In Between Description and Prescription: Analysing Metalanguage in Normative Works on Dutch 1550–1650

Abstract

This paper is the first to perform a systematic quantitative analysis of the arguments used to motivate selections in grammatical entries from normative works on Standard Dutch written between ca. 1550 and 1650. Thus, it aims to obtain insight into what language ideologies were characteristic of this early modern period, what these reveal about how Standard Dutch took shape in its initiating phase, and what the differences are between the codification of Dutch in the early modern period (16th/17th century) and the (post)modern period (20th/21st century; analysed in earlier studies). Although certain issues within the annotation method need to be addressed in future research, the results indicate that the following principles were particularly characteristic of the early modern period: for Dutch to be a good language in terms of its grammar, it ought to differentiate, display consistency, mirror Latin and Greek, and reflect the use of certain authorities. These linguistic principles form the roots of the part of the Dutch standard language ideology (SLI; which, as previous research has shown, came into existence in the decades around 1800) that connects ‘language’ with ‘norm’ and that bestows value on the language’s regularity. However, the additional connection to social identity, that forms a second and crucial part of the SLI, played no major part in the arguments used in this time period yet. Moreover, two important differences between the early modern period and the (post)modern period were found: (1) the latter period showed a higher degree of consensus and therefore of canonisation of the normative discourse than the former period; (2) the nature of the metalanguage used in normative publications was explicitly prescriptive in the later period but mostly ostensibly descriptive/implicitly prescriptive in the earlier period. This indicates that, in terms of the metalanguage used, the normative discourse in the formative period of Standard Dutch was in between description and prescription.

1. Introduction

During the Renaissance, several Western European standard languages were devised, including Standard Dutch. From around 1550 onwards, normative works such as spelling guides, dictionaries, and grammars were written in which tentative rules for this new standard language were formulated. Literary writers and grammarians who engaged in this normative discourse dismissed certain form variants, suggested others for use in the new standard, and documented these choices. In doing so, they selected and codified the form of the standard language on all its levels: pronunciation, spelling, grammar, and lexicon (Haugen 1966; Milroy and Milroy 1999). These selection and codification processes have been thoroughly examined (e.g., Van der Wal 1995; Van der Sijs 2021; Zwaan 1939; Dibbets 1995; Rutten et al. 2014b). As was illustrated by Van der Sijs (2021), the choices made in this period formed the foundation of modern Standard Dutch. This period can therefore be seen as formative for the Dutch standard language.

If we are to view the normative works written in this period as a normative discourse in a more Foucaultian sense of the word (e.g., Foucault 1966), the question rises whether this ‘debate’ was also formative in terms of the language ideologies it exuded (cf. Blommaert 1999, pp. 10–11). Language ideologies, in their broadest sense, have been defined as ‘a set of beliefs about language articulated by users as a rationalization or justification of perceived language structure and use’ (Silverstein 1979, p. 193; repeated in Woolard 1998, p. 4). One particular type of language ideology attested in previous research is the standard language ideology, which has been defined as ‘a set of beliefs about language correctness and a general intolerance towards non-standard variants and varieties’ (Deumert and Vandenbussche 2003b, p. 461; paraphrasing Milroy and Milroy 1999). Often, language ideologies are seen as views on the role of language in society or, as Kathryn Woolard defines them, as ‘representations, whether explicit or implicit, that construe the intersection of language and human beings in a social world’ (Woolard 1998, p. 3). As the study of language ideologies thus connects the linguistic to the social, much of the research in this field is devoted to the connection between language and power and to related practices such as identity-making and nation-building (e.g., Irvine and Gal 2000; Blommaert 1999; Gal and Woolard 2001b; Kroskrity 2000; Schieffelin et al. 1998). The same counts for the standard language ideology, which, to many researchers, is intrinsically connected to those systems of power. Grondelaers & Van Hout, for example, (re)define the standard language ideology as ‘a normative ideology imposed and sustained by institutions such as (formal) education and the media, but maintained by (silent) agreement between the language users’ (Grondelaers and Van Hout 2011, p. 114). Since, in the Dutch-speaking language area, language only became connected to nation-building and consequent institutionalisation and language policy-making around 1800, Rutten (2016, 2019) convincingly argues that this period marks the emergence of the Dutch standard language ideology. This raises the question whether the preceding early modern period, which did show the emergence of Standard Dutch, had a language ideology to speak of and, if so, how it can be characterised, if not as a standard language ideology.

Previous research indicates that the early modern period can be said to have had language ideologies of its own. Hüning et al. (2012), for example, argue that, with the ‘discovery of language’ (Burke 2004, p. 15ff; cited in Hüning et al. 2012, p. 9), meaning that language became a countable, discrete, and nameable entity, came the emergence of a ‘correctness ideology’ in the early modern period (Hüning et al. 2012, pp. 9, 22). They oppose this ‘language and norm’ type of ideology with the later ‘language and nation’ ideology, which marks the ‘instrumentalization of ‘correct languages’ as vehicles of identity politics and politics of democratization in the 18th/19th century’ (Hüning et al. 2012, p. 22), i.e., the standard language ideology as understood by Rutten (2016, 2019). Rutten, too, acknowledges the existence of what he calls, following Watts (2011), ‘language myths’ in the Dutch language area before 1800 and focusses on the ‘myth of neutrality’ (Rutten 2016, 2019). Although this ‘myth’ and the ‘correctness ideology’ from Hüning et al. (2012) were not institutionalised and therefore did not constitute a standard language ideology, these early modern ideals do constitute a language ideology in the broader definition from Silverstein (1979) cited above, i.e., a set of beliefs about language that serves as rationalisation or justification of its structure and use.

Notwithstanding these valuable analyses, our knowledge of language ideologies in the early modern period and, in particular, in the formative period of Standard Dutch deserves to be expanded. As (language) ideologies are notoriously difficult to access, for they can be expressed explicitly but more often remain implicit (Woolard 1998), Deumert and Vandenbussche propose, amongst other approaches, comprehensive corpus studies of secondary sources in which ‘standardizers’ outline and defend their proposals (Deumert and Vandenbussche 2003b, pp. 461–62). For the early modern period, then, this would entail an analysis of the normative discourse or ‘debate’ (cf. Blommaert 1999) mentioned in the opening paragraph of this introduction.

Even though Deumert and Vandenbussche (2003b) seemingly suggest to focus only on the prefaces of these grammars and dictionaries, in which the main principles of the normative works are outlined, researchers such as Kostadinova and Van der Meulen have shown that a different way to operationalise and, importantly, quantify the study of language ideologies is analysing the arguments used to motivate specific linguistic choices in prescriptive utterances (e.g., Van der Meulen 2020; Kostadinova 2018). Even though some earlier research did take prescriptive argumentation into account, they mostly did so only as a factor related to the main topic of the study (e.g., Anderwald 2012; Ebner 2016; Chapman 2019; see Van der Meulen 2020 for a review of these studies; see also Kostadinova 2018, pp. 101–5). It is particularly the study of arguments in their own right—connecting epithets to the values they stem from, mapping out annotation schemata for their categorisation, and providing analyses of patterns in their use as indicators of language ideologies—that is promising in this respect. Moreover, as the micro-selection process for Dutch can be characterised as polycentric (Deumert and Vandenbussche 2003a, pp. 4–5), this type of analysis seems particularly apt for determining language ideologies in the normative discourse on Standard Dutch.

The quantitative analysis suggested by Van der Meulen (2020) focusses specifically on the linguistic values underlying arguments that are used to justify linguistic micro-selections in prescriptive utterances. This type of utterance, as Moschonas (2020) claims, regularly takes on the form ‘one should neither say nor write X; one should say and write Y instead, because Z’. ‘Z’ represents what Moschonas calls the ‘explicative’: an argument motivating the micro-selection presented. These arguments, although greatly varying in formulation and type, can be categorised into different types corresponding to the values underlying them (Van der Meulen 2020). For example, when Petrus Leupenius, who published his Aanmerkingen op de Neederduitsche Taale ‘Comments on the Dutch Language’ in 1653, discusses the use of ‘&c’ or ‘etc.’, he states that he wants to banish these forms, as they are van buiten … ingevoert, … onse taale niet eigen, maar van anderen ontleent ‘imported from outside, not an inherent part of our language, but borrowed from others’ (Caron 1958, p. 16; Leupenius 1653, p. 21). The value this explicative is based on is that language should be pure, free of influences of other languages: a puristic argument (cf. Van der Meulen 2020; Thomas 1991). Systematically and quantitatively analysing patterns in the use of (types of) such arguments—or, as Van der Meulen calls them, ‘epithets’—and the values underlying them can provide valuable insight into the language ideologies of grammarians shaping the standard language. For instance, if our results show that Leupenius often uses puristic arguments, this indicates that he believed that, for Standard Dutch to be a good language, it ought to be free of foreign influence, and if more grammarians in this time period regularly use this type of argument, then this substantiates the claim that purity was part of the language ideology of the time.

Studies of this kind have thus far solely focussed on language advice publications from the 20th and 21st century; publications from a period that, for Standard Dutch, can be characterised, ideologically speaking, as having the standard language ideology, and that has even been argued to show demotisation or re-standardisation processes from the end of the 20th century onwards (e.g., Grondelaers and van Hout 2011; Grondelaers et al. 2016). Normative works written during earlier centuries have not been subject to such systematic analyses yet. We therefore cannot determine whether Van der Meulen’s (2020) conclusion that ‘the Dutch language should be pure’ and ‘the Dutch language should be grammatical, should obey by the rules’ are the dominant linguistic values applies to the 20th and 21st century alone or to Dutch prescriptivism in general.

This paper intends to fill this gap by applying the annotation schema developed by Van der Meulen (2020) to the metalanguage encountered in normative works on Standard Dutch written between ca. 1550 and 1650. Previous qualitative analyses of this time period indicate that the principles that the early modern grammarians based their tentative rules on differ depending on the level of language the rules address; the main principles underpinning spelling rules, for example, have been identified as cultivated pronunciation, uniformity, and etymology (e.g., Van der Sijs 2021, p. 208), whereas lexical choices were made mostly in puristic terms (e.g., Van der Wal 1995, pp. 28–29), and morphological choices often on the basis of a Latin ideal and the ‘one form, one function’ principle (e.g., Dibbets 1995, pp. 331–33; Van der Sijs 2021, pp. 370–71). The current study therefore solely focusses on arguments used for, and values underlying, grammatical choices. The research questions addressed in this paper are: What arguments are used to motivate the selection and codification of certain grammatical forms in normative texts on Dutch from the early modern period? What linguistic norms and values do these appeal to, and how do they fit into Van der Meulen’s annotation schema? Based on this analysis, what language ideologies were present in the early modern period, and what do these reveal about how Standard Dutch took shape in its initiating phase? Our quantitative approach provides a novel perspective on this already thoroughly studied material and thereby significantly enhances our understanding of the codification of Dutch in the Early Modern period and the language ideologies underlying it.

