Central South Slavic Linguistic Taxonomies and the Language/Dialect Dichotomy: Rhetorical Strategies and Faulty Epistemologies

Introduction

To investigate the politics of the language/dialect (L/D) dichotomy, this article analyzes debates about the proper linguistic classification of a former Yugoslav region we will refer to as the “Central South Slavic” region. Since the 1990s, many such debates on this issue have invoked the L/D dichotomy. Debates imagined in terms of the L/D dichotomy have two opposing sides: “splitters” and “lumpers”. Splitters emphasize the importance of linguistic differences. At the time of writing, splitters in this particular debate mostly contend that Central South Slavic contains four distinct “languages”: Bosnian, Croatian, Montenegrin, and Serbian. The lumper position contends that there is a single “language”, most often denoted “Serbo-Croatian”. Lumpers acknowledge linguistic diversity in both speech patterns and literary traditions, but consider such diversity less important than commonalities, and refer to subordinate varieties with terms such as “dialect”, “variety”, or “variant”. Since the 1990s, debate between splitters and lumpers has been so passionate and voluminous that various textbooks treat the Central South Slavic region as a standard example illustrating the dynamics of the L/D dichotomy (e.g., Trudgill 1995, 145; Simpson 2019, 12; Wardhaugh and Fuller 2021, 28; Jones and Themistocleous 2022, 21). As a case study, then, Central Slavic L/D politics holds more general lessons.

Debates about linguistic classification, however, cannot be analyzed solely in terms of the debate between splitters and lumpers because numerous theoretical linguists reject the L/D dichotomy on ontological grounds. Linguists taking what we call an “agnostic” approach argue that the L/D dichotomy is linguistically meaningless and conceptually problematic, often relegating lumper/splitter disputes to a stigmatized realm described as “politics”. In contrast to agnostics, we propose that lumpers and splitters be grouped together under the heading of “assertionists”, since they are willing to make assertions about linguistic status in terms of the L/D dichotomy.

This article thus examines assertionist linguists who have taken a stance on Central South Slavic. Focusing particularly on arguments presented as “linguistic”, we analyze assertionists from an epistemological perspective. What linguistic evidence do assertionsts adduce in justification? Assertionist studies are rarely theoretical: most seek to win support for either the splitter or lumper position. How do they seek to persuade? We find that, when viewed from an agnostic perspective, the arguments made by splitters and lumpers are quite similar. We emphasize that we take an agnostic approach: we do not take sides in Central South Slavic lumper/splitter debates and adopt no position on the “correct” number of Central South Slavic languages. Our goal instead is to analyze assertionist arguments as rhetorical strategies, so as to better understand L/D disputes as a form of politics.

The Language/Dialect Dichotomy and Central South Slavic

Our main theoretical contribution to the study of debates on language (cf. Berthele 2008; Blommaert 1999) is to analyze the L/D dichotomy not as a problem requiring a solution, but as a cognitive and rhetorical framework. Fundamentally, we analyze the L/D dichotomy as a mental construct and rhetorical device. We argue that analyzing how linguists use the L/D dichotomy to frame debates about linguistic classification can shed light on political attitudes.

We offer no definitions of the terms “language” and “dialect” because we do not use these words as analytical terms. Taking an agnostic approach, we consider “language” and “dialect” poor categories of analysis. We suggest instead that they are most usefully analyzed as “categories of practice”, which an influential article by Rogers Brubaker and Frederick Cooper defined as “categories of everyday social experience, developed and deployed by ordinary social actors”, and specifically “used by political entrepreneurs to persuade people to understand themselves, their interests, and their predicaments in a certain way […] and to organize and justify collective action along certain lines” (Brubaker and Cooper 2000, 4–5).

Since the categories “language” and “dialect” are highly contested, the various actors who invoke the L/D dichotomy ascribe diverse meanings to these key words. Indeed, as Raf van Rooy documented in his outstanding study of the “conceptual pair”, the development of these terms has been characterized by “terminological rather than conceptual continuity” since their invention in the sixteenth century (Van Rooy 2020, 299, for a taxonomy of different meanings, see pages 147–51). We suggest that understanding how the L/D dichotomy functions as a political argument requires the analysis of three binary oppositions: (1) the superordinate variety vs. the subordinate variety, (2) splitters vs. lumpers, and (3) assertionism vs. agnosticism. We consider each in turn.

Firstly, the L/D dichotomy typically posits a binary linguistic taxonomy juxtaposing a relationally supraordinate variety, typically called a “language”, with relationally subordinate varieties, typically called “dialects”. In the specific case of Central South Slavic, however, assertionist terminology shows considerable variation. Central South Slavic authors positing binary linguistic taxonomies containing one relationally supraordinate variety and relationally subordinate varieties have described the subordinate variety not only as a dijalekt (or dijalekat), but also as a narječje, govor, varijanta, or inačica.

Since the 1960s, the terms dijalekt, narječje, and govor have mostly denoted putative linguistic varieties unrelated to nation-state boundaries. Lumpers do not use them to describe written literary traditions, i.e., what splitters present as “national languages”. The term narječje most commonly refers to Štokavian, Čakavian, and Kajkavian, putative varieties respectively defined through the isogloss separating the words što, ča, and kaj (meaning “what”). In recent Central South Slavic dialectological taxonomies, a dijalekt typically denotes a subdivision of a “what” isogloss narječje, while a govor generally refers to a linguistic variety of a small and usually rural area (Lisac 2003; Okuka 2008).

Hoping to avoid taking sides with either lumpers or splitters, we use the term “Central South Slavic” to avoid specifying the number of supraordinate varieties. We use the term as an approximate areal designator defined to exclude Slovene, Macedonian, and Bulgarian. Splitters might, for example, imagine Bosnian (or Bosniak), Croatian, Montenegrin, and Serbian as distinct “languages”, each containing subordinate varieties. Lumpers imagine these varieties as subsumed within a superordinate “language”. The overarching “language” has most frequently been referred to as “Serbo-Croatian”, but has also been called “Croato-Serbian” (Bugarski 2004), “BCMS” (Simonović, Milosavljević, and Arsenijević 2023), “standard Štokavian” (Kapović 2011, 54), “Zajednički” [common language] (Bugarski 2018), and colloquially as “naški” [our language]. Lumpers typically describe the subordinate varieties Bosnian, Croatian, Montenegrin, and Serbian with the terms varijanta or inačica, both meaning “variant”.

At the time of writing, splitter assertionism dominates Central South Slavic political discourse: Bosnian, Croatian, Montenegrin, and Serbian currently enjoy widespread recognition as separate “languages” (cf. Greenberg 2004, 159–67). Lumper assertionism, however, was hegemonic before 1990 (cf. Jonke 1964, 101). During the Cold War, the Yugoslav state recognized “Serbo-Croatian” as an official language, even though the name of the putative common language as Roland Marti observed, proved “a bone of contention, srpskohrvatski ‘Serbocroatian’ alternating with the hyphenated srpsko-hrvatski ‘Serbo-Croatian,’ and both existing with the two elements in reverse order (hrvatskosrpski, hrvatsko-srpski)” (Marti 2012; see also Bugarski 2004). Some have also claimed a common language with two official “variants”, sometimes called “Eastern” and “Western” instead of “Serbian” and “Croatian” (Greenberg 1998, 719).

Several studies have examined the debate between splitter assertionism and lumper assertionism as a problem of political, intellectual, or ideological history (Bugarski 2010; Greenberg 2004). This literature has formulated the splitter/lumper dichotomy in different ways. Tomas Magner contrasted “separatist” and “unitarist” tendencies (Magner 1967, 344). Daniel Bunčić divided Central South Slavic “linguistic ideologies” into “ideologies of togetherness” and “ideologies of separateness” (Bunčić 2021, 344–52). Pieter Troch juxtaposed “Yugoslav” feeling with “particularist” sentiments (Troch 2015, 14). These studies, however, focused on the debate between lumpers and splitters and ignored agnosticism.

Indeed, the dichotomy between assertionists and agnostics has gone lamentably unexamined and untheorized, even though many theoretical linguists have expressed ontological doubts about the L/D dichotomy. Agnostics argue that the terms “language” and “dialect” have no objective linguistic referents. Henry Gleason’s textbook of descriptive linguistics claimed that “the problem of classification into such categories as language and dialect is intrinsically difficult or impossible. Several criteria can be proposed, no one of which is satisfactory” (Gleason 1961, 441). Cognitive linguist Ron Kuzar declared that “linguistic theory does not provide us with a clear definition of these terms. All attempts to base a clear classification of languages and dialects on objective criteria, such as mutual understanding or concentration of isoglosses, have failed” (Kuzar 2001, 234). Leading structuralist linguists, including Ferdinand de Saussure and Noam Chomsky, have also taken an agnostic stance (Chomsky 1988, 190; De Saussure 1959, 200–5; on Chomsky’s use of the dichotomy, see also Maxwell 2022). Many sociolinguists concur. Joshua Fishman judged that “the language/dialect issue […] is not resolvable on linguistic grounds alone” (Fishman 1985, 6). Richard Hudson’s sociolinguistics textbook insisted in italics that “there is no real distinction to draw between ‘language’ and ‘dialect’ ” (Hudson 1980, 36).

Indeed, some agnostics even proclaim agnosticism a universal linguistic consensus. Pioneering sociolinguist William Labov, for example, declared it “the general linguistic position that there is no substantive difference between language and dialect” (Labov 2011, 389). Indian scholar K. Das Gupta argued that “there is no difference between a language and dialect for a linguist” (Das Gupta 1991, 18). Sociolinguist Leonie Cornips wrote that “all linguists are convinced that the distinction between a language and a dialect cannot be made on the basis of linguistic criteria” (Cornips 2012, 17). Such categorial claims to universal consensus are untenable, since in practice numerous linguists have attempted to distinguish languages and dialects in terms of various linguistic criteria (Boga 2020; García and Sandhu 2015; Hammarström 2008; Koryakov 2017; Melinger 2018; Tamburelli 2021; Weiß 2009; Wichmann 2019). As we will see below, many professional linguists have also taken an assertionist position regarding Central South Slavic.

This article studies Central South Slavic assertionists who (1) present themselves as linguists and (2) argue for their preferred stance on linguistic grounds. A few scholars meet only the first of these two criteria: the Zagreb-based linguist and journalist Domagoj Vidović, for example, argued in 2016 that “the Croatian language is separate because its speakers consider it a separate language” (Vidović 2016, 6). Our analysis excludes Vidović because appealing to the supposed consensus of “Croatian speakers” is not a linguistic argument.

Before presenting our findings, we remind the reader that this analysis is not an exercise in linguistic classification, but a theoretical critique of assertionism. We have not studied assertionist linguists as purveyors of potential linguistic insight; we are not interested in the disagreements between lumpers and splitters. Instead, we are interested in the ontological gap separating assertionists from agnostics. We are investigating assertionist epistemological claims. How do assertionists claim to know what they are asserting? We analyze their arguments not as correct or incorrect, but as attempts to persuade. How do assertionist linguists attempt to convince others? What rhetorical strategies do they employ?

