. Eor an early discussion of issue publics, see Philip Converse, "The Nature of Belief Systems in Mass Publics," in Ideology and Discontent, ed. David Apter (NY: Free Press, 1964), 206-61. 475 14. Fora review, see Steven H.ChaffeeandStacey F.Kanihan, "Learning Ahout Politics from the Mass Media," Political Commiwication 14 (1997): 42130; see also Rohert D. McClure and Thomas Patterson, "Print vs. Network News," journal of Communication 26 (1976): 23-28; Russell W. Neuman, Marion R. Just, and Ann N. Crigler, Coiumon Knowledge: News and the Con^^truction of Political Meanini^ (Chicago, IL: University of Chicago Press, 1992); John Stauffer, Richard Frost, and William Rybolt, "Recall and Learning from Broadcast News: Is Print Better?" journal of'Broadcast ing 25 (1981): 25362. 15. Melvin L. DeFleur, Lucinda Davenport, Mary Cronin, and Margaret DeFleur, "Audience Recall of News Stories Presented hy Newspaper, Computer, Television, and Radio," journalisni Quarterly 69 (1992): 1010-22. 16. Mickie Edwardson, Kurt Kent, and Maeve McConnell, "Television News hiformation Gain: Videotex Versus a Talking Head," journal of Broadcasting & Electronic Media 29 (1985): 367-78. 17. Federico Fico, Carrie Heeter, Stan Soffin, and Cynthia Stanley, "New Wave Gatekeeping: Electronic hidexing Effects on Newspaper Reading," Communication Reu-arch 14 (1987): 335-51. 18. Fora comparisonof the use of graphics in online and print versions of newspapers, see Xigen Li, " Weh Page Design and Graphic Use of Three U,S, Newspapers," journalism & Mass Communication Quarterly 75 (summer 1998): 353-65. 19. Pew Research Center, The Internet Nezus Audience Goes Ordinary: Online Newcomers More Middle-Bnni', Less Work-Oriented (Washington, DC: Author, 1999), 20. Gans,DecidingWhat'sNen'S-Craber,Mass Mediaand American Politics; Page, Who Deliberates? 21. Lee Berton, "Whaddya Mean, Gray?" Columbia journalism Reviezo (September/October 1997): 41-44. 22. New York Times on the Web, Site Statistics. New York Times [Online, 1998]: 1. Available: Http://www.nytimes.com/adinfo/stats.html, 23. Graher, Processing the News. 24. Subjects in this condition read the Times on computers that had an ether-net connection to the university computing system, and that system was connected to the Internet through T-1 lines. These relatively swift connections meant there were few reading delays caused by image download times. 25. One of the subjects in the newspaper group attended the posttest administration but did not complete the questionnaire on the final day of exposure. Another subject in that group completed the final day instrument but did not return for the posttest. To conserve numbers, the data for both of these subjects are used in the appropriate analyses. 26. An example of a study using this design is found in Shanto lyengar and Donald Kinder, News That Matters: Television and American Opinion (Chicago, IL: University of Chicago Press, 1987). The present study was conducted over the last week of February 1998. This turned out to be a relatively active news week. On the first day of the project, U.N. Secretary General Kofi Annan carried out eleventh-hour negotiations with Iraq in an attempt to avert a threatened military strike by the United States. Naturally, attention to this issue was high. Thus, it was a challenge to convince students to limit tbeir news exposure to the controlled sessions. 27. During the debriefing, all of the subjects were probed regarding their suspicionsabout the true nature of the research. None of the subjects was able 4/6 ]mm.^usM & M/iss CoMMUNicjaioN Qi.i.\R[\:na to anticipate the hypotheses, and only a handful of the experimental group subjects had heen able to draw any conclusions from the questionnaire they completed on the final day of exposure. 28. Comparisons between the subjects who dropped out of the study and those who completed it were done using a number of variables gathered in the pretest. None of the analyses revealed a bias in the mortality. 29. Another advantage of conducting this study over time was the differences in time-frame for the two media. A good many of the stories that appeared in the online version of the Times on the afternoons of expc)sure appeared in the following day's print version. Looking at news exposure over time meant that audience recall was less likely to be inordinately influenced by a lack ofcontent overlap between the two versions on any one day. 30. It deserves mention that subjects were allowed to regulate their own exposure duration. None of the subjects read for less than 30 minutes, and many stayed for the entire hour. No large systematic group differences were observed. 3L Vincent Price and Edward J. Czilli, "Modeling Patterns of News Recognition and Recall," journal ofCominiiuicatkm 46 (1996): 55-78. 32. As this example illuj^trates, the application of criteria of this sort is imprecise. The categorization was done using a decision rule that at least 2 out of the 3 criteria should clearly indicate relative placement. 33. Not at all surprisingly, there was a relatively clear association of overall recognition with decreasing prominence. In paired-sample t-tests, the comparison between the high and middle-level events was almost significant (t{122) - 1,81,/'- .07). The comparison between middle and low prominence was significant (t(l21) -8.10,;' < .01). A similar pattern can be observed for event detail recall, below. 34. Unfortunately, three of the events were sufficiently significant as to have inspired related articles. For example, a number of articles about the U.N. accord with Iraq appeared that week. Although these other articles were not about the specific topic in question, they included related information. Therefore, analysis of recall accuracy for these items considered statements about all related stories as valid responses. 35. As is the case with tbe other recognition and recall data, responses about the Israeli spy situation were omitted from these variables. 36. Another way of looking at the changes in news consumption is to note how many subjects in each group reported no exposure in the previous week. For the control condition these percentages remained fairly stable over time (23'^ to 31% for television and 58';^. to 62"/;. for radio), but they changed considerably for the treatment groups. The portion of subjects reporting no television news exposure in the prior week jumped from 22% to 95"/.i in the online group and from 24"'n to 92'^. in the paper group. A similar pattern emerged for radio exposure. 37. The success of the random assignment to groups was checked by examining a number of relevant individual-level variables in the pretest for their even distribution across groups. None of the analyses detected significant differences between groups. However, there did seem to be a slight imbalance in gender distribution. Although the chi-square statistic was not significant, there was a slightly higher percentage of women in the paper group (76'!^o) than in the other groups (60"/;. in each). Consequently, all of the analyses reported here were rerun with gender added as a covariate. There was no effect of gender on the relationship between exposure group and first 477 section read. 38. Adding gender as a covariate fo this equation changes the significance level of the effect of exposure condition somewhat, F{1,78) - 3.46, p = .07. 39. Adding gender as a covariate to this equation changes the significance level of the effect of exposure condition somewhat, F(l,78) = 2.64, p .11. None of the other relationships reported in Table 2 was affected by the addition of gender. 40. This difference remained significant when story recall and gender were entered as covariates in an ANCOVA. 41. Re-categorizing these stories as national versus international events revealed the same pattern. Online readers recognized significantly fewer of both event categories. 42. These differences remained significant when gender was used as a covariate in ANCOVAs. 43. Because planned comparisons based on a priori expectations were used to determine differences among means, the control group can be reliably distinguished from the other groups on this variable. This is the case even though the overall F ratio falls just short of conventional levels of statistical significance. 44. Significant differences remained significant when gender was used as a covariate in ANCOVAs. 45. Though not originally hypothesized, it seemed sensible to expect that subjects who have well-developed media use habits prior to the beginning of the study may have been less vulnerable to the format effects studied here. That is, it may be that prior news exposure or some correlate of that concept could act as a conditional variable, moderating the effects of news channel. A good way to test this measure would have been to assess prior exposure to a newspaper like the New York Times. Unfortunately, the only available measure of prior newspaper exposure invited subjects to include the campus newspaper in their estimates. Thus, that would not be an appropriate measure. Using available data, all of the analyses of the hypotheses were rerun with a dichotomous index of general news exposure at the pretest and another of general political knowledge separately entered as independent factors. Neither of these variables interacted with exposure group in a consistent fashion. 46. The recognition and recall means represent the portion of respondents able to recognize the topic or able to accurately recall one or more detail about the events. 47. Eor a discussion of the importance of recency in online news presentation, see Frank Houston, "What I Saw in the Digital Sea," Columbia journalism Review (July/August 1999): 34-37. 48. Graber, Processing the Neius. 49. Graber, Processing the Neivs. 50. One reviewer suggested that subjects in the exposure groups may have been unusually sensitized to current affairs. That is, heightened exposure to public affairs information may have increased subjects' interest in certain topics and this may have resulted in other behaviors {such as talking about current events with friends) that increased their ability to recognize and recall those topics. Unfortunately, it is not possible to determine whether this sort of sensitization occurred in this study and, if it did, what role it played in the observed patterns. 51. The subjects used in this study are unlikely to have been frequent 4:78 JOUIWAUSM & Miss COMMUNKAnON readers of either the online or print versions of the New York Tlmi'!^. One could argue that their behavior in this study might not resemble the habits of regular users. However, it may be that these subjects' unfamiliarity with the Times is beneficial. Because they do not have pre-existing habits, they may rely on the layout of the Tirues as a guide in their reading. Thus, subjects in this study may be particularly likely to develop their reading patterns in response to the layout and cues of the format to which they were assigned. In that way, these subjects provide a good test of whether layout can impact reading habits, the central question of this study. 52. See, for example, Jeffery B. Abramson, F. Christopher Arterton, and Gary R. Orren, The Electronic Commonwealth: The Impact ofNeii'Media Technologies on Democratic Politics (NY: Basic Books, 1988); Gina M. Garramone, Allen C. Harris, and Ronald Anderson, "Uses of Political Computer Bulletin Boards," loitrmi of Broadcast ing ami Electronic Media 30 (1986): 325-39; Merrill Morris and Christine Ogan, "The Internet as Mass Medium," jciurtial of C(i;f;HHni/a)f/()f; 46(1996): 39-50; John E.Newhagenand ShtfizafRafaeli, "Why Communication Researchers Should Study the Internet: A dialogue," journal of Communication 46 (1996): 4-13. 479