Cultural and Linguistic Variation in (Geo)Spatial Conceptualization David M. Mark University at Buffalo

Categories "Of all the countless possible ways of dividing entities of the world into categories, why do members of a culture use some groupings and not use others? What is it about the nature of the human mind and the way that it interacts with the nature of the world that gives rise to the categories that are used? (Malt, 1995, p 85)

(Geo)Spatial Categories The question applies both to categories for spatial relations and to categories for spatial objects It also applies to the delimitation of object-like features from continuous geospatial fields

Where are the categories? Categories in the world (Cutting nature at its joints) or Categories from the mind: "When I use a word, … it means just what I choose it to meanneither more nor less." (Humpty Dumpty, in Lewis Carrolls Through the Looking Glass)

Categories: concepts or sets? Malt points out that psychologists usually define categories in terms of "concepts", "the mental representation that underlies the observed categories", but anthropologists tend to see categories as "sets of objects that are treated as equivalent and given a common name." (Malt, 1995, p. 134)

Biological Taxonomic Categories Biological classes may form a source domain for category systems for other domains For example, sea food can be classified into several phyla, classes, and families:

Trattoria Corte Sconta

Class Chordata

Class Molusca

Class Arthropoda å

Semantics What do you mean by semantics? The meaning of meaning The meaning (referent) of words? –or The meaning of things?

ReferentSymbol Concept The Semiotic Triangle

Geographic Categories: Dividing a Continuum? How are spatial relations, that form a (metric) continuum, grouped together into qualitative relations and considered to be equivalent? How are sizes and shapes, that form a (metric) continuum, grouped together into shape classes and considered to be equivalent? How are feature instances such as hills and valleys delimited from continuous elevation fields?

Schematization Len Talmys How Language Structures Space described how spatial situations are schematized (modeled?) in various ways before spatial terms are chosen The same continuous subset of reality can be schematized in different ways, leading to different conceptualizations, and different equivalences

(The scope of the remainder of this talk is narrowed by my own interest) Natural phenomena Natural language General speakers of the language I do not fully trust the results of introspection No immediate goal of engineering or building systems

Remainder of this Presentation History of the cognitive thread in geographic information science Spatial relations in language Ontology Geographic entity types Concluding remarks

History of the Topic, Part 1 In 1987, the U.S. National Science Foundation requested proposals from Universities that wanted to operate a National Center for Geographic Information and Analysis (NCGIA) UC Santa Barbara, the University at Buffalo, and the University of Maine submitted the winning proposal

The NCGIA solicitation The NCGIA described a research agenda using five bullet points Bullet #2: –General theory of spatial relations and database structures

NCGIA Proposal Team at work in Santa Barbara, December 1987

NCGIA Initiative #2 Our proposal outlined 12 Research Initiatives NCGIA Research Initiative 2 was entitled Languages of Spatial Relations and had Andrew Frank and David Mark as co-Leaders

Evolution of the topic: The topic evolved through several stages: –General theory of spatial relations –Languages of spatial relations –Cognitive and linguistic aspects of geographic space

NATO ASI and Book To close Initiative 2, Mark and Frank obtained a grant from NATO to conduct an Advanced Study Institute Two weeks in a castle near Avila, Spain, July 8-20, 1990 Produced an edited book Established foundations for the cognitive stream within GIScience research

COSIT (Conference on Spatial Information Theory) In September 2002, Andrew Frank and colleagues organized an international symposium entitled GIS From Space to Territory: Theories and Methods of Spatio-Temporal Reasoning in Pisa, Italy Papers were published in Lecture Notes in Computer Science

COSIT (Conference on Spatial Information Theory) In September 2003, Andrew and colleagues organized another international symposium entitledSpatial Information Theory: A Theoretical Basis for GIS on Elba, Italy Papers again published in Lecture Notes in Computer Science COSIT became a conference series, meeting every second year

COSIT93 included: D. M. Mark: Toward a Theoretical Framework for Geographic Entity Types. Spatial information theory: A theoretical basis for GIS (COSIT93), edited by A. U. Frank and I. Campari. (Topic not followed up until 2002!)

