Wednesday, December 1, 2010

Agent, Causer, and Instrument Subjects Are Not Distinct

There is a long history of distinguishing external arguments that seem to be volitional agents from both inanimate causers and instruments. Concentrating on Alexiadou and Schafer (2006) here (see their paper for more references), I show that the arguments for distinguishing them do not go through. The alternative is to say that all of them are the same, and are distinguished only by world knowledge. Alexiadou and Schafer call this view the "underspecification" view of external arguments. Borrowing the term from Ramchand (2008), let us refer to this underspecified external argument role as "Initiator." So the two views we are contrasting are the following:

A. All external arguments are Initiators.
B. External arguments must be divided into agents, causers, and instruments, which behave differently.

Alexiadou and Schafer (2006) present several arguments against a version of the Initiator view. However, the version they argue against is so simplistic as to be unworkable even without arguing against it. Their arguments against such a theory have no force against a more sophisticated version. Here is the crucial addition:

Every theory has to allow semantic selection for properties of the subject. Consider the following contrast:

(1) A rampaging elephant killed the Prime Minister.
(2)#A rampaging elephant murdered/assassinated the Prime Minister.

Unlike kill, murder and assassinate require subjects that are capable not just of volitional action, but also awareness of morality and politics. Similarly, the verb eat requires a subject that is capable of ingesting:

(3) The sea witch ate the captain's heart.
(4)#The sea ate the beach. (vs. The sea ate away at the beach, no such requirement)

These selectional requirements are over and above the distinction between agents and causers (or instruments). Every theory that I am aware of would treat the subjects of eat, kill, murder, and assassinate as agents.

So, every theory requires semantic selection in addition to whatever thematic roles it assigns to external arguments. In the Initiator theory, then, external arguments can all be assigned the broad role of Initiator, but individual verbs (or prepositions, or VPs) might impose additional selectional restrictions. With this in mind, let us examine Alexiadou and Schafer's arguments for distinguishing agents from causers.

Their first argument is that Greek passives only allow animate agents and not inanimate causers. So the Greek equivalent of The clothes were dried by the sun is reported to be ungrammatical (their example 2b). Similarly, some languages, like Jacaltec, are reported to only allow animate subjects of transitive verbs. Conversely, certain prepositions only allow causers:

(5) The window cracked from the pressure / *from Will.

Although one would want to know much more about Jacaltec, it is very easy for the Initiator theory to account for it: through selectional restrictions. Suppose transitive Voice in Jacaltec uniformly selects only animates; then the facts are accounted for, without any need to refer to a role of "agent." Similarly, the preposition from in the relevant sense in English selects nouns that denote events. Hence nouns like pressure, dancing, thermal expansion but not Will are good complements of from.

As for Greek, Alexiadou and Schafer go on to show that, in fact, inanimates are allowed in Greek passives; their examples in (42b-c) have The door was opened by THIS key/ the ELECTRONIC key (with focus intonation). I will return to this momentarily.

The last argument Alexiadou and Schafer give against the Initiator account is that, according to them, it cannot account for why some apparent instruments make good subjects while others do not. In fact, though, it accounts for the facts straightforwardly.

Here is the basic contrast:

(6) John loaded the truck with a crane/pitchfork.
(7) The crane/*pitchfork loaded the truck.

The Initiator theory accounts for this easily: external arguments have to be capable of being Initiators. That is, they have to be able to initiate events. In our conception of the world, cranes can do that, because they are self-propelled and capable of motion. Pitchforks are not. Change it to My new autopitchfork and (7) becomes fine. The same account explains the entire range of data they give in English, German, Dutch, and Greek, shown here only for English:

(8) The crane picked the crate up.
(9)#The fork picked the potato up.
(10) The falling axe broke the windowpane.
(11) The storm broke the windowpane.
(12) The chamomile cured Mary.
(13)#The scalpel cured Mary.

Again, cranes are capable of independent motion, whereas forks are not. (However, I personally find fork fine with modals or focus: "These plastic forks are just going to break. Only this titanium fork can pick up such a heavy potato." Again, see below.) Falling axes can initiate events by virtue of their motion; so can storms. Chamomile can because of its chemical properties. Scalpels, on the other hand, cannot do anything on their own. (Again, though, a scalpel is fine as a subject in certain circumstances: "THIS scalpel can rid/cure Mary of her unsightly blemishes!")

Alexiadou and Schafer also make an unmotivated distinction between "tools" and "secondary tools" (following Nilsen 1973). This distinction is supposed to be the following (and note that modals and focus do not help with 16b and 17b):

(14)a. Ashley cut the melon with a knife.
b. This knife cuts melon easily.
(15)a. Casey opened the door with the key.
b. This key opened that door.
(16)a. Cathryn ate spaghetti with a fork.
b.#This fork ate spaghetti.
(17)a. Denis is drinking juice with a straw.
b.#This straw is drinking juice.

There is no independent way of determining whether a tool is a "tool" or a "secondary tool." Moreover, the distinction is a spurious one; as shown above, the verbs eat and drink require external arguments that are capable of ingesting. It is the verbs that are behind the ill-formedness of (16b) and (17b), not the particular tool involved. Changing the verb changes the judgments:

(18)a. Cathryn scooped up spaghetti with a fork.
b. This fork can scoop up lots of spaghetti.

