Language Log

« previous post | next post »

Ian Preston has pointed me to Aleksandra Cichocka et al., "On the Grammar of Politics—or Why Conservatives Prefer Nouns", Political Psychology (published online 1/26/2026):

Previous research indicates that political conservatism is associated with epistemic needs for structure and certainty (Jost et al., 2003) and that nouns elicit clearer and more definite perceptions of reality than other parts of speech (Carnaghi et al., 2008). We therefore hypothesized that conservatives would exhibit preferences for nouns (vs. verbs and adjectives), insofar as nouns are better suited to satisfy epistemic needs. In Study 1, we observed that social conservatism was associated with noun preferences in Polish and that personal need for structure accounted for the association between ideology and grammatical preferences. In Study 2, conducted in Arabic, social conservatism was associated with a preference for the use of nominal sentences (composed of nouns only) over verbal sentences (which included verbs and adjectives). In Study 3, we found that more conservative U.S. presidents used greater proportions of nouns in major speeches, and this effect was related to integrative complexity. We discuss the possibility that conservative ideology is linked to grammatical preferences that foster feelings of stability and predictability.

You can read for yourself about their Study 1, in which 189 Polish undergraduates were given "incomplete sentences describing different people, which they were asked to complete with either a noun or an adjective (in six cases) and either a noun or a verb/adverb pair (in four cases)",  with the subjects' choices correlated against their scores on scales of "policy preferences" and "Need for Closure".

And their Study 2 was similar —  100 "young adults in Beirut" were scored on "political orientation" and on "grammatical preferences". For the "grammatical preferences", subjects were again given a choice of sentence completions:

To gauge preferences for nominal sentences, participants were asked to choose one of two options that would best complete three sentences describing different people. Response options featured either nominal (all-noun) or verbal sentences. For example, a sentence “Hanan likes to paint. Hanan …” could be completed with “(is) an artist” (nominal sentence) or “has artistic skills” (verbal sentence).

[I presume that in the Arabic versions, the descriptions "nominal sentence" and "verbal sentence" actually apply, as they clearly do not in the English versions…]

I'll just mention that the effects in these two studies, though statistically significant, are not predictively overwhelming. In the Polish study, general/social conservatism accounted about 3% of the variance in preference for nouns (r=0.16). In the Arabic study, political orientation accounted for about 10% of the variance in sentence preference. These numbers are consistent with the generally weak strength of statistical relationships in social psychology research, and thus should not be interpreted as a flaw in this paper.

I'm going to focus on Study 3, for which they describe the materials as follows:

In Study 1 [sic], we used archived transcripts of speeches delivered by U.S. presidents hosted by the website of the American Presidency Project (http://www.presidency.ucsb.edu/index.php) as of January 16, 2015. We coded transcripts of inaugural addresses (n = 21) as well as state of the union addresses (or the corresponding replacement speeches that were listed, total n = 80).10 These two types of speeches are comparable to the extent that they are regularly scheduled and addressed to very broad audiences. We began with the First Inaugural Address of Franklin D. Roosevelt and finished with the January 2014 State of the Union address delivered by Barack Obama, producing a database of 101 speeches delivered by 13 presidents in total.

The featured results:

Results revealed that Republican presidents employed a higher proportion of nouns in speeches (M = 0.26, SD = 0.02), in comparison with Democrats (M = 0.25, SD = 0.01), t(77.61) = 2.46, p = .02 (equal variances not assumed), Cohen's d = 0.50. There were no differences in the use of adjectives, t(99) = −0.45, p = .65, Cohen's d = 0.09, or verbs, t(99) = −0.15, p = .88, Cohen's d = 0.03.

In other words, 26% of the R presidents' words were nouns, while for D presidents, the figure was 25%.

I thought I'd check this result using somewhat different data, a different part-of-speech tagger, etc. So I compared 50 of Barack Obama's Weekly Radio addresses from 2010, to 52 of George W. Bush's comparable addresses from 2008. I tagged the texts of these addresses ("as prepared for delivery") using the Stanford maxent tagger, which uses the Penn Treebank P.O.S. tags.

Somewhat to my surprise, the results basically replicated the findings of Cichoka et al.  In fact, I found slightly stronger effects:

Here about 13.69% of Bush's words were category NN (singular common nouns), compared to about 12.21% of Obama's, for a difference of 1.47% absolute (or about 12% relative).

Adding up NN (singular common nouns), NNS (plural common nouns), NNP (singular proper nouns), and NNPS (plural proper nouns), we find that 26.60% of Bush's words were nouns, while 22.82% of Obama's words were nouns, for a difference of 3.78% absolute (or nearly 17% relative).

What categories of words did Obama use instead?

Sorting by the absolute difference in percentages, we find that the top four were RB (adverbs), PRP (personal pronouns), VBP (Verb, non-3rd person singular present), and WP (Wh pronouns).

At the other end, Bush's greatest wins were in proper nouns, common nouns, adjectives, and possessive pronouns.

To forestall the cries of "See, Obama is a narcissist!" in response to the difference in personal pronouns, I'll note that Obama actually uses fewer first-person singular pronouns than Bush does  (0.97% vs. 1.15%) — the PRP difference is mostly in first-person plural pronouns (1.86% vs. 1.00% we, 3.22% vs. 2.25% overall), with a slight edge to Obama in second-person and third-person pronouns as well.

I'll leave it to readers to evaluate the journalistic responses: Tom Jacobs, "Why Conservatives Prefer Nouns", Pacific Standard 2/8/2016; Adam Boult, "Can this 30-second test determine whether you're liberal or conservative?", The Telegraph 2/25/2016; Abigail Beall, "The simple trick that can reveal if you're a conservative or liberal: Study discovers which words give away political beliefs", Daily Mail 2/25/2016; "Grammar of Politics: Conservatives Prefer Nouns", Business Standard 2/25/2016.

The idea that the relatively weak sentence-completion effects are a "simple trick that can reveal if you're a conservative or liberal" was not promoted by the University of Kent's press release, and it would be unusual for the journalists to actually read the paper, so I suspect that it must have come from interviews with the authors. (Though of course the whole "one weird trick" meme has become a button that's easily pushed by accident…)

 

 

February 28, 2016 @ 5:59 am · Filed by Mark Liberman under Language and politics

Permalink