Research Methods in Practice: The Darrell Anderson Archives
← Blog

Research Methods in Practice: The Darrell Anderson Archives

Larry W Jordan Jr (Larouex)September 23, 202617 min read

In 2005, a Seattle chef named Darrell Anderson gave his collection to Renton Technical College. It runs to dozens of boxes of menus, programs, clippings, letters, photographs, cookbooks and scrapbooks, gathered across a working life, and it documents something no other collection does: how the Pacific Northwest built a culinary profession, trained it, and connected it to the national associations.

The college holds it as the Darrell Anderson Culinary Archives. The material is physical and available by appointment, and the finding aids are published online. What has not existed until now is the contents in digital form, item by item, with the people in them identified and the claims about them sourced.

I am doing that work on behalf of the Washington State Chefs Association, whose own history is what the collection mostly records. The stories, biographies and collections go up at history.washingtonstatechefs.com. I came to it as a software architect rather than a trained archivist, and my instinct was to treat it as a data problem: photograph everything, run text recognition over it, extract the names, publish. That was about half right, and the wrong half cost me the most time.

I am currently working through Box 20. Four biographies are finished and about 600 pages have been read in full. What follows is what the work has taught me, organized around the decisions that turned out to matter.

The Collection and the Man

Darrell Lester Anderson was born in Hood River, Oregon, in 1929 and died in Tacoma in 2011. He enrolled in the culinary program at Edison Technical School in Seattle in January 1953 under the GI Bill, graduated in 1955, and ran the culinary arts department at the Sea-Tac Occupational Skills Center from 1967 to 1991. He was president of the association twice, in 1970-71 and 1980-81.

The association has one continuous history under three names: the Seattle Culinary Association, founded in 1923, the Pacific Northwest Chefs de Cuisine Association from 1949, and the Washington State Chefs Association from 1970.

Box 20 is the biographies box, and photographing it produced 1,793 images. More than 500 were blank backs of pages, which were catalogued and then removed. The four completed biographies are of Hubert J. Eickman, Claude Sedenquist, Josef Sir and Clarence "Jay" Culter.

The Arrangement Was Already Done Twice

The most useful thing in Box 20 is three sheets of paper in Anderson's handwriting: a list of what is in each box. The header reads, in part:

Boxes 1-18 are recipes. Box 41-46, and 55-56 are Magazines. Boxes 54 & 57 are menus. Boxes 58-60 are plates. Boxes 47 & 48 contain sensitive items: Avail. in 2058.

The body of the list enumerates each unit inside Box 20, labelled "Chefs – Biographies and clippings" and then a surname.

Read that header again. Anderson did not merely keep his papers in order. He arranged them by subject, wrote out what each box contained, and imposed an access restriction with a date on it, all before handing the collection to an institution in 2005. That is a donor instruction, not a private memo, and it is six years older than he is dead.

The college then did its own work on top of his. Their published finding aids are the layer I index against, and my box index is built from theirs rather than from scratch. Anderson's own list corroborates it at points: his header says boxes 58 to 60 are plates, and the college's digitized commemorative plates are exactly that.

So the arrangement exists twice over, and my job is not to invent a third one. It is to go down a level, from box and folder into the individual item, and to attach to each item the people, dates and claims that a finding aid has no room for.

The Booklets Are Copies of Other People's Books

I assumed at first that the dossiers in Box 20 were Anderson's own papers. They are not.

Each booklet opens with the same printed holiday greeting card, on which Anderson typed a note of thanks. One is dated 31 December 1993, addressed to a chef and his wife, and thanks them for letting him research the association's history. It ends with a line that reframes the whole box: "These zerox copies are only a small part of your book." The spelling is his.

He borrowed scrapbooks from chefs and their families over about six weeks in the winter of 1993-94, photocopied a selection from each, bound the copies, and returned the originals. What Box 20 holds is his edit. The books he worked from were larger, and some of them may still be in family hands.

That has two consequences. Every item record needs a field distinguishing Anderson's copy from the original, because the two have different owners, different locations and different rights. And some of the best material in this story is not in the collection at all.

He used the same method again in 1995. After one of the association's chefs died that June, Anderson wrote to the man's former apprentices asking each for a résumé and a paragraph on what the apprenticeship, and the chef, had meant to their careers. About twenty signed tributes came back. He bound them into a memorial volume and the American Academy of Chefs presented it to the widow that July. It is an oral history project conducted by mail, thirty years before I got here.

Paper Written at the Time Beats Memory Written Later

The clearest methodological lesson came out of a disagreement about a date.

Four separate retrospective accounts in the collection say Eickman became head of Edison's culinary department in 1957. Tributes, career summaries, a program note. They agree with each other, which makes them look reliable.

The letters written at the time say otherwise. Paul Muellet, the department head, died in September 1956. The Washington State Restaurant Association wrote to congratulate Eickman on 26 October 1956. A letter dated 30 October 1956 records that his acting appointment had been made permanent the previous week.

