The standard assumption of any constructional approach to language is the notion that the basic unit of language as well as of linguistic description is the construction, i.e. a conventionalized form-meaning correspondence. This reading of construction comprises linguistic units of variable size ranging from morphemes to larger units. It is reflected in Goldberg’s definition of constructions:

According to Construction Grammar, a distinct construction is defined to exist if one or more of its properties are not strictly predictable from knowledge of other constructions existing in the grammar: C is a construction iffdef C is a form-meaning pair <Fi, Si> such that some aspect of Fi or some aspect of Si is not strictly predictable from C’s component parts or from other previously established constructions. (Goldberg 1995: 4)

This definition is a guideline for deciding what types of entities are to be treated as constructions, which, as far as sheer size is concerned, by definition range from morphemes to multi-word strings. Again, Goldberg states this quite clearly:

Phrasal patterns are considered constructions if something about their form or meaning is not strictly predictable from the properties of their component parts or from other constructions. That is, a construction is posited in the grammar if it can be shown that its meaning and/or its form is not compositionally derived from other constructions existing in the language. In addition, expanding the pretheoretical notion of construction somewhat, morphemes are clear instances of constructions in that they are pairings of meaning and form that are not predictable from anything else […]. (Goldberg 1995: 4)

This notion is unanimously agreed upon, as can be checked by a look at the great number of similar definitions in studies on construction grammar (see e.g. Kay and Fillmore 1999: 2-3; Croft 2001: 18-19 Michaelis 2004: 8), and there is no need for further discussion here. What is important for the purpose of this paper is the fact that this definition accords with one of the main tenets of grammaticalization studies, namely the gradience between lexicon and syntax and the intermediate stages that continuously arise while (more) grammatical items develop out of (more) lexical ones. It is this type of gradience that is dealt with in the well-known scales and clines of grammaticalization studies which are set up for grammaticalization in general, for single categories or for particular aspects of the process. To illustrate this very prominent concept in grammaticalization theory two examples may be in place here. Bybee, Perkins and Pagliuca (1994: 40) suggest a scale representing “degrees of fusion”, which proceeds from syntactic realizations to non-bound grams to inflection, derivation and lexical realization, thus spanning all formal levels of linguistic structure. A category-specific example is Lehmann’s cline for the development of case affixes, which has the following shape: “Relational noun > secondary adposition > primary adposition > agglutinative case affix > fusional case affix” (Lehmann 1985: 304).

As these clines, which lie at the core of grammaticalization processes, represent typical stages of an otherwise continuous, gradual shift through several levels of formal linguistic structure, they presuppose a view of language that does not draw a sharp line between syntactic structure and lexical entities, or between lexical entities and morphemes. Examples like these show that the notion of construction easily lends itself to such scales. And, conversely, the definition offered by constructional approaches can help isolate the relevant syntagmatic strings in a particular grammaticalization process.

In contrast to other syntactic models, construction grammar is not only concerned with the regular and productive part of language but gives equal consideration to its idiomatic structures and thus is devoted to a much larger range of phenomena than other models. This aspect, too, is generally agreed upon (see, for example, Fillmore, Kay & O’Connor 1988; Goldberg 1995: 7; Michaelis 2004: 8; Wildgen 1990: 69). It is highlighted by Kay and Fillmore in the following way:

To adopt a constructional approach is to undertake a commitment in principle to account for the entirety of each language. This means that the relatively general patterns of language, such as the one licensing the ordering of a finite auxiliary verb before its subject in English […], and the more idiomatic patterns […], stand on an equal footing as data for which the grammar must provide an account. (Kay & Fillmore 1999: 1)

The authors go on to illustrate their point by discussing the English expression what’s X doing Y (as in What am I doing reading this paper? Or What is it doing raining? ), which combines conventional as well as idiosyncratic linguistic features of English, and thus is a typical instance of a (partially) idiomatic construction (Kay & Fillmore 1999).

In studies of language change, the question of how to treat idiomatic or otherwise non-compositional structures is of the greatest importance for one simple reason. In the initial stages of change, any innovation – by definition – is not part of the regular and productive segment of language. Instead, it is marginal and irregular, and – regarded from the perspective of the existent linguistic system – deviant or even wrong. This applies, for example, to the stage of the critical contexts, as will be explained in Section 4. In order to describe these crucial stages in language change, a linguistic model is called for that a priori is prepared to account for this type of unprecedented structure in a principled way.

