Grant Ian Thrall, Ph.D. Software, Data and Business Geography Column Editor
GeoSpatial Solutions www.GeoSpatial-Online.com
email: thrall@geog.ufl.edu or GrantThrall@geo-tel.com
personal url: www.afn.org/~thrall

(c) Grant Thrall. 1995-2002. All rights reserved. Gainesville FL.

... what others have said about this published article...
"Grant Thrall ... in Geo Info Systems ... do[es] an excellent job of bringing the philosophy and theory of our discipline to the broader GIS community. Grant's February article "The Stages of GIS Reasoning" lays out in considerable detail his arguments for a five stage GIS model, Description, Explanation, Prediction, Judgment and Implementation- management, that incorporates a wide range of classical geographical literature and debate." Newsletter of the Association of American Geographers, GIS Specialty Group [David J. Cowen, University of South Carolina, Newsletter Editor]

... for further reading, please refer to ...
Wofford, Larry and Grant Thrall, 1997. "Real Estate Problem Solving and Geographic Information Systems: A Stage Model of Reasoning." Journal of Real Estate Literature, vol. 5, number 2, July, 177-201.

Thrall, Grant Ian, 2002. Business Geography And New Real Estate Market Analysis. Oxford University Press: London and New York. ISBN # 0195076362.

Introduction

The technological revolution of georaphic information systems (GIS) has triggered an intellectual revolution in the way we approach scientific analysis. We should occasionally reflect on where we have been and where we are going with our collective interests in GIS, so in this issue I offer my general philosophy of GIS analysis and management.

As a practical matter, I will provide some insight into how and why I select topics for the "ShopTalk" column. My goal is to prepare columns about each of the general GIS categories summarized below. More importantly, compartmentalization of GIS analysis is a first step toward understanding what we do as practitioners and students and teachers of GIS. Knowing that GIS technology is capable of more than what you are doing with it can help you begin to realize increased value with your GIS.

In 1993 I introduced a four-stage compartmentaliztion of GIS activity (Thrall and Elshaw-Thrall, 1993, p.65)

* Stage one: Description
* Stage two: Explanation
* Stage three: Prediction
* Stage four: Judgment

In the last two columns (Thrall and Ruiz, 1994; Thrall, Bates, and Ruiz, 1994) a fifth stage was implicitly added to my earlier list:

* Stage five: Implementation and Management.

All stage theories have a problem with what comes next. What is there after stage five? Is there a stage six? I have no doubt that other stages will be added to the list as GIS technology matures.

DESCRIPTION

Most GIS specialists focus mainly on the first stage, the accurate description and representation on a map of spatially distributed phenomena. Emilio Casetti has suggested that "[tlhe scientific knowledge of any facet of reality starts with descriptions and inventories of facts, observations, measurements, and with explanations based on narratives." Surveying may be considered an extreme case of geography as legal description. GIS practitioners are generally concerned with the descriptions of phenomena that lie above, below, or on those survey lines.

But the phenomena that we are interested in describing are not what make us geographers; rather, it is the manner in which we describe, analyze, or classify the phenomena that makes us geographers. I am deliberately inclusive by implying that all GIS practitioners are applied geographers. Agreeing that we are doing geography is a first step toward classification differentiating spatial phenomena from all other phenomena. GIS is currently the best vehicle we have for integrating the various geographic data operations that are required for geographic analysis. Although I am interested in urban land markets, and another GeoInfo Systems reader may be interested in the natural environment, we are both geographers because of our shared approach to analysis regardless of the divergent themes that may be our particular areas of interest. Geography is not distinguished from other disciplines because of its development of ideas concerning specific themes. What differentiates geography from thematic disciplines like physics, economics, and psychology is that geography - like history and geology - is an integrative discipline (see Casetti, 1993).

