London Underground Map Myths

A diagrammatic London Underground map can and should be correct in all geographical and technical details

On the web and elsewhere, whenever the London Underground map is discussed, a 'geographical correctness' contingent often surfaces to complain vociferously about the geographical anomalies on the particular design in question.Bayswater is too far from Queensway, Lancaster Gate is not near enough to Paddington, Hampstead is too far from Finchley Road, North Acton is too far from Willesden Junction, Temple is too far from Holborn, South Ealing is too close to Acton Town, Northwick Park is too far from Kenton, Edgware Road Bakerloo should be North of Edgware Road Circle Line. Whinge whinge whinge whinge! Don't these people ever stop? Well yes, very occasionally such a critic declares that he/she will attempt to produce a geographically accurate diagrammatic Underground map that is attractive and easy to use. Such people are then never heard from again.

First, note that The London Underground map is what is known as a Journey Planner. In other words, its purpose is to assist a user in planning the most efficient route between two stations. It is not intended to assist in identifying the two stations, you need an A to Z for that. It was never intended to be geographically accurate, nor (as we will see) could it be.

Second, note that all these complaints are coming from people who actually know London very well, so although they may find the anomalies irksome, they are not fooled by them. They therefore seem to be complaining on behalf of unknown users, perhaps some tourists, who either arrived at Paddington on the Heathrow express and failed to realise that if they lugged all their luggage over to Lancaster gate, they could get the Central Line directly to their hotel at Holland Park, or perhaps took a shopping trip from their Hotel at Bayswater, via Notting Hill Gate in order to visit some tourist tat shops at Queensway. Elsewhere, thousands of daily travellers From Bushey to Uxbridge fail to realise that they could walk from Kenton to Northwick Park, and thus easily travel by train rather than car.

It would be nice to have some evidence of large numbers of naive users having problems like this, and thanks to Oyster Card, technically the data do exist in order to find out how many people make ill-advised journeys. In the meantime, until we have clear evidence of people being misled, we need to be careful. It is precisely this nannyish 'attempting to help the user without attempting to understand the user' that has led to the erosion in design quality of the official Underground map. Instead, let's think about whether there might be costs to a geographically accurate London Underground map, or whether one really would be desirable or even practical.

In the well-known film, A Few Good Maps, Jack Nicholson famously said "You can't handle geography" [sorry]. Actually, you literally can't handle geography. The London Underground network is so extensive, and its centre so compact, that in order to have a legible centre, even with some of the extremities lopped off, a scale geographically accurate map would need to be approximately one metre by one metre. Try unfolding that on the Central Line during the rush hour

In fact, during the 1930s and also in 1947, London Transport did produce a geographical Underground map, but the fact that we don't have one today suggests that demand for it was limited and that it did not justify the expense of printing. Here it is:

It is a rather nice map, and is not too unwieldy to use, and at least it clarifies which stations are in easy walking distance of each other, but it does have shortcomings. For a start, it's only of limited use in identifying the best station. Because individual streets cannot be shown, you have to know roughly where you are/want to be, or the nearest main road. Also, there is insufficient space to show the centre, so the map has a cutaway on the back. This is absurdly inconvenient. Every single journey from suburbs to centre involves consulting two separate maps.

Users (that is, real people who plan real journeys using the Underground map) do not want or need a scale geographical map. Nonetheless, the London Transport Museum will happily sell such maps to anyone who wishes to have a go at trying one out.

Perhaps one solution to this problem would be a partial distortion of scale, magnify the centre but not the suburbs. This is, of course, what every diagrammatic Underground map does, and the results are plain to see on the current official map. The most obvious knock-on effect is that station distances in Zone 2 tend to get very exaggerated, especially horizontal distances. Finchley Road and Hampstead stations are in reality just over 1/2 mile apart, but you wouldn't guess that from the current official map. Also note the exaggerated triangle defined by Waterloo-London Bridge-Kennington, and the enormous distance between (nearby) Morden and Wimbledon stations. Stretching the Circle Line is partly to blame for all of these.

In fact, although Henry Beck is sometimes credited with devising the technique of scale distortion, there is evidence of Underground map designers making use of this decades before his work. Scale distortion need not be confined to diagrammatic maps, as this rather nice example from Bartholomew in the 1950s shows. The Bartholomew map doesn't suffer so badly from the problems highlighted in the previous paragraph. Perhaps there is a lesson to be learnt here.

Distorting scale without becoming diagrammatic (I)

It's an interesting map, but it doesn't go far enough. There is too much unnecessary detail. There is absolutely no need to know that, for example, the Central Line swerves a bit between Stratford and Leytonstone, and the centre of London still looks a bit awkward and compressed. What we are really trying to do is capture the spatial relationships between adjacent stations, so that two stations are nearby on the map, they are nearby in reality, and if one is due north on the map, it is due north in reality. So, let's try that one again.

