Altering the Foundations of Support for the President Through Priming

The disclosure that high officials within the Reagan administration had covertly diverted to the Nicaraguan Contras funds obtained from the secret sale of weapons to Iran provides us with a splendid opportunity to examine how the foundations of popular support shift when dramatic events occur. According to our theory of priming, the more attention media pay to a particular domain—the more the public is primed with it—the more citizens will incorporate what they know about that domain into their overall judgment of the president. Data from the 1986 National Election Study confirm that intervention in Central America loomed larger in the public's assessment of President Reagan's performance after the Iran-Contra disclosure than before. Priming was most pronounced for aspects of public opinion most directly implicated by the news coverage, more apparent in political notices' judgments than political experts', and stronger in the evaluations of Reagan's overall performance than in assessments of his character.

JL he disclosure that high officials within the Reagan administration had covertly diverted to the Nicaraguan Contras funds obtained from the secret sale of weapons to Iran provides us with a splendid opportunity to examine how the foundations of popular support shift when dramatic events occur. According to our theory of priming, the more attention media pay to a particular domain-the more the public is primed with it-the more citizens will incorporate what they know about that domain into their overall judgment of the president. Data from the 1986 National Election Study confirm that intervention in Central America loomed larger in the public's assessment of President Reagan's performance after the Iran-Contra disclosure than before. Priming was most pronounced for aspects of public opinion most directly implicated by the news coverage, more apparent in political notices' judgments than political experts', and stronger in the evaluations of Reagan's overall performance than in assessments of his character.
L residents who are popular in the country tend to have their way in Washington. Popularity is a vital political resource, perhaps the president's single most important base of power (Neustadt 1960;Ostrom and Simon 1985;Rivers and Rose 1985). Popularity, in turn, depends on the prevailing economic, social, and political conditions of the times. Unemployment, inflation, economic growth, flagrant violations of public trust, the human toll of war, sharply focused international crises, dramatic displays of presidential authority-all affect the president's standing with the public at large Vasilatos 1982a, 1982b;Kernell 1978;MacKuen 1983;Ostrom and Simon 1985). A president's popularity (and therefore his power) is shaped by large events played on a national and international stage.
Our purpose here is to illuminate in greater detail the foundations of public support for the U.S. president by taking an approach different from, but complementary to, the one that now dominates research. In the dominant approach, time series statistical methods are applied to aggregated public opinion data. The typical model includes a handful of macroeconomic measures, an indicator or two to reflect the costs of war, and a miscellaneous set of measures to stand for crises, scandals, domestic unrest, presidential initiatives, and more. Although this approach can point with authority to the important national and international events that drive change in popular support in the aggregate, it cannot tell us about the AMERICAN POLITICAL SCIENCE REVIEW VOLUME 84 NO. 2 JUNE 1990 dynamics of individual change. The aggregate time series results may tell us that scandal weakens the president's support with the public but not how scandal affects the thinking of individual citizens. Other than making the analytically convenient but highly unrealistic claim of homogeneity (that all citizens react in exactly the same way), the time series work is silent on what, exactly, citizens are doing. It does not and cannot tell us what is happening at the individual level. 1 Our approach is to examine processes of change in popular support for a president at the level of the individual citizen, with the goal of informing and enriching aggregate studies of presidential popularity. We pursue this ambition by looking closely at citizens' responses to a single event-the highly publicized and dramatic revelation, on 25 November 1986, that funds received by the United States from the sale of arms to Iran had been secretly channeled by members of President Reagan's National Security Council to the Nicaraguan Contras. We treat the Iran-Contra revelation as a critical test for a theory-which we call priming-that claims to provide a comprehensive and psychologically plausible account of how citizens formulate and revise their views of presidential performance.