Construction grammar assumes that the constructions of a language are not just a collection of unrelated items but are hierarchically ordered, so that similar or common features among constructions can be motivated by their relationship to other constructions. These relations are described via the notions of unification, inheritance, and coercion. [5] In order to solve the problem of polysemy of constructions, Goldberg (1995) postulates inheritance relations whereby both formal and semantic features may be transmitted from one construction to another, and thus similarity relations as well as polysemy may be accounted for (Goldberg 1995: 67-72). Furthermore, as inheritance links are uni-directional, this model accounts for “motivation” in grammatical structure. As Goldberg (1995: 70) – drawing on suggestions by Lakoff – points out: “A given construction is motivated to the degree that its structure is inherited from other constructions in language.”

A further concept that adds to the flexibility of the descriptive tools is the notion of coercion, as applied by Michaelis (2004). She uses the term coercion to account for the possibility to unify constructions, which due to mutually exclusive semantic restrictions, should not allow unification in the first place. The author argues that constructional meaning under certain conditions induces a shift in “the designations of content words” used in the construction (Michaelis 2004: 1). This means that conflicts of compatibility are solved by the reinterpretation of certain lexical items in terms of their fit for the constructional meaning. This mechanism does not merely serve to resolve semantic conflicts or conflicts between constructional meaning and lexical meaning. Instead, it is employed to interpret regular as well as irregular combinations of constructions with lexical items – i.e. it “is responsible for both coerced and compositional meaning” (Michaelis 2004: 1) –, and thus allows a powerful generalization. [6]

So far, these concepts have been applied to synchronic variation only, and the question whether they could be fruitfully used in the description of diachronic variation has not yet been taken up. However, it is obvious that in this area there is a particularly close relationship between diachronic linguistics and construction grammar. The principle of inheritance as used by Goldberg has a clear resemblance to analogical change, which in historical linguistics has long been treated as one of the central mechanisms of language change. Paul (1995 [1920]: 106-120) attributes analogical change to the effect of “proportional equations” (“Proportionsgleichungen”), i.e. the transposition of a relation from one domain into another, which, under certain conditions, may be conceived of as a kind of inheritance relation.

In addition, coercion, understood as the use and reinterpretation of lexemes in previously incompatible constructions, is based on cognitive and pragmatic procedures like metaphorical extension (which, too, may be treated as a kind of analogical transfer) and conversational implicatures in Grice’s sense. Speakers employ metaphorical (or metonymic) extension in first using a particular linguistic item in a new way (in a new construction); hearers interpret “new” linguistic utterance, i.e. use conversational implicatures, which are based on the same cognitive procedures as those used for production.

Though Michaelis does not explicitly mention this connection, it becomes quite evident in quotes like the following, which (focussing on the hearer’s side) defines coercion as a pragmatically motivated interpretation by the participants of the speech event: “Coercion effects are triggered when the interpreter must reconcile the meaning of a morphosyntactic construction with the meaning of a lexical filler” (Michaelis 2004: 7).

These cognitive and pragmatic procedures constitute the basis of the general and ubiquitous linguistic creativity of language users (who are always speakers and hearers at the same time), which has long been known to be one of the driving forces of linguistic change (cf. Lehmann 1985). Thus, to give just one straightforward example, the metaphorical extension of a word, which first becomes visible in a “coerced construction”, may become generalized and independent of that construction (via the creativity of the speaker/hearer), which results in semantic change of that linguistic item.

The concept of “coercion”, as it is understood here, does not refer to a newly discovered cognitive or pragmatic procedure; instead, it seems to be a useful notion to refer in a summarizing way to the linguistic results of the cognitive and pragmatic procedures speakers and hearers resort to in using their language. It is this area in particular where important synergetic effects between constructional concepts and principles of language change and grammaticalization are to be expected.

Two aspects will be treated under this heading: first, the possibility to describe constructions with varying degrees of depth and detail and, second, the principle for handling syntactic, semantic and pragmatic features in a uniform way.

In contrast to most of the more rigidly formalized grammatical descriptions, construction grammar allows for structural descriptions with varying specification, including the possibility to assign particular features to constructions in a holistic way. As Wildgen (1990: 69) points out, this means that the principle of compositionality, which in most models of grammar is strongly adhered to, in constructional approaches is restricted in favour of a holistic conception of constructional meaning.

