Review of "On Realizing External Arguments" by Koring et. al. (to appear, Linguistic Inquiry)

Review of “On Realizing External Arguments: A syntactic and implicature theory of the disjointness effect for passives in adult and child grammar” by Loes Koring, Eric Reuland, Nina Sangers and Ken Wexler. (to appear, Linguistic Inquiry)

(https://direct.mit.edu/ling/article-abstract/doi/10.1162/ling_a_00520/117701/On-Realizing-External-Arguments-A-Syntactic-and?redirectedFrom=fulltext)

This article (henceforth OREA) argues that the implicit external argument of the short passive is syntactically projected. That conclusion converges very nicely with a similar conclusion in Collins 2005, and Collins 2023 (MIT Press, forthcoming). While OREA’s argument is based on a disjointness implicature, Collins 2005, 2023 is based on principles A and B of the binding theory, the distribution of Helke expressions (e.g., ‘on my own’) and the distribution of secondary predicates. It is striking that two completely different sets of data yield the same theoretical conclusion, namely that the implicit argument in the short passive is syntactically projected. Such converging results (based on different data and different methodologies) should give one confidence in the theoretical conclusion.

In this blog post, I will outline the argument in OREA, and point out an issue that is unclear to me. I will also offer an alternative way to understand their child language data.

Consider the following pair of the sentences. 

(1)

a. John was seen.

b. John was seen by somebody other than John.

For adult speakers it is generally the case that whenever (1a) is true, so is (1b). But (1b) seems like an implicature instead of an entailment, since it can be canceled:

(2)

John was seen, namely by John himself looking through a security camera.

OREA derives the implicature in (1b) using the syntactic theory of implicatures in Fox and Katzir (2011). As I understand them, he basic assumptions are as follows: 

(3)

a. In the short passive, there is a syntactically present EA (‘external argument’)

b. The implicit EA is existentially bound (pg. 47).

c. The EA generates implicatures.

d. The implicatures are negated stronger alternatives.

e. The strong alternatives are formed by replacing the EA by contextually salient constituents.

Now consider (1a) to see how this works. By (3a,b), the representation of (1a) is something like the following (see Collins 2005, 2023 for a similar assumption):

(4)

John was seen EA.

By (3c,d,e), the following alternative is negated, assuming ‘John’ is contextually salient:

(5)

John was seen by John.

(4) and (5) together entail (1b), which is the disjointness implicature.

My concern with this argument is the following. By the assumptions in (3), any replacement of EA by a contextually salient proper name should be negated, since replacement by a proper name will always yield a stronger alternative. But (5a) entails (5b):

(6)

a. John wasn’t seen by John, Bill, Mary….

b. John wasn’t seen by anybody.

But (6b) contradicts (4). So (6b) is a contradictory implicature. 

One could argue that in (4) only ‘John’ is contextually salient, and so only ‘John’ can be substituted for the implicit EA. But the authors clearly state that the children are familiarized with all the characters in the experiment at the beginning of the session (pg. 17): “Each experimental session started with a warm-up phase in which the child was familiarized with all the pictures, the four characters, and the verbs used.” So all the relevant characters should have been contextually salient, and available for substitution. 

To patch this up, some more specific notion of contextual salience is needed. Perhaps this issue was discussed in the Fox and Katzir (2011), which I have not yet consulted. But it seems like a big enough issue that the authors should have addressed it in the paper itself.

This discussion brings up a different possibility for understanding the lack of the disjointness implicature in child language. OREA notes that in sentences like (7) for child speakers there is no disjointness implicature:

(7) Bart was washed.

In other words, children are happy to use sentences like (7) even if Bart washes himself. For this OREA concludes that passives in child language lack a syntactically projected EA. Lacking such and EA, the disjointness implicature in (3) does not go through (since the calculation of the disjointness implicature requires a syntactically projected implicit EA, see 3e above).

