Romanian painting · 1840–1989
The colour of Romanian painting, year by year: 1,655 canvases and no clear drift that is Romania’s own
Every dated painting in the CIMEC register of classified heritage (a selection from Romanian museums) was reduced to five colours. Put in year order they form a ribbon with no clear direction: from 1840 to 1959 median saturation stays between 11.6 and 12.5, and lightness rises by 6.9 points between 1860–99 and 1920–39, about as much as in paintings from Sweden and Austria.
Your year’s palette
Type a year between 1840 and 1989 to get its palette, with a few paintings from that year, as a picture to send.
#What changed and what did not
The first test was written before any colour was computed: between 1860–99 (402 paintings) and 1920–39 (587 paintings), a change counts if it exceeds twice the noise between museums, if its confidence interval excludes zero, if it has the same sign in the two large museums and without any one of the five most numerous painters, and if the foreign paintings do not show the same change. Not one of the three measures (lightness, saturation, tint warmth) passes all five.
Lightness rises by 6.9 points, from 41.7 to 48.5, and the interval 4.8–8.5 excludes zero. But the noise between museums is 4.1, so the rise is 1.7 times the noise, under the threshold of 2. And it appears abroad as well: +10.9 at Nationalmuseum (787 paintings), +9.0 at Belvedere (1,177). In the National Museum of Art of Romania the rise is +2.8, in the Bucharest Municipality Museum +7.3: the direction is the same, the size very different.
The result depends on the noise chosen. With the spread of the Romanian museums (4.1) the rise is 1.7 times the noise; with that of the foreign museums (2.8) it is 2.4 times and would pass the first test, but Romanian paintings lightened less than those abroad (6.9 against 10.0), so the change is not Romania’s own. The plan written beforehand did not say how to combine the two noises; the larger was chosen, and the fifth test accepts a difference in either direction.
Saturation stayed at 12.5 (1860–99) and 12.3 (1920–39), a difference of −0.2, about 9 times smaller than the noise. Tint warmth (b*) falls by 2.2, about as much as in the foreign paintings (2.9) and at the edge of the noise (2.0).
By painter rather than by country there is no common direction: Theodor Aman and Nicolae Grigorescu get lighter by about five points a decade, Camil Ressu and Corneliu Baba get darker. The average rise may come from who is in the museum in which period, but the data cannot separate that from varnish or from how the paintings were photographed.
Median lightness by period
Median saturation by period
Solid line: paintings in Romanian museums (between 86 and 587 per period). Dotted lines: Nationalmuseum (Sweden) and Belvedere (Austria), shown only where they have at least 25 paintings.
| Measure | Over 2× the noise | Interval excludes zero | Same sign in both museums | Without the top 5 painters | Different from abroad |
|---|---|---|---|---|---|
| Lightness (L*) | no | yes | yes | yes | no |
| Saturation (C*) | no | no | no | no | yes |
| Tint warmth (b*) | no | yes | yes | yes | no |
Other cuts tried: portraits lighten by 10.2 points, landscapes by 6.3, still lifes have too few paintings (14 and 42) for a conclusion. By museum: 2.8 and 7.3. A cut by medium (oil or tempera) was not made, because the records have no medium field.
What was not tested
After 1959 there are almost no control paintings (Belvedere 0, Nationalmuseum 1 in 1960–89). The Romanian paintings of 1960–89 (113) have saturation 14.8 and tint warmth 12.2, higher than in the earlier periods, but this jump was not in the plan and has no control, so it stays an observation.
For the pigments they used, see Painters’ colours.
#Why it is not a discovery
Paintings are photographed under different lamps, in different campaigns, and a museum photographs its collection in one way. For that reason the colour of a single painting cannot be corrected, and the noise was measured between museums: how far apart, in the same period, the medians of museums with at least 25 paintings fall.
Two limits do not go away. Yellowed varnish darkens and warms old paintings whatever the painter chose. And the sample is what the museums classified, not all painting: 699 of the 1,655 paintings come from one museum.
#Questions
How was the colour of a painting chosen?
From the photograph the central rectangle is kept, 70% of each side, to drop frames and edges. Colour is computed in CIELAB. The five colours shown come from a k-means grouping of the pixels of the same area.
Which paintings are included?
Only those classed as easel painting, without icons, ceramics, drawings or objects, with an explicit year, painted in 1840–1989. Of 1,713 with an explicit year, 1,655 remain: 11 near-monochrome photographs, 7 duplicates and those outside the years were removed.
Why is there no pigment?
This measures the colour visible in the photograph. The pigments it was made from are the subject of the page on painters’ colours.
#How it was measured
Source: the CIMEC “Clasate” records, published through Europeana, with an explicit year. Paintings were kept (1,895 records typed as painting; 1,715 after removing icons, ceramics, prints and objects; 1,713 with a single explicit year). The files were downloaded, unreadable or near-monochrome ones (11) and photographic duplicates (7) were dropped, then 1840–1989 was kept: 1,655.
Colour of each painting: the CIELAB mean (L*, a*, b*) of the central rectangle and saturation C* = √(a*²+b*²). Periods: 1840–59, 1860–99, 1900–19, 1920–39, 1940–59, 1960–89. Confidence interval: 2,000 resamples. Noise N: the average over periods of the standard deviation between museums of the medians of the same period (museums with at least 25 paintings), then the larger of the Romanian museums’ and the two controls’. Controls: 787 Nationalmuseum and 1,177 Belvedere paintings, 1840–1989, under the same conditions.
Files with an embedded colour profile (Adobe RGB and others, 131 of 1,713) were converted to sRGB; the control photographs carry none. The year in the records was compared with the signed year quoted in the description: 280 of 288 agree. The tests, measures and thresholds were written before the calculation, with one dated amendment in the work plan (the choice of noise). The year ribbon uses for each year the five dominant colours of that year’s paintings and, where there are few, of neighbouring years (up to four on each side, at least eight paintings).
#Sources
- CIMEC “Clasate”, classified heritage, via Europeana
- Nationalmuseum Sweden, via Europeana
- Belvedere Vienna, via Europeana
Images
Photographs: CIMEC, through Europeana, from the Romanian museum collections; licence CC BY-SA 4.0. The mosaic and the picture to send contain reduced and cropped versions; each record gives the artist and the museum.
The images are CIMEC photographs, resized and cropped into small squares; each painting links to its record.