October 08, 2004

A Milli Vanilli president?

An article by Dave Lindorff in Salon today, "Bush's mystery bulge", seems designed to bolster the rumors that in last week's debate, "President Bush was literally channeling Karl Rove" via a hidden earpiece. Technically, Lindorff just reports on the rumors, but the article has been widely cited as adding to their credibility.

The "evidence" cited includes Bush's "peculiar" pauses ("On several occasions, the president simply stopped speaking for an uncomfortably long time and stared ahead with an odd expression on his face."), the now-famous "let me finish" aside, and a bulge under Bush's coat, in back, thought to represent an electronics module of some sort.

Last Sunday I compared the pausing patterns of Bush and Kerry , and on Monday, I discussed the "let me finish" passage. At that time, I was skeptical of the "secret audio prompter" theory as an explanation for the "let me finish" aside, and argued that the aside seemed more plausible, in context, as a remark addressed to Lehrer. However, given that these rumors are now seeping into "big media", I thought I'd give the question another look (and listen). This prompted me to make some additional phonetic measurements, which turned out to bear on the question in a way that I didn't expect. It's by no means determinative, but it's enough to lead me to a tentative conclusion about these charges.

The main theory that Lindorff reports (taken from blogger Joseph Cannon among others) is this:

[T]he president and his handlers may have turned to a technique often used by television reporters on remote stand-ups. A reporter tapes a story and, while on camera, plays it back into an earpiece, repeating lines just after hearing them, managing to sound spontaneous and error free.

There are reports of several earlier news clips of presidential speeches where the audio is said somehow to have picked up the audio prompting as well as the live speaking, for example this CNN clip. Attempts to use this technique might certainly result in unusual pausing patterns, or other oddities of presentation. However, I have remained rather skeptical that the president used such a technique during last week's debate. At least, many of his answers seem spectacularly unlikely to have been recorded in advance by the kind of process that Lindorff suggests.

Geoff Pullum cited a characteristic example a couple of days ago: Bush's answer to Lehrer's question

Do you believe the election of Senator Kerry on November the 2nd would increase the chances of the U.S. being hit by another 9/11-type terrorist attack?

which consisted of the following sequence of phrases (with linking verbiage removed):

  • I don't believe it's going to happen.
  • I've shown the American people I know how to lead.
  • I understand everybody in this country doesn't agree with the decisions I've made.
  • People out there listening know what I believe.
  • This nation of ours has got a solemn duty to defeat this ideology of hate.
  • We have a duty to protect our children and grandchildren.
  • Ten million citizens [in Afghanistan] have registered to vote.
  • They're given a chance to be free.
  • They [the Afghans] will show up at the polls.
  • Forty-one percent of those 10 million [Afghans who have registered to vote] are women.
  • It's a phenomenal statistic.

Geoff wrote that "my reading of the whole answer is that we're looking at a man in a panic who has no idea what to say to the question. He has been taught a whole slew of tough-sounding clauses to reiterate, but can think of nothing to do but hurl them around at random." Whether or not you agree with Geoff's evaluation, and whether or not you like what Bush has to say, does it seem plausible to you that Bush would have recorded this particular meandering sequence as a package to echo back during the debate? By all accounts, he has excellent speechwriters who know how to craft a rhetorical structure.

The other trouble with the "echoing prepared material" theory is that it doesn't actually explain Bush's "let me finish" aside. If he's just echoing what's coming over the air, why not go on echoing? There's another theory, about handlers giving more indirect advice through the hypothetical earphone, which fits that piece of evidence better. But now the earphone-evidence argument is turning into a whole set of subtheories, one for each observation -- sometimes the earphones were feeding pre-recorded speechlets, which accounts for the pauses; and sometimes they were feeding advice, which accounts for the "let me finish" aside; and sometimes they were were feeding nothing, which accounts for the meanders. At this point, though, the theory has lost its explanatory force, as we linguists say.

This is not enough to prove that there was no earphone, or that an earphone played no role in Bush's pauses or his "let me finish" aside. But it seemed much more likely to me that his pausing reflected the normal cognitive stress of trying to select, arrange and reproduce chunks of memorized material in an appropriate sequence, and that his "let me finish" aside was meant to prevent Lehrer from going on to the next question.

So I decided to look in more detail at the timing of Bush's pauses and speech segments in the answer containing the "let me finish" aside. To start with, I reproduce below my transcript of the passage leading up to the "let me finish" business, with a couple of additional turns added, and adding duration measurements for the speech as well as the silences.

Jim Lehrer: Ninety seconds, Mr. President.
George W. Bush:
[pause 0.030]
[speech 2.156] Uh my opponent just said something amazing, he said [pause 0.465]
[speech 2.524]Osama Bin Laden uses the invasion of Iraq [pause 0.951 ]
[speech 2.180] as an excuse to spread hatred for America [pause 0.811]

[speech 2.833] Osama Bin Laden ((isn't)) gonna determine how we defend ourselves [pause 2.231]
[speech 1.952] Osama Bin Laden doesn't get to decide [pause 1.428]
[speech 1.233] The American people decide. [pause 0.762]
[speech 0.730] I decided [pause 0.306]
[speech 1.889] the right action was in Iraq. [pause 0.542]
[speech 2.068] My opponent calls it a mistake -- it wasn't a mistake. [pause 0.330]
[speech 1.355] He said I misled on Iraq. [pause 0.274]
[speech 4.117] I don't think he was misleading when he called Iraq a grave threat in the fall of 2002. [pause 1.303]
[speech 3.685] I don't think he was misleading when he said [pause* 0.126] that it was right to disarm Iraq [pause 0.965]
[speech 1.779] in the spring of 2003. [pause 0.609]
[speech 1.797] I don't think he misled you when he said that, [pause 0.565]
[speech 0.476] you know if you- [pause 0.372]
[speech 5.731] anyone who doubted whether the world was better off without Saddam Hussein in power sh- didn't have the judgment to be president, I don't think he was misleading. [pause 1.375]
[speech 1.128] I think what is misleading [pause 0.233]
[speech 1.239] is to say you can lead [pause 0.680]
[speech 1.096] and succeed in Iraq [pause 0.604]
[speech 1.680] if you keep changing your positions [pause 1.390]
[speech 1.335] on this war. And he has. [pause 1.256]
[speech 4.625] As the politics change, his positions change. [pause* 0.126] And that's not how a commander in chief acts. [pause 1.482]
[speech 3.273] I- t- I- uh uh w- let me finish ((here)). The intelligence I looked at [pause 0.807]
[speech 1.999] was the same intelligence my opponent looked at, [pause 1.520]
[speech 1.603] the very same intelligence. [pause 0.779]
[speech 6.879] And when I stood up there and spoke to the Congress, I was speaking off the same intelligence he looked at to make his decisions to support the authorization of force.

You can see that both the speech segments and the silent pauses vary quite a bit in duration. The question that I'm going to ask is how they co-vary: does Bush tend to pause longer before longer speech segments? If he's being fed lines in an earphone, I'd expect that he should have to listen to a longer stretch of prompt before launching into a longer phrase. Of course, if he's making it up as he goes along, it also might be taking him longer to plan a longer speech segment.

However, the correlation coefficient between his silences and the following speech segments in this passage is -0.05. This is not statistically distinguishable from zero. So there's no effect.

If there were an effect, we'd still have to decide whether this was due to prompting or just to the effects of compositional effort. But the fact that there's no effect makes it seem less likely, at least to me, that is is really being fed his lines. (I also have to say that it's hard to read this transcript and believe that the answer was composed as a whole in advance).

However, if we ask whether Bush tends to pause longer after longer speech segments, we get a very different answer: in this case, the correlation coefficient in this passage is 0.54. That's a respectable effect.

I'd confess that I didn't expect this at all. When I looked at the list of numbers, though, it kind of jumped out at me, and the correlation coefficient confirms the relationship. I have no idea why the effect holds -- maybe it's a rhetorical move, to give the audience more time to assimilate the longer phrases; maybe it's a compositional effect, with some short-term memory buffer taking longer to clear after finishing a longer phrase. It could be an accidental consequence of something about this particular answer.

Is this a general effect in public speaking? I don't know, and based on a quick literature search, I can't find any phonetic research that addresses the question. If some such research exists, I'll try to find out about it and let you know. If no one's ever looked at this simple question, it's a black eye for us phoneticians, I think.

So what's the conclusion? The rather meandering rhetorical structure of many of Bush's answers leaves me skeptical that he was getting his lines fed to him through a hidden microphone. The lack of any correlation between pause length and the following speech segment length tends to support this doubt, in my opinion.

The positive correlation between pause duration and the preceding utterance duration is fascinating, and worth following up on. If it's a general fact about debating, it's interesting to learn that. If it's specific to George W. Bush, or to a certain class of people including him, I'd like to know why. I can't think of any way to explain it in terms of the effects of a secret audio prompter, though.

As for that bulge in the back of Bush's coat, I have no clue what it is. I imagine that we'll find out, though.

[Note: the "Milli Vanilli" line is from Lindorff's Salon piece. It's kind of unfair, since what made Milli Vanilli a sacrificial llamb for pop inauthenticity, back in 1990 was lip syncing, not getting prompted.]

[Update 10/9/2004: an acquaintance who worked for some time as a producer at CNN told me that he has never heard of what Cannon describes as the "technique often used by television reporters on remote stand-ups", to have a pre-recorded version of their remarks played in their earphones as a prompt. That doesn't mean that it never happens, but at least it's not a routine method. And I haven't seen any validation of Cannon's assertion on this point by someone in a position to know. ]

[Wonkette pretty well sums it up: "Yes, we've seen the pictures. But we also watched the debate. If Bush was listening to some kind of radio signal, it was between stations." ]


Posted by Mark Liberman at October 8, 2004 05:46 PM