The option to choose descriptions of varying degrees of specification is of a great advantage in diachronic investigations, because of the trivial but far-reaching fact that for past stages of a language there are no longer speakers available whose competence one could draw on. Therefore, in any linguistic analysis of historical language, no matter how carefully it is conducted, gaps and uncertainties remain. Descriptive models requiring a complete analysis of the whole linguistic structure inevitably run the risk of anachronistic distortions, as the linguist is forced to transfer his or her own system onto the historical data.

In contrast to that, the descriptive practice favoured by constructional approaches provides for analytical solutions that avoid over-specified, non-provable descriptions and that better reflect the actual state of linguistic knowledge about an imperfectly known, past linguistic system.

A second merit with regard to formal aspects is the fact that constructional approaches do not make a sharp distinction between syntactic, semantic and pragmatic features, but postulate a “gestaltlike interaction of formal, semantic and pragmatic constraints” (Michaelis & Lambrecht 1996: 215). This again is conducive to the investigation of grammaticalization processes. The rise of grammatical markers is fundamentally based on the interrelation and gradual transition from one level of linguistic organization to another: new meanings arise out of stereotypical pragmatic inferencing, lexical forms acquire grammatical functions, the free ordering of discourse elements may become obligatory and syntactically constrained, syntactic structures become morphologized, morphemes erode and become phonological features, etc. The essence of this process was first described by Givón (1979: 209) in the famous grammaticalization scale, which is reproduced here: “discourse > syntax > morphology > morphophonemics > zero”.

In summary, in a study of the rise of grammatical markers, a neat separation along the lines of the traditionally distinguished linguistic levels is as impossible as it would be counter-productive.

The six German modals dürfen ‘to be allowed to’, können ‘can, to be able to’, mögen ‘to like, may’, müssen ‘must, to have to’, sollen ‘shall, to be to’ and wollen ‘to want’ are a paradigm case of grammaticalization. They all have a less grammaticalized use, in which they are stative verbs with typically narrow scope, and a highly grammaticalized wide-scope use as factuality markers, where they approach the stage of auxiliaries. The two uses are illustrated with the modal müssen in (1) and (2):

In (1) mußt predicates the state of ‘being obliged’ of the subject, i.e. the modal has narrow scope and lexical meaning. This is a prototypical example of the lexical use, which in the literature is often called deontic or agent-oriented. In (2) muß does not contribute to the propositional content of the sentence, i.e., unlike mußt in (1) it does not express the obligation of the subject. Instead, it has wide, i.e. propositional scope and expresses uncertainty on the part of the speaker concerning the factuality value of the whole proposition. The degree of uncertainty can be made explicit by a paraphrase with a sentence adverbial that roughly conveys the meaning of the modal. For the meaning of muß in (2) this is illustrated in (3):

In their use as speaker-based factuality markers the modals are integrated into the grammatical paradigm of verbal mood, i.e. they participate in a deictic grammatical category (Diewald 1999: 167-248).

It is well known that the uses of the modals illustrated in (1) and (2) are context dependent, insofar as there are contexts that favour one use to the exclusion of the other. The most important factors here are structural ones, or to be more exact, the morphological categories of the modal verb and the morphological categories of the infinitive. These factors are linked to the two readings in the following way: the grammaticalized modals do not allow periphrastic tenses, which means that, if a modal is used with a periphrastic tense, it can only have the less grammaticalized meaning. This is shown in (4), which allows only a lexical reading of hat müssen , etc.): [7]

On the other hand, there is a type of context in which the lexical reading is virtually excluded and the grammaticalized reading is highly favoured. This is the combination of the modal with an infinitive perfect, as in (2). (5) gives two further examples of this context type with other modals. They, too, have a grammaticalized reading only:

Summarizing this section, it can be said that for both of the central uses there is a specific morphologically marked context type that favours one reading to the exclusion of the other. As these contexts have the capability to distinguish the most grammaticalized reading of the modals from their prototypical lexical reading, they are called the isolating contexts here. Table II gives an overview:

As can be derived from the description above, an isolating context for the different readings of the modals must consist of at least three verbal elements, which are represented by the examples in Table II. To give an overview, the schematic structures of these two isolating contexts are rendered in Table III as set a) e.g. loben hat können, and set b) e.g. gelobt haben kann, respectively: [8]

It is suggested here that these two isolating contexts are constructions in the construction grammatical sense. More precisely, they belong to a subtype of idioms which in Fillmore, Kay & O’Connor (1988) are classified as “formal or lexically open” idioms. The authors define these idioms as “syntactic patterns dedicated to semantic and pragmatic purposes not knowable from their form alone” (Fillmore, Kay & O’Connor 1988: 505). Formal idioms are constructions whose compositionality is reduced, i.e. at least some part of their form-meaning correspondence has to be treated holistically and cannot be derived in its totality from other constructions or from a combination of other constructions. However, they are still fully productive, as their syntactic positions are not filled with lexically fixed items.

