Cycnos | Volume 17 n° Spécial Journées Charles V sur les propositions relatives et l'aspect "be+ing" - 

Naomi Malan  : 

The relative clause: An antecedent with a will of its own and a pronoun Ø one cannot ignore


Texte intégral

1As suggested in the title*, I propose to discuss (A) the syntactic, semantic and pragmatic features distinguishing the indefinite head relative from the definite head one; (B) the status of the marker Ø as opposed to that of the wh- relative pronoun and that. In both cases, I shall go on to discuss what bearing these findings could have on the choice of the relative pronoun. Finally, in the light of the above discussion, I shall conclude by (C) very briefly presenting a system for the above choice.

2I shall be using the label 'attached relative clause' to refer to what is known in English as the defining or restrictive relative clause; and the label 'detached relative clause 'to refer to what is called the non-defining, non-restrictive or explanatory clause. My choice of labels is not an arbitrary one. The terms 'attached' and 'detached' refer exclusively to the syntactic properties of the two types of clause: the former is a constituent of the noun phrase corresponding to its antecedent; the latter is not. The labels in general use refer to the semantic properties attributed to these structures and this can introduce confusion. For example, if a relative clause does not restrict it is sometimes thought to belong to the non-restrictive category (i.e. to be preceded by a comma); if a relative clause has an explanatory function, then equally it is sometimes thought to belong to this latter category. However, this is not the case. A distinction in terms of purely syntactic criteria enables one to avoid such pitfalls.

3R.L. Allen points out that a refers to what is unidentified whereas the refers to what is identified. There is, of course, no one-to-one correspondence between the use of the indefinite article and what is unidentified on the one hand and the definite article and what is identified on the other. For example, with cataphoric determination, the definite article does not refer to an identified element. Generic utterances provide another exception, and so on. However, in this discussion, I propose to simplify and to say that:

  • (a) the reference of an antecedent accompanied by the indefinite article is here considered to be non-retrievable in context / situation;

  • (b) the reference of an antecedent accompanied by the definite article is here considered to be retrievable in context / situation.

4Many if not most linguists and grammarians tend to use definite head examples when analysing the properties of the attached relative clause. This leads one to assume that no distinction need be made between attached clauses with definite heads and those with indefinite ones. One assumes that both play the same role but in fact this is not the case. The non-retrievable vs. retrievable parameter just mentioned has semantic and pragmatic implications which, in turn, trigger off not only semantic and pragmatic distinctions but also syntactic ones.

5It is a well-known fact that in written English the comma ought to be used if the utterer wishes to indicate that the relative clause is a detached one; this necessity is of course explained by the fact that, with a definite head, there can be referential ambiguity if the comma is omitted:

1.The storm was particularly cruel to the Spanish ships which had taken the worst punishment in the battle. (E. Thomson, Sir Francis Drake)

6According to whether the comma was omitted or whether none was intended, the storm was particularly cruel to all the Spanish ships (detached); or else it was cruel only to those which had taken the worst punishment (attached).

7However, for a non-generic reference, there can be no referential ambiguity when the comma is omitted if the head is indefinite. It is not possible for there to be any confusion of reference for an element presented for the first time.

8There seems to be only one exception to this, the case where the antecedent comprises more than one noun:

2. The Abbot Gregor Mendel in Brunn had, during the sixties, not only obtained the same result through extensive experiments with peas, which lasted for many years [...] but had also given the same explanation. (www)

9Here, the comma indicates that the whole noun phrase is referred to; in other words, it is not the peas that lasted for many years but the extensive experiments on them.

10Since, for indefinite heads, the comma is normally not required to solve problems of referential ambiguity, this means that for the most part they (relatives with indefinite heads) have the syntactic and suprasegmental features of the attached relative clause. Example (2), just mentioned, provides an exception. Another exception is the continuative relative clause1:

3. He appeared the last night of the carnival, danced with the two elderly protestant girls, excelled himself at the rifle range, and won a jug, which he gave to Mabel. (E. O’Brien, A Fanatic Heart, p. 88)

11Here, the giving of the jug to Mabel cannot enter into the description of its referent at the time of the event referred to in the matrix clause (the winning of the jug); only on a subsequent occasion can it do so (the jug he won at the carnival which he then gave to Mabel).

12This type of exception, where a comma indicates the detached status of the relative clause, helps one to understand why, for pragmatic reasons, it is necessary for there to be two main types of relative clause: attached and detached. The opening of a new tone unit2, marked by the comma and indicating the detached status of the clause serves as a warning to the addressee. Apart from the type of relative clause exemplified by (2), this warning can be interpreted as follows: the information contained in the relative clause does not contribute towards the description of the antecedent for purposes of identification: “what are we talking about?”

