The Syntax And Semantics Of Focus-Sensitive ParticlesIn German

DANIEL BÜRING and KATHARINA HARTMANN

THE SYNTAX AND SEMANTICS OF FOCUS-SENSITIVE PARTICLES IN GERMAN

ABSTRACT. We propose a comprehensive account of both the distribution and the interpretation of German focus particles such as nur, auch and sogar (‘only’, ‘also’, ‘even’). We argue that they always adjoin to non-arguments (in recent terms this means that they can adjoin to VPs, IPs, APs and root CPs, but never to argument DPs or argument CPs), and that they do not undergo LF raising. Presenting a range of mostly new data and observations, we show how this theory accounts for a variety of puzzling distributional facts about adverbial and ad-adjectival particles, in particular, their specific interpretations and their behavior with respect to scope and reconstruction.

1. I NTRODUCTION

English so-called ‘focus particles’ such as even, only and also, henceforth FPs, can occur in adverbial position, as in (1a), as well as in adnominal position, (1b). In terms of the Government and Binding Theory, the adverbial only in (1a) is adjoined to the verb phrase (VP), while adnominal only in (1b) is adjoined to the determiner phrase (DP) in object position, cf. (1a ) and (1b ), respectively (capitals indicate main stress, []F marks the syntactic focus): (1)a. a .

I only read [a NOVEL]F . [IP I [VP only [VP read [DP a NOVEL]]]].

Earlier versions and parts of this paper have been presented at the 10th Comparat-

ive Germanic Syntax Workshop in Brussels, the 13th West Coast Conference on Formal Linguistics in Los Angeles, and at the Generative Grammatik im Süden Workshop 1996 in Berlin. We thank the audiences for discussion and comments. We would also like to thank the Syntax Reading Group at UC Santa Cruz, in particular Jim McCloskey and Jason Merchant. Many thanks to Summer Kern, Roland Pfau and Ede Zimmermann, as well as to the editor of NLLT and our three anonymous reviewers, for their help and suggestions. Katharina Hartmann is grateful for the financial support provided by the Deutsche Forschungsgemeinschaft, project GR 559/5-2. Natural Language & Linguistic Theory 19: 229–281, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.

230

DANIEL BÜRING AND KATHARINA HARTMANN

b.

I read only [a NOVEL]F .

b .

[IP I [VP read [DP only [DP a NOVEL]]]].

A number of authors, among them von Stechow (1991) and Bayer (1990, 1996), assume that other Germanic languages, in particular German, have both adverbial and attributive FPs as well. The only author explicitly departing from this assumption is, to our knowledge, Jacobs (1983, 1986), who argues that the German focus particles nur, ‘only’, auch, ‘also’, and sogar, ‘even’, are exclusively adverbial. In this paper we provide various sets of previously unnoticed data that strongly support Jacobs’ position. The structure of the paper is as follows: After a preview of the general claims and arguments, we present our adaptation of Jacobs’ (1983, 1996) analysis of German focus-sensitive particles in section 2, which posits only adverbial FPs, and show how it accounts for the distribution of adverbial FPs in German, in particular the absence of ‘FP DP’ sequences inside of DPs and PPs. A surprising consequence of this move is that all sentence-initial occurrences of ‘FP DP’ must be analyzed as occurring in verb-third sentences with the FP attached to the root CP node, rather than to the DP in SpecC. In section 3 we provide the explicit semantics for FPs and show how the structures we assume lend themselves to a straightforward and adequate interpretation. In section 4 we discuss (and dismiss) an apparent argument for DP-attached FPs, namely that DP and FP form a constituent that can undergo quantifier raising. We show that alleged cases of [DP FP DP]-raising in German can, and in fact should, be reanalyzed without LF movement. Subsequently, we show in section 5 that attested cases of LF movement, namely scope and binding reconstruction of DPs, provide more evidence against DP-adjoined FPs. While the DP undergoes reconstruction, the FP retains wide scope. Section 6 introduces data parallel to those presented in section 5 which show that sentence-initial ‘FP CP’ sequences do not form a constituent either. This leads to a reformulation of the earlier prohibition against adjunction to DP in terms of a prohibition against adjunction to arguments in general. It is shown that this reformulation derives a number of additional facts not considered in the previous sections, among them adjunct-argument asymmetries and, in particular, the well-known but illunderstood fact that ‘FP CP’ sequences are systematically absent from the German ‘Nachfeld’ (i.e., the position occupied by extraposed elements). Finally, section 7 broadens the database to account for FPs inside of DPs, or more precisely, FPs adjoined to AP.


231

1.1. Adverbial and Adnominal Attachment of Focus-Sensitive Particles In this subsection we take a first look at the distribution of FPs in German and possible ways of accounting for it. For expository purposes, let us call ‘adverbial FPs’ all those FPs immediately dominated by a node within the clausal projection line, or, in Grimshaw’s (1991) terms, by a node which is an ‘Extended Verbal Projection’ (henceforth EVP). This class includes FPs adjoined to CP, IP, and VP. Thus, (1a) and the German (2) below both contain instances of adverbial FPs. Sentence (1b), where only is attached to DP (hence immediately dominated by the higher DP segment, which is not an EVP) does not (FPs in the German examples will be set in boldface): (2)

Peter kann sogar [VP KOCHEN]. cook P. can even

In many cases, however, word order does not clearly indicate whether we are dealing with an adverbial or an adnominal FP in a given German sentence. Because German VPs are strictly head-final, the left boundary of VP is not easily identifiable. As an illustration consider (3). It is compatible with an adverbial analysis, (3a), parallel to (1a), as well as with an adnominal analysis, (3b), which is parallel to (1b). (3)

Ich have nur [einen ROMAN]F gelesen. novel read I have only a

a.

ich habe [VP nur [VP [DP einen Roman] gelesen]]

b.

ich habe [VP [DP nur [DP einen Roman]] gelesen]

The semantics does not help us decide either, because (1a) and (1b) have identical truth conditions (cf. section 3 below). Thus, we cannot simply ask whether (3) is the translation equivalent to (1a) or to (1b) in order to decide on a structure for it. Analogous remarks apply to the case where a FP precedes the subject: The FP could be adjoined to the IP (adverbial, (4a)) as well as to the subject DP (adnominal, (4b)): (4)

Gestern habe sogar ICHF einen Roman gelesen. a novel read yesterday have even I Yesterday, even I read a novel.

a.

[CP gestern habe [IP sogar [IP [DP ich] einen Roman gelesen]]]

b.

[CP gestern habe [IP [DP sogar [DP ich]] einen Roman gelesen]]

232


We could thus conclude that sentences like (3) and (4) are structurally, but not semantically, ambiguous in German and leave it at that. Alternatively, we could seek to eliminate this ambiguity by invoking additional constraints which exclude one of the two structures, namely the adnominal one. The inverse – that German doesn’t allow for adverbial FPs at all and that only adnominal structures like (3b) and (4b) are admitted by the grammar – is not an option. First, there are cases like (2), in which VP is in focus and the FP is clearly attached to a verbal, rather than a nominal projection. Second, there are sentences in which the entire VP (rather than just the object DP) is the focus. To give an informal idea, sentence (5), on its pragmatically prominent reading, has focus on the VP as in (5a) (the other, dispreferred reading has focus on the DP as in (5b)):1 (5)

Ich wollte nur einem MISSVERSTÄNDNIS vorbeugen. I wanted only a-DAT misunderstanding forestall

a.

VP-focus: [VP [DP einem Missverständnis] vorbeugen]F interpretation: all I wanted was to forestall a misunderstanding (I didn’t mean to start a fight)

b.

DP-focus: [VP [DP einem Missverständnis]F vorbeugen] interpretation: all I wanted was to forestall a misunderstanding (I didn’t mean to forestall a fight)

A FP cannot adjoin within the focused constituent.2 That means that the VP focus reading is possible only if the FP is outside of VP, implying that (5), on the reading indicated (5a), must contain an adverbial (here, VP-adjoined) FP; it cannot be analyzed as an instance of adjunction to DP. We are thus left with the choice between admitting a spurious ambiguity in sentences like (3) and (4) and prohibiting adnominal DP in German altogether. We will refer to these choices as the ‘mixed analysis’ (because 1 Note that in either case, the focus will be realized by a pitch accent on the object DP (cf. e.g., Selkirk’s (1984) rule of ‘focus projection’); that is, both sentences sound the same. 2 That a particle cannot adjoin within a focused constituent can easily be seen in the parallel English examples:

(i)

I only wanted to forestall a MISUNDERSTANDING.

(ii)

I wanted to forestall only a MISUNDERSTANDING.

Adverbial only, as in (i), is compatible with either VP or DP focus, just like the German sentence (5). But adnominal focus as in (ii), where only sits within the VP, can only be interpreted as DP focus.


233

they assume both adverbial and adnominal FPs) and the ‘adverbial-only’ analysis respectively. In this paper we take the latter route. In particular, we follow Jacobs (1983, 1996) in claiming that the adverbial analyses in (3a) and (4a) are the correct ones, and that (3b) and (4b) are not possible structures in German. In our analysis, German does not allow adnominal FPs at all. In other words, German lacks a structural counterpart to English sentences like (1b). Concretely, we claim that (6) holds in German (we will argue in section 6 below that (6) follows from a more general principle): (6)

FPs can attach only to elements of the verbal extended projection.

Our goal in the rest of this paper is to present and motivate an adverbialonly analysis that assumes (6), and to demonstrate its advantages over a mixed analysis (e.g., Bayer 1990, 1996). While one might argue for an adverbial-only analysis on grounds of theoretical simplicity alone, we will focus on a wealth of empirical arguments in favor of the latter option. We will show that the mixed analysis overgenerates and that allowing for DP-adjoined FPs wrongly predicts various unattested distributional and interpretive possibilities. We will address the interpretation facts in subsequent sections, once we have presented our adverbial-only account in sufficient detail. Let us start with an argument based on the distribution of FPs: If FPs could adjoin to DP, we would expect ‘FP DP’ sequences within NPs and PPs. Since NP and PP are left-headed in German, such sequences should be easy to detect, because the FP should follow the N/P and precede the DP. But such cases are utterly unacceptable in German, as the following examples, partly from Jacobs (1983) show: ∗

[PP P FP DP]

a.

∗

mit nur Hans with only Hans

b.

∗

gegen sogar den Präsidenten against even the president

∗

[NP N FP DP]

∗

der Bruder nur des Grafen the brother only the-GEN count-GEN

(7)

(8) a.

234


b.

∗

die Proklamation sogar der Unabhängigkeit the proclamation even the-GEN independence

An adverbial-only analysis, incorporating (6) above, directly blocks the examples in (7) and (8). Unlike the cases (3) and (4), those in (7) and (8) do not allow for an analysis in which the FP is adjoined to an extended verbal projection. It is correctly predicted that the sequence ‘FP DP’ will occur only if DP is directly dominated by an EVP. The reader might wonder at this point why Jacobs’ adverbial-only analysis of German focus-sensitive particles has not been widely accepted in the light of data like (7), if its sole empirical cost is the elimination of a spurious ambiguity. We turn to this question in the next subsection. 1.2. Arguments Against an Adverbial-Only Analysis We are aware of four arguments against the adverbial-only analysis, which we will outline now along with a preview of their refutations: The first argument we call the ‘adjacency argument’. If we assume that FPs adjoin directly to the F-marked DP, we get a direct explanation for why (9a) is acceptable, but (9b) is not: The FP in (9b) simply did not attach to the focused DP: Gestern hat Rufus sogar [dem MÄDCHEN]F Blumen flowers yesterday has Rufus even the-DAT girl

(9)a.

geschenkt. given b.

c.

∗

Gestern hat sogar Rufus [dem MÄDCHEN]F Blumen geschenkt. Gestern hat sogar RUFUSF dem Mädchen Blumen geschenkt. Yesterday, Rufus even gave flowers to the girl.

An adverbial analysis, the argument continues, would demand that both (9a) and (9b) involve adjunction of a FP to an EVP (VP in (9a), IP in (9b)), so no difference in grammaticality is predicted. Note that it is not the pre-subject placement of FP which causes the ungrammaticality of the sentence, as witnessed by the grammatical (9c).3 3 Cf. e.g., Bayer (1996: 20): ‘In Jabobs’ [1983] theory it is an accident that sogar associates with [Rufus] in [(9c)]. There is no reason why it should not associate with some other focus constituent.’ As far as we can tell, the theory in Jacobs (1983) actually does account for these facts in much the same way as ours does, rendering Bayer’s criticism unjustified.


