Any model of language aspiring to be successful has to deal with figurative language, pervasive in any type of discourse, including our everyday conversations. In earlier models of grammar this creative aspect of language was often ignored or relegated to the black box of semantics or, even more often, pragmatics. In the cognitive framework, as developed by, among others, Langacker (1987,1991) and Lakoff (1987), the flexibity of linguistic coding has been integrated as one of the cornerstones of the model. Linguistic structures are regarded as form/meaning pairs, where the latter is defined as conceptualization "tailored to the specifications of linguistic convention" (Langacker 1987: 99). Conceptual categories are said to be structured around a prototype, a conceptual centre that gives the category some stability yet allows for enormous flexibility as well. Metaphorical and metonymical uses are thus defined as extensions from the prototype.

We will discuss different kinds of such extensions for the Dutch posture verbs "liggen" ('lie'), "zitten" ('sit'), and "staan' ('stand'). We will show that there are different extension patterns at work and, more importantly, we will illustrate how some of these may function as local prototypes, leading to a model that pairs flexibility to economy.

Our analysis is based on a considerable collection of contemporary Dutch non-fictional prose, approximately 2.5 million words, from which some 10.000 occurrences of the posture verbs have been extracted, excluding derivations like "bestaan" ('exist'), "ontstaan" ('originate'), or "bezitten" ('possess'), some of which quite frequent and possibly still have a semantic link with the base verb. Roughly, this means that 3.2% of the total number of lines in the corpus contains one of the three verbs. (Notice that there may be other posture verbs as well, e.g. "hangen" ('hang') or "leunen" ('lean'), but they are still of a somewhat different nature.). All extractions have been considered, but for the time being only the first 500 of each verb have been analysed more elaborately and coded with tags for elementary statistical analysis.

As is intuitively clear, the prototypical meaning of the posture verbs refer to human beings in a sitting, lying or standing position. Strikingly, however, the frequency of these uses is quite low, between 10% and 15%. All other uses are either "locational expressions", i.e. those that situate concrete entities other than humans in space (notice that the prototype is a subcase of these) or "metaphorical expressions", i.e. those no longer characterized relative to the spatial domain. The type of domain is what distinguishes "locational" from "metaphorical" expressions. However, from another point of view, the locational expressions situating entities other than humans could equally well be regarded as 'metaphorical' extensions. This in fact reveals one of the major difficulties of defining a metaphor, even within the Cognitive Grammar framework in which the conceptual domain is often taken as something established or defined. There is, however, considerable difficulty in delineating conceptual domains and the interpretation of an expression as metaphorical often seems post hoc. While relevant in itself, we will not go into this issue here (see, e.g., Feyaerts 1997 for some discussion). We use the locational / metaphorical distinction as a practical device to present our analysis. Moreover, we will not be much concerned with locational expressions here (see Lemmens, ms.), but focus on metaphorical uses instead.

Our data reveals that different types should be distinguished. A first case of metaphors concerns those where the posture verb occurs in a 'normal' collocation but the whole expression becomes metaphorical (idiomatic). For instance, the utterance "ik lig daar niet wakker van", literally, 'I lie not awake of-that', can be interpreted as a normal usage, since it refers to a human in a lying position in bed, awake. However, while the literal interpretation may still apply in some cases, the whole is often used in indiomatic reading, "it does not worry me", a relatively straightforward extension based on our experience that problems prevent us from sleeping well. In principle, the context provides clear clues to signal that the idiomatic reading applies.

More problematic are cases where the use of the posture verb no longer have a weaker experiential basis; these contexts prove often problematic for non-native speakers. A difference is to be made between those metaphorical extensions that have a link with locational expressions and those that do not. Consider, for instance, the fact that in Dutch one can say that one 'sits' in a depression. This use is a direct metaphorical extension from the spatial domain, since "zitten" is used in locational expression situating a referent in close containment, e.g., "ik zit in het water", 'I sit in the water' (no sitting posture is implied here). Among others, Lakoff & Johnson (1980) have shown the pervasive nature of metaphors based on the notion of a container: many abstract entities (time periods, situations, events, etc.) are commonly conceptualized as containers. Logically, then, whenever the CONTAINER metaphor applies, zitten becomes a plausible coding option.

In still other cases, the link with the prototype or other locational expressions may no longer be transparent, even to native speakers. For instance, Dutch commonly conceptualizes causes as 'lying' even in context where metaphorical containment is at issue. Some coding variation can be expected here, as is also the case, for instance, with references to absence of (abstract) activity, which can occur with any of the three posture verbs ('standing', 'sitting', or 'lying still'). The same is also true for values on a scale (they can 'sit', 'lie' or 'stand' at a certain position). As we will show, all of these uses impose their own image on a given situation, and each has its own set of inferences and/or related codings (e.g., the respective transitive equivalents).

A special subcase that we will consider in our talk is that of local prototypes, where a given metaphor has only a weak link with the prototype but has become quite entrenched itself and thus functions as a kind of local source for new extensions. Two nice examples are "printed text" and "open doors", both coded with "staan" ('stand'). The reason why printed text is coded with "staan" is probably no longer transparent to most native speaker, but it has laid the basis for extensions to all kinds of non-textual imprints, of either temporal or permanent nature, e.g. pictures, text on a screen, marks on the body. That open doors and windows are coded with "staan" is presumably due to the combination of their typical vertical dimension and their being perceived as being able to self-sustain a certain posture, an attribute that is part of the prototype of "staan" as well. Its high entrenchment motivates extensions to open coats or open discussions, all allowing a coding with "staan".

Our corpus analysis shows that the conceptual categories coded by the posture verbs takes the form of a semantic network in which nodes not only differ in salience but also in how they are related to the prototype. In the case of local prototypes, different extensions relate to one particular well-entrenched extension rather than to the general prototype. We will show what such networks can look like, for the three Dutch posture verbs. In some contexts, the posture verbs will be competitors in the coding contest. This view on lexical categories lines up with Tuggy's (1993), who addresses the vexing problem of distinguishing polysemy and vagueness and who also draws on the notions of "elaborative distance" and "relative entrenchment of (sub)meanings" to show how there is in fact a cline ranging from homonymy to vagueness.

REFERENCES

Feyaerts, K. 1997. Die Bedeutung der Metonymie als konzeptuellen Struktur-prinzips. Eine kognitiv-semantische Analyse deutscher Dummheitsausdrücke. Unpub. PhD. Dissertation, K.U.Leuven.
Lakoff, G. 1987. Women, Fire and Dangerous Things. Chicago: Chicago University Press.
----. & Johnson, M. 1980. Metaphors We Live By. Chicago: Chicago University Press.
Langacker, R. W. 1987. Foundations of Cognitive Grammar. Volume I: Theoretical Prerequisites. Stanford: Stanford University Press.
----. 1991. Foundations of Cognitive Grammar. Volume II: Descriptive Application. Stanford: Stanford University Press.
Lemmens M., "Tracing referent location in oral picture descriptions". Paper at the Corpus Linguistics 2001 Conference , Lancaster U.K., 29/03 - 02/04.
----. (ms.) "The semantic network of Dutch posture verbs", in review to appear in Newman, John. (ed.) The Linguistics of Sitting, Standing, and Lying. Amsterdam & Philadelphia, John Benjamins.
Tuggy, D. 1993. "Ambiguity, Polysemy, and Vagueness". Cognitive Linguistics 4: 273-290.