The last portrait of W.A. Mozart
Text & images
A biometrical statistical comparison - Martin Braun, April 2005
Text only

In 1999 Michaelis and Seiller published a report suggesting that a neglected painting, which had been locked away in the stack-rooms of a large public gallery in Berlin (Gemäldegalerie) for 65 years, most probably was the last, and also the best, portrait of Mozart, painted one year before his death.

After a necessary restoration of the painting, Michaelis and Stehr published a restoration report in 2004, and the picture was displayed in one of the public exhibition rooms of the gallery. In January 2005, one year before Mozart's 250th birthday, the matter was also presented to the mass media resulting in a wave of sensational news articles around the world.

In the communities of Mozart friends and Mozart experts the picture caused both enthusiasm and fierce rejection. The latter immediately resulted in questioning its authenticity. Michaelis and Seiller (1999), however, had presented plausible evidence for their suggestion. It is based on biographical data of Mozart and Edlinger, as well as on a detailed comparison of the portrait with an earlier one. The latter, the so-called "Bologna" Mozart from 1777, had been praised by the composer's father, Leopold Mozart, because of its exact resemblance to his son.

Here, a biometrical statistical comparison of the two paintings is reported. The result is that their subjects have to be considered as the same person, with an error probability of well below one in 10,000,000.

Three biometrical tests

A. Landmark test on non-identity
Face identification is largely determined by proportions and angles of landmark distances, such as length of nose. Comparison of these parameters between the two faces does not reveal one significant difference. Thus, the landmark test on non-identity is negative, and identity of the subjects remains possible.

B. Feature test on non-identity
Most faces have a number of non-general features, such as an indentation between the eyebrows. Comparison of such features between the two faces shows one single difference. The eyebrow curves on the Edlinger painting have a smaller radius. In many faces, however, eyebrow radius decreases, when eyebrows move up, as they have done in the Edlinger face. Thus, also the feature test on non-identity is negative, and identity of the subjects still remains possible.

C. Digital feature test on identity
Many faces have a number of digital features. These are features that are either obviously present or obviously absent. If two portraits, for which the non-identity tests have been negative, have a sufficient number of digital features in common, the probability of subject identity can be determined statistically. The following digital features appear in both portraits:

1) A nose tip with two tip-defining points.
2) An elevation on the ridge of the nose.
3) A horizontal indentation across the root of the nose.
4) A deep fold between nose and upper half of right eye.
5) A vertical indentation between the eyebrows, asymmetrically (!) toward the right.
6) A thinning-out in the lateral third of the right eyebrow.
7) An elevation on the left side of the forehead at an identical location.

Because the three features on the nose and the two in the eyebrow line are clearly visible at almost all light conditions, their frequency in the general population could be determined by feature counts in public portrait galleries.

A corpus of 103 adult Caucasian male portrait paintings was established by extracting all naturalistic style portrait that were stored in sufficient resolution from the internet archives of the Gemäldegalerie in Berlin/Germany (yielding 47 pieces, limited to 1700-1850) and the National Gallery of Arts in Washington D.C./USA (yielding 56 pieces) .

A corresponding corpus of 103 portrait photographs was established by extracting, in order of listing, the results from Google picture searches that included the search term "portrait". The feature frequencies were as follows:

1) Nose tip with two tip-defining points: 7 % in paintings, 7 % in photographs.
2) Elevation on the ridge of the nose: 5 % in paintings, 1 % in photographs.
3) Indentation across the root of the nose: 7 % in paintings, 2 % in photographs.
4) Indentation between the eyebrows, asymmetrically toward the right: 2 % and 6 %.
5) Thinning-out in the lateral third of the right eyebrow: 2 % and 2 %.

Next, these five features were tested on correlations. Because all tests were negative and because there is also no biological rationale to assume any correlation, the five features have to be considered as stochastically independent. Thus, the frequency of their joint occurrence is computed by multiplication of the single frequencies. The results for the probability that two non-relatives have the five features in common are below one in 11,800,000 re the painting data base, and below one in 69,000,000 re the photo data base.

Further, it should be noted that the probability estimate would have become even more extreme, if the other two digital features and the many non-digital features that the two faces have in common had entered the calculation.


1) There is no longer a need for an archival confirmation of the "Edlinger" Mozart. Both in court and in research, visual evidence generally has a higher rank than all other evidence.

2) Possible archival indications that the Edlinger painting might show an other person than Mozart would not be compatible with the visual evidence and could therefore only be based on errors or misunderstandings.

3) The possibility that somebody imitated Edlinger's style of painting from around 1790 and produced an age-shifted copy of the "Bologna" Mozart seems highly unrealistic. Edlinger was an outstanding and highly individual painter, and thus very hard to imitate. A false "Edlinger" Mozart without a name of the portrayed subject and without a signature of the painter would also not have made much sense, neither for the painter nor for the owner. As long as there is no evidence for this extremely remote possibility, the authenticity of the "Edlinger" Mozart has to be considered as proved.

Michaelis, Rainer, and Seiller, Wolfgang. Ein unbekanntes Bildnis Wolfgang Amadeus Mozarts in der Berliner Gemäldegalerie. In: Mozart-Jahrbuch 1999 des Zentralinstitutes für Mozartforschung der Internationalen Stiftung Mozarteum Salzburg. Bärenreiter, Kassel, 2000, 1-12.
Michaelis, Rainer, and Stehr, Ute. Das Berliner Bildnis des Wolfgang Amadeus Mozart. Museumsjournal, Berlin, Jahrgang 2004, Heft 3, 32-33.

After the publication of this report a hypothesis appeared that the portrait by Edlinger might not show W.A. Mozart, but instead a Munich businessman called Steiner. It turned out, however, that this suggestion was based on a technical error. It was found that apparently both Mozart and Steiner were portrayed by Edlinger and that the two paintings cannot be confused, due to their strongly differing sizes. See special reports: in English or German.

See also:
- Identification of a new Mozart portrait from the mid-1780s
- The Greuze Mozart - rediscovery of a portrait painting
- The Delahaye Mozart - rediscovery of a portrait of the 16-year-old composer
Free counter and web stats
Author's website