Apparent Gapping in NPs (Orth & Yoshida 2023)

Orth & Yoshida (2023) note the existence of something that looks like gapping inside NPs, where the head noun can be missing in the second conjunct, stranding an A-PP sequence:

(1)  I interviewed every candidate possible in this group and promising in that group.

Orth & Yoshida (2023) analyze this as ATB movement of N to D, with coordination of NPs below a single D.  As they show, the A must linearly follow the head noun in the first conjunct, which they analyze as movement of N to D across the A.  In their footnote 15, they dismiss a possible alternative analysis, one based on the prosodic deletion analysis that I proposed for non-constituent coordination in Bruening (2015).  Their dismissal is based on two statements, both of which seem to be false.

First, Orth & Yoshida (2023) state that there is no contrastive prosody in the examples they examine, like (1).  Nothing could be further from the truth.  The two As are contrastive, as are the two PPs, and, in my judgment at least, heavy contrastive stress has to go on possible and promising in (1).

Second, they state that the examples of non-constituent coordination in Bruening (2015) involve deletion of all the non-head elements of the coordination, whereas in their examples, it is the head noun that would have to be deleted.  This seems to be based on a confusion of syntactic and prosodic categories.  In Bruening (2015), the deletion rule targets the first phonological phrase after and and deletes all but the head of that phonological phrase, where the head is a prosodic constituent (a prosodic word, typically).  This rule actually seems like it would give a perfect analysis of examples like (1): (every candidate (promising)) would be a phonological phrase with head (promising), and all but the head would be deleted. Subsequent material, here the PP, is not affected, so the output would be exactly what is observed after and in (1).

This analysis would explain all of the facts discussed by Orth & Yoshida (2023).  The word order requirement falls out, because prosodic constituents in English are right-headed.  The limitation to coordination falls out, because non-constituent coordination is also limited to coordination contexts.  

The two analyses also make different predictions about the co-occurrence of prenominal and postnominal adjectives.  In my judgment, the deletion is acceptable when both are present:

(2) I interviewed every available candidate possible in this group and promising in that group.

Orth & Yoshida (2023) predict this sentence to be ungrammatical, since the head noun candidate has evidently not moved across the first A available in the first conjunct.  I conclude that the prosodic deletion analysis in Bruening (2015) is actually a better analysis for this set of data than the ATB movement analysis proposed by Orth & Yoshida.

References

Bruening, Benjamin (2015). Non-Constituent Coordination: Prosody, Not Movement.  In Proceedings of the 38th Annual Penn Linguistics Conference.  Philadelphia: University of Pennsylvania Working Papers in Linguistics.  Available at https://repository.upenn.edu/bitstreams/a1437ea1-75de-4b2d-9b01-d6364582c35e/download.

Orth, Wesley and Masaya Yoshida (2023). There is Something Missing in NP and Moving in DP. Natural Language and Linguistic Theory 41: 1509-1527.

Selectional Violations in Coordination and Processing

Sentences like You can depend on my assistant and that he will be on time (Sag et al. 1985) have been heavily discussed in the literature on coordination.  They are interesting because the second conjunct, a CP, is not allowed as the object of a preposition (for many speakers of English).  So *You can depend on that he will be on time is judged to be deviant (see Bruening 2025 for experimental verification of these judgments).  Bruening & Al Khalaf (2020) and Bruening (2025) analyze the saving effect of coordination as arising from a null N.  In You can depend on my assistant and that he will be on time, the CP is actually an NP, with a null N head.  The null N head is not allowed when the CP occurs in object position by itself, because it is semantically contentless; see the two works cited for details.

Kim & Lu (2024) propose instead that the acceptability of You can depend on my assistant and that he will be on time is a grammaticality illusion.  The idea is that the language processor basically forgets the selectional requirements of depend on by the time it gets to the CP.  Kim & Lu (2024) point out that in coordinations of three phrases, like The success of the project depends on [a good engineering design], [the diligence of the workers], and [that the contractors will do their part], the null N analysis predicts that the CP should be equally grammatical as either the second or the third conjunct.  In contrast, the processing theory would predict that acceptability would increase with distance, so that a CP would be judged better as the third conjunct than as the second.  Kim & Lu (2024) run an acceptability study and find that, indeed, a CP is judged better as the third conjunct than as the second.  They argue that this is evidence for the processing account.

There are two reasons that this is not correct.  First, it is true that in the null N analysis, there is no difference in grammaticality between the second and third conjuncts: a CP would be equally well-formed in both positions, according to the grammar.  This does not mean that naive participants in an experiment will judge them that way, though.  We know from decades of research that acceptability judgments in experiments involve many factors besides the grammar.  Increased complexity is known to lower acceptability judgments, for instance.  So Kim & Lu's (2024) findings are not direct evidence against the null N theory (or any grammar theory).

