How to explain linguistic variation and its role in language change.

Linguistic variation is a feature that characterizes every natural language. In this paper, I focus on linguistic variation and its importance and its role in language change (LC). To determine its importance and role, one must distinguish between emergence and diffusion of variants because both dimensions provide different contributions to LC. The emergence and further development of new variants is a process that often (though not always) displays the form of a cycle, whereas the diffusion within a speech community often follows an S-curve form. Both dimensions are important for LC, but they relate to very different aspects. In this paper, I will treat variation with respect to its role in LC. Variation, in general, is a crucial factor in LC in two respects: First, it is the result of LC and second, it can trigger LC.

Keywords: language change; pronoun cycle; reanalysis; emergence versus diffusion; iceberg principle

1 Introduction

Natural languages display significant variation, i.e., there is often more than one possibility to express the same content. In German, for instance, the statement MY BROTHER'S HOUSE can be expressed at least in three ways (1a)–(1c):

(1)

a.

das Haus meines Bruders

the house my- gen brother- gen

b.

das Haus von meinem Bruder

the house of my brother

c.

meinem Bruder sein Haus

my- dat brother- dat his house

Different linguistic subdisciplines are concerned with variation of this kind. Sociolinguistics takes for granted the existence of variation and investigates the sociolinguistic significance of variants because they "convey very different social meanings" ([9]: 4). Applying such an approach to the examples in (1), one can identify (1a) as a prestige variant since it occurs in Standard German, while (1b) and (1c) are substandard variants belonging to more colloquial or dialectal varieties. Concerning diachrony, sociolinguists are interested in, for instance, questions regarding the replacement of one variant by another in the standard and the social conditions for this to occur.

Current formal approaches do not take the existence of variation for granted. They assume that variation is constrained by grammatical principles and may therefore reflect different structures and/or grammars. Formal linguists may be interested in investigating which structural and grammatical or morphosyntactic changes underly the development of certain variants, cf. [34], [35] for the historical development and synchronic variation of (1c). Variation is therefore a very important topic in formal linguistics, especially microvariation since [28], because it allows giving an answer to the question of "what are the minimal units of syntactic variation" ([28]: xiii).

However, variation is not only treated differently in different linguistic subdisciplines or approaches; it is itself a highly heterogenous phenomenon ([36]). A variation on the language level typically involves cases where, for instance, different varieties or styles of a language use different constructions to express the same content, as exemplified with (1a) to (1c) above. That does not necessarily imply variation on the individual level because it is possible that, to stick with our example (1a)–(1c), some speakers of German always use the standard form (1a) and others (1b) for inanimate possessors and (1c) for animate ones. Both groups of German speakers would then have a homogeneous competence that does not display any variation (with respect to possessive constructions). Therefore, inter-speaker variation (i.e., on the level of speech communities or languages) does not necessarily involve intra-speaker variation (i.e., on the level of I-grammars).

Another crucial difference is between the emergence and diffusion of variants. The emergence of new variants concerns (I-)grammars because, for example, the development from (1a) to (1c) in the history of German is an instance of grammatical change that involves developments concerning genitive attributes and possessive pronouns (for details cf. [35]). The diffusion of variants within speech communities is in some cases a topic for sociolinguistics since replacing one variant with another often has to do with prestige and with adaptation to our interlocutors. As [10]: 370) puts it: We are "selecting linguistic variants contingent upon the setting in which we are speaking and on not only our own class, sex, age, ethnicity, style and much more, but also contingent upon all those things in the people we are speaking to."

However, linguistic change cannot always be explained by sociolinguistics. The reason why one variant wins over a competitor could also be that it has certain advantages over the latter. According to [25], the reason why do-insertion replaced the older variant with verb fronting in English had to do with computational and morphological simplifications (cf. also [42]).

In this paper, I will treat variation with respect to its role in language change (LC). Variation, in general, is a crucial factor in LC in two respects: First, it is the result of LC, and second, it can trigger LC. In the following, I will discuss an example of synchronic variation that is the result of LC (Section 3) and cases where synchronic variation that gives rise to structural ambiguity is an indirect source of LC (Section 4). In Section 2, I will introduce some basic concepts regarding LC and variation and discuss what formal and functional approaches can and cannot contribute to explaining LC. In particular, I argue that only formal approaches can explain the emergence of variation, while the spread of individual variants within a linguistic community (also) depends on sociolinguistic aspects.

Here, a note on sociolinguistics is in place: Since this volume compares formal and functional approaches to variation, it needs to be explained why I am contrasting formal approaches with sociolinguistics. In treating sociolinguistics as a functional approach I follow, among others, [31], who argue that functional linguistics (in their case systemic functional linguistics) is in reality functional sociolinguistics because it develops "a functional model of language [which] is strongly implicated in the design of a model of the social" ([31]: 12). According to [31]: 120), functional linguistics "is concerned with explaining language in relation to how it is used" and sociolinguistics is defined by [9]: 1) as "the study of the social uses of language," so sociolinguistics can be seen as a special kind of functional linguistics. Both functional linguistics and sociolinguistics see themselves as usage-based, and they also have in common their opposition to formal approaches of a generative nature.

2 Variation and LC: general remarks

LC and variation are inextricably linked. With respect to LC (especially grammar change, cf. e.g. [26]; [19]), we have to distinguish between the three steps in (2) ([42]: 75):

(2)

a.	the emergence of a variant,
b.	its (possible) spread within the grammatical system (extension into new contexts),
c.	its diffusion within the speech community.

