Psycho ... Linguistics ... Debates

As if you needed one more way of looking at the presidential debate... on behalf of a few of my linguistics students, my about-to-fail-me PC, the skip-back button on my iPhone and my children (who agreed to pretend like we didn't exist while we got this up and running) ...

I am pleased to present the psycholinguist's take on the third 2012 presidential debate: that is, an analysis of the slips of the tongue made by our candidates. (Data on the other two debates hasn't been sorted yet because I do sometimes still grade papers, eat dinner with my children and stop to enjoy the fall leaves.)

It's sort of startling how rarely we really make complete, grammatical utterances. Spontaneous human speech is full of sentence fragments, hesitations, restarts and pauses with with "um"s, "uh"s, "ya know"s and the ever-beloved "like". It's normal: we don't talk like the evening news.

A linguist would say that an "Error" in speech is a different kind of thing than any of those pedestrian speech disruptions. An error represents a systematic kind of failure: some thing that is both unintentional and counter to plan: not your sixth grade English teacher standing over you with a ruler to police your me-and-him(s), but your own mind watching in awed fascination at the things you say which you did not intend to say. It's like this:

These are probably not speech errors:

The short answer to the question: 'Why aren't those speech errors?' ... is that they were things that someone said intentionally. The speaker had a plan, and executed that plan ... as planned. A real speech error is a failure of your complex (pretty near miraculous) speech production system which surprises even you, if you notice it. You had that plan, but what came out of your mouth was something wholly unexpected.

So this, although inelegant to some extent, is not a speech error:

It's also not a speech error if you have a plan, execute it, and then decide that this wasn't what you'd meant. Usually, this amounts to a rephrasing of the same idea or a refinement of that idea. Like this... again, thanks to the Governor.

On to Mr. Schieffer: Here, the plan is replaced by something new, but the refinement is in the slight shift in meaning of a single word. The first thing he said was not wrong ... or most likely not wrong. You can use the speaker's intonation to tell whether, upon hearing what he'd said, he thought he'd made a mistake. In this case, I think he did not view it as an error... so neither do we!

Here's an instance where the plan stalls. The hesitations and repeats are attempts to make the plan go (darn it!), but go it eventually does. No real new or surprising elements butt their way in. This is not a speech error.

But, here finally, is a good, clean case of an error we can all agree on:

There's lots more on speech errors (or slips of the tongue as they are called for variety) here, so I'll wait if you want to read that and play some of the MP3s there with examples ... There's a bibliography there, too!It's not my intention to recreate that whole resource here, but rather to expand it in a slightly new direction.

Linguists and psychologists use errors in speech to try to learn about how the speech production system works. Linguists are nice people, and our mothers raised us not to call names or make fun of the way you guys all talk ... So, we're not generally on the prowl for utterances that reveal either your lack of facility with standard grammar or your squishy, deep, dark repressed secrets; we mostly just want to know how the human brain works: which, if you ask me, is squishy and deep and dark and secret enough really. These utterances are natural, stemming from the same sort of human fraility as whatever made you set the milk on the kitchen counter and leave your keys in the fridge ... come on ... admit it!

Errors in speech can tell us about the structure of human language and about the mechanisms through which we use that structure when we talk. We can tell what words are related to each other in the brain, how much of a sentence we plan at a time and, to some degree, how your environment impacts your ability to communicate. We might even eventually learn how one person's speech production mechanisms differ from another's ... and that is how we find ourselves here, looking at speech errors from the debate.

Let's Get Started!

Soooo, how did the fellas do with that debate?

All in all, over the merry course of the 90 minutes, we heard 64 genuine speech errors ... Of course we heard countless stumbles, restarts, grammatical errors, pauses and maybe even a few actual lies, but only 64 speech errors.

That's 64 errors over a 90 minute debate: I was writing something down every 85 seconds or so. But they weren't evenly distributed in time; they tended to occur in little flurries, giving me ample opportunity to get fast on the pause button.

And they weren't evenly distributed between participants either:


For most speech errors made in a publicized presidential debate in front of millions of people on international TV:

Would the candidates rise and approach the podium while I cue the band...

As the band plays "All Hail to Massachusetts":

Other Kinds of Results

That was fun: now, let's get into some more detail. Yes there were 64 errors made, but what does that mean?

Just to be contrary, we'll take the last question first:

Yes, that is a higher-than usual level of occurrence of errors. You can actually go quite a while before you hear an error in normal speech. You'll usually hear one or two within an hour of listening well (if you avoid listening to TV and movies and other genres where errors are systematically edited out. It's rare to get an error every 85 seconds ... unless you're a parent with a hungry cat and more than one child trying to organize breakfast:

"Sofia ... um ... Claire, would you get your hair out of the cereal, I mean the eggs. It's really trime to gro... has anyone fed the fitty?"

(Doubt me? Follow me around some Monday morning and you'll have enough data for two psycholinguistics dissertations and some left over for laughs.)

So, yes: 64 errors in an hour and a half... roughly an error every 85 seconds, is fun time error mania!

And what were those errors like?