The Iran-Contra Connection
The Iran-Contra drama began to unfold on 3 November 1986, when a Lebanese magazine reported that Robert C. McFarlane, the President's National Security Advisor, had secretly visited Tehran and that the United States had subsequently sent arms to Iran. In the face of mounting pressure from the news media, President Reagan went public on 13 November, disclosing that a "diplomatic initiative" with Iran had in fact been underway for some 18 months. The purpose of the initiative, he said, was to forge a new relationship with Iran, to bring an honorable end to the Iran-Iraq war, to eliminate state-sponsored terrorism, and to secure the safe return of the U.S. hostages held in Lebanon. Reagan went on to say that as part of this diplomatic initiative, he had authorized "the transfer of small amounts of defensive weapons and spare parts." He assailed the "wildly speculative false stories about arms for hostages and alleged ransom payments" and concluded with the emphatic declaration, "We did not-repeat, did not-trade weapons or anything else for hostages nor will we," a claim he repeated in a nationally televised news conference on 19 November.
On 25 November, the focus of the brewing scandal shifted abruptly away from Iran and the arms-for-hostages question. At noon that day, Attorney General Meese announced to a national television audience that funds obtained from the secret sale of weapons to Iran had been channeled to the Contras fighting to overthrow the Sandinista government in Nicaragua. The diversion of funds had been accomplished through a covert operation managed by members of the president's National Security Council (NSC). President Reagan then disclosed that Vice Admiral John Poindexter, director of the NSC, had resigned and that staff member Oliver North had been dismissed.
These remarkable revelations immediately took over the national news: suddenly, and dramatically, Nicaragua and aid to the Contras were the focus of frontpage stories (see Figure I). 2 Such news was not good for President Reagan's popularity. Figure 2 presents results from polls conducted by Gallup, ABC with the Washington Post, and CBS with the New York Times. All three register sharp declines in public support for President Reagan's performance, roughly coincident with the Iran-Contra revelation. It is impossible to estimate from these data how much of the decline in Reagan's popularity should be traced directly to disclosure of the Iran-Contra connection alone, but it is clear that the events of November significantly shook citizens' confidence in their president.

A Theory of Priming
Equipped with our theory of priming, we believe that Reagan's declining popularity can be explained, in part, by the conjunction of two facts: (1) the media's newfound fascination with covert aid to the Contras and (2) the public's opposition to intervention in Central America. According to the priming theory, when faced with a judgment or choice, people ordinarily do not take all plausible considerations into account, carefully examine and weigh all their implications, and then integrate them all into a summary decision. People typically forgo such exhaustive analysis and instead employ intuitive shortcuts and simple rules of thumb (Kahneman, Slovic, and Tversky 1982). One such heuristic is to rely upon information that is most accessible in memory, information that comes to mind spontaneously and effortlessly when a judgment must be made (Fischhoff, Slovic, and Lichtenstein 1980;Higgins and King 1981;Taylor 1982;Tversky and Kahneman 1981). When asked to evaluate a president's performance, U.S. citizens generally focus only on the aspects of their knowledge that happen to be most accessible at the time of judgment.
In turn, what information is accessible for presidential evaluations is determined 1987 Month by the prevailing economic, social, and political conditions of the times. For their knowledge about such conditions, most citizens of course rely on information and analysis provided by mass media. This means that the standards citizens use to judge a president may be substantially determined by which stories media choose to cover and, consequently, which considerations are made accessible. The more attention the news pays to a particular domain-the more frequently it is primed-the more citizens will, according to the theory, incorporate what they know about that domain into their overall judgment of the president. Hence, by calling attention to some matters while ignoring others, news media may alter the foundations of public opinion toward the president. (For a more detailed discussion of priming, see Iyengar and Kinder 1987, chap. 7.) This central claim of priming has been supported handsomely in a series of realistic experimental tests (Iyengar and Kinder 1987;Iyengar et al. 1984;Iyengar, Peters, and Kinder 1982 Priming and the Iran-Contra Connection The Iran-Contra disclosure provided us with a perfect opportunity to undertake such a test. We do so by exploiting the serendipitous fact that as the attorney general was making his announcement on 25 November, Survey Research Center interviewers were busy questioning citizens all across the country as part of the 1986 National Election Study. That the attorney general's announcement came roughly midway through the 1986 study enables us to see whether, as the theory of priming requires, citizens who happened to be interviewed after the Iran-Contra disclosure evaluated President Reagan more in line with their views on U.S. intervention in Central America than did those (otherwise comparable) citizens who happened to be interviewed before the story broke. If public assessments of the president's performance do indeed depend upon which pieces of political memory come most readily to mind, surely the Iran-Contra disclosure should have enhanced the impact of Central American policy on the public's view of Reagan.
