Download Papier HT_verbessert

Notes&on&stress%reconstruction!
and$syntactic$reconstruction!
Hubert Truckenbrodt
1. Overview
Bresnan (1971, 1972) has observed interaction of stress-assignment with syntactic movement. This occurs with wh-movement and in relative clauses. This paper investigates such
effects in German in terms of stress reconstruction.
It is not unlikely that this phenomenon is connected to syntactic reconstruction in a principled theory of grammar (Chomsky 1993). This paper assesses interactions of stress reconstruction with syntactic reconstruction to see whether (a) they provide independent support
for this connection or (b) they are compatible with this possibility or (c) they speak against
it. Interaction of the kind that could support (a) exists with reconstruction of a class of more
opaque idioms and with Condition C reconstruction. The result is that (b) obtains, and that,
in one of two ways (we cannot decide), the structural source of stress reconstruction is coextensive with that of syntactic reconstruction: Either one entails the other in a case-by-case
basis. Or both rely on (potentially separate) obligatory reconstruction mechanisms (as was
suggested for syntactic reconstruction by Chomsky 1993 in the form of the preference principle). This result about co-extension of the two may be taken as a weak argument of the
kind in (a) in favor of a shared structural source of stress reconstruction and syntactic reconstruction.
Section 2 reviews Bresnan’s observation about English. Section 3 establishes different
aspects of the effect and its analysis in German. Section 4 addresses the interactions with
more opaque idioms. Section 5 discusses the interactions with Condition C. Section 6 sums
up the results.
2
2. Bresnan’s observation
Bresnan (1971, 1972) employed the Nuclear Stress Rule (NSR) (Chomsky & Halle 1968)
which (leaving aside cyclic effects) assigns stress rightmost as in (1a). As Bresnan notes,
final pronouns must be exempt from the NSR as in (1b,c). This paper employs the term
sentence stress (nuclear stress) and marks its position by double underlining.
(1)
a. Mary teaches engineering.
b. Mary teaches it.
c. The boy bought some.
Throughout this paper the concern is with stress-patterns that are not affected by narrow
focus, since narrow focus on any of the elements in (1) will attract the sentence stress to the
focused item. The sentences in (1) and other sentences we will be concerned with may be
thought of as entirely new. They are focused in their entirety, or as not carrying focus at all.
Bresnan's crucial observation is another class of systematic exceptions to the NSR, next
to pronouns. This involves cases of syntactic movement from sentence-final position. She
credits (2) to Newman (1946). The a.-examples of (2) – (7) are the exceptions from the
NSR. These all involve movement from final position. In Bresnan's analysis the structure
before movement involved a lexical object that would receive the sentence stress, for example to leave plans in (2a) or had written what books in (6a). The (b)-examples are cases for
minimal comparison that do not involve this configuration and that show the final stress that
is predicted by the NSR. In (2) and (3) the comparison cases do not involve movement from
final position. In (4) – (7) the comparison cases are examples with movement of a pronominal element from final position. For example, in (6b) the underlying structure is had written
what. Here it is plausible that the final pronoun, like the ones in (1b,c), is independently
exempt from the NSR. Thus movement only interferes with the NSR if the element that
moves from final position would receive sentence stress before movement, as in the (a)examples in Bresnan's analysis.
(2)
a. George has plans to leave t.
b. George has plans to leave.
(3)
a. Mary liked the proposal that George left t.
b. Mary liked the proposal that George leave.
(4)
a. George found some friends he'd like you to meet t.
b. George found someone he'd like you to meet t.
3
(5)
a. Let me tell you about something strange I saw t.
b. Let me tell you about something I saw t.
(6)
a. John asked what books Helen had written t.
b. John asked what Helen had written t.
(7)
a. I can't help noticing how serene he is t.
b. I can't help noticing how he is t.
Following Truckenbrodt & Büring (in preparation) the effect will be cast in terms of stress
reconstruction in this paper. The following formulation is a starting point that brings the
relevant descriptive generalization into view:
(8)
Stress reconstruction
For the rules of stress-assignment, stress on an element coindexed with a trace
counts as stress in the position of the trace.
The examples (2) – (7) are repeated in (9) – (14) with coindexing and bracketing of the
coindexed element. Op stands for an empty relative pronoun operator. This is coindexed
with the head NP preceding the clause in the relative clause examples (10) – (13). Even
though this head NP is outside of the relative clause on most analyses, its stress reconstructs
into the trace position in the relative clause. The wh-questions in (13) and (14) involve only
wh-movement.
(9)
a. George has [[plans]1 to leave t1].
b. George has plans to leave.
(10) a. Mary liked the [proposal]1 Op1 that George left t1.
b. Mary liked the proposal that George leave.
(11) a. George found some [friends]1 Op1 he'd like you to meet t1.
b. George found [someone]1 Op1 he'd like you to meet t1.
(12) a. Let me tell you about something [strange]1 Op1 I saw t1.
b. Let me tell you about [something]1 Op1 I saw t1.
(13) a. John asked [what books]1 Helen had written t1.
b. John asked [what]1 Helen had written t1.
(14) a. I can't help noticing [how serene]1 he is t1.
b. I can't help noticing [how]1 he is t1.
As (8) brings out, in all cases of non-final stress, the stress that should be in final position
by the NSR is instead found on an element earlier in the clause that is coindexed with the
final trace.
4
Bresnan's analysis of her observation involves a cyclic application of the NSR, which in
her time was formulated in terms of weakening of the non-final elements. The NSR thus
applied to a fully stressed structure akin to (15a). At that stage there is still a copy of the
relevant element in final position. In the first application of the NSR this final element retains its strength while the preceding elements of the first cycle are weakened as in (15b).
The final object is made non-overt in (15c), leaving all remaining material from the first
cycle with subordinated stress. When the NSR reapplies at the highest cycle in (15d), it
reduces everything except for the rightmost element among the strongly stressed elements,
and thus derives the correct position of strongest stress.
(15) a.
b.
c.
d.
Helen left directions [for George to follow directions]
Helen left directions [for George to follow directions]
Helen left directions [for George to follow Ø]
Helen left directions [for George to follow Ø]
NSR 1st cycle
syntax
NSR 2nd cycle
3. Aspects of the effect and its analysis in German
In this section different aspects of the effect and its analysis in German are established.
Section 3.1. shows an analysis of the effect in the context of a general prosodic analysis of
German. Section 3.2. shows a restriction on the effect. Section 3.3. establishes the effect in
some detail in German relative clauses. Section 3.4. discusses how stress reconstruction
may fit with Chomsky’s copy theory of movement and reconstruction. Section 3.5. shows
the possible relevance of the head matching analysis of relative clauses to stress reconstruction.
3.1. The effect in German in a cross-linguistic prosodic analysis
In German, the effect plays out in the context of stress placement on a clause-final verb. It is
1
here illustrated using the theory of stress-assignment in (16) and (17). The account is a
1
This account builds on the two-level theory of Selkirk 1984, 1995b and Gussenhoven 1983, 1992
and ties this theory to the use of XPs in the prosodic structure of other languages (Chen 1987;
Hale & Selkirk 1987; Selkirk 1986; Selkirk & Shen 1990; Truckenbrodt 1999, 2006, 2007a). The
account was previously applied to the interaction of movement and stress in Truckenbrodt &
Darcy 2010. Stress-XP is from Truckenbrodt 1995 and was also applied to Italian in Samek-
5
modification of that of Gussenhoven (1983, 1992) which was first applied to German in
Uhmann (1991).
(16) Stress-XP: Each lexical XP must contain a beat of phrasal stress (accent).
(17) NSR-I: The rightmost phrasal stress (accent) in the intonation phrase is strengthened.
There are two levels of prosody-assignment above the prosodic word. At the lower of these
two levels, Stress-XP requires a beat of phrasal stress in each lexical XP. This is here
marked by single underlining. Consider (18). The arguments and adjuncts will typically
contain such XPs, as in [DP die [NP Lena]] 'Lena', [DP ein [NP Lama]] or [PP im [DP [NP Januar]]], where the NP is the lexical XP in these examples. These elements are therefore assigned phrasal stress by Stress-XP as in (18a). The NSR-I in (17) strengthens the rightmost
of these. This is shown by double underlining in (18b).
(18) a. Die Lena will dem Werner im Januar ein Lama malen.
b. Die Lena will dem Werner im Januar ein Lama malen.
the Lena wants the Werner in January a llama paint
‚Lena wants to paint a llama for Werner in January.’
Here and throughout this paper stress in an English translations, where indicated, shows the
stress assigned by the same rules and same analytical assumptions, rather than word-byword transfer of stress from the German example. Stress-XP and NSR-I (like Gussenhoven's
account on which they build) are intended to also work for English, without parameterization. Stress reconstruction applies in both languages. Stress effects of focus and givenness
are likewise comparable.
In German an unmarked rendition of (18b) shows typically rising (L*+H) accents on
2
non-final syllables with phrasal stress, as in Figure 1.
The claim that the final accent is the strongest of the sentence is based primarily on the
intuitions of native speakers and connects this account to earlier reports of sentence stress
3
that did not employ the two-level model (e.g. Höhle 1982b, Cinque 1993).
Lodovici 2005 and to English in Féry & Samek-Lodovici 2006. The NSR-I desends from the NSR
of Chomsky & Halle 1968 and the understanding of nuclear stress as being relative to the intonation phrase in Pierrehumbert 1980. For the formulation that strengthens the last accent of the intonation phrase, see Uhmann 1991 for German and Selkirk 1995b for English. In its application to
German the account builds on, and captures results of early important work like Höhle 1982a,
Krifka 1984, von Stechow & Uhmann 1986, Uhmann 1991 and Jacobs 1993.
2
The successive phonetic lowering among the peaks (downstep) is a phonetic process that does not
reflect relative strength of stress, see e.g. Truckenbrodt 2007b.
3
On the phonology and phonetics of this final accent with Southern German speakers, see
Truckenbrodt 2007b. The fact that it is falling in Figure 1 is an inherent feature of the strongest stress
of an intonation phrase but ocurs only where this sentence-stress is also utterance-final.
6
Die
Le-na
L*+H
will dem
Wer-ner
L*+H
im
Ja-nuar ein
L*+H
La-ma malen.
H+L*
Figure 1. F0 contour for Die Lena will dem Werner im Januar ein Lama malen, ‘Lena
wants to paint a llama for Werner in January.’ Speaker TL from Baden-Württemberg. The
sentence was read as an answer to the question Was gibt’s Neues? ‘What’s new?’. Adapted
from Truckenbrodt (2007b).
The crucial issue for stress reconstruction is when the verb receives stress. Let us therefore begin by establishing in some detail when it receives stress independently of stress
reconstruction. The verb does not receive stress when preceded by an (unscrambled) direct
object, as in (19a). However, it receives stress when preceded by an adjunct as in (19b). See
Gussenhoven (1983, 1992, Krifka (1984) and Jacobs (1993) on argument-adjunct contrast.
(19) a. sein Fahrrad putzen
ones bicycle clean
‚to clean ones bicycle’
b. am Wochenende putzen
on-the weekend clean
‚to clean on the weekend’
In the account of this distinction I adopt the classical view of the unmarked word order
‚adjunct before direct object’ in the German head-final structure in (20a): The adjunct is not
genuinely inside of VP, the direct object is genuinely inside of VP for reasons of theta-role
assignment, and must therefore be closer to the verb. A structure with only the object is
therefore as in (20b), a structure with only the adjunct as in (20c).
(20) a. am Wochenende
b.
c. am Wochenende
on-the weekend
[VP sein Fahrrad
[VP sein Fahrrad
[VP
ones bicycle
putzen]
putzen]
putzen]
clean
‚to clean (ones bicycle) (on the weekend)’
The stress contrast between arguments and adjuncts is captured in the application of StressXP to VP. In (20a,b) the object receives phrasal stress independently on [NP Fahrrad]. Now,
this phrasal stress also has the effect that the VP satisfies Stress-XP, since the word Fahrrad
7
is within the VP. Therefore the VP contains phrasal stress in (20a,b) and there is no need to
assign phrasal stress to the verb. (It is not detrimental that the same phrasal stress satisfies
Stress-XP for NP and VP, and the effect is also not cumulative in this account.) In (20c),
4
however, the phrasal stress assigned on the adjunct is not genuinely within the VP. The
application of Stress-XP to VP requires phrasal stress genuinely inside of VP, which is
therefore assigned to the verb. (The effects of the subsequently applying NSR-I are not
shown in (20). It strengthens the object in (20a,b) and the verb in (20c).) Another way of
putting this analysis of the argument-adjunct asymmetry is that the verb does not invoke
Stress-XP in (20a,b) where it is just a syntactic head, but that the verb invokes Stress-XP in
(20c), where it is a VP, hence an XP.
A further relevant observation holds for German default verb stress (with the exception
of the interaction of stress-assignment with movement that is put aside for now). In the
configuration [argument pronoun verb] in wide focus contexts, the verb always requires
stress of its own, in addition to stress on the argument. In other words, an argument separated from the verb by an overt pronoun cannot exempt the verb from requiring stress. (21)
shows this for indefinite pronouns:
(21) a.
dass vorhin ein Kind etwas essen wollte
b. # dass vorhin ein Kind etwas essen wollte
(unless essen is given)
that earlier a child something eat wanted
‚that a while ago a child wanted to eat something’
c.
dass du einem Kind etwas vorgelesen hast
d. # dass du einem Kind etwas vorgelesen hast
that you a child something read have
‚that you have read something to a child’
(unless vorgelesen is given)
The analysis of this restriction requires a syntactic structure like (22) in which the direct
object and the verb are joined in a VP that excludes higher arguments. (These might be in
higher verbal projections.) The verb heads that VP and may not raise to a higher position in
the presence of the direct object (see Kratzer & Selkirk 2007, Truckenbrodt & Darcy 2010).
In such a structure the stressed argument is outside of the VP that contains the verb. Stress
on that argument is not stress inside of VP. Satisfaction of Stress-XP for the VP requires
phrasal stress in the VP. It is assumed here, as in Bresnan's account, that pronouns are inherently stress-rejecting. The phrasal stress of the VP is then assigned to the verb.
(22) dass vorhin ein Kind [VP etwas essen] wollte
that earlier a child something eat wanted
The restriction on intervening pronouns also holds for definite object pronouns. Personal
pronouns arguably move to a higher position, so that arguments preceding them are all the
more outside of VP:
4
The account assumes that theta-role assignment and Stress-XP use the same strict standard for inclusion in the VP. Assuming that the adjunct is adjoined to VP, it counts as outside of VP for both theta-role and stress assignment.
8
(23) a.
dass das Kind es essen wollte
b. # dass das Kind es essen wollte
that the child it eat wanted
‚that the child wanted to eat it’
(unless essen is given)
A further relevant case is that of multiple final verbs. A stressed object of the inner verb
exempts all following verbs from requiring accent as in (24a). Where no such stressed object is present, as in (24b), the lowest verb receives phrasal stress. The current analysis assumes successively embedded VPs.
(24) a. [VP1 [VP2
ein [NP Fahrrad] reparierenV2] wollenV1]
a
bicycle
fix
want
‚to want to fix a bicycle’
b. [VP1 [VP2
etwas
reparierenV2] wollenV1]
something
fix
want
‚to want to fix something’
[NP Fahrrad] receives stress by Stress-XP in (24a). Stress on Fahrrad satisfies Stress-XP
also for VP2 and for VP1, since both contain the word Fahrrad. VP2 in particular (and the
word Fahrrad with it) is contained in VP1 as the complement of V1. Therefore the verbs do
not separately receive stress in (24a). In (24b) reparieren receives the phrasal stress of VP2.
This being the case, reparieren also satisfies Stress-XP for VP1 since VP1 then also contains
the stressed word reparieren.
With the basics of this stress-account in place, let us turn to the interaction of stressassignment and movement in German. The effect is found with wh-movement as in (25) and
with relative clauses as in (26). Note that all German relative clauses have an overt relative
pronoun.
(25) a. Ich frage mich [was für Bücher1 sie t1 geschrieben hat].
I ask myself what for books she written has
‚I wonder what books she has written.’
b. Ich frage mich [was1 sie t1 geschrieben hat].
I ask myself what she written has
‚I wonder what she has written.’
(26) a. Ich kenne alle Bücher1 [die1 sie t1 geschrieben hat].
I know all books which she written has
‚I know all the books she has written.’
b. Ich kenne alles [was1 sie t1 geschrieben hat].
I know all which she written has
‚I know everything she has written.’
9
In all four cases the VPs are [VP1 [VP2 t1 geschriebenV2] hatV1]. We find the expected stresspattern for these VPs in (25b) and (26b): Both VP2 and VP1 require stress by Stress-XP and
they both receive this by assignment of phrasal stress on the lower verb geschrieben.
The unexpected cases are (25a) and (26a). They show interaction with syntactic movement. We expect stress on the inner verb by Stress-XP; empirically, however, it is not required. In both cases Bücher is related to the trace in the VPs [VP1 [VP2 t1 geschriebenV2]
hatV1]. It is as though the stress on Bücher counts as stress in the position of the trace. In
other words, stress in the structure (27) counts as though Bücher was in the position of the
trace as in (28). In (28) the VPs contain stress on Bücher and Stress-XP does not require
stress on the verb.
(27) [ ... Bücher]1 (...) sie [VP1 [VP2 t1 geschriebenV2] hatV1]
books
she
written
has
(28) sie [VP1 [VP2 [ ... Bücher] geschriebenV2] hatV1]
she
books written
has
In that sense that we can think about this phenomenon as a case of stress reconstruction.
Notice that the argument cannot be undermined by postulating that the expression containing Bücher can have this effect from its surface position, without the mediating effect of
the trace. Empirically, as we have seen, such a stress-exempting effect is not normally possible across a pronoun. In the analysis, it is also ruled out: In (25a) for example the whphrase is in Spec,CP and thus clearly outside of the VP to which it contributes its stress.
In sum, we account for the assignment of phrasal stress and sentence stress in German in
terms of Stress-XP and NSR-I (rightmost strengthening). Narrow focus may override their
effects (though this is not discussed here). As in Bresnan's English cases, syntactic movement creates a class of exceptions to this default stress pattern. These exceptions can be
looked upon in terms of stress reconstruction: Stress-XP is satisfied for a VP (allowing a
stressless verb) when there is a trace in direct object position that is coindexed with a
stressed element earlier in the structure. For the purpose of Stress-XP it is as though this
earlier stress was within the VP.