Moreover, as Van der Meulen’s annotation schema was developed to annotate evaluative epithets in Dutch prescriptivist/language advice publications from 1917 to 2016 (Van der Meulen 2020), applying this schema to the earlier period will provide additional insight into the differences between the codification and micro-selection of Dutch in the 16th and 17th century and that of the 20th and 21st century. Although we refer to the former period as the formative period of Standard Dutch, we would here like to stress that it is not periods of standardisation we intend to compare but time periods: early versus (post)modernity. All too often, Haugen’s (1966) model of standardisation is misunderstood as linear, with clearly delineated and subsequent periods in the standardisation process (see, e.g., Rutten et al. 2014a, pp. 7–8; Joseph et al. 2020; Rutten and Vosters 2020 for more thorough discussions of this misunderstanding). By comparing these two time periods, we do not intend to perpetuate this misunderstanding but merely want to do justice to the fact that, as Deumert and Vandenbussche (2003a, pp. 9–10) put it, ‘what “standardizers” had in mind in the seventeenth century differs from their goals during the nineteenth and twentieth centuries’. The comparison performed in this paper should therefore not be viewed as a comparison of two different periods of standardisation—for selection, codification, diffusion, and acceptance were happening in all centuries studied—but as a comparison of two different historical periods, each with their own context and characteristics (also in line with Blommaert 1999’s ‘historiography of language ideologies’). Ideally, this paper will form the starting point of an overarching framework that can be used to analyse and compare all time periods and languages in terms of the arguments used to motivate linguistic choices; one framework that allows for comparison yet leaves room for specific characteristics. Such an approach would considerably further our understanding of the characteristics of each language or period and of the differences and similarities in standardisation processes.

To answer the research questions, this paper is set up as follows: Section 2 first explains the selection of materials and the annotation method. In Section 3, the results are presented and analysed, both in terms of the patterns in the use of metalanguage in the 16th and 17th centuries itself, as well as in comparison to the later centuries. These results are discussed in Section 4, followed by a summary in Section 5.

2. Materials and Methods

The following subsections explain the methodology used to obtain insight into what language ideologies were present in the early modern period in chronological order: Section 2.1 explains the selection process of normative works included in this analysis; Section 2.2 describes how these works were processed for analysis and what entries were selected for further annotation; Section 2.3 presents the annotation schema developed by Van der Meulen (2020), which was used to identify types of arguments in these entries, and explains how this schema was adapted for the purposes of this study; and finally, Section 2.4 explains how the annotated results were analysed.

2.1. Building the Corpus of Codification of Standard Dutch ca. 1550–1650

Several inclusion criteria were used for the selection of normative works for this study. Firstly, only works written in the first century of the formative period of Standard Dutch were selected. This period roughly coincides with the early modern period, ranging from around 1550 until the publication of the first officially sanctioned spelling guide (Siegenbeek 1804) and grammar (Weiland 1805) in the early 19th century. As our focus is on the shaping of Standard Dutch in its initiating phase, only works written in the first century of this period were included: between ca. 1550 and 1650. The latter date was prompted by the fact that no new normative grammars were written in the second half of the seventeenth century, thus providing a natural cut-off point (Van der Wal 1995, p. 40; Van der Sijs 2021, p. 376). Moreover, the focus is on works written, and not (necessarily) published, within this time period, because our interests lie in the ideas taking shape in this early period of standardisation, which is better reflected in writing than in publication dates. Secondly, works written for second language speakers and for educational purposes were excluded, and only works written in the Northern Netherlands were included. This was done to accommodate comparability between works but the latter also because, as the Southern Netherlands came under Spanish rule after 1585, Standard Dutch was mostly constructed in the Northern Netherlands (Van der Wal 1995, p. 29; Van der Sijs 2021, p. 48; Willemyns 2003). Finally, only works defined as grammars were included for analysis, thus excluding, e.g., spelling guides and dictionaries (following divisions made in Van der Sijs 2021). This was done for two reasons: first, again, for reasons of comparability and, second, to be able to focus our investigation on prescriptive utterances on grammatical topics.

In total, ten normative works, written by nine different authors, met these criteria and were thus selected for examination in this study (see Appendix A). One author wrote and published more than one book: Christiaen van Heule. His 1633 work De Nederduytsche spraeck-konst Ofte Tael-beschrijvinghe ‘The Dutch Grammar or Language Description’ was presented as an improved edition of his De Nederduytsche Grammatica ofte Spraec-konst ‘The Dutch Grammar’ (1625), but it differs from it in such fundamental ways that it has been included as a separate normative work (Caron 1971b). For all other works, later editions were examined for differences in metalanguage. When differences were found, entries from other editions were added to the study (following Sundby et al. 1991, p. 148). This resulted in the inclusion of two entries from the 1649 edition of Ampzing’s ‘Nederlandsch tael-bericht’ and two entries from Hooft’s Waernemingen op de Hollandsche tael from its 1700 edition (Van Hoogstraten 1700, pp. **6r–**8v).1

2.2. Selecting Entries

In order to find the utterances containing arguments for grammatical choices, all works were first analysed in their entirety and subdivided into entries. Following Van der Meulen (2020), the internal structure of the works determined what made up an entry, so entries ranged from large paragraphs to smaller units such as (parts of) sentences. Our material presented particular challenges in this regard, as what demarcates an entry is much less straightforward in this early period of standardisation, for reasons extensively enumerated in Poplack et al. (2015, pp. 23–25), e.g., the (inconsistent) use of outdated terminology, the lack of subsections, the structure following that of classical grammars instead of the language in question, and the resulting scattering of information on one phenomenon over multiple sections. Due to these challenges, merely following the internal structure did not suffice; the contents of the text additionally determined the entry demarcations. This is illustrated in Figure 1; despite its continuous appearance, this page contains eleven different entries. An entry can therefore be defined as a unit discussing roughly one topic. The data-driven nature of our approach thus somewhat circumvented the demarcation challenges, as it ensured that nothing was overlooked, but the division process remained challenging. We return to this point in Section 4.

In the second step, all entries were classified into six groups based on the topic they discussed. Since our interests lie in the grammatical realm, only entries discussing a grammatical topic were selected for analysis, excluding roughly 600 entries on spelling, 70 on pronunciation, 30 on lexicon (all more general comments concerned with purism), 30 on other related topics, and 90 more general remarks. In their ‘Dictionary of English Normative Grammar 1700–1800’ (Sundby et al. 1991, pp. 4–12), Sundby, Bjørge, and Haugland analyse the definition of ‘grammar’ as expressed in several eighteenth century grammar books and the challenges this poses to researchers with a current understanding of the topic (in line with the challenges from Poplack et al. 2015 mentioned above); the interpretations of the concept diverge so greatly, with consequences for what was discussed, where in the book and in what way, that ‘grammar’ cannot be clearly delimited (Sundby et al. 1991, pp. 13–14). This similarly counts for the grammars from the 16th and 17th century included in this study. We have therefore followed Sundby et al. (1991) in allowing more scope for lexical data than modern handbooks of grammar would. Fundamental to the selection, and the vast majority of it, were entries on the parts of speech, morphological inflection, and syntax, and excluded were entries that clearly concerned topics of spelling, lexicon, and pronunciation only, but we additionally included sporadic entries on spelling that intersected with, for example, inflection or clisis and, also, entries on more lexical topics, such as lexical morphology (e.g., compounding and derivation) and lexical choices (e.g., the distinction between na ‘after’ and naar ‘to’), as they could be interpreted as having grammatical bearings. This leads to grammatical entries on a wide range of topics being included in the selection for this study. In the end, 1237 such broad grammatical entries were included in this study.

In the third step, these 1237 entries were annotated for type of entry. This annotation was not present in Van der Meulen (2020) but was added to this study to accommodate the peculiarities of the time period in question, which deserve some elaboration. Whereas all entries from the 20th and 21st century either explicitly or implicitly took a stance on the acceptance of optional variability within the usage items discussed (Van der Meulen 2020), this was pertinently not the case with the entries in normative works from the 16th and 17th century. These often contained only definitions or paradigms of the topic discussed; for example, Kók (1649), on the topic of second person personal pronouns, just gave the following paradigm:

Een-voudt.	‘Singular.
N. Du.	Nominative: Du.
B. Dijns óft Dijnes, óft Dijner.	Genitive: Dijns or Dijnes or Dijner.
G. Dy.	Dative: Dy.
A. Dy.	Accusative: Dy.
Meêrvoudt.	Plural.
N. Ghy.	Nominative: Ghy.
B. Uwer.	Genitive: Uwer.
G. U.	Dative: U.
A. U.	Accusative: U.’
(Kók 1649, p. 21; Dibbets 1981, p. 31)

These types of entries form a problem for the current study, as they do not contain an explicit linguistic judgement but could still, e.g., by processes of erasure (cf. Irvine and Gal 2000), contain an implicit one (similarly argued in other works that deal with the problematic dichotomy between description and prescription, e.g., Amorós-Negre 2008; Greenbaum 1988; Pascual Rodríguez and Prieto de los Mozos 1998, who name this prescripcion encubierta, ‘covert prescription’). With his positing of Du for singular and Ghy for plural, for example, Kók implicitly dismissed the use of the latter form for the singular, a use well-established by this time (Van der Sijs 2021, pp. 417–18). This raises the question whether such entries should be considered Ipse dixits, entries that pose a judgement on a particular linguistic phenomenon without an argumentation to motivate it.

However, it proved impossible to distinguish such ostensibly descriptive yet covertly prescriptive judgements from entries that truly did not contain a judgement on variation (see also Chapman 2021). For example, multiple grammars (Van Heule 1625, 1633; Spiegel 1584; Leupenius 1653) include the rule that, in questions, the word order changes: ick spreeck, Jan redenkavelt ‘I speak, John polemicises’ becomes spreeck ick? redenkavelt Jan? ‘Do I speak? Does John polemicise?’ (Spiegel 1584, p. 94; Dibbets 1985, p. 275). Does the inclusion of this rule mean that some Dutch speakers were posing questions without changing the word order, which these grammarians wanted to proscribe? Or was this word order a given for the Dutch language, with no variation to speak of, and did this rule merely describe how Dutch works, perhaps only included because one had to discuss word order in questions in classical grammars?

Multiple ways to solve this issue were considered. The first was to ascertain the (non-) existence of prior variation. In theory, this should nowadays be possible by performing an analysis on the topic in a usage corpus that contains language use prior to the codification of the normative entry. In practice, however, due to our data-driven approach, the topics discussed were too varied and, sometimes, as in the case of word order, hard to operationalise in corpus studies, which would have made such analyses very time-consuming, if not impossible.2 The second and preferred remedy was less time-consuming and consisted of selecting only entries that contain explicit linguistic judgements for further annotation, thus excluding entries such as Kók’s paradigm and the word order rules above. These judging entries, which received the tag argumentation, were 490 in total (see Appendix B for the full annotation schema and definitions of the different categories). This solution is in line with the focus of the current study: the annotation of evaluative arguments and the analysis of their patterns. However, as the absence of argumentation is also a factor to be taken into consideration, the other types of entries, containing 747 entries in total, are also taken into account in further analyses and in the interpretation of the results (see Section 3 and Section 4).

2.3. Annotating Arguments

The 490 entries that contained a variation judgement, labelled argumentation, were further annotated according to Van der Meulen’s (2020) annotation schema for annotating evaluative epithets in Dutch prescriptive works from 1917 to 2016 (inter-annotator agreement score (Cohen’s kappa): κ = 0.83). We used this annotation schema as a starting point, firstly, to facilitate a comparison with that study, and, secondly, because we would ideally like to arrive at one framework to analyse and compare all time periods and languages in terms of the arguments used to motivate linguistic choices to be able to pinpoint the characteristics of each language or period and thereby the differences and similarities in standardisation processes. Therefore, the annotation schema used in this study was kept as similar as possible to the original, maintaining its classification into main categories subdivided into lower-level categories (see

Table 1. Annotation schema for arguments with underlying values, adapted from Van der Meulen (2020) (newly added tags in bold).

Top-Level	Lower-Level	Subcategory	Value
Authority	_Unspecified		Good language is determined by what an authority says
	_Dictionary		Good langue is determined by what a dictionary says
	_Frequency		Good language is determined by what a number of people say
	_Grammar		Good language is determined by what a grammar or grammarian says
	_Literary		Good language is determined by what an author says
	_Socio		Good language is determined by what a certain group of people says
Purity	_Unspecified		Language should be pure, free of influences of other language
	_Anglicism		Language should be free of English influence
	_Gallicism		Language should be free of French influence
	_Germanism		Language should be free of German influence
	_Latinism		Language should be free of Latin influence
	_Other_Language		Language should be free of the influence of another language
Quality	_Unspecified		Language should be qualitative
	_Beauty		Language should be beautiful
	_Care		Language should be well taken care of
	_Differentiation		Language should differentiate
	_Ease		Language should be easy
	_Effect		Language should have good effects
	_Logic		Language should be logical
	_Quantity		Language should be used in the right quantities
System	_Unspecified		Language should adhere to the system
	_Analogy	_Unspecified	It should (not) work the same as in another (unspecified) language
		_Dutch	It should (not) work the same as in another Dutch construction
		_German	It should (not) work the same as in German
		_Greek	It should (not) work the same as in Greek
		_Hebrew	It should (not) work the same as in Hebrew
		_Latin	It should (not) work the same as in Latin
	_Grammatical		Language should be grammatical
	_Immutable		Language should not be changed
	_Nature		Language should be used according to the nature of the language
Use	_Unspecified		Good language is determined by what is done
	_Authority		Good language is determined by what an authority does
		_Dictionary	Good language is determined by what a dictionary does
		_Grammar	Good language is determined by what a grammar or grammarian does
		_Literary	Good language is determined by what an author does
	_Bible		Good language is determined by what the Bible does
	_Frequency		Good language is determined by what a number of people do
	_Historical		Good language is determined by what was done in the past
	_Socio		Good language is determined by what a certain group of people says
Variety	_Unspecified		A specific variety of language is the right one
	_Genre		Language should be used in the proper genre (e.g., poetry vs. prose)
	_Geographic		The language spoken in certain geographic regions is right/wrong
	_Mode		Language should be used in the proper mode (i.e., written vs. spoken)
	_Register		Language should be used in the proper register (i.e., formal vs. informal)
	_Standard		Language should be used in accordance to the standard
Ipse dixit			No argument is given
Optional variation			Language should (not) contain variation
Other argument			There is some other reason why language is good or bad

). For reference, Table 1 contains a brief description of the values underlying the categories in the schema.3 Examples of the different types of arguments are given in Section 3 of this paper. The details of the schema will be discussed where relevant; for a full explanation of this categorisation, see Van der Meulen (2020).

We have, like Van der Meulen (2020), tagged lower-level categories in principle, in order to be as specific as possible, the only exception being unspecified or uninterpretable arguments; such arguments received the tag _Unspecified within the appropriate top-level category.

Despite wanting to stay as close as possible to the original, adaptations had to be made to accommodate the annotation schema to the earlier period of the standardisation of Dutch. These modifications included two changes at the top level. First, a new separate tag was added: Ipse dixit (cf. Van der Meulen 2018). This label was applied when an opinion was expressed without argumentation to motivate it. Spiegel provides an example of this argument type when he advocates the use of zich ‘himself’ over hem ‘him’ only by stating that this is what it should be na myn achting ‘in my opinion’ (Spiegel 1584, p. 84; Dibbets 1985, p. 255). Second, the tag Optional variation was put to a different use than in Van der Meulen (2020). There, this tag was used when a judgement was made with no explicit argumentation. We use the tag Ipse dixit for such utterances instead, because it is not always the case that variation is suppressed in entries that contain no arguments (de Vos and van der Meulen 2021). Moreover, the existence of optional variability was sometimes explicitly used as an argument justifying a linguistic choice in the early modern period. An example can be found in Spiegel (1584, p. 95), Dibbets (1985, p. 277), when he claims it appropriate om metter tyd … wat meer verandering in te voeren ‘to, over time, … introduce some more variation’ in the order of adjectives and nouns (e.g., ‘brother mine’ next to ‘my brother’). This prompted a more appropriate use of the tag Optional variation.

On the level of the more specific lower-level categories, the following changes were made. First and foremost, eight labels were added; see the bold categories in Table 1. In addition, the lower-level category Quality_Quantity was slightly stretched in its interpretation: in Van der Meulen (2020), this tag solely referred to using the right quantity of words; in this study, it also includes comments on the right quantity of letters within a word. Lastly, the lower-level tag _History/_Historical was removed from the top-level category System and moved to Use, as this top-level category better reflects the arguments that invoke history in the early modern period (cf. Van der Meulen 2018, 2019). Importantly, no lower-level categories were deleted; following Poplack et al. (2015, p. 22), we argue that unattested lower-level categories are just as important as attested ones and add valuable insight regarding the topic of this paper.

The exact nature of the added lower-level categories and the implications of our adjustments in general are discussed in the Results section, as all of the adjustments to the schema provide insight into the differences between the views on Standard Dutch in the early modern period and the later (post)modern period that the original schema was built for.

2.4. Analysis

After annotation, the types of arguments used to motivate linguistic judgements were counted and the patterns in their use, and the values underlying them, analysed to help determine what language ideologies prevailed in this early modern period and what this reveals about how Standard Dutch took shape in its initial phase (16th/17th century), especially when we compare the patterns found in arguments and values to those from the 20th and 21st century. The results of these analyses are presented in the next section.

3. Results

The following subsections discuss which types of arguments were encountered in normative works on Dutch from the early modern period (16th/17th century), how frequently they were found, what patterns can be discerned in their distribution, and how these results compare to those from the later (post)modern period (20th/21st century, from Van der Meulen 2020), focussing first on the top-level categories (Section 3.1) and, then, on the lower-level ones (Section 3.2). Examples of the different argumentation types, for illustration purposes, are given from Section 3.2 onwards.

3.1. Arguments: Top-Level Categories

Figure 2 show that, when the selection and codification of certain forms was motivated with arguments, early modern grammarians often used more than one argument; on average, two arguments per entry were used, with a minimum of one (in 259 entries) and a maximum of 16 (in one instance only). Van Heule (1633), for example, used three arguments to motivate his choice for him for the third-person masculine personal pronoun for singular dative: (1) the word has been used voor duyzent jaren, by onze Voorouders ‘for a thousand years, by our ancestors’ (tag: Use_Historical), (2) de Hoochduytsen hebben ‘Im’ in plaetse van ‘Him’ aengenomen ‘the Germans have replaced ‘Him’ by ‘Im’’ (tag: System_Analogy_German), and (3) Ziet van der Mijle de lingua Belgica ‘See Van der Mijle’s On the Belgian Language’ (tag: Authority_Grammar) (Van Heule 1633, pp. 72–73; Caron 1971b, p. 52).

In total, the 490 argumentative grammatical entries included 1011 arguments. It is important to note that having taken unsampled data from 10 normative works that range about a century inevitably resulted in a skewed distribution (see

Table 2. Number of arguments per grammarian (in order of absolute frequency).

Grammarian	Number of Arguments (Absolute Frequency)
Van Heule 1633	278
Van Heule 1625	159
Resolutiën 1628–1633	156
Ampzing 1628	124
Hooft 1635–1641	114
Leupenius 1653	91
Spiegel 1584	49
de Hubert 1624	38
Kók 1649	1
Radermacher 1568	1
Total	1011

): in Radermacher (1568; in Bostoen 1985) and Kók (1649), only one argument was encountered (of the types System_Grammatical and System_Nature, respectively), whereas Van Heule (1633) used 278 arguments (27.5% of all arguments encountered). This should be taken into account when interpreting the results presented in this section.

As can be seen in Figure 3, arguments from eight top-level categories were attested, with varying frequencies. Only Other argument was unattested, and the two most frequently found categories were Quality and System; together, they account for ca. 61% of all arguments found. This indicates that, according to these formative grammarians, for Dutch to be a good language grammatically, it ought to be qualitative and adhere to the system.

Figure 4 illustrates that, despite the inequal distribution of data, the observation that the top-level categories Quality and System made up most of the arguments found held up for most normative works (used in 22–53% and 5–41% of the times, respectively, in works where the number of arguments was n > 1). Moreover, Figure 5 shows that the presence of most top-level argument-types cannot be ascribed to just one of the normative works included in this study; they were encountered in 6.5 normative works on average. This together forms a strong indication that the values underlying the most frequently found argument types indeed belonged to the language ideology of the time.

These figures show one notable exception to this: the Resolutiën (1628–1633), which contained a relatively high number of arguments of the type Ipse dixit (n = 78, 50% of the total number of arguments (n = 143) found in that work; see Figure 4), accounting for over half of all Ipse dixit arguments (65% of all 120 Ipse dixits; see Figure 5). Section 3.2.5 further explores this result.

3.1.1. Comparison between the 16th/17th and the 20th/21st Centuries

When we compare the results from the early modern period to those from the (post)modern period, some similarities can be discerned (see

Table 3. Types of arguments used in normative works on Standard Dutch from ca. 1550 to 1650 (total number of arguments: n = 1011) versus 1917–2016 (data taken from Van der Meulen (2020); total number of arguments: n = 2322).

