- 1 Written English is not the only register allowing for subject omission. As discussed, for instanc (...)
- 2 The availability of non-referential subject omission argues against an analysis in terms of null (...)
1Even though English is not a null subject (‘pro drop’) language, in some written registers subjects of finite clauses are sometimes left implicit.1 Attestations from the relevant registers are illustrated in (1)-(6): (1) illustrates the diary register (Haegeman 1990, 1997, 1999, 2013, 2017), (2)-(3) illustrate what Matushansky (1995) calls ‘global topic’ texts, found in journalistic prose (2), encyclopaedia entries (3) etc., (4) illustrates instructional writing, such as recipes (4a), stage directions (4b), and parenthetical comments (4c), and (5) illustrates written notes (Janda 1985). The empirical focus of the present paper is mainly on subject omission in diary style and in global topic texts. (6) shows that in the relevant registers, non-referential subjects may also be omitted.2
- 3 As pointed out by Andrew Radford (p.c.) the question arises whether article drop as illustrated i (...)
- 4 The second occurrence of finite rained also lacks a subject, illustrating the phenomenon usually (...)
(1) a. Day3 before yesterday came on some men waiting with a she-camel which had fallen in the middle of the bridge over the Oued. (George Orwell, 7 January 1939, Online Orwell project)
b. Studies under [David] Daiches. (Plath, 1956, 126)
c. Sat there for 3/4 hours. (Woolf, 1940, 334)
(2) Solti’s timeline: The rise of a maestro
1932 Goes to Karlsruhe to assist Josef Krips. Within a year ___ is sent home as Krips anticipates the Nazis’ rise to power. …
1997 Conducts his last symphony with the Zurich Tonhalle Orchestra.
Shortly after, dies in his sleep, aged 84, in France. (Observer 9.9.12, p.17 c. 2)
(3) Common Sandpiper
Teeters hind part of body almost constantly, bobs head; when flushed flies low over the water, alternating rapid, shallow wingbeats. (Collins Nature Guide, Birds of Britain and Europe, 1994, Page 112)
(4) a. Serves four.
If refrigerated, will keep a week to 10 days. (Kennedy, Ellen, Things my mother told me. Boston, Branden Press Publishers, 1975: 105)
Contains caffeine (Coke can)
b. Ouch! Too sharp...
(Tosses the pencil. Starts on another.) (http://freedrama.net/deargod.html)
c. - Are you happy in your career?
- Are you saying I am a late bloomer? [Laughs.]
(Observer 26.6.16, interview Anna Chancellor, page 26, col 1)
(5) Wish you were here!
(6) a. Up at 6:45 - Rained all day & rained4 hard at times. (Sunday, October 4, 1964)
b. Started raining this P.M. [….] Poured down rain tonite. (Monday, August 31, 1964)
(a-b: http://mymothersdiary.blogspot.be/)
2It would seem plausible that these examples involve omission of a subject. In this paper, I pursue this intuition focusing on the distributional restrictions on the omission of subjects. The discussion will, among other things, lead to concrete proposals about the syntactic position of initial adverbial adjuncts. From now on, I refer to subject omission in written registers by the abbreviation WSO, and I will also use this abbreviation to refer to the omitted (and implied) subject.
3The paper is organized as follows: Section 2 inventorizes the distributional properties of WSO, Section 3 focuses on a specific coreference restriction of WSO, Section 4 provides the ingredients for a syntactic analysis of WSO, focusing specifically on the position of initial and medial adjuncts, Section 5 elaborates the syntactic analysis of WSO, Section 6 shows how the analysis elaborated may also account for the coreference restriction and Section 7 summarizes the paper.
4Throughout, the discussion is inspired by my own earlier work on subject omission, which was formulated in terms of formal syntax. For the present purposes, I will endeavour to keep the discussion at as theory neutral a level as possible in the hope that this may encourage readers from various theoretical frameworks to engage with the data.
5In most of the WSO examples above, the non-overt subject can be spelt out as a pronoun, as illustrated in the examples in (7)-(11), in which a subject pronoun is inserted in the attested WSO examples listed earlier. For completeness’ sake, I point out that, even though in the diary register subject omission typically affects the first person pronoun, third person subjects may also be non-overt, as shown by (6) and by (1b) and (1c):
(7) a. Day before yesterday I came on some men waiting with a she-camel which had fallen in the middle of the bridge over the Oued. (=(1a))
b. He studies under [David] Daiches. (=(1b))
c. They sat there for 3/4 hours. (=(1c))
(8) Solti’s timeline: The rise of a maestro (= (2))
1932 He goes to Karlsruhe to assist Josef Krips. Within a year he is sent home as Krips anticipates the Nazis’ rise to power. …
1997 He conducts his last symphony with the Zurich Tonhalle Orchestra.
Shortly after, he dies in his sleep, aged 84, in France.
(9) Common Sandpiper (= (3))
It teeters hind part of body almost constantly, it bobs its head; when flushed it flies low over the water, alternating rapid, shallow wingbeats.
(10) I/we wish you were here! (=(5))
(11) Up at 6:45 - It Rained all day & rained hard at times. (=(6a))
- 5 As pointed out by an anonymous referee for FREL, this raises the question as to what licenses omi (...)
6Like all types of ellipsis, WSO is governed by a pragmatic condition on recoverability. As far as referential subjects are concerned, only those that are highly salient and contextually accessible can be non-overt. Starting with seminal work by Ariel (1988: 88, table 7), a strong correlation has been postulated between the relative weakness of the anaphor and the degree of accessibility of the referent. In the hierarchy developed by Ariel, the zero pronoun is the ultimate type of weakness, which entails that its antecedent is maximally accessible.5 The default interpretation in a diary style text is that the null subject is referring to the writer, who is the topic of the diary, but other referents are possible, once these have been made suitably accessible in the context.
- 6 Thanks to Philip Miller for signalling this important difference and to an anonymous reviewer for (...)
7In some of the attested cases, a demonstrative nominal would appear to be more appropriate as the overt alternative for the WSO. This is particularly true for WSO in instructional registers (4a), as shown in (12). Observe that these examples involve deictic reference to an entity not represented by a contextual antecedent but whose referent would be in a physical space close to the reader of the recipe.6 In such contexts, the proximal demonstrative is preferred to a pronoun (see Ariel 1988: 68).
(12) This recipe serves four.
If refrigerated, this preparation will keep a week to 10 days.
This product contains caffeine.
8In the WSO pattern, the finite verb agrees with the non-overt subject (13). However, English being a languages with relatively poor inflectional morphology, the person and number features of the non-overt subject are not always fully recoverable from the verb inflection. While it is clear that the implied subject in (13a) is first person, the only pronoun compatible with the form am, in (13b) and (13c) the forms could and got are compatible with any person number combination:
(13) a. Am reading the book of Job. (Plath, 1959: 290)
b. Couldn’t even read. (Plath, 1956: 108)
c. Got up as brash, nerve-raking alarm ground off at 6. (Plath, 1956: 54)
9 The content of the WSO is usually contextually recoverable, but it must be pointed out that occasional instances of WSO are not uniquely recoverable: in (14), the WSO of didn’t eat and (perhaps less likely) of watched tv tonite could be construed as ‘I’, i.e. the diary writer, as ‘we’, i.e. the writer and ‘Ruby’, and possibly even just as ‘Ruby’.
(14) I sat down & didn’t even start supper – Didn’t feel good - Think I got too much sun - Ruby came over for a while - Didn’t eat supper until 7:15 - Watched tv tonite - Wrote a note to Vera & sent a birthday card. (“My mothersdiary.blogspot”: (1964), Friday, August 21, 1964)
- 7 The following is a BBC strapline on 27 June 2016: ‘Merkel says understands that Britain needs tim (...)
10For most speakers, WSO is restricted to root clauses (for embedded null subjects see Haegeman and Ihsane 1999, 20027). In the attested (15), the subject of the root clause is omitted, but all subjects of non-root clauses are overt and could not be omitted without loss of acceptability:
(15) a. Dreamt that *(I) picked up a New Yorker. (Plath, 1982: 304)
b. Says *(he) has been struck by the number of more or less ordinary Conservatives *(he) has met who are becoming perturbed by the Government’s foreign policy.
(Orwell diaries, Villa Simont, 22.11.38; http://orwelldiaries)
c. Have dinner at Palace where *(I) make a speech in reply to the Mexican President.
(Truman’s diary: http://www.trumanlibrary.org/diary/7 January 1947)
d. Separates from wife and moves into the Savoy hotel, where *(he) meets Valerie Pitts (above right with Solti), whom *(he) marries three years later. (Observer 9.9.12 page 17 col 2)
11 Observe that these data also show that one cannot simply postulate that the subject can be omitted provided that all relevant subject information can be recovered from the context. For instance, if the subject of dreamt in (15a) can be recovered from the context, then one can plausibly assume that the same holds for the subject of picked up and yet the latter cannot be omitted.
