It sometimes happens that different scholars arrive at similar conclusions more or less independently, and such cases are probably particularly good indications that they are on the right track. It seems that Scott Grimm’s recent study of count-mass, singular-plural and singulative-collective noun pairs (published in 2018) is a good example of this, as its results converge amazingly with a study published by Andres Karjus and myself in 2017.When he wrote his paper, Grimm seems to have been unaware of our study (though it was cited in Kurumada & Grimm 2017), just as we did not know his earlier unpublished work (Grimm 2012) when we wrote our paper (originally presented at the DGfS conference in Potsdam in 2013).
The basic idea of both papers is that the different ways of expressing count-mass and singular-plural contrasts are neither fully determined by the nature of the entities in the world, nor are they completely arbitrary. Different languages have different conventions, but they are limited by a scale that can be represented (in abbreviated form) as in (1).
(1) (the individuation scale) substances < aggregates < individuals
In Haspelmath & Karjus, we do not talk about a scale and focus on aggregates and individuals, but we say clearly what kinds of languages should exist and what kinds of languages should not exist:
(2) individuals universally tend to be expressed by plurative pairs (= ordinary zero singulars and marked plurals), while aggregates tend to be expressed by singulative pairs (in languages that have singulatives)
We looked at five languages with singulative pairs (Welsh, Maltese, Arbore, Murle and Krongo; the two first are also in Grimm’s sample) and found that they show the singulative pairs mostly with small fruits or vegetables, small animals, but sometimes also with paired body-parts and groups of people, e.g.
Maltese: dubbien ‘flies’ – dubbien-a ‘a fly’
Welsh: mwyar ‘blackberries’ – mwayar-en ‘a blackberry’
Arbore: moyde ‘eyebrows’ – moyde-nte ‘an eyebrow’
Murle: ŋooru ‘beans’ – ŋooru-woc ‘a bean’
Krongo: tuliŋ ‘leaves’ – n-tuliŋ ‘a leaf’
So these might all be subsumable under “aggregates”. But how do we explain that nouns denoting aggregates tend to appear as singulative coding pairs? Both Grimm and Haspelmath & Karjus (the latter in much greater detail) look at text frequencies of different types of nouns and find that aggregates are (relatively) more frequent with plural (multiplex) meaning, and individuals are more frequent in with singular (uniplex) meaning. Thus, we get special coding when a grammatical meaning is unexpected given the lexical item, and Haspelmath & Karjus argue that the patterns are instances of the much larger pattern of frequency-based coding asymmetries, which are explained by the functional-adaptive factor of coding efficiency.
Grimm (2018: 551) first invokes the old idea of “morphological markedness” being rooted in “semantic markedness”, and says that individuals are expected to “designate singular entities by default”, while aggregates would be “plural by default” – but it is unclear here where exactly these defaults come from, and why there should be a correspondence with formal marking. In a later section (560-562), Grimm turns to dicussing frequency, and while downplaying its importance (“countability classification cannot be reduced to textual frequency”), he eventually recognizes that
“zero-coded plurals or collective countability classes are beneficial from the point of view of economy of expression”
and that in languages such as Welsh, “the default form simply corresponds to the more frequently used form”, because “there is an intimate link between which word forms correspond to the default situation in the world and textual frequency”. Grimm stops short of saying, as we do, that the frequency asymmetries provide a causal explanation for the form asymmetries, but at least he comes close to it.
It appears that the major difference between Grimm’s overall approach and our (or my) approach is that Grimm is concerned with characterizing what is going on in particular languages, while our paper limits its aims to explaining cross-linguistic tendencies. Of course, language-particular phenomena are often much richer than what can be captured by cross-linguistic comparisons that aim to find universal tendencies. A lot of what happens in languages is idiosyncratic, and this is also the case in the domain mass/count contrasts, which take diverse forms in different languages. But one nice aspect of Grimm’s paper is that while he seems to start out from mass/count contrasts (‘dog’ vs. ‘water’), he actually looks primarily at the way languages code singulars and plurals. His term “countability class” really refers to classes of nouns with respect to singular and plural marking (he also uses “grammatical number category” in the same sense, and simply “grammatical number”, as in the title, which may be confusing). It is hard to know what the exact meaning of words that translate as ‘water’, ‘bread’ or ‘mosquito’ is in languages which do not have any plural marking (or optional plural marking) – maybe all nouns really refer to “masses” in such languages? But this is probably a Eurocentric question, which may not have an answer, and it is clear that cross-linguistic comparison with respect to singular and plural marking is not only possible but also yields clear universal tendencies with good functional-adaptive explanations.
Only one thing is clearly missing from Grimm’s paper, and it is present only marginally in our paper (at the very end): The full comparison of substance nouns (‘water’) and aggregate nouns (‘beans’) with respect to uniplex marking. In Murle, as we saw above, the aggregate noun ŋooru ‘beans’ has the singulative counterpart ŋooru-woc ‘a bean’, and in English, the substance noun water has the minimal-part counterpart a drop of water. Nouns like drop (of water), grain (of salt), head (of cabbage) are not grammatical markers, but they represent additional coding, as expected from the perspective of frequency asymmetry and coding efficiency. Thus, if we explain the overt marking of singulative aggregates by coding efficiency, we should probably also explain the contrast between water and drop of water by coding efficiency. In other words, perhaps the nature of mass nouns is not primarily their peculiar meaning, but the fact that they have very special counterparts to express uniplex meaning. In any event, from our perspective, substance nouns are perfectly in line with the explanations of the coding asymmetries that we developed primarily with respect to aggregate and individual nouns, and Grimm’s more detailed study seems fully compatible with this perspective.
Grimm, Scott. 2012. Number and individuation. PhD dissertation. Stanford: Stanford University.
Grimm, Scott. 2018. Grammatical number and the scale of individuation. Language 94(3). 527–574. doi: 10.1353/lan.2018.0035
Haspelmath, Martin & Andres Karjus. 2017. Explaining asymmetries in number marking: Singulatives, pluratives, and usage frequency. Linguistics 55(6). 1213–1235. doi:10.1515/ling-2017-0026.
Kurumada, Chigusa & Scott Grimm. 2017. Communicative efficiency in language production and learning: Optional plural marking. http://kinderlab.bcs.rochester.edu/papers/optional-number-marking_CogSci_final.pdf