[Lingtyp] "grammatically encoded" - reply to Riccardo

Mon Mar 27 18:30:48 UTC 2023

Dear all,

thank you for this highly interesting and thought-provoking discussion. In this context, I would like to point your attention to a recent volume edited by Andrej Malchukov and I in 2020 (Grammaticalization Scenarios: Cross-linguistic Variation and Universal Tendencies, 2 Vols. Berlin: Mouton de Gruyter). One of our aims was to quantify potential covariation effects between different parameters in pathways of grammaticalization from SOURCE to TARGET.

In cooperation with some 30 colleagues and renowned experts, who contributed data and chapters on 29 languages / families / areas, we collected data on some 1.000 pathways across the globe. Each pathway was checked for the following parameters (strongly inspired by Christian Lehmann’s work, with some differences):

Semantic Integrity
Phonetic Reduction
Paradigmaticity
Bondedness
Pardigmatic Variability (obligatoriness)
Syntagmatic Variability
Decategorization
Allomorphy

Each of these parameters were associated with predefined values (1, 2, 3, 4, see questionnaire in the volume). Then, we assigned one these values to each TARGET and we calculated if there was a value change from SOURCE to TARGET ([+] or [–]).

The results clearly show a more fine-grained pattern than is often assumed in general statements on (i) the coevolution of meaning and form and on (ii) the primacy of pragmatic inference.

What seems to be robust across different tests is, roughly summarized, that there are two clusters of covariation between parameters, which we called function-related and form-related. While there are interactions within these clusters, covariation across these clusters is much less frequent.

function-related:
Semantic Integrity, Paradigmaticity, Syntagmatic Variability, Decategorization

form-related:

Phonetic Reduction, Bondedness, Paradigmatic Variability, Allomoprhy

Moreover, there are also family-related issues which need further attention. We think that we just covered the peak of the iceberg with our study. More research, including in particular research beyond the familiar European languages, will be needed for understanding what underlies processes of linguistic change as we recurrently find them in pathways of grammaticalization cross-linguistically.

All the best,
Walter

________________________________
From: Lingtyp <lingtyp-bounces at listserv.linguistlist.org> on behalf of Kasper Boye <boye at hum.ku.dk>
Sent: Monday, March 27, 2023 12:41:29 PM
To: rgiomi at campus.ul.pt
Cc: Peter Harder; lingtyp at listserv.linguistlist.org
Subject: Re: [Lingtyp] "grammatically encoded" - reply to Riccardo

Dear Riccardo and all,

Thank you, Riccardo, for sharing your thoughts on “degree of grammaticity”.

You say that you “see no contradiction between ‘consider[ing] grammaticity features as symptoms of grammatical status’ and ‘score degree of grammaticity in terms these symptoms’”. In medicine, there is a contradiction, as witnessed by the existence of asymptomatic infections, and you wouldn’t find a good doctor who takes, say, the number of little red spots to indicate degree of severity of a rubella infection.

On the view of grammatical status that Peter and I have advocated (and it seems that at least you consider this view one option), degree of grammaticity is excluded by definition because grammatical status is defined relatively, in terms of conventionalized discursively secondary status. That is, a given linguistic element is grammatical if it is by convention devoid of the potential for being primary. On this relative understanding, it does not make sense to speak about degree of grammaticity: a given element cannot be more or less devoid of the potential for being primary

This of course, does not entail that it is always an easy matter to distinguish grammatical and lexical status empirically. Moreover, it may be that linguistic elements differ in terms of absolute discourse prominence potential (that is, they may differ in terms of their potential for attracting attention), but such differences are not differences in terms of degree of grammaticity, as long as grammatical status is understood in terms of relative prominence potential.

Similarly, linguistic elements may differ in terms of the number symptoms they show of secondary status, but just like in medicine, an increased amount of symptoms may be a function of time, rather than of the size of the underlying cause. For instance, phonological reduction is motivated by secondary status (alongside frequency or predictability) along the following lines: the speaker does not expect the listener to pay much attention to secondary elements, and may therefore invest less articulatory energy in them. It is natural to expect that the longer this motivation applies, the higher is the probability for phonological reduction.

