Wagner research group

Giving talks

There are many types of talks that one must give. It’s important to understand why you are giving the talk. It’s also important to understand who you are giving the talk to. Ultimately, you are trying to communicate something to your audience, so you must understand what you are trying to get them to understand and what it will take to get those particular people to understand that thing.

Novice speakers are often tempted to try to communicate their experience in the research project. The problem with this is that people can only absorb a certain amount of new information in a given sitting. Think of each of your audience members as having a limited amount of attention (let’s call it an attention pie) they can give you. Each new concept, fact, graph, etc that you introduce fills up that pie a little bit. When the pie is filled, that audience member is done, and will not understand much more about your talk.

Audience members will have their pie filled at different rates depending on how familiar they are with the subject. Familiar topics and graphs consume less attention pie. Unfamiliar topics and graphs consume more. This is why you often see people getting lost as a talk progresses, until only the experts are left following that particular talk.

There are two strategies to managing your audience’s attention pie. The first and most important is to cut out extraneous information. The best way of doing this is to understand where your audience can start from, and draw the most direct path possible to the main point you want to make. Anything not on that direct path should be removed. The second is to use familiar concepts so that the attention pie is less consumed by each statement in your talk. This method can include using familiar graphs, familiar methods, or even a light dosing of analogies.

10 minute talk (APS March Meeting, for example)

You get to say one thing. Template for slides:

  1. The question you are trying to answer. How it connects to larger goals in the field.
  2. The answer you will give (do not try to surprise your audience!).
  3. What your strategy is to answer the question.
  4. A clear explanation of your technique and model, including any parameters or approximations.
  5. A graph showing that your method is validated.
  6. A graph showing how your results answer the original question.
  7. What your results mean in the broader context.
  8. A reiteration of the logic of your talk, from question to answer.

The APS talk functions mainly as an advertisement for people who may want to look up your paper for more information, or to let people know what you’re working on for further conversation/collaboration. Typically you will not be able to explain your results in detail due to the short time frame.

20-30 minute specialized talk (Workshop talk, APS invited talk, group meeting talk)

Here you get to say 1-2 main points. I would suggest to keep it to one point, until you are very comfortable with more and the two points are related.

To a large extent, this will be an expanded version of the APS talk above. You get to show a little more detail and perhaps include a few more results.

I put the group meeting talk in this category because, while the group meeting is typically one hour, there is usually a lot of conversation.

1 hour specialized talk (condensed matter seminar, workshop talk)

You get 1-3 points here. I recommend fewer points if the audience is more diverse (see colloquium below), and more points if the audience is exactly in your specialty.

Prelim talk (20-30 minutes without questions)

These talks are a little different from the others; the point that you are trying to get across is that you have a reasonable plan for the rest of your PhD. Keep in mind that the prelim is in place to protect you from spinning your wheels unnecessarily. It is often a positive experience. That means:

  • The plan is well-researched and will answer an important question in the field. The successful completion of this plan will advance knowledge in a meaningful (doesn’t have to be huge) way.
  • You have the skills necessary to execute the plan. This is often shown through previous work.
  • The plan is feasible, and you know this either through your own preliminary work, or through other work which you could plausibly replicate.
  • You have mapped out the steps of the plan well enough that it is plausible you could complete it. Any potential showstoppers are known and you have a backup in case you encounter some.

Since the Q&A for this talk is often extensive (prelims are scheduled for 2 hours, while your actual talk should be 20-30 minutes), backup slides can be useful, but don’t feel like you need to show them if they don’t come up.

Defense talk (~30 minutes without questions)

Your objective here is to show that you have advanced scientific knowledge in a meaningful way. This means that you must establish:

  • There was a scientific question to be answered.
  • You answered the question in a satisfactory way.
  • You have communicated the question and the answer in a clear way to the scientific community.

Part of your talk is to prove that last point. Since it’s a short talk, you likely will not be able to talk about all the work you’ve done. Therefore, you should choose roughly one project to present, and mention the others (they will be in the dissertation). That project should be one that you’ve led.

Job talk (for postdoc)

You want to convince your prospective employers that you will successfully complete research projects (see the defense talk for how to establish this), and in particular successfully complete the projects your employers want completed. This looks a lot like the defense talk; you’re trying to show that you are a well trained researcher. It’s important that you remember the communication part of research; if the work has not been communicated, it may as well have not happened.

1 hour general talk (colloquium, diverse audience)

You get one point here. You might think that you get more, but since the audience is more general, you have more work to get them to the point. The attention pie of your audience gets filled more quickly because the concepts are unfamiliar to them, so you have to spread things that would be a quick mention over more time for such a talk. A colloquium/general talk will have a lot more background information and aids to understanding to avoid overloading the audience. A general rule is to try to aim the talk at the upper level undergrad level.


Set up a feedback loop in which you can judge for yourself whether you are effectively communicating. Strategies:

  • Write out what you plan to say. Read it back aloud and see if there are things left unexplained.
  • Describe in detail all the aspects of a plot you want people to understand. Do you lose patience? If so, there is too much on the plot.
  • For each slide, write a single short sentence that explains the takeaway message. If you can’t, then the slide is unfocused.

Succeeding at Q&A

Answering questions takes practice. Some notes:

  • Let the person answering the question finish their question. Don’t cut them off or try to guess what they’re asking before they’ve stated the question.
  • If you are miked and the questioner is not, it is often polite to repeat the question, possibly in abbreviated form. You may rephrase it to be a ‘better’ question while you do this, which can make answering easer. If you do rephrase the question, ask for confirmation that was actually the question.
  • Try to smile and give the questioner your full attention. 99% of the time, they are not trying to ‘get’ you, and even if they are, they will look bad if you treat them with dignity and kindness.
  • Give the questioner as much credit as possible.
  • It’s OK not to know the answer to a question. If you don’t know the answer to most questions, or you don’t know the answer to simple factual questions about your work (“Did you optimize the structure in DFT,” for example), then your preparation may not have been sufficient.
  • Try to answer only the question that was asked, as briefly as possible. If the question requires a subtle answer, it may be better discussed offline. Give a brief summary answer and say it’s subtle and the details can be discussed in person.