2.1 Medial position adjunct PPs: The literature

As pointed out by Haegeman (Reference Haegeman2002), there is a tendency in the generative tradition to consider medial adjunct PPs (such as (3c) and (4c) above) unacceptable in absolute terms, in contrast to medial adverbs. For instance, commenting on (5), Jackendoff (Reference Jackendoff1977:73) says: ‘First let us deal with the differences between AdvPs and PPs in V′′. The most salient difference is that AdvPs may appear preverbally as well as postverbally, whereas PPs may only be postverbal’.

1. (5)

This type of judgment is reiterated in the literature, for example in Emonds (Reference Emonds1976), who treats medial PPs such as those in (3b) and (4b) as parentheticals, and in Nakajima (Reference Nakajima1991), Rizzi (Reference Rizzi and Haegeman1997:301), Frey & Pittner (Reference Frey and Pittner1998:517), Pittner (Reference Pittner1999:175, Reference Pittner, Austin, Engelberg and Rauch2004:272), Cinque (Reference Cinque2004:699–700), Haumann (Reference Haumann2007), Belletti & Rizzi (Reference Belletti and Rizzi2010), and elsewhere. Reproducing the judgment in (5), Cinque (Reference Cinque1999:28) writes:

While we take no issue with the actual judgments of specific examples, the authors’ extrapolation that all medial PPs are ruled out does not correspond to the empirical data.

As a matter of fact, there is no agreement among authors that medial adjunct PPs are unacceptable. For instance, on the basis of the judgments presented below in (6), McCawley (Reference McCawley1998:207) does confirm the general tendency for adjunct PPs to resist medial position, but he also provides the examples in (7) below, with acceptable medial adjunct PPs. He comments: ‘I don't know of any neat way to distinguish between the P's in [6] and the ones in [7]’ (McCawley Reference McCawley1998:214, note 25).

1. (6)

2. (7)

Focussing on journalistic prose, Haegeman (Reference Haegeman2002) shows that medial PPs are regularly attested. The following illustrate a medial adjunct PP in a finite clause without an auxiliary, in (8a), a finite clause with an auxiliary, in (8b), as well as a non-finite clause, in (8c):

1. (8)

Several authors (Quirk Reference Quirk, Greenbaum, Leech and Svartvik1985:492, 514, 521; Ernst Reference Ernst2002a:504, Reference Ernst and Di Sciullo2002b:194; Mittwoch, Huddleston & Collins Reference Mittwoch, Huddleston and Collins2002:780) signal that weight considerations play a part in restricting the availability of non-parenthetical medial PP adjuncts. For a discussion of a definition of weight in determining word order, see e.g. Ernst (Reference Ernst and Di Sciullo2002b:194) and the references cited there.

2.2 Medial position adjunct PPs are rare

While the claim that medial PPs are categorically unacceptable is definitely incorrect, medial adjunct PPs are not as frequent as medial adverbs. Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985) provide an overview of the distribution of a range of adverbial expressions in the various positions in a sample of the Survey of English Usage corpus (see their description in Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985:489). Tables 1 and 2 are based on their Table 8.23 and summarize the percentages of adjunct PPs and adjunct adverbs in initial, medial and postverbal position. While Quirk et al. distinguish a number of medial and postverbal positions, our tables simplify their Table 8.23 in that we have grouped their distinct medial positions into one position and we have also collapsed their postverbal positions into one. Medial PPs are systematically outnumbered by postverbal PPs, both in writing and in speech. For adverbs, the opposite relation holds: medial adverbs are slightly more frequent than postverbal ones. That medial PPs are rare is also occasionally signalled in pedagogically oriented grammars such as, for instance the Collins COBUILD grammar (Sinclair Reference Sinclair1990:283) and Lambotte (Reference Lambotte1998).

Table 1. Distribution of PPs in the Survey of English Usage corpus (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985:501).

^a The discrepancy between the figures for spoken and written material and the totals are not accounted for in Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985).

Table 2. Distribution of adverbs in the Survey of English Usage (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985:501).

^a Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985) do not account for the discrepancy in the totals. Note that we only report on open class adverbs. We don't include in the count closed class adverb such as then, just, etc.

