Article body

1. Subtypes of deletion/silence

Deletion/silence seems to come in (at least) three guises. The most studied is what can be thought of as ‘deletion under identity’,[1] that is, deletion that depends on the presence of an antecedent. Some familiar examples are:

In all of these, any lexical item can be contained in the deleted phrase, as long as the requisite identity condition is met with respect to the antecedent.

A second type of deletion, well-known but less systematically studied, is one in which the notion of antecedent relevant in (1) is not relevant at all; typically, very specific lexical items are at issue. The following examples of such antecedentless deletion are mostly taken from Kayne (2005). In each example, the italicized part is what is visible/audible in the language in question;[2] the non-italicized capitalized word (or words) indicates what has arguably been deleted.

2. Toward characterizing deletion/silence that is not antecedent-based

The challenge is to find a maximally unified theory for deletions of the sort seen in (2) –(4)—and then to ask to what extent the deletions of (2)–(4) can be unified with those of (1). Van Riemsdijk (2003) clearly sets out the task for specified (antecedentless) deletion in these terms:

...a set of more general questions about specified deletion (or specified ellipsis, or specified zero formatives) whose answers will ultimately constitute a theory of such elements...What is the range of elements that can be represented by empty formatives?...What are the ingredients of Licensing Theory? In particular, what counts as a potential licenser and what are the conditions on the syntactic proximity between the licenser and the licensee?...

Van Riemsdijk 2003: 260–261

Although van Riemsdijk (2003: 261) takes his silent GO (in Swiss German; cf. in part earlier English We must away) to be an instance of a non-functional element, let me take light verbs (and light nouns) to be functional elements, in which case we can consider the following restriction as a first approximation:[11]

The restriction stated in (5) may seem odd for inning in (2c) above, since inning is, in American English, essentially limited to baseball contexts. However, Jean-Yves Pollock (personal communication) has made an arguably relevant suggestion involving silent PART for the case in (6a), where the idea would be that we have (6b).

If so, then two home runs in the seventh in (2c) might be as follows in (7), with INNING a modifier of PART, with PART directly satisfying (5), and with INNING itself perhaps counting as functional in a baseball context.

Silent INNING is of additional interest when it comes to the question of characterizing licensing conditions for deleted/silent elements (of the antecedentless type). Although the following contrast in (8)–(9) might suggest a right-branch versus left-branch effect, note (10).

The contrast between (9) and (10) suggests, instead, that what is at issue is whether or not there is a definite article present that immediately precedes seventh, as there is in (10) (and in (8)), but not in (9). If so, there may be a link to the Italian–French contrast seen in:

by virtue of which Italian, in the context of the definite article le allows silent ORE:

In contrast, French does not allow silent HEURES in (12), which contains no definite article.

As for the question why the presence of a definite article would make a difference in these cases, it might be that the definite article reflects the presence of an additional phase into the Spec of which inning and ore can move and become silent in the manner of Kayne (2006).

3. Ever silent elements

A third type of deletion/silence involves elements that, unlike those previously discussed, are associated with no phonology elsewhere, that is, elements which are never otherwise pronounced. There are two subcases. In the first, the silent element in question is not pronounced anywhere in the language in question, though it may be pronounced in other languages. In the second subcase, the silent element in question is not pronounced in any language at all.

The first subcase is exemplified by the Topic head that is pronounced in Gungbe but not ever pronounced, though present, in Italian (or English), along the lines of Rizzi (1997).[13] The second subcase, perhaps illustrated by little n, makes one wonder if the element in question is well-founded (unless a principled reason can be discovered as to why it remains unpronounced in all languages).

4. Inalienable possession with a definite article

Returning to what is central to this paper, namely to antecedentless deletions/silent elements, let us consider sentences involving inalienable possession, in particular those with a definite article apparently in place of a possessive pronoun, as in:

(There are a wider range of possibilities for such sentences in French and Italian, probably for reasons [having to do with datives and with reflexive clitics] that are at least partly independent of the definite article question.)

In examples such as (14), it is understood, even though there is no visible possessive pronoun, that the ankle in question is John’s, i.e. is a part of John’s body. It is therefore natural to think in terms of possessor raising or possessor deletion.[14] But that by itself is not sufficient to account for the following contrast:[15]

When the definite article the is present, a non-restrictive relative is allowed that has a kind of generic interpretation. This is much less, if at all, possible in the presence of his. The generic interpretation of the relative in (15) is evidently, at first glance surprisingly, compatible with our understanding the ankle in question to be one of John’s.