Thematic disciplines focus on comparatively narrow classes of phenomena; and, because of their narrow focus, they developed analytical knowledge earlier and to a greater extent than such integrative disciplines as geography. Instead, the integrative disciplines address broad classes of phenomena, perhaps even overlapping with phenomena of interest to the thematic disciplines. In the past this has relegated geography to "low status" as a discipline. Casetti (1993) writes that "[tlhe status of disciplines such as economics and psychology results partly from their ability to isolate several related clusters of relationships earlier on, and in the process to acquire de facto 'property rights' over them. Also, in these disciplines the development of theories, models, and validation techniques began earlier, went further, and culminated to a greater extent in well recognized bodies of 'analytic' scientific knowledge." Geography is no longer a low-status discipline. This change in status is parallel to the changing awareness of what geography can do for practitioners. The integrative technology of GIS has changed the rules of the game in the competition between thematic and integrative disciplines. Geographic information technology has taken away the competitive advantage previously enjoyed by narrowly focused thematic disciplines.

Consider Brian Berry's (1964) suggestion that classification in geography be conceptualized as a matrix (see Figure 1). He proposed that phenomena be categorized in terms of their type of activity, their location, and their time (as quoted in Fielding, 1974). This threefold division provides alternative differentiating characteristics exhibited by all phenomena.

Examples that fit within Berry's matrix can be found in the "ShopTalk" series.

I have represented homelessness (Thrall, 1993a) as an activity. There are homeless people, and there are nonhomeless people. Homelessness can be a long- or short-term situation. And the locations of homeless people can be identified. In another column, I represented an entirely different phenomenon - property-tax assessment as the activity(Thrall, 1993b; Thrall, Ruiz, and Sidman,1993). In the property-tax assessment analysis, time is held constant by selecting from all observations of property only those single-family dwellings that sold during the same market period. As geographers we are interested in the locational concentration or spatial dispersion of phenomena and the implications and ramifications of such geography. GIS is a technology for geographic description, a technology tool for the display of geographical information, and a vehicle for data visualization. The technology can make it easy to summarize large volumes of data and offer a new, higher level of productivity. I reported before (Thrall, 1993b) that in the 1970s I used computer technology to evaluate and describe the spatial pattern of property-assessment quality (Thrall, 1979). Data descriptive analyses like these, which required almost two years too accomplish in the 1970s, can now be completed in a few hours. Powerful data visualization coupled with higher productivity has brought a new way of approaching geographic problems. The technology truly allows the data to speak for themselves and allows for greater exploration of the data and greater opportunities for experimentation and hypothesis testing.

One school of thought within the discipline of geography maintains that geography's main objective is description: each place is unique, and we cannot make generalizations that allow us to apply what we learn from one location to understand the geography of another location (Hartshorne, 1939; Martin, 1994; however, see Lukermann, 1989 as quoted in Mart¡n, 1994). Indeed, to allow generalization would infer the natural evolution to the second stage, explanation.

This so-called idiographic school of thought dictates that what we learn about spatial patterns of property-tax assessment inequities in Hamilton, Ontario, Canada (Thrall, 1979) does not allow us to anticipate or predict the pattern of property-tax assessment inequities in Fort Pierce, Florida, USA (Thrall, 1993b). Without generalization, there can be no theory. Without theory, explanation is highly limited if not impossible. If we accept that each place is entirely unique, then information about geographic phenomena in one place cannot be drawn upon to anticipate such problems in another place. If each place is entirely unique then there can be no generalizations, no geographic theory, no anticipation of geographic patterns - regardless of how much observations from data explorations lead you to believe the contrary. But, if we believe that there are commonalities of like-kind geographic processes between locations, then each place is not entirely unique; the door is then open for geographic generalizations, or geographic theory.

EXPLANATION

Consider Waldo Tobler's First Law of Geography: "[E]verything is related to everything else, but near things are more related than distant things." By implication, if you change your location by a small increment, the resulting change in the geographic environment is expected to be so small that we can anticipate what the geographic environment will offer there. If you believe in Tobler's first law then you believe in a geographic theory, which implies a willingness to accept more geographic theories.

I believe that the best argument to the challenge that "geography can have no theory" is the creation and presentation of geographic theory itself; namely, the "proof is in the pudding" argument. There are many examples of geographic regularities and geographic theory. For illustration, the human side of the geography discipline offers explanations for

* the spatial distribution of settlements referred to as central place theory (Christaller, 1966; King, 1984),
* the spatial distribution of land values and land-use patterns (von Thanen, 1842; Hoyt, 1933, 1939; Alonso, 1964; Thrall, 1987,1990),
* human migration (Clark, 1984), and
* the manner in which people perceive and learn about their spatial environment (Golledge, 1978; Gould and White, 1986).

Many geographic generalizations are static. Static spatial analysis attempts to explain the state of affairs of a particular spatial phenomenon, independent of time and without regard to how the state of affairs came about. One example - cited from my own work only because it is most familiar to me (Thrall, 1987, Ch. 1-4) - of a static spatial theory from the above list, is how urban land values vary from one location to another. Urban land values generally decline with increased distance from urban commercial activity centers. The explanation for this general trend is that people are willing to pay more for the benefits of a shorter commute and pay less for the disadvantages associated with longer commutes. This generalization can be applied to explain the trend of land values in Vail, Colorado, USA; Gainesville; or any other place. The magnitude of the spatial trend of land values depends on the aggregate behavior of those individuals who make up the market, their ability to pay to avoid the disadvantage of commuting, and the strength of their preference for avoiding the commute. Households in Vail are overall more wealthy than households in Gainesville, and Vail residents generally place a greater premium on being near the urban core than do residents of Gainesville. Therefore, the rise in land values with proximity to Vail's urban core is expected to be greater than the rise in land values with proximity to Gainesville's urban core.

The evolution of geographic theory has followed the evolution of theory of its related disciplines. First, in the description phase, the collection and inventory of "factual information" is coupled with loose literary and/or philosophically based general- izations and explanations (Casetti, 1993, p.528) The manner in which I presented the

narrative of land values in Gainesville versus land values in Vail would fall within this tradition.

The next phase of theory development moves to the precise expression of concepts, often relying on mathematical constructs. Such mathematical constructs in the social sciences often fall within a class of models that may be referred to as heuristic theory; heuristic means providing aid in the solution of a problem.

In actuality, the above Gainesville versus-Vail narrative summarizes the implications of my own geometry-based mathematical construct (Thrall, 1987); a mathematically precise, general heuristic methodology that can be applied to evaluate the land use and land values of almost any urban situation.

What does the world of GIS have to do with the world of mathematical constructs designed to tell a story of relationships and causation? I believe that the general story gives purpose and direction to GIS analy- sis. Theory links GIS to the larger body of knowledge. GIS does not stand in isolation of the larger body of knowledge; rather, its utility hinges on its linkages to the larger body of knowledge. At the same time, GIS is now shaping the questions we expect our general theory to answer. And, because GIS is so revolutionary in its spatial data visualization, it gives rise to an awareness of phenomena in an unprecedented manner. GIS can inform us about the landscape of the spatial data base at hand. The modern geographer explores the virtual landscape residing in the computer, armed with the belief that within the particular spatial data base may lie the outcomes of general spatial processes. Not to be overlooked is the pizzazz that accompanies GIS; the mathematical heuristic approach looks rather dusty in comparison.

GIS makes basic spatial data description so powerful and productive that users who specialize in creating heuristic explanations simply cannot keep pace. The technology of description has outpaced our collective ability to explain. Consequently, those who are creating the vast array of well-documented, descriptive analyses that require explanation are now, for lack of anything better, falling back upon the earlier imprecise forms of explanation: loose literary and/or philosophically based generalizations and explanations. The burden falls on the mathematically precise theorist to keep up, to supply the breadth of the GIS community with explanation of their newly visualized landscapes. I expect the near future will offer greater disparity between our ability to describe and our ability to explain what we have documented.

