Full intensity, and a corner.Pieter Bruegel the Elder (attributed) — Landscape with the Fall of Icarus · Royal Museums of Fine Arts of Belgium, Brussels

Orientation, for a cold reader

The painting, the poem, and the instrument

In Ovid's telling of the myth (Metamorphoses VIII), three bystanders witness Daedalus and Icarus overhead: a ploughman, a shepherd, a fisherman, each astonished at the flying men. The painting keeps all three and deletes the astonishment. The ploughman drives his furrow, the shepherd leans on his staff gazing at an empty patch of sky, the fisherman watches his line; the nominal subject is a pair of legs disappearing into the water beside a departing ship. The picture became the canonical statement of catastrophe ignored, fixed in that role by W. H. Auden's Musée des Beaux Arts (1938): "how everything turns away / Quite leisurely from the disaster," and by William Carlos Williams's Landscape with the Fall of Icarus (1960), "unsignificantly / off the coast / there was // a splash quite unnoticed." Period literature adds a Netherlandish proverb the composition is usually read against: no plough stands still for a man who dies.

The object itself is contested, and materially compromised. The Brussels canvas is widely discussed in the literature as an early copy after a lost Bruegel original of about 1558; a second period version (the Van Buuren Museum panel, with Daedalus still in the sky) survives across town. The doubts rest partly on the support itself: oil on canvas is anomalous for Bruegel, one technical reading holds the picture was transferred from panel with the paint layer damaged and heavily overpainted through two relinings, and the most recent technical study (Currie and Allart, 2012) discounts the transfer while leaving the condition history real. Every fine-surface number in this essay therefore reads paint plus restoration; the skeleton-level readings (islands, mass grid, tonal planes) are the tier predicted to survive that, and the versions section tests the prediction as far as one phone photograph allows. This essay reads the Brussels picture through its reproduction, takes no side on the hand, and the versions question gets its own measured section near the end.

A word on the instrument, since the rest of the essay leans on two of its measurements. The Image Structure Reader passes over the whole picture and, for every point, records a handful of plain physical quantities: how sharp the local edges are, how strong the light-to-dark contrast is, how saturated the colour is. It blends the first three into what this essay calls structural weight (or “mass”): a single map of where the picture presses hardest, where the paint is doing the most work to hold your eye. Separately it computes saliency, a standard, training-free model of where a viewer's gaze is pulled on a first glance, derived from the image alone. Both are deterministic and repeatable, and the instrument recognises nothing while producing them: it does not know a ship from a shepherd, only that one patch carries more weight or pull than another. It has now read seven canonical paintings, so every figure below can be set beside Matisse, Degas, Pollock, Caravaggio, Soejima and Klimt in the same units. Its usefulness here is blunt: the whole Icarus tradition rests on a claim about attention and neglect, and attention and neglect are exactly what this machine can put a number on.

What kind of picture this is

Crowded like a drip painting, ordered like nothing in the corpus

Left to sort the picture on its own, the instrument breaks it into twenty separate clumps of weight (it calls them islands), and it finds no king among them: the weight is spread almost evenly across ten active clumps rather than piled onto one. In its own vocabulary the picture comes back “counterweighted,” a composition held up by opposing weights that balance rather than by a single dominant mass, and it files it in the family it labels gradient-field / controlled instability, distributed: built from soft gradients, tense but not chaotic, its weight spread out rather than concentrated [stable]. Put plainly, this picture is as busy as a canvas that has dissolved into pure texture, and yet it has not dissolved at all. Two regions act as load-bearing counterweights. The weight pools into the lower-left, onto the ploughman's ground (the left half of the picture carries 41% of the total mass against the right's 28%, the bottom 39% against the top's 25%), and nearly three-quarters of the frame, the whole sweep of sea and sky, is near-empty space that the loaded band leans across.