Here's the breakdown:
That [b] carries merrily over from 'bad' to 'wolf'. the most fun sort of error of this type is the spoonerism: where the first bits of words change places.
The big wad bolf
Click Here for More Minformation
Error type What the heck does that mean? Non-debate example
Phonological Mistakes which have to do only with sounds of words Utterance: big bad bwolf
Target: Big bad bwolf
Morphological Mistakes with meaningful subparts of words Utterance: N.P.H. and I talk all nighted.
Target: N.P.H. and I talked all night
That -ed (which should have stayed put on talked, is a morpheme: it has meaning: it tells you the verb is in the past tense. It moved over onto 'night', creating a pretty odd sounding sentence. When there's an error with a prefix or a suffix, it's a morphological error.
click Here for More Informations
Lexical Errors with whole words Utterance: thank you for inviting me; it was a lovely funeral.
Target: a lovely wedding
Happens all the time... to everyone! Two formal, stressful, unusual events which mark transitions in a human life accompanied by ceremony. You don't need Dr. Freud here: there's good semantic reasons for mixing these up.
Click Here for More Intuition
Syntactic Errors in the construction of sentences These are often combinations of two sentences, or a sentence in which words rearrange themselves unexpectedly. Casting our minds back a couple years, we might remember this little gem:
Utterance: When wings take dream.
Target: When dreams take wing.
Click Information to Get more Here

Generally, we expect more-or-less 60% of errors to be phonological. Because of the way English is built (with not very many suffixes and prefixes and such) we don't expect many morphological errors at all: call it 5%. That leaves 35% of errors to be divided between lexical and syntactic errors. Here's a breakdown for you:

You can argue with those last couple of numbers if you like: that's how it has seemed to sort itself out for me though. I'll be glad to thumb wrestle the heck out of any linguist who wants to challenge my particulars.

There's usually, though, another column which we call 'combinations' or 'multiple sources' or 'Errors of Unknown Genesis', depending on how fancy you're feeling. Those are utterances you know to be errors but whose source and proximal cause you simply can't determine either because it's ambiguous (there is more than one source) or because the error is just too complex or weird or unintelligible and you can't work it out. In the age of enlightenment, we might expect this combination category to be quite small. If we just quietly toss handfuls of notecards bearing problematic data into the recycling... then it generally is pretty small. But whenever I do that, somebody always sees them there and helpfully returns them to my desk ... so I've given up that strategy ... at least until I move the recycling bucket somewhere less obvious.

In any case, we can expect to find a certain subset of errors which we cannot readily categorize, and those go into the mix as well, at least in terms of counting totals. They usually come out of the phonology/morphology end of the continuum. Like this one:

So without further fussing around

Here are the numbers from the debate. Out of 64 errors:

Those porportions don't look quite right: let's sweep those pesky 'combination ' errors back into the recycling and do the percentages again. So this time, out of (64-9 = ) 55 categorized errors we have:

Closer, but the phonological number is still too small and the lexical number, arguably, too big.

We have seen similar sets of numbers in other situations though: When asked to narrate a silent movie, subjects do tend to produce more lexical errors. You can make up whatever story you want about how a debate is similar to narrating a silent movie, (and whatever story you want about why I know what happens when 90 people (NINETY PEOPLE) narrate the first 12 minutes of the classic Cat and the Hat video wit the sound turned off...) but the take-home might be more or less the same: Pressure to communicate inflates the number of lexical errors? It could be.

The fixty-sore million dollar question

But what everyone really wants to know is ... Did the two candidates differ from one another in any interesting ways? Beyond the overall numbers, might we perhaps 'blame' the skewed category porportions on one or the other candidate? Let's see:

Gov. Romney's Errors

Total errors = 36

Removing the uncategorized: Total errors = 31;

Pres. Obama's Errors

Total errors = 20

Removing the combination errors we have 18 errors total;

Using the adjusted numbers in each case, here they are together, with Gov. Romney's errors first:

A word about Direction

For some errors, you can pinpoint a direction. Like this:

Show off now by thinking to yourself ... before I name it ... what kind of error this is?

It's phonological: a problem with sounds. Two smash-ins of the same target, but in the first, the problem is that the /p/ from purple gets re-used later in the sentence again when it should not. We say it "perseverates". In the second, the first sound in crocodile gets used early: we say it's "anticipated".

In English, we generally have more anticipations in speech than we do perseverations: probably due to something about English intonation. Here though, we find the numbers are about equal. Of the 19 cases where you could name a direction for an error, 9 were anticipations and 10 were perseverations.

Directions for Gov Romney's errors could be determined in 12 cases, and half were anticipations and half perseverations. Only 4 of Pres. Obama's errors could be said to have an unambiguous direction, and all four of those were perseverations. (If you're fast with the math you've noted 3 missing anticipations (12+4 does not equal 19 even using Microsoft chips.... These missing anticipations all belong to Mr. Schieffer, who will receive similar scrutiny when he runs for president...)


What's it all mean?

Well, first off, probably nothing. There aren't tons of errors here, which might explain why no statistician has returned any of my calls since October 23. And, well, uh, be sure to read the the disclaimer.

Now, if you are willing to grant that both candidates had the same amount of time to talk, and both were presumably under the same amount of pressure, you might ask yourself what mad conclusions might be drawn from these observed differences.

Frequently Asked Questions

(Or fraternally or fondly or fortuitously asked questions ... people don't actually ask me about speech errors all that frequently really to be honest.)


So, there you have it. I don't honestly know if I should call this debate in favor of the candidate who graced us with the most errors or for the one whose error profile looked more standard and who made fewer errors. Maybe, we oughta take a poll! Next time your phone rings, be kind: it might be a linguist.

Contact info; Email Dr. Sheri Wells-Jensen; swellsj at
Friday Night Linguistics Homepage

Find your debate audio and transcripts: I used the audio and transcript from NPR Here. Nice, clean downloadable audio and a clear transcript. Thanks, NPR: I pledged my local radio station, too... mostly because I'm afraid Ira Glass will call if I don't, of course.

So, you ask, how much of this is true? Hmmm... well... go ahead and spoiile a good thing if you have to... but, actually, quite a lot of it is true: yes, that's what a speech error is and that's what linguistss do with errors. And yes, as best as we can determine, those are the amounts and kinds of errors from the third presidential debate. It's all true about down to the 'conclusions' section at which point it makes a hairpin swerve off the straight and narrow into more interesting territory for a while just for fun.