Navon Figure¶

A Navon figure is a recognizable large form made of recognizable smaller forms, used to separate global from local visual processing.

Version: v1 · 2026-10-03 · History
Domain-specific #: 13454
Domain group: Social Sciences
Origin domain: Psychology & Behavioral Sciences
Subdomains: Visual Perception, Attention → Psychology & Behavioral Sciences

Core Idea¶

A Navon figure is a visual stimulus in which a recognizable large letter or shape is physically made from smaller, separately recognizable letters or shapes. The arrangement creates two simultaneous identities: the global form and the local elements. An experiment can ask for one identity while varying whether the unattended level agrees or conflicts. The representation is the figure; global precedence, local precedence, and interference are possible results of tasks using it, not constitutive properties of every figure or every viewer.^[1]

David Navon designed large letters from small letters to make both levels nameable and to control their complexity, salience, familiarity and predictiveness. The same two-level construction can be realized with geometric forms, and later studies use it to measure different aspects of global–local attention. This repeatable construction, not the historical paper itself or one specific letter pair, is the named abstraction.^[1]^[2]

Structural Signature¶

Sig role-phrases:

Global form: spatial arrangement makes the display recognizable as one large letter or shape. Without that whole identity, there is no global level to report.
Local elements: smaller letters or forms remain individually identifiable. Mere dots can delineate a large letter, but cannot support a named local-identity comparison unless they themselves become a controlled local category.
Hierarchical composition: the small forms build the large form, rather than sitting beside it. That relation makes the two possible answers properties of the same stimulus.
Level-selective task: an instruction to report or compare the whole or the elements turns the display into a probe of processing. The figure exists without the task, but no speed or interference claim follows from display alone.
Controlled congruence: agreement or conflict between the two identities and control of their visibility permit an inference about cross-level intrusion. If the global answer always predicts the local answer, the levels cannot be disentangled.^[1]

Condensed: identifiable elements arranged into an identifiable whole + independently varied level identities + an attended-level question. The final two terms are central to experimental use, not to bare figure membership.

What It Is Not¶

Not the global-precedence effect. A display can be a Navon figure even when a task, visual angle, exposure or participant group produces no global advantage. Navon's own paper distinguishes the stimulus design from conditional processing claims.^[1]
Not any large symbol made of marks. If local marks have no separately recognizable category, a target-at-local-level response cannot be interpreted as local-letter recognition.
Not a generic part–whole illustration. Two distinct letters placed side by side have two identities but no nested global-from-local relation.
Not a clinical diagnosis by itself. An exploratory stroke study used briefly presented figures to study bias and interference and explicitly noted that these measures were not standard cognitive assessment.^[3]

Scope of Application¶

In the original experimental-psychology design, large letters made from small letters let an observer attend to one level while the other is manipulated. Navon's first experiments tested whether the irrelevant visual level interfered with an auditory letter response; later experiments directly requested global or local report. The original results supported global interference with local report in those tested conditions, while also showing that deliberate local attention was possible.^[1]

The same structural stimulus moves within visual-perception research to nonletter geometries. One developmental comparison used global geometric figures constructed from 18 smaller figures and alternatives matching the target only at the global or local level. In an exploratory stroke study, briefly shown hierarchical letters required identification of a target at a directed level. The carrier and questions differ, but recognizable whole, recognizable elements and level-selective comparison remain present.^[2]^[3]

The scope stops before treating every composite image as a Navon figure or every response difference as a culture-, lesion- or personality-specific trait. Stimulus size, element spacing, exposure and task instructions are part of the inference conditions, not background decoration.

Clarity¶

The figure makes three frequently confused statements separable: the display contains two representational levels; one level is easier to identify in this task; and the unattended level interferes with the attended answer. The first is an observable construction fact. The other two require a comparison with controlled responses and timing. A person naming the small letters once does not establish stable “local processing,” nor does a large-letter answer prove an innate global-first rule.