Top-Level Category	% of Total Number of Arguments
	16th/17th Century (ca. 1550–1650)	20th/21st Century (1917–2016)
Quality	33.0%	15.1%
System	27.8%	28.2%
Use	17.7%	4.6%
Ipse dixit4	11.9%	NA
Authority	4.4%	8.1%
Variety	3.2%	7.7%
Purity	1.3%	25.1%
Optional variation (see note 4)	0.8%	10.6%
Other argument	NA	0.7%

; results from the latter period taken from Van der Meulen (2020)). First, the top-level category System was of equally large importance in both time periods: 27.8% of the arguments found in the 16th and 17th century are of this type versus 28.2% in the 20th and 21st century. Second, the percentages for Ipse dixit in the earlier centuries and for Optional variation in the later centuries do not differ much: 11.9% vs. 10.6%, respectively. As was explained in Section 2.4, Van der Meulen (2020) used the tag Optional variation for instances that we tagged Ipse dixit for the earlier time period: judgements not motivated by argumentations. This means that, in both periods, there appears to be an equal amount of prescriptions posed without being explained. Third, arguments of the types Authority, Variety, and Other arguments played a minor role in both periods (respectively, 4.4% vs. 8.1%, 3.2% vs. 7.7%, and 0 vs. 0.7%), indicating that these were never at the fore of either period’s language values.

However, the comparison also yields important differences between the two periods. First, the top-level category Quality was more popular in the early modern period (33.0%, vs. 15.1% in (post)modernity), indicating that the use of this type of arguments is characteristic of the initiating phase of Standard Dutch. Second, arguments of the Purity type appear to be more prominent in the 20th/21st than in the 16th/17th centuries (25.1% vs. 1.3%, respectively). However, this is a reflection of entries selected more so than a viable result; for the 16th and 17th century, only grammatical entries were selected for analysis, whereas the data from the later time period included all types of entries (Van der Meulen 2020). This comparison therefore does not hold. Third, arguments of the top-level category Use were more often encountered in the early than in the late period (17.7% vs. 4.6%, respectively). This indicates that the values underlying Use arguments are more important to the language ideology of the early modern period than to that of the (post)modern period.

In order to unravel what these broad claims entail, the next subsection focusses on the lower-level categories of arguments that fall within these top-level categories.

3.2. Arguments: Lower-Level Categories

Taking the lower-level categories into account,

Table 4. Absolute frequency (and percentage of total number of arguments) of top- and lower-level categories of arguments used in normative works on Standard Dutch from ca. 1550 to 1650 (in order of frequency).

Top-Level Category	Lower-Level Categories	Absolute Frequency (% of Total)
Quality (334, 33.0%)	_Differentiation,	123 (12.2%)
	_Beauty,	93 (9.2%)
	_Effect,	63 (6.2%)
	_Quantity,	39 (3.9%)
	_Care,	7 (0.7%)
	_Ease,	6 (0.6%)
	_Logic,	3 (0.3%)
	_Unspecified	NA
System (281, 27.8%)	_Analogy,	113 (11.2%)
	_Grammatical,	107 (10.6%)
	_Nature,	30 (3.0%)
	_Unspecified,	24 (2.4%)
	_Immutable	7 (0.7%)
Use (179, 17.7%)	_Frequency,	70 (6.9%)
	_Authority (incl. _Dictionary, _Grammar, _Literary),	64 (6.3%)
	_Historical,	24 (2.4%)
	_Bible,	11 (1.1%)
	_Socio,	6 (0.6%)
	_Unspecified	4 (0.4%)
Ipse dixit		120 (11.9%)
Authority (44, 4.4%)	_Grammar,	26 (2.6%)
	_Unspecified,	14 (1.4%)
	_Frequency,	4 (0.4%)
	_Dictionary, _Literary, _Socio	NA
Variety (32, 3.2%)	_Genre,	18 (1.8%)
	_Geographic,	10 (1.0%)
	_Mode	4 (0.4%)
	_Register, _Standard, _Unspecified	NA
Purity (13, 1.3%)	_Unspecified	7 (0.7%)
	_Latinism,	5 (0.5%)
	_Gallicism,	1 (0.1%)
	_Anglicism, _Germanism, _Other_Language	NA
Optional variation		8 (0.8%)
Other argument		NA

shows that, out of the 40 categories in the annotation schema, 11 argument types were unattested in the formative period, indicating that the values underlying these argumentations did not play a role in the codification of Dutch in this time period.

Additionally, Table 4 illustrates, on the one hand, that the majority of the 29 argumentation types that were attested were found very infrequently: 12 lower-level categories were encountered fewer than 10 times (each representing <1% of the total number of arguments found), four categories in between 10 and 20 times (1 to 2%), and four other categories 20–30 times (2 to 3%). On the other hand, it shows that the top five encountered argument types together made up over half of the total number of arguments (55.0%): Quality_Differentiation (12.2%), Ipse dixit (11.9%), System_Analogy (11.2%), System_Grammatical (10.6%), and Quality_Beauty (9.2%). This indicates that the values underlying these frequently used arguments in particular were part of the language ideology in this early modern period.

3.2.1. Comparison between the 16th/17th and the 20th/21st Centuries

When focussing on the lower-level categories within the comparison between the two time periods, we notice that the inequal distribution of attested arguments found for the early modern period is mirrored in the (post)modern period (see

Table 5. Percentages of lower-level categories of arguments used in normative works on Standard Dutch from ca. 1550 to 1650 (total number of arguments: n = 1011) versus 1917–2016 (total number of arguments: n = 2322; table copied from Van der Meulen (2020)).

Category	% of Total Number of Arguments		Category
16th/17th Century (ca. 1550–1650)		20th/21st Century (1917–2016)
Quality_Differentiation	12.2%	22.4%	System_Grammatical
System_Analogy	11.9%	12.1%	Purity_Germanism
Ipse dixit (see note 4)	11.2%	10.6%	Optional variation (see note 4)
System_Grammatical	10.6%	9.7%	Purity_Unspecified
Quality_Beauty	9.2%	5.6%	Quality_Effect
Use_Frequency	6.9%	4.3%	Variety_Mode
Use_Authority	6.3%	4.1%	Quality_Unspecified
Quality_Effect	6.2%	3.9%	Use_Frequency
Quality_Quantity	3.9%
System_Nature, Authority_Grammar, Use_Historical, System_Unspecified	2 to 3%	2 to 3%	Authority_Grammar, Authority_Dictionary, Quality_Logic, System_History
Variety_Genre, Authority_Unspecified, Use_Bible, Variety_Geographic	1 to 2%	1 to 2%	Variety_Geographic, Purity_Gallicism, Variety_Register, System_Unspecified, Authority_Unspecified, Quality_Quantity, Purity_Anglicism, Authority_Freq, System_Nature
Purity_Unspecified, Quality_Care, System_Immutable, Optional variation (see note 4), Quality_Ease, Use_Socio, Purity_Latinism, Authority_Frequency, Use_Unspecified, Variety_Mode, Quality_Logic, Purity_Gallicism	<1%	<1%	Authority_Socio, Authority_Literary, Purity_Other_Language, Use_Socio, Use_Unspecified, Variety_Unspecified, System_Standard, Quality_Beauty, Other argument, Quality_Care
Authority_Dictionary, Authority_Literary, Authority_Socio, Purity_Anglicism, Purity_Germanism, Purity_Other_Language, Quality_Unspecified, Variety_Register, Variety_Standard, Variety_Unspecified,Other argument	NA	NA	Purity_Latinism, Quality_Differentiation, Quality_Ease, System_Analogy (and all its subtypes), System_Immutable, Use_Authority, Use_Bible, Variety_Genre (, Ipse dixit) (see note 4)

): in the earlier period, 20 out of the 29 attested lower-level argumentation types each made up less than 3% of all arguments encountered; in the later period, this counts for 23 out of 31 lower-level categories attested. The most striking difference between the two periods is the dominance of System_Grammatical in the 20th/21st centuries (22.4% of all cases) compared to its shared prevalence with Quality_Differentiation, Ipse dixit, System_Analogy, and Quality_Beauty in the 16th/17th centuries. This more balanced top five of the lower-level arguments indicates that there was less consensus within and therefore less canonisation of the normative discourse itself in the formative period of Standard Dutch as opposed to the later time period.

Moreover, in both periods, certain lower-level categories remained unattested, yet what categories were unattested differed between the two time periods, as Table 5 shows. This indicates that the values underlying these unattested argument types did not play a role in the language ideology of the time period they were unattested for. At the same time, their presence in the other period indicates that they were characteristic to the language ideology for that time period in particular if frequently used.

To discern these characteristics for the early modern period in particular, the following subsections zero in on the relevant lower-level categories encountered in each top-level category (in order of frequency) and give examples to illustrate the argument types encountered. All presented results hold for all normative works studied where the number of arguments was n > 1, unless stated otherwise.

3.2.2. Quality

Figure 6 shows that the lower-level categories of Quality arguments encountered in the early modern period were (in order of frequency): Quality_Differentiation, _Beauty, _Effect, _Quantity, _Care, _Ease, and _Logic. These types of arguments can be characterised by the following examples (our translations; the argument in question is underlined, as most cited entries contain multiple arguments):

