Mäda PrimavesiGustav Klimt, 1912–13. The Metropolitan Museum of Art, New York (accession 64.148). Oil on canvas, 149.9 × 110.5 cm.

Orientation, for a cold reader

Vienna, 1912: a banker’s daughter, a designer’s dress, a painter of ornament

In February 1912, Eugenia Primavesi came to Gustav Klimt’s studio to commission a portrait of her daughter. The Primavesis, Otto (a Moravian banker and industrialist) and Eugenia, were among the deepest patrons of the Wiener Werkstätte, the Vienna workshops that wanted every made thing, from a teaspoon to a house, to belong to one designed world. Klimt was that world’s painter. The sitter, Mäda Primavesi (1903–2000), was about nine. The literature says Klimt made a great many preparatory sketches (the figure usually given is around two hundred), and that nearly all of them chase one thing: the stance. The white dress she wears was made by Emilie Flöge, Klimt’s lifelong companion and a working fashion designer, at Klimt’s suggestion; the Met’s own account adds that Klimt introduced the band of coloured roses at the chest to make the dress more colorful. Keep that detail; it returns with numbers attached.

This is the painting’s first fact: the literature describes it as Klimt’s only major full-length portrait of a child, and its reputation rests almost entirely on body language. The Met calls out her “confident” stance; popular writing reads defiance, self-possession, “a nine-year-old who has already decided who she is.” Critics reach for Velázquez’s infantas. Every published reading we consulted, museum, popular, and formalist alike, locates the picture’s force in pose and gaze, and treats the lavender wall and flowered carpet as backdrop: decorative, charming, subordinate.

The instrument used here cannot see pose or gaze. It computes fields (where luminance changes, where colour is, where committed edges sit, where structural mass accumulates) and reads regions and relations in one coordinate space, with no idea that a girl is a girl. That blindness is the point of the exercise: it reads the half of the painting the literature waves through. What follows is the story of what happens when you measure both halves, the address everyone sees and the substance nobody audits, and find that Klimt built them as two separate machines.

Field summary

A portrait the instrument reads as a decorative field

Run the full structural read and the classifier does not return a portrait state at all. It returns diffuse_field, the label built for compositions whose mass is spread, weakly packed, without a dominant body, with flags for low packing density and peripheral pull. The numbers behind the label: mass cohesion μ 0.194 (the library’s dissolved Degas reads 0.15; a unified Matisse anchor reads 0.81); the mass mask breaks into twenty-five components (the largest holding 19%), and in the relation system no island carries more than 10.6–17% depending on edge mode; and, on one estimator, the mass centroid sits above the girl, up toward the wall (island ny −0.174), though a second estimator on the same field reads the other way (kernel δy +0.093), so the mass field is split by register rather than cleanly displaced. A later test settles which way to read it: remove the chroma term from the mass formula and the centroid returns to centre (Entry 43), so the “mass up into the wall” reading is substantially a colour fact restated, the durable version is that colour is displaced into the wall, not mass. The structural anchor the instrument does find is not her face: it is the dense edge-knot at her elbow, where the akimbo arm breaks the silhouette (a location partly method-shaped, the densest edge-knot is where this instrument tends to anchor). [narrowed]: the diffuse verdict holds in both edge modes (entropy 0.63–0.82); the displacement direction does not.

Held in the library’s shared coordinates, this places the painting (a placement, not a measurement of it — the compare mixes three reproduction chains and a crop at n=3, so read it as “who sits where”): it reaches the Degas pole (dissolution, order without an anchor) by the opposite route (not an all-over worked weave but a void at the centre of a decorated field), while splitting eye from weight the way the Matisse pole does (saliency and mass diverge by 0.080; the Degas’s coincide at 0.02), all in the corpus’s first high key. So the first measurement already states the painting’s paradox: a single standing figure fills the frame, and structurally there is no figure there: no packed body, no tonal anchor, no chromatic center. Where did Klimt put everything?

