What is documentation and why should you care?

Nothing you see on this web site is old. Almost everything has been created in the past decade, some of it in the past year or two. To be sure, we used old maps, sources and images in our work, but we transformed each one into something qualitatively new. Now any digital version of a text, painting or map is fundamentally different from its analogue counterpart, but when it comes to an historical document this difference takes on a much greater significance, for its very being has changed.

This raises a big problem. Both the form and content of an historical document are essential to understanding what it has to tell us about the past. By digitizing a document we change both, so necessarily it will be telling us different things about the past. Some of those things will not be true. Others will be misleading. While yet others will be telling us things that the people who created the document would not have known about themselves or their world.

To make sense of all of this, you need to know what it is we did. Documentation records what we did. Being aware of what we did and why, will help you to assess how the meaning of the many traces from the past found in this site may have changed. This will make you more critical of our material. It will also make the material more meaningful.

What were we thinking?

MAP is a way for people to share their data. Our guiding hope has been to create a modular system that will grow by continual add-ons - new layers, new variables, and new links among them - like what you might build with Leggo. The system is conceived to serve many users, many purposes, to tap various disciplines, and to portray events geographically, socially, and temporally.

Differing visions within MAP generated creative tensions and fuel on-going discussions about where we should go from here.

MAP brought together people who had been active in three major research centres on 19th century Montréal. Their desire to build collectively on the work of the many individuals and groups active in the field reflected their diverse experiences with the past and differing visions of the future.

Understandably, the most influential vision has been that of our founding co-ordinator Sherry Olson. Inspired and taught by her graduate students, Sherry made the geography department at McGill a centre for socially-informed historical geography.

Sherry had developed an approach and a technique which proved essential to MAP. The approach privileged median rents as the most sensitive social indicator for historical urban geography. These medians were calculated from municipal tax rolls. The technique identified streetscapes, generally both sides of a street for several blocks, as the most accurate mode of spatially representing this data. A 1983 article in Cahiers de géographie, that she co-authored with David Hanna, explained the theory and method that they designed for their plates on Montréal in the Historical Atlas of Canada, (Vol II and Vol III.)

The second major influence was that of Jean-Claude Robert, a leading social historian of 19th century Montréal and the author of the standard reference work on historical maps of Montréal. In the 1970s, working with the Groupe de recherche sur la société montréalaise au dix-neuvième siècle at the Université du Québec à Montréal, Jean-Claude co-authored highly influential articles on occupational structure and the importance of urban land. Under his co-direction, the GRSM conducted pioneering research on census returns to challenge the prevailing traditionalist interpretation of 19th-century Québec society.

This long experience with routinely generated nominal series, such as tax rolls and census returns, meant that when MAP started, we could build on a considerable body of completed research on industrializing Montréal. It also explains one of the central methodological choices influencing the architecture of our system. Although we did extensive work with period maps, the key source of reference for the 1846 and 1880 layers were the 1848 and 1880 tax rolls, because they alone provide both land registry numbers and street addresses. This meant that whenever there was a disagreement between two sources, the tax roll was considered to be the governing source. The result is an historically coherent geographic information system.

Privileging nominal series was at odds with a third influence on MAP's development. Initially working with the Montreal Business History Group, and subsequently at Memorial, Robert Sweeny had developed a quite different theory and method for understanding the past. He proposed a 'cubist' portrait of pre-industrial Montréal, wherein each source is treated as distinct because endowed with its own historical logic. From his perspective, one should not privilege one source over another. Contradictions between sources are not problems to be resolved, but rather evidence of the complexity of the past. In a article on initial lessons learned from MAP, he explored how disciplinary conventions hamper our seeing the differing nature of our sources.

