Karlien Franco*, Dirk Geeraerts, Dirk Speelmanand Roeland Van Hout
Concept characteristics and variationin lexical diversity in two Dutchdialect areas
https://doi.org/10.1515/cog-2017-0136Received 10 December 2017; revised 27 August 2018; accepted 07 September 2018
 Lexical diversity, the amount of lexical variation shown by a particularconcept, varies between concepts. For the concept
, for instance, nearly 3000 English expressions exist, including
 blitzed, intoxicated
, and
.For the concept
, however, a significantly smaller number of lexical itemsis available, like
. While earlier variation studies have revealedthat meaning-related concept characteristics correlate with the amount of lexical variation, these studies were limited in scope, being restricted to one semanticfield and to one dialect area, that of the Limburgish dialects of Dutch (Geeraerts,Dirk & Dirk Speelman. 2010. Heterodox concept features and onomasiologicalheterogeneity in dialects. In Dirk Geeraerts, Gitte Kristiansen & Yves Peirsman(eds.).,
 Advances in Cognitive Sociolinguistics
 (Cognitive Linguistics Research 45),23
39. Berlin: De Gruyter Mouton; Speelman, Dirk & Dirk Geeraerts. 2007. Destructuur van lexicale onzekerheid [The structure of lexical insecurity].
 Taal & Tongval
 20. 47
61, Speelman, Dirk & Dirk Geeraerts. 2008. The role of conceptcharacteristics in lexical dialectometry.
 International Journal of Humanities and Arts Computing 
2). 221
 ). In this paper, we investigate whether the impactof concept characteristics, viz. vagueness, lack of salience and proneness toaffect, is manifest in a similar way in other dialects and other semantic fields. Inparticular, by extending the scope of the earlier studies to other carefully selected semantic fields, we investigate the generalizability of the impact of concept characteristics to the lexicon as a whole. The quantitative approachthat we employ to measure concept characteristics and lexical diversity metho-dologically advances the study of linguistic variation. Theoretically, this paper
*Corresponding author: Karlien Franco,
 QLVL, University of Leuven, Leuven, Belgium,E-mail: karlien.franco@kuleuven.behttp://orcid.org/0000-0001-6499-265X
Dirk Geeraerts:
 E-mail: dirk.geeraerts@kuleuven.be,
 Dirk Speelman:
E-mail: dirk.speelman@kuleuven.be, QLVL, University of Leuven, Leuven, Belgium
Roeland Van Hout,
 Department of Linguistics, Radboud University Nijmegen, Nijmegen,Netherlands, E-mail: r.vanhout@let.ru.nlCognitive Linguistics 2018; aop
contributes to the further development of Cognitive Sociolinguistics by showcas-ing how meaning can be a source of lexical diversity.
 lexical diversity, lexical variation, dialectology, cognitive sociolin-guistics, Dutch
1 Introduction
Lexical diversity, the amount of lexical variation a particular concept shows, candiffer between concepts. For the concept
for instance, nearly 3000English expressions exist, including
 blitzed, intoxicated, hammered
m not as think as you drunk I am
 (Dickson 2009
 , cited in Lillo 2009). For the concept
, however, a significantly smaller number of lexical items is available, like
. Generally, studies in variational linguistics assume that thistype of lexical variation is influenced by lectal features, like the region of thelanguage user (e.g. Grieve etal. 2011; Wieling etal. 2014). However, as is appar-ent from this example, variation in lexical diversity can also be influenced by characteristics of the concepts to be expressed: for instance, concepts that areprone to affect show more variation (Allan and Burridge 1988, 2006).Crucially, the perspective on meaning and categorization in CognitiveLinguistics (CL) suggests that other types of meaning-related concept
character-istics may also affect lexical diversity. Geeraerts and Speelman (2010) andSpeelman and Geeraerts (2007, 2008) already found evidence for the interactionbetween features related to the prototype-theoretical organization of the lexiconand lexical diversity. First, they demonstrated that concepts that are moresalient, i.e. highly familiar concepts that are wired in very strongly in themind of a language user, show significantly less variation. Second, their resultsalso indicated that concepts that are vague, i.e. fuzzy at the edges and difficultto distinguish from associated concepts, show significantly more variation.