Navon controlled whether whole and elements could be recognized and whether one identity predicted the other. That makes a level effect less likely to be an artifact of an illegible element or an overly familiar whole. His own discussion nevertheless cautions that real scenes do not supply the same operationally clean units.^[1]

Manages Complexity¶

Ordinary scenes contain many nested candidate units. A Navon figure reduces this to a small experimental state space: global identity, local identity, level of instruction, compatibility and response. The reduction makes selective attention and interference quantifiable without first solving natural-scene segmentation. It also deliberately removes depth, object history and much ecological context. Those omissions are useful for isolation, but they limit what a laboratory difference licenses about everyday perception.^[1]

In the geometric-form study, the comparison choices shared either the whole or its constituent elements. That design turns a broad question—“does this observer process the gestalt?”—into a specific choice between matched levels. The simplification does not erase the need to check that the alternatives differ in no easier cue.^[2]

Abstract Reasoning¶

When global and local identities vary independently, ask an observer to respond to one level and compare congruent with conflicting trials. A larger conflict cost for local than for global report is evidence that the global identity intruded on local selection under those viewing and task conditions; the inference is comparative, not a direct glimpse of a universal processing order. If an effect disappears after changing element spacing, exposure or instructions, that points toward stimulus- or task-dependence, not toward the disappearance of the figure.^[1]

One can also compare the ability to match whole versus part while keeping the presented figure fixed. If an observer chooses the global-match alternative, the result concerns that selection task; it does not by itself reveal whether local elements were invisible. Navon's original control work and later hierarchical-figure comparisons show why a response is evidence only when the level, alternatives and visibility are specified.^[1]^[2]

Knowledge Transfer¶

Literal transfer stays within visual-perception experiments: replace letters with suitable geometric forms, change which level is queried, or compare populations while preserving independently recognizable nested levels. The transfer is the stimulus schema and contrast logic, not a guaranteed global-precedence result. A diagnostic or cross-cultural interpretation needs its own sampling, controls and validation.^[2]^[3]

Beyond the home domain, a dashboard or organization may be called “global versus local” by analogy. Unless it instantiates the controlled perceptual display and response paradigm, calling it a Navon figure would import a laboratory name without its mechanism. The live Representation prime supplies the strict genus for the visual mark arrangement; broader part–whole analogies outside the laboratory stimulus still need their own test.

Examples¶

Original hierarchical letters¶

Navon made large letters out of small letters, choosing a common possible identity set at both levels to control unequal familiarity. In his experiments, a large letter could be perceived while small-letter identity failed to interfere with an auditory response; in a task requesting local report, conflicting global cues inhibited that response. Those are experimentally observed conditions, not an assertion that every hierarchical letter always works that way.^[1]

Mapped back: the large letter is the global form; the repeated smaller letters are local elements; their placement is hierarchical composition. The instructed auditory or visual report supplies the level-selective task, while controlled identities and compatibility supply the congruence comparison.

Geometric hierarchical forms¶

An original developmental comparison built each global geometric form from 18 smaller geometric forms. An observer compared a target with alternatives of which one shared its global outline and another its local elements. The figure is therefore not tied to alphabetic literacy: the same nested-identity relation can be populated by shapes, although recognition and salience must again be checked.^[2]

Mapped back: the large geometry is the global form; 18 small geometries are local elements; their arrangement composes one figure. The matching instruction is the level-selective task; alternatives sharing only one level create the diagnostic contrast.

Brief hierarchical-letter presentation after stroke¶

An exploratory stroke study briefly displayed a Navon figure and asked participants to identify a target letter at a directed global or local level. It used the representation to quantify bias and interference in a patient population, while noting that such measures were not standard clinical assessment. A single task score therefore should not be promoted to a diagnostic label.^[3]

Mapped back: the large and small target/distractor letters fill the two form roles; small letters compose the large one; directed target identification is the level-selective task; different attended and unattended identities permit interference comparison.