In addition to testing this general hypothesis, we also evaluated three more detailed claims. The first pertains to the degree of correspondence between the news stories that constitute the prime and the opinions that are the target of priming. Several experimental results suggest a specificity to priming-that news coverage influences only the aspects of public opinion that are directly and immediately implicated by the story (see, e.g., Iyengar and Kinder 1987, 106-10). In the Iran-Contra context we would expect priming to be most pronounced on the questions of aid to the Contras and U.S. intervention in Central America, less evident on the abstract principle of whether the United States should generally intervene in other countries, still less apparent on judgments of U.S. power and prestige in the world, and invisible on matters completely unconnected to the Iran-Contra disclosure, such as the health of the national economy or the desirability of federal programs that assist blacks.
We also used the Iran-Contra revelation to examine whether some citizens are more susceptible to priming than others. We focused in particular on expertise. Compared to novices, experts know more about a particular domain; and their knowledge is better organized (Fiske and Kinder 1981). Political experts and novices may react differently to changes in the media's agenda for a number of reasons: (1) because their knowledge is denser and better organized, experts possess a greater and more flexible ability to deal with new information and to interpret it in ways consistent with their prior convictions (Fiske, Kinder, and Larter 1983); (2) because experts possess more informational support for their beliefs, they may be harder to budge; and (3) drawing attention to a particular aspect of national life may only remind experts of what they already know. In a pair of early experiments, experts were indeed relatively immune to priming by television news (Iyengar et al. 1984). But in subsequent experiments (reported in Iyengar and Kinder 1987, chap. 10) this result disappeared, so the significance of expertise in conditioning the impact of news coverage is presently unclear. Here we will see whether novices were more primed by the Iran-Contra revelations than were experts.
Finally, we examined whether news coverage altered the foundations of the public's judgments of President Reagan's character, particularly judgments of his competence and integrity in addition to judgments of his performance. The exper-501 imental results suggest that the impact of priming on judgments of presidential performance is greater than on judgments of presidential character but that judgments of character also seem to depend to some degree on which aspects of national life news media choose to cover (Iyengar and Kinder 1987, chap. 8). Because the 1986 National Election Study included questions measuring the public's view of President Reagan's competence and integrity, we can pursue these results in the context of the Iran-Contra affair.

Data
Our investigation draws on the 1986 National Election Study (NES) carried out by the Center for Political Studies of the Institute for Social Research at the University of Michigan. Following the November national election, face-to-face interviews were conducted with a national probability sample of 2,176 U.S. citizens of voting age. We confined our analysis to the 1,086 individuals who received Form A of the questionnaire, which included an elaborate assessment of views of President Reagan and a rich battery of questions on foreign affairs in addition to standard questions about the campaign, the candidates, the parties, serious national problems and pressing policy choices, and registration and voting.
In order to test priming, we partitioned the Form A sample into two groups: the first was comprised of the respondents who happened to have been interviewed before the 25 November revelations (N = 714); the second was made up of those who happened to have been interviewed afterward (N = 349). Respondents interviewed on 25 November were excluded. Initial comparisons revealed that the prerevelation and postrevelation groups were essentially indistinguishable across a variety of demographic and political comparisons, including education, race, age, gender, employment status, family income, partisanship, interest in politics, and ideological self-identification. This means that whatever differences between groups we might detect in public opinion toward President Reagan can reasonably be attributed to the Iran-Contra revelations.
We focused on three related but distinct aspects of the public's support for President Reagan: evaluations of his overall performance as president, assessments of his competence, and assessments of his integrity. Table 1 shows that after the Iran-Contra revelations, public support for President Reagan declined across all three. These results are in rough accordance with those reported in various commercial polls bracketing this period (see Figure 2).
Our special interest in these public troubles of President Reagan has to do, of course, with whether such troubles can be connected-in the way specified by the theory of priming-to the Iran-Contra revelations and the avalanche of pictures and stories that they precipitated. Did the revelations enhance the political importance of foreign affairs for the public's assessment of the president? In operational terms, we measured foreign affairs opinions with four questions: (1) whether federal spending on aid to the Contras in Nicaragua should be increased, decreased, or kept about the same; (2) whether the United States should become more or less involved in the internal affairs of Central American countries; (3) whether the United States would be better off not getting involved in the affairs of other nations; and (4) whether the United States' position in the world had grown weaker, stronger, or stayed about the same during the previous year. Table 1 reveals that public opinion on these matters changed hardly at all in response to the Iran-Contra revelation. Isolationism as a broad stance, attitudes toward U.S. involvement in Central America in general, and willingness to support more assistance to the Contras were all apparently unaffected by news of the diversion of funds (p > .35 in all cases), although more citizens than before claimed afterward that the United States' position in the world was weakening (45.1% vs. 37.6%, p < .02). This one shift probably has less to do with the Iran-Contra disclosure than with the widespread perception in the public that the Reagan administration had undertaken an arms-for-hostages deal with Iran. In any case, priming may equally occur in the absence or presence of change in the foundational elements of the public's presidential evaluations.

Results
To examine priming empirically, we must first specify a model of public support for the president relevant to the case at hand. In formal terms, the model is given by the following equation: Overall Performance = b 0 + b x (Contras and Central America) + b 2 (isolationism) + b 3 (U.S. strength) + b 4 (national economic assessments) + b 5 (aid to blacks) + be-bu (control variables).
Notice that equation 1 includes three variables to represent the domain of foreign affairs, not four. In preliminary analyses, we found that attitudes toward U.S. involvement in Central America and attitudes toward aid to the Contras were highly correlated. People who opposed aid to the Contras were likely to oppose U.S. involvement in Central America, and those who favored aid to the Contras were likely to favor U.S. involvement in Central America (r = .42). These two attitudes correlated more weakly with views on isolationism and on U.S. strength (.10 < r < .25), and these latter two attitudes were uncorrelated with each other (r = .04). Therefore, in the analysis of priming, we averaged attitudes toward aid to the Contras and toward U.S. intervention in Central America into a single measure.
In addition to the three measures of opinion on foreign affairs, equation 1 also includes a measure of the citizen's assessment of national economic conditions (an average of the citizen's perception of change over the past year in unemployment, inflation, and the general economy) and a measure of the citizen's opinion regarding the desirability of federal programs that provide assistance to blacks (averaged across two questions). We included national economic assessments and race policy views because both are highly relevant to presidential evaluations (e.g., Fiorina 1981;Kinder, Adams, and Gronke 1989;Rosenstone 1983) and because they are utterly unrelated to the Iran-Contra revelation. We expected that the impact of national economic assessments and race policy views on evaluations of President Reagan should either be unaffected by the revelation or should decline, a reflection of the media's sudden preoccupation with Central America. Finally, equation 1 also includes a standard set of background variables important for control purposes though of little substantive interest in their own right: employment status, age, race, gender, region, education, income, and party identification. 3 To test the basic claim of priming, we estimated equation 1 twice, first based on respondents in the prerevelation group, then based on respondents in the postrevelation group. Priming insists first of all that the impact on assessments of President Reagan's performance due to foreign affairs opinions-indexed in equation 1 by by b 2 , and b,-be greater in the postrevelation group than in the prerevelation group. Priming also requires that the impact of economic assessments and racial attitudes on evaluations of Reagan-indexed by b t and b s -should remain the same or decline across the two groups.
