Composition - Perception.

The Gestalt laws

Founded in the 1910s-20s by Wertheimer, Köhler and Koffka, Gestalt psychology holds one thesis: the perceptual whole differs from the sum of its parts - the brain actively constructs the image. The governing principle is the law of Prägnanz: among all possible organisations of a stimulus, the brain picks the simplest, most regular, most symmetric. All other laws follow: proximity (close elements form a group - spacing governs structure), similarity (alike elements group - the basis of all typographic hierarchy), continuity (the eye follows smooth paths - alignment produces unity), closure (the brain completes incomplete shapes - the Kanizsa triangle, fragmentary logos), figure-ground (the separation can be stable or ambiguous - Rubin's vase), symmetry (symmetric configurations look stable and complete), and common fate (elements moving together group - animation and transitions).

These laws are neither aesthetic rules nor cultural conventions: they are regularities of perception, very widely shared and largely pre-attentive. A composition that respects them reads as clear to anyone; one that violates them reads as disordered without the viewer being able to name why.

Immediate evaluation criteria for a composition: are related elements close? Do same-rank elements share one appearance? Can the eye follow clear paths? Are shapes complete or intentionally fragmented? Is the figure-ground separation crisp? Is the overall rhythm balanced or dynamic as intended?

Arnheim - visual forces

Rudolf Arnheim (1904-2007), a student of Wertheimer and Köhler, published Art and Visual Perception in 1954 (revised 1974). His thesis extends and radicalises Gestalt: the visual field is dynamic, traversed by forces. A disc near the edge “wants to fall”; at the centre it is “at rest”; on a third it is “active”. These sensations are not metaphors - they are reproducible psychological forces. Every frame has an invisible structural skeleton (centre, axes, diagonals) acting as a field of attraction: aligned with it, an element looks stable; misaligned, it generates tension - sought or avoided by intent.

Two rules of perceptual asymmetry: visual gravity (an element placed high looks heavier - the bottom of a balanced composition must carry more) and left-right weight in Western reading (an asymmetry exists, linked to attention and handedness, but its direction is still debated). Arnheim distinguishes symmetrical balance (calm, formal, hieratic) from dynamic balance (asymmetry compensated by forces - the principle of the Bauhaus, the Swiss school and contemporary design). McManus, Stöver and Kim (2011) measured that art photographs place their centre of mass closer to a structural axis than control images; their other tests do not confirm the theory literally: a fruitful intuition, not a law.

Itten - the wheel and the seven contrasts

Johannes Itten (1888-1967), master of the Bauhaus Vorkurs (1919-1923), published Kunst der Farbe in 1961 - a teaching reference of colour theory. His 12-segment wheel at 30° orders three primaries (red, yellow, blue), three secondaries and six tertiaries. The idea of a colour circle goes back to Newton (Opticks, 1704, in seven colours); Itten stabilised this twelve-part circle as a teaching tool. This site's chromatic wheel (12 segments at 30°, in OKLCH) belongs to that lineage. Harmonies are geometric relations on the wheel: complementary dyad (180°, maximal tension), equilateral triad (120°, dynamic balance), tetrad (90°, rich - needs one dominant), analogous (adjacent, gentle harmony), split-complementary (a compromise of richness and tension).

The seven contrasts: hue (distant pure colours - maximal with the three primaries), light-dark (the fundamental contrast of any hierarchy - yellow lightest, violet darkest), warm-cool (physiological before being aesthetic), complementary (opposites intensify each other and “vibrate”), simultaneous (a colour projects its complement onto its neighbours - deepened by Albers), saturation (pure against greyed), and extension (surface proportions: Goethe proposes yellow/violet ≈ 1:3, orange/blue ≈ 1:2, red/green ≈ 1:1 for balance).

Albers - the interaction of colour

Josef Albers (1888-1976), Itten's student then a master at Black Mountain College and Yale, published Interaction of Color in 1963: “In visual perception a color is almost never seen as it really is.” Where Itten builds the system, Albers demonstrates relativity: one colour looks like two depending on its ground; two different colours can look identical on compensating grounds; neighbours project their complement at the boundary; prolonged fixation yields a complementary afterimage (retinal fatigue). His 150 exercises unfold four demonstrations: one colour looks like two, two look like one, subtraction (a colour “cancelled” by context), optical mixtures (pointillism, the pixel principle).

Itten and Albers complete each other: the ordered structure (wheel, contrasts, harmonies) and the perceptual dynamics (simultaneous contrast, relativity). The operational consequence: a palette cannot be judged out of context - actual perception depends on proportions and juxtapositions, and juxtaposed complements intensify each other. Document generation must therefore run in two passes: theoretical selection by harmony rules, then perceptual validation in real context.

