1. Construction Grammar as a theory of the co-existence of the generalized and the item-specific

1When, in the first quarter of the 21^st century, one is looking for a theory that is likely to increase our understanding of the nature of language, Construction Grammar – in the sense of the research sailing under the names of Cognitive Grammar, usage-based approach and the various versions of explicitly constructionist approaches that are united by a number of key positions (Beckner et al. 2009; Hilpert 2020) – seems a prime candidate. This is not only because these positions are grounded in an overwhelmingly large body of empirical research ranging from corpus studies and phonetic analyses to psychological and neurolinguistic experiments, but also because it does not divide observed phenomena into those that are worthy of theoretical consideration and those that are not.

2In fact, constructionist models of language have moved idiomaticity and idiosyncrasy from the periphery of interest (back) to a center of linguistic theory (Croft & Cruse 2004: 225). The reason why the let alone and the Xer the Yer studies by Fillmore (Fillmore, Kay & O’Connor 1988) have had such an impact on the development of constructionist theory is not because of an interest in these constructions as such, but rather because they revealed the enormous complexity of “grammatical” knowledge associated with such low-level constructions and thus contributed to giving rise to a comprehensive theory of language that is able to encompass the findings of research on collocation, valency and other item-related phenomena alongside providing room for abstract schemata and generalized constructions (see also Gries 2008 and Gilquin & De Knop 2016)1.

3In this contribution, we would like to shed some light on a construction to be found in present-day German, which is to be located at the more idiosyncratic pole of the constructionist spectrum – the so-was-von construction. What we find so remarkable about this construction is that – unlike many other idiomatic low-level constructions – synchronically it is not particularly transparent. We aim to show here that historically the so-was-von construction can indeed be explained as an expression that is formed on the basis of established patterns of today’s language and has fossilized in some respect or another, as we will outline in Section 4 after providing a description of its formal and functional properties in Section 2 and a possible path of motivation in Section 3.

2. The prototypical so-was-von construction

4The following examples are rather typical of the German so-was-von construction:

5In such uses, we can describe so was von as functioning as an adjectival premodifier. Its semantic function is one of intensification of a gradable adjective or a past participle to an extreme degree. This is equally true when so was von is used with nouns (often denoting qualities and without an article) in examples such as:

6In many cases, however, nouns following so was von are part of a complex or idiomatic expression and it is this complex expression rather than the noun itself that is being intensified; compare:

7The same observation can be made with respect to prepositional phrases such as

8There are, however, also uses outside idiomatic expressions:

9The fact that the German preposition von requires a dative, but the NPs underlined in (4) are accusative provides further support for the construction status of so was von because von quite clearly cannot be seen as the head of a prepositional phrase, the underlined NPs in (4) and (5) being complements of auf and an respectively. This also shows in examples such as (6), where so was von is followed by a nominative NP:

10Other uses worth noting include so was von before verbs, adverbs or numerals (which however one might interpret as adjectives according to Duden 2016: 392, § 513):

11The so-was-von construction does not seem to occur freely on its own, although the so was von part can be placed after the quality that is being intensified as in (8), as in spoken dialogue or blogs:

12The construction is often used with modal particles such as ja and aber as in (8ab), which both refer to preceding context and support the evaluative meaning of the construction reflecting the speaker’s attitude (with aber being stronger than ja). This attitudinal function is also expressed by und zwar in (8c).

13Cases such as (9) are to be interpreted as performance errors when a speaker is looking for the right word:

14This impressively large diversity of contexts in which so was von appears – or, in constructionist terms, the diversity of formal units that can function as slot fillers of the construction – puts so was von in line with a small number of German emphasizer subjuncts such as wirklich, eh, ohnehin, offensichtlich, zweifellos. However, there is a difference in scope between such adverbs and so was von in that offensichtlich in (10a) extends to the whole clause, whereas so was von in (10b) clearly modifies the adjective:

15In the light of cases such as the ones under (7), it seems appropriate to describe so was von (like allenfalls, höchstens) as a modifier of the predicate of a clause, but not of the whole clause.

16The so-was-von construction fits into a group of intensifiers of degree such as etwas, einigermaßen, sehr, ausgesprochen, ungemein, überaus, zutiefst, zu, which can be stressed and also occur before adjectives (Das ist überaus schön), adverbs (Es regnet ausgesprochen oft), numerals (über 20 Jahre) and verbs (Sie hat sich sehr gefreut)2.

