2.4 Anaphors
2.4.5 Null complement anaphora
Though they don’t contain an explicit anaphor of any sort, sentences with null com- plements can also be understood as anaphoric, referring to events, as in (80), or to propo- sitions, as in (81).
(80) a. I asked Bill to leave, but he refused.
b. Sue was attempting to kiss a gorilla, and Harry didn’t approve.
c. We needed somebody to carry the oats down to the bin, but nobody volun-
teered. (Hankamer & Sag 1976: (56))
27Some speakers find even (79a) to be marked, but the pattern in (79) is fairly robust. Explaining such
speakers’ objections to (79a) will remain a topic for future work. 37
(81) a. Paul cheated on the test, and his teacher knows. b. Paul cheated on the test, and his teacher found out.
In both of the examples in (81), we interpret the null complement of know/find out as referring to the proposition introduced by the first clause: ‘Paul cheated on the test’.
Null complement anaphora (NCA) can be interpreted as having an event or proposi- tional referent, but never an individual referent, as noted in Shopen 1972, 1973; Grimshaw 1979; Saeboe 1996; Depiante 2000; Haynie 2009. Depiante 2000 gives the following illus- tration:
(82) a. The children know [N P the song ]
b. The children know [CP that it is time to leave ] (Depiante 2000: 60: (7))
(83) a. The teacher told the children that it was time to leave even though they already knew
b. * The children learned the song on Monday but by Friday they know longer
knew (Depiante 2000: 60: (8))
Even though know can take either a nominal or clausal complement, as shown in (82), only the clausal—propositional—version is compatible with NCA, while the nominal— individual—version is not, as in (83).
Combined with a Kratzer-Moulton view of attitude verbs, this non-individual restric- tion might explain why NCA seems to be incompatible with many verbs which otherwise are thought to take propositional complements. For example, compare (81b) to (81c).
(81) b. Paul cheated on the test, and his teacher found out. c. * Paul cheated on the dest, and his teacher discovered.
If a verb like discover can only take (propositional-content-bearing-)individuals as argu- ments, and NCA disallows individual-denoting antecedents, then we can explain the un- grammaticality of (81c). If find out, on the other hand, takes propositions—propositions, and not individuals bearing propositional content—as arguments, then the grammati- cality of (81b) is not problematic. This might also be an answer to the puzzle posed by Williams 2012 regarding the behavior of predicates like notice—which like find out can take truly propositional arguments, and thus participate in NCA28—in contrast to predi-
cates like win and ready, which take individual-type arguments—like discover—, and thus don’t participate in NCA.29 If correct, this raises a question about what else might dis-
tinguish notice/find out verbs from discover/believe verbs, which might explain, or at least confirm, this argument-type distinction. One possible answer might lie along the lines of the semantic characteristics discussed in White, Hacquard & Lidz in press: both types of verbs are representational and factive, but only the latter are assertive.
Hankamer & Sag 1976 discusses NCA in distinguishing two types of anaphora, sur- face anaphora and deep anaphora. On this account, surface anaphors are derived trans- formationally, and so have internal syntactic structure—and with it, certain syntactic requirements—, while deep anaphors show “no sign of having been syntactically com- plex at any stage” (i.e., have no internal syntax) (Hankamer & Sag 1976: 406). As a result, surface anaphors require a linguistic antecedent, while deep anaphors can be controlled pragmatically. Hankamer & Sag 1976 argues that NCA is an example of deep anaphora, as (following Bresnan 1971) is “sentential it”, i.e., the propositional anaphor it discussed above. In contrast, Hankamer & Sag 1976 argues that so is a surface anaphor.
Because NCA involves an ‘invisible’ component, I want to be careful to distinguish it
28Williams 2012 (knowingly) “stretch[es] the term somewhat”, using NCA more broadly than Hankamer
& Sag 1976 use it, or than I am using it here. For Williams 2012, NCA includes implicit arguments of verbs, as in Ron won. I will not argue that such a “stretch” is problematic, nor the opposite. That said, the difference is worth being aware of. Future comparison of these perhaps-distinct classes, e.g., with the extraction facts noted in Depiante 2000, would be worthwhile.
29They don’t fit under the Hankamer & Sag 1976 definition of NCA, that is; see previous footnote.
from other constructions which might appear similar. As Hankamer & Sag 1976 show, NCA is distinct from intransitive uses of verbs (where there is no object, null or other- wise, and thus no anaphora). The intransitive use of eat in (84), for example, tends be interpreted as having “general (unspecified))” object: i.e., it’s not that he just won’t eat soup and potatoes, but rather, he won’t eat anything.30
(84) I bring him soup and potatoes, but he won’t eat.
(Hankamer & Sag 1976: fn. 21: (i))
In contrast, the null complement of (85) is interpreted such that the speaker’s wife doesn’t approve specifically of the speaker’s gambling, not that she doesn’t approve of anything in general.
(85) I play cards and shoot dice, but my wife doesn’t approve.
(Hankamer & Sag 1976: fn.21: (ii))
NCA is also distinct from the null objects—which, unlike NCA, can be individual-typed— in a language like Spanish (which allows for implicit objects much more widely than English), as in (86).
(86) A: Quer´es caf´e?
Do you want coffee? B: S´ı, s´ı quiero
Yes, yes I want (Depiante 2000: 61: (10))
Depiante 2000 demonstrates that Spanish null objects are subject to island restrictions in a way that NCA is not. Finally, NCA is distinct from Verb Phrase Ellipsis (VPE), which
Hankamer & Sag 1976 classifies as surface anaphora on the basis of arguments by Ross 1969 and Grinder & Postal 1971.