Thirty-five manuscript logbooks, read page by page and transcribed word for word: 33,820 days at sea, each in the keeper’s own words, with his position as he wrote it and a link to the page itself. Press play and watch the fleet sail.
The only thing drawn under the data is a faint coastline. Every mark is a noon position a keeper worked out with quadrant and chronometer, and the tracks fill in the oceans themselves: the Atlantic drop south, Cape Horn, the Pacific line, Hawai‘i, and the crowded grounds of the Okhotsk Sea, Bering Strait and the western Arctic. In motion, a ship is drawn between her recorded positions in a straight line, and disappears when her log gives none for six weeks.
The voyages
Each line below is one voyage as its log records it: three or four years from a New England wharf to the far side of the world and back. Orange ticks are whale-strike days. The corpus is not a random sample. Most of these ships met each other: a log was chosen because a ship already transcribed had spoken her at sea.
Click a ship’s name to open her log in the explorer.
When two whalers met they hove to, traded news and oil counts, and each keeper wrote the other ship into his log. Twenty of those meetings survive here in both books. Laid side by side they check each other: the hour, the place, the barrels, and sometimes a name one hand left unreadable. A ship’s day ran noon to noon, so the two dates can differ by one.
Ice & climate
Whaling logs are among the only instruments that measured the 19th-century Arctic: dated positions with ice in sight, freeze-up and break-up, gales given hour by hour. This corpus holds 1,767 ice days (the Okhotsk Sea in June, the Chukchi edge in September, a winter frozen into Repulse Bay) and the eyewitness record of the days before the great fleet disaster of September 1871, when thirty-three ships were abandoned to the ice off Point Belcher.
Whales & nature
Every species had a name from the masthead: polar whales (bowheads), right whales, California greys and devil fish, sperm whales, finbacks, blackfish, grampus. The logs record where each was found, chased, struck, and lost. In the quiet stretches they record what else lived out there: walrus on the ice, goneys astern, terrapin taken at the Galápagos.
A catch record counts the whales a ship brought home. The logs also record the ones that got away with an iron in them. Each square below is one whale struck. About one in three was lost, and the share fell with every decade, from 41 percent in the 1840s to 18 percent in the 1870s.
People & events
Fires and mutinies, cholera quarantines, a boy lost at Tahiti, an Arctic explorer coming aboard for tea, four hundred and fifty emigrants carried from Hong Kong to California. Some of these events survive in two logs at once, told from both decks.
Explorer
Every entry in the corpus. Pick a voyage, choose what happened, or search the keepers’ own words, spelling and all. Open a day to read it as he wrote it, line for line, then follow the link to the manuscript page and check our reading against his hand.
The shore record
The logbooks say where a ship was. Dennis Wood, a New Bedford merchant, kept the other half of the record: four ledgers in which, from 1831 to 1876, he abstracted every whaling voyage the port could learn of: sailing, master, agent, each report that came home by letter or newspaper or a ship spoken at sea, the oil sent ahead, the arrival and the cargo “turned out.” The four volumes, held by the New Bedford Free Public Library and scanned on the Internet Archive, were read page by page in September 2026: 2,552 pages, 7,980 voyage entries (7,158 voyages once entries carried across pages are joined), 79,955 dated reports. No prior machine-readable transcription was found. Eight reports in ten can be placed on a map (one in ten by a latitude and longitude Wood copied down, seven in ten by a named port, island or ground), and every voyage is keyed to the whalinghistory.org registry, so Wood’s reports can be laid against the ships’ own logs.
Same map: what a New Bedford counting-house could see of the ocean, decade by decade. Watch the Hawaiian anchorage move: in the 1840s the fleet reported from Lahaina and Maui; by the 1860s the bubble is Honolulu alone. Fayal and Talcahuano hold steady as the sperm fleet’s post offices; Hakodate and Yokohama appear only after 1855. Hover a point for the vessel and the report as Wood wrote it. Reports with no year (3,132) and those with no place written are not drawn.