You say that you would not in principle expect to find elements that are focalizable, but not modifiable, since modifiability and focalizability both follow from lexical status. This ignores that linguistic elements differ in numerous other ways, however. For instance, the fact that some demonstrative pronouns are focalizable (what I want is THAT!), but not regularly modifiable (*little that) follows straightforwardly from the fact that they form whole constituents on their own (in other words, they do not come with a modifier slot). This is again motivated by their function, which is distinct from that of noun-based noun phrases, and also for this reason, it makes no sense to take the lack of modifiability as an argument that (focalizable) demonstratives are less lexical or more grammatical than noun phrases.

A strong argument for degree of grammaticity would be that grammaticalization displays a lexical-grammatical cline. But in many cases, at least, this is not the case when you look closely at the basic change, namely the meaning change. For instance, the development of English gonna from go does not display a gradual conceptual change from ‘motion’ to ‘future’. Rather the ‘future’ meaning is from the very beginning conceptually clearly distinct from the motion meaning. It arises as a construction-dependent discursively secondary inference in the context of purposive infinitives (going (in order) to swim), and is subsequently conventionalized (cf. the paper attached in an earlier reply).

With best wishes,

Kasper

Fra: Riccardo Giomi <rgiomi at campus.ul.pt>
Sendt: 22. marts 2023 18:56
Til: Kasper Boye <boye at hum.ku.dk>
Cc: William Croft <wcroft at unm.edu>; lingtyp at listserv.linguistlist.org; Peter Harder <rgl226 at hum.ku.dk>
Emne: Re: [Lingtyp] "grammatically encoded" - replies to Christian, Riccardo and Bill

Dear Kasper and all,

As far as I can see, you only get "silly" results if you pick the wrong parameters. I have never claimed that phonological reduction should be a measure of grammaticality -- as already pointed out by Christian in his seminal work, phonological reduction is ubiquitous in language, on the one hand (being involved in so many cases of lexical change), and is not involved in all cases of grammaticalization, on the other -- i.e. it is neither a sufficient nor a necessary criterion.
As regards closed-class membership, that is also not a criterion in itself for me, but it is of course a precondition for the, I believe, much more interesting notion of paradigmaticity (which I would define in terms of systematic semantic opposition, e.g. in pronoun or auxiliary systems, and not just in terms of syntactic behaviour -- lest we reach the paradoxical conclusion that virtually all nouns, verbs, etc. in a language form a sort of super-paradigm).

That said, I agree that we need to gain a clear understanding of "what the symptoms are symptoms of". If you take modifiability and focalizability, for example, which you also discuss in your papers with Peter Harder, both seem to be very reasonable parameters to me, because they are both manifestations of an element's potential for discourse-primary status. Yet, you do have cases of elements that are focalizable but not modifiable: since both criteria depend on the same fundamental property, you would in principle not expect this to happen: if it does, the only way out I see is to say that elements that pass one test but fail the other are more lexical (=less grammaticalized) than those that fail both. And this is precisely the reason why, methodologically, I would maintain that it makes perfect sense to (i) formulate a definition of grammatical vs. lexical status (which is bound to be a theory-driven decision, I'm afraid, such as your primary vs. secondary discourse status or Christian's notion of autonomy); (ii) identify several criteria that reflect that single underlying property (and I say "several" because, as pointed out in my earlier message, there might be functional and/or morphosyntactic classes for which the same underlying property manifests itself in different ways); (iii) use all the criteria that apply to a given class to *calculate* the degrees of grammaticality of its individual members (and this is where things get more empirical). In the end, this should lead to a continuum of more-to-less grammatical(ized) elements, which is consistent with most models' take on the lexical/grammatical opposition as a matter of gradience. Needless to say, the results may differ enormously, depending on your initial defition and the criteria that you pick as a consequence of that definition.

This is nothing particularly new, as you can see, just my (certainly imperfect) explanation of why I see no contradiction between "consider[ing] grammaticity features as symptoms of grammatical status" and "score degree of grammaticity in terms these symptoms".

Best wishes,
Riccardo

Kasper Boye <boye at hum.ku.dk<mailto:boye at hum.ku.dk>> escreveu no dia segunda, 20/03/2023 à(s) 20:07:

Replies in order of appearance below (again, apologies to those not interested in this discussion!)

