Oral Presentations and the Bandwidth Illusion

Which contains more information: a five-minute video or a five-page document?

As anyone who’s had to pay for excess cellular bandwidth knows, the video contains far more raw data than the text file. That doesn’t mean it has more information, though.

What if the video shows someone reading a Dr. Seuss book aloud? Most of the raw data in the video would just be a binary encoding of the speaker’s picture, and a tiny bit of motion around their mouth. The audio track would contain moderately paced speech in delightfully simple rhymes. In five minutes of viewing, the audience would easily get the point. Now imagine that the five-page document we’re comparing it to is a scientific research paper. Comprehending it – fully receiving its information – might require an hour of patient study, or longer if the reader isn’t an expert in the field.

The video contains a lot more data, but the paper contains a lot more information.

This fundamental distinction trips up many people when they try to give oral presentations. With slides and audio, the presentation seems as if it should have a lot of bandwidth. It would occupy gigabytes as a video, after all. It must be possible to cover at least a few research papers fully in a 45 minute presentation, right?

Wrong. Anyone who’s been to a scientific conference has seen the result of this conceptual error: rapid-fire talks, screens full of minuscule type splayed across multiple panels of data, and a set of conclusions impossible to evaluate. The problem is that despite appearances, an oral presentation is what the IT folks would call a very narrow pipe.

Let’s start with the words. Most conversations in English run around 100-150 words per minute. At that rate it would take about 20-30 minutes to read the text of a 3,000-word paper aloud. That will give the audience nothing but blank stares, though, because written and oral language aren’t interchangeable. The subject also matters. Scientific presentations are not Seussical rhymes. By its nature, research uncovers entirely new information in entirely new ways. If we know that the next thing someone says will be a name that sounds like “pinch,” it’s much easier to interpret than if we have no idea what’s coming. Just from the textual limitations, we can see that fitting multiple papers into a talk is going to require some compromises.

Adding graphics compounds the problem. The figures in a scientific paper are not frivolous. They are a means of compacting as much data as possible into as little space as possible while (usually) remaining comprehensible to a sufficiently dedicated reader. As Edward Tufte has demonstrated, one can cram several additional dimensions of information into a two-dimensional image, and indeed that’s now standard practice in many fields. Information-dense graphics work very well when the audience can unpack them at leisure, or when someone is willing to talk through the data step by step. They don’t work at all when flashed on a screen for 30 seconds during a jargon-laden spiel.

How do we avoid overstuffing our talks, then? While the subject of presentation design is deep and broad, I like to simplify it by reversing the process; start with the goal. In one sentence, what should the audience understand at the end of the talk? What information do they need to achieve that understanding? Implicit in these questions is some idea of who the audience is.

For example, if I want to explain how a newly discovered mechanism allows a virus to evade its host’s innate immune response, the audience will need to understand several things. They’ll need some background information on this virus and what’s already known about it, some idea how a specific aspect of innate immunity works, what experiments uncovered the new mechanism I’m discussing, and why I believe the results. For an audience of virologists and immunologists, I could abbreviate much of the introductory material and dive more deeply into the data. When speaking to a less specialized group, the emphasis might be the other way around. In either case, though, starting with the goal gives me an obvious outline to follow.

Speaking of outlines, that’s the next step: a brief outline listing the main sections of the presentation, and the amount of time each should take. The total time should be no more than the allotted time for the talk, of course. Everything else flows from the outline and those time limits.

Crucially, graphics come last. Slides should illustrate concepts that can’t be easily explained otherwise, or show data the audience is likely to care about. A corollary is that slides should contain mostly images, not text. Given the explanation above, it should be obvious that figures originally formatted for publication may need some adjustments for this more streamlined medium.

If you start building a talk this way and feel like you’re forced to leave out a lot of interesting information, then you’re doing it right. Not everything will fit. Remember that the goal is to communicate, not just talk. If it’s important to tell the audience more, point them toward a text document with the details.

It shouldn’t take them long to download it.