How to weigh what follows. Several of the measurements below are the same two fields, luminance and chroma, read over the same regions in different vocabularies. That the wall is smooth and chromatic and the dress smooth and neutral is one observation, whether it is phrased as colour-in-the-void, boundary independence, or a displaced mask centroid; it is stated once here, across the next two sections, and should be counted once. The study’s genuinely independent lines, each resting on a different measurement, are five: the address channel (luminance symmetry), facture (high-frequency luminance), the ground-dynamism ablation (luminance torque), the Eugenia comparator (a second painting), and grayscale persistence read against the corpus. Those carry the weight.

The dress

A void, manufactured to advance

The dress is where the paradox is engineered. It is the largest form in the painting and its brightest: the picture is the library’s first genuinely high-keyed work (37% of tonal mass in the highlights, τ 0.598; the Matisse is midtone-keyed, the Degas and Caravaggio shadow-keyed), and the dress is the crown of that key. In absolute terms it sits at the 75th percentile of luminance, among the brightest regions in the frame, while its structural mass sits at the 28th percentile. It advances by light and recedes by substance in the same silhouette: a gap of 0.47 between how far forward it comes and how little it weighs. Two audits stand behind that number. Rebuilt on a hand-drawn, geometry-only silhouette (no colour information anywhere in the mask), the gap holds at 0.39. And it is not a general property of smooth bright paint: run the same measure on the corpus’s other candidates and Caravaggio’s lit wall reads +0.02 (bright and massive: that painting’s anchor), the Degas tutu −0.15, the Matisse piano top −0.56. The Klimt dress’s gap is unique in sign and size. [stable] (One mechanical disclosure: the mass field is defined as 0.42·edge + 0.32·tone + 0.26·chroma, so a chromatic void is a quarter of the way to a mass void by construction; the edge-and-tone 74% carries the reading. An earlier correlation version of this test was discarded: it measured a global property of smooth paint, not the dress.)

A hostile first reading of that number is “you defined the dress by low chroma and then reported low chroma.” The mask that produced these values was built the opposite way, a silhouette drawn by hand, from geometry, with no colour information at all, and every load-bearing number holds on it:

What holds this void together is drawing. Read over its own fabric, the dress carries chroma of just 0.03–0.08, near neutral, and is defined almost entirely by its contour edge (edge mass 0.383 against colour 0.069). This is a construction the library has seen once before, in the Matisse study: a figure that is an edge enclosing a void. Klimt inverts it from midtone gray to high-key white, and scales it up to the largest form in the picture. The instrument’s perturbation field adds the uncomfortable corollary: the stable substrate of this painting is the grounds; the figure is what the composition is fragile to.

The stance

Everything the critics see is real, and measurable

If the instrument stopped there, it would flatly contradict a century of viewers. It doesn’t, because pose has its own geometry, and the geometry can be measured without knowing what a pose is. Frontality, how bilaterally symmetric a form is about its own vertical axis, is the pre-semantic trace of a body squared to you. Mäda’s face scores 0.71, her shoulders 0.86; the patterned wall, as a control, scores 0.07. Her face, shoulders, and full figure column each choose a symmetry axis within 0.1% of frame width of the exact frame centre, an axis-lock so tight it reads as a decision. She spans 92% of the frame’s height and addresses the viewer from its top ninth. And the small asymmetry that remains is the liveliest part: her shoulders are squarer than her face (0.86 vs 0.71 — two conventions of one mirror-symmetry method, agreeing in sign though not magnitude), the old device of planting the stance and turning the head, plausibly what two hundred sketches were for. [stable], with the head-turn [narrowed] to the upper body.

Run the same frontality read across the corpus (six studies, seven works, since the Eugenia portrait is a within-study comparator) and Mäda is not just high; she is alone. Central-column frontality 0.663; the next work of any kind reads 0.27; her own mother’s portrait by the same painter reads 0.09. The critics’ unanimous instinct, that this painting’s force lives in the stance, is not just supported, it is the single most extreme measurement the picture produces.