However, these studies only examined the influence of these meaning-relatedfeatures in one semantic field (viz. the human body) and in one dialect area (viz.the Limburgish dialect area of Dutch). As a result, some further questions remain
 In this paper,
 are used to indicate concepts, while
 are used for the lexicalitem used to refer to these concepts.
 We use
 in a slightly broader sense than
: while the latter may be definedas the denotational sense of a word, the former also includes additional semantic characteristicsthat may be associated with the concept (such as its degree of vagueness).
 Additionally, Speelman and Geeraerts (2008) showed that taking these concept characteris-tics into account improves the quality of dialectometric distance measures.
 Karlien Franco etal.
unanswered: Do concept characteristics like vagueness, salience and affectinfluence lexical diversity in semantic fields other than the human body? Isthe effect of the concept characteristics stable and equally strong in eversemantic field? Can we establish that such features impact lexical variation inthe same way in dialect areas other than the Limburgish region as well?The aim of this paper is, therefore, to examine to what extent the influenceof the concept characteristics distinguished in the previous studies is general-izable, in a double sense: by including concepts from other types of semanticfields in the analysis, and by broadening the geographical scope of the investi-gation through the inclusion of the Brabantic dialect area of Dutch. We willinvestigate the generalizability of the impact of concept characteristics in thefollowing way: is their impact stable and independent from both semantic fieldand dialect area and, if not, are differences categorical (the effect of conceptcharacteristics is absent or present) or gradual (their effect differs in strengthdepending on semantic field and/or dialect area)?That the effects should be pervasive is indeed not self-evident. On the onehand, with regard to the geographic dimension, it is possible that the amount of lexical diversity differs between dialect areas. For instance, the Limburgishregion, which forms the area of investigation of the above-cited studies by Geeraerts and Speelman, is a rural area with, on average, large geographicaldistances between the locations. Thus, it may be the case that in regions wheregeographical differences are generally smaller, or where there is more urbaniza-tion, we find less lexical diversity overall due to lexical levelling (e.g. Milroy 2002; Trudgill 2011). On the other hand, with regard to the semantic dimension,different fields may behave differently. For instance, names for children
s gamesare notoriously diversified, due to the fact that child language is highly imagi-native and creative (Weijnen 1966: 336; Pickl 2013). In contrast, a field likereligion is probably prone to less variability, because it is based on a highly standardized framework, accompanied by specific concepts and their corre-sponding names which are often directly borrowed into (dialectal) varieties.Thus, the impact of the concept characteristics may be unstable or may shift if other types of concepts are included in the analysis. More specifically, thehuman body concepts that were included previously are relatively concreteand universal (Lakoff and Johnson 1980 and see below). However, the degreeof concreteness and of universality of a concept may in itself interact with theway these concepts are processed and, thus, possibly with the amount of lexicaldiversity in dialect data. Consequently, there are a number of interaction effectsthat we may find: the impact of the concept characteristics may be insignificantin other types of semantic fields, their strength or effect size may be larger orsmaller, or their impact may be of a different nature. The use of a larger and
Concept characteristics and variation
more diverse dataset and a more elaborate modelling procedure allow us toexamine the generalizability of these effects to the lexicon as a whole. Ourresults will show that, although there are general differences in the amount of lexical diversity between the two dialect areas and differences in the strength of concept characteristics between semantic fields, overall, the concept character-istics significantly affect variation in the lexicon as a whole.More generally, we intend to contribute to the further development of Cognitive Sociolinguistics (Kristiansen and Dirven 2008; Geeraerts etal. 2010;Pütz etal. 2014) by adopting an onomasiological perspective to examine varia-tion in the dialect lexicon. On the one hand, the added value of complementingsemasiological research into lexical variation (i.e. research on
: exam-ining the meanings that a linguistic construction can take) with an onomasio-logical approach (i.e. research on
: investigating which names can beused to refer to a particular referent), was first made explicit in Geeraerts etal.(1994; further developments include Daems etal. 2015; Geeraerts etal. 1999;Heylen and Ruette 2013; Soares da Silva 2010; Speelman etal. 2003; Zenneretal. 2012). We contribute to this research paradigm by showing that investigat-ing a linguistic variety that has been under-researched in CL, viz. the dialectlexicon (exceptions include Swanenberg 2000; Szelid and Geeraerts 2008), usingan onomasiological approach offers new insights into the structure of linguistic variability. On the other hand, we open up a new line of research for studies inCognitive Sociolinguistics and CL at large. More specifically, the study of lan-guage-internal variation from a cognitive linguistic point of view has beencharacterized as focusing on the
 variation of meaning
 (the lectal and contex-tual variability of the meaningful categorization and construal phenomena thatlie at the core of CL) and the
meaning of variation
 (the categorization andcognitive representation of linguistic variation; see Geeraerts etal. 2010: 6
10).With the present paper, we want to draw attention to a third perspective on theinteraction between meaning and variation, viz.
