Selective attention is an important concept in communication, both because of its role in the limited effects paradigm and its intuitive power in describing how the media are used. It is what Chaffee and Miyo call an "enduring generalization" from the earliest studies about political campaigns (e. g., Lazarfeld, Berelson, & Gaudet, 1948). Because it is intuitively compelling and broadly understood, it has endured in textbooks and everyday thinking despite inconsistent empirical evidence and critical research reviews (most notably, Sears & Freedman, 1967).
Like most ideas in communication research, selective attention is repeatedly studied in the same way. Of particular interest for this study is the consistent use of obtrusive measures. Many selective attention studies use self-reported measures of attention or exposure (e. g., Chaffee & Miyo, 1983; summarized in Sears & Freedman, 1967; Sweeney & Gruber, 1984). In other words, subjects typically were asked how closely they watched a political event on television, or how closely they paid attention to a political candidate. Some studies even used self-reports of prospective behavior, such as asking someone if they planned to read a brochure in opposition to their beliefs or to a decision they just made (Lowin, 1969).
This paper is a report about a series of experiments that used unobtrusive measures of attention to news stories on a Web page, a computer program that determines which Web page a subject chooses and then records how long each subject is exposed to the page. Such measurement techniques have not yet been widely used (one example is Knobloch, Carpentier, & Zillman, 2003). The current authors believe that in contrast to many prior studies, this operationalization of media attention has greater internal validity because it does not rely on self-reports, and the experiments have greater external validity because the protocol uses ordinary news stories in their typical context of an online publication. In addition, this richness of measurement informs one's conceptual thinking about attention and lends itself to a unique conceptualization of selective attention. Finally, it is argued that this study offers a good test of the selective attention hypothesis.
Selective attention is, broadly speaking, the preference for information that is consistent with previously held beliefs together with the avoidance of information counter to those beliefs. The selective attention hypothesis is widely accepted, at least tacitly, in the face of inconclusive evidence. Sears and Freedman called selective attention "one of the most widely accepted principles" in communication studies despite "unsatisfying" empirical results that show no preference for consistent information (1967, p. 194). (1) Donohew and Palmgreen called the empirical evidence "equivocal" and reported, "Both supporters and opponents of the hypothesis are able to marshal impressive empirical evidence in support of their positions" (1971, p. 412). Despite these reviews, many textbooks and general reviews assert that selective attention is widespread. (2) Ironically, this intuitive belief in selective attention contrasted with little supporting evidence and a series of critical research reviews may have stifled further research (Sweeney & Gruber, 1984). In the last decade selective attention has not been the subject of much research in communication publications (some few examples are Chaffee, Saphir, Graf, Sandvig, & Hahn, 2001; lyengar, Hahn, & Prior, 2001; Knobloch et al., 2003).
Selective attention has a crucial function in the limited effects model, which has played an important role in the history of communication research and in an understanding of the mass media (Chaffee & Hochheimer, 1985; Gitlin, 1978). In the early voting studies, Lazarsfeld and his colleagues found that political partisans reported being exposed to more of "their own side's propaganda" than from the opposing candidate, and Lazarsfeld argued that people insulated themselves from opposing viewpoints with "high tariff walls against alien notions" that increased with the degree of partisanship (Lazarsfeld et al., 1948, p. 89). This observation supported Festinger's influential theory of cognitive dissonance, which gave selective attention a central role in the process of dissonance reduction when a subject is faced with competing beliefs (1957). Several influential communication scholars, especially Klapper, argued that this selectivity in exposure, perception, and retention meant that the mass media did not persuade the audience but primarily reinforced preexisting beliefs (Klapper, 1949, 1960). This was a central thesis of the limited effects model.
In 1967 Sears and Freedman published a devastating review of selective attention research that argued that the evidence was weak and could be explained by other mechanisms. They argued that rather than selective attention, the results from several early studies were really a form of "de facto" selectivity. De facto selectivity was when an audience agreed to an unusual extent with the communicator's position, such as when more Republicans than Democrats report they watch the Republican political convention. Sears and Freedman further argued that the methodology of much of the research was flawed. Often the measure of predispositions did not predate the measures of exposure, so it was not clear if selective attention or some sort of attitude change had taken place. In addition, Sears and Freedman criticized work in this area for its reliance on retrospective self-reports of media exposure, a criticism to which this study pays particular attention.
Selective attention research has encompassed media as diverse as pamphlets in a public space (LaVoie & Thompson, 1972) and political letters mailed to registered voters (Bartlett, Drew, Fahle, & Watts, 1974). As yet, however, little research has looked at new media technology, such as online newspapers, which may be more conducive to selective attention. Online media permit greater control by the reader to progress beyond a headline of an article; the reader must consciously choose to view an article, then click to it on the Web site. This degree of intention is unusual among the media, and it even challenges notions of what defines a mass medium.
Conceptualizing and Measuring Attention
The methodology here is uniquely suited to unobtrusively measure attention within a realistic setting, and the authors' conceptual thinking is guided by this. Attention was measured by the amount of time subjects spent perusing each Web page of an online publication about political issues. The dependent measures involve actual behavior, not self-reported attention, and the attention measured is specific to the material: Each story has its own page. It was determined exactly where a subject gravitated first when presented with information in agreement with and in opposition to their views. The authors then timed how long each subject spent looking at each type of information and how far into the news story they read, thus measuring the concept of attention along three dimensions: time spent, initial attention (or where a subject looked first), and depth of attention (or how far into a story the subject read). By combining attention to stories both consonant and counter to the subject's political beliefs a measure of selective attention was derived. This is attention paid to one position relative to the attention paid to another.
The authors' conceptual thinking is informed by this rich measure of attention. This concept of selective attention is clearly not selective avoidance, but rather selective approach (Frey & Wicklund, 1978). It is also clearly selective attention, and not selective exposure. Attention here is not "de facto," or what Sears and Freedman described as without cause or intention: This conceptualization is more akin to their "preference for supportive, rather than nonsupportive information" (Sears & Freedman, 1967, p. 196). It is intentional, and the material must be actively sought...