method and records
The Laboratory Notebook and Contemporaneous Records
Why recording at the time is not the same as recording accurately later, the attributable-legible-original-accurate vocabulary, how to correct an entry without destroying it, bound versus electronic notebooks, and how long to keep them.
A laboratory notebook is not a diary of what happened. It is the evidence that what happened is what you say happened, written at the time by the person who did it. Everything distinctive about notebook practice follows from that one requirement, including the parts that look pedantic: the ink, the single line through an error, the initials in the margin. This page covers what to record, when, how to correct it, and how long to keep it — written for a laboratory of two or five people rather than a site with an archive department.
Contemporaneous, and why reconstruction fails
Contemporaneous means recorded as the activity occurs, not at the end of the session and not the following morning. The rule is often read as bureaucratic fussiness. It is not: memory does not degrade evenly. It preserves what was expected and loses what was incidental, so a reconstruction is systematically biased towards the intended procedure and away from what was actually done. The value you meant to weigh survives; the second attempt at the balance does not.
That bias matters because the incidental details are the ones consulted later. Nobody rereads a notebook to confirm a run went as planned. They read it because something is wrong, and the useful entry is almost always the anomaly — the reagent from a different lot, the interruption, the reading taken twice. Those are precisely the entries a reconstruction drops.
The scale of the consequence is easy to underestimate. A substantial proportion of irreproducible results trace not to fraud or to poor experimental design but to ordinary errors in recording, transcription and calculation, which persist undetected because the record is too thin to expose them 3. A contemporaneous record is the cheapest available control against that class of error, and unlike most controls it costs nothing but the discipline of writing while the work is in front of you.
Two practical corollaries. Write directly into the notebook rather than onto a scrap to be transferred later — the scrap is the original record, so transferring it creates a copy and destroys the original. And record the raw observation before the calculated result: balance reading before mass, absorbance before concentration. A result whose inputs were never written down cannot be rechecked, only repeated.
Attributable, legible, original, accurate
The regulated world compresses documentation practice into a short list of adjectives, usually given as attributable, legible, contemporaneous, original and accurate, with completeness, consistency, endurance and availability appended 1. The vocabulary is worth borrowing even by laboratories that will never be inspected, because each word names a specific failure mode rather than a general aspiration.
| Attribute | Requirement | Failure it names |
|---|---|---|
| Attributable | Every entry identifies who made it and when | An unsigned entry nobody can be asked about |
| Legible | Readable by someone else, permanently | Pencil, faded thermal print, private shorthand |
| Contemporaneous | Written as the activity occurs | Evening reconstruction that omits the anomalies |
| Original | The first record is retained, not just a transcription | The scrap discarded after copying into the notebook |
| Accurate | Values match observation, corrections are visible | Overwritten digits and silent rounding |
| Complete | Includes repeats, failures and deviations | Only the run that worked is written up |
| Enduring and available | Survives storage and can be found on request | A notebook that exists but nobody can locate |
Completeness is the one small laboratories fail most often, and it fails in a specific direction. Failed runs go unrecorded because they feel like wasted effort. They are the opposite: three failed dissolutions followed by one success is a solubility observation, and it is invisible if only the success was written down. Record the attempts that did not work, with the reason if it is known and an explicit note if it is not.
Corrections without obliteration
The convention is old, universal and takes five seconds. Draw a single thin line through the incorrect entry so that it remains readable. Write the correct value beside or above it. Initial and date the correction. Where the reason is not obvious, add a short one — transcription error, wrong column read, balance not tared.
- Strike through with one line only. The original value must stay legible.
- Write the replacement value adjacent to the strike-through, not over it.
- Initial and date the change, in the ink used for the rest of the entry.
- Add a brief reason where the change is not self-explanatory.
- Never use correction fluid, never overwrite a digit, never remove or replace a page.
The reason obliteration is treated so seriously is that it destroys information which cannot be recovered and, worse, it destroys the distinction between error and alteration. A crossed-out 4 replaced by a 7, initialled and dated, is an ordinary transcription correction that any reader will accept. A digit written over another digit is indistinguishable from a value changed after the fact to suit a result, and no explanation offered later can undo that ambiguity. The same is true of a removed page: the gap proves nothing about the content but proves that content is missing, which is worse than an embarrassing entry left in place.
Two related habits are worth adopting. Rule a diagonal line through unused space at the foot of a page before starting the next, so that nothing can be added later into a gap. And never leave a page blank between entries; if one is skipped, strike it through and initial it. Both remove the possibility of retrospective insertion, which is the exact question an unfriendly reader will ask about a record with space in it.
Bound, loose-leaf or electronic
Bound notebooks and electronic systems both produce defensible records. Loose sheets in a folder do not, and the reason is worth stating plainly: with a bound book, a missing page is visible; with a file of loose sheets, a missing sheet is invisible. Completeness has to be demonstrable, and only sequence demonstrates it.
| Format | Strength | Weakness |
|---|---|---|
| Bound, pre-numbered | Sequence is inherent; a missing page shows | No search, no backup, one physical copy at risk |
| Loose-leaf in a binder | Convenient to write and file | Nothing establishes completeness; avoid for primary records |
| Electronic notebook with audit trail | Searchable, backed up, changes logged automatically | Only as good as the audit trail and the backup regime |
| Instrument output files | The genuine original for instrument data | Worthless if not linked to the notebook entry that produced it |
If you use a bound book, number the pages before use if the supplier has not, keep a contents page at the front, and photograph or scan completed pages periodically — a fire or a flood is a more realistic threat to a small laboratory record than any dispute. If you use an electronic system, the single question to ask is whether it keeps an audit trail that records who changed what and when, and that the user cannot switch off or edit. Without that, an electronic notebook is a word processor, and a word processor holds no original: the current text has silently replaced every earlier version 1.
Hybrid working is normal and needs one rule. Instrument files are the original data for instrument results, so the notebook entry must carry the file identity — filename, run identifier, date — and the file must be kept. Copying a peak area into a notebook and deleting the chromatogram converts an original into a transcription, and no amount of care in the transcription restores what was lost 4.
What a notebook proves, and how long to keep it
In a dispute — a challenged result, a question of who did what and when, a query from a collaborator or a funder — the notebook is not merely helpful. It is usually the only evidence, because everything else is recollection. What it can establish is narrow but valuable: that a particular person performed a particular operation on a particular date, using material with a stated identity, and observed particular values.
What it cannot establish is that those values were correct. That distinction disciplines the whole practice. A notebook is credible to the extent that it looks like a record kept for its own sake rather than one assembled for the occasion, and the markers of that are unglamorous: consistent ink, entries in date order, failures recorded alongside successes, corrections visible and initialled, no gaps. A tidy notebook containing only successful runs is less convincing than a messy one containing everything, and a reader assessing it will weigh the corrections as evidence of good faith rather than of carelessness 2.
Retention runs from the last use of the record, not from the date the notebook was filled. Practically: keep notebooks for at least the period any resulting data may be questioned, which for published or externally reported work usually means years rather than months, and check whether an institution, funder or client imposes a longer minimum — those obligations are commonly six or ten years and they override any internal preference. Keep them somewhere dry, indexed well enough to be found, and readable by someone who has left. A record that cannot be located on request has already failed, whatever its contents.
None of this requires a system. It requires one book per person, numbered pages, a pen, and the habit of writing while the work is in front of you. The laboratories that keep good records are rarely the ones with the best software; they are the ones where writing things down is part of doing the work rather than something done afterwards, when it has already become reconstruction.