• Quality_Differentiation: Liever seggen wy ook ‘gy bent’, dan ‘gy syt’, tot onderscheid van den tweeden persoon in het meervoud, dat men altyd moet betrachten, wanneer men het krygen kann. ‘We also rather say gy bent ‘you are [sgl.]’ to distinguish it from the second person plural gy syt ‘you are [pl.]’, a distinction one should always observe when one can.’ (Leupenius 1653, p. 65; Caron 1958, p. 46).
• Quality_Beauty: Maer noch af-sienelicker is het dus-danige Tael-spreuken in plaetse van het twede geval te gebruyken, als ‘Mijn oom Zijn kint’, ‘Mijn Vader zijn Broeder’, ‘Der vrouwen Haer dochter, &c.’ ‘But it is even more disagreeable to use such phrases instead of the genitive, for instance: Mijn oom Zijn kint ‘My uncle his child’, Mijn Vader zijn Broeder ‘my father his brother’, Der vrouwen Haer dochter ‘the woman her daughter’, etc.’ (Van Heule 1633, p. 56; Caron 1971b, p. 42).
• Quality_Effect: De oorzaeke dat het woordeken ‘Zich’ zomwijlen voor ‘Hem’ gestelt wort, is om deze volgende twijffelachticheden te vermijden, als ‘Hy heeft hem daer mede gemoeyt’, uyt deze woorden en kan men niet verstaen of hy eenen anderen ofte zich zelven gemoeyt, heeft, maer alle twijffel wort weg genomen, als men zegt ‘Hy heeft zich daer mede gemoeyt’. ‘The reason that the word Zich ‘himself’ is sometimes used for Hem ‘him’ is to avoid the following dubiousness, for example in Hy heeft hem daer mede gemoeyt ‘He has involved him therein’. From these words, one cannot know whether he has involved another person or himself, but all doubt is removed when one says Hy heeft zich daer mede gemoeyt ‘He has involved himself therein’.’ (Van Heule 1625, p. 92; Caron 1971a, p. 74).
• Quality_Quantity: Jacop van der Schure, oordeelt de laetste E, in de Deel-woorden des generleyen geslachts heel overtollich, ic en zoude het zelve niet wel durven tegen-spreken. ‘Jacop van der Schure judges the final e in participles of the neutral gender entirely superfluous, I would not dare contradict it.’ (Van Heule 1633, p. 26; Caron 1971b, p. 23).
• Quality_Care: ‘Kennen’ noscere, ende ‘konnen’ posse, ‘ik ken’, ende ‘ik kan’, sijn van seer grooten onderscheyd, gelijk wy alle weten: doch worden dickwils onbedacht van den gemeynen man vermengd. ‘Kennen ‘to know’ and konnen ‘can’, ik ken ‘I know’ and ik kan ‘I can’ greatly differ from each other, as we all know, yet they are often carelessly confused by the common man.’ (Ampzing 1628, p. F1v; Zwaan 1939, p. 182).
• Quality_Ease: …: het onderscheid van beteekenisse kann uit de reeden lichtelyk gemerkt worden. Noch dat achten wy geen onvolmaaktheid, maar een groot gemakk voor onse taale: want hoe sy minder veranderinge onderworpen is, hoe sy lichter vallt om geleert te worden. ‘… the difference in meaning can easily be inferred from the sentence. Yet we do not see that as an imperfection, but as a great convenience for our language: for the fewer inflections it is subject to, the easier it is learnt.’ (Leupenius 1653, p. 44; Caron 1958, p. 31).
• Quality_Logic: Daar het een groot misbruik is dat ‘en’ somtyds genoomen wordt voor een ontkenninge, gestellt synde by ‘geen’ of ‘niet’: soo wordt gemeenlyk geseidt, ‘gy en sullt niet dooden’, ‘gy en sullt niet steelen’, ‘gy en sullt geen overspel doen’: doch dat is teegen den aard der ontkenningen: want daar twee ontkenningen by een komen, doen sy soo veel als eene bevestiginge: nu ‘geen’ en ‘niet’ syn ook ontkenningen, daarom kann ‘en’, als een ontkenninge, daar by geen plaatse hebben. ‘For it is a great abuse that en is sometimes added to geen ‘none’ or niet ‘not’ and so taken for a negation: thus it is commonly said gy en sullt niet dooden ‘you shall not kill’, gy en sullt niet steelen ‘you shall not steal’, gy en sullt geen overspel doen ‘you shall not commit adultery’: yet that is contrary to the nature of negations: for when two negations are combined, they add up to a confirmation: well, geen and niet are also negations, therefore en, as a negation, cannot be put next to it.’ (Leupenius 1653, p. 70; Caron 1958, p. 51).

Moreover, Figure 6 shows that the lower-level category most frequently used in the early modern period was _Differentiation (n = 123), followed by _Beauty (n = 93) and _Effect (n = 63). This indicates that, when we say that the quality of language was important to the grammarians shaping Standard Dutch in its initial phase, this mostly refers to the language’s ability to differentiate, to be beautiful, and to reach the intended effect.

As the qualification _Differentiation was added to the annotation schema for this study on metalanguage from the 16th and 17th centuries, this lower-level category is especially interesting when pinpointing the language ideologies in this early time period.5 The tag refers to the ability of the language to differentiate a distinction in meaning through a distinction in form (cf. Sundby et al. 1991, p. 22). Here, the ‘one form, one function’ principle surfaces: a preference for a certain variant is regularly justified by the argument that its form more accurately distinguishes for example its case or gender from other cases or genders. Often, the entries that contain such an argument are examples of reallocation; they take two optional variants and allocate each to a different function (de Vos and van der Meulen 2021; de Vos, forthcoming; see also Poplack et al. 2015 and their strategies for factoring out variability; Britain and Trudgill 2005). Both of these functions could previously be expressed by one and the same variant, so, by this reallocation, one-to-one correspondence between form and function is reached, and both optionality, as well as multifunctionality, are suppressed.

Additionally, the vast majority of the examples that fell into the category _Effect (54 out of 63) targeted the unwanted effect that certain language use evokes ambiguity and thereby possible confusion, which connects this tag directly to _Differentiation; if languages perfectly differentiated, such ambiguities would never emerge. This link increases the already dominant position of the tag Quality_Differentiation in the formative period (see Table 4), thereby confirming that the value that good language ought to differentiate was part of the language ideology of the time.

3.2.3. System

Within the top-level category System, the following lower-level types of arguments were encountered (in order of frequency, see Figure 7): System_Analogy, _Grammatical, _Nature, _Unspecified, and _Immutable. Examples of these arguments are (my underlining):

• System_Analogy: ’Ten is ook so vremd niet in onse tale, als wel sommige meenen, dat het woordeken ‘gij’ eens ende onveranderlick soude blijven in sijn enkel ende veelvoudig getal: want dit en is niet alleenlick in dit woordeken ‘Gij’, maar ook in eenige andere woorden gebruijkelick: so seijtmen; ‘die looft God’, ‘die loven God’, ‘sy looft God’, ‘sy loven God’. blijvende de woorden ‘die’ ende ‘sy’, gelijk het woordeken ‘gij’, onveranderlick in haare buijginge, ende nochtans niet te min onderscheijdelick in haar getal door het gevolg der t’samenvouginge. ‘And it is also not so strange that the word gij ‘you’ in our language, as some argue, has one form for the singular and plural: for this is not only the case in this word Gij, but also common in some other words: for one says die looft God ‘they [sgl.] praise God’, die loven God ‘they [pl.] praise God’, sy looft God ‘she praises God’, sy loven God ‘they praise God’. The words die and sy, just like the word gij, remain unaltered in their inflection, and are nonetheless distinguishable in number because of what follows.’ (de Hubert 1624, pp. *4r–*4v; Zwaan 1939, p. 394).
• System_Grammatical: XVI. EEN VOORTVAEREND MAN, EEN MAN VOORTVAERENDE VAN AERDT, Liever een MAN VOORTVAEREND VAN AERDT; om dat VOORTVAEREND hier geen Participum is, maer Nomen. ‘XVI. Een voortvaerend man ‘an expeditious man’, een man voortvaerende van aerdt, ‘a man expeditious in nature’. It is better to use een man voortvaerend van aerdt, because voortvaerend is not a participle here, but an adjective.’ (Zwaan 1939, p. 283).
• System_Nature: …, maer al hoewel deze maniere, niet geheel hart en valt, zo wijkt het nochtans van den aert onzer sprake. ‘…, but although this is not completely displeasing, yet it deviates from the nature of our language’ (Van Heule 1633, p. 35; Caron 1971b, p. 29).
• System_Unspecified: XLV. D’ANDER staende zonder Substantyf kan bequaemelijk in Dativo Singulari hebben DEN ANDRE. ‘XLV. D’ander ‘the other’ can, when it is used independently without a noun, appropriately have den andre in dative singular.’ (Zwaan 1939, p. 246).
• System_Immutable: Want wy en seggen niet, ‘het saed Jans’, ‘het huys Pieters’, maar ‘Jans saed’, ‘Pieters huys’, ofte ‘het saed van Jan’, ‘het huys van Pieter’. …: niettemin wanneermen wat sal schrijven, ende het licht laten sien, het welke de nakomelingen ook gebruyken sullen, so behoren wy vlijtig acht te nemen, ten eersten, dat wy Duytsch, ende daer na, goed ende suyver Duytsch mogten spreken: op dat de nakomelingen sich met der tijd daer aen mogten gewennen, ende onse tale toekomstig niet vervalscht, ende gebroken, maer oprecht, ende naer haeren oorspronkelijken aerd mogte uytgesproken worden. ‘For we do not say het saed Jans ‘the seed John’s’, het huys Pieters ‘the house Peter’s’, but Jans saed ‘John’s seed’, Pieters huys ‘Peter’s house’, or het saed van Jan ‘the seed of John’, het huys van Pieter ‘the house of Peter’. …: nevertheless, when we write and publish something which will also be used by our descendants, then we should diligently take into account, firstly, that we use Dutch, and secondly, that we use good and pure Dutch: so that, over time, our descendants may get accustomed to it, and our language will not be garbled and mispronounced in the future, but will be pronounced in the right way according to its original nature.’ (Ampzing 1628, p. A3r; Zwaan 1939, p. 141).

The lower-level categories most often employed during the 16th and 17th century were _Analogy (n = 113) and _Grammatical (n = 107) equally (see Figure 7), thus indicating that adhering to the system means to conform to the workings of other constructions or to grammatical rules.

When compared to the 20th and 21st centuries (see Table 5), it is especially the lower-level category _Analogy that appears typical for the 16th and 17th century, as it was not present in the later centuries but highly frequent in the early modern period.6 This lower-level category can be subdivided further into what languages constructions needed to be analogous to. Most frequently, an analogy is made with another Dutch construction (n = 36), suggesting that the early modern grammarians valued a certain consistency in the rules Standard Dutch should abide by. These references to Dutch are quickly followed by references to Latin (n = 25) and Greek (n = 23 vs. German: n = 12, French: n = 5, Hebrew: n = 3, Italian: n = 3, Spanish: n = 1, and Unspecified: n = 5). This indicates that Dutch was often mirrored to the classical languages and that Standard Dutch was sometimes perhaps even modelled on it (similar to, for example, the Spanish normative tradition; see, e.g., Amorós-Negre 2020, p. 586).

