I was wondering what arguments there were to know that P is the head of a phrase [P + N].

As far as adjunct phrases are concerned, we can clearly see that as Ps select Ns (*during the rock; *in the Internet (English isn't my first language, I just assume that's the case based on what I've read and heard), etc.).

But what about Ps being selected by Vs? The following question can be posed: What selects Ns? Let's take the phrase to discriminate against rocks for example. We can find its counterparts without a preposition in other languages. They will presumably all exhibit quite identical selectional properties (e.g. an object being discriminated against can't be inanimate), so I find it unreasonable to assume that some semantic features that define "discrimination" are part of the verb in one language but part of the preposition in the other. For if we assume that, we'll be forced to posit lots of homonymous prepositions to account for various sets they combine with.

The same argument can be used in respect to theta-roles (a beneficiary is unlikely to be a non-living thing).

At the same, there are situations in which the same theta role can be introduced by more than two prepositions for a given verb (to think about v. to think of), and these prepositions may have different sets of selection. If we assume that Vs select Ns via Ps then we will have to create homonymous verbs.

The third option that I find appealing is to assume two levels of selection, the first one being that of Vs, the second one - of Ps.

P.S. One indication of Ns not selectiong Ps comes from causative prepositions in Russian. There's a pair of them (из-за & благодаря) that expresses the speaker's attitude to the desirability of the cause. If Ns determine Ps then their selection occurs beyond them as "desirability" is common for every N so that it doesn't differentiate anything, and is distant from not only their semantics but also from their connotations (which is directed at what they denote).

I hope more competent people comment on that and give a more elaborate explanation if mine wasn't complete and comprehensive.

  • Not an answer, but we can discriminate between rocks which still involves discriminating but not against anyone/thing. In your think about and think of example, there is a difference in meaning. Thinking about something involves some kind of prolonged consideration, where as thinking of merely means summoning something to mind. Sep 8, 2023 at 13:10
  • @Araucaria-him I took to discriminate against as an example but between which includes, I believe, a different theta-role. Also to think of/about: I had in mind a Russian example: говорить о/про. Of course one can find differences between them but they basically involve the same theta-role. That's what I intended to put the emphasis on.
    – Shpekard
    Sep 8, 2023 at 15:58
  • discriminate and discriminate against mean different things.
    – Lambie
    Sep 9, 2023 at 16:07
  • Heads are identified in terms of distribution. The word/element that most determines the distribution of the phrase is the head of the phrases. Prepositional phrases have a distribution that is distinct from that of noun phrases, hence the preposition must be the head. If the noun were the head instead, prepositional phrases would not be prepositional phrases, but rather they would be noun phrases. That would make no sense. Oct 10, 2023 at 0:37
  • Concerning the example discriminate against, one can assume it is not necessarily verbs per se that select their arguments, but rather predicates do that. Thus, the predicate is not discriminate alone, but rather it is the two-word combination discriminate against. Oct 10, 2023 at 0:45

1 Answer 1


Let me answer your question from the point of view of German grammar. German, like English, has so-called "prepositional objects", i.e. the verb selects a particular preposition, which has to introduce the object. In a language like German, you see that this preposition in turn assigns case to the noun. So there is a selection chain, and nothing speaks against having the P as the head of the verb's complement, since P visibly governs its own complement in turn. There is also a relation between the verb and the participant expressed by the noun, but this is in terms of semantic roles, and case government is distinct from that. The preposition can be a head and still be meaningless. So the meaning relation is between the verb and the PP, in which the P does not interfere. This is more or less the definition of a prepositional object; a phrase with a fixed and governed preposition, the semantics of the preposition is not active.

Having said that, there is of course a grey area between prepositions and case markers, and this indeed means blurring the difference between the directions in which dependencies run: case is a feature of the N, and a preposition is a governor that embeds the N. The grey area starts with English "of". I understand that the "of-genitive" is indeed analysed as a PP with "of" as a P-head. It just functions like a genitive of the noun. But ok, we can distinguish the syntactic structure from the function. It is still true that the function is performed by a word "of" that embeds a whole NP, in contrast to what a case feature would do (or often be taken to do).

  • @Alazon thanks for the reply. So you main point is government. I thought of it too but forgot to include in my original post. There are languages, like Welsh, that have inflecting prepositions. Suppose agreement is initiated by nouns. How are we then to make sense of the headedness of P?
    – Shpekard
    Sep 10, 2023 at 4:47
  • Welsh P agrees with a following pronoun, which, however can be dropped: "ar" (= on), "arno" or "arno fe" (= on-it it). It is the typical thing for person agreement to be triggered by a phrase that is governed by the agreeing head. So this case is entirely in line and supports the analysis that P is the head in the combination "on it".
    – Alazon
    Sep 10, 2023 at 9:25
  • Indeed, when you speak of an "inflected preposition" (inflected for person, a pronominal category), you are presupposing that P is the head -- otherwise you would call "it" a pronoun in some adverbial case (for example, "on it" with "it" as the head would have to be a pronoun "it" in the "superessive" case, as in Hungarian :)
    – Alazon
    Sep 10, 2023 at 9:36
  • I understand why the analysis of inflecting Ps as heads doesn't contradict a previous account of Ps but it's not the only option: adjective-noun agreement but now that I'm thinking about it, the head triggers agreement on the adjunct which is not the case for verbal agreement. But here we're dealing with Head-Complement, so what kind of agreement can we choose for it?
    – Shpekard
    Sep 10, 2023 at 9:54
  • The generalisation was about person agreement. Adjective-noun agreement is a different construction from head-complement, yes, but lo & see: there is no agreement in the feature person.
    – Alazon
    Sep 10, 2023 at 9:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.