Skip to main content
biology business computer-science geography library-and-museum-studies psychology visual-arts

​
Proximity

​
Description

Proximity names the structural claim that nearness on some dimension is itself a grouping signal. Items close together get treated as related — perceived as a unit, processed together, credited or blamed together, routed together. The claim is interesting because the items need not have any other relationship: proximity alone is enough to induce grouping, and removing the proximity weakens the group even if every other property is preserved. The dimension along which “near” is measured varies by domain. In Wertheimer’s original gestalt formulation it was visual space. In Constantine and Yourdon’s structured-design treatment of cohesion, it was textual position in source code. In Allen’s organizational research, it was physical office distance between engineers. In Tobler’s first law of geography, it was geographic distance. In information retrieval, it is embedding-space cosine distance. The concept generalizes the moment one such dimension is established: along that dimension, nearness implies belonging, and observers default to treating proximate items as a unit until told otherwise. The diagnostic question — “are these items related, or are they merely near each other on this dimension?” — names the trap. Proximity-as-grouping is a useful default but a fallible inference. When designers want grouping, they place items near each other and the inference produces the right interpretation. When proximity is accidental (the bug fix happens to commit alongside unrelated changes; the politician happens to be photographed with the corrupt donor; the citation happens to neighbor an unrelated paper), the grouping inference becomes a misattribution.

​
Triggers

User-initiated: User notices items being grouped, or wants to design for grouping, or worries about an unwanted grouping inference. Vocabulary cues: “they’re near each other,” “co-located,” “next to,” “adjacent,” “grouped together,” “the same neighborhood,” “guilt by association,” “context-by-position.” Agent-initiated: Agent notices the user is reasoning about a group of items as if the group were real, when the actual signal binding them is mere proximity on some dimension. Candidate inference: “your group is being held together by proximity alone; remove the proximity and ask whether anything else holds them together.” Situation-shape signals: Discussions of layout, organizational design, code structure, retrieval, attribution, blame, or credit. Reasoning about communities, schools, or movements where membership is established by association rather than by explicit declaration. UI / UX design choices about spacing. RAG and retrieval-system tuning. Statistical analyses where spatial autocorrelation needs to be accounted for.

​
Exclusions

  • No relevant proximity dimension exists — if there’s no natural distance metric on the items in question (e.g., enumerated abstract categories with no embedding), proximity-as-grouping has no purchase. Imposing a fake metric to invoke the concept retro-fits the situation rather than describing it.
  • Items are genuinely related via explicit non-proximity relationship — when two functions are related because A calls B, the relatedness lives in the call edge, not in their textual nearness. Calling that relatedness “proximity” misclassifies it; the call-graph relation is doing the structural work.
  • Distances are uniform — if all pairs of items are equidistant on the relevant dimension, proximity carries no grouping information. The concept needs variance in nearness for the grouping signal to be informative.
  • The grouping has been explicitly declared — when a set is defined by membership (“the engineers on Team Alpha”), proximity is at most a downstream consequence, not the constitutive signal. Treating an explicitly-declared group as proximity-based confuses the basis of the grouping.

​
Structure

Internal structure of proximity: a table of its component slots and the concepts that fill them.

​
Relationships

Relationship neighborhood of proximity: a graph of the concepts it connects to and the concepts it is a part of.
  • similarity — paired gestalt grouping principle. Items group by either or both; designers can choose which axis to deploy, and observers should ask which axis is doing the work.
  • seam — opposite design move. Seam separates by deliberate boundary; proximity groups by deliberate placement. A given design typically uses both at different scales.
  • seeding — the seed’s disproportionate influence partly traces to proximity-grouping: subsequent additions cluster around the seed in space, time, or genre, and get grouped with it.
  • catalysis — structural analogue at a different substrate. Catalysis brings reactants into spatial proximity at the active site to enable a reaction; proximity in cognition brings items into perceptual co-location to enable a grouping inference.
  • grain — proximity is meaningless without a chosen grain. The concepts compose into a discipline: “specify the grain, then invoke proximity.”
  • red-herring — a red-herring often works by being placed proximate to the load-bearing element (in narrative time, in spatial layout, in a list). The proximity-grouping bias of the observer is what the misdirection exploits.

​
Examples

Wertheimer's classic visual demonstrations · psychology

dot arrays where dots arranged in columns get perceived as columns, the same dots arranged in rows get perceived as rows. The dots themselves are identical; proximity along one axis is the load-bearing signal.

