Narrative, storytelling and rhetoric

A data story asks the reader to follow a path the author chose. Annotation states which feature matters, sequence decides what is seen first, and interaction is offered or withheld. That design carries a cost. The more the author controls the path, the less the reader explores. This sixth of ten units, after the fifth tested interaction, finds the field has documented that trade rather than settled what it buys. The controlled comparisons that exist measure preference, engagement and recall of the story better than they measure whether the reader takes the data accurately from it. Whether a story improves comprehension of the data, or chiefly persuades the reader that the data has been understood, therefore remains open.

The trade-off the genre was built on

Segel and Heer (2010) give the vocabulary. They analysed 58 published narrative visualisations and derived seven genres, among them the magazine style, the annotated chart, the comic strip and the video. They named three patterns for interactive stories: the martini glass, the interactive slideshow and the drill-down. Their organising idea is a balance between narrative flow imposed by the author and story discovery left to the reader. The comic strip hands much of the reading to pictures, which is a design answer for a reader working in a second language.

Kosara and Mackinlay (2013) pushed the balance towards the author. They argued that presentation, and specifically storytelling, is the next logical step for visualisation research, and deserves attention equal to exploration and analysis. That is an agenda rather than a finding, and the field largely adopted it as one.

Annotation is rhetoric, and the authoring tools inherit the trade-off

Hullman and Diakopoulos (2011) named the mechanism. They proposed visualisation rhetoric as a framework for how design techniques that prioritise a particular interpretation shape what readers take from a story. They drew a parallel with framing studies in political messaging, decision making and literary studies. Their devices work at the level of the data, the visual representation, the text and the interaction, and each is an addition or an omission. Framing is therefore unavoidable and accountable.

Amini and colleagues (2015) analysed 50 professionally designed data videos. They then ran workshops in which cinematographers, designers and screenwriters storyboarded data videos. The techniques that came back were imported from film, so the mechanisms that shape interpretation in a static story reappear in motion, where they are harder to notice.

Satyanarayan and Heer (2014) interviewed journalists and built Ellipsis, in which a story is authored as a state-based scene structure with dynamic annotations and decoupled coordination across views, expressed in a domain-specific language. The reader's freedom becomes a parameter the author sets at each step. Lee and colleagues (2015) added the process account around such tools: find insights in the data, make a story from them, and tell it to an audience.

The catalogue grew faster than the evidence

The design knowledge accumulated by example. Stolper and colleagues (2016) analysed 45 curated data-driven stories and derived 20 techniques under four headings: communicating narrative and explaining data, linking story elements, enhancing structure and navigation, and providing controlled exploration. Bach and colleagues (2017) proposed data comics as a genre whose panel sequence can carry an analyst's reasoning. Bach and colleagues (2018) turned that into design patterns, each a set of panels with a named narrative purpose, drawn from infographics, data videos and the authors' own practice. Each contribution is a catalogue built by reading examples, which records what authors do rather than what works.

Tong and colleagues (2018) classified the literature by who authors the work, how narrative and transitions are built, and why storytelling matters, setting mature sub-fields beside open ones, with interpretation among the unsolved problems.

The controlled comparisons and what they measured

Two studies test the trade directly. Wang and colleagues (2019) compared data comics with infographics in two studies, one controlled and one in the wild, and report that participants largely preferred comics for enjoyment, focus and engagement, and that comics improved understanding and recall of information in the stories. Boy and colleagues (2015) ran three field experiments on a news outlet and a visualisation gallery, testing whether an introductory story increases engagement with an exploratory visualisation. On the first experiment, roughly four thousand sessions on a carbon dioxide emissions explorer, visitors in the story version made 42.7 meaningful hover and click interactions on average, against 43.7 without it. The intervals overlap, so the story did not increase exploration. The measure is interaction rather than understanding, so the result limits the engagement claim and leaves the comprehension one open.

Narrative improved memory of the story and did not increase exploration of the data. Neither study isolates the accuracy with which a reader extracts a value, compares a series or detects a trend from the story's hold on the reader.

Garreton and colleagues (2024) measured the attitudinal side directly. Comparing versions of a long-form article, they found that the version using only data visualisation produced a stronger emotional impact and a larger change in the reader's initial attitude than illustration-only support. The outcome is attitude, not accurate reading of the data.

Contested question: does a story improve comprehension, or chiefly persuade?

The first position is that narrative communicates. Kosara and Mackinlay (2013) treat presentation as the field's next step, and Wang and colleagues (2019) report improved understanding and recall alongside higher engagement, which reads as evidence that the narrative form helps readers. The genre's spread is offered as validation. The design space rests on a corpus of published examples drawn from journalism, design, business, art and research, rather than on a measurement (Segel and Heer, 2010).

The second position is that the evidence measures persuasion more than comprehension. Hullman and Diakopoulos (2011) show that narrative devices shape interpretation and can mislead, so a reader's sense of having understood may come from the framing rather than the data. Boy and colleagues (2015) found no exploration gain across three experiments, and the two studies that measure narrative form at all (Wang et al., 2019; Boy et al., 2015) measure recall, preference and engagement before accurate reading of the data.

What would settle it is a study that holds the data constant and varies only the narrative framing. It would then measure three separate outcomes: data-reading accuracy, retention after a delay, and the reader's explicit belief about what the data shows. The field has the authoring models (Satyanarayan and Heer, 2014) and the reading-level methods (McKenna et al., 2017) to run it, but not the result.

The two positions rest on different evidence. The narrative position rests on a design-space analysis of published stories (Segel and Heer, 2010) and a position piece (Kosara and Mackinlay, 2013). It also rests on one comparison that measured preference, understanding and recall together (Wang et al., 2019). The persuasion position rests on three field experiments that found no exploration gain (Boy et al., 2015) and on a survey that leaves interpretation open (Tong et al., 2018). While the dispute is open, test the story on data reading rather than engagement, and hold the data constant while the framing varies.

The strongest counter-argument

The strongest objection is that the thesis understates what has been shown, since Wang and colleagues (2019) reported that comics improved understanding as well as recall. The reply keeps the boundary rather than denying the result: their outcome is understanding and recall of information in the stories, measured alongside a strong preference effect and not separated from the reader's investment in the narrative that carried it. The result establishes a benefit in memory and engagement, which is close to persuasion, and leaves accurate data reading unmeasured.

Where the area is heading

The direction is towards studying readers. McKenna and colleagues (2017) analysed 80 existing stories for seven flow factors, including navigation input, level of control and story layout, then ran exploratory studies and a crowdsourced study with 240 participants on the factors shaping engagement, treating narrative visualisation as a reading experience with measurable structure. Chevalier and colleagues (2018) studied how teachers instruct and how children acquire visualisation principles in elementary school, so reading a chart appears as a skill that has to be taught. Wang and colleagues (2019) pair a controlled study with one in the wild, the shape a decisive comparison would take.

The sources this unit rests on

Each note states what the cited document did and found, from the resolved record and the abstract or full text.