Notice that Bresnan's English examples are analyzed in very similar terms in the current
account. In Mary teaches engineering, Stress-XP assigns stress to the two NPs Mary and
engineering and NSR-I strengthens the rightmost of these. In Mary teaches or Mary teaches
it, stress on the verb is required by the application of Stress-XP to the VP. Likewise for the
interaction with movement. In [what books1 she has written t1] the unexpected observation
is that the VP [written t] does not show the stress expected due to Stress-XP. Stress reconstruction is observed insofar the stressed what books acts prosodically as thought it was still
inside of that VP, allowing a stressless verb.
3.2. A restriction on the effect
The effect shows an interesting prosodic restriction, which is mentioned here for completeness. The restriction is that stress reconstruction is blocked by an intervening element with
phrasal stress. In Bresnan's cases stress reconstruction obtains in (11a), (12a) and (14a)
10
across intervening unstressed pronouns and unstressed verbs. In (10a) and (13a) it obtains
across a name, which, being an NP, would normally carry phrasal stress by Stress-XP. As
mentioned by Selkirk (1995b) stress reconstruction in the sense of the current paper only
works in these cases if the name is contextually given and thus stressless. Also the detailed
observations of Gussenhoven (1983, 1992), who seeks an account without stress reconstruction, show nicely that the effect only obtains in the absence of intervening stressed (accented) material. The examples in (29) illustrate the blocking effect. In (29a) we see stress
reconstruction across an unstressed pronoun. In (29b) there is an intervening stressed element, unsere Kollegen 'our colleagues'. It receives phrasal stress by Stress-XP and it is
stressed by NSR-I. If there was an unobstructed effect of stress reconstruction, the stress on
was für Bücher should still reconstruct into the VP and license stressless final verbs. Empirically, however, this stress-pattern is not a neutral stress-pattern. It is possible only where
unsere Kollegen 'our colleagues' is contrastive or where ausgeliehen 'borrowed' is contextually given. (29c) is the only possible neutral stress-pattern in this case. It is a stress-pattern
without stress reconstruction: The VP contains stress on the verb, as required by Stress-XP
and regardless of the trace it contains.
(29) a. Ich frage mich [was für Bücher1 sie t1 ausgeliehen haben].
I ask myself what for books she checked-out has
‚I wonder what books she has checked out.’
b. # Ich frage mich [was für Bücher1 unsere Kollegen t1 ausgeliehen haben].
I ask myself what for books our colleague checked-out have
‚I wonder what books our colleagues have checked out.’
c. Ich frage mich [was für Bücher1 unsere Kollegen t1 ausgeliehen haben].
I ask myself what for books our colleagues checked-out have
‚I wonder what books our colleagues have checked out.’
Cases like this motivate the conclusion that an intervening stressed element like unsere
Kollegen 'our colleagues' in (29b) block the effect of stress reconstruction.
An analysis of this blocking effect is developed in Truckenbrodt and Büring (in preparation). The analysis is compatible with, and orthogonal to the current paper. It involves (a) an
effect of stress reconstruction of some kind or other, like the suggestions discussed in this
paper and (b) additional restrictions on the syntax-prosody mapping, which are not crucial
for other purposes in the current paper. The relevance of this restriction for the current paper
is thus only that we need to choose examples in which the intervening material is stressless.
3.3. The effect in German relative clauses
This section provides examples of stress reconstruction in German relative clauses to solidify the existence of this phenomenon in relative clauses. An analysis is provided in later
sections.
Examples of German stress reconstruction in relative clauses are shown in (30).
11
(30) a. Dort drüben liegt ein Buch1 [das1 ich t1 mitgebracht habe].
there yonder lies a book which I brought have
‚Lieing over there there is a book that I brought.’
b. Peter kennt jedes Buch1 [das1 du t1 geschrieben hast].
Peter knows every book which you written have
‚Peter knows every book that you have written.’
c. Das ist die Schublade1 [in die1 ich es t1 gelegt habe].
that is the drawer in which I it put have
‚That is the drawer in which I put it.’
It is remarkable that stress reconstruction allows entire relative clauses that are contextually
new but nevertheless stressless. The stress on the head noun is reconstructed into the VP of
the relative clause. This allows these examples to show this stress-pattern without contextual givenness of the verbs in the relative clause. If it were not for stress reconstruction, the
verbs in the relative clause would need to be stressed due to the application of Stress-XP to
their VPs.
Further examples are shown in (31a), (32a) and (33a), with minimal comparisons in the
b.- and c.-examples. The b.-examples do not have a lexical head noun that would carry
stress and could thus provide the basis for stress reconstruction. With no stress to reconstruct, we find the expected stress on the verbs of the relative clause. (31c) and (32c) show
that it won't do to stress the first lexical word (noun) that precedes the relative clause. Rather, the stress really needs to be on a head noun that can reconstruct for stress into the rela5
tive clause.
(31) a.
Hier ist eine Liste einiger Bücher [die ich t verkauft habe].
here is a list some books’ which I sold have
‚Here is a list of some books I sold.’
b.
Hier ist eine Liste von dem [was ich verkauft habe].
c. # Hier ist eine Liste von dem [was ich verkauft habe].
here is a list of that which I sold have
‚Here is a list of that what I sold.’
(32) a.
Hier ist eine Liste der Gäste [die ich t gewinnen konnte].
here is a list the guests’ which I win could
‚Here is a list of the guests I could to win.’
b.
Hier ist eine Liste derer [die ich t gewinnen konnte].
c. # Hier ist eine Liste derer [die ich t gewinnen konnte].
here is a list those which I win could
‚Here is a list of those that I could win.’
5
(31c) and (32c) would wrongly be allowed by Bresnan’s analysis that is illustrated in (15).
12
(33) a.
b.
Nenn mir bitte einige Papiere [die du magst].
name me please some papers which you like
‚Please name some papers that you like.’
Nenn mir bitte einige [die du magst].
name me please some which you like
‚Please name some that you like.’
The stress reconstruction pattern emerges with additional clarity in the context of narrow
focus on the DP that contains the relative clause, as in (34). Here the narrow focus on the
complex DP is a way of controlling that focused material, including the relative clause, is
new at least in this explicit context.
(34) [What is lying over there?]
Dort drüben liegt
[ ein Buch [das ich t mitgebracht habe] ]F
there over-there lies a book that I brought have
‚A book that I brought is lying over there.’
Examples very similar to the a.-examples in (35) – (37) are employed in the a.-examples in
(31) – (33), though they are here presented in the context of narrow focus on the complex
DP. This helps sharpen the contrasts with the b./c.-examples in (35) – (37). In the b./c.examples the trace in the relative clause is separated from the verb by the indefinite pronoun
etwas. Stress reconstruction does not affect the stress of the verbs in these cases. The site of
reconstruction, separated from the verb by the indefinite pronoun, could not exempt the
verbs from carrying stress (cf. (21)). Therefore the stress pattern without verb stress in the
relative clauses in the b.-examples is not possible. Stress is required on the verbs in the
relative clause, as in the c.-examples.
(35) [Of what is this a list?]
___________focus________________
a.
Es ist eine Liste einiger Bücher [die ich t verkauft habe].
it is a list
some books which I sold have
‚It is a list of some books that I sold.’
___________focus______________________
b. # Es ist eine Liste einiger Bücher [die t etwas eingebracht haben].
c.
Es ist eine Liste einiger Bücher [die t etwas eingebracht haben].
it is a list
some books which something brought have
‚It is a list of some books that brought us some (money).’
13
(36) [Of what is this a list?]
_____________focus_____________
a.
Es ist eine Liste der Gäste [die ich t gewinnen konnte].
it is a list
some guests which I win could
‚It is a list of some guests that I could win.’
_____________focus______________
b. # Es ist eine Liste der Gäste [die t etwas gespendet haben].
c.
Es ist eine Liste der Gäste [die t etwas gespendet haben].
it is a list
the guests which something donated have
‚It is a list of the guests that donated something.’
(37) [What do you want me to name?]
___________focus_________
a.
Nenn mir bitte einige Papiere [die du t magst].
name me please some papers which you like
‚Please name some papers that you like.’
___________focus______________
b. # Nenn mir bitte einige Papiere [die dir t etwas sagen].
c.
Nenn mir bitte einige Papiere [die dir t etwas sagen].
name me please some papers that you something say
‚Please name some papers that tell you something.’
Notice also that the stress pattern that shows stress reconstruction is typically optional
(Truckenbrodt & Darcy 2010). What is interesting about it is that it is available at all in the
presence of a contextually new VP. However, a stress-pattern that looks like stress is not
reconstructing is typically available in addition. This is true both for wh-questions and relative clauses. For example, next to (37a), an alternative is (38).
(38) [What do you want me to name?]
_________focus____________
Nenn mir bitte einige Papiere [die du t magst].
name me please some papers which you like
‚Please name some papers that you like.’
An analysis of this optionality is developed later in this paper.
In sum, there is good evidence that stress from the head noun of a relative clause can
reconstruct into the relative clause. Where it reconstructs into the VP of the relative clause,
it satisfies Stress-XP for the VPs of the relative clause and the verbs do not require additional stress. The relative clause is then stressless but new in its entirety.
