Examples:
Example | Instruction stream (clickable) | Notes |
---|---|---|
(23a) the book which inspired them (Example (24a); Sag 1997: 448
(23b) *the book which that inspired them |
[they,d,inspire,'v*',[book,which_rel],'Tpast',c_rel,the] | Subject relative. DP [book, whichrel] is
attracted to the edge of Crel by uRel on
Crel. By economy, whichrel values
both uT and uRel on Crel, and the T-to-C option is
impossible, i.e. *which that in (23b).
However, whichrel cannot value uD so book will raise further and uD is valued during the merge with the. |
(24) the person whose mother died (Example (24b); Sag 1997: 448)
|
[mother, '\'s', [person, who_rel], die, 'v~unacc', 'Tpast', c_rel, the] |
Assume whose here is whorel +
's and die is unaccusative,
i.e. [[whorel person]['s mother]] is
theta-merged at the internal argument position.
Furthermore, permit uRel on Crel to find iRel on whorel at the edge of possessive 's. [[whorel person]['s mother]] is attracted to the edge of Crel. However, whorel cannot value uD on person. Next, person raises and its uD is valued when merged with the. |
(25) the person whose mother's dog died (Example (24d); Sag 1997: 448) | [dog, '\'s', [mother, '\'s', [person, who_rel]], die, 'v~unacc', 'Tpast', c_rel, the] | Analysis is the same as for (24) above, except we permit uRel on Crel to recursively look for iRel at the edge of possessive 's. Internal argument is [[[whorel person] 's mother]['s dog]]. |
(26a) the person whose mother's dog we were (all) fond of (Example (24d); Sag 1997: 448)
(26b) the person whose mother's dog that we were (all) fond of (26c) *the person of whose mother's dog we were (all) fond (26d) *the person of whose mother's dog that we were (all) fond |
[dog, '\'s', [mother, '\'s', [person, who_rel]],of,fond,[we,d],v_be,'Tpast',c_rel,the] |
[[[whorel person] 's mother]['s dog]] is
attracted to the edge of Crel by uRel on
Crel. We assume, following (25), that uRel on
Crel can find recursively iRel at the edge of
possessive 's.
whorel cannot value uD on person. Next, person raises and its uD is valued when merged with the. (26b) is predicted as [[[whorel person] 's mother]['s dog]] is not a nominative subject, so economy does not apply. The option of T-to-C generates the complementizer that. (26c-d) are blocked by a spellout rule *of who, designed to block *the girl of who friends bought the cake. |
(27a) the person to whom they dedicated the building (Example (24e); Sag 1997: 448)
(27b) the person whom they dedicated the building to |
[person,who_rel,to,[building,the],dedicate,'v*',[they,d],'Tpast',c_rel,the] |
We assume dyadic to, i.e. the VP is [dedicate [[the
building] [to [whoRel
person]]]]. [whoRel person] is attracted to the
edge of Crel by uRel on Crel.
whorel cannot value uD on person. Next,
person raises and its uD is valued when merged with
the.
Pied-piping is permitted as whorel is overt. Note: after the two successful derivations, at step 41, the system tries unsuccessfully to employ monadic to with [dedicate [to [whoRel person]]]. The problem is the merge of [the building]. At step 48, the attempted XP-YP pair merge of [the building] fails. At step 59, the corresponding XP-YP set merge fails. After that, there are a further two parallel attempts pushing [to [whoRel person]] onto the stack. |
(28) [Give me] the phone number of the person whose mother's friend's sister's dog's appearance had offended the audience (Example (31a); Sag 1997: 450)
|
[audience,the,offend,'v*',[appearance,'\'s',[dog,'\'s',[sister,'\'s',[friend,'\'s',[mother,'\'s',[person,who_rel]]]]]],perf,v,'Tpast',c_rel,the,of,number,phone,the] | The DP with the relative pronoun whorel is
deeply embedded recursively within the specifier of
possessive 's for the head noun
appearance. However, it must be visible to uRel
probing.
