Reporting and Comment Clauses

(1)

Reporting and Comment Clauses

A cross-linguistic study

Name: Lieke van Maastricht Student number: Address: Tel.: Email: Programme: Primary Supervisor: Secondary Supervisor: Submitted: s1461486 Lijsterstraat 79 7523 ES Enschede +31633722103 LJ.van.Maastricht@student.rug.nl University of Groningen

Research Master Linguistics Dr. M. de Vries

(2)

1. Introduction

1.1 Foreword

In human communication we make use of language (among other things) to convey a certain message to our partner in conversation. At times, the mere act of uttering aforementioned message is sufficient to get our point across. However, sometimes we also wish to add more details, for instance on how sure we are of the message we are conveying, or how we came across the information we are sharing with our interlocutor. There are many ways in which this can be done, and not all languages make use of the same processes to do so. For example, in Gitksan, the language of the indigenous population of British Columbia, Canada, affixes are used to express whether the information presented has been obtained through hearsay (reportative), or because the speaker had sensory proof (they heard, saw, smelled or felt it), or because they inferred it (proof having been provided within the direct surroundings, by general knowledge, or by personal experience), or because they merely assumed that it is true. For instance the affix =kat is used to indicate that the information was provided by another person in a hearsay context, while =ima indicates that the speaker is not completely sure about the information he or she is communicating. Both affixes are shown in the following examples, which have been taken from Peterson (2010:58).1

The first one is the original message, without any additional details. In (2) the affix =kat is added to specify that the speaker heard the information from someone else. And in (3) =ima indicates that the speaker thinks that the original message might be true, but that he is not completely sure.

(1) mukwhl maaý

mukw =hl maaý ripe =CND berries

‘The berries are ripe.’ (2) mugwigathl maaý

mukw =kat =hl maaý ripe =REP =CND berries

‘[I heard] the berries are ripe.’ (3) mugwimahl maaý

mukw =ima =hl maaý ripe =MOD =CND berries

‘The berries might be ripe.’

This system, which uses affixes to encode the evidentiality2

(or, “the grammatical category that has source of information as its primary meaning”, Encyclopedia of Language & Linguistics 2006: 320) and the epistemic modality (which deals with the speaker’s “judgements about the factual status of the proposition.” (Palmer 2001:8) of a sentence, however practical and economic it might be, has not found its way into most of the languages spoken in Europe. Instead, in Germanic and Romance languages for instance, a more analytical way of encoding evidentiality and epistemic modality is used, as can be deduced

1_{In these examples the abbreviations}

CND, REP and MOD, taken from Peterson (2010), mean ‘common noun determiner’, ‘reportative’ and ‘modal’, respectively.

(4)

from the translation of the Gitksan examples provided above. In other words, the speakers of these languages must use words such as might and expressions such as I heard to describe the source of their information and the attitude they might have towards it.

One of the constructions that can be used to do this is the one of reporting and comment clauses. These clauses, which can be adjoined to another clause that represents the original message that one is trying to convey, can contain additional details about the way in which the information presented in the message was obtained, how it is conveyed, to whom it was conveyed, and whether the speaker is sure about the information shared with his interlocutor or whether he feels slightly reluctant to accept it as true. In the following examples this is demonstrated, the first one containing a reporting clause and the second one a comment clause (underlined).

(4) ‘John is a nice guy’, Mary said timidly to Clara. (5) John is a nice guy, Mary supposed.

Traditionally, the literature chooses to treat the two constructions separately, referring to those clauses that contain a reporting (or verbal activity) verb as reporting clauses and to those that contain a comment (or mental activity) verb as comment clauses. However, on closer study, the two constructions might have more in common and may well benefit from being analysed in the same way. In the present study, the properties of Germanic and Romance comment and reporting clauses will be looked at and compared in order to establish a uniform way of analysing them in German, English, Dutch, French and Spanish. The analysis proposed in the present work will add to the overall research on parenthetical constructions, but more importantly, it will form the first cross-linguistic study that treats comment and reporting clauses as one single construction type.

1.2 Research questions

Since the literature has generally treated these two constructions separately until now, while focussing mostly on the Germanic languages, the main research question in the present work will therefore be how comment and reporting clauses can best be analysed at a cross-linguistic level. Or more specifically:

(6) What are the main characteristics of reporting and comment clauses in Germanic and Romance languages, and how can these constructions be analysed syntactically, while also accounting for their semantic and discourse-related features?

In an attempt to answer the main research question, the following sub-questions have been formulated, which will enable us to expand the focus of this study to the different sub-topics related to comment and reporting clauses.

(7) What are the features that reporting and comment clauses possess and which of these features do they share?

(8) What is the nature of the dependency between the comment or reporting clause and the associated clause to form a complex sentence?

(9) Do reporting and comment clauses always have to be combined with another clause or can they also be attached to other projections?

(5)

(11) What are the existing theories on reporting and comment clauses? And do they adequately account for the differences and similarities that exist among comment and reporting clauses?

(12) Which way of analysing them at a syntactic as well as a semantic level would ultimately prove to be most adequate?

In order to answer the questions formulated above, the present study has been structured as follows: first, an empirical background is provided in which the main properties of both comment and reporting clauses will be discussed and compared (Section 2). Next, the theoretical background of both constructions will be presented, in which the existing literature, which includes De Vries (2006), Schelfhout (2003, 2004, 2005), Suñer (2000), Ross (1973), Reinhart (1983) and Corver & Thiersch (2002), will be reviewed in order to present the ways in which comment and reporting clauses have been analysed in the past, followed by the analysis proposed in the present study (Section 3). In contrary to former analyses, the one presented here, which will assume both the associated clause and the comment or reporting clause are independent main clauses that are joined paratactically, will focus on the modality expressed by the comment or reporting clause. It will be come clear that they can both express evidential or epistemic modality and the syntactic analysis proposed in Section 3 will account for these semantic properties by adding two projections to the structure of comment and reporting clauses, namely an Evidential Phrase (EvP) and an Epistemic Phrase (EpP).

(6)

2. Empirical background

As explained in the Introduction, this study will focus on two constructions that are both considered to be members of the parenthetical construction, but generally treated as unrelated to one another: reporting clauses and comment clauses. In (13) and (14) an example of each is given in which I have underlined the reporting and the comment clause.

(13) ‘I will be late’, John said. (14) He would be late, John feared.

In this section, I will introduce both constructions and describe their general properties, after which I will present the resemblances that exist between them. The next section will then present an overview of the way in which both constructions have been analysed syntactically in the literature, followed by my proposal as to the most adequate analysis.

