The Pew Survey Respondents Were Not Reacting to “Words Without Context”

In his response to my last post, Orin argues that some of the seemingly anomalous reactions of respondents to the Pew survey I analyzed might have been the result of the fact that the researchers asked them to react to “words without context” rather than “ideologies” or political ideas.

I highly doubt this accounts for more than a tiny fraction of the results. The words in question are not actually “without context,” as Orin suggests. Rather, any minimally informed person will quickly realize that they are a list of fairly standard political terms. A word like “progressive” or “socialism” may be ambiguous if stated in isolation from other words. But in a list with other political and economic terms, the context becomes a lot clearer. The instruction Orin quotes is actually for Question 39 in the survey, and comes after a list of 38 previous questions that were mostly about political subjects. That makes it even more likely that respondents understood the context.

In context, people tend to react to what they perceive as the meaning of the word, not merely to its sound or to vague associations with it. Of course, a person who doesn’t know the meaning of the word may indeed base his or her perception primarily on whether it “sounds good” in some vague sense. Orin’s example of the “french cruller” illustrates both points well:

For example, imagine I ask you whether you have a positive or negative reaction to the phrase “apple pie.” You’re very likely to report a positive reaction. After all, apple pie conjures images of July 4th picnics and American flags. On the other hand, imagine I ask for your reaction to the phrase “french cruller.” The results are likely to be mixed. Cruller is a pretty ugly word, and you might have mixed views of the french. But those results won’t necessarily mean that if you’re trying to figure out what dessert to bring to an event that you can expect more people will like apple pie than french crullers.

Most people know what an “apple pie” is. As a result, their responses to the question of whether they have a “positive” or “negative” view of apple pies is likely to be highly correlated with whether they like the taste of that kind of pie or not. By contrast, many people (myself included, until I read Orin’s post) don’t know what a “french cruller” is, and so may react primarily to the sound of the word or base their answer on how they feel about the French.

If presented with a list of different kinds of foods and asked to give their opinions of them, most people who know what those foods are will tend to give responses that correlate with their underlying view of those foods’ taste and nutritional value.

I will, lastly, mention that the formulation used in this poll (asking for reactions to “words” and “phrases”) is not unusual in survey research, and pollsters often get fairly rational and consistent answers when using it with respect to terms whose meanings are at least to some extent known to respondents. For example, Republicans are much more likely than Democrats to have a negative reaction to “liberal.” Democrats are more likely to have a negative reaction to “religious right,” and so on.