Preposition Stranding in Danish and English

advertisement
On Preposition Stranding
in English and Danish
Brad Penoff
1. Introduction
The phenomena of preposition stranding is very rare when examining various
languages around the world. The most distinct class of languages that demonstrates this
is that of the Germanic languages. Behavior within the Germanic languages varies from
one language to the next. On the one end, Danish, as Herslund (1984:58) puts it,
“…allows almost completely free stranding of prepositions.” Hornstein and Weinberg
(1981) believe that English can only be stranded from a PP that is immediately dominated
by a VP. This analysis, however, is does not account for several cases. Later,
Chomsky’s (1986) Barriers approach explains some unaccounted for cases. However, he
still leaves some English phenomena unaccounted, as Takami (1988) points out. It is
pointed out by Maling and Zaenen (1985) that Icelandic demonstrates preposition
stranding in certain constructions. On the other hand, stranding in Dutch is limited to a
certain brand of pronouns while it is restricted altogether in German (Herslund 1984).
Accounting for such a phenomena has been tried by numerous theorists (see
Hornstein and Weinberg (1981), Herslund (1984), Maling and Zaenen (1985), Chomsky
(1986), Takami (1988)). Despite these attempts, many individual accounts have posited a
theory that has explained a certain subset of the phenomena of preposition stranding. At
the same time, each has left a hole in their own accounts, at one point or another. In this
paper, we attempt to point out these holes, one by one. Since the papers we will explore
mostly account for English phenomena, we will try to get a larger picture by exploring
preposition stranding in Danish. Next, working from existing theories, we explain how
the phenomena of preposition stranding behaves in English and Danish. Finally, we
explain how the insights of the data and theories presented can be used to explain what is
happening in cases of preposition stranding in English and Danish.
2. Holes in existing theories
2.1 Holes in Hornstein and Weinberg (1981)
2.1.1
Under this theory, Hornstein and Weinberg account provide a universal account of
preposition stranding, as stated in (1).
(1a) a universal Case-marking convention:
NP -> [ + nom] if it is the subject of a tensed S
NP -> [ + obj] if governed by V
NP -> [ + obl] if governed by P
(1b) a universal filter blocking oblique traces: *[NP eoblique]
(1c) a language-specific rule of syntactic reanalysis:
V -> V* (where V c-commands all elements in V*)
Rule (1c) creates complex verbs out of any VP internal constituent(s). Under this
reanalysis, the NP complement of the old PP is now governed by the V under the new
VP. It can then be said that this NP is now marked with objective case since V governs
it. This NP can then be subject to movement now that its trace it would leave behind
would pass through the filter in (1b).
Under this analysis, the following behavior can now be accounted for:
(2a) [Which chair]i did you [[put your coat on]V ti]VP?
(2b) [Who]i did Robin [[talk to your mother about]V ti]VP?
(3a) *[Which year]i did Shakespeare [complete writing Hamlet]VP in ti? (Takami
(1988:300))
(3b) *[What inning]i did the Yankees [lose the ballgame]VP in ti? (H. & W.
(1981:56))
The grammatically of the examples in (2) have NPs governed by Vs and, therefore,
are subject to reanalysis. The trace left behind has objective case assigned to it (see (1a)),
and passes through the filter (1b) successfully. On the other hand, the examples in (3)
have NPs which are governed by Ps, so the trace left behind has oblique case assigned to
it. This is blocked by (1b).
Several arguments against this theory can be made. To begin, for reanalysis in
general, it is undesirable to lump together multiple VP internal constituents into one
complex verb. Not only is this undesirable, but it appears to be incorrect. Notice the
following examples:
(4a) John [likes]V apples and Nancy [ ]V oranges. (Takami (1988:304))
(4b) *Janet [put her engagement ring on]V the sofa and Sue [ ]V the table. (Takami
(1988:304))
(5a) [What]j did you [[talk to that guy ti about]V tj [who was here yesterday]S’,i]VP?
(Levine (1984:24))
(5b) [[talk to that guy [who was here yesterday] S’ about]V what]VP (Levine (1984:24))
Takami (1988) points out that if the complex verb in (4b) was to be considered a
single lexical item, then it should be able to be gapped in the same manner as in (4a).
This is not the case, though.