The colour

Who holds the colour, and what happens when you take it away

Now audit the half the literature waves through. Measure colour over each named object’s own paint (not a loose rectangle; the mask is verified and the segmentation sensitivity was swept), and the ordering is blunt:

The bare wall out-colours the dress fabric by 9–27× in every configuration where the probe verifiably sits on fabric, and the ordering, wall above dress, survives every segmentation tried, including the most favorable one the dress can be given (worst case still 1.3×). [stable] as an ordering; the multiplier is quoted as a range, never a soundbite. That mask-verified ordering is the primary evidence here; a field-level metric echoes it — colour-in-the-void β reads 0.378, the highest in the library (Degas 0.24, Matisse 0.14) — but β measures chroma where edges are sparse, which is the smooth wall by construction, so it is cited as corroboration of the zone probes, not the other way round. Colour and drawing have been pulled apart and housed separately: drawing to the girl, colour to the room.

One caution about that last claim, since it is easy to over-read. In this instrument the mass field is 26% chroma by construction, and the grayscale test zeroes exactly that term, so a colour-keyed work is partly built to fail it. The honest test is against the corpus, and it was run: degrade all seven works and Klimt’s grayscale persistence (0.455) is the lowest of the seven, while the two most value-structured works, the midtone Matisse (0.89) and the tenebrist Caravaggio (0.75), are the most robust. Blur (0.82) and low-contrast (0.85) leave Klimt’s anchor fixed; grayscale relocates it (drift 0.32). So the defensible statement is comparative: Mäda Primavesi is the most colour-dependent work in this corpus, and the value-keyed portraits are the most colour-robust, not a universal law about portraiture, and partly a property of the mass formula. [narrowed]

Two grammars

The figure is drawn; only the wall is coloured into being

The split runs deeper than where colour sits; it reaches how things are made. Score each region by how much of its content is committed in each channel (the share of its pixels above the whole frame’s 85th percentile):

Two findings sit in that table. First, the figure and the wall are built in different, non-overlapping languages: the girl is a value drawing, the wall is colour doing the structural work (chromatic-boundary independence 0.191, the corpus high: colour drawing lines that light does not draw). Second, and this corrects our own earlier framing, the split is not figure-versus-ground. The floor is value-drawn too, and the three blue zigzags at the base, which a formalist critic reads as “securing” the figure, turn out to be the most value-committed marks in the painting. They secure the base by extending the figure’s own language to the frame edge, not by translating between systems. The painting is a value drawing with exactly one colour-drawn zone: the upper field the girl stands against. [stable]

The punctuation

Two accent systems, one per language

The grammar split repeats at the scale of individual marks. Ask the instrument for the painting’s accents (small, high-leverage events) through its default lens, which is weighted toward edge and tone, and it returns only dark marks: the deep blue-green blobs and bird-forms scattered on the pale floor, and the belt’s dark leaves (leverage 0.83–0.87). A sparse, descending punctuation of value events below the waist. Switch the lens to the chromatic channel, coalesced to brush scale, and an entirely different set lights up: the rose belt, the flush of the cheeks, and the distributed flower-flecks of wall and floor: the ornament’s own grammar. The two systems barely overlap. The value language gets dark full-stops on the floor; the colour language gets scattered blooms; and only at the belt and the face do the two punctuation systems touch the girl. (Accent counts are capped by the lens and are never abundance measures; what is claimed here is the segregation of the two systems by channel, which survives the cap.) [stable as segregation]

The rivets

Two welds hold the void to the field, and the record marks one of them

Two separated systems still have to make one picture, and the joints are visible in the numbers. Trace chroma down the figure’s column and it is a flat, near-neutral line (median 0.087) with one dominant spike: the rose belt, at 0.314 of frame height, above the figure’s centre of mass, 0.07 of frame height shy of the upper golden section, at the empire waist of the dress, the pivot of the standing form. On its own paint the belt reads 0.42–0.91 across segmentation, reaching the wall’s own level at the favorable end and roughly half of it at the worst: less an accent on the figure than a piece of the wall’s language, the one place the subject approaches the ground’s full saturation. Only 5 of 229 measured rows on the figure ever reach ground-level colour, and all five belong to the belt band; take the belt away and the figure’s ceiling drops to 0.48, at the scattered flower ornaments on the skirt. [narrowed]: the belt is the dominant, banded weld; the ornaments are sparse secondary points below the ground’s level.

