On the one hand, they deal with different domains: pragmatics vs. grammar, so they don't strictly need to be compatible. It's hard to imagine any (reasonable) theory of grammar being completely incompatible with Relevance-theoretic pragmatics.

On the other hand, Construction Grammar, and Cognitive Linguistics in general, does not treat the Semantics/Pragmatics distinction as strictly as other grammatical theories, so that usage and context may either directly or indirectly affect the grammar. In that sense, Relevance may determine which constructions are used and even how they develop. This is not an area I've specifically read much about, but certainly seems like a good one for research, to determine the extent and nuances of the relationship. You might even ask if the constructions found in a language can be substantially explained by Relevance.

At the same time, a challenge is that neither of these is a theory per se in the most specific sense: they are both ideas that can be incorporated into specific theories. It seems extremely likely that "relevance" has some bearing on pragmatic interaction, whether or not formulated in precisely the way current publications say. Constructions are a little bit more controversial (some extreme Generative theories might attempt to deny that they exist entirely, although even then I think they would need to admit that idioms exist, which are the most extreme kind of construction, and then the question just becomes where to draw the line, not whether constructions exist). So these are really just approaches to analysis, and there's a lot of variation out there in specific proposals, with no consensus yet.

Can you clarify what you mean by not a theory per se with an example of a specific theory?

Good question! --

If you could provide an example of something accepted as a full fledged theory that would be helpful.

Unfortunately most so-called "theories" in Linguistics (it's a new science, after all) really don't have the rigor or specificity of something like the Theories we know from physics (e.g., the Special Theory of Relativity). Partly it's because they're not fully worked out, but mostly because they're still in flux, from on linguist to another, and even from one paper to another by the same author. If you look at Chomsky's work over the past ~70 years, he has substantially changed directions multiple times. Still we talk about "Generative Grammar" as if it is one thing-- and it is, in that it's an approach, but it's not a specific theory. Perhaps one of the most notorious non-theories is "Universal Grammar", which is Chomsky's proposal that all languages share some inherent, genetically-based grammatical core, but you really can't look up with UG is, because it has simply never been proposed in full, and it varies almost every time it's discussed, so you can't call it a theory (but many people still do). Instead, you can find many ideas about what might be part of UG, and many arguments that UG must exist (just by definition-- it's whatever is shared by all languages, but also that there must be some common core to human languages for children to learn to speak similarly). But no real "UG theory" has been proposed, because that would require it to be fully worked out. And most serious scientific theories also have reached some level of consensus-- at this point linguists agree about almost nothing.

One thing that is close to a theory is any long book (whether a textbook, or a more technical book) proposing a specific approach. You could call that the Author-2018 Theory, or whatever. But it will probably change next year, so it's not really a Theory in the common scientific sense.

What we have instead in Linguistics is many ideas that are proposed as good parts of (eventual) theories. For example, Construction Grammar centers around the idea that a good grammatical theory of human language has Constructions (in a technical sense).

Actually, one of the more worked out theories for a specific domain in Linguistics probably is Relativity Theory. It's still not quite to the level of various Theories in other fields, but it's been worked out to a relatively explicit level and is not as controversial as many other "theories" out there.

One way to tell a "Theory" from a non-theory is that a "Theory" can be falsified by showing it is incompatible with data. So if you find that Special Relativity does not apply to certain stars in the galaxy, then that model of physics is wrong. Theories are therefore very fragile (and probably all wrong, at a sufficient level of detail). But at the same time, a good theory that stands the test of time is strong, in another sense. What makes most "Linguistics theories" different from this is that they're falsified all the time, sometimes in the same paper where they are proposed. They're really just partial theories, or "flexible theories" that don't really hold themselves accountable to falsifiability in a strict sense. They're ideas about how to get to a theory eventually, but when some part of the proposal is falsified, then the theory is just adjusted a bit to make it fit better, and so science continues, toward Theories, but not there yet. You could, if you wish, consider every single proposal to be a Theory, but that's pointless because in 99% of the cases they're transparently falsifiable by looking at more data-- and they're not even really "Theories" to begin with because they're not complete (I don't mean as "a theory of everything" but even just within whatever domain they're trying to cover).