235

Our answer to the adjacency argument will be that a plausible version of the adverbial theory will rule out (9b), along with many other illicit cases of FP placement that do not involve DP attachment at all, by an adjacency requirement on a FP and its focus. Since a theory of adverbial FP placement is required anyway, the adjacency argument does not support a mixed theory. A precise and comprehensive account of adverbial FP placement, including the examples in (9), will be given in subsection 2.1. The second argument against the type of analysis we advocate here we will refer to as the ‘verb second argument’, which calls attention to sentences like (10): (10)

Nur HANS war betrunken. only H. was drunk

Since sentence (10) has the word order FP-DP-finiteV, and the finite V in German is always the second constituent in a main clause, one might conclude that FP and DP together must form a constituent. Under an adverbial analysis, the argument goes, nur would have to be attached to the clause (presumably CP), yielding an (illicit) verb-third structure.4 Our reply here will be to grab the bull by the horns, as it were, and agree that sentences like (10) are in fact verb-third. We will elaborate on this in subsections 2.3. and 6.2, arguing that no account of the data can simply prohibit adjunction of FPs to CP; our analysis is thus no more problematic than any other proposed. The third argument we call the ‘semantic argument’. It is based on the observation that in sentences like (9a), the particle sogar semantically associates with the DP dem Mädchen. The meaning of the entire phrase sogar dem Mädchen is a generalized quantifier which denotes something like the set of all properties that the girl and someone else have, such that it is very unlikely that the girl alone has these properties. In order to arrive at that meaning compositionally, FP and DP would have to form a constituent. Our reply will be that this argument confuses syntax and semantics. Semantic association of a FP with a focused constituent does not require adjunction of the FP to the focus, or any sort of syntactic constituency of the two. This point has been forcefully demonstrated in Rooth (1985), where a semantics for association with focus is developed which derives the correct interpretations for all the examples. We will demonstrate this in section 3. 4 Cf. e.g., Bayer (1996: 21f).

236


The fourth and final argument for a mixed analysis could be named the ‘quantifier raising argument’.5 It can be summarized like this: If a FP is attached to DP, it will undergo quantifier raising along with it. This will have semantic consequences, given that the FP is now ambiguous in scope. Adverbials do not undergo quantifier raising, so no scope ambiguity is predicted. In reply to this argument we will show in section 4 that the observed scope ambiguities are really ambiguities in surface structure rather than arising at the level of Logical Form. Given that surface structure neither need nor should be linked to DP-scoping, the quantifier raising argument lacks an empirical basis.

2. A N A DVERBIAL -O NLY A NALYSIS OF FP D ISTRIBUTION IN

G ERMAN In this section we present an account of the placement of FPs in German. At the heart of our analysis lies the notion of an ‘f-node’ (reminiscent of ‘focus-node’), as defined in (11). The principle governing the placement of a FP relative to its focus is simply (12).6 Henceforth we use the expressions ‘node α is F-marked’ and ‘α is the/a focus’ synonymously. The Particle Theory (preliminary version) (11)

For any node α marked F in a phrase marker P, let the set of f-nodes of α consist of all nodes β in P such that a. b. c. d.

(12)

β is an EP (extended projection) of some V γ β is a maximal projection β dominates α or is identical to α there is no EP β of γ such that β dominates β and β meets (11b) and (11c).

A FP must be left-adjoined to an f-node of its focus.

5 Arguments 3 and 4 have not generally been proposed contra Jacobs, but simply as a welcome consequence of the alternative, mixed analysis (although we have encountered in particular argument 3 in many discussions of the issue). It is mainly argument 2 that has been considered fatal for the adverbial-only kind of analysis we advocate in the present paper. 6 Throughout this paper we consider only examples with one FP and one focus. But the definitions are designed to carry over to cases with mutiple FPs, each associated with a different focus.


237

Informally, the Particle Theory says four things about the positioning of FPs: (11)a. FPs must be adjoined to an extended verbal projection b. FPs must be adjoined to a maximal projection c. FPs must c-command the focus d. FPs are as close to the focus as possible (11a) has been discussed in the previous section and will be elaborated upon in sections 5, 6, and 7. (11b) further restricts the class of adjunctions sites in a way we will justify in subsection 2.2. Finally (11c) and (11d) concern the relation between a FP and its focus, the former requiring a FP to c-command its focus, the latter requiring it to be as ‘low in the tree’ as possible while still c-commanding the focus, in other words: to be as close to the focus as possible. (11d) is a close kin of Jacobs’ (1983) ‘Prinzip der Maximalen Spätstellung’ (cf. also König (1991)) and will be crucial to our analysis of the German data. We will informally refer to (11d) as the ‘Closeness Principle’ or simply as ‘Closeness’. 2.1. FPs Adjoin As Close to the Focus As Possible As an illustration of the workings of Closeness, consider (9), repeated here with our proposed structures given underneath: (13)a. Gestern hat Rufus sogar [dem MÄDCHEN]F Blumen flowers yesterday has Rufus even the-DAT girl geschenkt. given a . [CP gestern hat [IP Rufus [VP sogar [VP [DP dem Mädchen]F Blumen geschenkt]]]] b. ∗ Gestern hat sogar Rufus [dem MÄDCHEN]F Blumen geschenkt. b . [CP gestern hat [IP sogar [IP Rufus [VP [DP dem Mädchen]F Blumen geschenkt]]]]

238


c. Gestern hat sogar RUFUSF dem Mädchen Blumen geschenkt. c . [CP gestern hat [IP sogar [IP [DP Rufus]F [VP dem Mädchen Blumen geschenkt]]]] Yesterday, Rufus even gave flowers to the girl.

In (13a) and (13b), the indirect object (IO) dem Mädchen, ‘the girl’ is focus-marked. A FP is acceptable following the subject, as in (13a), but not preceding it, as in (13b). So (11) predicts that VP, and only VP, is an f-node of the indirect object. By (11b), CP, IP, VP and all three argument DPs, but not C , I , V , V0 etc. are potential f-nodes. (11c) rules out the subject DP and the direct object DP Blumen, since neither dominates (or is identical to) the IO. (11a) rules out the indirect object DP itself, because it is not an EVP, leaving us with VP, IP and CP as potential f-nodes. Finally Closeness in (11d) rules out IP because VP qualifies as a β which is dominated by IP, a maximal projection, an EP of the V geschenkt, and dominates the focus. Closeness rules out CP for the same reason with β being either IP or, again, VP. Thus VP is the only permitted f-node for a focus marked indirect object DP, and by (12), the FP must adjoin to it, accounting for the impossibility of (13b). Contrast the analysis of (13a) and (13b) with (13c), in which the FP can and in fact must precede the subject. Maximal EVPs are VP, IP and CP again. But this time, VP drops out because it does not dominate the focus in SpecI, leaving IP and CP. CP is again blocked because IP is later/closer to the subject; IP itself, however, is not because the only EVP it dominates, VP, does not dominate the focus and so does not qualify as a β in the sense of (11d). Therefore, IP is the only f-node with subject focus, predicting (13c) to be possible. What the Particle Theory does in these examples is to force the FP to left-adjoin to the closest maximal EVP dominating the focused DP, rather than to the focused DP itself. So if the DP is left-peripheral in that EVP, a FP will end up adjacent to it, though it does not form a constituent with it. But the Particle Theory has a more general application. If the focus is on V or VP, it correctly predicts that a FP adjoins to VP, as in (2) and (5), repeated and structured here as (14) and (15) respectively.

(14)

[CP Peter1 kann2 [IP t1 sogar [VP [V KOCHEN]F] t2 ]]. P. can even cook


[CP

(15)

Ich1 I

wollte2 [IP t1 wanted

nur [VP [DP only

239

einem a

MISSVERSTÄNDNIS] vorbeugen]F t2 ]]. forestall misunderstanding In (14) the verb is focused. The maximal projections which dominate the focus and are EVPs are CP, IP, and VP. Among these, VP is closest to the focus and therefore qualifies as the f-node of the focus where FP adjoins to. In (15), the focus is on VP. The maximal projections which dominate (or are equal to) the focus and which are EVPs are CP, IP, and VP, among which VP is closest to the focus again. Finally, if a focus is contained within an argument, the prediction is that a FP will left-adjoin to the EVP immediately dominating that argument, i.e., to VP in (16): . . . weil [IP Kim sogar [VP [PP gegen [DP MARGRETSF against M.’s K. even . . . because

(16)

Mutter]] t1 ] aussagte1 ]. mother testified . . . because Kim testified even against Margret’s mother. The focus on Margrets is narrow (it cannot project from the specifier of the DP Margrets Mutter). Notice that adjunction of FP to DP which would make the particle adjacent to the focus would yield the unacceptable (17): (17)

∗

. . . weil Kim [PP gegen sogar [DP MARGRETSF Mutter]] against even M.’s mother . . . because K. aussagte. testified

This is again predicted by the Particle Theory: DP is not an EVP and therefore not an f-node of the focus. Parallel predictions are made for foci contained in DP-arguments, including subjects, and adjuncts. Given that the Particle Theory is required in the grammar of German in order to handle focused Vs, VPs and other focus EVPs, we conlude that the adjacency argument of section 1.2 loses its force. The adverbial-only theory of FPs does not predict that a FP can attach in arbitrary distance to the focus. Furthermore, since the Particle Theory directly accounts for the impossibility of PP and DP internal ‘FP DP’ sequences (exx. (7) and (17) above) and moreover, correctly predicts

240


the placement of FPs in cases where DP and PP internal DPs are focused, it is empirically more successful than a theory which allows for adjunction of FPs to DP. Before closing this subsection we want to note that Particle Theory categorically excludes post-nominal FPs such as in (18), where the FP follows (rather than precedes) the constituent containing the focal accent within a sentence-initial DP: (18)a. (∗ )Seine SCHWESTER nur überlebte den Unfall. his sister only survived the accident Only his sister survived the accident. b. (∗ )Für die ZUSCHAUER sogar war das Spiel unerfreulich. for the spectators even was the game unpleasant Even the crowd considered the game unpleasant.

Most speakers, including both authors of this article, indeed find sentences like (18) unacceptable. Standard grammars of German likewise do not mention this placement of FPs as a possibility (cf. Eisenberg (1989), Grebe et al. (1973), Heidolph et al. (1981), Zifonun (1997)). However, as an anonymous reviewer points out, Helbig (1988: 193/218) – from which the examples are taken – classifies these and similar examples as rare, sporadic and limited (‘selten’, ‘vereinzelt und in eingeschränkter Weise’), but possible. We do not wish to speculate about the differences between the registers that permit (18) and those that do not, but simply note that we do not intend our analysis to cover cases like (18). 2.2. Against Adjunction to Non-Maximal Projections (11b) above requires that FPs adjoin to maximal projections only. In this subsection we show how this requirement is crucial in analyzing some further cases of FPs associated with non-DP foci. Let us start by addressing the theoretical status of the restriction that FPs can only adjoin to maximal projections. It follows from standard X-bar theory or any of its reductionist cousins, if we assume that FPs are themselves maximal projections. This assumption seems justifiable, at least in the case of nur, ‘only’, and auch, ‘also’, which can occur alone in the sentence initial ‘Vorfeld’ position of verb second sentences. This position


241

is normally considered a specifier (SpecC on most views), and as such can only be filled by maximal projections:7 (19)a. [CP Auch [C war ich sehr MÜDE]]F . also was I very tired Also, I was very tired. b. [CP Nur [C WEISS das keiner]F ]. only knows that nobody It’s just that nobody knows about it. But even if (11b) cannot be deduced from general principles of phrase structure, we believe it should be maintained for a number of empirical reasons. It blocks adjunction of FPs to V, V , C, and C , a restriction which we will now demonstrate to be desirable.8 Let us start with V/C/C adjunction. Though we find FPs directly preceding the verb in verb final clauses like (20a), no such immediately preverbal FP is possible in verb second clauses with a sentence-initial XP: (20)a. . . . weil Peter Maria nur KÜSSTEF. . . . because Peter Maria only kissed b. ∗Peter nur KÜSSTEF Maria.9 Maria Peter only kissed 7 We do not have an explanation for why sogar, ‘even’, is blocked in this position. An alternative hypothesis, favored by two reviewers, would regard auch and nur in (19) as conjunction-like elements which happen to be homophonous to the focus particles. If so, then no conclusions about the phrasal status of the latter could be drawn and (11b) would have to be stipulated. Note, though, that the meanings of (19) can quite plausibly be paraphrased as ‘it was also the case that . . . ’ and ‘the only thing is that . . . ’, which could conceivably be derived from the meaning of the focus particles. We leave further investigation of this matter for future research. 8 In addition, it blocks adjunction to I and I , which, however, is not easy to see in German. 9 The grammatical version of (20b) would be (i), in which the FP is adjacent to the V-trace.