13What has just been said explains why an indefinite head relative generally has the syntactic features of the attached clause. As the head (plus the post-modification) is presented for the first time, the contents of the relative clause automatically build up its identity: “What are we talking about?” In other words, the above type of warning is not required, it serves no purpose.

14An important consequence of the syntactic behaviour peculiar to the indefinite head relative as opposed to the definite head one concerns their respective roles. As is known, the latter relegates its explanatory role to the detached clause. Indeed that is the sole function of the detached clause. But what happens when there is no detached clause to assume this function? I am talking about the indefinite head relative. Here, the explanatory role has to be conveyed somehow; so, despite its attached nature, the indefinite head relative has to and does perform this function. And not infrequently, that is its sole raison d'être. This is what sets it apart from the definite head type as regards its semantic facet:

4. His car pulls up at the college gates. [...] A correct porter in a bowler hat leads him to my staircase. (I am watching all this from an upper window that overlooks the court) (Graham Swift, Ever After)

15In the above utterance, everything in parenthesis is of an explanatory nature, including the contents of the relative clause. Without the latter, there could be a communications breakdown: how is it possible for the narrator to be seeing what goes on in the courtyard?

16Note that the relative pronoun in this utterance is that, despite the fact that the information conveyed by the relative is new information (confirmed here by the indefinite head). More often than not one does indeed find that or which for an attached relative clause of this nature. This seems to call for a reappraisal of any system whereby wh- signals a relationship that is non-retrievable in context; and whereby that signals a relationship that refers back. Such a distinction would imply that all definite heads call for that; and all indefinite heads call for wh-. However, Table 1 does not confirm this, even if it is true some of the time:

Table 1. - Attached relatives in the corpus A Far Cry from Kensington by Muriel Spark. Which and that, subject position, compared for definite vs. indefinite heads.













17The table shows that for definite heads, almost half (45.24%) of the relative pronouns correspond to which. And, for indefinite heads, more than half (56.67%) are that. Even allowing for phenomena such as cataphoric determination, generic reference etc. these correlations need explaining.

18The points raised in this section,and the statistics discussed in Table 1 suggest the following conclusions:

19Firstly, any definition of the semantic facet of the attached clause must make a distinction between indefinite heads and definite heads. Secondly, with two exceptions, the indefinite head relative has the syntactic features of an attached relative clause. Finally, the value of wh- as opposed to that might have to be reconsidered. The table suggests that even if all the occurrences of (that + definite head) indicate a retrievable reference, the occurrences of (wh- + definite head) would need explaining; as would the occurrences of (that + indefinite head).

20The table, as has been seen, refers only to relative “pronouns” in subject position. How are these percentages likely to change if one ignores the parameter of definiteness and, this time, includes relative pronouns in complement position?

21Before doing so, let us consider the status of Ø. It is a well-known fact that most grammarians consider that Ø corresponds to the omission of that. Such an analysis is probably based on what they have in common as regards syntactic constraints: neither is compatible with the phenomenon of pied piping (i.e. with both of them the preposition has to remain stranded):

5.a. *The man to that Silvia spoke (The man that Silvia spoke to)
5.b. *The man to she spoke (The man Ø she spoke to).

22Nevertheless, such a constraint does not provide sound enough evidence to confirm the hypothesis that Ø can be substituted for that without changing the message. Even though the propositional meaning remains the same, this is not necessarily true for the thematic meaning3.

23Moreover, it should be stressed that Ø can replace not only that but also the wh- pronoun. If preposition stranding is considered acceptable (less formal register), Ø becomes a possible choice:

6.a. Today, the consequences of a crisis in the Gulf are no less ominous. But the structure on which its stability depends is even more brittle. (Newsweek, 27/1/97).
6.b. Today, the consequences of a crisis in the Gulf are no less ominous. But the structure its stability depends on is even more brittle.

24Does Ø correspond to an apparently arbitrary omission of the wh- pronoun or that? Or are we dealing with a pronoun whose signifier is Ø and which is in opposition, in a system, to the other relative “pronouns”? For Saussure “la langue peut se contenter de l’opposition de quelque chose avec rien". He gives examples in Czechoslovakian where, Ø, corresponding to the genitive plural, is opposed to other forms. The same can be said for English:

7.a. Mary likes my cat.
7.b. Mary likes my cats.

25Here, the opposition between the signifiers Ø / -s corresponds, of course, to the singular / plural system.

26Another example in English is given by M. Ariel. This author points out that for T. Reinhart, in sentences where bound anaphora (with “deletion” of the pronoun) is an alternative and the utterer does not choose this alternative, then the addressee does not get the co-reference reading. No examples are given but the following manipulation illustrates this:

8.a. Mary hates having to get up early to be on time for work. (NM)
8.b. Mary hates her having to get up early to be on time for work.