12WSO is subject to what has come to be known as a ‘left edge restriction’ (Wilder 1994/1997). Simplifying somewhat, this means that material that would precede an overt subject is incompatible with WSO. Concretely, WSO is incompatible, among other things, with subject auxiliary inversion (16a, 16b), with wh-fronting (16b,16c) and with argument fronting (whether the fronted argument has a topic or a focus reading (16d)): in all of the examples in (16), the pronominal subject cannot be deleted:
(16) a. Will *(I) ever see her again?
b. How many presents will *(I) receive this year?
c. What a nice present *(I) received in the mail this morning.
d. This book *(I) didn’t like. (Wilder 1994/1997)
- 8 In spoken English, the availability of subject omission with an initial adjunct is less clear. We (...)
13 Yet, the left edge restriction might at first sight appear not to be absolute and there seems to be an argument/adjunct asymmetry. In the following attested examples, WSO coincides with adjuncts that would be expected to precede a pronoun in the canonical subject position: in (17) I have inserted a parenthesized subject in the slot in which an overt pronoun subject would be expected to appear. To the best of my knowledge, the availability of adjuncts with WSO holds across the various register types that deploy WSO. 8
(17) a. With a sigh of relief (I) saw a heap of ruins. (Woolf, V, 1940: 330; Ihsane 1998, (40j))
b. After Dr. Krook, (I) had good lunch at Eagle with Gary [Hamp]. (Plath, 1982: 126)
c. This morning (I) woke to get a letter in the mail from the estimable Dudley Fitts. (Plath, 1982: 304)
d. If no better by Friday, then (I) might have to go back to the UK for an MRI scan.
(Observer 23.8.9 page 10 col 2/page. 11 col. 1 (diaries of soldiers in Afghanistan)
e. 2007 After several drink driving and drug charges, (he) attended mandatory and voluntary rehab spells.
2010. While in Cannes for the film festival, (he) missed a progress hearing and a warrant was issued for her arrest. (Observer 26.11.12 page 31 cols 1-2)
f. Gannet
When fishing (it) plunges into sea from height of up to 40m. (Collins: page 18)
White Stork
When searching for food (it) walks majestically. (Collins: page 28)
g. If refrigerated, (this preparation) will keep a week to 10 days.
(Kennedy, Ellen, Things my mother told me. Boston, Branden Press Publishers, 1975: 105)
14Observe that an initial wh-adjunct is incompatible with subject omission:
- 9 Thanks to Andrew Radford (p.c) for help with these examples.
(16) e. How skilfully *(he) dodged the reporter’s questions. 9
15 The left edge restriction on WSO observed above offers evidence against treating the patterns of WSO in diary registers as a manifestation of the so-called pro drop phenomenon illustrated for Italian in (18). As is well known, in pro drop languages, the pronominal subject of a finite clause may be non-overt (18a). In some of the formal literature in the Government and Binding tradition (Chomsky 1981), this non-overt subject is analysed as a non-overt pronoun and is represented as pro (18b). Crucially, pro has a wider distribution than WSO, being available in embedded clauses (18c) and being compatible with various kinds of fronted material, such as fronted wh-phrases (18d) or focused arguments (18e).
(18) It a. Parlo /parli /parla etc italiano.
Speak-1sg /speak.2sg /speak.3sg Italian
b. pro Parlo italiano.
c. I ragazzi cantano [quando pro lavorano].
the boys sing.3pl when pro work.3pl
‘The boys sing while they are working.’
d. Quando tornerà pro?
when return.fut.3sg pro?
‘When will he/she come back?’
e. Giulia pro hanno invitato, (non Marina).
Giulia pro have.3pl invite.part-masc.sg (not Marina)
‘Giulia they invited (not Marina).’
- 10 The intuition is shared by Linda Badan and Ciro Greco, whom I thank for discussing this exampl (...)
- 11 Thanks to Andrew Weir for the judgment.
16There is also an interpretive difference between pro and WSO. Samek-Lodovici (1996: 31) signals that in Italian, the nominal complement of an agent by-phrase (19a) cannot be an antecedent to pro.10 In (19) co-reference is indicated by co-subcripting. This restriction does not govern the interpretation of English WSO, which can take the agent nominal, the department secretary in (19b), of a passive sentence as its antecedent:11
(19) It a. Questa mattina, la mostra é stata visitata da Giannii
This morning the exhibition is been visited by Gianni
Piú tardi, *proi/egli/lui ha visitato l’università.
Later on, he has visited the university
(Samek-Lodovici 1996: 31, (3))
b. Was contacted by the department secretaryi in the morning.
Told me she urgently needed the financial reports.
