Advice for Academic (Job) Talks

I am deeply passionate about good visual design and making interesting talks.  

Short guidelines for any talk

• Develop and use stylesheets

• Telegraph what you're doing 

• Stick to your allotted space

• Use hidden notes

• Give a sense of something bigger

• Be authentic, personable, and wise.

• Tailor your talk to your audience

The point of a talk is not to give a powerpoint of a paper (even if that's what many people do anyways). The point of a talk is to give an engaging perspective on your material. Slides help if you use them well; they hinder if there's extraneous visual information or too much text. Every time you want to put text on a slide, ask yourself, "Could I do this better with a picture?" Every time you want to put a graph on a slide, ask yourself, "How could I build this graph to more effectively draw attention to the salient features?".

What is a job talk?

A job talk is more than just a research talk.  A job talk is also an indicator of how well you lecture.  It's used to tell the department what you've already done, and where you're going next: a sense that you can situate your research within your field(s), use your dissertation to build a successful research program, and perhaps become a leader (or a notable person) eventually.   At a teaching school, they want you to treat student questions with respect and show evidence that you can conduct your research with their limited facilities using undergraduates as researchers.  At a research school, they want you to engage intellectually with profs and show evidence that your research program will generate lots of grants.

You want to give them the sense that they would be lucky to have you, and you would feel honored to be there. 

Job talks in my field are meant to be (and often are used as, at least at smaller schools) departmental colloquia.  I aim my talks at the average student member of the audience (undergrad at liberal arts schools, grad at universities).  They are usually meant to be 50 minutes with 10 for questions at the end; no one cares if you go a little under, but everyone hates it if you go over. (We still call them hourlong talks, even if they're really 50 minutes.)

A job talk is not a business talk.  You should be able to give 30-second, 2-minute, and 10-minute overviews of your research. However, none of them are the hourlong job talk.  

How do I make it beautiful?

In my field (Physics), we have slides.  We don't read papers.  If your field reads papers, this might be a little outside what you do.  Honestly, I don't know how I would give a talk without slides.  How would I talk about my graphs?  

When it's time to make a new talk, I think about audience.  Which research project are they most interested in?   Are they cognitive scientists, physicists, or educators?  Will there be sound (I have small video clips related to some projects)?  Generally, I give background information close to the slides it pertains to, rather than a big chunk at the beginning.  You'll have to spend more time on motivation than on actual background material -- you can slip in background information ("This is exactly like we expect because of prior research") as you go.

I have three ways to think about beauty.  There's the beauty of each slide and the beauty of all them put together, and then the way they interface with your patter and timing.  (I'll leave aside the inherent beauty of your research.  Your research is beautiful, right?)

Per-slide beauty

Do not make ugly slides.  

• Use pleasing, harmonious colors.  Projectors are different from your computer screen, and different from each other.  I find that black, white, and blue or purple are best on most projectors.  Yellows are often greenish grey; greens are either sickly or neon; reds are either brown or pink.  Grays are nice, but hard to calibrate well because projectors are often darker than your screen.  

• Black text. White background. Do not be tempted to reverse this, because it will inevitably look bad on the projector when the room isn't dark enough or the black isn't black enough.  Do not be tempted to use other, lesser contrast schemes.

• Use a readable font.  I use 32 for most text, 24 for small text, and 16 for references and slide numbers.  I like Gill Sans or Helvetica light (except for numbers because 1 and l look the same, so I use Courier).  Times New Roman looks fussy.  Never use Comic Sans.

• In some (many?) rooms, you can't see the bottom quarter of the slides.  You may put references, the bottom parts of diagrams, and slide numbers down there.  Do not put important conclusions at the bottom of the slide, no matter how tempted you are to finish a slide with a conclusion.  Consider perhaps having it appear in the middle part of the slide as a build.

• If there is too much information on a slide, you must assist your audience with interpreting it.  Build large diagrams from smaller pieces.  Highlight the relevant parts of the graph in order as you talk about them.  Get rid of your long bullet points.  Make some whitespace.

• Do not do spiffy animations just because they are spiffy.  Ask yourself, does this build add information? Does it draw focus inappropriately from my message?  Generally, the appear and disappear builds are sufficient.  Perhaps also fade in and fade out.

Many slide beauty

Your slides are all in one presentation.  Make them look like they belong together, even if you stole them shamelessly from prior talks.

• Master slides are your friend.  Know them.  Use them.  Love them.  Put them together into a stylesheet and use it consistently.

• Builds are your friend.  Don't just duplicate slides with more info.

• Powerpoint themes are not your friend.  They are ugly and distracting.  The key to true beauty in a talk is functionality.

• Whatever your color scheme, keep the palette small and consistent.  I usually have five or six colors: background (white), text (black), highlighting (baby blue), data color 1, and data color 2.  If you can't keep it small, at least keep it consistent.

• Think about making a follow-along ball or progress meter.  In a talk with five sections, I often make five icons.  I put them in the lower left of the talk, and highlight the one I'm on.  It's small and unobtrusive, and I replicate it (larger and centered) on the structural slides that introduce each section.  This is more important in longer talks than shorter talks because it orients your audience to their location in the talk. 

• If you're part of a big collaboration, you're probably going to get slides from your collaborators.  You can use them, but try to make them fit visually with all your other slides.

Beautiful patter

It's your presentation.  Don't look like you don't know what's coming. 

• Rehearse your talk.  Rehearse it from the beginning.  Rehearse it from some place in the middle because they will interrupt you with questions.

• Set up your presenter display so that it shows the current slide, the next slide, the time you've spent (or the time remaining), and your presenter notes.

• Keep your presenter notes short.  Two lines are good; one line is better.  Or just skip them.

• If you're pointing out features on the graph (with highlighting and builds), look at the graph that everyone else is looking at.  It makes them feel like they are part of an experience with you instead of just being talked at.

• Deflect questions that come at inopportune times.  My favorites are, "Great question!  It leads me directly to the next part!" and "You are skipping ahead!  Hold your question for a few minutes, and if you still have it at the end, ask it then"  They let the audience feel clever for guessing where I'm going next (academics like to feel clever), and let me answer the question in the beautiful way I have planned.

• Encourage people to ask questions at the start, and maintain a respectful tone when answering (or deflecting) them.

• Make sure that that patter needs the slides and the slides need the patter.  It's one integrated talk, not a collection of pictures to accompany your speech!

Talks are not papers

In a talk, you want to give as many cues to listeners as possible about how to react especially if you're going to deprive them of visual cues.  Think about the cadence of speaking as separate from the cadence of writing.

It's ok -- desirable, even -- to use phrases like:

• This is surprising, because

• That makes sense, because

• This is exciting, because

• Recall that (thing you said earlier)

• If that were true, we would expect X.

And to tell them where they are in the talk:

• So, we talked about X,Y,Z.  Let's see how that can play out in context A.

• The central problem is X.  Before I talk about X, a brief digression into Y.

• Altogether then, the primary research questions are X,Y,Z.  I'm just going to focus on X and Y in this talk.

And to ask questions deliberately:

• Why should we care about X?

• What would be the instructional implication for Y?

If your questions are important enough, put them on a slide by themselves.

Think about your style as a speaker.  Find a way (within your style) to put funny parts into your presentation. I like to deliver mine deadpan.  I like to have one laugh per ten minutes of talk; more is a fun audience, less is a flat audience.  

Enough about style! Tell me what to say!

All that said, here's a basic outline from one of my talks with two research projects (numbers of slides in parens).  I gave this talk for the job interview at a faculty job that I accepted, so I know it was effective.  It's a little shorter than normal in terms of slide number, but a lot more of the slides have more builds.  If each build were its own slide, there would be 105 slides (which is about right for my style -- one build per 20-30 seconds on average).

• Title (1)

• What is my field (4)

• What this talk is about (2)

• Research into student understanding structural slide (1)

  •  my postdoc (6)

  •  Transition slide (1)

  •  my dissertation (12)

• Curriculum design (7)

• Conclusions (3)

• What I can do for you (1)

Total: 38 slides

The slides for this talk, color-coded by their function in the talk, are part of the meta-talk I gave titled "Framing your research to get a job", a pdf of which is posted in the paper repository under "Advice". 

If you ask me nicely, I will send you my original Keynote presentation of the job talk. You will find that the slides do not stand alone -- they need the patter just as much as the patter needs the slides.  The paper repository on zaposa has pdfs of selected posters and papers, but no research talks.  

Where else can I find information about giving talks?

Matt Might is a computer scientist with lots of advice for academics entering the job market.  Here's his piece on academic talks.

xBlog is full of visual design tips.  I like their piece on making movies in Keynote, as one way to think of a good talk is as a dynamic, live-action movie.