What makes that worth a section is that it is genuinely new. The instrument has a small museum of prior cases to compare against, and this fits none of them. It is not the Degas, where an all-over weave melts every anchor into one even, shadowy field. It is not the Pollock, uniform everywhere by design. And it is not even the Caravaggio, which tames a busy scene the older way the instrument already knew, by lighting one figure far above the rest so a single anchor wins by a wide margin. Bruegel does something the corpus had never shown: dozens of separately vivid incidents, none of them dominant, held together not by a hierarchy of finish (some passages loud, others hushed) but by pure arrangement, weights that balance and planes that stack. Call it distributed order, and note that the instrument arrived at it unprompted [stable].

The skeleton: an anchor, and a triangle over emptiness

The picture's structural spine is the ship, not any person

Among those twenty clumps of weight, the instrument singles out the heaviest and most connected one, the piece the rest of the composition hangs from, and it is not a human being. It is the great ship at the right: the picture's structural anchor is a boat [stable]. The other load-bearing clumps, the ones that hold the balance taut, are the pale cliff and harbour town at upper left, the ploughman's dark spur of headland, and the band of sheep. Draw a line between the heavy ones and they make a triangle, cliff to ship to plough, thrown across the wide emptiness of the water. The corner where Icarus is drowning hangs off the edge of this network as what the instrument calls a satellite: attached to the structure, out at its rim, bearing none of its load.

Lay a three-by-three grid over the picture and the imbalance turns into plain arithmetic. The bottom-right cell, Icarus's cell, is the emptiest of all the foreground squares (it holds 10.2% of the weight, and is nearly half void), while the three squares along the lower left carry 14 to 15% apiece [stable]. The composition does not simply forget to feature its subject; it stacks its weight along the diagonal that runs away from him.

And it holds that weight without any rigid scaffolding of lines. It is worth explaining what was tested here, because the test had to be corrected in the doing. Renaissance perspective works by ruled lines, the edges of floors and buildings, that all rush toward a single vanishing point; find that point and you have found the picture's geometric cage. The instrument hunts for such a cage by taking the painting's longest firm lines and asking where they cross. In Bruegel the roughly twenty long lines it finds, the plough furrows, the hill slopes, the ship's rigging, lean mostly on the diagonal, and crucially they do not meet at a point: the place where they cross drifts across a third of the picture depending on exactly how long a line has to be to count. To be sure that was the painting and not a flaw in the method, the same tool was run on a textbook one-point perspective work as a control, Perugino's Christ Handing the Keys. And that control had to be read carefully, because it exposed a limit of the tool: its lines do meet, but every one of the 19 crossings is a horizontal line meeting a vertical one, the corners of the temple's tiled facade, not the receding lines that actually define the vanishing point (those never survive as long firm lines at all). So the honest statement is narrower than “the tool can find vanishing points.” What it can do is register a stable grid where one exists: on the Perugino the crossing-point holds its height to within 0.003 of the frame across repeated runs, while on the Bruegel nothing holds still. The line directions confirm the same contrast from the other side. Because the measurement is a little different each run, the essay takes it as the record does, from a frozen batch of 40 runs: on the Bruegel the lines are diagonal about 36% of the time (median) and level only about 5%, roughly 7× more diagonal than level; on the Perugino they are diagonal 0% of the time in every single run and level about 91%. The two pictures are exact opposites, and the absence of a ruled cage belongs to the Bruegel [narrowed: one control; the tool is shown to catch a grid where it exists, not to locate vanishing points]. This is not a discovery so much as a confirmation of what art history already says: Bruegel's kind of “world-landscape” was never built on Renaissance perspective, and the instrument recovers that fact on its own. The eye is led here locally, by the plough's diagonal and the dark screen of trees at the left, then handed out across the emptiness to the ship. Order without a skeleton of lines: the same distributed logic the weights and the tonal planes carry, showing up once more in the directions of the brushwork.

