The talk that forms the basis for this post was written for the annual Gerald Aylmer seminar run by the Royal Historical Society and the National Archives, and was delivered on 29 February 2012. The day was given over to a series of great projects, most of which came out of historical geography, and I was charged with providing a capstone to the event, and presenting a more general overview of the relationship between history and geography. There was a good audience of academics, archivists and librarians, all with a strong digital cast of mind and I very much enjoyed it. Unusually, I am also pretty sure I still agree with the majority of it even some six months after I sat down to write it. At the same time, I have put off posting it until now because it just did not quite feel like a blog post - too much history, too much text, too much of an internal discussion among academics. I have also recently found myself wary of blogging, having discovered (who knew?) that blogs form a sort of publication that people occasionally read. But, the work of people like Andrew Prescott has also reminded me of just how important it is to continue having the discussion. In the nature of a public talk, the text is rather informal, the notations slapdash, and the links non-existent.
Place and the Intellectual
Politics of the Past
Currently there is a rather wonderful raproachment between historical
geographers and historians; with archivists and librarians (as usual) providing
the meat, gristle and spicy practical critique. This is brilliant. These are cognate disciplines
which need to be in constant dialogue. The
habits of mind and analytical tools of geographers need to inform our understanding
of the past; while the mental ticks of the historian, and the authority of
history as a literary genre, are necessary tools for communicating all kinds of
memory to a wider audience.
Frustratingly, over the last half century, this
cross-fertilisation hasn’t always flourished.
Geography Departments (and even historical geographers) have not done a
lot of talking to history departments, and vice versa. Many geographers, through the 70s and 80s, in
particular, became ever more engrossed in the technical manifestations of their
field; while many historians have fallen for the joys of the linguistic turn.
In part, this has been about funding and the structures of
higher education. In the great taxonomy
of knowledge inherent in the creation of the notion of STEM subjects vs the Humanities,
Geographers naturally gravitated towards the areas with more secure
funding. While historians, frustrated by
the changing politics of their field – the collapse of ‘modernity’ as a form of
historical explanation (both right looking and left looking) – pursued theory,
gender and language, to the exclusion of positivist measurable change – they
chose places of endless debate and disagreement – usually in the form of a
various versions of identity politics - as a route to a politicised audience. Similarly, whereas historical geographers
looked to Europe and a thriving institutional base; historians tended to more
frequently look westward to North America, where historical geography is almost
unknown – or at least denied the security of an extensive network of independent
academic departments.
All of which is simply to say, that we are confronted with
two fields that should be in constant dialogue, but which simply have not
passed much more than the odd civil word in the last few decades. In the process, they have developed different
technologies of knowing and different systems of training and analysis. It strikes me that an event celebrating
Gerald Aylmer’s cross genre engagement with history and archives, with the
structures of the archive, and the stories that can be told with them, is just
the right place to bring these disciplines back onto the same map, or at least
to reconsider where in a rapidly changing technical environment, that necessary
dialogue might take place.
Of course, for most of the last decade or so we have had the
‘spatial turn’ in history; and for longer than that, the creation of a
post-modern geography. Historians have
struggled to define ‘space’ and context in ever more material (if still rather
flabby) terms; while some historical geographers have taken theories of
discourse and language seriously and extended them to the clear air enclosed by
the mind-forged boundaries symbolically represented on every map.
But this has been little more than a casual rapprochement –
driven largely by academic fashion; and has not fundamentally changed the centre
ground of either discipline. Most historians still trade in text mediated by
uncertainty and theory; while historical geographers, strive to tie data to a
knowable and certain fragment of the world’s surface.
What I want to suggest today, is that something rather more
profound than the ‘spatial turn’ is also
happening in the background, and that it promises to force these disciplines
(and several others) back into a more direct relationship.
And this change, this possibility, is being driven by
technology; both in the form of the ‘infinite archive’ – the Western Text
Archive second edition; and also through the direct public access to a newly
usable online version of GIS-like tools, in the form of Google Maps and its
many imitators.
