Dialects, registers and intraindividual variation: Outside the scope of generative frameworks?

Abstract

This article explores intraindividual microvariation in dialect syntax. We argue that in many cases the speaker has internalized a different (sub)grammar for each dialectal variety, in line with the hypothesis of universal bilingualism and parallel grammars argued for by Roeper (1999 et seq.). We discuss the question of how we can distinguish parallel grammars from optionality within one grammar, suggesting that the identification of correlating contextual factors might be a promising criterion. However, we also explore a more subtle type of variation, namely cases where a standard variety influences a potentially more vulnerable non-standard variety in a way that makes it exceedingly difficult for the language user and even for a trained linguist to discern what is what. We discuss whether or not these properties should be analysed as properties of another subgrammar (the standard grammar) or as fully integrated (albeit acquired) properties of the non-standard dialect.

1. Introduction

Studying almost any contemporary society readily reveals that examples of individual multilingualism seem to be ubiquitous. Not only do most people master more than one language, but even monolingual speakers master different registers. Many people also master different dialects of the same language, and sometimes they seem to mix different linguistic systems, both different dialects and different languages.

The classical Principles and Parameters framework (P&P; see e.g. Chomsky 1986, 1988, 2000) may not seem adequate for describing the syntactic system in the mind of an active multilingual speaker, especially since classical P&P and its predecessors were designed to accommodate an idealized monolingual language user. Chomsky (1965:3) claims that ‘Linguistic theory is concerned with an ideal speaker–listener in a completely homogeneous speech community’. Multilingual speakers or speakers with what seem to be mixed linguistic systems would be excluded as research objects in this theory: ‘The language of such a speech community would not be “pure” in the relevant sense because it would not represent a single set of choices among the options permitted by U[niversal] G[rammar] but rather would include “contradictory” choices for certain of these options’ (Chomsky 1986:17).

During the last 20 or 30 years, generative grammar has developed various theoretical and analytic devices in order to accommodate different phenomena of multilingualism, including phenomena that seem to be evidence for grammars with features from several linguistic systems co-existing in the mind of the same language user. Modifying some of the fundamental assumptions of classical P&P has been necessary in order to develop adequate theories for the study of non-standard dialects as well as intraspeaker variation, that is, for the ubiquitous linguistic variation we observe in the same individual. To some extent, any individual is multilingual; to quote Cook & Newson (2007:223), ‘The issue is really whether it is proper to set universal bilingualism to one side in linguists’ descriptions of competence or whether it should in effect form the basis of the description from the beginning’.

It would be a misrepresentation to claim that even classical P&P linguistics denies the existence of multilingualism or that the theory cannot in principle describe and explain the multilingualism of individuals. On the contrary, Chomsky (1988:187) says:

This [i.e. multilingualism] is a very important question which I have been pretending all along does not arise. The question is a very mysterious one. … The child learns different languages, say Spanish at home and English in the streets. But in fact, the problem is really more general, because every human being speaks a variety of languages. We sometimes call them different styles or dialects, but they are really different languages, and somehow we know when to use them, one in one place and another in another place. Now each of these different languages involves a different switch [parameter] setting … Somehow, young children have a theory of society and a theory of language, and they are able to link them up in some fashion to indicate that you speak this language in this social situation.

The ‘linking up’ mentioned towards the end of this quotation is important. It means that a multilingual language user typically selects the appropriate linguistic variety for whatever context he/she happens to be situated in.

In this article, we want to emphasize that any language user knows (i.e. recognizes and accepts), and even actively uses different styles, registers, dialects, languages, and they are able to use the right variety in the appropriate contexts, seemingly without much effort. In what follows we will discuss the phenomena of intraindividual variation and dialect syntax, and we will try to pinpoint some of the problems encountered in that specific area. In the remainder of this article, we will first (Section 2) present some data of individual microvariation illustrating different syntactic registers. In Section 3, we briefly discuss Roeper’s theory of universal bilingualism (Roeper 1999, 2016; Amaral & Roeper 2014) as a way of making sense of the variation mentioned. In Section 4, we focus on dialect syntax, we discuss the contemporary theoretical status of parameters in Section 5, and relate the more general notion of ‘contamination’, a subtype of cross-linguistic influence, specifically to dialect syntax in Section 6. Section 7 concludes the article.