The buried subject

Painted at full strength, given the room of a background detail

This is the measurement the whole study was built to make. For each named thing in the picture, the instrument reads four qualities over the object itself rather than over a loose rectangle around it (a box drawn round the legs would mostly measure water; the lab book documents how the reading is kept honest). The four are the ones introduced earlier plus two of their ingredients: structural weight, the sharpness of the edges, the strength of the colour, and saliency, the pull on the eye. Compare Icarus to the ploughman on these, point for point, and the easy version of the “buried subject” story falls apart at once [stable, jitter-checked]:

Point for point, Icarus is not marginalized at all. The splash is a crisp, locally vivid mark: harder-edged than the ploughman, brighter to the eye-pull model, and never more than a fifth softer in weight. Because those margins are small and the reading depends on exactly where the measuring box is placed, both boxes were nudged around a grid of positions, scaled to each box's own size (so the small legs and the big ploughman are perturbed by comparable amounts, not by a fixed pixel count that would jostle the legs far harder), and re-read at every position. The edge advantage survives in 24 of 25 of them, the eye-pull advantage in 20 of 25. The disaster, in short, is not painted quietly. A reading that stopped at strength would announce that Icarus is not marginalized, and it would be wrong.

The marginalization is entirely a matter of size and place. There is an obvious objection to that, and it is the right one: a pair of half-sunk legs is smaller than a man with a horse and a plough in any painting that holds both, so the bare fact that the ploughman carries twenty-five times more weight is partly just a fact about legs. The way to control for it is to compare Icarus not to one big figure but to the whole cast, ranked by how much of the frame each one occupies. Do that, and Icarus lands at the bottom of the middle rank [stable]:

In its own numbers, then: Icarus takes up 6.8–35× less of the frame than any of the large foreground figures, sits at the bottom of the middle rank alongside the fisherman, and still stands an order of magnitude above the true background specks, the distant town, the partridge, the sun. He is not a speck on the horizon, and he is not one of the leads: he is treated as the least of the minor incidents. Two honest caveats come with the table. The grouping into “leads, incidents, specks” is a human judgement (the instrument supplied the raw frame-shares; a person sorted them into three bands), and the close call in that sorting is the fisherman: promote that small worked figure from incident to lead and the “less than any lead” gap narrows from 6.8–35× to 2.1×. What holds no matter how the bands are drawn is the plain pair of facts: Icarus's share of the frame is 6.8–35× below the five large figures and 2.1× below even the smallest other human figure, and he is rendered at full local strength in the emptiest corner of the foreground.

There is a second thing the picture buries, and it needs stating with care because the numbers can be over-read. In the myth the sun is the cause of the fall: it melts the wax that holds the wings. Bruegel sets it low on the horizon rather than high (already a puzzle for scholars, since Ovid's sun is at noon), and of all the named objects it comes back the faintest, last on eye-pull, last on edge, last on weight, present only as a small warm patch of colour. It is tempting to call the agent of the tragedy its least-present object and stop. But a soft, low-contrast disc against a soft hazy sky would score low on those measures in any painting whatever, so the reading was checked against a baseline: the sun's eye-pull and edge come out almost identical to plain patches of the open sea and the sky, and a full order of magnitude below any worked surface. The low scores are the physics of a hazy disc, not a compositional act the instrument can claim. What does survive is modest and honest: the picture gives the myth's cause no structural presence beyond a warm note in the empty upper register. And even that is a fact about this canvas only, because the sun is one of the two things that differ between the two surviving versions (the other panel shows it overhead), so it cannot belong to the shared underlying design [narrowed: geometry-confounded, version-specific].