2.1 Reporting clauses: general characteristics

Reporting clauses are used to report the language of others. The literature generally distinguishes between two different modes of reporting that can be used to do this: the primary modes of reporting are direct speech and indirect speech, and free direct speech and free indirect speech constitute the secondary ones.3

The primary modes of reporting are formed by combining a reporting clause, which includes information on the person who is uttering a certain message and the act of communicating itself, with a reported clause, which is a representation of the aforementioned message.4

The reporting clause may also include information conveying to whom the communication act was directed, how it was carried out and the circumstances in which this was done.

Direct speech, almost exactly reproduces some previous speaker or writer’s utterance, while indirect speech is the reproduction of an original message in the words of a subsequent speaker, and is not a verbatim copy of the original utterance. Note that the original speaker, which is the subject of the reporting clause, and the person uttering the message which consists of the reported and reporting clause may be the same person in both construction types, as shown in (17). Examples (15) and (16) represent direct and indirect speech respectively.

(15) ‘I will behave myself’, Caroline assured me yesterday. (16) Caroline assured me yesterday that she would behave herself. (17) ‘I’m going home’, I said.

Apart from the fact that the reporting clause of a sentence in indirect speech includes the subordinating conjunction that, the most striking difference between (15) and (16) is that the deictic elements, which I have underlined, have changed perspective; they have been adapted from the original speaker’s point of view to that of the current speaker’s.

3_{Free indirect speech is also referred to as Erlebte Rede, free indirect style or discours/style indirect libre.} 4_{Quirk et. al. (1992) state that reporting clauses only include written and spoken messages. However, it seems to}

me that utterances performed mentally, as shown in examples (a) and (b), must also be included. (a) ‘It will be difficult to be polite to this nasty woman’, Caroline thought.

(7)

The same process takes place in the secondary modes of reporting. Yet the difference between the primary and the secondary modes of reporting is that the latter do not contain a reporting clause. Only the deictic elements can indicate whether we are dealing with free direct or free indirect speech. Example (18) represents free direct speech (the deictic elements reflect the original speaker’s point of view, here John) and (19) concerns free indirect speech (the deictic elements represent the current speaker’s perspective).

(18) [The party was so boring; John wished he could be anywhere else but there.] I will be bored to death at this stupid party. Why did I let Caroline talk me into coming?

(19) [Pete read the note and realized he had been framed.] So that was their plan, was it? He would show them he wasn’t to be fooled that easily!

In my opinion, it would be more advantageous to distinguish the primary and the secondary modes of reporting by looking at their embeddability, instead of looking at whether a reporting clause is added or not, since this is optional. I propose a division between constructions in which the reported clause is not embedded into a main clause on the one hand (which is the case in free direct and free indirect speech and in those cases where a parenthetical clause is added to the associated clause), and their embedded equivalent (direct or indirect speech) on the other hand. So from now on, we will distinguish between embedded and non-embedded speech, which can be oriented towards the original speaker (direct) or the current speaker (indirect), as is shown in Table 1.

EMBEDDABILITY DIRECT INDIRECT

Embedded John said: ‘I’m going home.’ John said that he was going home. Non-embedded ‘I’m going home’(, John said). He was going home(, John said).

Table 1: Examples of embedded and non-embedded direct and indirect speech.

The fact that the reporting clauses are optional in non-embedded contexts indicates that they are independent main clauses, in other words: parentheticals. However, there are two more factors that are of relevance here. After having presented the general characteristics of comment clauses, I will explain what they are and how they are all interrelated.

2.2 Comment clauses: general characteristics

Comment clauses either express the speaker’s comments on the content of the clause they appear with or give information on the way in which it is expressed. Quirk et al. (1992:§15.53-15.56) present the following six subtypes, each of which I also present an example:

(i) Finite comment clauses that have the form of a main clause: John, I believe, was late again.

(ii) Clauses that have the form of an adverbial finite clause that is introduced by as: What was more annoying, John was late again.

(iii) Clauses that have the form of a nominal relative clause: John, as you know, was late again.

(iv) to + infinitive clauses that function as a style disjunct: To be honest, John was late again.

(8)

Speaking as an objective judge, John was late again. (vi) –ed clauses that function as a style disjunct:5

Stated bluntly, John was late again.

In this study, only the first type of comment clause is of relevance, since it is the only one that shares most characteristics with reporting clauses. Therefore, only this first type is referred to, when speaking of comment clauses from now on.

Like reporting clauses, these finite comment clauses can be positioned at the beginning, in the middle and at the end of the associated clause, as shown in examples (20)-(22). However, it should be noted that the structure of (20) differs fundamentally from the one of (21) and (22). In (20), we are simply dealing with a complex sentence containing the main clause I suppose and the embedded clause Andrew doesn’t believe me, in which the subordinator that has been omitted. Consequently, this construction receives the standard analysis for complex sentences. Examples (21) and (22), though, have a very different syntactic structure; they consist of two main clauses that have been joined paratactically. As is reflected in their prosody; example (20) has one intonation contour, but in (21) and (22) the intonation contour of the comment clause is independent from the one of the associated clause.

(20) I suppose Andrew doesn’t believe me. (21) Andrew, I suppose, doesn’t believe me. (22) Andrew doesn’t believe me, I suppose.

Since comment clauses are only used to comment on the content of the clause it occurs with, the verb in the comment clause cannot always express all of its usual meanings. Often, the comment clause is used to mitigate the clause it occurs with and so the strength of the verb used is weakened. Compare for example (23) and (24), where the former can mean I assert the believe that John isn’t coming as well as John may not be coming, but I am not sure, while the latter can only receive the second reading.

(23) I believe that John isn’t coming. (24) John isn’t coming, I believe.

Comment clauses resemble main clauses in that they do not require a subordinator to introduce them. However, Quirk et al. (1992) state that they do not appear to be independent clauses either since, when used on their own, they lack syntactic structure; the verb lacks its (sometimes obligatory) complement. In what follows, it will become clear that this argument is not sufficient to contradict the arguments in favour of a main clause analysis.

According to Quirk et al. (1992:§15.53), which deals with the English language exclusively, comment clauses of the type (i) presented above may have four different semantic functions, listed below as I-IV:

(I) They express the speaker’s tentativeness over the truth-value of the matrix clause, or in other words, they can mitigate its content. These comment clauses allow several different subjects, tense types and might have a modal auxiliary or appear in a passive construction. Some of them may also be negated or added to a question.

5_{In my opinion, the last two subtypes can be combined into one group that contains all comment clauses that are}

(9)

(25) John would fail the exam, he feared. (26) John loves Mary, we might assume. (27) John hates books, I’ve been told.