17 An overt subject pronoun may be preceded by a left peripheral circumstantial adjunct containing a coreferential noun phrase. This is illustrated in (20a), in which the subject pronoun he can be coreferential with the nominal Mourinho contained in the adjunct PP during Mourinho’s first year in London. In what would look like the WSO counterpart to this example (20b), the coreference relation is unavailable: (20b) is acceptable, but it cannot have the reading according to which the implicit subject of became is coreferential with Mourinho.
(20) a. During Mourinho's first year in London, he became famous for his grey Armani coat.
b. During Mourinho's first year in London, became famous for his grey Armani coat.
The same contrast in referential relations is found with clausal sentence-initial adjuncts: in (21a) coreference between he and Mourinho is possible, but in (21b) the implicit subject of became cannot be coreferential with the nominal Mourinho in the adverbial clause.
(21) a. When Mourinho was in London, he became famous for his grey Armani coat.
b. When Mourinho was in London, became famous for his grey Armani coat.
- 12 Thanks to Andrew Weir and Chris Wilder for judgments.
18 The observed ban on coreference does not arise when a left peripheral adjunct contains a pronoun: such a pronoun can be coreferential with both an overt pronominal subject and with a non-overt subject in the matrix clause. This is shown in (22) and (23).12
(22) a. During his first year in London, he became famous for his grey Armani coat.
b. During his first year in London, became famous for his grey Armani coat.
(23) a. When he was in London, he became famous for his grey Armani coat.
b. When he was in London, became famous for his grey Armani coat.
19 These subtle interpretive differences suggest that care is needed when WSO sentences are equated with their analogues with a pronominal subject. In the next sections I develop a syntactic analysis for WSO that aims at capturing the observed argument/adjunct asymmetries illustrated in Section 2.3 and which can account for the interpretive effect of subject ellipsis signalled in the present section.
20 Though guided by the generative approach to syntax, I have attempted to formulate the analytical proposals in an informal and intuitive manner, keeping the implementation of theoretical machinery minimal. In this way, I hope to be able to provide a relatively theory neutral analysis that can be made compatible with many theoretical frameworks. Readers who are more familiar with the formal frameworks will be able to supplement more technical details of implementation.
21This section provides an outline of the constituent structure of the clause that will be the backbone of my analysis.
22There is a long-standing and fairly uncontroversial intuition that in essence a clause can be viewed as a combination of the ‘subject of predication’ and the material predicated of it. Attempts have been made in the generative literature to represent this intuition formally. The cartographic generative tradition is an approach to clause structure which decomposes the structure of the clause in an array of specialised structural layers each of which with a designated interpretive value (e.g. topic, focus etc) (see also Sections 4.2 and 4.3). In that literature (Cardinaletti 1997, 2004, Rizzi and Shlonsky 2005, 2006, Shlonsky 2014) it has been proposed that the subject-predication relation be a core component of clause structure and that the clause be formally represented as in (24), in which the subject of predication is a constituent that combines with the remainder of the clause, labelling the resulting unit SubjP. Interpretively, the ‘subject of predication’ serves as the anchor for the new information in the clause.
23In the representation, the label SubjP stands for the clausal layer which combines the subject and the clausal predicate. The label ‘clausal predicate’ in the diagram is a shortcut, I refer to the generative and the cartographic literature for precise proposals. I will not go into the internal structure of the clausal predicate itself. For full motivation and more examples see the references cited.
24The area to the left of the subject position contains a specialized slot for connecting devices such as subordinating conjunctions (25) and which is the landing site for the inverted auxiliary in subject auxiliary inversion (26). (25b) and (26b) are schematic representations in which I do not specify the label of the resulting unit, but crucially the constituent creates the interface between the propositional content of the clause and the context and it is taken to be categorically distinct from SubjP.
25Observe that the tree diagram format to be used here and below remains informal and for expository reasons, departs in a number of ways from standard formalisms: some constituents are labelled on the basis of grammatical function, such as ‘subject’ or ‘clausal predicate’, rather than category, and because hierarchical relations are the crucial component in the trees, label specifications that are not relevant to the points at issue have been omitted.
(25) a. (I think) that the prime minister will make his own position explicit.
(26) a. Will the prime minister make his own position explicit?
26Sentence-initial wh-constituents, both arguments and adjuncts, which encode the illocutionary force of the clause, also target the area to the left of the subject, (27a) contains an embedded interrogative with an argument (which points) in pre-subject position and a gap in the corresponding clause-internal object position. The representations of the relevant section are as in (27b) and (27c). In (28), the pre-subject wh-constituent in the embedded clause is an adjunct. As before, I assume that the constituent created as a result of wh-fronting is categorically distinct from SubjP.
(27) a. (I wonder) which points the prime minister will make explicit.
b. … [which points [the prime minister will make explicit]].