(i)

Peter KÜSSTEF Maria nur tV . Maria only Peter kissed

As Jacobs (1983) shows, adjacency of a FP to the trace of the focused element is an option with topicalization/wh-movement, head-movement and extraposition. We will not attempt to change the Particle Theory so as to cover these cases here, although the amendments are straightforward.

242


Every conceivable structure for (20b) will be excluded by (11b). First, the FP cannot attach to V and consequently move along to the verb-second position C, because V is non-maximal. Second, by the same token, the FP cannot attach to the verb-second position C directly. Finally, attachment of the FP to C , which would also yield the order in (20b), will be excluded too, because C is not maximal. But how is (20a) to be analyzed, then? We propose that the FP is attached to VP here, with the DO Maria scrambled out of VP, as in (21): (21)

weil [IP Peter [VP Maria [VP nur [VP tMaria KüSSTEF]]]]

If this analysis is correct, we predict that elements which cannot scramble should be allowed (in fact forced) to intervene between a FP and a focused V. The four sets of minimal pairs in (22) show that this prediction is borne out for directional PPs ((22a)), AP arguments ((22b)) and resultatives ((22c)), all of which are independently known to disallow scrambling (cf. Grewendorf and Sternefeld (1990), Webelhuth (1987)): (22)a. . . . weil man den Wagen nur in die Garage FAHRENF darf. man den Wagen in die Garage nur a . ∗ . . . weil . . . because one the car (only) into the garage (∗ only) FAHRENF darf.10 drive may because you may only drive the car into the garage.

b. . . . weil sie sich nur traurig FÜHLTF. sie sich traurig nur FÜHLTF. b . ∗ . . . weil . . . because she self (only) sad (∗ only) feels because she only feels sad. c. . . . weil sie sich nur ungeschickt ANSTELLTF. sie sich ungeschickt nur ANSTELLTF. c . ∗ . . . weil . . . because she self (only) clumsy (∗ only) acts because she only acts up clumsy. 10 Some speakers allow scrambling in these cases if the scrambled constituents are

contrastive topics.


243

d. . . . weil ich sie sogar in den Schlaf SINGENF würde. in den Schlaf sogar SINGENF d . ∗ . . . weil ich sie . . . bec. I her (even) into the sleep (∗ even) sing würde. would because I’d even sing her to sleep. Likewise, a FP cannot intervene within sentence-final verb clusters, owing to the fact that neither Vs nor VPs can scramble: (23)a. Siglinde wird sogar kommen

MÜSSENF.

b. ∗Siglinde wird kommen sogar MÜSSENF. Siglinde will (even) come (∗ even) must Siglinde will even have to come. Finally, infinitival complements to the raising verb scheinen, ‘seem’, cannot undergo scrambling either (in fact, they resist any kind of movement).11 As expected, in this case a FP cannot intervene between the infinitive and the main verb, as in (24a); instead the FP must precede the infinitival as seen in (24b). The unacceptability of (24a) contrasts in the predicted way with control verbs like versprechen, ‘promise’, in (24c), whose complements are known to be movable: Peter zu kommen nur SCHIENF . (24)a. ∗ . . . weil . . . because Peter to come only seemed b. . . . weil Peter nur zu kommen SCHIENF . seemed . . . because Peter only to come c. . . . weil Peter zu kommen nur VERSPRACHF. . . . because Peter to come only promised Peter only seemed/promised to come. 11 This is illustrated by the contrast between (i) (control) and (ii) (raising):

(i)

(ii)

. . . weil [zu kommen]1 niemand t1 versucht hat. nobody tried has . . . because to come ∗ . . . weil

[zu kommen]1 niemand t1 schien. nobody seemed . . . because to come

244


In sum, it seems well motivated to analyze ‘FP V’ sequences as derived from the movement of all XPs out of VP, as we do in (11b) by restricting adjunction sites for FPs to maximal projections. 2.3. V3 or Not V3? – The Verb Second Argument Let us now turn to the most surprising syntactic consequence of an adverbial-only analysis. Consider (25): (25)

Nur die HARTENF kommen in den Garten. come into the garden only the hard Only the tough ones make it into the garden.

In (25) nur precedes a sentence-initial subject. The adverbial analysis we propose implies, contrary to standard assumptions, that nur in this case is attached to CP, rather than to the topicalized DP. Schematically, it has to be assumed that (26a) rather than (26b) is the structure of (25): (26)a. [CP FP [CP DPF C . . . ]] b. ∗ [CP [DP FP DPF ] C . . . ] This assumption follows from the Particle Theory. Maximal projections which dominate (or are identical to) the focus in (25) are the subject DP and the root node, CP. Since only the latter is an extended verbal projection, it must be the f-node and the FP must therefore adjoin to it. The extraordinary consequence of the claim that FPs adjoin to CP, rather than to the element in SpecC, is that (25) (and in fact almost every sentence that begins with a FP) must be analyzed as a verb-third (V3) sentence. This flies in the face of practically every analysis of verb second-languages like German, for which it is usually held that any kind of adjunction to CP must be strictly ruled out in order to guarantee the verb second characteristic. In other words, as proponents of the adverbial-only analysis, we can no longer assume a general constraint preventing adjunction to CP. Instead, we have to stipulate a weaker constraint allowing only FPs to adjoin to CP.12 (It should be pointed out that in the absence of any deeper reasons for this prohibition, both constraints are mere language-specific stipulations.) 12 In fact, this class may very well be larger than just the FPs and include the negation element nicht, and, at least in interrogatives, other adverbials, cf. (i) (and the corresponding Swedish sentence (ii)), discussed in McCloskey (1998: 7):

(i)

Wenn wir nach Hause kommen, was sollen wir kochen? when we to home come what shall we cook

(G)


245

It is worth noting that no IP-internal focus will ever allow a FP to adjoin to CP. That is, even though we allow a FP to adjoin to CP in general, we still rule out sentences like (27), brought up by a reviewer, in which the focus sits within IP and SpecC is occupied by a non-focused element: (27)

∗

Nur es kommen die HARTENF in den Garten. into the garden only EXPL come the hard

The reason here is Closeness: The ‘lowest’ EVP that dominates the focus in (27) (i.e., the only one that does not dominate another EVP dominating the focus) is IP, not CP. Hence IP is the f-node in this example (as in the acceptable Es kommen nur die HARTEN in den Garten), and CP adjunction of FPs as in (27) is ruled out. Generally, Closeness will allow FPs to adjoin to CP only if the focus is outside of IP.13 There is distributional evidence which favors adjunction of FP to CP over adjunction of FP to SpecC. The argument presented below has the (ii)

In en stad som Fremont vem skulle inte vara uttråkad? in a town like Fremont who would not be bored

(Sw)

An exhaustive discussion of (what appear to be) exceptions to the verb second property in German can be found in Müller (in prep.), sec. 2.8.3. 13 As two reviewers point out, adjunction of FP to CP also predicts association of a CPadjoined FP with a focused verb in C0 as in (i) and (ii) to be possible, contrary to fact. In this case FP must adjoin adjacent to the verb’s base position rather than its derived position, as in (iii) and (iv): (i)

∗ Sogar sie KÜSSTEN sich.

even they kissed (ii)

REFL

∗ Sogar KÜSSTEN sie sich?

even kissed

they REFL

(iii)

Sie KÜSSTEN sich sogar. they kissed REFL even

(iv)

KÜSSTEN sie sich sogar? kissed they REFL even

It seems that with head movement, unlike with phrasal movement, only the foot of the chain is visible to Closeness. Like the interaction of FP placement and movement in general, we leave a detailed investigation of this matter for further investigation (cf. also note 9). Note though that (ii) will not be blocked by a prohibition against adjunction to CP (the FP could be in SpecC here), and that (iii) and (iv) would not follow from an alternative theory which simply claims that a FP adjoins to the constituent containing the focus.

246


following general form: If a sequence ‘FP XP’ can occur in sentence-initial position but cannot occur in the base position of XP, ‘FP XP’ should be analyzed as [CP FP [CP XP. . . ]. . . ] rather than [CP [XP FP XP]. . . ]. While this argument is hard to make for ‘FP DP’ (because only DPs which are dominated by the EVP in their base position can be moved to SpecC in German to begin with), it can be made for ‘FP CP’, ‘FP AP’ and ‘FP PP’. Let us illustrate the latter case here. Consider (28) (from Bayer (1996: 145)) and (29): (28)a. Nur [PP vom GRAFEN]1 habe ich [DP jeden Sohn t1 ] only of-the count have I every son bewundert. admired b. ∗Ich habe [DP jeden Sohn nur vom GRAFEN] bewundert. every son only of-the count admired I have I admired only the count’s every son. (29)a. Sogar [PP gegen die REGIERUNG]1 hat sie [DP eine even against the government has she a Proklamation t1 ] unterzeichnet. proclamation signed b. ∗ [DP Eine Proklamation sogar gegen die REGIERUNG]1 hat has a proclamation even against the government sie t1 unterzeichnet. she signed Even against the government did she sign a proclamation. (28) and (29) show topicalization of a PP out of an (object) DP. Adhering strictly to the verb second analysis, (28a) and (29a) are ‘source-less’ topicalizations. The topicalized PP is extracted from the DP-internal position marked by the trace. However, in this base position, ‘FP PP’ is ungrammatical, as witnessed by the (b)-examples. Our analysis offers an immediate explanation for this contrast: While the ungrammatical (b)-sentences would require attachment of a FP to PP, a non-EVP, the (a)sentences display adjunction to CP. On the alternative strict verb second analysis, the (a)-sentences, too, must involve the constituent [P P FP PP]; it would thus remain unclear what blocks the in situ counterparts in (b).


247

Summing up this section, we have presented an adverbial-only account of the distribution of focus-sensitive particles in German, i.e., one that generally bans adnominal FPs. We have shown how it accounts for the core data, including a number of otherwise unexplained distributional facts, in particular the absence of FPs in DPs and PPs and certain source-less topicalization. A surprising consequence of the proposal is the need to posit V3 structures in otherwise strict verb second languages. 3. T HE S EMANTICS OF ‘FP XP’

In this section we show how adverbial FPs are interpreted. It turns out that the correct interpretations are assigned to all the examples we have looked at so far, and that moreover, no semantic difference corresponds to the syntactic difference between the adverbial structure schematized in (30a), which we claim is the only one available in German, and the attributive structure like (30b), which we claim does not exist in German, but does, e.g., in English. (30)a. [EVP FP [EVP [XP . . . ]F . . . V ]] b. [EVP [XP FP [XP . . . ]F ] . . . V ] The framework we adopt for our analysis is the theory of Alternative Semantics as developed in Rooth (1985, 1992). The apparatus we assume is general enough to capture both structures in (30). Though this is more than is needed for German, it is advantageous if we want to apply it to languages that do have adnominal FPs, such as English. The seminal insight behind this approach is that there is a fundamental distinction between semantic association of a FP with an (F-marked) constituent, and syntactic adjunction of a FP to a constituent. In an adverbial example like the English I only read [a NOVEL]F , the FP only semantically associates with the object DP, but is syntactically adjoined to the VP. The common misconception underlying the semantic argument sketched in section 2.1. above is that semantic association should go hand in hand with syntactic adjunction (or some other local relation such as specifierhead agreement between the FP and its focus) in order to get the semantics right. That is, if a FP is associated with a focused DP, it should be in a local syntactic relation (such as adjunction) to it. This is not the case. Alternative Semantics provides the tools to get semantic association with focus in the absence of syntactic adjunction (for reasons to prefer such a purely semantic account of association with focus, see Rooth 1985, and the subsequent literature in the field).