27In (8.a), Ø is co-referential with the subject (Mary) of the verb hates; in (8.b) the pronoun her refers back to someone else already mentioned. So here, Ø is in opposition to a personal pronoun or name, in a paradigm - signifying co-reference with the subject.

28A final example is one given by Janine Bouscaren when she presents an analysis of the opposition between Ø and that with declarative verbs :

9.a. He said she liked it. (J. Bouscaren, 1992)
9.b. He said that she liked it.

29In (9.a), suggests the author, the utterer does not endorse the reported statement (she liked it) but only his own (he said). Whereas in (9.b), the utterer presents the reported statement as a fact: he endorses it. The above opposition shows that Ø does not have the same value as the conjunction that when it comes to the pragmatic dimension of discourse.

30The various examples just given indicate that Ø, as opposed to another operator, has a distinct value. The choice of one or the other is not arbitrary. If this is the case elsewhere, in English and in other languages, why should it be different for relative pronouns? Why should the domain of the relative clause represent an exception?

31By way of an answer, I quote P. Erades as reported by R. Allen (1966):

It may safely be said that in language a difference of form always corresponds to a difference in meaning and that whenever more than one construction is – theoretically – possible, they never wholly and under all circumstances denote the same thing. The first axiom of all valid linguistic thinking is that in language nothing can serve as a substitute for something else.

32So, if Ø has a distinct value, what are the implications? Firstly, Ø has to be a signifier in its own right, in opposition to wh- and that in a system. This in itself has further implications. In that, in standard English:

- when the “pronoun” is in subject position, the “work” is done by two operators, wh- and “that”4
- but when the pronoun is in complement position, the same “work” is done by three operators, not two (wh-, that and Ø),

33it necessarily follows that the distribution of roles has to be different depending on whether the relative “pronoun” is in subject or complement position (direct, prepositional or adverbial complement) in the relative clause. In other words, the function in the relative clause — of the “pronoun” — has to be a significant parameter. If this were not so, one would expect to find roughly as many occurrences of which in subject position as in complement position; and as many occurrences of that in subject position as of that, (or that and Ø) in complement position. But in fact Table 2 does not confirm this:

Table 2. - Attached relatives in the corpus Out of this World by Graham Swift.
(1) Which, that and compared in subject and complement position.
(2) Proportion of that vs. Ø occurrences in complement position.










Direct Object





That vs. Ø in direct object position




  • i) There are 10 times more occurrences of which in subject position (35.9%) than there are in complement position (3.1%).

  • ii) Virtually all (96.9%) of the pronouns in complement position are “that” or Ø, cf. only 64.1% of that occurrences in subject position.

  • iii) And if Ø is given a distinct value, and one looks only at the occurrences of that in both positions, then you have 64.1% in subject position as against only 9.4% in complement position, i.e. 7 times more in subject position.

34These figures seem to speak for themselves.

35The arguments put forward in this section and the statistics shown in Table 2 suggest the following conclusions:

36Firstly, Ø cannot correspond to an arbitrary omission of wh- or that. It has to be a pronoun whose signifier is Ø and which is in opposition, in a system, to the other relative “pronouns”. Secondly, the function of the relative pronoun has to be taken into consideration when determining its role, in other words the value of the pronoun cannot be considered independently of its syntactic site.

37To recapitulate briefly the substance of the last two sections, I have suggested that an explanation for the choice of the relative pronoun solely in terms of the opposition between backward reference and newly presented information is not a complete one, furthermore that Ø is a signifier in its own right and that the function of the relative pronoun in the relative clause is significant. Under the circumstances then, I should be able to come up with a system that accounts for the choice of the relative pronoun whilst taking the above-mentioned facts into consideration.

38The system I propose takes these factors into account and is anchored in the pragmatic dimension of discourse. What exactly is pragmatics? R. Huddleston5 defines it as follows: “Pragmatics deals with those aspects of utterances that go beyond the meaning of the sentences uttered.”

39My system draws its inspiration from one described by M. Ariel (1988), this itself being derived from the analyses of a dozen or so linguists, including T. Givon. It is based on the notion of hierarchy ordered as to degree of accessibility (of the reference) to the addressee. “Natural languages,” she suggests, “primarily provide speakers with means to code the accessibility of the referent to the addressee.”

40M. Ariel is not concerned with the choice of the relative pronoun, nor with new information (cf. indefinite heads). I used her analysis as a point of departure. Table 3 enables one to visualise my system.

Table 3. - Distribution of relative pronouns according to their function and the degree of reference accessibility.
-P: The identifying of the referent is considered to be non problematic for the addressee.
P: The identifying of the referent is considered to be problematic for the addressee.
+P: The identifying of the referent is considered to be particularly problematic for the addressee.












