This Construction Looks Like a Copy is Optional

Potsdam and Runner (2001), following Rogers (1974), distinguish between two copy-raising constructions in English. In genuine copy raising (GCR), a matrix DP is linked to a "copy" pronoun in the highest embedded subject position. In apparent copy raising (ACR), the associated pronominal copy is not the highest embedded subject.

(1)

1. a. GCR

   Richard seems/looks/appears like/as if he is in trouble.

2. b. ACR

   Mary seems/looks/appears like/as if her job is going well.

In certain English dialects, GCR exhibits the hallmarks of standard (noncopy) raising, while ACR does not. For example, only the former allows nonthematic arguments.

(2)

1. a. %There looks like there's gonna be a riot.

2. b. *There seems like John expects there to be an election.

Potsdam and Runner propose that GCR involves a nonmovement A-chain, subject to standard locality conditions, while ACR involves simple coindexation between two thematic positions.

Discussing the proleptic object (PO) construction in Madurese and English, in which an embedded pronominal copy is linked to a matrix DP, Davies (2005) observes that despite their superficial similarities, GCR and PO constructions cannot be conflated. Thus, the pronominal copy in the PO construction need not be a subject and the matrix DP must be thematic.

(3)

1. a. Sheryl thought about/of Tim that the police would never catch him.

2. b. Kelsey believed about the cat that it would be out of the bag. (literal)

On the basis of these and related facts, Davies concludes that PO constructions must be kept typologically distinct from GCR constructions. Nothing in Davies's study, however, argues against conflating PO and ACR constructions. This squib presents an argument that the two should nonetheless be kept distinct.

Heycock (1994) observes that ACR is possible even in the absence of an overt pronominal copy (contra Lappin 1984), although "it is of course clear that the matrix subject is interpreted as binding some 'understood' position in the complement" (p. 292). However, even the latter condition is not necessary. Consider a situation where John is watching TV in the other room and all of a sudden rushes in, pale and shaking. We can then utter (4a), with the intended reading (4b). Alternatively, we can use (4a) to mean (4c).

(4)

1. a. John looked like something terrible had happened.

2. b. John looked like he just learned that something terrible had happened.

3. c. John looked like something terrible had happened to him.

In fact, with enough contextual information, we may relate John to the embedded event in any imaginable way. This is even clearer in (5), which could be uttered after hearing a long motivational speech by Mary.

(5) Mary sounded like there's nothing hard work and good faith can't solve.

Given that there is no identifiable syntactic "empty slot" in the complement that is saturated by the matrix subject, we may conclude (with Asudeh and Toivonen (2007)) that the complement does not denote a semantic predicate, contra Lappin 1984 and Heycock 1994. Rather, it denotes a closed proposition, which, in turn, must be construable as capable of generating the perception event reported in the matrix clause (hence the oddness of, say, John looked like Alexander had defeated the Persians). Importantly, this is a pragmatic condition, susceptible to world knowledge.

In contrast, the PO construction is not similarly liberal; a pronominal copy must be present in the complement. Compare (6a–b) with (4a) and (5).

(6)

1. a. We thought about John that something terrible had happened *(to him).

2. b. *We believed about Mary that there's nothing hard work and good faith can't solve.

A natural way to understand this restriction is to view the PO predicate as a three-place relation between a property (the complement) and two individuals (the matrix subject and the PO), type <<e,<s,t>>,<e,<e,<s,t>>>>. The property denoted by the complement is predicated of the individual denoted by the PO (in the subject's belief worlds).

Suppose that the clausal complement is turned into a predicate by merging it with a null operator (a λ-abstractor) that binds the pronominal copy. Then (6b) and the copyless...