248


Let us start by presenting the basic idea of this approach. As previously noted, focus on a constituent Y is indicated via a focus feature F, which is realized as a pitch accent on the main stress-bearing syllable of Y. Semantically, each node X receives two different interpretations, its ordinary semantic value, [[X]], and its focus semantic value, [[X]]f . The focus semantic value of any node X, [[X]]f , consists of a set of alternatives to its ordinary semantic value [[X]]. This set is derived by substitution of the meaning of the focused constituent by (contextually plausible) alternatives. For example, if X is a VP containing a focused DP, as in [V P dance with [LUISE]F ], [[VP]]f is the set of alternatives {[[dance with Marlene]], [[dance with the wolf]], [[dance with the vampire]],. . . }. If X consists of just the focus, e.g., DANCEDF , [[X]]f will be the set of all alternatives to [[X]], i.e., the (possibly contextually restricted) set of things in the denotation domain of X. If X contains no F at all, [[X]]f is the singleton set containing X’s ordinary semantic value, e.g. [[danced]]f = {[[danced]]}. Particles like only are focus-sensitive quantificational elements, which can be interpreted in situ. Adopting Roothian semantics, we let [[only]] take an element of an arbitrary type as its argument, as long as this type ends in t. A general rule is sketched in (31). (Unlike Rooth (1985) we interpret directly, without intermediate translation into intensional logic; we give predicate calculus paraphrases below for perspicuity). (31)

If A is of type α, t, only A is of type α, t, too, and [[only A]] is the set of all B of type α such that B has the property [[A]] (i.e., B ∈ [[A]]), and no other property that is an alternative to that (i.e., in [[A]]f ); [[only A]]f = {[[only A]]}.

By the semantics given in (31), only can attach to VPs and DPs alike, as illustrated in the English examples in (32): (32)a. [[only DANCEDF ]] = the set of all individuals x who danced and have done no alternative thing to dancing: λx.danced(x) & ∀p ∈ ALT(danced): p(x) → p = danced b. [[only BILLF ]] = the set of all properties p that Bill has and that no alternative to Bill has: λp.p(bill) & ∀x ∈ ALT(bill): p(x) → x = bill14 14 Applying (31) literally to this case derives the set of all p such that p is a property Bill has, and if someone else has that property, that someone else’s properties equal those that Bill has. By Leibniz’ law this means that someone else is Bill: λq.[λp.p(bill)](q) & ∀r ∈ ALT(λp.p(bill)):r(q) → r = λp.p(bill). We will gloss over these details. Furthermore we ignore intensionality where possible.


249

To illustrate the workings of (31) for an adverbial FP associated with DP internal focus, we will look at the German case in (33): (33)

. . . weil nur FRITZF gekommen ist. come is . . . because only Fritz

The interpretation of the basic phrases, together with a standard predicate logic rendering of them, is given in (34): (34)

[[FritzF ]] = the set of all properties p which Fritz has: λp.p(fritz) [[came]] = the property of having arrived: λy.y came [[FritzF ]]f = the set of all sets P of properties p such that there is an alternative to Fritz who has precisely the properties p in P: λP.∃x ∈ ALT(fritz) & P = λp.p(x)

Let us now interpret (33), assuming, as our analysis does, that it is an adverbial FP: (35)

(weil) [IP nur [IP FritzF gekommen ist]] [[ [IP FritzF gekommen ist] ]] = the set of worlds in which Fritz arrived: ∧ arrived(fritz) [[ [IP FritzF gekommen ist] ]]f = the set of sets q of worlds w, such that some alternative to Fritz arrived in w: λq.∃x ∈ ALT(fritz) & q = ∧ arrived(x) [[ [IP nur [IP FritzF gekommen ist]] ]] = the set of worlds in which Fritz arrived and anyone else who arrived is identical to Fritz: ∧ [arrived(fritz) & ∀x ∈ ALT(fritz):arrived(x) → x = fritz]

For comparison, we now calculate the meaning of the parallel English example, which arguably involves an adnominal FP: (36)

(because) [IP [DP only [DP Fritz]F ] arrived] [[ [DP only FritzF ] ]] = the set of all properties p which Fritz has and noone else has: λp.p(fritz) & ∀x ∈ ALT(fritz): p(x) → x = fritz [[ [IP [DP only FritzF ] arrived] ]] = the set of worlds in which the property of having come is one which Fritz and noone else has: ∧ [arrived(fritz) & ∀x ∈ ALT(fritz):arrived(x) → x = fritz]

250


The last lines in (35) and (36) describe the same set. Hence our analysis derives the same truth conditions for German (adverbial) and English (adnominal) (as well as for a hypothetical language German’, which is just like German, except that it has adnominal FPs). The semantic argument, wherein the FP and its focus have to form a constituent in order to derive the correct interpretation, has thus been proved invalid. It can be shown that the equivalence in (30) holds in general between structures of the form (37a) and (37b), provided that the meanings of Y and Z in (37a) are within the semantic scope of the meaning of XP: (37)a. [EVP FP [EVP Y [XP . . . ]F Z ]]] b. [EVP Y [XP FP [XP . . . ]F ] Z ] In (35) and (36) above, XP was the subject DP, Y was zero and Z was the VP gekommen ist/arrived, respectively. Since the meaning of the subject takes the meaning of the VP as its argument, the latter is in the semantic scope of the former.15 Another instantiation of this same schema is the synonymy of the English examples in (38a) and (38b) (the lead-in sentence in parentheses is given to exclude the irrelvant VP-focus reading): (38)

(She didn’t kiss Egon.) a. She only kissed KURT. b. She kissed only KURT.

Here, XP is the object DP, Z is zero, and Y is the verb kissed. Semantically, the object DP takes the verb as its argument,16 which means that the meaning of Y is in the scope of XP; accordingly, the synonymy of (38a) and (38b) is predicted. The same argument goes through if the VP is assumed to contain the subject trace, given that the meaning of the trace (a variable of type e) becomes an argument of the verb meaning, which in turn becomes an argument of the object meaning. The reader can verify that all the examples discussed so far instantiate this general schema. 15 Subjects which take the meaning of a VP as their arguments are interpreted as generalized quantifiers. This means that the subject DP Fritz in (36) is a generalized quantifier, a necessary assumption in order to interpret [nur Fritz]. But even if Fritz is interpreted as individual-denoting, the equivalence holds, given that neither the individual nor the verb meaning are scope-bearing elements. 16 That is after ‘geaching’ from type et,t to type e, et, et. See also note 15 above.


251

4. T HE (A BSENCE OF ) A MBIGUITY A RGUMENT

In this section we will address the quantifier raising argument against an adverbial-only theory mentioned in 1.2. As we have demonstrated in the previous section, there is no need to have the FP and its focus be a constituent to get the association with focus effect. According to the quantifier raising argument the [DP FP DP] constituent is needed nevertheless, because [DP FP DP] can undergo quantifier raising (henceforth QR) together, which in turn has detectable semantic consequences. It will turn out, however, that while English provides a case for QR [DP FP DP] (and thereby a fortiori for the existence of [DP FP DP]), German does not. The conclusion to be drawn from this is that either German has no QR or that German has QR for DP, but not for [DP FP DP]. In the former case, the argument that a constituent [DP FP DP] must exist because it undergoes QR turns out to be a non-argument. In the latter case we are dealing with an argument against [DP FP DP], given the problem of how to block QR of [DP FP DP] under such an analysis. In subsection 4.1 we present the observation of Taglicht (1984) that a FP followed by a focused DP gives rise to ambiguities of scope in English. As von Stechow (1991) notes, a similar ambiguity holds in German. Both authors assume quantifier raising of [DP FP DP] in order to account for the observed ambiguity. In subsection 4.2 we show that the pertinent scope ambiguity can be read off the surface structure in German. Hence QR does not have to be assumed, rendering von Stechow’s implicit argument for [DP FP DP] in German invalid. Finally, in subsection 4.3 we turn to discuss a possible extension of our analysis to English. According to Kayne (1998), the scope ambiguities with focus particles follow exclusively from various steps of overt movement. 4.1. Scope Ambiguities and the QR-Approach Taglicht (1984) notes that in English a focus particle followed by a focused DP gives rise to ambiguities of scope. Consider our example (39) (= Büring and Hartmann’s (1995) ex. (1)): (39)

They were advised to play only Rock’n Roll. a. They were advised not to play anything but Rock’n Roll. b. Rock’n Roll is the only music they were advised to play.

This sentence has at least two readings, paraphrased in (39a) and (39b). The readings differ with respect to the uttered advice. In the first one the

252


advice is like this: Don’t play anything but Rock’n Roll! Don’t play Reggae, don’t play Drum&Bass, don’t play Mozart, etc. In the second reading, Rock’n Roll is the only kind of music they were advised to play. So the advice goes: Play Rock’n Roll! No other advice is given. You can play other music if you like, as long as you play Rock’n Roll. For reasons to be explained presently, we will refer to these readings as narrow scope, (39a), and wide scope, (39b), reading of only DP, respectively. To derive the narrow scope reading of (39), [DP only DP] either remains in situ or adjoins to the embedded IP. The latter option – which is more perspicuous – is illustrated in the LF in (40) (throughout this section we assume a CP cum PRO analysis of infinitival clauses, but nothing in the argument hinges on this): (40)a. they were advised [CP [IP [DP only Rock’n Roll]1 [IP PRO to play t1 ]]] b. they were advised to make Rock’n Roll the only music with the property: λx.they play x Adjunction to the matrix IP gives us the wide scope reading, as illustrated in (41): (41)a. [IP [DP only Rock’n Roll]1 [IP they were advised [CP PRO to play t1 ]]] b. Rock’n Roll is the only music with the property: λx.they were advised to play x On the face of it, a similar situation holds in German. As von Stechow (1991) notes, a sentence like (42) is equally ambiguous between a narrow and a wide scope interpretation: (42)

(weil) ich nur GERDAF geküsst zu haben bereue kissed to have regret (because) I only Gerda

a. I regret to have kissed nobody but Gerda. b. Gerda is the only person that I regret to have kissed. Again, we paraphrase the two readings. In (42a) the speaker’s regret has the implication that (s)he should have kissed other people. In (42b) (s)he did kiss other people, but Gerda is the only person (s)he didn’t enjoy kissing; all others were fun to kiss.


253

In addition, von Stechow (1991) observes that if the embedded sentence is extraposed, the reading corresponding to the wide scope interpretation (42b) is no longer available. (43)

(weil) ich es tCP bereue [CP nur GERDAF geküsst zu regret only Gerda kissed to (because) I it haben] have

a. I regret to have kissed nobody but Gerda. (I should have kissed other people.) b. ∗ Gerda is the only person that I regret to have kissed. (The other ones were fun to kiss.) Von Stechow suggests the following explanation for the German data. First, the ambiguity of (42) exists for the same reason as that of the English example (39): [DP FP DP] may QR to different positions, resulting in different scopes. Second, the extraposed infinitival clause is an island for movement out of it. Therefore, QR of [DP FP DP] can only target the embedded IP, making the wide scope reading impossible. Note that this explanation crucially builds on the idea that [DP FP DP] forms a constituent. In the following subsection we show that QR is not necessary in order to account for the ambiguity observed in (42) as well as for the lack of ambiguity in (43). 4.2. The in situ Analysis In contradistinction to von Stechow’s analysis, we claim that the two readings in (42), as well as the absence of the wide scope reading for (43), follow directly from surface interpretation of adverbial FPs. The cornerstone of our analysis is that the ambiguity in (42) is due to different adjunction sites of only at s-structure. Due to the OV nature of German, the sequence ‘FP CP V’ is syntactically ambiguous; two possibilities are schematized in (44):17 (44)a. ich [VP nur [VP [CP PRO GERDA geküsst zu haben] bereue]] I only G. kissed to have regret 17 A third option is to adjoin the FP to the embedded CP. For reasons to become clear

later, we will not discuss this option, which results in the same interpretation as that of IP adjunction.

