Academic journal article Linguistic Discovery

Semantic Maps as Metrics on Meaning

Academic journal article Linguistic Discovery

Semantic Maps as Metrics on Meaning

Article excerpt

1. Measuring Meaning

Meaning is a particularly elusive property to measure. The central problem is that the meanings of linguistic expressions are variable across languages, and it is still mostly unknown how large this variability is. It does not really help to analyze the meaning of a language-specific expression (for example the English verb to walk) by saying that it expresses a general concept (like WALK). Such a change in typography still leaves open the question as to what the relation is between WALK and, for example, the meaning of the German word spazieren or the Spanish word andar. Actually, without a more explicit definition of the concept WALK, asking whether andar expresses the concept WALK is not much different from asking whether andar means the same as to walk. Yet, individual linguistic expressions across languages never convey exactly the same range of senses, making such a simplistic approach to comparing meaning across languages devoid of content.

In this paper, I will defend the view that a much more profitable operationalization of cross-linguistic variability of meaning is achieved by defining the meaning of a language-specific expression as the collection of all contexts in which the expression can be used. This definition represents, to some extent, a reversal of the intuitive notion of meaning. Meaning is typically thought of as some kind of property of a linguistic expression that governs its potential appearance in a particular context. In this conventional view, the main difficulty is how to express this property called "meaning". The approach to meaning proposed in this paper simply defines this property as the sum of all actual appearances. It is of course practically impossible to ever collect all appearances of a particular linguistic expression (be it a lexical or a grammatical item) in a living language--though this is possible for a dead language by including all documentation available--but samples of contexts can be used for any empirical question at hand (cf. Croft 2007; Walchli & Cysouw 2008 for a similar approach to meaning).

Samples of the actual occurrences of expressions in concrete contexts can be used to compare the variation in meaning between different language-specific expressions. So, instead of assuming that we know what the English expression walk means, I propose to sample its meaning by considering various contextualized occurrences of walk-like situations. To compare expressions across languages, ideally the same sample of contexts should be used for all languages investigated. The parallel collection of such occurrences across languages can take various forms. It is possible to use extra-linguistic stimuli, like pictures (e.g. Levinson & Meira 2003) or video sequences (e.g. Majid et al. 2007), and investigate the linguistic expressions used to describe them. The contexts can also be defined purely linguistically, using descriptions of situations (e.g. Dahl 1985) or examples from parallel texts (e.g. Walchli 2005).

In the practice of grammatical typology it is often impossible to collect sufficient parallel expressions because of the limited amount of material available and because of the difficulty of finding native speakers for all the languages to be investigated. So, instead of concrete occurrences of language-specific expressions in context, normally somewhat larger domains of contexts are used in which an expression can occur (e.g. Haspelmath 1997). These domains are (more or less) explicitly defined as "chunks" of meaning, large enough to be identifiable from reference grammars, and small enough to capture the main distinctions of the cross-linguistic variation. (1) Both parallel expressions in context as well as the somewhat more abstract domains of meaning as used conventionally in linguistic typology are called analytical primitives in Cysouw (2007). (2)

One of the consequences of comparing languages on the basis of an (empirical) selection of analytical primitives is that such a selection strongly reduces the range of possible meanings that can be identified across languages. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.