Five of our thirty-five logbooks are voyages Wood also followed. Below, each ship’s own track in blue, and on it, numbered, what reached New Bedford about her: the positions Wood copied from a newspaper’s “spoke” column and the ports a letter said she was in. Where the log has a position within three days of Wood’s date, the distance between the two records is given. The port distances are the interesting ones: a letter dated at Honolulu is often written weeks after the ship sailed.
Most of the ledger is routine. Twice the routine breaks and Wood writes a list. In June 1865 the Confederate cruiser Shenandoah burned the Arctic fleet a month after the war had ended, and Wood gave the victims a page of their own, with his running tally in the margin; twenty of the twenty-one also carry the loss on their voyage page, usually in red ink. In September 1871 the fleet was caught by the ice off Wainwright Inlet, and Wood wrote almost the same sentence on 23 vessel pages, each with the oil she had aboard and the bone left blank. He was working from a source that had one figure and not the other.
Wood almost never names the ship that “spoke” a whaler; the vessels he names are the ones that carried her oil home. Of 2,471 reports naming another vessel, 2,205 are freight (“shipped 400 Sp per bark Syren”). Resolving them maps the packet lines that ran between the Pacific fleet and New Bedford: Honolulu’s Syren and Iolani, the Talcahuano and Fayal carriers, and the clippers that took bone round the Horn. Where the whaler’s page and the carrier’s both record the shipment, the figures agree eight times in nine.
For 962 voyages Wood’s reports can be laid against the ship’s own daily positions in the American Offshore Whaling Logbook dataset: 9,028 reports compared.
Counted where the two are not in gross conflict; about a third of Wood’s explicit positions are, and that third includes the sea-day and civil-day dating problem (a ship’s day ran noon to noon; allowing one day of slack moves 11 points of them into agreement), Wood’s own slips, and the logbook transcriptions’ errors. The figures are an upper bound on Wood’s error, not a measure of it.
Every entry in the four volumes, by vessel. Pick a ship to read what Wood wrote about each of her voyages (heading, reports and arrival), with the volume and page to check the scan. Spelling is his; uncertain readings are in brackets.
The checks
Every reading on this page was made by an AI reader looking at an image of the manuscript, and AI readers make mistakes. These are the checks that have been run, and what each one found.
| What | The check | What it found |
|---|---|---|
| Whales struck, saved and lost | A data editor at whalinghistory.org read one whole voyage, the Arab, line by line against our data. In answer, 4,993 entries were read again blind, from the keeper’s words alone, and disagreements went to a third reader who saw both answers unlabelled. | Half of all strike days had carried a wrong or blank count. They are corrected here. On his 37 strike days the blind reading agrees with him on 37 of 37; the old data agreed on 21. |
| Positions | Every decimal position was worked out again from the position as written, then tested against a coastline, the length of each day’s run, known ports, and ships met at sea. About 810 rows went back to the page. | Positions falling on land dropped from 476 to 33, and day’s runs over 1,000 miles from 1,605 to 31. The rest are flagged in the data. |
| Dates and ships | The weekday each keeper wrote was checked against the date we assigned. Every voyage was then reconciled with Dennis Wood’s shore ledger. | Several bands of entries were a month or a year off and were re-read from the page. Nine stretches of log had been assigned to the wrong voyage and were moved. |
| The words themselves | No blind test has been run yet on the logbook text. While the page links were being checked, a second AI reader compared 67 entries with their pages. | In about half it read at least one word or figure differently from us. Most differences are small, a compass point or a spelling. A few change the sense. See the page before you quote a line. |
| Links to the manuscript | Two pages in each of the 35 volumes were opened and compared with the entry that cites them. | 67 of 70 opened on the right page. In three volumes our page numbers drift, so those links open the volume instead. |
| The shore ledger | Sixty pages were keyed a second time, blind. All four volumes have since been read a second time. | About one report row in twenty carries a wrong digit somewhere. In volume 1, where the second reading is finished, that falls from 5.9 to 1.3 percent. Those fixes are not yet in the data shown here. |
What none of this can see. All the readers are the same kind of AI. Where two of them share a mistake, no comparison between them will find it. One voyage checked by one careful person is the only outside test so far. If you find an error, that is the most useful thing you can send.