Same Fire, Different Green

Remington's night skies are the most varied material in the plate set. Sampled in two of them, one measures Δu′v′ 0.0085 from D65 — almost a neutral grey, with only a faint blue-violet lean — and the other 0.0479, a definite deep blue. Between two paintings the same material is five times more coloured in one than the other, and both readings are far above the reproduction floor, so that spread is Remington's, not the printer's.

First Cow has one clear open night sky in its 466 photographed frames: a star field above a treeline, shot 307. Measured, it returns Δu′v′ 0.2552 at a direction of 263 degrees — an enormous, deep blue. That number is not a fact about the film. The region's mean 8-bit code value is 5.9 out of 255, and the whole patch is described by 45 distinct levels. At that level the encode's own quantisation is larger than the colour being measured, and dividing by the frame's very small diffuse white multiplies the error along with the signal.

What can be said is narrower and still worth saying. The direction of the film's night sky, 263 degrees, falls between the two Remington skies measured here, at 252 and 285 degrees. Both media put their night sky in the blue-violet quadrant, which is the least surprising agreement in this report — it is where night skies go. Whether the film's is as deep as Remington's deepest, or as nearly neutral as his palest, cannot be settled from an 8-bit distribution encode. It would take the master.

Both bodies of work spend a great deal of time under flat cloud, and both keep that cloud close to grey. The difference is which way each of them leans off grey, and the answer is: opposite ways.

Remington's overcast, sampled above the snow slope, sits Δu′v′ 0.0134 from D65 at a direction of 16 degrees — a warm, faintly pink-grey. First Cow's overcast sits at 0.0127 and 0.0249 in the two frames sampled, at directions of 187 and 213 degrees. The magnitudes are almost identical; the directions are 171 and 162 degrees apart, which is as close to exactly opposite as two directions get.

The separation between them, Δu′v′ 0.026 and 0.038, is two and three times the reproduction floor, so this is a real difference and not an artefact of printing. Corpus-wide the same thing shows up: across Remington's nocturne plates the sky band's mean direction is 6 degrees, warm; across the film it is 167 degrees, cyan. The film's sky is not a desaturated version of Remington's sky. It is the other side of the wheel.

The control run is unusually clean here: not one of the 39 corpora measured falls inside the reproduction floor of Remington's nocturne sky, including his own daylight plates, which sit 0.0240 away. His skies are too varied for a corpus median to be close to anything. First Cow ranks 22nd of 39, at Δu′v′ 0.0473. The claim defended here is the direction, not the rank: whatever else is true of Remington's skies, they do not lean cyan, and the film's do.

This is the material the film is made of and the one Remington uses least: the dark green-to-cyan rule finds a qualifying mass in 88% of First Cow's frames and in 12% of the nocturne plates. Remington's nocturnes are open country; First Cow is inside a wood for most of its length. Any comparison here is between a corpus staple and a corpus exception.

Where both have it, the direction is broadly shared and the amount is not. Remington's unlit timber measures Δu′v′ 0.0239 and 0.0163 in the two printings of one painting; the film's fern and moss measure 0.0399 and 0.0667 in the two frames sampled. Across the corpora the same gap holds: the plate median is 0.0181 against the film's 0.0307. First Cow's unlit green is roughly 1.7 times further from grey than Remington's.

The directions differ too, and by more than the floor: 101 degrees across the plates against 136 degrees across the film. Remington's dark green leans toward yellow; the film's leans toward cyan. Of the seven other films checked here that have enough dark green to measure, First Cow's is the most cyan of all, which is a point about this grade rather than about forests.

Against controls the film's green is not especially Remington-like either. Measured the same way, First Cow is 23rd of 38 corpora for closeness to Remington's unlit green, outside the reproduction floor, while 9 corpora sit inside it — Point Blank, Gangubai Kathiawadi, Leviathan, 1917 and Let the Right One In among them. The film's forest is further from Remington's than most films' shadows are, and it is further in a specific direction: toward cyan, and away from grey.