254


b. ich [VP [CP PRO [VP nur [VP GERDA geküsst zu haven]]] I only G. kissed to have bereue] regret To derive scope ambiguities with only we make use of these different structures. If only is attached to the embedded VP, as in (44b), we get the narrow scope reading: ‘I regret that the only person. . . ’. Adjunction to the matrix VP, as in (44a), accounts for the wide scope interpretation. In the latter case, the particle c-commands the matrix verb, yielding ‘I only regret that. . . ’. We will calculate these interpretations presently, using Rooth’s in situ theory of focus interpretation outlined in the previous section. Before doing so we want to address an issue the reader might have wondered about, namely, whether the wide scope structure proposed in (44a) is compatible with the Particle Theory, and the Closeness principle in particular. For convenience, we repeat the Particle Theory as it stands: The Particle Theory (preliminary version, repeated) (45)


(46)

β is an EP (extended projection) of some V γ β is a maximal projection β dominates α or is identical to α there is no EP β of γ such that β dominates β and β meets (45b) and (45c).


Since the structures in (44a) and (44b) are identical (except for the attachment site of the FP) and have the same focus, how can they both be possible? To answer this question, let us calculate the f-nodes for (43): (47)a. maximal projections in the sentence: CPmatrix, IPmatrix , VPmatrix , CPembedded, IPembedded, VPembedded, DPsubject, DPPRO , DPobject b. EVP among these: CPmatrix, IPmatrix , VPmatrix, CPembedded, IPembedded, VPembedded c. EVPs dominating the focus: CPmatrix, IPmatrix , VPmatrix , CPembedded, IPembedded, VPembedded


255

So far so good. The remaining candidate nodes in (47c) are ordered by dominance from left to right. Accordingly, Closeness in (45d) will exclude the two higher EVPs of the verb bereuen, ‘regret’, CPmatrix and IPmatrix because there is an EVP of bereuen, VPmatrix, which they dominate and which dominates the focus (β in the sense of (45d)). By the same reasoning, within the EVP of küssen, ‘kiss’, CPembedded and IPembedded are excluded because of VPembedded, which they dominate. This leaves us with VPmatrix and VPembedded as candidates for f-nodes. Does the latter exclude the former (given that it is dominated by it and hence closer to the focus)? No, because, crucially, (45d) compares only extended projections of the same verb, γ . Since VPmatrix is an EVP of bereuen, but VPembedded is an EVP of küssen, neither can block (i.e., qualify as β with respect to) the other. The phrase marker for a sentence like (43) thus contains two f-nodes, allowing for two different FP attachments.18 Generally, a structure will have as many f-nodes for a given focus as it has distinct EVPs dominating that focus. As we will see later, this is a very general and, we believe, fundamental property of FP placement in German. Going back to our examples, let us first look at the narrow scope reading. Bereuen ‘to regret’ is a control verb which takes an infinitive CP complement. According to (44), the focusing particle is adjoined to the VP of the embedded clause: (48)

18 An alternative that comes to mind is that phonetically null elements do not count in terms of Closeness. However, as we will see below, a FP can be attached within the matrix clause even if the focus is separated from it by overt material, provided that material belongs to the embedded clause.

256


The meaning of VP1 can be derived step by step: (49)

[[VP3 ]] = the property of having kissed Gerda (→ ordinary value) 3 f [[VP ]] = the set of properties p, where p = to have kissed some alternative to Gerda (→ focus value) 2 [[VP ]] = the property of having kissed (Gerda and) no alternative to Gerda 1 [[VP ]] = the property of regretting that one kissed (Gerda and) no alternative to Gerda

The denotation we worked out corresponds exactly to the narrow scope interpretation in (42a) (I regret to only have kissed GERDA). To get the wide scope reading, nothing but the attachment site of only changes: it is adjoined higher in the tree, namely to the matrix VP: (50)

Again, the meaning of (50) can be derived strictly compositionally: (51)

[[VP2 ]] = the property of regretting that one kissed Gerda [[VP2 ]]f = the set of properties p, such that p = to regret to have kissed some alternative to Gerda 1 [[VP ]] = the property (of regretting that one kissed Gerda and) of not regretting to have kissed any alternative to Gerda

Note that neither (48) nor (50) require the focusing particle or the focused DP to move in order to occupy the relevant scope position. And notice again that both sentences are string-identical. The difference between them lies in the scope of the particle, not in the focus. We now return to the examples with extraposition. Recall that these sentences lack the wide scope reading. How can the in situ theory account


257

for this fact? Above we claimed that in order to get the wide scope reading in (50) or (42b), the focusing particle only has to c-command the matrix verb at s-structure. This is clearly not the case in (43), the example with the extraposed infinitival clause. (52) gives the structure of this example. Only is adjoined within the extraposed CP, more precisely to the VP therein (by Closeness). Hence, it cannot take scope over the matrix verb. The missing wide scope reading thus follows without any further assumptions. (52)

[[VP]] = [[VP1 ]] in (48): the property of regretting that one kissed (Gerda and) no alternative to Gerda To sum up this section, we have shown that there is an alternative to LF raising of [DP only DP] in German, namely in situ interpretation of the FP. Exploiting the peculiarities of the German OV-structure, we accounted for the narrow and wide scope readings in (42a) and (42b) and also for the lack of the latter with extraposition, without invoking LF movement. Therefore, the data discussed do not present an argument in favor of a mixed theory.19 4.3. Kayne’s Non-LF Analysis The heart of the analysis proposed here is that the scope ambiguities found with German focus particles depend on the particle’s adjunction site. Given the German OV-structure, we showed that the ambiguity arises at s-structure. Our theory does not say anything about English, although 19 One might ask whether an argument against [ DP FP DP] could be made from the

absence of wide scope readings for FPs. Such an argument would show that in a sentence with the surface structure [. . . FP DP. . . ], DP can raise to gain wide scope, but the FP cannot raise along with it. We are not aware of any such construction, which is, we believe, due to the fact that QR is not attested in German in the first place (see e.g., Frey (1993), Zimmermann (1997)).

258


as mentioned, English exhibits a similar kind of ambiguity. A theory that bans adnominal only universally, and hence presents another alternative to LF-raising of only DP is proposed in Kayne (1998).20 Kayne analyzes a variety of different instances of scope ambiguities as resulting from various steps of exclusively overt movement. The discussion includes ambiguities of focus particles of the type found in (39) and (42). His central claim is similar to our proposal for German: only is an exclusively adverbial particle, and the ambiguity arises from a different (merging) position of the focus particle, rather than from QR. In the wide scope reading, the FP is merged with VP in the matrix sentence, in the narrow scope reading it is merged into the embedded clause. Hence there is no [DP FP DP] constituent. The lack of the wide scope reading in the German example (43) follows from the fact that since the particle cannot be analyzed as being part of the matrix clause, it necessarily belongs to the extraposed embedded clause. In all these respects Kayne’s analysis converges with the one proposed in Büring and Hartmann (1995) and in the present paper. The pertinent scope ambiguities in VO-languages like English cannot be read from the surface structure directly, which led to the assumption of LF-movement. Kayne’s proposal extends an adverbial-only analysis to such cases without invoking covert movement. The derivation of the different readings instead involves an excessive amount of (mostly vacuous) overt movement operations. We illustrated this for the English example (39), repeated here: (53)

They were advised to play only Rock’n Roll.

The following steps of Kayne’s analysis – which we will refrain from trying to motivate – are necessary to derive (53). In order to get the wide scope reading, only is merged into the matrix clause ((54a)). The FP then attracts the focus constituent to its specifier ((54b)). Subsequently, the focus particle, which is the carrier of a ‘word order feature’, is raised to the respective head for feature checking, (54c). Finally, VP preposes, (54d): (54)a. They were only [VP advised to play Rock’n Roll]. b. They were [Rock’n Roll]1 only [VP advised to play t1 ]. c. They were only2 [Rock’n Roll]1 t2 [VP advised to play t1 ]. d. They were [VP advised to play t1 ]3 only2 [Rock’n Roll]1 t2 t3 . 20 We are grateful to an anonymous NLLT reviewer for bringing this to our attention.


259

The narrow scope interpretation is derived from adjoining the FP to the embedded VP. The derivation is given step-by-step in (55): (55)a. They were advised to only [VP play Rock’n Roll]. b. They were advised to [Rock’n Roll]1 only [VP play t1 ]. (focus attraction) c. They were advised to only2 [Rock’n Roll]1 t2 [VP play t1 ]. (FP movement) d. They were advised to [VP play t1 ]3 only2 [Rock’n Roll]1 t2 t3 . (VP preposing) ‘Focus attraction’ and ‘particle movement’ are also involved in Kayne’s derivation of the German data, though mostly string-vacuously. While we think that extending our general line of argument to English is tempting, and may help shed some light on the parallels noted in Bayer (1996) and our note 28 below, we feel a certain concern about the loss in predictive force that the unrestricted nature of movement operations assumed in Kayne (1998) brings along with it (cf. Büring and Hartmann (1997)). A more detailed comparison of the two approaches to the German data has to await a later occasion.

5. T HE L ACK OF R ECONSTRUCTION OF FP S A DJACENT TO DP

So far, we have seen that banning adnominal FPs in German explains a number of distributional facts, in particular the absence of FPs within DPs and PPs; furthermore the correct interpretation for all the attested cases can be derived by positing only adverbial FPs. We also showed that arguments for LF movement of the alleged [DP FP DP] do not stand closer scrutiny. In this section, we will show that a FP adjacent to a DP cannot undergo reconstruction at LF, even where the DP alone clearly can. The structure of the argument is the mirror image of that discussed in the preceding section: DPs can undergo reconstruction in German, but FPs, even in cases when associated with and adjacent to a DP that undergoes reconstruction, cannot. Thus it follows that the FP and the DP do not form a constituent. (56) shows that topicalized object DPs in German can reconstruct. In (56a) the indefinite object einen Fehler, ‘one mistake’, on the predominant reading of this sentence, takes scope below the subject jeder, ‘everyone’, so the sentence can mean ‘Presumably everyone made some mistake’.

260


This reading is derived if the topicalized object can reconstruct to its base position (indicated by the trace in (56a)), which is below the subject position: gemacht. (56)a. [Einen Fehler]1 hat vermutlich jeder t1 a-ACC mistake has presumably everyone-NOM made Presumably, everyone made a mistake. (amb.) b. [Seinei Frau]1 respektiert jeder Manni t1 . his wife respects every-NOM man Every man respects his wife. The same kind of reconstruction happens in (56b), where the pronoun seine, ‘his’, in the topicalized object DP is interpreted as a variable bound by the subject quantifier jeder Mann, ‘every man’ (yielding a reading according to which every man respects his own wife). The prerequisite c-command relation between jeder Mann and seine is obtained after reconstruction of the topicalized object DP to its base position, marked t1 in (56b). Let us now turn to cases of object topicalization with a FP in sentenceinitial position: (57)a. Nur MARIA liebt jeder tobject . only M. loves everyone-NOM Only Mary is loved by everyone. NOT: Everyone loves only Mary. a . Jeder liebt nur MARIA. gelesen. b. Nur das ABSTRACT hat jeder tobject . only the abstract has everyone-NOM read Only the abstract was read by everyone. NOT: Everyone read only the abstract. b . Jeder hat nur das ABSTRACT gelesen. These sentences have only an interpretation in which the FP has scope higher than the subject quantifier, as indicated in the glosses.21 (To express 21 But an anonymous reviewer reports consistent judgements of inverted scope on these

and similar examples like Nur Fleisch aß niemand (‘Only meat was eaten by no-one’),


261

a reading according to which Mary is the only one loved at all, and no one read anything but the abstract, the primed sentences, with the subject overtly c-commanding the FP must be used.) These data follow directly from the adverbial-only analysis, according to which the particle must be attached to the root CP, as illustrated in (58a), rather than to the DP in SpecC, as in (58b). Even if the DP is reconstructed to its base position, analogously to the cases in (56), the particle remains adjoined to CP, where it inevitably gets wide scope: (58)a. [CP FP CP] derivation, yielding the correct reading even with reconstruction: SS: [CP FP [CP DP1 C0 Subj. t1 V]] C0 Subj. DP V]] LF: [CP FP [CP b. [DP FP DP] derivation, yielding the unattested reading: SS: [CP [DP FP DP]1 C0 Subj. t1 V] C0 Subj. [DP FP DP] V] LF: [CP Our argument, then, is that a mixed theory does not explain why the FP cannot reconstruct together with the DP in a configuration like (58b), even though we know that DP alone in principle can (as evidenced in (56)). As a last step we will strengthen this argument by showing that the FP cannot reconstruct, even if the DP associated with it actually does. That is, we will present sentences which clearly involve narrow scope objects alongside wide scope particles. Consider (59), where we find the sequence ‘FP DPobject’ sentenceinitially. Here the object DP reconstructs at LF, as shown in (59a), so that the pronoun contained in it can be bound by the subject quantifier. But a construal by which the FP, too, reconstructs beneath the subject position, (59b), is just as impossible as in (57) above: (59)

Nur ein Bild von seiner FRAU besitzt jeder Mann t. only a picture of his wife possesses every man

especially if Fleisch, ‘meat’, receives a secondary accent, and niemand, ‘nobody’, bears the main accent. We believe that the oddity of such an example on the FP wide scope reading can be explained if the secondary accent is taken to indicate an s-topic along the lines of Büring (1997) (the effect is even more striking with an accent on nur itself). But even with this intonational pattern we still fail to get an inverted reading. We leave this issue open for further investigation.

