The Forgotten Scribe of the 1890 US Census
In the grand narrative of data loss, the 1921 fire that decimated the 1890 US census is a tragic monument. We speak of the catastrophe—the water and flame that destroyed 99% of the schedules—in terms of aggregated loss, a statistical wound on the national archive. But to understand what was truly lost, and what open data means in a pre-digital world, we must look not at the fire, but at the hands that came before it. We must look at Julia Connelly.
Julia was a clerk-scorer, one of hundreds hired by the Census Office to tabulate results using Herman Hollerith’s revolutionary punch-card system. For months, she sat in a vast Washington, D.C. room, a human node in a proto-computational network. Her job was specific: she received stacks of cards, punched from the original handwritten returns by other clerks. Using a precise metal stylus, she would feel for the holes, reading the data not as numbers but as tactile patterns, and drop each card into the correct sorting box based on the combinations punched into it.
The data she handled was profoundly human: a card might represent a Civil War widow in Tennessee, a Chinese immigrant laborer in California, a child born into slavery now listed as a farmer in his forties. Julia’s task was to sort these lives into categories—by occupation, by parentage, by marital status—so they could be fed into Hollerith’s tabulating machine and become official statistics. She was, in effect, a human API, translating the analog scrawl of thousands of census takers into a machine-readable format.
This is the open data story we seldom tell. Long before servers and APIs, public records were made "readable" through immense, manual human labor. The "openness" of the 1890 data wasn't about a public URL; it was about its successful passage through this gauntlet of clerks and machines to become a published, authoritative report. Julia and her colleagues were the essential, fragile layer of preservation and access. They were the ones who decided if a smudged entry for "daguerreotypist" became "photographer" or was tossed into a pile for review. Their consistent, tedious work was what made the data coherent enough to be useful.
When the fire came, it didn't just burn paper. It burned the only comprehensive source that had passed through Julia Connelly's hands. The published aggregates survived, but the path back to the individual record—the raw, messy, human data that she had helped encode—was severed. The fire destroyed the primary dataset, rendering the work of the scribes permanently irretrievable.
Today, as we automate data pipelines and celebrate frictionless access, Julia’s story is a grounding reminder. Digital preservation often focuses on the medium—the disk, the format, the server. But the first act of preservation is the act of faithful translation, of making a record reliably readable for the next system, human or machine. It is meticulous, undervalued work. The 1890 census fire is a lesson in physical vulnerability, but the silent labor of Julia Connelly teaches us about the human vulnerability in every data chain. The data was open only because her hands, and hundreds like them, had first made it so.
Notes & further reading
A few pages I came back to while writing this: