Evidence in numbers
How sure are the positions on the maps, and what do the pages behind them rest on? This page counts the site’s own pages and sources. It counts what I read and cited, and it is not a verdict on the places or on the scholars whose work I read.
How sure the positions are
The site has 134 place pages. Each marks its position with one of three flags, explained on The evidence: site where a record fixes the position of the thing itself, approximate where the marker stands for a place or building near it, and uncertain where the position is only a candidate or an estimate. Of the 134, 11 are site (8%), 97 approximate (72%) and 26 uncertain (19%). The flags are my judgement, made page by page, and they are not measured against any outside standard.
The figure sorts the pages by the kind of play they serve. A place that serves plays of more than one kind is counted in each, so the four kinds add up to more than the 99 places that are settings. Only the histories have a site position among their settings: Berkeley Castle, Flint Castle, Roxburgh Castle, Sandal Castle (Wakefield) and Swinstead Abbey. No tragedy, comedy or romance place is site. The other 6 site positions belong to pages that are not settings in a play: Anne Hathaway’s Cottage, Cade Street, Heathfield, Shakespeare’s Birthplace, The Rose, The Theatre and The first Globe. One reason is that the comedies and tragedies are mostly cities and regions, which are not points, while the histories have castles, abbeys and battlefields that heritage bodies have surveyed. I have not tested that reason.
Where the positions come from
Every site position, 11 in all, was taken from a heritage record, and none of the other 107 positions is site. That is how I set the flag: a position is site only when a record fixes the structure itself, so a position from Wikipedia, Natural Earth or OpenStreetMap is approximate or uncertain at best. Heritage records fix 27 positions, of which 11 are site, 13 approximate and 3 uncertain, because a record can fix a ruin or a monument without fixing the thing a play names (a battlefield, say, or a lost castle).
Natural Earth supplies the most positions, 47 of the 134. These are city or town points, or points I set by hand against its coast, and they can only mark a place approximately. Wikipedia supplies 30. Of the 27 heritage-record positions, 16 cite a source the registry marks as an official record and 11 were read on mirror sites that reproduce the record, such as British Listed Buildings. Outside Natural Earth, 68 of the 87 positions rest on a source I read only through a search extract, so a position is only as good as the extract was.
Real, historical, disputed and fictional
Each place also carries a flag for what kind of thing it is. 111 pages are real places, 13 are disputed identifications, 8 are historical events and 2 are fictional places, Belmont and Illyria. Both fictional places are flagged uncertain, because a place that does not exist has no position to be sure of; the markers show where an editor or a tradition puts them, and the pages say so. Of the 13 disputed pages, 6 are uncertain, 5 approximate and 2 site; the two site ones, Roxburgh Castle and Cade Street, have known positions; what is disputed is whether the castle is the play’s and where Cade died.
Where the doubt comes from
Why is a place doubtful? The reasons differ, and they bear on a different question from the one the flags answer. A battlefield can be uncertain because nobody agrees where the armies stood, which says nothing about what the play knew. A place can be uncertain because the play gives it no site at all. I read the 23 places flagged uncertain, disputed or fictional and sorted each by the reason its page gives, and I sorted the 39 plays by the main source of the story: chronicle, Plutarch, romance, novella, poem or legend, or no single source. Both sortings are my judgement. King Lear and Macbeth go with the chronicles, Cymbeline and Hamlet with the romances, and moving any of them changes the shares by a point or two and not the order.
The bars count play-place pairs, as Patterns does, so a place used by two plays counts twice: 124 pairs in all, 24 of them flagged. Among the chronicle plays, 13 of 69 pairs are flagged, and 7 of those are about the ground (a battlefield or a building whose site is lost or contested) and 5 about a later tradition or my reading; only 1, Macduff’s Fife, has no site to show. Among the romance, novella and poem plays, 10 of 31 pairs are flagged, and 6 of those are places the play does not site, against none that are about the ground. The 17 Plutarch pairs have none flagged, which is partly because Rome and Athens are real towns at map scale; the one-source group has 1 of 7, Navarre.
That gives a cleaner reading than the flags alone, and a limited one. Taking the places themselves, 8 of the 23 are doubtful because no site or only a region can be shown, and the other 15 because of the ground or a tradition. The 8 are:
| Place | Plays | What the page says |
|---|---|---|
| Bohemia | The Winter’s Tale | A real kingdom with no sea coast, where the play puts a shore. Greene’s story already had the coast. |
| Macduff’s Fife | Macbeth | The peninsula; I found no fixed site for Macduff’s castle. |
| Navarre | Love’s Labour’s Lost | The country, marked at Pamplona; the king’s park has no site, and the guides call the court partly invented. |
| Roussillon | All’s Well That Ends Well | A region on the Spanish border by my reading of one reference text; no palace site. |
| Sicilia | The Winter’s Tale | The kingdom, marked at the middle of the island; no court site is given. |
| Belmont | The Merchant of Venice | Fictional. The source story already has a Belmonte. No source I read ties it to a place; my marker is a placeholder near Venice. |
| Illyria | Twelfth Night | The ancient Adriatic coast, but the duchy has no recorded site and reads as Italian. |
| Pentapolis | Pericles, Prince of Tyre | Two readings, Cyrene or a Greek island; the old tale the play follows gives no site. |
The places doubtful for another reason are, for the ground: Agincourt, Barnet, Bosworth Field, Crécy, Iden’s garden, Mortimer’s Cross and Ravenspurgh; and for a tradition or my reading: Cawdor, Cyprus, Elsinore, The Forest of Arden, Gaultree Forest, Glamis, Roxburgh Castle and Troy. These say something about the record or about later tellers, and little about what Shakespeare knew.
Three things limit this. A region-only pin may mean the sources I read are silent, not that the play is vague, and I did not read the plays. The numbers are small, 17 and 7 pairs in two of the four groups. And the same person made the flags, the sorting by source and the sorting by reason, so the result shows how I read the pages and not a measurement of the plays’ geography. It fits the idea that a play’s places are firmer where its source is a chronicle or Plutarch and looser where its source is a romance, and it does not prove it.
Where two positions disagree
On 17 pages, a note or the text gives a second position for the same place, or a second candidate for the ground, and states how far apart they are. I collected these by hand from the pages; they are not found by a script, and a page may disagree with a source without giving a distance, so this is a floor and not a complete count. Of the 17, 11 are under 400 m apart and 6 are a kilometre or more. On a wide map, where a pixel can be several kilometres, even the largest gap would barely show; on a town frame, where a pixel is a few metres, the small ones show.
| Place | The marker uses | The other position | Apart | Figure from |
|---|---|---|---|---|
| Crécy | the traditional battlefield beside Crécy, Wikipedia’s infobox position | Livingston’s case for high ground by the Forest of Crécy | more than 5 km | the page, from his argument |
| Iden’s garden | the village of Hothfield, named by Holinshed | Ripple Court, claimed in Kent’s heritage record | about 6 km | my arithmetic |
| Halae (Timon’s tomb) | Pleiades’ position for Halai Aixonides | another gazetteer’s position, to the south | about 2 km | the page |
| Mortimer’s Cross | the 1799 monument at Kingsland | the other candidate site for the fighting, at or just south of Mortimer’s Cross | about a mile | the Battlefields Trust, as the page reports it |
| Misenum | Wikipedia’s position for Cape Miseno | a survey record’s position for the Roman port | about 1.4 km | my arithmetic |
| Tewkesbury | the centre of Historic England’s battlefield report | where other historians put the armies | about 1,000 m | Historic England’s report |
| Charlecote Park | the Gazetteer for Great Britain’s grid reference | an encyclopaedia article’s position for the house | about 370 m | the page |
| Barnet | the Hadley Highstone obelisk | the obelisk’s earlier position, before it was moved around 1840 | about 180 m | the page |
| Thebes | Wikipedia’s position for modern Thiva | Wikivoyage’s position for the Cadmea | under 200 m | my arithmetic |
| Temple Grafton | the listed church of St Andrew | Wikipedia’s position for the village | about 120 m | the page |
| Temple Garden | Wikipedia’s position for the Temple | Historic England’s position for the Inner Temple Garden | about 100 m | my arithmetic |
| Baynard’s Castle | the scheduled monument of the first castle | the site of the 1483 mansion | about 100 m | the monument record |
| Mary Arden’s farm, Wilmcote | Glebe Farm, the listed Mary Arden’s House | Palmer’s Farmhouse, long shown as her house | about 80 m | the page |
| 35 Great St Helen’s | the editor’s point for 35 Great St Helen’s | the parish point | about 46 m | the page |
| Guild Chapel and school | OpenStreetMap’s centre of the Guild Chapel | an encyclopaedia article’s position for the Guildhall | about 25 m | the page |
| Leicester Cathedral | the editor’s position for the cathedral | the header of Wikipedia’s article | about 13 m | the page |
| Silver Street and the Mountjoys’ house | OpenStreetMap’s memorial plaque beside the garden | a mirror of Wikipedia’s position, measured from the garden’s centre | about 10 m | the page |
The distances are as each page states them. Where the figure is from my arithmetic, it is the distance between the two positions as the page gives them, and the page says so; where it is from a source, the page names it.