So, why do modals or focus help to make inanimates better external arguments? I submit that it is because they focus on inherent properties of the inanimate entity, properties that make it capable of carrying out the action of the verb. For instance, in (18b), using the ability modal focuses on the abilities of the fork, namely its tensile strength and scooping capacity (and retention, without spillage). These abilities, once they are brought to the conceptual fore, make the fork a good Initiator.

Finally, a major problem for Alexiadou and Schafer arises from their claim that inanimates can only be good external arguments if they can be construed as agents (like self-propelled cranes) or as inanimate causers (like falling axes, which can affect things in the same way storms can). They claim that an inanimate like a rag can never be either (see their complete set of examples in their 34a-e):

(19)#THIS rag cleaned the dishes. (personally I find this fine)

So, rag should never be a good subject. This is false:

(20) The rag muffled the sound.
(21) The rag smothered the fire.
(22) A rag covered the body.

These are clearly external arguments, because they can passivize (the sound was muffled by the rag, the fire was smothered by the rag, the body was covered by a rag). But this should not be possible, if rags cannot be agents or causers. In contrast, the Initiator view predicts these facts without difficulty. Rags can muffle sound, they can smother fires, and they can cover things. That is, their physical properties are such that they can initiate muffling, smothering, and covering events.

In sum, none of Alexiadou and Schafer's arguments for distinguishing agents from causers goes through. More broadly, I have seen no good reason to divide external arguments into different categories. The "underspecification" theory, which I have called the Initiator view here, accounts for all of the facts straightforwardly.


Alexiadou, Artemis and Florian Schafer (2006), Instrument Subjects Are Agents or Causers. In Proceedings of the 25th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla Proceedings Project: 40-48.

Nilsen, Don L.F. (1973), The Instrumental Case in English: Syntactic and Semantic Considerations. The Hague: Mouton.

Ramchand, Gillian Catriona (2008), Verb Meaning and the Lexicon: A First-Phase Syntax. Cambridge: Cambridge University Press.

Wednesday, November 17, 2010

Information-Structure Approaches to Islands

There is a strand of research that tries to explain island constraints on movement entirely in information-structural, or discourse, terms. Goldberg (2006) and Erteschik-Shir (2007) are recent examples of this. Both point to some phenomena that they claim are non-syntactic and yet are sensitive to islands, and conclude from this sensitivity that island constraints are not syntactic.

Citing Morgan (1975), Goldberg (2006, 132-133) claims that direct replies to questions are sensitive to islands. In all of the following, try to understand the answer to be "she is dating someone new." This is possible when that phrase does not occur inside an island, but not when it does:

(1) Q: Why was Laura so happy?
a. The woman who lives next door thought she was dating someone new.
b. #The woman who thought she was dating someone new lives next door.
c. #That she's dating someone new is likely.
d. It's likely that she's dating someone new.
e. #John shouted that she was dating someone new.
f. John said that she was dating someone new.
g. #John was hysterical 'cause she was dating someone new.
h. John left Manhattan in order that she could date someone new.

However, as Goldberg herself notes in footnote 3, this phenomenon is not fully general. Direct replies can occur inside complex NPs; example (2) is from Goldberg's foonote, and I add example (3). They can also occur inside initial if-clauses, which are very strong islands (4):

(2) Q: Why was Laura so happy?
A: I heard [a rumor that she was dating someone new].

(3) Q: Why was Laura so happy?
A: I talked to someone who said she was dating someone new.
(cf. * Who did you talk to [someone who said she was dating ---]?)

(4) Q: Why was Laura so happy?
A: If she was dating someone new, I would have heard about it.
(cf. *Who did you say that [if she was dating ---] you would have heard about it?)

It is therefore simply false that direct replies are sensitive to the same island constraints as movement.

Citing James (1972), Goldberg (2006, 133-134) also claims that exclamatives are sensitive to islands. Again, in all of the following try to take what is being remarked upon to be "Laura/she is dating someone new." This is possible if that phrase is not inside an island, but impossible if it is:

(5) a. Ah! The woman who lives next door thought Laura was dating someone new!
b. *Ah! The woman who thought Laura was dating someone new lives next door!

(6) a. Ah! It is likely that she was dating someone new!
b. *Ah! That she is dating someone new is likely!

(7) a. Ah! John said she was dating someone new!
b. *Ah! John shouted that she was dating someone new!

(8) a. Ah! John left Manhattan in order that she could date someone new!
b. *Ah! John was hysterical 'cause she was dating someone new!

Again, though, exclamatives are possible inside complex NPs when they are on a right branch:

(9) a. Ah! John talked to someone who said she was dating someone new!
b. Ah! John heard a rumor that she was dating someone new!

According to Erteschik-Shir (2007, 164), Morgan (1975) actually discussed fragment replies, not direct replies. Erteschik-Shir claims these are also sensitive to islands:

(10) Q: Did the man who Tricia fired leave town?
a. *No, Thelma.
b. No, the man who Thelma fired (left town).

This test seems to work a little better, since fragment replies are impossible with complex NPs on a right branch (11), but the correlation is still not perfect. It seems to me that a fragment reply is possible inside an indefinite subject (12):

(11) Q: Did you see the man who fired Tricia?
a. *No, Thelma.
b. No, the man who fired Thelma.