In order to assess the status of medial adjunct PPs in present-day English, we undertook a pilot search of the Corpus of Contemporary American English (henceforth COCA; http://corpus.byu.edu/coca/, COCA 2010) and the British National Corpus (henceforth BNC; http://corpus.byu.edu/bnc/, BNC 2010) in which we examined the distribution of the following temporal adjunct PPs: on three occasions, on those occasions, at one time, at a time, at some time, at this time, at that time, on many occasions and also of the manner adjunct in this way. For adjunct PPs occurring at a very high frequency (at one time, at a time, at some time, at this time, at that time, on many occasions, in this way), we based our study on a sample of the first 100 entries. We present our results in Tables 3 and 4. Obviously, the numbers in these tables in no way represent the full and final picture of the distribution of adjunct PPs, nor does our paper offer a statistical analysis of such data, but our findings suffice to show (i) that sentence-medial adjunct PPs are certainly attested, and (ii) that, fully in line with the literature, such medial adjunct PPs are outnumbered by postverbal adjunct PPs. In Section 3 we will see, however, that for a well-defined class of PP adjuncts, medial position is not just an option but is actually strongly preferred over postverbal position.

Table 3. Pilot study: distribution of PPs in medial position, COCA sample.

^a We have discounted occurrences of at a time followed by a temporal clause (at a time when . . .) which are final for weight reasons and for which medial position is unavailable.

Table 4. Pilot study: distribution of PPs in medial position, BNC sample.

3.1 Sentential negation in English

In English, negation can be expressed in a number of different ways, the most common of which are illustrated in (9). For recent analyses and a survey of the literature we refer to Zeijlstra (Reference Zeijlstra2004), Christensen (Reference Christensen2005, Reference Christensen2008), Moscati (Reference Vincenzo2006, Reference Moscati2011) and Tubau (Reference Tubau2008).

1. (9)

   1. a. The police did not talk to any witnesses.

   2. b. No one talked to the police about any crime.

   3. c. The police associated no one with any of these crimes.

   4. d. The police talked to no one about any of these crimes.

   5. e. The police never talked to any witnesses about the crime.

   6. f. Never had the police talked to any witnesses.

The canonical marker of negation is the particle not (or its contracted form n't) adjacent to the finite auxiliary. Alternatively, an argument of the verb is realized as a negative nominal constituent, such as no one in (9b) or (9c), or as a PP containing a negative nominal as in (9d), which also conveys negation (but see Section 4 for discussion). Finally, and most relevant for our purposes, in (9e) and (9f) a negative adjunct expresses sentential negation. In (9e) the adverb never is medial and in (9f) it is initial, triggering subject–auxiliary inversion (henceforth SAI; see Rudanko Reference Rudanko1987; Haegeman Reference Haegeman2000; Sobin Reference Sobin2003).

Negative adjuncts with sentential scope can also be realized as PPs. In (10a) the negative quantifier no contained inside the initial temporal PP at no time has sentential scope, witness the fact that it triggers SAI and licenses the negative polarity item any in the complement of the verb.Footnote ^1, Footnote ² The negative PP differs from its non-negative counterpart at that time, which does not, and cannot, trigger SAI, as is shown in (11).

1. (10)

   1. a. At no time had the police talked to any witnesses.

   2. b. * At no time the police had talked to any witnesses.

2. (11)

   1. a. At that time the police had interviewed the witnesses.

   2. b. * At that time had the police interviewed the witnesses.

Like negative adverbs, negative adjunct PPs with sentential scope can appear in sentence-medial position, as in (12). The availability of the polarity item any in (12a) confirms that at no time has sentential scope. Though we will mainly focus on temporal PPs like (12a), other medial adjunct PPs can also express sentential negation, see (12b).

1. (12)

   1. a. The police had at no time talked to any of the witnesses.

   2. b. The FQ at no level forms a constituent with the DP it modifies.

      (Will Harwood, p.c.)

In relation to the discussion in Section 2 above, the data in (12) obviously also challenge claims according to which medial adjunct PPs are categorically unacceptable. We go into these patterns in more detail here.

