Here's a sentence I'm working on now.

  • Tom saw [Jane reading the book].

I think 'saw' takes two arguments, "Tom" and "Jane reading the book". But I'm not sure what phrase that is, and what the head is in the phrase. And is it okay to consider that 'see' theta-marks Tom and the phrase? I cannot think of any appropriate theta role to assign to the phrase.. I've just started learning linguistics by my own so please help me out. Thank you in advance!

  • 2
    Jane reading the book is what Tom saw; so is the book in Tom saw the book. They should have the same theta roles, then. Of course the subordinate clause (not "phrase" -- there's a predicate with arguments here) has its own cycle, with Jane as subject, and book as object, and the whole clause marked by the Acc-ing gerund complementizer. Context will determine whether Jane is raised in addition.
    – jlawler
    Jun 26 '17 at 17:25
  • @jlawler Jane cannot be subject here, because the name could easily be replaced with "her" without changing any of the sentence's grammar. You could NOT say, "Tom saw she reading the book." You would say, "Tom saw her reading the book." "Her" is never a subject. Thus it is not a subordinate clause, as a clause requires a subject. Mar 26 '18 at 9:19
  • She is the pronoun used as subject of tensed clauses. But her is the pronoun used as subject of gerund clauses. You're right that they can't be substituted for one another.
    – jlawler
    Mar 26 '18 at 14:36
  • In "Tom saw [Jane reading the book]", Jane is only the understood subject of the bracketed catenative complement clause. Syntactically, "Jane" is object of the matrix clause, i.e. object of "saw".
    – BillJ
    Oct 14 '19 at 13:18
  • @BillJ Now I disagree. The gerund phrase/clause is a constituent. It passes the tests for constituents, e.g. clefting, It was Jane reading the book that Tom saw; pseudoclefting, e.g. What Tom saw was Jane reading the book; Answer fragments, e.g. What did Tom see? -- Jane reading the book, passivization, e.g. Jane reading the book was seen by Tom, proform subsitution, e.g. Tom saw that (that = Jane reading the book). I would agree with you if the test sentence were Tom saw Jane read the book. In such a case, the string Jane read the book no longer behaves as a constituent. Jun 10 '20 at 16:06

The complement of saw, Jane reading the book, is, categorially speaking, an IP whose I(nfl) head is the non-finite affix -ing, with a subject Jane in Spec I position (but not a CP, i.e., not a complete 'clause', against what jlawler's comment claims). That IP is, indeed, theta-marked by saw with the role traditionally called 'Theme'. It is not, however, Case-marked by saw, because IPs cannot absorb Case, which entails that the accusative Case feature that saw can assign remains available to be absorbed by some available DP (i.e., Jane, in this case).

As non-finite I(nfl) -ing cannot assign Case (except 'null' Case to PRO in constructions like I enjoyed [ __[PRO reading the book]], where enjoyed takes a full CP complement), in I saw Jane reading the book, Jane then 'exceptionally' receives accusative Case from saw (one of the 'exceptional Case marking' (ECM) verbs - which, at bottom, means a small number of verbs that may select IP instead of CP complements). The difference is that the DP occupying the Spec I of IP complements is not 'protected' by the CP barrier and remains accessible to Case-marking by an external verb, saw in this case, whereas the PRO subject of full CPs is inside the CP barrier and inaccessible to the higher verb (enjoyed, in my example above).

Evidence for the classical analysis just offered comes from at least two simple facts: one is that * I saw [PRO reading the book] is ungrammatical (whereas e.g. I enjoyed [__ [PRO reading the book]] is not), and it is ungrammatical because its PRO subject is doubly Case-marked (as 'null' Case by the IP-internal -ing Infl, and as accusative by the governing verb saw; the other is that when the higher verb see appears in passive form and so loses its transitivity and its capacity to assign accusative Case, as in __ was seen [__ reading the book], the lower subject Jane can no longer receive accusative Case from see and, to avoid violating the Case Filter, must raise from Spec IP into the Spec T of the main clause and become its subject, as in Jane was seen [ t reading the book]. This, in turn, entails that the ing-clause cannot have a PRO subject after ECM verbs like see, since it has a trace-of-Jane one, and so that it must be an IP, not a CP, Q.E.D.

  • this is good, i'm glad you answered this. one minor clarification question i have is the distinction between these ECM cases and the usual infinitival ones, eg with persuade. it seems there are some differences, for instance an adverbial can intervene between the ECM subject with persuade (I persuaded John quite easily to give the lecture). is this possible with -ing examples, eg maybe: I saw Bill clearly reading the book (where clearly is supposed to modify saw, not reading). Jun 10 '20 at 19:13
  • @BillJ The evidence from tests for constituents reveals that the string John to give the lecture in the sentence I persuaded John to give the lecture is NOT a constituent, whereas the string Bill reading the book in the sentence I saw Bill reading the book IS a constituent. See my message above in this regard. In a phrase structure grammar that assumes strict binary of branching, the difference is a problem, because the relevant strings are likely analyzed as constituents in both cases. Jun 10 '20 at 23:18
  • @tim modern approaches to syntax are more sophisticated than just "does this string pass certain constituency tests?" most of the traditional tests (substitution, coordination, etc) are super unreliable. further, not all constituents are even predicted to pass constituency tests. single-bar-level projections for example can't be targeted for movement Jun 12 '20 at 10:19
  • @one-off-post Tests for constituents are on the whole a big problem for phrase structure syntax, so your comment is not surprising for me, since you seem like a true proponent of Chomskyan phrase structure. The tests are, though, more consistent with dependency syntax. I have worked on the issue in detail. If you can be open-minded about the matter, have a read here: journals.helsinki.fi/lud/issue/view/15 Jun 12 '20 at 12:05

Your Answer

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

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