(28) a. (I wonder) when the prime minister will make his own position explicit.
b. … [when [the prime minister will make his own position explicit.]]
27 Fronted arguments precede the canonical subject position (29a). By hypothesis, such fronted arguments have specific discourse properties (topic, focus) and they end up in a position to the left of and structurally higher than the canonical subject position: (29a) has the partial representation in (29b), in which the square brackets show that the fronted object his own position is added to the full-fledged clause, which has a gap in object position. (29c) is a schematic tree diagram in which again I do not specify the label of the resulting unit, but, as before, it is taken to be different from SubjP. See Rizzi (1997) for proposals concerning the syntax of the left periphery of the clause.
(29) a. His own position the prime minister will make explicit tomorrow.
b. [His own position [the prime minister will make explicit tomorrow]].
28 Adjuncts may also precede the canonical subject position as in (30a), in which case, like topicalized and focused arguments, they are added onto SubjP (30b). Representation (30c) is again non-committal to the label of the resulting unit, though it is taken to be a projection distinct from SubjP.
(30) a. Tomorrow, the prime minister will make his own position explicit.
b. [Tomorrow [the prime minister will make his own position explicit]].
29In the attested examples in (31), a temporal adjunct is located between the canonical subject of the clause and the finite auxiliary: in (31a), the nominal this time separates the subject the critics from the finite auxiliary can; in (31b), the PP in a decade’s time separates the subject Cadbury from the finite auxiliary will, in (31c) the nominal next week separates the subject the president and the auxiliary will. See Haegeman (1984, 2002).
(31) a. The critics this time can only award their prize posthumously. (Observer 30.5.10, page 20, col 1)
b. Cadbury in a decade’s time will be a ghost of itself. (Observer 29.11.9, p. 24, col 3)
c. and I hope the president next week will announce that he’s going to support the general and give him all the troops he needs for that strategy. (http://edition.cnn.com/TRANSCRIPTS/0911/24/lkl.01.html)
30As argued extensively in Haegeman (2002), (31) should not be analysed as the result of the combination of (i) fronting of the adjunct (this time, in a decade’s time) to the left periphery (as discussed in Section 4.2, example (30)) and (ii) fronting of the subject argument to an additional peripheral position as a topic or a focus, because the pattern in (31) is compatible with contexts in which arguments are known to have a reduced access to the left periphery, such as interrogative if clauses (32), and conditional if clauses (33) (cf. Haegeman 2002, 2012). As shown in (32a) and (33a), these contexts resist fronting of an argument to their left-periphery, but (32b) and (33b) show that they remain compatible with the pattern in (31), in which a subject precedes an adjunct.
(32) a. *Asked [if his position the Prime Minister had already made more “explicit” the PMOS replied: no.
b. Asked [if the Prime Minister yesterday had made his position more "explicit" regarding the rebate and its negotiability], the PMOS replied: no. (www.number10.gov.uk/output/ Page7713)
(33) a. *[If this news the government had announced…]
b. [If the government last year had said] ... (www.leg.bc.ca/hansard/37th5th/h40311p.htm)
31I conclude that in (31), (32b) and (33b), in which the subject is separated from the finite auxiliary by an adjunct, the subject has not moved to the left periphery but that it occupies the canonical subject position of the clausal domain (i.e. ‘subject’ in (34) below). I propose then that the derivation of the clause proceeds as follows: the clausal predicate is formed, then the medial adjunct, as a modifier, is combined with the clausal predicate prior to the insertion of the subject, finally the subject is added. By hypothesis, the position labelled adjunct in (34) is specialised for modifiers. I will not assign a label to the combination of the adjunct and the clausal predicate, crucially it must be a constituent able to combine with a subject. For additional discussion of ‘medial’ adjuncts in relation to the position of the subject, I refer to Haegeman (2002) and De Clercq, Haegeman and Lohndal (2012).
- 13 Because double fronting of an argument is by and large unacceptable in English, a derivation by w (...)
32 This hypothesis that there is a specialised medial adjunct position captures the fact that while adjuncts may intervene between the subject and the auxiliary, arguments cannot be found in the same position as shown by the contrast in (35). In the ungrammatical (35b), the focussed argument his own position would have to be inserted in the clause-internal adjunct slot to the right of the subject, but, by hypothesis, in English this position is reserved for adjuncts, i.e. it is not available for arguments.13
(35) a. The prime minister tomorrow will make his own position explicit.
b. *The prime minister his own position will make explicit tomorrow.