17There also exists a slightly more formal variant of the construction (with etwas instead of was), which, however, does not occur nearly as frequently3.

18We also find variation within the so-was-von construction with respect to spelling. As you can see, there are considerable differences with respect to the spelling of sowas as one word in the newspaper corpora of DWDS (Figure 1) and the informal corpus Wikipedia.de 2011 Diskussionen (Figure 2):

Figure 1. Spellings so was von (red line) and sowas von (broken line) in the DWDS4 newspaper corpora between 1990 and 2020

Figure 2. Spellings so was von (red line) and sowas von (broken line) between 2006 and 2011 on the basis of data from Wikipedia.de 2011 Diskussionen5

19Over the last few decades we can observe an increase in one-word spellings of sowas: This increase shows in both corpora analyzed; in the newspaper corpus, the spelling so was is still the more frequent variant, whereas the informal texts of the Wikipedia.de 2011 Diskussionen corpus shows a clear preference for sowas. Needless to say, this one-word spelling is a clear indication of the combination being perceived as a chunk, i.e. an indication of a tendency towards idiomatization. It is reasonable to assume that the one-word spelling will become the established variant.

20The corpus data suggest that the so-was-von construction is typically used in informal language. Given the small size of spoken language corpora of present-day German, it is difficult to arrive at any general conclusions on the basis of 35 hits (11,7 items per million) in the FOLK- and 18 hits in DH-corpus (2,9 ipm)6. Nevertheless, we would argue that it is more than likely that the construction has its origins in the spoken language. One indication of this is the higher number of instances found in spoken corpora such as Wikipedia.de 2011 Diskussionen (see Figure 2). The use of the construction in general tends to be linked strongly to contexts of informal speech, because the evaluative meaning is closely related to situations typical of spoken language (including e-mail, chat)7 involving the interaction between speakers in debates, discussions etc. We are not aware of any corresponding construction with a similarly highly evaluative function established in written German.

21These various aspects of usage on the so-was-von construction show in a number of its properties:

1. Semantic function: As pointed out above, the German _{so-was-von construction} expresses an evaluation or, to be more precise, it serves as an intensifier within an evaluation. What makes the construction rather special is that one might to be tempted to argue that – and this may be due to the fact that it ends in a preposition – it construes the (unmodified) evaluative expression (i.e. egal or zum Kotzen in (11), the _{basic evaluator}) as a Thing in the sense of Langacker (2008). Such a view may, however, be instigated by a desire to relate the partitive uses of von discussed below as contributing to the construction’s meaning.

2. Usage conditions: The construction occurs predominantly in spoken-language situations and is rather informal.

3. Formal realization/variants: The most established form of the construction contains the pronoun was, which (in non-interrogative and non-relative uses) can be regarded as an informal variant of etwas. (This supports our hypothesis that the construction originated in the spoken language.)

4. Formal and semantic specification of slots: It is obvious that the _{type of evaluation} must represent a gradable concept or one that can – in the sense of Goldberg’s (2019: 37) coercion – be interpreted as gradable, as in (7c) and (7d) – which, of course, is true of any concept that can be combined with intensification constructions.

22The construction can be described in terms of three slots (partly following the architecture of the CASA-approach developed for English by Herbst & Hoffmann 2018) – one schematic slot that can be labelled attributee (i.e. the Thing evaluated), the lexically fixed intensifier slot (so was von) and a slot for the basic evaluator, which can have different formal expressions. intensifier + basic evaluator specification together form the attribute that is being assigned to the attributee. A construct such as so was von interessante Ideen can then be seen as a blend of the attributive construction and the so-was-von construction, as shown in Figure 3:

23We can thus suggest the following representation for the so-was-von construction:

Figure 4. The German so-was-von construction8

24Figure 4 illustrates a format for the representation of constructions that has been employed to the description of a number of English constructions (Goldberg & Herbst 2021; Herbst 2020; www.constructicon.de). Not only does it contain information on the formal and semantic properties of the construction (doing justice to a construction being a pairing of form and meaning/function), but it also indicates a number of typical slot fillers, which we refer to as the collo-profile of the construction. Such frequency-based collo-profiles can be regarded as the basis of the generalizations about constructional slots that we expect speakers to arrive at on the basis of previous language experience, just as they will be able to abstract over the contexts in which they have encountered the construction, i.e. the usage conditions9. This kind of description of a construction is intended to cover the various aspects that form part of our constructional knowledge as defined by Goldberg (2019: 7):

… constructions are understood to be emergent clusters of lossy memory traces that are aligned within our high- (hyper!) dimensional conceptual space on the basis of shared form, function, and contextual dimensions.