The arguments for classifying the isolating contexts as formal idioms are the following. First, the opposition between the two constructions exists only for the modal verbs (and for brauchen). No other German verb has this constructional choice. It is a relation between constructions that is specific to the modals. [9] This is illustrated in the examples (6) to (10):

(6) repeats the opposition of the two isolating contexts for the modals. (7) to (10) show that other verbs taking the so-called “Ersatzinfinitiv” do not allow the second construction (finite verb and infinitive perfect of a main verb) that is available to the modals. (10) illustrates the difference between the modals and the verb werden, which has the construction with finite werden and the infinitive perfect of the main verb, but no construction with an “Ersatzinfinitiv”. In short, only the modals have access to both constructions constituting the isolating contexts.

The second argument for regarding the isolating contexts as constructions rests on their semantics. As already mentioned and as noted in table II, both structures are associated with a discrete and unambiguous constructional meaning, i.e. a meaning which is not present in any other construction involving modal verbs. The isolating context for the lexical reading has a constructional meaning which predicates a modal state on the subject of the sentence. The isolating context for the grammaticalized reading, on the other hand, has a constructional meaning that can be paraphrased as “speaker attributes an uncertain factuality value to the whole proposition”. The distinctive semantic features added by the lexical content of each modal (the meaning of können versus müssen, etc.) are not part of the overall constructional meaning but part of the lexical construction of the modal itself.

Thus, though there are partial similarities with other constructions which indicate partial inheritance relations, the isolating contexts of the modals display a unique form-meaning correspondence which cannot be reduced to combinations of other constructions. They therefore qualify as partially productive, idiomatic constructions in the constructional sense of the term elaborated above.

The isolating contexts of the modals developed as the third stage of the grammaticalization process in the Early New High German period, roughly between 1500 and 1650 (Diewald 1999: 379-384). The preceding second stage, which represents the triggering of the grammaticalization process, is characterized by another type of context, the critical context. As already described in section 2, the critical context is a highly ambiguous structure, which through morpho-syntactic opacity allows several options for its interpretation, among them the new grammaticalized meaning. It is important to note that, in contrast to the other context types, the existence of the critical context is restricted to a fairly narrow time span in history and does not exist before or after that critical period. For the grammaticalization of the modals it is found in Middle High German, around 1200. It is the construction shown in table IV:

The critical context consists of the modal with the dental suffix -t- (“DS”) plus an optional nominal object plus haben/hân/sîn plus a past participle of a lexical verb. Although this construction is not attested before the middle of the 12th century, it is already found with all six modals around 1200 with a relatively high frequency (Westvik 1994; Deeg 1948), which means it spread rapidly. What is most important for the topic of this paper is the fact that this construction is only attested for the modals. There are no other verbs found in this construction at that time (Paul, Wiehl & Grosse 1989: 295f.). (11) gives an example with the modal können, which will be used here for further illustration: [10]

The interpretation of this construction is difficult, to say the least. It is highly ambiguous (which is not rendered fully in the semantic paraphrase in (11). Thus, disregarding the further linguistic context, (11) can be translated into Present-Day German (PDG) in at least the following three ways (compare Deeg 1948 and Westvik 1994):

This three-way ambiguity is not present in every instance of the construction, and even if it is, not all of the three readings are equally plausible in a particular context. However, and this is the decisive point, whenever this construction is used in that period, there is ambiguity, which includes the deictic reading. The next passage, for example, is ambiguous between the deictic reading (12a) and the (older) reading of the participle as an object complement, which finally developed into the periphrastic perfect tenses (12b):

The opacity of this MHG construction is not only a semantic one. The morphological forms that build it are highly opaque themselves. First, the non-finite structure hân + past participle is ambiguous between a reading in the sense of the infinitive perfect of PDG, in which the participle is the main verb and hân or sîn is the auxiliary, and a reading as a complex predicative structure where ‘have’ is the main verb and the past participle functions a predicative adjective (like in 11b: der kunde sie gelobet hân’ der konnte/könnte sie als Gelobte haben’).

Second, the morphological form of the modal itself is ambiguous, too. For MHG modals with a dental suffix it is not possible to distinguish between the indicative and the subjunctive of the past (Westvik 1994), because the formal marking of the subjunctive by “Umlaut”, which today provides a clear and systematic opposition between mochte vs. möchte, konnte vs. könnte, had not yet developed at that time and there was a large amount of formal syncretism (see Birkmann 1987: 194). Therefore it is reasonable to assume that the MHG modals with a dental suffix express a distal value which – depending on context – can be interpreted as modal and/or as temporal distance.

Thus, the critical context is characterized by the coincidence – or rather the clash – of two verbal forms that were both morphologically and morpho-syntactically ambiguous. These forms could not therefore mutually disambiguate each other. Being confronted with this construction, the recipient had several possibilities for interpreting it without gaining a clear indication from the construction itself.

It is proposed here that it is this structure, the critical context, in which the deictic reading and its necessary wide scope was not only one alternative interpretation among equally plausible others, but the most likely, the favoured, reading. [11] It could be reached from different starting points, i.e. from different ways of interpretating the opaque morphological forms, by similar conversational implicatures. However, as this is not the topic of this paper, it will not be followed any further. Instead, table V very briefly summarizes the different possibilities of resolving the opacity of the critical context:

Again, it is proposed here to treat this kind of multiple structural and semantic opacity as an idiomatic construction in the sense of Fillmore, Kay and O’Connor (1988). This time, however, we are confronted with what the authors call an extragrammatical idiom, and which they define as follows: “Such expressions [i.e. extragrammatical idioms] have grammatical structure, to be sure, but the structures they have are not made intelligible by knowledge of the familiar rules of the grammar and how those rules are most generally applied” (Fillmore, Kay & O’Connor 1988: 505).

As the critical context of the modals in MHG is a new and peripheral structure restricted to a small group of verbs and not fully analysable by the rules of the relevant linguistic system, it clearly is an instance of this type of idiom.

Having reached this conclusion, we are still presented with a problem. This construction neither shows an unambiguous though idiomatic form-meaning correspondence (like the isolating contexts of PDG), nor can it be treated as a simple case of polysemy or vagueness, because the different meanings correlate with different structural analyses. Instead, we are confronted with a complex structure containing several potential meanings, each of them associated with a distinct structural analysis. This raises the question which semantic and structural analysis should be assigned to this construction in the first place. Without being able to provide a complete answer to this question, I would like to end with a brief indication of the direction a solution could take.

As for the structural description, I propose to model it closely according to the attestable distinctions of the surface forms, i.e. to those features already mentioned in the description of the critical context. In particular, the dental suffix of the modal should be interpreted as a marker of a distal value without assigning a particular mood or tense, and, second, one should refrain from stipulating a particular interpretation as a periphrastic verbal phrase for the non-finite verbal forms (infinitive, past participle) present in the construction. As far as the meaning of the construction is concerned, there are indications that this construction may have had a specific stylistic function, namely that of reinforcement or emphasis. Thus, it seems reasonable to assume a “pragmatic” meaning which is assigned holistically to the construction. The content of this pragmatic meaning might be circumscribed as an iconic relation between the structural and semantic layer on one side and the pragmatic and stylistic layer on the other. [12] This suggestion rests on the assumption that the extraordinary semantic and structural opacity (which surely originates in conscious creative choices of the authors) is meant to iconically indicate an extraordinary stylistic salience, and thus serves as an instruction to the hearers or readers to recognize and evaluate the interpretational options and choose a suitable one themselves. This, however, needs further empirical investigation, and is beyond the scope of this article.

[FN] = Fürther Nachrichten, 21/22 January 1995.

[Texte] = Texte gesprochener deutscher Standardsprache. Erarbeitet im Institut für deutsche Sprache Forschungsstelle Freiburg i. Br. Volume 3. Alltagsgespräche. Fuchs, Harald P. & Gerd Schank (eds.) (1975).München: Hueber.

[Zeit] = Die Zeit, Nr. 5, 27 January 1995.

