Feature Borrowing in Language Contact

Abstract

In this paper, we consider mood selection in embedded clauses by focusing on a German-based minority language, Cimbrian, which is spoken in a northern Italian enclave. Mood selection in Cimbrian relies on the presence of two different complementizers, az and ke (the latter being borrowed from Romance varieties), each of which selectively require a specific mood. Az selects the mood subjunctive in modal sentences introduced by non-factive verbs, whereas ke co-occurs with the indicative in purely declarative clauses introduced by factive and semi-factive verbs. However, this binary distribution is challenged in the two following contexts, and it is precisely at this point that feature borrowing comes into play: (i) with the verb gloam ‘to believe/to think’, the expected binary pattern appears (irrealis az + subjunctive and the realis ke + indicative), but, crucially, a third construction emerges, namely ke + subj.; (ii) surprisingly, az + subj. displays some ‘gaps’ in its paradigm, specifically in the first person, which appeared in the data we collected via translation tasks from Italian into Cimbrian. Both phenomena shed light on how language contact works, not in terms of structural borrowing but rather in terms of the transfer of the specific features of a given lexical item.

1. Introduction

Cimbrian is a dialect of Bavarian origin which has lost the well-known Germanic linear V2 restriction allowing more than one constituent on the left of the finite verb (cf. 1):

(1)	[In balt]	[haüt]	[dar pua]	hatt	gesek	in has
	in-the woods	today	the boy	has	seen	the hare
	‘In the woods today, the boy saw the hare’.

In the literature devoted to Cimbrian1 syntax, there is, nevertheless, general agreement regarding the assumption that it still maintains structural V2, i.e., mandatory finite verb movement into the C domain, even after centuries of isolation from the German-speaking area in a Romance environment. This assumption is based on two traditional arguments:2

(i)

Subject–finite verb inversion, although this is limited to the pronominal subject:

(2)	a.	Gestarn	in balt	hatt=ar		nèt	gesek	in	has
		yesterday	in-the wood	has=he.cl		not	seen	the.acc	hare
		‘Yesterday, he didn’t see the hare in the wood’.
	b.	*Gestarn	in balt	hatt	dar pua		nèt	gesek	in	has
		yesterday	in-the wood	has	the.nom boy		not	seen	the.acc	hare

(ii)

Residual asymmetry between the root and the embedded word order pattern, essentially based on the relative order of the negative particle (nèt) with respect to the finite verb (Vfin NEG versus NEG Vfin):

(3)	Haüt	geat=ar	nèt	ka	Trento
	today	go=he.cl	not	to	Tria
	‘Today, he doesn’t go to Trento’.
(4)	(I speràr),	az=ar	nèt		gea	ka Tria	haüt
	I hope	that=he.cl	not		go.sbjv	to Trento	today
	‘I hope that he doesn’t go to Trento today’.

Of interest, Cimbrian is characterized by a dual system of lexical complementizers.3 In fact, Cimbrian presents both a class of low complementizers such as az that realize the lower functional field within the C domain, i.e., FinP (see Rizzi 1997), and a second class of higher complementizers that realize the highest portion of the C-domain (let us say SubordP, following Bhatt and Yoon 1991)4 to which the loanword ke belongs to. The most important diagnostic feature to detect the different position of the two classes of complementizers is the asymmetrical word order pattern triggered by az as already shown above in (4) and exemplified again in (cf. 6a) in clear contrast to sentences introduced by ke, which are characterized by the root word order pattern (cf. 5 with 6b):

(5)		Dar	geat	nèt	ka	Tria
		he	go	not	to	Trento
		‘He doesn’t go to Trento’.
(6)	a.	(I speràr),	az=ar	nèt			gea	ka Tria
		I hope	that=he.cl	not			go.sbjv	to Trento
		‘I hope that he doesn’t go to Trento’.
	b.	(I boaz),	ke	dar	geat		nèt	ka Tria
		I know	that	he	go		not	to Trento
		‘I know that he doesn’t go to Trento’.

The opposition between az and ke does not simply rely on the maintenance versus the dismantling of the asymmetry between root and embedded word order, but is reinforced by a clear specialization of two lexical items:

-

az always requires the subjunctive mood, and typically introduces a declarative clause selected by non-factive (volitional) verbs such as bölln ‘to want’ and non-assertive (affective) verbs such as speràrn ‘to hope’ (see 6a above).