41It should be noted that in this system, as in any, who, which preceded by a preposition, when and where cannot be taken into account as they are not systematically significant. Nor can whose be either; indeed it is never significant: the only other choice in the paradigm is which preceded by of, i.e. another wh- element.

42Firstly then, why is who not considered to be systematically significant? It is because of the (+Human) factor. When that (e.g. – P in subject position) would normally be called for, this choice is neutralised, overridden by another parameter; generally speaking if the referent is human, this calls for the choice of who, irrespective of whatever other parameters may be involved6. Further along, I hope to make it clear, why, in the following utterance, the choice of pronoun would have been that had it not been for the (+Human) factor:

10.a. Toby Stephen [Virginia Woolf’s brother] had been to Cambridge and his university friends came to his new home. This was Toby’s contribution to Bloomsbury but it was a vital one.
[ ... ] But if those evenings [at Fitzroy Square] in conversation and discussion with young people were a new experience for her [Virginia Woolf], they were not new to Lytton Strachey and his Cambridge friends. Strachey and several others of the Cambridge men who came to Fitzroy Square had been members of the “Cambridge Society”.

10.b. Strachey and several others of the Cambridge men that came to Fitzroy Square had been members of the “Cambridge Society”.

43As regards which preceded by a preposition, if this is not systematically significant, it is because formal register will avoid preposition stranding, which of course precludes the choice of that or where one or other of these signifiers might otherwise have been the suitable choice had the register been less formal. The manipulation of the following utterance demonstrates this:

11.a. Since Anne Boleyn, no less than her father, was first and last a phenomenon of the court, we need to explore the milieu to which she belonged. (E. Ives, Anne Boleyn)
11.b. Since Anne Boleyn, no less than her father, was first and last a phenomenon of the court, we need to explore the milieu Ø she belonged to.

44Finally, the recourse to when and where can be explained by various constraints precluding other choices in the paradigm of “pronouns”. So even if the wh- is sometimes relevant it is not consistently so. The following represent some of the constraints:

  • (i) Cases where no preposition is suitable but a “localising indicator” is required: the unsuitability of a preposition is explained by the fact that the location / site (corresponding to the referent designated by the antecedent) and the element that is located in it are not adapted to each other:

12.a. The pool of candlelight that bathed the lawyers, the prisoner’s chair and the empty place where the witnesses stood (H. Mantel, A Place of Greater Safety).
12.b. ?The pool of candlelight that bathed the lawyers, the prisoner’s chair and the empty place Ø the witnesses stood in.

45Here, the “container” corresponding to the empty place does not have enough spatial structure for the witnesses to be situated in relation to it.

  • (ii) Metaphor slipping (glissement métaphorique). Benjamin Lee Whorf (1956) points out that most of our prepositions are borrowed from the spatial domain. So that in an utterance referring to time (or anything non-spatial) the preposition is in fact a metaphor.

46My research has shown that in the domain of the relative the metaphor does not always work (as is the case elsewhere). Sometimes it is too literal:

13.a. It was to be a while before I found myself being addressed by my first name. This certainly coincided with the time when I was moved to lose my great weight. (M. Spark, “A Far Cry from Kensington).
13.b. ?This certainly coincided with the time Ø I was moved to lose my great weight in.

47With in, the time becomes a machine the narrator sits down in in order to lose weight.

  • (iii) Syntactic constraints can also necessitate the use of when in those cases where otherwise the pronoun would be indicated without a preposition, or even that. In the manipulation of the following utterance, the removal of the interpolated clause makes it possible for Ø to be chosen:

14.a. I thought of that day when, after one of her long, terrible cries, Milly and I had run up to Wanda’s room and found her in bed. (M. Spark, A Far Cry from Kensington).
14.b. I thought of the day Ø Milly and I had run up to Wanda’s room and found her in bed.

48I now propose to look at a few examples in each category of Table 3. Unfortunately it will by no means be possible to present here all the types of discourse situations covering each category7.

49(1) -P (the identifying of the referent is considered to be unproblematic)

  • (a) Backward reference where there is no antecedent competition (that in subject position, Ø in complement position)

50This type of utterance confirms current theories, especially if Ø is considered to be a deletion of that; but, as suggested, backward reference is not the only explanation for the choice of that or Ø, as we shall see further along.

15. Excessive speed is one of the primary causes of accidents, but there are also other things that cause accidents. (Central News 2, Oxford English Video)

51The element to which the contents of the relative refer back is in the first part of the utterance (causes of accidents). But even if it had been further back, or even if it had been merely implicit in the previous context that the antecedent’s reference was the only possible one expressed by the relative [as with (10.a) above] then the choice would still have been (-P).

52Note the presence here [as with (10.a)] of the marker other, confirming that it is indeed a case of backward reference. Other linear indicators of backward reference are markers like still, despite, something else, the only.

53Note, finally, that the reference is generic. Whatever the determiner accompanying it, a generic reference is always retrievable – but more or less so, depending basically on whether or not it has been reactivated in context. This is the case here, which also explains why we have (– P).

54In the following utterance, the relative pronoun is in complement position, hence the choice of Ø:

16. Dear God, I must warn people about her, he thought. He made a mental list of the people Ø he must warn. (H. Mantel, A place of greater safety, p. 450).

55Here again, the construction to which the contents of the relative refer back is to be found in the previous sentence (I must warn people about her).

  • (b) attached relative clauses with an explanatory role

56This type belongs to a larger category of what I shall term “False Defining” relative clauses.

57It was mentioned in the first section of this paper that an indefinite head is in general not accompanied by a detached relative clause. And that the explanatory role characterising the latter when it comes to definite head types has to be conveyed somehow. So that, even though it has the syntactic properties of an attached relative clause the indefinite head type assumes that explanatory role when necessary. And not infrequently, that is it’s sole function. This was illustrated by example (4):

4. His car pulls up at the college gates. [...] A correct porter in a bowler hat leads him to my staircase. (I am watching all this from an upper window that overlooks the court) (Graham Swift, Ever After)

58If the reference here is unproblematic despite the fact that it is presented for the first time, it is because the utterer is not asking the addressee to establish the reference of the “modified” antecedent, (window) for future retrieval. The (-P) choice is a way of indicating to the reader that the reference has no thematic importance and can be ignored. The function of the relative is purely an explanatory one.

17.a. Joleen came back with a bottle of Spanish wine. A cigarette Ø she lit was again snuffed after a few intense puffs. (R. Cooper, The last to go, p. 279)

59The same goes for this utterance, where the relative pronoun is in complement position (with Ø, so still in keeping with the table). Without the relative clause, there is a communications breakdown:

17.b. #Joleen came back with a bottle of Spanish wine8. A cigarette was again snuffed after a few intense puffs.

60The reader does not understand whose cigarette it is and consequently why the cigarette has suddenly been introduced into the situation.

  • (c) If the contents of the relative clause are not verifiable at the time of the “event” referred to in the relative, then the utterer will choose a (-P) operator.

61This concerns utterances with world-creating verbs, modals, relatives corresponding to metaphors, hypothetical utterances. Here the choice of (-P) is explained by the fact that the referent is not known to exist in the extramental extralinguistic9 world; so its identification is unproblematic in that the utterer does not expect the addressee to retrieve a reference considered to be irretrievable. The pragmatic message is “don’t bother to look for it, I have no reason to believe it exists.”

62The following represent a few examples:

  • (i) World creating verbs:

18. “ What we need is a yellow pages, a directory that tells us the job of each gene.” (Newsweek, 27/1/97)10

63The utterer cannot assert the truth value of the contents of the relative clause (a directory tells us the job of each gene). Indeed, as we know, such a directory does not yet exist.

19. “Funnily enough,” said Milly, “he’s looking for a book-keeper, “and he’s looking for someone Ø he can trust, with a recommendation.” (M. Spark, A Far Cry from Kensington).

64Here, Milly cannot assert the truth value of “he can trust someone”.

  • (ii) Modals, metaphors and hypothetical statements:

20. And there was a third party present, inhibiting the good row Ø they might have had. (H. Mantel, A Place of Greater Safety, p. 186)

65They had a good row is not verifiable: they did not have one.

21. And he was afraid of his mother, too. How she would wrap him if he came out, in the big yellow towel like egg yolk, how she would want him to get close to her smooth, sticky body, like a mouth that would swallow him. (G. Swift, Learning to Swim, p. 16)

66It is not only because of the modal here (would) that the contents of the relative are not verifiable. This is also a hypothetical statement (if he came out of the sea, a mouth would swallow him). Furthermore, we have a metaphor here (Her body was like...). So we have three different phenomena explaining the choice of (-P) indicating that the reference is unproblematic. It does not have to be looked for because it does not exist.

67It should be noted that most of the utterances looked at in this category have indefinite heads (hence correspond to new information) but that despite this, the choice is that or Ø, depending on the syntactic site. So this type of utterance helps to explain the figures in Table 1 and helps one to understand why, even if traditional analyses are right part of the time, factors other than backward reference also explain the choice of that and Ø.

68(2) P (the identifying of the referent is considered to be problematic)

69Here, as can be seen in the table, we have wh- entering the picture and Ø dropping out (this last operator referring systematically to unproblematic references).

70So, in this category, we have wh- in subject position and that in complement position. This means to say that that has a floating value: in subject position (as seen for the (-P) category), that indicates an unproblematic reference. But in complement position, that indicates that the reference is problematic.