Differing visions within MAP generated creative tensions and fuel on-going discussions about where we should go from here. We have, however, resolved the question of according priority to a particular source. In 1825, no standardized street numbering system for Montréal existed, and so there was no easy way to link nominal series to the map. Not being able to do the linkage, Sweeny might be said to have made a virtue out of necessity. When, with the collaboration of Jean-Claude Robert, he used the Sulpician's terrier to create a GIS for the town centre for 1825 to 1840, he too privileged his textual source over the map. He argues this confirms his earlier assessment that geographic information systems, as one of the quintessential technologies of contemporary capitalism, tend to hide important features of the city as people experienced it.

The method to our madness

Building the base map

Our entire system is erected upon a digital map that we constructed from files lent to us by the Service de géomatique de la Ville de Montréal, They provided us with a set of MapInfo files of the city for the year 2000, known as the SIURS geobase, and an extensive set of CAD files. Working in consultation with Michel Reney (Service de géomatique), Rosa Orlandini converted these into ArcView files. As the original geobase did not include adequate metadata, or documentation, we prepared a handbook, as well as metadata files, FGDC standard, for each layer. The city geobase included present-day streets, blocks, lot lines and building footprints with addresses, providing a framework anchored in modern surveying, metric measurement, air photos, and an architecture of geodesy or geometry of the earth.

To use maps from earlier dates, we stretch them like a quilt in a frame, pegging them to the "official" or standard framework. This makes it possible to compare maps made under quite different conditions and for various purposes. The base map we created linked together hundreds of different CAD files into a single multilayered system, which we used as the basis for a new map for 1880. There was no single Goad map; it was originally published as an atlas of 44 two-page plates. The closest thing to a view of the whole city in Goad was the index plate reproduced here. To see how really new our map is, compare this image to the detail in the banner at the top of this page. The Goad plates were published at a scale of one inch to 200 feet (or 1:2400). Our task was to recreate this series of images as a connected system consistent with the projection used by the city (nad83 zone 8). Orlandini and Kevin Henry tested alternative procedures: ArcInfo8 steps; ArcView with script ImageWarp; and ArcView 3.2 with extensions Image Analyst and IMAGINE Image Extension. We took the third option. Each plate was scanned by the McGill Library one page at a time and saved as tiff files. Smaller, partial plates meant that Sherry Olson with a team of graduate students had 89 images to rectify.

Rectifying Goad

Rectifying a map means changing it so that the map shares the same co-ordinates as another map. This involves identifying a point on each map that you believe to be the same location and treating it as a control point or anchor. After you have identified a number of anchors, the software warps the overlay map to fit the co-ordinates of your base map. Initially we thought we could use evidence from the built environment as our anchors. So, for example, we would anchor the corners of Notre Dame church.

This assumption explains our decision to build our 19th century layers by starting with the Goad. Once we had the Goad rectified, we thought we could use it as the basis for the rectification of the Cane 1846 map and then use the Cane as the base map to rectify the Adams 1825 map. Working backwards, we would maximise the number of buildings appearing on both maps and so ensure the most reliable rectification by increasing the number of potential control points.

To our considerable surprise, rectifying to property lines proved to be considerably more accurate than using buildings as our anchors.

Our initial rectifications highlighted the challenge of variability from one plate to another and showed that we would need many anchors. The margin of error was running from five to ten meters on most plates; where the comparable error in a modern document created to current engineering standards from arial photographs of contemporary buildings is in the order of one meter. Our high level of inaccuracy was disturbing, as minimally we wanted the centre of any lot to actually be within that lot, because this was where data points linked to the lot would be placed. We thought an acceptable margin of error would be approximately 3 meters, or less than half the width of almost all lots in the city.

In many parts of the city there were no buildings from 1880 that had survived to the present and so, faute de mieux, we used existing property lines as our anchor points. To our considerable surprise, rectifying to property lines proved to be considerably more accurate than using buildings as our anchors. Although property lines are invisible abstractions, corner lots, as well as intersecting property lines in the interior of a city block, proved to be relatively stable features over 120 years.