(12) Q: Was a statue of Tricia built in Poughkeepsie?
a. No, Thelma.
b. No, a statue of Thelma (was).
(cf. *Who was a statue of built in Poughkeepsie?)

Fragment replies also work with clause-initial if-clauses:

(13) Q: Did you say that if you see Tricia at the party, you'll leave?
a. No, Thelma.
b. No, if I see Thelma (I will).
(cf. *Who did you say that if you see at the party you'll leave?)

Once again, then, phenomena that are claimed to be sensitive to the same islands as movement are actually not.

Moreover, it seems to me that any discourse-based account of islands faces an insurmountable problem from sluicing contexts. Consider the following dialogs:

(14) A: Yesterday I met a man who claimed that John stole something.
B1: What?
B2: *What did you meet a man who claimed that John stole?

(15) A: John was furious because his wife had taken something of his.
B1: What?
B2: *What was John furious because his wife had taken?

(16) A: We've been assigned to read a book and write a certain kind of paper.
B1: What kind of paper?
B2: *What kind of paper have we been assigned to read a book and write?

In each of these cases, A makes a statement containing an indefinite. B seeks to determine the referent of that indefinite. B can do that with sluicing in all of the (B1) cases. B cannot do that by repeating the entire sentence, with extraction of a wh-phrase corresponding to the indefinite (the B2 cases). Yet the discourse contexts for B1 and B2 are identical. It is simply impossible to explain the contrast between B1 and B2 as being due to discourse.

I conclude from all of the above that discourse approaches to islands simply do not work, and should be abandoned.


Erteschik-Shir, Nomi (2007). Information Structure: The Syntax-Discourse Interface. Oxford: Oxford University Press.

Goldberg, Adele E. (2006). Constructions at Work: The Nature of Generalization in Language. Oxford: Oxford University Press.

James, D. (1972). Some Aspects of the Syntax and Semantics of Interjections. Paper presented at the 8th Regional Meeting of the Chicago Linguistic Society.

Morgan, J.L. (1975). Some Interactions of Syntax and Pragmatics. In P. Cole and J.L. Morgan (eds.), Syntax and Semantics Volume 3: Speech Acts. New York: Academic Press, 289-304.

Wednesday, October 6, 2010

A Note on Levine (2010) and So-Called Movement Paradoxes

Levine (2010) criticizes an analysis of the ass camouflage construction (ACC) proposed by Collins, Moody, and Postal (2008). In the ACC, the possessor of the noun ass acts as a binder and controller while the head noun does not, but at the same time the head noun strictly controls agreement:

(1) John and Mary's ass is (*are) making theyself mad. (Levine 2010, example 21)

In Collins, Moody, and Postal's (CMP) analysis, the possessor starts out in a lower position (the base position of the subject) and moves into the possessor position (where the NP [__'s ass] occupies the higher subject position, Spec-IP). As CMP themselves point out, this analysis requires giving up basic assumptions about movement; namely, that the derived position must c-command its base position.

Levine (2010) rightly criticizes CMP's analysis, and offers a different one in the framework of HPSG, where binding and agreement are determined by different features. However, he then goes on to draw an unwarranted conclusion regarding theory comparison. Basically, his argument goes like this: (1) the movement analysis does not fit in with the basic architecture of the theory within which it is couched; (2) the theory he offers follows straightforwardly from assumptions of the HPSG architecture that he adopts; (3) therefore, the HPSG theory is to be preferred.

However, this conclusion does not follow at all. I can adopt a movement theory architecture (as I usually do in my own work), but reject CMP's analysis of the ACC (which I did, as soon as I read it). It is true that some researchers working within a movement architecture make the methodological mistake of trying to analyze all grammatical phenomena as movement, but this is a mistake in some instances and not an inherent flaw in the architecture. Other researchers working within the same basic architecture offer competing non-movement accounts of the same grammatical phenomena (witness binding and control, both of which some have tried to analyze as movement, and others have argued are not movement at all). In fact, working within a movement-style architecture, I could even adopt Levine's own analysis of the ACC!

So, while Levine's criticisms of CMP are right on target, his broader conclusion is not. In fact, one could go further and point to an issue that proponents of non-movement theories often raise in support of their theories, but which in fact undermines them. This is the issue of what some have called "movement paradoxes." These are often cited as showing that movement is entirely the wrong approach. Basically, a constituent that has moved is ungrammatical in the position it is claimed to move from:

(2) (Sag 2010, example 55)
a. That Kim is ready, you can rely on __.
b.*You can rely on that Kim is ready.

Proponents of non-movement theories embrace this, claiming that in filler-gap dependencies, the filler and the gap do not have to match in category.

However, this mismatch has been shown to be quite limited, and works in only one direction. Recent work by Alrenga (2005) and Takahashi (2010) has shown that the generalization is the following (adapted from Takahashi 2010):

(3) A clausal constituent (CP) is allowed to move only if its base-generated position is one in which an NP is allowed to appear.

Both Alrenga and Takahashi offer movement-based accounts of this generalization.

The point that I wish to make is that the very limited nature of this mismatch screams out for an explanation. Mismatches should not be embraced as something that the very architecture of a theory should allow, as non-movement approaches do; rather, they should be banned, and an explanation for the one exception that exists should be sought. Non-movement approaches that embrace category mismatches predict that they should arise all over the place, when in fact they do not. For instance, you never see the converse of the generalization and example above: a case where a CP but not an NP is allowed, but suddenly under movement the NP is allowed. This never happens:

(4) a. *I hoped a good result.
b. *A good result, I hoped.
c. I hoped that a good outcome would result.