Dear Christian,

This concerns your important remark that on the account advocated by Riccardo and by Peter and me, it might seem as if grammatical status were devoid of positive properties.

At a theoretical level, the basic idea is that grammatical can only be understood relative to lexical elements, which they presuppose and upon which they depend. I think that at least traces of this idea are found in several attempts to understand the lexical-grammatical contrast. For instance, “function words” presuppose “content words” in relation to which they have a function; “procedural meaning” presupposes “conceptual meaning” in relation to which the procedures apply. On our account, this asymmetry basically has to do with attentional potential: grammatical elements are conventionalized as background, and hence depend on combination with lexical elements, which have the potential to be foreground.

Still at a theoretical level, the implication of this is that grammatical elements do indeed share positive properties: they represent a special kind of conventions (they are conventionalizations of background status), and they show special types of dependencies. In Boye & Harder (2012), we suggested ‘ancillariness’ as a cover term for these properties. As captured by this term, however, these properties can only be understood in relation to the properties of lexical elements.

Empirically, the implication is that grammatical elements can only be identified relative to lexical ones. Since, the basic difference between lexical and grammatical pertains to attentional potential, grammatical elements can be identified in terms of lack of potential for prominence. There is, however, also what may be seen as a positive criterion: grammatical elements are dependent on combination with (ultimately lexical) elements in relation to which they are secondary.

Dear Riccardo,

While there is no necessary conflict between continua and discrete cutoff points, I think it is a mistake to score degree of grammaticity in terms of number of ‘grammaticity features’ like closed class membership and phonological reduction.

You may of course define grammatical status in terms of these features, but then you run into problems: if you don’t require presence of all features, you get silly results (e.g. Kingston is more grammatical than Georgetown); if you do require presence of all features, you exclude a number of things you don’t want to exclude (e.g. going to has a grammatical variant prior to phonological reduction to gonna).

Alternatively, you may consider grammaticity features as symptoms of grammatical status. This is perfectly fine as long as the symptoms can be derived from an understanding of the phenomenon of which they are symptoms (cf. Boye & Harder 2012). However, if you then score degree of grammaticity in terms these symptoms, you confuse the number of symptoms with an unwarranted assumption about a property of what the symptoms are symptoms of.

Dear Bill,

Thanks for bringing the question what makes a concept grammaticalizable into the discussion. I think Peter’s and my own account is compatible with yours. From our point of view, the alternative to grammatical meaning – i.e. meaning conventionalized as discursively secondary – is discursively secondary meaning that must be arrived at through context-dependent inference. Thus, grammatical meaning represents a shortcut (by way of convention) to discursively secondary meaning. This means that what makes a meaning, function or concept grammaticalizable must be that it is convenient as discursively secondary information across different contexts, and useful enough to license conventionalization. I think this is exactly the case with the concepts or functions you point at in your 2007 paper.

With best wishes,

Kasper

Fra: Lingtyp <lingtyp-bounces at listserv.linguistlist.org<mailto:lingtyp-bounces at listserv.linguistlist.org>> På vegne af William Croft
Sendt: 18. marts 2023 00:17
Til: lingtyp at listserv.linguistlist.org<mailto:lingtyp at listserv.linguistlist.org>
Emne: Re: [Lingtyp] "grammatically encoded" - answer to Christian

I have read this discussion with interest. I have found analyzing the lexical-grammatical distinction difficult; I made several attempts. I started from Talmy's observations about the semantic categories that recur in the inventories of grammatical elements across languages. I should mention at the outset that the distinction Talmy was considering is grammaticalizable (or "grammaticizable" -- Slobin 1997) concepts vs. other, "lexical-only" concepts. This is different from grammaticalized vs. not grammaticalized concepts. Grammaticalizable concepts may be expressed lexically, but may then undergo grammaticalization. Grammaticalized concepts are not only grammaticalizable but have undergone grammaticalization and hence have acquired additional properties -- more on that below.

