Structural neutrality in Frisian-Dutch interaction
In this paper I will discuss the phenomenon of neutrality in Frisian-Dutch interaction. I will show how neutrality plays a role in code-switching between the two languages, and how it is sought to extend by interlingual conversion rules. Finally I will show how a number of structural changes in the Frisian system can be accounted for by assuming an internalization of these conversion rules by the bilingual child.
1. Typological background
Frisian and Dutch are the two official languages in the Netherlands' province of Frisia (Fryslân, Friesland). Everybody masters Dutch, the state language. Over half of the population is native bilingual Frisian-Dutch.
Although superficially rather distinct, the two languages show a high degree of resemblance. First, there is an important degree of objectively observable resemblance on all levels of grammar:
(1) objectively observable resemblances (OOR):
Secondly, there is a type of resemblance that is less obvious at first sight. There are many regular morphonological differences between Frisian and Dutch:
(2) many regularly corresponding differences (RCD):
An extensive list of these RCD type resemblances is given by Sjölin 1976. A few examples are given in (3):
(3) Du: ee ~ Fr: ie (steen/stien, heel/hiel, keel/kiel)
Du: aan ~ Fr: ean (gaan/gean, laan/leane, staan/stean)
Du: sch ~ Fr: sk (schande/skande, schat/skat, schip/skip)
Du: tegen ~ Fr: tsjin (tegenstelling/tsjinstelling)
The aim of this paper is to show how these resemblances (OOR and RCD) play a role in several forms of Frisian-Dutch language interaction. They underlie various forms of neutrality. Neutrality results from a relatively strong resemblance between the two lexico-grammatical systems involved. I will contend that it is a precondition for several forms of Frisian-Dutch interaction. I will also demonstrate how some specific aspects of the interaction of this particular language pair are determined by the search for extension of neutrality along existing lines. In this paper I will concentrate on the following forms of interaction:
Structural models of (intrasentential) codeswitching show a lot of variation. Yet, there are roughly two mainstreams to be observed (although these are not necessarily mutually exclusive):
(5) -Linear or neutrality models: these models presuppose several (neutral) points in the sentence that are potential switchpoints. Neutrality can be triggered a.o. by linear equivalence (Poplack 1980; equivalence constraint), and more specifically by borrowings or nonce-borrowings (Clyne 1972), affixation (Muysken 1987) and homophony (Crama & Van Gelderen 1984).
(6) -Embedding models: these models presuppose a matrix language, in which under specific structural conditions elements from a second language can be embedded (Di Sciullo, Muysken & Singh 1986, Myers-Scotton 1995).
Neutrality models as defined in (5) are largely construed on the basis of language pairs that show a high degree of lexical or grammatical resemblance (Poplack 1980 for Spanish-English, Crama & Van Gelderen 1984 for Dutch-English).
The lexico-grammatical resemblances between Frisian and Dutch are possibly even larger than those between the above-mentioned language pairs. If the universality claim for neutrality models is warranted, it is in the line of expectations that Frisian-Dutch codeswitching can be described with a neutrality model.
In this paper I will address several neutrality models that are not always equally easy to discriminate. One of the most important constraints in a neutrality framework is the equivalence constraint (EC), which was developed by Poplack 1980 on the basis of Spanish-English code-switching material. This constraint is based on linear similarity between the translation equivalents in both languages: 'Code-switches tend to occur at points in discourse where juxtaposition of L1 and L2 elements does not violate a syntactic rule of either language, i.e. at points around which the surface structures of the two languages map onto each other'. The word order in Frisian and Dutch is very similar, so in a model based on the EC, many potential switchpoints would be expected. Van Hout & Muysken 1995 also indicate that linear equivalence is more typical of 'alternation and congruent lexicalization' than of embedding ('insertion' in their terms).
Neutrality may also result from homophony, as found for Dutch-English codeswitching by Crama & Van Gelderen 1984. When the translation equivalents of a word are (nearly) homophonous in both languages, a switch may occur at that point. In the same way, borrowings or nonce-borrowings, possibly completely or partly adapted to the grammar of the host language, may trigger sufficient neutrality (Clyne 1972, Muysken 1987).
Now let us turn to some examples of Frisian-Dutch code-switching, taken from several corpora used by De Jong & Riemersma 1994, Sjölin 1976 and Wolf 1995b. Undercast indicates Frisian, capitals indicate Dutch.