2. Individual microvariation: Some data

The kind of multilingualism mentioned above is typically the focus of dialect and (other) microvariation research, and has been a pivotal point of research in sociolinguistics. In generative grammar frameworks, until quite recently, it has however often not been heeded to the same extent. This might be due to the tacit assumption that a language user’s choice of register or style or any other ‘minor’ variation is at most relevant at a lexical, not a syntactic level, i.e. that such phenomena are mainly concerned with what words and lexical elements the language user chooses to employ, which is usually considered to be beyond the scope of generative grammar.

However, microvariation also includes syntactic phenomena, as has been made evident by several recent large-scale projects recording dialectal syntactic variation; for the Nordic countries we mention NORMS, ScanDiaSyn and N’Clav.¹ Syntactic variation also clearly comes into play as regards different registers. For instance, one set of syntactic rules applies to written standard norms, another set to other contexts. In standard written prose, a Norwegian language user would hardly ever drop the subject in a finite clause (*Fortsetter bombingen ‘continues the bombing’), but the phenomenon is a hallmark of newspaper headlines (Fjeldstad 2000). Moreover, in so-called ‘diary syntax’ (Haegeman & Ihsane 2002), subject drop is very common even in languages where this would yield ungrammatical results in ordinary written prose (Traff ham igjen i dag ‘met him again today’). In informal speech, both the subject and the auxiliary may easily be left out ([Har du] Sett’ n Ola nå? ‘[Have you] seen Ollie around?’); the same is the case for the object in topic drop ([Den filmen] Så æ i går, lit.: ‘[That film] saw I yesterday’), see e.g. Nygård (2018).

One of the more peculiar facts is that Norwegian hymns may use ‘German’ word order: in embedded clauses, the finite verb is clause-final (Om du deg skjuler når angsten meg plager, lit.: ‘if you yourself hide when fear me torments’, Barstad 2000), although in standard register the verb would appear before its complements in Norwegian. A similar fact can be observed in what we refer to as ‘festive syntax’, observable in greetings and salutes in newspapers (Hurra for Mia som på Elveng bor! ‘Hooray for Mia who on Elveng lives!’). Clause-final finite verbs, or more generally, non-V2 word order even in main clauses is obviously considered more ‘festive’ than ordinary V2; this type of syntax is accepted in these contexts, but rejected in most others, see e.g. Eide & Åfarli (2007), Eide & Sollid (2011). In poetry, we find unusual word order in phrases (Tre katter små ‘three cats small’). All these syntactic phenomena are often lumped together under ‘poetic free word order’. A thorough investigation of the data, however, reveals that poetic word order does not open the door to syntactic anarchy. On the contrary, the syntax of these registers is in fact subject to quite strict rules. The relevant syntactic system is simply different. For instance, in the hymn example given above, the verb cannot be clause-initial in the når-clause: *Om du deg skjuler når plager angsten meg, lit.: ‘if you yourself hide when torments fear me’ (Barstad 2000). Likewise, Små katter tre ‘small cats three’ would be ungrammatical however poetic the context (Eide & Sollid 2011).

A different syntactic system is also at play in a language user using a non-standard dialect. A number of syntactic features of Norwegian dialects are only acceptable in dialectal contexts. When lexical items from the standard dialect are used, those features become unacceptable and ungrammatical. The dialect spoken in Skedsmokorset not far from Oslo is particularly illuminating in this regard (Reite 2011). Speakers of this dialect as a rule master the ‘standard’ Oslo dialect in addition to their local dialect. When they use the standard wh-word, they use obligatory V2 in direct wh-questions, as expected: Va sier’u? ‘what say you?’ vs. *Va ru sier? ‘what you say?’. However, when they use the wh-word of the local dialect, they allow non-V2 in such cases: Å sier’u? ‘what say you?’ and Å ru sier? ‘what you say?’ The use of non-V2 word order in direct wh-questions (which is a special feature of many Norwegian dialects, but not of standard varieties) is licensed only by the local dialect setting, and crucially, by the corresponding local wh-word. This shows that both dialect systems co-exist in the mind of the speaker.²

Underlying this flexible use of different (syntactic) varieties are corresponding grammars. This means that the individual language user has internalized several different grammars that he/she is able to put to use in the appropriate contexts. This is the thesis of universal bilingualism (Roeper 1999 et seq.), to which we now turn.