3.4. An approach to stress reconstruction in the copy theory
Consider reconstruction for reflexive binding as in (39a). Chomsky (1993:34ff) suggested to
analyze reconstruction of wh-movement in terms of a copy theory of movement and reconstruction. An underlying structure for the embedded question before movement is shown in
14
(40a). Movement creates a copy of the wh-phrase in clause-initial position as in (40b), the
structure at spell-out (the branching point to LF and PF). Only the upper copy is spelled out
at PF as in (40c). At LF the operator which is separated as in (40d) and either the lower copy
is retained as in (40e) or the upper copy as in (40f). The sentence in (39) allows two options
of binding the reflexive as shown. Binding theory applies at LF, so that a lower copy at LF
in (40e) leads to Bill as the antecedent of himself. Retention of the upper copy as in (40f)
leads to John as the antecedent of himself in (39).
(39) John1 wondered [which picture of himself1/2] Bill2 saw
(40) a.
b.
c.
d.
e.
or f.
[which picture of himself]
PF:
[which picture of himself]
LF: which x [x picture of himself]
→ which x
→ which x [x picture of himself1]
Bill saw [which pictures of himself]
Bill saw [which pictures of himself]
Bill saw
Bill saw [x pictures of himself]
Bill2 saw [x pictures of himself2]
Bill2 saw t
This copy theory of movement and reconstruction is appealing insofar it makes reconstruction sit comfortably in the theory of syntax. It is assumed for independent reasons that the
wh-phrase originates in object position (in this case). The copy theory of movement and
reconstruction retains a silent copy of this original structure and employs it to account for
‚reconstruction’ effects of different kinds like reflexive binding in the example (39). The
syntactic account of reconstruction is strengthened by phenomena in which reconstruction
for one phenomenon entails reconstruction for another phenomenon, as shown in Chomsky
(1993), Heycock (1995), Fox (1999) and others.
The analysis of stress reconstruction would seem to be particularly straightforward given
a cyclic derivation in terms of vP and CP phases suggested in Chomsky (2008). For each
phase, elements that have moved to its edge are ignored; the remaining phase constitutes the
spell-out domain, which is processed at PF and at LF before further structure is added at the
top of the phase. (40b) could here also stand for a representation at the vP phase where the
wh-phrase has left-adjoined to the edge of the phase and the remainder, Bill saw [which
pictures of himself] is sent to PF and LF. At PF, Stress-XP would apply, stressing the object
but not the verb. We might assume that the NSR-I applies as well, assigning sentence stress
to the object. These properties, a stressless verb and an object with sentence stress, could
then be retained though the derivation.
In this implementation, it would be important that the wh-object does not loose its syntactic structure at PF. Instead, it would be reasonable to adopt the architecture assumed in
prosodic phonology for independent reasons (see e.g. Selkirk 1995a, Truckenbrodt 2007),
namely that syntax and prosody are co-present at PF. This would be important to allow
upper and lower copy of the wh-phrase to stand in a chain relation upon the return from PF:
The lower copy must have serveed for the computation of stress (exempting the verb from
being stressed), the upper copy must inherit the stress of the object. The chain-relation is
furthermore relevant for later cycles at PF where the grammar must spell out the upper copy
and leave the lower copy silent. This theoretical choice is related to what forces cyclic
movement to the edge of the phase. Chomsky’s idea is that this is required because of what
happens to the rest of the phase at PF. His implementation of this idea is that loss of the
15
syntactic structure at PF requires cyclic movement to the edge of the phase. If we want to
integrate stress reconstruction into this picture, we may be interested in alternative PF processes (alternatives to loss of syntax) that force cyclic movement to the edge of the phase.
Stress reconstruction would be compatible with a minimally modified version of the idea of
Fox & Pesetsky (2005). They suggest that linearization is crucial for forcing cyclic movement. While they see linearization as a phenomenon of syntax with consequences for PF, it
might also be a PF phenomenon.
Overall, stress reconstruction would seem to fit well into the copy theory of movement
and reconstruction of Chomsky (1993), and in the architecture of derivation by phase of
Chomsky (2008). A minor adjust that would seem to be needed is the effect of PF on the
syntactic representation. It must of course supply stress-assignment. Next to that, it would
need to involve co-presence of syntax and phonology, as assumed in prosodic phonology. If
it supplies linearization (Fox and Pesetsky 2005), it could still force successive-cyclic
movement to the edge of the phase.
3.5. Approach to stress reconstruction in relative clauses
I concur with Salzmann (2006, this volume) on the overall assessment of theories about
reconstruction in relative clauses:
(41) a. head-external analysis: Buch [das1 ich das1 gekauft habe]
b. head raising analysis:
Buch1 [das ich Buch1 gekauft habe]
c. head matching analysis: Buch1 [[das Buch]1 ich [das Buch]1 gekauft habe]
NP-matching DP-movement
(a) The classical head-external analysis assumes that a relative pronoun moves alone and
semantically forms a predicate that is set-intersected with the head NP, in (41a) Buch
‚book’. As Salzmann points out, this does not lead to an interesting syntactic theory of reconstruction in relative clauses. It is, however, compatible with a semantic account of reconstruction as shown in Krifka (this volume) and Grosu & Krifka (2008).
(b) An alternative is the head-raising analysis of Vergnaud (1974). In this analysis the
head NP raises from inside of the relative clause to the relative-clause external position.
(c) The head matching analysis of Munn (1994), Citko (2001), Hulsey & Sauerland
(2006), and Salzmann (2006), this volume is illustrated in (41c). The relative pronoun is an
article with an NP complement, i.e. a regular DP that fronts in the relative clause. Its trace is
silent like other traces are silent. Furthermore, a matching relation is established between
the external head NP and the NP of the fronted DP. This is a relation similar to a chain created by movement insofar identity is required and the lower of the two elements is not
spelled out.
Salzmann (2006, this volume) points out that the head-matching analysis provides a
restrictive perspective on syntactic reconstruction in relative clauses. For analyzing stress
reconstruction in these terms, we require that the NP matching process shares with copies
created by movement that the upper copy that suppresses the spelling out of the lower copy,
spells out the stress of the lower copy. That seems to be a very natural assumption.
16
4. Obigatory stress reconstruction with more opaque idioms
It seems very likely that stress reconstruction is a consequence of a mechanism of syntactic
construction like that suggested in Chomsky (1993) in terms of the copy theory of movement. This section and the following section explore interactions of stress reconstruction
with syntactic reconstruction to assess whether (a) there is independent evidence for this
hypothesis (b) whether the facts are compatible with this option or (c) whether there is evidence against it.
6
This section discusses stress reconstruction in connection with idiom chunks. A class of
idioms here called more opaque idioms will be seen to require stress reconstruction. It is
suggested that more opaque idioms resist overlaid information structure and that the apparent optionality of stress reconstruction in other cases comes from the option of overlaying a
topic-comment structure. Stress reconstruction is then suggested to be obligatory on the
surface with more opaque idioms because these idioms do not allow the overlaid topiccomment structure. However, we will not be able to determine whether stress reconstruction
is obligatory because syntactic reconstruction for the idiom is obligatory, or whether the
mechanism underlying stress reconstruction is more generally obligatory.
4.1. More opaque idioms: observations
Putting a chunk of an idiom into a wh-phrase or into the head NP of a relative clause requires some amount of transparency of the idiom such that that chunk can be referential or
semi-referential (see below). Idioms differ in the degree in which they allow this.
(42) illustrates the more opaque idiom jemanden sticht der Hafer lit. ‚the oat is stinging
someone’, in my dictionary translated with the English idiom ‚someone is feeling their
oats’. The German idiom consists of the nominative subject (der Hafer ‚the oat’) and the
verb (stechen ‚stings’). The accusative object obligatorily precedes the subject as in (42a).
The idiom requires stress on the preverbal subject der Hafer ‚the oat’ as shown. (42b) is
impossible on the idiomatic reading. I assume that it is ruled out because it would require
contrast on the following verb. However, since this verb is part of the idiom, such contrast is
not possible.
6
Effects of reconstruction of idiom chunks has been described in the literature for both whmovement and relative clauses (Chomsky 1993, Munn 1994, among others).
17
(42) a.
Ich glaube, dass dich der Hafer sticht.
b. * Ich glaube, dass dich der Hafer sticht.
I think that you the oat stings
‚I think that you are feeling your oats.’
This idiom is generally marked when the oat is questioned as in (43a). To the extent that this
is acceptable the reference of the oat in the wh-phrase is the impulse alluded to by the idiom. The oat may be relativized in what we may call a bleached reading as in (44a). In this
bleached reading the oat might stand as much for the impulse as for the situation of having
7
such an impulse, i.e. possibly for the meaning of the entire idiom. Importantly, both cases
require stress reconstruction as in (43a) and (44a) for idiom chunk reconstruction. The stress
patterns without stress reconstruction in (43b) and (44b) are not compatible with idiom
chunk reconstruction.
(43) a. ?
Ich frage mich welcher Hafer dich sticht.
b. ??/* Ich frage mich welcher Hafer dich sticht.
I ask myself which oat you stings
lit. ‚I wonder which oat is stinging you.’
(44) a.
b. ??
Wir kennen alle den Hafer der dich sticht.
Wir kennen alle den Hafer der dich sticht.
we know all the oat which you stings
lit. ‚We all know the oat that stings you.’