I have written elsewhere that "[r]egrettably, one of the weaker aspects of ... [spatial analysis] has been linking the formal theory with formal empirical-statistical analysis. The audience for ... theory has been largely limited to other theorists. Empiricists are all too often unfamiliar with the theory; their research agenda has been based largely upon their personal experiences without the benefit of the theory. The burden for this weakness does not fall only upon the empiricist; too often theorists have made little attempt to communicate their findings to empiricists. More important, there have been too few ... analysts who have made the effort to contribute to both theoretical and empirical analysis" (Thrall, 1987).

I don't believe the state of affairs I commented on in 1987 has changed, at least in my own subdiscipline, with the exception that in 1987 mathematical theory had the advantage over most empirical work, especially data description. GIS technology has recently reversed the advantage; data description is much more useful and more accessible today than heuristic theory.

PREDICTION

We nudge toward prediction when we ask "what if' comparative questions of our static explanatory analysis. This is known as comparative statics, or borrowing from the previously discussed terminology, what I refer to as heuristic prediction.

For illustration, we can pose the question, How would land values be expected to change if a new urban activity center were to arise (Thrall, 1987, Chapter 7) in addition to a single urban central core as in the Gainesville-versus-Vail narrative? Comparative static analysis describes and explains the state of affairs of each scenario: land values before the new activity center and land values after the introduction of another activity center. The comparative static analysis is the evaluation of the two different states of spatial affairs. If the new activity center has not been built then the comparative static analysis allows us to anticipate - a heuristic prediction - the trajectory of land values as a result of the introduction of the new commercial node.

What is the difference between this "heuristic prediction" and a numberic prediction? Much of mathematically based geographic theory can be classified as being of the heuristic type; that is, the theory tells a general story and can be related to the specific circumstance only by analogy. The value of heuristic theories is in their ability to provide anticipation of the trajectory of change and not the expected numeric magnitude of the change.

Consider the following example of heuristic prediction, which should be familiar to anyone with a modest knowledge of economics. Again, this is an example of the economists' (heuristic) theories of demand and supply. Say there is a hard freeze in Florida that kills many orange trees. Given: the supply of oranges declines. Given: the schedule of how much orange juice people are willing to buy at a specified price - namely the demand schedule - does not change. Heuristic: the supply curve moves upward and to the left, shifting along the demand schedule from S to S'. Conclusion: the price of orange juice is anticipated (predicted to rise. (See Figure 2). The heuristic prediction is the trajectory "prices rise." The heuristic explanation for the rise in the price of orange juice is the hard freeze that reduced the supply or orange juice.

Knowing that land values three miles from the new urban activity center will increase by 15 percent following the introduction of the activity center is more valuable than knowing only that land values will increase. Why then bother with the keuristic prediction when the numeric prediction can be much more valuable to our judgment? Why not "just give the facts?"

One answer is that the heuristic prediction

can be referred to as a "first prediction estimate." It is a cheap forecast: no computer time is required, no expensive data are necessary, no software has to be purchased. The heuristic prediction depends entirely upon investment in human capital: education and training, and time and imagination focused on creating abstract heufistic methodology. The heuristic prediction can be the "first out the door," and moreover it is often the only forecast estimate we can get. A trajectory is better than no information at all.

The heuristic estimate also plays a more important role. It provides anticipation and explanation for the outcome, places boundaries on the problem at hand, helps establish the procedure for the desired numeric prediction, provides the first line of defense for evaluating the credibility of the numeric prediction, and improves our general intuition - our reasoning - so that we can better anticipate in our environment what-next and what-where.