Light and dark: how the deep space is built

Not one graded wash but a stack of distinct tonal planes

Measured as a whole, the spread of lights and darks is unremarkable: mostly midtones, a good deal of shadow, little bright highlight. What is striking is how differently lit the four quarters of the picture are from one another. The instrument scores this plane-to-plane difference and it comes back at the ceiling of the scale (the maximum, 1.0) [stable]: the upper-right quarter, sun and sky, is pale and high-keyed, while the lower-left quarter, the ploughed earth, is deep and dark, and the two other quarters sit in between. This is how the picture opens its enormous depth, not with a single smooth wash from dark foreground to bright distance but by stacking separately keyed planes one behind another. Of the five paintings in the comparison set, Bruegel is the extreme case of this kind of local-versus-global contrast, standing clearly apart from the other four. And it is worth noting where Icarus lands in that scheme: he is down in the dark near-plane, while the blaze of light he supposedly fell out of sits a whole tonal world away, on the far diagonal.

Colour: a servant of the drawing, pooled in the empty spaces

Colour traces the forms rather than composing on its own

One useful question to ask of any painting is whether its colour does independent work, laid down in places the drawing ignores, or whether it simply fills in and reinforces shapes the drawing has already defined. The instrument answers it by checking how often strong colour and strong edges fall in the same places. In Bruegel they coincide: colour here is not an autonomous layer with its own agenda, it rides the drawing and describes objects [stable]. That is a distinctly pre-modern habit, and the comparison set shows it: of the five paintings, only Bruegel and the Caravaggio bind colour to edge this way, while the three modern works, Degas, Pollock, Klimt, all let colour come loose from the drawing.

There is one apparent exception that turns out to be a mirage, and it is worth airing because it is the kind of number that can mislead. By one scalar measure, this picture has a lot of colour sitting in its empty areas, where there is no drawing to speak of, which might suggest colour doing something on its own after all. But when the picture is scrambled a hundred different ways and re-measured, that same number barely moves: it reflects the plain fact that the picture contains large soft expanses of coloured sea and sky, not any deliberate arrangement of them, so it earns no interpretation [narrowed, I-15, null-checked at N=100]. What is real, and visible in the colour map rather than in any single number, is where the colour actually lives. The most saturated surfaces in the whole painting are the teal water and the golden sky, the very emptiness the loaded band of figures leans across; the figures themselves are drawn dark against them. The colour gathers in the void.

The punctuation: many small accents, no single exclamation

Nowhere does one mark shout above the rest

“Accents” here means the small, sharp marks that punctuate a surface, the equivalent of stresses in a sentence. The instrument looks for them two ways. Its default tool, which hunts for marks that stand out by contrast, finds none at all, and that null is informative: on a surface this evenly and thoroughly worked, no single spot is enough of an outlier to clear the bar (the Degas taught this lesson the hard way, and here it is applied on purpose rather than stumbled into) [narrowed, I-5]. Its second tool, tuned to catches of pure colour at brush scale, finds the real system: about seventeen small colour-accents, and they are scattered everywhere, the ploughman's red sleeve (the single most saturated figure-surface in the picture), the shepherd's blue, the flock, the sails, the far town. It is a punctuation with no exclamation point anywhere: the field of small incidents that makes the picture feel busy, spread evenly from edge to edge rather than gathered into a climax [stable]. (The count is capped by the tool and is not a tally of every fleck; the finding is the spread, not the number.)

Attention: where the picture sends the eye

The one measurement that speaks directly to Auden's poem

This is the section nearest to what the poem is actually about. Recall the two maps the instrument keeps: structural weight, where the picture presses hardest, and saliency, where a first glance is pulled. In most scenes those two land in roughly the same place. Here they come apart, and, more to the point, the eye-pull map settles down-left on the ploughman's side of the picture, not on the corner where Icarus is drowning. The obvious worry is that such a result could be an accident of the palette rather than the composition, so it was put through a stiff test: the picture was scrambled into a hundred versions that keep its colours and textures but destroy its arrangement, then a hundred more scrambled a second way, and the real painting's eye-away-from-corner effect was stronger than every single one of the 200 scrambles. It is a property of how the picture is arranged, not of what colours it happens to use [stable, null-checked at N=100]. Auden's line is that the human witnesses do not look at Icarus. The instrument cannot read a painted shepherd's gaze, that requires knowing a face is a face, which is past its limit, so this essay's remarks about the shepherd staring at empty sky are things one sees, not things measured. But one level up from the figures' eyes, the picture's own machinery of attention agrees with the poem: the composition steers the viewer's eye away from the disaster, and it does so demonstrably.

