Why the axial line?

Thank you to everyone for the fulsome replies to my little question.  In my various sporadic contacts with the Spacesyntax community I have noticed great generosity in dealing with dabblers like myself.    (Incidentally, the three people I remember being most vehement in their distrust of spacesyntax were all former Bartlett students of architecture  – what do you do to them?)

On Sat, 10 Feb 2001 Alan Penn <[log in to unmask]  wrote:
 Well - in the light of what I've said above, it should be clear that a curve can be represented by a direct connection, if that is what you are trying to represent, and so long as you are being consistent. The question is, if I represent it one way (direct connection) or another (broken up somehow) which version of the analysis tells me something interesting (for instance, which correlates best with an observed human behaviour or what ever). The choice of representation is part of the analysis, but the empirical data about the world 'talk back' and tell you which one to think about. This is just the beginning of course. If, say, the connected version correlated best with movement, you are now left with a new and more interesting question, that is 'why?'. This is why I think the question 'why the axial line?' is the more interesting question. 

This paragraph seems to get to the heart of my confusion over procedure.  You say that spacesyntax is a theory (or theories) about the effects of space on society which are falsifiable.  Doesn't this mean that you postulate a causal mechanism between features of the world and the phenomena being studied (observed human behaviour)?  The mechanism allows you to predict what type of behaviour will be observed in another situation which is not obvious from the first (non-trivial difference).  It also specifies  _exactly_ which features are relevant and must be measured or recorded.

If you try different representations until you find a correlation,  surely you are searching FOR a mechanism, not applying one.  And you might find that the axial integration correlates with the number of peppermints eaten by the researcher - probably not a causal relation!  You say that a curve can be represented by a straight link "if _that_ is what you are trying to represent"; but what is "that"?  Presumably it is some mechanism by which physical features cause variations in human behaviour;  but _what_ mechanism?

In talking about finding a suitable representation  for a problem,  you seem to be dealing with a problem which has two variables (probably independent ones).  You do not know what it is that is affecting people’s behaviour (is it configuration in this instance?), and you also do not know what combination of environmental features reliably  generate particular behaviours;  so how can you make falsifiable predictions?    

Kuhn refuted Popper?   Well the scientific method still gets along in an empirical sort of way doesn’t it?  I understand it is usual to undertake experiments where there is only one uncertainty.  This allows you either to discover whether particular measurable features cause predictable changes in something else, OR to predict the phenomena of the real world from the set of causes which appear to be present (or vice versa)

I seem to be ‘teaching my grandmother’ here – I haven’t even completed a research degree – and doubtless the relevant mechanisms for Spacesyntax have been worked out over the last twenty five years, but it is not very evident from the publications.  Are the mechanisms kept in the background for fear of frightening the public?  Or are they implicit in the accumulated wisdom at UCL?

I really think it is here that spacesyntax (as published) lacks clarity.  The measurements made are technical and hard to understand; as you say, this is the nature of science.  The behaviour studied is controversial and open to argument - only to be expected.  It is the casual mechanisms between them which are interesting,  which might be understood at a 'popular level , and which is able to provide evidence for the importance of spatial configuration which ordinary mortals could appreciate.


I am interested to hear that ‘minimum axial line’ maps can be generated by computer from the all-line map.  I gather this is not done in practice,  since the all-line software has not been updated for modern machines.   However, this rather misses the important question of what information goes INTO the maps.  Alan says that measures are generated from a dxf file of a plan, but unless you know what EXACT features cause the behaviour of interest, how do you map the right thing?  Always a danger of rubbish-in  = rubbish-out,  so I am puzzled that Sheep writes that "fractional analysis will make such map based observation questions irrelevant".    Especially as he follows this with "One thing we do not understand clearly from traditional space syntax is when the visibility matrix (where I can see) and the permeability matrix ( where I can go ) differ. For example an office with half-height partitions, or an office with glass walls."       This is surely a critical point,  in particular because on first impressions it would seem that visibility maps reflect the experience of strangers, and permeability maps reflect the experience of ‘familiars’ (residents?).  

The ‘fractional’ analysis of the axial lines that make up a curve is most interesting – but why draw short, straight lines around a curve in the first place?  Is it because the curve restricts sightlines (as most streets do) AND because sightlines matter to the users in question?  If the users are local residents, perhaps it only matters that it is a continuous ‘walkable’ path.  If the path curves through the lawns of the local park then sightlines are not restricted – there is no point at which strangers would loose sight of their destination.  The same goes for the underground walkway – do regular users find it confusing?  Don't they just KNOW which way to go?  Regular commuters find their way through a maze of tunnels in the London Underground stations without lifting their eyes from their newspapers!