33 The medial adjunct can only be interpreted as modifying the predicate of the clause that immediately contains it. In this respect, it contrasts with initial adjuncts (30). In (36a) the initial adjunct this week may modify the matrix event (‘has said’), but when contrastively stressed it may also have ‘low construal’ and modify the embedded clause, in which case (36a) is equivalent to (36b) (cf. Postal and Ross 1970: 145). In medial position in (36c), low construal is unavailable: this week cannot be interpreted as modifying the embedded clause.
(36) a. This week the president has said [that he will support the motion].
b. The president has said [that he will support the motion this week].
c. The president this week has said [that he will support the motion].
34Summarizing and simplifying somewhat, while adjuncts may be inserted outside SubjP (30c) or inside SubjP (34), fronted arguments of the verb such as direct objects are always outside SubjP (29c). I will now explore the ramifications of this typology of adjuncts for the syntax of WSO.
35 Consider the distributional properties of WSO outlined in Section 2: WSO is incompatible with the following elements:
- subordinating conjunction (15).
- inverted auxiliary (16a, b);
- fronted (wh) argument (16b, 16c, 16d);
- fronted wh adjunct (16e).
36 Leaving aside the compatibility of WSO with non-interrogative adjuncts for a moment (17), the data suggest that what enables the WSO configuration is that the WSO subject is the constituent that is added last to the clausal projection. Adding material to the projection after the subject has been inserted makes WSO impossible. Put differently: the creation of SubjP by adding the subject to the clause terminates the projection, i.e. the projection of the sentence with WSO in effect terminates at the layer SubjP. So, the left edge condition on WSO can be rephrased as a hypothesis that structure building terminates at SubjP.
(37) Truncation Hypothesis
The availability of WSO depends on the possibility of terminating the projection of the clause at the level of SubjP: the subject is the final constituent added to the clause.
37I assume that (37) is a marked option of the grammar of English and that it is register-specific, I will speculate briefly on this point below. The hypothesis that clauses can terminate at the layer SubjP and hence lack any higher layers has independently been first elaborated to account for subject omission in acquisition (Rizzi 1994, 1995, 1999, 2006).
38The root character of WSO follows from (37), because a clause featuring WSO cannot be embedded. In order to embed a clause, further projection is required. The hypothesis also predicts that all left peripheral constituents are banned, since there simply is no ‘room’ for them to project.
39(16a), with an inverted auxiliary, is incompatible with WSO, because the fronted auxiliary is external to SubjP, and thus the subject, I, does not terminate the structure (38a). Along the same lines, (16d), for example, with a fronted argument, is also incompatible with WSO, because the fronted argument is external to SubjP, and thus the adding of the subject, I, does not terminate the structure (38b).
40Concretely, the truncation stipulation means that in the context of WSO the peripheral layer(s) of the clause are systematically unavailable.
41 At this point, of course, many questions arise in relation to the availability of structural truncation as a marked option. For instance, the question arises why truncation is not generally available in the grammar of English, i.e. why it is that only in the specialised registers should the clausal structure be allowed to terminate after the insertion of the subject.
42 Recall that the peripheral layers of the clause are associated with encoding illocutionary force and with hosting discourse-functional material such as topicalized and focalized constituents (Rizzi 1997, Haegeman 2012 for a survey). It so happens that registers instantiating WSO are restricted to highly specialised communicative situations. In the diary style, for instance, there is no interaction between discourse participants. The speaker/writer is not addressing an external hearer/reader and there is no turn taking. In addition, the discourse topic is also invariant: the diary is about the diary writer and his life. Similarly, in global topic texts (cf. (2) and (3)) the discourse topic is held constant across the entire text. In journalistic prose, the latter texts are also set off typographically from the main body of the newspaper. We can speculate that while in the unmarked case, each utterance comes with its own discourse coordinates, the specialised communicative contexts can set up a global and invariable interface with the discourse with fixed and shared coordinates and that these can be held constant across a number of (declarative) sentences. Segments with such invariant discourse coordinates may then allow for the projection of clauses to terminate with the insertion of the subject, resulting in a truncation of the left periphery, i.e. the interface between the proposition and its discourse context.
43 Another question that arises is why, in a language like English, which is not a null subject language, the structural truncation itself allows for subject ellipsis. This issue is not addressed in the present paper. For a proposal that strongly rests on theory internal considerations, see Rizzi (2006) and Haegeman (2016).
- 14 An anonymous reviewer asks about a potential intonation difference between (39a) and (39c). The p (...)
44 While all left peripheral material is incompatible with WSO, WSO is compatible with adjuncts, as discussed in Section 2.3 (see the attested (17)) and illustrated with a constructed example in (39a). At first sight, based on the analogy with a sentence with an overt subject, (39a) might be represented as (39b). But this would raise a problem: by the truncation hypothesis, WSO is only compatible with a structure terminating at SubjP, and in order to project a left peripheral adjunct the structure must project beyond SubjP: as soon as we have inserted tomorrow in the left periphery in (39b), SubjP is not the topmost projection, and hence subject omission should be ungrammatical, contrary to fact.14
(39) a. Tomorrow will make his own position explicit.
