Delivering Effective Presentations in Science & Medicine

Content by Mike Urbanowski MD, PhD

As physicians and scientists, communicating complex concepts and ideas is central to our profession. In residency and beyond, you will likely deliver grand rounds presentations, case presentations, morbidity and mortality conferences, and conference posters. You will also lead inpatient teams and teach critical skills to your junior colleagues. Clear and effective communication is important to all of these endeavors.

Scroll

Three core philosophies to consider when approaching presentation design.

Clarity of Content is the Goal

Your audience is lending you their time and attention. Delivering a presentation that is intentionally, for instance by lack of effort, unclear IS rude. To avoid rudeness, try to place yourself in the position of an audience member who does not know much about your topic. From this perspective, consider what clarity of content would look like to you.

Create a safe space

Reassure your audience that it is alright "not to know." You’re here to help them understand — not to test them. No one learns well in a hostile room.

There are no absolutes in presentation design

The goal is always clarity, but there are multiple ways to achieve a clear and aesthetically appealing presentation. Experiment, fail, get better. To emphasize this point, this final bullet point is set to Comic Sans, long demonized in the font world and among designers of professional presentations - yet it is fairly readable and there is no reason that a good presentation could not be typeset in Comic Sans.


Topics covered in this presentation.

  1. Learning Theory
    Concepts in learning theory can inform the fundamental architecture of a presentation.
  2. Graphical Design
    Standards for turning data-points into accurate and effective graphics.
  3. Presentation Engineering
    Leveraging digital presentation software for effective storytelling
  4. Typesetting
    Be kind to your audience and gentle on their eyes.
  5. Deploying
    Ideas for preparing to deliver your presentation

1 | Learning Theory

Elements of learning theory can be used to structure a presentation.

Films like Inception are very complicated, but they are well received by audiences. Contrast this with a highly technical presentations where many professionals struggle to engage audiences and facilitate learning. I call this "the Inception paradox." Filmmakers and writers use a rich tradition of story design to draw in attention and facilitate learning. As physicians and scientists, we can learn from these cross-disciplinary practices.

Cover of the book Style: Lessons in Clarity and Grace

A lot of inspiration for the content in this section comes from the writing guide Style: Lessons in Clarity and Grace. Although this is a resource for better writing, the ideas described in this book have a lot of relevance in the presentation preparation space.

Element #1: Continuity Our brains learn better with continuity of topics because the context surrounding a learning point helps trigger retention.

Comparison of a bulleted list versus a continuous, linked flow of content.
Take a look at this figure. On the left is a highly simplified format that many technical presentations take. A presenter will say "you need to known that the first three letters of the alphabet are...". Now lets consider the right part of the image. Realize that each statement is couched in continuity with the topics around them. Now the presenter may say "A is followed by B. Remember B, well B is the second letter, and if you have gotten this far, the C is the third letter to learn.”

Element #2: Repetition Repetition in the introduction, content, and summary phases further facilitate learning and encourage retention.

Diagram of the repetition principle across introduction, content, and synthesis phases.
  1. Introduction — tell them what you’re going to tell them.
  2. Content — tell them.
  3. Synthesis — review, summarize, and put it in context.

So now you have two tools based on principles of learning theory that you can use to help with the initial planning of your presentations:

  1. Continuity of topics — This provides a microstructural framework. Use this concept to develop-out slides and transitions between slides.
  2. Repetition — This provides a macrostructural framework. Use this concept to plan the overall flow and components of your presentation.

2 | Graphical Design

Standards for turning data-points into accurate and effective graphics.

Our world (medicine, biology, chemistry, physics) is filled with complex and repeatable patterns driven by many factors and variables. These relationships are not always apparent to the viewer. However, graphics are leveraged to help our minds link these variables and make predictions. In fact, data charts and graphics are the most effective tool for telling complex data-centric stories.

Here are a couple historical examples...

