Describing spatial scenes in Russian: universal and language specific properties

The paper discusses the main results of a cognitive linguistic study, aimed at modeling the process of transforming visual information about space into a NL text. As far as the language of the studied text was Russian we can look at these results in a typological perspective and put forward some plausible hypotheses about universal properties of this process and those that may be characteristic of the way speakers of Russian conceptualize space.

The source of the data for the analysis is the corpus, gathered by A. R. Armeeva [Armeeva 1994]. It consists of 15O expository space-describing texts, produced by 25 informants, that were asked to describe 6 pictures each (2 interiors, 2 landscapes and 2 still lives) and do it in such a way, that a person who does not see the picture could figure out spatial relations between objects.

Analyzing these texts we derived a discourse grammar, which underlies them. This grammar embodies an ordinary mundane conception of space as opposed to its scientific conception. According to this conception space is a configuration of objects connected with spatial relations. Such a conception of space is widespread among languages and cultures, but it may be not universal (here I can point to data of L. Talmy about an American Indian language where the world, and specifically some spatial scenes, are conceptualized not so much in terms of objects, as in actional terms).

If the space is conceptualized as a configuration of objects, the goal of a speaker, describing a scene is to successively introduce objects and connect them with spatial relations in such a way that the hearer could form in his mind a 3-dimensional representation of the corresponding scene as similar to the original scene as possible. The structure of space-describing texts reflects the sequence of cognitive operations performed by the Speaker in order to achieve this goal. It can be represented in a form of a grammar in (1):

(1) The grammar underlying space-describing texts:

Space description => Exposition + Elaboration
Exposition => Categorization + Specification + (Characterization) + (Localization) + (Commentary)
Elaboration => Sequencei*
Sequence => Shot i*
Shot => Configuration + (Specification) + (Characterization) + (Commentary)
Configuration => Topos + Copula + Figure
Topos => Relator + Landmark
(An asterisk* means iteration)

Each textual category is represented at the sentential level by a limited number of syntactic or lexico-syntactic patterns. E.g. a category "Categorization" is represented by the following sentence types:

(2) a. nominative [ e.g. Komnata. /room/ "A room."]
b. classificational [e.g. Eto muzej. /this (neut.sg.nom) museum / "This is a museum."]
c. existential [e.g. Pered nami bol'saja zala /before us big hall/ "There is a big hall in front of us."]
d. perceptional [e.g. Vidim inter'jer doma / see (1pers. pres.pl.) interior house (gen.) / "We see an interior of a house."]

Patterns are language specific, but reflect a presumably universal cognitive operation, performed by the speaker at the very beginning of describing a perceived scene. Although in some texts this operation and a corresponding textual category has a null linguistic expression, it is necessarily there and is indirectly represented in linguistic structure of a text. When Categorization is expressed overtly it activizes a corresponding frame in hearer's mind, e.g. a frame of a room, and this frame helps him/her to figure out some aspects of the scene even when they are not mentioned explicitly in the description.

The next universal cognitive operation reflected in our text is dividing the scene into fragments for more elaborate description of space. These fragments correspond to main textual units of space description called Sequences (Sq). Sqs are chosen in different modes, that we call compositional strategies and each of these modes has its own verbal cues in a text. On the basis of our data 4 strategies can be distinguished: 1) Frame Strategy, when Sq = domain, pertaining to some slot of a corresponding spatial frame (e. g. frame "Room" or "Landscape"); 2) Salient Object Strategy, when Sq = domain, surrounding some salient object; 3) Path Strategy, when Sq = (segment of an) imaginary itinerary of the speaker around the scene; 4) Scanning Strategy, when Sq = domain, surrounding some vector of scanning the scene. These strategies are presumably universal too, but the preferences among them may be culture specific.

The most mysterious cognitive operation, reflected by the textual category Configuration, is the choice of the of Landmark or (Reference Object) among two (or more) objects, whose spatial relations are to be described in this textual unit. In cognitive linguistics (e.g. in [Talmy 1983] and [Jackendoff 1987]) various factors, affecting this choice were discussed. Some of them have to do with objective physical properties of perceived objects (e.g. their size, movability, etc.), others are subjective, that is connected with the speaker (=perceiver) (e.g. the time when s/he first perceived an object). In [Kobozeva 1995] we have shown that in cases of conflict among different factors either of the two kinds of factors mentioned above may be given priority, depending on the overall discourse compositional strategy the speaker initially chooses for converting non-linear visual information into linear verbal form. In [Kobozeva 1996] we discussed objective physical factors and supported the idea that objects are mentally ranked according to some hierarchy (the notion "rank of object" was first introduced for modeling the describing of space in [Ilyin, Ignatova 1992]). Addressing the problem of determining the rank of an arbitrary object in a perceived scene we argued that ranks of objects are not "cognitive primitives", being calculated on the basis of information, stored in several slots of the two kinds of frame structures, associated with this object: frame-prototype that belong to knowledge base and current frame that is created as a result of perception of an observed object - a token of a given object type.

In Russian as in many other languages Configuration is represented linguistically by locative prepositional or adverbial construction. The choice of preposition or adverb expressing spatial relations (Relator) is dependent on a typological type of object - a cognitive semantic category introduced in [Talmy 1983]. The number of typological types and membership in them is language specific. In languages with noun classifiers this category is grammatical, in Russian it is a so called covered category, revealed by
semantic compatibility restrictions of spatial prepositions and words, denoting size (see [Rakhilina 1995]).

A characteristic feature of space conceptualization by Russian speakers is reflected by an extensive use of verbs denoting position (e. g. stand "stand", "lie" and the like) as Copula in Configurations. It means, that the attention is paid not only to spatial relation of a Figure to Landmark, but to Figure's own absolute orientation.

An optional feature in space descriptions, but definitely connected with conceptualization of space, is a characterization of the shape of perceived objects. In our paper we discuss two ways of overtly expressing this property of physical object in Russian: 1) by words whose primary semantic function is to denote shape per se, such as pryamoj "strait"; 2) by words whose primary semantic function is to denote a specific kind of objects, but whose secondary semantic function is to denote specific shape, characteristic of this kind of objects such as kolokol "the bell" or kolbaska "small sausage". For the second kind of words we propose a term "templates of shape". The existence of "templates of shape" in a language reflects the fact that in our minds words denoting objects are connected with image schemes of particular shapes. Presumably this cognitive semantic category is just as universal as topological type, but the membership in it is definitely language specific and specific are the diagnostics for this covered category in each language. We a going to discuss "templates of shape" in Russian.

References

[Armejeva 1994] - Armejeva A.R. Lokativnye predlozhnye konstruktsii kak sredstvo opisanija prostranstva. Diplomnaja rabota. Moskva: MGU, 1994.

[Jackendoff 1987]- Jackendoff R. On Beyond Zebra: The relation of linguistic and visual information // Cognition, 28. 1987.

[Ilyin, Ignatova 1992] - Ilyin G.M., Ignatova V.N. Sistema "RISUNOK <=> TEKST" // Programmnye produkty i sistemy. N2. 1992.

[Kobozeva 1995] - I. M. Kobozeva. Kak my opisyvaem prostranstvo, kotoroe vidim: problema vybora "orientira" // Trudy mezhdunarodnogo seminara "Dialog'95" po komp'juternoj lingvistike i ee prilozhenijam. Kazan', 1995.

[Raxilina 1995] - Raxilina E.V. Semantika razmera.// Semiotika i informatika. Vyp. 34. M., 1995.

[Talmy 1983] - Talmy L. How language structures space // Spatial Orientation: Theory, Research, and Application. Plenum Press, 1983.