1The meeting of the European Colloquium in Quantitative and Theoretical Geography (ECQTG) held in Durham during the summer of 1999 marks the 21st Anniversary of such meetings. In this paper, and largely for the benefit of posterity, I outline the history of these meetings, showing how they owe their evident success to the informal way in which they are organised. A very informal quantitative analysis on the nature of the presentations given indicates that although there have been substantial changes in the immediate concerns, there has also been a strong concentration on central concepts within geography. In interpreting what follows, four qualifications need to be made. First, for all kinds of reasons, history is slippery stuff, especially when we look back from our present state. The temptation is to trace lines that lead to the present and to ignore the false starts, blind alleys and so on. They may not have an influence on what we do today, but at the time they were important. For this reason, rather than attempt a deconstruction of supposed 'trends' as shown by the series of meetings, the paper resorts to some basic quantitative arguments. Second, several office moves since 1978 have led to my losing vital materials. The paper relies heavily on the assistance of several people that has made it possible to compile a reasonably complete archive of attendee and papers presented. Finally, and despite my making strenuous attempts to avoid it, the view I present inevitably has a UK perspective (see also Longley, 1986 ; 1990).
2Two apparently unrelated events led to the initial meeting of the ECQTG. In 1977 Professor S. Gregory, author of the vastly influential Statistical Methods for the Geographer (1963), was invited by our colleagues in the French Group du Pont to address them. Those who have read about the supposed Bristol/Cambridge axis in UK quantitative geography and think that it was all to do with Haggett and Chorley on a supposed 'Cambridge (Chorley) to Bristol (Hagget) Axis' should note that there also was another axis from Manchester to London with two unsung and largely forgotten influences from climatologists Crowe and Gregory. I recommend Crowe's review of the 1967 Cambridge/Bristol book Models in Geography as an absolutely classic of deconstruction. Gregory came back to England excited by what he had heard and seen and decided to try to promote further contacts. The second event was a visit at more or less the same time to the little town of Lampeter in West Wales by Kilchenman (Karlsruhe). There he met Dawson and together they came up with the idea of some meeting of European quantitative geographers. I happened to visit Dawson the weekend after for some work on a joint project, and we carried the discussion forward.
3At the time I was the secretary of an Institute of British Geographers Study Group in Quantitative Methods and, following the conversations in Lampeter, Gregory asked me if I could make some progress with the idea. At the time I think it is true to say that we in UK were the best organized, and, with a membership of around 450, certainly numerically the largest, grouping of quantitative geographers in Europe. In UK I approached our main social science funding agency (then called the Social Science Research Council) and a modest award to fund around 15 researchers to attend the first meeting resulted. Various letters ensued until an exciting offer to organize a venue from Rimbert at CNRS Strasbourg arrived. The result was a meeting - by invitation only, and with limited numbers - of French, Austrian, German and British quantifiers in Strasbourg in the autumn of 1978. Interestingly, in view of later changes, this meeting was conducted making use of all three languages and seemed none the worse for it.
4In my report to SSRC (Unwin, 1978a) I concluded
It would be foolish to pretend that there is not still a wide gap between the research traditions in France Germany and UK that rises in part from the academic history outlined above but is also related to the very different research traditions in which it is embedded. Further it was recognized that it will be many years before a genuinely European dimension is added to quantitative work'
5A major outcome of this first meeting was an arrangement for a second meeting held in Cambridge in 1980. In his review of this second meeting Bennett (1981) was a bit more explicit :
The review papers highlighted in remarkable fashion the degree to which geographical outlooks differ between the participating countries.
6Bennett went on to recognize six themes of contrast :
-
Differences in the support of national agencies for quantitative geography and the extent of its integration within any geographical 'establishment'.
-
Differences in the degree of human/physical interaction.
-
Differences in the extent to which spatiality was a unifying theme.
-
Differences in the extent to which academic geography was a cover set for all the spatial sciences.
-
Differences in the extent of the involvement with planning.
-
Differences in the extent to which quantitative work was seen to be under attack from other ideologies within the subject.
7Table 1 shows the number of attendees at each of the eleven meetings to date, as well as the number of formal presentations. The attendance data seem to show that the perceived pleasantness of the venue rather than the science is important in determining the meeting size. The most popular remains Chantilly (Note that in his book The Geographer's Art, Haggett used a photograph of the attendees at the Chantilly colloquium to illustrate international networking), with the Italian Alpine resort of Bardonecchia and Cambridge in England close seconds. The fine old cathedral city of Durham has also proved popular.
Location
|
Year
|
#people
|
#papers
|
Strasbourg
|
1978
|
49
|
29
|
Cambridge
|
1980
|
120
|
21
|
Augsburg
|
1982
|
82
|
45
|
Eindhoven
|
1985
|
83
|
65
|
Bardonecchia
|
1987
|
124
|
95
|
Chantilly
|
1989
|
165
|
74
|
Stockholm
|
1991
|
90
|
66
|
Budapest
|
1993
|
78
|
52
|
Spa
|
1995
|
80
|
59
|
Rostok
|
1997
|
111
|
79
|
Durham
|
1999
|
95
|
72
|
Table 1 : The Colloquia
* In addition to plenary paper sessions both Strasbourg and Cambridge featured Table Ronde discussions. This is a total of over 650 papers.
8Table 2 shows the numbers of attendees to each of the first ten colloquia subdivided by their nation of origin. The original data are not as unequivocal as might be imagined. From time to time, a stated origin is not the delegate's nationality, some of the listed delegates weren't actually present, and, equally some local people do not appear in the delegate lists. However, you can see some obvious trends in these data that parallel and reflect the development of Europe itself over the same period. A basic model would be that the host country is usually, but not always, the best represented and normally - if the venue isn't in France - the next most numerous group are those from France, followed by UK and Germany. You might like to speculate on this relative to the numbers of academic geographers in each of these countries and perhaps also the notion of intervening opportunities in the various countries in the form of other related conferences. Proximity in space and time of some ECQTG meetings with those of the Regional Science Association also helps explain some of these data. Until the 1999 meeting there was clear evidence of widening participation. Cambridge saw the addition of colleagues from the Netherlands, Veldhoven a group from Sweden, and so on. The three most recent meetings have seen the very welcome addition of delegates from Russia and from countries in the former Soviet Union. The 'other's' group is of some interest and the attractiveness of these meetings has been reflected in a pretty steadily rising number of delegates from Canada, USA, Israel and so on.
Table 2 : Origins of Colloquium delegates as % of the meeting totals
Deductive modelling of systems
|
15
|
Spatial Statistical analysis
|
12.0
|
AI/IKBS and 'geocomputation'
|
4.5
|
Geographical Information Systems
|
10.0
|
Spatial interaction modelling/networks
|
10.8
|
Classification and taxonomy
|
4.8
|
General statistics, linear models
|
8.3
|
Geometry and fractals
|
2.3
|
Choice/Location/Allocation modelling
|
10.6
|
Time series modelling
|
2.9
|
Remote sensing, sampling and data
|
4.4
|
Cartography and visualization
|
3.2
|
'Introspection'
|
8.0
|
Unclassified
|
3.0
|
Total n = 657
|
|
Table 3 : Overall proportions of papers in 14 topic categories
9It can be seen that the most popular topics have been in mathematical modelling, with spatial analysis, spatial interaction modelling, and location/allocation modelling more-or-less equal second. I suspect that our post-modern colleagues would find this summary of over 20 years work surprisingly 'geographical' in its methods and objectives. These overall figures show some changes in relative frequencies over time which seem fairly obvious but which are none-the-less of interest. Table 4 presents the relevant data. There are some clear trends in these data, of which maybe six are worth recording :
-
The obvious decline in 'introspection'. The first two meetings were dominated by presentations and discussions relating to the state of quantitative geography in each of the participating countries. Inevitably, this type of discussion has declined.
-
2.The rise in deductive and mathematical modelling of systems of interest.
-
Local influences also play a part, such that it is possible to detect the influence of the nature of geographical research in the host country that, perhaps inevitably, has usually provided a heightened proportion of the papers.
-
'GIS' as such came in at Augsburg in 1983 but in no sense does it dominate. It seems to me that participants at these meetings have rather more sense than to allow them to be dominated by materials that are well-reported elsewhere, notably in the GISDATA series.
-
The rise and fall of discrete choice modelling.
Table 4 : Percentage of papers in each of 14 topic categories (columns represent meetings from 1978 across to 1999)
10Of possibly greater interest than these overall percentages is the fact that they are correlated. Table 5 shows the correlation matrix, R, between these proportions. To an extent, these correlations arise because we are dealing with ratios of a 'closed' numbers in the range 0-1. However, there are some relatively strong correlations which indicate that, relatively speaking, the topics rise and fall in popularity together. This is best illustrated in an informal way by an examination of the principal components shown as Table 6. The four components with associated eigenvalues greater than 1.0 generate the loadings matrix shown as Table 7. . These loadings are 'normalised' to 1.00 and, just as the results of virtually every published PCA make 'sense', so these can be forced into some kind of pattern. Component I has high positives on geometry, GIS, and AI with negatives on Time series. Its positive direction is towards Studying (spatial) form rather than process. Component II represents general statistical concerns versus spatial ones and is labelled General vs spatial statistics. Its positive side reflects work using standard methods and its negative spatial statistics. In contrast, Component III represents Time series and Classification. I am unable to give an interpretation of the loadings on Component IV and it is labelled I haven't the faintest idea !
|
Modelling
|
|
|
|
|
|
|
Spatial Stats
|
0.20
|
|
|
|
|
|
|
AI/IKBS
|
0.46
|
0.19
|
|
|
|
|
|
GIS
|
0.25
|
-0.19
|
0.56
|
|
|
|
|
Interaction
|
-0.43
|
-0.12
|
-0.03
|
0.21
|
|
|
|
Classification
|
-0.01
|
-0.62
|
-0.32
|
-0.07
|
0.07
|
|
|
G.L.M
|
-0.48
|
-0.32
|
-0.50
|
-0.28
|
0.24
|
0.24
|
|
Geometry/Fractals
|
0.15
|
-0.05
|
0.76
|
0.75
|
-0.02
|
-0.13
|
|
Location models
|
-0,46
|
-0,44
|
-0,18
|
-0,04
|
-0,17
|
-0,13
|
|
Time series
|
-0.24
|
0.18
|
-0.73
|
-0.79
|
-0.07
|
-0.11
|
|
RS/samp/data
|
0.10
|
-0.24
|
0.29
|
0.38
|
0.24
|
0.06
|
|
Cartog/ViSc
|
-0.16
|
0.12
|
-0.55
|
-0.73
|
-0.37
|
0.29
|
|
Introspection
|
-0.08
|
-0.06
|
-0.46
|
-0.65
|
-0.21
|
0.11
|
|
Unclassed
|
0.22
|
0.49
|
0.61
|
0.32
|
-0.20
|
-0.72
|
|
Geometry/Fractals
|
-0.40
|
|
|
|
|
|
|
Location models
|
0.03
|
-0.01
|
|
|
|
|
|
Time series
|
0.50
|
-0.91
|
-0.00
|
|
|
|
|
RS/samp/data
|
-0.00
|
0.23
|
-0.20
|
-0.12
|
|
|
|
Cartog/ViSc
|
-0.15
|
-0.45
|
0.19
|
0.41
|
-0.47
|
|
|
Introspection
|
0.02
|
-0.40
|
-0.44
|
0.49
|
-0.35
|
0.59
|
|
Unclassed
|
-0.43
|
0.31
|
0.40
|
-0.32
|
-0.03
|
-0.40
|
-0.54
|
Table 5 Correlations between the topic relative frequencies, (n=11)
Root
|
%of trace
|
%Cumulative
|
4.8977
|
35.0
|
35.0
|
2.6061
|
18.6
|
53.6
|
1.9750
|
14.1
|
67.7
|
1.4905
|
10.6
|
78.8
|
0.9516
|
6.8
|
85.1
|
Table 6 : Principal Component Analysis of the paper topics
|
I
|
II
|
III
|
IV
|
Modelling
|
0.18
|
-0.18
|
-0.46
|
-0.18
|
Spatial Stats
|
0.05
|
-0.48
|
0.08
|
-0.29
|
AI/IKBS
|
0.40
|
-0.09
|
-0.12
|
-0.07
|
'GIS'
|
0.39
|
0.21
|
0.01
|
0.06
|
Interaction
|
0.03
|
0.35
|
0.25
|
-0.29
|
Classification
|
-0.14
|
0.40
|
-0.29
|
0.32
|
G.L.M
|
-0.22
|
0.30
|
0.30
|
-0.17
|
Geometry
|
0.37
|
0.09
|
-0.14
|
0.21
|
Location
|
-0.01
|
-0.13
|
0.50
|
0.53
|
Time series
|
-0.37
|
-0.12
|
0.16
|
-0.34
|
RS/samp/data
|
0.16
|
0.29
|
0.00
|
-0.31
|
Cartog/ViSc
|
-0.32
|
-0.22
|
-0.19
|
0.35
|
Introspection
|
-0.31
|
-0.10
|
-0.36
|
-0.10
|
Unclassed
|
0.29
|
-0.37
|
0.28
|
0.00
|
Table 7 : Principal component loadings, paper topics
11Figure 1 shows a time series plot of the scores on PC1 Studying form for each of the eleven meetings. It shows a remarkable and steady increase over time that reflects an increasing concern in the colloquia for analysis using fractal geometry, basic GIS operations such as map overlay and visualisation. The meeting scores on PC III, shown in Figure 2 and labelled General vs spatial statistics, oscillate quite a bit but peak in 1982 (Augsburg) and again around 1990 (Chantilly and Stockholm) since when there has been a steady decline. This seems to indicate an increasing emphasis on analysis using methods specifically developed for spatial data. Figures 3 shows that the scores for Component III, labelled Time series and classification, peak in the mid 1980s (Veldhoven) and have declined since. Even with the benefit of the time series plot of the scores on it for each meeting I am still unable to interpret Component IV shown for completeness in Figure 4. As is usual with any analysis of this type these results are equivocal, but they seem to indicate that there are some interesting structures uncovered.
12Finally, as Table 8 shows, topics of a largely human geography nature have dominated the meetings with around 3/4 of all the papers presented at these meetings having had human geography subject matter. It is clear that, with some notable exceptions, our physical colleagues do not see these meetings as addressing the core of what they do. However, it should be noted that this wasn't true for the first two meetings where there may well have been deliberate attempts to balance. To an extent it also reflects the general demise of genuine physical geography in some countries and the more recent meetings in Eastern Europe, where there are different intellectual traditions, have to an extent redressed the balance.
|
Phys
|
Hum
|
Env& Mixed
|
Strasbourg
|
14
|
48
|
38
|
Cambridge
|
19
|
38
|
43
|
Augsburg
|
16
|
70
|
14
|
Veldhoven
|
14
|
76
|
9
|
Bardonecchia
|
9
|
78
|
13
|
Chantilly
|
7
|
76
|
17
|
Stockholm
|
7
|
73
|
20
|
Budapest
|
8
|
59
|
33
|
Spa
|
9
|
80
|
11
|
Rostok
|
18
|
65
|
17
|
Durham
|
18
|
58
|
24
|
Table 8 : Physical/Human/Environmental as % of presentations
13From time to time these meetings have been supported by small grants from our various national agencies, but they have never had seriously large funding from any EU, NATO, or similar programmes. In this respect they contrast with some more recent ventures with a geographic information science tag to them. So, in terms of the 'deliverables' how have we done ? In my view the answer is really quite well :
14First, a great deal of work in progress that has subsequently been published in the formal literature has been reported at these meetings for the first time. In particular, many young scholars have had their first experience of an international meeting at these colloquia and have gained immeasurably from the experience.
15Second, there has been a series of edited books containing selections from the papers given, usually organised around a particular specific area of theoretical or quantitative work (see for example Bennett, 1981 ; Bahrenberg, St.blein and Taubmann, 1982 ; Timmermans and Wrigley, 1988 ; Pumain, 1991 ; Holm, 1997 and Fisher, Sikos and Bassa, 1995)
16Third, the journal Geographical Systems and its reincarnation Journal of Geographical Systems grew out of discussions during an ECQTG meeting and, of course, CyberGeo itself was a direct result of resolutions made at the Bardonecchia meeting.
17Finally, there has been some limited joint work arising from contacts made at ECQTG meetings, but nowhere near as much as I would have liked or expected. A curious benefit that perhaps we did not expect in 1978 when we organised the first meeting has actually been that workers from individual countries have got to know their colleagues from elsewhere in the same country rather better. In the specific example of theUK I am aware of numerous research contacts and collaborations that have resulted. This unexpected result has a lot to do with the traditional informality of ECQTG meetings and I am reasonably sure that it also applies outside of UK.
18It is now 21 years since that Strasbourg meeting. It is traditional in many societies to mark the age of 21 in some way. In Britain we used to get what we call the key to the door, our independence from our parents, and usually had a party to celebrate. In the 21 years since the first Colloquium, academic fashion within geography has shifted many times. At the moment in many European systems the fashion is to find not just quantitative analysis, but almost any work based on real data problematic (see, for example the essay by former UK quantifiers Johnston and Taylor in Pickles, 1995). As all geographers know this is often the case, but the inference that is sometimes attached, that all quantitative and mathematical work using geographic data is either misguided or at best 'old hat' does not follow from it. Attendance at any of the recent colloquia in this series would demonstrate that quantitative and mathematical-theoretical geography is very much alive and well.
19Acknowledgement : Numerous people have been kind enough to supply me with missing papers and it is a pleasure to thank Paul Longley, Joost Hauer, Henri Chamussy, Denise Pumain, Stewart Fotheringham, Robert Rojeck, Bob Bennet and Reiner Schwartz for filling in some of the gaps.
© CYBERGEO