To deal with historians and the infinite archive first - I
don’t think that historians have quite twigged it yet – though librarians and
archivists certainly have - but the rise of the ‘infinite archive’ has
fundamentally changed the nature of text.
It has turned text in to ‘data’, with profound implications for how we
read it, and deploy it as evidence. My
guestimate is that between fifty and sixty per cent of all non-serial
publications in English produced between Caxton and 1923 – between the first
English press and Mickey Mouse – has been digitised to one standard or another;
with a smaller percentage of serial materials thrown in.
This has ensured that the standard of scholarship has in
many ways improved. It is now possible
to consult a wider body of literature before setting out to analyse it. But, it has also pushed us to the point where
it is no longer feasible to read all the material you might want to consult in
a classic immersive fashion. Instead we
are moving towards what Franco Moretti has dubbed ‘distant reading’, and towards
the development of new methodologies for ‘text mining’ – or the statistical
analysis of large bodies of text/data. Stephen
Ramsay’s new book, Reading Machines, illustrates four or five examples of what
he describes as a new form of ‘Algorythmic Literary Criticism’, but is simply a
taster for a wider series of practical methodologies. That Tony Grafton, president of the American
Historical Association could recently and hyperbolically claim that textmining
was simply the ‘future’ of history, and that it was already here; reflects a
truth most digital humanists (whose ranks are dominated by librarians and archivists)
have been struggling with for the last three or four years; but which most
historians are only now becoming aware of.
The Google Ngram viewer, which allows you to rapidly chart
the changing use of words and phrases as a percentage of the total published
per year is just the most high-profile online tool in a wider technological landscape. The associated, and ill-named, ‘culturomics’
movement being built on the back of the ngram viewer is another. I love the ngram viewer, and spend my Sundays
charting the changing use of dirty words decade by decade. But it also forms the basis for a newly statistical
approach to language. And lest we
forget, recorded language is the only evidence most historians use.
I am not entirely convinced by the culturomics work, which
focuses on describing social and linguistic change through consistent mathematical
formulae, but in related studies by people such as Ben Schmidt, Tim Sherratt
and Rob Newman, one can find the beginnings of a pioneering analysis of large
scale texts that promise to remodel how we understand cultural change, and the
relative influence of events.
These graphs simply illustrate that the word ‘outside’ both
grew in commonality over the course of the nineteenth century (perhaps
understandably as people’s lives migrated inside); and that if this phenomenon
were a reflection of naturally evolving language (embedded in people’s
vocabulary in youth), its adoption according to the age of the authors whose
work has been published, would look like the first graph; but that in fact it
looks like the second. In other words,
these visualisations created by Schmidt suggest a history of the adoption of
the use of the word ‘outside’ in response to events; and in the process give us
a way of measuring the cultural impact of specific happenings – its import as
measured by individual responses to them.
Or look at Tim Sherratt’s visualisations of the use of the
terms ‘Great War’ and ‘First World War’ in 20th century Australian
newspapers. While entirely
commonsensical, the detailed results of the 1940s in particular mark out the
evidence for a month by month reaction to events; allowing both more directed
immersive reading (drilling down to the finest detail), and a secure
characterisation of large scale collections.
Or to bring this back to a British perspective, we can look
at work Bill Turkell and I have done on the Old Bailey Online – simply charting
the distribution of the 125 million words reported in 197,000 trials, to
analyse both the nature of the Old Bailey Proceedings as a publication, and
their relationship to words spoken in court – in this instance to illustrate,
among a few other things, that serious crimes like killing were more fully
reported than others in the 18th century proceedings, and that this
pattern changed in the 19th century.
This stuff works and is important; and will necessarily form
a standard component of the research of anyone who claims to understand the
past through reading.
But it points up a further issue. If each paragraph in the infinite archive,
all the trillions of words, is simply a collection of data, it immediately
becomes something that can be tied to a series of other things – to any other
bit of data. A name, a date, a selection
of words, or a phrase, or most importantly in this context, a place – defined
as a polygon on the surface of the earth.
In other words, the texts that form the basis for western history can
now be geo-referenced and tied directly to a historical/geographical
understanding of spatial distribution, which can in turn be cross analysed with
any other series of measures of text – textmining makes text available for
embedding within a geographical frame.
I can’t emphasise this enough: the creation of a digital
edition of the western print archive means that it can be collated against all
the other datasets we possess. The
technology of words, and how we engage with them, has changed; creating a new
world of analysis. With a bit of Natural
Language Processing, and XML tagging; and a shed load of careful work, a
component of text that hitherto has been restricted to human understanding
becomes subject to precise definition: “he walked for twenty minutes from St
Paul’s westward, coming first to Covent Garden, and then onwards to Trafalgar
Square’, changes from a complex narrative statement reflecting an individual’s
experience, into three individual locations, a journey’s route, and a rate of
travel; each capable of being expressed as a polygon, a line, a formulae – a
specific bit of translatable data.
What I want to say next might not sound quite right in this
context. But, I am hoping this
development of text as data, and by extension, text tied to place, will have a
more profound impact on our understanding of the past precisely because, for
the most part, it has not emerged from historical geography. It has been driven by people interested first
in text, and only then, in data such as place.
As a long-time admirer of the work of historical
geographers, and avid reader of it; I believe that the rise of a highly
sophisticated form of desktop GIS, requiring substantial training and expertise
to make work, has contributed to the evolution of a widening gap between
disciplines, and has in some ways distanced historical geographers from the
kind of audience historians have traditionally courted. The rise of the geographer as ‘expert’ has
been both impressive and excluding.
But, in the last few years a real alternative has
emerged. I understand geographers are
sniffy about it, and I know full well that it doesn’t provide the kind of
powerful analytical environment that a fully functioning GIS Editor, Analyst and Viewer package
can generate in combination with a Spatial database management system. But it is usable and it is continually
getting better. And I mean Open Street
Map, Google Earth and Google maps, and the range of open source browser side
services that build on it, like BatchGeo.
Together they make available to everyone, a good and growing
proportion of the tools previously only available to a technocratic elite. In the process and in combination with the
transition of text in to data; we are suddenly in a position to do something
different.
My favourite exploration of what can actually be done online
in an intuitive and accessible way, is Richard Rodger’s collaborative project
with the National Library of Scotland: Visualising Urban Geographies, and the associated
Addressing History sites:
The important thing about these projects is that they allow
a wider audience to use historical maps in the way a historical geographer
would, and to upload their own KLM files, and to explore the data, and relate
it to a modern map. It is historical
geography made user friendly. And that
is important.
I am also a great fan of the New York Public Library Labs,
Map Rectifyer Project, which crowd sources the kind of warping of maps that
people just could not previously do.
And more recently, the British Library’s adaptation of the
same methodology to their own map collections.
As much as anything projects like these educate a wider
public (including historians) about the methodologies and issues traditionally
faced by historical geographers, and generally hidden behind a beautifully
presented set of final maps designed to make a point, rather than to allow an
intellectual journey. These sites form
open invitations to discover all the issues associated with the underlying maps
and all the problems with the data.
Together, text as data and user friendly GIS make it newly
possible to imagine an environment in which geographical information, and
display, form a natural and unproblematic component of every other analytical
process. It makes possible a situation
in which historians cease to be mere text merchants, obsessed with the perfect
quote, and compelling (if largely un-evidenced) argument; and where geographers
have a new access to the subtle mappings of the marks of ‘culture’ in its
broadest sense – a new way of thinking about the geographical distribution of
behaviours and ideas, that bring within a geographical fold questions
traditionally preserved for others.
By extension, In other words, I want to suggest that it is
very much the moment for a bonfire of the disciplines, and that while history
and geography can now begin to speak in new terms, the same forces are also making
it possible for literature, and art history; for all the disciplines of memory
and explanation, to speak in new ways to each other. We quite suddenly share a new culture of data
– and data can be translated.
I will return to both these developments in a minute. But, by way of illustrating the kinds of
things that we are now able to do as a result of this newly open and analytical
framework – through the mash-up of text and space - I want to spend a little time
discussing a project that Bob Shoemaker, Matthew Davies and I, and a large team
of other people, recently completed, called Locating London’s Past. Please excuse me for spending the next couple
of minutes on something that sounds a little bit too much like ‘me and my
database’ to be entirely appropriate.