45 However, (39b) is not the only representation for (39a). As shown in some detail in Section 4.3, there is also a medial adjunct position for adjuncts to the right of the canonical subject position:
46If we can assume that in the case of WSO the adjunct makes use of the medial position, then the projection will be able to terminate with the projection of the subject and the subject can be non-overt.
47Following the discussion of the interpretation of (36c), the prediction of this analysis is that an adjunct co-occurring with WSO can only have a local interpretation. This prediction is correct: in (40) the adjunct yesterday modifies the time of the main clause predicate (i.e. confirmed) and not that of the bracketed embedded clause (i.e. would clarify).
(40) Yesterday had confirmed [that the Prime Minister would clarify his position].
48Weir (2012b: 13) discusses similar data: Weir’s judgment of his (26) is reproduced here in (41): while (41a) is acceptable with omission of the subject, (41b), in which the adjunct tomorrow is fronted from the embedded clause to the left periphery of the root clause, is not compatible with subject ellipsis (but see Cinque 2002, 2006, Edelstein 2012 for adverb climbing).
(41) a. ___ Is likely [that John will win tomorrow].
b. *Tomorrow, ___ is likely [that John will win].
49However, as the reader will have observed, a problem arises in relation to a representation such as (39c), which is taken to underlie WSO with medial adjuncts. In the relevant examples, an overt pronoun would be dispreferred in the position to the left of these adjuncts, unless it receives emphatic stress:
(39) d. ??He/OK HE tomorrow will make his own position explicit.
50At this point, I can only formulate a tentative answer to this problem. I will assume that the restriction on the distribution of the subject pronouns is due to some kind of weight requirement on constituents, the weak unstressed subject pronoun being unable to be followed immediately by a heavier constituent. Several authors (Quirk 1985: 492, 514, 521, Ernst 2002a: 504, 2002b: 194, Huddleston and Pullum 2002: 780) discuss how weight considerations constrain the availability of (non-parenthetical) medial adjuncts. To the extent that such weight considerations are strictly viewed in terms of phonological weight, then they should not impact on the distribution of a non-overt subject.
51 We have seen that there are restrictions on the distribution of referential expressions in relation to WSO patterns. An overt subject pronoun may be preceded by a left peripheral circumstantial adjunct containing a noun phrase coreferential with that subject but coreference is not available for the WSO counterpart to this pronoun. This was illustrated in (20), repeated here as (42). The subject pronoun he can be coreferential with the nominal Mourinho contained in the adjunct during Mourinho’s first year in London in (42a); (42b) is acceptable, but it cannot have the reading according to which the implicit subject of became is coreferential with Mourinho.
(42) a. During Mourinho's first year in London, he became famous for his grey Armani coat.
b. During Mourinho's first year in London, became famous for his grey Armani coat.
52 The observed ban on coreference does not arise when a left peripheral adjunct contains a pronoun: such a pronoun can be cofererential with both an overt pronominal subject and with a non-overt subject in the matrix clause. (21), repeated here as (43), is a relevant pair:
(43) a. During his first year in London, he became famous for his grey Armani coat.
b. During his first year in London, became famous for his grey Armani coat.
53The representation of (42a) is as in (44a):
54The correct representation of (42b) cannot be as in (44b), with he representing WSO, because this example is not compatible with the truncation hypothesis according to which for the subject to be deleted SubjP must terminate the projection.
55Rather, following the discussion in Section 5, we must assume that the representation of the example is as in (44c)
- 15 Thanks to Alexandra Simonenko and to Philip Miller for insightful discussion of this point. Needl (...)
56 As mentioned, like all types of ellipsis, WSO is governed by a pragmatic condition on recoverability. If we assume Ariel’s (1988: 88, table 7) correlation between the relative weakness of the anaphor and relative accessibility of the referent, the zero-form used for anaphora is the ultimate form of weakness and hence its antecedent must be maximally accessible. Pursuing this intuition, then, where licit, the use of WSO, i.e. a zero form, entails that the antecedent – here intended to be interpreted as the person referred to as Mourinho - is topmost on the accessibility hierarchy.15 But if this referent is highly accessible, as implied by WSO, then it is unexpected that when next referring to this entity, the speaker uses a low accessibility marker, i.e. the nominal Mourinho, rather than an appropriate marker of higher accessibility, i.e. the (possessive) pronoun.
