Montréal: The Pleasure of Maps Made by Hand, or by Eye

Google is getting us all used to the idea of automatically generated maps, which sacrifice many opportunities for clarity and beauty in order to be instantly available and automatically up-to-date. But Anton Dubrau, who writes the intelligent Montréal transit blog Catbus, asks:

I guess there is this general question whether frequent network maps should be automatically generated, or made by hand. Which is probably related to the question whether they should be abstract and compact, or geographically accurate. Or whether they should be published today, or … later. It took me more than a solid week to make a map of Montréal’s network by hand.

He certainly did, and it’s a beauty.  But you’ll have to go to his post to see the thing full-size, and to see how it compares to the transit agency’s map and a street-map overlay.  It’s a great piece of work.


There’s no question that the most beautiful network maps will continue to be those made by hand, with great care and thought, by people who know the city.  I’d like to say “most useful” as well, but that will be true only as long as they can be kept up to date.

Of course, it wasn’t that long ago that “by hand” meant with paper and ink, as opposed to on a computer.  Perhaps “by eye” is a better term for this kind of work, which rests on seeing not just the literal geography of the city, which is easy to automate, but also its structure and character, which still appear only to the talented local eye.

34 Responses to Montréal: The Pleasure of Maps Made by Hand, or by Eye

  1. Zoltán September 6, 2010 at 5:31 am #

    I like the way in which the routes that one can rely on all day are significantly more prominent than those which only offer frequent service at peak times.
    I also notice that this map has resisted the temptation that tends to afflict map-makers to show commuter ral services prominently even when they only offer a few journeys in the peak direction at peak times. That’s almost certainly a good thing, as more than once such maps have allowed me to excitedly plan out a trip involving the fast-looking train line that I can see, only to look up the schedules and find that train service doesn’t exist when I need it.
    That said, in the case of Montreal, I might consider making the Deux-Montagnes commuter rail line slightly more prominent than the others (making the dashed grey line solid and black, perhaps) and noting in the key that it offers service at least every hour on weekdays, as if you’re in the very north of the city it’s often worth waiting for the 5-10 minute trip through its long tunnel straight downtown. Then again, it costs more, doesn’t accept metro+bus passes, and provides the most minimal weekend service, so I’m not quite sure about whether that would be a good thing to do.

  2. samussas September 6, 2010 at 6:19 am #

    Great map. One of the best Frequent Network we have seen here for the time being. It’s concise and highly readable.
    Three remarks:
    1. Won’t it be better to give their names to the Metro lines? Line numbers are (at least for me) easier to remember than colors.
    2. I find the use of shades of gray for the directional network a bit confusing. There is places where there is too much gray to easily follow the lines.
    3. Still on the directional network. You might want to think about adding orientation arrows in the middle of the lines. It’s not so much practical to have to refer to the end of the lines to know in which one each line flows.

  3. Danny September 6, 2010 at 10:31 am #

    Excellent map, great work.

  4. GD September 6, 2010 at 10:34 am #

    absolutely brilliant map. Though the physical shape of Montreal and its grid help, this is an astounding piece of work.
    I really wish I’d had that map when I visited this April, as the frequent buses run on extremely convenient routes, only I didn’t take them, because I had now way of telling which bus routes would really be useful to me.
    This map is the best illustration for Jarrett’s larger points about frequent network maps.

  5. Zoltán September 6, 2010 at 12:19 pm #

    “the frequent buses run on extremely convenient routes, only I didn’t take them, because I had now way of telling which bus routes would really be useful to me”
    I had exactly the same experience in Montreal myself. This is what keeps tourists on the subways, from which they see nothing of the city they’ve visited!

  6. rhywun September 6, 2010 at 5:49 pm #

    Fantastic map, and a real pleasure to look at. I would love to see something like this for my city (NYC/Brooklyn).

  7. Alon Levy September 6, 2010 at 6:51 pm #

    Rhywun, I started making one for New York, starting from the easier to handle crosstown buses. I still don’t know what to do with interlined segments like Fifth/Madison.

