Corpus

The Barbados Mercury, 1783–1848

2,334 issues of a colonial commercial paper, from the Barbados Archives Department. Half of it is advertising, and that is the point.


What the paper was

Of the 119,700 items read so far, 46.2% are advertisements and 9.5% are articles. Set beside the Madras Courier, which runs 15.9% advertising and 25.4% articles, the two papers are different animals under one crown: a commercial sheet and a government news-paper. Barbados counts in pounds, dollars, hogsheads and Portuguese moidores; Madras in pagodas and rupees.

The existing OCR breaks on the long s, so a search for verse misses verfe and most of the eighteenth century is unfindable.


The advertisements name enslaved people

This is the most consequential material in the corpus and it should be stated plainly rather than buried in a statistic. 12,291 of the items read so far use the vocabulary of slavery, and 11,200 of those carry at least one name. Those names — Harry and Jemmy-Guy, absconded, with a warning against harbouring them; Oxford, a creole of Antigua — are in many cases the only surviving record that the person existed.

Runaway advertisements name people who appear in almost no other surviving record. Harry and Jemmy-Guy, absconded, with a warning to others against harbouring them. Jack, a cooper and a creole of the island. Oxford, a creole of Antigua, run away from a Mr Connor in Bridgetown.

We are not publishing that material here. It has been offered to the archive that holds the originals and to the scholarly projects built for exactly this class of document, whose infrastructure and care it deserves more than ours.


Provenance and honesty

Source images: Barbados Archives Department, via the Digital Library of the Caribbean. Public domain by age. Fetching was serial and paced.

A first count of the slavery vocabulary gave 29.7%. That figure was wrong: the keyword list included for sale and harbour, which match ordinary goods and shipping and matched 5,119 items on their own. The corrected figure was 10.6% then and is 10.3% now, on four times the corpus. The correction is enforced in the script by a two-sided control: the excluded terms must still be findable, or the count refuses to print. They are — for sale matches 27,123 items and harbour 1,865 — which is what makes their exclusion a judgment about meaning rather than an absence of data. It is recorded here because a number that moves by a factor of three under scrutiny should be visible, not quietly replaced.