How to Read an Old Ship's Log: Columns, Conventions, and the Sea Day

This is a working guide to how to read an old ship's log — the sea-day date, the course and distance columns, observed versus dead-reckoned position, bearings, wind, and the remarks. It matters because the errors that survive into published work and into climate datasets are almost always conventions misread as plain facts: a date that is not the calendar date, a longitude measured from a meridian that is not Greenwich.

An old ship's log is a ruled ledger to be read column by column, not a narrative to be read line by line. Work across the page in order: the date at the head (which, under the sea-day convention, runs from noon to noon, so the PM entries come before the AM entries), the compass course steered, the distance run as measured by log and line, the noon latitude and longitude — distinguishing observed positions from dead reckoning — the bearings taken to land or other vessels, and the remarks column, which carries the day's events. The single most important caution is that no standard logbook format existed; Mystic Seaport's guide to an 1828 logbook page makes the point plainly, and the printed column headings on any given volume are your best guide to what that particular keeper meant to record.

First, establish what you are actually holding

Before reading a column, identify the genre and the hand that produced it. A log is an official operational record; a journal may be a personal one. The Nantucket Historical Association's guide to its ships' logs collection draws the line usefully: logs, generally kept by captains or mates, record daily position, weather, vessels sighted, ports visited, and — on whaling voyages — whales seen and taken; journals, often kept by captains' wives, passengers, or crew, record accidents, illness, pastimes, and port encounters. The two genres frequently sit in the same box and answer different questions.

Provenance matters just as much within official series. The National Archives notes that Royal Navy captains' logs (ADM 51) were compiled from the masters' logs, with the captain adding what he judged relevant or was obliged by regulation to give; lieutenants' logs were derived in the same way and are now held by the National Maritime Museum, which describes them as regulatory records rather than private diaries. From 1799 onwards these officers' logs were superseded by ships' logs in ADM 53. The practical consequence: two logs from the same ship and date are often copies at one remove from a common source, and their divergences are evidence about the copying, not necessarily about the weather. Treating them as independent witnesses — and collating them as versions rather than merging them — is the honest procedure.

Mystic Seaport adds one more link in the chain: data were often collected on watch by the mate in charge and chalked onto a log slate, then written up into the book at the end of the sea day. The fair copy you are reading is already a transcription.

The date at the head of the page is not the calendar date

This is where datasets get corrupted. Under the traditional nautical convention, the sea day ran noon to noon, so the day entered in the log as 10 May began at civil noon on 9 May and ended at civil noon on 10 May — and, as the Naval Marine Archive puts it, PM thus came before AM. A page's afternoon entries belong to the previous civil date.

Three qualifications keep this from becoming a mechanical rule you can apply everywhere.

The Admiralty order of 11 October 1805

The order moved Royal Navy log reckoning from noon-to-noon to midnight-to-midnight, but implementation was delayed and uneven. Masters in particular went on starting the day at noon for years afterwards, which produces genuine discrepancies between a captain's and a master's log of the same ship on the same date. There is no safe blanket transition date for masters, captains, lieutenants, merchant services, East India Company vessels, or non-British navies. Establish the practice of your specific volume from internal evidence — the position of the noon observation relative to the day's events is usually decisive.

The noon sight is not the same claim as the day boundary

The meridian altitude at noon supplied the daily latitude check and structured the page around it, but the surviving evidence supports the noon-to-noon convention far better than it supports any universal formal rule that the sight itself reset the day for every service.

The astronomical day is a separate story

Astronomers agreed to start the astronomical day at midnight only in 1925, to coincide with the civil day. Do not import that date into naval log practice; they are unrelated conventions that happen to share a vocabulary.

Reading the columns

Course

The courses column records the compass course steered by the helmsman — not a true bearing. The National Archives Foundation's column-by-column logbook guide is explicit on this, and the same page defines the variation column as the difference between magnetic and true north. If your log records variation, it is telling you how to convert; if it does not, you are looking at magnetic courses and must reconstruct variation from period charts before plotting. Deviation from the ship's own iron is a further, usually undocumented, term.

Courses are written in the point system — NNE, SbE, WSW — and are frequently abbreviated further, with by rendered as a bare b. Transcribe the abbreviation as written; expand it, if at all, in a separate layer, on the same principle that governs manuscript abbreviations generally.

Distance and speed

The knots-and-fathoms column records speed as measured by a chip log or, later, a taffrail (patent) log. A knot is one nautical mile per hour: a line knotted at fixed intervals was streamed astern and the knots run out in a timed interval counted, using a half-minute glass in the classic arrangement described by the Shipwreck Centre and Maritime Museum. Royal Museums Greenwich's how to make a log line explainer works from a nautical mile of 6,080 feet.

