Abstract

Transhierarchy is a method of navigating nonlinear hypertext within a stable, hierarchically arranged, graphical user interface. The method preserves connectivity, provides context during navigation, and is visually compact, making it a viable method for navigating link-based hypertext on small-screen devices without removing context. This design preserves full connectivity, provides persistent navigational context, and remains visually compact—enabling fluid exploration of link-based hypertext without requiring additional semantics, multiple windows, panes or tabs within windows.

In a transhierarchy, a standard tree view is generated by traversing outbound links from a given node. Inbound links and their descendants are transcluded into the hierarchy, allowing continuous navigation to cross-referenced nodes.

This article makes three main contributions. First, we explore the history of navigating hypertext via hierarchical views situating transhierarchy within a broader design tradition of hierarchical views for hypertext navigation. Second, we present an implementation of transhierarchy in the first author's open-source software, em, and survey its use in several pieces of commercial outliner software today. Third, we discuss implications for representing search history in WWW browsers, suggesting that transhierarchy offers a principled alternative to linear history stacks.

1 Background

We identify transhierarchy as a user interface technique for use with graphical user interfaces (GUIs) for link-based hypertext. Strict hierarchies are often insufficient for directly representing the nonlinear structure of hypertext [19]. The disorientation problem (also known as the navigation problem or the lost-in-hypertext phenomenon) is defined by Conklin [2] as ‘the tendency to lose one's sense of location and direction in a nonlinear document’. Hierarchies have been criticised, appropriately, on epistemological grounds for constraining knowledge representations to categorical, typological relationships that discount or deny the fluid, associative nature of semantic connections. Nevertheless, hierarchies provide a predictable structure that suffers less from the disorientation problem of nonlinear text [2, 4, 19] and some genres of writing rely on hierarchy. Perhaps this is because users necessarily perceive lexias in a linear order (although possibly with repeats) and also as an entire constellation. Hierarchy provides ‘powerful information hiding capabilities’ [4, p. 206] through nested levels that can expand and collapse. The frequency of appearance of hierarchical structure across diverse interface elements—tables of contents, menus, chapters, heading levels—suggests the widespread use and generalisability of hierarchy, despite its limitations.

(Mathematical) graphs are a natural representation of link-based hypertext; edges are links and lexia are vertices. The hierarchical subtrees that are implicit in graphs are made explicit in the multitree data structures which are ‘directed acyclic graphs (DAGs) that have large, identifiable substructures that are trees. [...] In a multitree, the descendants of any node form a tree. As multitrees themselves are DAGs, and not trees, a node in the structure can have more than one parent’ [5, p. 86]. Multitrees provide a conceptual framework for the subtrees that can be traced in a graph, but do not specify a user interface. We describe such an interface in the Section §2 (Transhierarchy) below.

Transclusion [11] is the process of embedding content from a different location such that modification of the original will be visible in the embed. Nelson emphasises the importance of context when he describes transclusion as ‘reuse with original context available, through embedded shared instancing’ [12, p. 32]. Preserving context is a key motivation for the use of transclusion within hierarchy. Nelson does not reject hierarchy, but recommends including it within a broad array of structures as he writes ‘It would be hyperarchical, permitting the same material to be organized into simultaneous alternative structures—hierarchies, sequences, hyperplexes’ [12, p. 32]. He refers to all such structures as compound text.

1.1 Three prescient systems

An early example of compound text [11, pp. 1/15–1/16] — in which documents are composed of sequential lexias drawn from a database — is Symbolics' Document Examiner1 [20]. Document Examiner is a read-only link-based hypertext based on the Zmacs editor's link model but with substantial enhancements: the visual interface has multiple panes including a textual list of visited lexia (termed ‘bookmarks’), and the ability to show a graphical representation of the tree of which the current document is composed. The tree view presents lexias' identifiers horizontally (the tree branches to the right) while the list of visited lexias are presented vertically (from top to bottom). Lexias are included one of four ways: completely in-place, as cross-references (ancillary texts), by synopsis (a brief summary in place of the document), or implicitly with link text (what is termed a ‘link marker’ in the Dexter Model). It is not clear how the order of lexias to be included in place is determined; it seems that they are presented either by in-order traversal of an imposed hierarchical structure or by a topological sort of that imposed structure. Document Examiner has innovative features and its design demonstrates great forethought, but its creator notes the difficulties of allowing annotation or alterations to base lexia due to difficulty of transcluding different versions [20, p. 321].