The verb hope allows a CP but not an NP; it still does not allow an NP when the NP moves. If category mismatches were fully general, we should find such cases. That we do not indicates that the theory that bans them is on the right track. A theory that embraces category mismatches will have to stipulate conditions that ban all but the one described in (3).


Alrenga, Peter (2005). A Sentential Subject Asymmetry in English and Its Implications for Complement Selection. Syntax 8: 175-207.

Collins, Chris, Simanique Moody, and Paul M. Postal (2008). An AAE Camouflage Construction. Language 84: 29-68.

Levine, Robert D. (2010). The Ass Camouflage Construction: Masks as Parasitic Heads. Language 86: 265-301.

Sag, Ivan (2010). English Filler-Gap Constructions. Language 86: 486-545.

Takahashi, Shoichi (2010). The Hidden Side of Clausal Complements. Natural Language and Linguistic Theory 28: 343-380.

Tuesday, September 14, 2010

Implicit Objects of Find-it-Psych-Adj

Landau (2010) attempts to give an argument that implicit arguments must be represented syntactically. His argument mostly uses a construction he calls find-it-Psych-Adj, illustrated below:

(1) The Professor found it annoying to listen to that speech.

As Landau shows, the subject of find is interpreted as the experiencer of the psych adjective, in this case annoying. It is this implicit experiencer that then controls the PRO subject of the infinitive (to listen to that speech).

What is interesting about this construction, not noted by Landau, is that it has all of the properties of Obligatory Control. First is the obligatory coindexing: the subject of find must be identical to the experiencer of the psych adjective. Second, as Landau shows, the implicit argument cannot be spelled out:

(2) John finds it amusing (*to him) to watch the prisoners suffer. (Landau 2010, ex. 38a)

As Landau shows, this is not true of a finite complement to find, just as it is not true of finite complements of control verbs:

(3) The researcher found that it was annoying to her/the subjects to have to hear the instructions again.
(4) compare control:
a. Billy hopes (*him/his friend) to become Captain Marvel.
b. Billy hopes that he/his friend will become Captain Marvel.

Third, only sloppy readings are available in ellipsis:

(5) The professor finds it amusing to race ants, and his assistant does too. (only, amusing to the assistant)

This is not true of finite complements to find, which allow strict readings:

(6) The professor found that it is amusing to him to race ants, and his assistant did too. (either, amusing to the assistant, or amusing to the professor)

Fourth, this construction only admits de se readings, just like control. The usual context for this involves an amnesiac war hero. Suppose this amnesiac hero is watching a TV show about his own exploits, but doesn't realize that the person he sees on the TV is him. He sees that the person on TV is embarrassed about discussing his own exploits. The following cannot describe this scenario:

(7) The amnesiac finds it embarrassing to discuss his war exploits.

The sentence in (7) requires belief de se, which is missing from this context. In contrast, a finite complement to find does not, and is felicitous in this context:

(8) The amnesiac finds that it is embarrassing to him to discuss his war exploits.

The overt him can refer to the person the amnesiac is watching on TV, without him realizing that it is himself. This is exactly like control into non-finite clauses versus overt pronouns in finite ones.

So, it appears that the relation between the subject of find and the experiencer of the psych adjective is one of obligatory control (which must be exhaustive, and cannot be partial; see Landau 2000). If all obligatory control is control of PRO, then the implicit experiencer of the psych adjective must be PRO:

(9) The amnesiac1 finds it embarrassing PRO1 [PRO1 to discuss his war exploits].

This is interesting in its own right, because PRO is generally restricted to subject position in English. But it also severely weakens Landau's argument that implicit arguments must be syntactically represented, because for the purposes of this argument he was excluding PRO. Most researchers recognize the syntactic reality of PRO, and basically view it as an unpronounced pronoun. PRO can participate in all the syntactic relations that an overt NP can; so it is not surprising that the implicit experiencer of the find-it-Psych-Adj construction can. What Landau really needs to make his argument go through is an implicit argument that his diagnostics classify as ``weak,'' but which can nevertheless do partial control. (Personally, I am not convinced that there are any ``weak implicit arguments'' in English).

One last comment on Landau's theory: In his theory, strong implicit arguments have D, but weak ones do not. This makes weak implicit arguments unable to serve as the subject of a secondary predicate, because, according to Landau, only NPs with D can be arguments. But then how could a weak implicit argument be an implicit argument in the first place? By definition, an implicit argument is an argument, of some predicate. This part of Landau's account appears to me to be inherently contradictory.


Landau, Idan (2010). The Explicit Syntax of Implicit Arguments. Linguistic Inquiry 41: 357-388.

Landau, Idan (2000). Elements of Control: Structure and Meaning in Infinitival Constructions. Dordrecht: Kluwer.

Update to post (9/17/2010):

Idan Landau sends the following comments. I agree with him that implicit goals of communication verbs are convincing cases of implicit arguments that can control but not bind.

Hi Ben,

Thanks for this interesting commentary. Here are some comments of my my own.

1. The ban on lexicalization of the argument is not special to this construction. Quite a few constructions involve an implicit position which can't be lexcicalized:

a. The topic merits discussion (*of it).

b. John gave me a/*his call.