3.2 Negative adjunct PPs and the expression of sentential negation

Sentences with preposed negative constituents such as the pair in (13a,b) have been discussed extensively (see, among others, Rudanko Reference Rudanko1987; Haegeman Reference Haegeman2002; Sobin Reference Sobin2003; Radford Reference Radford2004; Haumann Reference Haumann2007 and the references cited there). In (13a), without SAI, the negative quantifier no contained in the PP in no clothes encodes constituent negation (‘without clothes’) and does not take sentential scope; in (13b), with SAI, the PP-internal negative quantifier has sentential scope (‘there are no clothes such that . . .’).

1. (13)

   1. a. In no clothes Mary looks attractive.

   2. b. In no clothes does Mary look attractive.

Less attention has been paid to the distribution and interpretation of postverbal negative PPs. We briefly consider here some discussions in the literature.

Tottie (Reference Tottie1983) studies the alternation between S[ynthetic] negation (he said nothing) vs. A[nalytic] negation (he did not say anything) in American English, using both informants’ questionnaires and corpus material. However, her data do not include many relevant examples of PPs. Summarizing her conclusions on the basis of the informants’ questionnaires she writes:

1. ([14])

Observe that in the three examples in (14), the medial negative adjunct PP is not set off prosodically. Indeed, in spite of its relative weight, even the PP by no stretch of the imagination occupies medial position in (14c). Inserting commas in (14c) would entail that the negative PP cannot scope over the clause and would render the sentence unacceptable, as is shown in (14c′).

1. (14) c′. *It is, by no stretch of the imagination, a happy choice.

In their discussion of negative markers in English, Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985:783) systematically compare a positive sentence with its negative alternative. Their example set in (15) below is of interest in the light of our discussion. While in the positive (15a) the adverb somehow is in postverbal position, the negative adjunct PP is placed medially in (15d). Quirk et al. do not comment on this shift in position.

1. (15)

   1. a. They'll finish it somehow.

   2. b. They won't in any way finish it.

   3. c. They won't finish it at all.

   4. d. They will in no way finish it. (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985:783, ex. (8))

Pullum & Huddleston (Reference Huddleston and Pullum2002) distinguish ‘verbal’ negation, expressed by medial not or n't associated with an auxiliary, as in (9a) or (15b, c), from ‘non-verbal’ negation, expressed by means of a negative constituent such as a negative quantifier (no, nothing, no one, etc.) or a negative adverb (never, no longer, no more). Relevantly, they provide (16a) as an instance of a non-verbal sentential negation. In this example negation is encoded in a postverbal adjunct PP. Following Klima (Reference Klima, Fodor and Katz1964), McCawley (Reference McCawley1998), Horn (Reference Horn1989), Haegeman (Reference Haegeman2000), De Clercq (Reference De Clercq, An and Kim2010a), and others, the standard diagnostics to detect negativity (16b–e) show that the postverbal negative constituent in (16a) can take sentential scope.Footnote ³

1. (16)

   1. a. We were friends at no time.

      (Pullum & Huddleston Reference Pullum and Huddleston2002:788, ex. [5iia])

   2. b. We were friends at no time, not even when we were at school.

      (Pullum & Huddleston Reference Pullum and Huddleston2002:789, ex. [10ia])

   3. c. We were friends at no time, and neither were our brothers.

   4. d. We were friends at no time, were we?

   5. e. At no time were we friends.

Along the same lines, Haumann (Reference Haumann2007:230) provides (17a), in which postverbal on no account negates the sentence and Kato (2002) presents (17b) as an instance of sentential negation expressed by a postverbal negative PP (but see the discussion concerning (22) below):

1. (17)

   1. a. She will go there on no account, not even with John.

      (Haumann Reference Haumann2007:230, ex. (130b))

   2. b. He will visit there on no account. (Kato 2002:67, ex. (14a))

However, native speakers often consider sentences with postverbal negative adjunct PPs as less than perfect. And indeed, while they present (16a) without comments, Pullum & Huddleston (Reference Huddleston and Pullum2002:814) themselves signal that in fact postverbal negative PPs lead to a lower acceptability. They illustrate this point by means of the (weak) contrasts in (18) and (19): the examples in (18), with a negative adjunct PP in postverbal position, are more marked than the corresponding sentences in (19), which contain a combination of the negative marker not with a postverbal adjunct PP containing a negative polarity item (NPI).

1. (18)

   1. a. ?I am satisfied with the proposal you have put to me in no way.