Therefore, [[[[[[whorel person]['s mother]]['s friend]]['s sister]]['s dog]]['s appearance]] is attracted to the edge of Crel by uRel on Crel. whorel cannot value uD, so person will raise further to form a new head. Determiner the merges with the NP headed by person, valuing uD on person. |
(29a) [This is] the senator to whose mother's friend's sister's favorite charity the lobbyist had donated a small fortune (Example (31b); Sag 1997: 450)
(29b) [This is] the senator to whose mother's friend's sister's favorite charity that the lobbyist had donated a small fortune (29c) [This is] the senator whose mother's friend's sister's favorite charity the lobbyist had donated a small fortune to (29d) [This is] the senator to whose mother's friend's sister's favorite charity that the lobbyist had donated a small fortune |
[charity,favorite,'\'s',[sister,'\'s',[friend,'\'s',[mother,'\'s',[senator,who_rel]]]],to,[small,fortune,a],donate,'v*',[lobbyist,the],'Tpast',c_rel,the] |
Similar derivation to example (28). We wh-relativize
senator from the lobbyist donated a small fortune
to the senator's mother's friend's sister's favorite
charity.
[[[[[whorel senator]['s mother]]['s friend]]['s sister]]['s favorite charity]] is attracted to the edge of Crel by uRel on Crel. whorel cannot value uD, so senator will raise further to form a new head. Determiner the merges with the NP headed by senator, valuing uD on senator. There are four parses as the options of inserting the complementizer that and the pied-piping of prepositional to are both freely available. |
(30a) the time when I got drunk
(30b) *the time when that I got drunk |
[drunk,get,v_unerg,[i,d],'Tpast',[time,when_rel],c_rel,the] |
Assume the time expression [whenrel time] is
merged at TP level for semantic reasons. Both pair merge and
set merge options are tested. Because extraction must take
place for relativization, only the set merge option can
succeed. (Pair merged adjuncts are inaccessible to probing.)
[whenrel time] is attracted to the edge of Crel by uRel on Crel. whenrel cannot value uD on time. Next, time raises and its uD is valued when merged with the. However, whenrel can value uT (as well as uRel) on Crel. By economy, no T-to-C, that in (30b), is possible. |
(31a) the town where I grew up
(31b) *the town up where I grew |
[up,grow,[town,where_rel],'v_unerg',[i,d],'Tpast',c_rel,the] |
Assume the location adverbial [whererel town]
is merged at the VP level. Both pair merge and
set merge options are tested. Because extraction must take
place for relativization, only the set merge option can
succeed. (Pair merged adjuncts are inaccessible to
probing.)
[whererel town] is attracted to the edge of Crel by uRel on Crel. whererel cannot value uD on town. Next, town raises and its uD is valued when merged with the. However, whererel can value uT (as well as uRel) on Crel. By economy, no T-to-C (that) is possible. Finally, up does not head a PP with complement [whererel town] (merged at VP level instead). Therefore no pied-piping, as in (31b), is possible. |
(33a) someone to rely on
(33b) *someone on to rely |
[someone, d_rel, on, rely, 'v*', ['PRO', d0], 'Tinf', c_rel, d] |
[drel someone] is attracted to the edge of
Crel by uRel on Crel.
drel cannot value
uD, so someone will raise further.
Tinf checks uT on Crel.
(drel cannot value uT on Crel, uT is
valued by Tinf instead.
Pied-piping is blocked by the empty determiner drel, cf. overt whorel in (35a). (Note: this lexicon contains both simple prepositional on and dyadic on, e.g. put the book on the shelf. Step 17 onwards using dyadic on fails to derive.) |
(34a) someone for you to rely on
(34b) *someone on for you to rely |
[someone, d_rel, on, rely, 'v*', [you, d], 'Tinf', for, c_rel, d] | Assume complementizer for generally licenses an
overt subject, you, by assigning Case, cf. PRO in
(33a). Crel, containing uRel, piggy-backs onto
for, attracting [drel someone] to its
edge.
