Thomas Hoffmann's article proposes a cognitively viable theoretical framework for explaining how constructions can emerge in the history of a language. The case which Hoffmann discusses is the coming into being of the comparative correlative construction of the type The smaller a car is, the easier it is to park in late Old English and early Middle English. While car sizes and parking spaces were perhaps not a matter for discussion among speakers of that time, it was for them as important as it is for us to be able to say that if any two things differ in one respect, then they also differ in another (for details on the construction's semantics, see Beck Reference Beck1997; Cappelle Reference Cappelle2011). In this particular comparative correlative construction, the first part (C1, for ‘clause 1’) is interpreted as standing in a sort of protasis relation to the apodosis-like second part (C2, for ‘clause 2’). That is, this C1C2 construction reflects, in an iconically appropriate way, the order of a hypothetical statement followed by its consequence.
Having some means of talking about correlations was no doubt also an indispensable communicative need for speakers of earlier Old English, but as Hoffmann reports from the linguistic literature, these speakers couldn't avail themselves yet of the structure just mentioned. Indeed, this C1C2 comparative correlative construction arose later in English than what could only anachronistically be called the ‘inverted’ comparative correlative construction, of the type A car is (the) easier to park the smaller it is, a pattern still available today. In this C2C1 structure, the first part is clearly a main clause and the comparative phrase is not fronted but finds itself in situ.
This less iconic ‘consequence–before–conditional’ C2C1 correlative was only one of the correlative patterns that existed in Old English. Hoffmann convincingly shows how these Old English patterns together formed a close-knit correlative constructional network – a network with one node crucially missing. On the one hand, there was a C2C1 proportional comparative correlative pattern of the type . . . swa . . . swa . . . and, as just pointed out, also a C2C1 comparative correlative pattern of the type . . . the . . . the . . . (or actually, . . . ðy . . . ðy . . .). On the other hand, there was also a C1C2 proportional comparative correlative pattern of the type swa . . . swa . . . but, for all the formal and semantic links among these patterns all the more surprisingly, no C1C2 comparative correlative of the type the . . . the . . . (or ðy . . . ðy . . ., to be again historically accurate). This asymmetry in the set of constructions, it is now claimed by Hoffmann, led to the genesis of the up-to-then missing C1C2 the–the (ðy–ðy) structure.
Hoffmann breathes new life into European Structuralism by claiming that in the emergence or change of constructions there is an important role played by the constructional network as such – that is, by its very structure at a given time. Saussure and Meillet famously saw language as a system in which everything holds together and in which all parts have to be considered in terms of their synchronic interconnectedness. Hoffmann embraces this Structuralist idea of ‘language as a system’ and reinterprets it as ‘linguistic knowledge as a mental network of stored constructions’, thus accommodating it within the cognitively plausible contemporary theory of Construction Grammar. But the big question is: how can such a system/network change?
Busting the ghost of the Saussurian paradox
I am hardly the first to point out that there is something fiendishly paradoxical in conceiving of language as a well-oiled machine with all its parts in place like cogs and wheels and in acknowledging that there is undeniably such a thing as language change. For the greater part of the twentieth century (for some discussion, see, e.g., Weinreich, Labov & Herzog Reference Weinreich, Labov, Herzog, Lehmann and Malkiel1968; Detges & Waltereit Reference Detges and Waltereit2008; McColl Millar & Trask Reference McColl Millar and Trask2015: 250ff.), linguists puzzled over the question, now known as the Saussurian paradox, how a language can keep on being communicatively efficient as a structured system of speech sounds, lexical items, grammatical forms and larger constructions, when it was also obvious that at any given time this system is in the middle of a great many ongoing changes to its interrelated parts. Didn't these constant changes inevitably throw the system into disorder, rendering it unsuitable for communication?
Meanwhile it is deemed that this paradox can be resolved by stressing, as notably Labov first did in the 1960s, that language is not a completely homogeneous whole and that speech preferences of some speakers may be copied by others, for instance to mark group identity. Far from being peripheral and anecdotal, variation is in-built in language and is now recognized to drive language change. Moreover, some of the variation among individual speakers may be influenced by their differing familiarity with or exposure to another language, in a situation of bilingualism or language contact for example. All of this is well known. However, the emergence of the C1C2 the–the construction in English discussed by Hoffmann does not involve variation. Rather than gaining frequency – which, as Hoffmann proposes, happened later, thanks to its relative processing ease – the C1C2 the–the construction was introduced into the system, ex nihilo (or so it would seem – more on which shortly). And what's more, this change to the system didn't come from any outside influence but was driven by the system itself. Hoffmann does briefly mention how the Latin pattern . . . tanto . . . quanto . . . is reflected, only very occasionally, in a special C2C1 correlative pattern in Middle English, but this is irrelevant to the emergence of the C1C2 the–the correlative under discussion here.