I'm not sure PRO is motivated in all of them. Just like in your proposal, the implicit position in (a) is an object position; in fact, in the find-it-Adj. cases, it's an oblique position. For PRO to occur in it would be very unusual. In fact, so unusual - also crosslinguistiucally - that this alone is a strong reason for me to doubt your conclusion.

This would imply that the signature properties - sloppy reading, ncessary se se etc. - are a necessary corollary of PRO, but not a sufficient indicator for its presence; other grammatical elements may induce the same signature. Note that simple reflexives are also necessarily sloppy and de se:

c. John hates himself and Bill does too.

d. The unfortunate hates himself.

I somehow sense that the find-it-Adj. construction involves some "complex predicate", in which the argument structures of 'find' and 'Adj.' coalesce' (note that the subject of 'find' is also an experiencer), and that this process is responsible for the effects you observed. If true (I don't know yet how to formalize this), the situation would be quite different from standard OC, where the argument structures of the infinitive and the main predicate remain distinct (ignoring the small class of restructuring).

2. There are many examples of WIA controllers in English - other than in the find-it-Adj constructions. I cite an example with 'say', and one can add other communication verbs:

a. John said/signaled/shouted (to the kids) to stay alert.

Dative controllers are easily omissible with gerundive complements:

b. He recommended/suggested/offered [filing a complaint] (to the tenants).

Experiencer arguments of adjectives are also possibly implicit:

c. It is fun/amusing/disturbing (for Peter) to watch this scene.

And so on. So I don't think your skepticism is warranted.

3. Your last comment rests on the ambiguity of 'argument'. I tried to be clear on this term, but maybe I wasn't clear enough. Anyway, the contradiction is illusory.

One sense of argument is the logico-semantic sense: an argument of a predicative function, or more linguistically, a slot in a theta-grid. This sense is completely a-syntactic, and has no implications for syntactic representation. It is in this sense that WIAs are arguments.

A second, syntactic sense of 'argument' is relevant for the saturation of secondary predicates. This is where [D] is crucial, in definining SIAs. It is an empirical observation that two such notions of 'arguments' must be entertained (the split between WIA and SIA is theory-independent). One can think of various ways to localize it in the grammar. Perhaps it is in the nature of the predicator itself - verb or adjective; maybe it is in the distinction between primary and secondary predicates; or maybe it is a by-product of a Pred head, present in secondary predicates only (and not in verb-argument combinations). Whatever the solution is, the two senses will be available in different contexts and hence would not create a contradiction.

Wednesday, September 1, 2010

Locality as Linearization

Fox and Pesetsky (2005) make a number of interesting proposals regarding the locality of movement and linearization. In a nutshell, they propose that locality constraints on movement can be derived from the nature of linearization, which takes place cyclically. As Sabbagh (2007) points out, this theory predicts that string-vacuous movement should not be subject to any locality constraints. Sabbagh claims that this is true in the case of rightward movement (specifically, right node raising). How about leftward movement? Is string-vacuous leftward movement impervious to locality conditions?

The following are some attempts to find out. Although constructing relevant examples is not easy, it appears that the answer is no.

The first attempt involves though-preposing. This can front a predicate adjective inside an adjunct clause:

(1) Intelligent though he is, he still doesn't understand simple logic.

Exclamatives can also front predicate adjectives, and they can do so over clause boundaries:

(2) How intelligent he thinks he is!

Now we can try to form an exclamative from though-preposing. This would involve fronting the predicate adjective out of an adjunct clause (the though clause), but this movement would be string-vacuous. The result is ill-formed:

(3) *How intelligent though he is, he still doesn't understand simple logic!

If locality reduced to linearization, (3) should be well-formed. The fact that it is not seems to indicate that fronting out of an adjunct clause is not allowed, even when it would not change the linear order of any of the syntactic terminals.

One might object that maybe though-preposing would not be able to front a wh-predicate in the first place, so that the input to the exclamative fronting would be what is ill-formed here, not the fronting to form the exclamative. So let's try something that does not mix wh- and non-wh-phrases/operations. Here is a free relative, apparently headed by the wh-phrase what, which has undergone wh-movement from object position (object of ate):

(4) What he ate is still available.

The phrase what he ate is the subject of the clause. Subjects are known to be islands to movement. But if islands (locality) were derived from linearization, they should not have any effect if the movement is string vacuous. So suppose we embedded (4) under a question-embedding verb, and tried to question what:

(5) *I wonder what he ate is still available.

The result is ill-formed. It appears to be ill-formed because we have attempted to move what out of the subject what he ate. But this movement does not change the linear order; so if locality were derived from linearization, (5) should be well-formed.

Here is another example, which is analogous to Sabbagh's (2007) right node raising examples. According to him, right node raising is really across-the-board movement to the right. This only needs to be string-vacuous in the rightmost conjunct (but the gap has to be rightmost in all other conjuncts, too). So, if things worked like they should, across-the-board leftward movement should be able to violate all locality conditions if it is string-vacuous in the leftmost conjunct (and the gaps in the other conjuncts are also leftmost). So let's go back to though-preposing, which can front VPs as well as adjectives:

(6) Sit on a nail though he did, he still can't get out of class.