The results of estimating equation 1 with Ordinary Least Squares (OLS) regression are presented in Table 2. To interpret the coefficients shown there, keep in mind that all variables were coded to range from 0 to 1, with 1 indicating (1) favorable evaluations of Reagan's performance, (2) support for the Contras and for U.S. involvement in Central America, (3) support for interventionist foreign policy in general, (4) the view that the U.S. had grown stronger in the world, (5) belief that national economic conditions had improved over the previous year, and (6) opposition to federal aid to blacks, respectively. As Table 2 reveals, public opinion on foreign affairs did indeed become more important for assessments of the president's performance in the immediate aftermath of the Iran-Contra disclosures. Foreign affairs attitudes were relatively unimportant to the public's view of Reagan's performance prior to 25 November (see Table 2, col. 1). After 25 November, however, the story is very diferent-foreign affairs loomed rather large in the public's presidential assessment (see Table 2, col. 2).
Moreover, the sharpest increases in Table 2 appear for the aspects of public opinion on foreign affairs most immediately implicated by the revelations. The importance of public opinion on the question of assistance to the Contras and U.S. intervention in Central America increased substantially from the prerevelation period to the postrevelation period (the unstandardized regression coefficient went from .18 to .29), as did the importance of the public's view of the general choice between intervention and isolationism (from .02 to .10). Meanwhile, the public's view of the strength of the United States around the world was evidently unaffected by the revelation (.14 vs. .15). This pattern of results corroborates the experimental findings noted earlier. Both suggest that priming requires a close correspondence between the news stories that do the priming and the opinions that are primed. 4 The theory of priming predicts not only that public opinion on foreign affairs will become more important for presidential assessments after the Iran-Contra disclosure, but also that aspects of public opinion relevant to the president's success 505 but unrelated to the disclosure will not become more important. The results in Table 2 confirm this prediction as well. First, citizens' assessments of national economic conditions contributed sizably to their view of President Reagan's performance-but did so equally before and after 25 November (.33 vs. .35). Second, citizens' views on race policy were evidently shunted aside (if momentarily) by the media's sudden preoccupation with Central America. Prior to 25 November, citizens who opposed federal programs designed to help blacks were more likely than citizens who supported them to support President Reagan. After 25 November, however, this political difference over race no longer contributed to public differences over the president's performance (.22 vs. .00). It would appear that the disclosure of 25 November altered the foundation of support for the president both by bringing certain aspects of public life to center stage and by pushing other aspects of public life off the stage altogether.
In order to illustrate the magnitude of the priming effects documented in Table  2, we generated predicted evaluations of Reagan, first using the prerevelation group's regression coefficients and then using the coefficients estimated with the postrevelation group. This required us to specify values of all the predictor variables in equation 1. For this purpose we chose to represent an average, middle-ofthe-road U.S. citizen: a white, female, forty-year-old high school graduate from the Midwest with an annual family income of 22 thousand dollars, a political independent, who believed that national economic conditions had changed little over the previous year and who neither favored nor opposed federal programs for blacks. We carried out this exercise twice: once assuming that our hypothetical average citizen held views on foreign affairs that would predispose her to support Reagan (favored aid to the Contras and U.S. involvement in Central America, favored international interventionism generally, and believed the United States was maintaining its international strength) and once assuming she held views on these issues that would move her in the opposite direction (opposed aid to the Contras and U.S. intervention in Central America, favored isolationism, and believed that the U.S. had lost international strength).
The results of this simulation suggest that the consequences of priming for presidential support are contingent on the citizen's prior views. Among (typical) citizens predisposed to support Reagan on foreign policy grounds, the effects of priming were negligible. Equation 1 predicts evaluation of President Reagan's performance by such people to be .73 (on the zero-to-one scale) prior to the Iran-Contra revelation and .75 afterward. But among typical citizens predisposed to oppose Reagan on foreign policy grounds, the story is very different: equation 1 now predicts a Reagan evaluation of .53 before the revelation and only .38 after, a steep falloff in support. Thus, the effect of priming was to reduce assessments of President Reagan's performance among critics of U.S. policy in Central America substantially.
