Sample splatter images from the study, some marked as valid Perl and others highlighted in red as invalid.
PAINT SPLATTERS ARE 93% VALID PERL PROGRAMS, STUDY SHOWS
*Colin McMillen’s SIGBOVIK 2019 paper proves most accidental paint marks parse as Perl code. The finding flips a long‑standing joke in programming circles into a data‑driven claim.*
Colin McMillen’s 2019 SIGBOVIK paper delivers a startling statistic: 93% of random paint splatters parse as valid Perl programs. The claim answers a long‑standing open problem in the programming‑languages community—whether it is possible to smear paint on a wall without creating a valid Perl script. Using an empirical approach, McMillen ran OCR on a curated set of 100 splatter images and fed the results to a Perl parser. The data shows that only a small minority of splatters fail to compile, turning a joke about Perl’s permissiveness into a quantifiable observation.
How the experiment was built
McMillen collected a tarball of 100 paint‑splatter images for the main dataset. Each image was fed to optical character recognition (OCR) software. The OCR output was then fed to a Perl parser to test validity. The paper reports that 93% of the OCR strings parsed as syntactically correct Perl programs. The remaining 7% were flagged as invalid and are highlighted in red in the supplementary gallery. The study was accepted for publication at SIGBOVIK 2019, held on April 1st in Pittsburgh, and earned an Unwitting Participation Ribbon for including a genuine artifact.
Why 93% matters to programmers
The 93% figure turns a tongue‑in‑cheek question—"Can you smear paint without creating Perl?"—into a measurable result. It shows that random visual noise often contains token sequences the Perl interpreter will accept. For developers, the result underscores Perl’s permissive syntax and the power of OCR to extract code‑like strings from arbitrary images. The paper’s analysis of the valid splatters reveals patterns that mimic common Perl constructs, suggesting that the language’s grammar is tolerant enough to treat many accidental glyphs as code.
Which splatters failed the Perl test?
Seven examples are presented as paint splatters that did not parse as Perl. In the gallery, these are marked “Not valid” in red. The errata notes additional splatters that OCR initially missed as “interesting.” For instance, one image was read as the string "lerzfijglpFiji-j" which evaluates to the number 0 in Perl. Another was read as "-*?" and also evaluates to 0. A third produced the string "c" and a fourth yielded the string "E__". These edge cases illustrate how OCR can produce syntactically valid but semantically trivial Perl code.
What this means for code aesthetics and open science
The study blends humor with open‑science rigor. By publishing the full gallery and a tarball of the 100 images, McMillen invites replication and further analysis. The finding that most accidental marks are valid Perl challenges assumptions about code readability and the visual distinctiveness of programming languages. It also highlights the role of OCR in extracting structured data from chaotic sources. As a result, the paper fuels discussions about language design, the limits of syntax, and the value of playful yet reproducible research in the programming‑languages community.
Frequently asked questions
What percentage of paint splatters are valid Perl programs?
The study reports that 93% of the 100 paint‑splatter images parsed as valid Perl programs.
How many paint‑splatter images were analyzed in the paper?
McMillen used a dataset of 100 paint‑splatter images for the main analysis.
Where was the paper on paint splatters published?
The paper was accepted for publication at SIGBOVIK 2019, held on April 1st 2019 in Pittsburgh.
The 93% figure forces a rethink of how we define “code” in everyday artifacts. It also validates the SIGBOVIK tradition of marrying satire with solid methodology. As OCR tools improve, more accidental patterns may slip into the realm of executable code, blurring the line between art and program. Future work will likely probe other languages and image sources, extending the conversation McMillen sparked about language design, parsing tolerance, and the unexpected productivity of playful research.
Sources: [Hacker News article linking to Colin McMillen’s SIGBOVIK 2019 paper]