Branching Narrative Structures Explained

A branching narrative is a story written in advance as a set of paths: the material — scenes, clips, dialogue — exists before publication, and what varies is the order and the subset the audience meets. That is the boundary with open-world games, where events are computed from simulated systems rather than selected from a fixed set. The base terms are collected in my pillar article on interactive cinema; here we look at the shape of the paths themselves.

The typology I use is from my article “Interactive Cinema on DVD in the Discourse of Audiovisual Art” (Baltic Journal of Legal and Social Sciences, 2025; DOI 10.30525/2592-8813-2025-3-12). It distinguishes four strategies — nodal branching, parallel, labyrinth, environment-based — by two questions: who assembles the sequence, the author in advance or the participant during the viewing, and how much of the authored material a single pass can ever contain:

For each structure: how the paths are organised, at least one verified example, and the trade-offs I keep in mind when designing one.

Nodal branching

The story pauses at decision points; each choice switches to a pre-made branch, and branches usually rejoin in shared “hub” scenes. This is the oldest and most common structure. Kinoautomat (1967) let the hall vote on what happens next; in the DVD era Scourge of Worlds: A Dungeons & Dragons Adventure (2003) asked viewers to decide at forks, and Choose Your Own Adventure: The Abominable Snowman (2006) turned the gamebook tree into DVD menus; today the same scheme runs Bandersnatch and Late Shift. Final Destination 3 (2006) applied it to a mainstream home-video release with its “Choose Their Fate” option. In my FMV-01 prototype the hire-or-walk fork is exactly this: two branches out of one decision.

• + Legible choices: the audience always understands that a decision happened; bottlenecks keep production sane; works on every carrier from cinema to streaming.
• − Scene count grows with every fork; deep trees get expensive fast; thin branches make “wrong” choices feel punitive.

Parallel / multi-perspective

Several storylines run side by side, and the work is built for switching between them rather than following one: the audience assembles the whole picture from perspectives. Linear film knew the grammar — the Rashomon effect retells events from several viewpoints — but there the retellings followed one another; interactive carriers make the tracks co-present and switchable. The clearest DVD-era example is Late Fragment (2007), an interactive multi-protagonist film from the Canadian Film Centre and the National Film Board of Canada, whose three protagonists’ stories the viewer moves between — the “hypermovie” discussion around multi-protagonist DVD films described exactly this structure. In games, Detroit: Become Human runs three protagonists’ threads in parallel.

• + The whole is richer than any single pass; switching is cheap — new perspectives, not new endings.
• − Viewers may miss whole tracks; the design must reward switching without punishing a single-track viewing.

Labyrinth

A web of scenes where paths cross, loop and rejoin; there is no single spine, several entries lead into the same passages, and the reader can wander. The canonical example is hypertext: Michael Joyce’s afternoon, a story (1990), the founding work of Storyspace fiction, analysed in detail in Aarseth’s Cybertext and Murray’s Hamlet on the Holodeck. In games the logic survives in The Stanley Parable (2013), where corridors loop and the narration comments on the routes you take. Film rarely affords a true labyrinth — too much footage — so on DVD and in FMV it appears in diluted forms: alternate endings, re-crossing subplots.

• + The structure itself becomes the experience: wandering, recognising loops, building a mental map.
• − Hard to QA; easy to produce dead ends and orphan scenes; “completion” is hard to define.

Environment-based

No default route at all: the work is a database or a space, and the participant assembles the sequence by searching, inspecting, exploring. Her Story (2015) turns a police interview archive into the whole narrative; Immortality (2022) does it with a director’s lost filmography — in the DVD article’s typology these are the environment-based structures among modern hybrids. The DVD era had its own step in that direction: Tender Loving Care (1998) mixed film playback with psychological questionnaires and case files, treating the disc as material to explore rather than a line to follow. My FMV-01 borrows the idea locally: the nodal spine carries still frames whose hotspots reward inspection.

• + Replay value without new footage; the viewer’s own assembly becomes part of the authorship.
• − No guaranteed dramatic arc; discovery depends on the participant, so exploration itself must be designed as meaningful.

A naive tree doubles the material with every fork: ten binary choices in a row would need over a thousand paths. Production reality demands the opposite. Four techniques keep branching affordable:

• Bottlenecks (hubs) — branches diverge and rejoin into a shared scene, so each fork costs its own segment, not its own universe.
• State instead of forks — one scene reacts to accumulated flags (what you said, what you took) instead of existing as separate variants.
• Foldback — branches start differently and reconverge on the same key beats, preserving the arc while faking breadth.
• Reordered access — the same clips revisited in different orders: the environment-based trick; a database keeps branching after release, in the viewer’s head.

In FMV-01 the working answer is a graph of five node types — dialogue, choice, still, cutscene, ending — where scenes do not own the story; the graph does. The full walkthrough with the scheme of the vertical slice is in the first devlog. Editing matters as much as topology: in interactive cinema the cut itself is where variable paths are assembled — the subject of my article on montage as a tool for creating a variable narrative.

Every structure above can be drafted on paper — index cards for nodal trees, a grid for parallel tracks, a graph sketch for labyrinths — and for small projects paper is still the fastest tool. When branches outgrow the wall, dedicated software takes over: Twine drafts link-based hypertexts for free; Arcweave is built specifically for branching story maps; my own FMV Builder is an authoring studio for interactive films and visual novels — a branching editor with FMV mechanics and export to a web or offline build, grown out of the FMV-01 production (project page).

The tool matters less than the discipline: model the graph before shooting, mark the hubs, keep one source of truth for the state. A scheme that exists only in the director’s head does not survive contact with a film crew.

There is a reason branching keeps returning: it is not only a software pattern but a cultural technique. In my chapter “Branched Narrative: From Technical Device to the Art of Modeling the Future” I argue that multi-path storytelling participates in how culture imagines contingency, choice, and the image of the future — each fork is a small model of a possible tomorrow. AI tools will make producing more variants cheaper, which makes the structural discipline above more important, not less: the wider the field of branches, the more valuable the architecture that keeps it meaningful.