Summary report on the results of the project
Modelling syntactic change: a case study in Pennsylvania German
ESRC grant number R000237820
Background
The data
Comparison with other Germanic languages
Grammaticalisation of fer
The change as a contact phenomenon
Theoretical modelling of the change
A related issue
References
Background
The background for the research funded by this grant was the realisation that two related changes had taken place recently in Pennsylvania German (PG) as spoken in Waterloo County, Canada and that these changes had conspired to create a major change to the non-finite complementation system. The details of this change were badly documented. The change is particularly fascinating since no other variety of Continental West Germanic appears to have undergone a similar change, but at the same time, there is no obvious way in which the change can be shown to have resulted from contact with English. One dimension of the change involves the spread of a purposive construction, the fer...zu construction, from its original, adverbial, function to complement clauses. At the same time, there has been a loss of the infinitival marker, zu, so that many non-finite clauses now contain only one overt functional element associated with (the extended projection of) the clause.
The data
On the basis of the data recorded, we have been able to establish that the fer...(zu) construction is still standardly used in purposive constructions, as it is in a number of non-standard varieties of German. At the same time, our initial impression that fer...(zu) has spread widely through the complementation system has been confirmed. We have found the construction in complement clauses with a wide variety of matrix verbs, as complements or modifiers to adjectives and nouns and in subject position.
The database also gives clear evidence that the loss of zu which was evident in earlier material has progressed. This loss of zu is evident in all age groups and with both denominations. Younger speakers will not use zu at all in translation sentences or story telling; the oldest speaker who did not use zu in any environment was in her early thirties. All of these speakers will, however, accept sentences with zu where they would themselves never use it. Similarly, the older speakers with a higher frequency of zu will not describe as ungrammatical the equivalent sentences with fer as used by younger speakers. In environments where zu previously occurred in conjunction with fer, we now get fer as the only functional element. A few examples are given in (1).
(1) a. Er hot des gduh fer sei dankbarheit weise.
he have3.sg this do.part FER his gratitude show. inf
'He did this in order to show his gratitude.'
b. Es is gebreichlich fer de huut ab duh.
it be.3sg customary FER the hat off do.inf
'It is customary to take the hat off.'
c. Ich hab versproche fer iem helfe.
I have.1sg promised FER he.dat help
'I have promised to help him.'
d. Ich bin ready fer gee.
I be.3sg ready FER go.inf
'I am ready to go.'
A striking fact is, however, that fer is now also used in constructions which appear never to have had fer&zu, but only zu. In these environments then, fer has actually replaced zu. One example of such an environment is so-called wh-infinitivals.
(2) a. Er hot nett gwisst was fer duh.
he have.3sg not know.part what FER do
'He didn't know what to do.'
b. Ich wees nett wu fer gee.
I know.1sg not where FER go
'I don't know where to go.'
In recordings from ten years ago, such constructions obligatorily contained zu and never fer. Interestingly, this environment is the last stronghold of zu, so that many speakers who would not otherwise use zu still have this as the only option for wh-infinitivals. The speakers who do not use zu at all, however, now use fer in this construction.
A further indication that the role of fer within the complementation system has changed is the fact that it now occurs in environments in which a bare infinitive was previously the only option.
(3) a. Ich wees er gee wil fer schwimme.
I know.1sg he go.inf want FER swim.inf
'I know that he wants to go swimming.'
b. Ich hab ien gsehne fer die daer reikumme.
I have.1sg he.acc see.part FER the door come.through.inf
'I saw him coming through the door.'
Amongst the older speakers, who make a distinction between fer-infinitivals and zu-infinitivals, we have been able to establish a subtle but consistent contrast between the use of the two types. With zu infinitivals, there is an implication of immediate future, whereas fer-infinitivals imply potentiality. Hence for these speakers, the example in (4a) implies that the speaker is on his way to milk his cows, whereas (4b) is a statement about the type of farm that he has.
(4) a. Ich hab fufzich kien zu melge.
I have.1sg fifty cows ZU milk.inf
'I have fifty cows to milk.'
b. Ich hab fufzich kien fer melge.
I have.1sg fifty cows FER milk.inf
'I have fifty milking cows.'
This distinction can be related to the original prepositional source of the two elements, zu as a preposition still is directional and direction in time can be seen as a metaphorical extension of this. The origin of fer is the benefactive, from where it has been extended to the purposive and a remnant of this meaning leads to the preferred reading of (4b). It is indeed possible that some similarities between the use of to-infinitivals and the -ing form in PG which have been put down to contact with English, are in fact due to similarities in the grammaticalisation path of the elements rather than contact (cf Louden 1988:209-212).
