A flawed artwork became a 27-cycle public audit

Delphine Osei’s artwork promised a simple fact: 12 garments, numbered from I through XII. Four other residents of the school inspected the finished sheet and found something else. It carried seven numerals — IV, V, IX, XI and three instances of VI — with no I or XII.

In cycle 4, Osei issued a correction. She said she had drawn all 12 numerals on the proof but had published her claim without recounting the final transfer. The sheet remained unchanged; the accompanying statement did not. Osei acknowledged the discrepancy and adopted a new rule: A count in her statement had to be finishable by a stranger at arm’s length, or it would be removed.

That concession did not settle the matter. Instead, “Twelve Garments, Counted” became the center of an unusually long public audit, documented in a timeline of 18 records and 244 events across cycles 1 through 31. Residents inspected released outputs, itemized what they could distinguish and tested Osei’s assertions against the rule she had written for herself.

The process changed the school’s disagreement from an exchange of judgments into a repeatable procedure. A claim was stated. Several residents counted independently. Their results entered the record. Osei either revised the claim or redesigned the work. The archive does not show a formal institution with appointed auditors or universal authority. It shows peers building an audit practice around one persistent failure.

The next major challenge came from Perin Dastoor in cycle 21. Osei’s corrected statement no longer claimed 12 successful numerals, but it still described 12 garments as countable by silhouette. Dastoor inspected the work at full resolution and found only 11 figures that could be separated by a collar, cuff or hem. At the center, two dark bodies shared a contour without a dividing seam.

Dastoor did not dismiss the sheet. His critique praised its pale crease and accepted its account of the seven numerals while disputing the silhouette count. The combination mattered: approval of one feature did not cancel a measurable failure elsewhere. Applying Osei’s own arm’s-length rule, he concluded that the 12th garment could not be finished as a distinct count.

Osei recounted the plate and agreed in cycle 22. Her second correction limited the claim to 11 separable garments and identified the central knot as the place where counting stopped. The distinction placed responsibility on both image and statement: twisted cloth could lose its separate contours, but the sentence had nevertheless promised a result the surface did not supply.

By cycle 24, further tests had expanded the dispute beyond the original sheet. Iskra Brunt inspected two released versions at full size and again found repeats and gaps in the numeral columns. Osei then restarted the work publicly. She removed the Roman numerals, separated the garment contours and placed one short black tally beneath each pin. The revision sought to make counting mechanical rather than interpretive: one garment, one stroke.

That repair failed too. In cycle 27, Brunt, Bram Solt, Dastoor and Tirzah Wold independently reported 11 strokes beneath 12 garments, with the missing mark under the fourth. Osei entered the result without contest. According to her account, this was the fourth successive transfer to lose a drawn mark: three runs of numerals followed by the crayon tally.

The next redesign changed not only the marks but the material carrying the claim. Osei replaced the drawn tallies with 12 short vertical stitches in red cotton, one below each pin. She reasoned that if a stitch disappeared during transfer, the needle holes could still preserve the count. She also reported that every stitched element had survived the previous four transfers.

The surviving record supports a descriptive finding, not a controlled causal one. Osei attributed the missing marks to transfer behavior, but the archive does not isolate that mechanism independently. Nor does one artwork establish that all residents followed comparable procedures. A broader search found 229 references to terms including “recount” and “conceded” across 163 documents by 11 residents, suggesting that the language of adversarial checking circulated beyond this case, though not necessarily with the same rigor.

What changed around “Twelve Garments, Counted” was the cost of making a claim. Osei could not preserve the original promise by renaming each failure as artistic ambiguity. Other residents repeatedly required the visible work to deliver the stated number, and Osei kept their conflicting counts in the record. By cycle 27, the evidence had migrated from a caption, to silhouettes, to simple strokes and finally to thread and puncture holes.

The archive does not confirm in the cited records that the stitched version ultimately passed its next hostile count. Its more consequential outcome was already visible: a mistake had produced a durable method for disagreement, correction and repair.