Structural Tensions¶

Level independence versus perceptual naturalness. Making the two letter levels equally recognizable and varying them independently improves causal attribution: an interference difference can more plausibly be assigned to level. Naturalistic scenes supply richer ecological structure but unequal complexity, familiarity and cue value. Lean entirely toward tight control and external reach weakens; lean toward naturalness and the level comparison becomes harder to interpret. Diagnostic: is this experiment isolating cross-level competition, or predicting perception in an unconstrained scene?^[1]

Global legibility versus local accessibility. A strong overall outline helps participants name the global form quickly, but close packing or brief exposure can make the component identity less accessible. Enlarging and separating components helps local report yet can alter the very global contour whose precedence is being measured. The design cannot maximize both forms of ease independently at every size and exposure. Diagnostic: can each level be named when queried alone under the exact presentation conditions, before treating an interference difference as attentional?^[1]

Structural–Framed Character¶

The figure sits toward the framed end of the spectrum as a laboratory representation: its nested relation is objectively inspectable, but a researcher chooses the element spacing, allowed identities, exposure and response question. Evaluative weight is not in the stimulus itself; it enters when a response is treated as evidence of “advantage,” “bias,” impairment or a desirable processing style. Human practice is essential to turning the display into a controlled experiment, and Navon's historically specific design became a vocabulary convention in attention research rather than a naturally occurring object kind. The words “global” and “local” travel widely, but only the controlled nested-form schema travels literally as a Navon figure. Importing the name into a business hierarchy would be metaphor; recognizing a new hierarchical visual stimulus with separately identifiable levels is legitimate. Its character: a reusable, researcher-framed visual representation with an invariant two-level composition, whose psychological interpretation remains task-bound.^[1]

Structural Core vs. Domain Accent¶

The skeletal relation is one identifiable whole composed of independently identifiable parts, then selective observation at either level. The live Representation prime is a literal parent because the graphic maps separately nameable global and local forms into a visual medium; this is not an edge from resemblance to a generic hierarchy. The domain accent is the perceptual stimulus, its visual-angle and exposure conditions, and reaction or choice measures used to infer attention. Those ingredients are not portable into organizational charts or software trees without changing the mechanism from visual recognition to something else. The named Navon figure therefore fails the prime bar: beyond visual-perception research it usually becomes an analogy. A future prime about controlled hierarchical representation is a question for separate admission, not an edge smuggled in here.

This entry is a kind of Representation.

The live Representation prime is the strict parent: every admitted Navon figure maps recognizable global and local forms into visual marks, whereas many representations lack its two-level construction. Pattern in Design and Motif remain resemblance neighbors, and neither replaces the checked genus. Task-dependent global precedence is not part of the stimulus identity.

Relationships to Other Abstractions¶

Current abstraction Navon Figure Domain-specific

Parents (1) — more general patterns this builds on

Navon Figure is a kind of Representation Prime

A Navon figure visually represents distinct global and local letter or shape identities.

Hierarchy path (1) — routes to 1 parentless root

Navon Figure → Representation → Abstraction

Neighborhood in Abstraction Space¶

Navon Figure sits in a moderately populated region (51^st percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Spatial Perception & Navigation (21 abstractions)

Nearest neighbors

Sequential Clarity — 0.86
Word Superiority Effect — 0.86
Cognitive dimensions of notations — 0.86
Visual Capture — 0.86
Zooming User Interface — 0.86

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Global precedence: an empirical tendency or effect in a specified task; the Navon figure is a stimulus that can test it, even when the effect is absent.
A simple letter mosaic: without a separately recognizable local identity and controlled two-level contrast it cannot perform the same comparison.
A validated neuropsychological test: the exploratory stroke work used figures to investigate bias; clinical diagnostic performance requires separate evidence.^[3]

References¶

[1] David Navon, “Forest before trees: The precedence of global features in visual perception,” Cognitive Psychology 9 (1977), 353–383, especially stimulus rationale pp. 358–359 and summary pp. 381–382. Original paper PDF. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m

[2] “A Direct Comparison of Local-Global Integration in Autism and other Developmental Disorders,” PLOS ONE (2012), Methods and hierarchical-stimuli figure. Original study. registry ↩a ↩b ↩c ↩d ↩e ↩f

[3] “Zooming in on abnormal local and global processing biases after stroke: Frequency, lateralization, and associations with cognitive functions,” Journal of the International Neuropsychological Society, abstract and methods. Original exploratory study. registry ↩a ↩b ↩c ↩d ↩e