Two errors follow from misreading this column. The first is treating a knot figure as a distance; it is a rate, and the day's run is the aggregation of watch-by-watch entries. The second is assuming units are stable. The International Hydrographic Review's history of the nautical mile records that historical leagues varied — a British sea league of 2.43 nautical miles among them — so any conversion belongs in a documented derived field, never applied silently to the transcribed number.

Latitude and longitude, observed versus dead reckoning

The noon observations sit in the middle of the page, and the Archives Foundation guide lists the usual set: distance per log, latitude DR, longitude DR, latitude observed by sextant, longitude often by chronometer, current where available, and variation. The distinction between the DR pair and the observed pair is the most consequential thing on the page.

Dead reckoning is an estimate propagated from the last good fix using course, speed, and elapsed time — the standard explanation of DR as an estimated position makes clear that errors accumulate until an observation resets them. CLIWOC found accumulated DR error severe enough to place reconstructed tracks over land. Record which is which; a merged position column destroys information you cannot rebuild.

Longitude carries a second trap. Do not assume Greenwich. In building its database, the CLIWOC team identified 646 different zero meridians in use across the logs — Start Point, Ushant, Cape Roxent, Madeira, Isle of May Bay and many more — with a single voyage sometimes shifting reference as new landfalls were sighted. The meridian is often stated only in a heading, a preface, or implicitly by the landfall named in the remarks.

Bearings

Bearings locate land, other vessels, or fixed objects relative to the compass, and two or more of them can support a fix. They are usually recorded in the remarks column rather than a ruled field, typically as an object, a compass direction, and an estimated distance: Cape Clear NNE 5 leagues. Read them alongside the DR positions — a bearing to a known headland is frequently the only independent check on a run of estimated positions, and the moment where a track can be corrected.

Wind and weather

Wind is given as direction blown from plus force. Force may be a Beaufort number or a pre-Beaufort word — "fresh," "moderate," "a hard gale." Resist mapping the words onto modern numbers. The Met Office's history of the Beaufort scale dates its creation to 1806 with thirteen forces, revision to the familiar twelve in 1807, and the final 1810 version adding the effects of each force on a Royal Naval frigate. Adoption by particular navies, merchant services, and nations came later and unevenly, and needs source-specific checking. The weather column often uses Beaufort's letter code, where b is blue sky and bc blue sky with cloud.

The remarks column

For most historians this is the part that carries the research. The Archives Foundation describes it as the place for the important events of the day: setting and striking sail, soundings, various navigational measurements, and the mundane accounting of wood, coal, and fresh water. It is also where illness, punishment, deaths, sightings, encounters with other vessels, and landfalls appear — the material that CLIWOC's compilers logged as miscellaneous facts alongside the meteorology. The remarks column is the narrative spine of the page and, characteristically, the least ruled and most cramped part of it.

Transcribe before you interpret

The discipline that protects a logbook project is the same one that protects any manuscript edition: produce a faithful reading first, and put every conversion in a separate, labelled layer. The TEI guidance on levels of manuscript transcription is a useful frame here — a philological transcription follows the orthography and registers abbreviations, insertions, and deletions, making clear which letters stand in the manuscript and which are supplied editorially. For a log page that means keeping the sea date as written, the compass course as written, the knots as written, and the longitude as written against its own meridian, then deriving the civil date, true course, distance in nautical miles, and Greenwich longitude as separate fields. CLIWOC did exactly this, preserving original calendar and time fields alongside normalised UTC positions and DR indicators. If you are turning transcriptions into CSV or JSON, the faithful text stays the master and every derived value stays recoverable. The general case for keeping normalisation as a documented layer rather than a replacement holds especially firmly here, because a corrected date or converted longitude is indistinguishable from an observed one once it has overwritten the original.

When it is not one page but four hundred

Reading a single page carefully is a paleography problem. Reading a series is a throughput problem, and it is the reason logbooks sit at the centre of historical climate and scientific data rescue: ICOADS now spans 1662 to the present, CLIWOC alone digitised 1,674 logbooks covering 4,942 voyages and 280,195 reports, and NCEI holds images of 1,235 East India Company logbooks from 1786–1834. Volunteer transcription carried much of this work and remains slow and supply-limited; Old Weather, the best-known effort, is currently resting. The tradeoffs between volunteer and machine routes are worth thinking through series by series, as is the specific problem of getting the numbers off the page without losing the grid.