COSIT95 included: D. R. Montello: How significant are cultural differences in spatial cognition? M. J. Egenhofer and D. M. Mark, Naïve Geography D. M. Mark and others, Evaluating and Refining Computational Models of Spatial Relations Through Cross-Linguistic Human-Subjects Testing –Spatial information theory: A theoretical basis for GIS (COSIT95), edited by A. U. Frank and W. Kuhn.

"Looking for (cultural differences in spatial cognition) in all the wrong places…" In the early 1990s, Mark, Frank, and others speculated that GIS might be biased toward an Anglo-Germanic conceptual system and world view

"Looking for (cultural differences in spatial cognition) in all the wrong places…" Mark & Egenhofer conducted research to calibrate the relationships between mathematical models of spatial relations and natural language expressions The road and the park Tested English, Spanish, French, Norwegian, Mandarin

Defining Spatial Relations Following mathematical developments by Max Egenhofer and John Herring, we used the 9-intersection formalism to define what constitute different or same topological spatial relations Note that real instances (such as roads and parks) must be schematized in a certain way before the formalism can be applied

In the 9-Intersection Model... Each spatial object has an interior, a boundary, and an exterior These three parts of one object are tested for intersections with the three parts of the other object The empty/non-empty status of these 9 intersections defines the spatial relation

Numbers of Different Spatial Relations The 9-intersection model defines –8 distinct spatial relations between two regions –33 distinct spatial relations between two unbranched lines –19 spatial relations between an unbranched line and a region

8 relations between two regions

The 19 line-region relations

Mathematics is Not Enough Mathematically, each of these 19 relations is equally distinct from each of the others Each is unique The mathematical model does not help determine whether some relations are more important or salient than others

Mathematics, Cognition However, data from human subjects suggests that some differences are more salient than others Some adjacent pairs are kept in distinct linguistic spatial relations, while others, equally distinct mathematically, are often grouped together

Testing With Human Subjects Grouping Task –40 diagrams, 28 subjects, 3 languages Drawing Task –32 English subjects drew for 64 sentences –19 Spanish subjects drew for 43 sentences Agreement Task –64 diagrams, 11 sentences, 3 languages, 600 subjects, 36,000 judgments

Testing With Human Subjects Across languages, similarities were more striking that differences Mark & Egenhofer published 9 papers on this topic and did not report significant cross-language differences

Agreement: Cross and Cruzar Each dot represents one road-park diagram; 135 to 144 subjects per language

Ontology Around 1997, DMM started working with Philosopher Barry Smith Smiths approach focuses on realism, and reduces emphasis on language and cognition EURESCO meeting at La Londe (France) in 2000, organized by Stephan Winter and Andrew Frank, led to IJGIS special issue on ontology edited by Stephan Winter Andrew Turk attended that meeting

Geographic Categories Geographic categories are often standardized in gazetteers and SDI –Feature codes in the US Geographic Names Information System –Entity types in the US Spatial Data Transfer Standard –Etc.

Here are some English-language descriptions of some landscape feature types that are referred to by a single word in some other language, but have no single word in English: An area from which you cannot see the sea A landmass containing an area from which you cannot see the sea An area of agricultural land reclaimed from a water body or wetland An island of land completely surrounded by one or more younger lava flows An island of grassland left unburnt after a surrounding wildfire …

Example: Hawaiian has a word Kipuka Kipuka: A Hawaiian word for an island of land completely surrounded by one or more younger lava flows

Icelandic has a Word for an island in a lava flow too! –Óbrinnishólmi –Literally, un-burnt-hill The Walmajarri (in Australia) language has a word with a similar meaning: –Nyirirr: an island of grass left unburnt by surrounding fire So, a concept of a patch of habitat not destroyed, when general destruction sweeps through an area, might be a general template that we find in several languages but not in English, German, French, etc…

If we based our ideas about geographic concepts only on the meanings of words in English (plus French, German, Spanish), we would miss out on a lot of conceptual variation! Yet current Geographic Information Systems and Spatial Data Infrastructures are based mostly on English and other dominant European languages…

French and English Categorize Standing Water Bodies Differently

English

French

Hills and Mountains in English Mountain: 1. a. A large natural elevation of the earth's surface, esp. one high and steep in form (larger and higher than a hill) (OED)

Hills and Mountains Hill: 1. a. A natural elevation of the earth's surface … after the introduction of the word mountain [into English], gradually restricted to heights of less elevation; … (OED)

Hills and Mountains: Not Only Size! Hill: a more rounded and less rugged outline is also usually connoted by the name (OED) HillMountain ??