One of the clearest examples of navigating graphs hierarchically is by Wang & Rada [22, p. 422] describing their MUCH system. They explain, ‘While the logical model of the MUCH system is a semantic net, a user sees at any time a view of this net as a tree. The tree (an indented list of node names, or an outline) can be folded2 and unfolded.’ The MUCH system partitioned links into hierarchical (‘organizational’) and associative (‘referential’) links.

In MUCH, a virtual tree is generated from the root DAG using a depth-first traversal algorithm. Each node is traversed and thus rendered only once. Due to the dynamic ordering of the traversal, deleting a link ‘results in a hierarchical view drastically different from the initial view’ [23, p. 377]. Corroborating concerns about the disorientation problem, Wang & Rada [22] summarise valuable feedback from users of MUCH, saying, ‘Users wanted “stable” documents. We take those comments to refer to navigational stability, viz. a way for users to avoid disorientation while “moving” from lexia to lexia. Other views upon the same document can be appreciated only after the stable view exists’ [23, p. 375]. Wang & Rada address this problem directly in the subsequent RICH system by allowing nodes to be traversed more than once, thus insulating non-adjacent parts of the hierarchy from local changes. This technique of duplicating nodes to maintain a stable tree structure is seen in modern outliners as well (described below). Furthermore, the duplication of nodes prompts an inquiry into transclusion, as the hypertext system designer is confronted with the problem of how to keep the nodes synchronised while also exposing additional context. Transclusion, context, and navigational stability were all important considerations in the formalisation of transhierarchy.

In the subsequent RICH system, hierarchical and associative structure are kept separate in the user interface: The hierarchical view consists of a standard indentation view with collapse/expand functionality. The associative view, or ‘spider view’, shows inbound and outbound links of the active node. The associative view is limited to adjacent links. These views are optionally kept in sync such that changing the active node in one view will change it in both. Content is ‘shared’ but not ‘embedded’, thus only partially fulfilling the criteria of transclusion. The limited amount of visible context could be a drawback of this design3. Also, this model is likely limited to desktop applications given its use of four separate windows: one for hierarchy, one for association, and two corresponding windows for displaying respective node contents. Despite these limitations, the dual views provide an effective method of navigating graphs hierarchically and deserves recognition as 1 of the first implementations of ‘hierarchical views of a graph’[23, p. 377].

1.2 Outline processors

One of the ways that strict hierarchies can be made explicit is by recursively indenting child nodes beneath their parent nodes (termed an indentation view). Outline processors, or outliners [6], are specialised text editors that primarily utilise such indentation views. While the indentation view is most often associated with strictly hierarchical data, it has also been considered as a potential alternative to traditional graphical layouts that visualise nodes and arcs. The indentation view ‘eliminates the need for explicit tree arcs [arrows], and it also provides a clearer distinction between the hierarchical and the arbitrary graph structure. However, the indentation view suggests a sequential ordering of the children of each node’ [5, p. 82]. The ordering problem led Feiner [3] to pursue nested set notation, which they implemented in their hypermedia system, Interactive Graphical Documents (IGD), veering away from the traditional outliner structure. Nested set notation, originally described by Knuth [10], utilises a graphical layout of nested boxes of nodes. Utting & Yankelovich [19, p. 73] explain that ‘systems with data organized in a hierarchical structure can easily mitigate the size problem by collapsing and expanding branches of the hierarchy as desired’, but note: ‘Intermedia links, however, form a network rather than a hierarchy. Our attempts to design a way to collapse and expand sections of our network were unsuccessful’. Others have also turned away from hierarchical information hiding, developing techniques for collapsing network structure directly, such as the node clustering of Schaffer [18]. The authors of the MUCH and RICH systems took the approach of adding additional semantic constraints to facilitate tree generation, while traditional outliners typically disallowed associative connections altogether.

