The Science Of A Great Subway Map

Researchers at an MIT lab have devised a way to determine how well straphangers can comprehend a subway map in a single glance. Massimo Vignelli really did know what he was doing.

The most famous New York City subway map is the one New Yorkers rejected. That would be Massimo Vignelli's system map from 1972. Vignelli's modernist design stressed visual clarity over geographical precision; all the lines ran vertical or horizontal, for instance, and rectangular Central Park was rendered as a square. Public pressure led subway officials to replace the map in 1979 with one much less distorted in style but also far more cluttered to the eye.

Transit maps have a considerable impact on the everyday lives of people in cities. Traveling in and around busy urban environments can be tough for anyone--from the tourist visiting for the first time to the native heading into an unfamiliar part of town. So establishing whether a map like Vignelli's has merit despite its detractors is a matter of real consequence: the quicker people process information on a subway or bus map, the easier their lives will be.

Recently, some vision scientists at MIT developed a remarkably direct way to perform just this type of map evaluation. The research team, led by Ruth Rosenholtz of the Computer Science and Artificial Intelligence Laboratory, devised a computer model capable of determining how well people will comprehend a subway map (or any other complex diagram) in a single glance. The model spits out alternate visualizations called "mongrels"--twisted images that represent how our brains actually process the maps in front of our eyes.

The MIT mongrels draw on new scientific insights into peripheral vision. Research by Rosenholtz and others has suggested that peripheral vision operates by pooling together information outside a person's direct line of sight. These peripheral pools sacrifice detail for overall impression to reduce the amount of data we process; they're a little like a low-resolution JPEG in that sense. So the mongrels effectively show what visual elements--color, text, space, line orientation, among them--have been condensed into pools during the map's journey from eye to brain.

"What these mongrels try to capture is this qualitative information about what you lose in the periphery," Lavanya Sharan, a postdoc who collaborated on the work, tells Co.Design. "Looking at these mongrels is a way of confirming the designer's intuition."

Being located in Boston, the researchers took a particular interest in the Massachusetts Bay Transportation Authority’s recent contest to redesign the subway map. Earlier this month, the M.B.T.A. announced the winning entry, designed by Michael Kvrivishvili, which will begin to appear in stations next year. Rosenholtz, Sharan, and graduate student Shaiyan Keshvari created mongrels of both the current map and the contest winner to see whether or not the city was getting a visual upgrade.

Here's the current M.B.T.A. subway map:

Image: Courtesy MBTA/Massachusetts Department of Transportation

And here's the mongrel, computed for a person who's looking directly at the Kendall/MIT stop on the red line:

Image: Courtesy of the Computer Science and Artificial Intelligence Laboratory at MIT

The mongrel betrays several problems with the current map design. The silver line has disappeared into the Atlantic Ocean; it was far too light. Southern branches of the green and red lines, while perhaps geographically accurate, have become difficult to track. Many of the angled station names are a terrible blur, and the busy area between the orange and red lines is a total smear.

Now here's the contest winner:

Image: Courtesy of Michael Kvrivishvili

And here's its mongrel, also based on someone looking right at Kendall/MIT:

Credit: Courtesy of the Computer Science and Artificial Intelligence Laboratory at MIT

A few things stand out right away. The subway lines take sharper turns that are easier to follow, especially the four now-parallel green line branches. Major transfers are also a bit more crisp as a result. The station names, now nearly all horizontal, can be distinguished (if not read). The map isn't perfect--the silver line remains hard to spot at first--but from a perspective of peripheral vision the map does seem like an improvement.

"You can see that it preserves a lot of the information from the original map," says Sharan, "but it lays it out in a way that's slightly more soothing for the eye."

Of course, unless people are running for a train, they generally don't have to absorb everything about a subway map in a single glimpse. But the basic lesson still applies: a map need not stay geographically faithful to be visually useful. In certain other real-world applications, this understanding might be critical; designing in-car navigation maps that take peripheral vision into account could potentially save lives.

As for Vignelli, public opinion does finally seem to be swinging his way. In late 2011, the Metropolitan Transportation Authority of New York introduced a weekend service map designed by Vignelli in his 1972 style. The MIT lab has shown the wisdom of this choice by creating mongrels that compare Vignelli's weekend map to the regular weekday map:

Image: Courtesy of the Computer Science and Artificial Intelligence Laboratory at MIT

The daily map, geographically inclined just as the public wanted back in the 1970s, is a mess. The diagrammatic Vignelli weekend map, meanwhile, hardly looks like a mongrel at all--a sign of the designer's preternatural understanding of visual processing. Intuition confirmed.

[Image: Courtesy MBTA/Massachusetts Department of Transportation]

Add New Comment


  • dc guy

    I stopped paying attention to this article when I read the claim that on the Vignelli map "all the lines ran vertical or horizontal, for instance," This is true if you only look at Manhattan, such as the map subset featured in the New Yorker article linked to above. I always thought the prevalence of the diagonal lines in Brooklyn were what was most objectionable about his map.

    I also agree with the sentiment that does not seem very scientific to have computers tell us what humans experience.

  • CrankyFranky

    I liked what I read about the original London Underground map - a breakthrough was ignoring distance (close central stops could be spread apart for readability - distant suburban stops could be compressed to fit - as the key information riders needed was the sequence of stops and which line had their desired start and end points - thus simplified the information could be fit into an easy-to-read graphic which has become world-famous as a design classic

    I'll guess that these days young people are more likely to punch in icon on their small screen - and wait for the GPS to tell them which direction to walk - as in recent attempted meeting - 'Which McDonalds are you at ?' - 'I dunno - where my GPS told me ...'

  • malachite2

    I use the MTA map that's a "mess" and don't find it to be difficult at all. I spend time in the city several times a year and find the "messy" map to be invaluable in using the subway system & getting where I want to go. I also use a bus map.

    I am really tired of people using computers --which is just using a model w/the programmer's biases built in--as the final word on what works best for humans, Who often have more variables & variations built in then the software does. But then, I find Microsoft's OSs since DOS to be annoying & often poorly designed, while supposedly using it is "intuitiive." Ditto for some free (i.e., not Microsoft designed) mail client & brower "upgrades."

    Sometimes all people don't want or find useful the same thing.

    Forced standardization is a real problem of the 21st century if not the 20th, because of course it's cheapest for multi-nationals to manufacture and sell one size--and to increase their profits, all of the rest of us have to believe that one size fits all (while the obscenely overcompensated CEOs & upper management of the corp can afford custom made).

    Vignelli's model looks less easy to understand to me, given the way the subway lines in NY are laid out and the usefulness of knowing where you can change from one to the other underground. Be interesting to see how the crosstown shuttle is handled in that map or stops at which there are transfers to multiple other lines. The current map shows those aspects of the system very well/clearly.

    Not mentioned at all is one of the NYC subway improvements made at some point in the 1990's--which was to improve the quantity & quality of the SIGNS in the subway system itself. That, in & of itself, has made finding one's way around underground easier.

  • Arun Bhatia

    Looks so similar to what any mind maps software throws up when one puts in much did you say MIT spent on the 're'-search ..?? :-P

  • Merv Humphreys

    Is this a case of reinventing the wheel?
    London's subway map is regarded as a design classic.
    Why not simply adapt it to local conditions in the USA?
    We don't always need original solutions to our challenges if there are available resources out there. We simply need to research the existing range of solutions and see if there are any cultural barriers to overcome before implementing them with suitable adjustments.

  • bertoa

    How often does a rider need to comprehend a subway map in a single glance? That is just not a useful metric of real-world usability; not many people (if any, really) will need to process an entire map at once. Who is traveling across town on all lines at the same time? No one.

    What a rider needs to do with a map is to figure out where they are (geographic context) and how to get to (readability and clarity of map symbolism) where they need to go (geographic context again). Additionally, the geographic context gives the user opportunity of choice of modes of transit ("oh, this station is really pretty close, I can just walk or take a bike if I need to").

    The Vignelli map, while really pretty, isn't very useful in a real world navigational sense (I say this as a 20 year NYC resident). Where the merits of his design do come into play is with the (online only) Weekend Subway Service Change map, because it DOES allow you to scan in one go ALL the lines & stations (and, most importantly, which are in and out of service). Mind you, this map/action is mostly for people who are already familiar with the system and just need to figure out what obstacles are in the way of them getting from A to B, not newbies trying to figure out where A and B are and then how to navigate between them.

  • Alex

    You took the words right out of my mouth. Heaping praise on the Vignelli map for its simplicity completely misses the fact that people found it very hard to use because of its lack of geographic context. It was TOO simple. And it even had geographic inaccuracies for the sake of the design. It showed 50th St on the 1 train west of the A/C/E line when it's actually east! That's a horrible inaccuracy that surely left at least a few tourists of the day perplexed or lost. The current map is messy. I won't argue that. But a good transit map balances usefulness with simplicity.

  • vicente ocana

    The great problem with subway maps is taking account of the millions of people who has already learnt the visuality and have aquired the visual patterns. If you try to change too much, all of them will complain because they are not eager to learn a new pattern. However, for the new users (and usually the brand designer who attempts to redesign the map is one of them) will have no problem to accept the new visual language.

    As always, a profound user research is a must in order to design an interface (and a map indeed is one)

  • jbfoxlee

    The claim that "(MIT)...devised a computer model capable of determining how well people will
    comprehend a subway map (or any other complex diagram) in a single
    glance," seems dubious. I think it's more accurate to say it predicts how well a diagram can be visually processed. Note: i don't make maps, this is a layman analysis.

    Comprehension is not something you can determine solely on the visual layout of the diagram. Context matters. Especially with maps.

    I bet you could design an NYC subway map that looks perfect to this model that makes absolutely no sense to anyone. Just keep an appropriate amount of space between lines and make sure they all tail away in relatively neat ways from the origin. The choice of the origin in the samples really sets it up. Not saying the current map is perfect, just questioning the model's validity for the purpose.

    Having a diagram represent the natural state of the subject does matter, and is why Viginelli's map was criticized. The current 1979 map does balance space with physical representation, and something this MIT model does not understand as it clearly has no geographical component. (If it does I'll eat my lunch).

    To me a good map would be the optimization of the visual findings of this model vs geographical accuracy.

  • Kenneth Field

    So this is really just a way to 'see' the impact of cartographic design...the relationship between form and function and the way the map-maker has used visual hierarchy, figure-ground, contrast etc to focus attention. It gives us a picture (which we then subjectively evaluate) but can it also lead to some quantifiable measure of a map's success? I'd be interested to see a full write-up with methods and also a way to use the techniques myself. This would be useful to explore alternative map design proposals.

    Can you also post some mongrels of Beck's classic 1933 map given he is most commonly credited with producing the most elegant map design of this type. Of course jen Lea (below) is right...Vignelli was by no means the first.

  • Jen Lea

    Don't you mean Henry C Beck and George Dow knew what they were doing? It was those two men who initially proposed non-geographical linear diagramatics for underground transport, using horizontal, vertical and diagonal lines and making the distances between stops on the maps equal (regardless of actual geography). They were doing this as far back as the 1920s. Vignelli applied these ideas for the New York maps.