To keep the scope of the paper feasible, our case study focuses on assertionist claims about Central South Slavic written after 1990. A longer history would transcend the scope of a single article: Central South Slavic lumper/splitter debates have evolved considerably since the nineteenth century, when layered linguistic taxonomies first became intertwined with nationalist fantasies (Saenko 2022; Maxwell 2023; Vukotić 2024). We gathered material via the search functions of online databases and university libraries using the keywords “dialect”, “Serbocroatian”, “Serbian”, “Croatian”, and “Bosnian” in English, German, Russian, and both Latin and Cyrillic incarnations of Central South Slavic. We cannot claim to have exhaustively documented all assertionists, even those within our recent timeframe. Nevertheless, after studying the sources we collected (18 different scholars represented in 23 sources: 12 journal articles, 3 monographs and textbooks, 4 book chapters, 2 conference papers, and 2 media interviews), we find that the various scholars employ a limited range of rhetorical strategies. We suggest, therefore, that our sample presents saturated data for the post-1990 period.

We have documented three rhetorical strategies, which we call (1) the avalanche of trivia, (2) the appeal to hypothetical evidence, and (3) the denigration of the political. We find that both splitters and lumpers have invoked all three strategies, and indeed that splitters and lumpers sometimes adduce very similar evidence and/or reasoning to reach diametrically opposed conclusions. While in this study we have restricted our attention to Central South Slavic, we suspect that the rhetorical strategies documented here would also appear in similar sources from other parts of the world.

Assertionist Rhetorical Strategies (1): The Avalanche of Trivia

Let us begin with assertionists who have justified their stance with long lists of data illustrating linguistic difference, typically arranged under headings such as phonology, morphology, etc. We define the “avalanche of trivia” as a collection of data presented without a theoretical definition of what the data is supposed to prove. Splitters proclaim or insinuate that the differences presented are important; lumpers that they are insignificant. In the absence of a definition of what distinguishes a superordinate variety (e.g., a “language”) from a subordinate variety (e.g., a “dialect”), such data constitutes a logical non sequitur. The avalanche of trivia does, however, lend the author the appearance of erudition and expertise while intimidating lay readers.

Consider an article that linguist and translator Marko Kovačić wrote as a visiting scholar in Indiana. Kovačić began with the question: “Are Serbian and Croatian one language or two?” and concluded that “the two languages are closely related but not identical” (Kovačić 2005, 195, 203). He thus espoused splitter assertionism.

Kovačić presented linguistic data under three headings: “phonology and morphology” (as a single category), then “vocabulary”, and finally “syntax”. Kovačić noted for example that the phoneme /h/ “is more often omitted in the east, so hoćeš “you want” becomes oćeš” (Kovačić 2005, 197). Syntactically, he mentioned that Croatian uses infinitives where Serbian uses the word da to introduce a conjugated verb, e.g., ne mogu to jesti vs. ne mogu to da jedem (“I can’t eat this”), hereafter referred to as the issue of da-clauses (Kovačić 2005, 201). Lexically, Kovačić listed over one hundred word pairs, listing “Slavic words” (e.g., vlak – voz “train”) and “loan words” (kafa – kava “coffee”) in separate categories. Lexical differences also dominate seven sample sentences that supposedly illustrate the differences between Serbian and Croatian, e.g.,

Sr. Voz je stigao na stanicu tačno u pet časova.

Cr. Vlak je stigao na kolodvor točno u pet sati.

‘The train arrived at the station at exactly five o’clock.’

(Kovačić 2005, 203)

Kovačić concluded that the “most conspicuous differences between Serbian and Croatian are […] the vocabulary, mostly among nouns” (Kovačić 2005, 198).

Miro Kačić, then professor of linguistics at the University of Zadar, made a rather more exhaustive case for splitter assertionism in the 1995 book Hrvatski i srpski: Krivotvorine i zablude, written with the help of Ljiljana Šarić (Kačić 1995). The book attracted political support and was subsequently republished in English, German, and French translation (Kačić 1997a; 1997b; 2000). Chapter 14, entitled Inačice ili jezici? [Variants or languages?] listed numerous linguistic examples. Syntactically, Kačić contrasted Croatian infinitives with Serbian da-clauses (e.g., Hoću te prebiti – Hoću da te pribijem “I want to beat you up”) (Kačić 1995, 129). Phonologically, he contrasted Serbian e or i with Croatian ije, (e.g., dete – dijete “child” or zalivati – zalijevati “to water”) in order to claim that literary Croatian has seven vowels, (/a, o, e, i, u, r̥, iē/), unlike Serbian, where “there is no /iē/ sound” (Kačić 1995, 122–3). Kačić, like Kovačić, saw vocabulary as “the largest category” of differences (Kačić 1995, 130), contrasting puristic Croatian neologisms with international words, to which he ascribed a Serbian character (e.g., nogomet – fudbal “football”; topništvo – artiljerija “artillery”) (1995, 130). Elsewhere, he emphasized phonetic differences in both Greek and Slavic loanwords: the Greek sound marked with the letter β supposedly appears in Croatian words as b but in Serbian words as v (barbarin – varvarin “barbarian”), while in recent loanwords Croatian supposedly uses c where Serbian uses s (financ – finans “finance”) (Kačić 1995, 124). Kačić concluded that “lexical differences between Croatian and Serbian exist on several levels” (Kačić 1995, 130, cf. 1997a, 135).

The sorts of linguistic data that Kovačić and Kačić adduced to support splitter assertionism, however, have elsewhere been presented as evidence for lumper assertionism. John Frederick Bailyn, professor of linguistics at Stony Brook University in New York, tested what he called “the Single Language Hypothesis” by asking informants from Zagreb to translate “nine short texts from Serbian into Croatian” (Bailyn 2010, 189). He then studied the discrepancies between the texts, analyzing any phonological, morphological, syntactical, and lexical differences.

In a few cases, Bailyn adduced exactly the same evidence as did Kovačić and Kačić. All three, for example, refer to the same word pair: kemija – hemija (“chemistry”). Kovačić (2005, 1999) saw evidence of “different reflections of the same loanwords”, while Kačić (1995, 128) that “the Greek letter ch [χ] becomes k in Croatian and remains h in Serbian”. By contrast, Bailyn thought that such contrasts “are lexically specified and do not represent any phonemic distinction” (Bailyn 2010, 193, emphasis in original). Bailyn also thought da-clauses “the best-known syntactic division between the Eastern and Western variants” and reports that “five of the sixteen speakers transformed the clause in [an example sentence] into an infinitival clause”. He then declared them “a construction choice rather than a real parametric syntactic distinction”, partly because “Serbian speakers are perfectly comfortable with both variants”, but also because “the Western variant does not disallow the da + tensed V clauses, despite the common lore to the contrary” (Bailyn 2010, 204, emphasis in original). Finding that “all variants” accept the statement Naredili su mu da ide (“they ordered him to go”) and reject Naredili su mu ići, Bailyn concluded that “Serbian and Croatian speakers share the same judgements about this unusual restriction, a similarity far more intriguing and potentially important than the differences in construction choice” (Bailyn 2010, 207).

Since Kovačić, Kačić, and Bailyn examined the same linguistic examples and reached diametrically opposed conclusions, it seems that those examples mean nothing on their own. Linguistic evidence cannot resolve a L/D dispute without a theoretical context: raw data is relevant only in reference to clear linguistic definitions of “language” or “dialect”. Kovačić and Kačić failed to provide any such definitions.

Bailyn, however, presented criteria for evaluating linguistic data in terms of the L/D dichotomy. Some of his criteria are absolute: his Single Language Hypothesis is supposedly confirmed if two varieties have “the same phonemic inventory”, “identical derivational morphological devices”, “identical distinctly represented morphological categories”, and the “identical form of the actual bits of inflectional morphology” (Bailyn 2010, 188). These tests permit no hedging: the phonemic inventory must be “the same” not “nearly the same”; the various morphological features must be “identical”, not “nearly identical”. Perhaps because Bailyn’s tests are so absolute, they require no quantification: the slightest difference would evidently falsify the Single Language Hypothesis.

When examining phonemic and morphological linguistic evidence, Bailyn claimed that “all variants” shared the same five vowel phonemes (/a, e, i, o, u/) and the same 25 consonant phonemes (2010, 191, 192). When adducing word pairs such as suvo – suho (“dry”), Bailyn maintained that “these systemic distinctions do not introduce any phonemes” (Bailyn 2010, 189), which supports the Single Language Hypothesis. In his analysis of morphology, however, Bailyn hedged. He found “almost no variation” in derivational morphology (Bailyn 2010, 195, emphasis added) identifying “only instances of a minor change in morphological sub-category of nominals” (2010, 198, emphasis added). He concluded that “Croatian and Serbian (and Bosnian and Montenegrin) share 99 % of morphological categories expressed overtly and 99 % of the specific morphophonemic forms” (Bailyn 2010, 200). Had Bailyn applied his stated thresholds of absolute similarity, his evidence would have led him to splitter assertionism: a “99 % similarity” is not “identical”. In practice, however, Bailyn argued for lumper assertionism. While it might be defensible to confirm the Single Language Hypothesis with a level of similarity less than 100 %, a non-absolute threshold ought to be specified quantitatively. Would the Single Language Hypothesis be satisfied if the similarity was 95 %? 80 %? Bailyn does not propose a numerical threshold, so his data are merely impressionistic.

As concerns syntax and lexicon, however, Bailyn proposed vaguer criteria. In terms of syntax, he required “identical settings of major syntactic parameters”. Lexically, he demanded a “degree of identity” (Bailyn 2010, 188). He neither explained how to decide which syntactic parameters qualified as “major” nor quantified what “degree” of lexical identity sufficed.

Bailyn himself judged his syntactic evidence inconclusive. Comparing reflexive possession constructions, he found “0 % of changes” in one situation, but “a fairly high percentage of changes” in another, a conclusion that he judged “may or may not represent some sort of syntactic difference between the two variants”. He suggested that “a follow up study […] might illuminate this issue” (Bailyn 2010, 203).

Bailyn deemed his lexical evidence more decisive: “The lexical distinctions that do exist come nowhere near to reaching the level of distinguishing two languages.” Bailyn did not, however, specify what that level would be (2010, 187, emphasis in original). He presented his evidence by word type, finding that nouns changed “8.92 %” of the time, implying a similarity of 91.08 %. He found that “functional and grammatical lexical items” remained the same “in 99.79 % of cases,” concluding that “clearly, the Single Language Hypothesis is strongly supported” (Bailyn 2010, 209). Nevertheless, Bailyn did not actually specify what percentage similarity would satisfy the Single Language Hypothesis. He concluded merely that “a far wider set of lexical distinctions are commonly found between dialects considered the same language” (Bailyn 2010, 207).

When discussing thresholds of lexical similarity, Bailyn might have referred to lexicostatistics, a subfield of linguistics where many scholars have specified thresholds of linguistic similarity to differentiate “languages” from “dialects”. Pioneering lexicostatician Morris Swadesh proposed in 1954 that varieties sharing fewer than 80 % of cognates were separate “languages”. Other lexicostatisticians have proposed other thresholds; a survey of lexicostatistal studies conducted in New Guinea documented thresholds as low as 54 % and as high as 91 % (McElhanon 1971, 134). While Bailyn did not follow lexicostatistical procedures when calculating lexical similarity, applying thresholds would have supported his lumper stance. Bailyn (2010, 208) calculated an overall lexical change of 3.91 %, implying a lexical similarity of 96.09 %, which is well above the most popular lexicostatistical thresholds. Nevertheless, he did not actually specify any threshold.

In short, while Bailyn tried to define the Single Language Hypothesis, his definition proves problematic. His phonemic test worked, but his morphological tests were so absolute that he himself chose not to apply them. His syntactic test proved inconclusive. His lexical test was insufficiently quantified. Yet Bailyn still outperformed Kačić and Kovačić, neither of whom offered even implicit theoretical definitions. Kačić simply proclaimed things like “phonetic differences are evident” or “syntactical differences become evident” (Kačić 1997a, 127, 133), without analyzing or commenting on the relative importance of those differences in relation to similarities. Kovačić wrote of Serbian and Croatian that “similarities between them are superficial, while differences are found more deeply” (Kovačić 2005, 203), without explaining how to decide whether individual similarities or differences qualify as superficial or deep.

While linguistic data presented without reference to clear theoretical definitions is merely trivia, it nevertheless serves a rhetorical function. In sufficiently large quantities, linguistic minutiae establish the author’s credentials as a knowledgeable expert, and can impress and intimidate nonexperts. Even the pedantry and tediousness of presenting such data may serve a purpose: doubters may prefer to abandon the debate. Overall, the avalanche of trivia seeks to persuade by overwhelming its audience.

Assertionist Rhetorical Strategies (2): The Appeal to Hypothetical Evidence

Gathering data for an avalanche of trivia requires time and effort, and many assertionists choose an easier path. Instead of listing sound shifts, word order differences, or lexical pairs, some authors appeal to hypothetical evidence described as phonological, syntactical, lexical, or some other linguistic adjective. Alternatively, some assertionists invoke abstractions such as “linguistics”, or an imaginary linguistic consensus. As with the avalanche of trivia, adducing hypothetical evidence ultimately rests on establishing linguistic expertise. Instead of establishing that expertise with an extensive recitation of linguistic trivia, however, the appeal to hypothetical evidence relies on expertise proclaimed with bravado.

An appeal to hypothetical evidence appears in the work of Ranko Bugarski, a professor of linguistics who characterized himself as “born and bred in Sarajevo” (Bugarski 2004, 16). Bugarski declared in 2012 that in Bosnia “the Muslims, Serbs and Croats all spoke the same language, with occasional minor variations which could not possibly affect communication” (Bugarski 2012, 228). He justified his lumper assertionism on the grounds that “deep-seated similarities, in terms of structure and everyday lexicon frequently reaching the point of identity” outweighed “the structural and lexical differences among these variants” (Bugarski 2012, 232). Bugarski did not, however, actually discuss any structural and lexical features.

Discussing “three variants of Serbo-Croatian” in a 2012 essay, Nudžejma Obralić and Azamat Akbarov, both then teaching at the International University of Sarajevo, adduced only two pieces of linguistic evidence when arranging evidence under four linguistic headings. As differences supposedly “understood and well known by everyone”, they specifically invoked “phonetic / orthographic differences such as mleko / mlijeko” and lexical differences “such as veljača – februar”, (“milk”; “February”). They then declared that “practically there are no grammatical differences” without actually discussing any grammatical features, and asserted that “differences in spelling, syntax and pronunciation are not significant enough to be treated as separate languages” without discussing spelling, syntax, or pronunciation (Obralić and Akbarov 2012, 33).

Hypothetical evidence also features in the work of splitter assertionists. In 1996, the Croatian Academy of Arts and Sciences, Department of Philological Sciences (Hrvatska akademija znanosti i umjetnosti, Razred za filološke znanosti), hereafter HAZU, issued a declaration published in the journal Jezik, a self-proclaimed “Periodical for the Culture of the Standard Croatian Language” (Babić 1996, 161). The HAZU declaration insisted that “the Croatian language is a particular Slavic language when seen from either the linguistic or sociolinguistic viewpoint, and equally from other points of view, such as cultural, historical and political” (HAZU, Razred za filološke znanosti 1995, 162). Rather than presenting linguistic, sociolinguistic, cultural, political, and historical data, however, the HAZU declaration simply alluded to “levels” of linguistic analysis:

[T]he contemporary Croatian language differs in features of phonology, morphology, word construction, syntax, lexicon, and style, and in its Latin graphic system. Therefore, it differs from Serbian on all linguistic levels, so that it has never been possible, reading and writing in ‘Serbo-Croatian’ (‘Croato-Serbian’) has never been possible, and remains impossible to this day (HAZU, Razred za filološke znanosti 1995, 163).

HAZU vehemently rejected any subordinate status for Croatian claiming for it the status of a “language”, but without proposing any linguistic criteria for defining the term. It merely insinuated that the definition depended solely on linguistic criteria, and that linguistic evidence, if gathered, would demonstrate language-defining differences.

Other assertionists dispensed with linguistic subcategories entirely. When neurolinguist Karel Diller claimed that “outsiders all recognize that Serbo-Croatian is one language in spite of the socio-political and cultural factors that would separate them”, he simply alluded to the “linguistic grounds that help us recognize Serbo-Croatian as one language” (Diller 1999, 172). The aforementioned Obralić and Akbarov similarly declared that “according to Slavic as the linguistic discipline, there is no a single reason that the Serbian, Croatian and Bosnian might be considered as separate languages but as variations of one language” (Obralić and Akbarov 2012, 33).

