When a complement PP goes missing: a study on the licensing of Swiping
Overview One of the mysterious properties of the so-called Swiping construction ([1,2,3]) is that
adjunct PPs, but not complement PPs, can license Swiping ([1]). Contrary to this observation, we
show that complement PPs can, in fact, license swiping, but only when they have undergone Heavy
Shift (HS). Based on this novel observation, we make the following two claims. First, following
[4,5,6], we claim that there are two types of movement, one leaving a copy behind and the other not.
We will show that HS falls into the second class. Second, Gapping and Pseudo-gapping (PG) are
necessarily derived by "unreconstructing" movement: HS for Gapping ([7,8,9]; cf. [10,11]) and
A-movement for PG ([12]).
Observation [1] shows that Swiping is licensed with no antecedent PP ((1a)) or an adjunct PP
antecedent ((1b)), but not with a complement PP antecedent ((1c)). One prominent account of this
paradigm is that Swiping is licensed when the PP is not "given" in the antecedent of IP-deletion ([2],
based on "Avoid F" in (2) by [13]). According to [2], a VP that excludes the adjunct PP, which is a full
proposition under the VP-internal subject hypothesis ([14,15,16]), can serve as an antecedent for
IP-deletion. As a result, adjunct PPs are not "given" in the VP ((3)). On the other hand, complement
PPs are necessarily dominated by the innermost VP and must be "given" in the VP. Thus, [2]
concludes that complement PPs do not license Swiping. There are, however, cases where complement
PPs license Swiping; according to our informants, Swiping with a complement PP is acceptable when
it is heavy shifted ((4b)). As far as we know, such data has never been reported. This suggests that HS
somehow makes the complement PP "not given" in the VP. Additionally, Swiping is also licensed by a
complement PP that is a Gapping/PG remnant ((5b,c)), which indicates that Gapping /PG also makes
complement PPs "not given". This parallelism between HS and Gapping/PG suggests that, in both
cases, complement PPs are excluded from the VP that licenses IP-deletion in Swiping. The fact that
in-situ complement PPs do not license Swiping indicates that only an overt operation can make the
complement PP "not given" and that there is no string-vacuous HS.
Analysis First, the asymmetry between in-situ and heavy shifted PPs is accounted for by the so-called
reanalysis operation ([17]). If a complement PP stays in-situ, it undergoes reanalysis and the P and V
are amalgamated as in (6). Assuming that Swiping is generated by adjoining a wh-head to P at PF
((7); [2]), the P melded with V should not be available for this adjunction operation. On the other hand,
we assume that HS frees a P from the peril of reanalysis. Thus, the P in heavy shifted PP becomes
available for the adjunction operation.
     Our analysis implies that the movement that licenses Swiping with complement PPs (i.e. HS) must
not leave a copy. If it does as in (8), the PP should still be "given", assuming that the copy in the VP is
used for the calculation of "given"ness; we would incorrectly predict no contrast in (4). (9) gives a
strong support to this view; the PP containing a bound variable in (9b) cannot be interpreted below the
QP every book when the QP remains in the scope of someone, which shows HS does not reconstruct.
In contrast, a topicalized PP, which is presumably reconstructed, does not license Swiping ((10)). To
capture this, we assume that rightward movement such as HS does not leave a copy, in accord with
the line of argument offered by [4,5,6] for A-movement. In sum, there are two types of movement: the
kind that leaves a copy (i.e. topicalization) and the kind that does not (i.e. HS and A-movement).
     Our account has an implication for the analysis of Gapping and PG. Given the above argument that
Swiping with a complement PP is licensed only by "unreconstructing" movement in the antecedent of
IP-deletion, the contrast in (5a) versus (5b-c) suggests that Gapping/PG remnants also undergo such
movement. Combined with [18]'s argument that Gapping is derived by ellipsis (based on island-repair
phenomena), our analysis supports the "movement-and-deletion" type of analysis of Gapping ((11);
[7,8,9]), where the gapping remnant (Harry) undergoes HS. On the other hand, [12] convincingly
shows that A-movement, and not HS, underlies PG ((12)), based on the fact that PG allows
P-stranding, while HS does not ((13)). Our analysis is compatible with these accounts, assuming
neither HS nor A-movement leaves a copy. If we employ analyses where Gapping/PG remnants stay
in-situ, the contrast in (5) would remain mysterious.
     Finally, our analysis gains support from data involving P-stranding. Recall our claim that
reanalysis blocks Swiping. Given that P-stranding in PG is possible in reanalysis configurations ([19]),
we predict that P-stranding of PG and Swiping are incompatible. This prediction is borne out ((14), in
contrast with (13a)), confirming our reanalysis hypothesis.
Conclusions This study provides novel data where Swiping is licensed with a complement PP when it
undergoes "unreconstructing" movement. As an empirical consequence, we claim that Gapping/PG
remnants undergo such movement. This study also has a significant implication for the theory of
movement and reconstruction: some types of movement (HS and A-movement) do not leave a copy.

