The Indo-European languages, as well as many other languages around the world, have a count/non-count distinction as a part of their grammar. Some uses of nouns or noun phrases (NPs) with number modifiers are grammatical and others are not; some quantifiers are grammatical when used with some nouns/NPs while others are not; some nouns can appear with indefinite articles, but others cannot (or only with great difficulty). For instance:
On the opposite hand, some nouns/NPs do not have plural forms, and so all agreement is singular; some uses of nouns/NPs allow for comparison with “measure phrases” but others don’t (in the singular).
The uses of car and similar nouns/NPs is count; those of water, air, and blood are non-count (also known as mass).
When this phenomenon is seen as a grammatical restriction in the language, it is natural to make it become a syntactic rule, as done in grammars like Duden, Quirk et al, Huddleston/Pullum, and others. But there is a bit of a paradox here, since one would have to know whether a noun or NP is mass or count in order to know whether it can be used in the plural, or with an indefinite determiner, etc. There should be some special as well as visible (to the linguist at least) ingredient "in the noun or NP" that demonstrates its membership into the mass or count category.
So, this makes one look “into the noun” to find something relevant, and a most plausible candidate would seem to be something “ontological” or “in the meaning” that distinguishes count from mass nouns or NPs – perhaps some notion of “individual” or “object” or a process of “individualization”. However, this approach has also seen difficult challenges, ranging from the fact that some relevant languages make (translations of) the same word differ in their count/mass status. The word "lightening" possesses a mass reading in English while the German counterpart "Blitz" is interpreted as count although they refer to the same concept.
Even within the same language words that seem to denote the very same type of “ontological item” can be mass or count.
An example is the different countability class of onion (Allium cepa) and garlic (Allium sativum). Both are subspecies of the same category of plants (Allium) and thus very similar on an ontological basis. Nevertheless, onions are countable and garlic is not.
Another problem is the existence of “universal grinders” and “universal packagers” that can change an object's countability class.
In considering all kinds of “semantic” options, one very basic fact has been overlooked: most nouns are ambiguous (polysemous). In some cases, all “senses” of the noun fall into the same class, but this is not correct for all classes. Consider the innocuous word marble.
If interpreted as a hard crystalline metamorphic rock that takes a high polish; used for sculpture and as building material it behaves like a mass noun, if interpreted as a small ball of glass that is used in various games, it behaves like a count noun (the interpretations are taken from WordNet).
There might be a hidden rule that makes instances out of material, but this might not be the whole story (it applies to iron (material and club), but what about love (the feeling, and the persons towards the feeling is expressed)? And how do the processes underlying these ambiguous nouns to so-called dual life nouns, items that seem equally to be mass and count, as e.g. cake?
Another issue that is well known (at least since Allan (1980)) is that the distinction between mass and count is not actually binary. Allan develops eight classes, the empirical analysis carried out in BECL contains at least four classes.
So: what will work? Has there been some unappreciated presupposition made here that should be uncovered? Possibly, and that is something this network is to wonder about. Furthermore, we wish to study the following issues: