Trying to apply key concepts from CxG to modality in English, I came across research by Boogaart (2009). He particularly looks at the modal verb 'kunnen' in Dutch and raises an interesting issue:
On the basis of cross-linguistic evidence, he finds that 'kunnen' is used in Dutch with an epistemic interpretation in questions only (in an affirmative context, the verb 'zullen' has to be used instead). Although he does not use the term, it therefore seems that there is a particular collostructional preference between the epistemic reading of 'kunnen' and the Dutch interrogative construction. The problem is, as he nicely points out, that this preference seems to be true of the epistemic reading only and not the other senses that 'kunnen' typically conveys (i.e. dynamic / deontic possibility).
The challenge, therefore, is to understand how this type of knowledge can be captured by the construction 'kunnen'. If we assume that we have one form 'kunnen' which gives access to a polysemy network (dynamic/deontic/epistemic possibility), then it is not clear where the collocational preferences associated with the epistemic reading only are stored. It cannot be part of its formal properties (otherwise, this preference should apply to all the different senses identified). But can it be considered a functional property? (Boogaart assumes it cannot). Or is it yet stored elsewhere?
For this reason, Boogaart argues that either we need to postulate a distinct, 'epistemic kunnen' construction or we need to argue for a monosemous account of the verb. (He favors the latter approach.)
I am not sure exactly what the perspective in CxG might be. Any thoughts on how CxG would analyse the collostructional preferences of the epistemic reading of 'kunnen' ?
Thanks !
Hi Benoît,
That's an interesting question. Here's my two cents. CxG allows one to integrate collostructional preferences in a hierarchical network of constructions.
At the highest level of generality, I agree with Boogaart that there's probably just one 'kunnen' construction in Dutch, whose abstract meaning relates to possibility and whose formal specifications have to ensure, minimally, that this word is used as a verb. This general construction can be linked to a 'kunnen' construction at a lower level of generality, whose semantic value is something like 'epistemic possibility' and whose form is interrogative. Stating that the form is interrogative can be done by making this specific construction a bit 'larger' than just consisting of 'kunnen': one might, for instance, include a position for a subject NP appearing after 'kunnen', so that there is S-Aux inversion, and one might include the requirement that there is some sort of interrogative intonation pattern.
As an interrogative sentence is only a typical, rather than necessary, environment for 'kunnen' to be used epistemically, we also need at least one other low-level pattern, where 'kunnen' is semantically still specified as being of the epistemic possibility kind but where there is no formal specification (on top of what is inherited anyhow by the maximally general 'kunnen' construction). In the constructional network, we can then give different 'weightings' to these two lower-level constructions. I've seen construction grammarians play with the thickness of the frames around constructions, with thicker frames marking a higher level of entrenchment, using a convention that goes back to Langacker's network representation of symbolic units.
In fact, there are many complications, but I believe these can still be dealt with. One complication is that 'kunnen' has several inflected forms ('kan', 'kunnen', 'kon', 'konden' and also 'heeft ... gekund', 'had ... gekund', etc.), and so I wonder whether or not one needs different epistemic constructions for all these forms individually or whether the epistemic use with each of them can be inherited from the more general knowledge that 'kunnen' has these inflectional forms. In all likelihood, there are again different degrees of preference for different inflectional forms, which then would necessitate a similar solution as suggested above.
All best,
Bert
Hi Bert,
Thanks a lot for the answer!
It actually confirms a couple of ideas I had in mind. The only thing I might ask, I think, is this: is it correct to assume that the type of analysis you present is still compatible with a polysemous view of 'kunnen', regardless of there probably being a lower level/more specific "epistemic kunnen" construction? (Or at least, should I say, that it provides more than evidence that 'kunnen' is necessarily monosemous)
Best, Benoît
Hi again Benoît,
Yes, I believe the analysis is fully compatible with a polysemous approach. If the maximally general construction has a very low level of entrenchment, then in effect, what you get is constructional polysemy. The lower-level constructions are then cognitively more accessible than the single abstract construction that binds them all. Yet, the fact that they ARE still bound by such a general construction, however faint its presence in the speaker's mind, means that there is polysemy, not homonymy.
All best,
Bert
Right, I see. Thank you for the answers, they'll be very helpful!
