All 4 books by Edward Tufte now in
paperback editions, $100 for all 4
Visual Display of Quantitative Information
Beautiful EvidencePaper/printing = original clothbound books.
Only available through ET's Graphics Press:
catalog + shopping cart
All 4 clothbound books, autographed by the author $180
catalog + shopping cart
Edward Tufte e-books
Immediate download to any computer:
Visual and Statistical Thinking $2
The Cognitive Style of Powerpoint $2
Seeing Around + Feynman Diagrams $2
Data Analysis for Politics and Policy $2catalog + shopping cart
Edward Tufte one-day course,
Presenting Data and Information
Boston MA, April 24, 25, 26
Seattle WA, May 13, 14
Portland OR, May 16, 17
Cambridge MA, June 3, 4
Not a question, just an observation about a practical system that does a great job of erasing irrelevant ink from maps.
MapBlast has been testing (since March 2001, apparently) a new type of map for driving directions, under the name LineDrive. The maps don't include useless features not on your route; instead you get just the roads you're supposed to take, distances, and major cross streets, simplified and with proportional distances adjusted so that turns are easier to follow. In contrast, most driving directions just highlight a route on an traditional road map. It makes for a vast improvement in the readability of the direction maps.
The technology is described in a paper by Maneesh Agrawala and Chris Stolte, Rendering Effective Route Maps: Improving Usability Through Generalization. My take-away is that they modeled the maps on how people draw maps by hand when giving directions.
-- Paul Haahr (email)
I took a look at the LineDrive maps and have a few observations.
LineDrive adjusts the scale in ways that can be disconcerting. I made a map showing the route from the airport to my parents' house. One 23 mile section of highway is shorter than a 5 mile section on a different part of the route.
LineDrive only shows the route you are on. If you make a wrong turn or miss an exit, you'd be on your own. LineDrive maps have only one purpose. They are only useful for a specific route.
I don't think details on a map are "irrelevant ink." A decent highway map is full of information. You can use it to find your way from point to point. You can change your route in the middle of a trip. You can use it again to find your way to another spot.
I'd much rather use a decent highway map to find my way around than a LineDrive map.
I love to look at maps. I'd much rather use something like a wonderfully detailed map of the Taconic Range (large file), than a LineDrive map of the same area.
-- Michael Eglinski (email)
Usually for information displays, less is less. Indeed, to clarify, add data. See chapter 2, micro/macro designs, Envisioning Information on the issues here.
Clutter is not an attribute of information, clutter is a failure of design. So fix the design rather than stripping all the detail out of the map.
A design solution for the route map is to overlay (with a magic marker, say) the particular route on a detailed map. Then the particular is highlighted within a rich context. Internet computer resolution, in practice at least, is generally too low for good map design, so MapBlast may be making a virtue of a necessity.
Chapter 6 of Envisioning Information shows a classic Ogilby strip map from the 1600s, which is similar to, but better than MapBlast. The MapBlast maps resemble the maps on the widely purchased but little used car navigation systems.
Finally, the names "MapBlast" and "LineDrive" are, for me anyway, unpleasant and uncartographic, more about marketing than about real guide-maps.
-- Edward Tufte
Michael Eglinski has provided a superb comparison of two map designs for the same area. This comparison improves over design-chat analysis.
-- Edward Tufte
Keep in mind that the goal of LineDrive maps is to present a specific route to a navigator. Our assumption is that navigators will print the LineDrive map and then refer to it while driving to the destination. Often the navigator is also the driver. In such situations, LineDrive maps are designed to be much easier to use than a detailed road map. My experience is that it can be difficult to read a detailed road map and concentrate on the road at the same time.
A common problem with overlaying/highlighting the route on a detailed road map is that the scale of most road maps is usually either too large or too small. A road map that is designed to show every road in every neighborhood in the area often requires a large sheet of paper (difficult to peruse while driving) to show long routes. A smaller-scale map that can show longer distance routes will force all the short neighborhood roads to shrink to a point and essentially vanish. Compare the the standard MapBlast overview map to the LineDrive map shown at Rendering Effective Route Maps: Improving Usability Through Generalization. Notice that it would be impossible to follow the route shown in the MapBlast overview map to reach the destination. All the short roads at the beginning and end of the route are invisible. Although the LineDrive map is sparse it provides enough information to follow the route and its small size makes it convenient to use while driving.
