# Hypermedia and the Line > [!abstract] What this note is > An argument: **hypermedia is not better than linear media for narrative or sustained argument.** It won completely as infrastructure and lost as a form. Sources checked 19 September 2026; where the research is conditional or contested, this note says so. See also [[What is Hypermedia]] for the software sense of the word. ## The claim that was made The dream is older than the web. In 1945 Vannevar Bush described the **memex**, a desk of microfilm with "associative indexing," so that any item could summon another and a reader could build named **trails** through a library [1]. Around 1963 Ted Nelson coined **hypertext** and **hypermedia**, publishing the idea in 1965, and spent decades on Project Xanadu: two-way links, transclusion, versions, provenance [2]. By the early nineties the idea had acquired a theory. George Landow argued that hypertext was "a strikingly literal embodiment" of what Barthes and Derrida had described: the text without a centre, the reader promoted to co-author [3]. Jay David Bolter argued that electronic writing replaced print's hierarchy with multiple equal orders under the reader's control [5]. And in June 1992 the novelist Robert Coover announced in the *New York Times Book Review* that the novel as we knew it had ended, killed by "the tyranny of the line" [4]. That phrase is the whole argument in four words. Sequence was cast as coercion. Choice was cast as freedom. ## What reading research found instead The trouble arrived early, from people building the systems. In 1987 Jeff Conklin named **disorientation**, "the tendency to lose one's sense of location and direction in a nonlinear document," and **cognitive overhead**, the mental cost of deciding which link to take [6]. By 1989 the failure had a nickname, "lost in hyperspace" [7]. Two decades of experiments followed. DeStefano and LeFevre's 2007 review is the careful summary: hypertext structure raises the cognitive demands of decision-making and visual processing, and that added load impairs reading performance [8]. **The effect is conditional, not universal.** It falls hardest on readers with lower working memory and less prior knowledge, and some structures help some readers. The honest sentence is that linking a text taxes the reader in a way a line does not, and that tax is regressive: it costs most from those who know least. Two more findings sharpen it. Salmerón and colleagues found that the **order** in which a reader traverses linked sections predicts inferential comprehension, with a semantically coherent path helping most for low-knowledge readers [9]. So the sequence an author chooses is not a cage; it is information. And eye-tracking work has found that when words are links, readers fully process **the linked words** while unlinked text gets less attention, and rate unlinked sentences as less important [10]. A link does not add a possibility to a sentence. It reweights the sentence. > [!warning] The mistake this argument must not make > Screen-versus-paper research is **not** hypertext research. Delgado and colleagues found a real screen disadvantage in 2018 (Hedges' g = −0.21 across 54 studies and 171,055 participants), Clinton confirmed it in 2019, and both found the penalty under time pressure and for **informational** text — with **no significant difference for narrative** [11][12]. Those studies compare media and conditions, not linked versus linear structure. Borrowing them as evidence against hypertext is the single easiest way to overstate this case, and the narrative result cuts the other way besides. And the other side is real. Chen and Rada's 1996 meta-analysis of 23 experiments found hypertext users' performance tended to be more **effective**, while efficiency consistently favoured non-hypertext readers [13]. Hypertext buys accuracy and spends time. Prior knowledge moderates the whole question, and not always in the same direction. The claim here is not that hypertext is bad. It is that it is not **better**, and specifically not better at the thing its champions claimed for it. ## What happened to the form Literary hypertext was made, seriously and well. Michael Joyce's *afternoon, a story* was shown in 1987 and published on diskette in 1990; Stuart Moulthrop's *Victory Garden* followed in 1991, Shelley Jackson's *Patchwork Girl* in 1995. They are taught, anthologised, and studied. They were never read at any scale outside the field, and the medium they were made for stopped being a medium. A 2024 assessment puts it plainly: Storyspace hypertext "fell out of popularity relatively quickly, and as a literary medium was dead by the turn of the century" [14]. The most telling witness is Coover himself. In a 1999 keynote he observed that "in terms of new serious literature, the Web has not been very hospitable," that the discrete crafted work had been replaced by "this vast disorderly sprawl," and that hypertext had become a way to add enhancements to "more or less linear narratives" [15]. It was not a recantation; he still bet on digital literacy. It was a man who had announced the end of the line, seven years later, describing the line's survival. Espen Aarseth had already taken the theory apart from inside. His 1997 *Cybertext* proposed **ergodic literature**, text requiring non-trivial effort to traverse, a category that includes the I Ching and MUDs as well as hypertext, cutting across the digital divide that the theorists had treated as decisive. Reading *afternoon*, he found a classic work of **modernist** literature [16]. Not a rupture. A continuation, in a new container. ## Where hypermedia actually won Completely, and somewhere else. Tim Berners-Lee's 1989 CERN memo proposes "a distributed hypertext system" [17], and that system now carries nearly everything. But look at what it carries: articles, posts, documentation, video, feeds. Overwhelmingly **linear** works, delivered by hypertext plumbing. On Wikipedia, the most densely linked reading surface ever built, one study found only about 4% of links are clicked more than ten times a month [18]. The links are the road, not the journey. Nelson's own verdict on what won is worth keeping: "HTML is precisely what we were trying to PREVENT" — links that break, point one way, and cannot be followed back to their origin [2]. Meanwhile the idea kept its engineering meaning and thrived. Roy Fielding's 2000 dissertation made **hypermedia as the engine of application state** a constraint of REST [19]; the htmx community revived it for applications, with *Hypermedia Systems* in 2023 [20], and the ACM Hypertext conference formalised "hypermedia controls" in 2024 [21]. That is [[What is Hypermedia]] in this vault: hypermedia as architecture, where it is genuinely superior to the alternatives. **So the split is the finding.** As a way to build systems, hypermedia won on the merits. As a way to tell a story or make an argument, it lost to the line, and the line never actually noticed the fight. ## Why the line holds Beyond the experiments, three plain reasons. These are argument, not measurement. **Sequence carries meaning.** A narrative is causes in order. Rearranged, the causes become coincidences. An author's ordering is not a restriction placed on the reader; it is most of the work. **Every branch pays for itself.** In branching text, each section must re-establish who is present and what is known, because the writer cannot know what came before. The result is either repetition or incoherence, which is why nonlinear fiction is so hard to write well and so rarely worth the cost. **Agency is not meaning.** Choosing where to go next is a real pleasure, and games earn it honestly. But being the reason something happened is different from being moved by what happened. Hypertext theory treated the second as a lesser version of the first. Readers disagreed, by leaving. ## What the dream got right Reference works. Documentation. Encyclopedias. Legal and scientific literature, where citation is the point. Wikis. Anything an expert enters with a question rather than an appetite for a beginning. In that territory the associative trail beats the line, exactly as Bush described, and this vault is built on the assumption. Bush was right about libraries. Landow and Coover were wrong about novels. Both facts belong in the same sentence. ## The current question The nonlinear reading of 2026 is not a Storyspace fiction; it is a chat window, and a search that ends without a visit. Industry analysis puts the majority of US searches ending with no click through to a website [22], though that is traffic data, not peer-reviewed work on comprehension, and no study yet measures how well anyone understands what a chat interface tells them. The field has noticed. ACM Hypertext ran 2025 and 2026 under the theme "The World as Hypertext," and recent work argues for reintroducing **hypertextual friction** — provenance, associative thinking, the deliberate cost of a decision — against interfaces optimised for engagement and efficiency [23]. Which turns the old argument over. Hypertext's critics said the reader was made to work too hard. The new complaint is that the reader is not being asked to work at all. --- **A note on where this came from.