 variation through meaning
:how the amount of variation in the expression of a concept is influenced by thecharacteristics of that concept. Finally, our quantitative approach can methodo-logically advance the study of linguistic variation in CL because the data andtechniques we use to gauge characteristics of concepts can also be applied toother types of research questions.Before turning to a more precise definition and operationalization of theconcept characteristics in Section 3, Section 2 sketches the datasets used in thispaper. Section 4 presents the quantification of lexical diversity, the dependent variable in the analysis. The results are discussed in Section 5. Section 6 roundsoff this paper with a conclusion.
 Karlien Franco etal.
2 Data
To examine whether the influence of concept characteristics on lexical diversity is generalizable across different dialect regions, we include data from thedigitized databases of two large-scale dialect dictionaries of Dutch, viz. the
Woordenboek van de Brabantse dialecten
Dictionary of the Brabantic Dialects
(WBD) and the
 Woordenboek van de Limburgse dialecten
Dictionary of theLimburgish Dialects
 (WLD). The latter dictionary was also used in the studiesby Geeraerts and Speelman (2010) and Speelman and Geeraerts (2007, 2008). Abig advantage of these dictionaries is that they are onomasiologically organized,containing all the lexical dialect variants available in the Brabantic andLimburgish dialect areas for a large number of concepts. Each dictionary con-sists of three large parts, (1) agrarian terminology, (2) specialized non-agrarianterminology and (3) general vocabulary, but for practical reasons we only include data from the third part. Part 3 is further subdivided into four sections,which each consist of three or four thematic volumes that each represent asemantic field, like
the human body 
the house
society, school & education
. To select the semantic fields that are included in the analysis, wefully rely on the way the concepts and semantic fields are delineated in thedictionary, i.e. on the manner in which the editors of the WBD and WLD carveup the lexical data into concepts (entities of meaning) and semantic fields (setsof thematically related concepts).
As we are examining the effect of concept characteristics on the number of lexical variants that are available per concept, it is important that the data werecollected systematically. For this reason, we restrict attention to the dialectal
 The same procedure was followed in the studies by Geeraerts and Speelman (Geeraerts andSpeelman 2010; Speelman and Geeraerts 2007, 2008). However, a reviewer correctly points outthat we could have used some other way to examine the relationship between the lexicaldiversity in the basic material (i.e. the lexical items, regardless of which concept/semanticfield they are attributed to by the lexicographers) and the structure of the lexicon, characterizedby differences in salience and vagueness. However, there are two reasons why our approach hasmerit. First, the fieldworkers and lexicographers working on the WBD and WLD were well-trained dialectologists, who were very familiar with the Brabantic and Limburgish language andculture. As a result, we trust that their intuitions about the organization of the material they worked with, reflect the natural patterns to a large extent. Second, as the data were collecteddecades ago, we fear that imposing some other
 a priori
 defined structure on the data may resultin a bias we wish to avoid (e.g. towards modern-day society). Furthermore, this would entailtaking an additional step (viz. not only analyzing the relationship between variation in lexicaldiversity and the structure of the lexicon, but also examining the conceptualization of a conceptin a dictionary). This falls outside the scope of our paper.
Concept characteristics and variation
of 38