The second weld is not colour but paint. Measure facture (fine, high-frequency surface energy) and the two raised events on the figure are the belt (0.048) and the blue hair-bow at the crown (0.043), each about 1.5× the smooth dress; the wall between its flecks is as smooth as the dress (0.029). So the figure is pinned at waist and crown by its own two raised events, the belt a double rivet, highest chroma and highest facture at once. [narrowed]

Here the commission record and the measurements meet. The dress was Emilie Flöge’s work; the rose band, by the Met’s own account, was Klimt’s addition to it. On the canvas as measured: the garment attributed to the designer is the chromatic void, and the one concentrated colour event the figure carries, pitched at the level of the painter’s own wall field, is the painter’s documented insertion. Scope this carefully: the account covers the belt; it does not settle whether the roses were sewn or only painted, and it says nothing about the sparse skirt ornaments, whose authorship is open. What the pixels and the document jointly support is narrower and still sharp: the figure’s single ground-level colour event is the one the record attributes to Klimt’s intervention in the designer’s garment. [narrowed, documented]

The aperture

The gesture of command is where the field gets in

The single most-quoted feature of this painting is the arm akimbo: the hand planted on the hip, the elbow thrown out, the child’s claim of authority. Structurally, an akimbo arm does something no writer we found mentions: it encloses a triangle of negative space between forearm and torso. In this painting that triangle is open; the wall passes through it, and floor blooms show at its base. Measure the colour inside the aperture and it reads 0.50–0.56 (two independently hand-drawn geometry masks): 58–90% of the wall depending on which wall read you divide by (65% against the same mass-locked wall value as the table above), and 7–17× the dress fabric. On every basis the same fact: the decorative field does not stop at the figure’s silhouette; it enters the figure’s convex hull through the gesture that asserts dominance over it, and the arm that encloses that gap is the densest edge-knot in the painting, the point the instrument selects as its structural anchor. The pose of command and the breach in the outline are one construction: the same arm draws both (the anchor is partly method-shaped, but the aperture chroma is not — it reads the same on any mask). [stable]

Three observations sharpen the moment (a reader’s, two of them tested). Both hands are hidden: the akimbo hand disappears behind the hip, the other arm runs behind the back. The pose everyone reads as command shows no hands, a tension the instrument cannot adjudicate. The deep violet a viewer registers behind the wrist and hip is the aperture’s own interior; around the rest of the silhouette the wall actually pales (a lighter, less saturated band against the figure edge: the same push-forward effect by the opposite mechanism). And the sleeve rosette with its two green leaves sits exactly on the elbow: an ornament planted on the anchor knot itself, the first link of a chain (elbow, belt garland, skirt flowers, the floor’s strewn islands) that carries the ornament’s grammar across the body and out into the room. [aid]

The division of labor

The icon refuses to move; the field moves for it

Stand back from the parts and the painting’s economy becomes visible. The figure is locked to the frame’s axis to a tenth of a percent, bilaterally squared, nearly weightless: an icon. Yet the whole image carries a strong asymmetric pull, up and to the left. Where does the movement live? The clean way to ask is to remove one part and leave the frame intact. Inpaint the figure out (no crop, no flat fill) and the torque is unchanged: 0.285 against 0.288 whole. Inpaint the ground out instead, same frame, and it collapses to 0.046. The movement is entirely the ground’s. And this measure — unlike the mass metrics — is computed on luminance, not on the chroma-carrying mass field, so it is a genuinely independent line, not the colour observation restated. [stable]

A candidate measure the library is still testing puts the same fact a different way. R_spatial asks whether an image’s global asymmetry is carried by some region or emerges only from the arrangement: this painting reads 2.9–5.0, the strongest and widest such reading in the corpus. Cut the picture anywhere and every piece is nearly balanced; only the whole leans. No region owns the movement; the arrangement does, which is what a composition looks like when its icon is symmetric and its energy is distributed through the field around her. [aid: a candidate measure, quoted as a range; its cut-sensitivity is part of the reading]