3. Motivation of the meaning of the so-was-von construction

25As we have seen, the construction status of so was von X can be established both on semantic and formal grounds. The sequence clearly represents a unit of meaning in the sense of Sinclair (2004), although stressed so can occur in more or less the same contexts with a similar meaning.

26If we were to analyze was von as a postmodifier of so, then this would be a very unusual sort of postmodifier, and one which exclusively occurs with so. Outside the construction, the sequence was von does not seem to occur as a phraseological unit, but it can be found in sequences that can clearly be analyzed as pronoun and preposition as in (13):

27While (13b) is a prepositional object of hören with a semantic role reference/topic, the role of the prepositional_von objects of haben in (14) express a partitive meaning10.

28One could thus speculate that this partitive use of von may have been instrumental in the emergence of the so-was-von construction. Another case of semantic overlap could be seen in the expression na so was, which indicates surprise and, like so was von, has the main stress on so and expressions of the type so etwas Tolles etc. It might thus be conceivable to make out a case for the so-was-von construction representing a kind of blend (Fauconnier & Turner 1996; Sweetser 1999; Turner 2018) of so was! expressing surprise, i.e. an extreme degree of unusualness, and was von in terms of a partitive construction.

Figure 5. The so-was-von construction as a blend of so was and hat was von

29However, Figure 5 may lend some cognitive plausibility to the so-was-von construction for speakers of present-day German in the sense that it brings together meaning elements of two other constructions. Whether this is how we can imagine that it came about historically, we will discuss in the next section.

4. The historical dimension of the so-was-von construction

30It has to be said that we have only got sparse historical data. We could not find any evidence for the so-was-von construction in Middle High German. The Mittelhochdeutsche Begriffsdatenbank contains four occurrences of sô was von, where was is 3^rd person past tense. Furthermore, Middle High German has a relative pronoun (with definite personal reference) of the type swaz (sô waz), which takes a partitive genitive as an attribute (Paul et al. 2007: § M 50, § S 79).

31In the Kern- und Erweiterungskorpus des deutschen Textarchivs (DTA, via DWDS), which contains material from 1473 to 1969, the earliest occurrences of the sequence so was von stem from the 18^th century, which may be due to the meagre text basis of the corpus:

32None of these corresponds to the so-was-von construction: in (15a), so is a temporal-conditional conjunction (‘if, when’), and was von Wildpräth means ‘something of deer’, with the Middle High German partitive genitive having been replaced by a PP_von after a pronominal use of was. The so was in (15b) (‘something like this’) is the accusative object and von meinem Bruder Anselmo a prepositional object of the verb hören. (15c) again shows a partitive PP_von, but without the evaluative meaning of so-was-von construction in present-day German, although (15c) can be seen as pointing in the direction of an evaluative marker.

33It is not before the late 18^th century that we find examples to which today’s construction can be traced back. Such examples typically occur in dialogues within narrative texts. Since the 1770s, one can observe a trend of making use of language that shows characteristics of “normal” spoken language so that we can assume an increase of informal style in these texts. While these are quite obviously examples of “language written to be spoken”11, we can only speculate about a possible use of the so-was-von construction in the spoken German of earlier periods.

34The examples under (16) appear in ambiguous contexts, which allow a partitive as well as an evaluative interpretation (in the sense of expressing surprise). In (16ab), this is underscored by the fact that they are used in exclamations, as indicated by the exclamation mark. In (16a), stress on von would allow a partitive reading, but with the stress on so, one could imagine an evaluative interpretation. (16b) is an example of a noun used after so was von, but, interestingly, although adjectives are probably the most frequent slot fillers in the construction in present-day German, we could not find any uses of adjectives in that position prior to (16b), despite the fact that adjectives do not usually occur after the preposition von. (16c) comes close to today’s usage, although a partitive reading cannot be totally excluded either.