      (Pullum & Huddleston Reference Huddleston and Pullum2002:814, ex. [24ib])

   2. b. ?As far as I can recall, I have purchased food at the drive-through window of a fast-food restaurant on no street in this city.

      (Pullum & Huddleston Reference Huddleston and Pullum2002:814, ex. [24iib])

2. (19)

   1. a. I am not satisfied with the proposal you have put to me in any way.

      (Pullum & Huddleston Reference Huddleston and Pullum2002:814, ex. [24ia])

   2. b. As far as I can recall, I have not purchased food at the drive-through window of a fast-food restaurant on any street in this city.

      (Pullum & Huddleston Reference Huddleston and Pullum2002:814, ex. [24iia])

As shown in the extract below, the authors account for the above contrasts in terms of processing load, rather than in terms of grammaticality:

Though Pullum & Huddleston do not pursue this point, their account of the contrasts between (18) and (19) leads to the correct prediction that medial position will be preferred for the negative adjunct PP: (18a) and (18b) are definitely improved with the negative PP in medial position. Observe that even for the slightly longer PP on no street in this city in (20b), considerations of weight do not lead to a degradation.

1. (20)

   1. a. I am in no way satisfied with the proposal you have put to me.

   2. b. As far as I can recall, I have on no street in this city purchased food at the drive-through window of a fast-food restaurant.Footnote ⁴

De Clercq (Reference De Clercq2010a, Reference De Clercqb) reports the judgments in (21)–(24). The examples in (21) show that while the non-negative PP at that time is accepted both in medial (21a) and postverbal (21b) position, its negative analogue remains acceptable in medial position (21c) but postverbal position (21d) is rejected. In contrast with the judgment reported by Kato in (17b) above, postverbal on no account in (22b) is also considered unacceptable by De Clercq's informants. The examples in (23) and (24) provide additional judgments along the same lines.

1. (21)

   1. a. The police had at that time interviewed the witnesses.

   2. b. The police had interviewed the witnesses at that time.

   3. c. The police had at no time talked to the witnesses.

   4. d. ?*The police had talked to the witnesses at no time.

2. (22)

   1. a. You should on no account move to Paris.

   2. b. ?*You should move to Paris on no account.

3. (23)

   1. a. She should at no time reveal the secret.

   2. b. ?*She should reveal the secret at no time.

4. (24)

   1. a. They would under no circumstances reveal the problem.

   2. b. *They would reveal the problem under no circumstances.

A fully acceptable alternative to a sentence with a postverbal negative adjunct PP is one in which sentential negation is expressed by the canonical marker of sentential negation not/n't and in which an NPI any replaces the negative quantifier no in the postverbal PP. The contrast between the perfect (25) and the contrasts in acceptability observed for degraded (22b), (23b) and (24b) suggests that it is the negative component of the postverbal PPs that causes the degradation.

1. (25)

   1. a. She should not reveal the secret at any time. (De Clercq Reference De Clercq2010b:9)

   2. b. You should not move to Paris on any account.

   3. c. They would not reveal the problem under any circumstances.

3.3 The distribution of negative PP adjuncts

In Section 2.2, we saw that as far as non-negative adjunct PPs are concerned, postverbal PPs outnumber medial PPs in the English corpora considered. To assess the distribution of their negative counterparts, we examined the distribution of the negative adjunct PPs at no time, on no account, by no stretch of the imagination, on no occasion, in no event, at no other N and in no way (see Quirk et al.'s (20) above). Our pilot study reveals an asymmetry between negative PPs and non-negative PPs. Medial non-negative PPs are less frequently attested than postverbal non-negative PPs. Medial negative PPs are far more frequent than postverbal negative PPs, which are in fact very rare indeed. These findings offer further support for Haegeman's (Reference Haegeman2002) claim that medial adjunct PPs are not categorically excluded. On the other hand, while non-negative adjunct PPs are easily available in postverbal position, postverbal negative PPs with sentential scope, while available, are the marked option.

Tables 5 and 6 summarize the results of our searches for the negative PPs at no time, on no account, by no stretch of the imagination, on no occasion, in no event, at no other N (see (26e, f, g)) and in no way.

Table 5. Distribution of negative adjunct PPs, COCA sample.

SAI = subject–auxiliary inversion

Table 6. Distribution of negative adjunct PPs, BNC sample.