Best, Benoît
Hello,. I think I agree with Bert. But I come from a somewhat different perspective, formalized CxG. My particular stripe is Sign-based Construction Grammar. Unlike some of the other constructional approaches, SBCG doesn't view words or lexemes as constructions. Constructions are licensors of sign combinations rather than signs themselves. But because this is a highly lexicalist framework, lexemes and lexical classes are at the heart of the grammar, and constrain their 'collocational preferences' (we call these preferences the word's or lexeme's 'combinatoric potential'). That is, the words and lexemes tell us in their feature values what constructions they can unify with. For example, auxiliary verbs in English have features that enable them to unify with the Inversion, Negative and Ellipsis cxns in English, but non-auxiiiaries lack these features. I'm trying to find an analogy to the Boogaart case. Here's a possible one: in English an epistemic modal must have a stative complement while a deontic verb does not have this aspectual selectivity (e.g., 'She must leave' can only have a deontic reading, but 'She must live here' has both readings). To capture this we could say that modals are polysemous, but all that means is that there are a couple of different lexical entries that share a FORM value. Each has distinct frame semantic and valence properties. We don't need to posit a special 'must' construction to capture this.
Hello--
It seems the problem raised in the question is only a problem because we assume that polysemy is a given. But is it? Where does kunnen get its multiple meanings? The answer is it means different things in different constructions. So, the notion of polysemy is not at all theory-neutral and should not be taken for granted. Polysemy makes sense in a words-and-rules framework where sentence meaning is mainly attributed to the meaning of the verb as it is interpreted in THAT sentence but the meaning of the construction in which it is used is ignored. Within CxG, the notion of polysemy is redundant if not meaningless because construction meaning (e.g. interrogative construction) is responsible for sentence meaning and by constructional coercion kunnen acquires the epistemic interpretation.
This kind of problem is not new and has already been addressed in CxG. Recall Goldberg's (1995) example of sneeze. In words-and-rules grammar one would have to argue that sneeze as in "He sneezed the napkin off the table" is a transitive or even causative verb, thus creating a quasi-polysemy situation. But within CxG, it is the CAUSED MOTION construction that is responsible for the interpretation of the sentence.
Hello,
Thank you Laura for your answer. I must admit I hadn't looked into what SBCG could offer before posting this question, but it does provide an interesting account for how to explain this particular 'issue'. (I mostly know Cognitive Construction Grammar, but I definitely need to read more about SBCG).
Thank you also Zhuo. Of course I agree with you that positing polysemy might be part of the issue. Like Boogaart, then, you would favor a monosemous approach. Yet at the same time, I am not really convinced that 'kunnen' gets its epistemic interpretation via a coercion process either. (What construction exactly would the epistemic meaning of 'kunnen' be coerced from? It can't be the "interrogative construction" mentioned above, so one needs to posit at least a third cx.) In a section of my thesis, I focus on the notion of coercion and also address the question of whether there exist such abstract constructions that 'coerce' a deontic or epistemic interpretation on modals (as suggested both by Wärnsby 2002 and Boogaart 2009), and the answer seems to be no. (Wärnsby 2002 provides a solid argumentation against this view) From that perspective, I therefore find more attractive Laura's perspective in terms of different entries that share a form value (whereby the semantic difference - if I understood correctly - still occurs at the lexical level somehow).
Again, thanks for both your answers, it's given me much food for thought!
Hi Benoît, I agree the selection of a complement by a modal is no more a constructional concern than the pairing of any head and complement. Modals are polysemous and this is a lexical rather than constructional fact. The semantic selectivity of a modal verb is a fact about the particular modal verb and not about the construction that pairs a modal and its complement. If I may recommend something I've written on aspectual coercion, I think it could help here (ref below). Evidentiary epistemic modals have an aspectual constraint: the complement must be stative (which stands to reason, as such sentences describe the evidentiary basis on which the speaker is concluding that a given state holds at speech time). Since every type selector is also a type shifter, an evidentiary epistemic modal can shift the reading of a dynamic complement to a stative (habitual) reading. For example, 'You must polish the floors frequently' has both a deontic reading and an epistemic one in which the speaker is inferring a state that holds between episodes of floor polishing. But of course again this is merely evidence that semantic selectors like the indefinite article are also potential semantic type shifters, not evidence that there is a special deontic or epistemic construction.
Michaelis, Laura A. 2011. Stative by Construction. Linguistics 49: 1359-1400. http://spot.colorado.edu/~michaeli/documents/Stative_by_Construction-corr.pdf
Yes, I would assume that we have to talk about modal constructions. They mark facticity (meaning) and have a recognizable form: V + VINF. In addition, the class 'modal verbs' is not closed. In German, for example, we could not adequately analyze "Ich brauche nicht essen" ('I do not need to eat' in the meaning 'I do not have to eat') without claiming a modal construction of obligation.
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking