This job talk talk (PDF slides) by Arthur Spirling is really good. I agree strongly with 95% of it, am aggressively indifferent to 4% of it, and strongly disagree with 1% of it. This post is about that last 1%, but it should not be missed that I weakly agree with 99% of what’s in there.
My disagreement is with the statement “For the love of all that is holy…do not write a ‘script’, do not learn a ‘script’.” My view is unusual enough that it warrants a fuller explanation, and that’s what this post is about. I think that nearly everyone should write out their entire job talk in advance.
First, some background. I wrote out my job talk for a particular personal reason which is probably not relevant in and of itself: I stuttered as a child. That stutter has almost entirely disappeared (the exceptions are when I try to speak German or when I try to read long passages of text aloud–more on that later) but it was significant enough that I have never forgotten it. One fear that I had when going on the job market was that my stutter would reappear, a result of nervousness while “on the spot” during a job talk. Similar things happened on a couple of memorable occasions when I was a child.
To ease my nerves, I decided to write out my entire job talk. Yes, like a script. The idea was that if I found myself incapacitated by a bout of stutters that prevented me from speaking extemporaneously, I would have a print out of the talk available and I could just read it, or worst case scenario I could email the text to people later. (That I know that I still stutter when I read text aloud never occurred to me as a problem.) The point is, for reasons having not much to do with any strategy for preparing a proper talk, I wrote out what I would say.
Doing so had such clear benefits that I recommend this to anyone. I can summarize why easily: writing out your job talk makes it a real object that you can manipulate.
Writing out a job talk script allows you to do a couple of things that even practicing it 30 times (which I also recommend) will not do.
- Identify and master tricky key passages. There are parts of every job talk where precise wording matters. They are also the parts that are easiest to flub. Everyone has heard someone try and fail to describe what the identification conditions of a difference-in-differences model are. Perhaps your talk requires you to define a term like “mimesis” in a particular way. In my case, I needed a quick and clear explanation of why nominal exchange rate depreciations across a region don’t have real effects on exports in any individual country. I didn’t know this at the time, but I had to practice that to get the wording right… no weasel words, no “you-knows,” no speeding up to cover as much as I could and losing everyone as a result. The best way to get around this is to sit down and write out how you want to utter those three or four sentences. If you’re doing that, you’re already writing out your talk.
- Separate the talk from the slides. I am a firm believer that slides—Beamer, PPT, whatever—should be spare enough that they do not need to be read. Do you know what people often use slides for? As notes to themselves, like a crib sheet or an incomplete transcript-in-advance. The slides are there to facilitate or support the talk, not to summarize it. But if the talk does not exist except for in your mind (when you’re not practicing it) or in your stream of consciousness (when you are in the midst of practicing it), there is a tendency for slide creators to do something worse, which is to effectively write out the talk on the slides.
- Adjust your talk at the level of the phrase, sentence stream, or section, at your leisure. If you have written out your talk, you can edit it too, just like you can edit your slides (which you will do obsessively). If you have a text, you need not require yourself to remember that the next time you practice you wan to try to make point A in a slightly different way, or to move it after point B. Editing a talk that only actually exists as you are giving it is hard. Your technology to do so is basically confined to editing the slides (see previous).
There may be other benefits—it gives you a sense of control, perhaps, and provides you with the ability to visualize how much longer an additional paragraph’s worth of text will make your talk. But the key thing is that the talk becomes an object that you can control. Think of it this way. Many of us have had good ideas that were incomplete until they have been written down, and we could not have known it any other way but by writing it down. Practicing a job talk for the first time will tell you what’s wrong, but without an object to manipulate it is hard to respond.
Two caveats, one big and one small. The small one is that you probably only need to write out your first job talk. It gets easier and you get better at it.
The big caveat is that writing out a job talk is different from reading it, and your job is not to memorize each and every word. I do not recommend reading a job talk if you are a political scientist. If you memorize each and every word, you are likely to deliver your talk like a robot, so this is not the goal either. Your goal is to have done the talk so many times that you do not need to memorize it. Arthur is right that the talk is not a recital.
So my advice is to write out your job talk. This probably applies to you, although it might not. Do not read that script, do not memorize it, but do write it. And don’t worry, if all goes well, no one will ever know.
 I am specifically indifferent to the advice on questions. “You can signal weakness and unsuitability for employment by…stating explicitly that you won’t be taking questions during the talk (why not? because you can’t cope?).” It is true that you might not want to do this at NYU or about 10 other departments. Nearly everywhere else it is fine to say “I’d prefer to save all substantive questions until the end, but I am happy to take any clarification questions at any time.” There’s nothing weak about not wanting to be interrupted except for clarifications which you are happy to provide. But in fact, this normally doesn’t matter, because you usually won’t be interrupted in the first place, and you might even be asked by the search committee chair how you would prefer to handle questions, in which case you should say what you’d prefer.
I am also indifferent to “giving a bad one will guarantee you will not get the job,” which is false, but we live in a world in which we don’t want to see bad talks, so if you are a student you should act as if it is true.
And finally, I am indifferent to the advice to present results with “use + or – signs, and embolden/color for significance.” I am old fashioned about this, I actually want to see numerical results. Others insist on graphs (which is fine I suppose but these do not add information [PDF]). I just don’t think that there is enough of a consensus in our field about what you should do to be normative about this.
 It was not ideal to have a last name with repeated bilabial stops.
 I believe that this norm does not generally apply in political theory.