The key idea in LineDrive is variable scaling. We ensure all the short roads are long enough to be visible but we also try to maintain the relative ordering of the roads by length. So shorter roads should appear shorter than longer roads in the final LineDrive map(In rare circumstances the relative ordering is violated). This is exactly the approach most people use when they are asked to quickly sketch a map showing a friend how to get to a particular location.
An underlying assumption in LineDrive is that navigators have some familiarity with their environment. Often the users of sites like MapBlast are requesting a route within their own metropolitan area and they have a basic mental model of that area. However, they might not know how to get to a specific address. The LineDrive maps provide just enough detail to help these navigators reach that specific destination. While a little more context information (like nearby roads) might be useful it must be balanced against the clutter it might create for navigators who may only be able to take quick glances at the maps as they are driving.
If you scroll down to the bottom of the LineDrive map webpage it is possible to include a standard MapBlast overview map and a zoomed in detail map of the region around the destination. We do this to provide a little more context in the regions that navigators are most likely to need it.
For navigators who are unfamiliar with their environment (such as tourists) LineDrive maps are not the right choice. But even the standard MapBlast maps are not the right choice for such navigators. Notice that the standard MapBlast maps don't provide very enough much detail beyond the route. If you make a wrong turn it will be difficult to get back onto the route regardless of whether you are using LineDrive or the standard MapBlast map. In such situations I too would prefer a large detailed road map.
Our paper describes all of our design decisions in much more detail.
While Michael's example of the map of the Taconic Range is beautiful I don't think it can't be compared to the LineDrive map because they were designed for very different purposes.
For what its worth I'm not a big fan of the names MapBlast and LineDrive either.
-- Maneesh Agrawala (email)
This answer and your paper are very thoughtful.
How about at least some context around the route map; there is enough space, and the context might rescue the driver/navigator in the event of a wrong turn. What might be the mistake rate in following a map: one wrong move in 15 or 20? Context will enable recovery.
Also context allows more than just navigating to one place; maybe the driver/navigator will learn or see something else! Also context enables memory, so the next time they go to the same place it will be easier rather than rote following directions. Context allows learning.
Modelling a design solution after what people actually do is a good idea; but when people draw by hand they are in a profoundly low-resolution display sitation. So again, push back visually a good contextual map and then overlay the route and instructions on top. Thus the grid of the route is not blank page, but rather is a real map in the background.
When done well, with appropriate layering and separation, the overlay method dominates all solutions.
-- Edward Tufte
Make the route maps large, with big type, since they are often being read by the driver (perhaps while wearing driving glasses, rather than map-reading glasses).
Also think about in-out locators; that is, something to make it easy to return quickly to the current location/sequence on the map. See the older editions of The Joy of Cooking; the recipes have little marks, in the middle of sentences, to assist in returning to the recipe at the appropriate point.
Some of these ideas are discussed in my Visual Explantions, in the material on field guides beginning at page 114.
Layering and sepration techniques are in chapter 3 of Envisioning Information.
-- Edward Tufte
After playing with several routes familiar to me, I agree that LineDrive maps would be greatly enhanced with some, but not much, additional context information.
The LineDrive maps and the methodology utilized are very similar to the method I used when developing a street directory for my local fire department. Fire department personnel use this guide to "navigate" from fire houses to emergency scenes, which can be confusing in non-grid New England suburbs. Firefighters have no need for context information- they are only interested in the best/fastest route. Note that these directions are entirely text- no maps! This guide is useable to someone not familiar with the area, but, like the LineDrive maps, is much easier to use with some local knowledge.
The addition of limited context information would seem to assist LineDrive maps just as it would improve my street directory: If a particular route is blocked/impassible for some reason, or a wrong turn is taken, the user should not be sent into a blank void. The addition of some parallel alternate routes, perhaps by grayscaling surrounding roads or offering a context overlay, would allow for some margin of error and not leave the user feeling the prescribed route exists in a vacuum.
-- Mark Kasinskas (email)
Layering is potentially problematic in this situation due to the mathematics of variable scaling. You generally can't draw a "map in the background" without running into trouble.
If you overlay a full map on a LineDrive diagram it will have to be distorted to fit with the main, variably scaled route map--and the full 2D distortion can be much more confusing than the scaling on the route. For instance, the "north" direction in the background map may be forced to change significantly from place to place, even if it changes only a little along the route map. And the changing distance scale may be a nightmare to label on the background map, compared to putting numbers on route segments.
I think the LineDrive solution works very well as is... true, more detail would be nice, but the variable scaling is an extremely helpful innovation and is worth the loss of context.
