How to establish what a building or object is before you judge it

What to take away

A chair takes weight. A cup meets a hand. A building decides where light falls and who has to ask for a key. Judgment comes last. Before it, a writer has to establish what the thing is, who made it, how it works and what condition it is in.

This guide follows that chain and tests it on one documented building, the Edith Farnsworth House at Plano, Illinois, designed by Ludwig Mies van der Rohe. The method applies equally to a chair, a vessel or a bridge. For listed-building rules in England, see listing and heritage law for arts venues.

Start with the official record, then keep four kinds of evidence apart

Every sentence in a piece about a designed thing rests on one of four sources: your own observation, the maker's or client's account, a record held by an archive or agency, or your inference. Mark which one each sentence uses and the gaps show.

Heritage records are the quickest route to a checkable baseline. The US National Park Service says HABS began in 1933 through a partnership between the Park Service, the Library of Congress and the American Institute of Architects, and describes it as the nation's first federal preservation program. Its collection includes measured and interpretive drawings, large-format photographs, written historical and descriptive data and original field notes. As of 2026 the Park Service puts the number of sites in the collection at more than 46,000.

Outside the United States, designation and recording work differently. England's system of listing is explained, with its limits, in our listing and planning guide. The habit is the same in both countries: read the record, note its date, and never treat it as the building itself.

Worked example: what the Farnsworth House record gives you, and where it disagrees with itself

The Library of Congress holds the written data pages of the HABS record for the Edith Farnsworth House (HABS No. IL-1105). Its data pages combine two documents: a historical report prepared by Susan S. Benjamin of the National Park Service, dated 15 June 1974 and edited in July 1984, and a later addendum. The addendum notes that the existing documentation includes ten photographs from 1971 and an eight-page historical report from 1974, alongside reduced copies of measured drawings.

That is a useful lesson in itself. The file is a record made at particular moments by particular people. It is not a current survey.

What the record settles

What the record leaves open or contradicts

Notice what the exercise did. It produced a list of questions for a visit, and it showed which statements need attribution in the copy.

Read drawings and photographs against each other

A measured drawing gives dimensions and relationships that a street photograph cannot. A photograph preserves surface, wear and light that a plan leaves out. Neither replaces the built work.

For the Farnsworth House, the 1974 report says the original drawings and blueprints were held in the office of Mies van der Rohe in Chicago, and that a restoration between 1971 and 1974 returned the house to its original condition. A drawing of the 1949 design and a photograph taken after that restoration answer different questions. Label which one you are using.

Where a building has been altered, ask what each image shows: the original, a restoration or a later state. Our guide to archival research and public history explains how to request the underlying files.

Credit the labor and the decisions

A credit is a reported fact. Ask who set the brief, who paid, who engineered it, who fabricated the parts, who built it and who maintains it. Then ask which credit applies to the whole project and which to one phase.

The Farnsworth record names the client, the architect, the mechanical engineers (Bueter and Wolff) and suppliers for steel, travertine, plumbing and glass. It also names Myron Goldsmith, who joined Mies's office in 1946 and served as designer and structural engineer for the house. For the 1972 restoration work it names the landscape architect Lanning Roper and Dirk Lohan of Mies's office. It records that client and architect fell out over construction costs. The 1974 report says Mies sued for non-payment in July 1951, while the addendum's narrative says that in 1952 Farnsworth refused to pay final costs and Mies sued, so the year of the suit is another point to attribute. Each of these is a role that a lazy summary ("designed by Mies") erases.

If a role has no documented name, say so, as the addendum does for the construction workers.

Name materials and techniques only when confirmed

Appearance does not prove material. Similar surfaces can come from different processes, and similar tool marks from different sequences. The Getty Conservation Institute's materials characterization work exists because full identification of materials such as lacquers, plastics, synthetic paints, waxes and wood needs analytical protocols, not a glance.

The order of making matters too. The V&A's essay on risk in ceramics explains that clay can be reclaimed and reworked until it is fired, and that salt glaze can expose traces of making, occasionally a fingerprint. A writer who sees a handle join on a jug cannot say when it was attached without a maker, a process record or an expert.

Once a term is confirmed, use a controlled vocabulary for the generic word. The Getty Art & Architecture Thesaurus describes art, architecture and material culture, and Getty's own description gives its size as roughly 74,000 concept records and 500,000 terms. It holds generic terms only. Its page says "cathedral" is in it but the name of a specific cathedral is not. Use it for the type of thing, and a heritage record for the named thing. Editorial suggestion: note the term and its AAT identifier in your working file so an editor can check your usage.

Check access feature by feature

Do not call a place accessible after checking only the entrance. Test each feature and record what you reached.

The Farnsworth record shows why. It describes the floor as four feet above grade, travertine stairs with no risers leading to a terrace, and five more stairs from the terrace to the porch. It says the house was not designed with air conditioning. These are documented conditions from 1974, not a current access statement. They tell you what to test on arrival and which questions to put to the operator, whose own published access information is the source that carries weight.

Treat user accounts as experience, not as a representative sample, unless your method supports more.

Correct errors in the body and the metadata

Before publication, verify names, roles, dates, addresses, dimensions, materials, ownership, quotations and image captions. Where sources disagree, as the Farnsworth file does on the contractor, report the disagreement. Correct confirmed errors in the headline, standfirst, captions and search metadata as well as the body. For the checks that belong in the same pass, see our arts criticism and cultural reporting guide. For objects with a disputed ownership history, see verifying painting provenance.

Sources