(7) witst noch wol wat se dan seine, wat waar, wat WEER IS HET, BEWAARDER
do you remember what they then said, what weather, what WEATHER IS IT, GUARD
("what type of weather")
(8) en de partij dy't hy derby blaasde, is net foar HERHALING VATBAAR
and the part he played, is not for REPETITION APT ("apt for repetition")
(9) yn dit gefal gjin DADEN, MAAR WOORDEN
in that case no DEEDS BUT WORDS
(10) myn HOOFDBEROEP DAT IS P.T.T.
my MAIN PROFESSION THAT IS POSTMAN
(11) ha we dus eh ien ha we in eh VUILNISsek ha we VOOR in rút GEbrocht
have we so er one have we a er GARBAGE bag have we AT a window INFL-placed
("so we have placed a garbage bag at a window")
The sentences (7-10) show several instances of neutral switchpoints. It must be noted that in all four examples, the underlined word is Frisian and not Dutch. In neither case are the Dutch and Frisian cognates exact homophones, but in all cases phonological and lexical identification is obvious. In the examples in (7) and (10) this identification is even more likely, as they are instances of RCD; the Frisian form can be derived from its Dutch cognate by applying regular phonological conversion rules.
Now let us consider some alternative analyses for (7-10). I will discuss two embedding models, respectively the Matrix Language Frame Model (MLFM) and the Government Constraint (GC).
The MLFM in its most recent shape states that there is a Matrix Language (ML) that provides all the system morphemes within the boundaries of the CP.
(12) Matrix Language (ML) vs. Embedded Language (EL)
The Matrix Language is the more dominant one and supplies the morphosyntactic frame of any CP containing morphemes from both languages participating in CS.
(13) Are there restrictions on ML+EL constituents? Yes
The ML provides the sentential frame of CPs with intrasentential CS. Specifically for mixed constituents, the ML supplies all system morphemes and morpheme order.
The EL can only supply content morphemes.
(12-13) imply that all systeme morphemes within any CP should be in the same language. This poses serious problems for (11). The verb GEbrocht is construed from a Frisian content morpheme, which carries Frisian participial inflection, and the Dutch participial prefix GE-. The ø-inflection on the auxiliary ha is Frisian, though. The status of the article in is ambiguous. In (8) the negation element net is Frisian, while the bound morpheme -baar is Dutch. For the other sentences things are less clear, as Myers-Scotton is not specific as to whether prepositions, adverbs and determiners should be considered system morphemes or not.
The GC is based mostly on DiSciullo, Muysken & Singh 1986. In their theory, code-switching is not allowed between elements that show a government relation, here practically defined as a constraint on code-switching between case-assigner/assigned element or subcategorizing/categorized element. This model poses a problem for sentences like (8), where a Frisian preposition selects a Dutch noun. In (11) the language status of the article is ambiguous, which may neutralize the switch, in terms of DiSciullo, Muysken & Singh.
(7) poses a problem also. The Frisian phrase wat waar ("what type of weather") is grammatical, whereas its Dutch counterpart wat weer is not. In Dutch the prepositional element voor needs to be inserted between wat and weer. As wat is reported to be Frisian (which can be told from the pronunciation), it would select a Dutch noun, which is a violation of the GC. If wat would have been Dutch, the phrase would not be interpretable in an embedding model, as we would have Dutch lexical filling of a Frisian structure.
In (8-9) we find instances of idiomatic constructions in Dutch. One might wonder if a (nonce-)borrowing analysis would be applicable here. This is highly unlikely, however, as one would expect the entire idiom to be in Dutch: VOOR HERHALING VATBAAR, GEEN DADEN, MAAR WOORDEN, or in the case of (9) both functional elements to be in the same language: gjin DADEN, mar WOORDEN. Therefore I will assume that all above-mentioned examples are instances of code-switching.
Although the exact nature of neutrality in Frisian-Dutch code-switching is yet unclear, we may conclude from the limited data set used here that this corpus of Frisian-Dutch code-switching càn be described within a neutrality model. Neutrality is easily found because of a high degree of resemblance (OOR and RCD) between the two lexico-grammatical systems involved. As a result, there are many potential switchpoints.
[...] een volledige competence in het Fries wordt [...] normaliter verkregen via het Nederlands.
a full competence in Frisian is normally acquired through Dutch (Sjölin 1976)
The term conversion is taken from Sjölin 1976 and refers to strategies that allow the formation of neologisms in Frisian from Dutch. RCD (and OOR) is redefined as interlingual rules. Some examples are given in (14).
(14) oandacht (Du: aandacht, Fr: omtinken) attention
sondachskool (Du: zondagsschool, Fr: sneinsskoalle) Sunday school
waterskap (Du: waterschap, Fr: wetterskip) type of regional administration
ienichsins (Du: enigszins, Fr: wat) somewhat
opstean (Du: opstaan, Fr: derôf komme) get up
maaltyd (Du: maaltijd, Fr: miel) meal
The words in (14) have been converted from Dutch into Frisian. RCD based rules such as [Du: aa ~ Fr: oa] and [Du: ee ~ Fr: ie] were used.
The examples in (14) have acquired a certain independent status in Frisian through frequent use, but generally it can be stated that converted forms do not have an independent status in the Frisian lexicon. They are formed straight from Dutch through conversion. As such they share some properties with nonce-borrowings that do not have any independent status in the host language either. Consequently, it is to be expected that Frisians are rather uncertain as to what exactly the status of conversions is. This fact might trigger neutrality.
Conversion should not be mistaken for phonological adaptation of (nonce-)borrowings. A word like enigszins does not in any way violate Frisian grammatical rules. Yet, the form is often found in Frisian in its converted form, as a result of the phonological RCD [Du: ee ~ Fr: ie].
Apparently, neutrality is not only found in Frisian-Dutch interaction, but it is also sought to extend. This becomes even clearer if we have a look at the cases in which conversion is used. First, this often occurs when there is a lexical gap in Frisian. Frisian may in so far be characterized as an incomplete language, that it lacks sufficient terminology in many fields. Secondly, it happens when there is a Frisian equivalent, which has acquired connotations that make it unfit for use in a certain context. Relatively many Frisian words are found unfit for spoken use, as Sjölin states, because of their literary connotations. He does not specifically address how these connotations were acquired, but one may generalize and say that these words are characterized by a relatively low frequency and a form that is rather distant from its Dutch translation equivalent. In such cases, the Frisian word is apparently marked as "insufficiently neutral". The alternatives are to use the Dutch word or to use the converted Dutch word. Sjölin states that when that choice occurs, the latter possibility is often preferred, as it is considered "more Frisian". In other words, neutrality is sought in RCD rather than OOR type resemblances.
A very similar situation seems to exist with regard to the language pair Low German-High German, as reported by Hansen-Jaax 1995. She links the occurrence of certain types of code-switching and conversion ("Analogiebildungen" in her terms) to the typical diglossic situation that characterizes Northern Germany. Although not completely comparable, the sociolinguistic relationship between Frisian and Dutch may to a certain extent also be termed as one between a High and a Low variety. In fact this non-diglossic functional distribution of the H and L variety may well be a precondition on the certain applications of conversion, as it makes it possible that Frisian is used for discussion of predominantly H domains. Furthermore, as in Northern Germany, the two languages spoken in Frisia are rather closely related and have a small 'perceived linguistic distance', which makes it easy to find neutrality. I would therefore not be surprised if much of what is said in this paper, is typical of language pairs that share those two characteristics.
4. Structural change
In the following I will contend that neutrality strategies underlie some structural changes in Frisian. The changes discussed below are typical of the youngest generation of speakers and discussed more extensively in Wolf 1995a; 1996.
Let us have a look at (15-17).
(15) Fr: *wy gean iten conversion of auxiliary selection
Du: wij gaan eten --conv--> Fr: wy gean iten
we go eat ("we are going to eat")
(16) Fr: *ik wol it litte meitsje conversion of word order
Du: ik wil het laten maken --conv--> Fr: ik wol it litte meitsje
i want it let fix ("i want to have it fixed")
(17) Fr: ik hear him rinnen/*rinne conv. of final /n/ deletion
Du: ik hoor hem lopen/lope --conv--> Fr: ik hear him rinnen/rinne
i hear him walk
Fr: ik lit him *rinnen/rinne
Du: ik laat hem lopen/lope --conv--> Fr: ik lit him *rinnen/rinne
i let him walk
Frisian and Dutch traditionally have some rather divergent systems of future auxiliary selection. Dutch uses gaan (to go) in all cases, whereas the Frisian translation equivalent gean (RCD type resemblance) is restricted to a small number of verbs. More recently, however, Frisian gean has come to be used with all verbs, as in (15), thus severely reducing the much richer Frisian system.
A striking difference in word order between Frisian and Dutch is the order of the verbs in the sentence-final verbal complex. In Dutch the main order is head-complement. In Frisian this is the reverse. A certain percentage of the youngest generation of speakers seems to allow both word orders in Frisian, though. An example can be found in (16).
