Languages are systems that link forms (or shapes) to meanings, so in this sense, linguistic analysis consists in establishing meaning-form correspondences. And of course, such correspondences explain speaker behaviour. What I’m talking about in this post is a more ambitious kind of explanation: Explaining language structures by meaning-form correspondences. One well-known label for kinds of meaning-form correspondence is “iconicity” – so in a way, this blogpost continues the theme of my (2008) paper, which apparently has not lost its relevance. The occasion is two recent typological studies, on differential locative coding (Matushansky 2019), and differential adpossessive coding (Ortmann 2018). Both work in the tradition of formal-logical semantics, and both share my interest in linguistic diversity and worldwide tendencies.
Matushansky (2019) discusses differential place marking in Latin, or what she calls “restricted locatives”. While spatial relations (locative, allative, ablative) are generally marked with prepositions (in, ad, ab), Latin uses Locative/Accusative/Ablative marking with names of towns and small islands, as well as a few topo-nouns such as domus ‘house’ and rus ‘countryside’. For example, one can simply say Alb-ae [Alba-LOC] ‘in Alba’ or Rom-am [Rome-ACC] ‘to Rome’.
A key aspect of Matushansky’s story for this differential treatment of noun subclasses is the idea that there are two semantic types of nouns: ordinary object-denoting nouns, and a small class of locus-denoting nouns. With object-denoting nouns, one needs a special preposition to shift the noun to make it locus-denoting (e.g. in horto ‘in the garden’). At first blush, this looks very plausible – there is indeed a general tendency for place names to require shorter or no locative markers, documented in great detail by Stolz et al. (2014). And if we claimed that these nouns occur with “underlying” or “zero” prepositions (as some authors have said, such as McCawley and Kayne), we would have to say why it is these nouns, and not others, that idiosyncratically require (or allow) the deletion of the preposition (or a zero adposition). The “locus-denoting” idea makes the Latin pattern appear more natural.
However, this idea is too “natural”: Matushansky has no explanation for why the locus-denoting nouns are restricted in such a curious way in Latin (towns and small islands, but not large islands and countries, and domus ‘house’ but not hortus ‘garden’). This has long been a fundamental issue in grammatical analysis: Morphosyntactic categories often follow conceptual/semantic classes to some extent, but not fully. Verbs often denote actions, but not always. Females are typically in feminine gender, but there are exceptions. And so on… Many have concluded from this that morphosyntax is autonomous, and that there can be no atomistic functionalism (Newmeyer 2001; 2005). In the 1980s, the rejection of generative grammar’s overly formalistic approach led to reactions such as Givón (1984-1990) and Langacker (1987), who emphasized the need to look beyond pure syntactic forms and take into account meanings and pragmatics – these are the views to which Newmeyer reacted, asserting the autonomy of morphosyntax (and rejecting atomistic explanations that links meanings directly to forms). Nowadays, it seems to be semanticists working in the logical-formal tradition that are trying to reestablish the link between meaning and form, by adjusting the assumed meaning that is then linked to the forms. But this strategy works only if there is independent justification for the meanings that are posited. However, Matushansky’s claim of locus-denotation specifically for this idiosyncratic set of Latin nouns seems to rest entirely on their formal behaviour. The same problem recurs in other languages with “restricted locatives”.
By contrast, in my recent (2019) paper on differential place marking, I have not made any speculative claims about particular languages. Instead, I focused on universals, and I provided a functional-adaptive explanation of the universals in terms of the efficiency theory of asymmetric coding. This does not give us elegant language-particular analyses, but we get clear a cross-linguistic predictions: On the scale “human noun > common inanimate noun > place name/topo-noun”, the likelihood of (longer) locative marking decreases as we go down the scale (Haspelmath 2019: 323). The Latin place names and topo-nouns are evidence for this cross-linguistic generalization (which is functionally motivated), and they are illuminated by the efficiency theory, but we still need to accept the idiosyncrasies of Latin. Morphosyntax is autonomous, so the functional explanation cannot consist in atomistic pairing of meanings and forms. (Whether Latin place names are analyzed as locus-denoting need not play a role; maybe it is different for different speakers.)
Moving on to Ortmann’s (2018) paper on alienability contrasts, we see a similar mode of argumentation, but Ortmann’s paper is much more thorough, building on many years of study, taking into account a large number of studies from the typological and logical-semantics tradition (my 2017 paper on the same topic seems to have come too late for him to take into account).
