Presidential pronouns, one more time

« previous post | next post »

Yesterday, Marc Cenedella did a sort of  Breakfast Experiment™, and reported the results in "I, Obama: The President and the personal pronoun":

President Obama has taken criticism in some sectors for his use of the personal pronoun in describing, and applauding, the nation’s success in covert operations. So I’ve spent my Saturday morning at the outstanding website of The American Presidency Project at the University of California, Santa Barbara to find out what happens when Presidents speak to the CIA.

I picked twelve notable addresses from Presidential speeches at the CIA’s Langley headquarters over the past 52 years and run [sic] the numbers.

This is a big step forward, relative to most of the discussion of this overblown media obsession that we've seen over the past couple of years. As I wrote the last time the business of Obama's first-person singular pronouns came up,

There are two problems with this meme. The first problem is that frequency of FPS pronoun use is not in fact correlated with the personality traits that they associate with it (and with Obama). (For discussion and references, see "What is 'I' saying?", 8/9/2009.) The second problem is that Obama's empirical frequency of FPSP use is in fact on the low side, when you compare his speeches, press conferences, interviews, etc. to similar performances by other recent presidents. (For discussion and references, see "Fact-checking George F. Will, one more time", 10/6/2009.)

In this context, there are two good things about Mr. Cenedella's post. The first good thing is that he actually counted pronouns. The second good thing is that he compared this result to the counts in a set of roughly comparable speeches.

This is in striking contrast to (for example) George F. Will, Peggy Noonan, Charles Krauthamer, and Stanley Fish, who made similar sorts of claims about specific speeches without counting or comparing. In those and other similar meme-bleats from the media herd, the facts were objectively embarrassing: the rate of first-person-singular pronoun use in the speeches they cited was in fact comparable to or lower than the rates found in speeches over the years by other politicians in similar situations.

Now in fact, this is pretty much what Cenedella found. Here's his graph:

He finds Obama's I-usage rates to be high enough to motivate the question

So is President Obama the most egotistical president since Nixon, or simply a leader who believes in direct and personal communication?

But the differences are not all that large, and Cenedella's evaluation even mentions the idea that we ought to go beyond mere blind counting, and look at how and why people use first-person singular pronouns:

Wearing my partisan hat, I’d have to say it is rather jarring to read Obama’s addresses after all those who came before him.  There is clearly a different tenor, a different tone, and a different conception of the relationship between the man of the Oval Office and the Agency.

But putting a longer lens on it, we must also note that the present situation is one in which the nation’s security is primarily secured by the type of specialized, professional operations conducted by the CIA and special forces teams.

And with my CEO hat, I’d note that leadership sometimes requires the ‘I’ and not the ‘we’.  In times of crisis, or in situations where significant or substantial organizational hurdles block the path to success, it can be the case that the leader needs to switch from the congratulatory ‘we’ to the responsibility and accountability of the ‘I’.

So far, so good.

But there are two problems with the whole exercise, one strictly numerical and the other interpretive.

The numerical problem is that some of Cenedella's counts, and also the corresponding arithmetic, seem to be wrong. And correcting the numbers pretty much destroys his case, such as it is. Thus for Bush 2's first CIA speech, he counts 10 I's in what looks from his graph like a bit more than 500 words. He plots this as about 1.6%. My FPSP-counting script gets:

12 i
3 my
2 me
2 i've
1 i'm
599 words, 15 I's (2.50 percent), 20 FPSP (3.34 percent)

(I don't know why Cenedella decided to count only I and not my, me, mine, etc. — but the effect is pretty much the same either way.)

Running the same script on Obama's most recent speech to the CIA, I get

35 i
7 my
3 me
2 i'm
1 i've
1742 words, 38 I's (2.18 percent), 48 FPSP (2.76 percent)

So comparing these two speeches, Obama's rate of I-usage was 13% less than W's (2.18% vs. 2.50%), and his overall rate of first-person-singular-pronoun usage was 17% less (2.76% vs. 3.34%).  Does this mean that Obama is less egocentric or less responsible and accountable or whatever? I don't think so — but it does suggest to me that Cenedella's subjective impression of Obama's speech might be telling us more about Cenedella than about Obama:

Wearing my partisan hat, I’d have to say it is rather jarring to read Obama’s addresses after all those who came before him.  There is clearly a different tenor, a different tone, and a different conception of the relationship between the man of the Oval Office and the Agency.

Checking a couple of Cenedella's other counts, I found some even larger discrepancies. Thus for Clinton's second address to the CIA, Cenedella's graphs show about 36 I-words in a bit more than 2500 total words, for a rate of a bit less than 1.5%. My script gets very different counts, and therefore a very different percentage:

15 i
3 me
1 i've
1641 words, 16 I's (0.98 percent), 19 FPSP (1.16 percent)

And in JFK's address, which Cenedella counts as around 7 I's in 500 words, for a rate of a bit more than 1.5%, almost a third of the transcript is JFK reading the citation for Alan Dulles's National Security Medal. (Awarding that medal was the central point of that presidential visit and speech.) Removing the quoted citation, which of course entirely lacks first-person pronouns, I get

9 i
2 my
2 i'm
1 i'd
344 words, 12 I's (3.49 percent), 14 FPSP (4.07 percent)

Which seems to make JFK the champion egotist by far — if we believed that these numbers meant anything in themselves, which I don't.