-

ke introduces embedded declarative clauses in the indicative mood that are selected by strongly assertive verbs such as khön ‘to say’ or semi-factive (knowledge) verbs such as bizzan ‘to know’ (see 6b above), perceptive verbs such as seng ‘to see’ and weakly assertive (epistemic) ones such as pensàrn ‘to think’.

The specialization of az versus ke is extremely interesting with regard to both the way in which the borrowing of functional words first affects the target system (cf. Section 2) and how it could further develop favoring the occurrence of ke and hence the gradual dismantling of the root-embedded word order asymmetry. In Section 3, we discuss a potential innovation in the occurrence of ke that was produced during the translation tasks by some native speakers which unexpectedly admit ke + subjunctive mood; in Section 4, we consider an evident weakness of az in connection with the first-person plural in the subjunctive paradigm. In the concluding remarks (cf. Section 5), we emphasize the coherence of our observations with recent assumptions about the relevance of heritage language studies to a theoretical approach to language contact in the bilingual brain and, ultimately, to a theory of language change (see Polinsky and Scontras 2020; Lohndal et al. 2019).

2. The Specialization of az versus ke

The specialization of az versus ke per se represents clear evidence of how the borrowing of functional words5 impacts on the target system from both a structural and a morphological point of view:

A. From a structural point of view:

(i) az represents the [-WH] counterpart of [+WH] ob ‘if’ in the introduction of argumental subordinate clauses and competes with finite verb movement with regard to the same structural position within the C domain; that is, Fin⁰. This is completely in line with the well-known German pattern whereby dass ‘that’ and the finite verb compete for the lexicalization of the so-called linke Satzklammer; that is, the head of the CP resulting in the root-embedded word order asymmetry.

(ii) ke does not compete with az with regard to the low C head but enters the Cimbrian Split-C domain lexicalizing its highest portion; that is, SubordP, and requiring the root V2 word order pattern (see also Padovan et al. 2016).6

B. From a morphological point of view:

(i) ke does not interfere with either the position or the morphology of the finite verb, as the latter displays the default indicative mood and always raises to Fin⁰, thus dismantling the root-embedded word order asymmetry (see 4a above).

(ii) az always requires the subjunctive mood and inhibits verb movement by reproducing the traditional word order asymmetry displayed by V2 Germanic languages.

The following graphical representation (see Bidese et al. 2013, p. 55) illustrates the structural difference between az and ke:

(7)

Argumental clauses introduced by az imply the selection by a non-factive verb, and the embedding is overtly marked by the following phenomena:

(a) the lexical complementizer;

(b) the morphology of the finite verb, which displays the subjunctive mood; and

(c) the word order pattern; that is, the delayed position of the finite verb (=no verb movement = no subject–verb inversion).

By contrast, argumental clauses introduced by ke are generally selected by factive or semi-factive verbs, and embedding is signalled only by the lexical complementizer as there is no word order asymmetry nor subjunctive mood; that is, the finite verb raises to Fin and displays the default indicative value, as in the root declarative clause.

These clear-cut differences between the syntactic behavior of az with respect to ke force us to assume a double system of complementizers—in line with a large body of literature (see, among others, Grewendorf and Poletto 2009, 2011; Padovan 2011; Bidese et al. 2012, 2014; Padovan et al. 2016). There is, indeed, no historical evidence for a high az, as we would coherently expect for a Germanic V2 variety; on the contrary, high complementizers has been assumed for medieval Romance languages (see Wolfe 2019). The insertion of che into the Cimbrian C-system must be interpreted as a consequence of the splitting of the C-domain—i.e., the loss of linear V2 which opened the door to the borrowing of the Romance functional word.

The specialization of ke with regard to az shows that the borrowing of a functional word in the Cimbrian complementizer system resulted in a situation that is uncommon in both the Italian and the German systems, as both rely on a single lexical complementizer (that is, che—dass). Nevertheless, the choice of a different lexical complementizer determined by the class of factive versus non-factive verbs is well known in the literature, since it is attested in different language systems.7

If our analysis proves to be correct, we can assume that the borrowing of the lexical complementizer ke favoured the specialization of the paradigm that relied on a single lexical complementizer for the introduction of an argumental clause (the lack of a ‘high’ complementizer; that is, no lexical item in the highest structural position in the C domain), thus necessitating a restructuring of the system via the specialisation of the autochthonous az.8

The actual situation reveals two signs of a potential diachronic development in favour of ke (see Section 3 below) with a parallel reduction of the occurrence of az (see Section 4 below).