The precise expression of concepts relying on mathematical constructs is not without critics. Almost a decade ago I wrote inresponse to criticism from two Cambridge University geographers (Driver and Philo,1985) that ". . . some extremes in mathematical content [have been] taken by some in our discipline, where the spirit of rigor in mathematics has led much pure theoretical research to be sterile in substantive content; ... it is not sufficient to evaluate research by using the criteria that (a) no mathematical errors are made, (b) only the most minor of assumptions are required, and (c) great attention is given to logical consistency, elegance, and rigor. Such criteria on their own may allow the geographer to theorize complacently about imaginary worlds ... implications about which are neither intended for nor capable of empirical verification ... [such work is] substantively vacuous." (Thrall, 1985).

British geographer Openshaw (1989) takes these comments further: "[T]he deductive route to theory formulation has been taken too far and has not been particularly successful. Far too many so-called 'theories' have never been tested and many more are untestable! [Tlhey are 'massively theoretical' rather than applicable, and of dubious relevance." (as quoted in Johnston,1993)."

A second brand of predictive models falls into the category of stochastic, where interrelationships are expressed and measured using probability-based measures. These geographic models are generally less accepted for their ability to explain and more accepted for their ability to give rise to reasonable and acceptable forecasts. Examples of this class of models - again from the human or social science side of the discipline - would include:
* geographic market evaluation known as gravity and spatial interaction models (Haynes and Fotheringham, 1984), and
* Spatial diffusion (Monte Carlo) models that forecast the spread of phenomena from limited origins to other locations through time (Hagerstrand, 1953; Brown, 198 1; Morrill et al. 1988; Thrall, Sidman, Elshaw-Thrall and Fik, 1993).

Notwithstanding the successful use in forecasting with geography's gravity and diffusion models, Casetti (1993) writes that "[t]he social sciences [overall] are collectively under attack ... because of their inadequate performance. These [attacks] focus upon their creating models and theories that do not come to grips with reality, upon their inability to produce reliable predictions and forecasts..." Openshaw (1989; as quoted in Johnston, 1993) writes, "No one has seemingly ever asked who are the potential users for geographical models. Previously it never mattered; today it does"' Moreover, Openshaw responds to the new large market for good geographic predictive models by offering analyses that may lack ". . . any strong theoretical justification" but are immensely successful. The cost of quickly achieving good predictive models in response to market demand ". . . might well be a reduction in the emphasis on explanation and the tacit acceptance of a different and inferior form of understanding ... [to] give models in human geography a greater degree of relevance and marketability."

The mixture of academia and the market bothers some geography academics. Geographers Driver and Philo write in response to some of my musings in 1985 that ". . . computer-aided analysis of complex data sets ... technical wizardry cannot itself supply us with explanations." I agree with this statement. They continue ". . . [T]he widespread concern about government cutbacks in education budgets or current directions in research funding ... reinforce the increasingly technocratic nature of geographic research [T]echnocratic geography effectively marginalizes all other modes of interpretation and explanation. It encourages geographers to seek out the very latest in digital mapping, data processing, or remote sensing facilities . . ."

From Driver and Philo's point of view this is very bad. In conclusion they write that "[e]ven if such techniques enhance the 'image' of our subject, as Thrall contends, they do little to make it more 'relevant' in either academic or social terms." I agree. It is not the GIS technology sitting on the shelf that creates value. Rather, it is what practitioners ultimately do with the technology that creates value. My response was that "I applaud if the technology ... can enhance our ability to obtain resources . . ."

Ron Johnston, one of the great British pundits of geography philosophy, subsequently categorized my stand as being among the "radical approaches" to geography (Johnston, 1991). 1 believe this underscores a problem with academia today, as well as a decade ago when this exchange occurred. It was considered radical by the academic geography discipline to emphasize that in addition to being rigorous, geography should be relevant - that it should be applied. It was considered radical to note (and to take advantage of the fact) that relevant high-tech geography can be both beneficial and profitable. I must admit that being classified by Ron Johnston as a "radical geographer" for these reasons is an appellation that I am honored to have received.