As researchers of the 1980s and 1990s began pursuing other approaches to hypertext, the early vision of integrating hierarchy with nonlinear hypertext was not actively pursued. It appears that the problem of how to induce a hierarchy from general text without losing context did not seem worth the effort. We see that there are specific application domains where transhierarchy is highly desirable and feasible: For example, the visual display space limitations of old hardware has a modern analogy in mobile devices. It seems that some commercial outline processing software is using transclusion, although likely without recognising its antecedents. Promising developments in commercial outliners offer new solutions to this problem. A new class of hypertext systems, including the first author's system, em [15], have adapted outliners to work with nonlinear text using a technique we call ‘transhierarchy’.

2 Transhierarchy

Transhierarchy is a method of transclusion that allows the reader to go beyond hierarchical boundaries without losing the overall tree structure and its benefits. It offers a means of navigating nonlinear hypertext within a familiar and understandable view.

In this article, we focus on read-only navigation, and thus refer to the user as ‘reader’, omitting details of the authorial process. However, it should be noted that the model is easily implemented as a What You See Is What You Get (WYSIWYG) system that allows reading, navigating, and editing in the same view. Indeed, all of the outliners we studied allowed editing in this way. In such systems the reader and author are the same person, or, in a collaborative environment, are equally capable of reading and editing the shared content.

Nonlinear hypertext may be represented as a directed graph, which consists of any number of nodes and edges, where the edges each have a start and an end node. Given an edge that connects node to node (in that direction), we may say that node has an outbound link to node, and node has an inbound link from node, as depicted in Figure 1. Many hypertext models include typed links, typed nodes, multimedia, or other layers of complexity. The definition of transhierarchy used in this paper may be extended to those models, but is presented in its most general form, where it only requires a simple, directed graph of text-based nodes.

Figure 1: Two views of a tree

part (a): a tree graph using ovals for vertices and arrows for directed edges; only vertices A, B and C are highlighted. There is an edge leading into vertex A. Vertex A has outgoing edges to each of vertex B and vertex C. part (b): the label ‘A’ is at the left of the first line, there is an ‘open’ arrow to its left indicating that the indented material below it is subordinate. below label ‘A’ and indented are labels ‘B’ and ‘C’ each on their own lines and aligned with each other. there are ’closed’ arrows beside each of ‘B’ and ‘C’.

Figure 2: Inbound links to a given node

part (a): a visual presentation of the previous figure's part (b). There are two labels indented and below the label B: B1 and B2. The other labels appear in the same relative positions as before but their colours are visually less prominent. part (b): similar to (a) but with changes to the label B and what is below it. label B has a different style of ’open’ arrow and also a small superscript digit 2. below that label is the heading ’Contexts:’ indented. below that heading is the list of vertices that have outgoing edges to B: A and C1. each of those labels have ‘closed’ arrows.

In transhierarchy, a directed graph (Figure 1 a) is used to generate a collapsible, text-based tree view (Figure 1 b). Starting at a given node in the graph, each outbound link is rendered as a child in the tree. This process is repeated for each child to an arbitrary depth. Cycles (loops within the graph structure) are permitted as long as a maximum depth is defined, at which point the node will be collapsed and no additional descendants will be rendered. This algorithm works equally well on a Directed Acyclic Graph (DAG). The precise user interfaces to control expanding and collapsing are not germane here; they may provide manual control, or heuristically determine which nodes should be expanded or collapsed.

The primary problem to be addressed by transhierarchy is how inbound links are rendered within the hierarchical structure (see Figure 2). The indented structure of outliners does not allow more than one parent (inbound link) to be rendered in the usual manner unambiguously. Simple jump links to other parts of the hierarchy provide nonhierarchical navigation but do not qualify as transhierarchy as they do not utilise transclusion.

