Reading chronometer rate records: error, correction and uncertainty

A table of time comparisons can look authoritative before its subject and origin are known. Record the archive, document identifier, instrument designation, writer where known, date range and source of the time reference. Keep a photograph or faithful transcription alongside your interpretation. Do not attach a loose table to a watch because a maker’s name matches: makers supplied many instruments and numbers may refer to case, movement, inventory or later issue systems.

This guide presents explicit analytical definitions and synthetic arithmetic examples. It does not reproduce an archival rating book, and the illustrated Barthet chronometer has no verified rate series here. The museum catalogue supports its identity and described construction, not the numerical examples below. Maintaining this boundary makes the guide useful without manufacturing a testing history for a real collection object.

For this guide, define error as indicated chronometer time minus reference time. A positive value means the chronometer reads fast; a negative value means it reads slow. Rate is the change in error divided by elapsed time. A chronometer can have a sizeable offset but a small interval rate, or show almost no offset at one comparison while changing quickly. A single accurate-looking reading therefore does not establish stable behaviour.

These definitions are an explicit convention, not a claim that every historical author used identical headings. Preserve terms such as fast, slow, gaining, losing and correction exactly as written. Look for a worked example or explanatory heading before converting signs. An error in transcription of the convention can reverse an interpretation while leaving every copied digit intact. If the original remains ambiguous, mark the ambiguity rather than select whichever sign makes the result seem plausible.

Under our convention, correction equals reference time minus indicated time: it is the negative of the error. If an instrument reads 12 seconds fast, its error is plus 12 seconds and the correction applied to the reading is minus 12 seconds. Adding that correction produces the reference-time estimate for that observation. A historical writer may describe the same situation with a different sign label, so the original heading must travel with the copied value.

Retain the date and instant to which a correction belongs. An old correction does not apply indefinitely and certainly does not diagnose an instrument preserved centuries later. If a documentary analysis extrapolates using a stated rate, label the result as calculated and identify the assumptions. Do not silently turn a prediction into an observed reading. Missing entries, changes of reference or interventions can interrupt the basis for extrapolation.

Consider invented teaching readings, not measurements of the illustrated object. At the first comparison, error is plus 12 seconds. Exactly 48 hours later, error is plus 18 seconds. The change is plus six seconds across two days, so the interval average rate is plus three seconds per day. The chronometer was already fast and gained further during that interval. At the second observation, correction under our convention is minus 18 seconds.

If the second comparison occurred 36 hours later instead, the same six-second change would average four seconds per day. Counting two printed date headings would produce the wrong denominator. Use the actual elapsed interval where available. When the times of observation are absent, state whether the assumed interval is one, two or another number of days, and do not present the derived rate with precision the document cannot support.

A result printed to several decimal places can be less reliable than the original observations. Rounding, uncertain transcription, the reference clock and the timing of the comparison all limit what the numbers establish. If readings are recorded only to whole seconds, extra calculator digits do not create better evidence. State the observational precision where known, and use “not recorded” where it is absent. Do not invent confidence intervals without a justified measurement model.

Two endpoint readings reveal an average across the interval; they do not reveal the complete path between them. A temporary disturbance, varying conditions or compensating changes can remain invisible. Keep blank entries blank and preserve annotations about transport, winding, adjustments or repairs if the source supplies them. An unexplained change is an observation to investigate, not automatic proof of a particular broken component.

The Barthet catalogue describes a two-arm compensation balance and a dial thermometer. Those facts make temperature a relevant research question, but the presence of compensation is not evidence of perfect temperature independence. Likewise, Arnold No. 92’s helical spring and OZ balance are catalogued construction details, not a numerical performance guarantee. A table without temperatures cannot establish a temperature coefficient merely because the instrument includes compensating parts.

Keep documentary evidence and mechanical diagnosis separate. A change in average rate can raise a question, but it does not uniquely identify lubrication, balance-spring condition, escapement behaviour or a replacement part. Physical examination belongs with a qualified specialist, and the historical state may differ from the present one. A museum photograph, even an exposed-movement view, cannot certify that an instrument is ready for safe operation.

RMG’s explanation gives 15 degrees of longitude per hour of time difference. The magnitude of an uncorrected four-second time error therefore corresponds to one arcminute of longitude in the idealised time-angle relationship. This is an angular sensitivity, not a complete accuracy estimate for navigation. Local observation, astronomical reductions, reference conventions and other uncertainties remain outside that simple conversion.

Do not apply a universal distance conversion: longitude spacing depends on latitude. Do not automatically assign east or west from the sign of a chronometer error without specifying how local and reference time enter the particular calculation. Our learning tool limits itself to magnitude and deliberately does not calculate a navigational position. Document the reference meridian in a historical record; modern convention should not erase a source’s original coordinate system.

Use columns for document location, instrument identifier, comparison instant, indicated time, reference time, original recorded wording, interpreted error, elapsed interval and calculated rate. Add a note column for uncertainties and interventions. This structure keeps observations, transcription and interpretation distinct, consistent with the source-critical approach used throughout the encyclopedia. A later reader should be able to reproduce the arithmetic without accepting every historical inference.

Store the original image, transcription and revised interpretation as separate versions with dates. Keep a readable export and a backup beyond the same device. Cite the exact archival item and page if verified; a generic institutional home page is not a substitute. If only a catalogue summary was consulted, say so. Publishing an incomplete but honestly scoped record is preferable to inventing a document title, trial result or missing serial number.

A documentary calculation does not require operating the object. Do not wind an unassessed chronometer, move an exposed movement, release a transport device or attempt a detent adjustment to fill a spreadsheet. Ask an appropriately qualified conservator or specialist about safe examination, especially where service history and current condition are unknown. Conservation of historical evidence takes priority over restoration, and restoration takes priority over repair solely to make an object run.

For a proposed acquisition, separate authenticity, documented performance, present mechanical condition and price. None is established by this guide’s arithmetic. Historical rate records can contribute to provenance when their connection to the object is demonstrated, but they cannot alone prove originality of every component or present reliability. Retain unanswered questions and link them to the evidence that would be needed to resolve them.