In Cybertext, Espen Aarseth defines “ergodic literature” as that which requires “nontrivial effort…to allow the reader to traverse the text” (1997: 1). Over the course of the study, he explores hypertext novels, adventure games, generated poetry, and multi-user dungeons (MUDs) as examples of ergodic literature. What has always struck me as unusual is that he all but ignores the codex despite acknowledging that many cybertexts “are indeed books” (1997: 8). Put differently, why does Aarseth ignore gamebooks?

Gamebooks are more commonly known as choose your own adventure (CYOA) books, so named from the series published by Bantam Books from 1979-1998. Over those two decades, 184 CYOA books appeared, each offering—per the books’ covers—“you” the chance to be “the star of the story.” While addressing the reader in the second person set the books apart, the true hallmark of the series was their forcing the reader to choose what “they” would do next, effectuated by turning to a particular page.

Edward Packard developed the concept of CYOA in the 1960s while telling bedtime stories to his two daughters (see Kraft 1981; Edward Packard, ‘Edward Packard, Interviewed by Brian Croxall’, 9 Sept. 2025, Phone). He—and most of the publishing world—was unacquainted with Doris Webster and Mary Alden Hopkins’s 1930 novel Consider the Consequences, which is the first known CYOA-style gamebook (see KZSC Santa Cruz 2018). This novel was republished in 2025 by Pushkin Press.

These choices then led to subsequent forking paths, with the some volumes having more than 40 different endings for the reader to explore. The stories, targeted at children between the ages of 10-12, proved incredibly popular despite (or, more likely, due to) their unconventionality: it is the sixth best-selling book series of all time (‘List of best-selling books’ 2025).

Despite their popularity, the CYOA form of ergodic literature has been mostly dismissed in academic accounts of interactive fiction (see Niesz and Holland 1984; Randall 1988; Aarseth 1997; Montfort 2005).

CYOA receive more attention in other venues. The books were treated as objects of interest by the popular press at the time of their arrival (see Harmetz 1981; Kraft 1981; Scaffetti 1986). There was also clear early interest in CYOA gamebooks on the part of K12 educators (see Newman 1988; Sargent 1991; Angileri 1994). Some academic treatment of CYOA has emerged in recent years (Mawhorter et al. 2014; Ashwell 2015; Cova and Garcia 2015; Cook 2021). And CYOA continues to be a popular topic in mainstream/Internet journalism (Jamieson 2015; Laskow 2017a, 2017b; Schwartz 2020; Jamison 2022; Shepherd 2024).

One reason for the difficult reception to CYOA in academia might be the challenge of producing a “reading” of a multicursal text, where one cannot be certain everything has been encountered. Again, Aarseth: “it is possible to explore, get lost, and discover secret paths in these [ergodic] texts, not metaphorically, but through the topological structures of the textual machinery” (1997: 4). An obvious solution is one familiar to the digital humanities: visualization of the stories’ network structure.

When Chooseco LLC began republishing many of the original CYOA novels in 2006, many of them included network visualizations on their rear covers (Laskow 2017b).

In 2009, Mark Sample catalogued as many “CYOA maps” as he could find, most of them being for single CYOA books.

Christian Swinehart’s “One Book, Many Readings” is an exception as it examines 12 CYOA-like books.

Since these are small explorations, the authors, many of whom were students, tend to foreground conclusions about the particular book they plumb. While these maps partake of the DH drive to visualization, they miss another DH trend: working at scale. The clear departure from this trend is Jeremy Douglass, whose 2017 Transverse Reading Gallery contains network graphs and data for 834 gamebooks (including the CYOA series) from the Demian Katz collection at UC Santa Barbara. Completed in partnership with students, Douglass’s work is impressive, but it also currently lacks conclusions about what these data reveal.

As per Moretti, “Not, of course, that the map is already an explanation; but at least it shows us that there is something that needs to be explained” (2004: 84).

In this project, I try to strike a balance between Douglass’s scale and the one-off interpretations that appear in Sample’s bibliography. Over the last year, I have conducted both archival research in the Katz collection and oral interviews with Edward Packard, the author of the first and the last CYOA books Bantam published (as well as fifty-four others). The goal of my research is to extract what Mawhorter et al. define as each narrative’s “choice structure” (2014). In brief, I extract the various exits from each page into a structured data format that is then tractable to exploration through Python’s NetworkX package (Hagberg, Schult, and Swart 2008) and Gephi (Bastian, Heymann, and Jacomy 2009). This method allows me to compare the topology of individual CYOA stories and to explore large trends within the corpus (see Figure 1).

For example, a tactic employed by Bantam to market the books was to emblazon the cover with the number of endings in each volume. Such an approach makes sense given the fact that multiple endings were one of the ways the series differentiated itself. Over time, however, my data show that the number of endings per book decreases sharply, suggesting that it was ultimately not the ending count that mattered nearly as much as the unconventional narrative structure. Another decision made by marketers was to ignore the fact that the books’ authors steadfastly refused to gender the main character of each adventure. Instead of leaning into the non-gendered you of a second-person narration, the press overwhelmingly chose to illustrate the main character of the books as a white male. I discuss these patterns and many more: how many paths are there through these books? which page is most likely to have an ending? how long does it take to get to a “bad” ending or a “good” one?