Ortmann argues for a “radically compositional” analysis of the widespread difference between alienable and inalienable constructions. He discusses a semantic difference between semantic possession (as with relational nouns, especially kinship terms and body-part terms), which involves an inherent possessor argument, and pragmatic possession (as with nouns like ‘dog’), where the relation to the possessor argument must be established by a POSS operator (so ‘John’s dog’ has the semantic analysis λx. [DOG (x) ∧ POSS (John’, x), p. 121). Then he argues that alienable relators (as well as possessive classifiers of the Oceanic type) are “morphological exponents of the relation POSS” (p. 129). He discusses a wide range of constructions in great detail, and takes into account a lot of the literature, even relying on a “scale of expressing possession” proposed a long time ago by Seiler (1983; 2001). He seems to endorse the old understanding of the alienability contrast in terms of “iconicity” (Haiman 1983; 1985):
“Greater conceptual proximity between possessor and possessum corresponds to greater morphosyntactic proximity, that is, to less overt “effort” and explicitness in marking (Seiler 1983). It is in this sense that alienability contrasts are said to involve iconicity: less conceptual distance is mirrored by less morphosyntactic complexity.”
It did not become fully clear what role this plays for Ortmann, who focuses on the idea that the semantic analysis should match the formal patterns very closely. But again, as in Matushansky’s paper, it is clear that the semantic analysis is adjusted to fit the formal contrasts, and is not fully motivated by the meanings. It is of course plausible that body-part terms like ‘finger’ and kinship terms like ‘sister’ are “inherently relational”, while nouns such as ‘dog’ need a POSS component – and this has often been said in the past. However, for artifact nouns like ‘house’ and ‘book’, one may wonder whether they do not have an “inherent possessor”, too, because humans to not produce artifacts without an owner in mind. And even ‘dog’ is a very human-centred concept – unlike wolves or whales, dogs do not occur in nature but always coexist with humans, and typically have human owners that feed them.
So again, an explanation of the differential coding tendencies in terms of efficiency of asymmetric coding (as in Haspelmath 2017) is more explanatory. Relational nouns (especially kinship terms and body-part terms) occur more frequently with possessors, so there is less need for overt coding of the adpossessive relationship because it is more expected. As a result, we get universal trends (for which massive evidence exists), but we need not make claims about particular languages. Ortmann’s paper is sufficiently thorough to note some of the language-particular idiosyncrasies: For example, in Ewe, only kinship terms are inalienable, while in Acholi, only body-part terms are inalienable. If the proposal is that the coding differences directly correspond to semantic differences (in the spirit of “radical compositionality”), then how can such differences be dealt with? Ortmann says that such cases are “exceptionally overwritten by idiosyncratic lexical items” (p. 121), but what does this mean? That these are exceptions to radical compositionality? Or that kinship terms in Acholi are not relational, in a “semantically exceptional” way? Again, Ortmann’s attempt to establish a close correspondence between meanings and forms works beautifully for many core cases, but there is no full match. The autonomy of moprphosyntax reasserts itself, in a way that makes the whole enterprise very questionable. Meaning and form do correspond in an approximate way, but not sufficiently to support a true explanation.
Perhaps the strangest consequence of Ortmann’s approach is that the English Genitive ’s ends up having no meaning, because it cannot denote the POSS relation: After all, this is required only for pragmatic possession (“Kim’s house”). But English requires Genitive ’s also for inherent possession, like “Kim’s sister”. There is no POSS here (as we can infer from languages with alienability contrasts), so ’s must do something else here. Quite sensibly, Ortmann does not want to say that there are two different sorts of Genitive markers in English, so he ends up saying that the Genitive ‘s has no meaning, and that POSS has no counterpart in English (§7). So there are two types of nothingness here: a semantically empty marker, and an uncoded meaning. To his great credit, Ortmann takes pains to be consistent, and his proposal is refreshingly counter-Anglocentric – it makes English appear weird. But this situation highlights the fact that it is apparently impossible to come up with a compositional story that is equally intuitive for all languages.
Now what about antipossessive markers, like -tsil- in Yucatec Maya, which signal that a relational noun is not possessed? (e.g. in taatah ‘my father’, tatah-tsil- ‘father’, Lehmann 2002). I took these to show that markers in adnominal possessive constructions play the role of signaling an unexpected situation, in line with the efficiency theory of asymmetric coding (Haspelmath 2017: §6.3.1). Ortmann is thorough enough not to leave these out of consideration, and he says that they “signal a de-relational operation” (§6). This is fairly close to what I said – but is it in line with the spirit of “radical compositionality”? I understood this spirit to be the idea that each form should correspond to a meaning. In the case of “derelationalizers” (what I call antipossessive markers, following Stiebels 2006), the form does not correspond to a meaning – it “signals an operation”. This is not unlike what we find in verbal marking, where voice markers may signal such things as agent suppression (passives) or patient suppression (antipassives). Here, too, the role of the marking seems to be to indicate an unexpected situation – not to denote a particular meaning.