[Parz] = Wolfram von Eschenbach. Parzival. 2 volumes. Nellmann, Eberhard (1994) (ed.). Frankfurt/Main: Deutscher Klassiker Verlag.

Birkmann, Thomas (1987). Präteritopräsentia: Morphologische Entwicklungen einer Sonderklasse in den altgermanischen Sprachen. Tübingen: Niemeyer.

Bisang, Walter (1998). Grammaticalization and language contact, constructions and positions. In Ramat, Anna Giacalone & Paul Hopper (eds.) The Limits of Grammaticalization. Amsterdam/Philadelphia: Benjamins. 13-58.

Bybee, Joan, Revere Perkins & William Pagliuca (1994). The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press.

Croft, William (2001). Radical Construction Grammar. Oxford: Oxford University Press.

Deeg, K. (1948). Der Infinitiv Perfekt im Frühmittelhochdeutschen. Dissertation Universität München (Typoscript).

Diewald, Gabriele (1999). Die Modalverben im Deutschen: Grammatikalisierung und Polyfunktionalität. Tübingen: Niemeyer.

Diewald, Gabriele (2002). A model for relevant types of contexts in grammaticalization. Wischer, Ilse & Gabriele Diewald (eds.). 103-120.

Fillmore, Charles & Paul Kay (1995). Construction Grammar. Stanford. CSLI Publications (CSLI Lecture Notes) (Manuscript).

Fillmore, Charles, Paul Kay & Catherine O’Connor (1988). Regularity and idiomaticity in grammatical constructions: the case of let alone. Language 64: 501-538.

Givón, Talmy (1979). On Understanding Grammar. New York: Academic Press.

Goldberg, Adele E. (1995). A Construction Grammar Approach to Argument Structure. Chicago: Chicago University Press.

Grice, Paul (1989 [1975]). Logic and conversation. In Grice, Paul Studies in the way of words. Cambridge, MA: Harvard University Press. 22-40.

Heine, Bernd (2002). On the role of context in grammaticalization. Wischer, Ilse & Gabriele Diewald (eds.). 83-101.

Hopper, Paul & Elizabeth ClossTraugott (2003 [1993]) Grammaticalization. 2nd edition. Cambridge: Cambridge University Press.

Kay, Paul & Charles P. Fillmore (1999). Grammatical constructions and linguistic generalizations: The What's X doing Y? construction. Language 75: 1-33.

Lehmann, Christian (1985). Grammaticalization: Synchronic variation and diachronic change. Lingua e Stile 20: 303-318.

Lehmann, Christian (2004). Theory and method in grammaticalization. Zeitschrift für germanistische Linguistik 32(2) (special issue “Grammatikalisierung” edited by Diewald, Gabriele): 152-187.

Michaelis, Laura & Knud Lambrecht (1996). Toward a construction-based theory of language function: The case of nominal extraposition. Language 72: 215-247.

Michaelis, Laura (2004). Type shifting in construction grammar: An integrated approach to aspectual coercion. Cognitive Linguistics 15: 1-67.

Paul, Hermann (1995 [1920]). Prinzipien der Sprachgeschichte. 10th edition. Tübingen: Niemeyer.

Paul, Hermann, Peter Wiehl & Siegfried Grosse (1989). Mittelhochdeutsche Grammatik. 23th edition. Tübingen: Niemeyer.

Traugott, Elizabeth Closs (2003). Constructions in grammaticalization. In Joseph, Brian & Richard D. Janda (eds). The Handbook of Historical Linguistics. Oxford: Blackwell. 624-647.

Westvik, Olaf Jansen (1994). Die Struktur hätte + Infinitiv + Modalverbinfinitiv: Aspekte von Vorgeschichte und Geschichte. In Leirbukt, Oddleif (ed.) Proceedings of the 11th international Tromsø symposium on language: “Modalität im Deutschen” (Nordlyd: Tromsø University working papers on language and linguistics 22). Tromsø. 142-161.

Wildgen, Wolfgang (1990). Konstruktionsgrammatik. In Wagner, Karl Heinz & Wolfgang Wildgen (eds.) Studien zur Grammatik und Sprachtheorie. Bremen: Univ. Bremen (Bremer Linguistisches Kolloquium 2). 65-84.

Wischer, Ilse & Gabriele Diewald (2002) (eds.). New Reflections on Grammaticalization. International Symposium, Potsdam, 17-19 June, 1999. Amsterdam: Benjamins.