(28) John isn’t a very nice guy, I don’t think. (29) Where is he going, I wonder?

(II) They express the speaker’s certainty concerning the truth-value of the matrix clause. That is, they may ascertain its content. These comment clauses allow different subject and tense types as well and they too can take an auxiliary and be negated.

(30) John would fail the exam, he claimed. (31) John loves Mary, I will admit.

(32) John hates books, I don’t doubt.

(III) They express the speaker’s emotional attitude towards the content of the matrix clause. Some of them are followed by to and the infinitive of a verb of speaking.

(33) John will be here at eight, I’m happy to say.

(IV) They are used to claim the hearer’s attention or agreement. The subject of this type of comment clause is usually in second person singular. When used as a negated question they can be added to a declarative sentence to ask for the hearer’s agreement, while positive questions can be added to interrogative associated clauses to call for the hearer’s attention.

(34) John is a nice guy, wouldn’t you say? (35) Is John a nice guy, do you think?

Since the characteristics of type (I) and (II) overlap and both their semantic functions are related to expressing one’s attitude concerning the truth-value of the matrix clause, I propose to view them as one function type, which, incidentally, has strong resemblances to the reporting clauses, with which I will be comparing them. Subtypes (III) and (IV) are hereby excluded from this study.

2.3 Resemblances between reporting clauses and comment clauses

While there are some obvious differences between reporting and comment clauses, such as the kind of verb they occur with, I would like to argue that there are also several important similarities between the two constructions, such as those listed in the paragraphs below in which I use the abbreviations REP-CL and COM-CL for reporting clause and comment clause

respectively. Positional liberty

Both reporting and comment clauses can occur in initial, medial and final position with respect to the associated clause.6

6_{However, keep in mind though that the first example involves subordination, but the second parenthesis, as}

(10)

(36) He said: ‘your house is very old.’ [REP-CL]

‘Your house’, he said, ‘is very old’. ‘Your house is very old’, he said.

(37) I suppose your house is very old. [COM-CL]

Your house, I suppose, is very old. Your house is very old, I suppose.

The apparent lack of a direct object

Both clause types contain a so-called object gap. In other words, the sub-categorisation feature of the reporting or commenting verb is not always fulfilled. Often, its internal argument is left empty, as shown in the following examples.

(38) [‘Your house is very old’]clause 1 [he said ø ]clause 2 [REP-CL] (39) [Your house is very old]clause 1 [I believe ø ]clause 2 [COM-CL]

The option of having an overt marker

They both allow for the use of an overt marker at the beginning of the comment or reporting clause, which is used to connect the parenthetical clause to the associated clause it occurs with. For instance, in Dutch, the adverb ‘zo’ (so) can be used.

(40) ‘Jouw huis is erg oud’, zo zei hij. [REP-CL]

your house is very old so said he ‘‘Your house is very old’, so he said.’

(41) Jouw huis is erg oud, zo lijkt het. [COM-CL]

your house is very old so seems it ‘Your house is very old, so it seems.’

They both involve propositional modality

Palmer (2001) argues that evidential modality and epistemic modality are both instances of propositional modality, since each of them is concerned with the speaker’s attitude towards the truth-value of the proposition.7

Reporting clauses are usually an example of evidential modality, since they provide information about the source of the reported clause. In other words they can express hearsay. Comment clauses, though, generally form an example of epistemic modality because they can serve to mitigate or ascertain the faith the speaker has in the truth value of the proposition in the associated clause.

The inability to occur as an independent clause

Neither comment nor reporting clauses can occur as an independent clause, since in that case they would both lack a direct object.

7_{There are linguists that argue that evidentiality is a sub-type of epistemic modality (Cf. for example Kiefer}

(11)

(42) * [He said ø]clause [REP-CL]

(43) * [It seems ø]clause [COM-CL]

The fact that they can have an impersonal-passive counterpart

The active form of both reporting and comment clauses can be transformed into an impersonal passive construction.

(44) ‘John is a nice guy’, Mary said to me. [REP-CL]

‘John is a nice guy’, it was said to me (by Mary).

(45) John is a nice guy, Fred led me to believe. [COM-CL]

John is a nice guy, I’ve been led to believe (by Fred).

Furthermore, as De Vries (2007) notes, non-embedded quotations, which are usually combined with a reporting clause, can also be combined with a finite comment clause. This is shown in (46), which contains several (underlined) deictic elements reflecting the point of view of the original speaker. This contrasts with the free indirect speech in (47), where the tense of the verb, the pronouns and the adverb of place reveal the perspective of the current speaker (who is not Ferdinand).

(46) ‘If nobody comes looking for me, I will be stuck here for a long time’, Ferdinand feared.

(47) If nobody came looking for him, he would be stuck there for a long time, Ferdinand feared.

So it seems that, apart from sharing many important features, both reporting clauses and comment clauses can be added to a non-embedded associated clause, whether it is a quotation or not. This is important because it entails that both comment and reporting clauses are independent main clauses, in other words parentheticals. As will become clear in the next section, when comment and reporting clauses are used in a non-embedded context, they must be analysed syntactically as parentheticals as well. Now that it is clear that comment clauses as well as reporting clauses can be used in the same contexts, it appears that they both fulfil similar semantic roles as well. Reporting clauses cannot only be used to report on the language of others, but also to express the speaker’s attitude concerning the associated clause. In other words, reporting clauses can function as an epistemic modal, which is a function usually attributed to comment clauses. On the other hand, comment clauses, aside from their usual function (to express epistemic modality and consequently ascertain or to mitigate the associated proposition) can also be used to report on the source of the information that is given in the associated proposition, which means that they can function as an evidential, which is generally taken to be a characteristic of reporting clauses.

(12)

SEMANTIC FUNCTION

MODES OF REPORTING

EPISTEMIC MODALITY EVIDENTIALITY

direct speech ‘I am going crazy’, Ben thought.

‘I am going crazy’, Ben said.

Non-embedded

indirect speech He was going crazy, Ben thought.

He was going crazy, Ben said.

direct speech Ben thought: ‘I am going crazy’.

Ben said: ‘I am going crazy’.

Embedded speech

indirect speech Ben thought that he was going crazy.

Ben said that he was going crazy.

Table 2: The different contexts in which comment and reporting clauses can occur, in combination with the semantic functions they can fulfil.