Hornstein and Weinberg’s approach requires that reanalysis “crucially applies in the
base preceding all transformations”. Levine (1984) points out that extraposition, in the
case of (5a), must then apply after reanalysis. This means that (5a) would have the deep
structure in (5b). If this is the case, then extraction for extraposition must occur out of
the lexical item “talk to that guy who was here yesterday about”. Reaching into a lexical
item is a concept that is also undesirable.
The arguments so far have been against the main tool that Hornstein and Weinberg
(1981) used, namely reanalysis. However, if we grant them the correctness of their
framework, counter examples can still be found. For example, they claim that
preposition stranding can only occur in VP internal contexts. This apparently is not the
case.
(6a) What day did it rain [on t]?
(6b) Which World War did John lose his arm [in t]? (Takami (1988:305))
(6c) Which act did John leave the theater [after t]? (Hornstein and Weinberg
(1981:79, fn. 25))
(7a) Which park did you see the baseball game [in t]?
(7b) Which library do you usually study [at t]? (Takami (1988:305))
(7c) Which city did the candidate tour [through t]?
It appears that PPs that are not VP internal, such as temporal PPs (as in (6)) and
locative PPs (as in (7)), will allow stranding in certain cases, but not in others (as in (3)).
Hornstein and Weinberg say that the processing strategy of such examples can “bypass
the grammar” (H. & W. (79:fn. 25)). This explanation seems ad hoc, considering that
examples with similar structures (as in (3)) can not bypass the grammar, as well.
It can be argued that there truly is no such construction as “VP-internal”, but rather
that the PPs in (6) and (7) adjoin to a higher VP node. We will reconsider this argument
in section 4.
2.1.2
But what about extraction from complex NPs? Hornstein and Weinberg’s theory
claims that the following behavior, as explained by Anderson, is simply “a subcase of a
more general phenomena” (Hornstein and Weinberg (1981:70)):
(8a) the picture of John (Hornstein and Weinberg (1981:69))
(8b) the performance of the play (Hornstein and Weinberg (1981:69))
(9a) [the [picture]N [John]NP]NP (Hornstein and Weinberg (1981:69))
(9b) [the [performance]N [of peace]PP]NP (Hornstein and Weinberg (1981:69))
(10a) Who does he [[have the picture of]V t]VP?
(10b) *What is that the performance [of t]PP?
As put by Hornstein and Weinberg (1981), Anderson claims that complex NPs
including PPs can be base generated in two ways (see (9)): either the PP is present (as in
the (b) cases) or the preposition is added by another rule further in the derivation (as in
the (a) cases). (10b) would violate the filter in (1b) since the trace is assigned oblique
case since it is governed by a P (see (1a)). On the other hand, the trace in (10a) is
assigned objective case since a V governs it(see (1a)). As we will discover in the next
section, Chomsky’s “Barriers” framework essentially takes this idea and formalizes it
more cleanly.
2.2 Holes in Chomsky (1986)
NP
N’
DET
the
NP
N
N’
DET
PP
the
picture
N’
N
P
NP
of
t
PP
P
NP
performance of
Figure 1
t
Figure 2
To begin, let us review how Chomsky determines grammaticality differences in cases
of preposition stranding within NPs like those in (10). Note Figure 1 and Figure 2 above.
(11a) Who does he have [the picture [of t]PP]NP? (= (10a))
(11b) *What did he see [the performance [of t]PP]NP?
The details and definitions of Chomsky’s notions of barriers, blocking categories,
L-marking, Subjacency, and barrier inheritance are not discussed here (For a more
detailed account see Chomsky (1986) or Culicover (1997).). However, assuming a
familiarity with these concepts, it can be said is that Chomsky argues that since the PP in
figure 1 is L-marked by the N “picture”, it is not a blocking category, and therefore not a
barrier. This means that the object of the PP can be subject to movement, as in (11a). On
the other hand, the PP in figure 2 is not L-marked by the lexical N “performance” since
the PP is a sister to the projection N’ instead. This means that the PP forms a barrier to
extraction. The NP inherits barrierhood from the PP, so would need to cross two barriers
in order to extract as in (11b). This sentence is starred since Subjacency is violated.
The data presented thus far has only accounted for NP internal cases involving the
preposition “of”. This might suggest that “of” is more open to stranding than other
prepositions. This is explored below.
(12a) He saw [the bomb [with red lettering]PP]NP.
(12b) *Whati did he see [the bomb [with ti ]PP]NP?
(13a) He saw [the picture [with us in it]PP]NP.