In an interesting variant of the appeal to hypothetical evidence, some assertionists invoke an imaginary unanimous consensus. Splitter assertionist Radoslav Katičić, professor of linguistics at the Universities of Vienna and Zagreb until his death in 2019, proclaimed in 1995 that “we all know there is a special Croatian language” (Katičić 1995, 20). By 1997 he was attributing such knowledge to the general public:

After a lecture of mine, a Croat living abroad ‘hit the nail on the head’. He said in the discussion which followed: ‘We all know that there is a Croatian language of our own! Why is it so difficult to make others see it?’ Well, really, why? Why are some domains of linguistic reality not being taken into account when South Slavic languages are listed and classified? (Katičić 1997, 180).

Katičić made no reference to any linguistic evidence, or even to categories of linguistic evidence. He simply appealed to a supposedly self-evident “linguistic reality”.

Lumper assertionists also invoke an imaginary consensus. Obralić and Akbarov invoked the authority of “most of linguists [sic]” (Obralić and Akbarov 2012, 32), and particularly the opinions of “Prof. Dr. Riđanović” and “Prof. Dr. Dzevad Jahic [sic, Dževad Jahić]” (Obralić and Akbarov 2012, 33). In a 2012 interview, Miloš Kovačević, professor of linguistics at the University of Kragujevac, appealed to the opinion of a generic “linguist and philologist”. He specifically claimed that

any basically competent linguist and philologist knows that […] they are not grounded on any scientific criteria for measuring a language identity, but they stand for ‘language policy’. And nobody considers these – which are languages only in name – to be ‘languages in a linguistic sense’.^[1]

Kovačević evidently intended his “basically competent linguist” to represent all linguists, inaccurately depicted as unanimously sharing the same opinion. The fact that a hypothetical consensus can be adduced by both lumpers and splitters demonstrates, of course, that any appeal to consensus is fundamentally dishonest. Both lumper assertionists and splitter assertionists face innumerable opposing opinions whose existence they must pretend not to notice.

How does the appeal to hypothetical evidence function as a rhetorical strategy? We suggest that it is essentially a bluff. Assertionists hope that hypothetical evidence invoked with sufficient bravado will pass unchallenged. The bluffing hypothesis explains why assertionists tend to present their claims with such unwarranted authority: bluffing is more effective when done confidently. The hypothesis also explains the rhetorical ubiquity of the imaginary consensus: if everybody supposedly agrees, then extensive documentation is no longer necessary.

Assertionist Rhetorical Strategies (3): The Selective Denigration of the Political

Assertionists who acknowledge opposing opinions, finally, often claim that their opponents are tainted by “political” factors or interests. Assertionists employing this strategy present themselves as imparting objective facts, asserting or insinuating that opponents are biased or partisan. The denigration of the political, an incarnation of the ad hominem fallacy, rhetorically depends on a Manichean binary juxtaposing wicked “politics” with veracious “linguistics”. Some explicitly conflate “linguistics” with “science” and its truth claims.

Assertionists denigrating the political typically juxtapose two opposing epistemologies: a “linguistic” one and a “political” one, associating the former with their own position, and the latter with that of their opponents. Andrew Simpson’s linguistics textbook, for example, acknowledged that “nationalist leaders decided to refer to the same set of variants of the single language Serbo-Croatian as three fully distinct languages – Serbian, Croatian and Bosnian”, but insisted that such subdivision “into distinct ‘languages’ has consequently clearly been made on the basis of political not linguistic grounds”. Regardless of political factors, however, Simpson insisted that “from a linguistic point of view […] the speech of Serbs, Croats, Bosnian Muslims and Montenegrins should still be taken to be dialectical variants of a single language” (Simpson 2019, 12). Anita Pandey, a professor of linguistics in Maryland, more concisely listed “Bosnian, Serbian and Croatian” as “examples of languages that are really dialects”, mislabeled “for political reasons” (Pandey 1998, 76). The aforementioned Ranko Bugarski suggested that “Serbo-Croatian might be characterized as one linguistic language in the guise of several political languages” (Bugarski 2005, 322), later arguing that Central South Slavic “may be regarded as one language linguistically but several languages politically” (Bugarski 2020, 165). Bugarski also thought “the objective scientific truth of the matter when stripped of politics and emotions” showed that “on the linguistic and communicational level, Serbo-Croatian can still legitimately be regarded as a single entity” (Bugarski 2004, 18). In these cases, the adjective “political” functions, in effect, as a synonym for “ignorant”, “biased”, or some analogous affront to intellectual credibility.

In other cases, the adjective “political” implies hypernationalism, evoking the specter of violence or genocide. Lumper assertionist Srđan Mladenov Jovanović proclaimed that in “the warring sides in Yugoslavia (Serbs, Croats, Bosnians) did speak (and still do) the same language”, and described splitter assertionism as “an artificial discursive creation of difference led by linguistic elites and heartily supported by political players”. He concluded that “language was, for lack of better words, butchered by the nationalist linguistic elites” (Jovanovic 2017, 157, 158). Denouncing “the imposed unification of these two languages in the former Yugoslavia” (Kačić 1997a, 164), Kačić characterized lumper assertionism as “politically damaging” and responsible for “linguistic violence on the Croatian language” (1997a, 147). In 2022, splitter assertionist Mikhail Sergeevich Xmelevskij of St. Petersburg State University promised to give an objective evaluation of the linguistic status of the “four idioms” in the South Slavic dialect continuum by including the “synchronic and diachronic aspect”, the “naming of idioms by the ethnic groups”, and “the religious factor” (Xmelevskij 2022, 1–2). He concluded that there has never been a common superordinate “language”, suggesting that “the name ‘Serbo-Croatian language’ was artificially created during the period of the SFRY [socialist Yugoslavia] for purely political and ideological purposes, aimed at erasing linguistic and national differences between Serbs and Croats” (Xmelevskij 2022, 11). The aforementioned HAZU declaration saw lumper assertionism as a plot “to destroy all Croatian peculiarities and the Croats as a people” (HAZU, Razred za filološke 1995, 162).