(I don't know how Mr. Cenedella's numbers came out so far off — in some cases, I suspect that he counted by hand, and simply missed some instances; in other cases, such as the puzzling Clinton numbers, he may have clicked over to the wrong speech. In any case, he's to be commended for giving us not only his numbers but also links to the specific documents that he derived them from.)

The second problem is interpretive.

Cenedella has some plausible thoughts on why a president might use first-person singular pronouns ("… leadership sometimes requires the ‘I’ and not the ‘we’.  In times of crisis, … the leader needs to switch from the congratulatory ‘we’ to the responsibility and accountability of the ‘I’").  This is a useful  addition to the distinction that James Pennebaker draws in "What is 'I' saying?", 8/9/2009, between the "graceful I", which is polite, hedging, respectful of multiple perspectives, through more reportorial phrases like "I saw", to "egotistical, controlling sledgehammer I's". And there are other kinds of first-person singular pronoun uses as well —  references in informal opening remarks ("Someone told me that Secretary Cohen was here, but I haven't seen him yet"); promises of action ("I intend to continue to work with Congress to make sure that our law enforcement officials at home have got the tools necessary"), and so on.

Give these large interpretive differences among different sorts of FPSP usage, simple counts are going to be a very weak signal at best. So there's an obvious next step. If you think that such counts are useful indicators, then you really should decide on a taxonomy of FPSP types, assign individual instances to those categories, check inter-annotator agreement rates, and draw conclusions — if any — based on statistics over these disambiguated categories.

My own opinion, FWIW, is that this would turn out to be a difficult annotation task, requiring a lot of careful work to define the categories and to train annotators. I'm not convinced that we would learn much from applying the process to political speeches. And at this point, the commentariat's obsession with this aspect of President Obama's speeches still lacks any empirical plausibility, despite Mr. Cenedella's efforts.

Past LLOG posts on related subjects:

"Recommended reading", 5/3/2011
""A sociopath and narcissist and manipulator"", 8/9/2010
"Open fraud as Op-Ed discourse", 7/10/2010
"Them there I's", 2/11/2010
"Fact-checking George F. Will, one more time", 10/6/2010
"What is 'I' saying?", 8/9/2009
"'I' is a camera", 7/18/2009
"I again", 7/13/2009
"Another pack member heard from", 6/9/2009
"Royal Baloney", 6/9/2009
"Inaugural pronouns", 6/8/2009
"Obama's Imperial 'I': spreading the meme", 6/8/2009
"Fact-checking George F. Will", 6/7/2009


  1. Bruce Rusk said,

    May 22, 2011 @ 11:32 am

    To be really thorough you'd have to think about how to compare short and long sentences with first-person subject, and decide whether it matters whether the pronoun is repeated ("I will dredge the lagoon and capture the swamp-beast that has been terrorizing our citizens" vs "I will dredge the lagoon. I will capture the swamp-beast…"). The latter has many Is but no more verbs whose subject is I, and in this case all that a raw count would be measuring is, arguably, a stylistic choice and the level of formality.

  2. Kylopod said,

    May 22, 2011 @ 1:31 pm

    Most presidential addresses aren't written by the president but by speechwriters. So even if you do find a thorough, precise method of ego-counting (for lack of a better term), it still would probably be meaningless as to assessing the president's temperament.

  3. Victor Mair said,

    May 22, 2011 @ 1:32 pm

    My wife, Li-ching, was nearly obsessive about avoiding the use of the first person pronoun, both in Chinese and in English, but especially in Chinese. In speech, though more so in writing, she would go to any length to avoid using the first person pronoun, unless it were absolutely necessary, and then she would only do so begrudgingly. She'd much rather use a passive construction or some other circumlocution to escape having to refer directly to herself. In the hundreds of letters I received from Li-ching during our marriage, I doubt that she used the FPSP as much as a dozen times. In fact, I don't remember her ever using it in writing and she only rarely used it in speech, such as when she would say something like "You take out the garbage and I'll was the dishes."

    Naturally, I noticed this extreme FPSP avoidance early on in our marriage and finally asked her about it. The answer she gave was that she considered it egotistical and almost immoral to refer to oneself. I suppose that she picked this up along with a whole package of other familial and social values she grew up with.

    Li-ching was the oldest child in her family, and it is interesting that — as I started to observe later on — the use of the first person singular pronoun increased among her siblings in reverse proportion to their ages. Li-ching was decidedly of the old school, having been born and experienced her formative years on the mainland before the founding of the PRC, while the others spent all or nearly all of their formative years in Taiwan, which was a hothouse of social change, with many American forces based there at various times. Still, none of them would ever refer to their aunts, uncles, and other elders, for example, by their first names, but only by their relational designations (e.g., jiujiu, bobo, shushu, laoye, laoniang, and so forth).

    Li-ching's abhorrence of the FPSP was also manifested in her behavior, for — in all respects — she was the most unselfish person I ever met.