(13b) *Whati/*Whoi did he see [the picture [with ti (in it)]PP]NP?
(14a) He enjoyed [the story [about the blue boat]PP]NP.
(14b) *Whati did he enjoy [the story [about ti ]PP]NP?
(15a) He liked [the man [in the car]PP]NP.
(15b) *Whati did he like [the car [in ti ]PP]NP?
Apparently, prepositions other than “of” are more resistant to stranding when they
are NP internal. Another point is that if (11a) says the lexical entry of the N “picture”
takes a PP complement then (13b) raises the question whether N lexical entries, in
general, can only accept certain PP complements or not. We will interpret Chomsky
(1986) to mean that this selection can be encoded into the lexicon.
Chomsky’s approach captures VP internal cases, as well as extraction from NPs.
However, Chomsky (1986) fails to account for VP external cases such as those presented
as counter examples to Hornstein and Weinberg (1981) in (6) and (7). This hole is
apparently filled in Takami (1988).
2.3Holes in Takami (1988)
2.3.1
Takami takes a functional approach to explain VP internal as well as VP external
cases of preposition stranding. He notes the failures of purely syntactic accounts of the
phenomena, especially in VP external cases. His approach accounts for this previously
unexplained class of stranding by taking into consideration pragmatic effects and, in
some cases, prosodic factors. His so-called “more/less important” approach also has a
sort of gradient effect build into it that accounts for “the fact that the phenomenon under
discussion is not a matter of complete acceptability or complete unacceptability” (Takami
(1988:333)). No syntactic theory has fully explained this.
As with any other functional account, Takami’s approach has a certain degree of
vagueness built into it. In one sense, his solution creates new problems since the answer
to the phenomena wanting to be described is a vague method that cannot make discrete
grammaticality discussions. Regardless, his approach brings up some interesting points
about the phenomena at hand and makes some observations about cases of preposition
stranding that are intuitive to native English speakers.
Below is stated the hypothesis that Takami (1988) proposes:
(16) An NP can only be extracted out of a PP in which the NP may itself be
interpreted as being more important than the rest of the sentence.
Using the pragmatic condition for stranding in (16), Takami analysis focuses on the
notion of importance. This notion is decided in (17)-(21) by reformulating the (a)
examples into interrogatives in the (b) and/or (c) forms. By analyzing what the focus of
these reformulated questions is, the importance of the stranded PPs in the (a) examples
can be determined. The focus of questions is to the “more important” part of the
sentence, so if the “more important” part is the PP then the preposition can be stranded.
(17a) Which year did the couple get married [in t]? (Takami (1988:315))
(17b) Was the couple getting married in 1950?
(17c) Did the couple get married in 1950?
(18a) ?Which year did you finish school [in t]? (Takami (1988:315))
(18b) Were you finishing school in 1950?
(18c) Did you finish school in 1950?
(19a) ??Which year did you go on a holiday [in t]? (Takami (1988:315))
(19b) Were you gone on a holiday in 1950?
(19c) Did you go on a holiday in 1950?
(20a) ??/*Which date did he die [on t]? (Takami (1988:315))
(20b) Was he dead on January 1, 1950?
(20c) Did he die on January 1, 1950?
(21a) *Which year was John still a small boy [in t]? (Takami (1988:311))
(21b) Was he still a small boy in 1950?
(21c) *Did he still a small boy in 1950?
In (17), the reformulated interrogatives are definitely asking about the particular year
in question rather than if they were married or not. So, the PP is the most important part
of the sentence and (17a) is ruled grammatical. On the other extreme, in (21), the
reformulated interrogative in (21b) is definitely asking whether he was a small boy or
not. Therefore, the PP is not the most important part of the sentence and (21a) is ruled
ungrammatical.
On the other hand, in (18)-(20), it is not as clear where the focus of the (b) and (c)
questions is. In particular, in (19b) and (19c), using different intonations and contexts,
either question could be interpreted to be asking either if he went on a holiday or not in
1950, or if he went on a holiday in 1950 or in some other year. For this reason, we get an
intermediate grammaticality judgement.
2.3.2
Despite the apparent success Takami has in explaining the behavior of VP related
cases of preposition stranding, Takami fails to explore how (16) would work in NP
internal cases of preposition stranding. Let’s take a look at some data and see if his
hypothesis can be applied.
(22) Who did he give Brian [a picture [of t]PP]NP ?
(22b) Was it a picture of Sue that he gave Brain?