3.2.4. Use

Figure 8 shows that the lower-level types of arguments within the top-level category Use found are (in order of frequency, argument underlined):

• Use_Frequency: Anderzins so wanneer het Gevall blijkt uijt het voorgaande ledeken, ofte dat het met eenige vvoorden meer bekleed vvord, so en is sulks so seer niet van noode, immers in dit woord: so seijt men, ‘den goeden God sij lof’, ‘doet des hemels God belijd’: ende niet so seer gebruijkelick, ‘den goeden Gode’, ‘des hemels Gode’. ‘Something else happens when the case becomes apparent from the article in front of it, or when it is accompanied by a couple more words, for then this [inflection] is not necessary in the noun: for one says den goeden God sij lof ‘praise be to the good God’, doet des hemels God belijd ‘confess to the heavenly God’: and not so commonly used is den goeden Gode, des hemels Gode.’ (de Hubert 1624, p. *8r; Zwaan 1939, p. 127).
• Use_Authority: Dit waernemen der Gevallen in het Veelvoudig, is by de Ouden altijt gebruykt geweest, ook by Koornhert, Aldegonde, Grotius, Ampsingius, ende andere. ‘Observing cases in plural was common use for the Elders, also for Koornhert, Aldegonde, Grotius, Ampzing, and others.’ (Van Heule 1625, p. 24; Caron 1971a, p. 25).
• Use_Historical: Ende so wenschte ik wel, dat de woordekens ‘dij’, ende ‘dijn’, in heur oud gebruyk mogten hersteld worden, …. ‘And so I wish that the words dij ‘you’, and dijn ‘yours’, may be restored in their old use, ….’ (Ampzing 1628, p. E4v; Zwaan 1939, p. 180).
• Use_Bible: DE PAT, DEN PAT, an HET PAT, ut nostra et vulgatum est.(HET PAT) ‘De pat, den pat or het pat ‘the path’ like in ours [i.e., the Deux-Aes Bible] and in the Vulgate. ([decision:] Het pat)’ (Zwaan 1939, p. 216).
• Use_Socio: D. MY (nunquam MYN, ut vulgus hic loquitur) ‘Dative: my ‘me’ (never myn ‘mine’, like the common people say here)’ (Zwaan 1939, p. 212).
• Use_Unspecified: NEVEN, BENEVEN, NEFFENS, BENEFFENS, promiscue: spectetur usus. ‘Neven, beneven, neffens, beneffens ‘next to’, indiscriminately: it depends on its use.’ (Zwaan 1939, p. 209).

As Figure 8 also illustrates, most references to Use were either to _Frequency (n = 70) or to _Authority (n = 64), indicating that something was pre- or proscribed because of either its frequent use or its use by someone of authority. While Use_Frequency arguments were distributed fairly equally, it was mostly Van Heule (1625, 1633) who employed the latter subtype: out of all 64 Use_Authority arguments used, 88% were used by him (n = 56).

Again, comparison to the (post)modern centuries points us towards the relevance of the category Use_Authority for this early time period. In order to accommodate the annotation schema used for the 20th and 21st centuries to the early modern period, the lower-level categories _Dictionary, _Grammar and _Literary were replaced by the umbrella term _Authority, encapsulating all three lower-level tags. This was more appropriate for this time period, as the authorities invoked often combined the roles of grammarian and literary author, making it impossible to determine whether they were praised for their literary language use or that in their grammars or dictionaries. Use_Authority was unattested in the (post)modern period (both the umbrella term, as well as the subtypes _Dictionary, _Grammar, and _Literary; see Table 5), yet highly frequent in the early modern period, indicating that referring to the language use of an authority was typical of the language values of the 16th and 17th centuries.7

The authorities cited were mostly other grammarians, literary authors, scholars, or bible translators (or men who combined these roles), such as Philips of Marnix Lord of Saint-Aldegonde, Samuel Ampzing, Jacob Cats, Hugo Grotius, Cornelis Kiliaan, or Simon Stevin.8 This indicates that these men were seen as examples not only worth following in their scholarly work but also worth imitating in their use of the Dutch language. However, it remains to be seen whether they actually employed the uses they were cited for (cf. Van Hardeveld 2000, who proved that this was often not the case for usages cited in Lodewijk Meijer’s (1629–1681) lexicographical work; similar discrepancies between reception and reality have been attested for German; see, e.g., (Salmons 2018, pp. 279–80) on Martin Luther’s use of verbal frames).

3.2.5. Ipse Dixit

As was illustrated in Section 3.1, arguments of the Ipse dixit type in the 16th and 17th centuries were mostly found in the Resolutiën (1628–1633), where 50% of all arguments used were Ipse dixits, which together accounted for 65% of all Ipse dixits used in the early modern grammars studied (see Figure 4 and Figure 5, n = 78). The presence of this type of argument in the top five (see Table 5) is thus heavily skewed by the inclusion of the Resolutiën in our analysis; if we were to exclude it, Ipse dixit-arguments would only make up 4.9% of all arguments used. This seems to suggest that the Ipse dixit type should be seen as a less important argument type for this time period than previously stated.

However, there is a caveat to this. The profuse presence of arguments of the Ipse dixit type in the Resolutiën can be explained by taking the nature of this work and an important part of our annotation method into account. To do this, we first need to present the results for the annotation that preceded the annotation for type of argument, namely the annotation for type of entry.

The type of entry annotation was the result of our data-driven approach and an adjustment to Van der Meulen’s (2020) method (see Section 2.2) and, therefore, also reveals peculiarities of the views on Standard Dutch from the 16th and 17th centuries. While Van der Meulen (2020) concluded that all entries included in his study took a stance on the acceptance of optional variability within the usage items discussed, only in ca. 40% of all 1237 entries found in normative works from the 16th and 17th century such a stance was taken (n = 490, presented by the entry tag argumentation). The majority of entries did not contain a linguistic judgement but only a definition, an enumeration, a paradigm, a question, a rule, or a combination of these (ca. 60%, n = 747). This seems to suggest that grammars written in the early modern period were for the most part descriptive instead of prescriptive.

However, Figure 9 shows that there are differences in this respect between the normative works included in this study: some of the works indeed have a majority of other types of entries (definitions, rules, etc. combined in Figure 9 as other), whereas others are mostly argumentative. Of the former type, Radermacher (1568; Bostoen 1985) and Kók (1649) stand out especially; they each contain over 90% of other-type entries. Note, however, that the work by Radermacher contains 10 entries in total, as it is a manuscript that is only partly preserved (Bostoen 1985); this is bound to result in a skewed distribution. Kók’s Ont-werp der Neder-duitsche letter-konst ‘Design of the Dutch Grammar’ (Kók 1649), however, contains 109 entries in total, out of which only one is of the type argumentation. This would seemingly make this work particularly descriptive in nature.

Of the latter type, the Resolutiën (1628–1633) is a true outlier; with 100% of its entries containing a linguistic judgement (n = 111), it is the only normative work in which this type of entry forms the majority of entries. This is where the nature of the Resolutiën comes in; even though the work discusses grammatical topics, it is not a grammar as such. It contains the notes of discussions between the many Dutch translators of the Bible, meant to ensure uniformity within the translation (more a type of ‘style guide’; see, e.g., Straaijer (2018) for a classification of different genres). This means that the work provides decisions made on specific topics of discussion only and not a full grammar of Dutch. Importantly, where there was discussion, there ought to have been variation; therefore, all its entries could be interpreted as containing a judgement on linguistic variation. While, for the entries in the Resolutiën, a judgement on variation could be presupposed, the absence or presence of such a judgement could not be adequately determined for certain entries taken from the other works (see Section 2.2). This explains the high frequency of the argumentation type of entry in the Resolutiën, as opposed to the other normative works.

Returning, then, to the Ipse dixits; as was shown before, the many argumentative entries in the Resolutiën were only sometimes accompanied by argumentations but often just included unmotivated rules of thumb: Ipse dixits. It is particularly these linguistic judgements without argumentations that, in the normative works that did not presuppose discussion, could not be unequivocally distinguished from other types of (ostensibly descriptive) entries. For example: had the entry a BEGINNEN, BEGONNEN ‘Of BEGINNEN ‘to begin’, BEGONNEN ‘began/begun’’ from the Resolutiën (Zwaan 1939, p. 201) been encountered in another normative work, it would have been counted as a rule, not as an Ipse dixit within the argumentation type of entry. After all, the entry only states that begonnen is a form of the verb beginnen ‘to begin’; on the surface, there is no mention of variation, no rejected or preferred use, and no judgement, just a descriptive rule (see Appendix A for the full definitions of the different types of entry). However, because of the nature of the Resolutiën, we were able to say that there must have been variation in the verb beginnen to start with, as it would otherwise not have been included in these discussion notes. From the entry itself, we cannot tell what the other variants were, nor what the arguments against or for certain variations were; we only can tell that begonnen is the variant chosen after discussion on the topic.

This same conclusion could not be drawn for the entry beghin, beghon óf beghan, beghonnen ‘begin, began, begun’ encountered in the Twe-spraack (Spiegel 1584, p. 89; Dibbets 1985, p. 265). For the Twe-spraack, a judgement on variation could not be presupposed, as this work was not a type of ‘style guide’ but a traditional grammar, detailing the inner workings of the Dutch language on a range of different topics. The entry on the verb to begin in this work has therefore been counted as an other type of entry, for, at the surface level, it does not meet the requirements for an argumentation type of entry; there is no variant or use discouraged or advocated. However, the entry could, in hindsight, perhaps also be interpreted as a covert linguistic judgement on the same topic as the entry from the Resolutiën (the paradigm for beginnen), seemingly deserving of the annotation Ipse dixit within an (implicit) argumentation type of entry instead.

Due to our approach, some of the entries that were tagged as others in the other normative works might, after closer inspection, turn out to be argumentations after all and, more specifically, Ipse dixits, albeit with an implicit linguistic judgement. The descriptiveness of most of the grammars in this study should therefore be taken with a large pinch of salt; they are likely only ostensibly so. We will return to this more general point in Section 4; for this subsection, we can conclude that the amount of Ipse dixits is more likely to have been underestimated in the other works than overestimated in the Resolutiën.

3.2.6. Authority

Three lower-level categories were present within the top-level Authority (in order of frequency, argument underlined):

• Authority_Grammar: … het welke de E. Christiaen van Heule, Mathematicus, in sijne Spraekkonste ook seer wel gemerkt, ende aengeteykend heeft. ‘…which was also noticed and noted by the honourable Christiaen van Heule, Mathematician, in his Spraekkonste ‘Grammar’.’ (Ampzing 1628, p. A2v; Zwaan 1939, p. 140).
• Authority_Unspecified: …zo en hebben wy ons tegens dat out ende noch het tegenwoordig gemeyn gebruyk niet dorven stellen, alhoewel het zelf eerst ons voornemen geweest is, te doen, maer door het oordeel eeniger hoochgeleerden, hebben wy bewogen geweest, dat naer te laeten, …. ‘…thus we have not dared to disapprove of that old nor of the present-day common use, even though it was our intention to do so first, but the opinion of a couple of scholars has persuaded us to leave it be, ….’ (Van Heule 1625, p. 11; Caron 1971a, p. 17).
• Authority_Frequency: Wy vougen hier in het out gebruyc der Werc-woorden, zonder welc gebruyc wy in onze Sprake zeer veel verliezen, ooc zijn tot de herstellinge des zelven, alle Tael-geleerde gheneycht geweest, als Koornhert, Aldegonde, Ampsingius, …. ‘We add here the old use of verbs, because without it we would lose a lot in our language, and all grammarians too have been in favour of revitalizing these, such as Koornhert, Aldegonde, Ampzing, ….’ (Van Heule 1633, pp. 87–88; Caron 1971b, p. 61).

