Appearance
Levels of description are the nested tiers — fonds, sub-fonds, series, sub-series, file, item — that let you describe a body of records from the general to the particular without repeating yourself. Apply them by describing the whole first (the fonds), then only the lower levels that genuinely exist in the material, recording each piece of context exactly once at the highest level where it applies. The two rules that matter most: describe general-to-specific, and never repeat inherited information.
What are the levels and which ones do I need?
The full ISAD(G) model offers six levels, but it is a toolkit, not a checklist:
| Level | Describes | Typical use |
|---|---|---|
| Fonds | All records of one creator | Almost always present |
| Sub-fonds | A subordinate body or major function | Large or complex creators only |
| Series | Records of one activity/form | Very common |
| Sub-series | A subdivision of a series | When a series is large |
| File | A folder/unit of related items | Common for access |
| Item | A single document | Selective, high-value material |
Use only what your material supports. A small personal archive might be just fonds and file.
How do the two core principles work?
General to specific. Describe the fonds as a whole first; its title, dates, creator and administrative history set the context for everything beneath. Then move down, each level adding only what is new.
Non-repetition (ISAD(G) 2.4). Do not restate at file level what you already said at series level. If the creator is stated at the fonds, lower levels inherit it. This is what keeps multilevel finding aids readable instead of bloated.
text
FONDS: Records of Thornbury Engineering Co., 1886-1974 (creator stated here)
SERIES: Minute books, 1886-1970
FILE: Board minutes, 1923
(do NOT repeat "Thornbury Engineering Co." here — inherited)How do I apply levels step by step?
- Identify the fonds boundary by provenance — what one creator produced.
- Describe the fonds fully: scope, dates, extent, creator, administrative history.
- Identify series by the creator's functions/forms (see arrangement first).
- Add lower levels only where they exist and add value.
- Set the level element on each record (ISAD(G) 3.1.4 / EAD
level). - Strip repetition — review each lower record and delete inherited context.
- Vary depth by value — describe heavily used series deeper, lightly used ones shallower.
How do I encode the level in EAD?
EAD3 carries the level on the component element:
xml
<archdesc level="fonds">
<did><unittitle>Records of Thornbury Engineering Co.</unittitle>
<unitdate normal="1886/1974">1886-1974</unitdate></did>
<dsc>
<c level="series">
<did><unittitle>Minute books</unittitle></did>
<c level="file">
<did><unittitle>Board minutes 1923</unittitle></did>
</c>
</c>
</dsc>
</archdesc>Use the controlled values your platform expects (fonds, subfonds, series, subseries, file, item); a typo here breaks faceting and display.
Why is uneven description depth acceptable?
It is not just acceptable — it is good practice. Resources are finite, and research value is uneven. Describing a heavily consulted correspondence series to file level while leaving routine financial vouchers at series level concentrates effort where users benefit. This aligns with minimal-processing thinking: get baseline access across the whole fonds first, then deepen selectively. Uniform item-level description of everything is usually a misallocation, not a virtue.
What pitfalls should I avoid?
- Forcing all six levels onto material that has three.
- Repeating inherited context — the classic cause of bloated, unreadable finding aids.
- Confusing fonds with collection — provenance versus artificial subject grouping.
- Inconsistent level values that break catalogue faceting.
- Item-level everything when series-level access would serve users just as well for far less effort.
Key Takeaways
- The model is fonds, sub-fonds, series, sub-series, file, item — use only the levels your material needs.
- Describe general-to-specific and never repeat information given at a higher level (ISAD(G) 2.4).
- Record the level explicitly (ISAD(G) 3.1.4 / EAD
level) using controlled values. - Vary description depth by research value and resources, not uniformity.
- A fonds is defined by provenance; a collection may be an artificial subject grouping.
- Strip inherited context on review to keep multilevel finding aids readable.
Frequently Asked Questions
What are the standard archival levels of description?
The common hierarchy is fonds, sub-fonds, series, sub-series, file, and item. ISAD(G) treats these as a flexible model — you use only the levels your material needs, not all six.
Do I have to describe at every level?
No. You describe at the levels that exist meaningfully in your material. Many collections go fonds to series to file and never use sub-fonds, sub-series or item level.
What is the rule of non-repetition?
ISAD(G) 2.4 says information already given at a higher level should not be repeated at lower levels. You record context once at the top and let lower levels inherit it.
How do I show which level a record is at?
Use the ISAD(G) 'level of description' element (3.1.4), and in EAD3 the level attribute on the component, with values like fonds, series, file and item.
Can I mix levels of detail across a collection?
Yes. You can describe one well-used series to item level and leave another at series level only. Description depth should follow research value and resources, not uniformity for its own sake.
What is the difference between a fonds and a collection?
A fonds is the whole of the records created by one creator through their activities, defined by provenance. A collection is often an artificial grouping assembled by subject; the terms are not interchangeable in strict practice.