3. A theory of universal bilingualism

According to the theory of universal bilingualism developed by Thomas Roeper, all humans are massively multilingual (Roeper 1999; see also Kroch 1994; Yang 2002, 2016; Roeper 2007, 2016; Amaral & Roeper 2014 for similar and/or related ideas). The notion of multilingualism is thus extended beyond mastering multiple national or otherwise clearly sociologically distinct languages, which has been the traditional understanding of what being multilingual amounts to. The quote from Chomsky (1988) given above signals a similar shift from the traditional understanding, see also Chomsky (2000:59): ‘Whatever the language faculty is it can assume many different states in parallel’.

Roeper’s extension seems to accommodate microvariation data like those discussed above very well. Thus, parsing (or writing) hymns involves a different I-grammar from the one involved in reading scientific prose. Likewise, speaking one’s local dialect involves a different I-grammar from the one we use when speaking or understanding the national ‘standard’ dialect. It is the extralinguistic context that triggers which I-grammar is activated, ‘somehow we know when to use them’ (Chomsky 1988:187), and a given variety also typically involves characteristic lexical choices, which supports the idea that different grammars or I-languages are involved.

Roeper also claims that more ‘hidden bilingualism’ is found in many cases where free variation between syntactic structures is encountered. While other theories resort to the notion of optionality or optional rules within one grammar in such cases, Roeper argues for two contradictory subgrammars. He supports his theory with the optional V2 rule in English (among a range of other phenomena). As is well known, English does not employ an obligatory V2 rule; in most cases, putting the finite verb in second position leads to ungrammaticality (*He reads often Nietzsche). However, there is a context-triggered V2 rule in children’s stories and narratives with most verbs of telling or saying (‘I am melting,’ shrieked the witch). Verbs of saying thus trigger a different set of grammatical rules than other verbs, e.g. other clause-embedding verbs, in this context. Such examples, Roeper says (1999:6), show that

A person has numerous grammars: every lexical class with rules that are incompatible with another class should constitute a separate grammar. It sounds unwieldy and implausible to argue that a person has a dozen grammars. The essence of this assertion may, nonetheless, be true. It implies that the notion of a grammar should change to a more local conception.

Language development is another area where different grammars co-exist, according to Roeper. Most children acquiring English go through a stage where structures with an accusative subject in main clauses (Me want) co-exist with the ‘adult-like target grammar’ (I want). Roeper (1999:170–171) argues that there is no optionality with respect to the morphological case of the subject; instead, it would be more correct to describe this situation as a stage where two different, and contradictory, grammars compete. One is specified to always give the subject accusative case (in the relevant clause type), while the other always dictates a nominative subject (in the corresponding clause type). All development of this type should be seen as competition between grammars; diachronic studies should be analysed from the same perspective: ‘I proceed from the assumption that wherever one finds a continuum, or historical gradualism, a more refined level of analysis will reveal discrete phenomena’ (Roeper 1999:4).

We will return to Roeper’s approach to bilingualism in Section 6 below. In particular, there we will discuss one question that Roeper’s approach raises, namely how to distinguish two closely related grammars from one grammar with optionality.³ However, at this point we will focus on microvariation pertaining to dialect syntax.

4. Dialect syntax

A dialect is often understood to be a linguistic variety common to a group of people in a geographically (or socially) defined area. According to this understanding, a dialect is a collective phenomenon, i.e. something that is common to a group of people. In generative grammar, however, a language is defined as an individual phenomenon; it is an object generated by an internalized language or grammar, an I-language or I-grammar. The implication is that a dialect is an individual phenomenon since a given dialect is a language. In other words, a dialect is an idiolect, which means that there are as many dialects as there are individuals. Based on this, Kayne (1996:xiv–xv) presents some controversial ideas (our italics):

[I]t is often estimated that the number of languages presently in existence is 4000–5000. Such estimates must evaluate the contribution of Italy as one. […]

[…] that in Northern Italy alone one can individuate at least 25 syntactically distinct languages/dialects solely by studying the syntax of subject clitics. […]

A very conservative estimate would be that present-day Italy has at least 500 syntactically distinct languages/dialects. 500,000 would in consequence, I think, then be a very conservative extrapolation to the number of syntactically distinct languages/dialects in the world at present.

It is possible to arrive at a much more radical reevaluation based on the following question: Can anyone think of another person with whom they agree 100 percent of the time on syntactic judgements (even counting only sharp disagreements)? […]

[I]t is entirely likely that no two speakers of English have exactly the same syntactic judgements. In which case there must be many more varieties of English than is usually assumed.

If we accept the characterization of a dialect as an I-language or idiolect, a dialect understood as a collective phenomenon is a delimited set of I-languages/idiolects that share certain traits (see also Eide & Sollid 2011). This is in line with the way a dialect is characterized in Chambers & Trudgill (1998:5): ‘“Dialect” […] refers to varieties which are grammatically (and perhaps lexically) as well as phonologically different from other varieties’.

The definition of a dialect as an I-language/idiolect naturally opens the possibility for including second language varieties into the class of dialects; see the following quotation from Berggren & Tenfjord (1999:21) (our translation):

Det gjør det rimelig å omtale mellomspråk som varieteter […]. Mellomspråk er imidlertid ikke gruppespråk, de er unike for hver innlærer. Slik sett har de karakter av idiolekter. Dette aspektet ved mellomspråk lå til grunn for Corders gamle omtale av mellomspråk som “idiosynkratiske dialekter”.

[This makes it natural to talk about interlanguages as varieties […]. Still, interlanguage is not group language; it is unique to each individual learner. Thus, interlanguages may be seen as idiolects. This feature of interlanguage was fundamental to Corder’s characterization of interlanguage as ‘idiosyncratic dialects’.]

In other words, interlanguages are I-languages that typically exhibit even more drastic idiolectal variation than do ordinary dialects.

Given the discussion above, we can characterize a dialect as an individual I-language/idiolect. A dialect is one of several I-languages in the mind of the individual language user. Thus, linguistic data from the same informant can be generated by several different grammars. Distinguishing between the grammars can be relatively easy when it comes to multilingualism involving languages that are clearly different, such as Norwegian and English, for example. However, it can be difficult when a standard dialect co-exists with a local dialect, the two varieties have largely the same lexical base, and the pronunciation of the local dialect might not differ significantly from that of the standard dialect (i.e. if it is easy to ‘translate’ between the two dialects with respect to pronunciation). In such cases, where the two varieties are closely related, we employ the term ‘subgrammars’ for the two I-languages involved.

Distinguishing between different subgrammars in the same individual also brings up the question of parameters, a topic that has been hotly debated lately.

5. Parameters in dialect syntax?

In classical P&P theory, parametric variation was seen as the effect of grammatical macroparameters being built into the structure of UG and thus linked to general properties of the grammar, for instance word order. It was soon proposed (Borer 1984) that the source of parameter variation is restricted to the functional lexicon, i.e. that parametric variation should be explained as the effect of different languages employing different functional categories (or the different make-up of functional categories in different languages). Thus, word order variation was no longer stipulated as an effect of grammatical macroparameters built into UG, but as an effect of functional categories triggering movements that yield certain word orders.

A third stage in the understanding of parametric variation was reached at the beginning of this century when it was proposed that parametric variation may be the effect of so-called third factor properties (Chomsky 2005) of linguistic interfaces, not part of UG or core grammar. For instance, word order variation may be seen as an effect of linearization being imposed on a hierarchical linguistic representation, not because UG says so, or because the relevant core grammar employs certain functional categories, but simply because the medium of speech or writing demands that the linguistic string is linearized as either AB or BA. Also at the beginning of this century, Newmeyer (2004, 2005) argued that parameters should be jettisoned from generative grammar, and that their job should be done by language-particular rules, a move that was criticized in Roberts & Holmberg (2005), among others.

How does the microvariation typically encountered in dialect syntax fit into these different conceptions of the nature of parametric variation? Henry (1996:79) says the following about the expectations about dialectal differences based on classical parameter theory:

[C]lassical parameter theory, under which the effects of parameter setting are seen in a range of areas of the grammar, predicts that differences between dialects will be comparatively large-scale at the surface level; a difference in the setting of one parameter will have a range of effects throughout the grammar.

However, one seldom encounters such large-scale differences in dialect studies. Dialects belonging to the same ‘language’ often differ only with regard to one or a few quite minor syntactic properties; see Kayne (1996:xiii):

To the extent that one can find languages that are syntactically extremely similar to one another, yet clearly distinguishable and readily examinable, one can hope to reach a point that the number of observable differences is so small that one can virtually see one property covarying with another.