  8. rhywun September 6, 2010 at 7:29 pm #

    New York is hard because so many routes run at 10-minute or even more frequent intervals – and the network of such routes does not seem to form any sort of comprehensive “plan” in the way that ant6n’s map of Montreal makes so clear. For example, a dense neighborhood like Bedford-Stuyvesant has lines every few blocks – compounded by the fact that all the streets are one-way – and most if not all of those routes are “frequent”. Maybe “Select Bus” will some day make more sense out the situation by concentrating more service in major corridors….
    One simple change I would love to see on the NYC maps (there’s a separate map for each of the five boroughs) would be a special indicator for what SF calls “community” routes on their map. These are routes which seem to run (in SF) at 30-minute intervals, and serve more as neighborhood “circulators” than long-distance routes. In effect, these are the routes that are NOT frequent, and in a dense city like SF or NY you can (nearly) count on all other routes to be at least frequent enough to disregard the schedule (during the daytime). I think this would be a quick fix for the problem that the official maps make service look “better” than it really is, without (yet) having to develop a “frequent” network.

  9. Alon Levy September 6, 2010 at 7:49 pm #

    Rhywun: I actually couldn’t do 10 minutes – it would leave too much out, even in Manhattan. NYCT calibrates service to demand instead of to fixed intervals, so 10-minute service ends surprisingly early on a lot of routes. I decided to do two classes of frequent service – 15 and 7.5, identified by line width.
    Doing it more cleanly, with just one 10-minute cutoff, might also work, but that would run into the interlining problem on 5th/Madison.

  10. Brent September 6, 2010 at 8:09 pm #

    That’s a nice map and done very professionally (angles, line spacing, etc.). It does raise a question, though: is it better to show these types of maps geographically or schematically?
    In general, most systems tend to use schematic diagrams for rapid transit service, but geographic maps for surface routes. I can understand the desire to map a frequent network schematically (similar to rapid transit routes) for “branding” purposes (oh, how I hate that word!) — but, in my view, surface route maps need to be geographic.
    A subway line, being underground, is much more abstract to the user, whereas a bus or streetcar line (especially one with standard stop spacing) obviously has a close relationship with the surface street network and needs to be illustrated more closely to scale.
    One exception might be a very complex downtown network (e.g., London; Washington), where a schematic diagram might help make some sense out of the network.

  11. rhywun September 6, 2010 at 8:16 pm #

    Yeah, that’s another problem in New York – the frequency varies wildly on every route. I had a similar goal to create a high-frequency map and while I never made it beyond the planning phase, IIRC I was using the “Noon” frequency indicated for each route on the reverse of the map as a sort of “average” frequency. It’s an interesting topic to me – calibrated service vs. rigorous headways. I lived in a (German) city once where all the routes had predictable headways and found it very useful, but I suppose it could lead to waste.

  12. Westnorth September 6, 2010 at 8:47 pm #

    I was just in Montreal and saw ads for this service on many bus stops along the frequent-service network, as well as at metro stations. It is indeed a pretty high bar to set, but one that matters in a city with extreme weather.
    I just love that this map shows the green line’s S-curve (which facilitates cross platform transfers) at Lionel-Groulx, too!

  13. Anon256 September 6, 2010 at 9:36 pm #

    As a starting point for New York, an hour or two with the MTA GTFS files, Perl, and yielded this map of stops served by an NYCT bus every 10 minutes or better from 6:30 to 20:00 (i.e. comparable to daytime subway routes).

  14. rhywun September 6, 2010 at 10:02 pm #

    Hey, that’s pretty neat, and as I suspected, the coverage seems inexplicably spotty in some areas, like my Bay Ridge (the SW corner of the map within that “ring” of expressways) or large parts of downtown Brooklyn and Park Slope – perhaps that’s because they feel the subway service is sufficient enough that “high-frequency” buses aren’t needed.

  15. Alon Levy September 6, 2010 at 10:32 pm #

    Anon256, a few things on your map look off to me. For example, the M15 runs every 5 minutes or better in both directions on the entire route, not just in the southbound direction. Even its northern end has 10-minute northbound service on each of the local and limited until 9.