Covert drel in [on [drel someone]] incompatible with pied-piping of on, so (34b) is blocked. As described in (33a), someone will raise further and merge with covert d. |
(35a) the baker in whom to place your trust (Example (63a);
Sag 1997:461)
(35b) *the baker whom to place your trust in (Example (63d); Sag 1997:461) |
[baker, who_rel, in, [ trust, '\'s', [you, d]], place, 'v*', ['PRO', d0], 'Tinf', c_rel, the] | Both (35a-b) are predicted to be grammatical here,
i.e. pied-piping is permitted. (Judgements shown are from
Sag.)
Assume in is dyadic, i.e. VP structure is [place [your trust [in [who baker]]]]. (Note: the derivations with dyadic in begin at step 51 and end at step 103. Simple preposition in is tried first, but fails to converge at step 50.) Either [in [whorel baker]] (for pied-piping) or [whorel baker], the internal argument of in, may be attracted to edge of Crel by uRel on Crel. NB: whorel is defective, i.e. it doesn't value the uD feature on baker. Therefore, baker raises from edge of C and its uD feature is checked by determiner the. |
(36a) the baker to place your trust in
(36b) *the baker in to place your trust |
[baker, d_rel, in, [ trust, '\'s', [you, d]], place, 'v*', ['PRO', d0], 'Tinf', c_rel, the] | No overt whorel here,
cf. (35a-b). Covert drel is used instead,
i.e. internal argument of in is [drel
baker]. [drel baker] is
attracted to the edge of Crel by uRel on
Crel. Pied-piping is not an option since
drel is an empty category, cf. (35b).
drel cannot value uD so baker will raise further and its uD will be checked by the. Assume uT on Crel is valued by Tinf. (Note: the derivation with dyadic in begins at step 28 and concludes at step 54. Simple preposition in is tried first, but fails to converge at steps 16 and 27 for pair and set merge, respectively.) |
(37) the person for us to visit (Example (75b); Sag 1997: 464)
|
[person, d_rel, visit, 'v*', [we, d], 'Tinf', for, c_rel, the] |
Similar to (34a), assume complementizer for
generally licenses an overt subject, we, in a
tenseless clause by assigning Case. Crel,
containing uRel, piggy-backs onto the complementizer
for, attracting [drel person] to its
edge.
drel cannot value uD on person. The head person raises further and merges with the, which values person's uD feature. |
(38) the person to visit (Example (75c); Sag 1997: 464) | [person, d_rel, visit, 'v*', ['PRO', d0], 'Tinf', c_rel, the] |
Compared with (37), there is no for with an overt
subject; instead, we assume the covert subject is PRO.