In em [15], the hypertext system developed by the first author, the reader may activate a transhierarchical view that renders all inbound links to a given node as child nodes (Figure 3).

Figure 3: Two transhierarchical views

part (a): the same graph image as before except for two changes: vertex C1 now has outgoing edges to new vertices X, Y and Z; also some vertices are highlighted. the highlighting is: vertices A and C1—which are the only vertices with outgoing edges to vertex B— are strongly highlighted; the vertices that are connected to vertex B are also highlighted though not as strongly. part (b): a larger visual presentation than in part (a); because vertex B2 has in-edges from both vertex A and vertex C1, the relevant part of the list shows the label B with two changes: (1) an ‘open’ arrow indicating that it is expanded and (2) a small superscripted digit 2. Immediately below the label is the heading ’Contexts’ and below that (and not further indented but one their own lines) is label ‘A’ with a ‘closed’ arrow and label ‘C1’. ‘C1’ is accompanied by an ‘open’ arrow and below it, and indented, is a list of the 3 vertices it has outgoing edges to.

Outbound links are not modified, but merely replaced visually. A special subheading ‘Contexts’ along with a hollow triangle bullet indicate to the reader that they are viewing inbound links rather than outbound links.

A descendant is any node that can be reached by following outbound links from a given node. The path that the user navigated before activating the transhierarchical view is preserved in the user interface. Figure 3 shows an example: in part (b), 's child nodes (,,) are rendered hierarchically beneath, but (along with its siblings and ancestors for additional context) is still visible. The inbound links that are displayed in the transhierarchical view utilise transclusion so that the reader can navigate further without losing the surrounding context. For example, in Figure 3 b, the reader is able to explore node and its descendants, while, its siblings and its ancestors continue to be visible.

Figure 4: In em: A transhierarchical view of a node titled ‘History Tree Interface’. The three inbound links (representing articles that mention ‘History Tree Interface’) are rendered as child nodes. Additional context is provided by rendering descendants of the inbound links via transclusion (‘Foss NoteCards extension’ and ‘ “The purpose of this display...” ’).

as described by the caption

Figure 5: Transhierarchy in Roam (using the same information depicted in Figure 4 ).

Transhierarchy in Roam (using the same information depicted in Figure 4). Each of the articles that references ‘History Tree Interface’ is listed under the ‘Linked References’ heading. The text ‘History Tree Interface’ is repeated under each article, and descendants are rendered beneath, via transclusion. The most notable differences are in visible cues and styling.

Caption references: 4

Figure 6: Transhierarchy in ThinkTool. This presentational style most closely matches that of Roam (Figure 5 ) although the order of references differs.

Each of the articles that references ‘History Tree Interface’ is listed under the ‘References’ heading. The text ‘History Tree Interface’ is repeated under each article, and descendants are rendered beneath, via transclusion. This presentational style most closely matches that of Roam (Figure 5) although the order of references differs. Most of the differences are in visible cues, styling and the order of the child nodes.

Caption references: 5

Figure 5 depicts an actual example, showing a sample of the authors' research notes for this article. Transhierarchy is activated on the ‘History Tree Interface’ node, showing all of the inbound links from nodes representing articles that mention that term. Transclusion under each of the three inbound links provides additional context resulting in a compact summary of the use of ‘History Tree Interface’ within the hypertext.

Transclusion allows readers to easily retrace their steps and provides visual cues that mitigate the disorientation problem. If the system supports WYSIWYG editing, then the transcluded descendants can also be edited (which would modify the originals). In some systems (e.g., Roam [16], ThinkTool [7]) deleting a transcluded node deletes only the original node, while in others (e.g., WorkFlowy [24]) all instances throughout the hypertext are deleted. The deletion behaviour as well as more complex implementations of transclusion, such as those that include versioning, filtering, or typed links, ultimately depends on the judgement of the system designer. Users' needs and preferences are areas for future investigation.