My proposal will adopt a familiar sort of idea, namely that (15) contains a deleted/silent his.[16] Yet I will crucially take this silent his not to be part of the head of the non-restrictive. This is shown for (15) in the following (with capitals again indicating silence/deletion):

In (17), the ‘head’ of the non-restrictive relative is the phrase the ankle.[17] There is a silent/deleted HIS present in (17), but that HIS is the possessor not of pronounced ankle, but rather of silent TOKEN (or INSTANCE or COPY, with a possible link to classifiers).[18]

The analysis in (17) provides an account of certain facts having to do with number. Consider the following contrast:

With the definite article in (18), hands is possible but not heads (unless one allows for two-headed beings). Whereas in (19), with possessive their, plural heads is fine (even with one head per person). Given (17), the contrast in (18) reduces to the contrast seen in the following (cf. Vergnaud and Zubizarreta, 1992: 619):

This is so as follows. In (18) with hands, we have:

This is well-formed, and could be extended with a non-restrictive relative of the sort seen above in (15), yielding the sentence:

which is in turn associated with:

However, the version of (18) with heads would correspond to the ill-formed:

whose ill-formedness rests on that of:

(As before, strictly speaking, (25) is (irrelevantly) well-formed, but requires that human bodies have more than one head.)

Possible, on the other hand, in contrast to (18), is:

with singular head, and with the analysis:

The analysis suggested in (17) also provides an account of the restriction concerning adjectives seen in:

With the definite article, a non-restrictive adjective is not possible,[20] in a way that now reduces to the impossibility of such non-restrictive adjectives in generic sentences of the sort seen in:

On the other hand, restrictive adjectives like left, right, upper, lower are possible with the, as in:

This can now be seen to be tied to:

The adjective left in (31) need not be stressed. In this respect, (31) contrasts with:

where to my ear broken must be stressed, suggesting that the use of the in (33) is not quite the same as in (31) or (26), as it must not be, given:

Rather there must be, in a way that remains to be spelled out, a link to:

and/or to (in a hospital context):

with (33) perhaps not involving TOKEN.

5. Ringo

The proposal of the previous section, exemplified in (17), repeated here in (37), elucidates the apparent ambiguity concerning ankle found in (38).

Rather than thinking that ankle in such sentences is ambiguous in the sense of simultaneously being generic and specific, we can now, following (17) and (37), take ankle itself in such sentences to be neither generic nor specific. The generic facet of such sentences depends on the the that is present preceding ankle, and the specific facet on the silent HIS TOKEN. In effect, if (17) and (37) is on the right track, the apparent ambiguity in question is to be attributed to different parts of the associated syntactic structure.

It may be that a similar approach is called for in sentences of the sort studied by Jackendoff (1992), for example:

in which himself is understood as a statue of Ringo. Jackendoff notes the interesting contrast with the reverse case illustrated in:

which is impossible if Ringo is the statue and himself the person. Sense can arguably be made of this contrast, if we take (39) to contain a silent STATUE/COPY, as in:

From this perspective, (40) would be possible only if the following were possible:

The key difference is that in (39) and (41) Ringo c-commands himself, whereas in (40) and (42) Ringo does not c-command himself. Thus himself in (39) meets the demands of Condition A of Binding Theory, whereas himself in (40), despite appearances, does not.

6. Book

The initially paradoxical character of (38), in which ankle seems simultaneously generic and specific, has a partial counterpart in sentences like (43), in which book appears to simultaneously be both concrete (in weighing more than two pounds) and abstract (in taking almost a year to write).[22]

In the spirit of the analysis of (38) suggested in (37), we can take (43) to be analyzable as:

In (43) and (44), then, the non-restrictive relative has as its ‘head’ the phrase Mary’s latest book. At the same time, the VP headed by weighs has as its subject the larger, distinct phrase A TOKEN/COPY OF Mary’s latest book. The content facet of book in (43) is associated with the phrase Mary’s latest book; the physical object facet, on the other hand, is associated with the distinct phrase A TOKEN/COPY OF Mary’s latest book. If this proposal is on the right track, then there is no need to attribute ambiguity of the content versus physical object type to the noun book itself.

In effect, as in all the antecedentless deletion examples of (2)–(4) above, the presence of silent elements (whose cross-linguistic licensing conditions, needless to say, need to be looked into in much more detail) simplifies the interpretation, in the sense that the interpretive component can now read, in a direct way, more off the syntax than it could have in the absence of those silent elements.