Distorting scale without becoming diagrammatic (II)

With this map, I've tried to tame all those meanders, which add complexity without adding assistance in journey planning. Also, to fit in the station names slightly more easily, I ended up applying a 15 degree tilt to the centre. This has distorted the spatial relationships a little bit (e.g. compare Sloane Square on the above two maps) but I suggest that this does not affect the relative positions of nearby stations and so the benefits outweigh the costs. The result obviously isn't a scale geographical map, it's more a spatially representative map. Even so, in a few places (I will leave it to readers to identify them) the exact spatial relationships of nearby stations could not be shown. Many people fail to appreciate that in real life, Underground lines do not have to swerve round station names up to a scale 1/2 mile long. On a map it is different, either you make the scale impossibly large so that legible names can be fitted in, make the routes of the lines so contrived that they are difficult to follow, or else overlap station names over lines, muddling up details at parts of the map that are already congested.

So, if you really must have a map that keeps the geographical anomalies to an absolute minimum, then this is the map for you. You could try to convert it to Beck's rules, but I don't recommend it. You will just end up with something ghastly like the map discussed in Myth Number 11. Sized to fit A4 paper, the map is easily legible, and therefore usable. On the other hand, because of the scale distortion, a user will always have to bear this in mind when planning a journey. Tottenham Court Road to Holborn is easily walkable, but not Chorleywood to Rickmansworth.

But when I show this map to people, most don't like it. It looks a bit wayward and untamed, perhaps even visually disturbing. A proper diagram seems to be preferred, one that gets rid of as many changes of direction as possible, even if this introduces geographical anomalies. On the geographically accurate map, compare the correctly-awful route that the Central Line takes from Shepherd's Bush to Ealing Broadway with the simple straight line on even the current official map. The cost is that North Acton is a long way from Willesden Junction (one mile in reality), the benefit is that the line is easier to follow.

Assuming that people prefer diagrammatic maps which keep changes of direction to an absolute minimum, even if this results in geographical anomalies, what can the designer do? Global accuracy is impossible, but maybe we can at least help the user a little bit by pushing certain stations within easy walking distance together so that he/she can see how easy it is to change between them if necessary. Of course, if the user already knows which stations are close together, or has an A to Z at hand, then the information is superfluous. We will also have a more serious problem: the naive user cannot know which station locations on the map constitute travel hints, and which constitute accidents of the distortion process. Which looks easier on the current official map? Liverpool Street to Moorgate, or Farringdon to Kings Cross?

Which looks easier, Kings Cross to Farringdon, or Liverpool Street to Moorgate?
Can you redesign the whole map to eliminate every anomaly such as this? Or is this futile?

This gives us our first conclusion: It is pointless complaining about local geographic anomalies on a usable diagrammatic map because the presence of any anomalies at all means that the user cannot make any use of the geographical information. The user cannot distinguish between (1) accurate geographical information, and (2) inaccurate geographical information that results as an inevitable consequence of the distortion process entailed in making the map usable.

This then mires us with issues of how to show between-station interchanges and which ones to show, and leads to complex tube-walkline maps and so on. People start expressing strong opinions on the importance of showing that Northwick Park to Kenton is easily walkable, without considering whether this really is a strategic interchange, whether a novice user without an A to Z really wants to chance a street interchange that may not be adequately signposted, and if the user actually does have an A to Z, why is there a need to duplicate information that he or she is already in possession of.

Let's move on to a more specific issue; preserving the relative spatial positions between stations on a diagrammatic map, even though it will inevitably contain geographical anomalies elsewhere. Sometimes this will be impossible in any case. Harringay Green Lanes is almost directly above the Piccadilly Line, between Turnpike Lane and Manor House, but it can't be placed in the 'correct' location, because then it would look like an interchange with the Piccadilly Line. Even when relative spatial position can be shown, we will face the same problem as before: how can users know which spatial layouts are 'correct' on the map, deliberately shown, and which are incorrect because of constraints imposed by the designer following diagrammatic rules? But even if users could tell the difference, how would this information be useful?

Imagine you are changing from Euston Square to Euston. The current official map is not quite right, Euston is northeast of Euston Square, not north. Even so, do you see armies of bemused people clutching maps trying to see which direction to turn? Of course not, people follow the signs. And suppose someone makes a mistake and emerges on the south side of Euston Road without realising it. If this person went just by the Underground map, the error would be compounded and the person would end up at Warren Street.

Relative spatial location implied by an Underground map can only be used by a person in possession of a compass. Such a person will first need to be able to identify the orientation of the station exit. Not all stations oblige. Just because one station is north of another, it does not mean that the exit will be facing north. Second, the person can then identify the general direction needed to head, although if the local streets are unhelpful (none available in the exact direction required) then an A to Z will be needed, although if the person has an A to Z then the compass and the spatial information on the Underground map can be dispensed anyway.