So instead in Linguistics we have many ideas (some of them called "theories", but also by other names), and these are good suggestions for components of theories. Chomsky has spent the past 70 years or so searching for a Theory of grammar but hasn't really gotten there yet. His latest enterprise (for the past 20+ years) is the Minimalist Program, which very explicitly is not a theory at all, but an approach to designing a theory: specifically, he suggests the best theory is the simplest one. So Minimalists seek to eliminate unnecessary components from earlier proposals, in order to figure out which components are essential.

As for Construction Grammar, it's far from a Theory just because there really isn't "one way" to do Construction Grammar-- far from it: there are many shared ideas across different works, but basically everyone working in it has slightly different ideas, and therefore is imagining a slightly different theory.

At the same time, all of this is still theoretical because it is being proposed in the context of developing a theory. But there's so much variation and shifting, we're just not there yet.

Out of interest, would you classify the recent Oxford Handbook on Construction Grammar as such a text book as you mentioned above?
https://www.amazon.co.uk/Oxford-Handbook-Construction-Grammar-Handbooks/dp/0195396685/ref=sr_1_1?ie=UTF8&qid=1543752827&sr=8-1&keywords=oxford+handbook+of+construction+grammar

Really helpful again. Thanks Daniel
Just out of curiosity, and I guess this could be a separate post in its own right, but do you think that experimental pragmatics will help things along and help decide between the debates? I recognise that there will still be differences, just curios as to whether you think this has potential to move us towards developing theories akin to those in physics that you mentioned earlier.

The question "what is a 'theory'?" is one of those unresolved issues in epistemology. There is no theory of "theory". The word "theory" is so vague, taken out of context, that all you can say is that it's a mental object of some kind that relates epistemological objects to what they refer to. When you specify "theory in this context", you can get more specific and the word becomes less useless. Because linguists think they are scientists (not clear to me if that is really true but let's say it is), then you start reading the literature on what scientific theories are. Of course, you have to know what "science" is, in particular what kinds of theories are not scientific theories.

Typically, physical scientists approach the problem via exemplification rather than definition, and the examples are drawn from the history of physics. A typical bottom-line criterion is that statements in a theory are either true or false, and the truthiness or falsity of the theory is determined by observation. Moreover, the statements of a theory are sufficiently specific that they unambiguously refer to a particular state of existence – for example, "there will be an electron at this position in space-time".

However, people also like to add other things to the theory of theory. One is the idea that a theory should be well-supported with many strands of evidence; or, it should be the most parsimonious account of a phenomenon. Usually, people add a requirement of explanatoriness, that something is a theory only if it explains an aspect of nature. The problem with those add-ons is that they are subjective and not subject to empirical disconfirmation and quantification. In linguistics, the "explanatory" card is played quite often, with one side asserting they they find theory X to be more explanatory, and rejecting theory Y because it doesn't "explain". The "many strands of evidence" criterion also fails to say how many is required to be "many". The many-strands add-on moves "theory" from being a logical object to being a social object (which is indeed one theory of theory).

One simple characterization of "theory" that I have encountered is that it is a collection of assumptions about a field, which, given a model of an experiment in the field, predicts the outcome of that experiment. Although I disagree with this, it's not completely horrible. It has the disadvantage that it depends on the sociological concept "field", but it's not worse than other accounts of "theory" which still implicitly embrace the field-relative nature of theory. In fact, there are no scientific theories other than those covering fundamental forces and entities in physics: under strict versions of what a theory is, "theory" would be generally useless for the conduct of science. Rather than reject the concept "theory" itself, I think we have to embrace the idea that there are theories in linguistics, that linguistic theories are not as powerful as those of particle physics, but they are still theories.