Similarly with the idiom x ist eine Laus über die Leber gelaufen lit. ‚a louse has run
across x’s liver’, translated in my dictionary as something is biting x. The idiom consists of
the nominative subject (eine Laus ‚a louse’), a PP (über die Leber ‚across the liver’) and the
verb (laufen ‚walk/run’). The dative object, denoting the person to whom this happened,
obligatorily precedes the subject. The idiom can only be stressed on the subject as in (45a),
not on the PP as in (45b). When the idiom is put in a wh-question as in (45) and in a relative
clause as in (46) stress reconstruction is required as in the a.-examples. A stress pattern
without stress reconstruction is not available, as seen in the b.-examples
(45) a.
Ihm ist eine Laus über die Leber gelaufen.
b. ?? Ihm ist eine Laus über die Leber gelaufen.
him is a louse across the liver walked
‚Something is biting him.’
7
The bleaching referred to here is not a reduction of the idiomatic reading, which is fully intact in the
acceptable examples. Bleaching here refers only the extent to which the idiom chunk is construed
as referential.
18
(46) a.
Ich frage mich was für eine Laus ihm über die Leber gelaufen ist.
b. ?? Ich frage mich was für eine Laus ihm über die Leber gelaufen ist.
I ask myself what for a louse him across the liver walked is
lit. ‚I wonder what kind of louse walked across his liver.’
‚I wonder what is biting him.’
(47) a.
Wir wundern uns über die Laus die dir über die Leber gelaufen ist.
b. ?? Wir wundern uns über die Laus die dir über die Leber gelaufen ist.
we wonder ourselves about the louse which you across the liver walked is
lit. ‚We are wondering about the louse that walked across your liver.’
‚We are wondering what is biting you.’
Further idioms that require stress reconstruction for the idiomatic reading are illustrated in
(48) – (50). In these examples the direct object is the first element of the idiom and requires
the stress in the idiom. As in the preceding cases stress reconstruction is required for idiom
chunk reconstruction.
(48) a.
Es gibt da noch ein Hühnchen das ich mit ihm zu rupfen habe.
b. ?? Es gibt da noch ein Hühnchen das ich mit ihm zu rupfen habe.
it exists there also a chicken that I with him to pluck have
lit. ‚There is still a chicken that I have to pluck with him.’
‚There is still a bone that I have to pick with him.’ (dictionary translation)
‚There is still a matter that I have to discuss with im/scold him for.’
(49) a.
Wir staunen alle über die Eulen, die du nach Athen trägst.
b. ?? Wir staunen alle über die Eulen, die du nach Athen trägst.
we maze all about the owls which you to Athens carry
lit. ‚We are all amazed about the owls you are carrying to Athens.’
‚We are all amazed about the coals you are carrying to Newcastle.’ (dictionary)
‚We are all amazed about this superfluous thing you are doing.’
(50) a.
Nicht von ungefähr kommt der Frosch den du im Hals hast.
b. ?? Nicht von ungefähr kommt der Frosch den du im Hals hast.
not from broadly comes the frog which you in-the neck have
‚No accident is the frog you have got in your neck.’ (dictionary translation)
‚It is no accident that you have difficulties speaking.
4.2. More opaque idioms: analysis
I turn to some analysis of these more opaque idioms. All their stress-patterns are arguably
default stress patterns in German. Because of the idiom status the stress pattern is here not
distorted by information structure. The structurally simplest case is the idiom in (48), which
is analyzed in (51). Stress-XP requires stress in the NP Hühnchen ‚chicken’ of the object,
which also does duty for both VPs that contain this NP. Both verbs are part of the idiom.
19
(51) mit jemandem [VP [VP ein Hühnchen zu rupfen] haben ]
with someone
a chicken to pluck have
Where the idiom includes the subject, I assume raising of the verb from V to v in the absence of an overt DP object, following Kratzer & Selkirk (2007) and Truckenbrodt (2012).
This is shown in (52). The NP Hafer receives stress by Stress-XP which also satisfies
Stress-XP for vP. The verb therefore does not receive stress. VP is exempt from Stress-XP
because it is phonetically empty (Truckenbrodt 1999, 2012).
8
(52) (dass) jemanden1,ACC [vP der Hafer [VP t1 t2] sticht2 ]
(that) someone
the oat stings
It is interesting that German has a range of idioms that involve the subject but not the
object (the accusative jemanden ‚someone’ in (52)) and that the object obligatorily scrambles out of the idiomatic vP in these cases. We can account for the obligatory scrambling in
terms of two assumptions: (a) as suggested by Chomsky (1993:39) an idiom needs to be a
unit at LF and (b) scrambling the object out of the idiom is not an option at LF because sstructure and LF generally match in German (movement to the left periphery apart), see
Diesing (1992. Chomsky suggests (a) as the motivation for idiom chunk reconstruction at or
before LF. It is highlighted in (53).
(53) An idiom needs to be a unit at LF (Chomsky 1993:39).
A structure for (44a) is shown in (54). The VP contains the reconstructed NP Hafer and
the idiomatic verb sticht, in one LF constituent in accord with (53).
(54) Wir kennen alle den Hafer [CP der [NP Hafer] dich [VP der [NP Hafer] sticht ] ].
we know all
the oat
which oat
you
which oat stings
The fate of the higher copies is left open here.
Let us turn to the most interesting point, the obligatory stress reconstruction in the wake
of syntactic reconstruction. At first sight it may be an unremarkable observation that the
same word of the idiom always carries the sentence stress, even under wh-movement and
relative clause formation. However, this stress-pattern is a default one that is assigned by the
default rules (here: Stress-XP) for all the idioms. With the wh-phrase in Spec,CP and even
more with the relative clause head external to the relative clause, we would expect that
Stress-XP assigns the stress of the VP to the verb instead (or the PP if there is one):
8
The stressless PPs that occur in some of these idioms also exist in non-idiomatic utterances
as discussed in Uhmann 1991, Kratzer and Selkirk (2007) and Truckenbrodt (2012). They
are unexpected for the theory of Stress-XP, which otherwise achieves detailed empirical
coverage in German. For concreteness, I assume that these PPs semi-incorporate into the
verb, forming a complex verb like [V [PP über die Leber] gelaufen], lit. ‚across the liver run’.
It is possible that the verb strives to be a prosodic word, suppressing the effect of Stress-XP
inside of it (Truckenbrodt 2012).
20
(55)
oat
which you
stings
* Wir kennen alle den Hafer1 [CP der1
dich [VP t1 sticht ] ].
|
stress wrongly expected by default rules (Stress-XP)
The only way of preventing this is to assume that the displaced stressed idiom chunk still
counts as inside of the VP for stress-assignment, i.e. stress reconstruction.
Furthermore, stress reconstruction is obligatory in these cases. The optional alternative of
a non-reconstructing stress pattern is not available in these cases. We will see that there is
no other phenomenon in German where syntactic reconstruction forces visible stress reconstruction in the same way. This motivates the following suggestion. The optional alternative
to stress reconstruction is due to a topic-comment structure. It is plausible that what is special about the more opaque idioms are that they are the only corner of the language in which
overlaid information structure, such as a topic-comment structure, is not allowed. The topiccomment analysis is developed in the following sections. For now, let us ask what this
means for the analysis of the more opaque idioms, where we do not find this overlaid effect.
There are two possible conclusions:
(56) a. Stress reconstruction in A-bar movement is generally obligatory. Or:
b. Stress reconstruction is obligatory where syntactic reconstruction is obligatory.
By stress reconstruction being obligatory under circumstances X in (56) I mean that the
mechanisms underlying it are present obligatorily under circumstances X, allowing that the
visible effect of stress reconstruction on the surface can be hidden by an additional topiccomment structure in structures other than more opaque idioms.
The idioms will not help us decide among the two hypotheses in (56). Since the absence
of the topic-comment structure is only found in a corner of the language in which we also
want to assume obligatory syntactic reconstruction for the idioms, the reason for obligatory
stress reconstruction may be either one of the two.
4.3. A topic-comment structure overrides stress reconstruction
First, I follow Frey (2004) in assuming that German clauses can have an aboutness-topic,
here also called sentence topic, in an early position of the clause and that this position is
independent of Spec,CP. This topic is also close to what Diesing (1992) sees as material
outside of the VP and close to the topic of Jäger (2001) adopted also in Kratzer and Selkirk
(2007). It is not crucial here whether every clause requires such a topic (with the assumption
that it can be constituted by an implicit parameter), but it seems at least that clauses like to
have such a topic, as we will see in a moment. The resulting topic-comment structure has
consequences for stress assignment. I spell them out in (57) and call them topic stress.
(57) Topic-stress
In a topic-comment structure the topic must not contain stronger stress than the
comment.
21
Consider first (58a). With an initial topic preceding the subject the subject prefers a stress
pattern with a stressless verb. Without the adverb, the most natural stress pattern is one in
which sentence stress is assigned to the verb (in addition to the phrasal stress on the subject). Why is this so?
(58) a. dass draußen ein Mann gegeigt hat
b. dass
ein Mann gegeigt hat
that (outside) a man fiddled has
‚that (outside) a man has fiddled’
Building on the suggestions of Jacobs (1993) and Frey (2004) and following Jäger (2001)
and Kratzer and Selkirk (2007) the difference is here analyzed as shown in (59) and (60). In
(59) the initial adverb is the topic and the rest of the sentence the comment, as shown. Topic
stress is trivially satisfied and has no effect on the stress relation between the subject and the
verb. Therefore we find default stress within the VP, the stress that we also saw in the idi9
oms. In the current account the stress on the subject but not the verb is derived by StressXP as discussed in connection with (52). In (60), on the other hand, there is no initial adverb
and so the sentence will tend to take the subject as its topic, as shown. Here topic stress
rules out the default stress pattern *[ein MannTOPIC gegeigtCOMMENT] in which the topic
would have the strongest stress. Topic stress correctly derives that additional stress is as10
signed to the verb and that this is the strongest stress of the sentence.