SAI = subject–auxiliary inversion

The lower frequency of postverbal negative adjunct PPs sets them off sharply from postverbal non-negative adjunct PPs, which, as shown in Tables 3 and 4, are well-attested. To complete the picture, Tables 7 and 8 provide the relevant figures for medial and postverbal position of the corresponding adjunct PPs containing an NPI: at any time, under any circumstances, on any account and on any occasion. For at any time and in any way, we have again used a reduced sample of 100 examples. As was the case for the non-negative PPs discussed in Section 2 above, postverbal position is more easily available.

Table 7. Distribution of NPIs: medial and postverbal position, COCA sample.

Table 8. Distribution of NPIs: medial and postverbal position, BNC sample.

Some of the (rare) postverbal occurrences of negative PPs are illustrated in (26).

1. (26)

In preparation for the next section we need to add one ingredient to the discussion, which we have not touched upon so far: whereas negative adjunct PPs resist postverbal position, the canonical position of negative complement PPs is postverbal, see (27a). Indeed there is no medial position available for negative complement PPs, as is shown by (27b). However, the postverbal position of the negative complement PP is felt to be a marked option in comparison to encoding negation medially by means of the canonical marker of negation n't/not, where the corresponding postverbal PP contains an NPI, as in (27c).

1. (27)

   1. a. Mary has talked to no one.

   2. b. *Mary has to no one talked.

   3. c. Mary hasn't/not talked to anyone.

4.1 Question tags and negative clause-typing

Ever since Klima's (Reference Klima, Fodor and Katz1964), reversal tags or question tags as illustrated in (28) have been used as a diagnostic to determine whether a sentence is affirmative or negative (Horn Reference Horn1989; McCawley Reference McCawley1998):Footnote ⁶

1. (28)

   1. a. John is working on a PhD, isn't he?

   2. b. John isn't working on a PhD, is he?

Standardly, it is proposed that a negative question tag identifies an affirmative sentence (28a) and that a positive question tag identifies a negative sentence. Let us adopt the tag test as a diagnostic to determine the polarity of the clause, focusing on sentences containing a negative PP. Informally, we will say that clauses are typed for polarity as either negative or positive. Needless to say, clause-typing for polarity ([+/–negative]) is orthogonal to clause-typing for interrogative/declarative ([+/–wh]) since the value [+/–negative] may combine with the value [+/–wh]. Along these lines, a sentence negated by medial not/n't is negative, and so is a sentence which contains medial never, e.g. (29a). A sentence containing a medial negative adjunct PP is compatible with a positive question tag, e.g. (29b), and hence is also ‘negative’ in the intended sense.

1. (29)

   1. a. Mary has never talked to anyone, has she?

   2. b. She had at no point talked to anyone, had she?

As discussed above, postverbal negative adjunct PPs are rare, but to the extent that they are acceptable, such sentences are only compatible with positive tags. The example in (30a) is from Pullum & Huddleston (Reference Huddleston and Pullum2002), (30b) is based on Pullum & Huddleston's [24i]. We conclude that postverbal negative adjunct PPs also type the clause as negative.

1. (30)

   1. a. We were friends at no time, were we?

   2. b. As far as I can recall, we have purchased food at the drive-through window of a fast-food restaurant on no street in this city, have we/*haven't we?

      (based on Pullum & Huddleston Reference Huddleston and Pullum2002:814, ex. [24ii])

When it comes to sentences containing negative complement PPs though, the pattern of question tags is reversed for our informants. As can be seen in (31), while sentence-medial not induces a positive tag, the sentence with the postverbal negative complement PP to no one is compatible with a negative tag (see also Horn Reference Horn1989:185, citing Ross Reference Ross, Gross, Halle and Schützenberger1973 for a similar example with a negative nominal complement).Footnote ⁷

1. (31)

   1. a. Mary has talked to no one, *has she/hasn't she?

   2. b. Mary hasn't/not talked to anyone, has she/*hasn't she?

We conclude, then, that there is an argument–adjunct asymmetry: while postverbal negative adjunct PPs may be rare, to the extent that they are possible they type the clause as negative. On the other hand, we can see that postverbal negative complements do not type the clause as negative, since they are not compatible with a positive question tag.

4.2 Clause-typing and sentential negation

Our hypothesis is that clauses are typed for polarity: they are either positive or negative. Polarity determines the choice of question tag. In line with the cartographic approach (Rizzi Reference Rizzi and Haegeman1997; Moscati Reference Vincenzo2006), we assume that polarity typing is syntactically encoded on a head in the C-domain such as Laka's (Reference Laka1990) ΣP, or Progovac's (Reference Progovac1993, Reference Progovac1994) PolP. We propose that in the case of negative sentences, this head must establish a local checking relation with a negative constituent. From the distribution of the tags, we conclude that the medial negative marker not and the medial adverb never are able to license the clause-typing negative head in the C-domain and that postverbal negative PP complements cannot do so.

1. (32)

   1. a. Mary hasn't talked to anyone, has she?

   2. b. Mary has never talked to anyone, has she?

   3. c. *Mary has talked to no one, has she?

We interpret the contrast in (32) as deriving from locality conditions on clause-typing. Putting this first at an intuitive level, the negation in (32c) is ‘too far’ from the C-domain to be able to type the clause as negative and hence to license the positive tag. Various implementations can be envisaged to capture these locality restrictions. In terms of Phase theory (Chomsky Reference Chomsky and Kenstowicz2001, Reference Chomsky, Freidin, Otero and Zubizaretta2008), for instance, one might say that being contained within a lower phase (vP), the postverbal negative complement PPs cannot establish the required licensing relation with the relevant head in the C-domain.

To make this proposal more precise, let us propose that the polarity-related head in the C-domain contains an unvalued feature, [pol:__], which has to be assigned a value through a local checking relation. In (32a) and (32b), with the medial negative markers not and never, the feature [pol:__] in the C-domain can be valued through an agree relation with the interpretable negative feature on never.Footnote ⁸ If the C-polarity head is typed as negative, then the clause will be compatible with a positive tag.

In (32c), on the other hand, the negative quantifier no one in the VP-internal argument PP is contained in the vP phase and hence it is too low to be able to value the clausal polar head by an agree relation. We assume that in the absence of a negatively valued checker, the polarity feature of the clause is typed as positive by default and will hence not be compatible with the positive reversal tag.

1. (33)

   1. a. [_CP [_Cpol: neg] [_TP Mary has not_[neg] [_{v P} talked to anyone]]]

   2. b. [_CP [_Cpol: neg] [_TP Mary has never_[neg] [_{v P} talked to anyone]]]

   3. c. [_CP [_Cpol__] [_TP Mary has [_{v P} talked to no one_[neg]]]]

A final remark is in order here. Though it does not lead to a positive tag, (31a)/(33c) is still felt to be a ‘negative’ sentence due to the presence of the negative DP. For instance, like (32a) and (32b), (32c) will combine with a neither tag rather than with a so tag.Footnote ⁹ Klima (Reference Klima, Fodor and Katz1964) considers neither tags also to be a diagnostic for negativity (see also (16c) above):

1. (34)

   1. a. Mary has not talked to anyone, and neither/*so has Jane.

   2. b. Mary has never talked to anyone, and neither/*so has Jane.

   3. c. Mary has talked to no one, and neither/*so has Jane.

As discussed already by McCawley (Reference McCawley1998:604–612), the reversal tag-diagnostic which we used previously and the neither/so tag gives different results. It is not clear to us at this point how to capture this in terms of our discussion. De Clercq (Reference De Clercq, Nouwen and Elenbaas2011b) proposes that in examples such as (34c) the negation encoded in no one within the complement of V takes scope by virtue of its quantificational properties, in the same way that, for instance, the universal quantifier encoded in everyone can scope over the clause in (35). The precise implementation of this proposal would lead us too far and it also depends on the assumptions regarding the syntactic encoding of scope, see De Clercq (Reference De Clercq, Nouwen and Elenbaas2011b) for one proposal. Crucial for us is that, syntactically, the postverbal vP-internal argument cannot establish a local checking relation with the polarity feature, which by hypothesis is in the C-domain: polarity checking is different from the operation that determines the scope of the quantifier in (35).

1. (35) Mary has talked to everyone.

We tentatively assume that the neither tag is sensitive to the scopal/quantificational properties of the negative quantifier in a way that the reversal tags are not.