The stage for that division is flat, the premise on which this whole reading rests, and, in candor, the weakest-supported of the study’s structural claims (one of its three tests, a mark-scale gradient, came back too noisy to use; the other two hold). [aid] The floor shows no recession: its colour and edge content are uniform front to back (variation ~13%), the zigzags are a frontal repeat with no vanishing convergence, and after discarding the canvas edge the painting contains no straight line longer than 6% of the frame’s height, not even the wall/floor junction, which never resolves into a committed horizontal, so no armature, no perspective box, nothing for depth to hang on. (Compare the library’s Degas, with its 22-line convergence node, or the Matisse’s rectilinear grid.) A formalist critic’s claim that carpet and figure form a stabilizing pyramid resolves into channels, after a sensitivity sweep over the triangle’s shape: the mass field anti-correlates with every version of it (−0.29 to −0.32 across nine apex/base configurations; the weight sits in the upper corners, outside the triangle), while the contour support does not survive the same sweep (edge NCC −0.00 to +0.04, indistinguishable from zero). What survives of the critic’s intuition is real and specific: the base band genuinely is the painting’s strongest horizontal seam. The pyramid is not built, and barely drawn; the base is secured. [stable]

Put the three refusals together, a flat stage, a weightless centred icon, a field carrying all the motion, and the centered-ness itself gets its audit: the Default Gravity Index reads 52, “leaving the basin”: centering near-maximal (0.79) while inner mass is low (0.176). Read those two facts separately rather than through the composite, of the index’s four inputs only inner-mass is mass-derived, so it carries the same chroma caveat as the displacement claim above; the honest statement is the pair, centred by frame and light, thin at the core. The composition performs centrality and, at least by the one contaminated term, declines to fund it. The eye, meanwhile, goes where the apparatus points, to the girl (saliency and mass diverge by 0.080, the Matisse pattern, not the Degas one). Where you look and where the picture’s weight sits are two different places, and the painting works because they are.

The pair

The same painter, the same family, the opposite policy

Is this simply how Klimt painted? The controlled comparison exists: a year later, Otto Primavesi invited Klimt to paint his wife. Eugenia Primavesi (1913–14, Toyota Municipal Museum of Art) is the same painter, the same patron family, the same ornament-saturated world, and the opposite decision on every axis this study measures. (Its reproduction is an art-book scan, so these are within-image, directional comparisons; decimals are not claimed.)

The mother’s dress is the most chromatic surface in her portrait; the daughter’s is the least in hers. The mother’s weight sits on her body; the daughter’s is displaced to the wall field. On the address axis they split maximally: the frontal-faced Eugenia lands, in the corpus map, next to a Pollock and the masked test (isolating her figure from its setting) found that low score largely real, not a setting artifact: her figure, about its own axis, genuinely reads less bilaterally symmetric than her daughter’s. Two independent channels, the same split, one painter, one family, a two-year window. The chromatic-void treatment of Mäda is a choice, not a formula, a dial Klimt turned one way for the child and the other for the mother. (One trait the pair shares: both read superadditive on the candidate arrangement measure; the emergent-whole signature may belong to the ornament-ground genre itself, while the figure-chroma policy is the variable.) Sources call Eugenia’s portrait a “counterbalance”; no document confirms the two were conceived or hung as a pair, so intent stays unclaimed. What is measured is that the two canvases are poles of one opposition. [stable as measurement; the designed-diptych reading is undocumented]

Interpretation

Ornament and a child, Vienna 1912: the argument the numbers walk into

Everything above is measurement; this section is reading, and says so.

