How can a reading of a textual description of a landscape be expressed as a map? Maps form a medium different from verbal texts, and the differences have consequences not only for how things are said, but also for what can be said at all using maps. Where are these limitations to be found?
In this abstract, I will discuss the relationship between verbal and map based geographical communication. I have created a model of the geographical information read from a source text, then tried to express the contents of the model as maps. I will show that types of geographical information exist that can be stored in and read from verbal texts, but which are impossible to express as geographical maps without significant loss of meaning.
I used a set of Scandinavian border protocols from the eighteenth century (Schnitler 1962) as source material for this research. The text is based on interrogations about geography with more than 100 different persons, of whom many presumably did not use maps very much if at all. It was created in a society, or a set of societies, on the brink of the transformation from oral to written cultures, where some were firmly placed within the written culture, while others had only been exposed to the activity of reading texts for a few decades. The voices in the text represent persons coming from different ethnic and professional backgrounds, e.g., Sami reindeer herders, Norwegian farmers, and Danish military officers, thus bringing a set of different perspectives into the geographical conversation.
Modelling is to create and manipulate models in order to learn from the experiences we gain through the process. Models are in this context representations of something which is created for the purpose of studying what is modelled more closely (McCarty 2005: 24). In the work reported on here, a computer based model was constructed based on the source text. Through experiments on this model, new knowledge about the text was gained. The creation of the model itself also led to problems giving new insights, so it was part of the series of experiments. The same can be said of the development and tailoring of the software used.
The whole experiment was performed in a qualitative manner. Occurrences in the text were not counted and compared, they were rather evaluated as individual cases, in accordance with traditional humanities research where seeking local knowledge about specific cases is the main activity (Galey 2010: 95-100).
The point of a modelling exercise lies in the process, not in the model as a product. Modelling mediates epistemologically between modeller and the modelled, that is, between the researcher and the data. The model was built through a combination between computer based tools and a thorough close reading in which all geographic information I was able to extract from the text was included. The model was stored as a set of connected facts in a computer application I developed in Java for this research.
In the application the results of the close readings are stored as triplets in a fact database. Then the ‘raw’ triplets go through several steps of interpretation, leading to more and more formal structures. These steps eventually result in the standard format RDF1, then further interpretation leads to geographical vector data in a simplified version of the GML format2 which can be imported to GIS software to be viewed as maps.
The modelling process thus involves several steps, each formalising the data in a stricter sense. For each step, what ‘falls off,’ what is difficult to avoid loosing, is the interesting part. This ‘falling off’ will include things that cannot survive a transfer from one system to another, which misses the added level of formality. Through this process I attempted to translate geographical information from one medium to the other. The ‘fall-offs’ show us what was lost in the process; this was not always because of differences between the media, but candidates for further examination were found in the ‘fall-offs’. Examples from this modelling work will be shown in the paper, including some situations where ‘fall-off’ occurs.
As the goal is to understand how texts express geographical information, I made an effort to base the maps on what I read from the text only. While it is impossible to exclude all context information, I tried hard to exclude any local context based on geographical knowledge of the area. This makes the process different from mapping place name information in literary texts, as done by e.g. Smith and Crane (2001). The latter process includes adding a significant amount of contextual knowledge to the process through what is learned from the pre-existing map.
The issues faced in the modelling experiments led to the development of the following typology of textual expressions:
1. Under-specified texts. Based on such a text, more than one map can be drawn, and at least two of these maps are significantly different.
This situation occurs when the geographical information in a text can be expressed as more than one significantly different map. This happens when directions such as ‘east’ or measurements such as ‘1-2 miles’ are used: given lack of other evidence in the text, a number of different geometrical interpretations of the statements are possible.
2. Fully specified textual descriptions. Only one map can be drawn based on the description. If the text mentions something, it is fully specified geometrically.
Texts in formal languages, such as GML, are typical cases. I have not found examples of this type in natural language texts.
3. Ambiguity and negation. The text includes expressions in the forms ‘A or B is located at C’ (ambiguity) or ‘There are no A’s in B’ (negation).