Now, if there had been an initial stage of competition between, say, the C1C2 the–the and the C2C1 the–the pattern, this could have served as a kind of linguistic Petri dish in which we could then have observed constructional growth in the form of a shift in usage frequency. But the absence of any C1C2 the–the construction in the early Old English network poignantly brings back to life the ghost of the Saussurian paradox. How can the system itself be held responsible for its own change? Or, in modern Construction Grammar parlance, how can grammar, as an ordered network of constructions, bring about a change to itself by adding something where there was nothing?
Hoffmann doesn't problematize this act of creation as much as I am doing here, but he does offer a solution. Echoing McMahon's (Reference McMahon1994) discussion of structural approaches to sound change, he argues that paradigmatic gaps in the system – or missing nodes in an ordered constructional network – are disfavoured. Adopting recent insights and notions from grammaticalization studies and diachronic Construction Grammar, in particular analogical thinking (e.g. Fischer Reference Fischer2007; De Smet Reference De Smet2009) and ‘analogization’ (Traugott & Trousdale Reference Traugott and Trousdale2013), he furthermore suggests that speakers first identify the gap in the network and subsequently form a construction which ‘restores’ the balance. (I use scare quotes, because there was no balance initially.)
The system/network thus strives to a state of maximal symmetry, in which it is filled out evenly over the constructional space without displaying any paradigmatic gaps. And it is the relations among the constructions already present at a previous state of the system/network that co-determine what the construction filling a gap should look like. In other words, the new C1C2 the–the correlative didn't appear out of the blue: the other Old English correlative constructions, mentioned above, served as partial models.
Make a change
Obviously, the system doesn't change by itself; it needs speakers to make it change. In Hoffmann's Cognitive Structuralism, speakers aren't introduced as a deus ex machina. Rather, they are part and parcel of the theory, insofar as one cannot meaningfully talk about a cognitive network of constructions without making at least implicit reference to someone who has linguistic cognition, that is, a speaker. The same is not true for classical Structuralism, where the language system could be – and in fact was – seen as detached from those who make use of it; recall the langue/parole distinction.
In his article, Hoffmann suggests that analogical thinking and analogization happen unconsciously in the mind of speakers. This may well be the case for the sort of constructional genesis at hand. However, I would like to propose that this may be different for other kinds of constructions. I have argued elsewhere against the constructionist tenet that phrasal/clausal syntactic structures and word-level lexical constructions such as derivation and compounding are different in size but, since they are all essentially form–function pairings, not in kind. My dissenting view may now find some support from analogical formation within Hoffmann's constructionist framework of Cognitive Structuralism. The way speakers fill a perceived gap in the network of lexical constructions might be much more conscious and purposeful than how they create larger syntactic constructions. Such a difference would highlight the distinctness of lexical and syntactic constructions.
Consider the following extract, taken from the novel The Bone Clocks by British author David Mitchell, more specifically from a section set in post-apocalyptic Ireland in the year 2043. I apologize for the coarse language towards the end.
Dónal Boyce calls after us, ‘I'll be seeing you, Lol.’
He's a leery, muscly, horny threat. We leave the village, passing the Slán abhaile sign and the old eighty k.p.h. speed-limit sign. ‘I didn't like the way Dónal Boyce was looking at me, Gran,’ says Lorelei.
‘Good,’ I tell her. ‘I didn't, either.’
‘Me neither,’ says Rafiq. ‘Dónal Boyce is a jizbag.’
I open my mouth to say, ‘Language,’ but don't.
(David Mitchell, 2014, The Bone Clocks, London: Sceptre, p. 570)
The word jizbag is not to be found (yet) in the Oxford English Dictionary (OED), and though jizzbag, with two z’s, is mentioned in the Urban Dictionary (www.urbandictionary.com) and has two occurrences in the 1.9 billion Corpus of Global Web-Based English (GloWbE; http://corpus.byu.edu/glowbe/), it is a fair assumption that Mitchell coined jizbag, to refer to a contemptible person, as a new word, one that in the foreseeable future may be common usage among youngsters. This is a not altogether implausible prediction, given that today we already have the words scumbag, douchebag, sleazebag, dirtbag and shitbag, among others. Scumbag originated as an offensive slang word for condom, with scum referring to semen. Mitchell may have seized on the opportunity to complete the lexical network by combining the head word bag with another slang word for semen. (Jiz(z) is attested in the OED, s.v. jism.) So, I believe Mitchell also used analogical thinking and analogization, but in a much more deliberate way than in the case discussed by Hoffmann.
To be sure, swearwords hold a special status among lexical items. Not all novel compounds and derivations are the result of conscious coining; some of them come into the world quite imperceptibly and without the creator being aware of adding anything to the stock of words. Still, one of the merits of Thomas Hoffmann's Cognitive Structuralist account of analogy-based language change is that it unintentionally raises the question whether clausal constructions are ever created consciously.