In the 1986 NES data, the prerevelation group's average evaluation of Reagan (on the zero-to-one scale) was .602 whereas the postrevelation group's was .548. How much of this .054 decrease can be attributed to priming? We used each of the 1986 NES respondents' actual demographics and attitudes to generate two predicted Reagan evaluations: one using the prerevelation regression weights, the other using the postrevelation regression weights. The average predicted prerevelation evaluation was .532 and the average predicted postrevelation evaluation was .499, a difference of .033. Thus, almost two-thirds of the decrease in this sample's Reagan approval ratings can be attributed to priming.
Our next move was to investigate whether political novices were more susceptible to priming than political experts. Following Zaller's (n.d.) advice, we assessed expertise using measures of objective knowledge about political affairs. In particular, we partitioned the 1986 NES respondents according to their success at identifying six political figures: George Bush, Caspar Weinberger, William Rehnquist, Paul Volker, Robert Dole, and Tip O'Neill. The 65% who correctly identified three or fewer were considered novices, and the 35 % who correctly identified four or more were considered experts. Then we simply repeated the analysis summarized in Table 2, separately within each group.* Among political novices, the Iran-Contra revelation had a substantial priming effect (see the upper panel of Table 3). Opinions on foreign affairs were more important in novices' assessments of Reagan's performance after the revelation than before, whereas national economic assessments were no more important, and opinions on federal programs to aid blacks were less important. Particularly noteworthy was the dramatically enhanced prominence of Central American policy in novices' presidential evaluations after 25 November (the regression coefficient nearly tripled, from .12 to .35). As a general matter, the pattern of priming effects noted for the public as a whole is maintained and sharpened among novices alone.
Meanwhile, priming was less apparent in the assessments of President Reagan's performance offered by political experts (see the lower panel of Table 3). Indeed, priming among experts seems confined to a single aspect of opinion. In the wake of the Iran-Contra disclosure, experts were more likely to evaluate President Reagan according to their general views on U.S. intervention in international affairs (.06 vs. .20). It is interesting both that experts seem generally less susceptible to priming and that the character of priming among experts is different. Novices appear to be primed on those aspects most directly and concretely implicated by the news coverage while experts, insofar as they are primed at all, are influenced at a more abstract level. 6 The distinction between novices and experts apparent in Table 3 is consistent with our experimental results (reported in Iyengar et al. 1984) and with other aspects of the NES survey data as well. In particular, the decline in support for President Reagan's performance in the immediate aftermath of the attorney general's disclosures registered in the public as a whole was greater among novices than among experts. Of the novices, 64.8% approved of Reagan's performance prior to the Iran-Contra revelations while just 57.4% did so afterward, a net change of 7.4 percentage points. Meanwhile, 67.6% of the experts approved of Reagan's performance before the revelations, and 65% did so afterward, a net change of 2.6 percentage points. Novices were more likely than experts to be swept away by the avalanche of stories and pictures set in motion by the 25 November revelations.
Finally, we examined whether media coverage of the Iran-Contra connection altered the foundations of the public's assessments of President Reagan's character. To do so, we reestimated equation 1, first with assessments of Reagan's competence, then with assessments of his integrity, as the dependent variable. 7 The results shown in Table 4 replicate, in a somewhat attenuated fashion, those reported earlier regarding the public's assessments of President Reagan's per-508 formance. For judgments of character, as for judgments of performance, opinions on Central American policy became more important after the Iran-Contra revelation than before while assessments of national economic conditions and views on government race policy became, if anything, less important. The effects of priming were a bit more pronounced in the public's judgments of Reagan's competence than in judgments of his integrity. Indeed, apart from the increased importance of views on Central American policy, the public's assessment of Reagan's integrity appears quite unmoved by the Iran-Contra revelation. The greater susceptibility of competence assessments apparent in Table 4 may reflect, in part, how the news media began to frame the Iran-Contra story, namely, as exposing Reagan's disengagement from U.S. foreign policy. Framed in this way, the Iran-Contra story naturally implicated the president's competence more than his integrity.

Discussion
The disclosure in the fall of 1986 that funds received from the sale of arms to Iran had been secretly channeled to the Nicaraguan Contras provides an excellent opportunity to test the theory of priming in a politically consequential setting. By and large, the theory stands up well to this examination. Citizens questioned after the revelations held President Reagan to an altered set of standards, and these alterations can be directly traced to the changes in the media's agenda provoked by the Iran-Contra revelations. As expected, priming was particularly pronounced for the aspects of public opinion most directly implicated by the news coverage (aid to the Contras and involvement in Central America), was more apparent in the judgments of political novices than in the judgments of political experts, and showed up more clearly in the public's evaluations of President Reagan's overall performance than in assessments of his character. Taken together, these results strongly support the theory of priming and are important not least because they escape the artificiality that inevitably accompanies experimental laboratory research, which had provided priming's sole empirical support up until now. Our confidence in priming is fortified by the close convergence between previous experimental results on the one hand and the results reported here, based on personal interviews with a national sample of citizens responding to a real crisis, on the other.
The comparative advantages of the present investigation-representative sampling, professional interviewers, careful and elaborate pretesting, the serendipitous intrusion of a dramatic and heavily covered event-are real enough; but we should also acknowledge some comparative disadvantages as well. As we noted earlier, the prerevelation group closely resembles the postrevelation group in terms of basic demographic and political characteristics. However, the two groups could still differ from one another in consequential ways that we missed. Given the present design, we cannot be certain that the differences we observed between the prerevelation and postrevelation groups in their assessments of President Reagan were actually due to priming and not to some preexisting and unmeasured difference or to some event other than the Iran-Contra revelation. Notice that this worry is swept aside by the procedure of random assignment that is the heart of the experimental method. It is the convergence of results across different methods of testing that is crucial in science in general and crucial to the standing of the theory of priming in particular.
Priming provides an empirically grounded, psychologically plausible account of how individuals form and revise their views of presidential performance.
Priming therefore aspires to complement the dominant tradition in research on support for the president, which applies time series methods to estimate the impact of national and international events on change in the public support. This research has been enormously informative about the aggregate effects of such events while revealing little about the diversity among citizens' reactions that seem certain to underlie change and stability in the aggregate. From such research we know a great deal about how the public as a whole will respond to a change in unemployment or to a dramatic international crisis but virtually nothing about which citizens are most likely to increase their support for the president, which are likely to decrease their support, and which are likely to be unmoved. The literature is even less prepared to tell us why different citizens respond differently to the same event. Priming provides a general framework to answer such questions.
Priming also carries an implication for the study of political change more generally. Most quantitative studies of this sort assume that the effects of economic and social change are constant over t i m ethat, say, the impact of a one-percentagepoint change in the unemployment rate on the incumbent's reelection chances is invariant across history (or, less dramatically, the length of the time series). Priming collides head-on with this assumption. According to priming, shifts in news media content alter the political importance that the public attaches to the flow of events. If priming is pervasive, the assumption of constant effects seems dubious. This does not mean that the typical time series analysis of political change is worthless, only that such analysis would be enhanced by incorporating the fundamental insight of priming-that through its monopoly over the immediate telling of political history, media possess the power to influence what the public considers and what it ignores.
A final implication of our findings involves stepping back and taking a normative stance on the matter of political change. Exposure to political information through mass media varies enormously across the U.S. public. Some citizens are constantly preoccupied with the flow of political news while others are utterly indifferent to it. Not surprisingly, those who are most heavily exposed to political news also accumulate the most political knowledge. The devoted viewer, listener, and reader becomes society's political expert. With this in mind, we might expect that it would be the expert who would be most influenced by changes in the media's agenda. In the case of the Iran-Contra affair, the news media's daily updating would presumably have the greatest impact on citizens who absorbed the complete story and the least impact on those exposed only to fragments of it. Our results suggest just the opposite. Citizens with the least knowledge (and presumably the least exposure) manifested the largest priming effects. Thus, change in support for the president in response to the Iran-Contra revelations appears to have been dominated by the least-informed, a result that has troubling implications for the exercise of power in Washington. Presidents who enjoy popular support typically have success in shaping the political agenda of the nation (Kernell 1986;Rivers and Rose 1985). Our findings suggest that change over time in popular approval-and thus the waxing and waning of presidential power-may depend the most on the citizens who know the least. their help in preparing the manuscript; Janet Attarian and Konnie Toth for their careful coding of the New York Times; and Shanto Iyengar, Steven Rosenstone, and Janet Weiss for their good advice. The public opinion data reported here were originally collected by the Center for Political Studies of the University of Michigan, Institute for Social Research, on behalf of National Election Studies. Krosnick and Kinder are equally responsible for the analyses and interpretations presented here.