262


a. LF: only

possesses every mani [a picture of hisi wife]

The only person every man possesses a picture of is his wife.22 b. ∗ LF:

possesses every mani [only a picture of hisi wife]

Every man only possesses a picture of his wife. The same is illustrated by (60), where we find a topicalized object DP with a relative clause containing a negative polarity item (NPI). (Witness the contrast between niemand, ‘nobody’, and jemand, ‘somebody’.) Again, the DP has to reconstruct in order for the NPI to get into the scope of the negative subject. Yet the only reading for the sentence is one where the FP retains its wide scope:23 (60)

Nur die Hoffnung, dass wir je wieder GEWINNEN1 hat has only the hope that we ever again win niemand/∗ jemand tDP behalten. nobody/∗ somebody retained

a. LF: only [CP win again]]

has nobody retained [DP the hope that we’ll ever

The only thing nobody retained was the hope that we ever win again. 22 This example contains enough remarkable features to deserve a paper of its own, most

of them, however, orthogonal to the issue at hand. Seiner Frau, ‘his wife’, is focus, and [[his wife]]f must range over (something equivalent to) skolem functions from individuals to individuals. In semi-technical parlance the sentence means (i)

the only individual-valued function f, such that every man x possesses a picture of f(x) is the wife-function

The non-attested FP-reconstruction reading would be (ii)

for every man x, the only individual-valued function f, such that x possesses a picture of f(x) is the wife-function

(i) but not (ii) is compatible with every man possessing more than one picture. See Jacobson (1994) and Sharvit (1997) for discussion of such functional readings, and Krifka (1992) for the technical machinery needed to make this work in a framework with free variables. 23 Incidentally, examples like (59) and (60) also provide direct and compelling evidence that non-surface scope in these examples is due to reconstruction, not QR. Raising the subject quantifier across the object would again yield the impossible scoping subject [ FP [ object . . . .]].


263

has nobody retained [only the hope that we’ll ever b. ∗ LF: [CP win again]] Nobody retained only the hope that we ever win again. These scopings are as predicted by our adverbial-only analysis. Topicalized DPs can freely undergo reconstruction, regardless of whether or not they are associated with a FP. FPs, being base-generated in adverbial positions, cannot.24 The scopings are completely unexpected given a mixed analysis. We thus conclude that the reconstruction facts provide clear and strong evidence for our adverbial-only analysis, including the assumption that FPs can adjoin to root CPs.

6. CP C OMPLEMENTS

In this section, we will extend our analysis to CP complements. We will see that ‘FP CP’ does not form a constituent any more than ‘FP DP’ does. This will lead to a reconception of the reason why [DP FP DP] is excluded. 6.1. FPs Cannot Adjoin to Argument CPs We start by observing that the same tests we applied to ‘FP DP’ initial constructions in the last section indicate that ‘FP CP’ initial constructions also need to be analyzed as involving attachment of the FP to the matrix clause. First, they display only a wide scope reading for the FP:

24 FPs can of course take non-surface scope, if they are within a bigger constituent which reconstructs. Thus in (i), the NPI FP auch nur (lit. ‘also only’, which means the same as even in negative context) is sitting within a topicalized object clause. For it to be licensed it needs to be c-commanded by the subject at LF, which means that the object clause reconstructs.

[PRO Auch nur Spanisch zu sprechen] fiel niemandem t leicht. even only Spanish to speak fell nobody easy

(i)

Our adverbial-only analysis accurately draws the line between acceptable cases like (i), where the topicalized constituent contains an EVP, and those like (59), or the minimally contrasting (ii), in which it does not. Again, a mixed analysis would predict them to be on a par. (ii)

∗

Auch nur [Spanisch] fiel niemandem t zu sprechen leicht. even only Spanish fell nobody to speak easy

264


(61)

[Nur [CP only

dass that

MARIA Maria

Hans Hans

geküsst kissed

hat] has

wussten knew

wir [VP tCP tv ]]. we a. LF1: only [CP that Maria. . . ] knew we tCP The only thing we knew is that MARIA kissed Hans. b. ∗ LF2:

knew we [only that Maria. . . ]

NOT: We knew that only MARIA kissed Hans. The absence of the reading in (61b) follows if we assumed that the embedded CP can never be the adjunction site for the FP.25 (For a different view, which assumes LF-raising of the FP, cf. the appendix in 9.) The parallelism to the DP case extends to cases where the complement CP is forced to undergo reconstruction. The FP retains wide scope, as is demonstrated in (62), where a pronoun in the complement clause is bound by the subject, and (63), where the NPI je needs to be licensed by the subject: (62)

Nur [CP dass eri MARIJUANA raucht]1 versucht jederi only that he marijuana smokes tries everybody zu t1 verheimlichen. to hide

tries everybodyi PRO to [that hei marijuana a. LF: only smokes] hide The only thing that everybody tries to hide is that they smoke marijuana.

25 In fact, unlike in the case involving ‘FP DP’, the scoping facts are hard to explain under the assumption that the FP is adjoined to the embedded CP. Interpreting CP as a set of worlds, (61) would be predicted to mean the same as (i):

(i)

We know that only MARIA kissed Hans.

To get the correct interpretation one would have to assume that sentences are obligatorily type-lifted to a semantic functor of e.g., type st, et, et (for example λpλx.p(∧ [kissed(hans)(mary)])(x)). However, this interpretation also predicts obligatory semantic reconstruction of both CP and the FP, an unwanted result, as we will demonstrate.


265

tries everybodyi PRO to [only that hei marijuana b. ∗ LF: smokes] hide Everybody tries to only hide that they smoke marijuana. (63)

Nur [CP dass wir je wieder GEWINNEN]1 wagt niemand only that we ever again win dares nobody zu t1 hoffen. to hope

a. LF: only

dares nobody to hope [that we’ll ever win again]

The only thing nobody dares to hope is that we ever win again. b. ∗ LF:

dares nobody to hope [only that we’ll ever win again]

Nobody dares to only hope that we ever win again. We conclude that even though the complement CP can – and sometimes must – undergo reconstruction, the FP cannot. Evidently, it cannot attach to the embedded CP, just as it cannot attach to a DP. The prohibition of FPs to adjoin to complement CPs, however, is not expected given what we said so far. After all, CP is an Extended Projection of V, to which attachment is allowed. What is more, the pertinent restriction cannot be expressed in terms of syntactic category at all. If we tried to block LFs such as (61b), (63b) and (62b) by prohibiting adjunction to CP in general, we would lose along with them the LFs in (61a), (63a) and (62a), where the FP attaches to the matrix CP, leaving no permitted structure for those sentences. This apparent dilemma, however, leads us straight to a significant change in the theory. We propose that the prohibition at work here is in fact one against adjunction to arguments. This prohibition is formulated in (64), the final version of the Particle Theory:

266


The Particle Theory (final version) (64)


(65)

β is a non-argument β is a maximal projection β dominates α or is identical to α there is no EP β of the same head that β is an EP of such that β dominates β and β meets (64b) and (64c).


The major change between the preliminary Particle Theory in (11) and (12) and the final Particle Theory in (64) and (65) concerns clause (64a). While clause (11a) of the preliminary Particle Theory stated that an f-node must be a verbal extended projection, (64a) says that it must not be an argument. Note that we no longer need to restrict adjunction to verbal categories. The reason that FPs cannot adjoin to DPs is that the latter are always arguments in German.26 Like clause (64b), discussed in subsection 2.2 above, the new clause (64a) presumably does not need to be stated as part of the Particle Theory, because it holds more generally. A prohibition of adjunction to arguments was first proposed in Chomsky (1986) for movement-created adjunction. McCloskey (1998) presents a detailed study in which he shows that it should also hold of base generated-adjunction.27 The crucial difference between root CPs and subordinate CPs, then, is that only the latter are arguments. Therefore, attachment to the root CP is possible, whereas attachment to XPs in SpecC, including CPs in SpecC, is impossible if XP is an argument. The absence of readings (61b) and (62b) is thus captured.28 Let us explore a number of further consequences of the hypothesis that adjunction to argument CPs is banned. 26 In contradistinction to the old version of the Particle Theory, (64) wrongly permits adjunction of FPs to NPs within DPs such as ∗ der sogar MANN mit dem Hut, ‘the even

man with the hat’. We have to leave this problem open. 27 As a reviewer points out, this generalization of the prohibition against adjunction to arguments implies that appositive modifiers to argument DPs must be analyzed to be DP-internal, rather than DP-adjoined. 28 English seems to obey the same prohibition against adjunction of FPs to argument clauses. Adjunction to DP-internal CPs is impossible, as shown by (i) and (ii): (i)

∗ The fact only that JOHN came. . .

(ii)

∗ The question only who LOVES her. . .


267

First note that while CPs following the finite verb in C0 (occupying the so-called ‘Mittelfeld’) are often preceded by FPs, these cases can be reanalyzed along exactly the same lines as those of ‘FP DP’ in section 3 above, together with the remarks about the semantics made there. That is, in all those cases we can assume that the FP is attached to VP or IP containing the CP. Second, just as with DPs, it can be observed that unambiguous [CP FP CP] structures yield ungrammatical sentences. Thus, [DP N FP CP] is impossible, (66a), as is [AP A FP CP], (66b), and [P P P FP CP], (66c): (66)a. ∗ [DP die Behauptung [nur [CP dass MARTHA gekommen ist]]] the claim only that Martha come is the claim (only) that Martha came b. ∗Ich bin [AP froh [nur [CP dass MARTHA gekommen ist]]]. that Martha come is I am glad only I am glad (only) that Martha came. c. ∗ Sie ging [PP ohne [nur [CP dass ICH wusste, warum]]]. without only that I knew why she left She left without (only) me knowing why. All these cases follow straightforwardly, assuming that CP here is an argument of N, A, and P, respectively, and can thus not be adjoined to. On the face of it, examples like (iii) show adjunction to a sentential object: (iii)

John said only that Peter LOVES Mary.

It may be argued, however, that (iii) should rather be analyzed as right-adjunction of only to VP as in (iv) (thanks to G.K. Pullum for pulling this and 68 similar examples from the 87/9 Wall Street Journal corpus for us), combined with CP extraposition: (iv)

Bank of America is believed to have received more than 40 inquiries since it decided to sell the unit last November, although it chose to deal seriously with Mr. Schwab only.

This analysis gains some plausibility from the fact that the ‘VP FP CP’ pattern does not occur with complementizer-less CPs, which are known not to extrapose: (v)

∗ John said only Peter LOVES Mary.

Clearly, right-adjunction of FPs is not an option for German. Thus, (66b) and (67c) below cannot be analyzed in the same way.

268


Third, as already noted by Jacobs (1983), prohibiting FPs adjoined to argument CPs provides the key to solving a notorious problem of German syntax, namely that ‘FP CP’ sequences are systematically excluded in extraposed position: (67) a.

Spec C C0 [IP nur dass sie kommt habe ich

b.

ich

c.