Of these three, _Grammar was most prominent in use during the 16th and 17th centuries (n = 26; see Table 4). Similar people were cited as in examples of Use_Authority-arguments (e.g., Ampzing, de Hubert, Aldegonde, Coornhert, Van der Schuere, Kiliaan, and Van der Mijle). Even though these men have been tagged as grammarians, they may well have been cited for their literary or lexicographical works, prompting the tags _Literary and _Dictionary instead. However, as these distinctions were less clear in the early modern period (see Section 3.2.4), we combined them in the tag Authority_Grammar, thereby rendering the other two tags empty.

Six out of ten grammarians used Authority arguments, but this type was mostly, like with arguments of the type Use_Authority, used by Van Heule (1625, 1633). Interestingly, Van Heule showed a shift in his use of this argument-type: in 1625, he used Authority arguments of the _Unspecified kind relatively more frequently, referring to rules stated by others without making explicit who had posed them, whereas, in 1633, he went on to cite them by name. This suggests that, by 1633, the normative discourse on Standard Dutch might have shifted from a collection of individual, mostly unrelated works towards a discussion between grammarians.

3.2.7. Variety

Examples of the lower-level types of arguments encountered within the top-level argument type Variety are (in order of frequency, argument underlined):

• Variety_Genre: Dit onderscheyt der geslachten en behouft in den rijm altijt niet nagevolgt te worden, want om die oorzaeke zouden de Rijmers al te nouw gebonden zijn, in het waernemen der voeten, dewijl dan dat eenen afbreuk van onze spraekx cierlickheyt zoude veroorzaeken, zo wort den Rijmers, eene volle vryheyt gelaten, om de byvouglicke worden somtijts te verkorten. ‘This distinction of gender does not always need to be followed in poetry, because Poets would then be too restricted whilst observing poetic feet, which would then cause an impairment of our language’s grace, thus Poets are given full freedom to shorten adjectives sometimes.’ (Van Heule 1625, p. 16; Caron 1971a, p. 20).
• Variety_Geographic: In het verkleijnen der woorden hebben de Brabanders de meeste volkomenheyt. ‘In the diminutives, the people from Brabant have the most perfection.’ (Van Heule 1633, p. 161; Caron 1971b, p. 110).
• Variety_Mode: … om kortheids wille wordt somtyds voor ‘my’ en ‘wy’, ‘me’ en ‘we’ gebruikt, insonderheid daar sy achter een werkwoord gestellt worden, ‘sy hebbenme’, ‘soo seggenwe’: Wy laaten toe, datmen om de soetvloeijentheid soo spreeke: maar de letteren ten vollen uitschryve, ‘sy hebben my’, ‘soo seggen wy’. ‘… for brevity’s sake, me ‘me’ and we ‘we’ are sometimes used for my ‘me’ and wy ‘we’, especially when placed behind a verb: sy hebbenme ‘they have me’, soo seggenwe ‘so say we’. We allow speaking thus for euphonic reasons, but in writing, the letters need to be written out in their entirety: sy hebben my ‘they have me’, soo seggen wy ‘so say we’.’ (Leupenius 1653, p. 48–49; Caron 1958, p. 35).

Out of the 32 instances of the argumentation type Variety encountered in normative works from the 16th and 17th century, 18 are of the subtype _Genre (see Table 4). As the example above illustrates, this subtype contains references to poetic license (see, e.g., Sundby et al. 1991, p. 42), where certain variants are or variation is allowed in poetry but not in prose. Although this type of argument was not included in the (post)modern period (see Table 5), it was too rarely observed in this early time period for its underlying value to have formed an important part of the language ideology of this time.

3.2.8. Purity

Three lower-level categories were attested within the top-level category Purity: _Gallicism, _Latinism, and _Unspecified. Examples of these types of arguments (in order of frequency, argument underlined):

• Purity_Unspecified: Geen weyniger misslag word-er in het stellen der by-worpige woorden begaen, als de selve niet voor, maer achter hunne selfstandige woorden geplaetst worden: als ‘een man groot’, ‘een kind kleyn’, ‘een paerd sterk’, voor ‘een groot man’, ‘een kleyn kind’, ‘een sterk paerd’, het welke te gansch ongerijmd is, ende onder deckzel van rijm-verlof geenzins te lijden: want men en mag so niet rijmen datmen de tale geweld doet: ende dat geen goed Duytsch en is, kan dat wel goed rijm wesen? Wy en mogen ons hier niet behelpen met andere spraken. Ygelijke tale heeft haere eygenschap: maer de onse en kan dit niet lijden. ‘There is no greater mistake in the use of adjectives than to place them not in front, but after their nouns: e.g., een man groot ‘a man great’, een kind kleyn ‘a child small’, een paerd sterk ‘a horse strong’, for een groot man ‘a great man’, een kleyn kind ‘a small child’, een sterk paerd ‘a strong horse’, which is entirely preposterous and cannot be suffered under poetic license: for even in rhyme the language must not be violated: and that which is not good Dutch, can that even be good poetry? We cannot make do with other languages here. Every language has their own characteristics: but ours cannot suffer this.’ (Ampzing 1628, p. F3r; Zwaan 1939, pp. 185–86).
• Purity_Latinism: Hier uyt is ook ontstaen, dat men de aengenomene vreemde woorden in de eyge namen in hunne gevallen naer de Latijnsche wijse liever heeft willen buygen, als naer onse eygene: so seggen onse geleerden, Petrus, Petri, Petro, Petrum, Petre, Petro: Samuel, Samuelis, Samueli, Samuelem, Samuel, Samuele, &c. dat dan immers geen goed Duytsch en kan wesen, ‘From this it has also arisen that people have preferred to use Latin case inflections in accepted foreign proper names rather than our own: so our scholars say Petrus, Petri, Petro, Petrum, Petre, Petro: Samuel, Samuellis, Samueli, Samuelem, Samuel, Samuele, etc. which cannot be good Dutch.’ (Ampzing 1628, p. A2v; Zwaan 1939, p. 140).
• Purity_Gallicism: VVant onder deckzel van rijmen, den VVaal te spelen, is ganz ongerijmd, ende te seggen; … ‘Hij is genegen niet’, voor; ‘Hij is niet genegen’. ‘De gunste goed’, voor; ‘de goede gunst’’. …, ende diergelijke wijse van spreken meer; is het Nederduijtz VValzelick, ende valzelick verdraijd…. ‘For to play the Walloon disguised by rhyme is entirely preposterous, and to say: … Hij is genegen niet ‘He is inclined not’ for Hij is niet genegen ‘He is not inclined’. De gunste goed ‘The favour good’ for De goede gunste ‘The good favour’. …, and such ways of speaking is twisting Dutch Walloonishly and falsely….’ (de Hubert 1624, p. *6r; Zwaan 1939, p. 123).

Purity arguments were, like Variety arguments, used very irregularly in the 16th and 17th centuries: only 13 of the kind were found. This would suggest that the purity of the Dutch language was not important to the grammarians shaping its standard language in the early modern period. However, here, we are deceived by our selection of entries for analysis (see, also, Section 3.1.1); we can conclude that purity was of no major importance for grammatical topics, but this could be completely different for topics of a different kind, for example, lexicon or spelling. This inkling is confirmed by the fact that many explicitly puristic dictionaries were published in this time period (see, e.g., Van der Sijs 2021).

3.2.9. Optional variation

The use of arguments of the type Optional variation is, again, negligible; only eight instances were attested (see Table 4). However, this type is all the more interesting because of its content; it stresses the advantages of variation (not necessarily optional; see de Vos and van der Meulen 2021). This advocacy of variation goes directly against Van der Meulen’s (2020) conclusion that the prescriptive value ‘language should not contain variation’ plays an important role in Dutch prescriptivism; this may be the case for publications from the 20th and 21st centuries, but in the early modern period, at least in some cases, variation was also seen as a virtue.

4. Discussion

This paper is the first to perform a systematic quantitative analysis of the arguments used to motivate the selection of certain forms in grammatical entries from normative works on Standard Dutch written between ca. 1550 and 1650. We annotated 1011 arguments used to motivate a linguistic judgement in 490 entries that contained such a judgement, and we analysed the patterns in their use and the values underlying them. Our results indicate that, according to grammarians in the early modern period, in order for Dutch to be a good language in terms of its grammar, it ought to be qualitative, adhere to the system, and reflect use. The lower-level argument categories used within these top levels informed further interpretation of these broad statements; in order for Standard Dutch to be a good language grammatically speaking, it ought to differentiate, be beautiful, conform to grammatical rules, display consistency, mirror Latin and Greek, and reflect frequent use or use by authorities. These are the principles that shaped Dutch grammar in its initial phase. Their absence in Van der Meulen’s (2020) annotation schema, constructed for his study on prescriptivist/language advice publications on Standard Dutch from the 20th and 21st century, combined with their high frequency indicates that, especially, the importance bestowed on its ability to differentiate, to be consistent or to mirror Latin and Greek, and to reflect the language use of certain authorities was characteristic of the codification of Dutch grammar in the 16th and 17th century.

Based on this analysis, we can determine what this reveals about what language ideology prevailed among linguists and grammarians in this formative period of Standard Dutch. In the Introduction to this paper, we cited Silverstein’s (1979) definition of a linguistic ideology as a set of beliefs about language that serves as rationalisation or justification of its structure and use. Hüning et al. (2012) argue that the early modern period saw the emergence of a ‘correctness ideology’, connecting ‘language’ to ‘norm’ for the first time. The values mentioned above, then, used to justify micro-selections for Standard Dutch, together constitute what the early modern Dutch grammarians saw as the norm: grammar that was able to differentiate, be beautiful, conform to grammatical rules, display consistency, mirror Latin and Greek, and reflect frequent use or use by authorities.

