The LINGTYP e-mail list is an old-fashioned way of communicating – it has existed for 23 years in its present format, as an unmoderated e-mail list for the typology community (made possibly by the LINGUIST List), and one might think that it should have been replaced by something more modern. But it continues to be very useful, I find, and again this month (July 2021), there were some really interesting discussions, on the meaning of the term “personal pronoun”, as well as on the nature of grammatical concepts.
It seems to me that we need to have these discussions, but they don’t easily happen at conferences, where the main purpose is to tell each other about our recent work. And they don’t happen in our research papers either, because there we have to convince the reader that we solved the main problems.
In this post, I’d like to address a point made by Bill Croft in a thread about the term personal pronoun that was started by Ian Joo. After several specific definitions had been proposed, Croft said:
“All grammatical concepts are theory-laden, and are interconnected with other theory-laden concepts, including semantic, pragmatic, and discourse concepts. Since we have different theories of all of these things, we won’t agree with a definition of “pronoun” that assumes a theory that we don’t accept.”
Christian Lehmann seconded Croft’s point: “terms are not defined in isolation; definitions have their place in a theory. The concepts defined have their place in a taxonomy and/or meronomy.
Lehmann and Croft have been role models for me throughout my career, so I want to take their statements very seriously, but here I take a different view. I use this blogpost to articulate it, because it allows me more space and I don’t want to fill everyone’s mailbox with too many unsolicited messages.
The general perspective that I take here (following other work such as my 2021 paper about standardization of grammatical terminology) is that discipline-wide definitions of common grammatical terms are possible, even without general agreement on the basic issues of theoretical linguistics. Linguists agree much more on the basic concepts and issues than they agree on the terms in which to talk about the phenomena.
I do recognize that there is a widespread feeling in the discipline that terminological issues are not important, or that in any event we can have good terms only if we have a good understanding of the phenomena: “Terminology depends on understanding.” Bill Croft says:
“All of these issues depend on one’s theoretical stance regarding deixis and anaphora, accessibility, definiteness, meaning and use, reference and modification, nouns and adverbs, and what, if anything, these concepts have to do with pronouns.”
It seems to me that this is currently the majority view among grammarians, and it is particularly widespread in generative grammar, where nobody discusses terminology, as far as I am aware. (This makes sense, because generative grammarians do not make the distinction between comparative concepts and descriptive categories, and their goal is to analyze languages in terms of the innate basic building blocks of languages.)
Here I would like to make the case for the opposite: Understanding (also) depends on shared terminology. I do not, of course, deny the general point that we need a shared conceptual basis to understand the meanings of the words we use, so at the highest level, it is certainly true that we cannot agree on terms without agreeing on SOME theoretical ideas. However, in actual practice, we do agree on a lot of concepts, and I think we can build definitions of many important terms on what we agree on.
For example, I recently proposed a definition of the term existential clause construction, which is different from other definitions used by some other authors, but not because of any theoretical discrepancies – as I explain in the blogpost, my definition is more in line with the way the term is actually used in the literature. Similarly, I recently proposed a definition of the term morph, which is very simple (“a morph is a minimal linguistic form”), and which again does not derive from any general-theoretical commitment. (It is true that I do not make the generative-grammar assumption of a rich innate grammar blueprint, but I do not exclude this possibility either – it just isn’t relevant for much of the work that grammarians do, because so little is certain about it.)
These definitions will not make everyone happy, but I do not think that this is because my colleagues disagree with “the theory” behind them. Most of the time when people are not happy with my proposed definitions, it is because they have the feeling that they are too narrow or too broad – and they are right! Our grammatical terms are often used in fairly disparate ways, and any (retro-)definition that delimits a term in a clear way will be too broad and/or too narrow from an alternative perspective. Completely smooth retro-definitions are typically impossible.
Bill Croft lists the following things that “we have different theories of”:
deixis and anaphora, accessibility, definiteness, meaning and use, reference and modification, nouns and adverbs…
But how do we know that we have different theories of them? If we do not have clear definitions of these terms, then we cannot even know whether our views differ or not. We often experience difficulty in understanding what our colleagues mean, but is this because of disagreements of “different theories”? Or is it because we do not understand clearly what they mean?
Consider anaphora, a term that was likewise discussed recently on the LINGTYP list (e.g. here). Does it refer narrowly to backward-looking reference, contrasting with cataphora, or is it meant as a general term, covering both backward-looking and forward-looking reference (i.e. is it used as a synonym of endophora)? It would seem that given the context, Croft means endophora (rather than anaphora in a narrow sense), and indeed, in a given context, our terms often become clearer than they would be out of context.
And consider meaning and use: as Arie Verhagen has pointed out repatedly (e.g. here), the English term meaning is broader (at least in its connotations) than German Bedeutung because in English, one can say that “a word means something” (German ein Wort bedeutet etwas) or that “a speaker means something” (German *ein Sprecher bedeutet etwas). When a speaker means something, German must use a special verb (meinen, cognate with mean). So here, too, it may be that a simple consideration of definitions of terms would move us forward, and that we need not rely on the hope that we will eventually find the true “theory” of semantics and pragmatic.