An example will further illustrate the overlap in semantic function. Consider (48), in which a non-embedded associate clause in indirect speech is added to a reporting clause. In this case she and Angela could be the same person. Angela might have originally said I will be late and the current speaker is reporting on that. In this case the reporting clause functions as an epistemic modal: the fact that Angela will be late is ascertained by the fact that she said so. In (49), though, Angela and she cannot be the same person because the associated clause is a direct quote. In other words, Angela can only have said She will be late and in that case, she cannot be Angela. Here, the reporting clause functions as an evidential because it mainly serves as an indication as to the source of the information expressed in the associated clause.8 (48) She will be late, Angela says. [Epistemic reporting clause] (49) ‘She will be late’, Angela says. [Evidential reporting clause] These same functions can also be performed by comment clauses, as shown in (50) and (51). This means that the modality of the reporting or comment clause depends on the context provided by the associated clause.

(50) She will be late, Angela fears. [Epistemic comment clause] (51) ‘She will be late’, Angela fears. [Evidential comment clause] In brief, I propose that there are (at least) two semantic functions that can be fulfilled by both reporting and comment clauses. Firstly, they can report on the language of others, which is an example of hearsay and therefore evidentiality, and secondly, they can also be used to express the attitude of the speaker concerning the truth-value of the matrix clause by mentioning the source of the information or expressing doubt or certainty, which makes them a part of the domain of epistemic modality.

It seems reasonable to assume that this distinction in modal function would also lead to a distinction in syntactic structure, and so I expect that, in the end, we will arrive at one syntactic analysis to account for parenthetical clauses with an epistemic function (independently from whether they contain a reporting or a comment verb and whether they are attached to a quote), and another one for parenthetical clauses with an evidential function (again, without distinguishing between comment or reporting verbs and the context they occur in).

8_{Often, an epistemic reporting or comment clause simultaneously indicates the source of the information given}

(13)

In sum, there are four parameters that play a role in both comment and reporting clauses: embeddability, direct vs. indirect speech, the possibility of being added to a direct quote and modality. However, the distinction between direct and indirect speech coincides with the feature +/- quote. When a comment or reporting clause is added to a direct quote, the deictic elements in it automatically reflect the perspective of the original speaker and when a comment or reporting clause is not added to a direct quote, the deictic elements reveal the point of view of the current speaker. Since these two parameters overlap, only the feature +/- quote is presented in Table 3, which is the result of combining the features that are shared by comment and reporting clauses into one overview.

MODALITY

EMBEDDABILITY +/-QUOTE

EPISTEMIC EVIDENTIAL

Embedded +Quote John thought: ‘I’m going home.’ John said: ‘I’m going home.’

Non-embedded -Quote He was going home, John thought He was going home, John said.

Embedded -Quote John thought he was going home. John said he was going home.

Non-embedded +Quote ‘I’m going home’, John thought. ‘I’m going home’, John said.

Table 3: Relevant parameters in comment and reporting clauses.

Concluding, comment and reporting clauses are hereby defined as independent clauses that can be added to an associated projection (that can optionally have the form of a direct quote) in order to either have an epistemic function (i.e. communicate the speaker’s attitude or certainty towards the associated clause) or to express evidentiality (i.e. indicate the source of the information presented in the associated clause).

(14)

3. Theoretical background

Since both construction types have not yet been treated as one and the same, different syntactic analyses have been proposed for comment and reporting clauses in the literature. In this sub-section I will first present the syntactic theories postulated for reporting clauses that accompany direct quotes, followed by those that have been proposed for comment clauses, either with a traditional comment verb or with a reporting verb. Finally, I will explain how I intend to account for the syntactic structure of both clause types in this study. The preliminary conclusions drawn in this section, which are mostly based on Germanic data will then have to be confirmed by the Romance data obtained from the two questionnaires presented in Sections 5 and 6.

3.1 Reporting clauses: syntactical analyses

De Vries (2006), which deals with reported direct speech in Dutch, follows Schelfhout (1998) in arguing, as I do, that quote-final constructions have a different syntactic structure than those where the quote is either followed by the reporting clause or interrupted by it. Not only do they differ in their intonation pattern, but quote-initial constructions also have more freedom when it comes to inversion and the verb types they can be used with. Subject-verb inversion is impossible in quote-final constructions,9

while it is completely acceptable, and even obligatory in Dutch and German, in quote-initial constructions, as shown in examples (52)-(55).

(52) John said: ‘I’m going home’. (53) * Said John: ‘I’m going home’. (54) ‘I’m going home’, John said. (55) ‘I’m going home’, said John.

Obviously, all reporting clauses, independently of where they are positioned, can make use of reporting verbs to introduce the direct quote. They can contain verbs of speech and writing, verbs of emotional speech, verbs of arguing, verbs that express direct emotion, sound or thought and even verbs of observation and explanation. But quote-initial constructions can also make use of verbs that are technically not related to the reporting of language, as long as they can be explained by the context, which is impossible in quote-final constructions as shown in examples (56) and (57):

(56) ‘Let me go! I don’t want to go home!’, Jenny pulled herself free. (57) * Jenny pulled herself free: ‘Let me go! I don’t want to go home!’10

Reporting clauses usually contain transitive verbs, which might be why it is so tempting to think the quote is simply its direct object, but, in contrast with quote-final constructions, quote-initial constructions can also take intransitive verbs, verbs that do not have an available object position, or even those that contain no verb at all:

9_{De Vries (2006) justly notes that subject-verb inversion can, of course, be triggered by other mechanisms, like}

a yes/no question, imperatives or connected discourse.

10_{Obviously, when the two clauses are presented as two separate sentences (Jenny pulled herself free. ‘Let me}

(15)

(58) ‘I’m tired, can we go home?’, Lisa yawned. (59) * Lisa yawned: ‘I’m tired, can we go home?’.

(60) ‘I am very mad at you! You stole my diary!’, Sandra told off her little brother. (61) * Sandra told off her little brother: ‘I’m very mad at you! You stole my diary!’. (62) ‘The war is over’, thus the president.

(63) * Thus the president: ‘The war is over’.

It appears to be so that in quote-final constructions the quote has to be related to a complement position of the reporting verb, otherwise the sentence is ungrammatical. Because of these fundamental differences between quote-initial and quote-final constructions De Vries (2006) proposes that we are dealing with two different syntactic structures as well. He follows Luif (1990) in stating that quotes are autonomously used substantives11

and so quote-final constructions can simply be analysed as a matrix clause with an embedded quote.

However, for quote-initial constructions he postulates a different analysis. He argues that the relevant reporting clauses are parentheticals, and therefore main clauses, which are adjoined either at the end or in the middle of the direct quote. To account for the fact that subject-verb inversion can take place, despite of the fact that the parenthetical is a main clause, and the fact that the reporting clause contains an object gap, he assumes that a Dutch reporting clause in a quote-initial construction contains an abstract quotative operator. This operator has the meaning of the adverbial ‘zo’ (so). This would mean that a reporting clause in quote-initial sentences would have the following structure:

(64) [CP OPi V [IP … ti … tv … ] ]

The abstract quotative operator can be replaced by an overt adverb zo (so), which also obligatorily triggers subject-verb inversion. Whether overtly present or not, in reporting clauses with a transitive verb, the operator also serves to license the empty object, as shown in example (65).