  4. Spell Me Jeff said,

    May 22, 2011 @ 3:30 pm

    Would not some comparison between "graceful" I and "sledgehammer" I add depth to this discussion?

  5. Trimegistus said,

    May 22, 2011 @ 6:31 pm

    I remember how this 'blog spent so much time and effort debunking criticisms of President Bush's speechmaking. Oh, wait — no I don't.

    Why are you reprinting press releases from the DNC?

    [(myl) One of the links in the list above, which you obviously didn't read, happens to be a defense of Sarah Palin against a similar FPSP argument. As for your "failure to remember" our dozens of posts "debunking criticisms of President Bush's speechmaking", links have been provided for you several times before, for example on 1/9/2011 — or you could find them yourself with a trivial Google search. I conclude that you're either willfully obtuse or dishonest, i.e. a troll; in consequence, future comments from you will be deleted.]

  6. Craig Russell said,

    May 22, 2011 @ 6:56 pm


    I'll refresh your memory:

  7. Craig Russell said,

    May 22, 2011 @ 6:58 pm


    These links each contain further links — LL argued that rhetoric attacking Bush's speech was not based in fact dozens or even hundreds of times.

  8. Baba Ganoush said,

    May 22, 2011 @ 7:13 pm

    Would not some comparison between "graceful-I" and "sledgehammer-I" add depth to this discussion?

    Agreed. If the rates of "I-usage" don't differ that much from one president to another, perhaps we might examine the tone in which the I, me, my words are spoken, which does make a difference to the listener. Obama may think he is projecting self-confidence. However, a lot of his listeners seem to think he is projecting arrogance, which may have only a little to do with the actual I, me, my word counts. Would a subjective analysis of Clinton's/Bush's/Obama's delivery of a speech yield the same finding as an objective analysis of the text of the same speech? I think not….at least, as I recall, neither Clinton nor Bush were so badly prone to the shifty tennis-swiveling-head speaking style as Obama.

  9. Dan Lufkin said,

    May 22, 2011 @ 8:20 pm

    Swedish used to avoid both the first and the second-person pronouns. We've discussed using the third person for the second person several times. The reader can avoid using the first person (jag = I) by substituting "en annan" (= somebody else). Example: Får en annan titta på tidningen? (Could someone else look at the newspaper?) to mean "Could I look at the newspaper?).

  10. Jenny said,

    May 23, 2011 @ 1:34 am

    Cenedella thinks "it is rather jarring to read Obama’s addresses after all those who came before him." Nixon and Reagan were Californians, and the rest of the recent have been from the Southern US (at least by long residency, if not birth). President Obama is from Illinois. Could this be the discontinuity Cenedella hears?

  11. etv13 said,

    May 23, 2011 @ 2:28 am


    Nixon was from California, in the sense that he grew up here, but Reagan settled in California, and Obama in Illinois, as adults. I doubt Illinois has had a profound impact — or much at all — upon Obama's way of speaking.

  12. Steve F said,

    May 23, 2011 @ 6:10 am

    Presumably, those who consider use of the first person pronoun a sign of narcissism would rate Julius Caesar – notorious for referring to himself in the third person – as the most modest and self-effacing man who ever lived.

  13. [links] Link salad goes back to the work week | said,

    May 23, 2011 @ 7:44 am

    […] Presidential pronouns, one more time — Language Log on yet another weird conservative obsession with Obama. This reminds of the conservative outrage over the Obama 'feet on the Oval Office desk' photo a while back, which was in no wise tempered by the existence of a nearly identical Bush 'feet on the Oval Office desk' photo that had never bothered anyone. Not even liberals. Confidential to GOP in America: all the manufactured outrages make your genuine outrages a lot harder to take seriously. […]

  14. Bloix said,

    May 23, 2011 @ 9:45 am

    I happened upon an article today that summarizes research done of corporate quarterly earnings conference calls in an effort to determine whether there were language cues foe deceitful statements. Among the findings: CEO's who were presenting fraudulent results "use notably fewer first person pronouns," perhaps because use of them "implies a individual's ownership of a statement" whereas liars may "try to dissociate themselves from their words."

    The research paper, which appears to be unpublished, is "Detecting Deceptive Discussions in Conference Calls," by David F Larker & Anastasia A Zakolukina, Stanford University Graduate School of Business. The article is called "Liar, Liar" and appears in the April 2011 issue of ACC Docket, the Journal of the Association of Corporate Counsel.

  15. chris said,

    May 23, 2011 @ 10:21 am

    Presumably, those who consider use of the first person pronoun a sign of narcissism would rate Julius Caesar – notorious for referring to himself in the third person – as the most modest and self-effacing man who ever lived.

    And, of course, when he did say something in the first person, like "Veni, vidi, vici", he expressed his essential humility by omitting the pronouns.

  16. Julie said,

    May 24, 2011 @ 9:15 am

    I have spent long stretches of time in both France and the United States, (amongst others) away from my native UK. I do believe I have perfected the art seeming overly blunt and forthright to at least half the people I talk to whilst simultaneously coming across as a dithering annoyance to the rest.

    I am available for coaching sessions …….well, not really.

RSS feed for comments on this post