Second, the grammaticality illusion analysis incorrectly predicts that all kinds of selectional violations should be acceptable in coordination.  If the processor basically forgets what categories are allowed by the time a non-initial conjunct is reached, a non-initial conjunct should be able to be any category.  Bruening & Al Khalaf (2020) showed that this is not true.  For instance, the semantically contentless preposition of that is allowed in nominalizations (the destruction of two of the towns) is not allowed in second or third conjuncts with conjoined arguments of verbs: *They destroyed all the cities and of two of the towns. *They destroyed the world famous library, the lighthouse that was an engineering marvel, and of the hanging gardens.  See Bruening & Al Khalaf (2020) for further examples involving other categories.  The processing theory would have to be modified so that the processor can forget selectional requirements just when a non-initial conjunct is a CP.  This does not seem like a reasonable modification, since forgetting A should not depend on what B is encountered later.

I conclude that, while there might be a role for processing in selectional violations in coordination, Kim & Lu (2024) have not shown that the best analysis is in terms of a grammaticality illusion.  Their findings are also not problematic for grammar-based theories like that of Bruening & Al Khalaf (2020) and Bruening (2025).

References

Bruening, Benjamin (2025). Selectional Violations in Coordination (A Response to Patejuk and Przepiorkowski 2023). Linguistic Inquiry 56: 439-483.

Bruening, Benjamin and Eman Al Khalaf (2020). Category Mismatches in Coordination Revisited. Linguistic Inquiry 51: 1-38.

Kim, Nayoun and Jiayi Lu (2024). Coordination of Unlike Categories Creates Grammaticality Illusion.  Korean Journal of English Language and Linguistics 24: 52-61.

Sag, Ivan et al. (1985). Coordination and How to Distinguish Categories. Natural Language and Linguistic Theory 3: 117-171.

A Late Note on Bobaljik (2008)

Bobaljik (2008) is a significant publication for two reasons: (1) for arguing that it is morphological case and not grammatical function that determines what is accessible for agreement; and (2) for showing that this explains a typological gap when case and agreement do not align. Bobaljik (2008) also argues for two other points that have not generally been adopted: First, that morphological case is assigned post-syntactically; and second, that agreement, since it depends on morphological case, must be a post-syntactic operation as well. In this short note, I question Bobaljik's motivation for making morphological case assignment be post-syntactic.

The basic argument that morphological case assignment is not syntactic is that it does not feed any other syntactic operation (other than agreement). According to Bobaljik (2008), nothing in the syntax depends on whether an NP is assigned, say, dative versus nominative case. However, the very data that he discusses show that this is false. As he notes, dative subjects in Icelandic have all the syntactic properties of subjects, but apparent dative subjects in German do not. If morphological case assignment does not take place in the syntax, then this difference is impossible to state. Take the verb `help'. In both languages, it assigns dative case to its complement. In both languages, the verb can be passivized. In Icelandic, this makes the dative NP the grammatical subject. In German, it does not. But the logical object of other verbs (those that assign structural accusative case) does become the grammatical subject in German. There is no way to explain the difference in German except to refer to the lexical dative case assignment, since both types of verbs can be passivized. I conclude that morphological case assignment does have syntactic consequences, and it therefore cannot be taken out of the syntax.

Moreover, the logic of the argument is flawed. There are many syntactic operations that apparently do not have any syntactic consequences, except perhaps for agreement. For instance, an adjective can be merged with a noun in an NP. In no language that has ever been described will this have any consequences for the distribution of that NP. The adjective might agree with the head noun, but that is the only syntactic operation that merging an adjective will feed. Should we then conclude that merging an adjective with a noun does not take place in the syntax? Of course not. Merging two things is exactly what syntax does. It would be redundant to have another module of grammar to do that just for adjectives and nouns.

Since most of the field has not accepted Bobaljik's (2008) conclusion that both morphological case and agreement are post-syntactic, this note is probably not important, but it is worth making. To my knowledge, no one has addressed his argument directly.

Reference: 

Bobaljik, Jonathan David (2008), “Where’s Phi? Agreement as a Post-Syntactic Operation.” In Daniel Harbour, David Adger, and Susan Bejar, eds., Phi-Theory: Phi Features Across Interfaces and Modules, Oxford: Oxford University Press, pp. 295–328.

"Binding" in short passives is logophoric

(This is an excerpt from my paper to be published in Glossa, "English middles and implicit arguments." I am excerpting it here because the data may otherwise go unnoticed, when they are quite important to the analysis of passives.)