35In the texts from the first half of the twentieth century, we found only one use with the evaluative meaning after the particle na as in (17a), which is typical of the spoken language:

36(17ab) show that von NP still has to be interpreted as a prepositional phrase in which the preposition determines the dative case of the noun. A few first uses with adjectives as in (17c) can be found since the 1920s, but it is only from the 1970 and especially 1980s onwards that so was von Adj gains in frequency and comes to dominate uses of the construction in 21^st century corpora of German12. There is a clear difference with respect to the occurrences of so was/sowas von in the pre-1970 and post-2010 data, as shown in Figure 6:

Figure 6. Types of phrases after so was / sowas von based on the analysis of data before 1970 from DWDS and IDS (black lines; n=116) and post 2010 from DWDS and IDS (green lines; n=4828)13

37The most important development to be observed here is a very clear increase in the share of uses of so was/sowas von with adjective phrases (given the different corpus sizes, we can only indicate percentages for the respective periods although data before 1970 are rare). Before 1970, we see a very low degree of idiomatization because we can observe a comparably high number of uses in which von governs a dative NP, whereas on the basis of the post-2010 data we can make a clear case for an idiomatic so-was-von construction for present-day German.

38Interestingly, this coincides with a rise of na so was in the corpora during the 20^th century, which suggests that this expression may have existed in spoken German before that. Whether this informal exclamative use has played a part in the coming about of the so-was-von construction, as speculated in section 3, is difficult to say14.

39Thus it is plausible to assume that during the 20^th century so was + PP_von gave rise to a construction of the type so was von + AdjP, NP, AdvP, PP, V. The idiomatization of so was von into an intensifier of an evaluation – and presumably also its predominant use in informal spoken German – may have been instrumental in the emerging flexible realizations of the slot expressing the evaluation as such.

40It can’t be emphasized too strongly that the data situation for early occurrences of the construction or possible precursors is very poor. Up to the year 1900, there are only sporadic instances that could be regarded as precursors; even up to 1970 the DWDS only contains 116 cases in which so was von can be analyzed in terms of the evaluative meaning associated with the construction. This can be taken to mean that the construction came to be used only gradually in the written language and that it has only developed into a full idiomatic construction over the last fifty years or so. However, a considerably larger number of data for the time between Early High German and the 1970s (including spoken German) would be required in order to outline the developmental path of the construction in an empirically sound way. The only line we can take in the light of this rather deplorable situation is to suggest a possible path of the emergence of the German so-was-von construction that arises from assumptions made on the basis of the sparse data available up to the second half of the 20^th century.

41We can thus provide the following sketch of the emergence of the so-was-von construction:

Figure 7. Possible emergence of the so-was-von construction in the history of German

5. Concluding remarks

42There thus seems to be considerable evidence for the claim that we are justified to postulate a so-was-von construction for present-day German:

1. The construction can be characterized as expressing an extreme degree of an evaluation expressed by a large variety of different slot fillers, which puts it in a line with other emphasizers in the German language. This means that the word von occurs in syntactic contexts in which it does not appear in its established uses as a preposition, e. g. before adjective or prepositional phrases. Furthermore, while von in its prepositional uses case-marks the following NP as a dative, this is not true of so was von – so in this case, the emergence of a new construction results in an increase of syntactic irregularity15.

2. The construction is a relatively new construction which seems to have emerged as recently as the second half of the 20^th century. This may explain why it is predominantly used in the informal spoken language and in related genres such as blogs.

3. There is no obvious constructional path that would provide a satisfactory story of how the construction has come about, although the diachronic development of partitive uses of von and was von can be seen as preparing the ground for the emergence of the so-was-von construction.

4. The construction seems to be unique to German in the sense that no constructions that could be considered “parallel in form” exist in Danish, Dutch, English, Icelandic, Norwegian or Swedish.

5. Given the frequency of so was von in present-day German, it is a construction that – like the English the-two-of-them construction (Herbst 2016) – definitely ought to be included in teaching materials in the spirit of Pedagogical Construction Grammar (De Knop & Gilquin 2016) if we want students to learn how to express themselves idiomatically in the target language.

43We thus hope to have shown that a constructionist account of phraseologisms is able to overcome the weakness associated with traditional accounts of phraseology of which De Knop & Mollica (2016: 54) say that it “implies that the meaning of phraseologisms is not motivated or that their motivation cannot be determined by speakers from what they already know”.