The terms relevance and marketability are key to understanding market-driven GIS today, particularly GIS-based spatial forecasts that transform geographic data by way of mathematical models. In the information age, mathematical transformation of data adds value. Openshaw writes that "[tlhe fundamental technical change [in today's society] ... is the transformation of knowledge into information which can be exchanged, owned, manipulated, and traded. This commodification of information in the form of computer data bases both has economic implications and creates all manner of new opportunities for models as a means of adding value to data" (1989; as quoted in Johnston, 1993).

The person using the geographic data (forecasts) perceives the data as having value. The value to the data user would be expected to be proportional to cost savings

or revenue gained as a result of improved decision making based on the data. In other words, the "so what" question is answered by how the descriptions, explanations, or predictions are used to improve one's judgment.

JUDGMENT

The integration of judgment into GIS analysis has only recently begun to receive the attention of scholars, producing university-oriented GIS literature; therefore, the technology of this component of GIS is not well developed. I have referred to this part of the technology as computer-assisted decision strategy (Thrall and Elshaw-Thrall, 1990), implying that GIS is one part of a larger information technology that may be drawn upon to improve our judgment. Other names have been used such as spatial decision support systems (Ralston, Tharaken and Liu, 1994), which implies that GIS is the main information input to the judgment process. Regardless of the terminology, using GIS to improvejudgment is a crucial stage in reasoning with GIS because it links GIS to the market economy. This stage ultimately gives GIS its value based on the willingness of GIS consumers to pay.

In the decision-making process, rarely is one individual capable of evaluating the importance of the wide array of all considerations that must be part of a final judgment. Groups or teams of individuals are allowed by GIS technology to have input into the process of judgment. I described in Geo Info Systems (Thrall and McCartney, 1991, see also McCartney and Thrall, 1991) an application of a Delphi-like procedure that - in application - has successfully built consensus among the various constituencies of the decision-making process; at least consensus in the inclusion of layers and the weighting of those layers in the GIS analysis.

As GIS technology takes advantage of artificial intelligence, judgment will be improved; however, I do not foresee technology developed for this stage of GIS replacing decision makers any time soon. That is why I have used the word "assisted" in the phrase "computer-assisted decision strategy."

IMPLEMENTATION-MANAGEMENT

Management and implementation take two forms:(a) the management and creation of the source geographic data as input into the GIS process; and (b) the management of geographic-based information that is output from the other four GIS stages.

I have dealt with item (a) in two "ShopTalk" columns (Thrall, Bates, and Ruiz, 1994; Thrall and Ruiz, 1994). These columns documented the problems inherent in earlier efforts to implement local government GIS, particularly countywide geographic data bases. Those local governments that adopted GIS early had problems because the early technology was often being developed as it was implemented. Later-adopting local governments benefited by the industry having gone through the process of leaming how to implement a countywide GIS and having already developed the technology required for that implementation.

Today, the market for new adopters has shifted to the private marketplace, to the business firm. Businesses want off-the shelf, turn-key, ready-to-use technology. GIS vendors are responding to this demand by offering more user-friendly software and by offering easy-to-access, GIS-ready data. Access to affordable data and quality software is a giant leap forward; however, many businesses that take the leap and adopt GIS today will be disappointed, as were early adopting local governments. The technology is not in place to bridge the gap between making a GIS map on the one hand and making a good business decision on the other. For illustration, the purchase of a popular spreadsheet software program does not make one an expert in finance; the spreadsheet is a vehicle for making financial analysis easier for those who already know finance. Likewise, GIS software today is designed to make geographic analysis easier for those who already are familiar with spatial reasoning.

As the business community adopts GIS, it will go through a process of self discovery and find that its management is not trained in basic geographic concepts and not ready to use geography to improve its decisions. The business market will respond in two ways.

First, the demand for trained geographers (spatial analysts) will increase. Some more innovative business schools, such as De Paul University in Chicago, Illinois, USA, are taking steps to include GIS in their curricula to prepare their students to take advantage of this market. Other business schools like American University in Washington, DC, view GIS as unimportant and choose not to prepare their students to take advantage of geographic information.