(59)
__topic__ _______comment________
dass draußen [vP ein Mann gegeigt] hat
that outside
a man fiddled has
(60)
__topic__ _comment_
dass ein Mann gegeigt hat
that a man
fiddled has
I now adopt the suggestion of Krifka (this volume) the a sentence topic in a wh-questions
can be the wh-phrase and that in relative clauses the head can have sentence topic status of
the relative clause. I pursue the idea that these are possible sentence topics in the sense relevant here to which (57) applies.
The more opaque idioms, then, are the ones that do not allow separation of the idiom into
a topic and a comment part. Their stress is as analyzed above: syntactic reconstruction provides the input structure to stress assignment so that we obligatorily find a stressless verb
following a reconstructed stressed argument (obligatory stress reconstruction).
9
More generally this default stress involves stress on the subject but not the verb so long as the subject is not followed by overt accusative or dative objects, see Kratzer and Selkirk 2007, Truckenbrodt 2012.
10
In the presence of more than one verb, the preferred position is the innermost one since it satisfies
Stress-XP for the upper VP at the same time.
22
In other cases, stress reconstruction is accompanied by the option of the topic-comment
structure. Examples are shown in (61) – (63). The postulated topic constituent is bracketed
and labeled T.
(61) a. Ich frage mich was für Bilder er gesehen hat.
b. Ich frage mich [was für Bilder]T er gesehen hat.
a. ‚I wonder what pictures he saw.’
b. ‚I wonder what pictures he saw.’
(62) a. Ich frage mich was für Bücher sie mag.
b. Ich frage mich [was für Bücher]T sie mag.
a. ‚I wonder what books she likes.’
b. ‚I wonder what books she likes.’
(63) a. die Bücher die sie mag
b. die [Bücher]T die sie mag.
a. ‚the books that she likes’
b. ‚the books that she likes’
(topic-comment)
(topic-comment)
The analysis of (61b) is shown in (64a) on the assumption that there is no syntactic reconstruction and in (64b) on the assumption that there is syntactic reconstruction. It turns out
that the topic-comment analysis allows both structures and predicts for both cases that the
sentence stress will be on the verb. In (64b) in particular it cannot fall on the silent copy in
the comment, for it would then emerge on the surface within the overt copy that is the topic,
and topic stress would not be satisfied.
(64) a.
_____topic____
... [was für Bücher]
what for books
__comment__
er t gesehen hat
he t seen has
b.
_____topic____
... [was für Bücher]
what for books
__________comment__________
er [was für Bücher] gesehen hat
he what for books
seen has
Thus the structural basis for stress reconstruction (a copy of the moved element in the VP)
may still be present or it may not still be present, but Topic stress will prevent it from
emerging in a visible stress effect so long as a topic-comment structure is assigned.
We are also interested in interactions between stress reconstruction and syntactic reconstruction. However, if the topic-hypothesis is correct, there is a sector of the language where
we don’t get to see anything because of the overriding topic option. This matches with the
facts, but it is not a very telling match. Let us consider some cases in turn.
First, there is range of less opaque idioms. They are here analyzed as transparent enough
to tolerate a topic-comment structure. They allow a reconstructed stress-pattern as well as a
non-reconstructed/topic-comment stress pattern:
23
(65) a.
b.
(66) a.
b.
(67) a.
b.
?
?
?
Ich frage mich in welchen sauren Apfel er gebissen hat.
Ich frage mich in welchen sauren Apfel er gebissen hat.
I ask myself in which sour apple he bitten has
lit. ‚I wonder in which sour apple he has bitten.’
‚I wonder which bitter pill he swallowed.’
Ich frage mich was für einen Bären sie ihm aufgebunden haben.
Ich frage mich was für einen Bären sie ihm aufgebunden haben.
I ask myself what for a bear they him tied-onto have
lit. ‚I wonder what bear they tied onto him.’
‚I wonder what lie they told to him.’
Sehr dünn ist der seidene Faden an dem das hängt.
Sehr dünn ist der seidene Faden an dem das hängt.
very thin is the silken thread on which this hangs
lit. ‚The thread if silk on which this hangs is very thin.’
‚This might easily go wrong.’
Second, anaphor reconstruction also allows both stress patterns:
(68) a. Ich frage mich was für Bilder von sich sie mag.
b. Ich frage mich was für Bilder von sich sie mag.
a. ‚I wonder what pictures of herself she likes.’
b. ‚I wonder what pictures of herself she likes.’
Similar observations obtain for scope reconstruction. Heycock (1995) pointed out reconstruction effects in connection with verbs of creation, which are further discussed by Fox
(1999); see also Heycock (this volume). I employ examples that provide a test case for
stress assignment below. While (69) has the two scope readings shown, only one of them is
possible with a verb of creation as in (70). The a.-paraphrases render readings in which nmany scopes over want. In the b.-paraphrases want scopes over n-many. Heycock and Fox
argue with the help of interaction of this phenomenon with Condition C that there is obligatory reconstruction in readings involving creation as in (70).
(69) Wie viele Geschichten will sie erzählen.
how many stories wants she tell
‚How many stories does she want to tell?’
What is the number n such that
a. there are n-many stories and she wants to tell them.
b. she wants to tell n-many stories.
(n-many > want)
(want > n-many)
24
(70) Wie viele Geschichten will sie sich ausdenken?
how many stories wants she herself invent
‚How many stories does she want to invent?’
What is the number n such that:
a. * there are n-many stories and she wants to invent them. (*n-many > want)
b.
she wants to invent n-many stories
(want > n-many)
However, we do not see obligatory stress reconstruction entailed by this syntactic reconstruction:
(71) a. Ich frage mich, wie viele Geschichten sie sich ausdenken will.
b. Ich frage mich, wie viele Geschichten sie sich ausdenken will.
a. ‚I wonder how many stories she wants to invent.’
b. ‚I wonder how many stories she wants to invent.’
The analysis of all the preceding cases is that there is syntactic reconstruction in these cases,
but that we do not get to observe whether the VP-internal syntactic copy entails prosodic
reconstruction because the option of the overlaid topic-comment structure. I this is correct,
we can learn two things from it. For one thing, the apparent absence of interaction of stress
reconstruction with syntactic reconstruction is not an argument against attributing them to
the same source, syntactic reconstruction. The effects may be hidden by the option of a
topic-commend structure. Second, we can learn that a topic-comment structure does not
prevent syntactic reconstruction.
In summary, the topic hypothesis accounts for the more opaque idioms. It also accounts
for the absence of interaction of syntactic and prosodic reconstruction in a range of cases.
The more opaque idioms show obligatory stress reconstruction. This is certainly compatible
with the possibility that syntactic reconstruction (here: for the idiom) and prosodic reconstruction draw on the same grammatical mechanism. However, we cannot exclude the possibility that they are separate mechanisms after all and that the mechanism underlying stress
reconstruction is inherently obligatory (only overlaid with the topic-comment structure
where this is allowed).
5. Condition C effects and stress reconstruction
I begin with wh-movement. There are predicates that reconstruct for Condition C like the
typical example from the English literature in (72a), here in its German translation. Other
25
predicates as in (73a) and (74a) do not seem to reconstruct for Condition C as readily. As
noted in Krifka (this volume) we will normally judge these on stress-patterns and contextual
assumptions that ameliorate the judgments as much as possible. Such stress-patterns are
employed in these a.-examples. Even the bad ones are far from reaching the unacceptability
we find with Condition C violations that do not require reconstruction as in (75). Now, as
we shift the sentence stress to the offending R-expression and employ a pattern of stress
reconstruction, as in (72b), (73b) and (74b), we get a strong Condition C effect, comparable
to (75), in all these cases. The c.-examples are controls that show that the problem with the
b.-examples is really a Condition C violation.
(72) a. ?(?) Ich frage mich welche Bilder von Maria sie mag.
b. *
Ich frage mich welche Bilder von Maria sie mag.
c.
Ich frage mich welche Bilder von Maria du magst.
I ask myself which pictures of Maria she/you like(s)
a. ‚I wonder which pictures of Mary she likes.’ (?(?))
b. ‚I wonder which pictures of Mary she likes.’ (*)
c. ‚I wonder which pictures of Mary you like.’
(73) [Let us return to the case of Vischnevsky1]
a.
Wir müssen rauskriegen, welche Bilder von Maria2 sie2 gesehen hat.
b. * Wir müssen rauskriegen, welche Bilder von Maria2 sie2 gesehen hat.
c.
Wir müssen rauskriegen, welche Bilder von Maria er1 gesehen hat.
we must find-out which pictures of Maria she/he seen has
a. ‚We must find out which pictures of Mary2 she2 has seen.’
b. ‚We must find out which pictures of Mary2 she2 has seen.’ (*)
c. ‚We must find out which pictures of Mary he1 has seen.’
(74) a.
Ich frage mich, welche von Marias1 Bildern sie1 mag.
b. * Ich frage mich, welche von Marias1 Bildern sie1 mag.
