Parts of speech

We all learned about nouns, verbs and their friends in elementary school.  A typical list is

• noun
• pronoun
• adjective
• verb
• adverb
• preposition
• conjunction
• interjection

(See here for more detail)

These are generally useful categories, but if you're really trying to figure out a language, you have to slice a little finer.  For example there are

• Transitive verbs (ones that take an object, like hit)
• Intransitive verbs (ones that don't take an object, at least not typically, like sit)
• Modal auxiliary verbs (like can, could, may and might, which some dialects can stack up into lovely constructions like might could and may can)
• Phrasal verbs, like get up
• Countable nouns, like tree
• Uncountable (or mass) nouns like water
• Pluralia tanta (always-plural nouns -- singular plurale tantum) like scissors
• Comparable adjectives, like tall
• Uncomparable adjectives, like dead, or NP-complete
• Determiners, including articles like the or an, but also adjectives like some, any, all, this or that
• Comparators, like more, most, less and least

Such distinctions go some way towards predicting what words can and can't be used together.  For example, you don't normally use comparators like more with uncomparable adjectives:

• Smith is more famous than Jones.
• *Graph isomorphism is more NP-complete than 3-Sat.

(The * at the beginning indicates something that wouldn't normally be said.  I'm fudging with "wouldn't normally be said" instead of "incorrect" or "ungrammatical" as it is notoriously easy in general to invent contexts in which a given construct would make sense.)

This being language, the boundaries aren't perfectly crisp.  Mass nouns don't generally appear as plurals, but there are a few exceptions, for example

• When referring to some standard serving, as in I ordered three waters.
• When referring to different types of a given substance, as in She preferred the wines of Bordeaux.

Some nouns can work both ways, for example

• Hand me a brick.
• We need five tons of brick.

And let's not even get into whether it's OK to say "more unique" even though unique is supposed to be an absolute and therefore uncomparable.

The word that got me thinking of all this was summit, as a verb meaning "to reach the summit of".  As the "of" would suggest, this verb is generally transitive -- it takes an object, as in Apa Sherpa summited Everest for the twentieth time (which he actually did, last May).  However, the object is often omitted, as in Apa Sherpa summited for the twentieth time.  In contexts where this would be said, it would be abundantly clear that Everest was the peak in question.  In particular, it doesn't matter how many other peaks he might have climbed how many times.

So is summit then acting as an intransitive verb, or a transitive one with an implied object?  I tend towards the latter, as would most grammarians, I believe.  But what about more common cases like sing?  In I sang, there is no implication that I sang any particular song, so one would think sing is acting intransitively.  But I must have been singing something.  Is it really acting transitively, but with an implied, unspecified object?  At some point, such qualifications cease to pull their own weight.  As the man said, volleyball is technically racketless team ping-pong, played with an inflated ball and raised net while standing on the table, but what does that buy us?

What interests me here is how grammar, which is by definition pure syntax, seems unable to stay cleanly separated from semantics.  For example, some mass nouns resist the plural

• * I would like three neutroniums.
• * He was a connoisseur of neutroniums.

In the first example, one does not serve neutronium.  In the second, there is only one kind of neutronium.  How would we detect such errors?  I would think the process is something like

• In a construction like three neutroniums, if the object is a substance, we expect it to mean a particular sort of container full of the substance.
• But that doesn't make sense in the case of neutronium.

In that view, the syntax is fine and the error is semantic.  Mass nouns, then, are syntactically nouns, but ones whose plural forms have particular semantic features.  Similarly, whether a verb is used transitively or intransitively is a syntactic distinction, but whether there is an implied object is semantic concern.

Except that "object" is a syntactic concept.  One way of reconciling this is to posit that the syntactic form Apa Sherpa summited, for example, is somehow transformed into the form Apa Sherpa summited Everest, with Everest as the object.  The choice of "transformed" here deliberately suggests transformational grammar, though I'm not sure that's completely appropriate.

Another would be to posit that the form Apa Sherpa summited gets transformed into some internal structure, in which the concept represented by summited requires something acting in the semantic role of "thing which is summited", which we may as well call an "object", albeit with some risk of confusion.  This putative internal structure would be describable in words, for example Apa Sherpa summited, or He summited, or Apa Sherpa summited Everest, or Everest was summited by Apa Sherpa and so on, but it would be an essentially different structure from any of those sentences.  As I very dimly undestand it, this is more along the lines of cognitive grammar.