(65) ‘I love you’, OPi

declared Romeo [e]i .

De Vries states that zo (so) can be used in almost all contexts, but that the further away the meaning of the reporting verb is from the canonical say, the more it is interpreted as the manner in which someone did or said something. I might add that the fact that zo (so) cannot be used in final constructions, as shown in (66), further corroborates the fact that quote-final constructions require a different analysis than quote-initial constructions. One might argue that example (66) is acceptable when the reporting clause serves as an introduction to an example of the type Jan is een aardige vent. Zo deed hij gisteren nog boodschappen voor me. (John is a nice guy. ‘So’ he did me groceries for me yesterday’). But it is clear that (66) and (67) cannot receive completely the same reading.

11_{Arguments in favour of this view are the following;}

‐ The fact that a quote can be in another language. (John said: ‘Je t’aime’ vs. *John said that je t’aime.)

‐ Quotes cannot be paraphrased, when intended as an explanatory apposition. (#John said ‘doe’, or ‘female dear’ vs. John said that it was a doe, or female dear.)

‐ Quotes can be incomplete, in the sense that only part of an utterance can be repeated, while leaving out the rest of it.

(16)

(66) * Zo verkondigde Romeo: ‘ik houd van je’. so declared Romeo I love of you ‘so declared Romeo: ‘I love you’.’

(67) ‘Ik houd van je’, zo verkondigde Romeo. I love of you so declared Romeo ‘‘I love you, so declared Romeo.’

According to Schelfhout (1998) the quotative operator functions as a free anaphor, pointing to the quote. Since it must have this referent to point to, the quote cannot succeed the reporting clause containing the quotative operator, which explains why (66) is ungrammatical.

The reporting clause and the quote are to be generated separately, after which the reporting clause is inserted parenthetically within or after the quote by using parenthetical merge (for more on parenthetical merge see De Vries (to appear)), which creates a paratactic relationship between the two clauses, instead of a hypotactic one.

Suñer (2000) proposes a similar analysis for Spanish reporting clauses, while making several adaptations to Collins & Branigan’s (1997) article on which her theory is based. Since Spanish is a null-subject language, and English is not, she demonstrates how the parenthetical approach in combination with a quotative operator, which is generated in the object position of the reporting verb and moved to the Spec of FP (the Focus Phrase), accounts for the different properties of both languages. Before briefly mentioning some of these differences, I will first introduce the syntactic structure she proposes for constructions containing a reporting clause. The tree-structure in Figure 1 represents the linear structure of the reporting clauses presented in (68) and (69), the first of which is an English example and the second a Spanish one. In Figure 1, the English reporting clause is presented in bold and the Spanish one in italics. Since neither Suñer (2000) nor Collins and Branigan (1997) mention how the direct quote and the reporting clause should be joined, I will adopt De Vries’s (to appear) theory and propose to add a Par(enthetical)P(hrase) at the beginning of the reporting clause, which will tie the two clauses together in a paratactic way.

(68) [[QUOTE] [ParPPar [CPC [FPQOpi [F’[+qu] ([+foc]) [TP[T’[T[AspP [Aspsaidk [vPJohnj [v’tk [VPtk ti]]]]]]]]]]]]]

(69) [[QUOTE] [ParPPar [CPC [FPQOpi [F’[+qu] ([+foc]) [TPproj [T’ [Tdijok [AspP[Asp[vPtj [v’tk [VPtk ti]]]]]]]]]]]]]

(17)

Figure 1: Syntactic structure proposed by Suñer (2000) for quotative inversion in English and Spanish reporting clauses.

Negation

In English, quotative inversion cannot occur when the reporting clause contains a negation that has scope over an adverb, while this is possible in Spanish, as is shown in the examples below. Suñer (2002) accounts for this difference by claiming that in English the negation is directly below T and so it has scope over the verb because the latter has not moved out of AspP and therefore the action of reporting is negated, which makes the sentence ungrammatical. Yet in Spanish the verb is above AspP because it has moved to T and so the negation can only have scope over the adverb, and the sentence does not become ungrammatical because it is not the action of reporting that is negated, but the way in which this is done, as is shown in examples (70)-(75), which have been taken from Suñer (2000:548).

(70) ‘Let’s eat’, said John just once. [En]

(71) * ‘Let’s eat’, not said John just once. (72) * ‘Let’s eat’, said not John just once. (73) * Let’s eat’, didn’t John say just once. (74) * ‘Let’s eat’, John didn’t say just once.

(75) ‘Comamos, no dijo Juan una sola vez. [Sp]

Complex verbs

(18)

complex verbs in English, as shown in (76)-(80). When quotative inversion does not occur in English, the auxiliary is placed in v and the participle in V.

(76) ‘What time is it?’, John was asking of Mona. [En]

(77) * ‘What time is it?’, was John asking of Mona. (78) * ‘What time is it?’, was asking John of Mona.

(79) ‘¿Qué hora es?’, le había preguntado Juan a Mona. [Sp]

(80) ‘Quelle heure est-il?’, a demandé Jean à Mona. [Fr]

Adverbials

One final difference in cross-linguistic properties that is the consequence of the fact that the verb moves to Tin Spanish, but stays in Aspin English, is the position of adverbials, which is much more free in Spanish than it is in English, as is shown in (81)-(84), which have been taken from Suñer (2000:536). When looking at the distribution of adverbials in Spanish in (85)-(88), it can be said that they might adjoin either to vP or to VP, since they can be added between the verb and the subject, but also between the verb and a non-fronted quote. In English, it appears the adverbial can only be adjoined at the right periphery of the sentence.

(81) ‘I am leaving’, John shouted abruptly. [En]

(82) ‘I am leaving’, shouted John abruptly. (83) ‘I am leaving’, * shouted abruptly John.

(84) ‘I am leaving’, ? abruptly shouted John.

(85) ‘Me voy’, Juan dijo de pronto. [Sp]

(86) ‘Me voy’, dijo Juan de pronto. (87) ‘Me voy’, dijo de pronto Juan. (88) ‘Me voy’, de pronto dijo Juan.

Now that it is clear that a parenthetical analysis functions best to account for the (cross-linguistic) properties of direct quotes in combination with a reporting clause, it is time to take a look at the structure of comment clauses in order to see whether there are any syntactic similarities between the two construction types, in addition to the descriptive resemblances mentioned in Section 2.