The emergence of a variant is a process at the level of I-grammars, and it is, as I assume, a sudden, abrupt change ([43]). Diffusion, by contrast, proceeds at the level of speech communities, and it can last very long because it is normally a slow and gradient process. Both dimensions of LC (i.e., emergence and diffusion) often show characteristic, yet clearly different courses. With respect to the emergence of new variants and their further development, LC often has the form of a cycle (cf. [22], [23], [24]; [7]): A variant A is replaced by a variant B that eventually gets replaced by a variant C, which resembles the original variant A in some respect. Thus, A, B, and C are variants that compete for the same purpose. Synchronic variation arises when the succession of the single forms shows some chronological overlap, so that original and new forms often co-exist (cf. [3a] vs. [3b]).

(3)

​

The prototypical instance of a cyclical change is Jespersen's Cycle (JC), which describes the development of negation particles (cf., among many others, [6]: 182). JC comprises three stages (4a)–(4c): A clitic neg-particle at Stage I is complemented by a free neg-particle at Stage II; at Stage III, the original neg-particle disappears, leaving the free neg-particle as the sole negative marker. This development is a cycle because in Stage III, there is again a one-part negation – exactly as in Stage I.

(4)

a.	Stage I: clitic neg-particle
b.	Stage II: clitic and free neg-particle
c.	Stage III: free neg-particle

The emergence of the German neg-particle nicht is traditionally explained with JC (cf. [16] among others). After weakening the original neg-particle ni to ne/en, it was strengthened with the particle niht, which ended up being the only neg-particle left.

More recently, however, researchers have argued that the second stage has not existed in Middle High German (MHG), cf. [6]: 192): "From our data, there is no evidence for a stable Jespersen Stage II period in the history of German." While early MHG is still predominantly a Stage I language, classical MHG is already largely a Stage III language. This means that Stage II overlaps in time with both Stage I and Stage III. However, this does not necessarily imply that Stage II was completely absent in the development of nicht as a neg-particle. That an intermediate step in a developmental cycle is not observable in the historical data could just mean that the loss of an old variant and the diffusion of a new variant proceed in a similar slow speed so that they overlap to a great extent. It is then an illusion that an intermediate step B does not exist because the emergence of a new variant C presupposes this stage, even if it is not visible or reconstructable in the data. A developmental cycle, such as (3b) above, could give the impression that Variant B is co-existent with both A and C so that it does not correspond to a separate stage of development. However, in a cyclical development, the emergence of Variant B requires the existence of Variant A, and the emergence of Variant C requires the existence of Variant B. W.r.t. JC, this means that the emergence of the new neg-particle nicht requires the bi-partite variant ne/en ... niht as predecessor; otherwise, it could not have developed as neg-particle, that is, Stage III requires Stage II.

As historical linguists, we have to distinguish between the reconstruction of the grammar-/language-internal development and its chronology. The latter also concerns the diffusion of a new variant within a speech community: For its reconstruction, we are dependent on the availability of an appropriate amount of reliable data which we do not always have – especially for more distant periods. This seems to me to be the reason why the internal grammatical reconstruction of a development is not always exactly reflected in its chronology.

When we come to the diffusion of a variant within a speech community, we also often observe a typical course that takes the shape of an S-curve (cf. [15]; [10], and many others). Initially, a new variant occurs only rarely, followed by a strong increase in frequency that eventually levels off. Chambers describes it as follows: The combination of these three stages – initial stasis, rapid rise, and tailing off – gives a characteristic shape in graphic representations that is known as an S-curve. [...] The S-curve has [...] been observed in diffusions of all kinds (Chambers and Trudgill 1998 : 162–4), and is now established as a kind of template for change. (Chambers 2002b : 361)

The S-curve describes the spread of a new variant in a speech community. According to [10]: 355), the three stages of the S-curve can be associated with the succession of generations "whereby some minor variant in the speech of the oldest generation occurs with greater frequency in the middle generation and with still greater frequency in the youngest generation." In my (but probably not in Chambers') view, this correlation is compatible with the assumption that language acquisition is responsible for grammar change (but not for LC in general). That means that new variants emerge during the language acquisition – e.g., via reanalysis, cf. [43]. However, since this aspect is irrelevant to my argument here, I will not go into it further.

The important point is that cyclical change and S-curve change do not contradict each other, but instead describe different dimensions of LC: The succession of variants often follows a cyclical shape (3a, b above), whereas the diffusion of a certain variant into the speech community displays an S-curve. Both types of change complement each other, for only both together capture a complete LC, i.e., its initiation (emergence) and completion (diffusion).

In the following two sections, I will focus on variation being the result of and variation triggering LC. In both cases, only a formal approach can give us the right explanation, but not a sociolinguistic one.

3 Synchronic variations as a product of LC

In most cases, we almost completely lack historical data for the emergence of a variant, and we see only synchronic variation. In these cases, we can conclude that the observed synchronic variation is the result of LC that, on the other hand, can be reconstructed based on the observed variation, i.e., the variation on the synchronic level allows us to draw conclusions with respect to possible LCs. An example of this is the pronoun cycle (PC) as reconstructed in [37]. In the case of the PC, we completely lack historical data, so we depend on the synchronic variation for its reconstruction.

The relevant variation shows up in the order of pronominal subjects and objects. As is widely known, in Standard as well as in dialectal German, pronominal arguments occur in the order of subject before the object, as in (5a), while the reverse order is only possible if the subject pronoun is emphasized and therefore stressed (5b). However, there are very few dialectal varieties where the reverse order object > subject pronoun is the grammatically accepted one. (5c) is an unmarked utterance in these varieties corresponding to (5a) in other varieties.

