My working model is thus: An adjective should be flexible in that it can describe a variety of nouns. A noun adjunct looks like an adjective but can only be connected to a limited number of nouns. So, hot oven and microwave oven show 'oven' described by an adjective and a noun adjunct, respectively. But may I apply the same method to words that appear to be adjective-like but function like a noun adjunct, and vice versa? Examples follow.

  1. Gordian. To me, Gordian is a noun adjunct, in spite of its suffix, because it basically can only be used to describe 'knot' (see Ngram Viewer data, refresh the search to see the results, if needed)

  2. Fermi. On the other hand, Fermi, when not used in reference to the person, can be an adjective because it can modify a variety of nouns (Ngram Viewer).

Is my reasoning sound? Do you have other criteria to distinguish these parts of speech?

  • 2
    Since there is no real difference between adjective and N' adjunct, you can make the terminological distinction any way you want. Who cares?
    – Greg Lee
    Dec 4 '17 at 9:20
  • 4
    There's also whether the word can be used predicatively (The oven is hot, *The oven is microwave), but then you also run into problems like 'The sequence is Cauchy, and therefore convergent', which seem to suggest that 'Cauchy' is an adjective. (This agrees with your prediction though, since a distribution, a series, the mean value theorem, etc, can also be Cauchy). Dec 4 '17 at 12:00
  • 2
    The parts of speech labels are not affected by the function of the word in the clause. They are simply modifiers and may be nominal ("a brick wall") or adjective ("a long letter"). Btw, I'd use the term 'adjunct' for modifiers and supplements in clause structure, not phrase structure.
    – BillJ
    Dec 4 '17 at 18:25
  • I care because I'm making a discrete list of eponyms which requires boundaries, and I've chosen adjectives. Cauchy, incidentally, is on my list, though I wasn't sure whether people would accept using it predicatively. Thanks! Yet it feels strange for me, a non-mathematician. Furthermore, can radiation be Hawking? Can a test be Turing? Hmm... It seems much easier for a curve to be Gaussian, or a network to be Bayesian, than a level to be Fermi, despite "Fermi level" occurring more frequently than "Bayesian network". I found no examples of Fermi used predicatively. The suffix matters, it seems. Dec 5 '17 at 2:33
  • @BillJ, sorry, I don't quite understand why PoS labels wouldn't be affected by the function. One reason this concerns me is I'm trying to understand how PoS taggers analyze corpora. If you wouldn't use 'noun adjunct' the way that I have, could you offer an alternative? For example, 'attributive noun'? Dec 5 '17 at 2:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.