3.2 Comment clauses: syntactic analyses

In the literature, the syntactic analyses proposed for comment clauses generally focus on extraction on the one hand and on parenthetical insertion on the other hand. The former theory assumes that the comment clause is actually the main clause and that (an element from) the subordinated clause is extracted and placed into the main clause. This analysis has been defended by Ross (1973) and McCawley (1982), among others. The parenthetical analysis has been argued for by Reis (1996), Corver & Thiersch (2002)12

, Schelfhout (2000) and Schelfhout (2004), among others and proposes that the comment and the associated clause are both generated separately and later joined together.

12_{In fact, Corver & Thiersch (2002) divide comment clauses up into two subtypes and propose a}

(19)

In what follows, I will first discuss Ross (1973) as a representative of the extraction analysis, demonstrating in which contexts it fails to account for the distribution and characteristics of comment clauses, while at the same time presenting evidence in favour of the parenthetical analysis, of which the main characteristics have already been presented in the preceding section for reporting clauses. I will then review Corver & Thiersch (2002) and Reinhart (1983), upon which the former article is based, to introduce one last perspective to the analysis of the structure of comment clauses.

3.2.1 Ross (1973), Bresnan (1968) and Jackendoff (1972)

In Ross (1973) arguments are given in favour of an analysis that argues that comment clauses are generated as a matrix clause in which the associated clause is embedded. The embedded clause is then ‘sentence lifted’, or ‘slifted’, into a sentence-initial or sentence-medial position to form a new sentence and the subordinator that is omitted. In Figures 2a and 2b, in which ‘S’ is used for ‘sentence’, this process is illustrated.

Figures 2a and 2b: Structural representation of Ross’s (1973) Slifting analysis.

With this extraction analysis he argues against the theories proposed by Bresnan (1968) and Jackendoff (1972), who state that comment clauses are generated as sentence adverbs in final position, as shown in Figure 3.

Figure 3: Structural representation of Bresnan’s (1986) and Jackendoff’s (1972) sentence adverb analysis.

Ross (1973) presents several arguments in favour of slifting, which I will go over now in order to demonstrate that neither Ross’s (1973) nor Bresnan (1986) and Jackendoff’s (1972) analysis can completely account for the distribution and characteristics of comment clauses, while the parenthetical analysis does.

Ross’s (1973) first presents four arguments in favour of the slifting analysis that deal with the notion of ‘influence’: he presents four phenomena (complementizer choice, factivity, sequence of tenses and non-specificity) in which one element in the syntactic structure influences the occurrence or interpretation of another. He then argues that the fact that this influence exists entails that at one time in the formation of the sentence, the influencing node must have dominated the influenced node.13_{For instance, Ross (1973) proposes that only}

13_{It can be noted that the other arguments presented in favour of the slifting analysis mostly deal with influence}

(20)

verbs that take that-clause complements can have their objects slifted, which, according to him, accounts for the ungrammaticality of (89) versus (90), which is acceptable, since the associated clause functioned as the object of the comment clause at some point in the derivation.

(89) Hilda has been brainwashed, I fear.

(90) * Hilda has been brainwashed, I eat/try/begin.

However, there are several examples of comment clauses that contain intransitive verbs, verbs that take a noun-phrase as its object or comment clauses without a verb at all, as is demonstrated in the examples below. In my view, this constitutes a strong argument against any extraction analysis, which of course is based on the fact that the associated clause has at some point been embedded into the comment clause. In the case of the verbs used in the examples (91)-(93) the sub-categorisation feature of the comment-verbs used could never be fulfilled by the associated clause and therefore the latter could never have been generated as the former’s complement clause. For example, it is clear that She doesn’t know what she is doing can never function as the object of I defended my sister, since there is no vacant object position available in the comment clause, but even if there had been one available, to defend takes a DP complement, instead of a CP complement. Therefore, slifting could not account for sentences such as (91)-(93), while the parenthetical analysis does.

(91) Hilda has been brainwashed, I grimaced.

(92) I want to go to MacDonalds, I badgered my sister. (93) Hilda has been brainwashed, thus the president.

Ross (1973) makes a similar argument concerning the use of factive verbs in the comment clause. He states that the proposition expressed by their complements must always be considered to be true by the speaker, which explains the awkwardness of both (94), a normal embedded sentence, and (95), which is its slifted counterpart.

(94) ? Max suddenly realized dogs can fly. (95) ? Dogs can fly, Max suddenly realized.

Yet, this appears to be more a matter of pragmatics than of syntactic structure, since there are several (imaginary) contexts one could think of in which (94) and (95) might be acceptable.

Concerning sequence of tenses, or tense agreement, it is argued that non-factive verbs in the past tense are incompatible with a present tense in its embedded complement-clause and that the same thing can be observed in slifted comment clauses, as shown in (96). This would constitute proof that the comment clause must have previously dominated the associated clause and has been slifted afterwards. However, as Kaltenböck (2007) duly notes, counterexamples such as (97) show that this phenomenon does not pose a problem with several other non-factive verbs.

(96) * There is something funny about Venus, it seemed to me. (Ross 1973:139)

(97) There is something funny about Venus, I thought/claimed. (Kaltenböck 2007:36)

(21)

One last phenomenon that, according to Ross, is related to ‘influence’ is the fact that some indefinite DPs in complement clauses can acquire a definite reading, depending on the main verb they occur with: according to Ross (1973), sentence (98) does not allow a definite, or specific, meaning, while (99) does. He has the same intuitions concerning the slifted equivalents of these sentences, as presented in (100) and (101), which again would prove that the slifted clause was once the main clause and the associated clause at some point embedded into this main clause.

(98) I hope that Don Giovanni has baptized a Communist. (Ross 1973:143)

(99) I notice that Don Giovanni has baptized a Communist. (100) Don Giovanni has baptized a Communist, I hope. (101) Don Giovanni has baptized a Communist, I notice.

Yet, in my opinion, (101) can easily be interpreted as non-specific when imagining, for instance, that one sees that a baptism has taken place (certain instruments used in baptisms are still laying near the baptismal font, for instance) and Don Giovanni told us earlier that he intended to baptize a Communist later that day. In that case, a non-specific reading of (101) appears to be completely acceptable. In short, most of the arguments made in favour of slifting when it comes to the four ‘influence’-phenomena presented above, do not add up and some of them, when looked at properly, actually constitute as proof against the slifting analysis. These phenomena do not pose any problems for the parenthetical analysis.