(1)    a. She fixed it, but I don't remember what with. (no antecedent PP; sprouting)
       b. (?)She fixed it with something, but I don't remember what with. (adjunct PP)
       c. She talked to someone, but I don't remember who (*to). (complement PP)
(2)    Avoid F: "The content of the focused P should not be given."
(3)    The lower VP segment serving as an antecedent of sluicing
       She [VP [VP tshe fixed it] with something], but I don't remember [CP [PP what WITH] [IP she
       fixed it tPP]].
(4)    a. *[IP John [VP tJohn talked [PP to someone]] yesterday], but I don't remember who to.
       b. [IP John [VP tJohn talked tPP] yesterday] [PP to someone], but I don't remember who to. (HS)
(5)    a. *John talked to Mary, and Bill talked to someone. I don't remember who to.
       b. ?John talked to Mary, and Bill talked to someone. I don't remember who to. (Gapping)
       c. ?John talked to Mary, and Bill did to someone. I don't remember who to. (PG)
(6)    John [VP talkedV+toP Mary].
(7)    X0-adjunction approach by Merchant (2002)
       [PP toP whoD] -> [PP whoD+toP twho]
(8)    [IP John [VP tJohn talked [PP to someone]] yesterday] [PP to someone], but ....
(9)    a. Someone gave every book1 [PP to its1 prize winning author]. (every>some, some>every)
       b. Someone gave every book1 tPP yesterday [PP to its1 prize winning author].
                                                                          (every>some, ??some>every)
(10) *[PP To someone in my department], John talked tPP. I don't remember who to. (Topicalization)
(11) The "movement-and-deletion" approach to Gapping
       Mary [VP [VP dated t1] Bill1] and Susan [VP [VP dated t2] Harry2].
(12) The "A-movement and VP-deletion" approach to Pseudo-gapping
       Mary hasn't dated Bill, but she has [AgroP Harry1 Agro [VP dated t1]].
(13) a. John talked to Mary and Bill did someone. (Pseudogapping)
       b. *John talked to t1 yesterday someone1 from our department. (Heavy Shift)
(14) *John talked to Mary and Bill did someone. I don't remember who to.
(The judgment in (4), (5), (9), (10), (13), and (14) was obtained from our informants through p.c.)

References
[1] Rosen, C. (1976) "Guess what about," NELS 6 Proceedings, 205-211.
[2] Merchant, J. (2002) "Swiping in Germanic," C.J.W. Zwart and W. Abraham (eds.), Studies in
    Comparative Germanic Syntax, John Benjamins, Amsterdam, 289-315.
[3] Craenenbroeck, J. van (2004) Ellipsis in Dutch dialects, Ph.D diss., Leiden University.
[4] Lasnik, H. (1998) "Some Reconstruction Riddles," UPenn Working Papers in Linguistics 5.
[5] Lasnik, H. (1999a) "Chains of arguments," S. Epstein and N. Hornstein (eds.) Working
    Minimalism, MIT Press, 189-215.
[6] Fox, D. (1999) "Reconstruction, binding theory and the interpretation of chains," LI 30, 157-196.
[7] Jayaseelan, K.A. (1990) "Incomplete VP Deletion and Gapping," Linguistic Analysis 20, 64-81.
[8] Lasnik, H. and M, Saito. (1991) "Curious correlations between configurations licensing (or failing
    to license) HNPShift and those for Gapping," ms. UConn.
[9] Yoshida, M. (2005) "The rightward movement analysis of gapping in NP and its structural
    implications," Paper presented at WCCFL 24.
[10] Johnson, K. (1996) "Binding the Gap," ms. UMass, Amherst.
[11] Lin, V. (2000) "Determiner Sharing," WCCFL 19 Proceedings.
[12] Lasnik, H. (1999b) "Pseudogapping Puzzles," S. Lappin and E. Benmamoun (eds.), Fragmants:
    Studies in Ellipsis and Gapping, Oxford University Press, Oxford, 141-174.
[13] Schwarzschild, R. (1999) "GIVENness, AVOIDF and Other Constraints on the Placement of
    Accent," Natural Language Semantics 7, 141-177.
[14] Kitagawa, Y. (1986) Subjects in Japanese and English, Ph.D. diss. UMass, Amherst.
[15] Kuroda, S.-Y. (1988) "Whether we agree or not," Lingvisticae Investigationes 12, 1-47.
[16] Koopman, H. and D. Sportiche (1991) "The position of subjects," Lingua 85, 211-258.
[17] Hornstein, N. and Weinberg, A. (1981) "Case theory and preposition stranding," LI 12, 55-91.
[18] Coppock, E. (2001) "Gapping: In Defense of Deletion," CLS 37 Proceedings, 133-148.
[19] Lasnik, H. (1999c) Minimalist Analysis, Blackwell, Oxford.