The spatial information read from the text cannot be represented as one single map; one will need two or more maps, a dynamic map or a map with its geometry corrected by a textual description.
The choice of the words ‘significantly different’ as opposed to merely ‘different’ in the description of type 1 is necessary and highlights how the border to type 2 is fuzzy. Two maps can always be made which are slightly different, but which will occur as similar to the reader. Small adjustments in location are routinely made in cartographical work in order to improve the readability of maps. The choice of symbols can also change without its leading to significantly different maps. Any text, also of type 2, could be represented by two slightly different maps.
In the case of Schnitler, type 1 situations are abundant, we see a handful of type 3, and none of type 2. The groups are partly exclusive: 1 and 3 can exist in the same textual description, but is hard to see how a textual description can be both 2 and 1 or both 2 and 3.
The discussion of the relationship between geographical maps and verbal texts fall into the long tradition of inter-art and intermedia studies. Two important oppositions established by Lessing (1893) can be summarised this way:
1. Actions in time should be applied in poetry, and bodies in space in painting.
2. What is hidden is not seen in painting, while things hidden can still be seen in poetry.
Whereas Lessing’s dichotomy between poetry and painting has been questioned, e.g., by Frank (1963) and Mitchell (1980) in connection to their discussions of spatial form in literature, it has never been eliminated. A recent way of formalising such distinctions is the model of Elleström (2010). Instead of starting from a set of different media or art forms, he takes a bottom-up approach, starting from a set of media modalities. His set includes four, namely, the material, sensorial, spatiotemporal and semiotic modalities. The differences between texts and maps fall mainly in the latter two categories.
The two problem areas found in the modelling of Schnitler, under-specified texts and ambiguity and negation, are both connected to Lessing’s thinking. When ‘east’ in a text can mean a number of different things, potentially covering half of the possible directions from a place, the direction between two features on a map is measurable, with only a limited inaccuracy. Space is expressed differently in the two media. And when disjunction is hard to put on a map, it is also connected to the fact that the sign systems used to refer to the external reality are different. This will be discussed in light of Lessing’s opposition and Elleström’s system in the paper. Some comments will also be made as to how this links to narratology.
Elleström, L. (2010). The Modalities of Media: A Model for Understanding Intermedial Relations. In L. Elleström (ed.), Media borders, multimodality and intermediality. Basingstoke: Palgrave McMillan, pp. 11-48.
Frank, J. (1963). Spatial Form in Modern Literature. In The widening gyre: crisis and mastery in modern literature. (First published: 1945). New Brunswick, N.J.: Rutgers UP, pp. 3-62
Galey, A. (2010). The Human Presence in Digital Artefacts. In W. McCarty (ed.), Text and Genre in Reconstruction: Effects of Digitalization on Ideas, Behaviours, Products and Institutions. Open Book Publishers, pp. 93-117.
Lessing, G. E. (1893). Laokoon: oder über die Grenzen der Mahlerey und Poesie. Erster Theil. In K. Lachmann and F. Muncker (eds.), Gotthold Ephraim Lessings sämtliche Schriften. Neunter Band (3., aufs neue durchges. und verm. Aufl.). Stuttgart: Göschen. Photographic reprint. Berlin : de Gruyter 1968, pp. 1-177. First published: 1766.
McCarty, W. (2005). Humanities computing. Basingstoke: Palgrave Macmillan.
Mitchell, W. J. T. (1980). Spatial Form in Literature: Toward a General Theory. Critical Inquiry 6 (3): 539-567.
Schnitler, P. (1962). Major Peter Schnitlers grenseeksaminasjonsprotokoller 1742 – 1745, Volume 1. Oslo: Norsk historisk kjeldeskrift-institutt.
Smith, D. A., and G. Crane (2001). Disambiguating Geographic Names in a Historical Digital Library. In Research and Advanced Technology for Digital Libraries: 5th European conference, ECDL 2001. Darmstadt: Springer, pp. 127-136.