There were two types of lots visible on the 1880 Goad: building lots and cadastral lots. In Goad, cadastral lots appeared even where no building had yet been erected, and they usually appeared in the SIURS geobase even where buildings have been demolished. Unfortunately, the extensive detailed information written on the Goad plates meant that it was not possible to use them as the basis for an automated drawing of the cadastral lines. We decided to create a GIS layer of the cadastral lots using as our base the cadastral maps of the city, parish and island of Montréal drawn by Louis-Wilfrid Sicotte between 1876 and 1878. A detail of Ste-Anne ward is shown here. We then rectified the Goad to Sicotte. Using the cadastral lots made it easier to compare maps from different dates, and to move between them with greater confidence, despite changes in house numbering and street names. There is a slideshow available from the side menu illustrating how we created this cadastral layer, as well as a technical paper explaining how we automated the drawing of the lines.

Jean-François Palomino created a five block test area for 1880. We deliberately chose an area in Saint-Jacques where all the buildings had been demolished - for construction of the Radio Canada skyscraper and its adjacent parking lots - to demonstrate the ability of the system to recover a lost world. This test area was linked to two tables we had created from the tax roll, one for the owners of properties and one for the occupants. It was at this stage that a key decision was made. Instead of using Goad's street addresses we decided to make the tax roll occupant table the fundamental key for address searching. Palomino's text describes the process, while his slideshow, available from the side menu, illustrates it.

Initially, the set of 1880 GIS layers we drew included: polygons for blocks, cadastral lots, wharfs and the canal; points for each building by type of construction, and lines for railways. streets and cartways. We then created polygons for each street and laneway from the block layer and a new lot layer of existing building lots, and finally, over several years, Robert Sweeny drew the polygon layers of the built environment.

Problems with what we did.

Our layers were all drawn by hand from raster images (jpgs) of the 89 rectified images. We would not recommend this method. Frequently a single line on the original image represented the side of a building, the limit of the lot and sometimes the edge of a block. By drawing the layers for buildings, lots, and blocks separately we necessarily created many inaccuracies. Frequently, the outline of a building spilt over the outline of the lot, or the outline of the lot spilt over the outline of the block. Correcting for these slivers was a time-consuming and fastidious task.

Put simply, the danger is that people using the system will believe what they see...

A better method, we used for our turn of the century layers, involves retracing lines, or more precisely "arcs", for every line on the map and then using software to combine the arcs into polygons. Considerable editing is still needed to assign all the resultant shapes into buildings, lots and blocks, but this method is faster, generates fewer inconsistencies, and creates fewer ambiguities. Buildings and lots share the exact same line, just as they do on the original map.

Problems arising from the nature of the orginal atlas were compounded by how we scanned the plates.The Goad plates were never designed to appear as a single huge map, and so edge matching problems were inherent in the printed version. Our scans were done using a single camera focused on the centre of the page, the resultant distortion introduced additional edge matching problems.

There remains a major problem, which we have not resolved. GIS systems create a two dimensional image of the world that appears to be remarkably accurate. The default settings for the ESRI software we used created an area for every lot and building in square meters down to the third decimal point! This illusion of accuracy is seriously misleading. Historical maps were never this accurate to start with.

Goad's atlas was not a legal document, such as an owner or buyer or insurance company might have a surveyor do for his or her lot. It was an everyday tool for firemen and fire insurance agents to recognize risk. We do not know how the atlas was compiled. Goad's surveyor and draftsmen sometimes presented projects not yet built, or streets the city anticipated opening or widening, as if they already existed. From what sources did they work? We have too little information to interpret many of the lines in this atlas.

Furthermore, as we have seen, our rectifications of these old maps introduced new distortions, with wide margins of errors. None of this is readily apparent to users of our GIS. Put simply, the danger is that people using the system will believe what they see, or what the system says, to be true. Considerable critical awareness on the part of users is required to avoid this trap.

Rectifying Cane and Adams

If property lines rather than existing buildings were they way to go, then we no longer needed to work backwards from the Goad to the Cane and then from the Cane to the Adams. As considerable work had already been done on Adams, we started with it.

In the 1990s, Robert Sweeny had created a non-rectified version of Adams using Micrografix Designer. This early work is on view in our gallery. Initially, we attempted to work with this facsimile. It proved impossible to rectify, particularly in the less developed areas where Adams had had far fewer reference points to anchor his original survey.

The MAP edition of the John Adams map was created in three steps. The original map was scanned at high resolution and in eight parts by the McGill Library. Jason Gilliland and Kevin Henry then rectified the tif files to match our Goad plates. Here we see the area around Place d'Armes after rectification. The rarity of control points in the south-west, near St Gabriel's Farm and the canal, meant that positioning in this area is at best a rough approximation. Finally, Sweeny drew shape files for each of the various layers.

Unlike Goad's atlas, the originals of both Adams and Cane were black and white engravings. In a GIS it is possible to use colour to convey information that was not readily apparent or not easily distinguishable in the original. This has an immediate impact on how the viewer sees the map. This is not just a question of one being flashier than the other, although a desire to enhance the 'gee whiz factor' probably did play a role in our choice of colours. The user can vary the colours to suit a particular purpose or aesthetic preference.

By choosing to represent vacant lots and farms in 1825 and gardens, farms, treed areas and Mount Royal in 1846 in various shades of green, we effectively convey an image of the city that is quite different from that of the originals. This is most in evidence when viewing the maps as a whole, where green predominates. A pastoral, romantic view of nature is a feature of many period drawings, paintings and etchings and a Euro-centric perception of North America was undoubtedly at play there. Our choice, however, was based on the desire to convey the importance of local gardens and orchards in 1825. The apparent "green belt" around the town in 1846 was qualitatively different. The gardens and copses drawn by Cane were private pockets of affluence in a segregated city, where topography offered visibility, enhanced views, fresh air. drainage and was believed healthier.

Rectifying the Cane map posed almost insurmountable difficulties. Unlike the other two maps, this was a commercial venture. It was a promotional vehicle, designed to be seen rather than to be a guide for military or corporate decision-making. Cane himself did not enjoy the legal authority of Adams or his training in military surveying. Nor did he enjoy the resources of Goad & Co. Correspondence with the Mayor indicates that he was not being paid to take into account the effect of topography (elevation) or to create accurate lot boundaries. Not surprisingly therefore, Cane's was the most inaccurate of the three original maps.

Rectifying large parts of the city at a time, as we had done with the Goad images and the eight sections of Adams, simply was not possible with Cane. In order to keep within a three meter margin of error, Sweeny divided the Cane map into many small parts. Most were only a few city blocks, sometimes not even that. Working primarily from the Goad, he cross-checked images with the already rectified Adams. He then tiled together the numerous rectified images to provide a mosaic as background for drawing the shapes files.

A key feature of our edition of Cane are the streetscapes. We cannot yet link household data from the 1840s to a specific lot, but we can with a fair degree of certainty place households within a particular street segment. Designed to represent data from the 1842 census returns and the 1848 tax rolls, each streetscape consists of the households on both sides of a street for several blocks. MAP inherited David Hanna and Sherry Olson's streetscapes for 1848 (see above). They had used the rental values from the tax roll to establish coherent units of approximately 50 households each. These were in a point file, with textual support defining the limits of each streetscape. Sweeny used these dimensions to transform the points into polygons based on the lots facing on to that particular stretch of street.

This visual representation of the built environment does not cover the whole city. The streetscapes represent only where people are known to have lived, in 1842 110, by 1848 120. In our application three major indices are shown using this technique: the number of servants per household in 1842, the number of households headed by labourers in 1848, seen here, and the median rents in 1848. In our beta release of this application we also had a streetscape of the franchise in 1842, but subsequent analysis showed that the census returns on this question were quite misleading. In our gallery of images for the 1840s, there is a graphic illustration and explanation of the problem.

Slideshows

How we built our test area for 1880 (en français).

Creating the cadastral layer (in English).

Technical papers

Rosa Orlandini's report on the cadastral layer (in English).

Extracts of Caroline Sauriol's report on the 1881 Census (en français).