What the pages cite
The 134 place pages carry 1,166 citations to 588 different sources, out of the 682 in the registry. A page cites a median of 5.5 different sources, from 1 to 15, and has a median of 8 citations, with the most heavily cited page at 32. These are the citations in the pages’ own text; the notes under figures and the notes on the positions are not counted.
| Kind of source | Citations | Share | Different sources |
|---|---|---|---|
| Secondary: reference work | 442 | 38% | 197 |
| Secondary: Wikipedia | 302 | 26% | 133 |
| Scholarly | 160 | 14% | 93 |
| Secondary: blog | 65 | 6% | 38 |
| Official record | 64 | 5% | 41 |
| Secondary: amateur site | 61 | 5% | 43 |
| Secondary: news | 61 | 5% | 35 |
| Primary | 11 | 1% | 8 |
How much was read is a separate question, and the answer is the weak point of the whole site. Of the 1,166 citations, 956 (82%) are to sources I read only through a search extract; 125 (11%) are to pages I fetched and read through a tool-generated summary; and only 77 (7%) are to sources I read in full. The remaining 8 are to an abstract or to a work I know only from another source’s citation. An extract is usually the passage the claim rests on, but I did not see what surrounds it. 59 pages rest wholly on extracts, abstracts or works not read.
Place pages keep three kinds of statement apart: what the plays say, what the documents and the ground show, and tradition, inference and open questions. Counting the words of the three, 22% describe what the plays say, 46% what the documents and the ground show, and 32% tradition, inference and open questions. 116 pages say in so many words that I did not read, or did not search, the play, and 7 mark a figure as my own arithmetic.
The thinnest pages
A page can be well sourced on reference works and still lack a primary source, an official record or a scholarly work, which are the three kinds the registry does not class as secondary. 53 of the 134 pages cite none of the three in their own text. A page can still rest its position on an official record: Roxburgh Castle’s does, from Canmore, but the page does not cite it in its text. The share is lowest for the comedies (2 of 16) and highest for the tragedies (12 of 26) and histories (24 of 50). I have not worked out why. It may reflect when each page was written and which extracts I could find, and not the places.
Six pages cite two sources or fewer: 35 Great St Helen’s, Cawdor, Hall’s Croft, Macduff’s Fife, New Place and Scone. This list is my own to-do list. A page on it is thin, not wrong, and a better source may exist that I did not reach, because I cannot fetch Wikipedia and I read most pages only through extracts.
The pages that cite no primary, official or scholarly source, each listed once under the first kind of play it serves:
- Histories (24)
- Amboise, Baynard’s Castle, Berkeley Castle, Blackfriars, Bordeaux, Calais, Coventry (the lists), Crosby Place, Crécy, Dartford, Flint Castle, Mortimer’s Cross, Orléans, Paris, Poitiers, Pomfret Castle (Pontefract), Rouen, Roxburgh Castle, Southampton, St Albans, Temple Garden, Tower of London, Westminster Abbey and Whitefriars.
- Tragedies (12)
- Actium, Alexandria, Birnam Wood, Caesar’s pyre, Cawdor, Elsinore, Forres, Forum and Rostra, Glamis, Inverness, Macduff’s Fife and Scone.
- Comedies (2)
- Belmont and Milan.
- Romances (5)
- Antioch, Mytilene, Tarsus, Thebes and Tunis.
- Pages that are not settings in a play (10)
- 35 Great St Helen’s, Bankside and the Clink liberty, Castillon, Hall’s Croft, Leicester, Leicester Cathedral, Lumphanan, New Place, Troyes and Wark on Tweed.
How this was made, and what it leaves out
- What is counted. All 134 place pages: the settings of the plays, the reference points flagged ‘Not in the play’, and the pages for London’s playhouses, Shakespeare in London, Stratford, Warwickshire and the marriage of 1582. The kinds of play are the ones on Patterns, and 99 pages are settings on a play map.
- The flags are mine. Site, approximate and uncertain, and real, historical, disputed and fictional, are my judgement, and the counts show how I applied them, not how accurate the positions are. A position flagged site can still be wrong by the width of a building.
- What the counts cannot show. Whether a claim is true. A page with many sources can be wrong and a page with two can be right. The counts say what I cited and how much I read, and not whether a cited source is right.
- Hand-set parts. The table of disagreements, the sorting of plays by the main source of the story and the reason given for each doubtful place are collected by hand. The two sortings are at the top of
tools/genevidence.py, and the script counts from them. - No tests. The page reports counts and shares. With pages written over many weeks, and the same person reading and flagging, I do not offer a significance test.
- Remaking it.
python3 tools/genevidence.py <site folder>(it needsphpon the path) rewrites the figures and the numbers fromdata/*.geojson,content/placesandinc/sources.php. After a new page, a new source or a changed flag, run it again and check this text against the new numbers.
The figures are drawn from the site’s own data; no map is used.