The reader's choice of activating the transhierarchical view is preserved historically in their path of navigation as the depth of the tree increases. Utting & Yankelovich [19, p. 72] express a special interest in ‘the coupling of the historical information, which [they] call the path, with the contextual information, which [they] call the map’. They cite Foss's [4] ‘History Tree’ NoteCards extension as a notable example of history-based navigational hierarchy. World Wide Web (WWW) users will be familiar with the difficulties arising from branches in browsing history (see Jones and Cockburn's study [9] for example) which the ‘History Tree’ was an attempt to address in NoteCards. Transhierarchy utilises a similar approach, though the path is less explicit. The reader's decision to open a transhierarchical view and follow an inbound link at a given node is effectively encoded into the tree's ancestors via transclusion. Because cross-references are pulled in transclusively without erasing the current context, the navigational path is preserved. Thus the chain of ancestors provides a compressed historical path of the reader's navigational steps. In this model, depth-based, vertical navigation is made visible in the hierarchy of expanded nodes, while lateral navigation is largely ephemeral. Transhierarchy differs in this regard from the ‘History Tree’ extension, which explicitly encodes each navigational step regardless of the structure of links that are followed.

3 Transhierarchy in the Wild

We review commercial outliner software that exhibits the use of transhierarchy. We highlight commonalities and possibilities for what could be a shared conception of the issues and methods, in the way that modern word-processing software implements the same functions in different ways using mostly the same vocabulary.

3.1 Roam

Roam [16] is an outliner that emphasises its handling of ‘networked thought’ through ‘bidirectional links’. Inbound links are rendered under two headings: ‘Linked References’, which are explicitly defined by the author (see Figure 5), and ‘Unlinked references’ [17], which are dynamically generated through a search query. Both types of references are rendered as fully transcluded tree views. Descendants are rendered up to an arbitrary depth, and editing any descendant in the embedded view modifies the original. The implementation of Linked and Unlinked references in Roam exemplifies transhierarchy by enabling navigation beyond hierarchical boundaries and using transclusion to preserve context.

3.2 ThinkTool

In ThinkTool's [7] implementation of transhierarchy, inbound links are rendered under the ‘References’ heading (Figure 6). This method allows for full transclusion of the parents and their descendants. Inbound links initially begin in a collapsed state to save visual display space, and can be further collapsed by clicking the ‘References’ heading itself (an excellent example of information hiding as described by Utting & Yankelovich [19]).

In ThinkTool, explicit references which cut across the hierarchy are differentiated from hierarchically nested nodes much like how RICH's ‘reference’ and ‘organizational’ links operate differently [22, 23]. Roam also differentiates associative references (‘pages’) from hierarchical parent-child relationships (‘blocks’) [17]. No such distinction is made in em [15], in which all connections between nodes are untyped, directed links. In an expandable outline view where nonlinear structures can be navigated hierarchically, the distinction may be unimportant but we cannot be confident without further study in situ.

3.3 WorkFlowy

WorkFlowy [24] supports shared instances of nodes with transclusion (which they term ‘mirroring’) [1] but does not have a way to view all inbound links. Obsidian [13] renders inbound links (‘backlinks’) but does not utilise transclusion to display additional context. As one user states, ‘It is very difficult to make out context from backlinks. I think the current implementation is similar to showing search results. I feel true power [sic] of backlinks only comes with Context’ [14]. This points to the need for both bidirectional linking and transclusion in nonlinear hypertext systems.

