Vermeer Did Not Just Copy a Camera. The Paint Layers Say So.
Three technical studies show Vermeer built his pictures in revised layers of black underpaint, lead white and ultramarine. That ends the "effortless copying" tale. It does not end the camera debate.
Four Layers of Paint, One Overworked Theory. That is the mock headline. Here is the serious one: the paint in Vermeer's pictures does not support the tale that he effortlessly copied a projected image. It also does not prove he never used a camera obscura. I came to this post expecting the first claim to be enough. The sources made me narrow it.
The question
The popular tale says Vermeer put a camera obscura in his room and traced what he saw. The tale is old. Lawrence Gowing argued from X-rays of Girl with a Pearl Earring that the pearl's round highlight could come from projected light, and Philip Steadman later proposed the device as "a constant reference" while Vermeer painted [1]. Steadman measured maps and floor tiles in the pictures to link them to real room sizes [2].
The question: what does the paint itself, read layer by layer, say about that tale? I hold that technical studies change the standard story of a master's method more often than critics expect. This is a test of that view on one master.
Data and where it came from
I read published material on three paintings. I did not open every full paper. Some journal pages blocked my fetch tool, so for some findings I rely on abstracts, search summaries and museum or catalogue pages. I mark those below.
- Girl with a Pearl Earring, c. 1665, oil on canvas, 44.5 by 39 cm, Mauritshuis. The 2018 "Girl in the Spotlight" project used infrared reflectography, macro X-ray fluorescence and microscopy on the whole painting, and microsamples [3][4][5].
- The Milkmaid, c. 1657 to 1658, oil on canvas, 46 by 41 cm, Rijksmuseum. I used the ColourLex pigment record [6].
- The Art of Painting (Vienna) and the four Vermeers at the National Gallery of Art in Washington. The NGA team wrote up A Lady Writing, Woman Holding a Balance, Girl with the Red Hat and Girl with a Flute [7]. For Vienna I rely on the NGA article and a catalogue summary, not the 2010 Vienna study itself [7][8].
For the camera side I read Essential Vermeer's survey of the debate, a Barnebys summary, and a Counting Vermeer page on X-rays [2][9][10].
Method
I did not run code. I made one hand tally. For each painting I ask whether a published finding fits a simple copying model, where a projected image is traced once and then filled in. The model predicts a few things: a plan fixed early, little revision, and paint that follows the outline. A builder model predicts the opposite: revision, shifting underlayers and colour made by stacking layers. I mark each finding as fits copying, fits building, or neutral. This is a reading of text, not a measurement. Another reader may score some rows differently.
What the paint shows
How it was built: the Girl's blue scarf
Here is the "how it was built" paragraph. The canvas of Girl with a Pearl Earring had a warm light grey ground. Vermeer drew fine black outlines around the figure and laid in light and shade on top [11]. Pigments were limited: red, brown and yellow ochre, natural ultramarine, indigo, lead white and the rare yellow dye weld [4]. In the blue headscarf, a thick blue layer, up to 70 micrometres, sits on a black underlayer on the shadow side. On the lit left side the blue sits directly on the grey ground [3]. The blue is natural ultramarine, ground from lapis lazuli, mixed with varying amounts of lead white [5]. The yellow part of the scarf uses yellow ochre with lead white, and the border adds ultramarine and lead-tin yellow [5].
So the blue is not one blue. It is two different builds, one for the light side and one for the shadow side. That is the first thing the "blue is simply blue" tale misses.
The background is another surprise. The dark space behind the Girl was not flat black. It has a black base with a later layer holding weld, chalk, red ochre and indigo [4]. The 2018 work found that Vermeer modelled diagonal fold lines on the right by varying the lead white in a semi-translucent green glaze. The indigo and weld in that glaze have faded [3]. The picture looked different when it was new.
The Milkmaid and the Art of Painting
The Milkmaid tells the same story. The ground holds chalk, lead white and umber. The tunic has black undermodelling that extends beyond the blue skirt, then a white underpainting, then the blue itself. Blues are ultramarine with lead white, often glazed again with ultramarine. The bodice is lead-tin yellow. The red skirt is madder lake and lead white over charcoal black [6].
In The Art of Painting, a catalogue summary reports fine black underdrawing at the chandelier arms and the map frames, and a copper-based underpaint covered by a semi-transparent ultramarine layer on the green-blue drape [8]. The NGA authors add that this painting shows summary white chalk notations on the unfinished easel canvas [7].