I ask myself which of Marias pictures she likes
a. ‚I wonder which of Mary’s pictures she likes.’
b. ‚I wonder which of Mary’s pictures she likes.’
(75) * Sie1 mag Bilder von Maria1.
She likes pictures of Maria
A similar, though weaker, effect obtains in (76) and (77). Here stress reconstruction is
forced by an element in the NP that is different from the R-expression. These seem to still
be degraded, even though the source without a reconstructing stress pattern in (73a) and
(74) is acceptable.
(76) ?(?) Wir müssen rauskriegen, welche Bilder von Maria2 sie2 gesehen hat.
we must finde-out which pictures of Maria she/he seen has
‚We must find out which pictures of Mary2 she2 has seen.’ (?(?))
26
(77) ?? Ich frage mich, welche von Marias1 Bildern sie1 mag.
I ask myself which of Maria’s pictures she likes
‚I wonder which of Mary’s pictures she likes.’ (??)
These examples could be analyzed by maintaining that stress reconstruction entails syntactic
reconstruction. In (72a), (73a) and (74a) we see the weaker or absent effects of reconstruction where stress reconstruction does not come into play. In the corresponding b.-examples
and in (76) and (77) stress reconstruction requires an underlying syntactic copy of the
moved element in its original position. Therefore Condition C effects emerge or are
strengthened. The argument would follow the argument in Heycock (1995) and Fox (1999)
who argue that scope reconstruction entails reconstruction for Condition C, demonstrating
that they both draw on the same mechanism of syntactic reconstruction. However, as will be
seen momentarily, there is another plausible interpretation of this degradation. That will be
the third interpretation discussed below.
Consider first a second approach. Chomsky (1993) proposed a preference principle by
which syntactic ‚reconstruction’ (leaving a copy in his copy theory of movement and reconstruction) is the norm. The preference principle accounts for the deviance of examples like
(72a). If the underlying copy is normally retained, the Condition C effect is correctly derived. (If the underlying copy was only optionally retained, the option of not retaining it
could be chosen and (72a) should be fully acceptable.) In addition, Safir (1999) has argued
that a mechanism of vehicle change affects reconstructed representations. Vehicle change is
an independently motivated analytical device due to Fiengo & May (1994). Applied to reconstruction it allows that an R-expression is replaced by a pronoun in the reconstructed
copy. This can explain the difference between (72a) on the one hand, and (73a) and (74a) on
the other. The reconstructed structures mirror the sentences in (78) – (80). The boldfaced
pronoun is the one that replaces the R-expression in the reconstructed copy.
(78) ?? Sie1 mag alle Bilder von ihr1.
(79) ? Sie1 hat alle Bilder von ihr1 (schon) gesehen.
(80)
Sie1 hat alle von ihren1 Bildern gesehen.
In this analysis there is always a lower copy of the moved element in the syntactic representation because of the preference principle. The a.-examples in (72) – (74) are improved
because of vehicle change. The corresponding b.-examples would now lead us to the conclusion that sentence stress on an element blocks vehicle change in its reconstructed copy.
Without vehicle change we obtain LFs with reconstruction that are comparable to (81) –
(83).
(81) * Sie1 mag alle Bilder von Maria1.
(82) * Sie1 hat alle Bilder von Maria1 (schon) gesehen.
(83) * Sie1 hat alle von Marias1 Bildern gesehen.
27
While this would be promising, it is not obvious how to analyze the improvements in (76)
and (77). The reconstructing R-expression is not stressed here, so we would not have the
same incentive for blocking vehicle change in the reconstructed copy. Without further ado,
the prediction of the account in terms of vehicle change would derive vehicle change in (76)
and (77) and therefore wrongly predict that these examples are fine.
Consider then a third interpretation. Krifka (this volume) suggests that independent anaphoric relations of the R-expression into the context ameliorate Condition C violations, in
particular under reconstruction. He relates this to the account of Condition C by Reinhart
(1983). The ameliorating anaphoric relations may be of two kinds and we will be concerned
with both here. First, they may be anaphoric relations of the R-expression itself. Second,
they may be anaphoric relations of the constituent containing it. Both are relevant here.
First, in (72a), (73a) and (74a) the R-expression can be construed as anaphorically related to
the previous discourse. Second, the fronted wh-phrase can be construed as a topic that is
given in the context. Now, sentence stress on the R-expression, as in (72b), (73b) and (74b)
blocks both these ameliorating effects: A contextually given element cannot carry sentence
stress (see e.g. Féry & Samek-Lodovici 2006; Ladd 1983). Therefore the stressed anaphor
itself cannot be construed as contextually given. At the same time the fronted constituent
containing the R-expression also cannot be construed as given, since it also contains sentence stress. So there is an independent conceivable source of the contrast within each of
(72a,b), (73a,b) and (74a,b). To make this a little more concrete, assume the following:
Assign one point of amelioration for each of the following: (i) anaphoric R-expression, (ii)
anaphoric constituent containing the R-expression and (iii) predicates favoring vs. not favoring reconstruction. For (iii) we will simply assign a point of improvement for (73) and (74)
without analysis. Now (72a) gets two points of amelioration, (73a) and (74a) each get three
points, while (72b) gets 0 amelioration and (73a) and (74a) one point (for the predicates).
This is a basis for modeling the judgments in (72) – (74) pretty well: 0 and 1 are *, 2 is ?(?)
and 3 is ok. This account also makes sense of (76) and (77). R-expression no longer contains sentence stress, so there is more point of amelioration relative to the ungrammatical
(73b) and (74b). On the other hand, sentence stress is still within the initial constituent, so
that initial constituent is still blocked from being construed as contextually given (or a topic), i.e. there is no further amelioration. We arrive at two points and predict ??, appropriately.
Notice that a combination of the last two theories is possible as well, if we assume that
vehicle change is a marked mechanism that allows for some smaller amount of improvement. Then the factor (iii) of the preceding discussion could simply be the option of vehicle
change. Let us assume this for concreteness.
That account looks very reasonable. The distinction between the a.- and the b.-examples
in (72) – (74) can therefore not directly be related to stress reconstruction. The distinction
might instead be due to the anti-givenness effect of sentence stress as such. All cases of
weaker or no violation of Condition C could be derived by amelioration due to contextual
givenness. Would the observations nevertheless be compatible with the assumption that
stress reconstruction and syntactic reconstruction draw on the same grammatical mechanism? It would. All cases in which stress reconstruction occurs, i.e. the b.-examples of (72)
– (74) as well as (76) and (77) must all involve reconstruction in the amelioration account.
The calculation of amelioration needs this as a starting point relative to which amelioration
takes place. However, we do not know with any certainty what forces syntactic reconstruc-
28
tion in these cases. It could be that stress reconstruction forces syntactic reconstruction here.
However, it is now also conceivable that the preference principle forces this reconstruction
more generally. Krifka assumes semantic reconstruction, which is also compatible with the
observations on Condition C above.
Consider also relative clauses. These show different behavior for reconstruction of Condition C (Munn 1994, Salzmann 2006, this volume). Predicates that inherently reconstruct
for Condition C (regardless of stress) like the one in (72) do not show such inherent reconstruction for Condition C in relative clauses as in (84). Notice that the examples in (72a,b)
and (84a,b) have a non-reconstructing stress pattern.
(84) [What can I show you?]
a.
Zeige mir die Bilder von Maria, die sie mag.
b.
Zeige mir die Bilder von Maria, die sie mag.
show me the pictures of Maria which she likes
a. ‚Show me the pictures of Mary that she likes.’
b. ‚Show me the pictures of Mary that she likes.’
Where a reconstructing stress-pattern is employed as in (85) the Condition C effect emerges
in one case, (85b).
(85) a. [What can I show you in connection with Maria?]
Zeig mir die Bilder von Maria, die sie mag.
show me the pictures of Maria which she likes
‚Show me the pictures of Mary that she likes.’
b. [What can I show you?]
?? Zeig mir die Bilder von Maria, die sie mag.
show me the pictures of Maria which she likes
‚Show me the pictures of Mary that she likes.’ (??)
This is also compatible with Krifka’s perspective. We might assign one additional point of
amelioration for relative clauses generally, so that we only see the stronger effects of the
wh-questions emerge, in a weakened form. The effects that are already weaker with whquestions would not appear at all. This would correctly describe these observations, both
those without stress reconstruction and those with reconstruction. It would raise the issue of
where the point of amelioration for relative clauses comes from. In an analysis of relative
clauses in terms of head matching this might be related to the additional mechanism of
matching the external head with the relative-clause internal copy.
I sum up the discussion of possible interaction with Condition C. It is plausible that
givenness amelioration and its being blocked by stress effects play a major role here. Even
this account, however, requires that reconstruction has taken place, to allow the Condition C
effects to be derived in their strong form where amelioration is blocked and in weaker forms
otherwise. Thus syntactic reconstruction might itself be forced by stress reconstruction, or it
might be generally obligatory, as per the preference principle. The original motivation for
the preference principle, (72a), suggest the latter. For the specific question of this paper, I
29
take the result to be that either syntactic reconstruction is required by prosodic reconstruction, or that syntactic reconstruction is always required by the preference principle.
Let us take this together with the results of the preceding section. For one thing, the evidence is compatible with the possibility that stress reconstruction and syntactic reconstruction are tied to each other and co-vary. For another, they may be separate mechanisms, in
which case more opaque idiom reconstruction requires the assumption that the mechanism
underlying stress reconstruction is obligatory. For the same scenario of separate mechanisms, the Condition C reconstruction cases would suggest that syntactic reconstruction is
then also obligatory, as per Chomsky’s preference principle.