Second, GIS software vendors will respond by offering market-niche software targeting specific needs for specific classes of businesses. This will represent a significant growth segment for the GIS industry during the next few years.

CONCLUSION

I have offered an overview of five stages of reasoning between which GIS can be divided. Although they are presented in a hierarchical manner, I do not believe that one stage is inherently more important or better than another. Furthermore, most applications involve a blending from one stage to another, along with a shifting back and forth between earlier and later stages.

A rationale for decomposing GIS into categories is the ability to view where past personal efforts have been concentrated, the ability to place those efforts within the context of the larger picture of GIS, and the ability to then identify where the greatest future benefits will originate if the appropriate resources are allocated.

NOTE

Portions of this paper were presented at the 1994 annual meeting of the Association of American Geographers in San Francisco, USA, and at the 1993 annual meeting of the American Real Estate Society in Key West, Florida. This paper was made possible by funding from the Herbert U. Nelson Memorial Fund, National Association of Realtors. I am very grateful for having received the award and the funding.

REFERENCES

Alonso, W. 1964. Location and Land Use: Toward a General Theory of Land Rent, Cambridge, Mass.: Harvard University Press.

Beffy, B.J.L. 1964. "Approaches to Regional Analysis: A Synthesis." Annals of the Association of American Geographers (54):2-1 1.

Brown, L. 198 1. Innovation Diffusion: A New Perspective. London: Methuen.

Casetti, E. 1993. "Spatial Analysis: Perspectives and Prospects." Urban Geography 14(6):526-537.

Christaller, W. 1966. Central Places in Southern Germany. Translated by C. W. Baskin, Englewood Cliffs, New Jersey: Prentice-Hall.

Clark, W.A.V 1984. Human Migration, (7), Scientific Geography Series, Newbury Park and Beverly Hills, Califomia: Sage Publications.

Driver, F., and C. Philo. 1985. "The Implications of 'Scientific' Geography." Area. 17(9):161-162. Fielding, G.J. 1974. Geography as Social Science. Harper & Row Publishers: New York.

Golledge, R.G. 1978. "Leaming About Urban Environments." T. Carlstein, D. Parkes, and N.Thrift (Eds.), Timing Space and Spacing Time, 1; Mciking Sense of Time. London: Edward Arnold,76-98.

Gould, P., and R. White. 1986. Mental Maps, 2d. ed. Allen & Unwin: Boston.

Hagerstrand, T. Innovation Diffusion as a Spatial Process. (A. Pred, Trans.) Chicago: University of Chicago Press. (Original work published 1953). Hartshorne, Richard. 1939. "The Nature of Geography: A Critical Survey of Current Thought in the Light of the Past." Annals of the Association of American Geographers 2:173-658.

Haynes, K.E., and A.S. Fotheringham. 1984. Gravi and Spatial Interaction Models, 2, Scientific Geography Series, Newbury Park and Beverly Hills, California: Sage Publications.

Hoyt, H. 1933. One Hundred Years of Chicago Land Values. Chicago: University of Chicago Press.

Hoyt, H. 1939. The Structure and Growth of Residential Neighborhoods in American Cities. Washington, D.C.: Federal Housing Administration.

Johnston, R.J. 1993. "Meet the Challenge: Make the Change." RJ. Johnston, ed. The Challenge For Geography: A Changing World, A Changing Discipline, Oxford: Blackwell Publishers, 151-180.

Johnston, R.J. 199 1. Geography and Geographers,4th ed. London: Edward Arnold.

King, L.J. 1984. Central Place Theory, I Scientific Geography Series, Newbury Park and Beverly Hills, California: Sage Publications.

Lukermann, R 1989. "The Nature of Geography: Post Hoc, Ergo Propter Hoc?" In Reflections on Richard Hartshome's "The Nature of Geography," J. N. Entrikin and S.D. Brunn, Eds. Occasional Publications of the Association of American Geographers: 61.