Of course, there is nothing wrong with attempting to preserve relative station locations on a map, unless this priority results in a reduction in usability in other parts of the map; extra kinks or worse. Baker Street to Paddington is an excellent example. The problem is that the Bakerloo Line really can't make its mind up, and Paddington station itself is a bit hopeless, with the Hammersmith & City Line at one end of the main line station, and the rest of the Underground lines at the other end. An attempt to show Paddington 'correctly' resulted in the following usability disaster in the late 1990s:

Travelling from Latimer Road to Kensal Green? Be careful!
Which Paddington Station should you change at?

The solution was to run the Bakerloo Line underneath the Circle, which is far less misleading, at least for Paddington:

Paddington is now correct, but Edgware Road is wrong. Does it really matter?

Designing a diagrammatic map for a complex network is a compromise. Paddington is now shown in a less misleading way, and provides a suitable warning about using this as an interchange station (although we might argue that the user also needs to be warned about Bank/Monument Central to Circle Line). Does it really matter about the relative locations of Edgware Road? The typical user will leave the platform, enter the ticket hall, see a big sign over the door pointing in the direction of the other station, cross the roads and find the other station. It's not difficult and it is very well signposted. No need for your compass and A to Z then.

This gives us our second conclusion: Try to get things 'technically correct' if you can, but not at the expense of the user. There are many different users who need to be catered for: focus on the less knowledgeable users who might be most easily misled or confused, pay less attention to the complaints about minor aberrations from more knowledgeable users. They know where everything is anyway, and hardly ever make use of the map.

The top of the Circle Line is interesting from another aspect of getting things technically correct: getting the interchanges right. Paddington, Baker Street, and Euston are 'correct' ish, but might that not lull the user into false sense of security when it comes to Kings Cross?

Paddington has no cross-platform interchanges, Baker Street has a cross-platform one way and a same-level interchange the other way between the Bakerloo and Jubilee, and a same-level interchange in one direction between the Metropolitan and the rest of the surface lines. Euston correctly shows the cross-platform interchange between Victoria Line and Northern Line City branch (but not the actual directions that are cross-platform), but what does that tell us about Kings Cross or Bank?

Here again, we have the classic problem for the novice user. How can he/she tell the difference between a configuration that is a travel tip (Euston) and a configuration that conveys no such information (Kings Cross). Indeed, given our experience with the other stations shown (including Bank), Kings Cross implies a cross-platform interchange between the Piccadilly and Northern Lines, and that the Victoria Line is a dead loss, so that no one should contemplate interchanging from this to the Circle Line. On the other hand, for people with intermediate levels of knowledge, getting Euston right probably serves as a good memory-jogger, and provided that novices are not misled by these configurations, this is probably a worthwhile piece of information: the costs are outweighed by the benefits.

Not so my last example, where the cost is plain to see. The way in which Underground lines relate to each other is now meticulously shown, with the 'surface lines' (Metropolitan, District, Hammersmith & City and Circle) always passing above the deep-level tubes. This doesn't seem to have become a rule until Harold Hutchison's 1960 map (a designer that most would otherwise think twice before emulating). Really, this isn't much more than an in-joke. It certainly has no utility when making a journey. Changing from surface lines to tube lines does not always involve looking for stairs or escalators down. At Kings Cross, it is first necessary to go up before going down, as is also the case, for example, at Liverpool Street (westbound surface lines to Central Line). The moral for this story is, as always, when changing trains look for the direction signs.

So, what is the cost? The top of the Circle Line includes two other lines, Metropolitan and Hammersmith & City, a very thick band of colour, and one that is not going to get lost easily. Running the more spindly tube lines underneath does run that risk, however, and their obliteration at Kings Cross and Euston can only make things harder for the user:

Which is easier to follow? The 'correct' version or the 'wrong' version?

This has been a long discussion, but I hope that you can see that the complexities of map design and usability mean that a usable map of the Underground cannot be geographically correct, but this probably does not matter, and that although it can be made technically correct, this must always take account of the limitations in how this information can be used, and must never be at the expense of usability. Design critics should be very careful about what they expect a map to show, and what they perceive to be mistakes. The trick is not to make life harder for the novice, because such users know the least, they are the most likely to make mistakes, and the least able to make sense of all the various travel hints and tips that sporadically appear on the map.


If you read Myth Number 10, you will see that I cautioned the map designer against producing a map that conflicts with the user's mental model of a city. But here, I seem to be saying that it's OK to go against people's knowledge if it gets rid of kinks and makes a map clearer. How can this be?

Beck's Paris map tilted the city on its side, and was a gross distortion; a map that would conflict in a major way with people's most basic conceptualisation. Stations were way off where people might expect to find them, and the paths the lines followed bore no resemblance with reality in many places. Changing the relative position of two stations is a minor misdemeanour compared with this, and those small numbers of people likely to know and be bothered by this are probably those who know a city so well that they rarely consult and compare maps in any case.

Last updated 18/01/08