Either way, notice that this means that syntactic reconstruction and stress reconstruction
are coextensive, be it on a case-by-case basis, or by their both or each relying on an obligatory mechanism. We may take this to constitute a weak argument for their going back to the
same mechanism of grammar, i.e. the copy left behind in syntactic movement in Chomsky’s
copy theory of movement and reconstruction.
6. Summary
These are the main results of this paper.
First, stress-assignment shows effects of reconstruction, as first shown in different terms
by Bresnan (1971, 1972). This phenomenon occurs both in wh-questions and in relative
clauses.
Second, more opaque idioms reconstruct for stress obligatorily. This suggests that stress
reconstruction is either inherently tied to syntactic reconstruction or more generally inherently obligatory. It also suggests that its apparent optionality of stress reconstruction in other
cases results from the option of overlaying a topic-comment structure that overrides the
visible effect of stress reconstruction.
Third, stress assignment heavily interacts with Condition C reconstruction. It blocks
amelioration effects due to givenness noted by Krifka this volume. At the same time, the
cases of stress reconstruction in which amelioration is thus blocked show reconstruction for
Condition C. This may be because stress reconstruction forces reconstruction for Condition
C, or because reconstruction for Condition C always applies (preference principle).
Fourth, when we take the preceding two points together, they suggest that either syntactic
and prosodic reconstruction are tied to each other, or they are both inherently obligatory.
Either way they are coextensive. This may be seen as a weak argument for assuming that
they do go back to the same grammatical mechanism. In addition, it was seen that the facts
are compatible with an analysis in which stress reconstruction and syntactic reconstruction
both draw on the same grammatical mechanism.
30
References
Bresnan, Joan. 1971. Sentence stress and syntactic transformations. Language 47, 257-81.
—. 1972. Stress and syntax: a reply. Language 48, 326-42.
Chen, Matthew Y. 1987. The Syntax of Xiamen Tone Sandhi. Phonology Yearbook 4, 10949.
Chomsky, Noam. 1993. A minimalist program for linguistic theory. In The view from
Building 20, K. Hale & S.J. Keyser (Hg), Cambridge, Mass.: MIT Press, 1-52.
—. 2008. On Phases. In Foundational Issues in Linguistic Theory. Essays in Honor of JeanRoger Vergnaud, R. Freidin, C.P. Otero & M.L. Zubizarreta (Hg), Cambridge,
Mass.: MIT Press, 133-66.
Chomsky, Noam & Morris Halle. 1968. The Sound Pattern of English. New York: Harper
and Row.
Cinque, Guglielmo. 1993. A null theory of phrase and compound stress. Linguistic Inquiry
24, 239-97.
Citko, Barbara. 2001. Deletion under identity in relative clauses. Proceedings of NELS 31.
Diesing, Molly. 1992. Indefinites. Cambridge, Mass: MIT Press.
Féry, Caroline & Vieri Samek-Lodovici. 2006. Focus projection and prosodic prominence
in nested foci. Language 82, 131-50.
Fiengo, Robert & Robert May. 1994. Indices and identity. Cambridge, Mass; London,
England: MIT Press.
Fox, Danny. 1999. Reconstruction, binding theory and the interpretation of chains.
Linguistic Inquiry 30, 337-62.
Fox, Danny & David Pesetsky. 2005. Cyclic linearization of syntactic structure. Theoretical
Linguistics 31, 1-45.
Frey, Werner. 2004. A medial topic position for German. Linguistische Berichte 198, 15390.
Grosu, Alexander & Manfred Krifka. 2008. The gifted mathematician that you claim to be:
Equational intensional 'reconstruction' relatives. Linguistics and Philosophy 30,
445-85.
Gussenhoven, Carlos. 1983. Focus, mode and the nucleus. Journal of Linguistics 19, 377417.
—. 1992. Sentence accents and argument structure. In Thematic structure, its role in
grammar, I. Roca (Hg), Berlin, New York: Foris, 79-106.
Hale, Kenneth & Elisabeth Selkirk. 1987. Government and Tonal Phrasing in Papago.
Phonology Yearbook 4, 151-83.
Heycock, Caroline. 1995. Asymmetries in reconstruction. Linguistic Inquiry 26, 547-70.
—. this volume. Relative reconstructions. Can we arrive at a unified picture?
Höhle. 1982a. Explikation für "normale Betonung" und "normale Wortstellung". In
Satzglieder des Deutschen, W. Abraham (Hg), Tübingen: Narr, 75-153.
Höhle, Tilman N. 1982b. Explikationen für "normale Betonung" und "normale
Wortstellung". In Satzglieder im Deutschen. Vorschläge zur syntaktischen,
semantischen und pragmatischen Fundierung, W. Abraham (Hg), Tübingen: Narr,
75-153.
Hulsey, Sarah & Uli Sauerland. 2006. Sorting out relative clauses. Natural Language
Semantics 14, 111-37.
31
Jacobs, Joachim. 1993. Integration. In Wortstellung und Informationsstruktur, M. Reis (Hg),
Tübingen: Max Niemeyer Verlag.
Jäger, Gerhard. 2001. Topic-comment structure and the contrast between stage level and
individual level predicates. Journal of Semantics 18, 83-126.
Kratzer, Angelika & Elisabeth Selkirk. 2007. Phase theory and prosodic spellout: the case of
verbs. The Linguistic Review 24, 93-135.
Krifka, Manfred. 1984. Focus, Topic, syntaktische Struktur und semantische Interpretation.
Universität München.
—. this volume. A direct compositional approach to Condition C effects under
reconstruction and their exceptions.
Ladd, D. Robert. 1983. Even, focus, and normal stress. Journal of Semantics 2, 257-70.
Munn, Alan. 1994. A minimalist account of reconstruction asymmetries. Proceedings of
NELS 24, 397-410.
Pierrehumbert, Janet Breckenridge. 1980. The Phonology and Phonetics of English
Intonation. Massachusetts Institute of Technology. Ph.D. thesis.
Salzmann, Martin. 2006. Resumptive prolepsis. A study in indirect A'-dependencies.
Utrecht: LOT.
—. this volume. In favor of a new version of the Matching Analysis.
Samek-Lodovici, Vieri. 2005. Prosody-syntax interaction in the expression of focus.
Natural Language & Linguistic Theory 23, 687-755.
Selkirk, Elisabeth. 1984. Phonology and syntax: the relationship between sound and
structure. Cambridge, Mass: MIT Press.
—. 1986. On Derived Domains in Sentence Phonology. Phonology Yearbook 3, 371-405.
—. 1995a. The prosodic structure of function words. In Papers in Optimality Theory.
University of Massachusetts Occasional Papers 18, J. Beckman, L.W. Dickey & S.
Urbanczyk (Hg), Amherst, Mass: GLSA, 439-69.
—. 1995b. Sentence prosody: intonation, stress, and phrasing. In The handbook of
phonological theory, J. Goldsmith (Hg), Cambridge, Mass: Blackwell, 550-69.
Selkirk, Elisabeth & Tong Shen. 1990. Prosodic domains in Shanghai Chinese. In The
phonology-syntax connection, S. Inkelas & D. Zec (Hg), Chicago: The University
of Chicago Press, 313-37.
Truckenbrodt, Hubert. 1995. Phonological phrases: their relation to syntax, focus, and
prominence. MIT. Doctoral thesis.
—. 1999. On the relation between syntactic phrases and phonological phrases. Linguistic
Inquiry 30, 219-55.
—. 2006. Phrasal Stress. In The Encyclopedia of Languages and Linguistics, 2nd edition,
Vol. 9, K. Brown (Hg), Oxford: Elsevier, 572-79.
—. 2007a. The syntax-phonology interface. In The Cambridge handbook of phonology, P.
de Lacy (Hg), Cambridge: Cambridge University Press, 435-56.
—. 2007b. Upstep of edge tones and of nuclear accents. In Tones and tunes. Volume 2:
Experimental studies in word and sentence prosody, C. Gussenhoven & T. Riad
(Hg), Berlin: Mouton, 349-86.
—. 2012. Effects of indefinite pronouns and traces on verb stress in German. In Prosody
Matters. Essays in Honor of Lisa Selkirk, T. Borowsky, S. Kawahara, T. Shinya &
M. Sugahara (Hg), London: Equinox Publishing, 487-513.
32
Truckenbrodt, Hubert & Daniel Büring. in preparation. Syntax-prosody correspondence at
the phrasal level.
Truckenbrodt, Hubert & Isabelle Darcy. 2010. Object clauses, movement, and phrasal
stress. In The sound patterns of syntax, N. Shir & L. Rochman (Hg), Oxford:
Oxford University Press, 189-216.
Uhmann, Susanne. 1991. Fokusphonologie. Eine Analyse deutscher Intonationskonturen im
Rahmen der nicht-linearen Phonologie. Tübingen: Niemeyer.
Vergnaud, Jean-Roger. 1974. French relative clauses. MIT. Doctoral thesis.
von Stechow, Arnim & Susanne Uhmann. 1986. Some remarks on focus projection. In
Topic, focus, and configurationality, W. Abraham & S.d. Meij (Hg), Amsterdam
and Philadelphia: John Benjamins, 295-320.