Especially the values of differentiation (based on the ‘one form, one function’-principle), conformation to grammatical rules and consistency tie in with what Hüning, Vogl, and Moliner argue was the purpose of the many early modern dictionaries, spelling guides, and grammars; they were, for the first time, striving for discreteness, uniformity, and stability of language (Hüning et al. 2012, pp. 20–21). The rules incorporated in them, Hüning, Vogl, and Moliner argue, were ‘descriptions of regularities which could be observed in the (use of the) language in question’ (Hüning et al. 2012, p. 14). Our analysis shows, however, that regularities were not just observed and described but also prescribed: Dutch was, at the same time, made more regular (cf. Silverstein 1979, p. 233, who claims that to ‘understand’ one’s own linguistic usage is potentially to change it, i.e., rationalisations can make it more regular; repeated in Woolard 1992, p. 241). The mirroring of Dutch to Latin and Greek also fits into this quest for a regular norm. Especially Latin was considered an ideal language, as it showed the regularities desired (e.g., Dibbets 1995, pp. 331–33). Thus, the attempts to match Dutch constructions to the classical languages had a similar regularising effect. Here, then, we see the roots of what Rutten (2019, p. 115) calls the typical 18th century ‘polishing’ of Standard Dutch, i.e., its normalisation and regularisation; even though this came to full fruition later, the connection of language with ‘norm’ and ‘regular’ was already made in arguments used to motivate micro-selections in the early modern period and was based on a Latin ideal (see, also, Dibbets 1995; Van der Sijs 2021; Van der Wal 1995).

However, also telling is what is missing from the frequently invoked values informing the norm for Dutch in the 16th and 17th century: social factors such as social class or geographic region. For example, the values ‘Good language is determined by what a certain group of people says’ (Use_Socio) and ‘The language spoken in certain geographic regions is right/wrong’ (Variety_Geographic) were only appealed to six and 10 times, respectively. Two other types of Use are more frequently appealed to: Use_Frequency (n = 70) and Use_Authority (n = 64, alongside 44 appeals to what these same authorities said about Dutch: top-level category Authority). The former indicates a norm based on use by the majority of people, which suggests a socially neutral interpretation (cf. Rutten 2019 on the ‘myth of neutrality’; see, however, Gal and Woolard 2001a for a warning against the deceiving ‘aperspectival objectivity’ that comes with invoking ‘the public’). The latter appears less neutral; value is bestowed on the language use of authority figures (one of the characteristic aspects of this time period, see Section 3.2.4). However, although the men cited were certainly part of the social and cultural elite of the time, it is not their social status they were referenced for; their authority came, importantly, from their scholarly or literary work, i.e., their expertise on language.

So, although a norm of regularity is developing in this time period, this ideal is not firmly connected to social identity yet (see, also, Hüning et al. 2012, pp. 20–21). Instead, the motivations that shape the direction of the standardisation process in the early modern period focus more on language internal factors; factors that proved formative for an important part of the later standard language ideology (the ‘correctness ideology’ part). This indicates that Rutten is right to claim that, even though many of the important ideas that are part of the Dutch standard language ideology were already present in the early modern period, it was only later, namely the period around 1800, that these ideas were brought together in a coherent framework, i.e., the framework of linguistic nationalism (Rutten 2016, p. 52; 2019, p. 128). Further research into other aspects of the normative discourse on Standard Dutch from the period before 1800 could help determine how the early modern view, with the norm for grammar as one that is able to differentiate, be beautiful, conform to grammatical rules, display consistency, mirror Latin and Greek, and reflect frequent use or use by authorities, can be truly characterised, ideologically speaking.

Next to these implications for the language ideology in the early modern discourse on Standard Dutch, the results presented in Section 3 also helped us to establish what the differences were between the codification of Dutch in the early modern period (16th/17th century) and the (post)modern period (20th/21st century; studied in Van der Meulen (2020)). The main linguistic values for the early modern period described above differ from the dominant idea in the (post)modern period that language should be pure and grammatical (Van der Meulen 2020), although it should be noted that this study looked at all levels of language, not just at grammar. Further comparison of the results for both time periods indicates an additional difference; whereas the later period was dominated by an appeal to adherence to grammar, there was not one dominant value in the early modern period. In this early period, four values were appealed to more or less equally frequently: the language’s ability to differentiate, its similarity to other constructions or languages, its adherence to grammar, and its beauty. This indicates that there was less consensus in the early modern period than in the (post)modern period and, therefore, less canonisation of the normative discourse itself.

Thus, the value of our systematic method was proven; using the same annotation schema constructed for normative works on Dutch from the 20th and 21st century for this study on works from the 16th and 17th century made it possible to quantify and compare the values present in the language ideologies of both time periods, yielding valuable results. If future research were to apply this method to prescriptive utterances from different levels of language (e.g., spelling), different time periods, and different languages, this would paint an even clearer picture of the characteristics of all time periods and languages studied.

However, we did have to adjust the method to accommodate it to the early modern period, indicating that the method as outlined by Van der Meulen (2020) was not yet fully equipped for such a wide application, and perhaps it still is not. While adjustments to the annotation schema are valuable and can be justified by the data-driven approach (argument types newly discovered in works from the 16th and 17th centuries had to be given a tag of their own), this has resulted in a very elaborate annotation schema consisting of 40 categories. Even though the challenges encountered and remedied in this study have considerably contributed to finetuning this method, future research has to determine how generally applicable it truly is and what further changes need to be made to accommodate it to other levels of language, time periods, and languages.

Moreover, one particular weakness of the data-driven approach outlined in this paper was uncovered in the analysis of the tag Ipse dixit (a linguistic judgement without explicit argumentation to justify it), which also was part of the top five of categories most frequently found in the 16th and 17th centuries. Despite meticulous attempts to ensure consistency throughout, the assignment of the appropriate categories to entries from the early modern period proved particularly challenging because of the differences between the 20th/21st century materials and earlier materials, and the difficulties posed by working with the latter (see, also, Poplack et al. 2015). These challenges were most prominent in what demarcated an entry but also in what constitutes grammatical entries and at the level of the type of entry annotations. Our results from these type of entry annotations seemed to suggest that grammars written during the formative period of Standard Dutch were, for the most part, descriptive (defining, enumerating, giving paradigms, posing questions, simply making explicit what rules governed the language) instead of prescriptive (containing linguistic judgements and argumentations motivating these). However, this was likely due to the decision to tag only entries that contained explicit linguistic judgements as argumentations, thereby excluding entries that could, by processes of erasure, contain implicit linguistic judgements (cf. Irvine and Gal 2000).

This strategy, on the one hand, remedied the impossibility of distinguishing such ostensibly descriptive yet implicitly prescriptive judgements from entries that truly did not contain a judgement on variation (see, also, Chapman 2021) but, on the other hand, likely led to an underestimation of the amount of Ipse dixits and, thereby, argumentations in most normative works included in this study. This indicates that grammars from the early modern period were presumably, for the most part, only ostensibly descriptive as opposed to the explicit prescriptiveness of the normative works from the 20th and 21st century studied in Van der Meulen (2020). Despite the issues surrounding this conclusion, even the contrast between (ostensible) descriptivism or implicit prescriptivism in the early modern period and explicit prescriptivism in the (post)modern period proves a valuable outcome of this study, as it substantiates the claim that the nature of normative publications changed from more normative grammar towards specific usage advice in the 20th century (Van der Wal and Van Bree 1992). Moreover, it indicates that the early modern Dutch grammars, often explicitly referred to as ‘prescriptive grammars’ (e.g., in Willemyns 2003, p. 98), show similarities to contemporary descriptive European grammaticography in their search for a good balance between description and prescription (Amorós-Negre 2020). This diverges from, for example, Spanish, where the prescriptive ideology and overt linguistic criticism were particularly present throughout the 16th and 17th centuries, continued by the Real Academia Española established in 1713, only slowly giving way to covert prescription in codification tasks from the 19th century onwards (Amorós-Negre 2020; note, however, that her focus is on grammaticography, whereas Van der Meulen’s (2020) conclusions are based on prescriptive/language advice publications; the divergence mentioned here therefore solely applies to the early modern period). Instead, the early grammars written on Dutch were certainly prescriptive in their reception but not so much in their use of metalanguage.

As a final remark, we want to point out that future research has to find solutions for the methodological issues mentioned here, which do not just touch upon this study but on prescriptivism research in general. The challenges encountered in this study indicate that, to truly grasp what happens in normative publications, we need to find a better way of eliciting the different stances towards variation encountered in normative works (see de Vos and van der Meulen 2021 for a first attempt at this), their implicit or explicitness, and their prescriptive or descriptiveness, which can only truly be determined by relating it to actual language use of the period in question, as attempted by Ayres-Bennett (2019). As a first step towards this goal, we argue that our annotation method is best integrated into Ayres-Bennett’s (2019) model for evaluating metalinguistic texts as a systematic way to assess the second aspect in her model, whether the work is prescriptive in its choice of metalanguage. Integrating both methods might even solve some of the issues mentioned above as, for example, the assumed ostensibility of the descriptive nature of the metalanguage in normative works from the formative period of Standard Dutch could be accounted for in other aspects the model (e.g., whether it is prescriptive in its effect). The overview of the types of arguments provided in this paper can additionally serve as a starting point for a study into the effect of prescriptivism on language use, for metalanguage has been described as a possible factor of influence (e.g., Anderwald 2012).

5. Summary

Although certain issues within the method need to be addressed in future research, our study has obtained valuable insight into the characteristics of the codification of Standard Dutch in the early modern period. We quantitatively substantiated the qualitative claim that morphological choices made in this formative period of Standard Dutch (ca. 1550–1650) were often motivated by an appeal to a Latin ideal or to the ‘one form, one function’ principle (e.g., Dibbets 1995, pp. 331–33; Van der Sijs 2021, pp. 370–71). On the basis of this study, we can add that appeals to consistency within the language, to beauty, to conformation to grammatical rules, and to its reflection of frequent use or the language use of certain authorities were also characteristic of the language ideology on grammar of this time period. In these linguistic values, we see the roots of the standard language ideology that came into existence in the decades around 1800 (Rutten 2016, 2019); more precisely, the part of that ideology that connects ‘language’ with ‘norm’ (i.e., Hüning et al. 2012’s ‘correctness ideology’) and that bestows value on the language’s regularity. This was the focus of the shaping of Standard Dutch in its initiating phase. The additional connection to social identity (‘language and nation’, Hüning et al. 2012, p. 22; see, also, Rutten 2016, 2019), that forms a second and crucial part of the standard language ideology, played no major part in the arguments used in this time period.

Our data additionally show that the distribution of arguments used in the early modern period considerably differs from the dominant idea in the (post)modern period (1917–2016)—albeit applicable to all levels of language, not just grammar—that language should be pure and grammatical (Van der Meulen 2020). Moreover, our results indicate two further important differences between the early modern period and the (post)modern period with regard to their codification of Standard Dutch: the latter period showed a higher degree of consensus and, therefore, of canonisation of the normative discourse than the former period, and the nature of the metalanguage used in normative publications changed from ostensibly descriptive or implicitly prescriptive in the 16th and 17th centuries to explicitly prescriptive in the 20th and 21st centuries (cf. Van der Wal and Van Bree 1992). This indicates that, in terms of the metalanguage used, the normative discourse in the formative period of Standard Dutch was in between description and prescription.