3. A First Exception to the System: ke with the Subjunctive Mood

The data obtained during the translation task from Italian into Cimbrian revealed an interesting exception to the general system discussed in the previous paragraph: On one hand, our informants never connected az to the indicative mood; on the other hand, the choice of ke could unexpectedly co-occur with a subjunctive verbal form. The first context in which this occurred was with the semi-factive and assertive-epistemic verb gloam ‘to believe/to think’, which can select both the sequences az + Subjv and ke + Ind depending on the grade of certainty/veridicality. Let us consider the following examples:9 (8) shows the stimulus sentence in Italian to be translated into Cimbrian and (9) shows the examples of the translated sentences:

(8)	Italian stimulus sentence:				Loro	credono	che	(lui)	sia	arrivato	tardi
					they	believe	that	(he)	be.sbjv	arrived	late
					‘They believe that he arrived late’.
(9)	a.	Sa	gloam	ke	dar	iz	gerift spet				(ke + Ind)
		they	believe	that	he	is.ind	arrived late
	b.	Sa	gloam	azz=ar		sai(be)	gerift	spet			(az + Subjv)
		they	believe	that=he.cl		is.subjv	arrived	late
	c.	Sa	gloam	ke	dar	sai(be)	gerift	spet			(ke + Subjv)
		they	believe	that	he	is.subjv	arrived	late
	d.	*Sa	gloam	azz=ar		iz	gerift spet				*(az + Ind)
		they	believe	that=he.cl		is.ind	arrived late

A further context in which we elicited the string ke + Subjv was with the negative expression ‘z iz nèt khött ‘it is not certain’ (see 10), which triggered a non-veridical interpretation:

(10)

Italian stimulus sentence:

Non

è

detto

che

Gianni

venga

con

noi

not

is

said

that

Gianni

come.subjv

with

us

‘It is not certain that Gianni will come with us’.

(11)

’Z

iz

nèt

khött

ke

dar Gianni

khemm

pit

üs

it

is

not

said

that

the Gianni

come.subjv

with

us

We are aware of the fact that, from a methodological point of view, the collected data are not statistically significant and that translation tasks are an extremely unusual way to collect data pertaining to bilingualism. Nevertheless, this approach allowed us to reproduce a strong ‘artificial’ attrition in the L1 that favored the borrowing of features. In addition, it should be noted that the data were confirmed by a follow-up questionnaire and that Tyroller (2003, p. 238), who questioned many more speakers, also mentioned that ke could select the subjective mood; unfortunately, he did not provide any examples.

The data presented above were notable due to at least three aspects:

(a) The exception to the az + Subjv − ke + Ind distribution (see 9c and 11 above) only goes in one direction, namely ke + Subjv, but *az + Ind;

(b) the possibility of combining ke with a subjunctive verbal form does not imply a word order variation. The subordinate clauses introduced by ke always display a root word order pattern; that is, no asymmetry root versus an embedded clause (see 7c and 9), unlike the sequence az + Subjv; and

(c) in our data, the combination ke + Subjv was only attested in translation tasks using the Italian model sentence and was not present in spontaneous production (but see Tyroller 2003, p. 238).

The innovation concerning the extension of ke within the specialization space of az is not as unusual at it may seem at first glance. On the contrary, it is coherent with the process of the gradual dismantling of the root versus embedded word order asymmetry, as noted previously in the literature on Cimbrian syntax (see Bidese and Tomaselli 2016), and acts as a further potential step towards the loss of V2. Furthermore, the pressure of Italian on Cimbrian morpho-syntax appeared to follow a precise path:

(i) The borrowing of a functional word, that is, the lexical complementizer ke, as a holistic lexical brick with no modal features that realises the highest position in the Cimbrian C domain; and

(ii) the functional feature [+MOOD] enters the system by replacing the default value of an empty (that is, unlexicalized) Fin⁰, which is c-commanded by ke. These two steps are represented in the following schema (see Bidese et al. 2013):

(12)

If we are correct, the impact of the model language on the functional C domain in the target language complies with the following requirements:

(a) It affects the highest functional layer, as a new functional word enters the system from above;10