71Let us consider a few examples of the types of utterance in this category, bearing in mind again that space constraints make it impossible to mention them all.

  • (a) Backward reference with antecedent competition (i.e. there is more than one possible referent for the antecedent).

72In the previous category, (-P), backward reference tended to confirm traditional theories (that/Ø), but here, these theories are only confirmed in complement position (with that). In subject position, with wh-, they are not.

22. While John Hawkins, Francis Drake and their companions were absent on the business which ended so disastrously at San Juan de Elua, events in Europe had taken a decisive turn for the worse. (E. Thomson, Sir Francis Drake).

73Here, the reader knows that Hawkins and Drake have been engaged in any number of activities. These have been described in detail. But which one is going to be selected? A host of them compete for the role of the antecedent’s referent. The choice of which in subject position indicates the problematic nature of the identification. In other words, the identifying of the referent is not plain sailing. Here we have a case of a reference referring back to one of a number of possible earlier occurrences.

23. Okay, I’m cutting the cards – that’s to say I’m taking off about half the pack – so I’m leaving you, the magician, with the bottom half and I’ve got the top half. [...]. Okay. I’m putting it back [the chosen card] on the top of the half that I’m holding. (S. Axbey, Soundtracks. Real Life Speaking).

74Here we have two competitors for the possible role of referent, the half held by the demonstrator and the half held by the apprentice magician, (who is also the presenter of the radio programme). Both halves, as we can see, have been referred to just before. Here again, the identifying of the referent is not plain sailing and the choice of that in complement position serves as a warning to the addressees (in this case, not only the apprentice magician but also those tuned into the radio programme, who are not able to see what’s being done).

75These last 2 utterances provide examples of what we could call contrastive reference.

  • b) Less accessible generic concepts

76In the (-P) category I touched on generic concepts considered to be easily accessible (ex 15), basically because they have been reactivated in the previous context.

77Let us now take a look at an utterance where the concept has not been reactivated:

24. But a great many people fell in with Hector’s pretensions, a surprising number, especially those simple souls who quell their doubts because they cannot bring themselves to discern a blatant pose. (M. Spark, A Far Cry from Kensington)

78I include who here to show that the wh- is indeed significant in this example; although, as we have seen, it is not systematically so because of the (+Human) factor. Note that the wh- pronoun is in subject position, in keeping with the table.

79In the above utterance, the concept referred to has not been presented, nor even hinted at in the previous context. It is indeed a hermetic concept. But with the choice of the demonstrative adjective those, the narrator indicates that she considers the reader to be in the know about people of the type mentioned.

  • c) The phenomenon of switch of subject

80T. Givon points out that subject continuity is the norm and that when this is not respected, the grammar provides the means to mark this11. The examples he himself gives are with personal pronouns (stressed vs. unstressed). But in the domain of the relative, I have found that it is the choice of the relative pronoun that marks this. Where the choice would normally be (-P), you will find (P). (And if the choice would normally be (P), then you will find a (+P) marker).

81The notion of a problematic identification is explained by the fact that the addressee is mentally prepared for an unproblematic reference: the same as the referent of the previous grammatical subject. The switch of subject is momentarily confusing and calls for a reshuffling process.

82For this type of phenomenon, the relative pronoun is always in complement position: it is not the identifying of the relative pronoun’s referent that poses a problem; the relative pronoun merely carries the warning concerning the problems to do with its “sister” argument, the subject:

25. Libyan officials have also written directly to Lockerbie victims. Libya, the letters assert, has tried to resolve the issue by offering to have the two alleged culprits tried in a neutral country. They accuse America and Britain of resisting this option solely to isolate a regime that they dislike. (The Economist, 25/10/97, p. 58).

83This is a particularly problematic example: the subject they in the relative clause is not the same as the preceding they, subject of the global utterance. It is not the Libyan officials who dislike the regime, it is America and Britain. Note that had it not been for the phenomenon of switch of subject, (-P) would have been indicated. Shared knowledge is one of the parameters calling for (-P) and the reader knows that the terrorist act destroying the PanAm plane over Lockerbie sparked off hostilities between America and Britain on the one hand and Libya on the other.

84A final example in this (P) category has to do with syntactic problems involving the identifying of the co-referent i.e. the antecedent. In the Noun Phrase corresponding to the antecedent there is more than one noun; and the choice of a (P) marker serves as a warning12: “It’s not only the noun just to the left of the relative pronoun that is co-referential with it. Go back further.”

26. No contract or transaction [...] shall be void or voidable [...] solely because the director or officer is present at or participates in the meeting of the Board of Directors or committee thereof which authorizes the contract or transaction. (legal contract)

85Here we have four nouns in the Noun Phrase, all of which are included in the co-reference. It is not just the noun committee that is co-referential.