Frisian has got two infinitives, one ending in -e (/@/), the other one in -en (/(@)n/). These are in complementary distribution. Perception verbs select the -en infinitive, while a verb such as litte (to let) selects the -e infinitive. In Dutch the same two endings on infinitives occur, but in this language they are free variants. As (17) shows, the distribution in Frisian appears to be changing. The two infinitives have not become free variants, though. They are still in complementary distribution, but the -en infinitive may also be realized as an -e infinitive. The -e infinitive must still be realized as an -e infinitive. This new situation is explicable from Dutch, however, if we assume that the Dutch infinitive is analyzed as underlyingly -en, with optional deletion of final -n. This (phonological) rule could well have been adopted into Frisian, without affecting the underlying complementary distribution of the two infinitives.
As these changes are characteristic of the youngest generation of speakers, one might wonder if they are not the reflection of errors due to the fact that the speakers are still in a language acquisition stage. This possibility is ruled out by research by Ytsma 1995 and Wolf 1996, however. On the basis of a real-time investigation over four years, they show that the number of non-Frisian word orders does not decrease significantly over time. They also show that there is a huge gap between the sixteen year old informants and their parents, the latter of whom never of hardly ever use Dutch word order. Finally, it was found that the non-standard constructions are used by some speakers who are in their twenties, who are way past the primary language acquisition stage.
Now there are at least two scenarios to account for these and other changes in Frisian. These can be described as respectively mixed output and mixed input.
The mixed output scenario is defended in De Haan 1990 for the word order change illustrated in (16). In this scenario a switch is made to the Dutch grammatical system for the production of the verbal complex. The lexical level is Frisian, however. This scenario may be called the grammatical counterpart of conversion in the lexicon.
Several objections can be made against this scenario, the most elementary one being that there are no independent motivations to deviate from the theoretically desirable principle to describe data from one language with one model. As it has shown from field research by a.o. Ytsma 1995 and Wolf 1995a; 1996 that speakers use both word orders, often without realizing which order is originally Frisian and which one is Dutch, it is probably preferable to consider this variety a new and independent variety of Frisian. This variety is usually referred to as Interference Frisian (IF).
Another reason why the mixed output scenario is not desirable is that there are various constructions in the IF verbal complex that do not occur in Dutch. The distinction between the two infinitives, as in (17), is one of these.
A third reason has to do with the motivation for this switch to Dutch. It is unclear what that would be. De Haan refers to observations that the Frisian complement-head order is perceptually more difficult than its reverse. That would trigger a code-switch to the Dutch system. That would not explain, however, why the Dutch order is also found in simple complexes with two verbs. Besides, in these simple complexes, the complement-head order is usually allowed in Dutch also, and even appears to be preferred in the spoken language by some speakers.
Finally, the mixed output scenario is unable to account for the intergenerational gap between speakers of standard Frisian and speakers of IF.
The mixed input scenario starts from the assumption that IF is a variety on its own. Although there is no grammatical conversion, I will claim that the occurence of neutrality and interlingual rules plays an important role in a number of structural changes.
If we assume that Frisian children acquiring their languages, observe the OOR and RCD in the Frisian situation and how language users apply those resemblances for interlingual rules, it is not unlikely that these children interpret these interlingual rules in such a way that Dutch language output may serve as input for the grammatical systems of both Dutch and Frisian. Only at a later stage they would be able to discriminate sufficiently between the two languages to tell them apart at the competence level. Elements of Dutch grammar have already been internalized at that stage, however. At the performance level, they will still use interlingual rules, of course, as described in the previous section.
Evidence for the claim that children apply interlingual rules in language acquisition is given in (18) and (19). Here we find a number of examples of conversion applied by a three year old child. Frisian-to-Dutch conversion is by far most prominent, as the child's first language is Frisian, but there are a few instances of Dutch-to-Frisian. These conversions are extremely abundant and unlikely to occur in adult speech. This child is apparently still at a stage at which interlingual rules are freely applied.
(18) ploatje (Du: plaatje, Fr: plaatsje) picture
in (Du: in, Fr: yn) in
(19) paraplui (Fr: paraplu, Du: paraplu) umbrella
puizeltje (Fr: puzeltsje, Du: puzzeltje) puzzle
bluid (Fr: bloed, Du: bloed) blood
buisduik (Fr: bûsdoek, Du: zakdoek) handkerchief
lijts (Fr: lyts, Du: klein) little
tijpmasijne (Fr: typmasine, Du: typmachine) typewriter
sijnasappels (Fr: sinasappels, Du: sinasappels) oranges
ijn (Fr: yn, Du: in) in
ondersijken (Fr: ûndersykje, Du: onderzoeken) investigate
The mixed output scenario is compatible with the intergenerational gap, if we assume that only recently a situation with sufficient bilingual competence, sufficient neutrality and a sociolinguistically proper functional distribution of the H and L variety has been attained. The lexicon of young speakers is definitely more neutral than that of older speakers, which indicates that the search for neutrality has gone much further. That makes it more likely that they have been exposed to mixed input.