In itself, Locating London’s Past is not particularly
important, but it illustrates one naïve attempt to play with these new possibilities;
to take text/data and accessible online GIS, and make something that facilitates
mapping words.
This project grew from the rich soil that is failure. Five or six years ago, as a final component
of the original Old Bailey project, we struggled to incorporate a mapping
feature on to the site that could be delivered online. But
in the last few years, we realised something was changing; and inspired in
particular by the Edinburgh project, we decided to try again. The outcome - Locating London’s Past – does three things that are new. First, it makes available a fully rasterised
and warped version of both John Rocque’s 1746 map of London; and the first
‘accurate’ OS map of the capital created between 1869 and 1880 – both of which
have been fully ‘polygonised’, and related to a modern Google maps
representation of London. And second it
brings together around 40 million words of text, and a raft of established
datasets – a couple of hundred million lines of data - in a newly geo-coded
form that can be ‘mapped’ against both area and local population, at the level
of streets, parishes and wards. And
finally, it relates both these resources to the first comprehensive, parish
level population estimates for the 18th century.
In the process it
brings together text and maps in a new way, delivered in a cut down Google maps
container that even a historian can understand.
For the maps and GIS, we turned to Peter Rauxloh of the Museum of London
Archaeological Service (MOLA), who worked with
scans and an index of place names drawn from Rocque’s map created by
Patrick Mannix, to develop the kind of resource that underpins the best sort of
traditional desk bound GIS project.
The 24 sheets of
the original map were turned in to a single image, and then warped onto
the first reliable Ordinance Survey map from 1869-1880, creating a direct
geo-referenced relationship between the first accurate modern representation of
London and Rocque’s eighteenth-century version.
Rocque 1746 After Georeferencing.
The geo-referencing
operation involved identifying some 48 common points between Rocque's original
map and a modern OS map; leaving us the task of defining all the streets,
courts, parishes and wards that made up 18th century London. In the end, this amounted to some 29,000
separate defined polygons.
Parish boundaries which intersect with the
street of Cheapside, London.
Completed street network for main area
covered by Rocque's map.
Street lines expanded to polygons based on
recorded width.
All of which gave
us something rather cool – a proper, interactive and accurate map of 18th
century London, that among a lot else, let’s you go from here:
To here:
And more
importantly, lets you go here – all those parishes securely defined:
And all those
streets and cul-de-sacs:
In the process it
makes, each parish and street, ward and cul-de-sac newly available as an
analytical category – defined as a specific area, and location – defined in
terms of its distance from any other place, and the route between them, its
size and importance in a hierarchy of streets – and defined securely against
the earth’s surface.
All of which left
us with just one more task – the, to us, more familiar job of providing the text/data
to put into these analytical polygons.
And for that, we
brought in the material available from the Old Bailey online for the 18th
century crime, from London Lives, fire insurance records, voting records for
Westminster, Hearth Tax returns and plague deaths; and finally a bunch of
archaeological material from Mola.
Along with the more
structured data, we ended up with a couple of hundred million words of text,
primarily reflecting crime and events – descriptions of behaviours given under
oath to magistrates, in court, at sessions and before a coroner; which we then
processed using a combination of automated methodologies, including Natural
Language Processing, and manual checking, to identify some 4.9 million place-name
instances, each tied to its own polygon.
All of this data
was then made available for search and mapping – including both structured and
keyword searches – so both the ability to search on the crime of ‘murder’, and
the word: teapot.
Inevitably, there
are problems with the data and the map; but, it nevertheless allows us to map
things like the number of small houses in the 1690s – defined as having one or
two hearths as recorded in the Hearth Tax Returns.
Or more
contentiously, to map the distribution of suicide cases in the coroners’
inquests found by a keyword search of 5000 inquests on ‘felo’ – as in felo de
se – and ‘suicide:
Or the distribution
of the mention of a horse, mare or gelding, in the Old Bailey.