There is an important point towards the end of Sheep’s mailing:  "think about the effects of a stair case if you do take it do you always come out facing the direction you want?’"    Orientation is a separate factor from wayfinding which I had not considered.  I assume that the important point is that those who are familiar with a space know where they are when they get there, and do not need to maintain a sense of orientation.  However, when I first went to Venice I stayed for a week without a map.  I ‘knew’ a number of ‘direct’ routes across the city by gradually-extended exploration.  I recognized my departure points and my destinations.  Sometimes, to my surprise, I found that one ‘direct’ route crossed another (or arrived at the same destination).  It was a surprise because my mental map had no global orientation:  I didn’t know which destination was near another.  When I returned another year with a map I was fascinated by the relative location of landmarks, and just how convoluted some of my ‘direct’ routes had been!

The converse of this point is Alan’s question: "the simplified 'visual' line correlates best - you have an interesting question to answer - 'why?"     Well surely this is because we are measuring "Routes along which people can keep their destination in sight."  Or as I put it before, 'continuous paths coincident with lines of sight.' (incidentally, Alasdair’s  question of defining ‘exactly’ the axial line surely refers to defining which lines are relevant to the ‘minimum-depth axial map’, not to any old axial line, doesn't it?)   In Alan’s next mailing,  he says "The procedure is both pragmatic and empirical, but axial maps are not arbitary. They reflect somthing about the world"  .  Surely we can be more definite than this,  the axial line reflects what I just wrote, or perhaps something better formulated by those with long research experience,  but how can you work with it if you do not make it explicit?  You can’t use ‘something’ as a research tool can you?  This question is different from defining which lines are maximal.

The following question, "Why should a map drawn one way correlate better with observed movement . . . "   seems to be another version of this same question:  if you know the mechanism, then ‘good maps’ show all (and only) the features relevant to that mechanism, bad maps confuse the issue.   Alasdair writes that "there is nothing wrong with there being more than one solution, or indeed, calling several different solutions all "good" axial maps."     Again, can’t we be a bit more definite about what types of axial map are possible? (leaving aside confused ones); permissible maps reflect the possible variations of mechanism. 

Surely there are only two types of mechanism available for human movement (at the whole body level):  Where am I able to go? and Where am I allowed to go?  

The first one deals with the constraints of the physical environment and of our understanding of it (wayfinding, and possibly the ‘fluid’ analysis of crowd movement  (by the way Alasdair, when did air cease to be a fluid?))  It seems to have two variable parameters,  prior knowledge and physical ability.    This might make four maps depending on what ‘watersheds’ are found in the variables, or perhaps one map with variable figures annotated to each line.  Prior knowledge probably only has values of ‘stranger’ , ‘visitor’ and ‘inhabitant’.   Physical ability  might perhaps be scaled as;
Wheelchair – average – desperate. (to finish up with Alan’s postulated pole-vaulting burglar).  It would need research to see where the variations along these parameters made a different to the features that need to be measured.  Has anyone done this?  Or am I wrong in this speculation of map-types?

Of course, there is also the second question; Where am I allowed to go? For instance, when I walk through a carpark  and glance down an alley to see a courtyard beyond, why does it look like "the way to go" in some instances, but not in others?  I would suggest that it is not just whether I can see my destination (axial wayfinding) but whether it looks ‘private’ (an issue of social control, not defensible space)

The two questions seemed to be equal partners in the first book (social logic of space) but since then spacesyntax seems to have split into ‘Hillier analysis’ (axial lines, natural movement, co-presence) and ‘Hanson analysis’ (boundary spaces, social control, observation).  I suspect that axial analysis works so well on city size problems because they 'average out' the variations in behaviour caused by personal differences.  At the opposite end of the scale, I was a little disappointed that ‘Decoding homes’ concentrated so much on rooms (boundary spaces). This seems to do the opposite, to make personal relationships of occupants so definite that the universal effects of public movement seemed to become secondary. 

I feel that spacesyntax has unique ways to explain human behaviour in small systems of semi-public space (which is continuous and ambiguous), because it is here that purely spatial relationships are so important.   However, the mechanisms probably lie beyond the axial line:  perhaps someone has researched this already?  There are hints at the end of 'Decoding homes' that it is being given attention.  Is there a way of seeing what work is being done?  I appreciate the UCL website, where I have found papers by the main players, but there must be lots of other stuff going on isn't there?   Alan  writes that I "should read Ruth Conroy's PhD"  How do I do that?  Are there any summary papers published?  Then there was the comment about "theory or hypothesis like in my book", I don't think I know about this either.

Once again, thank you for all the advice and information - on the website and on this list.  It has really been something to think about!

Sincerely

Tom Dine
[log in to unmask]

Chassay+Last Architects
Primrose Hill
London