This scenario is also in line with the theoretically more desirable principle of one model for one language. As Frisian appears to be one language at the psychological level, no matter how far-reaching the structural changes are, this is even more desirable.
Also, this scenario is compatible with the observation that non-neutral grammatical constructions, i.e. constructions that occur in the minority language only, and not in the state language, are slowly disappearing in both Frisian and Low German.
More specifically, finally, facts like the change in distribution of the two infinitives in IF, that are not accounted for in the mixed output scenario, can be accounted for in the mixed input scenario.
The interaction between Frisian and Dutch is characterized largely by neutrality. Because of the relatively high degree of resemblance between the two languages, a neutral switchpoint is easily found, which makes it easy to swich from one language to the other. As a result, code-switching between Frisian and Dutch is largely determined by neutrality, that can be triggered in various ways.
The resemblances between the two languages, especially those of the RCD type are reinterpreted as interlingual rules that make it possible to enlarge the neutrality between the two languages by conversion strategies at the performance level.
These interlingual rules may well be interpreted by children acquiring their bilingualism as applicable at the competence level. As a result, Dutch language data would serve as input for the Frisian grammatical system, which would result in the adoption of Dutch grammatical rules into the Frisian grammatical system, thus yielding a single new grammatical system.
Universiteit Utrecht/Fryske Akademy
Clyne, M.G. (1972), Perspectives on language contact; Based on a study of German in Australia, Melbourne: Hawthorne Press.
Crama, R. & H. van Gelderen (1984), Syntactic constraints on English-Dutch code-switching, M.A. Universiteit van Amsterdam.
Di Sciullo, A.-M., P. Muysken & R. Singh (1986), 'Government and code-mixing', Journal of linguistics 22, p. 1-24.
Haan, G. de (1990), 'Grammatical borrowing and language change: the Dutchification of Frisian', in: D. Gorter et al (eds.), Fourth International Conference on Minority Languages. Vol I: General Papers, Clevedon: Multilingual Matters.
Hansen-Jaax, D. (1995), Transfer bei Diglossie; Synchrone Sprachphänomene im Niederdeutschen, Hamburg: Verlag Dr. Kova….
Hout, R. van & P. Muysken (1995), 'Insertion, alternation, congruent lexicalization; corpus-based approaches to bilingual speech', in: Network on Code-Switching and Language Contact, Summer school code-switching and language contact, Ljouwert: Fryske Akademy.
Jong, S. de & A.M.J. Riemersma (1994), unpublished transcripts, Ljouwert: Fryske Akademy.
Klavans, J. (1984), The syntax of code-switching; Spanish and English, Ms. MIT.
Muysken, P. (1987), 'Neutrality in code-mixing', in: Eigen en vreemd; Identiteit en ontlening in taal, literatuur en beeldende kunst, Handelingen van het 39ste Nederlands Filologencongres, Amsterdam: VU Uitgeverij.
Myers-Scotton, C. (1995), 'Arabic and constraints on codeswitching', lecture Universiteit Utrecht.
Poplack, S. (1980), 'Sometimes I'll start a sentence in Spanish Y TERMINO EN ESPANOL: Toward a typology of code-switching', Linguistics 22, p. 99-135.
Sjölin, B. (1976), "Min Frysk"; Een onderzoek naar het ontstaan van transfer en "code-switching" in het gesproken Fries, Bijdragen en mededelingen der dialectencommissie van de KNAW, Amsterdam: Noord-Hollandsche Uitgevers Maatschappij.
Wolf, H. (1995a), Feroarings yn de Fryske tiidwurdlike einrige; In ûndersyk nei in tal resinte syntaktyske feroarings by Frysktalige jongerein, research report Fryske Akademy, Leeuwarden.
Wolf, H. (1995b), 'It wie kerstfakânsje, krystfakânsje bedoel ik; Een beknopte inventarisatie van codewisseling bij Fries-Nederlands tweetaligen', Ms. Utrecht.
Wolf, H. (1996), 'Changes in the Frisian verbal complex', Ms. Ljouwert.
Ytsma, J. (1995), Frisian as first and second language; Sociolinguistic and socio-psychological aspects of the acquisition of Frisian among Frisian and Dutch primary school children, diss. KUB, Ljouwert: Fryske Akademy.