Kayne (1996:footnote 11) also says: ‘In the extreme case, one may find an isolated property distinguishing two very close dialects’. As Henry (1996:79–80) points out, this doesn’t fit with classical parameter theory:

The possibility of two dialects differing from one another in that one applies a process (for example verb raising) obligatorily while in the other it is optional, sits uneasily with current approaches.

Also, dialectal microvariation doesn’t seem to fit easily with variation explained as the effect of third factor properties either, since such effects are expected to be general and ‘macro’, not small-scale and ‘micro’.⁴

On the other hand, variation explained as an effect of differences pertaining to functional categories seems to be more promising, since functional categories are sufficiently detailed, allowing for major or minor variation. The same goes for Newmeyer’s rule-based approach. Thus, there is nothing that precludes two I-languages from differing with regard to only one property, whether that is couched in the terminology of differences regarding functional categories or just rule differences.

In contemporary P&P theory (where authors gradually tend to adhere to the term ‘biolinguistics’ instead) there seems to be a general consensus that the classical grand-scale macro-parameters have little to offer when it comes to questions of multilingualism, bidialectalism, and language mixing, areas that the field is now willing and eager to address. Instead, proposals have been made that parameters come in different sizes; consider the microparameter approach of Kayne (2000, 2005) versus Baker’s macro parameters (2001, 2008a). More recently, Roberts & Holmberg (2010) and Roberts (2019), for example, advocate the idea that parameters are organized in hierarchies, where the size and the effect of a parameter is determined on its position in this hierarchy. Moreover, the effect of parameters is also determined by other factors, such as feature economy.⁵

It is not important for us in this study to take a stand in the general parameter debate; we will just note that Newmeyer’s rule-based approach or an approach that assumes that parameters are restricted to functional categories seem to best accommodate dialectal microvariation. Moreover, with approaches such as Roberts (2019) mentioned above, the opposition between parameter-based and feature-based approaches to variation is in a sense dissolved.

Now we turn to a perhaps still harder problem, namely how do we know when we are faced with one grammar with optionality as opposed to two closely related subgrammars? We will approach this problem via a notion we will refer to as contamination.

6. Dialect syntax and cross-linguistic influence

Cook & Newson (2007) refer to an earlier study by Balcom (2003) in Cook (2003) of a phenomenon that we might call contamination, a special subtype of cross-linguistic influence. This concept describes a situation where the grammaticality judgements of a language user are influenced by the acquisition of a second language:

[M]ulti-competence has raised the issue of how the second language knowledge in the [final state] affects the other component of the final state – the L1 [first language] […]: French speakers who know English react against French sentences using the middle voice compared to those who don’t know English […].

Un tricot de laine se lave à l’eau froide.

*A wool sweater washes in cold water.

(Cook & Newson 2007:239)

Sunde (2019) likewise demonstrates how young Norwegian speakers are clearly influenced by their massive English input when giving grammaticality judgments on Norwegian constructions, judgments clearly deviating from those of their elder peers. One may very well posit a comparable contamination between different varieties, dialects, and standards of a national language. This fits well with the findings of Leonie Cornips (see especially Cornips 2006) who has studied Dutch dialects for a number of years. In some cases, informants identify a construction as being part of their local dialect, while other aspects of the study reveal that the construction does not belong to the dialect, but to the standard norm. Similarly, Henry (2005:1599) mentions (in relation to her work on Belfast English) that

[…] for non-standard dialects, the judgements of those speakers are likely to be clouded by their knowledge that many of the structures in their dialect are considered ‘ungrammatical’ or ‘incorrect’ by speakers of the standard variety. This may have been reinforced within the education system, where such forms are often highly stigmatized.

Below, we will give two examples of such potential contamination in the I-grammars of Norwegian dialect speakers, and then we will speculate how we might deal with such phenomena theoretically.

The first example is illustrated in (1)–(3), where (1) exemplifies the word order pattern in wh-questions in Standard Norwegian (obligatory V2). The example is written using Nynorsk spelling.

Now consider (2).

The dialects in south-eastern Norway typically show the pattern in (2a), which has the same word order as standard written Norwegian (obligatory V2). The sentence in (2a) is written using Bokmål spelling. However, many other Norwegian dialects (especially along the western and northern coast) allow the word order in (2b), i.e. no V2, see also Lundquist, Westendorp & Strand (2020 – this volume). This is the case for the dialect in Halsa, Nordmøre (Åfarli 1986:98). The spelling used is adapted to the pronunciation of the local dialect.

Many of the dialects that allow (2b), i.e. non-V2 wh-questions, also permit the word order in (2a), i.e. V2. This is the case for the dialect in Halsa, Nordmøre, as shown in (3).

The two possibilities in (3) raise the issue of how to interpret such optionality: Is this genuine optionality within one grammar, or should we interpret this interchange as stemming from two different grammars, where (3a) is really an effect of an interfering standard dialect grammar?

The other example of potential contamination where we find two patterns that may potentially belong to two different subgrammars is presented by Norwegian verb particles, as discussed in Aa (2015). It is traditionally taken for granted in generative research that whereas Swedish has obligatory left-hand particles (Han kastade ut hunden ‘He threw out the dog’) and Danish has obligatory right-hand particles (Han kastede hunden ud ‘He threw the dog out’), Norwegian grammar has more or less free variation between the two patterns, as seen in (4).

However, based on an investigation of the traditional dialectological literature on verb particles coupled with his own thorough empirical investigation of contemporary Norwegian dialects, Aa found that the spoken dialects strongly prefer the pattern with left-hand particles in (4a), and that the right-hand pattern in (4b) is more common in standard written and spoken Norwegian. He then hypothesizes (Aa 2015:237–243) that what seems to be free variation is really the result of a mix-up and alternation between two separate grammars, where the right-hand pattern is the result of historical Danish influence during the centuries when Norwegian was a part of the Danish kingdom. It is particularly significant that during the time when Norwegian introduced compulsory schooling for all children in the 19th century, Danish was the language of instruction. Also, Danish was used during that time when Norwegian established a literary language. Thus, Aa (2015:243) hypothesizes that

[w]riting Norwegian today is thus to master the conventions of a syntax based on both NO_DI [Norwegian dialects] and Danish. When we take this into account, the alternation in [(4) above] can be considered a switch between these two separate and deterministic grammars, so that it is a grammar alternation rather than a particle alternation. In that case, the apparent free alternation found in NO_ST [standard Norwegian] is the result of bilingualism.

The similarity with the wh-question data shown earlier is obvious. In both cases, there is potential contamination in the sense that – pre-theoretically speaking – the dialect/language lets in alien constructions. This fits well with the findings of Cornips mentioned earlier. In her studies, informants sometimes identify a construction as being part of their local dialect, while other aspects of the study reveal that the construction does not originally belong to the dialect, but to the standard norm.

This of course raises the question of how we can know that a given construction or property really belongs to a dialect or a ‘norm’ (i.e. a standard), and moreover, how we should interpret such phenomena theoretically. For instance, are we faced with genuine optionality within one grammar, or should we interpret the data as stemming from two different deterministic grammars, where one pattern is an effect of an interfering standard dialect grammar?

In cases where the use of the construction or property (or more generally the register, variety) in question is correlated with a specific contextual setting, it seems reasonable to count the two varieties as stemming from different subgrammars.⁶ For instance, hymn syntax is used in hymn contexts, or festive syntax in festive contexts. Also, the dialect syntax switch in Skedsmokorset reported in Reite (2011), mentioned in Section 2, seems to be conditioned by contextual factors, and moreover, it is correlated with lexical accommodation (different for the two varieties), as well. In that case too it seems reasonable to assume that two different subgrammars are used. Such cases are not in principle different from two national languages being used by a speaker in different contexts.

The examples given in this section are different, since there seems to be free variation between constructions or properties without any obvious accompanying contextual factors being discernible. Assuming that such factors are absent, we could still be faced with ‘hidden multilingualism’ (Roeper 1999) with two different subgrammars involved, or, alternatively, we could be faced with genuine optionality within one grammar, even though the historical impetus for this optionality is, say, the influence that a dominating standard dialect or written standard have on a local dialect.

If we assume hidden multilingualism and different subgrammars in such cases, the most economical approach would be to assume that grammars are deterministic and do not permit optionality. Instead of assuming that we have potentially three different grammars (a deterministic one with V2 or right-hand particles, a deterministic one with non-V2 and left-hand particles, and a non-deterministic one that allows both), it is preferable to adopt the more restrictive hypothesis that only the two deterministic alternatives are available. Thus, considerations of theory economy favour the deterministic solution. If this is correct, we must conclude that the two options in free variation patterns really belong to different subgrammars. Still, such an analysis is driven by conceptual considerations of theoretical economy, but it might not be empirically and factually right. There is still the possibility that in fact only one grammar that allows optionality is involved.⁷

To solve this issue would require a much larger study (to determine the empirical facts on a larger scale) and a much more thorough debate on our current linguistic models. In short, this is a question that cannot be settled in the format of this short paper.

However, we will briefly offer some related thoughts on how to cope with grammaticality judgements and working with informants in dialect syntax. We suggest that it is a reasonable research approach to assume that two varieties that are correlated with specific contextual factors indeed reflect different (sub)grammars in the mind of the speaker. Thus, a speaker who switches to Swedish in a Swedish setting, and to English in an English setting, clearly invokes two different grammars. Likewise, a speaker using hymn syntax in a hymn setting, or a poetic syntax in poetry, invokes a different grammar from her or his typical everyday grammar, and a speaker of the two varieties in Skedsmokorset invokes two different grammars as well, since there are specific contextual factors (and lexical choices) that go along with the use of each variety. In such cases, we are faced with instances of universal bilingualism involving different subgrammars (Roeper 1999).

As for cases like the two sets of data discussed in this section where no contextual factors can be found, the researcher must nevertheless be open to the possibility that there are such factors after all. Thus, a reasonable strategy when working with informants will be to always be looking for such factors, in order to determine whether one grammar (with optionality) or two subgrammars are involved.⁸ This implies that a set of data from the same informant (recordings, written texts, grammaticality judgements) might in principle display features of contradicting subgrammars (see Eide & Åfarli 2007, Eide & Sollid 2011). This should make linguists proceed with caution in their work with informants and data (see also Henry 2005).

If a data set consists of data from different, and potentially contradictory, subgrammars, constructing a single coherent underlying grammar would be akin to constructing a single picture from a box of puzzle pieces, where each piece may in principle belong to a different puzzle. This would be a challenging and ultimately a futile task. On the other hand, if a grammar in fact allows optionality, constructing different puzzles where there is in fact only one would also be wrong and futile.

How to deal with informants and judgements from non-standard dialects is discussed in an illuminating and interesting way in Jamieson (2020). Jamieson discovered that younger speakers of Shetland Scots exhibit a kind of perceptual hyperdialectalism (see also Cornips & Poletto 2005), i.e. that the younger speakers of the dialect tend to extend acceptability of the variable in question to contexts where older speakers did not accept it. Jamieson further argues that this can be explained as an effect of the younger speakers being linguistically insecure and therefore unreliable in their judgements, which again highlights the importance of proceeding with caution when dealing with informants who are speakers of non-standard dialects.

7. Conclusion

We took as our point of departure the contrast between the ‘ideal speaker–listener’ of traditional generative grammar and the apparent chaos of linguistic variation that the researcher is faced with in the real world.

We very soon zoomed in on intraindividual microvariation and explored phenomena where the language user knows, masters and actively uses different closely related varieties, like different registers and dialects, which they are able to use in the appropriate contexts without effort. We argued that in many cases where we encounter such microvariation, the speaker has internalized a different (sub)grammar for each given variety, in line with the hypothesis of universal bilingualism and parallel grammars argued for by Thomas Roeper. In particular, we explored the question of how we can distinguish parallel grammars from optionality in one grammar, suggesting that the identification of correlating contextual factors might be a possible criterion.

However, we also explored a more subtle type of variation which is particularly relevant in dialect syntax, namely cases of what we call ‘contamination’ where a standard variety influences on a more vulnerable non-standard variety so as to impress some of its own I-language properties on the I-grammar of speakers of the non-standard variety in a way that makes it exceedingly difficult to discern what is what. In particular, should properties that can be traced back to such contamination be seen as properties of another subgrammar (the standard grammar), or have they become an integrated part of the influenced non-standard dialect, and should therefore be analysed as effects of the grammar of that dialect? This question is explored, but not really answered; although we recommend that the researcher exercise the utmost care when working with non-standard dialect data.

The relevance of the vexed notion of parameter for the analysis of dialectal variation and other types of microvariation has also been discussed. Although the degree of commitment to the old parameter notion was toned down, the concept of parameters is in the process of migrating towards feature-based explanations, and we believe feature theory has an important role to play in the understanding of I-grammars of dialects, syntactic registers and intraspeaker variation.

This article has only scratched the surface of the many problems concerning research in microvariation and dialect syntax. Hopefully, it has contributed to illuminating some of the pressing problems and placing them on the agenda for future research.