(22c) Did he give Brian a picture of Sue?
(23a) *What did he lend Brian [the performance [of t]PP]NP ?
(23b) Was it a performance of Hamlet that he lent Brian?
(23c) Did he lend Brian a performance of Hamlet?
In order for (16) to account for the difference between (22a) and (23a), it would have
to be the case that (22b) and (22c) are questioning if the picture was of Sue or if the
picture was of someone else. On the other hand, it would also have to be the case that
(23b) and (23c) are questioning what was lent to Brian. This desired effect is not
necessarily the case, as we will show below.
Below, the question in (17c) is asked and answered. Using (16), we are trying to
account for the grammaticality of (17a). Along the same lines, (22c) is asked and
answered. Again, using (16) in this case, we are trying to account for the grammaticality
of (22a).
(24a) Did the couple get married in 1950? (=(17c))
(24b) No, it wasn’t until 1951.
(24c) *No, they had lunch on Wednesdays.
(25a) Did he give Brian a picture of Sue? (=(22c))
(25b) ??No, it was Sharon.
(25c) No, it was a breath mint.
In the above, the (b) and (c) parts are possible answers to the corresponding (a) part.
The judgement in (24c) follows from the fact that since they are already a couple, we
know that they were had already married, according to Takami (1988:315). However,
the answers to question (25a) are not as clear since it sounds as if the question is asking
about what was given.
So it appears that the same tests that he uses to distinguish the grammaticality of
(17a) by asking (24a) can not yield the expected results by asking to (25a) to get the
grammaticality of (22a). Notice how the same confusion occurs when comparing (25a)
and (21a) to that of (26a) and (23a).
(26a) Was he still a small boy in 1950? (=(21b))
(26b) *No, it wasn’t until 1951.
(26c) No, he was a middle-aged teen.
(27a) Did he lend Brian a performance of Hamlet?
(27b) ??No, it was Macbeth.
(27c) No, it was a blender.
So what does this failed functional account of explaining preposition stranding within
NPs tell us? Let us divert this question to section 4. In the meantime, let’s make some
observations about the behavior of VP preposition stranding in Danish
3. Preposition stranding in Danish
Following Herslund (1984), we will show how the phenomena of preposition
stranding behaves in Danish. To begin, wh-movement from a PP within a VP is allowed
(see (28)). Also, topicalization (as in (29)), clefting (as in (30)), relativisation (as in
(31)), and infinitival constructions (as in (32)) are all grammatical in Danish, as are their
English counterparts. Note that (28)-(32) are all apparent VP internal extractions.
(28) Hvad tænker du på e? (Herslund (1984:49))
‘What are you thinking of?
(29) Olseni ville jeg aldrig stemme på ei. (Herslund (1984:49))
‘Olsen I would never vote for’
(30) Det er Olseni jeg stemmer på ei. (Herslund (1984:49))
‘It is Olsen I vote for’
(31) Pigeni somi jeg har fået den af ei er rejst. (Herslund (1984:50))
‘ The girl who I have got it from has left’
(32) Peteri er svær at blive klog på ei. (Herslund (1984:51))
‘Peter is difficult to figure out.’
Danish also has cases of stranding that involve VP external PPs. Again, this is
definitely a thorn in the side of syntactic accounts such as Hornstein and Weinberg
(1981), as Herslund (1984) points out.
(33) Hvemi var han gået i teatret med ei? . (Herslund (1984:52))
‘Who had he gone to the theater with?’
(34) Hvilken øi har han købt hus på ei? (Herslund (1984:53))
‘Which island has he bought a house on?’
(35) *Hvadi kom han efter ei? (Herslund (1984:54))
‘What did he come after?’
(35) is judged ungrammatical by Herslund, so it is true that Danish does not allow
stranding in all cases. However, to try to account for which cases they apply and which
cases they do not, let’s apply Takami’s hypothesis expressed in (16).
When applying Takami’s functional approach to these, it appears that the
grammaticality of the Danish examples can be explained much in the same way as we did
for the English examples (17) and (21) respectively in (24) and (26). (36) illustrates how
to account for the grammaticality of (34). (37) illustrates how to account for the
ungrammaticality of (35).
(36a) Did he buy the house on Fanø?
(36b) No, it was in Hawaii.
(36c) *No, it was a yacht.
(37a) Did he come after his crash?
(37b) ??No, it was after his appointment.
(37c) No, he had to take a week off.