Comparison with other Germanic languages
Within the Germanic language family, there are a number of other varieties which have fer...zu rather than um...zu in purposive constructions, e.g. Pfälzisch (Henn 1980) and Luxemburgish (Bruch 1973 [1968]). As far as we are aware, apart from PG, it is only Luxembourgish (L) that has undergone a spread of fer...zu clauses from the purposive into complement clauses. We have done some initial work with native speakers and written texts in Luxembourgish, and the spread of fer...zu seems to be quite far-reaching in this variety too. There is, however, a major difference between the two varieties and that is that in L, fer is optional. In a few environments, zu is optional as long as fer has been deleted, hence giving rise to a bare infinitive as an alternative to fer...zu or zu. In L then, even though a purposive element has spread to a general complementiser position, it shows no signs at the moment of developing further into an infinitival marker. Hence the L development is not parallel to the PG in this respect.
Even though these changes in PG appear to be unique in current varieties of CWG, in a general historical and typological perspective they are not. The historical development from a purposive marker to an infinitival marker is a common one typologically; Haspelmath (1989) gives examples of this change from a wide variety of language families; Turkic, Finno-Ugric, Semitic, Bantu, Nakho-Daghestanian and Dravidian. The development from a purposive marker to an infinitival marker can then be described as a common path of grammaticalisation. In fact, as Haspelmath (1989) demonstrates, the same change has already happened in Standard German, in that zu also originated as a purposive marker.
Grammaticalisation of fer
In the literature on grammaticalisation, it is standardly assumed that the process of grammaticalisation can affect all aspects of a linguistic element's properties: phonological, morphological, syntactic (structural) and semantic (functional). In the case of fer in WCPG, we cannot discern any phonological or morphological changes. The preposition fer, the purposive fer and the fer in complement clauses all have the same phonology, and none of the elements has morphological complexity. In order to study the grammaticalisation path of fer we need then to focus on structural and functional aspects of its use.
Since purposive fer is already a functional element, this is a case of grammaticalisation from a grammatical element into a more grammatical element (Heine et al 1991: ch6). This in turn means that we shall have to involve in our account features not of lexical semantics but also relating to functional properties of clausal structure. At the same time the changes documented here can be used as a tool to diagnose what that clausal structure must be. In this way, we can contribute to refining the models of syntactic description used within grammaticalization studies (Vincent 1999; 2001). [See section 6 below for the development of this argument.] We follow Lehmann (1982, 1995) and Haspelmath (1989) and view the changes in terms of semantic generalisation. In the case of fer, the origin is the benefactive preposition which takes on the use of a purposive marker. Purposive can be viewed as an abstract form of benefactive meaning and hence this is a first step in the direction of semantic generalisation. We have then followed Haspelmaths definitions of increased generality of clause type, which is paired with a decreased concreteness in meaning:
irrealis directive > irrealis potential > realis non-factive > realis factive.
Haspelmath shows that the German zu has progressed along this line and that it now occurs in all environments. In the extreme factive use zu is therefore almost entirely desemanticised (though it can retain some modality in the less general uses).
Considering the WCPG now in the same light, we find a very similar development to that of German zu. We find fer used in all clause types, including the realis factive:
(5) a. Ich glaab er deet gleiche fer danse.
I think.3sg he subj like FER dance.inf
'I think he would like to dance.' [Irrealis-directive]
b. Ich bin ready fer gee.
I be.1sg ready FER go
'I am ready to go.' [Irrealis-potential]
c. Sie ecschpect dich fer kumme frie.
she expect.3sg you.acc FER come early
'She expects you to come early.' [Realis-non-factive]
d. Ich bin verschtaunt fer dich sene do
I be.1sg astonished FER you.acc see.inf here
'I am astonished to see you here.' [Realis-factive]
The change as a contact phenomenon
It has been suggested in the literature that the changes that have taken place to the non-finite complementation system of PG are due to language contact (e.g. van Ness 1994). However, we are not aware of any author who has actually provided arguments to support the case for contact induced change. To the contrary, Louden (1988:205-212) argues against this conclusion on the basis that the fer...zu construction in its purposive origin clearly is native to the varieties of German from which PG originated as well as the fact that the variety of American English spoken around the original PG speakers in Pennsylvania is unlikely to have been a so-called for...to variety. As mentioned in section 3 above, the type of change that has taken place in PG is a common one typologically and there is no reason to assume that it cannot be a case of language internal change. The fer...zu construction of PG is also different in detail from both the Old English for...to construction and the use of the construction in modern for..to varieties of English. Furthermore, even though there are for...to dialects in Canada (Carroll 1983), there is no evidence that contact with these have played any role for the WCPG speakers. A comparison of our data with those of Louden shows that the details of the changes in the use of fer...(zu) are very similar across varieties of PG and hence a contact explanation that relies on any local variety of English is implausible. As part of the work on this project, we have considered two different theoretical accounts of modern English for...to varieties and their implications for the WCPG data, and found that there is littleto suggest that the accounts can be carried over to WCPG.
The theoretical modelling of change
In terms of the widely adopted Principles & Parameters account of clause structure, the changes we have identified concern the C(omplementizer) and I(inflection) positions in a hierarchically defined "spine" of extended projections. However, if we followed the strict logic of that model, any functional shift of an item such as fer would have to be accompanied by a positional shift since the material that was formerly associated with the C head now marks categories that belong to the I head. Indeed in that framework it is generally assumed that items like English for or German um are in C whereas to/zu are in I. If PG fer assumes the function of zu it ought therefore to shift to I, and this change ought to be discernible in its linear position relative to other items in the clause. We have applied a number of the criteria used in the literature to our data in an attempt to establish whether the syntactic scope of fer has been narrowed, i.e whether there has been structural lowering of fer. Thus, the expected order when both for/fer and to/zu are present would be
(6) for/fer subject negation to/zu
In an approach within which functional and structural changes are assumed to be parallel, refunctionalisation of fer ought to mean that speakers accept constructions in which it is preceded rather than followed by the subject and/or the negative particle. We were able to force acceptance of such an order by some speakers by direct elicitation. Many speakers were however still reluctant to accept them, and for all speakers the preferred pattern in naturally occurring discourse is with fer in clause-initial position.
Positional change, then, to the extent that it has happened, lags considerably behind functional change. This has led us to formulate an account in terms of an approach that can express functional and structural properties in separate dimensions, namely Lexical Functional Grammar (LFG, Bresnan 2000). In this approach, complementizing and inflectional functions can be represented without identifying them with specific structural positions. A clause in which fer and zu have their older functions in a purposive clause without zu -deletion can be represented schematically as in (7).
Here fer combines with a subjectless clause (XCOMP) to form a constituent that has purposive meaning and belongs to the set of adjuncts that its mother has. This ensures an adjunct interpretation, but allows for further adjuncts to be associated with the mother. The infinitival zu, on the other hand, is a particle which does not add meaning itself. This is a similar interpretation of the function of zu to that adopted in earlier versions of Head-driven Phrase Structure Grammar of English to as a raising verb (Pollard and Sag 1995, but compare Sag and Wasow 2000).
By contrast, the more recently evolved fully desemanticised general function, fer would have the same f-structure as zu in (7b), as in (8). Since it is also used as a purposive, it would retain an alternative f-structure as in (7a).
(8) fer = ¯
In terms of c-structure, the original use of fer with zu can be represented as in (9a) (cf example like (1a), but previous to zu loss), and the more recently evolved general use as in (9b) (cf example (5d)). The C node in (9a) would be annotated with the f-structure from (7a) and the C node in (9b) with that of (8).
(9) a. C''
C I''/ S
fer sei dankbarheit zu weise
b. C''
C I'' / S
fer dich sene do
In the context of this report, these representations are necessarily simplified, but they serve to indicate the general form of the analysis we have been developing. In particular, they show clearly how it is possible to provide a formal analysis in which function and structure are expressed on different planes.
A more radical, but in some ways more promising, alternative that we have begun to experiment with abandons the requirement that these functional elements need to be identified with a c-structure position. This leads to an analysis in the spirit of work by Anderson (1993, 1996) and following work on functional information associated with noun phrases by Payne and Börjars (2000). In this approach, the functional category can, but need not, be represented under a separate node in the tree. Whether it is or is not an independent category, its position is accounted for through alignment constraints. On this view items are assigned to a position in a linear string not by insertion into hierarchically defined tree but as a function of a special class of constraints --- so-called alignment constraints --- which determine an item's position relative to the right or left edge of the relevant domain, in the present instance the left edge of the clause. The attraction of this Optimality Theory view is that such constraints are "soft": i.e. they may either override or be overridden by competing constraints, such as those determining the scope of negation or the position of the subject. Marked situations then arise historically when the constraint governing a given item's alignment outranks other constraints that might have affected its distribution. This approach in turn requires adopting the model of OT/LFG developed for instance in Bresnan (1998) and applied in the historical domain in Vincent (1999, 2001).
A related issue
As a result of work done on this project, we have made an interesting discovery with respect to one further issue related to non-finite complementation, which we did not set out to study. In Continental West Germanic languages (CWG) such as German, subordinate clauses are verb final. If this verb selects as its complement another clause, then this clause occurs to the left of its matrix verb (this is obligatory in non-finite clauses, whereas in finite clauses such heavy constituents may be shifted to the right). Given that all the clauses involved are verb final, we would in principle expect to find a verb cluster at the right edge of the subordinate clause in which the matrix verb is the rightmost one, immediately preceded by the second highest verb etc, until on the left we get the non-verbal complements (if any) of the lowest verb. This is illustrated in (10).