OREA argues that the reason why the child passive lacks a syntactically projected EA is that the child passive is actually an adjectival passive, which always lack a syntactically projected EA, even in adult language. As they grow older, they acquire the distinction between the two kinds of passive.

An alternative is that children do in fact have a syntactically project EA in (7), but it is the following, where the implicit EA is a null pro.

(8) Bart1 was washed pro1.

On this theory, the difference between children and adults is that children are more forgiving of principle B violations, and so allow representations such as (8) (on such a difference, since footnote 11 of OREA).

In fact, Collins 2023 shows that the implicit EA in the short passive has three distinct syntactic realizations: (a) generic implicit argument (progen), (b) existential implicit argument (proun), (c) definite implicit argument (prodef). The implicit argument in (8) would fall under case (c): definite implicit argument.

OREA considers the possibility that the disjointness effect in the passive is caused by a principle B effect, but they reject the possibility in the following terms (pg. 6):

(9)

“We will now consider the disjoint reference effect in verbal passives in more detail. Is it the result of a syntactic violation, such as condition B? The answer is negative since it can be cancelled,…”. 

They add the following in footnote 11:

(10)

“In order for children to allow such a representation would then mean that, for children, as opposed to adults, passivization does not involve existential quantification.”

The problem with (9) and (10) is that they fail to recognize that there are actually three different types of implicit arguments involved in the short passive in English. When the implicit argument is definite, it triggers a principle B effect, as noted in Collins 2023 (chapter 4). When the implicit argument is existential, it triggers a disjointness implicature, as noted by the authors. Both explanations are needed to account for the data.

The advantage of my proposed principle B alternative is that it would not require such a drastic difference between the way children and adults represent passives. Rather, principles of UG (such as the Theta-Criterion) would apply at the beginning of acquisition and ensure the child acquires passive syntax (corresponding to the adult passive syntax) immediately upon hearing it for the first time (see Collins 2023, chapter 4 for relevant discussion and the ‘wedge’ argument).

A potential problem with my account is the following, noted by OREAL in footnote 11:

(11)

“Furthermore, the developmental pattern we observed differs from the developmental pattern observed in Condition B environments. Children allow illicit coreference up till the age of six (Chien and Wexler 1990), whereas in the present experiment most 4- and 5-year-olds did not display any difficulties.”

It is not clear to me how to account for this difference in development pattern under my suggested principle B alternative. However, the gap between age six and age four-five does not seem all that great to me. It may be that something about the passive is enabling them to correctly process principle B effects. Alternatively, they may be learning, like adults, that the dominant interpretation of the passive is existential/generic, as opposed to definite (see Collins 2023, chapter 4 for discussion). If so, they may cease to freely use the definite pro implicit argument as they get older.

While I criticize OREA in various ways above, I do agree that there is a disjointness implicature when the implicit EA is existential. How exactly to capture this implicature remains unclear to me. It is not really clear to me that a syntactically projected EA is needed to capture the disjointness implicature. In any case, the syntactic evidence in Collins 2005, 2023 for a syntactically projected EA in the short passive is already very compelling.

A minor criticism of OREA relates to an ongoing disagreement in the field as to whether the implicit EA in the passive needs to be syntactically projected. Collins 2005 and 2023 argue that the implicit EA of the short passive is syntactically projected. Bruening 2013 takes a contrary position. The paper would have been strengthened considerably if it had noted how its conclusions decided between current theories of the passive.

A challenge, not picked up in this review, is to develop an account of adjectival passives in the Merge-based framework of Collins 2023. I leave this to future work.

References

Bruening, Benjamin. 2013. By Phrases in Passives and Nominals. Syntax 16, 1-41.

Collins, Chris. 2005. A Smuggling Approach to the Passive in English. Syntax 8, 81-120.

Collins, Chris. To Appear. Principles of Argument Structure: A Merge-Based Approach. MIT Press, Cambridge.

Fox, Danny, and Roni Katzir. 2011. On the characterization of alternatives. Natural language semantics,19(1), 87-107.