John Snow's 1854 cholera map of Soho, London — deaths clustered around the Broad Street pump.
Here is John Snow’s Cholera Map. John Snow, known as the “Father of epidemiology”, used a graphical representation of deaths in the neighborhood around the Broad Street well (London) to deaths. Although he likely had little knowledge of the microbial etiology of Cholera, he was able to infer that the well was causing the deaths, and this knowledge also allowed him to create a life-saving experiment; when he shut down the well he stunted the spread of the pathogen and saved lives. John Snow, cholera map, London, 1854 (public domain).
Charles Minard's 1869 figurative map of Napoleon's 1812 march on Moscow and catastrophic retreat.
Here is another famous graphic, “Napoleon’s March on Moscow”. This graphic shows force depletion during offensive and retreat stretches of his campaign, which is widely regarded as one of the most devastating military campaigns in history. Notice that the width of each segment of time shows the relative size of his forces. Even more beautifully, the graphic overlays the map of his march itself with several salient geographic features enhancing our own human analysis of this event. Charles Joseph Minard, 1869 (public domain).

Four general guidelines for graphical excellence

These principles are drawn heavily from Edward Tufte’s famous The Visual Display of Quantitative Information. We’ll apply all four to a real dataset in a moment.

  1. The chart’s dimensionality should not exceed the dimensionality of the data used to generate the chart. Two variables → a 2-D chart, never a fake-3-D one.
  2. Area carries meaning. The eye reads the area of a mark — use bars only when height is the quantity.
  3. Maximize the ratio of information to ink. If an element adds no meaning, remove it.
  4. Use scales that reflect the real process. An exponential process wants a log axis.

Examples of these guidelines being put into practice. The data at the right shows number of E. coli bacteria present in a cell culture given a certain amount of time after inoculation. Data points are present for strains of E. coli; a wild-type (normal) strain and a mutant strain. If we put this information directly into microsoft excel, you will get an image much like the first one seen directly below.

Raw E. coli growth data plotted for wild-type and mutant strains.
Excel’s idea of “excellence”

Notice the automatic graphic software has taken two dimensional data (an independent time variable and a dependent variable for number of bacteria) and created a three dimensional chart. The problem is that our eyes and minds live in a three dimensional world and so perceive things which are further away as larger. Graphic software does not make the same correction. Therefore, we can misinterpret three dimensional data very easily.

Guide 1 · Dimensionality

The data has two dimensions — time and cell count. The fix is to reduce the dimensionality of the graph to two dimensions.

Guide 2 · Area carries meaning

Do we care about the height of each bar? — a couple hundred million or billion cells? Not really. We care about the trend. A bar areas do not contribute meaning and reduce the clarity thereby clouding the intention of the data chart.Instead, remove the bars and plot data-points instead.

Guide 3 · Remove excess ink

The current graph has a shaded background, which only lowers contrast with the data. Remove it.

Guide 3 · More excess ink

The horizontal gridlines invite us to read exact counts — but on this chart the exact number is not really needed to analyze trend. Remove them too.

Guide 4 · Appropriate axis scales

Bacteria grow exponentially — with lag, log, stationary, death phases. A linear axis makes it hard to see these biologically relevant phases. Switch to a logarithmic scale.

A little polish

Trim the axis, label the axes, size the type for a room, and distinguish the series by fill — not color (avoid red/green; use it sparingly).

Finally, just a couple more things to cleanup...

  1. Optimize the axis range to show a better spread of data-points.
  2. Add axis labels
  3. Increase the font size for better readability
Final polished chart standards: trimmed axis range, labeled axes, larger type.




Graphical traps.

As powerful as graphs are for revealing complex natural phenomena, when implemented in an inappropriate way, they can also mislead the viewer. Being able to identify some high-yield pitfalls is important.

Trap 1 · The pie chart

Avoid pie charts Social-network market share, 2017. Quick — is Pinterest bigger than YouTube? A pie encodes each value as an arc area, which the eye compares poorly. Illustrative shares, source: scc.ms.unimelb.edu.au (2017).

The fix

The same numbers as bars on a common baseline — now they rank themselves at a glance. When you need to compare parts, use bars, arcs.

Two ways to shrink a doctor icon to show a decline from 8,023 to 6,212 — scaling area exaggerates the drop.
Trap 2 · Misleading areas
Here is an example graphic originally highlighted by Edward Tufte in his book The Visual Display of Quantitative Information. It displays a story of how the number of doctors devoted to family practice is rapidly shrinking over time in California. The problem is that this graphic varies two dimensions at the same time so we see the relative comparison between Family doctors in 1964 (8023) as different from the number of Family doctors in 1990 (6212), though the actual raw numbers only represent a fall of about ¼ ; the 1990 graphical doctors appears perhaps 1/3 of the size of the 1964 family doctor.
Trap 3 · The baseline

Per-capita spending, 2017–2022. Spending like crazy! But look at the y-axis — it doesn’t start at zero. Illustrative figures.

Drop it to zero

Same data, honest baseline. When a bar’s area carries meaning in this scenario.

In Summary Be critical, be careful, and be truthful of your graphics.

Digital tools for graphic design: here are some tools that you can use to make data charts.

Software Pros Cons
Excel Highly accessible to most residents. Poor native graphical standards. Charts often require forced customization to achieve high graphical quality.
PythonR
Free, have options for integrated data analysis. Steep learning curve. Also, optimized for analysis, and sometimes have limited graphical styling options.
Prism Easy to use, best graphical standards Proprietary (expensive, less for students)
D3.jsChart.js
Expansive chart types, free, allows high standards, capable of interactive and dynamic graphs for power creators. Very steep learning curve

3 | Presentation Engineering

Leveraging digital presentation software for effective storytelling

PowerPoint offers dozens of options; when considering whether or not to use a transition or animation, ask whether the transition or animation will increase the clarity of the presentation.

PowerPoint is fundamentally a vector-graphics program. Every element is defined by math, not pixels — a position, a size, a color, and whether it’s visible. Magnify it 10,000× and it stays crisp. And because each of those properties is just a number, a single keystroke can change how an element looks.

PowerPoint's Format Shape panel showing a line object's numeric properties: position, height, length, color, and visibility.
Objects in powerpoint (lines, shapes, etc) are defined by parameters. By right clicking on the object and selecting "Format Shape", PowerPoint's format shape panel opens on the right. In this example, a line is featured which is defined mathematically by a position on the slide (x-cor, y-cor), a height, a length, a color, and a visibility variable. Use animations and transitions to changes these variables during the time the presentation is given.

Animations transitions control the flow of attention. Utilize animations and transitions that complement the theory of learning that we have already discussed.

Some suggested transitions and animations.

Transition/Animation Menu Parameter Action
Fade InFade Out
animations visibility Select this to make objects fade-in and fade-off of the screen. Often you will use "Fade" (green) to make objects appear in a certain order.
Fade transitions Softness of slide advancement Applies to an entire slide. Makes it so that one slide "softly" fades into another.
Line Move animations location of an object on the screen object moves from starting point to new location specified by a planning line placed on the screen.
Morph transitions many parameters Gracefully coordinates the movement of many objects on a slide by implementing multiple simultaneous animations. This is organized by creating a copy of the slide that you want to animate and setting the new copied slide next sequentially in the slide deck. Apply the morph transition to the slide. Finally, for each object that you want to animate (move the object, enlarge the object, shrink the object, change color... for instance) change the parameter in the second slide. When you click to transition slides, a complex "morph" or multiple simultaneous animations will occur.

Artificial intelligence in presentation design.

Artificial intelligence has come a long way in the last few years. With such a rapidly evolving, complex, and powerful technology, it is hard to know how best to apply it to your work. How do we judge where to spend extra time working "by hand" versus letting AI do work in the background?

Side-by-side comparison: an AI-generated anatomical illustration with garbled, nonsensical labels, next to a clean, accurate clinical algorithm diagram.
Artificial intelligence has come a long way in even the past three years. Shown on the left above is a slide copied from a published figure as cited. It's absurd and the end of a joke about how AI can hallucinate and cause mayhem. In 2024 we laughed at this, but more recently Dr. Zaki Ahmed has been applying AI and appropriately stringent human review to revise best-practice guidelines. On the right above is an example of the output of his work. This is an AI generated treatment flow diagram. While not perfect, it will very likely save time in the content creation process. There is a lot of optimism about how AI will save time and encourage high presentation standards in the future.

- But, CAUTION -

Do not use AI to analyze or visualize data directly. Instead, use AI to build tools to analyze and visualize data. Let's look at what this means...