27. It has long been the aim of this government to dismantle the system of protection for historic buildings that we enjoy in Britain. (Private Eye.)

86It is not the historic buildings that we enjoy, nor their protection. It is the system of protection for them.

87(3) +P (the identifying of the referent is considered to be particularly problematic)

88As can be seen in the table, in this section we have only wh-. Note that in subject position wh- indicates both problematic AND particularly problematic references: language has its limits.

  • a) The case of the antecedent presented for the first time which has thematic importance

89We are talking here about indefinite heads (or cataphoric determination) corresponding to the current theme or the upcoming theme. This is a case that is compatible with traditional theories (new information + which)

90Here we shall look at one in complement position:

28. There was a period between the ages of about thirteen and seventeen which he remembered as the happiest in his life. [...]. He would get up every morning at six and train for two hours in the baths, and again before lunch [...]. Once in the cool water of the baths, his arms reaching, his eyes fixed on the blue marker line on the bottom, his ears full so that he could hear nothing around him, he would feel quite by himself, quite sufficient. (G. Swift, Learning to Swim, pp. 14-15).

91Here the happy period in question is introduced for the first time (indefinite head) and represents the new theme, which is developed over a paragraph.

  • b) The case of the already identified antecedent which is reanalysed to give it a new identity in keeping with the current theme

92Here a definite head is retrograded by giving it an indefinite head. This strategy enables the utterer to focus on one aspect of its identity considered to be significant. In this way, it acquires a new identity. This explains the +P choice: the new referent is particularly problematic in that it is considered to be not yet retrievable. Unlike some of the examples looked at, this type of discourse phenomenon does not appear to be compatible with traditional analyses: we have which with an already presented item:

29.a. Sooner or later he [Mr Blair ] will not just have to talk about “tough choices”, he will have to begin to make some [...]. At that point, which is likely to arrive within no more that a year or three, the battle could become far more equal. Mr Blair has aroused great expectations, and has rested them on his own credibility. That credibility will soon be vulnerable to attack by a party which has restored its own reputation for competence, as well as soundness of motive and generosity of spirit. The Conservative’s task is to be that party. (The Economist, 4/11/97, p. 20)

93Here, the Conservative party has already been presented: the whole article is about this party and its potential for change and for becoming a threat to the Labour Party. So The Conservative Party becomes “A party + a virtual property” enabling it to become a threat in “one or three years’ time.

94The change of identity becomes apparent if we perform a manipulation restoring the definite head and deleting the relative:

29.b. (?) That credibility will soon be vulnerable to attack by the Conservative Party. The Conservative’s task is to be that party.

95What we are saying here is that the Conservatives’ task is to be the Conservative Party, which is not the intended message.

  • c) Finally, we have the situation where the identifying of the referent presents not one problem (where you would get P), but two, hence +P

30. The revelations of Mr Ecclestone’s dealings with Labour have been forced out of the government amidst a welter of denials, threats of legal action and back-pedalling. On the face of it, the connections are far more damning than the allegations of Conservative Party sleaze which Labour had so much fun with during the last election. (The Economist, 15/11/97, p. 17).

96The first problem is “switch of subject”. The previous subject, corresponding to the theme, is the connections between Mr Ecclestone and the Labour Party. But the subject of the relative clause is the Labour Party (Labour). The second problem is a syntactic one. In the chain, the identifying of the antecedent is not straightforward. The co-referent of the relative pronoun is not sleaze nor Conservative Party but allegations of Conservative Party sleaze.

97In this discussion, it is not possible to go into the placing of the nuclear accent but before closing I should just like to very briefly touch on my hypothesis about this, which the following table enables one to visualise.

Table 4. - Placing of the nuclear accent according to the degree of reference accessibility

+ A: Nuclear accent situated within the relative (large majority of cases).

- A: Nuclear accent situated outside the relative.




- A / +A



98My hypothesis is that the placing of the nuclear accent corresponds to a further degree of fine-tuning in the accessibility scale. If the accent is outside the relative clause (and this is virtually restricted to the –P category), then the retrieval of the reference is considered to be straightforward: example (16) provided an illustration of this (Dear God, I must warn people about her, he thought. He made a mental LIST of the people Ø he must warn).

99Here, as we saw, the construction to which the contents of the relative refer back is to be found in the previous sentence (I must warn people about her).

100At the other end of the spectrum, you have, for example, the antecedent presented for the first time, which has thematic importance. As the contents of the relative correspond to new information, the nuclear accent will obviously be within the clause. Example (28) illustrates this: (There was a period between the ages of about thirteen and seventeen which he remembered as the HAPpiest in his life).

