Finding
How long news actually took to cross the British empire, measured from two colonial newspapers rather than argued from one anecdote.
An eighteenth-century newspaper prints its own date, and most items carry a dateline naming where and when they were written: LONDON, 27th May. The gap between the two is how long that item took to travel. One item is an anecdote. Thousands give a distribution.
| origin | to Barbados | to Madras |
|---|---|---|
| London | 47 d (42–54) n=602 | 168 d (140–185) n=208 |
| Paris | 58 d (51–71) n=199 | 173 d (164–207) n=88 |
| Lisbon | 58 d (50–73) n=26 | 203 d (178–220) n=28 |
| Dublin | 59 d (50–76) n=38 | 151 d (143–222) n=11 |
| New-York | 45 d (38–59) n=50 | — |
| Calcutta | — | 22 d (20–30) n=119 |
| Bombay | — | 18 d (14–24) n=78 |
| local | 0 d (Bridge-Town) | 4 d (Madras) |
Median days from dateline to printing, interquartile range in brackets, n = items measured. Recomputed 6 September 2026 over 9,329 transcribed pages and 153,000 items, four times the corpus these figures were first published on.
Same empire, overlapping decades, and the Atlantic colony heard London more than three and a half times faster than the Indian one. The interquartile ranges are narrow, which matters: these are distributions, not a lucky pair of examples.
Paris trails London into both colonies, which is not a fact about the Channel. Paris news reached Madras and Bridgetown through London, and the lag is the reprinting. The size of that lag is itself informative: eleven days into Barbados against five into Madras, because on the longer route the reprinting delay is a smaller share of a much longer journey. Lisbon behaves the same way and more so, arriving in Madras thirty-five days behind London despite being closer to it.
Survival is not random: what was digitised is a sample of what survived, which is a sample of what was printed. A colonial paper also reprints selectively, so these are the lags of items an editor chose to carry, not of every ship that arrived.
The earlier version of this page said the numbers should tighten as the corpora grew, and they have — but one of them also moved. Barbados was published at 58 days on 116 observations and is 47 on 602. Two faults account for the difference. The Madras print-date parser was reading only about a quarter of issues, so the London figure there rested on 46 observations and now rests on 208. And the Barbados dateline parser was taking month abbreviations as place names: a dateline reading FEB. 12, 1803 names no place at all, but the abbreviating full stop satisfied the same delimiter a place name would, so seven of the paper’s nine commonest “origins” were months and the real London figure sat underneath them. Both are fixed in the published script, with a control that fails if either returns.
The direction of the finding never depended on either. 47 and 168 are further apart than 58 and 162, and the medians reproduce to within two days under an independently written implementation.