Perhaps the most prominent scholar to criticize the “political” is the linguist Snježana Kordić who, though educated in Osijek and Zagreb, has spent much of her career in Germany. In a 1997 interview, she quoted Robert Dixon to argue that

[t]he word ‘language’ is used in at least two rather different senses. There is the political sense where each nation or tribe likes to say that it speaks a different language from its neighbors. And there is the linguistic sense where two forms of speech which are mutually intelligible are regarded as dialects of a single language (Kordić 2002, 104)

Any approach taking this “political sense” into account, according to Kordić, “is not suitable for science” (Kordić 2001, 194), since it “would mean that it is possible that new foreign languages, for example Pulan or Slavonian, could emerge tomorrow, because everything would depend only on the inhabitants of Pula or Slavonia to state that they speak their own language” (2001, 194). Some scholars might find the sudden emergence of claims to a Pulan or Slavonian “language” an intriguing development worth investigating, but Kordić insisted on relying exclusively on “system-linguistic, genealogical, and communicative criteria” (Kordić 2002, 107-8).

Several assertionists explicitly claim that their position is simply “truthful” or “factual”. The aforementioned Radoslav Katičić, a splitter assertionist, proclaimed that lumper assertionism “blocks the way to valid knowledge”, because it is “simply not true, and no objective and serious understanding of Croatian language reality can be founded on it” (Katičić 1995, 17, 21). The notion of unitary Serbo-Croatian analogously provoked splitter assertionist Stjepan Babić to exclaim: “Whoever claims that it exists or that it existed does not know the facts, or if he knows the facts, he does not know which theory to apply to those facts” (Babić 2010, 151). Lumper assertionists make similar epistemological claims. Andrej Stojanović insisted in 2014 that Bosnian and Croatian cannot be labeled as “languages”, but only as the Zagreb and Sarajevo “variants” of a “Serbian literary language”, since “there is a discrepancy between the formal name and the factual ethnic affiliation” (Stojanovich 2014, 111, emphasis added).

While assertionist truth claims are typically problematic, assertionist accusations of political bias are often correct. Assertionism is an inherently political stance, and both lumper assertionism and splitter assertionism have political ramifications. As Ronelle Alexander put it,

When, and how, does one decide that a sub-variety of a language – whether it is called a dialect or a regional variant – qualifies as a separate language? One might think this is a purely linguistic matter, and indeed linguists are frequently called upon to make these judgement calls. At heart, however, the issue is a sociocultural and (above all) political one; this is due, of course, to the bond between language and national identity (Alexander 2013, 343).

Assertionists who denigrate their opponents as “political”, however, prefer to ignore the political ramifications of their own preferred stance. In the context of assertionist arguments, the word “political” typically serves as a pejorative epithet, irredeemably tainted with bias. Rhetorically, therefore, assertionists cannot afford to admit that their own stance has any political implications.

The denigration of the political and the desire to transcend politics combined in two extraordinary claims made by Peter Trudgill, a respected sociolinguist with five honorary doctorates.^[2] In 2017, Trudgill signed the “Declaration on the Common Language” (Kamusella 2021, 179–82), which proclaimed, among other things, that “a common language is used in Bosnia and Herzegovina, Croatia, Montenegro and Serbia”. Justifying his decision to sign the declaration in the New European magazine, Trudgill declared an imaginary consensus, claiming that “linguistic scientists are agreed that BCSM is essentially a single language”, even if “some nationalists are doing their best to make the four varieties more distinct from one another by artificially introducing differences where none existed”. As a lumper assertionist, Trudgill dismissed splitter assertionism as “rather silly”. He denigrated the political by supporting “defenders of linguistic common sense […] in their struggle against the linguistic unreason of the nationalists.”^[3]

Trudgill’s assertionism seems all the more remarkable given the eloquence with which he had previously articulated a principled agnosticism. Trudgill’s 1995 textbook declared that “there can be no linguistic answer to whether Serbian and Croatian are one language or two. The answer is a political and cultural one.” As a theoretical linguist, Trudgill adduced Central South Slavic to illustrate a more generalizable phenomenon:

Is Macedonian really a language? Is there a Bosnian language which is distinct from Croatian and Serbian? Are Moldovan and Rumanian the same language or not? Are Flemish and Dutch one language or two? Is Corsican a dialect of Italian or not? Is Swiss German actually a separate language?

Trudgill placed all such questions firmly beyond the reach of linguistics as a discipline: “there is no way we can answer these questions on purely linguistic grounds. Ironically, it seems that it is only linguists who fully understand the extent to which these questions are not linguistic questions” (Trudgill 1995, 145). Significantly, Trudgill justified both his agnosticism and his assertionism with reference to an imaginary linguistic consensus. He defended lumper assertionism on the grounds that “linguistic scientists are agreed”, and justified agnosticism on the grounds that “linguists” understand the insufficiency of “purely linguistic grounds”. Confronted with such contradictory invocations of linguistic consensus, however, nonlinguists must be forgiven skepticism that even prominent linguists can depict the opinions of their discipline accurately.