But once we take a broader view, it becomes clear that the rationale for grammatical forms in general is not to signal specific grammatical meanings. Instead, they help the hearer become aware of speaker intentions that they do not already expect anyway. We resort to names or full nominals only when the hearer cannot infer who/what we are referring to, and otherwise we use short anaphoric forms or nothing. We often use plural marking, tense marking and semantic-role marking only (or primarily) when needed. Differential coding is rife in languages. It requires a theory of expectation-based efficient coding, not a theory that assumes consistent meaning-form correspondences. And such a theory predicts cross-linguistic tendencies, not tight meaning-form correspondences at the language-particular level.
References
Givón, Talmy. 1984-1990. Syntax: A functional-typological introduction. Vol. 1-2. Amsterdam; Philadelphia: Benjamins.
Haiman, John. 1983. Iconic and economic motivation. Language 59(4). 781–819. https://doi.org/10.2307/413373.
Haiman, John. 1985. Natural syntax: iconicity and erosion. Cambridge: Cambridge University Press.
Haspelmath, Martin. 2017. Explaining alienability contrasts in adpossessive constructions: Predictability vs. iconicity. Zeitschrift für Sprachwissenschaft 36(2). 193–231. https://doi.org/10.1515/zfs-2017-0009. (https://www.academia.edu/37762885/Differential_place_marking_and_differential_object_marking)
Haspelmath, Martin. 2019. Differential place marking and differential object marking. STUF – Language Typology and Universals 72(3). 313-334.
Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol. 1. Stanford, CA: Stanford University Press.
Lehmann, Christian. 2002. Possession in Yucatec Maya (ASSidUE 10). 2nd, revised edition edn. Erfurt: Universität Erfurt. http://www.christianlehmann.eu/publ/ASSidUE10.pdf.
Matushansky, Ora. 2019. The case of restricted locatives. Proceedings of Sinn und Bedeutung 23(2). 161–178. https://doi.org/10.18148/sub/2019.v23i2.604.
Newmeyer, Frederick J. 2001. Where is functional explanation? In Mary Andronis, Christopher Ball, Heidi Elston & Sylvain Neuvel (eds.), Papers from the Thirty-Seventh Meeting of the Chicago Linguistic Society, 99–122. Chicago: Chicago Linguistic Society.
Newmeyer, Frederick J. 2005. Possible and probable languages: A generative perspective on linguistic typology. Oxford: Oxford University Press.
Ortmann, Albert. 2018. Connecting the typology and semantics of nominal possession: Alienability splits and the morphology–semantics interface. Morphology 28(1). 99–144. https://doi.org/10.1007/s11525-017-9319-6.
Seiler, Hansjakob. 1983. Possession as an operational dimension of language. Tübingen: Narr.
Seiler, Hansjakob. 2001. The operational basis of possession: A dimensional approach revisited. In Irène Baron, Michael Herslund & Finn Sørensen (eds.), Dimensions of possession (Typological Studies in Language 47), 27–40. Amsterdam: Benjamins.
Stiebels, Barbara. 2006. From rags to riches: Nominal linking in contrast to verbal linking. In Dieter Wunderlich (ed.), Advances in the theory of the lexicon, 167–234. Berlin: Mouton de Gruyter.
OpenEdition suggests that you cite this post as follows:
Martin Haspelmath (June 13, 2020). Why meaning-form correspondence is not explanatory: Differential coding in locative and adpossessive constructions. Diversity Linguistics Comment. Retrieved October 4, 2024 from https://doi.org/10.58079/nsvu
Indeed, Matushansky’s concept of object-denoting vs. locus-denoting nouns doesn’t explain the contrast between domus ‘house’ (which can be used without preposition) and hortus ‘garden’ (which can’t). Accordingly, this concept is not appealing as an explanation for the asymmetric encoding of the concept ‘inside something’ in Latin. The same is true for Ortmann’s theory of inalienable possession because it doesn’t account for the contrast between Ewe (only kin-terms but not body-parts) and Acholy (only body-parts but not kin-terms).
Your own theory of asymmetric encoding is – if I’m not mistaken – roughly the following. People expect some nouns, such as ‘house’, to be often used in utterance presupposing being inside. For this reason a special sign for ‘inside’, i.e. a preposition, is redundant and can be omitted. This generates the cross-linguistic tendency of omitting a special marker for ‘inside’ in such words as ‘house’. This explains why Lat domus can be used without preposition. The same is true for kin-terms and body-parts which people expect to be meant as belonging to someone. This is the reason for the cross-linguistic tendency to omit the possession marker on kin-terms and body-parts.
So much is clear. What I still don’t understand is how your theory accounts for such contrasts as Lat domus vs. hortus and Ewe vs. Acholy. If speakers of a particular languge routinely omit a redundant marker, why only in domus etc. but not, for instance, also in hortus? Why only in body-parts (Acholy)? But may be this is just a part of a more general question. If omitting the preposition in Latin is due to its redundant status with several nouns why don’t we see the same omitting in German, English etc.?