(b) the borrowing of functional elements implies the absence of specialization of the paradigm at both the lexical and the morphological level in the target system. In fact, our analysis suggests that

(i) there is no substitution of the lexical element (ke instead of az), but rather the filling of an empty structural position in the target language. In other words, a new item taken from the Romance dialect is inserted into the Cimbrian C domain in a high ‘free’ position (SubordP); and

(ii) the change in the feature specification in a functional head only applies to the default value. In other words, the subjunctive mood co-occurring with ke realizes a non-selected default value in Fin⁰. In fact, we consider the Cimbrian ke to be a lexical (that is, semantic) subordinator that does not select the mood feature in Fin⁰, thus allowing feature borrowing from the Italian system.

The assumptions in (a) and (b) allowed us to draw our first conclusion: the coexistence of two grammatical systems in the bilingual brain supports the borrowing of abstract features and their recombination in the bilingual mind (see Aboh 2015, p. 5).

4. A Gap in the Paradigm of the Subjunctive Mood: The First-Person Plural

We found an interesting confirmation of what we have developed thus far in a second phenomenon that concerns the embedded sentences selected by the non-factive (or, more precisely, the semi-factive) verb gloam ‘to believe/to think’. If the embedded verb selected by gloam is in the first-person plural, the embedded sentences are always introduced by ke. In other words, in our data, the speakers systematically choose ke with the first-person plural in the embedded sentence. Consider the following examples:

(13)	Italian stimulus sentence:				Credono	che	(noi)	siamo	arrivati	tardi	[+/− realis]
					believe.3pp	that	we	are	arrived	late
					‘They believe that we arrived late’.
(14)	a.	Sa	gloam	ke	bar	soin	gerift		spet
		they	believe	that	we	are	arrived		late
	b.	*Sa	gloam	az=par		soin	gerift		spet
		they	believe	that=we.cl		are	arrived		late

To understand the problem in depth, let us consider the stimulus sentence and the entire paradigm of both possible expected translations: az + Subjv and ke + Ind (see Bidese et al. 2013, p. 54):

(15)

Credo

di

essere

in ritardo

I

goabe

zo soina

spet

believe.1ps

to

be

late

I

believe

to be.inf.fl

late

(16)

a.

Credo

che

tu

sia

in ritardo

I

gloabe

az=to

sai(be)st

spet

believe.1ps

that

you

be.subjv

late

I

believe

that=you.cl

be.2s.subjv

late

b.

I

gloabe

ke

du

pist

spet

I

believe

that

you

are.ind

late

(17)

a.

Credo

che

lui

sia

in ritardo

I

gloabe

azz=ar

sai(be)

spet

believe.1ps

that

he

be.subjv

late

I

believe

that=he.cl

be.3s.subjv

late

b.

I

gloabe

ke

dar

iz

spet

I

believe

that

he

is.ind

late

(18)

a.

Credo

che

noi

siamo

in ritardo

*I

gloabe

az=par

soin

spet

believe.1ps

that

we

be.ind/subjv

late

I

believe

that=we.cl

be.1p.ind/subjv

late

b.

I

gloabe

ke

bar

soin

spet

I

believe

that

we

are.ind

late

(19)

a.

Credo

che

voi

siate

in ritardo

I

gloabe

azz=ar

sait

spet

believe.1ps

that

you

be.subjv

late

I

believe

that=you.cl

be.2p.ind/subjv

late

b.

I

gloabe

ke

dar

sait

spet

I

believe

that

you

are.ind

late

(20)

a.

Credo

che

loro

siano

in ritardo

I

gloabe

az=ze

soin

spet

believe.1ps

that

you

be.subjv

late

I

believe

that=they.cl

be.3p.ind/subjv

late

b.