Rather than claiming that Chomsky's theory isn't even a theory, I would (or might) say that it makes fewer claims than some other theory. That, b.t.w., is about the current theory, not the full 70-year cycle of theorizing associated with the gentleman. His current theory is vastly superior to his early-80's theory in deleting claims. The ontology of syntax is greatly reduced relative to the old days, although perhaps not as reduced as it was in the really old days.

I have to take issue with Daniel's claim that Generative Grammar isn't a theory, it is a collection of approaches. It is a theory, specifically that there exists a mental system of computations which constitute human linguistic competence (the ability to generate and evaluate utterances in a language), which can be symbolically modeled. HPSG, Aspects theory, LFG, GB and so on are instances of GG. I don't know if Construction Grammar is, but I suspect based on the Berkeley-HPSG connection that it is.











So people will say that a theory

Overall good points panini, and I agree with most of it. A couple comments:

Rather than claiming that Chomsky's theory isn't even a theory, I would (or might) say that it makes fewer claims than some other theory. That, b.t.w., is about the current theory, not the full 70-year cycle of theorizing associated with the gentleman. His current theory is vastly superior to his early-80's theory in deleting claims. The ontology of syntax is greatly reduced relative to the old days, although perhaps not as reduced as it was in the really old days.

I don't disagree that Chomsky has a theory in mind when he talks about language, but it hasn't been published anywhere. He's tinkering with ideas that are components of a theory, but at best I'd call it a "theory in progress".

I have to take issue with Daniel's claim that Generative Grammar isn't a theory, it is a collection of approaches. ... HPSG, Aspects theory, LFG, GB and so on are instances of GG.

Those are all different theories. I don't see how you can think of "Generative Grammar" as a theory. It's an approach, with some theoretical ideas. There's way too much internal disagreement to consider it to be a unified theoretical whole.

...It is a theory, specifically that there exists a mental system of computations which constitute human linguistic competence (the ability to generate and evaluate utterances in a language), which can be symbolically modeled...

I'm not sure what "it" it is you're speaking of! That's the problem. And what they have in common, if anything, is a hypothesis rather than a theory.
To look at this another way, "falling" isn't a theory of physics, and arguably not even "gravity", but some specific theory of how gravity works would be. Broad approaches like "Generative Grammar" fall in one of the first two categories, certainly not the third.

I don't know if Construction Grammar is, but I suspect based on the Berkeley-HPSG connection that it is.

CxG is explicitly not "generative" according to some, and certainly not "Generative" in the proper-noun sense. There are so many flavors of it that some may considered generative in some ways, but that's not the purpose of the approach. It's a functional approach, rather than a formal one, and therefore not about creating grammars that generate sentences. Fundamentally different things. You might be able to turn it into one, but it's not meant to be, at least not by the majority of linguists who use it.

--
Of course there's little point in us arguing about terminology here, so it may not be worth me debating these points-- overall you're right to say what you've written, because we don't know exactly what "theories" are

For example, you reject my claim that HPSG, GB, LFG, RG are kinds of Generative Grammar, proffering that each is a theory of its own.

No-- I agree those are GG, based on the same idea, but consider them to be distinct theories. I can't consider them to be the same theory, or even flavors of the same theory, because they have fundamentally different assumptions and mechanics, to the point where there is almost no prediction left. If it's extremely "fixable" (flexible) in that way, I don't see how it can be called a "Theory".

That suggests that you see a theory as being an isolated object that isn't logically connected to other such objects.

Correct: I understand Theories (capital T, with whatever qualifier is added) to be large, relatively well accepted proposals that are something of a coherent whole.

I, however, would maintain that there can be competing theories within a theory. Competing theories arise because, in the context about a certain level of agreement and shared principles, there are still details of implementation – that is, every theory is incomplete (except, I hear, optics), and therefore room for further investigation and disagreement. I refer you to Quantum Mechanics and the numerous theories within that theory – usually known as "interpretations".

"Interpretations" or "theoretical variants" would be fine, but, no, I wouldn't call them theories.
But again, it doesn't matter much what we personally call them, since I think we agree about the substance (e.g., different versions of GG are related and have some shared foundation).