Contribution
It gave the agenda-setting hypothesis its first systematic test and its name as a research programme. It matched what a sample of voters said were the key issues against a content analysis of the media those voters used, in one community during one campaign. It also set the hypothesis against its rival of the time, selective perception, with a comparison that could have favoured either.
Key insight
Influence on what an audience thinks about is a different thing from influence on what it thinks. The paper quotes Cohen: the press “may not be successful much of the time in telling people what to think, but it is stunningly successful in telling its readers what to think about.” The measurable trace of the first is a match between the ranking of issues in the news and the ranking of issues in the audience’s own judgements of importance. Such a match can arise with no change at all in anyone’s attitudes toward the issues.
Assumptions
- Aggregate units. Voters as a group are compared with the media as a group. “This study has compared aggregate units—Chapel Hill voters as a group compared to the aggregate performance of several mass media.”
- One community, one window. Five Chapel Hill precincts chosen as “economically, socially, and racially representative of the community”. Interviews 18 September–6 October 1968, and media content 12 September–6 October.
- Undecided voters only. A filter question kept those “who had not yet definitely decided how to vote—presumably those most open or susceptible to campaign information.”
- The issue question. “What are you most concerned about these days? That is, regardless of what politicians say, what are the two or three main things which you think the government should concentrate on doing something about?” (fn. 9, from Trenaman and McQuail).
- Content coding. Fifteen categories, split into “major” and “minor” items by space, position and time, with intercoder reliability above .90. The nine media were taken from a spring 1968 pretest: four regional dailies, the New York Times, Time, Newsweek, and NBC and CBS evening news.
- No alternative source of observation. The authors argue that voters cannot observe the campaign except through the media, so a common response to events cannot explain the match.
Key results
- Table 1. A large share of major campaign coverage was about the campaign itself, not the issues: “Campaign analysis” was 28% of all major items and “Campaign events” 19%. Thirty-five percent of the major coverage of Wallace was horse-race analysis.
- Table 2. The parties’ issue emphasis in the media differed widely. Foreign policy was 64% of Humphrey/Muskie issue items and 38% of Nixon/Agnew’s. Law and order was 52% of Wallace/LeMay’s.
- The headline correlations. Voters’ issue emphasis against the media’s emphasis on “the main campaign issues”: +.967 for major items and +.979 for minor items. The paper does not say which correlation coefficient was used, does not print the voters’ issue distribution, and does not say over how many issues the correlation runs. Table 2 lists five main issues. The figures could not be recomputed here.
- Table 3, selective perception against agenda-setting. Forty-five respondents leaned to a candidate without being committed. For four media, the correlation of their issue emphasis with all the news was compared with its correlation with news about their own party. “18 of 24 possible comparisons show voters more in agreement with all the news rather than with news only about their own party/candidate preference.” Rechecked here from the table: 18 favour all the news, 5 favour own party, and 1 is a tie (CBS, Democrats, major items, .83 and .83).
- Table 4. Correlations with individual media ranged from .30 (Time and Newsweek, major items) to .97 (New York Times, minor items).
- Table 5. The media agreed with one another, most among like media, and more on major than on minor items. The authors read the gap from perfect agreement as the “pseudo-environment” falling short of the campaign.
- Table 6, an aside. Voters for whom affect dominated their account of their most important issue reported less recent media use about that issue. The authors report no statistics “because of small Ns in some cells”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Undecided Chapel Hill voters’ aggregate ranking of issues closely matched the composite emphasis of the media they used | strong, as an aggregate correlation in one sample | +.967 and +.979; Tables 3–4 |
| C2 | The match is better explained by agenda-setting than by selective perception | moderate | 18 of 24 comparisons in Table 3 favour all the news over own-party news |
| C3 | The direction runs from media to voters, not the reverse or from a common cause | weak | argued, not tested: voters lack other access to the campaign; journalists’ and audiences’ news values usually diverge; five of the nine media are national (fn. 13) |
| C4 | Individual media reproduce the political world imperfectly, and agree more on major stories | moderate | Table 5 |
| C5 | Salience of affect predicts less information-seeking on one’s most important issue | weak | Table 6, no statistical test |
Concepts
- Agenda-setting: the media “set the agenda for each political campaign, influencing the salience of attitudes toward the political issues”, as distinct from changing “the direction or intensity of attitudes”.
- Major and minor items: a split of news by space, position and time, defined medium by medium in the method section.
- Selective perception (the rival): voters attend mostly to news about their own party, so their agenda should match that news best.
- Pseudo-environment (after Lippmann): the media’s composite picture of the campaign, which the voters’ agenda tracks.
Connections
It builds on Lang and Lang and on Cohen (1963) for the hypothesis, on Berelson, Lazarsfeld and McPhee for the limited-effects background, and on Trenaman and McQuail for the issue question. Entman (1993), which the record holds as LIT-1332, is the second media route the exchange named beside it, and could not be reached. Barnehl and Schumacher (2024) build their measure of capture on agenda-setting and cite the Luo et al. meta-analysis of the programme that followed this paper.
Bearing on the record
- CLAIM-404’s evidence route. The claim says an organization can be governed through the evidence its procedures receive. The paper is a documented case of the population-level form: an audience’s ranking of what matters tracked the ranking the media supplied, while nothing in the paper shows any change in what voters concluded about the issues. That is the claim’s distinction between influence on the inputs and influence on the decision rule, observed in a public. The paper does not identify the direction, so it supports the claim as an illustration and not as a demonstration.
- TERM-064. The term’s table says legibility is not salience. This paper measures salience in the term’s sense: which issues a public counts as important. It says nothing about whether those issues became more legible.
- QUESTION-253. Agenda-setting is one empirical candidate for “which facts become legible enough to count as evidence at all”. Here the media’s selection fixes which issues reach the public’s list of important problems. The paper has no time dimension, so it bears on neither of the question’s dynamic readings.
- THEORY-842 is filed Active from the correlations as measured, and THEORY-838 is filed Proposed for the causal hypothesis the correlations are offered for.
No instruction for machine-learning practice; nothing for the anthology.
Limitations
- Correlation over aggregates, in one sample. The authors say so: “The existence of an agenda-setting function of the mass media is not proved by the correlations reported here, of course, but the evidence is in line with the conditions that must exist if agenda-setting by the mass media does occur.”
- Few units. The correlations run over a handful of issue categories, so values near one carry little statistical weight, and no test is reported.
- Direction is argued, not identified. The two alternatives the discussion rejects, a common response to events and media catering to audience interests, are rejected by argument. The media’s emphasis reflects “to a considerable degree the emphasis by candidates”, so the agenda the voters matched may be the candidates’ agenda as relayed.
- The issue question is not neutral to the hypothesis. It asks what the government should “concentrate on doing something about”, which is close to asking for an agenda.
- Exposure is not measured per person. Media were matched to the community, not to each respondent’s use, except in Table 3’s party split.
Open questions
- Does the match hold at the level of individuals, with exposure measured per respondent? The authors call for this (“from a broad societal level to the social psychological level”).
- Does a change in media emphasis precede a change in public emphasis? A panel or an experiment would answer it.