• Amini, F., Riche, N. H., Lee, B., Hurter, C. and Irani, P. (2015). Understanding data videos: looking at narrative visualization through the cinematography lens. doi:10.1145/2702123.2702431 Analysed 50 professionally designed data videos and extracted their salient constituents, then ran workshops in which cinematographers, graphic designers and screenwriters storyboarded data videos. The techniques came from film, so the rhetoric of the static story reappears in motion.
• Bach, B., Riche, N. H., Carpendale, S. and Pfister, H. (2017). The Emerging Genre of Data Comics. doi:10.1109/mcg.2017.33 Proposed data comics as a genre that combines sequential panels with data visualisation to communicate data and its context, and set out its components. The authors argue the panel sequence can carry an analyst's reasoning in a way a single chart cannot, a claim about capability rather than a measured effect.
• Bach, B., Wang, Z., Farinella, M., Murray-Rust, D. and Riche, N. H. (2018). Design Patterns for Data Comics. doi:10.1145/3173574.3173612 Derived design patterns for data comics, each describing a set of panels with a specific narrative purpose, from an analysis of infographics, data videos and existing comics and from the authors' own experience. The patterns are offered as reusable solutions, which codifies craft rather than testing it.
• Boy, J., Detienne, F. and Fekete, J.-D. (2015). Storytelling in Information Visualizations: Does it Engage Users to Explore Data? doi:10.1145/2702123.2702452 Ran three web-based field experiments on a news and opinion outlet and a visualisation gallery, testing whether an introductory story increases engagement with an exploratory visualisation. Against the authors' hypothesis, the story did not increase user engagement in exploration.
• Chevalier, F., Riche, N. H., Alper, B., Plaisant, C., Boy, J. and Elmqvist, N. (2018). Observations and Reflections on Visualization Literacy in Elementary School. doi:10.1109/mcg.2018.032421650 Studied how teachers instruct and how students acquire visualisation principles and skills in elementary school, then used the lessons to propose research directions for visualisation literacy. Reading a chart appears as a taught skill.
• Garreton, M., Morini, F., Celhay, P., Dork, M. and Parra, D. (2024). Attitudinal Effects of Data Visualizations and Illustrations in Data Stories. doi:10.1109/tvcg.2023.3248319 Ran an experiment on long-form articles that compared data visualisation with illustration and found the data-visualisation version producing a stronger emotional response and a larger shift in attitude. It measures persuasion directly, the outcome the unit says the genre runs on.
• Hullman, J. and Diakopoulos, N. (2011). Visualization Rhetoric: Framing Effects in Narrative Visualization. doi:10.1109/tvcg.2011.255 Presented visualisation rhetoric as a framework for how design techniques that prioritise a particular interpretation affect what readers take from a story, drawing on framing studies in political messaging and decision making. The devices sit at the levels of data, representation, text and interaction, and each is an addition or an omission.
• Kosara, R. and Mackinlay, J. (2013). Storytelling: The Next Step for Visualization. doi:10.1109/mc.2013.36 Argued that presentation, and specifically the use of storytelling elements, is the next logical step for visualisation research and merits attention equal to exploration and analysis. The paper is a position statement, setting the field's agenda rather than reporting evidence that stories work.
• Lee, B., Riche, N. H., Isenberg, P. and Carpendale, S. (2015). More Than Telling a Story: Transforming Data into Visually Shared Stories. doi:10.1109/mcg.2015.99 Set out a visual data storytelling process in three steps, exploring data to find insights, making a story from those insights, and telling the story to an audience. They use it to name opportunities for research on visualisation as a storytelling medium.
• McKenna, S., Riche, N. H., Lee, B., Boy, J. and Meyer, M. (2017). Visual Narrative Flow: Exploring Factors Shaping Data Visualization Story Reading Experiences. doi:10.1111/cgf.13195 Analysed 80 existing data stories and identified seven flow factors, including navigation input, level of control and story layout. They then ran exploratory studies and a crowdsourced study with 240 participants, treating narrative as a reading experience to be observed.
• Satyanarayan, A. and Heer, J. (2014). Authoring Narrative Visualizations with Ellipsis. doi:10.1111/cgf.12392 Built Ellipsis, informed by interviews with journalists, in which a story is authored as a state-based scene structure with dynamic annotations and decoupled coordination across views, expressed in a domain-specific language, and evaluated it with award-winning journalists. The model makes the author's path explicit and the reader's freedom an explicit parameter.
• Segel, E. and Heer, J. (2010). Narrative Visualization: Telling Stories with Data. doi:10.1109/tvcg.2010.179 Analysed 58 published narrative visualisations and derived seven genres, among them the magazine style, the annotated chart, the comic strip and the video, plus three patterns for interactive stories: the martini glass, the interactive slideshow and the drill-down. The method is a design space analysis, so comprehension is characterised rather than measured.
• Stolper, C. D., Lee, B., Riche, N. H. and Stasko, J. (2016). Emerging and recurring data-driven storytelling techniques: analysis of a curated collection of recent stories. URL: https://www.microsoft.com/en-us/research/wp-content/uploads/2016/04/MSR-TR-2016-14-Storytelling-Techniques.pdf Reported a qualitative analysis of 45 curated data-driven stories and derived 20 techniques under four categories: communicating narrative and explaining data, linking story elements, enhancing structure and navigation, and providing controlled exploration. The catalogue records what authors do rather than what works.
• Tong, C., Roberts, R., Borgo, R., Walton, S., Laramee, R. S., Wegba, K. et al. (2018). Storytelling and Visualization: An Extended Survey. doi:10.3390/info9030065 Surveyed the storytelling and visualisation literature and classified it by who authors the work, how narrative and transitions are built, and why storytelling matters through memorability and interpretation. The classification sets mature sub-fields beside open ones, and interpretation sits among the unsolved problems.
• Wang, Z., Wang, S., Farinella, M., Murray-Rust, D., Riche, N. H. and Bach, B. (2019). Comparing Effectiveness and Engagement of Data Comics and Infographics. doi:10.1145/3290605.3300483 Compared data comics with infographics in two studies, one controlled and one in the wild. The authors report that participants largely preferred comics for enjoyment, focus and overall engagement, and that comics improved understanding and recall of information in the stories.