Ross (1973) continues by presenting two arguments that concern restrictions on person. He argues that there are certain predicates referring to internal mental states that are odd in questions in first person present tense which goes as well for certain assertions about internal states of the addressee, as in shown (102)-(103).

(102) Do {you/? I} think that extraterrestrials exist? (Ross 1973:149)

(103) It seems to {me/? you} that Marge is a witch. (104) Do extraterrestrials exist, do {you/? I} think? (105) Marge is a witch, it seems to {me/? you}.

This peculiarity occurs in slifted sentences as well, as shown in (104) and (105), which, according to Ross (1973), proves that slifting is an adequate way of dealing with comment clause constructions since the fact that the peculiarity occurs in both construction types proves that the latter are derived from the former. However, once again, this appears to be more of a pragmatic discrepancy than a syntactic one. It is unnatural to ask questions concerning one’s own mental state since people are generally aware of their own state of mind. The fact that this is the case in complex sentences as well as sentences to which a parenthetical has been adjoined, has, in my opinion, more to do with the pragmatic interpretation of such sentences that their syntactic structure. Since this problem therefore should be dealt with at discourse level instead of at the syntactic level, the parenthetical analysis does not have to account for the ungrammaticality of (104)-(105) (nor does the extraction analysis, for that matter).

Ross (1973) further claims that the slifting analysis also ties in nicely with Wh-island constraints. Since the associated clause is initially dominated by the comment clause, an element from a DP in subject position can never be moved to any position inside the comment clause, because it cannot go out of the CP it was generated in, which constitutes a barrier. Therefore sentences like (106), which were generated as (107), are justly ungrammatical under the slifting analysis.

(22)

(107) I believe [your claim [that Pete pointed out that Max has a tuba]CP]DP

However, these examples do not concern a real comment clause. We are dealing with a normal embedded sentence in which the object is complex as well. If the one tried to apply slifting to (107), the sentence would become ungrammatical, as shown in (108), in which Pete pointed out only has scope over Max has a tuba.

(108) # I believe your claim that [Max has a tuba, Pete pointed out].

Ross then turns to so-pronominalisation of the object of certain comment verbs. According to him, this can only take place in embedded contexts and is unacceptable in paratactic contexts, as shown in (109) and (110), which have been taken from Ross (1973:153). Though, when the first part of (110) is followed by a comment clause, as in (111), the sentence becomes grammatical.

(109) Max thinks that apricot paste has no calories, and his doctor thinks so too. (110) * Rufus is flatulent, and his wife thinks so too.

(111) Rufus is flatulent, I think, and his wife thinks so too.

Ross (1973) states that this provides evidence for the fact that the associated clause Rufus is flatulent was once the complement of the comment-verb. However, I think that the ungrammaticality of (110) might have more to do with the fact that too requires a referent, either in an extraction or in a paratactic context, which is not available in (110, but is provided by the comment clause in (111). Therefore, this says nothing about the adequateness of the slifting analysis.

Turning to anaphora binding, Ross (1973) states that sentence (112) is ambiguous because the pronoun his can either be bound by John, or by Ed, or by both. In other words, Ed might photograph John’s sister or his own. According to Ross (1973), this construction is formed by applying VP-deletion to the second part of sentence (113). He states that the only way in which (112) would be interpreted as John and Ed will both photograph John’s sister, would be if the rule of VP-deletion would ‘overlook’ the fact that the second his sister is bound by Ed and the first is bound by John. In Ross (1967) it is hypothesized that the only differences the VP-deletion rule can ‘overlook’ in order to delete the second photograph his sister, are differences in commanded pronouns. This would mean that in (114) Ed must at some point have c-commanded his in order to derive the (115) reading of (114), which would be the case in a slifted sentence.

(112) John will photograph his sister and Ed will too.

(113) Johni will photograph hisi sister and Edj will photograph hisj sister too. (114) [Hisi sister is annoying], [Johni feels], [and Edi feels so too].

(115) Hisi sister is annoying, Johni feels, and Edj feels hisj sister is annoying too.

(23)

Concerning negation, Ross (1973) states that comment verbs generally cannot be negated, unless the verb itself has negative force. However, there are certain verbs that do allow negation, but only if the associated clause contains a negative element as well, as shown in (116)-(119).

(116) * John is a very nice guy, I don’t think. (117) John isn’t a very nice guy, I don’t think. (118) John is a very nice guy, I don’t doubt. (119) * John isn’t a very nice guy, I don’t doubt.

He accounts for this distribution by applying the following three rules, in this order: not-copying (which places a copy of the negation in the embedded sentence into the comment clause, and which is optional), slifting (which is optional as well) and not-deletion (which deletes the original negation in the previously embedded clause, and is obligatory). Subsequently, if not-copying does not take place, neither does not-deletion. In examples (120)-(123) the entire process is demonstrated for the verb deny.

(120) [S1 I deny [S2 that John isn’t a nice guy.]] [Original embedded sentence] (121) [S1 I don’t deny [S2 that John isn’t a nice guy.]] [not-copying] (122) [S0 [S2 John isn’t a nice guy] [S1 I don’t deny]] [slifting] (123) [S0 [S2 John is a nice guy] [S1 I don’t deny]] [not-deletion] According to Ross (1973), this will account for all possible contexts. However, these additional rules that are needed to make the slifting analysis work, do not independently explain the distribution that is observed: there is nothing in the structure of these sentences that determines whether not-copying must take place in order to end up with a grammatical sentence or not. Furthermore, in many cases the transfer of the negation from the associated to the comment clause causes a change in meaning of the sentence as a whole, which is unwanted. In the parenthetical analysis a negation can easily be added if necessary, but it too cannot independently account for the distribution of negation in English and the fact that negative concord causes two negations to act as only one. Both analyses therefore have different problems when it comes to negation: the slifting analysis requires additional rules in order for it to work and then is nowhere nearer explaining why a negation is sometimes needed in the first place, and the parenthetical analysis can easily enough add a negation if necessary, but struggles with the link that must somehow exist between the associated clause and the comment clause in order to turn the two syntactic negations into one semantic one and cannot explain its distribution either. At best it can be said that the parenthetical analysis deserves our preference here, because no additional rules are needed to form the structures required in contexts with negative concord. An explanation as to the reasons behind this distribution will have to be studied at another occasion.

(24)

(124) I promise you to be home early, Max knows. (Ross 1973:160)

(125) I promise you to be home early, Max thought.