Martin, G.J. 1994. "In Memoriam: Richard Hartshorne, 1899-1992." Annals of the Association ofamerican Geographers 84:480-492.

McCartney, J.W, and G. 1. Thrall. 199 1. "Real Estate Acquisition Decisions With GIS: Ranking Property For Purchase." Proceedings of the Annual GISILIS Conference, October, Atlanta, Georgia.

Morrill, R., G. Gaile, and G.I. Thrall. 1988. Spatial Diffusion, 10: Scientific Geography Series. Newbury Park and Beverly Hills, California: Sage Publications.

Openshaw, S. 1989. "Computer Modeling in Human Geography," B. Macmillan (Ed.),Remodeling Geography. Oxford: Basil Blackwell, lewood 70-88.

Ralston, B., G. Tharaken, and C. Liu. 1994. "A Spatial Decision Support System for Transportation Policy Analysis in Bangladesh," Journal of Transport Geography 2: 101-1 10.

Thrall, G.I. 1993a. "Basic GIS Software Can Help Make Urban Policy." Geo Inft) Systems 3(i):5".

- 1993b. "Using a GIS to Rate the Quality of Property Tax Appraisal." Geo Info Systems 3(3):56-62.

- 1991. "The Production Theory of Land Rent.' Environment and Planning A. 23:955-967.

- 1987. Land Use and Urban Form. London: Methuen.

- 1985. "Reply to E Driver and C. Philo, A Rebuttal to Their Implications of Scientific Geography: A Comment on Thrall's September 1985 Commentary." Area 17(9):162-163.

- 1979. "A Geographic Criterion for Identifying Property Tax Assessment Inequity." Professional Geographer 31, 278-83.

Thrall, G.I., and A.R Marks. 1992. "Siting Hospitals to Provide Cost-Effective Health Care.' Geo Info Systems 2(8): 58-66.

Thrall, G.I., and JW.. McCartney. 1991. "Keeping the Garbage Out: Using the Delphi Method for GIS Critefia." Geo Info Systems l(l):46-52.

Thrall, G.I., and M. Ruiz. 1994. "A History of Implementing an Urban GIS, Part One: Design, Tribulations, and Failure." Geo Info Systems 4(7):50-58.

Thrall, G.I., B. Bates, and M. Ruiz. 1994. "A History of Implementing an Urban GIS, Part Two: Two Solutions Toward a Working GIS." Geo Info Systems, 4(10):46-51.

Thrall, G.I., C. Sidman, S. Elshaw-Thrall, and T. Fik. 1993. "The Cascade GIS Diffusion Model for Measuring Housing Absorption by Small Area with a Case Study of St. Lucie County, Florida." Journal of Real Estate Research 8(3):401-420.

Thrall, G.I., and S. Elshaw-Thrall. 1990. "A Computer-Assisted Decision Strategy For Evaluating New Satellite Hub Sites For a Local Utility Provider." Computers, Environment, and Urban Systems 14:37-48.

Thrall, G.I., and S. Eishaw-Thrall 1993. "Commercial Data for the Business GIS." Geo Info Systems 3 (7)63-68.

Thrall, G.I., and S. Elshaw-Thrall. 1991. "Reducing Investor Risk: A GIS Design For Real Estate Analysis." Geo Info Systems l(I 0):40-46.

Thrall, G.I., S. Elshaw-Thrall, M. Ruiz, and C. Sidman. 1993. "Using GIS to Visualize Spatial Data (TIN, Voronoi Polygons, Surface Models)." Geo Info Systems 3(5):59-65.

Tobler, W. 1970. "A Computer Movie Simulating Urban Growth in the Detroit Region." Economic Geography 46(2):234-240.

von Thunen, J.H. 1842. Der Isolierte Staat. P. Hall, Ed., von Thunen's Isolated State. London: Pergamon.