1. Most, but not quite all, the time series work presumes that citizens respond uniformly to whatever is happening in the country. The conspicuous and excellent exception is provided by Hibbs, Rivers, and Vasilatos (1982a), who found that citizens' reactions to events were conditioned by their class affiliations and partisan attachments.
2. Figure 1 displays the number of front page column-lines (text, headlines, and pictures) mentioning the Nicaraguan rebels, Nicaragua, or the Contras.
3. Employment status was coded 0 for respondents who were looking for work and 1 for respondents who were not. Age and educational attainment were coded in years. Income was coded 1 to 22, representing 22 separate income categories. Race was coded 1 for whites and 2 for nonwhites. Gender was coded 1 for males and 2 for females. Party identification was represented by two dummy variables: the first was coded 1 for Democrats and 0 for all other respondents; the second was coded 1 for Republicans and 0 for all other respondents. Region was represented by three dummy variables: the first was coded 1 for residents of central states and 0 for all others; the second was coded 1 for residents of southern states and 0 for all others; and the third was coded 1 for residents of western states and 0 for all others.
4. To test the statistical significance of the changes in the coefficients across the pre-and postrevelation groups, we estimated an enhanced version of equation 1. In particular, we added the following variables to it: (pre-post), (pre-post)(Contras-Central America), (pre-post)(isolationism), (pre-post)(U.S. strength), (pre-post)(national economic assessments), (pre-post)(aid to blacks). Pre-post is a dummy variable coded zero for the prerevelation group and one for the postrevelation group. Thus, the coefficients associated with each of the multiplicative terms tests whether the impact of each aspect of public opinion on evaluation of Reagan's performance differs from the prerevelation group to the postrevelation group. The significance levels of these coefficients appear in the text and in the far righthand column of Table 2. 5. We set the dividing line between experts and novices in this fashion because it generated as close to equal-sized groups as possible while making the experts more rare than the novices, a distribution that suits current wisdom about the distribution of political expertise (see Kinder and Sears 1985). Operationalizing expertise in terms of general knowledge about politics differs from the way we have operationalized it in our previous studies of priming. There, we used measures of domainspecific knowledge (e.g., Iyengar and Kinder 1987;Iyengar et al. 1984). In the present case, we would have preferred to use measures of knowledge about Central American affairs or about Nicaragua in particular, had such measures been included in the 1986 NES.
6. To assess the statistical significance of the differences associated with expertise, we estimated an enhanced version of the equation described in n. 4. To that equation we added six new terms: (expertise), (expertise)(pre-post)(Contras/Central America), (expertise)(pre-post)(isolationism), (expertise)(pre-post)(U.S. strength), (expertise)(prepost)(national economic assessments), (expertise)(pre-post)(aid to blacks). Here we treated political expertise as a continuous variable defined as the proportion of the six political figures correctly identified. The coefficients associated with the multiplicative terms then assess whether the magnitude of the pre-to postrevelation difference in each attitude's impact depends upon expertise. When we estimated this enhanced equation, we found that the three-way interaction involving attitudes toward Central American policy and that involving views on isolationism were both marginally significant (p = .09 and .14 respectively) but that the remaining three three-way interactions were not (p > .25 in each case).
7. The measure of competence is an average of respondents' judgments of how well the terms intelligent and knowledgeable describe Reagan; the measure of integrity is an average of comparable judgments regarding moral and decent.