(5)

a.

Tätst	du	ihn	besser	kennen!
did	you	him	better	know

b.

Tätst	ihn	DU/*du	besser	kennen!
did	him	YOU/*you	better	know

c.

Tätst	ihn	du	besser	kennen!
did	him	you	better	know
'If you knew him better!'

The existence of these variants has been known since the Wenker survey (cf. Figure 1), which was conducted from the 1870s into the 1930s – see [17] for more information. In Figure 1, the black dots symbolize locations with the reverse order object before the subject pronoun, which are mainly found in the Upper-German-speaking areas.

Graph: Figure 1: Wenker sentence 18 ([37]).

The Wenker survey provides evidence for the existence of the reverse order in the 19th century. The next time the reverse order appears is in two of the six regional atlases of the Bavarian Linguistic Atlas, namely in the Linguistic Atlas of Middle Franconia (Sprachatlas von Mittelfranken, BSA – SMF) and in that of Bavarian Swabia (Sprachatlas von Bayerisch-Schwaben, BSA – SBS). The linguistic material was compiled in the 1980s and 1990s and represents more or less the second half of the 20th century.

Figure 2 illustrates a map from the Middle Franconian Atlas and Figure 3 a map from the Bavarian-Swabian one. Red squares symbolize the reverse order.

MAP: Figure 2: BSA – SMF vol. 7, Map 126.

MAP: Figure 3: BSA – SBS vol. 9.2, Map 393.

In Bavarian Swabia, the reverse order (black squares) is attested mostly in the north where another pattern is also very frequent – namely sentences with null subjects (gray circles). Null subjects also occur in Middle Franconian, but to a lesser extent (see Figure 2). Interestingly, the areas of Middle Franconia and Bavaria, where reversed orders are attested, form a continuous region, since the north of Bavarian Swabia borders directly on the south of Middle Franconia ([37]).

The next source for the reverse order is the Syntactic Atlas of Hessian Dialects (SyHD), whose data were collected in the 2010s. In Hessian dialects, the reverse order is attested with the 2.SG and the 3.PL ([37], [38]). Figure 4 illustrates areal distribution of the 2.SG, which was tested with a modified version of the Wenker sentence 18. As seen on the map, the reverse order is a minority variant that shows no areal focus.

Graph: Figure 4: SyHD.

Although the reverse order is attested only in a very small minority of dialects or dialectal varieties, its existence cannot be denied. Overall, there is sufficient empirical evidence: It is attested with 1.SG, 2.SG, and 3.PL, and in three independently conducted dialect surveys.

The variation between the canonical and the reverse order that we observe in synchrony finds its explanation in the diachrony, namely in the PC as I will show in turn, after presenting some basic facts about pronouns in German dialects (based on [37]).

First, pronouns can have four forms, as shown in Figure 5. These forms differ from each other morphologically as well as syntactically. On the morphological level, pronouns can be full, reduced, clitic, or even null. However, each individual pronoun only has two forms – one full form and one of the non-full forms. Central Hessian is a so-called "distinct reduced" dialect where the non-full form is mostly a reduced one. The pronoun of the 3.SG.FEM 'she', for instance, has the two forms säi and se: The first one is the full form (6a), and the second one is a reduced form (6b)/(6c). Bavarian, a "distinct clitic" dialect, by contrast, shows a dichotomy between the full form and a clitic form in, for example, the 1.SG 'I': e (7a) versus i (7b). Reduced and clitic pronouns differ syntactically: Clitic pronouns (as well as null pronouns) are restricted to the Wackernagel position (WP). This is the position immediately following finite verbs in root sentences or the complementizer in embedded clauses, whereas reduced pronouns are additionally allowed in SpecCP (cf. [7a] vs. [6b]/[6c]), where, on the other hand, clitic pronouns cannot occur (cf. [7c]). Full pronouns are banned from the WP.

(6)

a

SÄI	singd	unn	daazd	de	gannse	Doag.
SHE	sings	and	dances	the	whole	day

b

Se	singd	unn	daazd	de	gannse	Doag.
she	sings	and	dances	the	whole	day
'She sings and dances the whole day.'

c

Dai	Kist	hoddse	de	Inge	gegäwwe.
Your	box	has.she	the	Inge	given
'She gave your box to Inge.'

(7)

a

Gesdan	han'e's	gmocht.
yesterday	have.I.it	made

b

I	han's	gesdan	gmocht.
I	have.it	yesterday	made

c

*e	han's	gesdan	gmocht.
I	have.it	yesterday	made
'I made it yesterday.'

Graph: Figure 5: Typology of pronouns.

As (7a) shows, the regular order of arguments within the WP is subject before object. This, however, only holds if all pronouns are of the same type. Independent evidence that the pronoun type can force a reversal of order comes from object pronouns. If both indirect and direct objects are clitic pronouns, as in (8a), the indirect object precedes the direct one. However, if the indirect object is a full pronoun, then the order is reversed, as in (8b). This is the case with the dative form of the 3.SG.MASC in Bavarian for which only the full form eam 'him' exists.

(8)

a.

Und	hod'a'da'n	geem?
and	had.he.you.it	given
'And did he give it to you?'

b.

Und	hod'a'n	eam	geem?
and	had.he.it	him	given
'And did he give it to him?'