So he was acting head from September 1956 and confirmed in the last week of October. The four later accounts rounded to the following year, and each one copied the rounding from the one before it. A letter written in the week something happened outranks a tribute written thirty years afterward, even when the tributes outnumber it.

The pattern repeated everywhere. Newspaper obituaries are written under deadline from family recollection. One of them calls Eickman a Minnesota native. The 1950 census, a 1960 appeal from his childhood parish school in Wisconsin, and the family record all say Wisconsin. He started work at thirteen in Winona, Minnesota, which is almost certainly where the error came from. The archive's version is published and the obituary is footnoted.

Names Are the Hard Problem

I expected transcription to be the bottleneck. It is not. Telling people apart is.

A single run of 97 pages produced 1,030 distinct name strings. Automated matching on string similarity gets this wrong in both directions at once. It merges six different men named Johnson into one person. It splits Oliver Hanson and Oliver Hansen, who were one man and served as the association's secretary under both spellings.

The forename is what distinguishes people, and the forename is exactly what varies. One chef appears in print as Clarence, as Clarence A. and as Jay, with the surname spelled three different ways across the trade press. His engrossed certificate and Anderson's own label agree on one spelling, so that is the authoritative form and the other two are printers' errors. All three are stored as aliases, because a search that knows only one of them misses two thirds of the man.

Two rules resolve most of the rest, and both are computable from data already extracted:

  1. Two names on the same printed roster are different people. A membership list does not name the same man twice. This single rule refutes most bad merges.
  2. Two spellings holding the same office in the same organization are the same person. That is what resolves Hanson and Hansen.

Everything left over goes to a person, as a two-column comparison with the source pages beside it. Rejected merges are recorded alongside accepted ones, or the next pass proposes them again forever.

What Machines Do, and What They Cannot

The pipeline splits photographs into volumes, detects blank pages, corrects orientation, runs text recognition, and extracts names and claims. Three failures taught me where the line falls.

Finding where one book ends. The camera timestamps show a pause whenever a volume was set down and the next one picked up, which finds most boundaries. It missed one entirely, because two booklets were photographed back to back without the camera being put down. The holiday greeting card that opens each booklet turned out to be the better marker: matching the top of each page against a known opener found the boundary the timing missed, and scored it well clear of anything else in the run. Timing suggests a boundary. It never proves one.

Recognizing blank pages. Most of these booklets are printed on one side, so every other photograph is a blank back. The hard cases are blank backs where ink from the front shows through the paper. Software measuring edge detail counts that ink as content, and a mirrored-text test catches only the strong cases. Seventy-nine candidates were checked by eye, and thirteen turned out to be real pages with show-through behind their own text. One was a roster of the association's past presidents from 1938 to 1959, which would have been a bad loss.

Orientation. Anderson mounted clippings sideways and photographed them as mounted, so in one run more than a third of the frames are rotated. Text recognition can score a page of text at four rotations and pick the best. A photograph scores nothing at any rotation, so the first pass left every photograph upright and reported it as undecidable. All 226 of those frames were then looked at on contact sheets. Forty-two were sideways. Once turned, nine revealed handwritten captions, including one naming Edison's principal and Eickman in 1952.

The costs follow from this. Machine steps scale with the number of pages. Name adjudication, source verification and rights decisions do not, and they are where the effort actually goes. Plan the work in reviewer hours per dossier, not pages per hour.

One correction belongs here, because it was mine. I read "Eighth Annual" off a small contact-sheet image of a 1954 banquet program. At full resolution it says "Fourth," which moves the first of those suppers from about 1947 to about 1951. Contact sheets are for triage. They are not for transcription.

Describing What You Find

Getting the facts right is half the work. The other half is recording them in a shape that stays usable after I am no longer the person doing this.

Archival description separates three kinds of decision, and separating them settled several arguments I had been having with myself.

Which fields exist. Structure. What an item record has to carry: identifier, title, creator, date, extent, the volume and box it sits in, a scope note, a rights class and a display image. Established structures exist for this, including Dublin Core for general description and Encoded Archival Description for collection hierarchy. My schema is currently local, which is fine to build against and a problem to publish from. The college's finding aids already live in a standards-based system, and regional aggregators such as Northwest Digital Heritage, which feeds the Digital Public Library of America, ingest standard metadata rather than bespoke tables. A crosswalk from my item table to a recognized structure is small work now and large work later.

How values are written. Content. How a date is expressed, how a personal name is formed, what an identifier looks like. Descriptive standards such as DACS exist to settle exactly this, and I had been carrying an unresolved version of it for months, with two different identifier formats in use across different parts of the work. Unifying them is cheap at 80,000 rows and expensive at 800,000.

Which terms are allowed. Value. Controlled vocabularies for the fields that repeat in every box: format terms for menus, programs, scrapbooks, clippings and photographic prints, plus occupations, places and personal names. Name authority files and vocabularies such as the Getty Art & Architecture Thesaurus exist for this. My people table, with its alias list, is already a local authority file. What it lacks is any link outward to the wider ones.