Diagram: data enters a trained neural network and a data visualization comes out, with a note that solutions from trained hidden layers are near impossible to deconvolute, raising the question of who is responsible for hallucinations.
The temptation is there to have some data in a spreadsheet, feed it into AIs trained neural networks, and AI produces analysis or a data-graphic. Fundamentally, this is bad. Neural networks are trained on large sets of data. This training reinforces complex patterns of hidden layers while weakening others to produce a result that is identified as appropriate or beneficial. Once the network is trained, your novel data is fed in and a result is produced. The troubling problem is, who or what would be responsible for a hallucination, even one that looks completely appropriate? Moreover, traditional “debugging” or “post-mortem” of the inappropriate output of an AI-neural network is computationally intensive and, in some cases, impossible. In essence AI’s path to the solution you see is a black box.
Diagram: instead of feeding data directly through a trained neural network to get a visualization, route data through tools and code for data analysis, which produce the results and visualization; a note reads that the world can hold you responsible for this.
A better approach is to allow AI to generate hard-coded tools for data analysis and presentation. Once those tools are coded in black-and-white script, you (who approved the tools) or even the creators of an AI platform that produced the tool, could be held responsible, or at least implicated for mistakes in the analysis. As physicians and scientists, this is a responsibility that we must willingly accept.

Bottom line: let AI generate hard-coded, inspectable tools — scripts you can read and repeat with different inputs — that do the analysis and graphing.


4 | Typesetting

Be kind to your audience and gentle on their eyes

(even if they don't know it)

Typesetting:To arrange type (characters) or process data so that the composition of elements can be publicly presented.

Let's take a look at an example. Here is a fundamental phrase from a high-school biology textbook. Many viewers of this presentation will agree that there is something wrong with how it is typeset. It's small, unreadable, and placed strangely on the page.

The mitochondria is the powerhouse of the cell.

Too small, cramped, oddly placed. Unreadable.

The mitochondria is the powerhouse of the cell.

Now it overpowers the eye — also poor.

Themitochondriaisthe powerhouseof the cell.

Words cascade the way we read — top-left to lower-right — with emphasis by size and weight. Typesetting is always an artistic experiment; make it your own.




A few “presentation pillows” to consider.

An anatomical engraving of the human heart, used as a moment of reflection.
A moment to reflect. Anatomical plate, public domain.

5| Deploying Presentations

Ideas for preparing to deliver your presentation

You can pour weeks into a talk and a poor delivery still undoes it. A few things that protect the landing.

Your slides are not your notes. People learn better when more senses are engaged. Don’t read the text on the screen — let the visual carry the structure while your voice carries the teaching, and the two reinforce each other.

Practice — out loud, in an empty room, and then in front of mentors or colleagues who’ll give you honest, critical feedback.

Acknowledge complexity, then organize it. When something is genuinely hard, say so — and then take responsibility for bringing order to it. Take this classic:

Deaths aboard the Titanic, 1912

Survivors and deaths, split by class and sex. There’s a lot going on. Rather than read every bar, let’s orient, then pull out the pattern. Counts: the classic Titanic dataset (Board of Trade inquiry, 1912).

The first pattern

Across every class, men died at far higher rates than women and children — “women and children first,” in the data.

The second pattern

And class mattered: first-class passengers were markedly more likely to survive than those in third. Now the audience can find their own trends.

On posters, put the most important thing — usually your novel data — at the visual center, where the eye is drawn. The layout can be a classic triptych, or something new; experiment.

Mike Urbanowski's award-winning ACP 2025 clinical vignette poster, laid out as a two-axis infographic.
A poster that broke the triptych mold — time flows down, modalities across — and won at National ACP 2025. Static displays still leave room to experiment. M. Urbanowski et al., ACP 2025 (author’s own work).

And don’t linger. It’s tempting to labor over a slide and then talk to it for five minutes. Most audiences won’t follow. George Lucas built the Return of the Jedi sand barge over months in the desert — and put it on screen for seconds. Effort spent building a thing doesn’t entitle it to the audience’s time. Show the important part, then move on.


Review and Synthesis

Learning Theory

  • Continuity of topics
  • Repetition

Graphical Design

  • Graphs reveal measurable patterns in nature
  • Clear and minimize clutter
  • Be truthful to the eye and to the process

Presentation engineering

  • Use visual presentation transitions and animations that complement learning theory
  • AI will be helpful and possibly time-saving.
  • Never directly analyze or visualize data with AI; use AI to build tools to analyze and graph data instead.

Typesetting

  • Set for clarity
  • Study masters
  • Create your own style.

Deploying presentations

  • Practice!
  • Engage multiple senses
  • Organize complexity for your audience
  • Do not linger on a slide.