In the same city, in the same years, Adolf Loos was prosecuting his war on exactly this aesthetic. Ornament und Verbrechen (written around 1908, lectured in 1910) framed ornament as regression, as the smothering of the honest modern individual under decoration. This painting is as close as Klimt ever came to staging that fight on one canvas, and the measurements land on both sides of it at once. In the channel Loos cared about, matter, weight, substance, the painting enacts his charge almost literally: the person is out-coloured by the empty field behind her (an ordering that is unbreakable, at a fabric-locked 9–27×), out-weighed by her own setting, breached at the elbow. In the channel the viewers use (address, frontality, scale, the locked axis) she overpowers everything in the library. And the pair shows the arrangement was elective: her mother got the opposite settings from the same hand.

Both period readings of the pose therefore stay open, and the instrument genuinely cannot pick: a girl sidelined by décor despite every formal honour, and a girl whose presence is so secure it needs no material backing, are the same set of numbers. What the numbers do add to a century of writing about her stare is the location of the machine: the stare works because the apparatus of address and the apparatus of substance have been split, tuned to opposite extremes, welded back together at two points, a rose belt at the waist and a painted bow at the crown, and breached at a third: the open triangle at her hip where the field she out-faces slips inside her outline. Klimt gave the child everything a picture can say and kept everything a picture can hold. The portrait lives at that seam, and the seam is measurable.

Public frame

Where this sits against existing readings

The pose-and-gaze consensus (the Met, Google Arts & Culture, popular writing): confirmed, and given an instrument. The address channel puts numbers on “commanding” (frontality 0.71/0.86, the frame-axis lock, 92% height, the head-turn inside the square stance) and finds Mäda alone at the address-forward extreme (0.663; the next work of any kind reads 0.27). What the consensus misses is that address is the only axis on which the figure dominates; on every material axis the ornament wins, and the consensus’s “decorative background” is the painting’s structural engine.

The formalist pyramid (Artists Network’s “Anatomy of a Painting”): adjudicated rather than contradicted. Their zigzag intuition survives measurement (the base band is the painting’s strongest horizontal seam); the pyramid itself is inverted in the mass channel (robustly negative across a shape sweep) and unsupported in the contour channel once swept. Two notions of “compositional stability” part ways in this painting; the critic used one, this study measured both.

The defiance / proto-feminist reading (HuffPost, Sartle): neither endorsed nor debunked. The measured seam supports it and its opposite with equal ease; the study’s contribution is to say precisely which half of the picture that debate is about (the address apparatus), and that the other half was deliberately withheld from her.

The provenance record: the family sold the portrait by 1928 (through Hugo Bernatzik to the Neue Galerie); it entered the collection of Jenny Pulitzer Steiner by 1937, was seized after the Anschluss in 1938, restituted to Steiner herself in 1951, and came to the Met in 1964 as the gift of her daughter Clara Mertens, “in memory of her mother.” The Open Access image this study is built on carries that history; unlike Steiner’s Wasserschlangen II, this chain closes with restitution to the owner in her lifetime. (The chain is reconstructed from the record’s search index and secondary sources; a verbatim reading of the Met’s own provenance field was still pending at study close, and the Lost Art entries were not machine-accessible.)