At any rate, I decided that the best way forward was to embed the issue in a broader theory of the verbalization of experience. I took Chafe's model of verbalization (Chafe 1977a,b are the primary original sources) and elaborated it in a 2007 paper (which cites other work of Chafe's). In the conclusion I suggested some reasons why certain concepts are grammaticalizable, based on the model of verbalization.

This is a functional model, so it is similar in aim to Kasper and Peter's proposal. But that is about grammaticalizable concepts. To understand the nature of grammaticalized concepts, it is possible, even likely, that at least some of the structural considerations that Christian proposes are relevant.

Interjections only partly fit in the verbalization model. But the verbalization model is just one part of a larger model of language and its role in communication, and the role of communication in joint action. I have largely followed Clark's version of the larger model (Clark 1996; he calls verbalization 'formulation'); see Croft (2009). Interjections are varied and serve a different functions in the larger model.

Bill

Chafe, Wallace. 1977a. Creativity in verbalization and its implications for the nature of stored knowledge. Discourse production and comprehension, ed. Roy Freedle, 41-55. Norwood, New Jersey: Ablex.

Chafe, Wallace. 1977b. The recall and verbalization of past experience. Current issues in linguistic theory, ed. Peter Cole, 215-46. Bloomington: Indiana University Press.

Clark, Herbert H. 1996. Using language. Cambridge: Cambridge University Press.

Croft, William. 2007. The origins of grammar in the verbalization of experience. Cognitive Linguistics 18.339-82.

Croft, William. 2009. Toward a social cognitive linguistics. New directions in cognitive linguistics, ed. Vyvyan Evans and Stéphanie Pourcel, 395-420. Amsterdam: John Benjamins.
Slobin, Dan I. 1997. The origins of grammaticizable notions: beyond the individual mind. The crosslinguistic study of language acquisition, vol. 5, ed. Dan I. Slobin, 265-323. Mahwah, N.J.: Lawrence Erlbaum Associates.

________________________________
From: Lingtyp <lingtyp-bounces at listserv.linguistlist.org<mailto:lingtyp-bounces at listserv.linguistlist.org>> on behalf of Kasper Boye <boye at hum.ku.dk<mailto:boye at hum.ku.dk>>
Sent: Tuesday, March 14, 2023 3:47 AM
To: Christian Lehmann <christian.lehmann at uni-erfurt.de<mailto:christian.lehmann at uni-erfurt.de>>; lingtyp at listserv.linguistlist.org<mailto:lingtyp at listserv.linguistlist.org> <lingtyp at listserv.linguistlist.org<mailto:lingtyp at listserv.linguistlist.org>>
Cc: Peter Harder <rgl226 at hum.ku.dk<mailto:rgl226 at hum.ku.dk>>
Subject: Re: [Lingtyp] "grammatically encoded" - answer to Christian

  [EXTERNAL]

Dear Christian,

Regarding focusability

The focusability criterion (like the other criteria) depends on language- and construction-specific means for focusing, and different means come with different limitations. Indeed, the criterion may in many cases be impossible to apply. Another criterion (not mentioned in Boye & Harder (2012), but in later publications) is modifiability; again, we are not the first to suggest that modifiability can be used to distinguish lexical from grammatical elements, but the criterion can straightforwardly be derived from the claim that grammatical elements are by convention discursively secondary.

As already mentioned, there are also limitations to stress as a means for investigating focusablity. For instance, stress is not always found exactly on the focused element. Your example with would seems to belong to the group of exceptions that fall under verum focus. Therefore, your example does not show that would can be focused, and hence it does not show that would is lexical. In principle, however, it is perfectly possible that some modal verbs, or variants thereof, are lexical. In Boye (2010), I argued that this is the case with some Danish modal verbs. In Boye & Bastiaanse (2018) we showed that even a rather course-grained distinction between lexical and grammatical modal verb variants in Dutch is significant for the description of agrammatic speech: the proportion of modal verb items classified as grammatical relative to items classified as lexical was significantly lower in agrammatic speech than in the speech of non-brain-damaged controls (see Boye et al. 2023 for an overview of similar studies, and a usage-based theory of agrammatism).

Regarding your proposal for a definition of grammatical status

I would be very interested in seeing your detailed proposal, but my basic problem with your proposal is not that it looks circular, but that it looks entirely structural. I prefer a functional-cognitive, usage-based definition that entails a rationale for the existence of grammar.