∗ich

I

V] gesagt

habe nur dass sie kommt gesagt habe have

gesagt nur dass sie kommt said only that she comes

According to our analysis, nur in (67a) and (67b) is attached to the matrix CP and VP, respectively, i.e., to an EVP of the matrix V. No such analysis is available for (67c): If the particle right-adjoined to some matrix EVP, it would violate (64c) by failing to c-command the focus (it would also violate (65), which only allows for left-adjunction). The alternative analysis for (67c) would have the particle adjoined to the extraposed CP, but that would violate (64a) by adjoining to an argument. Therefore, (67c) is correctly ruled out by our analysis. Note that our solution to the puzzling distribution of ‘FP CP’ sequences provides us with an unambiguous position to tell whether something is a constituent or not. A ‘FP XP’ sequence in extraposed position can only be an instance of a FP adjoined to XP. We are now in a position to check a further prediction made by (64a), namely that only non-arguments can be preceded by a FP if extraposed. Unfortunately, DPs do not generally extrapose in German, but CPs do (PPs do markedly, but hardly at all if they contain focal stress; cf. also section 7 below). We already saw in (67c) that argument CPs do not allow FPs in extraposed position. However, just as predicted, adjunct CPs do allow FPs if extraposed, as shown in (68) ((68d) is from Bayer (1996)).29 (68)a. Karl hat sein Fenster mit Styropor verklebt, nur damit er Karl has his window with styrofoam glued only so-as he Ruhe hat. peace has Karl has glued styrofoam to his windows only so as to have a bit of peace. 29 Surprisingly, it is impossible to adjoin FPs to relative clauses, be they extraposed or

not. We do not have an explanation for that fact. It is interesting, however, that McCloskey (1998) finds that relative clauses in English, too, do not allow for adjunction.


269

b. Peter will ihm nicht die Hand geben, sogar nachdem er sich Peter wants him not the hand give even after he self entschuldigt hat. apologized has Peter won’t shake his hand even after he apologized. c. Maria will kommen, auch wenn sie nicht eingeladen ist. Maria wants come even if she not invited is Maria wants to come, even if she is not invited. d. . . . weil Hans hereingekommen wäre, nur wenn alle . . . bec. Hans entered had only if all geschlafen hätten. sleeped had . . . because Hans would have entered only if everyone had been asleep. Summing up this section, we started out by observing that CP arguments behave just like DP arguments, in that a FP preceding a focus-containing CP cannot undergo reconstruction along with that CP, even if the CP clearly does. Likewise FPs cannot occur adjacent to CPs inside DPs, APs, and PPs, just as they cannot next to DPs in these positions. We proposed to generalize over both cases by alluding to a general ban against adjunction to arguments as proposed by Chomsky (1986). In addition to deriving the facts about CP distribution, this ban turned out to derive the right distinction between argument CPs and adjunct CPs, and, perhaps most importantly, derive the general ban on ‘FP CP’ sequences in extraposed position, as far as argument clauses are concerned. 6.2. A Return to the Verb-Second Argument As a last remark, we want to point out that taking CP-complements into consideration also sheds new light on the verb-second issue. Recall that the issue is that adjunction of a FP to CP creates structures in which more than one constituent precedes the finite verb in C0 . A disadvantage of allowing this is that one can no longer guarantee the otherwise strict verb-second nature of German by stipulating a general ban against adjunction to CP such as (69): (69)

Adjunction to CPs is prohibited.

270


Rather, the ban required under the adverbial-only analysis must be one against adjunction of most elements, but not all (in particular, not FPs), to CP. For concreteness, let us adopt (70) (cf. note 13, though): (70)

Only FPs can adjoin to CP.

As pointed out in section 1, we believe that much of the scepticism against Jacobs’ (1983) original proposal is fueled by a reluctance to give up (69) in favor of something like (70). Against this background consider (71): (71)

Nur um welche Zeit wir GEGANGENF sind verraten wir nicht. are tell we not only at which time we left The only thing we don’t tell is at what time we left.

Here we have a verb-second clause with an initial FP and a topicalized embedded wh-interrogative clause. According to our proposal, the FP must adjoin to the matrix CP, as in (72): (72)

[CP FP [CP [CP∗ [PP wh. . . ] C0 . . . VF ] C0 . . . ]]

Could a proponent of a mixed theory analyze (71) without adjunction to the root CP? (73a) and (73b) show conceivable alternative structures in which no adjunction to the root CP is assumed: (73)a. [CP [CP∗ [PP FP [PP wh. . . ]] C0 . . . ] C0 . . . ] b. [CP [CP∗ FP [CP∗ [PP wh. . . ]] C0 . . . ]] C0 . . . ] We can immediately dismiss (73a), in which the FP is attached to the whPP in the embedded SpecC, given that this PP does not contain the focus at all. Not only does this violate any conceivable generalization about where FPs attach, it would also give a completely wrong interpretation, given that the FP does not c-command the focus. That leaves (73b), with the FP adjoined to the embedded question in SpecC. But notice that while (73b) maintains that at most one constituent can precede the finite verb in a main clause, it does not maintain that CPs cannot be adjoined to. The reason is that the FP is attached to CP∗ (we deliberately chose an example in which the wh-phrase, being a PP, doubtlessly occupies SpecC). In other words, this analysis of a sentence like (71) must violate (69), too. The question then becomes which theory – mixed or adverbial-only – allows for a more natural refinement of (69). On either account, adjunction of FPs to CP must be allowed, i.e., (70) must replace (69). In our adverbialonly analysis, this is all that needs to be said, given that the only other


271

restriction is the independently needed prohibition against adjunction to arguments. The alternative line would have to invoke an explicit prohibition against adjunction of FPs to root CPs, which, as far as we can see, lacks independent motivation. Such a prohibition would furthermore be a curious instance of a restriction which regards root-clauses only; usually, embedded clauses are the more restricted ones, a generalization first observed in Ross (1973), and aptly named the Penthouse Principle. In sum, then, the fact that the present analysis predicts verb-third main clauses for German cannot be held as a theoretical argument against it. The alternative mixed analysis will have to allow for adjunction of FPs to CP, too, and neither analysis offers an explanation for why this option is restricted to FPs. We submit that the present proposal is preferable because it makes the more accurate predictions with respect to various phenomena, and because it requires no additional stipulations on top of (70). 7. OTHER O CCURRENCES OF FP S

In this section, we will look briefly at FPs occuring in APs, arguing that their behavior parallels that of the adverbial cases discussed so far. Let us start by pointing out that FPs can in fact be found within non-verbal constituents, namely within DPs modified by APs, as in (74): (74)a. eine nur an MUSIK interessierte Studentin a only in music interested student-FEM a student interested only in music b. der sogar mit KARL verfeindete Förster the even with KARL quarreling forest ranger the forest ranger who is quarreling even with Karl c. unser auch von ORIGAMI begeisterter Hausmeister our also of Origami enthusiastic janitor our janitor who is enthusiastic also about Origami Clearly, the FPs are inside the DP, as witnessed by the fact that they are ‘sandwiched’ in between the determiner and the head noun. Within the DP, FP placement follows the same rules seen above, in particular the Closeness principle: (75)a. dein mit Inbrunst nur an DICH denkender Wolfgang your with ardour only at you thinking Wolfgang

272


b. ∗dein nur mit Inbrunst an DICH denkender Wolfgang your only with ardour at you thinking Wolfgang your Wolfgang, thinking ardously of you only Our theory predicts that the FP must be adjoined to the whole AP, not to the PP within it: While the latter is an argument (to the adjective) the former is not. The structure of (74a) must thus be (76):30

30 This theory also predicts that unmodified APs, too, should allow for adjunction of

FPs. While this seems true for some cases, as in (i) and (ii), additional restrictions occur, as (iii), provided by a reviewer, shows: (i)

ein nur mittelmäßiger Student an only mediocre student

(ii)

eine nur unwesentliche Änderung an only inessential change

(iii)

∗ eine nur rote Tasche

an only red bag Interestingly, these restrictions are paralleled with VP-adjoined FPs: (iv)

Der Student ist nur mittelmäßig. the student is only mediocre

(v)

Die Änderung ist nur unwesentlich. the change is only inessential

(iv)

?∗ Die Tasche ist nur rot. the bag is only red

It thus seems that additional semantic restrictions on the scope of FPs are at work here, which we have to leave for further research.


273

(76)

That (76) is indeed the correct structure can be demonstrated by an argument parallel to that involving CPs in the last section. As a first step we note that predicative APs allow for local extraposition of the PP argument: (77)a. Der Student muss an Kunst interessiert sein. the student must in art interested be b. Der Student muss [AP [AP t interessiert] an Kunst] sein. interested in arts be the student must c. [CP [AP [AP t interessiert] an Kunst][C sollte er sein]]. should he be interested in arts The PP an Kunst, ‘in art’, in (77b) is right-adjoined to the AP. This can be seen more clearly in (77c), where the entire AP has been topicalized, showing that the ‘A PP’ sequence is in fact a constituent.31 31 We do not want to go into the arguments regarding why we assume PP-A to be the base-generated order. Suffice to say that A-PP is the more restricted pattern being impossible in adnominal APs; and that DP arguments to A can occur only pre-adverbially, which would make A and V completely parallel. See also Haegeman (1995) for further arguments in favor of that assumption.

274


As a second step, we then observe that a FP cannot occur extraposed together with the PP: (78)a. [AP Nur an KUNST interessiert] sollte er sein. only in art interested should he be b. ∗[AP Interessiert nur an KUNST] sollte er sein. interested only in art should he be (78a) shows that a structure like (76), in which the FP is attached to the AP (a non-argument) can be the input to extraposition. The unacceptability of (78b) on the other hand shows that a structure in which the FP is adjoined to the PP, which is an argument, must not be permitted. This is exactly what the theory advocated here predicts. It should be noted that (78a) is in fact ambiguous between the structures (79a) and (79b): (79)a. [CP [AP nur AP] C0 . . . ] b. nur [CP AP C0 . . . ] However, the ambiguity is purely structural, given that there are no other scope-taking elements in the clause. True semantic ambiguity can be found, though, in (80), where the readings can be facilitated by different intonations. Single (nuclear) stress on Eiern, ‘eggs’, promotes the [CP FP CP] reading, while nuclear stress within the IP (say on nicht, with a secondary stress on Eiern) favors the [AP FP AP] reading: (80)

Nur mit EIERN belegt schmeckt es nicht so gut. only with eggs topped tastes it not so good

a. nur adjoined to AP if there are only eggs on it it doesn’t taste as good b. nur adjoined to CP the only way it doesn’t taste as good is with eggs on it


(81)

275

Nur mit EIERN belegt ist es nicht. only with eggs topped is it not

a. nur adjoined to AP it doesn’t have only eggs on it b. nur adjoined to CP the only thing missing are eggs on it Let us now consider (81). As is well known, nur presupposes the truth of its complement. Hence, a sentence of the structure not(only(P)) implies P. This is the case in (81a), where nur is reconstructed beneath the negation, yielding the presupposition that there are eggs on it. A sentence of the form only(not(P)) on the other hand will presuppose not-P. This case is the reading in (81b), which presupposes that it has no eggs on it. Similar remarks apply to (80), which is more complicated in that it is a generic statement: On reading (80b), the FP has scope over the matrix predicate, but on reading (80a) is does not; different truth conditions result. In sum, we have shown in this section that PP-complements to APs behave in interesting ways parallel to CP-complements to V, in that they cannot carry along a FP when extraposed, suggesting that they do not form a constituent with them. Let us add a speculative note to this section. Presumably the analysis of AP-attached FPs can be carried over to numeral expressions, which can be modified by FPs as well: (82)a. mit nur EINEM Wagen with only ONE car b. mit nur ZWÖLF Schlägen with only TWELVE hits c. in nur WENIGEN Sekunden within only FEW seconds On the face of it, all the cases in (82) display adjunction of FPs to argument DPs, which is of course impossible under our proposal. Conceivably, however, the FP is attached to AP here as well, provided that we analyze numerals and quantifiers such as wenig, ‘few’, as adjectives. In favor of

276


this, note that these expressions can co-occur with demonstratives and definite determiners: (83)a. dieser/der eine Wagen this/the one car b. diese/die zwölf Schläge/wenigen Sekunden this/the twelve hits/few seconds Furthermore, they show the typical adjectival inflection, including the alternation between weak and strong declension. Notably, FPs are excluded from co-occuring with elements like definite determiners and demonstratives, which are not adjective-like in any respect. The contrast between (82a) and the ungrammatical (84a) proves this claim: attachment of the FP to a demonstrative is excluded. (84b) is provided to show that this restriction is not semantic in nature; it means exactly what (84a) would mean. (84)a. ∗ mit nur diesem/jenem/dem Wagen with only this/that/the car b. nur mit diesem/jenem/dem Wagen While these cases certainly deserve closer scrutiny, they seem to us to be at least compatible with the generalizations and claims put forward in this paper.