We have not even
begun to explore what the data tells us, nor was this created in the
expectation that we would be able to do so – but the important thing is that it
does allow us and everyone else, to explore this material in a new way – and do
so quickly enough to facilitate the testing of new hypotheses, and random
midnight thoughts. And to quickly test
words against spaces, text/data against spatial data.
Of course, all of
this is contentious, and I suspect will leave historical geographers rather dissatisfied. The original data is variable, the percentage
securely geo-referenced is inconsistent and I am waiting for a few proper
demographers to critique the population figures. As worrying, the data is not currently
available for the more subtle analytical approaches that have been so fruitful
in historical geography. We can’t easily
define networks, for instance. In other words, this is a rough starting
point, and all the critical skills of a true sceptic are needed when using
it. But this site does allow us to play
with all this data in a new way, and to come up with insights and
hypotheses for further investigation. To,
for instance, map all the instances of the words for the industrial colours of ‘blue, red and yellow’ against the natural
hues of ‘brown and green’ to explore an urban environment and to suggest different
ways of thinking about a wider cityscape.
It also means that we have 40 million words of
geo-referenced text that we can use as the basis for a new kind of text mining
– that incorporates space with linguistic change; and which will add to the
geographers’ toolbox all the rather wonderful methodologies of corpus
linguistics: Measures of Text Frequency and Topic Modelling to name just two.
We are, of course, nowhere near where we really want to
be. For that, we will need to have a lot
more text, and a lot more subtlety. I
want to be able to map all the places in a newspaper by subject and category of
article – to have a scrolling representation of places mentioned in a text as I
read (either immersively or distantly).
I want to be able to use corpus linguistics, semantic search and
syntactic analysis (ontologies and all the methodologies designed for
text/data) in combination with both secure place name data, and historically sensitive
boundary and population data. There is
not much point in comparing 18th century text with the modern road
network or county boundaries; or wondering why there is not a lot of text
coming from Greenland in the absence of population density figures. I want to be able to map networks defined by
individuals, defined in turn by the words they use; and networks defined by
geographical measures such as road width.
What percentage of London was made up of parkland at different stages? And what words and crimes are dominant in
those different parks (Hyde Park vs Moorfields?). And as importantly, I want to be able to test
the results against secure measures of statistical significance.
All of the components to make this happen are in place – we
all now work with data and data is interchangeable – subject to unending
automated translation - making the main technical hurdle essentially unproblematic. But there is still a long way to go. And there is also a clear and present danger
in the process. And it is that danger,
that I now want to turn to.
I spent last year co-directing one geographical project –
Locating London’s Past - and one text mining project – Datamining with Criminal
Intent. Both projects were
intellectually engaging beyond measure.
I learned more about history – and sources I already thought I knew well
- doing something else with them, than I could possibly have done in a year of
reading. But I also found myself
struggling against the run of the data I was helping to produce. I came to count myself among those who Lewis
Mumford had in mind in 1962 when he warned urban geographers that:
‘… minds
unduly fascinated by computers carefully confine themselves to asking only the
kind of question that computers can answer and are completely negligent of the
human contents or the human results.’ Lewis Mumford, “The Sky Line "Mother Jacobs Home Remedies",” The
New Yorker, December 1, 1962, p. 148
In other words, I found myself limping uncomfortably towards
a positivist abstraction in which there were few people, but much data;
beautiful graphs and compelling trends, but few of the moments of empathetic
engagement that make history so powerful and which form a little discussed
component of its authority as a genre of literature.
So, at this point, having waxed on the joys of technology
and what it allows you to do, I want to stand back for a minute and remember
the individual in the landscape. And in
this instance, just one individual – a man named Charles McGee or Mckay, who
stood just here for over forty years, from at least 1809, until his death in
1854; making a living as a one-eyed crossing sweeper – a black Jamaican refugee
from Britain’s wars of colonial expansion:
Or to put it differently, he stood just here, on a map
created just before he arrived:
Or, for a map that should have included him, here:
Or if we want to get down to street level, just here – the
obelisk he stood in front of, itself visible on the map:
MacKay became a part of the image of this cityscape almost
immediately. William Bennet was the
first to record his presence, placing him before the obelisk dedicated to John
Wilkes that stood at the top of New Bridge Street:
He was already missing one eye, but had not yet started
sporting his shock of white hair:
A year later, in 1810, he McKay was still there:
As he was when Ackerman published the same vista in 1812. Recognisable, even though his faced has been
scratched white by some later owner of this image, clearly made uncomfortable
by his presence in the landscape:
Five years later, in 1817, John Thomas Smith, the keeper of
prints and drawings at the British Museum, gave us our first detailed portrait,
and our first biography.