Eye paths and attentional hierarchy

Perception is a temporal journey: the eye moves in very brief saccades, between 200-300 ms fixations. The Nielsen Norman Group (2006, across thousands of pages) described the F-pattern on text-dense pages (full top sweep, shorter second sweep, descent along the left edge - opening words must carry), design practice adds the Z-pattern on visual pages (logo, navigation, diagonal, primary action bottom-right - the four corners of the Z are the zones of maximal attention), and the Nielsen Norman Group the Layer-Cake pattern, the desirable one: each heading triggers a full horizontal read. F is not fate - it emerges when structure is missing; strong hierarchy produces Layer-Cake.

Independently of position, some factors capture attention: contrast, size, saturated colour (and an isolated colour in a neutral field), motion (the most powerful factor), faces (a dedicated brain module; a depicted gaze steers the viewer's gaze), isolation in white space, and downward verticality. Every effective composition has a focal point - created by concentrated contrast, converging lines, isolation, a pattern break, or a single colour in a desaturated field. Without it, the eye wanders.

The main heading in the zone of maximal attention (top, or the Z entry); the chromatic focal aligned with the identity's dominant; progression steered by the typographic hierarchy (a heading = a return to the left, then Layer-Cake); the hero image anchored by a face or strong contrast.

Cognitive load and perceptual economy

The brain is a limited resource - three laws measure it. Hick-Hyman (1952): decision time grows logarithmically with the number of options - a 5-item menu is processed fast, a 50-item menu paralyses; prefer 5-7-item submenus. Miller (1956), the 7 ± 2: working memory holds 5 to 9 simultaneous items (Cowan, in 2001, puts the limit closer to about 4) - navigation, feature lists and hierarchy levels must stay under that limit. Least effort (Zipf 1949): between two equivalent interfaces, humans almost always pick the simpler - established conventions, recognition over recall, fewer steps.

Sweller (1988) formalises cognitive load: intrinsic (the content's difficulty), extraneous (difficulty added by poor presentation), germane (productive learning effort, added around 1998). Design's first function is to minimise extraneous load. The brain prefers regularity, repetition, clear hierarchy and chunks (06 12 34 56 78 is retained better than 0612345678).

Palette limited to 3-5 main colours (by analogy with Miller); 3 heading levels maximum; 2 typefaces maximum (least effort); a regular grid (predictability lowers load). A composition within these limits is processed without conscious effort.

Colour emotions - from theory to measurement

Itten and Albers were theoretical and qualitative; since the 2000s experimental psychology measures. Jonauskaite et al. 2020 (over 4,500 participants, 30 countries, 12 colours, 20 emotions): colour-emotion associations are strongly universal - red-anger/love, black-grief/fear, yellow-joy; warm colours carry activating emotions, cool ones deactivating; black and brown are the most negative, pink among the most positive; linguistic and geographic proximity modulate at the margin. Spence and Di Stefano 2022: colour-sound correspondences are not direct, they pass through emotional mediation (colour → emotion ← sound), validated by Palmer 2013 (PNAS), Palmer 2016, Isbilen & Krumhansl 2016, Whiteford 2018.

Palmer and Schloss 2010 (PNAS), the Ecological Valence Theory: colour preferences aggregate experiences associated with objects of each colour - blue (sky, water) universally preferred, khaki (decay) rejected, saturated red ambivalent (blood / ripe fruit). Three levels stand apart: universal (red-anger, yellow-joy, blue-calm, black-mourning - near everywhere), cultural (bridal white in the West, mourning white in Asia; festive red in China), individual (biography, synaesthesia - marginal in mass communication). This is precisely the empirical ground this site's crossmodal projection maps, zone by zone, in the OKLCH bicone.

Synthesis - articulating the three pillars

The three pillars form one system: the grid (the structural skeleton), typography (the signifying surface), perception (the relation to the viewer). Decision sequence for composing a document: 1. define the intent (register, dominant emotion, desired eye path); 2. choose the palette (dominants by intent, Itten's contrasts, Albers validation in context); 3. define the grid (format, margins, columns); 4. choose the typography (class by register, pairing, scale, measures); 5. compose (hierarchy on the grid, Gestalt laws, focal point, Arnheim balance); 6. verify (cognitive load, eye path, perceptual validation).

A ten-question validation grid: grid consistent across the page? margins in a harmonic ratio? 60-70 characters per line? three hierarchy levels maximum? typefaces consistent in register and metrics? chromatic dominants aligned with the intent? Gestalt laws respected (proximity, similarity, alignment)? focal point clear and well placed? load within Miller's limits (5-9 attention items)? eye path guided? Ten yeses: the composition is professionally publishable.