VP-fronting can front VPs across clause boundaries. So, if locality reduced to linearization, we should be able to front the VP sit on a nail out of the though clause. We won't be able to tell in (6), but we will be able to if we try to do it in an across-the-board fashion out of two conjuncts at once:

(7) *Sit on a nail though he did, he still can't get out of class, and though she might, it won't excuse her behavior.

The result is ill-formed. Again, it appears that string-vacuous movement, including string-vacuous across-the-board movement, is still subject to locality conditions (here, the ban on extracting from adjunct clauses).

I hope other people will try to construct other examples. If it is true that string-vacuous leftward movement still obeys locality constraints, as it appears to from the examples here, then it is not possible to derive locality conditions on movement from linearization.


Fox, Danny, and David Pesetsky (2005). Cyclic Linearization of Syntactic Structure. Theoretical Linguistics 31: 1-45.
Sabbagh, Joseph (2007). Ordering and Linearizing Rightward Movement. Natural Language and Linguistic Theory 25: 349-401.

Wednesday, May 26, 2010

Subject restrictions, particle verbs, and consumption verbs

Many people have noted that certain transitive verbs in English require intentional subjects:

(1) (Folli and Harley 2005, ex 1b)
a. The groom ate the wedding cake.
b.#The sea ate the beach.

(Example numbers from Folli and Harley 2005 come from the prepublication version available at Folli and Harley typically judge the inanimate subjects as "*," but I do not think it is appropriate to view them as ungrammatical. Rather, the sentence imputes intention to the subject, which is not usually appropriate with an inanimate subject. Hence, I prefer to mark them as "#.")

Both Folli and Harley (2005) and Ramchand (2008) claim that the class of verbs that require an intentional subject is the class of consumption verbs. This class is supposed to take an "incremental theme" object, the kind whose quantizedness in the sense of Krifka (1992) makes a difference for telicity:

(2)a. He ate three pieces of wedding cake in/??for an hour. (quantized object: telic)
b. He ate wedding cake for/??in an hour. (non-quantized object: atelic)

However, the class of verbs that require an intentional subject is clearly not the same as the class of consumption verbs. Here is the list of verbs given by Folli and Harley (2005):

(3) (Folli and Harley 2005, ex 24)
a.#The sea ate the beach.
b.#The wind carved the beach.
c.#Erosion nibbled the cliff.
d.#The washing machine chewed the laundry.

(4) (given for Italian by Folli and Harley 2005, ex 25)
a.#The sun drank the lake.
b.#Inflation sucked our savings.

While eat, carve, and drink take incremental theme objects, nibble, chew and suck do not:

(5)a. The child nibbled the cookie for/??in an hour. (quantized object but atelic)
b. The cowboy chewed two pieces of beef jerkey for/??in an hour. (quantized object but atelic)
c. The child sucked a lollipop for/??in an hour. (quantized object but atelic)

Conversely, there are consumption verbs like devour that take incremental themes, but have no subject restriction:

(6)a. I devoured two pies in/??for half an hour. (quantized object: telic)
b. I devoured pies for/??in half an hour. (non-quantized object: atelic)
c. The tsunami devoured the coastal resort.

Moreover, Folli and Harley (2005) claim that adding a particle or other resultative phrase makes the subject restriction disappear:

(7) (Folli and Harley 2005, ex 24)
a. The sea ate away the beach.
b. The wind carved the beach away.
c. Erosion nibbled away the cliff.
d. The washing machine chewed up the laundry.

(8) (given for Italian by Folli and Harley 2005, ex 25)
a. The sun drank up the lake.
b. Inflation sucked up our savings.

Folli and Harley therefore design a constructional theory of verb meaning wherein consumption verbs combine with a light verb, vDO, which requires animate agents. When they combine with a resultative particle, in contrast, they combine not with vDO but with vCAUSE, which requires only a causer, animate or inanimate.

However, it is not true that verb-particle combinations uniformly lack subject restrictions. Consider the consumption particle verb put away. With this verb, the subject restriction actually holds of the particle use, and not the non-particle use:

(9)a. His phone records put the suspect in Manhattan at 12:03 AM. (inanimate subject OK)
b. The groom put away thirteen helpings of wedding cake.
c.#The sea put away the beach. (=/= The sea ate up the beach.)

Other particle verbs also have subject restrictions:

(10)a. The authorities locked me up.
b.#The earthquake locked me up. (situation: I'm touring a jail with self-locking doors when an earthquake hits, causing the door of the cell I'm in to close)

(11)a. The gangster bumped off his rivals.
b.#The tsunami bumped off the sunbathers.

(12)a. The speaker explained away the exceptions.
b.#That theory explains away the exceptions. (OK if understand that theory to mean "advocates of that theory")

Therefore, the constructional theories designed by Folli and Harley (2005) and Ramchand (2008) will not work. They predict that all and only verbs that take incremental theme objects will show subject restrictions, and no verb-particle combination will. Both predictions are false. Moreover, Folli and Harley (2005) note that there are other verbs that require intentional subjects, but they offer no constructional account of these:

(13)a. Sue/#The tornado murdered someone. (Folli and Harley 2005, ex 21f)
b. The warden/#Sickness jailed Andrew. (Folli and Harley 2005, ex 21h)

Since these are not "consumption" verbs, they do not fall under Folli and Harley's theory. The quantizedness of their objects does make a difference for telicity, but in this they do not differ from close synonyms like kill and imprison, which have no subject restrictions:

(14)a. The tornado killed someone.
b. Sickness imprisoned Andrew.