  16. ant6n September 7, 2010 at 12:45 am #

    Thanks to all for the positive feedback. As Jarret points out, I guess the real value is in keeping this map up to date. And keep fixing errors as they are being noticed.
    Can you show some some draft for your map?
    The Deux-Montagnes line has indeed a lot of potential, being electrified and all, but right now it’s not a frequent service at all… When the amt will finally put an overpass at the junction de l’est, they may be able to upgrade the service towards a surface metro. The real keen observer might notice that I marked the tunnel as going through McGill and Edouard Montpetit. There were some plans here and there at times to create two underground intermodal stations there, which would be really neat.
    1: The metro lines in Montreal are only referred to by their color. Using their actual numbers would not be very useful to users. I spent some thought on how to refer to the lines, but just used the numbers in the end.
    2: I think the visibility of the gray lines depends on the contrast of your monitor/paper. I actually spent some time to figure out that I need only 4 shades of gray so that any lines that meet will have different colors.
    3: I am still pondering the directional arrows. Showing enough directional arrows, without cluttering the map, while making it fit the overall design is really not simple. I really liked the idea of only having arrows at the terminals, but it might not be enough.
    I think the schematic map is more useful here, because with its compactness it shows the network – on one sheet of paper. A geographically accurate map that would actually be useful as such would be waaay bigger, which in turn would make it hard to grasp the overall network quickly – and carry it around in your pocket.
    The Lionel-Groulx S-curve still confuses me after all these years – continuing on the other line in the same direction you have to transfer to a metro that goes the other way.

  17. Alon Levy September 7, 2010 at 1:01 am #

    Okay, I’ve made a 10-minute map for Manhattan. The map is based only on midday weekday frequency. I copied the colors from the official map to allow distinguishing routes; the colors could be changed for extra clarity.
    If a route is frequent because of interlining, then infrequent branches are ignored, even if they’re only near-misses. This turns out to avoid the interlining hell I was afraid of.

  18. GD September 7, 2010 at 1:21 am #

    “It does raise a question, though: is it better to show these types of maps geographically or schematically?”
    no question for me in the case of Montreal – this city has a pretty easy-to-grasp geography, especially in the parts interesting for the short-term visitor (i.e. downtown, the Guy-Concordia hood, and the entire Plateau) so the benefits of a schematic map have little downsides if any at all.
    Also, Montrealais maps are weird to begin with because they tilt the cardinal directions. In any case, if you have a geographical map of the vicinity at every transit stop, then an uncluttered schematic map of the system will be sufficient. IMHO.

  19. Anon256 September 7, 2010 at 2:24 am #

    @Alon: Regarding the M15, headways actually slip above 10 minutes on a few trips. The schedule shows a northbound M15 at 96th and 1st at 06:55, then none until 07:07, then none until 07:18 (even though these same trips run 10 minutes or less apart further south; presumably this arises from some estimate of traffic conditions). The routes that show up on your map and not mine all suffer from similar slight gaps (mostly in the early morning; a map based only on 8am-8pm looks a bit more like yours).
    The nature of an automatically-generated map is to be very sensitive to these sorts of minor details and deviations from strict criteria, in addition to the aesthetic and usability issues Jarett discusses above. There is also perhaps a question here of whether “every N minutes” should really mean NEVER waiting a second more than N minutes at any stop on the route, and if not where the line should be drawn.

  20. Alon Levy September 7, 2010 at 2:47 am #

    My rule was to look only at midday headways. But in a situation like that of the M15 at 7, I’d say that it just comes from a traffic estimate, so it’s not really a 12-minute gap. One advantage of midday headways is that they don’t fluctuate based on traffic estimates.
    I think the best way to turn this into a mechanical rule is to replace “10-minute headways” with a rule like “at least 3 buses every 32 minutes” or even “at least 6 buses every 62 minutes.” In both cases, the extra 2 minutes are there to provide flexibility. Unlike in LA, the only routes in NY that have rapid service are frequent anyway, or at worst near-misses (e.g. Bx15/Bx55).

  21. samussas September 7, 2010 at 2:48 am #

    1: Interesting fact. I actually checked maps of the STM and you are right. Numbering/Naming is not used. Which I find quite odd. Not the best way (in my mind) to advertize a service and I find it also a bit unconsiderate for (say) color-blind people.
    Anyway, what I find more interesting is how this difference of approach in nomenclature highlight the different ways brains function. I have myself a very photographic memory and after many years of conditioning, I need to tag a name to a service and, when a map doesn’t show that kind of references, I can’t apprehend it without thinking at least twice about what I’m looking at.
    Anyhow. I know someone that even after living in Paris for 15 years was not able to refer to the Métro lines to their correct local naming/branding but was always calling them by their colors.
    I think, the more information you have on a map (like clear naming/coloring) the more you catter for users. The don’t all work the same and might be conditioned by their own previous experience.
    2: Maybe. I also know that not eveyrbody perceived colors and shades the same way and your delicate balance of grays might be lost on some eyes.
    3: I was actually thinking that you could drown the arrows in the fabric of the line themselves. Don’t put them arround but hash the continuity of the line with them. I don’t know if I’m being clear?

