Priming and the plight of public opinion polling
Public opinion polling has, well, issues that it needs to work through.
Cengiz Erisen got his PhD in political science at SUNY Stony Brook in 2009. He recently published an article based on his dissertation in Political Psychology, co-authored with two of his Stony Brook professors, Milton Lodge and Charles Tabor. Titled “Affective Contagion in Effortful Political Thinking”, it is one of the best treatments I have read of priming in political attitude formation and presentation.
Priming is the process of being influenced by stimuli we are not consciously aware of. Malcolm Gladwell describes some great priming examples in his book Blink, such as the experiment in which people unscrambled sentences that contained words that primed “old age”, and then walked more slowly down a long corridor than people who were not primed with old-age words.
Erisen’s experiments were masterpieces of simplicity. A total of about 350 Stony Brook students, pretty evenly distributed among self-declared Democrats, Republicans, and Independents, were asked to record on a computer their attitudes on a number of public policy issues like the energy crisis and illegal immigration. They were then asked to record their free-form “thoughts” about one issue, like illegal immigration.
Unknown to the students, they were subliminally primed with simple happy faces, frowning faces, or neutral faces in between each of their recorded thoughts. These primes were displayed for 39 milliseconds and were backward and forward masked with random characters so they would not be detected consciously. No student reported seeing the primes or knowing they were present. Each participant was consistently primed with one type of image – happy, sad, or neutral – throughout their session.
This was a strong test of the subliminal priming hypothesis. The primes were extremely simple, abstract, irrelevant, and incidental. The attitudes being primed were highly cognitive – participants had to consciously think up and write down full sentences about the policy issues. People were forced to think, probably much more than they would be expected to in a public opinion polling phone call.
The results were highly significant statistically, and in the predicted direction. First, the number of thoughts generated was influenced by the primes. The average number ofnegative thoughts recorded on the immigration issue was 1.40, 2.53, and 1.69 respectively under the positive, negative, and neutral priming conditions (F(1,187)=7.15, p<.001). The average number of positive thoughts was 1.64, 1.22, and 0.90 under the positive, negative, and neutral priming conditions (F(1,187)=7.00,p<.001). People came up with more positive thoughts when subliminally primed with a cartoon happy face, and more negative thoughts when primed with a cartoon frowning face.
In addition, the expression of negative or positive valence in the thoughts was impacted by the primes. People who had expressed explicit negative attitudes toward illegal immigration in the questionnaires listed an average of 4.33 negative thoughts under the negativepriming condition. But under the positive priming condition, people with negative attitudes listed an average of only 1.82 negative thoughts. The authors note:
the subliminal presentation of smiling and frowning cartoon faces strongly and significantly promoted affectively congruent thoughts on illegal immigration, and this effect was not washed out by the power of prior attitudes. In support of the substantive power of affective congruence, participants in our first study listed on the order of twice as many thoughts that were congruent with the prime as those that were incongruent, regardless of their prior attitude on the issue.
In other words, an “invisible” little happy face makes it harder for you to think up negative thoughts about an issue you feel negatively about!
When are nonconscious factors most likely to have this kind of influence on our opinions and behaviors? The authors provide a concise list of conditions, and point out that these factors tend to be operative in most citizen’s relationship with politics:
The experimental literature presents clear evidence that implicit processes underlie all conscious processing and have been shown to be more valid predictors of top-of-the-head, like-dislike evaluations
- when affectively charged cognitions are available and strong;
- when explicit measures are tainted by social desirability, deceit, or prejudice;
- when one is under time pressure; when the costs of being wrong are low;
- when attention is otherwise engaged or distracted;
- when an environmental event is noticed but not recognized as being influential;
- and when one’s behavior is not so consequential as to trigger questions about “why did I think, feel, say, or do that?”
These situational and contextual factors characterize the world of politics for many citizens most of the time, where typically, the consequences of political action are distant and indirect, uncertainty reigns, rumination is rarely called for, rapid-fire media distract, and self-exposure to the stream of information routinely infuses one’s thoughts with congenial cues. (bulleting added)
What are the implications for public opinion polling? If a subliminally presented simple cartoon smiley face can significantly influence a person’s expression of political opinions and attitudes, how much priming is going on when you get a call from an opinion polling organization that wants to hear your opinions on a whole range of political topics? What Erisen’s research shows is that those opinions may be extremely volatile and are probably influenced by many sources you are not aware of. Some of those sources may be related to politics, like what you saw on the news earlier in the day, or what the late-tight talk show host said in his monologue last night, or what your colleague said in a conversation at the office this afternoon, but you could also be affected by things completely unrelated to politics, like stubbing your toe just before the phone rang, or hugging your child, or being thirsty. Is public opinion polling just one big echo chamber, creating an illusion of “opinion” in a giant feedback loop of polling, reporting the results of polling in the media, and then priming people’s opinions with those media reports?
OK, that’s probably too extreme a scenario, but I think it’s fair to say that assuming that polling results represent “real” opinions, derived from rational evaluations of options and alternatives, is naïve at best.
What opinion polling needs is a supplemental tracking mechanism that tests the depth and cognitive “intensity” of expressed opinions. It would be great if every public opinion result was accompanied not just by a “plus or minus 3.1% margin of error” statistic, but also by an “intensity index” that showed where that “47% approval rating” fell along a spectrum of “rock solid” to “totally squishy”, a very different kind of margin of error.
This is a slightly modified re-post of an article originally posted in my long-defunct previous blog on April 2, 2009.
About the Author: Steve Genco