Factually, then, subject restrictions do not line up nicely according to other verb classifications. Verbs that take incremental themes may or may not impose subject restrictions. Verb-particle combinations that encode something like a cause and a result state also may or may not impose subject restrictions. There seems to be no independent classification of verbal phenomena that correlates with subject restrictions.

Rather, it seems to me that a verb taking only an intentional subject is simply part of our lexical knowledge of that particular verb, in combination with world knowledge. Take the difference between kill and murder. Aspectually these two verbs are identical, but the former imposes no restriction on its subject while the latter does. This is clearly because of the lexical meaning of murder: it includes intentionality in its meaning. Turning to consumption verbs, verbs like eat also seem to require intentionality of their subjects; presumably eat means something like "consume with a mouth," which requires that its subject have a mouth that the subject controls. Adding a particle appears to refocus the meaning of the verb on the resulting state of the object, with the effect that the subject restrictions disappear. However, with put away the opposite happens: put by itself only involves changing (or specifying) something's location, with no intentionality involved. But adding away to create a resultative consumption verb adds the implication of desire and volition.

In summary, although there appear to be generalizations regarding which verbs impose restrictions on their subjects, this is not true; if one looks around, one discovers exceptions in every possible way.


Folli, Raffaella, and Heidi Harley (2005), "Flavors of v." In Paula Kempchinsky and Roumyana Slabakova, eds., Aspectual Inquiries, Dordrecht: Springer, pp. 95–120.

Krifka, Manfred (1992), "Thematic Relations as Links between Nominal Reference and Temporal Constitution." In Ivan Sag and Anna Szabolcsi, eds., Lexical Matters, Stanford, CA: CSLI Publications, pp. 29–53.

Ramchand, Gillian Catriona (2008), Verb Meaning and the Lexicon: A First-Phase Syntax. Cambridge: Cambridge University Press.

Monday, May 10, 2010

Resultatives: Exceptions to the Direct Object Restriction

Wechsler (2005) claims that the direct object restriction on resultative secondary predicates is not real, and that resultatives can be predicated of deep subjects if the conditions are right. His claimed counterexamples include unergative verbs of motion like the following:

(1) She danced/swam free of her captors. (from Levin and Rappaport-Hovav 1995, p. 186)
(2) However, if fire is an immediate danger, you must jump clear of the vehicle. (State of Illinois, Rules of the Road; cited in Levin and Rappaport-Hovav 1995, p. 186)
(3) The driver and the fireman had jumped clear before the crash. (Thomas the Tank Engine; Wechsler 2005 ex. 33c)

Resultatives can also be predicated of subjects of transitives, according to Wechsler:

(4) The wise men followed the star out of Bethlehem. (Wechsler 2005, ex. 34a)
(5) The sailors managed to catch a breeze and ride it clear of the rocks. (Wechsler 2005, ex. 34b)
(6) He followed Lassie free of his captors. (Wechsler 2005, ex. 34c)

These examples are very suspicious, however. First, example (4) is simply a directional PP: it can be replaced with any such PP (through the pass, under the archway). Second, the only adjectives that can be used like this are free and clear, and they can be used adverbially with a meaning that is clearly not resultative:

(7) Stand clear of the moving doors. (subject does not become clear as a result of standing)

Other adjectives that generally make good resultatives (as in example 8) cannot be used with these verbs:

(8) She wiped the table clean/dry.
(9)a. *The sailors rode the breeze dry.
(9)b. *The sailors jumped clean/dry.

(And note that clean and dry are closed-scale adjectives, so they should be fine as resultatives modifying selected arguments of the verb, according to Wechsler. See more below on closed- vs. open-scale adjectives.)

So, an alternative explanation for all of Wechsler's putative counterexamples to the direct object restriction is that they are directionals, basically like a PP. That is, free and clear can be used like a PP as directional modifiers. As such, they modify the main predicate: verbs of motion regularly allow specification of starting points, paths, and end points. But these are not resultatives at all, which are secondary predicates and not modifiers of the main predicate.

Evidence that this is correct comes from conjunction possibilities. Free and clear can be conjoined with directional PPs with these verbs:

(10) They jumped clear of the vehicle and through the hoop.
(11) She danced free of her captors and into the next room.
(12) They rode the waves clear of the rocks and onto the beach.

This is not possible with APs that are clearly resultatives, even when the PP by itself is fine with the verb:

(13) She pounded the metal through the hoop.
(14) She pounded the metal flat.
(15)*She pounded the metal flat and through the hoop.

A second piece of evidence comes from pseudoclefting. Wechsler's free and clear examples allow pseudoclefts with where, like directional PPs, but true resultatives do not:

(16) Free of her captors is where she danced (to).
(17) Clear of the rocks is where they rode the waves (to).
(18)*Flat is where they pounded the metal (to).
(19)*Clean is where they wiped the table (to).

A third piece of evidence comes from co-occurrence with starting point and path PPs. Endpoints regularly occur with both of these:

(20) The sailors rode the breeze off the rocks, along the shore, and out into the open sea.

So do clear and free, again acting like directional PPs:

(21) The sailors rode the breeze through the narrow gap and clear of the rocks.
(22) She danced out of her captor's arms and free of their grasp.