Tobler's first law of geography · geography

“everything is related to everything else, but near things are more related than distant things.” Empirically validated across spatial datasets in epidemiology, economics, ecology, and crime; the basis for spatial-autocorrelation methods in statistics.
papers cited together by many other papers get inferred as related, even when their content doesn’t directly engage. Co-citation networks are used to identify “schools” and research communities precisely on this proximity-as-grouping basis.
functions placed near each other in a file are read as related; functions in different files are read as unrelated. The proximity of source text drives the reader’s mental model of the system’s modular structure. The discipline of cohesion is in part the discipline of making textual proximity match logical relatedness.
being seen near a stigmatized person, group, or symbol triggers attribution of shared properties. The inference is logically weak but cognitively automatic; political consultants and PR strategists manage proximity-to-symbols actively.
Constantine and Yourdon’s Structured Design (1975/1979) introduced the formal taxonomy of module cohesion — the degree to which the elements grouped inside one module genuinely belong together. They rank cohesion from strongest to weakest: functional (every element serves one task), through sequential, communicational, procedural, temporal, logical, down to coincidental (the “junk drawer” module whose elements share no real relationship). Cohesion, in their framing, is the formal name for the discipline of making textual proximity match logical relatedness: keep all the code for one function physically together so a reader (and a future editor) can hold it in one place, and a change stays local instead of rippling across scattered sites.Inference: The cohesion taxonomy gives the proximity primitive a graded vocabulary in software. The elements are lines/functions; the proximity is co-location within one module; the structural claim is that why things are placed near each other matters — co-location justified by shared function (high cohesion) is good, co-location justified merely by “runs at the same time” (temporal) or “happens to be similar” (logical) is weak, and arbitrary co-location (coincidental) is harmful. The transferable lesson is that proximity is a signal of relatedness, so placing unrelated things together actively misleads: the reader infers a relationship the structure does not actually have. Good design makes the proximity honest.
Wertheimer’s 1923 paper introducing the gestalt grouping principles isolated Nähe (proximity) as one of a small set of perceptual rules: items near each other in space tend to be grouped, even when their individual features differ. The dots-and-rows demonstrations he used became the canonical illustration — a uniform grid of dots reorganizes into rows or columns depending on which inter-dot spacing is smaller.The same shape recurs in independent literatures. Constantine and Yourdon’s Structured Design (1979) made code “cohesion” — keeping logically related code physically close — a load-bearing principle of software organization. Thomas Allen’s Managing the Flow of Technology (1977) documented the “Allen curve”: engineer-to-engineer communication frequency drops sharply with physical distance, more than common sense would expect. Tobler’s first law of geography (“near things are more related than distant things”) states the same regularity for spatial phenomena.Inference: That so many independent fields rediscovered the same regularity — proximity in some dimension predicts grouping or interaction — is the catalog’s licence to treat proximity as a portable primitive rather than as a visual-perception phenomenon. The dimension can be space, time, taxonomy, semantic embedding, or social network — the structural move is the same.
Batesian mimics rely on proximity-and-resemblance: predators encountering the mimic near the model species apply the model’s avoidance learning to the mimic. The proximity-grouping bias is built into the predator’s evolved learning machinery.
Stephen E. Palmer, Vision Science: Photons to Phenomenology (1999) — modern textbook treatment of gestalt principles and figure-ground.
the probability of communication between two engineers drops sharply (roughly as 1/distance) with physical separation. Co-locate two teams and they collaborate; put them in different buildings and they don’t. The curve is robust enough to be a design constraint for R&D facility layouts.
Allen’s Managing the Flow of Technology (1977) established the “Allen curve”: the frequency of communication between two engineers falls off sharply with the physical distance between their workstations. The decay is steep within the first 30–50 meters and then flattens — beyond roughly 50 meters a colleague is almost no more likely to be talked to than someone in another city. Allen’s striking empirical detail is that two people sitting 6 feet apart are about four times as likely to communicate regularly as two sitting 60 feet apart, and that this distance effect governs all media: people who are physically close also call and email each other more, contradicting the assumption that telecommunication erases distance.Inference: Allen gives the proximity primitive a quantified organizational instance. The entities are people; the proximity is physical distance between desks; the emergent property is the rate of information flow — and therefore collaboration, problem-solving, and innovation. The transferable lesson is that proximity does not merely correlate with relatedness, it causes interaction: placing two teams near each other manufactures the collaboration that distance suppresses. This is why the Allen curve became a design constraint for R&D facility layouts — and why it generalizes to any system where the cost of an interaction scales with separation, from cache locality to neighborhood street life.
items grouped via small whitespace and separated via larger whitespace need no explicit grouping marker (boxes, lines). Proximity does the work. Refactoring a cluttered UI often consists of nothing more than adjusting margins.
Tobler’s 1970 paper stated what came to be called “the first law of geography”: everything is related to everything else, but near things are more related than distant things. The formulation arose in the context of simulating urban growth, where spatial autocorrelation — neighboring cells of land using similar characteristics — turned out to be the dominant statistical regularity in the data.Inference: The law independently rediscovers the gestalt proximity principle for spatial data, half a century later, in a quantitative form usable for prediction. The cross-substrate consistency — perception, software cohesion, organizational communication, geography — is what licenses the catalog to treat proximity as a structural primitive rather than as a domain-specific finding.