Semantic Maps

Abstract

A semantic map is a method for visually representing cross-linguistic regularity or universality in semantic structure. This method has proved attractive to typologists because it provides a convenient graphical display of the interrelationships between meanings or functions across languages, while (at the same time) differentiating what is universal from what is language-specific. The semantic map model was initially conceived to describe patterns of polysemy (or, more generally, of co-expression) in grammatical categories. However, several studies have shown that it can be fruitfully extended to lexical items and even constructions, suggesting that any type of meaning can be integrated in a map. The main idea of the method is that the spatial arrangement of the various meanings reflects their degree of (dis)similarity: the more similar the meanings, the closer they are placed—in accordance with the so-called connectivity hypothesis. Within the semantic map tradition, closeness has taken different forms depending on the approach adopted. In classical semantic maps (alternative terms: “first generation,” “implicational,” “connectivity” maps), the relation between meanings is represented as a line. This is the graph-based approach. In proximity maps (alternative terms: “similarity,” “second generation,” “statistical,” “probabilistic” maps), the distance between two meanings in space— represented as points—indicates the degree of their similarity. In this scale- or distance-based approach, the maps are constructed using multivariate statistical techniques, including the family of methods known as multidimensional scaling (MDS). Both classical and proximity maps have been widely used, although the latter have recently gained interest and popularity under the assumption that they can cope with large data more efficiently than classical semantic maps. However, classical semantic maps continue to be useful for studies aiming to discover universal semantic structures. Most importantly, classical maps can integrate information about directionality of change by drawing an arrow on the line connecting two meanings or functions. Beyond the choice between the two types of maps, one of the issues that has sparked debate and critical reflection among researchers is the universal relevance of semantic maps. The main question that these researchers address is whether semantic maps reflect the global geography of the human mind. Another much discussed issue is the identification of the factors that increase the accuracy of semantic maps in a way that allows for valid cross‐linguistic generalizations. Such factors include the choice of a representative language sample, the quality of the collected cross‐linguistic material, and the establishment of valid cross-linguistic comparators. Acknowledgments: The author wishes to thank one anonymous reviewer for their useful comments. For discussion of the material in this article, the author is grateful to Stéphane Polis.