Machine transcription is now a defensible first pass rather than an unattended pipeline. A 2025 study of context-aware tabular recognition on historical logbooks reports improved word and character error rates on its own dataset while still concluding that dense cells, mixed printed and handwritten material, and numeral confusions such as I/1 and 0/9 make image-led human verification the working standard.

This is the stage where Leo fits. Its transcription model, ATR-1, reads Latin-script material without per-corpus training — which matters for logbooks, since CLIWOC's holdings are English, Dutch, Spanish, and French vernacular records rather than a single-language corpus, and the writing system, not the language, is the constraint. It transcribes what is on the page: the abbreviated compass course, the archaic spelling in the remarks, the number as written rather than as it "should" read. Translation, if you need it for a Dutch or Spanish log, is a separate operation that writes to a new tab and leaves the base transcription untouched. The genuine weak spot is worth stating plainly: pages where dominant pre-printed structure meets dense handwriting — which describes many ruled logbook forms exactly — are the hardest case for any model, including this one, and complex tabular layouts vary. Treat machine output on such pages as a draft to be checked cell by cell against the image, which is what verifying transcription accuracy means in practice. What you gain is a searchable text layer over the whole series and a place to keep the images, transcriptions, and archival metadata together; what you still owe the record is the reading.

The habit that makes the difference

Read the printed column headings before you read a word of the entries, and read them on each new volume rather than assuming continuity. Establish the day convention from the page's own structure. Note the meridian. Keep observed and estimated positions apart. Copy what is written and derive everything else beside it. The particular value of a logbook is that its keeper recorded, hour by hour, what he could actually measure and see — and the reader's job is to carry that distinction forward intact, so that the next person to open the volume, or the dataset built from it, can still tell the observation from the inference.

Frequently Asked Questions

How do you read an old ship's log page column by column?

Read an old ship's log as a ruled ledger, not a narrative. Start with the date at the head of the page, remembering that under the sea-day convention the day runs noon to noon, so the PM entries precede the AM ones. Then work across: the compass course steered, the distance and speed run by log and line, the noon latitude and longitude — keeping observed positions separate from dead reckoning — any bearings to land or other vessels, wind direction and force, and finally the remarks, which carry the day's events. No standard format existed, so read the printed column headings on each volume first.

What is the sea day in a ship's log, and why does PM come before AM?

The sea day ran from noon to noon rather than midnight to midnight, so a page dated 10 May began at civil noon on 9 May and ended at civil noon on 10 May. That is why the afternoon entries appear above the morning ones: the PM material belongs to the previous calendar date. An Admiralty order of 11 October 1805 moved Royal Navy reckoning to midnight-to-midnight, but implementation was slow and uneven, and masters in particular kept starting the day at noon for years. Establish the convention from your own volume's internal structure rather than applying a blanket transition date.

What is the difference between an observed position and dead reckoning in a logbook?

An observed position comes from an instrument — latitude by meridian altitude at noon with a sextant, longitude often by chronometer — while dead reckoning is an estimate propagated from the last good fix using course, speed, and elapsed time. DR error accumulates until an observation resets it; CLIWOC found accumulated error severe enough to place reconstructed tracks over land. Logs commonly record both pairs side by side, and the distinction is the most consequential thing on the page. Keep the two in separate fields when you transcribe: a merged position column destroys information you cannot rebuild afterwards.

Was longitude in old ship's logs always measured from Greenwich?

No. Do not assume Greenwich. In compiling its database, the CLIWOC team identified 646 different zero meridians in use across the logbooks — Start Point, Ushant, Cape Roxent, Madeira, Isle of May Bay among them — and a single voyage sometimes shifted its reference as new landfalls were sighted. The meridian in use is often stated only in a heading or a preface, or implied by the landfall named in the remarks. Transcribe the longitude as written against its own meridian, then derive a Greenwich value in a separate, labelled field so the original stays recoverable.

Can AI transcribe old ship's logbooks accurately?

Machine transcription is now a defensible first pass on logbooks, not an unattended pipeline. Leo's model, ATR-1, reads Latin-script material without per-corpus training, which suits logbook series that mix English, Dutch, Spanish, and French, since the writing system rather than the language is the constraint. It transcribes what is on the page — abbreviated compass courses, archaic spellings, numerals as written. The honest weak spot is pages where dominant pre-printed structure meets dense handwriting, which describes many ruled logbook forms; complex tabular layouts vary. Treat output on such pages as a draft to be checked cell by cell against the image.