(10) & [ [ [ (NP) (AP) V3non-fin ]compl of V2 V2non-fin ]compl of V1 V1fin ]
Even though this order does occur in varieties of CWG, it is by no means the only attested order; the verb cluster can be "scrambled". Many different accounts for such data have been proposed, making different prediction as to the limitation to the variation. One of the influential theoretical articles on the subject, Haegeman and van Riemsdijk (1983), for instance, predicts that V2 V1 V3 should not be possible, still this order is commonly found in the WCPG data recorded for the purposes of this project. An example is given in (11).
(11) ... das sie mich gheert hot kumme.
that she I.acc hear.part have.past come.inf
'...that she heard me coming.'
Louden (1990) claims for the PG material that he has recorded this is indeed the standard order. However, our data shows that for the variety of PG that we have studied other orders also frequently occur. Even though we have not studied the variation in WCPG and its limitation in detail, it is clear that our data forms an interesting basis for a detailed study of the factors which influence what orders are possible.
References
Anderson, S. R. 1993. Wackernagel's revenge: clitics, morphology, and the syntax of second position. Language 69:68-98.
Anderson, S.R. 1996. How to put your clitics in their place, or why the best account of second-position phenomena may be something like the optimal one. The Linguistic Review 13:165-191.
Bresnan, J. 2001. The Emergence of the Unmarked Pronoun. In G. Legendre, J. Grimshaw and S. Vikner (eds), Optimality-theoretic syntax. Cambridge, MA: MIT Press. 113142.
Bresnan, J. 2000. Lexical Functional Syntax. Oxford: Blackwell.
Bruch, R. 1973 [1968]. Luxemburger grammatik in volkstümlichem abriss. [Beiträge zur Luxemburgischen Sprach- und Volkskunde] Luxembourg: Editions de la Section de Linguistique de l'Institut gr.-d.
Carroll, S. Remarks on for-to infinitives. Linguistic Analysis 12:415-451
Cinque, G. 1999. Adverbs and functional heads. Oxford: Oxford University Press.
Haspelmath, M. 1989. From purposive to infinitive - a universal path of grammaticalization. Folia Linguistica Historica X:287-310.
Heine, Bernd, Ulrike Claudi & Friederike Hünnemeyer (1991) Grammaticalization: a conceptual framework. Chicago & London: The University of Chicago Press.
Henn, B. 1980. Pfälzisch. Düsseldorf: Pädagogischer Verlag Schwann.
Lehmann, C. 1982. Thoughts on grammaticalisation. A programmatic sketch. Köln: Institut für Sprachwissenschaft, Universität zu Köln.
Lehmann, C. 1995. Thoughts on grammaticalisation. [Rev and expanded version]. München: Lincom Europa.
Lightfoot, D. 1999. The development of language: acquisition, change and evolution. Oxford: Blackwell.
Louden, M. 1988. 'Bilingualism and syntactic change in Pennsylvania German'. PhD dissertation. Cornell University.
Payne, J and K. Börjars 1999. On the justification for functional categories in LFG. In M. Butt and T. Holloway King (eds), Proceeedings of LFG 99, Stanford, Ca. Online proceedings at http://www-csli.stanford.edu/ publications/LFG4/LFG99.htm.
Payne, J and K. Börjars 2000. The exponence of definiteness in the noun phrase: and LFG account. Paper presented as part of the Workshop on m-structure. LFG2000, Berkeley, Ca, June 2000.
Pollard, C and I. Sag 1995. Head-driven Phrase Structure Grammar. Chicago: University of Chicago Press.
Rizzi,L. 1997. The fine structure of the left periphery. In L. Haegeman (ed) Elements of syntax. Dordrecht: Kluwer. 281-337.
van Ness, S. 1994. Pennsylvania German. In E. König and J. van der Auwera (eds), The Germanic languages. London: Routledge. 420-438.
Vincent, N. 1999. The evolution of c-structure: prepositions and PPs from Latin to Romance. Linguistics 37: 1111-1153.
Vincent, N. 2001. Competition and correspondence in syntactic change: null arguments in Latin and Romance. In S. Pintzuk, G. Tsoulas & A. Warner (eds), Diachronic syntax: models and mechanisms. Oxford: Oxford University Press. 25-50.
Vincent, N. 2001. Lexical historical grammar. In M. Butt & T. Holloway King (eds) Time over matter: diachronic perspectives on morphosyntax. Stanford: CSLI Press.
Vincent, N. In prep. Review of Cinque (1999). To appear in Journal of Linguistics.