This observation concerning object pronouns gives us a clue as to how the reversed order of subject and object pronouns came about. (9) shows the variation we observe with the order of pronominal subjects and objects. In addition to the regular order (9a) and the reverse order (9b), a pro-drop variant appears more often (9c). Sentences with only object pronouns are attested with 1.SG (see Figures 2 and 3) and with 2.SG (Figures 1 and 4).

(9)

a.	Subj > Obj
b.	Obj > Subj
c.	Obj

The areal distribution of the options in (9a) to (9c) reflects the diachronic development within the PC. The PC describes the development of pronouns and it involves the following steps (10):

(10)

Pronoun cycle (PC)

full > reduced > clitic (> null or agreement)

For the possibility that the different pronominal forms evolve out of each other along the scale in (10) through increasing weakening, a necessary precondition is that each form is syntactically in direct competition with its neighbor(s). Full pronouns can be used as weak pronouns in SpecCP (as well as reduced pronouns), where they may get segmentally reduced and would then be allowed in the WP also (just like clitic pronouns) and where they could develop into a clitic (and eventually become null). Therefore, each step in the PC is theoretically plausible and needs no stipulation. The reverse order can be explained as a certain stage in the PC. In the Linguistic Atlas of Middle Franconia (BSA – SMF, cf. Figure 2 above), where all data were transcribed phonetically, the reverse order is attested with full pronouns (11a)/(11b) and with reduced/clitic pronouns (11c). The full forms have a consonant [χ] (11a) or a long vowel (11b), whereas the reduced/clitic form only consists of a short vowel.

(11)

a.

Dęskhąndɑrįχgēm.	(021 Stübach)
that.can.you. dat.I.give
'I can give that to you.'

b.

Dëskhųndǝrīšọgēm.	(165 Dietfurt)
that.can.you. dat.I.already.give

c.

Dẹskhọndɑrįšọgēvɑ.	(078 Brundorf)
that.can.you. dat.I.already.give
'I can give that to you.'

(12a) to (12e) describe the development that led to the reverse order with 1.SG pronouns and to its eventual disappearance. At Stage I (12a), two clitic pronouns appear with the canonical order subject before object (scl > ocl); at Stage II (12b), the subject clitic has vanished and become null; at Stage III (12c), the null subject is replaced with a full pronoun that is non-clitic and thus appears to the right of the object clitic; at Stage IV (12d), the new subject pronoun is weakened to a reduced pronoun and eventually develops into a clitic that, however, initially stays in the position after the object clitic; at Stage V (12e), the canonical order subject before object clitic is restored again. Note that (11a) and (11b) correspond to Stage III and (11c) to Stage IV – thus providing empirical evidence for our explanation.

(12)

a.	/ǝ/ (=scl)	Vfin-scl1>ocl	(= Stage I)
b.	∅	Vfin-ocl	(= Stage II)
c.	redPron	Vfin-ocl>Pronfull/red	(= Stage III)
d.	scl	Vfin-ocl>scl2	(= Stage IV)
e.	scl	Vfin-scl2>ocl	(= Stage V)

The explanation of how the reverse order emerged is found in the PC, but the question that remains is why pronoun forms weaken at all. Weakening (or erosion, i.e., phonetic reduction) is a process that is often part of grammaticalization processes ([27]: 579). However, the fact that weakening occurs with grammaticalization does not yet explain why weakening occurs at all. [21]: 11) Head Preference Principle (HPP) may supply an answer: "Be a head, rather than a phrase." To see why the HPP is relevant here, one must consider that the pronominal forms of the PC differ structurally. The respective structures are given in (13a) to (13d): Pronouns are DPs containing a ϕP (or AgrD in the sense of [45]), and all pronouns start as heads of ϕP, where they may remain as clitic, or null pronouns (13b), or from where they may raise to D° as full pronouns, which then spell out the D part. As for reduced pronouns, I assume that they are ambiguous in that they can spell out both parts depending on the syntactic position they appear in. The evidence for this assumption comes from German dialects, where they are possible in SpecCP as well as in the WP (see [6b] and [6c] above). Being in SpecCP, reduced pronouns must be a phrase (13c), whereas in the WP they need to be a head (13d). Now we can see why the PC finds its explanation in the HPP: The weakening of the individual forms is a development from phrase to head – as predicted by the HPP.

(13)

a.	[DP Pronfull [ϕP Pron [NP ...
b.	[DP [ϕP Proncl/ø [NP ...
c.	[DP Pronred [ϕP Pron [NP ...
d.	[DP [ϕP Pronred [NP ...

The synchronic variation observed with the order of pronominal arguments is an example for the kind of variation that has no sociolinguistic basis or do not "convey [...] different social meanings" ([9]: 4). To understand why it exists and how it emerged, one has to investigate grammar properly.

4 Variation as a trigger of LC

In this section, I will focus on variation in the sense of different surface strings or manifestations that share the same underlying structure. This kind of synchronic variation can trigger LC rather than being itself the result of LC. This is possible because some of the surface strings are structurally ambiguous and can therefore trigger a particular type of LC, namely reanalysis involving rebracketing and relabeling ([43]).

Consider the following example: In MHG, the preposition seit 'since' could select for a clausal complement. This clause could have the form of a relative clause embedded under a d-pronoun (14a), which seems to be the 'maximal' surface manifestation. Alternative forms are first what seems to be an ordinary that-complement clause (14b) or second a clause where the complementizer is dropped for stylistic reasons as in (14c). Complementizer-drop is attested even then when the d-head of the relative clause was not absent (14d)/(14e).