#76 4 B Shiprock area buttes/ monoliths P Finger Rock Mitten Buttes Shiprock Picacho Peak For features too small to be mountains, yet too jagged to be hills, English relies on other terms, such as rock, butte, peak, mesa, etc.; more on this later…

Digging Deeper: Ethnographic Fieldwork on Landscape Categories Mark & Turks work with Yindjibarndi Mark, Stea, and Turks work with Navajo Interpretation

mountain hill marnda bargu burbaa Yindjibarndi English

mountain hill marnda bargu Yindjibarndi English marnda hill Most marnda are hills, but… Some marnda are mountains, and Some hills are bargu Smaller bargu are mounds…

There are no permanent or even seasonal rivers or creeks in Yindjibarndi country Larger watercourses have running water in them only after cyclones Permanent sources of water include permanent pools along the channel of the Fortescue River, as well as some permanent small springs, and soaks (where water can be obtained by digging)

Yindjibarndi ontology and terminology treat the channel and the water separately The channel is always wundu, is always present When water is flowing, the water is manggurdu (flood) or yijirdi (trickle)! wundu yijirdi wundu No Yindjibarndi word is fully equivalent to river in English

Likewise, Yindjibarndi has no word for waterfall

Thardarr The Yindjibarndi word thardarr refers to A place on a cliff where water sometimes or always falls, whether the water is flowing or not. When water is flowing, the water is manggurdu (flood) or yijirdi (trickle)!

Comparison of Yindjibarndi (Rows) and Navajo (Columns) Terms for Longitudinal Depressions in the Landscape

Future Work: Feature Extraction We need robust and flexible language- specific methods for –Extracting features from elevation and land cover data –Determining properties such as size, shape, etc. –Classifying the features Feature delimitation may be contingent on the definitions!

Future Work: Many languages There are about 5,000 languages in the world that still have 1,000 or more speakers There may be around 100 geographical terms per language More than 500,000 terms that need to be defined and implemented!

Deeper Issues: What and Where What and Where may be processed in different areas of the brain Landau, B., and Jackendoff, R., "What" and "Where" in Spatial Language and Spatial Cognition. Behavioral and Brain Sciences, 16,

Closing Remarks: Spatial Relations (more nearly) Universal that Entity Types ? Maybe spatial relations (in language) seem relatively 'universal', with few or weak cross- linguistic differences, because spatial relations are the same" everywhere, so Homo sapiens may have evolved hard-wired procedures to handle them in the brain and mind... (But what about Melissa Bowermans work on Korean?)

Spatial Entity Types On the other hand, perhaps spatial object categories, especially geospatial entity types, are NOT the same everywhere, so the cognitive mechanisms for classifying objects may be far more 'plastic (Or perhaps there just are fewer constraints on entity types…)

The More Things Change… Perhaps the categorization of spatial relations and of geographic features have important similarities after all…

The More Things Change… In each case, there seem to be low level components (primitives?) that, if not universally used, can be recognized as common across languages Then, basic level or commonly used terms (such as in, on, at; mountain, river, lake) refer to groups or combinations of these components Different languages group the components in different ways

Summary & Conclusions Geographic entity types (landscape features) show cross-language variation in conceptualization that will have to be documented using ethnographic methods Spatial relations also must be tested in a broad range of genetically-unrelated languages

Summary & Conclusions Research on geographic categories may reveal aspects of linguistic relativism of categorization that are sometimes difficult to document for other domains There are clear implications for geographic information retrieval, and perhaps for SDIs and GIS

Acknowledgments I appreciate the support for my research that has come from the U.S. National Science Foundation over the last two decades The research reported here would not have been the same without the contributions of many many people, including David Simonett, Mike Goodchild, Max Egenhofer, David Zubin, Len Talmy, Werner Kuhn, Barry Smith, Andrew Turk, David Stea, and especially Andrew Frank !