57 The same reasoning would also account for the unavailability of coreference between the subject of the main clause and the nominal Mourinho in the sentence final adjunct PP in (45): note that here overt pronoun and the zero-pronoun pattern in parallel: coreference is ruled out both with an overt pronoun (45a) subject and its non-overt alternative (45b) because both the pronoun and the zero-form signal that the antecedent is accessible.
(45) a. He became famous for his grey coat during Mourinho’s first year.
b. Became famous for his grey coat during Mourinho’s first year.
58 In the next section I develop a syntactic implementation of the coreference effects.
59 The contrast between the interpretation of the overt subject and that of the non-overt counterpart in (42) can also be reinterpreted in terms of the syntactic approach to coreference developed in the generative framework. Specifically, if we adopt representation (44c), the ban on coreference between the WSO subject and the nominal Mourinho in (42b) can be shown to be in line with what has come to be known in the generative tradition as Principle C of Chomsky’s 1981 Binding Theory, a module of the grammar according to which structural configurations play a role in determining referential dependencies of nominal expressions. I will first briefly outline the core ingredients and then show how they apply to the data under discussion.
60 Principle C of the Binding Theory is informally stated as in (46).
(46) Principle C (Binding theory: Chomsky 1981)
a referential expression cannnot be referentially dependent on a c-commanding nominal (pronoun or referential expression)
A constituent α c-commands the constituent β with which it is combined and all the constituents contained inside constituent β.
61 (47) illustrates the domain of application of this constraint. In (47a), the subject he corresponds to α, it combines with the constituent has told Bill’s brother to leave, β, and as a result the constituent he ‘c-commands’ the constituent told Bill’s brother to leave and its content. By Principle C the constituent α, here the pronoun he, and the referential expression Bill, contained in β, cannot be coreferential.
(47) a. Hei/*k has told Billk’s brother to leave.
62 Consider now (42a), repeated as (48a), in which the overt subject pronoun he can be coreferential with the nominal Mourinho, contained in the initial adjunct PP during Mourinho’s first year. The adjunct precedes the canonical subject, so, by the analysis proposed there, it must occupy a position external to SubjP, i.e. to the left of, or higher than, the subject. As shown in diagram (48b), in this example, the c-command domain of the overt pronoun he is the clausal predicate, the overt subject does not c-command the initial PP during Mourinho’s first year in London or material contained in it, i.c the nominal Mourinho. Because the subject he does not c-command the nominal Mourinho, coreference between he and Mourinho is possible.
(48) a. During Mourinho’s first year in London, he became famous for his grey Armani coat.
63 In (42b), repeated as (49a), coreference between the understood subject and the nominal Mourinho is not available. I proposed that in the derivation of this sentence, the adjunct PP during Mourinho’s first year first combines with the clausal predicate and then the subject combines with the resulting constituent (see (44c)). In this configuration, the subject nominal α c-commands β, the constituent containing the adjunct and –crucially– it also c-commands the nominal Mourinho which is contained inside β. As a result, coreference between α, the deleted subject, and Mourinho, contained in β, is not possible.
(49) a. During Mourinho's first year in London, became famous for his grey Armani coat.
64The Binding Theory also successfully rules out coreference between the (overt or non-overt) subject and the nominal Mourinho in the sentence-final adjunct PP in (45), repeated here in (50a). We assume that analogous to (44c)/(49b), the adjunct during Mourinho’s first year is a modifier combined with the clausal predicate (50b):
(50) a. (He) became famous for his grey coat during Mourinho’s first year.
- 16 As observed by an anonymous reviewer, if the analysis elaborated here is on the right track, the (...)
65 Again, as in (49b), α, the subject nominal, c-commands β, a constituent containing the adjunct; hence, the subject nominal c-commands the nominal Mourinho contained inside β. As a result, coreference between the subject, whether overt or deleted, and Mourinho is again unavailable.16
66 The empirical focus of the paper is the register specific ellipsis of subjects. After offering an overview of the major distributional restrictions of subject ellipsis in the written register, the focus is on the argument/adjunct asymmetry in the relevant pattern: while subject ellipsis is incompatible with fronted arguments in sentence-initial position, it is compatible with what at first sight appear to be sentence-initial adjuncts in the same position.
67 The paper develops a specific syntactic analysis of the position of adjuncts in English. It is proposed that the observed argument/adjunct asymmetry observed in relation to subject ellipsis comes about due to the fact that what at first sight appear to be sentence-initial adjuncts are inserted in a structural position below that of the deleted subject.
68 The observed restrictions on referential dependencies between the subject of the matrix clause and a nominal contained inside an adjunct can also be captured by the analysis.