Mondrian 1905–1917 — structural arc across three readings
Manovich · brightness × saturation | VTL · centroid offset × cohesion
VTL · cohesion over time (year × μ arc)
VTL · cohesion over time (year × μ arc)
ImagePlot 2 — Perceptual Space
x: brightness · y: saturation · color: year
brightness
saturation
VTL Kernel — Structural Space
x: centroid offset (signed) · y: cohesion · color: year
Δx± (centroid offset)
μ (cohesion)
VTL Kernel — Temporal Arc
x: year · y: μ cohesion · color: year
year
μ (cohesion)
Year
Thumb size
28px
19051917
Left: Perceptual: brightness rises, saturation compresses (147→208, 98→45) — the tonal story De Stijl tells the eye.
Center: Structural space: centroid drift tightens, mean |Δx| drops 44% by 1910 — mass consolidates before the visual grammar changes.
Right: Temporal arc: μ peaks in 1914 (Cubist synthesis, four paintings above 0.55), then drops in 1917 (Plus/Minus scatter, μ 0.18). The mature grid is not in this dataset.
128 paintings. Three readings.
Before the Grid
128 paintings · 1905–1917A perceptual and structural reading of Mondrian, 1905–1917
Russell Parrish · Artist Influencer · artistinfluencer.com
· Images and perceptual data: Lev Manovich / Cultural Analytics Lab
The Structural Finding
Between 1905 and 1917 Mondrian's compositions undergo a measurable structural shift that precedes the famous grid. The clearest signal is centroid consolidation: early works are laterally scattered: 38% lean left, 34% center, 28% right. By 1910–1913, 57% of paintings are center-balanced. Mean horizontal displacement drops by 44%. The compositions become gravitationally neutral before any visual grammar of De Stijl is in place.
A second signal follows in 1913–1914: peripheral pull (xp) collapses from ~0.18 to 0.07. Structural mass migrates away from frame edges. This is the Cubist synthesis moment; cohesion (μ) peaks in 1914 at mean 0.49, with four paintings above 0.55. The composition contracts inward and unifies.
Then 1917 reverses. The Plus and Minus works (all-over compositions with no single dominant region) score lower cohesion (mean μ 0.18) than the early landscapes. Peripheral pull rises again (xp ~0.20). The kernel reads structural fragmentation in exactly the works that feel most abstract. The grid itself falls outside this dataset; what is visible here is the structural preparation and the scatter that preceded it.
Centroid neutrality arrives before the grid. Fragmentation follows the Cubist peak. The grid is the resolution of a structural arc, not its beginning.
Structural Arc by Period
Early · up to 1909 · n=94
Scattered
μ 0.285 · xp 0.184 · |Δx| 0.081
Lateral scatter, edge-anchored mass. Trees, dunes, figures pull left and right. No structural center of gravity.
Cubist · 1910–14 · n=27
Consolidating
μ 0.372 · xp 0.122 · |Δx| 0.046
Centroid pulls to center. Peripheral mass drops. Cohesion peaks in 1914. Compositions contract inward.
Plus/Minus · 1917 · n=6
Scattered again
μ 0.175 · xp 0.204 · |Δx| 0.021
Lowest cohesion in the dataset. All-over distribution reads as fragmentation, not unification. Centroid stays centered, but mass disperses.
What to Look For
Early left-bias. The early landscapes (blue, right panel) lean left of center more than right: 36 paintings left vs. 26 right. Dutch landscape convention places sky on the right, mass on the left. The centroid neutrality that arrives in 1910 erases this asymmetry. The kernel is reading a compositional convention before it reads an artistic decision.
Left panel: perceptual story. Brightness rises, saturation compresses as Mondrian moves toward De Stijl: 147→208 brightness, 98→45 saturation by period. The tonal trajectory is monotonic and visible to the eye. Color encodes year.
Right panel: structural story. Horizontal axis is signed centroid offset: left of center = mass left, right = mass right, near zero = balanced. Vertical is cohesion (μ). Watch early blue points scatter wide and low; 1914 reds cluster high-center; 1917 reds drop back to low cohesion.
Hover for tooltip. Each painting shows image, year, and both panel's axis values simultaneously. Find the 1914 cluster in the right panel: these are the Cubist paintings with highest cohesion. The 1917 cluster sits low despite appearing visually advanced.
Navigation
Scroll wheelZoom toward cursor, each panel independent
Click + dragPan, each panel independent
HoverImage preview, year, axis values for both panels
Thumb sizeSlider at bottom scales thumbnails in both panels
Metric Definitions
| Metric | Range | What it measures |
|---|---|---|
| Left Panel: Perceptual | ||
| brightness | 0–255 | Median luminance. Rises monotonically across career (147→208). |
| saturation | 0–255 | Median color saturation. Falls monotonically (98→45). The De Stijl compression. |
| Right Panel: Structural (VTL) | ||
| Δx± centroid | [−1, +1] | Signed horizontal position of gradient-weighted structural mass. Near zero = centered. The centroid neutralization is the earliest and clearest structural signal. |
| μ cohesion | [0, 1] | Largest connected structural region as fraction of total. High = unified (one dominant field). Low = fragmented (many small regions). Peaks 1914, bottoms 1917. |
| xp (tooltip) | [0, 1] | Peripheral pull: structural mass near frame edges. Drops from 0.18 to 0.07 in 1913–14, then rises again in Plus/Minus works (0.20). |
| SDI (tooltip) | [0, 1] | Mean distance of structural mass from its centroid. Relatively flat across all periods (0.45–0.47); dispersal does not track the structural arc. |
Dataset and Scope
128 paintings, 1905–1917. Images and perceptual data from Lev Manovich / Cultural Analytics Lab, released with ImagePlot 2 (2026). Structural metrics computed by VTL kernel at 256px.
Dataset ends at 1917. The mature De Stijl grid paintings, what this structural arc was building toward, are not included. The arc visible here is the preparation and scatter that precede the grid, not the grid itself.
n is uneven. 94 early paintings, 21 mid, 13 late. Year-level means for 1910–1917 should be read cautiously (3–6 paintings per year). The period-level findings are robust; annual trends are suggestive.