The honest qualifier is large. This exact rectangle is the least stable in the whole reproduction control: the two printings of the same Remington painting disagree by Δu′v′ 0.0118 and 26 degrees of direction, which is about three quarters of the corpus-level hue gap between the plates and the film. The corpus medians are built from many images and are therefore steadier than any single pair, but a reader who wants to see this difference in one painting and one frame will be looking at something a printer could have caused.

There is no snow in First Cow. All 466 photographed frames were reviewed at thumbnail size and none contains any. The film is set in the Oregon wet season and its weather is rain, river mist and mud. Snow is Remington's single most distinctive material — it appears in roughly a third of the plate set and it is what most of his best-known nocturnes are made of — and the film simply does not have it. That asymmetry is the largest single fact in this report and no measurement can repair it.

What can be done is a substitution test, stated as such. If the question is what happens to the picture's largest pale, low-chroma field — the material that sets the ceiling and carries the light — then Remington's is snow and First Cow's is overcast sky and the river under it. Both were sampled and compared, not as the same material, but as the same job in the composition.

They land in almost the same place. Remington's sunlit snow measures Δu′v′ 0.0114 from D65 and his moonlit snow 0.0155; First Cow's overcast sky measures 0.0127 and 0.0249. The separation between the film's fort sky and Remington's sunlit snow is Δu′v′ 0.0082, and against his moonlit snow 0.0053 — both below the reproduction floor of 0.01179. On the chromaticity plane, this film's pale field and this painter's snow are indistinguishable at the resolution this method has.

The corpus-wide version agrees. The bright low-chroma field rule finds a qualifying field in 38% of the nocturne plates at a median Δu′v′ of 0.0186, and in 23% of the film's frames at 0.0166. The directions are 112 and 136 degrees, which is inside the floor as well.

That agreement is real and it is not distinctive. On the same measurement across every corpus here, First Cow is 15th of 39 for closeness to Remington's large pale field, and 21 corpora fall inside the reproduction floor. The closest of all is The Assassination of Jesse James by the Coward Robert Ford, at Δu′v′ 0.0011 — which is what one would expect from a film with real snow in it, and is the clearest illustration in this report of how little a single material agreement proves.

The lightness is a different story and cannot be compared honestly. Remington's snow reaches a median L* of 82 after exposure matching; the film's pale field reaches 57. In the delivered encode the gap is far wider still, 62 against 22. A painted snow field is the brightest thing in its picture by a long way, and First Cow's brightest thing is not.

This is the one place where the answer is yes, and it is the strongest agreement anywhere in the report.

The warm-light-core rule — the picture's own brightest, strongly warm pixels — was applied unchanged to both corpora, so the flame in a painting and the flame in a frame are selected by the same definition rather than by hand. Remington's campfire core measures Δu′v′ 0.0472 from D65 at a direction of 47 degrees; his second campfire 0.0583 at 44 degrees; First Cow's candle 0.0454 at 38 degrees. The separation between the film's flame and the nearer plate is Δu′v′ 0.0074 with 9 degrees of direction between them — inside the 0.01179 floor. Against the second plate it is 0.0140, marginally outside it.

Corpus-wide the same result holds and is better supported: 10 plates and 168 frames carry a qualifying core, at median Δu′v′ 0.0512 against 0.0449, in directions 45° against 38°. Both differences are inside the floor.

The conclusion needs care, because it cuts both ways. It is a real agreement: a painted campfire and a photographed candle put their hottest colour in the same place, which is not guaranteed — a grade that warmed its flames, or a painter who pushed them red, would show here immediately, and neither did. But it is also the least distinctive agreement available. Running the same measurement on every corpus in this repository puts First Cow 6th of 38 for closeness to Remington's flame colour, with 8 corpora inside the reproduction floor. Another Round, O Brother, Where Art Thou? and Portrait of a Lady on Fire are all closer. Fire is a black-body radiator, and everyone who photographs or paints one honestly lands in the same small region of the plane.

The flame is a small object; the light it throws is the picture. Sampling what each fire falls on separates the two questions, and they get different answers.

The direction survives. Remington's firelit ground measures a direction of 35 degrees and First Cow's candlelit boards 28 degrees — 7 degrees apart, comfortably inside the floor. Both put the warmth of a fire's falloff in the same orange.

The amount does not. Remington's firelit ground sits Δu′v′ 0.0535 from D65; the film's lit boards sit at 0.1404, and the separation between them is 0.0875 — seven times the reproduction floor and the second largest gap in this report. In the film, the light thrown by a candle is more strongly coloured than the flame itself: 0.1404 for the boards against 0.0454 for the core. In the painting the relationship runs the other way, 0.0535 for the ground against 0.0472 for the flame.

Both put more colour into the falloff than into the source, and that in itself is a shared instinct. The size of the effect is what separates them. In Remington's campfire the ground is 1.1 times further from grey than the flame — 0.0535 against 0.0472, a difference smaller than the reproduction floor. In First Cow the boards are 3.1 times further from grey than the candle — 0.1404 against 0.0454. Remington lets flame and firelit ground be nearly the same colour; the film keeps the flame core almost white, and its own five-cluster palette puts that core at a pale blue-white, then spends all the warmth on the wood. In a picture where nothing else is warm at all, that is where the warmth has to live.

A face across a fire in Remington measures Δu′v′ 0.0676 from D65 at a direction of 38 degrees. A face under lamplight in First Cow measures 0.1013 at 24 degrees. The separation is Δu′v′ 0.0395 with 14 degrees between the directions — a resolved difference in magnitude, and a direction difference of about half the floor, so the hue is shared and the intensity is not. The film's lit skin is half as far again from grey as the painting's, and leans further toward red.

The corpus-level version of this question could not be answered with the tools available. This repository's candidate-skin heuristic is designed for photographed faces; applied to the plates it selects 32% of the average painting, because dry grass, horsehide, deerskin and lamplit canvas all pass the same gate. On the film it selects 3%. The heuristic is not wrong so much as inapplicable, and the numbers it returns for the plates describe warm paint rather than skin. That is why this question is answered from two hand-placed, visually confirmed patches and not from the corpus.

The result is consistent with the falloff finding above and probably the same finding. Both media light a face with fire; the film puts more chroma into what the fire reaches.

Both bodies of work spend real time on water at low light: Remington's canoe nocturnes on still lakes, First Cow's opening river and the stream the men cross. The samples do not agree, and they do not agree in an interesting way.

Remington's lake below the canoe measures Δu′v′ 0.0333 at a direction of 279 degrees — a definite blue-violet, the same quadrant as his night skies, which is what a still lake reflecting a night sky should be. First Cow's river beside the paddle measures 0.0662 at 211 degrees, a strong cyan-blue; its broken river surface measures 0.0060 at 103 degrees, which is very nearly neutral. The separations from the plate are Δu′v′ 0.0621 and 0.0393, both far outside the floor.

The two film samples are 0.06 apart from each other, which is more than either is from most other things in this report. That is not a contradiction: a smooth deep river reflecting a teal sky and a broken riffle scattering white light are genuinely different materials, and the film treats them as such. Remington's water is more consistent because his water is nearly always the same thing — a flat reflecting surface at night.

The direction difference is the one worth noting. Remington's water goes violet; the film's goes cyan. That is the same split the sky question found, and for the same reason: the two put their ambient light on opposite sides of neutral.

Remington's dry foreground grass measures Δu′v′ 0.0566 from D65 at a direction of 44 degrees: a definite warm ochre, sitting in the same quadrant as his firelight. First Cow's ground measures 0.0193 at 86 degrees in the leaf litter of the prologue, and 0.0183 at 175 degrees in the fort mud — barely coloured, and in the second case leaning cyan. The separations are Δu′v′ 0.0443 and 0.0701, four and six times the reproduction floor.

This is the clearest single disagreement in the report and it is about ambient light rather than about ground. Remington's plains at night are lit by something he paints as warm, and his ground returns that warmth. First Cow's forest floor and fort mud are lit by overcast Oregon sky, and the grade lets them keep its coolness. The same split appears in the sky, in the water and here; three materials, one cause.

It also explains most of the whole-picture result. The largest areas in a Remington nocturne — ground, grass, snow, sky — carry his warm bias, and the largest areas in First Cow carry the film's cool one. That is what moves the two hue spectra 34 degrees apart while their fires sit on top of each other.