There is a sharper twist when this is set against the earlier studies. In the Matisse and the Klimt, the gap between where weight sits and where the eye goes was the very thing that pointed at a hidden or displaced subject, the tension flagging the buried figure. Here there is no such tension: weight and eye sit on the same side, the ploughman's, and together they simply leave Icarus out. The same instrument that once located a concealed spectator in a Degas by that disagreement finds Bruegel's burial performed the opposite way, by the agreement of the two channels rather than their conflict [stable].

How the calm is made

A stillness assembled out of small pulls that cancel

The instrument can also measure a picture's overall “lean” or twist, how much the whole arrangement pulls off to one side or rotates rather than sitting in balance. On this measure Bruegel is the calmest of the five compared paintings by a wide margin, far quieter than the turbulent Degas or the tilted Klimt: as a whole, the picture is nearly at rest. The interesting part is how that rest is produced, because it is not the flatness of an empty design. Chop the picture into halves and quarters and each piece, taken alone, leans harder than the whole does [aid, candidate register n=8]. The plough's diagonal, the ship's list, each little incident's own asymmetry, all of these pull, but they pull in different directions and very nearly cancel out into the global calm. (The Klimt was the exact opposite kind of picture, where the whole leans more than any of its parts.) And the busyness has a beat to it: the instrument rates this the most rhythmically self-repeating of the five, the pulse of sheep, ships and waves recurring at every scale. Auden's word “leisurely” gets a mechanical gloss here, then: not one large calm but many small urgencies summing to none [narrowed: this “cancellation” reading is a candidate measure, not a settled one; the register note keeps it honest].

Depth: recession that the instrument can only half-see

Each cue to distance is measurable; binding them into “depth” is not

The instrument does not perceive depth as such, by design: it sees a flat array of measurements and has no idea that one patch is meant to lie behind another. This painting is a fair test of just how far that limit goes, because its depth is the gentle, atmospheric kind, a whole landscape receding to a hazy horizon. Sample the open water at four distances from near to far and the textbook recipe for distance, things going paler, cooler and softer as they recede, holds only across the far half of the run; the near end refuses to follow, because it is overridden by local facts the instrument cannot discount, a corner in shadow and the churned wake behind the ship [narrowed, n=4 bands]. The one distance-cue that comes through cleanly is sheer shrinkage: take the same kind of object, a sailing ship, at three depths, and its area drops 27× from nearest to farthest [stable]. So depth in this picture is really several separate devices working together, the atmospheric fade, the stack of tonal planes from the last section, and this steady shrinking of like things. Each one is measurable on its own. What the instrument cannot do is fuse them into a single felt sense of space, because that fusion requires knowing that a dark corner is shadow and a white streak is spray, which is a judgement about meaning, not measurement. The old limit, that the instrument is depth-blind, is now stated as a mechanism rather than a shrug.

The two versions: one design, two objects

Different paint, the same underlying composition

Recall that this Brussels canvas is contested as a copy, and that a second old version survives in another museum, the Van Buuren panel, which still shows Daedalus aloft in the sky. The instrument can weigh in on that puzzle, as far as an honest global tool can, by measuring how structurally alike two images are, boiling each down to a compact signature of its composition and comparing. Three reproductions went in: the pinned Brussels capture; a casual phone photograph of the same Brussels canvas, kept on purpose as a control (it tells us how much difference a mere change of camera and lighting produces, with the object held fixed); and the Van Buuren panel. The distances come out like this, where a smaller number means more structurally alike:

