Lexical Diachronic Semantic Maps:

Representing and explaining meaning extension

Situated within the tradition of the semantic map methodology, the research project “Le Diasema” addresses the question of how semantic maps make significant predictions about language change at the lexical level. More specifically, this project has the following three main objectives:

To incorporate the diachronic dimension into semantic maps of content words;
To extend the method so as to also include information about the cognitive and cultural factors behind the development of the various meanings;
To create a web service that will allow researchers to generate diachronic maps themselves.

The empirical basis of “Le Diasema” consists of Ancient Greek (8th – 1st c. BC) and Ancient Egyptian (26th c. BC – 10th c. AD), two languages with significant diachronic material based on which we are able to describe the semantic extensions of lexemes belonging to different semantic fields. Using dynamic semantic maps, we then compare the observed meaning extensions in the two languages in order to identify shared cognitive motivations and to assess the potential impact of cultural factors on the evolution of the various lexical fields in both languages.

What are semantic maps?

A very short introduction

A semantic map is a common method in semantic typology to detect cross-linguistic regularities and recurrent patterns in semantic structure. Additionally, it is used as a heuristic technique to analyze the multiple senses of linguistic units in the languages of the world, as well as their possible meaning extensions (see, e.g., Haspelmath 2003; Cysouw et al. 2010).

Based on cross-linguistic polysemy data, one can infer a graph of connected senses (the nodes of the graph). There is a single constraint for plotting semantic maps: all the linguistic items, when mapped onto the graph, must cover a connected region (this is known as the connectivity hypothesis; see Croft 2001).

The figure on the right is an illustration of lexical semantic map with weighted edges. In order to plot this map, we processed data from CLICS for 13 meanings connected to the senses ‘tree’ and ‘wood’. We ended up with 245 ‘words’ in 168 languages expressing more than one of these 13 meanings. It is then possible to plot a map for the 13 meanings that respects the connectivity hypothesis. This means that each word of the data sample will cover a connected region of the graph.

Furthermore, this map is weighted: the thickness of the edges between the nodes depends directly on the frequency of a polysemy pattern. For instance, it shows that words expressing at the same time the meanings ‘tree’ and ‘wood’ (thick line) are much more frequent than the ones expressing simultaneously ‘tree’ and ‘woods, forest’ (thin line).

Some examples

Three common types of semantic maps

Classic semantic maps

Classic semantic maps are two-dimensional graphs with meanings connected by lines. The length of the edge is generally irrelevant, only the graph structure matters.

A semantic map of dative function

(Haspelmath 2003: 213)

Dynamic semantic maps

Dynamic semantic maps are semantic maps with arrows replacing the lines between the meanings. The oriented graph shows the possible pathways of meaning extension for any linguistic item.

A dynamic map of modal necessity

(van der Auwera & Plungian 1998: 95)

Weighted semantic maps

Semantic maps can have weighted-edges; the weight represents the frequency of co-occurrence of meaning pairs in linguistic items. It is usually captured graphically by thicker vs. thinner lines or by different types of lines (solid vs. dotted vs. dashed lines).