The NGA finding that matters most
The NGA team found "no consistent evidence of underdrawing" in its four paintings. It found a monochrome brown sketch, then coloured underpaint that was "surprisingly free and robust": coarse pigments, bold strokes, strong light and shadow, thick lead-tin yellow highlights. They also found Vermeer used copper driers to speed drying, and that he "resketched": he tested other hair arrangements and moved a quill between underpaint and final layers [7].
A Counting Vermeer page agrees from the X-ray side. It says a textured underpaint under a smooth top layer is a characteristic Vermeer feature, and that he "carefully constructed his compositions during the painting process by changing, adding and removing elements." The same page notes that the large Art of Painting shows no such changes, which suggests advance planning for a big work [10].
The tally
| Finding | Painting | Fits copying | Fits building | Neutral |
|---|---|---|---|---|
| Blue built differently on lit and shadow sides | Pearl Earring [3] | x | ||
| Glaze modelled with varying lead white | Pearl Earring [3] | x | ||
| Black undermodelling, white layer, then blue | Milkmaid [6] | x | ||
| Free, coarse underpaint, then smooth finish | NGA four [7] | x | ||
| Resketching, quill moved | A Lady Writing [7] | x | ||
| No consistent underdrawing | NGA four [7] | x | ||
| Few changes in a large planned picture | Art of Painting [10] | x | ||
| Round highlights "could have derived" from a camera image | NGA four [7] | x |
Result: 5 of 8 rows fit building, 2 fit copying, 1 is neutral. I do not trust the ratio. The rows are not independent, and I picked them. The direction is clearer than the number.
What the result does and does not show
Fact. The paint evidence supports a painter who revised, stacked and adjusted. It does not support "effortless" copying. The 2018 team found a careful order of work, from background to foreground, with the greenish glaze the first layer on the underlayers [11].
Dispute. Does a layered method rule out a camera? Not by itself. A camera image can only give a pattern of light and shade. Vermeer could use it for the first brown sketch and then build the rest by hand. The NGA authors say as much: the round highlights could come from a camera image, but he used them selectively on unreflective surfaces, which looks like adaptation and not mechanical copying [7].
Two critics disagree here, and both have a case. Steadman argued that geometry in the pictures matches projection, and about half of leading Vermeer specialists accept his tracing idea, according to Essential Vermeer [2][9]. Walter Liedtke wrote that the camera's role "may have been inspired by an interest in the camera obscura, but its importance to the artist has been greatly exaggerated" [9]. Alan Chong asks why Vermeer would plot perspective by pinpricks if he were tracing [9]. The Counting Vermeer page adds that pinholes at the vanishing point in seventeen paintings show a mechanical method for perspective lines, with a pin and chalked cord. That is a workshop method, not an optical one [10]. No camera is listed in his estate, and no document says he used one [9].
I keep a tally of critics who changed their minds. Arthur Wheelock Jr. goes in it: Essential Vermeer says he backtracked and argued that Vermeer "persistently departed from what he actually saw" [2].
This part is taste. I find the pure-copying story dull, and the builder story better. A man who moves a quill between underpaint and finish is making decisions, not tracing. But taste is not evidence. A tracer can also make decisions.
Sensitivity: which assumption moves the result most
The biggest assumption is what "the camera story" means. Under the strong reading (a traced image gives most of the picture, with little rework), the paint evidence cuts against it: 5 of my 8 rows. Under the weak reading (a camera gave a tonal start for some pictures), the paint evidence is nearly silent. The verdict flips on this one definition, more than on any pigment fact.
Second assumption: the sample. Three paintings plus the NGA four is a small part of the surviving work, and the studies differ in method. Some paintings (such as The Art of Painting) look planned, others look reworked. A different set could change the tally.
Third: my reading of secondary summaries. I saw several findings in abstracts and catalogue pages, not the full papers. Where a catalogue summary and a paper disagree, the paper wins.
My current view
My position that technical studies change the standard story survives, but narrower than I first wrote. They changed the story of the method (effortless, optical, one pass). They did not settle the camera question. I hold this at about 0.65, unchanged, because the evidence here is both for and against the claim that critics underrate technical studies. The studies overturned "effortless". They left "some optical help" open.
The detail I would look at first in the gallery: the right edge of the headscarf, where blue sits on black. If the shadow side looks built and the lit side looks direct, the picture is already arguing with the camera tale.