The sociocultural and political context exists whether or not linguists choose to be cognizant of it. Our analysis suggests, however, that assertionists denigrating the political disingenuously choose to be cognizant of the political context only when attacking their opponents. They ignore the political ramifications of their own stance, which they present as “linguistic” and untainted by “politics”.

Conclusion and Discussion

This article has analyzed rhetorical strategies in assertionist linguistic arguments. We did not take sides between splitters and lumpers, but analyzed assertionist arguments from an epistemological perspective. Exploring how assertionists justified their claims revealed three main rhetorical strategies: the avalanche of trivia, the invocation of hypothetical evidence, and the denigration of the political. While this study examined recent discussions of Central South Slavic, we suspect similar rhetorical strategies would characterize L/D disputes in other times and places.

All these rhetorical strategies rely on a performance of authority. The avalanche of trivia demonstrates erudition and mastery of technical language. Hypothetical evidence feigns an appeal to scholarly consensus. The denigration of the political essentially disparages alternate views. Whatever the rhetorical appeal of these strategies, however, all three are epistemologically flawed. Indeed, we suggest these rhetorical strategies serve in part to hide the flawed epistemological foundations of assertionism, whether of the lumper or splitter variety.

Assertionist arguments fail to address the fundamental epistemological issue: What distinguishes a “language” from a “dialect”? Many theoretical linguists have raised ontological doubts that the terms “language” and “dialect” are useful. Assertionists must provide some definition of the ontological objects whose existence they so vehemently assert. Of the scholars examined above, only Bailyn made any attempt to provide such a definition, and he did not actually apply his own definition in practice.

The fundamental problem with assertionism lies in the assumption that the dichotomy can be meaningfully analyzed on purely linguistic grounds. For around two hundred years, nationalist thinking in its great variety has transformed the entire world, including Eastern Europe and the Balkans. In Eastern Europe, however, nationalists have tended to espouse what Tomasz Kamusella memorably called “the normative isomorphism of language, nation and state”, and later summarized with “handy algebraic-like equation language = nation = state” (Kamusella 2022, 212, xiv, 66), an ideal Jan Blommaert (2006, 247) alternately characterized as the “Herderian cocktail of one language – one culture – one territory”. The ubiquity of normative isomorphism makes assertionism an inherently political stance in the Central Slavic Balkans. Since the existence of a “language” implies the existence of a nation, proclaiming or denying the existence of a “language” then affects the legitimacy of nationalist claims. Statements about the legitimacy of a nationalist claim, furthermore, cannot be meaningfully analyzed if scholars consciously ignore politics as a matter of principle. The denigration of the political does not merely impede analysis, it prevents it.

Nevertheless, the longing for a purely linguistic analysis has inspired curious fantasies of ignoring relevant political realities. John McWhorter once wistfully asked of Central South Slavic “if all these dialects were spoken in some uncharted region rather than artificially corralled into ‘countries,’ where would you draw the line between one ‘language’ and another one?” More generally, he fantasized about how dialect continua would be subdivided “if the world had not been partitioned into countries” (McWhorter 2001, 74, 84). McWhorter’s extraordinary desire to pretend state power does not exist may be atypical, but numerous other linguists have unhelpfully denigrated political factors as a matter of principle (Ahlquist 1997, 28; Myhill 2006, 10).

The impulse to ignore politics may in turn derive from the desire of linguists to attain for their discipline the prestige of the natural sciences. Chomsky, for example, has wondered whether the study of language, “can be reduced to physics […] or whether physics can be extended in some natural way to accommodate it” (Chomsky 1988, 173). Perhaps these aspirations partially explain Chomsky’s unusually contemptuous attitude toward the study of the L/D dichotomy, which he once declared irrelevant to “any approach to the study of language or meaning” (Chomsky 2000, 31). As concerns the L/D dichotomy, Chomsky is not an assertionist but an agnostic. As fellow agnostics, however, we observe that Chomsky’s dismissive attitude fails to shed light on the dynamics of the L/D dichotomy in a given political context.

While agnostics are right to be skeptical of assertionist claims, we suggest that agnostic skepticism does not itself constitute a strategy of analysis. Agnostics all too often simply dismiss the L/D dichotomy as unworthy of study. In a passage criticizing assertionist claims that mutual intelligibility can resolve lumper/splitter disputes, Charlotte Gooskens and Vincent van Heuven once wrote that “the choice of whether two language varieties are dialects of the same language is basically a practical, political matter and not a question with any scientific import or theoretical status” (Gooskens and van Heuven 2021, 57). We find it remarkable that any scholar can dismiss the “practical” and the “political” as lacking scientific and theoretical importance! Even linguists who do not personally wish to study the politics of the L/D dichotomy should accept that it raises questions of legitimate scientific curiosity.

Linguists seeking analytical strategies for analyzing the L/D dichotomy, or lumper/splitter disputes, should look to other disciplines for inspiration. The L/D dichotomy is a taxonomy, and thus can be studied a particular type of classification. The politics of classification has important social implications, particularly in the construction of ethnic and national groups, which are often imagined through linguistic categories (Maxwell 2018). Linguists analyzing the L/D dichotomy should thus employ theoretical tools developed for studying the social construction of nationalism. In an influential article, Brubaker and Cooper observed that “appeals and claims made in the name of putative ‘nations’ […] have been central to politics for a hundred and fifty years. But one does not have to use ‘nation’ as an analytical category to understand and analyze such appeals and claims” (Brubaker and Cooper 2000, 5). Similarly, appeals and claims made in the name of putative “languages” (or putative “dialects”, “variants”, etc.) can be studied without treating the terms “language” or “dialect” as analytical categories. Social constructivism thus provides agnostics with a methodology for studying the L/D dichotomy.