It is true that the terminological difficulties with some other terms are perhaps intractable (for example, the term reference has been used in multiple senses, and currently I have no idea how this situation might be remedied), but terms like “personal pronoun” do not defy definition by any means.
So yes: all our concepts are “theory-laden” (Croft), and yes: all our terms “have their place in a taxonomy (and/or meronomy)” (Lehmann). For example, if a morph is a kind of form, we should be clear about the meaning of form (so my 2020 paper has a section (§4) on the term form, and how it relates to expression and sign). Another example: An affix is a kind of bound form. Especially the proposed definition of affix is partly unintuitive (which is why I hesitated for several years before proposing it), but the problem with it is not that “we disagree about theories”. The term bound form is not widely understood by linguists (although many morphology textbooks define it clearly, as I note in §4 of my 2021 paper), but its definition is not dependent on any larger theory that we might have divergent views about.
Thus, I agree with Croft and Lehmann about the importance of “theories” and “theorizing” – this should go without saying when one proposes definitions of grammatical terms. However, I do not agree that we need to resolve many bigger issues in order to come up with better definitions of traditional grammatical terms that we use in our conversations all the time. There is no good way of achieving this goal (it is a collective action problem that has no easy solution), but it is beginning to be recognized even by some generative linguists (see this recent paper on polysemous terms by Evelina Leivada & Elliot Murphy). We will not make quick progress on the fundamental questions of our discipline (judging by the slow progress over the last centuries), but we can improve some of our methodologies, such as our technical terminology (and maybe even introduce clearer terms for really basic concepts such as Saussure’s faculté du langage).
OpenEdition suggests that you cite this post as follows:
Martin Haspelmath (July 14, 2021). What kind of theory do we need for meaningful conversations about grammatical concepts (such as “personal pronoun”)? Diversity Linguistics Comment. Retrieved April 21, 2025 from https://doi.org/10.58079/nswc
Another thought: doesn’t the idea of only using the same terminology for things that can be shown to be “the same” in the sense of comparative concepts (for example there should be some procedure for determining whether something is a noun or a verb across languages such that every linguist should, in theory, be able to reach consensus on whether something satisfies the conditions to “count” as a verb) inherently prohibit people from saying that something is “underlyingly” a noun, or whatever, without any clear definition for how to determine this? (Which, if I understand correctly, is a key factor of Generative work)?
Well, linguists have different goals, and one goal is language-particular analysis – for such analyses, abstract elements (schemas, underlying structures, zeroes, etc.) are necessary, even though there is no agreement about what the nature of these abstractions should be. So for p-linguistics, I think that underlying structures are often illuminating. But for g-linguistics, we need comparison of languages, and comparative concepts cannot ba based on such language-particular abstractions. P-analyses always have some esthetic component to them, it seems – which is why linguists often engage in “argumentation” (rather than the usual hypothesis-testing mode of other sciences).
A lot of people seem to think that their analysis of specific languages can be useful for understanding Human Language as a whole. For example, they might write an article titled something like Case: Evidence from X language. Am I understanding correctly that you think this sort of thing is an inappropriate conflation of P linguistics and G linguistics?
In general, how would you analysis of individual languages can help in understanding Human Language as a whole (if at all)?
Yes, I have a recent discussion of this issue here: https://ling.auf.net/lingbuzz/005158. As you say, I don’t think that a paper like “Case: Evidence from X language” is viable unless it is embedded in a clear comparative-universalist context.
Out of curiosity, why is it necessary to use abstract structures for language-particular analysis? I’m used to the Generative approach where abstract structures are assumed to tell you something about people’s “mental grammars”. But if you aren’t making claims about mental grammars, what is the purpose of positing such abstract elements? Is it just to make your analysis more compact, or to make it “look nice”?
Since language systems are productive, it is not sufficient to list all the forms (this is different from traffic signs, of which there isa closed class which one doesn’t have to analyze). So one needs abstract structures – in particular constructions and constructional relations (inheritance and sister relations). “Zero” and “movement” (etc) are just particular notations for types of constructions and construction relations. This point may not be obvious from the current literature, which tends to present these differences as ideological. I think they are largely notational, though the question of innateness is a question of methodological strategy (and this closer to an ideological question).
I think one issue is that when various linguists use a single term in different ways, and someone tries to get them to all use it in one way, the ones who are told to change their usage of the term seem to think that this means that something is wrong with the original concept which they used the term for, and thus refuse to change their usage. I think if a term is used in multiple ways, it would be helpful to suggest a new term for the various alternate meaning at the same time as a single definition is suggested for the polysemous term, so that they still have a term to describe concepts that are useful for them.
Yes, we need good terms for all the commonly used concepts. Of course, the number of diverse concepts is completely open-ended, so the temptation to use an old term for a new concept will always be there. The main purpose of my activities in this regard is to create greater awareness in the discipline. There are no easy fixes.