3.4 em

The primary difference between the design of em and that of Roam, ThinkTool, and WorkFlowy, is the aesthetic emphasis on a single, monistic graph structure. There is no differentiation between sentences, paragraphs, or pages in em. Each of em's lexia is a first-class node which can be linked and transcluded. In Roam, users must decide whether to add content as a stand-alone page or a nested descendant. Making nodes first-class affects subsequent navigation and editing. In em, all editing operations available in the hierarchical view are available in transclusion view. No distinction between associative references and hierarchical (parent-child) relationships is made in em. Hierarchy is formed by following directed links; associations are formed by using the transhierarchy to transclude the contents of an inbound link. The distinction is thus implicit and historical, based on the reader's navigational behaviour. This design choice comes at the cost of some stability, as the decision of which nodes are rendered as children cannot be determined without consulting the reader's navigational history. Although the usability [8] impact and learning curve of this compromise have not been thoroughly examined, initial user experience suggests that the stability of the hierarchy combined with reader-directed activation of transclusion may be enough to prevent disorientation. The benefit is an elegant, unified structure that can be freely navigated and edited in a compact space. The intention of em's design is to facilitate users' awareness of the semiotic nature of text, i.e., that text is interconnected, requiring interpretation through a web of connections. We intend to explore the relationship between structural and semiotic monism in future work.

4 Discussion

4.1 Transhierarchy by design

We speculate that the co-existence of associative and hierarchical structures in Roam, ThinkTool, and other outliners may be an artefact of adapting outliners to work with nonlinear hypertext. In contrast to those earlier approaches, the unification of associative and hierarchical structures, for example as implemented in em, provides transclusion of any linked node, whereas in other systems transclusion may be limited to nodes connected by certain link types.

There may be multiple reasons for the use of transhierarchy. Broadly, these are:

for the creation of documents for other people to read (e.g., Concord [21] and Document Examiner [20]);\

to help users make sense of thoughts or text by exposing a simpler structure, and to avoid confusion by limiting context (e.g., MUCH [22] and RICH [23]); and\

to simplify complex information tasks by exposing a simple organising sub-structure and limiting context (e.g., outline processors).\

We suggest that transhierarchy could also be helpful as a way to visualise and traverse one's browsing history in WWW browser software. The stack-based model of browsing history — i.e., the list of webpages the user loaded during the current session — that browser software typically presents is a linear sequence. This presentation loses valuable information; when a user revisits a webpage, part of the history is removed from the display [9]. Although not as complete a solution as possible in a closed system such as Intermedia [19], a transhierarchical presentation in a WWW browser could preserve the multiple contexts of loadings of particular webpages.

4.2 Limitations

While transhierarchy is highly appropriate for some common applications it has limited applicability for other uses. The four limitations to universal applicability that we perceive are:

First, all of the systems we studied present a tree view, yet violate the one-parent-per-child rule that is definitional of trees. This mismatch between the visual presentation and the underlying data structure may have implications for usability. The systems we studied mitigate this by placing inbound links under a clearly distinguished heading (‘References’ in Roam and ThinkTool; ‘Contexts’ in em). Roam and ThinkTool distinguish inbound links further through layout, colour, and other stylistic elements, at the expense of visual simplicity. Further study is needed to assess how much this violation reduces usability (if at all), and if it is only problematic during the initial learning of the tool.

Second, although the indentation view can functionally represent many types of hypertext, it is clearly more suited to some types than others. For example, it is less ideal for long-form narrative structure, where meaningful relationships are embedded within the prose and cross-references are primarily established through inline links. The collapse-expand mechanism is often less appropriate for long blocks of (linear) prose.

Third, transhierarchy shows outbound links by default, which favours (or gives preference to) navigation in that direction. A hypertext in which the reader follows more inbound links than outbound links would be more difficult to navigate with transhierarchy, since the reader would have to manually switch to view inbound links at each node.

Lastly, the simple tree structure described here does not provide a mechanism for typed links, which may be a requirement for navigating semantic networks and ontologies.

4.3 Future work

This article surveys the history of navigating hypertext via hierarchical views and contemporary implementations of transhierarchy in outliner software. As further research we plan to conduct a naturalistic field study to better understand how transhierarchy is being used and what features its users think would be useful.

5 Summary

Navigating nonlinear hypertext without losing one's sense of location or the ability to navigate effectively [2] remains a persistent challenge in hypertext system design, especially when conducting cognitively intensive tasks. Transhierarchy uses transclusion to seamlessly insert a subtree within an existing hierarchy to improve navigation without losing context thus aiding clarity of thought and expression for users.

Our contributions are threefold:

We have formalised transhierarchy as a distinct user interface technique, situating it within the historical trajectory from Document Examiner, MUCH and RICH to contemporary systems.\

We have surveyed its implementation across four modern outline processors—Roam, ThinkTool, WorkFlowy, and em—identifying common patterns and design variations.\

We have demonstrated that transhierarchy offers a viable approach to hypertext navigation on small-screen devices, where visual display space is at a premium.\

There are however unanswered questions about potential compromises between usability, particularly learnability, and expressive power. Transhierarchy is not without limitations: it favours outbound navigation, may be challenging to learn due to its apparent violation of tree semantics, and does not natively support typed links. Nevertheless, user-directed transclusion of inbound links provides a stable navigational structure that helps mitigate disorientation whilst maintaining the information-hiding benefits of hierarchical views.

The convergence of transhierarchy techniques across independently developed commercial outliners suggests that this approach addresses genuine user needs. Beyond personal knowledge management, transhierarchy offers a promising foundation for rethinking how WWW browsers present browsing history—moving from linear stacks to contextually rich, navigable trees. As hypertext systems continue to evolve, the bridge between hierarchical and associative structures that transhierarchy provides deserves sustained attention from both researchers and practitioners.

ORGANISATION

In §1, Background, we review the existing literature related to the hierarchical navigation of link-based hypertext, including methods of combining hierarchy with associative links, information hiding, indentation view, transclusion, and the disorientation problem of nonlinear text.

In §2, Transhierarchy, we describe the method of transhierarchy in detail, demonstrating its use of hierarchy, associativity, and transclusion as exemplified in the first author's open-source hypertext system, em [15].

In §3 Transhierarchy in the Wild, we survey the use of transhierarchy historically (in Document Examiner [20], MUCH [22], and RICH [23]) and in three commercial outliner apps (Roam [16], ThinkTool [7], and WorkFlowy [24]), and contrast them with em.

In §4 Discussion, we describe potential limitations of transhierarchy and contexts which may affect its applicability. We conclude with a summary of the key points presented in this paper and pointers to avenues for future investigation to draw greater attention to this rich yet understudied vein of navigation.

Notes

1The corresponding editor software for creating hypertexts is Concord [21].

2Note that what Wang & Rada [22] refer to as ‘folded’ we term ‘collapsed’.

3The trade-offs for users of having too little context and the potential for being overwhelmed with detail are areas for future study. The appropriate balance will no doubt depend on individual users and their needs in relation to specific uses.

Source

ACM full-text HTML

ACM Digital Library

Imported from ACM’s structured HTML source. ACM Reference Format: Raine Revere and James Blustein. 2026. Transhierarchy redux: New opportunities found in outline processors. In 37th ACM Conference on Hypertext (HT '26), September 14--18, 2026, London, United Kingdom. ACM, New York, NY, USA 7 Pages. https://doi.org/10.1145/3800935.3830840

References

[1] Workflowy Community. n.d.. Virtual Instances of Bullet point. Retrieved Sept. 04, 2020 from https://workflowy.zendesk.com/hc/en-us/community/posts/360001389746/comments/360011366531

[2] Jeff Conklin. 1987. Hypertext: An Introduction and Survey. IEEE Computer 20, 9 (1987), 17–41. https://doi.org/10.1109/MC.1987.1663693

[3] S. Feiner. 1988. Seeing the forest for the trees: hierarchical displays of hypertext structures. In Proceedings of the ACM SIGOIS and IEEECS TC-OA 1988 conference on Office information systems(COCS’88). Association for Computing Machinery, New York, NY, USA, 205–212. https://doi.org/10.1145/45410.45432

[4] Carolyn L. Foss. 1988. Effective browsing in hypertext systems. In User-Oriented Content-Based Text and Image Handling(RIAO’88). Le Centre De Hautes Etudes Internationales d'Informatique Documentaire, Paris, France, 82–98. https://doi.org/10.5555/3374430.3374437

[5] Peter Gloor. 1997. Elements of Hypermedia Design: Techniques for Navigation & Visualization in Cyberspace. Birkhäuser Boston, Germany. https://doi.org/10.1007/978-1-4612-4144-7

[6] Michael Heim. 1999. Electric language: A philosophical study of word processing. Yale University Press, UK.