What this does not prove

Limits of the evidence, and the corrections that shaped it

• Not intention. The separation of address from substance is measured; that Klimt meant it as meaning is a reading, and lives only in the section marked as one. The Mäda/Eugenia opposition is measured; a designed pendant is undocumented.
• One reproduction per painting. Every number is a fact about (painting × reproduction × scale window). The Met capture is museum-grade; the Eugenia is an art-book scan and is used only for within-image, directional comparisons. Resolution-sensitive values are quoted in their stable window (the structure-mask fractal dimension reads ~1.53 at 512–2048px; its native-res value was a documented artifact).
• The headline ratio is a range, not a number. “The wall out-colours the dress” is segmentation-proof (worst case 1.3×); “12×” is one configuration of many (fabric-locked: 9–27×). The audit that retired the soundbite also surfaced a finding: the probe escapes onto the dress’s own scattered ornaments, because the dress really is a void plus ornament points.
• Corrections are part of the record. A first-pass belt value (0.256) came from a probe box that missed the garland; corrected on the roses it reads ~0.88, a correction that strengthened the reading it corrected. Both values, and every refuted test (a bilingual-zigzag hypothesis, a counter-current hypothesis, a wet-over-dry confound, a proposed frame-padding experiment declined as methodologically circular), are kept visible in the lab book.
• The address channel has a ceiling. Its frontality is bilateral-symmetry-of-luminance: a good coarse proxy for facing, blind to gaze itself, and unrefinable into pose sub-signals without the object recognition this instrument deliberately does not have. Its head-turn reading is trusted here only because two conventions of the one mirror-symmetry method agreed in sign here (they differ ~3× on the face).
• Semantic labels are human. “Dress,” “belt,” “elbow,” “wall” are names attached to verified pixel locations; the instrument knows none of them. Whether the field behind her is wallpaper, a hanging, or pure painted surface is exactly the kind of identification this method cannot make, and this essay calls it only “the wall” or “the field.” The 92% figure-height value is likewise a measurement read off the reproduction by eye (crown to sole), not a detection.
• Every chroma number is relative to this reproduction. “Colour” here is OKLab chroma computed from the JPEG decoded as sRGB and robust-normalized per image (see the glossary below), so every chroma ratio inherits the museum photograph’s white balance and ICC handling, an uncontrolled error source on top of the single-reproduction bar. The wall-vs-dress ordering is robust to any monotone recalibration; the multipliers are not.
• The mass field is 26% chroma by construction (0.42·edge + 0.32·tone + 0.26·chroma, edge and tone from luminance). Two consequences, both tested rather than hidden: the grayscale-persistence result is partly formula-guaranteed (hence the corpus baseline above), and the “mass displaced into the wall” reading is substantially a colour fact — removing the chroma term returns the mass centroid to centre, so this essay states it as a colour claim, not a mass claim. Torque, by contrast, is computed on luminance alone, so the ground-dynamism result does not inherit chroma.
• Several sections are not independent evidence. The field summary, dress, colour, and two-grammars sections are the same two fields (luminance, chroma) over the same regions in different vocabularies — one observation, counted once. The independent lines are the five named at the top.
• One measure above is a candidate, not an instrument result. R_spatial (the emergent-whole reading) is on the library’s candidate register, quoted as a range; it has not earned the core and is graded [aid] here.

Glossary

Every number, operationally

All fields are computed from the reproduction decoded as sRGB and are robust-normalized per image (1st–99th percentile mapped to 0–1) unless noted. Every value is therefore a position within this reproduction’s own distribution.

Sources & record

• The Metropolitan Museum of Art, object 436819, accession 64.148 (Mäda Primavesi): dating, sitter, credit line, provenance. Open Access (CC0), image DP243354.
• The Met / Google Arts & Culture, “Klimt’s Portrait of Mäda Primavesi”: the commission (February 1912), the preparatory sketches, and the Flöge dress account (made by Emilie Flöge at Klimt’s suggestion; the rose band Klimt’s addition).
• Artists Network, “Anatomy of a Painting: Gustav Klimt Captures Mäda Primavesi”: the pyramidal-composition and zigzag claims tested here, and the wet-over-dry technique note.
• Adolf Loos, Ornament und Verbrechen (c. 1908; lectured 1910; French publication 1913).
• Jenny Steiner: biographical record (collection confiscated 1938, Dorotheum auctions from 1940); the Wasserschlangen II restitution literature; Lost Art database person entries (Steiner, Mertens).
• Rudolf Arnheim, Art and Visual Perception (California, 1954/1974): perceptual weight, the centric/eccentric frame.
• Instrument lineage: the ISR library: Matisse, Degas, Pollock, Caravaggio, Soejima, Klimt, Bruegel, Cartier-Bresson, Sora Pair, MidJourney Ensemble, Velazquez, and Malevich
• Studies (same coordinate space; the calibration poles cited throughout); instrument findings I-1…I-14, OBSERVATIONS.md.
• The full measurement record for this painting (43 lab-book entries including every correction, sensitivity sweep, refuted hypothesis, and the candidate-register updates): klimt_labbook_standalone.html.