Associative navigation offers lateral jumps across levels

What it is

Tasks often move sideways: from this sofa to a matching table, from this policy to the form it cites. On a classification tree those objects may live on different branches, so a hierarchical walk has to climb to a common ancestor and descend again. Associative navigation puts an entry to a related object on the current page, so people can jump laterally across levels without making that round trip.

It is a navigation move, not a conversion of the whole site into hypertext. Both ends may still have parents; the link only handles “from here straight to there.”

Why it happens

Hierarchical edges mean “belongs to”; associative edges mean “used together.” The next step is often driven by the second relation: compatible parts, mutual citations, another piece by the same author, a status view of the object just inspected. With only hierarchical edges, the shortest legal path for a cross-branch task is up and down again. Nodes in the middle have nothing to do with the current goal, scent breaks on the way, and people get lost on the ancestor page.

A lateral jump draws the task-co-occurring edge directly. Scent comes from the relation type itself—“goes with,” “cited by,” “next chapter”—which sits closer to the job than a parent class name. The jump works when both ends are already related in the user’s model. Chaining unrelated objects turns the lateral entry into noise that is learned away after the first miss.

Studying it

Compare cross-branch tasks under “tree only” versus “related links allowed.”

• Paradigms: add a “related” edge on top of a tree test and measure path length for cross-class tasks; in logs, rate of clicks on related / compatible / see-also from a detail page and subsequent success. In Pirolli’s information-foraging frame, associative links are scent channels between patches.
• Independent variables: presence of a lateral entry, whether the relation type is named (compatible versus a vague “more”), distance between the ends on the tree.
• Dependent variables: clicks on cross-branch tasks, whether the path detours via an ancestor, task completion after following an associative link.
• Methodological note: tree tests by default have no lateral edges and will overstate the cost of hierarchy as the only path. Design cross-branch tasks separately; do not evaluate associative navigation with “find an item under this category” tasks that already follow the tree.

Where it stops holding

In a single linear flow (checkout, account opening) the next step is prescribed; a lateral jump pulls people out of the flow. Strongly permissioned objects must not become reachable through “related” when the person has no right to the node. Algorithmic “you may also like” is a personalized patch, not necessarily an architectural associative edge; stability and explanation differ. Where objects have no real co-occurrence, forcing associative navigation only adds lostness.

Applying it

• Inventory relation types for cross-branch tasks first: compatible, cites, sequential next, another view of the same entity. Build entries only for those types, not a generic “related content” pool.
• Place the entry beside the current object and name the relation; do not offer only a row of thumbnails.
• The page landed on by a lateral jump must still connect back to the primary structure (parent class and global nav remain). Otherwise the jump is a one-way teleport.
• Verify with five real tasks that cannot be finished inside one class, search forbidden. Log detour depth without associative entries; if people still climb to an ancestor after the entries are added, scent is weak or the wrong objects are chained.

Related

• Within the group: G2.03.2 Network structure lacks a sense of place · G2.03.3 Associative entries must sit beside the primary structure, not replace it
• Adjacent: G1.09 Types of organization systems · G1.05 Metadata · G1.01 Definition and scope of information architecture
• Search terms: associative navigation · lateral jump · related links