Many thanks for this comment! Well, the idea is that the marker is less redundant in domus ‘house’ than in hortus ‘garden’, because, as you say, “people expect some nouns, such as ‘house’, to be often used in utterance presupposing being inside.” The expectation is not the same in other nouns, because these other nouns do not occur in locative contexts to the same extent. Likewise, body-parts occur in possessed contexts much more often than other nouns, leading to certain expectations. So it’s the asymmetries in expectation that lead to asymmetries in coding.
The other question is: Why is there a difference between Latin (with its differential place marking) and German (where we don’t have differential place marking)? Why do Acholi and Ewe differ in the way they do? The answer is: Nobody knows – these differences are the results of historical accidents. We can explain universal trends, but we cannot explain historical accidents. Much of linguistics tries to explain the phenomena of particular languages, but this is a category mistake. Only general properties of Human Language can be explained (in the sense of external explanation).
This is certainly true. It doesn’t make much sense to propose a cognitive (or some other kind of external) explanation for a feature in just one language because this particular feature in this particular language can always be due to a historical accident. It’s different with cross-linguistically recurrent features. Here an accident is a possible assumption in each particular case (like meeting a spider with only 7 legs) but this becomes more improbable with each new language sharing the feature.
So it’s perfectly possible that English once had a house/garden-contrast like Latin and hand/bicycle-contrast like Acholi but lost both by some accident (sound change etc.). The same assumption can be made for German, Russian, Latvian, Sanskrit, Ket etc. But then the lack of the constrasts predicted by your theory becomes recurrent. Correct me if I’m mistaken here but isn’t the lack of these contrasts a widespread feature?
If this is correct than we either need an explanation going beyond the idea of some historical accident or a different theory of differential location/possession marking.
The proposal is that there are two conflicting pressures (or competing motivations): the pressure for explicitness and the pressure for efficiency. Languages like English without a ‘house/garden’ contrast show explicit coding, while languages like Latin show efficient coding. What we don’t find (because no pressure favours it) is languages with counter-efficient coding (locative marking only with ‘house’, but not with ‘garden’).
I see, thanks for clarification. I somehow fogot that the second tendency, favoring explicitness, plays a role too. Your theory is clearly better than the alternatives you discuss in your post because it explains both coding asymmetries (and more) in the same simple and coherent way.
However, I think that languages with something close to counter-efficient coding might exist. What I think of is the so-called inessive case of Lithuanian which codes ‘inside’. The inessive forms are constructed with postposition -e (in dialects -i, i.e. pre-Lith *-en) which is added to the inherited locative. In Standard Lith the postp. -e is obligatory everywhere. But there are dialects in which the postposition is restricted to the singular of (all) nouns and not used in the plural (where the inherited locative just assumes the inessive function). In such dialects we have pievoj-e (loc.sg.-e) ‘in the meadow’ but pievosu-Ø (loc.pl.-Ø) ‘in the meadows’ etc.
Prof. Haspelmath, Regarding ‘s in English, and genitives in general, see several articles/presentations by Barbara H. Partee (joined or individual work), among others, the following ones (they should be available on Partee’s site at UMass, otherwise I can send them to you)
Partee, Barbara H. and Vladimir Borschev 2003 Genitives, relational nouns, and argument-modifier ambiguity, in E. Lang, C. Maienborn & C. Fabricius-Hansen (eds.): Modifying Adjuncts (Interface Explorations 4), 67-112
(is in the list of references of Ortmann A. (2015) Uniqueness and Possession: Typological Evidence for Type Shifts in Nominal Determination. In: Aher M., Hole D., Jeřábek E., Kupke C. (eds) Logic, Language, and Computation. TbiLLC 2013. Lecture Notes in Computer Science, vol 8984. Springer, Berlin, Heidelberg)
Barbara H. Partee, 2006 Lecture 8. Semantics of Possessives, continued. Arguments vs. Modifiers. Puzzles of predicate possessives. ((The Structure of Meaning, Lecture March 20, 2006 )
Partee, Barbara H. Lecture 9. Fragment 3 of English including Possessives, using Lambdas. (The Structure of Meaning, Lecture March 22, 2006 )
Barbara H. Partee 2006 Lecture 10. Compositionality, context-dependence, and meaning shifts. (The Structure of Meaning, Lecture 10 March 27, 2006)
Partee, Barbara H. 2006 A Note on Mandarin Possessives Demonstratives & Definiteness. In Drawing the Boundaries of Meaning. Neo-Gricean Studies in Pragmatics & Semantics in Honor of L.R.Horn, eds. Birner & Ward 263-280. Benjamins.
Thanks – yes, some of these papers were cited by Ortmann. He is more ambitious than Partee et al. in that he wants to link alienability contrasts to compositional interpretations. And he seems to imply that his approach somehow *explains* the formal marking, which it doesn’t (I think).