(36a) apparently is asking whether the house is on Fanø rather than if it was a house
that was bought. On the other hand, (37a) apparently is asking if he was coming or not
instead of asking what he was coming after.
So it appears that it is safe to say that the grammaticality of Danish VP internal and
VP external cases of preposition stranding are identical to that of English. Also, cases of
VP external stranding in Danish can be accounted for with Takami’s (1988) functional
approach, as well.
4. Interpretation
As shown in section 2, syntactic theories have only been able to account for cases of
stranding that are either VP internal or NP internal. On the other hand, Takami’s
functional approach was only able to account for all VP related cases of stranding in both
English and in Danish. For a pictorial representation, see Figure 4.
functional account
NP internal
VP internal
VP external
syntactic account
Figure 4.
NP internal cases are best accounted for by using Chomsky (1986) to say that certain
NPs L-mark certain PP complements. This information is in the lexicon for specific Ns.
On the other hand, VP external cases are beyond any lexical or syntactic account.
(38a) *Which party did John bury the letter [after t]? (Takami (1988:309))
(38b) Which attack did the pirates bury the treasure [after t]? (Takami (1988:310))
(39a) *Hvadi behandlede os med ei? (Herslund (1984:54))
‘What did he treat us with?’
(39b) Hvadi behandlede han bordplanden med ei? (Herslund (1984:54))
‘What did he treat the table top with?’
For syntactic theorists such as Chomsky (1986) that have accounted for VP internal
cases of stranding, they might argue that there is no such thing as VP external PPs. For
instance, they may support a structure like that in Figure 5 rather than Figure 4.
S
NP
VP
S
PP
NP
VP
VP
Figure 4
PP
Figure 5
Proof supporting Figure 5 follows.
(40)John left the theater after 10 PM and Susan left the theater after 10 PM.
(41)John left the theater after 10 PM and Susan did too after 11 PM.
(42)John left the theater after 10 PM and Susan did too.
In order for the structure in Figure 4, when we use a proform for the VP in (40) in
(41) then the PP must follow it. However, as shown in (42), it appears to be optional.
This would support the argument for the structure in Figure 5 since it could be said that
either VP could be replaced with the proform “did too”.
Regardless whether the stranding is VP internal or VP external, the data in (38) and
(39) cannot be explained syntactically. Figure 3 still applies in that there is a class of VP
stranding that we can explain syntactically and a class that we cannot. A possible
solution to this is if we had information in the lexicon about the particular brands of
complement that could be stranded. So we wouldn’t just have that it takes an NP but
rather an NP from some particular subset of NPs. But how would we encode this
information into the lexicon in light of (38)? Both NPs are objects.
This may tell us that, some cases of the actual account of preposition stranding lies
beyond the syntax. Examples like (38) can only be accounted for regarding native
speaker intuitions about pirates and letters. Takami’s approach may shed a little light on
this side of things. However, it cannot be the case that it lies entirely beyond the syntax
though, as shown by the difficulty of relating Takami’s functional account for NP internal
cases of stranding (see Section 2.3.2).
Other languages’ grammaticality judgements regarding preposition stranding can be
explained syntactically. Dutch, for example, only allows a certain brand of r-pronouns in
wh-constructions (Hornstein and Weinberg (1981), Herslund (1984)). German bars
preposition stranding altogether (Herslund (1984)). However, in languages like English
and Danish, it can be said the phenomena of preposition stranding falls outside of the
syntax, although not completely.
References
Chomsky, N., 1986. Barriers. Cambridge, MA : MIT Press.
Culicover, P., 1997. Principles and Parameters. Oxford : Oxford University Press.
Herslund, M., 1984. Particles, prefixes, and preposition stranding. In: F. Sørensen and L.
Heltroft (eds.), Topics in Danish syntax, 34-41. Copenhagen: Akademisk Forlag.
Hornstein, N. and A. Weinberg. 1981. Case theory and preposition stranding. Linguistic
Inquiry 12, 55-91.
Levine, R., 1984. Against reanalysis rules. Linguistic Analysis 14, 3-30.
Maling, J. and A. Zaenen. "Preposition Stranding and Passive" Nordic Journal of
Linguistics, 8(2), 1985, 197-209.
Takami, Ken-ichi. "Preposition Stranding: Arguments Against Syntactic Analyses and
Alternative Functional Explanation". Lingua, 76(4), 1988, 299-335.
Download