8. S UMMARY

We have presented a theory of focus particle placement in German which accounts for the distributional patterns as well as the semantic properties of constructions involving such particles. The main features of the analysis are: • German focus particles are maximal projections which can be adjoined to non-argument XPs. • The placement of the particles with respect to the focus is governed by a principle that requires particles to be maximally close to the focus within a given extended projection. • German focus particles can adjoin to matrix CPs allowing verb-third main clauses.


277

• German focus particles are interpreted in situ. They do not undergo LF-raising or reconstruction on their own. A number of hitherto unobserved data regarding reconstruction, NPI licensing and scoping have been presented and shown to support the analysis proposed.

9. A PPENDIX : O N LF-R AISING A NALYSES

In section 6.2 we pointed out that there are no conceptual reasons to prefer an analysis that does not assume adjunction of FPs to root CPs over the one proposed here (in fact, we suggested that they should be dispreferred). Conceptual issues aside, let us briefly ask whether and how an alternative analysis could be reconciled with the observed facts. Let us imagine an alternative mixed analysis which preserves the strict verb-second property for root clauses by assuming FPs to be adjoined to an element in SpecC, which we will call XP. The two most problematic facts for such an analysis would presumably be that a sentence-initial FP always takes scope over the matrix verb, even though it would not c-command it under such an analysis; and that a sentence-initial FP cannot undergo reconstruction, even though XP can. While the first problem could be handled by changing the semantic composition in the way sketched in note 25 above, it seems that the only way to meet the second problem at the same time is to assume that the sentences have essentially the structure they have in our analysis, not at s-structure, but only at LF. Such an analysis, which we will call the ‘LF theory’, is proposed in Bayer (1990, 1996). There it is assumed that the FP starts out adjoined to the embedded CP in a sentence like (63), repeated below as (85). (85)

Nur [dass wir je wieder gewinnen]1 wagt niemand zu t1 dares nobody to only that we ever again win hoffen. hope The only thing nobody dares to hope is that we ever win again.

The FP then obligatorily LF-moves to a position from which it ccommands the verb (i.e., a position immediately dominated by an EVP). This position would have to be SpecPrtP, the specifier of PrtP, a component of the IP system. By assumption, the movement would be triggered by a requirement of the FP to be in a Spec/Head relation with a Prt0 , the head

278


of PrtP, at LF. It is in this position that the FP is interpreted, accounting for the fact that sentence-initial FP always take scope over the matrix verb. Evidently, for the case of ‘FP CP’ in sentence-initial position, the movement of the FP would have to follow reconstruction of the assumed [CP FP CP] complex to a position lower than SpecPrt. So it would not immediately follow that a FP could not take scope from a reconstructed position. Let us assume, however, that the landing site of the FP, SpecPrt, is higher than the subject, such that the FP essentially takes scope as if it were adjoined to CP, even though it isn’t. Then a sentence like (85) would get the right interpretation, mediated by an LF as sketched below: (86)

LF:

wagt [PrtP nur [Prt niemand [tnur dass wir je wieder dares only nobody that we ever again

gewinnen] zu hoffen]] win to hope By the same token, though, we would expect the FP to outscope the subject if the assumed [CP FP CP] complex had not been moved in the first place. This, however, is wrong, as the interpretation of (87) illustrates. (87)

Hier wagt niemand nur [CP dass wir je wieder gewinnen] zu that we ever again win to here dares nobody only hoffen. hope

a. Here, nobody dares to only hope that we’ll ever win again. b. ∗ Here, the only thing nobody dares to hope is that we ever win again. Sentence (87) is not easy to understand, but if it has any reading it is (87a). Under no circumstances can it be interpreted to mean the same as (85), i.e., (87b). However, the latter meaning is wrongly predicted by the LF theory, as the two essentially identical LFs in (85) and (87b) illustrate. A second problem with the LF theory is that it does not follow that a FP has to take widest scope if there are embedded SpecPrt positions for it to move to. If PrtP were a regular part of the EVP, one would expect there to be two such positions in a sentence like (85), whose structure – as the LF


theory would have it – we show in (88) (here SpecPrtP positions):

279

marks the two available

(88)

Only the higher SpecPrt will provide the empirically correct scope position for the FP in (85), while the lower will wrongly derive a reading like (87a). Thus the LF theory must have the FP LF-move to that higher SpecPrt position, either from its embedded base position after reconstruction of CP∗ to tCP ∗ , or perhaps, more locally, from an intermediate landing position in the embedded SpecC to which CP∗ would be assumed to have reconstructed. Both of these movements seem unusual from the standpoint of traditional bounding theory. But most of all, it remains completely unclear what would force this movement rather than the apparently straightforward derivation in which the FP ends up in the embedded SpecPrt, yielding the unattested reading (87a) for (85). We do not think that it is impossible to devise some mechanism to remedy this in an LF theory. At this point, we conclude that the theory as it stands does not derive the fact that FPs do not reconstruct. On a more general level, we believe that the LF theory, and presumably any approach along these lines, suffers from the fact that it will always have to ensure that alleged LF movements of FPs never change the scope relations that the FP is engaged in at s-structure (a situation quite opposite to those that usually support arguments for LF movement). We think that the easiest way to derive surface scope is to assume interpretation in surface positions. In the absence of any syntactic or semantic argument to the contrary, we believe this simplicity to be an additional virtue of the theory advocated here.

REFERENCES

Bayer, Josef. 1990. Directionality of Government and Logical Form: A Study of Focusing Particles and Wh-Scope, Habilitationsschrift, Konstanz.

280


Bayer, Josef. 1996. Directionality and Logical Form, Kluwer Academic Publishers, Dordrecht, Boston, London. Büring, Daniel. 1997. ‘The Great Scope Inversion Conspiracy’, Linguistics & Philosophy 20, 175–194. Büring, Daniel and Katharina Hartmann. 1995. ‘Is it [Only Rock’n Roll] Or Just Like It?’, in J. Camacho, L. Choueiri and M. Watanabe (eds.), Proceedings of the Fourteenth West Coast Conference on Formal Linguistics, CSLI Publications, Stanford, pp. 63–77. Büring, Daniel and Katharina Hartmann. 1997. ‘Doing the Right Thing’, The Linguistic Review 14, 1–42. Chomsky, Noam. 1986. Barriers, MIT-Press, Cambridge, MA. Eisenberg, Peter. 1989. Grundriß der Deutschen Grammatik, Metzler, Stuttgart. Frey, Werner. 1993. Syntaktische Bedingungen für die semantische Interpretation (studia grammatica xxxv), Akademie-Verlag, Berlin. Grebe, Paul et al. (eds.) 1973. Duden. Grammatik der Deutschen Gegenwartssprache, 3. Auflage, Dudenverlag, Mannheim. Grewendorf, Günther and Wolfgang Sternefeld. 1990. ‘Scrambling Theories’, in G. Grewendorf and W. Sternefeld (eds.), Scrambling and Barriers, John Benjamins, Amsterdam/Philadelphia, pp. 3–37. Grimshaw, Jane. 1991. Extended Projections, Ms. Brandeis University. Haegeman, Liliane. 1995. The Syntax of Negation, Cambridge University Press, Cambridge. Heidolph, Karl Erich et al. 1981. Grundzüge einer Deutschen Grammatik, Akademieverlag, Berlin. Helbig, Gerhard. 1988. Lexikon Deutscher Partikeln, VEB Verlag Enzyklopäie, Leipzig. Jacobs, Joachim. 1983. Fokus und Skalen, Niemeyer, Tübingen. Jacobs, Joachim. 1986. ‘The Syntax of Focus and Adverbials’, in W. Abraham and S. de Meij (eds.), Topic, Focus, and Configurationality, Benjamins, Amsterdam, pp. 103–128. Jacobson, Pauline. 1994. ‘Binding Connectivity in Copular Sentences’, in M. Harvey and L. Santelmann (eds.), Proceedings of SALT IV, Cornell University, Ithaca, NY, pp. 161– 178. Kayne, Richard. 1998. ‘Overt vs. Covert Movement’, Syntax 1, 128–191. König, Ekkehard. 1991. The Meaning of Focus Particles, Routledge, London/New York. Krifka, Manfred. 1992. Focus, Quantification, and Dynamic Interpretation, Ms, University of Texas at Austin. McCloskey, James. 1998. ‘Adjunction, Selection and CP-Recursion’, Ms, UCSC. Müller, Stefan. in prep. Yet Another Theory of Complex Predicates, Ms, DFKI Saarbrücken. Rooth, Mats. 1985. Association with Focus, PhD dissertation, University of Massachusetts at Amherst. Rooth, Mats. 1992. ‘A Theory of Focus Interpretation’, Natural Language Semantics 1, 75–116. Ross, John R. 1973. ‘The Penthouse Principle and the Order of Constituents’, in C. Corum et al. (eds.), You Take the High Node, Chicago Linguistic Society, Chicago, pp. 397–422. Sharvit, Yael. 1997. The Syntax and Semantics of Functional Relative Clauses, PhD dissertation, Rutgers University. von Stechow, Arnim. 1991. ‘Current Issues in the Theory of Focus’, in A. von Stechow and D. Wunderlich (eds.), pp. 804–825. von Stechow, Arnim and Dieter Wunderlich (eds.) 1991. Semantics – An International Handbook of Contemporary Research, de Gruyter, Berlin. Taglicht, Josef. 1984. Message and Emphasis, Longman, London and New York.


281

Webelhuth, Gert. 1987. ‘Eine universelle Scrambling Theorie’, Paper presented at the annual meeting of the Deutsche Gesellschaft für Sprachwissenschaft at Augsburg. Zifonun, Gisela. 1997. Grammatik der Deutschen Sprache, de Gruyter, Berlin. (= Schriften des Instituts für Deutsche Sprache). Zimmermann, Malte. 1997. ‘An Empirical Study of Quantifier Scope in German’, in Groninger Arbeiten zur Germanistischen Linguistik 41. Rijksuniversiteit Groningen, pp. 205–225. Received 29 March 1999 Revised 25 August 2000 Daniel Büring Department of Linguistics UC Los Angeles 3125 Campbell Hall Box 951543 Los Angeles, CA 90095-1543 USA [email protected] Katharina Hartmann Institut für deutsche Sprache & Literatur II Johann Wolfgang Goethe-Universität Grüneburgplatz 1 D-60629 Frankfurt/Main Germany [email protected]

The Syntax And Semantics Of Focus-Sensitive ParticlesIn German

The Syntax And Semantics Of Focus-Sensitive ParticlesIn German

Suggest Documents

Syntax and Semantics

SYNTAX AND SEMANTICS - arXiv

THE SYNTAX-SEMANTICS INTERFACE

The Syntax and Semantics of ÂµCRL - CiteSeerX

THE SYNTAX AND SEMANTICS OF UNSELECTED ... - CiteSeerX

The Syntax and Semantics of Locating Adverbials

THE SYNTAX-SEMANTICS INTERFACE AND THE DOMINO ...

between semantics and syntax

Russian syntax and semantics

LOIS: Syntax and Semantics - MIMUW

(Syntax)âSemantics

Syntax-Semantics Interface

PSYCHOLOGICAL PREDICATES AND THE SYNTAX-SEMANTICS

Concord and the Syntax Semantics Interface

Construction Grammar and the Syntax-Semantics

Syntax and Semantics of Fluid Construction Grammar

Formal Syntax and Semantics of Programming Languages

The Effect of Syntax-Semantics Mismatch on

Formal Syntax and Semantics of Programming Languages

Syntax and Formal Semantics of Hybrid Chi

The Syntax and the Semantics of the Deontic Modals

Syntax-Semantics Interface - Google Sites

EXPLETIVES AND THE SYNTAX AND SEMANTICS OF COPY RAISING

Syntax-Semantics Interface - Google Sites

The Syntax And Semantics Of Focus-Sensitive ParticlesIn German