Read plainly: the two versions are structurally the same composition. Their distance, 0.079, sits inside the range the tool calls “same work,” only a shade more than the 0.06 that a mere change of camera produces on the identical canvas, and nowhere near the much larger gap that separates genuinely different paintings. Three honest qualifiers ride along. First, the “same work” threshold is itself a rough calibration from a handful of examples, so 0.079 clears it by a thin margin, not comfortably. Second, the Van Buuren photo is lower-resolution than the Brussels pin, so the comparison was repeated with the Brussels images shrunk to match; the versions still read 0.079, so the result is not an artifact of resolution. Third, and most honestly, at that matched resolution the neat layering breaks down, the different-object pair falls to the same 0.057 floor as the same-object control, so the tidy story that “the object difference adds a measurable notch above mere camera noise” holds only at the original scale and is flagged as such [stable, with the layering caveat]. (A second, cruder byte-level similarity check returns nothing for any pair, as expected for photographs of different physical objects; the kinship claim rests on the compositional signature alone.)

The scholarship supplies the frame the numbers sit in. Dendrochronology dates the Van Buuren panel to 1577 at the earliest, eight years after Bruegel's death, so that version is definitionally not by the Elder; and the technical literature (Currie and Allart's Brueg(h)el Phenomenon, 2012) treats both versions as copies of a lost original, with which copy is more faithful an open question. That is exactly the situation "one skeleton, two objects" describes, and the per-axis version of this delta (which channels each version deviates on) is the obvious next instrument contribution, logged as an open thread. No hand is attributed. One reading the literature makes, worth setting beside the numbers: in the Brussels picture the shepherd stares up at empty sky, and the Van Buuren panel, with Daedalus still aloft, shows what that gaze was once for.

Optional interpretation

Clearly marked, and one sentence long

Everything above is measurement; this paragraph is reading. The composition executes the proverb it was long read against (no plough stands still for a man who dies) not by staging indifference as darkness or distance but by a harder device: rendering the catastrophe with exactly the same care, the same local commitment, as the ploughing, the grazing and the sailing, and then giving it the room of a background detail and the weight of none. If the fields measured anything worth a sentence of interpretation it is this: the picture's cruelty is bookkeeping, an allocation of frame, and the allocation is precise. Whether that precision belongs to Bruegel or to a faithful copyist, the skeleton carrying it is the same in both surviving versions.

What this does not prove

• Not intention. The allocation of mass and room is measured; that it was "meant" is a human reading, marked as such above.
• Not authorship. The delta reads one shared skeleton at same-work distance (upper edge of a provisionally calibrated band); it cannot attribute a hand, the Van Buuren panel is dendro-dated 1577+ (not the Elder's), the chain-vs-object layering holds only at the pinned scale, and the Van Buuren reproduction is 1772px on an unknown chain.
• Condition chain. The Brussels surface is heavily overpainted after relinings (the transfer-from-panel reading is disputed; Currie and Allart 2012 discount it). Fine-surface numbers (edge means, facture, fragile-to-blur) partly measure restoration campaigns; the skeleton tier (islands, mass grid, tonal planes, counterweights) is predicted restoration-invariant, consistent with the phone-photo delta, but untested against documented restoration maps.
• Size is confounded; the finding routes around it. A pair of legs is small in any painting. The corrected claim: 6.8–35× below every foreground actor, bottom of the incident tier with the fisherman (2.1×), an order of magnitude above the background details, at full intensity (edge higher in 24/25 normalized jitters, mass never dimmed beyond 1.2×).
• The tier taxonomy is hand-made. The instrument produced the frame-shares; a person sorted them into actors, incidents and details. Reclassify the fisherman as an actor and the "below every actor" floor is 2.1×, not 6.8×.
• Null verdicts are percentiles, at the test's 1300px scale (N=100 per null): mass_divergence 100th/100th. DGI 1st/0th, which is a positive result: nearly every null re-arrangement scores HIGHER default-gravity than the real picture, i.e. scrambling pulls the work toward the generic centred basin, so its off-centre arrangement is compositional. β below both nulls but by ~0.6% relative, detectable and compositionally negligible: the colour-in-void reading stays the map's.
• The armature control validates grid detection, not VP detection. All the control's intersections are horizontal × vertical facade crossings, so "the estimator can find convergence" was never shown; what stands is a stable rectilinear structure on the Albertian control (y locked to width 0.003 over 10 runs) against nothing locking here (y wandering 0.24–0.61, span 0.37). One control, orientation bins disclosed in the record; the claim is no rectilinear structure (the documented Weltlandschaft expectation, recovered independently), never a located node.
• The sun finding is geometry-scoped and Brussels-only: its low saliency/edge match the picture's other smooth regions, and the sun differs between versions.
• The witnesses' gaze is not measured. Figural facing and gaze are the pre-semantic ceiling (C-4); the essay's shepherd-stares-at-empty-sky is observation. What is measured is the composition's saliency field (Attention), which is a different, weaker, but real correlate.
• R_spatial is a candidate (C-3, register n=8), and its subadditive point-verdict does not classify composition type; the claim used here is the cancellation reading plus the low measured torque, together.
• The depth probe is a proxy battery: four hand-verified bands, Spearman on ranks, directional not significant; "depth" as a bound percept is explicitly out of reach (C-1).
• One object's reproductions, pinned scales: all numbers from bruegel_working_2600.jpg; D values quoted only from their swept stable windows. The compare is placement, not measurement: at n=5 the maximum |z| is √(n−1)=2.000, so only ranks are read, never magnitudes or significance.
• Accent counts are lens-capped and are not abundance measures; the claim is channel segregation and distribution.