(14)

a.

sît	des,	daz	ich	von	lande	schiet	(Tristan 4123)
since	that. gen	that	I	from	country	departed

b.

sît	daz	ich	von	lande	schiet	(Tristan 4123, mss. W, N, O)
since	that	I	from	country	departed

c.

sît	ich	von	lande	schiet	(Albrecht von Johansdorf, MF 92,7)
since	I	from	country	departed
'since I departed from the country'

d.

sît	des	mîn	muot	betrogen	ist	(R. von Ems, Barlaam und Josaphat, l. 8402)
since	that. gen	my	courage	betrayed	is
'since my courage was betrayed'

e.

sît	dem	sîn	vreide	sî	ze	wege	(Ulrich von Singenberg, 20, 5, 7)
since	that. dat	his	delight	be	to	way
'since his delight was on the way'

Thus, we have four surface variants that share the same underlying structure in (15). They only differ in which parts of the structure are spelled out and which ones are not – as indicated in (15) with brackets. In (14a), all parts of the structure are spelled out, whereas in (14b), the d-head of the relative clause, and in (14c), additionally, the complementizer of the relative clause remain unpronounced. In (14d) and (14e), the d-head, but not the complementizer of the relative clause, reaches the surface structure.

(15)

[PP sît [NP (des/dem) [CP (daz) ...]]]

In MHG, all four forms seem to appear in free variation with no grammatical or otherwise functional difference between them. In particular, the types represented by (14b) and (14c), i.e., without the d-head of the relative clause, occurred very frequently in MHG texts. (16a) and (16b) give an example: Both variants are used within three lines. The reason why the variant without the complementizer is used in (16a) and the one with it in (16b) may simply lie in the meter – note that Meleranz is a verse epic –, but that means that both variants do not differ grammatically in any substantial way.

(16)

a.

sît	er	dar	inne	gehûset	hât	(Meleranz, l. 1747)
since	he	there	in	housed	has
'since he lived in it'

b.

sît	daz	ich	dir	hân	verjehen	(Meleranz, l. 1749)
since	that	I	you. dat	have	told
'since I have told you'

We will not know with certainty whether the following speculation is true or not, but we could speculate about the following scenario: Both variants had a different sociolinguistic meaning in everyday speech in MHG, with the variant with the complementizer being considered more conservative and the variant without as more innovative. The innovative variant has prevailed, as we know, while the conservative one has been lost so that only the former occurs in New High German. Now, it might be the case that a comprehensive corpus study would show that the increase in the complementizer-less variant corresponds to the S-curve. However, that would not be the whole story for two reasons. Firstly, the S-curve would only describe the spread of the innovative variant within the speech community. Secondly, it would not explain the actual development on the grammatical level: the emergence of the conjunction seit 'since' from the preposition seit. This development is the result of a structural reinterpretation (cf. [17]) in which the original structure was simplified and the lexeme seit was categorically reinterpreted. Therefore, the actual development consists in a reanalysis involving rebracketing and relabeling ([43]). What triggered the reanalysis of the preposition seit as a complementizer was that utterances like (14c) or (16a) were structurally ambiguous because they match both structural options in (17) – the original structure and the new one.

(17)

[PP sît [NP (des/dem) [CP (daz) ...]]] → [CP [C° sît] ...]

The syntactic behavior of other lexemes often contributes to making an utterance structurally ambiguous. This can be seen clearly in the behavior of pronouns: If an overt complementizer is absent, pronouns in the WP attach to (nearly) any material to their left – and this could even be the preposition outside the relative clause as in (18a) to (18c). Pronouns cliticize onto the preposition sît, giving it the 'flavor' of being a complementizer.

(18)

a.

seidu	mich	chenst	so	wol	(Der Münchner Oswald, l. 1240)
since-you	me	know	so	well
'since you know me so well'

b.

sitt	sô	grôze	gâbe	gîst	(Ortnit, Stanza 118, l. 1)
since-you	so	big	gift	give
'since you give such a big gift'

c.

sîts	ir	sô	smerzent	(Friedrich der Knecht, Lied 20, Stanza 6, l. 8)
since-it	her	so	hurt
'since it hurts her so'

A comparable case is the complementizer weil 'while' that emerged from the noun Weile 'while, space of time' (MHG wîle) ([40], [43]). Here again, the source structure involves a relative clause, and we find several surface variations (19a)–(19d) depending on which parts are spelled out and which ones are not. The different surface manifestations of the same underlying structure produce a structural ambiguity, which eventually triggered a reanalysis as noted in (20).

(19)

a.

alldiewil	das	ich	uwer	pflegen	sol	(Prosalancelot 224, 27)
all-the-while	that	I	you	care	shall
'as long as I shall care for you'

b.

al	die	wîl	du	bî	mir	bist	(Parzival 485, 9)
all	the	while	you	at	me. dat	are
'as long as you are with me'

c.

Dwil	ich	off	ertrich	on	sunde	nit	enmocht	gewesen
the-while	I	on	earth	without	sin	not	neg -could	been
'because I could not live on earth without sin'

d.

di	here	cristenhait ...	sal	loben ...	Wile	ummer	diese
the	noble	christianity ...	shall	praise ...	while	always	this
werlt	gestet
world	persists
'the noble christianity has to praise, as long as this world exists'

(20)

[DP wîle [CP ...]] → [CP [C° wîle] ...]

The emergence of both seit and weil are instances of LC that have their starting point in the synchronic variation of a special kind: The respective utterances differ slightly in their surface manifestation, but they are all the expression of one and the same underlying structure (more examples of this kind are presented and discussed in [43]). In order to trigger LC, this kind of synchronic variation must create a situation where the underlying structure of an utterance becomes ambiguous, and this structural ambiguity then triggers reanalysis involving rebracketing and relabeling. Therefore, synchronic variation of this kind is indirectly responsible for LC.

Synchronic variation is the starting point of LC in that one variant out of several similar ones prevails and spreads within a speech community. In this section, I tried to show that the actual LC does not consist in the suppression of the competing variants and in the spread of the 'winning' variant within the language community, but in the structural and categorical changes that happen 'under the surface'. Further, one can only take a look 'under the surface' within a formal framework by using an approach that provides for more than one structural level.

5 Conclusions

Linguistic variation characterizes every natural language. In this paper, I discussed linguistic variation and its role in LC. I argued for a strict distinction between emergence and diffusion of variants because both dimensions provide different contributions to LC. The emergence and further development of new variants is a process often (though not always) displaying the form of a cycle, whereas the diffusion within a speech community often follows an S-curve form. Both emergence and diffusion are essential aspects of LC, but they relate to very different dimensions. Grammar in a narrower sense, i.e., particularly (but not exclusively) morphosyntactic features, changes in the way it is determined by principles expressed in formal theories. Such principles are, for example, economy principles as formulated in [21] or [40], and these principles explain why and how grammar changes. The development of nouns or prepositions into complementizers is triggered by [40]Early Merge Principle, whereas [21] HPP is responsible for the emergence of complementizers out of wh-pronouns (to mention just two examples). All of these developments can thus be explained within formal approaches, but not within sociolinguistic ones. The domain of the latter ones is the diffusion of variants within speech communities that often has the form of an S-curve. Since (some) linguistic variants bear "sociocultural significance" ([10]: 369), sociolinguists can explain the observation "that rates of change fluctuate, and that periods of relative stability can be followed by periods of considerable flux" ([10]: 364). However, this is not LC in the strict sense, but the spread of innovations in a speech community – and these are two different things.

In sociolinguistics, "[s]tudying language variation proceeds mainly by observing language use in natural social settings and categorizing the linguistic variants according to their social distribution" ([9]: 3). In formal linguistics, however, the study of variation serves other purposes. In this paper, I focused on two aspects of the relationship between variation and LC: synchronic variation as result and as starting point of LC. I tried to show that only a formal approach can detect what is going on 'under the surface'. Thus, it is only possible to give a satisfying explanation for the observed variation between the canonical and the reverse order of pronominal arguments if one refers to formal concepts and principles (like the PC or Gelderens' HPP). Similarly, one needs to apply formal concepts and principles to see why variation in the surface manifestations of the same underlying structure can trigger LC in form of reanalysis involving rebracketing and relabeling. As I have argued in [39], language seems to resemble an iceberg of which seven-eighths are invisible because they are 'under the water'; thus grammar change often happens 'under the surface' and "not visible to the naked eye" ([33]: 237).

To summarize these findings: Formal linguistic and sociolinguistic approaches to LC are concerned with different aspects (emergence and diffusion), thus complementing rather than contradicting each other.

Acknowledgments

I would like to thank the organizers of the workshop and the audience for their insightful and fruitful discussion of the topic from which the paper benefited considerably. Special thanks to three anonymous reviewers for their very helpful comments and Maren Michel for proofreading. All remaining shortcomings or inconsistencies are my own.

1 BSA – SBS: König, Werner (ed.). 2003. Bayerischer Sprachatlas: Sprachatlas von Bayerisch-Schwaben. Vol. 9.2: Formengeographie II. Heidelberg: Winter. 2 BSA – SMF: Heyse, Thurid, Alfred Klepsch, Alexander Mang, Sibylle Reichel, Steffen Arzberger (eds.). 2007. Bayerischer Sprachatlas: Sprachatlas von Mittelfranken. Vol. 7: Morphologie und Syntax. Heidelberg: Winter. 3 MHDBDB: Mittelhochdeutsche Begriffsdatenbank. Universität Salzburg, Interdisciplinary Center for Medieval and Early Modern Studies, coordination: Katharina Zeppezauer-Wachauer. Since 1992 (ongoing). https://mhdbdb.sbg.ac.at/#/home (accessed 13 April 2024). 4 SyHD: Fleischer, Jürg, Alexandra N. Lenz & Helmut Weiß (eds.). 2017. SyHD-atlas. Conceptualized by Ludwig M. Breuer in cooperation with Katrin Kuhmichel, Stephanie Leser-Cronau, Johanna Schwalm and Thomas Strobel. Marburg, Frankfurt am Main & Wien. http://www.syhd.info/startseite/ (accessed 13 April 2024). 5 Axel-Tober, Katrin. 2017. The development of the declarative complementizer in German. Language 93(2). e29–e65. https://doi.org/10.1353/lan.2017.0030. 6 Breitbarth, Anne & Agnes Jäger. 2018. History of negation in High and Low German. In Agnes Jäger, Gisella Ferraresi & Helmut Weiß (eds.), Clause structure and word order in the history of German (Oxford Studies in Diachronic and Historical Linguistics 28), 181–219. Oxford: Oxford University Press. 7 Breitbarth, Anne, Miriam Bouzouita, Lieven Danckaert & Elisabeth Witzenhausen (eds.). 2019. Cycles in language change. Oxford: Oxford University Press. 8 Cardinaletti, Anna & Michel Starke. 1999. The typology of structural deficiency: On the three grammatical classes. In Henk van Riemsdijk (ed.), Clitics in the languages of Europe (Empirical Approaches to Language Typology Eurotyp 20.5), 145–233. Berlin & New York: Mouton de Gruyter. 9 Chambers, Jack K. 2002a. Studying language variation: An informal epistemology. In Jack K. Chambers, Peter Trudgill & Natalie Schilling-Estes (eds.), Handbook of language variation and change, 3–14. Malden, MA: Blackwell. Chambers, Jack K. 2002b. Patterns of variation including change. In Jack K. Chambers, Peter Trudgill & Natalie Schilling-Estes (eds.), Handbook of language variation and change, 349–372. Malden, MA: Blackwell. Chambers, Jack K. & Peter Trudgill. 1998. Dialectology, 2nd edn. Cambridge: Cambridge University Press. Craenenbroeck, Jeroen van & Marjo van Koppen. 2008. Pronominal doubling in Dutch dialects: Big DPs and coordinations. In Sjef Barbiers, Olav Koeneman & Maria Lekakou (eds.), Microvariation in syntactic doubling (Syntax and Semantics 36), 207–239. Bingley: Emerald. Croft, William. 2010. The origins of grammaticalization in the verbalization of experience. Linguistics 48(1). 1–48. https://doi.org/10.1515/ling.2010.001. Déchaine, Rose-Marie & Martina Wiltschko. 2002. Decomposing pronouns. Linguistic Inquiry 33. 409–442. https://doi.org/10.1162/002438902760168554. Denison, David. 2003. Log(ist)ic and simplistic S-curves. In Raymond Hickey (ed.), Motives for language change, 54–70. Cambridge: Cambridge University Press. Donhauser, Karin. 1996. Negationssyntax in der deutschen Sprachgeschichte: Grammatikalisierung oder Degrammatikalisierung. In Ewald Lang & Gisela Zifonun (eds.), Deutsch – typologisch (Jahrbuch des Instituts für deutsche Sprache 1995), 201–217. Berlin & New York: De Gruyter. Fleischer, Jürg. 2017. Geschichte, Anlage und Durchführung der Fragebogen-Erhebungen von Georg Wenkers 40 Sätzen: Dokumentation, Entdeckungen und Neubewertungen (Deutsche Dialektgeographie 123). Hildesheim, Zürich & New York: Olms. Freidin, Robert & Jean-Roger Vergnaud. 2001. Exquisite connections: Some remarks on the evolution of linguistic theory. Lingua 111. 639–666. https://doi.org/10.1016/s0024-3841(01)00045-6. Fuß, Eric. 2017. Language change. In Ian Roberts (ed.), The Oxford handbook of universal grammar, 459–485. Oxford: Oxford University Press. Fuß, Eric & Melani Wratil. 2013. Der Nullsubjektzyklus: Etablierung und Verlust von Nullargumenten. In Jürg Fleischer & Horst Simon (eds.), Comparing diachronies, 163–196. Tübingen: Niemeyer. Gelderen, Elly van. 2004. Grammaticalization as economy. Amsterdam: John Benjamins. Gelderen, Elly van (ed.). 2009. Cyclical change (Linguistik Aktuell 146). Amsterdam: John Benjamins. Gelderen, Elly van (ed.). 2013. The linguistic cycle: Language change and the language faculty. New York: Oxford University Press. Gelderen, Elly van (ed.). 2016. Cyclical change continued (Linguistik Aktuell 227). Amsterdam: John Benjamins. Haider, Hubert. 2021. Grammar change: A case of Darwinian cognitive evolution. In Maria Rita Manzini (ed.), Biological Evolution: More than a metaphor for grammar change [special issue]. Evolutionary Linguistic Theory 3(1). 6–55. Hale, Mark. 2007. Historical linguistics: Theory and method. Oxford & Malden: Blackwell. Heine, Bernd. 2003. Grammaticalization. In Richard D. Janda & Brian D. Joseph (eds.), The handbook of historical linguistics, 575–601. Oxford: Blackwell. Kayne, Richard S. 1996. Microparametric syntax: Some introductory remarks. In James R. Black & Virginia Motapanyane (eds.), Microparametric syntax and dialect variation, ix–xviii. Amsterdam: Benjamins. Kroch, Anthony S. 1989a. Function and grammar in the history of English: Periphrastic do. In Ralph W. Fasold & Deborah Schiffrin (eds.), Language change and variation (Current Issues in Linguistic Theory 52), 132–172. Amsterdam & Philadelphia: John Benjamins. Kroch, Anthony S. 1989b. Reflexes of grammar in patterns of language change. Language Variation and Change 1. 199–244. https://doi.org/10.1017/s0954394500000168. Martin, James R. & Glyn Williams. 2004. Functional sociolinguistics. In Ulrich, Ammon, Norbert Dittmar, Klaus J. Mattheier & Peter Trudgill (eds.), Sociolinguistics: An international handbook of the science of language and society, 120–129. Berlin & New York: De Gruyter Mouton. Roberts, Ian. 2010. Agreement and head movement: Clitics, incorporation, and defective goals. Cambridge, MA: MIT Press. Uriagereka, Juan. 2002. Derivations. London: Routledge. Weiß, Helmut. 2008. The possessor that appears twice: Variation, structure and function of possessive doubling in German. In Sjef Barbiers, Olaf Koeneman, Marika Lekakou & Margreet van der Ham (eds.), Microvariation in syntactic doubling (Syntax and Semantics 36), 381–401. Bingley: Emerald. Weiß, Helmut. 2012. The rise of DP-internal possessors. In Gunther de Vogelaer & Guido Seiler (eds.), The dialect laboratory: Dialects as testing ground for theories of language change (Studies in Language Companion Series 128), 271–293. Amsterdam: John Benjamins. Weiß, Helmut. 2013. UG und syntaktische (Mikro-)Variation. In Werner Abraham & Elisabeth Leiss (eds.), Dialektologie in neuem Gewand: Zu Mikro-/Varietätenlinguistik, Sprachenvergleich und Universalgrammatik (Linguistische Berichte Sonderheft 19), 173–207. Hamburg: Buske. Weiß, Helmut. 2015. When the subject follows the object: On a curiosity in the syntax of personal pronouns in some German dialects. The Journal of Comparative Germanic Linguistics 18(1). 65–92. https://doi.org/10.1007/s10828-015-9071-4. Weiß, Helmut. 2016. Pronomenabfolge: Subjekt, Objekt. In SyHD-atlas. https://www.syhd.info//apps/atlas/index.html#pronomenabfolge-su-o (accessed 14 April 2024). Weiß, Helmut. 2018. Das Eisbergprinzip – Oder wieviel Struktur braucht die Sprache? In Elisabeth Leiss & Sonja Zeman (eds.), Die Zukunft von Grammatik: Die Grammatik der Zukunft, Festschrift für Werner Abraham anlässlich seines 80. Geburtstags (Studien zur deutschen Grammatik 92), 429–447. Tübingen: Stauffenburg. Weiß, Helmut. 2019. Rebracketing (Gliederungsverschiebung) and the Early Merge Principle. Diachronica 36(4). 509–545. https://doi.org/10.1075/dia.00015.wei. Weiß, Helmut. 2020. Where do complementizers come from and how did they come about? A re-evaluation of the parataxis-to-hypotaxis hypothesis. Evolutionary Linguistic Theory 2(1). 30–55. https://doi.org/10.1075/elt.00014.wei. Weiß, Helmut. 2021a. Darwinian language evolution: Remarks on Haider's grammar change. In Maria Rita Manzini (ed.), Biological Evolution: More than a metaphor for grammar change [special issue]. Evolutionary Linguistic Theory 3(1). 73–82. Weiß, Helmut. 2021b. Reanalysis involving rebracketing and relabeling. In Ulrich Detges, Richard Waltereit, Esme Winter-Froemel & Anne C. Wolfsgruber (eds.), Whither reanalysis? [special issue]. Journal of Historical Syntax 5. 1–26. https://doi.org/10.18148/hs/2021.v5i32-39.147. Weiß, Helmut & Seyna Maria Dirani. 2019. Strong or weak? Or: How information structure governs morphosyntactic variation. In Antje Dammel & Oliver Schallert (eds.), Morphological variation: Theoretical and empirical perspectives (Studies in Language Companion Series 207), 311–342. Amsterdam: John Benjamins. Wiltschko, Martina. 1998. On the syntax and semantics of (relative) pronouns and determiners. The Journal of Comparative Germanic Linguistics 2. 143–181. https://doi.org/10.1023/a:1009719229992. One reviewer would not classify (1b) as substandard, but at most as colloquial. It is often assumed by more functional-oriented linguists or sociolinguists "that the variants that occur in everyday speech are linguistically insignificant" ([9]: 4). At a minimum, formal linguists do pay thorough attention to even the smallest linguistic variation. It is common to distinguish only two steps: innovation (or actuation) and propagation (diffusion) (cf.[13] among others). The second step is only available if an innovation first arises in one context and is then generalized to other contexts (e.g., when relative clause complementizers develop into general complementizers, as was the case with Germandass 'that', cf.[5]). In cases where there is no expansion in new contexts, Step 2 is not present (e.g., with adverbial complementizers likebis 'until' orseit 'since', cf.[40],[41],[43]). The spread of a new variant within the grammatical system and its extension into new contexts probably follows the S-curve, too. According to[15]: 64),[29],[30] has shown "that the S-curves for different syntactic contexts (e.g. interrogatives, negatives) show the same rate of change, his so-called Constant Rate Hypothesis." https://www.syhd.info//startseite/index.html (accessed 11 April 2024). This four-part typology modifies[8] tripartite system with strong, weak, and clitic pronouns. I will not go into it further here, but see[37], and[44] for arguments as to why[8] system is not sufficient theoretically or empirically. There are a few exceptions where a pronoun has only one form ([37]). The reverse order with the 2.SG and the 3.PL can be explained in the same way, cf.[37] for further details. This structural proposal combines assumptions made, e.g., in[45],[18],[14],[12],[32], or[20]. If not mentioned otherwise, all examples from MHG are quoted after theMittelhochdeutsche Begriffsdatenbank 'Middle High German Conceptual Database' (MHDBDB), which is available online and contains a large number of annotated MHG (and Early New High German) texts. As becomes clear in the structure (15), it is a relative clause with a 'mute' external head. Besidesdaz, (subordinating)und 'and' andso 'so' are attested as complementizers in such constructions ([43]). One reviewer points to the possibility "that the structural description of seit + clause remains in modern times as given to the left of the arrow in (17), such that there is no real syntactic change." As mentioned in [40], [43], this indeed seems to be the case in some German dialects. As far as non-dialectal German is concerned, seit also counts as a conjunction, i.e., it can be assumed that the development in (17) actually took place. Note that this objection does not apply to the following example weil 'while', so the argument as such would not lose validity if seit had indeed remained a preposition. See[40],[43] for more details on how and why structural ambiguity can trigger this type of reanalysis. Prosalancelot 40, 35–36. Leben V7780.

By Helmut Weiß

Reported by Author