Dear Peter - Thanks for this very interesting material. I didn't get
Noah's e-mail, but here are a few comments:
  - I suspect your cultural interpretation may be on the right lines.
The theory of 'natural movement' (the proportion of movement due to
the structure of the grid)  says only that to the degree that built
forms are distributed throughout the system and movement is from all
points to all others, then we can expect to find a relation between
grid configuration and flows. This suggests to me that you might look
for 'correlation contours' (Noah may have suggested this, as the idea
originated in a study he did of downtown Boston over the summer),
that is regions of the grid where there is is a correlation within
the region, but which gets much weaker if you add in the adjacent
regions. A similar effect is shown for the City of London in the
original 1993 'Natural movement'paper. You systems are rather small
of course, so there is limited scope for this, but it would be
interesting to try. Noah has talked about an algorithm for
identifying correlation contours, but at the moment the method is
trial and error with the data table and the map !
  - some of your systems have 'cut nodes' - that is, lines which if
eliminated cut the system into  two. These are unusual in settlements
and could have effects. Formally speaking, nothing that happens on
one side of the cut node can affect the order of integration values
on the other (though it can of course affect its degree). I'm not
sure if this will have an effect on movement patterns, but it's a
factor worth bearing in mind in trying to diagnose what is going on.
Also, I don't think long sequences of lines on the edge of
settlements have been looked at before from the point of view of
movement. They do have quite a strong effect on the more grid like
parts of the grid, and you would need to understand the impact they
had on movement. Are these sequences lined with houses ? Or are they
the sparser parts of the settlement.
  - have you, in addition to your data table for each settlement, a
data table for settlements - that is, the rows are individual
settlements and the columns mean first order values such as
integration, connectivity etc and second order measures such as
intelligibility and synergy, and of course the r-squares for real
movement. Analysing the settlement data table may provide some useful
clues. If you have it, I be happy to look at it. One things that
could be interesting. You use the term 'predictability', and I assume
you mean the correlations between syntactic 'integration' and
'choice' (in fact log choice as they are distibuted very
differently). Integration is the measures of the 'closeness' or
'accessibility' of each node to all others, and choice the measure of
'betweenness', that is the degree to which movement on simplest paths
from all nodes to all other passes through each node. Predictability
was the term we used in the original 1987 paper for the correlation
between the two. At that time, we explored this relation a good deal,
and found it was possible to devise systems in which the correlation
between the two was pretty poor. Most notable was the kind of system
we called a 'handbag': that is clutch of well interconnected lines
(the bag) and a single sequence of lines leaving the clutch of lines
at one point and re-entering it at another (the strap). Some of your
settlements look pretty handbag-like !  We haven't pursued this
recently, as integration type measures (especially restricted radius
measures which we didn't have in 1993) were much more successful on
line maps than choice measures. But we now have a segment based
analytic capability due to Shinichi Iida,  called SEGMEN, which
weights segment relations metrically, geometrically and
topologically, and 'finds' the line structures as sequences of
certain kinds of segments relations. This model does both integration
and choice-like measures, and brings out more complex inter-relations
between the two type of measure. We have completed our first tests of
this model with excellent results (a paper will be in the public
domain soon), and we are now looking to test it more widely on other
data sets. It occurred to me that your settlements might be an
excellent case study, both to look at the movement pattern with the
finer scale analysis and also to explore structural features of the
settlement more fully. At the moment, we would have to run the
analyses here, but as the settlements are small this wouldn't take
long. At the moment we are looking to set up a dedicated machine to
run these analyses more or less constantly, to we can accelerate the
wider testing programme.

In the meantime, the best next move might simply be to look at the
outliers in your scatter plots and identify which lines they are. You
may be able to diagnose quite a lot just by looking at this. - Bill




>To Alan, Bill and Noah,
>
>Thanks for your input re/ the spatial configuration of Canadian arctic
>communities. I've included a powerpoint file with the axial maps of
>Arviat and Rankin Inlet - two communities on the west coast of Hudson
>Bay. The scatterplots for ATC (all terrain cycle) and pedestrian
>movement are also included (both r and r^2 values). I've also included a
>sample of the axial maps of a few other Arctic communities as a pdf
>file. There seem to be two primary types of layouts - radial and linear.
>Of the 23 communties I've analyzed, intelligibility ranges from 0.079 to
>.410
>
>At the present time, I've only been able to collect movement data for
>Arviat. Gate positions were set up on every street - usually midway
>along. Observations were made for 8 minutes at a variety of times during
>the day. Each Gate was visited at least 5 times during the field season
>(summer). The pattern of movement may differ radically by season. For
>example, in winter snowmobiles are used instead of all terrain cycles
>and cold temperatures may alter where people go, how frequently they go
>there, and how they get there.
>
>My current thinking is that the low pedestrian movement correlations may
>have something to do with 'what' is attracting movement and how that
>'what' is distributed within the community. The land use pattern in
>Arviat is one in which retail outlets (primarily the Hudson Bay Company
>store) and Government services lie along the most integrated routes.
>People (households), on the other hand, are distributed along the edges
>of the community. I "shadowed" the movements of several Inuit (with
>their permission) and observed that they spent 75% of their time
>circulating around the outer edges of the community visiting with
>family, and 25% of their time accessing retail businesses and government
>services in the integration 'core'. This strikes me as opposite to what
>one would expect to see in a southern Euro-Canadian town where people
>spend far more time cirulating around integrated locations accessing
>retail outlets, workplaces, government services, etc, and far less time
>circulating around their neighborhoods systematically visiting neighbor
>after neighbor (as do Inuit).
>
>In Inuit society, the family is the essential socio-economic unit of
>production and an important network of mutual assistance. Many families
>obtain most of their food through hunting, and pool their labour and
>equipment (ATC's, rifles, GPS's) to do so. So, I'm wondering if the
>pattern of movement I observed in Arviat is used to sustain these types
>of relationships? If so, then this might represent an example of how
>'cultural effects' can occasionally cirumvent or at least reduce the
>influence that the configuration of the community might have on human
>movement. As Bill has suggested, the other possiblility is that the low
>correlations could simply reflect the fact that the landuse pattern in
>these implanted and predetermined communities has not been shaped over
>time by the natural movement pattern. Perhaps it is some combination of
>the two. Regardless, would the mis-match between human movement and
>spatial configuration eventually have social consequences? Could the
>natural movement patterns of Inuit be used to develop new urban planning
>principles for Arctic communities?
>
>I'd be interested in getting some opinions from the space syntax
>community on this. I'm fairly new to space syntax, but I'd like to
>expand this study to other communities in the Canadian Arctic to see if
>the pattern persists.
>
>Noah - I am familiar with Doug Anderson...small world! I had no idea
>that he was interested in space syntax!
>
>cheers,
>
>Peter
>
>
>Attachment converted: Macintosh HD:Arctic Communities.ppt
>(SLD3/PPT3) (0008CD28)
>Attachment converted: Macintosh HD:Arctic Axial Maps.pdf (PDF /CARO)
>(0008CD29)
>Attachment converted: Macintosh HD:pcdawson 1.vcf (TEXT/ttxt) (0008CD2A)