The Dog Whistle: Lee Atwater’s 1981 Confession

In 1981 Lee Atwater, then a young operative in the Reagan White House and later chairman of the Republican National Committee, gave an interview to the political scientist Alexander Lamis on condition of anonymity. He was quoted in Lamis's 1984 book as an anonymous Republican official. The full audio was released by The Nation in 2012, after Lamis's death, and it is more explicit than the published excerpt.

Atwater describes the evolution of Southern political strategy with unusual candor. You start out in 1954 by saying a racial slur, he says. By 1968 you cannot say it — that hurts you, it backfires. So you say things like forced busing, states' rights, and all that stuff, and you are getting so abstract. Now you are talking about cutting taxes, and all these things you are talking about are totally economic things, and a byproduct of them is that Blacks get hurt worse than whites. He adds that if it is getting that abstract and that coded, we are doing away with the racial problem one way or the other, and that obviously sitting around saying we want to cut this is much more abstract than the slur.

It is the clearest first-person account on record of how a coded appeal is designed, and what the designer thinks it does.

The mechanism is implicit associative priming. A coded term activates an associated schema in listeners who hold the association, without ever stating it, and the activation is not subject to the explicit belief-checking that an overt statement would receive. Anthony Greenwald and Mahzarin Banaji's work on implicit social cognition established that automatic associations can operate independently of, and sometimes contrary to, consciously endorsed attitudes — which means the appeal can reach listeners who would sincerely reject the explicit version.

Deniability is not a side effect of this design. It is the delivery mechanism. Because the term has a literal meaning that is entirely legitimate — and forced busing, states' rights and tax policy are all real subjects with real arguments — any listener who names the subtext can be accused of imagining it, of injecting race into a conversation about school administration. The accusation of paranoia is available to the speaker as a matter of design, and it costs the listener who objects more than it costs the speaker.

Which produces the characteristic argument that surrounds these appeals: one group hearing an obvious message, another group hearing nothing, and no way to settle it from the text alone.

Two things make the analysis tractable rather than a matter of assertion.

The first is pattern rather than instance. A single ambiguous phrase proves nothing. A speaker who reaches repeatedly for terms that carry the same association, in contexts where the literal meaning does not fit the argument being made, is producing a pattern that has an explanation.

The second is measurement. This is one of the few areas of political rhetoric where the effect is testable: present the coded and uncoded versions to matched audiences and measure the difference in racial-attitude activation and in policy preference. That work has been done, and finds effects. Which means the question of whether a whistle is being blown is, in principle, an empirical one rather than a matter of interpretive taste.