Smith claims Mckay, or McGee as he styles him, was already
old beyond credibility in 1817, though another account would put his age as 50
in that year. His hair ‘almost white’,
was tied back in a tail and Smith firmly locates him at his ‘stand… at the
Obelisk, at the foot of Ludgate-Hill’.
He also claims (as do most commentaries on well known street figures),
that he was secretly wealthy; and that he attended Rowland Hill’s Methodist Tabernacle
on Sundays; that he was lately seen wearing a ‘smart coat’ the gift of a city
pastry chef, and finally that his portrait, made in October of 1815, hung in
the Twelve Bells public house on Fleet Street – around the corner from the
obelisk.
And again, in 1821, in his depiction of Tom and Jerry,
‘Masquerading it among the Cadgers’:
And finally, in the same year, Cruickshank includes McKay in
his ‘Slap at Slop’, suggesting along the way that McKay was involved on the
edges of radical London:
And so to John Dempsey’s portrait from sometime in the 1820s
– which seems to me to speak of a man and a place, of a life lived in a
landscape, more powerfully than any other.
At the beginning of the next decade, he was still there,
depicted this time, from a different perspective:
And three years later, Mackay also became the model upon
which Charles Matthews based his depiction of a modern Othello in ‘the Moor of
Fleet Street’, first performed to
disastrous reviews, at the Adelphi in 1833.
In the play, Mackay is depicted as engaged in a battle of jealousy
and rage among the low characters of London; and is described as ‘the Moor who
for many a day hath swept Waithman’s crossing over the way’ from Ludgate Hill.[i]
His spotted red bandana, clearly visible in
Dempsy’s depiction, invested with gypsy lore, and gypsy power, to ‘keep woman
honest, or cure the worst cold’, and given a history steeped in London’s boxing
lore, and serving in the play, the role of Desdemona’s lost handkerchief.
The best account of his later life is from Charles Diprose’s
authoritative history of St Clement Danes, where McKay lived, off Stanhope
Street. Diprose describes McKay as ‘a short, thick-set man,
with his white-grey hair carefully brushed up into a toupee, the fashion of his
youth; … he was found in his shop, as he called his crossing, in all weathers,
and was invariably civil. At night, after he … swept mud over his crossing… he
carried round a basket of nuts and fruit to places of public
entertainment...’ And according to Diprose,
‘He died in Chapel Court, St Giles, in 1854, in his eighty-seventh year.’ A later historian, William Purdie Treloar
claims McKay was then replaced at his stand by a drunken soldier who ‘sometimes
made 8s to 10s a day’, and drank as much each evening.
All of which is simply to say, that some people stand in the
same place longer than many buildings; and have a greater right to appear on a
map, than many landmarks. As we move
towards that new data rich environment of text/data and intuitive GIS; as naïve
historians and the wider public, come to use the ideologically laden genre that
is a map as an interface for trillions of words of text; and as they step back
from their own text to view text/data from afar, I just think it is important
to remember that landscapes and cityscapes only exist between the ears of their
denizens – that we cannot map the subtleties of Ludgate Hill and New Bridge
Street without trying to know Charles Mackay.
With Lewis Mumford, we need to ensure that we are not ‘completely negligent of the human contents or the human results’ of
asking the questions only computers can answer.
[i]
Note that Waithman is a wealthy linen draper with a shop on Fleet Street. His daughter is reputed to have been
especially kind to McKay, and to have received a legacy from his on his death
of £7000. Treloar gives a more detailed
account of Waithman’s role as alderman and MP, and suggests his daughter
regularly took out soup and warm food to McKay, p.124.
