People by profession: research methods and useful context

Most occupation labels in a biographical database were invented after the people wearing them died. That is the first thing to understand about sorting people by profession, and it explains most of what goes wrong afterwards.

What to take away

Reference works overwhelmingly store a primary occupation, because interfaces need something to display and lists need something to sort on.
Notability and occupation are different variables and get merged constantly.
Occupation thesauri and authority vocabularies are useful, they give you consistent terms, hierarchies, and a way to search up and down a tree.
Honorary memberships, fellowships, chairs held in name, court appointments that were income rather than duty, and courtesy titles all look like occupations in a database and are not.

The label came later

Words for occupations have histories. "Scientist" is a nineteenth-century coinage; the people we now file under it described themselves as natural philosophers, or as physicians, or as mathematicians, or by whatever institutional post paid them. The modern split between artist and artisan hardened long after painters and sculptors were working as guild craftsmen who took commissions for altarpieces, painted furniture, and gilded signs without distinguishing between the jobs. Engineer, journalist, designer, curator, and composer all carry present-day meanings that would have been unintelligible to people the label is routinely applied to.

Applying a modern occupational term to an earlier life is not necessarily wrong. It is often the only way to make a person findable. But it is an editorial act, and the honest version of it says so:

Cataloged here as an astronomer. Contemporary records describe the post as [title in the source's wording].

That is a sentence a reader can check. "He was an astronomer" is a sentence that has thrown away the evidence and imported a modern job description.

One person is not one occupation

Reference works overwhelmingly store a primary occupation, because interfaces need something to display and lists need something to sort on. Lives do not cooperate.

People change careers, sometimes several times, and the change is usually invisible in a single-valued field.
People hold a day job that funded the work they are remembered for. The clerkship, the customs post, the teaching, the medical practice: these were the occupation for most of the waking hours and appear in none of the lists.
People work across fields that later separated into disciplines and were not separate at the time.
People hold appointments that were part administrative, part technical, part ceremonial, with no modern equivalent to map onto.

The result is that a person's entry in three different reference works can carry three different professions, all defensible, and a researcher who notices the discrepancy will suspect an error where there is only a forced choice.

Store occupation as a repeatable field with a date range on each entry, and the problem largely dissolves. What you lose is a tidy single-column sort. What you gain is a record that does not have to lie.

The record layers behind these labels sit in different custodies, described on the US state pillar, the city pillar and the country pillar.

Known for is not the same as worked as

Notability and occupation are different variables and get merged constantly.

A person is cataloged under the activity that made them findable, which is usually the one with surviving output: the published book, the signed painting, the recorded performance, the patent. The activity that occupied the working life may have produced nothing that survives and nothing anyone indexed.

This has a systematic direction. It favors occupations with durable, attributable products, and it under-represents teaching, administration, care, craft production, performance before recording, and every kind of work absorbed into an institution's output rather than credited to a person. When you count professions across a dataset, you are counting what leaves traces, not what people did.

Controlled vocabularies force choices you can see

Occupation thesauri and authority vocabularies are useful, they give you consistent terms, hierarchies, and a way to search up and down a tree. A published example is the Getty's Union List of Artist Names, which pins one identity to many name forms. They also impose their own shape:

Granularity is fixed by the scheme, not by the evidence. A source may say only "worked in the printing trade"; the vocabulary may require you to choose between compositor, pressman, and publisher. Whatever you choose is more specific than what you know.
Hierarchies encode a view. What sits under what, and what is a sibling of what, is a decision made by the vocabulary's maintainers for their own collection.
Crosswalks lose information. Mapping one scheme onto another rarely produces clean pairs. Terms merge, split, or land on an approximate parent, and after two hops the original distinction is gone.
Retired terms leave residue. Vocabularies are revised. Older records keep the old term, newer ones use the replacement, and a search on either misses half the collection unless the mapping is maintained.

None of this is a reason to avoid controlled terms. It is a reason to keep the source's own wording in a separate field, permanently, next to the normalized value.

Where an occupation label came from matters

The same word means different things depending on who wrote it down. Where entry to the work was controlled, the word was also a legal fact, and surveys of occupational licensing show how differently that control falls between places and periods.

Source of the label	What it establishes	Blind spot
Self-description	How the person presented their work	Aspiration, and the audience being addressed
Guild or professional admission	Formal qualification in a trade	Excludes everyone working outside the guild
Tax and rate records	The trade a household was assessed under	One label per household, chosen for revenue purposes
Census occupation column	What an enumerator recorded, under instructions	Unpaid, informal, seasonal, and women's work routinely omitted or entered as household status
Employment or payroll records	A post held, with dates	Says nothing about work done outside the post
Catalog or index heading	How a library or database made the person findable	A retrieval decision, not a biographical one
Obituary or biographical dictionary	What a summarizer thought was the point	Written to a house style, often decades later

The census column is worth dwelling on, because it feeds so many datasets. For long stretches the occupation of a household was recorded by an enumerator following printed instructions, one entry per person, with women's paid work commonly unrecorded and household status entered in its place. Counts of women in trades built from those returns measure the recording rules at least as much as the labor market.

Credit is a separate problem from occupation

Even where the work is known, attaching a name to it is not automatic:

Workshop and studio production distributed hands across a single object, with a master's name attached to the whole.
Collaborative and serial work has always had uncredited contributors, and crediting conventions differ by industry and period.
Assistants, apprentices, copyists, arrangers, translators, editors, and research staff appear in the record inconsistently and often not at all.
A name on a patent, a title page, or a founding document is a legal or commercial fact before it is a statement about who did the work.

If you are assembling people by profession from credits, you are assembling people who received credit. Say so, and treat conclusions about who did what as claims requiring their own evidence.

Titles are not jobs

Honorary memberships, fellowships, chairs held in name, court appointments that were income rather than duty, and courtesy titles all look like occupations in a database and are not. Check whether a title implied work, and in what period, before it becomes an occupation entry. Where you cannot establish it, record the title as a title.

Building a profession field that holds up

Make it repeatable. One person, several entries, each dated.
Keep the source's exact wording alongside any normalized term, and never overwrite one with the other.
Record what kind of source produced each label, using the categories above.
Separate three things that get merged: work done, post held, and heading cataloged under.
Flag anachronistic terms explicitly rather than pretending the person used them.
Where the evidence supports only a broad term, use the broad term. Precision you cannot source is a defect, not a feature.
Never derive an occupation from a name, a family, a place, or an association.

Living people

An occupation is less sensitive than an origin, but the same rule holds: what a living person does now is something they can state and you cannot verify from a stale database entry. Roles end, titles lapse, and people leave fields. Attribute and date any current-employment claim, or leave it out. This site's general position on assertions about living people is set out in the heritage guide, and it applies here too.

Bottom line

A profession field is a retrieval aid that readers treat as a biographical fact. The gap between those two things is where the errors live. Keep the original wording, keep the dates, keep the source type, allow more than one value, and be willing to write "described in the record as" instead of "was." Related sorting problems appear in the century pages and in the verification guide.

Common questions

How many occupations should a profile field hold?

As many as the sources support, each with its own dates and its own record. A single value is a summary somebody wrote, not something a document established.

Should the occupation be what someone was known for or what they did?

Store both, in separate fields. They differ for most people with a public reputation, and collapsing them is how a reference work loses a working life.

Is a controlled vocabulary worth using?

Yes, and store the source's own wording beside the code. The vocabulary makes searching possible; the original wording is the evidence.

What about a title that names a rank rather than a job?

Keep it as a title. Coding it as an occupation produces an occupational series that is partly a series about status.