But true resultatives do not:

(23) I kicked the box out the door, down the hall, and into the trash.
(24) I kicked the box to pieces.
(25) I kicked the box (*to pieces) out the door (*to pieces) and down the hall (*to pieces).
(26) I danced myself dry (*from soaking wet).
(27) I laughed myself silly (*from perfectly lucid).

I conclude, therefore, that Wechsler's putative counterexamples to the direct object restriction are not real counterexamples. They are not resultatives at all: they are directionals. When they specify the endpoint of a path, they might seem to have the semantics of a resultative, but they have very different grammatical properties from true resultative secondary predicates.

More generally, I conclude that there are no real counterexamples to the direct object restriction, which stands as a significant generalization about resultatives.

Moreover, Wechsler claims that examples like (28) are ungrammatical because resultatives that are predicated of arguments of the verb must be closed-scale adjectives. (29) is grammatical because there is no such requirement on non-selected NPs:

(28)*We danced tired.
(29) We danced ourselves tired.

This is not a good enough explanation for the contrast between (28) and (29), however. Sentences like (28) are still bad with closed-scale adjectives like dry and clean:

(30)*We danced dry/clean.

There is no way to predicate a true resultative of an underlying subject without a fake reflexive as in (29). Again, the direct object restriction stands.


Levin, Beth and Malka Rappaport-Hovav (1995), Unaccusativity: At the Syntax-Lexical Semantics Interface. Cambridge, MA: MIT Press.

Wechsler, Stephen (2005), Resultatives Under the Event-Argument Homomorphism Model of Telicity. In Nomi Erteschik-Shir and Tova Rapoport (eds.), The Syntax of Aspect: Deriving Thematic and Aspectual Interpretation. Oxford University Press. pp. 255-273.

Friday, May 7, 2010

The Adjacency Requirement on Appositives

This second post also addresses Christopher Potts' book, The Logic of Conventional Implicatures (OUP, 2005). In chapter 4, Potts discusses various kinds of supplements, including appositive relative clauses. A general feature of supplements, according to Potts, is that they must be strictly adjacent to what they modify (the example number is that from Potts, page 104):

(4.29a) *We spoke with Lance before the race, who is a famous cyclist, about the weather.

However, the adjacency requirement appears to me to be relaxed just when another supplement is embedded in the non-local one:

(1) We spoke with Lance before the race, who, as far as I could tell, was primed and ready to go.
(2)??We spoke with Lance before the race, who was primed and ready to go.
(3) Bob Scalpbender came by, who, as you know, endorses phrenology.
(4)??Bob Scalpbender came by, who endorses phrenology.

The non-local supplement also has to occur rightmost; adding "about the weather" to (1), as in (4.29a), renders the sentence unacceptable.

This relaxation of the adjacency requirement is a rather odd fact, one which I have no account of.

First post! Expressive attributive adjectives

I'll inaugurate Linguistics Commentary with some observations on Christopher Potts' book The Logic of Conventional Implicatures (OUP, 2005).

In Chapter 5 (section 5.3.1), Potts argues that Expressive Attributive Adjectives (EAs) like damn and f-cking have a completely standard syntax, and are garden-variety attributive adjectives. It seems to me that this is not true; EAs have a number of properties that distinguish them from other adjectives. (I use f-cking in all the examples below, but any other EA will work as well, like damn.)

First, EAs can modify other adjectives, but most adjectives cannot; they need to be turned into adverbs:

(1) That's a f-cking big insect!
(2) That insect is f-cking huge!
(3) That's a horrendously/*horrendous big insect!
(4) That insect is horrendously/*horrendous big!

Second, EAs can modify an adjective that appears before them, unlike other adjectives:

(5) That's a big f-cking insect! (what's shocking is how big it is)
(6) That's a big horrendous insect! (horrendous does not modify big)

Third, modifiers of adjectives (adverbs) allow the head adjective to elide, but this is not possible with an EA:

(7) Are you hungry? Terribly!
(8) Are you hungry? *F-cking! (vs. I'm f-cking hungry!)

Similarly, as adjectives, EAs do not allow one-replacement:

(9) He's a conservative Republican, and she's a moderate one.
(10)*He's a f-cking Republican, and she's a f-cking one too.

Fifth, EAs have a different intonation pattern from other adjectives/adverbs. They cannot take the main stress (stress indicated by CAPS):

(11) I'm TERRibly hungry!
(12)*I'm F-CKing hungry!
(13) I'm f-cking HUNgry!
(14) That's a big UGly bug!
(15)*That's a big F-CKing bug!
(16) That's a BIG f-cking bug!
(17) I was hit by a CRAzy madman!
(18)*I was hit by a F-CKing madman!
(19) I was hit by a f-cking MADman!

Relatedly, they cannot be contrastive, unlike regular adjectives:

(20) The CONSERvative Republicans tried to kill this bill, not the moderate ones.
(21)*The F-CKing Republicans tried to kill this bill, not the ones I like.

Hence, EAs cannot perform the function of restricting reference, unlike regular adjectives.

While some of these properties probably follow from the theory of EAs that Potts gives, some others don't: property (i), they can modify adjectives directly; (ii) in prenominal position they can modify the adjective that precedes them; (iii-iv) they do not allow ellipsis or one-replacement.

That's it for the first post! I hope someone can take up these properties and explain them.