-- Martin (email)
When I need a map to drive somewhere, this is what I do:
1) Print out several maps. First is a detail of where I am, the last is a detail of where I'm going, and then some intermediate maps if necessary.
2) Arrange them in order on a clipboard or something similar.
3) As I drive out of range of one map, I tear it off and toss it in the back seat.
This way even the usual Yahoo! or MapQuest maps are fine.
In the future I will try the LineDrive map first and use my regular method as a backup. I expect it will work out well.
-- David Person (email)
Does anyone remember AAA TripTiks? I think you can still get them by going to the local AAA office (if you are a member). Before the days of MapQuest, MapBlast, MapsOnUs, and [insert Map-related moniker here], there was always the reliable guy in the AAA office who would: 1. Ask you where you were going. 2. Ask you what you wanted to see along the way. 3. Go get all the strip maps that AAA published along your route. 4. Highlight the roads you should take (using a highlighter). 5. Stamp all relevant "events" along the road: construction, tolls, exits, etc. 6. Bind it all into a booklet for you to keep in your car.
This service was fantastic -- the strip maps often unfolded to show, on the reverse, much larger area maps with hotel and local service information. The highlighted route was easy to follow, even for the driver-navigator, because of an excellent use of color. But it was the person putting the map together who really made this system work. My local AAA office had a guy working there who just knew a lot about driving and what there was to see, so he could draw on his experience to generate a fast route, a scenic route, a cheap route, etc.
While the various driving direction services make getting these kinds of maps easier, they also lose the human element. Additionally, since the web-based mapping systems are all designed to be viewed at screen resolutions, they lack the information density generally available in most adequate printed maps. I generally find the online services great for text-based directions and for giving me a good idea of how to find a place in a region I already have experience navigating -- but useless for much of anything else.
-- Scott Zetlan (email)
I think the word everyone is dancing around is "landmarks." I believe that there's a fair amount of evidence from psychology that humans (and other animals, too) often navigate by orienting to particular landmarks rather than using all the information in the environment.
For instance, when giving directions, people give a series of landmarks, and the number increases near decision points (i.e., when you have to make a turn.) As in, "You'll see a gas station" (landmark) "just before the intersection," (decision point) "but if you go over the bridge, you've gone too far" (landmark to catch errors).
Standard maps are versatile, because they provide an abundance of landmarks, all equally emphasized. The LineDrive system is intended for one, specific, single task, and tries to provide only the landmarks relevant to the task -- but it seems to go too far by omitting all nearby landmarks.
Turn-by-turn maps in Microsoft's Streets and Trips almost like this (more by accident than by design, I think). It's a series of panels that shows each intersection where you turn, so you can often see a few nearby streets. I like them, but I always print off an overview map, with the route highlighted, to complement them.
One other issue: I'll bet that reading a map while driving increase the risk of an accident. Not just because of the reading, but because you're forced to fiddle with a piece of paper. I wish there was a safer way to get navigation info in a car.
-- Zen Faulkes (email)
There's always the talking GPS which tells you when to turn -- if only the onboard computer could be upgraded more frequently to account for road construction. It's that pesky slowly-changing-data-over-time problem that plagues databases of all kinds (paper included: books must be constantly reprinted and redistributed as human knowledge changes).
-- Scott Zetlan (email)
Regarding data bases containing old data, I recently had occasion to check the "tax land parcel map" of a track of land in an adjoining county. The county had only completed its online mapping initiative a couple of years ago, but alas, the map it spewed forth predated a map from the early 1970's.
-- Gene Prescott (email)
I think the LineDrive maps are an excellent first attempt. I tried several routes I regularly take and found them to give good local details that are hard to spot on mapquest-type maps, and made visible by the variable scaling.
The person that mentioned landmarks is on to something I think. The LineDrive maps do resemble ones that we might draw by hand, except for landmarks. Maybe MapBlast should consider getting input from their clients on what landmarks to include at local intersections (offer a lottery prize perhaps?). That way, the database would become a living entity kept current by the community of users (oops -- is this a bad word?)
I remember once seeing directions used by rally drivers, and they contained a sequence of individual intersections with textual instruction beside each one. Maybe LineDrive should consider something like this: it would then be easy to allow clients to print the individual intersections at a size suitable for easy viewing while driving, and would also allow the display of local context at variable scaling. MapQuest does this with its turn-by-turn feature.
Another idea might be to print the step numbers on the map itself, to link the verbal tabular instructions with the local geography of the map.