[7] Jonas Hvid. n.d.. ThinkTool: A tool for taking associative notes. Retrieved Sept. 04, 2020 from https://thinktool.io/

[8] International Organization for Standardization. 2018. Ergonomics of human-system interaction — Part 11: Usability: Definitions and concepts (2 ed.). Standard. ISO, Geneva, Switzerland. ISO/TC 159/SC 4.

[9] Steve Jones and Andy Cockburn. 1996. A study of navigational support provided by two World Wide Web browsing applications. In Proceedings of the seventh ACM conference on Hypertext(HT’96). Association for Computing Machinery, New York, NY, USA, 161–169. https://doi.org/10.1145/234828.234844

[10] Donald E. Knuth. 1973. The Art of Computer Programming (2nd ed.). Vol. 1. Addison Wesley, Reading, Massachusetts, USA. 309–310 pages.

[11] Theodore Holm Nelson. 1982. Literary Machines. Mindful Press, Sausalito, California, USA.

[12] Theodore Holm Nelson. 1995. The heart of connection: hypermedia unified by transclusion. Commun. ACM 38, 8 (Aug. 1995), 31–33. https://doi.org/10.1145/208344.208353

[13] Obsidian n.d.. Obsidian: A second brain for you, forever. Obsidian. Retrieved Sept. 04, 2020 from https://obsidian.md/

[14] Obsidian Forum. 2020. Fully transclude backlinks. Post #3, Retrieved 2020-09-04. https://forum.obsidian.md/t/fully-transclude-backlinks/325/3

[15] Raine Revere. n.d.. em: The personal sensemaking app. Retrieved Sept. 04, 2020 from https://emtheapp.com/

[16] Roam Research. n.d.. Roam: A note-taking tool for networked thought. Retrieved Sept. 22, 2020 from https://roamresearch.com/

[17] Roam Tips. 2020. What is an unlinked reference in Roam?https://www.roamtips.com/home/roam-unlinked-references. Retrieved 2020-09-04.

[18] Doug Schaffer, Zhengping Zuo, Saul Greenberg, Lyn Bartram, John Dill, Shelli Dubs, and Mark Roseman. 1996. Navigating hierarchically clustered networks through fisheye and full-zoom methods. ACM Transactions on Computer-Human Interaction 3, 2 (June 1996), 162–188. https://doi.org/10.1145/230562.230577

[19] Kenneth Utting and Nicole Yankelovich. 1989. Context and orientation in hypermedia networks. ACM Transactions on Information Systems 7, 1 (Jan. 1989), 58–84. https://doi.org/10.1145/64789.64992

[20] Janet H. Walker. 1987. Document Examiner: delivery interface for hypertext documents. In Proceedings of the ACM Conference on Hypertext (Chapel Hill, NC, USA) (HT’87). Association for Computing Machinery, New York, NY, USA, 307–323. https://doi.org/10.1145/317426.317448

[21] Janet H. Walker. 1988. Supporting document development with Concordia. In Proceedings of the Twenty-First Annual Hawaii International Conference on System Sciences, Vol. 2. IEEE Computer Society, Los Alamitos, California, USA, 355–364. https://doi.org/10.1109/HICSS.1988.11825

[22] W. Wang and R. Rada. 1995. Experiences with semantic net based hypermedia. International Journal of Human Computer Studies 43, 3 (1995), 419–440. https://doi.org/10.1006/ijhc.1995.1052

[23] Weigang Wang and Roy Rada. 1998. Structured hypertext with domain semantics. ACM Transactions on Information Systems 16, 4 (Oct. 1998), 372–412. https://doi.org/10.1145/291128.291132

[24] Workflowy. n.d.. WorkFlowy: A simpler way to organize your mind. Retrieved Sept. 22, 2020 from https://workflowy.com/