What happens in a museum when it suspects that a work of art in one of its exhibitions or its collection has been misattributed? What are a museum’s standard procedures and strategies for examining the attribution of works that enter its space? How can museums move beyond the opinion of a single expert or curator by engaging different kinds of impartial specialists? How can the museum make the public aware of its research processes, and ultimately, the results?
Today, there are still no shared standards for museums on how to conduct due diligence on art or how to communicate findings to the public. With the growing trend toward institutional transparency, some pioneering international museums are beginning to look for new ways to review and ask serious questions about their collections and the works displayed in their spaces.
Three recent cases show different ways museums can conduct due diligence on artworks.
Vermeer
The first is the example of the National Gallery in Washington, D.C., which recently sought to verify a work in its permanent collection attributed to Vermeer, Girl with a Flute. The museum then shared and made the results of its technical analysis available online for scholars to read and for public transparency, concluding that the work was not by the artist’s hand. The museum then went a step further by creating an exhibition that allowed viewers to examine the work along with another in the collection with a stable attribution, explaining the museum’s approach and conclusions to the public and allowing visitors to the exhibition to follow the process and form their own opinions. Leaving attribution as an open question is an excellent approach for cases where there is no definitive answer. It teaches us to accept the possibility of dubious attribution rather than seeking a conclusive, often falsely reassuring answer. It also allows others to form different opinions: when the same work was recently loaned to the Rijksmuseum in Amsterdam for its monographic exhibition on Vermeer, the Rijskmuseum interpreted the National Gallery’s findings differently and chose to cautiously display the painting to the public as “attributed here to Johannes Vermeer.”
Donatello
A second innovative example is the recent Donatello exhibition at the Victoria & Albert Museum in London, which included a final section on forgeries and copies. The exhibition was complemented by a catalog essay by curator Whitney Kerr-Lewis on “Imitators, Copyists, Forgers.” The museum openly warned the public about the many artists who copied Donatello for four centuries after his death and the difficulty today of securely attributing many works to him. The catalog, which incorporates uncertainty into the very fabric of the exhibition, points out that a marble panel is most likely the creation of a “16th-century devotee.” This approach demonstrates that despite our hunger for works of art securely attributed to a single artist, the reality of attributions is much more complex: there is still a great difficulty in separating the artist’s works, true tributes to him by followers, and well-made forgeries that were made to deceive.
Titian
The third example is the Kunsthaus Zurich. In 2018, the museum welcomed into its collection a painting attributed to Titian entitled Evening Landscape with Couple (c. 1518-1520). It was the only Titian in Switzerland and as such had gained a certain status around its attribution. In this case, it was not the museum itself that questioned its attribution. According to Artnet News, the claim first appeared in the press. It was made after the work was subjected to an artificial intelligence (AI) digital analysis. This new form of analysis was not done by the museum or with its permission.
Kunsthaus Zurich stated that: “The digital AI analysis was applied without our sending a high-resolution photo, without our order and without our consent.” One might ask whether this approach to de-attributing a painting from a museum collection is ethical, responsible, or even legal. Regardless of legal judgment, such an approach risks damaging the museum’s reputation. The AI company reported that it tested the work against digital images of 300 paintings by Titian, including Diana and Callisto, housed at the National Gallery in Scotland, and The Penitent Magdalene, at the Hermitage Museum in St. Petersburg. The result was that there was an 80% chance that the work was misattributed. The Kunsthaus Zurich rightly commented that, in addition to not authorizing this examination of the work, “we are not sufficiently familiar with the method, its possibilities and limitations, nor with the arguments concerning our painting.” According to the AI company, the method “is very accessible, meaning fast, hassle-free (no transportation, no insurance), cost-effective, but also objective….In our opinion, objectivity is one of our most important strengths, and potentially the one that really makes a difference, as we are looking at a market that is still dominated by subjectivity.”
In my opinion, removing the human from the equation seems counterproductive: it is well known that AI is still a tool in its infancy that requires the intervention and human expertise of specialists to interpret its results. A study of the actual work, in all its material facets, is not included in AI, which is a tool that has yet to be incorporated into the other tools of attribution, such as connoisseurship, provenance research, and forensic analysis.
According to Artnet.com, the Kunsthaus has now acquired another painting that is visually similar to the one already in its possession, titled Lovers with Lute in a Romantic Landscape, with an attribution to French artist Nicolas Poussin “or his circle.” The first work is an oil on paper, while the second is an oil on canvas; the dimensions are almost the same as are the subjects and landscape. The museum created a two-year authentication project to examine both works. It has also decided to hang the two works side by side at the Kunsthaus so that viewers can participate in the process and ensure that it remains open and transparent. It is hoped that the museum will publish its findings, generating further discussion so that a general consensus can be reached among scholars on both works