Hjelmslev, a forerunner of the semantic maps method in linguistic typology?

ABSTRACT

In this paper, we show that Hjelmslev’s approach to language description and crosslinguistic comparison, on the one hand, and the semantic maps model used in linguistic typology, on the other, differ significantly. Although Hjelmslev paved the way for employing graphic representations as a means to show how each language of the world subdivides the semantic continuum in its own way, he can hardly be considered as a forerunner of the semantic maps tradition. In a nutshell, Hjelmslev’s schemas are meant to compare the specific organisation of individual linguistic systems, but the semantic maps method aims at unveiling semantic regularities across languages. The former targets the particular ‘grid’ imposed by each language on a given semantic space, but the latter abstracts away from specific linguistic systems and posits universal atoms of sense that can be organised in cross-linguistically valid networks.

KEYWORDS:

• General linguistics
• structural linguistics
• typology
• semantic maps
• glossematics
• Hjelmslev

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Notes

¹ *We are very grateful to Sémir Badir (Liège), Hans Basbøll (Odense), Eitan Grossman (Jerusalem), Frederick Newmeyer (Washington), and an anonymous reviewer for insightful comments on earlier drafts of this paper.

See Section ‘Early Studies and Theoretical Foundations’ of Georgakopoulos (2019). In this paper, we focus on so-called ‘classical semantic maps’ (van der Auwera 2008). Many observations could also apply to the proximity maps (van der Auwera 2013), but they are not discussed here specifically. For a recent comparison of the two approaches, see Croft (2022).

² Wälchli refers to Schopenhauer (1913), who “used overlapping circles to illustrate the non-congruence of concept across different languages”, as an “early philosophical pioneer of the semantic map method” (2010, 336).

³ Gilbert Lazard, to whom reference is made by Cysouw, Haspelmath, and Malchukov (2010, 1), is an important intermediate figure between orthodox structuralism and modern functional typology. In his contribution about “La catégorie de l’éventuel”, for instance, he mentions explicitly Hjelmslev’s “substance du contenu” and “forme du contenu” (Lazard 1975, 358), while anticipating many aspects of the semantic map method. We are thankful to an anonymous reviewer for drawing our attention to Lazard’s early influence in this field.

⁴ This is also known in the literature as ‘Haiman’s isomorphism hypothesis’ (see Wälchli 2010, 337). To put it bluntly: identity of form reflects similarity in communicative function or meaning (see already Anderson 1982, 227–228).

⁵ Commutation is the most important requisite for comparison (Hjelmslev [1943] 1961, 74), but not the only one. Its role is discussed in detail in §14, Invariants and variants.

⁶ The linguistic description resulting from the projection of the structure of a language onto another (Hjelmslev 1935, 80).

⁷ The proper ‘cases’, as content-elements, not the case-markers, called formants, which are their corresponding expression-elements.

⁸ The reduction of linguistic units to minimal distinctive features is of course a shared methodological stance across structural approaches, yet the way glossematics carries out this reduction is specific to glossematics, because it follows Hjelmslev’s assumption of glossemes as pure forms. This contrasts with the Prague school, whose conception of ‘distinctive features’ was substance-based. In glossematics, the analysis into glossemes is universal: their number comes from a decomposition of the number of taxemes into prime factors (see Hjelmslev [1943] 1961, 100) and their definition depends on structural phenomena within a given class of taxemes (mostly neutralisations). Glossemes may receive a substance-interpretation, yet this is not indispensable and is carried out at a different stage of the analysis. For instance, the position β on the first dimension of the case-category (‘approximation – distancing’) can be interpreted as denoting ‘mostly proximity,’ while the position α on the first dimension of the vowel-system (‘anterior – non anterior’) can receive the interpretation ‘exclusively non anterior’ (see for instance Hjelmslev [1948] 1970, 220).

⁹ i.e., a government which can only occur between elements belonging to the same sentence (nexus), see Hjelmslev ([1938] 1970, 157–158).

¹⁰ The hypothesis of the meaningful content of categories, see Hjelmslev (1928, Ch. III, § I).

¹¹ This claim is not absolute, since the meaning depends on the value of every element within a given category.

¹² The semantic explanations provided by Hjelmslev are quite subtle and sometimes felt as subjective, as the process of combining (inferring) the particular meanings of the elements (polysemic state) from the abstract meaning of the category (monosemic state) is not always clear. See in this respect the description of the ‘attributive’ (Hjelmslev 1935, 139, 154) and ‘comparative’ cases (Hjelmslev 1937, 15–16; see also Hjelmslev 1937, 72). This might have been one of the reasons why the method was later adapted, with the introduction of purely formal criteria of distribution (such as ‘neutralisations’ or ‘syncretisms’).

¹³ In this sense, they are language-specific descriptive categories (as opposed to comparative concepts) in modern typological terms (Haspelmath 2010).

¹⁴ See Hjelmslev’s idea of “conversion” in Hjelmslev [1934] 1972, 111–sq., discussed in Cigana 2016; cf. also Hjelmslev [1957] 1970.

¹⁵ Deductive semantic analysis is assuredly used (see van der Auwera and Temürcü 2006, 132), but is not a core feature of this model.

¹⁶ Note that scholars who use the semantic maps method have different opinions about the concepts that they identify and plot on a map. Some linguists see them as cognitively salient (Croft 2001; Cristofaro 2010 is critical of this position), while others consider them to be merely comparative concepts (Haspelmath 2010) constructed for linguistic comparison (Haspelmath 2016, Haspelmath 2018) without any commitment to their cognitive reality.

¹⁷ Which goes back at least to the Humboldtian conception of languages as systems that make infinite use of finite means.

¹⁸ For this principle, see further Haspelmath (2003) and François (2008). Note that it cannot be easily tested and respected when large databases of colexicalization patterns (such as Clics³ [https://clics.clld.org], see Rzymski et al. 2019) are used as a source for plotting semantic maps. The concepts or meanings in such databases are indeed posited in advance (see the Concepticon [https://concepticon.clld.org]); they do not result from cross-linguistic comparison.

¹⁹ For a list, see Georgakopoulos and Polis (2018, 21).

²⁰ For an overview of the lexical maps, see Rakhilina, Ryzhova, and Badryzlova (2022).

²¹ A look at the issues visualised by Figure 11, for a single category, should make this point clear.