There is a fourth thing, which I did not expect to need and now consider load-bearing.

How certain each claim is. None of the standards above carry this, because they describe records rather than assertions about people. In this project a biographical fact is not a column on a person. It is its own row, pointing at the page it came from, carrying a confidence level of confirmed, probable, possible, disputed or not found, and a review state. A database constraint refuses to let a claim reach verified without a source attached, and the published footnotes are generated by walking those citations. A sentence with no source has nowhere to hang.

That constraint turns an editorial intention into something the system enforces. It is also what lets two documents disagree without either one silently overwriting the other, which happens constantly in this material.

Several things I was already doing turn out to have standard names, which matters the first time a funder asks what practices the work follows. Anderson's 2058 restriction is a condition governing access, and it lives in the database as a field on the box record so nothing can be published past it by accident. The clipping problem is a condition governing use. The question of Anderson's copy versus the original scrapbook is custodial history.

What the Record Does Not Say

Every finished biography has a hard edge, and being explicit about where it falls is part of the work.

One chef's booklet runs to 25 pages with content and gives no year of birth or death anywhere. What survives of him is what a man keeps: his certificate, the programs from two good dinners, the notices when he was elected treasurer, and the write-ups when he finally had a place of his own.

Another file begins in 1953, when the subject was 23 and already assisting the chef at a Seattle restaurant, and stops in 1968. He lived until 2019. Fifty years are missing.

A third is a memorial volume, so it is detailed about how the man taught and nearly silent about where he came from. A funeral card carries his birth date and birthplace, and it is the only document in 257 photographs that does.

Two of Eickman's own career summaries disagree about whether he served in the Army or the Navy. Twenty-six pages of newspaper coverage say nothing about military service at all. So the biography names no branch, and says why.

Publishing an unverified claim about a named chef is worse than publishing nothing. A short entry that states only what the evidence supports is a good archive page. An essay padded out to look complete is not.

The Research Is Not Solitary

Anderson understood this in 1993, and I keep relearning it.

The association's own published history and list of past presidents has independently confirmed several dates the collection could only estimate. The Museum of History and Industry holds a 1949 press negative whose sleeve lists nine men at Edison, including Eickman. MOHAI invites the public to identify people in its photographs, and this collection can very likely name the rest of them.

The limits of searching are just as clear. A 1981 committee roster names 24 people with their titles and employers. I searched ten of them online and nine returned nothing at all. Everyone in that room is now roughly 75 to 95 years old. Putting the list in front of the association's current membership will do more than any further searching, and the window on that closes a little every year.

Where living people are involved, the approach runs through the association and through public or institutional channels. No home addresses or telephone numbers are collected, and no people-search sites are used. Anyone interviewed will be asked for a signed release at the first meeting rather than later.

Rights and Privacy Come First

These decisions have to be made before publication, not after, and they are more restrictive than I expected.

In the first run of Box 20 read end to end, 45 percent of the catalogued items are newspaper and trade clippings, which are other publishers' copyright. Each will be cleared, cited without reproduction, or withheld.

In the memorial volume, 41 of 257 pages print a living person's home address, telephone number or date of birth, because the material was assembled by asking working chefs for their résumés. Those pages are flagged restricted and unpublishable.

A community cookbook in the collection credits 48 of its 298 recipes to children, by name and grade. Those children are adults now. Their names are transcribed as printed and stored with the grade field so they can be filtered at publication, and whether to show them is a decision still outstanding.

For the same reasons, this post names very few living people.

What Changed My Mind

I started out treating Anderson as the collector and myself as the historian. That is backwards.

He was secretary of Seattle's Escoffier society chapter in 1964 and secretary of the local Academy of Chefs chapter in 1995. He borrowed the books, made the copies, wrote the thank-you letters, annotated nearly half the pages in one booklet with dates, sources, continuation marks and research notes to himself, wrote the box list, and then placed the whole thing with a college so it would outlast him. Those annotations are a finding aid he built on top of the documents, and they are evidence about him as much as about the chefs.

There is also a line running through the collection that I did not see at first. A letter of February 1954 thanks Eickman for the training he was giving a young student. The student was Darrell Anderson.

When Eickman died in 1995, the Seattle Times asked Anderson about his old teacher. He said: "He never had a harsh word for a student. He was always calm, cool, collected."

Anderson is the spine of this collection, and his own biography is the one I have not written yet.

What Is Next

Three dossiers remain unread in Box 20, and the boxes after it are queued behind them. The identifier scheme is settled, the crosswalk to a standard description structure is next, and the biography of Anderson himself comes after the boxes he built are read.

If you trained at Edison Technical School, worked in Seattle's professional kitchens, or remember this association's earlier decades, I would like to hear from you through the Washington State Chefs Association.

Sources and Links

Share this postLinkedIn
See Where You Stand