I

gloabe

ke

se

soin

spet

I

believe

that

they

are.ind

late

As it can easily be seen, Italian verb credo ‘to believe’ only selects the complementizer che with the subjunctive form which is morphologically different from the indicative one in all persons of the paradigm but the first-person plural. On the other side, Cimbrian has two possibilities depending on whether the action in the embedded clause should be interpreted as [+irrealis] (az + subjunctive) or [+realis] (ke + indicative). In Cimbrian, the morphological difference between subjunctive and indicative is visible only in the singular persons, and in the plural ones it exhibits syncretic forms. Cimbrian manifests the contrast through the specialization of the complementizers in a very consistent way with one exception. The first-personal plural is the only person in which the subjunctive and indicative forms are homophonous in Italian (for example, siamo ‘we are.subjv/ind’, abbiamo ‘we have.subjv/ind’, arriviamo ‘we arrive.subjv/ind’, leggiamo ‘we read.subjv/ind’ and finiamo ‘we finish.subjv/ind’). This is the only context in which the Cimbrian complementizer system exhibits a ‘gap’; that is, the choice of az is ruled out and ke represents the only possible choice (cf. 16). It should be stressed that the problem is not the homophonous forms in the Cimbrian mood system, in which, actually, all plural forms are syncretic between the indicative and the subjunctive; the real point is the syncretism in Italian, which is exclusively in the first-person plural.11

We assume that the exclusion of az in the first-person plural depends on the choice of Ind induced by homophony in Italian. In other words, the lack of subjv was contact induced in the grammar of our bilingual speakers who assumed these homophonous verbal forms to be a ‘default Ind’, thus forcing the choice of ke, as represented in the following schema (see Bidese et al. 2013, p. 56):

(21)

The notion of ‘default’ was revealed to be crucial once again, despite its formal definition not yet having been provided.

In the translation task, the influence of che in the model language was the indirect result of the morphological homophony of the suffix in the first-person plural. Of the two morphological values, the indicative exponent prevails leading to the logical selection of the appropriate lexical complementizer, which is ke as opposed to az, which is the specialized form for the subjunctive mood.

5. Conclusions

Cimbrian is a German(ic) variety that has been spoken in isolation in a geographical and linguistic Romance-language context for centuries; furthermore, the number of its native speakers has decreased dramatically in recent decades. In fact, it is included in the list of ‘definitely endangered’ languages (see Moseley 2010, p. 24). Its status as a heritage language is not entirely canonical because a larger community that maintains this variety (or rather its historical evolution) in a different geographical area does not exist, as is the case for the native varieties spoken by immigrants, nor is there a standard variety for reference. In fact, the process that resulted in Standard German followed the Cimbrian settlements in the north-eastern Italian regions by at least three centuries.

Nevertheless, Cimbrian complies fully with the three fundamental aspects of heritage languages on which there is general agreement in the scientific literature (see Lohndal et al. 2019, p. 4):

A. Heritage speakers are minority language speakers in a majority language environment. In this regard, Cimbrian is a severely endangered minority language of Germanic origin that is spoken in northern Italy.

B. Heritage speakers are bilingual. This is also true for Cimbrian, as all Cimbrian speakers have full competence in both the Italian dialect of the surrounding environment and in Standard Italian, which is the language of school education.

C. By the time they are adults, heritage speakers tend to be dominant, or more proficient (in a productive way) in the language of their larger national community.

The results of our analysis of the Cimbrian complementizer system support this and provide new insights into the following theoretical generalizations about stable and vulnerable domains in language contact situations in terms of multilingual competence (see the surveys proposed and discussed by Polinsky and Scontras 2020 and Lohndal et al. 2019, among others):

(i)

Syntactic knowledge, particularly the knowledge of phrase structure and word order, appears to be more resilient to incomplete acquisition under reduced input conditions than is inflectional morphology. In other words, the structure of the C domain is not only maintained, but is even enriched due to the pressure of the model language. In fact, the borrowing of the lexical complementizer ke resulted in an innovation within the Cimbrian system that is not found in either Italian or German, namely a specialized introduction for embedded clauses selected by factive versus non-factive verbs; furthermore, an embedded sentence introduced by ke does not manifest root-embedded asymmetry, but maintains the root word order pattern characterized by structural V2.

(ii)

The relative stability of syntactic features compared to morpho-phonological exponents. Our research revealed that the feature characterization of C as [+Fin], [+Mood] was definitely more stable than were the corresponding morpho-phonological exponents (subjunctive versus indicative)

(iii)

A prevalence for ‘defaults’; that is, the prevalence of the indicative mood with respect to the subjunctive mood and the dismantling of the root versus embedded word order asymmetry in favor of the root word order pattern.

Even if the formal definitions of crucial notions such as ‘default’ and ‘unmarked’ are still far from being determined, it seems clear that this theoretical question will find an answer precisely in the test ground of language contact situations that manifest a sufficient degree of both complexity and transparency.