4.3 Clause-typing and adjunct PPs

Let us now return to the distribution of negative adjunct PPs. We have seen that the preferred position for such PPs is medial rather than postverbal. A sentence with a medial negative adjunct PP is compatible with a positive reversal tag, as shown in (36a) below, entailing that the negative PP must be able to type the clause. Pursuing our analysis, we will assume that, like the marker of negation not and like the medial negative adverb never, the medial negative adjunct PP is in a sufficiently local relation to the C-domain to value the polarity feature. We conclude from this that such PPs must not be contained within the vP phase. If they were, then we would not expect them to pattern with medial not and never. Depending on one's assumptions about functional structure, the negative PP might be vP-adjoined, as in (36b), or it might be taken to be the specifier of a medial functional projection, as in (36c), which we label FP.Footnote ¹⁰

1. (36)

   1. a. She had at no point talked to anyone, had she?

   2. b. [_CP [_Cpol:neg] [_TP she had [_{v P} at no _[neg] time [_{v P} talked to anyone]]]]

   3. c. [_CP [_Cpol:neg] [_TP she had [_FP at no _[neg] time [_{v P} talked to anyone]]]]

Postverbal negative adjunct PPs are marginal, but to the extent that they are available they were shown to be compatible with positive tags, see (16d) above, suggesting that they too type the clause. The analysis of such examples depends on one's general assumptions about the syntax of postverbal PPs (see Cinque Reference Cinque2004 and Belletti & Rizzi Reference Belletti and Rizzi2010 for overview of some options). If right adjunction is admitted in the theory (see Ernst Reference Ernst2002a, Reference Ernst and Di Sciullob), at no time in (37a) might be right-adjoined to vP. Hierarchically speaking, though postverbal, the PP in (37b) is outside vP and remains within the local checking domain of the polarity head in C. Given that, in terms of hierarchical relations, the relation between C and the postverbal adjunct in (37b) is identical to that between C and the medial adjunct PP in (35b, c), this approach does not offer any insight into the perceived degradation of negative adjunct PPs in postverbal position.

1. (37)

   1. a. She had talked to them at no time, had she?

   2. b. [_CP [_Cpol:neg] [_TP she had [_{v P} [_{v P} talked to them] at no_[neg] time]]]

On an anti-symmetric/cartographic view in which right adjunction is not available (Cinque Reference Cinque2004), one might propose that the negative PP occupies the specifier position of a functional projection, FP (as in (37b′)), and that its postverbal position is derived by leftward movement of the vP to a higher position. The movement could arguably be triggered by the need for the negative PP to receive focal stress (see Jayaseelan Reference Jayaseelan2008, Reference Jayaseelan2010).

1. (37) b′. [_CP [_Cpol:neg] [_TP she had

2. [[_{v P} talked to them] [_FP at no_[neg] time [_{v P}talked to them]]]]]

Assuming that the projection hosting the PP and the projection hosting the fronted vP do not themselves constitute phases, the polarity head in C can continue to establish a local checking relation with the postverbal negative PP in (37c). On a more speculative note, we add here that the representation in (37b′) may contribute to explaining the observation that the postverbal position of the negative PP in (37a) is degraded: the fronting of the vP to a position c-commanding the negative PP might be argued to create a weak intervention effect for the relation between C and the negative PP.

A correct prediction of our account is that a negative DP in the canonical subject position always types the clause as negative: (38a) is only compatible with a positive tag. This is so because the negative feature on no one is in a local relation with the polarity feature in C:

1. (38)

   1. a. No one talked to the police about any crime, did they?

   2. b. [_CP [_Cpol:neg [_TP no one_[neg] [_{v P} talked to the police about any crime]]]]

The proposal developed here, elaborating on De Clercq's work, also has further implications for the representation of clause structure and in particular for the demarcation of phases. Passive sentences with a postverbal negative by phrase take a negative question tag (39).

1. (39) The book was adapted by no one, wasn't it?

In terms of our account this entails that, as is the case for postverbal arguments, the negative component no one cannot value the polarity feature in the C-domain. This implies that, unlike postverbal adjuncts, the by phrase must be contained within a phase. We do not pursue this issue here as it hinges, among other things, on the analysis of passives (see Collins Reference Collins2005 for a relevant analysis).