Transparent s Contrar to m earlier conviction, a conversation with Michael Ellsworth has persuaded me that the transparent phenomenon has to be looked at as complementation to the t-nom, rather than modification b the PP[of], because of the rampant occurrence of DNI. (1) I ll have a cup, thanks. Chuck s observation that sometimes the of can be left out reinforces this: (2) that tpe event So I ll tr to develop a version of the semanticall inert ( case making ) PP[of] and a pumping cn to get transparent s out of regular ones. E.g., transparent cup out of regular cup. Since DNI is part of the stor, I ll take a shot at that, too. I am going to stick to m claim that determiners have to agree with the t- head, despite Chuck s e of (3) these kind of apples because I can t think of an other eamples of non-agreement with t- that don t involve these or those plus a tpe/kind/sort (see 4). So I m hoping (3)-tpe facts (determiner agrees with obj of of) can be relegated to a less productive cn of their own. (4)a. *several kind of apples b. *three kind of apples c. *those bowl of nuts d. *this kinds of apple Also, (3) seems to have a special kind of semantics, not well paraphrased b these kinds of apples, this kind of apples, or even these apples, but better b apples of this kind. So it isn t eactl as if kind were simpl transparent. Taking this tack, if it can be maintained, permits one the attractive generalization that the t--headed NPs are sntacticall normal and that therefore that the ordinar headcomplement-cn is at work in licensing t--headed NPs as well as the DNPs that have t- headed NP complements. That is, this approach lets us do all the work with a leical entr we need anwa (inert of), a pumping cn taking ordinar s to t-s, along with leical entries for the inherentl-t-s as needed. 1
Of leeme <of > prep CAT DET + [1] [1] This is the same of leeme that will be needed elsewhere: (5)a b c d top of the bo niece of the vicero aware of his faults deprived of her libert T- Pumping For a transparent pumping cn we need something with a saturated daughter and a mother that requires an empt PP[of] complement and which makes the inde and LTOP of that PP (which are the inde and LTOP of the PP s DNP comp) its own inde and LTOP (modulo DNI, which I ll come to later). Also the of the mother want to sa that the mother t- delimits the complement s interpretation in a contetuall appropriate wa. 2
3 [C] phrase of,... prep [B] cntt delimit fr -ER -EE transparent--cn [C] [B] (Please forgive the ugl processing.) I ve used the feature in a wa that won t scale to select a PP[of]. We once discussed at one time the old CG leical-head feature. I forget what the HPSG-stle alternative might be.
4 EXAMPLE: cup in tast cup of tea <cup> phrase of,... prep cup fr INST cntt delimit fr -ER -EE < cup > cup fr INST
DNI Here s a shot at an approach to null complement anaphora and DNI in particular. For simplicit, I ll talk just about dni first and then tr to generalize to null complements. Assume that ever sign with a list also has a (possibl empt) N(ull)COMP(lement)S list, each of whose members is a Null Complement Anaphor (nca). dni is a subtpe of nca. The FS of a nca contains a single feature [ inde]. dni inde nca inde / \ ini inde... Ever member of the NCOMPS is some subtpe of nca and is constrained to share an inde with a member of the list. (6) [NCOMPS <...[ ]...> [ <...[ ]...>] Here s a dni pumping cn. SYN CAT [1] [B] [D] NCOMPS [E] [F] the [G] BV REST [C] [2] CAT [1] SYN [B] [D] [C] NCOMPS dni [E] [F] the [G] BV [C] REST [2] (Sorr things got a little out of line, but I hope ou can read it. MathTpe is the pits. Haven t got time for nca generall now, but there are probabl enough mistakes in this so far to satisf anone s critical faculties.) 5