Regarding grammaticalization as a gradual phenomenon

I agree that grammaticalization is a gradual phenomenon, but not that this entails that grammatical status is a matter of degree. In the attached preprint (paper to appear in Transactions of the Philological Society), I argue that on a strict understanding, grammaticalization is embedded in at least three continua (a conventionalization continuum, a splitting continuum and a discourse prominence continuum), but does not presuppose or show any evidence of lexical-grammatical cline.

References

Boye, K. 2010. ‘Raising verbs and auxiliaries in a functional theory of grammatical status’. K. Boye & E. Engberg-Pedersen (eds.). Language usage and language structure. Berlin: Mouton de Gruyter. 73-104.

Boye, K. To appear. ‘Grammaticalization as conventionalization of discursively secondary status: Deconstructing the lexical-grammatical continuum’. Transactions of the Philological Society.

Boye, K., & R. Bastiaanse. 2018. ‘Grammatical versus lexical words in theory and aphasia: Integrating linguistics and neurolinguistics’. Glossa: a journal of general linguistics, 3.1, 29. DOI: http://doi.org/10.5334/gjgl.436<https://eur02.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdoi.org%2F10.5334%2Fgjgl.436&data=05%7C01%7Cboye%40hum.ku.dk%7C6888e70c1c1f438a0c4208db2afebf1e%7Ca3927f91cda14696af898c9f1ceffa91%7C0%7C0%7C638151045860446123%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=DqiFoR3Ae8iMT%2FOLgyT8e46%2F7NbDRy7ROhfpJ9g0E%2BE%3D&reserved=0>

Boye, K., R. Bastiaanse, P. Harder & S. Martínez-Ferreiro. 2023. ‘Agrammatism in a usage-based theory of grammatical status: Impaired combinatorics, compensatory prioritization, or both?’ Journal of Neurolinguistics 65, 101108.

Fra: Lingtyp <lingtyp-bounces at listserv.linguistlist.org<mailto:lingtyp-bounces at listserv.linguistlist.org>> På vegne af Christian Lehmann
Sendt: 10. marts 2023 21:07
Til: lingtyp at listserv.linguistlist.org<mailto:lingtyp at listserv.linguistlist.org>
Emne: Re: [Lingtyp] "grammatically encoded" - answer to Christian

Dear Kasper,

the clearest cases of focusing are cleft-sentences. At the same time, it is clear that many sentence components are not amenable to clefting, and many of those that are not are nevertheless lexical rather than grammatical. Thus I suppose focusability will not, for your purposes, be operationalized as amenability to clefting.

Thus you need to consider milder forms of focusing. If contrastive stress counts, then it remains true that many items that have otherwise been regarded as grammatical can bear contrastive stress. Think of exchanges such as this:

Will you do it? - I would do it if [so and so].

In my understanding, what is focused here is exactly the conditional modality, so what is stressed is its expression.

My attempt at a definition may seem circular until I spell out how constraints on the distribution of items and classes of items are formulated and quantified. (It has probably been done somewhere in the literature.) This is independent of a prior definition of 'grammar'; it just refers to cooccurrence of items in constructions. When I have spelled out some cases, I may take the liberty of sending you the URL.

Allow me to repeat that if you take grammaticalization seriously as a gradual phenomenon, then grammatical status, too, is not a yes-or-no matter, but rather one of degree. Consequently, no single binary criterion like focusability will suffice for its operationalization.

Best,

Christian

--

Prof. em. Dr. Christian Lehmann
Rudolfstr. 4
99092 Erfurt
Deutschland

Tel.:

+49/361/2113417

E-Post:

christianw_lehmann at arcor.de<mailto:christianw_lehmann at arcor.de>

Web:

https://www.christianlehmann.eu<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.christianlehmann.eu%2F&data=05%7C01%7Cboye%40hum.ku.dk%7C6888e70c1c1f438a0c4208db2afebf1e%7Ca3927f91cda14696af898c9f1ceffa91%7C0%7C0%7C638151045860446123%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=414C2MSreE3tjw3JNViq7JIamyYL91ngIt7W1xCIJw0%3D&reserved=0>