  22. Jarrett September 7, 2010 at 3:12 am #

    @Alon. Most of Lower Manhattan looks pretty bleak on your map!

  23. ant6n September 7, 2010 at 9:32 am #

    If one would align the grid with the sheet of paper, as opposed to rotate by 30 degrees, then the map would look surprisingly similar to Torontos.
    And nobody wants that.

  24. ant6n September 7, 2010 at 10:10 am #

    interesting idea. I should try that out. I’ve tried using tiny arrows instead of stop dots, but it’s too small. I also tried putting larger bubbles (like in the terminals), but i find it clutters too much. I also tried just putting arrows on the intersections with the metro (like in the terminals), but if lines meet going in different directions it makes a mess.
    In the end of the day the lines go towards the center; that should be fairly obvious.

  25. Anon256 September 7, 2010 at 2:59 pm #

    @Alon: A rule like “at least 3 buses every 32 minutes” runs into the problems discussed in Jarett’s post regarding averaging. As a transit rider, I don’t care how many buses are coming in the next 32 minutes, I care how long I’m going to have to wait for the next bus, and on the M15 that will sometimes be more than 10 minutes at some stops. That said, here is a similar map using the “3 per 32” rule (6:30 to 20:00 span).

  26. Alon Levy September 7, 2010 at 3:56 pm #

    If there are 3 buses coming in the next 32 minutes, then it’s almost certain a bus will come in the next 10-11 minutes. There are a few exceptions in the schedule, but they all come from padding to deal with traffic. They do not actually come from lower frequency, or from bunching.

  27. Theo September 7, 2010 at 4:06 pm #

    AND the tilt makes Montreal’s map look like Paris’s old map:

  28. Jeff Wegerson September 7, 2010 at 7:07 pm #

    Once a “hand” map is produced then updating it should not be a particularly time consuming process. Since even the best auto-generated map is going to require some “hand” tuning I’m not sure there is much of a controversy over “hand” vs “automatic” when it comes to a “finished” product.

  29. ant6n September 8, 2010 at 1:18 pm #

    Keeping all but a few lines tilted means that all labels of metro stations can be horizontal, which is also a deliberate design choice.

  30. GD September 9, 2010 at 1:59 am #

    @ant6n and Theo
    just for the record, I do like the way this map is tilted. I just wanted to comment on the Montrealais phenomenon of tilting it differently, and the issues that come along with that, especially if you try to align a google maps printout or the view on a smartphone with a municipal streetmap.

  31. Ben Smith September 9, 2010 at 8:03 pm #

    Slightly off topic, but does anyone else notice that on Google Maps it shows the Blue line not connecting to the Orange line at Snowdon?

  32. Nathanael September 13, 2010 at 9:05 pm #

    OK, I have strong opinions on this.
    Automatically generated, totally geographically accurate, maps should always be *STEP ONE* of map design. Even in the days when there was no automatic generation, the equivalent, careful purely mechanical and formulaic mapping, must come *first*.
    This provides the information for the *map designer*, who is capable of dealing with the clutter, capable of finding the details where the map obscures them due to too many features, and needs the full information for design purposes.
    The map for the casual user — or the map for any other specific purpose, as you may want several maps — is then generated by starting with a copy of this automated map, clarifying the areas where it may seem misleading, and removing the “clutter” so that the important parts pop out visually immediately, according to map design principles.
    This is the fundamental error made with redesigns of the London Underground map: they always started with the *previous* map. If there have been more than tiny changes to your network, you must start with the fully accurate, painfully detailed geographic map and “reduce” it to a diagram anew each time.

  33. Nathanael September 13, 2010 at 9:07 pm #

    “If there are 3 buses coming in the next 32 minutes, then it’s almost certain a bus will come in the next 10-11 minutes.”
    Maybe in Montreal. In Boston traffic may well mean that you won’t see a bus for an hour. 😛
    Buses in mixed traffic in congested cities have severe problems with schedule-keeping; unfortunately it is important for the map to reflect reality rather than aspiration. Tricky problem there….

  34. Montreal Island November 21, 2010 at 6:20 am #

    I will be needing such a map pretty soon:D I’m planning to come visit Montreal, Montreal Island more exactly this winter. I can’t wait to see all the tourists’ attractions I’ve heard so much about.