101The points made and the examples discussed in this third and last section suggest the following conclusions:

102The choice of the relative pronoun indicates degrees of reference accessibility, these being fine-tuned by nuclear accent positioning. M. Ariel, mentioned at the beginning of this section, refers to three categories of reference accessibility markers:

  • High accessibility markers

  • Mid accessibility markers

  • Low accessibility markers,

103the relative clause belonging to the category of low accessibility markers.

104This latter fact is derivable from the analyses of T. Givon: the less easy it is to retrieve a reference, the more information there has to be in the form of lexical content. M. Ariel goes on to point out that within each category there is a further fine-tuning of accessibility markers. Although her comments and examples do not refer to the relative clause or to supra-segmental features, nevertheless this section of my discussion suggests that her analysis is equally applicable to these areas.

Notes de bas de page numériques

1 I postulate that the continuative relative clause is characterised by the fact that it refers to an event whose time reference is posterior to that of the matrix clause. It can therefore not contribute towards the identification of the referent at the time of speaking.
2 Tone Units: “Chunks of utterances all of which have pitch patterns which accord with acceptable ‘whole’ intonation patterns: A. CRUTTENDEN, 1986, pp. 36-42.
3 Propositional meaning vs. thematic meaning: see R. HUDDLESTON 1984.
4 With the exception of pushdown relative clauses (N. Malan, 1999, p. 3).
5 R. Huddleston, 1985.
6 I am indebted to F. Benoit for having pointed this out to me in the first place. In N. Malan 1998 and 1999, this debt of gratitude was inadvertently overlooked. It is hoped that mention of it here will help make amends for the omission.
7 For an exhaustive presentation, see N. Malan 1998 or 1999.
8 The sign (#) indicates that the utterance is apragmatic.
9 Term proposed by F. Benoit.
10 In the Americcan media, there is virtually no opposition between "which" and "that", the latter marker having almost exclusively ousted the former. But in the case of the utterance in question, "that" would also have been the indicated choice in standard British English.
11 T. GIVON, 1993.
12 Note the different types of warning for noun phrases of this nature, depending on whether the head is indefinite [see comment re (2)] or definite as in the utterances under consideration in this section.

Notes de bas de page astérisques

* This paper is based on my thesis defended in 1998 at the Sorbonne-Nouvelle Paris III University under the supervision of Claude Delmas; and on its abridged version in book form. (See bibliography).


ALLEN, R. L., The Verb System of Present Day American English, The Hague, Paris : Mouton, 1966, p . 156.

ARIEL,M., “Referring and Accessibility”, Journal of Linguistics 24 (1), March 1988, pp. 65 – 82.

BOUSCAREN, J., Linguistique anglaise. Initiation à une grammaire de l’énonciation. éd. Rev., Paris : Ophrys, 1992, p. 113.

EMONDS, J., “Appositive Relatives have no Properties”, Linguistic Inquiry (10), 1979, pp. 211-243.

FABB, N., “The Difference between English Restrictive and Nonrestrictive Relative Clauses”, Journal of Linguistics (26), 1990, pp. 57-78.

FUCHS, C., MILNER., J., A propos des relatives. Etudes empirique des faits franÿais, anglais et allemands, et tentative d’interprétation. En collaboration avec LE GOFFIC, P., Paris : Sociéte d’études linguistiques et anthropologiques de France, 1979; pp. 126-133.

GIVON, T., English Grammar. A function based introduction, Amsterdam/Philadelphia: John Betjamins, 1993.

HUDDLESTON, R., Introduction to the Grammar of English. Reprint [1st ed. 1984]. Cambridge: CUP, 1985.

MALAN, N., La proposition relative en anglais contemporain : une approche pragmatique, Gap, Paris : Ophrys, 1999, 139p.

MALAN, N., Sémantique, syntaxe et pragmatique de la relative en anglais contemporain, Lille : Atelier National de Reproduction des Thèses (microfiche), 1998, 461p.

McCAWLEY, J., “Parentheticals and Discontinuous Constituents”, Linguistic Inquiry (13), 1982, pp. 91-106.

SAUSSURE, F. de, Cours de linguistique générale, 3e éd., Préf. De Ch. Bally et Alb. Sèchehaye. Paris : Payot 1971.

THOMPSON, S.A. “The Deep Structure of Relative Clauses,” pp. 79-98, in: FILLMORE, C., LANGENDOEN, D. (eds). Studies in Linguistics and Semantics. New York: Holt, Rinehart and Winston, 1971.

WHORF, B. LEE. Language Thought and Reality. Reprint. (1st ed. 1956). Massachusetts: MIT Press, 1993. 278 p.

Pour citer cet article

Naomi Malan, « The relative clause: An antecedent with a will of its own and a pronoun Ø one cannot ignore », paru dans Cycnos, Volume 17 n° Spécial, mis en ligne le 22 septembre 2008, URL :


Naomi Malan

Université de Paris III