-- Drew Knight (email)
Several responses mention the value of landmarks. I agree that this is probably the single most valuable addition that could be made to computerized driving directions. Certainly when we give someone oral or sketched directions we make heavy use of landmarks; e.g. "Texaco station", "K-Mart", or "big stone building with white pillars".
The main reason this isn't being done for computerized directions is lack of adaquate data. There is no database of visual appearances like "big stone building". It is possible to acquire a rich database of commercial locations, but this data is still difficult to use:
-- The location data is often not precise enough to place the structure correctly relative to an intersection. Probably at most 50-70% of business landmarks could be placed correctly as to being on the right or left and before or after a turn.
-- There is nothing in the data to indicate the visual prominence of the structure. Is it directly on the street with a large sign, or is it set back behind trees or hidden in a mall? The database will often contain a very large number of potential landmarks along the route but no basis for selecting the ones that actually stand out.
Assembling this kind of data for even a fraction of the roads and streets in the USA is a very expensive proposition and is not likely to happen any time soon.
-- Clifford Cary (email)
Thierry Guergiou wrote an interesting analysis of how he navigated through an orienteering course -- how he used a map to find his way. You can see an english translation here:
The French original is available at www.tero.fr
Guergiou pairs images showing a bit of the map he used during the event and a map with the specific features he used to navigate highlighted.
Orienteering is a sport that involves using a detailed topographic map to navigate through a series of checkpoints. The challenge is to find the optimal route and to execute it quickly. Guergiou is a two-time World Champion.
-- Michael Eglinski (email)
Google Maps (http://maps.google.com) has taken an approach quite the opposite of LineDrive by including satellite images in their maps. (Click "Satellite" in the upper right- hand corner of the main map page.) At the greatest magnification, the smallest object which can be clearly resolved is a car. Driving directions are given as a highlighted path above the detailed photograph, with accompanying text. The map is surprisingly responsive to direct manipulation, e.g. scrolling by "grabbing" the image with the cursor.
For a diversion, select driving directions from your house to a well known location (your workplace, the airport, or even the Washington Memorial -- "301 7th St SW Washington, DC 20407" puts one just south of The Mall, then look west for the characteristic needle). The abundance of detail is ... well ... fun.
Google's approach is akin to the bus route in Envisioning Information, p 108--109, though lacking the layering of street and landmark names. This certainly slows one's reading the maps, as some locations must be "calculated" from clearer nearby landmarks -- I know this funny intersection here is Claremont and Colby, so _that_ must be Telegraph Avenue. In foreign areas -- even in one's home town -- this calculation may prove impossible. Also, in some cases, the highlighted path does not coincide with the road in the photo. The highlighted path can be offset from the pictured road by as much as the road's width. Which is distorted, the photograph or the surveyor's data?
-- Sean Garrett-Roe (email)
An clever comparison of the readability of the three main online maps and why Google Maps is more readable than its competitors.
Interestingly the article starts with a quote by ET.
"It turns out that Google uses a variety of techniques and visual tricks to help make its city labels much more readable than those of its competitors. From the use of different shadings to the decluttering of areas outside of major cities, it sure seems like Google has put a lot of thought into how it displays the labels appearing on its maps. I have no doubt that little touches like these are among the many reasons why Google remains the web's most popular mapping site."
-- Andrew (email)
Apple has just patented a technology similar to LineDrive so this topic might soon be relevant again.
Ten years ago I was an enthusiastic user of LineDrive, and was disappointed when the technology was acquired and then buried by Microsoft.
My primary complaint with the technology was that it was geared towards the wrong output format at the time. It produced low resolution web images which I then printed on letter sized paper and took with me in my travels.
Given that the destination format was paper, MapBlast could have served LineDrive directions as pdf files to be printed on 8.5 x 11 paper. This would have allowed for sharp images, clearer text, and more detail. Detail could have taken the form of landmarks, parallel streets, cities passed on long drives, etc. All this while maintaining the "back of a napkin" advantages it possessed.
Now the primacy of paper in the car is a thing of the past, though if a pdf version of LineDrive were available I'd use it today. The question becomes, how to best make use of such a technology on a mobile device that is equipped with a GPS and allows for dynamic interaction.
I would argue that many of the same advantages a PDF format would have provided are a good starting point. Beyond that, an application that could intelligently scale detail such that you could see both your current surroundings (in terms of streets and landmarks) and then next two decision points at a glance would be enormously useful.
Such a thing on a high resolution tablet might even look nice if designed properly.
-- John Harrison (email)