[drel person] is attracted to the edge of Crel by uRel on Crel. drel cannot value uD on person. The head person raises further and merges with the, which values person's uD feature. |
Stream: | [[drunk],[get],[vunerg],[[i!D],[d!case!N]],[Tpast!phi],[[time!D],[whenrel!case!N]],[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | |
SO: | [] |
Stream: | [[get],[vunerg],[[i!D],[d!case!N]],[Tpast!phi],[[time!D],[whenrel!case!N]],[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | |
SO: | [drunk] |
Stream: | [[vunerg],[[i!D],[d!case!N]],[Tpast!phi],[[time!D],[whenrel!case!N]],[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | |
SO: | [get[get][drunk]] |
Stream: | [[[i!D],[d!case!N]],[Tpast!phi],[[time!D],[whenrel!case!N]],[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | |
SO: | [vunerg[vunerg][get[get][drunk]]] |
Stream: | [[i!D],[d!case!N]] |
Stack (⥥): | |
SO: | [] |
Stream: | [[d!case!N]] |
Stack (⥥): | |
SO: | [i!D] |
Stream: | [] |
Stack (⥥): | |
SO: | [d!case[d][i]] |
Stream: | [[d!case[d][i]],[Tpast!phi],[[time!D],[whenrel!case!N]],[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | |
SO: | [vunerg[vunerg][get[get][drunk]]] |
Stream: | [[Tpast!phi],[[time!D],[whenrel!case!N]],[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | [d!case[d][i]] |
SO: | [vunerg[d!case[d][i]][vunerg[vunerg][get[get][drunk]]]] |
Stream: | [[[time!D],[whenrel!case!N]],[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | [d[d][i]] |
SO: | [Tpast[Tpast][vunerg[d[d][i]][vunerg[vunerg][get[get][drunk]]]]] |
Stream: | [[time!D],[whenrel!case!N]] |
Stack (⥥): | |
SO: | [] |
Stream: | [[whenrel!case!N]] |
Stack (⥥): | |
SO: | [time!D] |
Stream: | [] |
Stack (⥥): | [time!D] |
SO: | [whenrel!case[whenrel][time!D]] |
Stream: | [[whenrel!case[whenrel][time!D]],[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | [d[d][i]] |
SO: | [Tpast[Tpast][vunerg[d[d][i]][vunerg[vunerg][get[get][drunk]]]]] |
Stream: | [[whenrel!case[whenrel][time!D]],[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | [d[d][i]] |
SO: | [Tpast[d[d][i]][Tpast[Tpast][vunerg[d[d][i]][vunerg[vunerg][get[get][drunk]]]]]] |
Stream: | [[crel!rel!T!phi],[the!case!N]] |
Stack (⥥): | [whenrel!case[whenrel][time!D]] 'Tpast?' [time!D] [d[d][i]] |
SO: | [Tpast?[Tpast[d[d][i]][Tpast[Tpast][vunerg[d[d][i]][vunerg[vunerg][get[get][drunk]]]]]][whenrel!case[whenrel][time!D]]] |
Stream: | [[the!case!N]] |
Stack (⥥): | [whenrel[whenrel][time!D]] [time!D] [d[d][i]] |
SO: | [crel[whenrel[whenrel][time!D]][crel[crel][Tpast[Tpast[d[d][i]][Tpast[Tpast][vunerg[d[d][i]][vunerg[vunerg][get[get][drunk]]]]]][whenrel[whenrel][time!D]]]]] |
Stream: | [[the!case!N]] |
Stack (⥥): | [time!D] b [whenrel[whenrel][time!D]] [d[d][i]] |
SO: | [crel[whenrel[whenrel][time!D]][crel[crel][Tpast[Tpast[d[d][i]][Tpast[Tpast][vunerg[d[d][i]][vunerg[vunerg][get[get][drunk]]]]]][whenrel[whenrel][time!D]]]]] |
Stream: | [[the!case!N]] |
Stack (⥥): | [time!D] b [whenrel[whenrel][time!D]] [d[d][i]] |
SO: | [time!D[time!D][crel[whenrel[whenrel][time!D]][crel[crel][Tpast[Tpast[d[d][i]][Tpast[Tpast][vunerg[d[d][i]][vunerg[vunerg][get[get][drunk]]]]]][whenrel[whenrel][time!D]]]]]] |
Stream: | [] |
Stack (⥥): | [time] b [whenrel[whenrel][time]] [d[d][i]] |
SO: | [the!case[the][time[time][crel[whenrel[whenrel][time]][crel[crel][Tpast[Tpast[d[d][i]][Tpast[Tpast][vunerg[d[d][i]][vunerg[vunerg][get[get][drunk]]]]]][whenrel[whenrel][time]]]]]]] |
Step: | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 |
Stack depth: | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 1 | 0 | 0 | 1 | 1 | 1 | 4 | 4 | 5 | 5 | 5 |