** Jeff started this as a multidisciplinary critical writing paper and did not finish the class. The conclusion he reached then was that he could only posit that hypermedia is not better than linear narrative. That is not a failure to land a thesis. Given the record, it is the thesis. ## Sources 1. [Vannevar Bush, "As We May Think," *The Atlantic*, July 1945](https://www.ee.columbia.edu/~dpwe/papers/Bush45-aswemaythink.pdf) 2. [Ted Nelson, "A File Structure for the Complex, the Changing and the Indeterminate," *Proc. ACM 20th National Conference*, 1965](https://dl.acm.org/doi/10.1145/800197.806036) · [Nelson on HTML](https://xanadu.com.au/ted/TN/WRITINGS/TCOMPARADIGM/tedCompOneLiners.html) 3. George P. Landow, *Hypertext: The Convergence of Contemporary Critical Theory and Technology*, Johns Hopkins, 1992 (later *Hypertext 2.0*, 1997; *3.0*, 2006) 4. Robert Coover, "The End of Books," *New York Times Book Review*, 21 June 1992 5. Jay David Bolter, *Writing Space*, Lawrence Erlbaum, 1991 · Janet Murray, *Hamlet on the Holodeck*, Free Press, 1997 (Murray argued for **authored** interactive narrative and was already sceptical of undirected link-hopping) 6. [Jeff Conklin, "Hypertext: An Introduction and Survey," *IEEE Computer* 20(9), Sept 1987](https://dl.acm.org/doi/10.1109/MC.1987.1663693) 7. D. M. Edwards & L. Hardman, "'Lost in hyperspace': cognitive mapping and navigation in a hypertext environment," in McAleese (ed.), *Hypertext: Theory into Practice*, 1989 — the standard citation for the phrase 8. [DeStefano & LeFevre, "Cognitive load in hypertext reading: A review," *Computers in Human Behavior* 23(3), 2007](https://www.sciencedirect.com/science/article/abs/pii/S0747563205000658) — a narrative review, not a pooled meta-analysis 9. Salmerón, Cañas, Kintsch & Fajardo, *Discourse Processes*, 2005; Salmerón & García, 2012 10. [Fitzsimmons, Weal & Drieghe, eye-tracking studies of hyperlink reading, *PLOS ONE*, 2019–2022](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0263669) 11. [Delgado, Vargas, Ackerman & Salmerón, "Don't throw away your printed books," *Educational Research Review* 25, 2018](https://www.uv.es/lasalgon/papers/Delgado%202018%20dont%20throw%20away%20your%20printed%20books.pdf) 12. [Clinton, *Journal of Research in Reading*, 2019](https://onlinelibrary.wiley.com/doi/abs/10.1111/1467-9817.12269) 13. [Chen & Rada, "Interacting with Hypertext: A Meta-Analysis of Experimental Studies," *Human–Computer Interaction* 11(2), 1996](https://www.tandfonline.com/doi/abs/10.1207/s15327051hci1102_2) 14. [Calvin Olsen, "Hypertext and its afterlives," *New Review of Hypermedia and Multimedia* 30(1–2), 2024](https://www.tandfonline.com/doi/full/10.1080/13614568.2024.2374286) 15. [Robert Coover, "Literary Hypertext: The Passing of the Golden Age," keynote, Digital Arts and Culture, Atlanta, 29 October 1999](https://nickm.com/edited/vox/golden_age.html) 16. [Espen Aarseth, *Cybertext: Perspectives on Ergodic Literature*, Johns Hopkins, 1997](https://williamwolff.org/wp-content/uploads/2013/01/aarseth-ergodic-ch1-1997.pdf) 17. [Tim Berners-Lee, "Information Management: A Proposal," CERN, March 1989](https://www.w3.org/History/1989/proposal.html) 18. [Dimitrov, Singer, Lemmerich & Strohmaier, "What Makes a Link Successful on Wikipedia?" WWW 2017](https://dl.acm.org/doi/10.1145/3038912.3052613) 19. [Roy Fielding, *Architectural Styles and the Design of Network-based Software Architectures*, UC Irvine, 2000](https://ics.uci.edu/~fielding/pubs/dissertation/fielding_dissertation.pdf) 20. Carson Gross, Adam Stepinski & Deniz Akşimşek, *Hypermedia Systems*, 2023 — [hypermedia.systems](https://hypermedia.systems/) 21. [Gross, Shaffer & Revelle, "Hypermedia Controls: Feral to Formal," ACM Hypertext '24](https://dl.acm.org/doi/10.1145/3648188.3675127) 22. Industry traffic analysis, 2025–26, on searches ending without a click — *trade analysis, not peer-reviewed* 23. [Liu & Almeda, "Agency Among Agents: Designing with Hypertextual Friction in the Algorithmic Web," Adjunct Proc. ACM HT '25](https://arxiv.org/abs/2507.23585) · [ACM Hypertext 2026, "The World as Hypertext"](https://ht.acm.org/ht2026/) ## Related - [[What is Hypermedia]] — the software sense: links and forms as the engine of application state - [[AI Reading List]] — [[Nexus]] on information networks, and what they are for - [[When You Need a Human]] — another note about a tool's limits