23 Oral narrative, data graphics
Until now, our discussions of narrative, visualization, and interactivity assume that the communication must stand on its own to deliver our messages to our audience.
But sometimes we may want to combine our data-graphics with a verbal narrative. Doumont (2009a) motivates oral presentations as,
Oral presentations are about having something to say to one’s audience and being able to say it in an articulate way while looking them in the eye. They are about engaging the audience with mind and body, conveying well-structured messages with sincerity, confidence, and, yes, passion. They are about presence, about being seen and being heard, not about having the audience look at slides that are in fact written pages, while hoping not to be noticed.
Perhaps we present orally for those reasons. But let’s consider two more. First, we are all social, and can benefit from having a direct, human connection to the material we care about. Second, from a communication efficiency standpoint, we have a direct feedback loop in real time through each aspect of the narrative we present, if we’ve prepared to enable that loop. And having direct feedback at the time of the communication can be extremely valuable, leading to more rapid progression of the ideas presented.
In those circumstances, we should adjust how we provide our explanations and visuals, and we have multiple options. First, we should consider whether our audience would benefit from pre-presentation review of important material and, if so, how we should craft that material. Second, we should consider whether, if at all, we should adjust how we visualize and explain our information during the presentation. Here, we consider both issues, in turn.
23.1 (Pre-)meeting preparation
Edward Tufte has advocated for beginning meetings with a detailed written (paper or electronic) document with the expectation that our audience reads them before a discussion, so that the meeting itself can become a more productive dialogue with deeper understanding. Most recently, he provides a chapter-length treatment of the idea in Edward R. Tufte (2020). Here, he sketches out constraints: the document “should be 2 to 6 pages long, written in sentences, with appropriate images and data displays. Do not send out your stuff in advance, people won’t read it.” He explains why:
Audience members read 2 or 3 times faster than you can talk. The document is in hand, everyone in the audience reads with their own eyes, at their own pace, their own choice of what to read closely.
In contrast:
In sllide presentations, viewers have no control over pace and sequence as the presenter clicks through a deck — viewers must sit in the dark waiting for the diamonds in the swamp.
Then why are you there?
Your job is to provide intellectual leadership, which is why you are making the presentation.
Doumont (2009a), too, recommends companion documents to presentations. Along with Tufte’s above advice, he offers five thoughts. First, content and credibility: “your audience seeks to learn What is the substantive content? What are the reasons to believe the presenter?” Thus, our document and presentation should explain the problem, its relevance and the so what, and our next steps for resolving the problem. To be credible all content must not just be truthful, but it should also cite sources, including data, quote experts on the specific issue, and reveal any potential biases.
Second, and his next point has been front and center in many of the earlier sections, we should think about our audience. Along with what we’ve considered, do not underestimate our audience. Do not “dumb things down” but, instead, respect their intelligence. Third, when you present after the audience has read your document, “do not repeat”do not merely repeat what they read.” Instead, go more in-depth, and refer back to the document paragraphs where relevant to tie together the two channels of information.
Fourth, practice your presentation ahead of time! More than once! Consider writing it out word for word, then speaking it aloud to get a feel whether the written sentences feel like natural spoken language, and to have an estimate of how the time required for presenting it verbally. And part of aloud rehearsal is to include white space — silence — for your audience to process your messages.
Finally, respect your audience’s time: finishing early is better than finishing late!
All these ideas can be helpful in particular contexts. But consider a course I teach in Columbia’s graduate program — Storytelling with data — and the pre-lecture study of primary source material, guided in part by this text. This occurs by necessity before discussion, before class. Primary-source material, for the most part, does not fit within Tufte’s recommended six pages for that which we cover during a weekly session. Certainly, one aim of the text before your eyes is to provide summary of various source materials, as well as highlight important ideas. Yet students are expected to prepare ahead of time.
Let’s turn, now, to the post-study presentation and, specifically, how to incorporate verbal communication and data-graphics.
23.2 Verbal with the (data) visual
23.2.1 Limitations with multi-modal communication
To motivate the value of time working on design of presentations, consult Edward R. Tufte (2008). In fact, after considering Tufte here, we may be wondering whether slides are even a good idea:
PowerPoint, compared to other common presentation tools, reduces the analytical quality of serious presentations of evidence.
This is especially the case for the PowerPoint ready-made templates, which corrupt statistical reasoning, and often weaken verbal and spatial thinking.
To answer that question, we may consider how Tufte himself has conducted presentations, as at E. Tufte (2016). It’s worth watching, and in it, you’ll find he does use presentation visuals! Edward R. Tufte (2008) provides reasoning. First, PowerPoint (still, today) has poor default formatting for any slide you intend to create, and encourages things like excessive bullet points or the equivalent to what we considered with data visuals: let’s call it slide-junk. Secondly, slides as presented can be very low resolution compared to paper, most computer screens, and the immense visual capacities of the human eye-brain system (id). With little information per slide, many slides are needed. Third, information stacked in time makes it difficult to understand context and evaluate relationships (id). Visual reasoning usually works more effectively when the relevant evidence is shown adjacent in space within our eye span. This is especially true for statistical data, where the fundamental analytical task is to make comparisons (id).
When designing visuals for a verbal narrative, use the right tool for the information. As Tufte explains, Many true statements are too long to fit on a slide, but this does not mean we should abbreviate the truth to make the words fit. It means we should find a better tool to make presentations. Secondly, increase data-ink on slides too, within reason. Doumont echos this point (as do I):
A major noise source on slides is text, especially a lot of text, for the audience cannot read one text and listen to another at the same time. An effective slide gets the message across (the so what, not merely the what) globally, almost instantly. There lies the challenge: to express a message unambiguously with as little text as possible. Visual codings being in essence ambiguous, effective slides almost always include some text: the message itself, stated as a short but complete sentence. Besides this text statement, the message should be developed as visually as possible: this development shoiuld include only whatever words are necessary for the slide to stand on its own.
Beyond including no superfluous information in their design, slides should include no superfluous ink in their construction. Optimal slides get messages across clearly and accurately with as few graphical features as possible: they are noise-free (Doumont 2009b).
Along with Doumont (2009b), Schwabish (2016) provides helpful general advice for designing visuals for presenting. To be sure, the broad strokes all paint what we’ve canvased previously. Again, per Doumont: get our audience to pay attention to, understand, and (be able to) act upon a maximum of messages given constraints. Indeed, as Doumont correctly notes, most slide visuals contain too much text. Period. When asking ourselves whether we’ve over-texted, consider whether we have created the slide for our own understanding, perhaps in part as speaking notes? Don’t do that. You may have notes, but they do not belong on your visuals, and do not show them to your audience. Second, are the slides to be presented to double as a written report? Don’t do that, either. Instead, if dissemination is important, create the aforementioned document, with full sentences, paragraphs, and a coherent narrative. If you must, you may also key the document to the slides. Or, less ideal, when circulating the slides after the presentation, include a (perhaps lightly edited) transcript of your verbal narrative. For examples of transcripts with visuals, see Corum (2018) and Corum (2016)1. But well-created presentation slides would not, in most cases, adequately explain their context without our verbal narrative. Third, if we find ourselves copying from written documents onto presentation visuals, perhaps because we’re running low on time, consider that this approach is not better than nothing. If the words distract your audience from listening to you, if your words are competing with those on your visuals, those visuals are worse than none at all.
23.2.2 Strategies for verbally communicating data-graphics
If text on visuals compete for attention with our verbal narrative, and the general advice above is use as few words as possible, how should we think about the lengthy advice in data graphics to annotate and explain, in this context?
That advice still applies! But some of the annotations and explanations are now verbal. We need strategies for how much explanation we place on data graphics that support verbal communication. And the first principle is pacing—understanding how quickly or slowly we reveal information to our audience.
23.2.3 Pacing and the visual narrative
When we present data visually, we face a fundamental constraint: eyes beat memory. As designer Amanda Cox notes, using our eyes to switch between different views that are visible simultaneously imposes much lower cognitive load than consulting our memory to compare a current view with what we saw before. This insight shapes how we should pace our presentations.
Rather than showing everything at once, we can reveal information progressively, letting the audience build understanding layer by layer. This principle applies whether we use static graphics revealed sequentially, animated transitions, or fully interactive displays. The key is controlling what the audience sees when, so they can focus on one concept at a time without holding complex comparisons in memory.
23.2.3.1 Animation as a pacing tool
“Animation” is an overloaded term that encompasses several distinct techniques for pacing visual information. We can distinguish three approaches, each suited to different communicative purposes:
1. Narrative storytelling
The first approach treats the visualization as a narrative journey, where data elements enter, exit, and transform to tell a story over time. This technique excels at showing how a system evolves or how relationships change across dimensions.
The canonical example is Hans Rosling’s 2006 TED talk, “The Best Stats You’ve Ever Seen” (available on YouTube). Rosling used animated bubble charts to show how countries’ life expectancy and GDP per capita changed over 200 years. Rather than showing all years simultaneously—which would create an unreadable tangle—he animated the bubbles moving across time, with his voice narrating the transitions. The animation served the narrative: viewers could watch the Industrial Revolution spread, see the impact of wars, and understand the AIDS epidemic in Africa as stories unfolding over time, not as static comparisons.
The key insight: narrative animation presents change as a continuous flow. The viewer experiences the process of change rather than comparing discrete snapshots. This technique is particularly powerful for showing temporal evolution, but can also illustrate other continuous transitions.
2. Transitions between states
The second approach uses animation to smooth the transition from one visualization state to another. Unlike narrative storytelling, which shows continuous change, transitional animation bridges discrete views, helping the audience track how individual data elements move from one configuration to the next.
A masterful example appears in the interactive introduction to machine learning at R2D32. The visualization guides viewers through how a decision tree splits data by smoothly morphing between views—showing raw data, then the first split, then subsequent refinements. Each transition is animated so the audience can follow specific data points as they move between states. Without these transitions, viewers would struggle to connect what they saw before with what they see now, forcing them to rely on memory rather than their eyes.
Transitional animation respects the “eyes beat memory” principle: instead of asking viewers to mentally compare a previous view with the current one, we show them the transformation directly. Their eyes track the change, eliminating the cognitive burden of remembering what was where.
3. Video-style playback
The third approach treats the visualization as a video sequence that viewers control. Rather than automatically animating, we provide controls for play, pause, stop, rewind, and step forward/back. This technique is useful when viewers need to explore at their own pace or revisit specific moments.
Tools like gganimate (in R) enable these approaches, though the workflow has evolved: we now typically specify the animation we want using natural language prompts to AI systems, which then generate the code using gganimate or similar libraries. The principle remains the same—we control what appears when—but the implementation now often begins with a clear specification rather than direct coding.
23.2.3.2 Progressive disclosure through layering
Regardless of which animation technique we use (if any), the fundamental pacing strategy remains: layer information progressively. We can apply this even with static graphics by revealing them incrementally during a presentation.
Start with the scaffolding: show the axes, labels, and legend first, explaining what the graphic will measure and how to read it. Only then reveal the data itself. For complex graphics, add data series one at a time, explaining each before introducing the next. This approach—sometimes called “layering” or “progressive disclosure”—ensures the audience understands the framework before confronting the data, and grasps each data element before adding complexity.
Practical example: When presenting a multi-series line chart showing quarterly sales by region, begin with empty axes labeled “Quarter” and “Revenue ($M)”. Explain what the chart will show. Then add just the first region’s line, describing its trend. Add the second region, comparing it to the first. Continue until all regions appear. By the final frame, the audience sees the complete picture, but they have built their understanding step by step rather than facing an overwhelming tangle of lines all at once.
This layering approach mirrors how we should verbally guide the audience: first explain what they are about to see, then reveal it in manageable chunks, using the sound of our voice to bridge the transitions.
23.2.3.3 Small multiples: Animation’s static alternative
Sometimes animation is impractical or inappropriate. Perhaps the data doesn’t involve temporal change, or the presentation format doesn’t support video, or we want viewers to compare states directly rather than remembering them. In these cases, small multiples offer an elegant alternative that respects the same “eyes beat memory” principle.
Small multiples display several versions of the same graphic side by side, each showing a different slice of the data. The viewer’s eye can jump between views instantly, making comparisons without holding one image in memory while looking at another. This approach is particularly powerful when we want to show how a pattern varies across categories, time periods, or conditions.
A compelling example appeared in the New York Times visualization “How Coronavirus Has Changed New York City Transit, in One Chart” (Penney 2021). Rather than animating a single map through time, the piece displayed small multiples—dozens of static maps arranged in a grid, each showing subway ridership at a specific moment during the pandemic. Viewers could scan across the grid, comparing March to April, weekdays to weekends, different boroughs simultaneously. The small multiples format allowed readers to see the full arc of change at once, choose their own path through the data, and make comparisons that would be impossible if the views were separated in time.
The choice between animation and small multiples depends on what you want the audience to do. Animation excels at telling a story in a specific sequence—the viewer follows your narrative path. Small multiples excel at enabling exploration and comparison—the viewer chooses their own path. When you need the audience to see relationships between states rather than experience the journey between them, small multiples often serve better than animation.
Both techniques, along with progressive layering and sound bridges, share a common goal: reducing the cognitive burden on the audience by letting their eyes do the work rather than their memory.
23.2.4 Sound bridges: Guiding attention across transitions
A powerful technique from cinema applies equally well to data presentations: the sound bridge. In film, a sound bridge occurs when audio from an upcoming scene begins while the current scene is still on screen, or when audio from the current scene continues after the visuals have cut to a new scene. The sound serves as a transitional device, smoothing the audience’s journey from one moment to the next.
Consider a classic example: as a character finishes speaking in one location, their voice continues over a cut to a different place or time. The audience hears the audio before understanding the new visual context, or continues hearing familiar audio while processing new imagery. This overlap creates continuity, preventing the jolt of an abrupt transition. The sound provides an anchor of familiarity that carries the audience through the visual change.
Application in presentations:
We can apply this technique when transitioning between data graphics or between a verbal explanation and a visual display. The principle is simple: let your voice bridge the gap.
Example in practice:
Imagine you are presenting a quarterly sales report. You have just shown a line chart of revenue trends, and now you want to display a bar chart breaking down sales by product category.
Without a sound bridge: You finish explaining the trend line, pause, click to the next slide, then begin: “Now, looking at this bar chart…” The audience is still processing the previous graphic while you are already showing the next, creating a moment of confusion.
With a sound bridge: As you finish discussing the trend, you begin the transition while the first chart is still visible: “What explains this surge in Q3? To understand the drivers behind this growth, let’s examine how different product categories contributed…” Your voice continues across the slide transition. By the time the new graphic appears, the audience is already oriented to what they should be looking for. Your voice has bridged the visual gap.
Key strategies:
Preview before you reveal: Begin explaining the upcoming visual while the previous one is still on screen. Describe what it will show, why it matters, and what the audience should notice. Then transition.
Continue across the cut: If you must switch slides before finishing a thought, let your voice continue uninterrupted while the audience processes the new visual. Your ongoing narration provides continuity.
Connect thematically: Use transitional phrases that link the previous content to what’s coming: “This trend becomes clearer when we break it down…” or “To understand why this matters, consider…”
Signal with your voice: Changes in pace, emphasis, or tone can signal transitions even as you bridge between visuals. A slight pause before a key point draws attention; increased momentum signals urgency.
Why it works:
Human attention cannot switch instantly. When we cut directly from one visual to another without warning or context, we force the audience to play catch-up. The sound bridge respects the cognitive reality that understanding requires continuity. By carrying our narration across the transition, we prevent the audience from becoming disoriented. They remain anchored to the flow of ideas even as the visuals change beneath them.
This technique transforms presentations from a series of disconnected moments into a continuous narrative, where each visual flows naturally from the last, guided by the thread of the speaker’s voice.