In short, the arguments provided by Ross (1973) in favour of his slifting analysis, which is an example of an extraction analysis (as are Bresnan 1968 and Jackendoff 1972, who assume that the comment clause is generated as a sentence-adverb that can be extracted and placed into a sentence-medial position as well), mostly do not add up, and while he tries to account for the syntactic structure of comment clauses, his arguments are often more related to pragmatics, which is not relevant here. While it is clear that there will need to be established some kind of (binding-) relationship between the comment clause and the associated clause that accounts for the ‘influence’ that certain elements from the comment clause sometimes have on elements from the associated clause, slifting does not nearly account for all of them. This is a complex matter since some of the data on comment clauses can be explained by a hypotactic relationship between the associated clause and the comment clause, while there are also phenomena that clearly cannot. The parenthetical analysis will have to account for this as well, and it is sure that this is no easy feat. Although it proves to be more adequate than the slifting analysis in many contexts, the transparency effects that do exist between the associated clause and the comment clause remain its weakness. A solution for this problem must still be found, but in the mean time, it has been demonstrated that there are to many counter-arguments in order to continue to accept the slifting analysis as adequate, while the parenthetical analysis proved to be more suitable in most contexts.

Aside from the fact that the parenthetical analysis is more adequate in accounting for the distribution and characteristics of comment clauses than the slifting analysis, as demonstrated above, one of the more general arguments in favour of the parenthetical analysis is the fact that it can account for the main clause word order in the comment clause as well as the associated clause, since it considers them both to be independent main clauses. In Dutch, main clauses have verb-second word order and subordinate clauses are verb-final, as shown in (126).

(126) [Jan weet] [dat Bob een kat heeft]]. John knows that Bob a cat has ‘John knows that Bob has a cat.’

As shown in (127), which is the slifted variant of (126), the associated clause cannot have the word order of a subordinated clause. In addition, the comment clause must undergo subject-verb inversion for which the slifting analysis does not account. In (127), the associated clause must have main clause word order in Dutch, as shown in (128), so this makes it impossible for the associated clause to have ever been embedded in the comment clause and eliminates the possibility of slifting having taken place.

(127) * Bob een kat heeft, Jan weet. Bob a cat has John knows (128) Bob heeft een kat, weet Jan.

Bob has a cat knows John

(25)

fact the covert operator that precedes the verb. This way, the comment clause still has verb-second word order, which confirms the fact that it is a main clause.

Another argument in favour of the parenthetical analysis is the fact that, when assuming that sentences containing a comment clause consist of two independent clauses, the associated clause no longer can be interpreted as the complement of the comment clause, which, as demonstrated above, is syntactically impossible in certain cases. (Cf. examples (91)-(93), here repeated as (129)-(131)).

(129) Hilda has been brainwashed, I grimaced.

(130) I want to go to MacDonalds, I badgered my sister. (131) Hilda has been brainwashed, thus the president.

In sum, the parenthetical analysis seems to be the most adequate way of accounting for the properties of comment clauses. In what follows, I will touch upon one last theory that has been proposed concerning their nature.

3.2.2 Reinhart (1983) and Corver & Thiersch (2002)

Reinhart (1983) argues that there exist two different kinds of comment clauses, and that both of them require a different syntactic analysis. This theory has been elaborated on by Corver & Thiersch (2002) and I will now present the main ideas introduced in both papers. Reinhart (1983) argues that a distinction must be made between subject-oriented (SuO) sentences containing a parenthetical clause (which include comment clauses) and speaker-oriented (SpO) sentences containing a parenthetical clause.14

In the first case the sentence is oriented towards the subject of the parenthetical-clause and in the second towards the current speaker.15

The difference between the two types is best explained by an example. The sentence in (132) can be considered ambiguous, as becomes clear when a context is added to it, as in (133)-(134).

(132) John’s car was stolen, Mary thinks.

(133) A: What does Mary think? (SuO)

B: John’s car has been stolen, Mary thinks.

(134) A: Why is John so sad? (SpO)

B: John’s car has been stolen, Mary thinks.

In (133) the entire sentence uttered by B is the answer to A’s question, while in (134) only the associated clause John’s car was stolen can be considered the answer to A’s question and the parenthetical Mary thinks is only added to mitigate the faith speaker B has in the truth-value of the proposition uttered in the associated clause.

14_{Even though Reinhart (1983) talks about sentences containing a parenthetical clause she will first propose to}

analyse both types according to an extraction analysis, which is based on the idea that the parenthetical clause has been generated in a subordinate position, in which case it can never be a parenthetical, since these are main clauses.

15_{The distinction made based on orientation must not be confused with the perspective of the associated clause}

(26)

Reinhart (1983), Corver (1998) and Corver & Thiersch (2002) present several syntactic tests in order to distinguish between SuO and SpO clauses. In Table 4, I give a brief summary of these tests, relying on the judgements made by the authors of the aforementioned articles concerning the grammaticality of the example sentences. Several comments can be made concerning their adequateness, but I will present the tests before turning to the objections that can be made against them.

TEST SPO SUO

I

Pronominalisation 16

Preferred forward pronominalisation: A: Where’s John?

B: Johni would be late, hei feared.

B’: ?Hei would be late, Johni feared.

Obligatory backward pronominalisation: A: What did John fear?

B: Hei would be late, Johni feared.

B’: #Johni would be late hei feared.

II Embeddability

Embeddable:

We’ll hire Mary [because Johni likes her,

hei thinks.]

Not embeddable:

*We’ll hire Mary [because hei likes her,

Johni thinks.]

III or so …

Can take or so…: A: Where’s John?

B: Johni would be late, or so hei feared.

Cannot take or so…: A: What did John fear?

B: *Hei would be late, or so Johni feared.

IV Subject verb

inversion

Subject-verb inversion is impossible: A: Where’s John?

B: *Johni would be late, feared hei.

Subject-verb inversion is possible: A: What did John fear?

B: Hei would be late, feared Johni.

V Quantifier

binding17

Q-binding is impossible: A: Is the ambiance good here? B: *No, hei would be fired, everybodyi

expected.

Q-binding is possible:

A: What did everybody expect?

B: Hei would be fired, everybodyi expected.

VI

Scope of negation

Negation has scope over part of the host clause:

A: The interviews have begun.

B: [Mary has a good chance] thought John, or maybe not.

Negation has scope over the whole host clause:

A: What did John think about Mary? B: [He liked her] thought John [or maybe not].

VI

Prosodic features

Has the prosodic features of a sentence with an adverb:

John would be late, he feared. John will be late, obviously.

Has the prosodic features of a direct quote: He would be late, John feared.

‘I will be late,’ John feared/thought.

16_{Test I-IV and VI are taken from Reinhart (1983).}

Reporting and Comment Clauses

Reporting and Comment Clauses

A cross-linguistic study

Table of contents

1.

Introduction

2.

Empirical background

3.

Theoretical background