Sources & record

• Pieter Bruegel the Elder (attributed), Landscape with the Fall of Icarus, oil on canvas, Royal Museums of Fine Arts of Belgium, Brussels. Source image: the Google Arts & Culture / RMFAB capture (12265×8000), public domain via Wikimedia Commons. The attribution and dating debate (early copy after a lost original, c. 1558) is the museum's and the literature's, reported here, not adjudicated.
• The Van Buuren Museum version (oil on panel, with Daedalus); Wikimedia Commons reproduction, 1772px. The museum phone photograph of the Brussels canvas: Wikimedia Commons, frame-cropped for the control.
• Ovid, Metamorphoses VIII (the ploughman, shepherd and fisherman as witnesses).
• W. H. Auden, Musée des Beaux Arts (1938); William Carlos Williams, Landscape with the Fall of Icarus (1960).
• Rudolf Arnheim, Art and Visual Perception (California, 1954/1974): perceptual weight, centric/eccentric composition.
• Christina Currie & Dominique Allart, The Brueg(h)el Phenomenon: Paintings by Pieter Bruegel the Elder and Pieter Brueghel the Younger (Brussels, 2012): both versions as copies of a lost original; the panel-transfer reading discounted; dendrochronology dating the Van Buuren panel 1577 at the earliest.
• Positive-control image: Perugino, Christ Handing the Keys to St Peter (Sistine Chapel), public-domain Wikimedia reproduction at the matched working scale.
• Study record: bruegel_labbook_standalone.html (Entries 01–20; labbook.md is the markdown twin), report_nopersist.json, icarus_probes.py/.json, jitter_probe.py/.json, null_percentiles.py/.json, depth_probe.py/.json, resolution_sweep.py/.json, delta_versions.json + _matched, sun_baseline.py/.json, armature_probe.py + armature_intersections.json + armature_node_spread.json, compare_out/. Instrument: the ISR library (findings I-1…I-15, candidates C-1…C-4); prior studies of Matisse, Degas, Pollock, Caravaggio, Soejima, Klimt; papers at parallaxmetrology.com.
• Instrument lineage: the ISR library’s Matisse, Degas, Pollock, Caravaggio, Soejima, Klimt, Bruegel, Cartier-Bresson, Sora Pair, MidJourney Ensemble, Velazquez, and Malevich