SNAPlab

Meanwhile, as I work on publications for my current research project, older pieces are finding their way online. In September, our article “Tiny Scratches Belong to Movies” was published in the Art & Perception journal.

It needs to be said that the initial idea for this research came from Johan Čapek, my student at the time. He approached me in 2024 with the intention to explore the perception of cinema in different versions and formats. Gradually, we developed this idea into specific research questions about the perception of restored and unrestored versions of old films. It all happened at a time when I was slowly gravitating towards eye-tracking research (although I didn’t have access to any reliable devices to collect data).

We were lucky that everyone we approached agreed to collaborate. When we asked restorer Tereza Frodlová and curator Matěj Strnad from the Czech Film Archive, they gave us a wonderful set of stimuli – clips from two versions of old Czech films, both restored and unrestored. In particular: The Cassandra Cat, Ducháček Will Fix It, Kristián, The Blue Star Hotel, Old Czech Legends, A Blonde in Love, and The Cremator. But the involvement of Tereza and Matěj wasn’t limited solely to the selection of stimuli. They were essential when we discussed the results and potential consequences for the restoration of old films.

The last member of the team was Stanislav Popelka. We didn’t know each other beforehand, but he generously agreed to lend us his lab and dedicated his time to helping with the design of the study. He also suggested the statistical methods we later used when analysing the collected data.

And I forgot to mention – we had no funding. Just enthusiasm.

Maybe rather than just reminiscing about all this, I should say something about our findings, too.

First, it needs to be stressed that eye-tracking is not a usual method utilised by film archivists and restorers. I guess their work is complex enough as it is, so there is not much space left for theories of cognition and perception. Therefore, we had to build our research questions on the foundations of eye-tracking studies in visual art restoration.

The procedure of the experiment was simple. Participants came to the lab, sat in a chair, and watched a computer monitor with a stationary eye-tracker. Then they watched 16 clips (the same 8 film scenes in two versions) where unrestored and clean versions alternated. After that, we asked them several questions to find out if they consciously noticed visual flaws and what kind of flaws were most annoying to them.

Just to explain why I am writing about general “flaws” instead of using a taxonomy and calling individual elements of the unrestored image a defect, damage, or error: To be honest, I don’t have enough information to tell those apart in certain specific cases. However, I think there might be some interesting hidden research exploring how individual categories and types of flaws change the perceptual ease of viewers. But back to our article.

Our findings regarding the specific questions can be summed up as follows:

• Are viewers perceptually or consciously disturbed by flaws in the unrestored version of the film?
  • Perceptually, yes: significantly more fixations in the clean version, significantly longer fixations in the unrestored version.
  • Consciously, not necessarily: some of the participants didn’t notice the flaws at all, and most of them weren’t disturbed by the presence of flaws in the image.
• What types of flaws disturb viewers the most?
  • If noticed, black frames (the total loss of the image as a result of a destroyed frame – see figure below) were mentioned as the most disturbing.
• How do viewers understand the ‘authenticity’ of unrestored films?
  • For participants, the flaws might be part of their authentic experience.

You might be surprised why we asked about authenticity. Well, participants talked about it even when we didn’t ask, so we made a research question out of it. Moreover, I really like to operationalise terms like that. And in the end, the question of authenticity is, in my opinion, the most interesting finding of the article (because gaze metrics acted similarly to those in visual art).

There is a paradox hidden in how authenticity is understood by viewers versus restorers. While restorers also respect authenticity, they might in some cases decide to use black frames when an image is damaged but the sound track survived. However, for participants, these were the most obvious and disturbing flaws. Unlike tiny scratches, which were understood as authentic, nobody appreciated black frames. Therefore, what is authentic for an archivist can, in a sense, destroy the impression of authenticity for the viewers.