How to label and catalogue a mineral collection
Five fields, one permanent number, three certainty states. The IMA list holds 6,239 valid species, so a name without its source is an opinion, not a record.
Record five fields for every mineral specimen: number, species, locality, source and date, and notes. The number is the only field that must never change. The IMA list of valid species carried 6,239 entries in September 2026, so a species name recorded without its basis is a dealer's opinion rather than a determination.
In short
- The number is the spine and it is permanent. Every other field can be corrected as you learn more. The number cannot, because it is what ties the specimen, the label, the box, the catalogue row and the photograph together. Assign it once and never reuse one.
- Five fields is the working minimum: number, species, locality, acquired from and when, and notes. Everything else — size, hardness, price paid, condition, analytical history — is useful and optional. A collection with five fields recorded is a documented collection; one with three is not.
- Record certainty in three states, not two. “As written” for what the old label says, “determined” for what you or someone else established and how, and “undetermined” for what is genuinely unresolved. Collapsing the three into one field is how false certainty enters a collection.
- Never write on the specimen in a way you cannot undo, and never overwrite an old label. An original label in a Victorian hand is a historical document; a transcription is not a replacement for it.
- A catalogue in one place is not a catalogue. Two copies, one of them somewhere else, in a format you could still open in twenty years — a spreadsheet exported to CSV beats any collection-management application on that test.
| Field | Example | Rule | Common mistake |
|---|---|---|---|
| Number | BSM-0417 | Permanent, unique, never reused | Renumbering when the collection is reorganised |
| Species | Fluorapophyllite-(K), by locality and habit, unanalysed | Name the basis, not just the name | Writing the species flat when it was never determined |
| Locality | Wheal Gorland, Gwennap, Cornwall, England | Most specific first, then up the hierarchy | Recording only the county, or only the country |
| Source and date | Ex coll. J. Smith; acquired 14 March 2019 | Who it came from and when it reached you | Leaving it blank because it feels obvious at the time |
| Notes | Label as written: 'Haematite, Cleator Moor'. Old label retained. | Anything a future reader would need | Recording judgements without recording evidence |
Why the number must never change
Every collection that has been reorganised has a version of the same disaster: the specimens were renumbered to match a new arrangement, the old catalogue still refers to the old numbers, and the photographs, the correspondence and the notebook now point at the wrong stones.
The number is not a position in the collection. It is the identity of the specimen. It has no meaning, it should carry no information, and it should never encode the species, the locality or the shelf, because all three of those can change and the number cannot.
A running serial is enough: a prefix if you like, then digits, assigned in the order specimens arrive. Never reuse a number, even for a specimen that has been sold, broken or given away — the record of what happened to it is part of the collection's history, and reusing the number destroys it.
Write the number where it can be read without handling the specimen: on the box, on the label, and — if you do it at all — on an inconspicuous part of the specimen itself in a reversible way. The traditional method is a small patch of removable white paint or clear lacquer with the number written on it in ink; the patch can be lifted with the appropriate solvent, and the specimen is unmarked underneath. Never write directly on a specimen, never use permanent marker, and never write on a display face.
Three certainty states, and why two is not enough
This is the part of cataloguing that most collections get wrong, and it is the part that determines whether the records are trustworthy in fifty years.
Most catalogues have one species field. Into it go three quite different things: what an old label says, what somebody established by test or analysis, and what the cataloguer thought was probably right. Once they are in the same field they are indistinguishable, and a guess made in 1974 reads exactly like an analysis done in 1974.
The three-state convention separates them:
As written. The label says “uvite, Brumado”. Record that, in quotation marks, with a note of the label's appearance and apparent age. This is evidence about the specimen's history and it is never edited.
Determined. Someone established the species and there is a basis: a hardness and streak test you did, a density measurement, an X-ray diffraction pattern, a curator's opinion. Record the determination and the method. “Kyanite, confirmed by directional hardness test” is a record; “kyanite” is a claim.
Undetermined. The honest state for most specimens. Apophyllite cannot be resolved to species without analysis. Langite cannot be separated from wroewolfeite without it. Uvite and fluor-uvite are an analytical distinction. In every one of these cases, “undetermined between X and Y, with reason” is a better record than a confident single name, because it tells a future reader exactly what work remains.
The quotable version: record what you know, record how you know it, and record what you do not know. The third one is the field that ages best.
Species names move, so record the date you recorded them
The IMA Commission on New Minerals, Nomenclature and Classification maintains the list of valid mineral species. Its master list, updated in September 2026, states that it covers 6,239 currently valid species. That number grows every year, and existing names are redefined, renamed and discredited as the mineralogy is revised.
Three examples from this site alone. Apophyllite stopped being a species name in 1978 and is now a group of six. Heulandite was split into a suffixed series in 1997. Uvite was reassigned in 2022, so a pre-2022 “uvite” label most likely describes what is now fluor-uvite. None of those specimens changed; the names did.
The practical response is small and completely effective: record the date alongside any species determination. “Heulandite (as at 1994)” is instantly interpretable by anyone who knows the nomenclature history; “heulandite” alone is not. It costs one column and it means your catalogue survives the next revision.
The second response is to leave old labels alone. A label reading “Haematite, Cleator Moor” in a nineteenth-century hand records the spelling, the terminology and the collecting culture of its moment. Correcting it to “hematite” tidies away evidence. Record the modern name in the catalogue and keep the label as it is — the argument set out on what ex coll. means on a mineral label.
The physical labels, and the system that actually survives
Each specimen gets its own label, in its own box, and the label never travels separately from the stone. That is the one rule on this page with no exceptions. If a specimen has several labels from successive owners, keep all of them in order; they are the provenance.
Write on acid-free card in permanent, water-resistant ink. Not on the back of a receipt, not in biro that will fade, not on paper that will go brown. If sulphides are stored nearby, keep the labels out of the closed box with them: sulphide decay produces acid, and the acid attacks paper before the specimen looks obviously bad.
The label carries: number, species, locality, and source. That is enough. The catalogue carries the rest, because a card that has to hold everything becomes a card nobody can read.
Photograph each specimen with its label in the same frame, once, when it arrives. This takes thirty seconds, it fixes the association permanently, and it is the single highest-return habit in cataloguing. If a box is ever tipped out, a folder of photographs reconstructs what the drawer cannot.
For the digital side, choose boring. A spreadsheet with one row per specimen, exported to CSV, will be readable in twenty years. A collection-management application may not be installable in five. Keep two copies, one of them in a different building or in cloud storage, and check once a year that you can still open the file. A catalogue that exists only on the machine in the same room as the collection does not protect against the event most likely to destroy both.
If you are starting from an inherited collection with labels already separated, our page on what to do with an inherited collection covers the triage, and the wanted list sets out the material we are actively looking for once the records are in order.