Short passives in English appear to permit binding of an anaphor by the implicit logical external argument (e.g., Baker et al. 1989; Roberts 1987; Collins 2005):

(93) a. Some poems can be read aloud to oneself better than others can.
b. Melissa was delighted to find that some of her poems could be read aloud to herself quite well.
c. We were delighted to find that some of our poems could be read aloud to each other quite well.
d. Randi was delighted to find that the butter could be spread easily on herself after she had microwaved it.
e. To the students’ consternation, tickets to the school play were sold more easily to themselves than they were to others.
f. The winter swimmers were delighted to find that the goose fat could be rubbed easily onto themselves once it was warmed up.

One might conclude from this that the logical external argument of the short passive is syntactically represented. This conclusion would be too hasty. Anaphors can sometimes be used logophorically, meaning that they can take as an antecedent an entity that does not locally c- command them and may not even be present in the syntax (Pollard and Sag 1992; Reinhart and Reuland 1993). Charnavel and Sportiche (2016) offer a way to rule out this possibility: using inanimate anaphors. Anaphors used logophorically can only take sentient entities as their antecedents. Inanimate anaphors require a local c-commanding antecedent and do not permit a logophoric use. When we use inanimate anaphors in short passives, we find that they are unacceptable:

(94) a. * The moon’s gravitational pull isn’t strong enough for asteroids to be attracted to itself very easily.
(cf. The moon attracts asteroids to itself quite frequently.)
b. * This machine’s programming has the result that oil is spread on itself once a day. (cf. This machine spreads oil on itself once a day.)
c. * This automatic thresher’s design allows spare blades to be stored inside itself. (cf. This automatic thresher stores spare blades inside itself.)
d. * This machine is designed so that X-rays are constantly shot at itself. (cf. This machine constantly shoots X-rays at itself.)

I take this to indicate that putative examples of binding by the missing external argument of a passive are not actual examples of binding. The missing external argument of a passive is not capable of binding an anaphor. An anaphor can be acceptable in the passive, but when it is, it is being used logophorically.

Further strengthening this conclusion, unaccusative examples that are similar to the examples of the passive in (93) also permit an anaphor. Unaccusatives do not have an external argument, so they must involve anaphors being used logophorically:

(95) a. Some tones sound better to oneself than they do to others.
b. Randi was delighted to find the butter spreading out all over herself without her needing to do anything.
c. To the hikers’ consternation, snow seemed to fall only onto themselves and not onto the bushes alongside the path.
d. The winter swimmers were delighted to find that the goose fat melted easily onto themselves once it was warmed up.

I conclude that apparent binding by the logical external argument of a short passive is not binding at all, it involves a logophoric use of an anaphor. It appears that, when there is no logical external argument in the syntax (in short passives, in unaccusatives, and in middles---see the paper), it is possible for any anaphor in the verb phrase to be interpreted logophorically rather than as a syntactic anaphor. Why this would be the case is not entirely clear.

(See the paper for complete references, and an analysis of middles and passives and why they differ with regard to depictive secondary predicates and control.)

Idioms and Other Fixed Expressions in Ditransitives

Finally, a new post! This one is about a recent paper by Larson (2017), on fixed expressions in ditransitive constructions. Larson (2017) aims to be clearing up confusion on this topic, but in fact he adds to it. He also completely misses the point of the data.

Larson argues that we need to distinguish between idioms and other types of fixed expressions, especially collocations. He claims that idioms must be constituents, but collocations do not have to be. His evidence regarding collocations, however, is exactly the same evidence that has been used to show that idioms do not have to be constituents (e.g., O'Grady 1998). His one example to show that a literal collocation like rancid butter is not a constituent is his example (25), rancid yellow creamery butter, where modifiers can come in between parts of the collocation. But this type of modification was shown long ago to be common with idioms, too. For instance, a phrase that Larson treats as an uncontroversial idiom, the cat (is) out of the bag, can be modified in exactly the same way: The Feminist Cat is out of the Marxist Bag (https://www.youtube.com/watch?v=374dTEMGii0). Idioms with open slots for possessors are also common, like get X's goat, but get and goat do not constitute a constituent. In fact, my own recent work (Bruening 2017) argues that idioms, collocations, and other types of fixed expressions all obey exactly the same constraints, and should be accounted for in exactly the same way. I am not the first to argue this; see Bruening 2017 for references.

The point of fixed expressions in ditransitive constructions made in Bruening (2010) is that they fall into certain patterns, excluding others. Specifically, there are expressions like give X the creeps, which include the verb and second object but not the first object; there are expressions like give rise to X, which include the verb, first object, and preposition, but not the object of the preposition; there are expressions like send X to the showers, which include the verb and prepositional phrase but not the first object. There is also an alternating class, read X the riot act, read the riot act to X. Most importantly, there are no expressions like throw the wolves X, which would include the verb and first object while excluding the second object. There are also no expressions that alternate with this frame (throw the wolves X, throw X to the wolves; throw X to the wolves exists but it does not alternate).

This is a striking pattern of data that requires an account. Whether you call any individual expression an "idiom" or a "collocation" is irrelevant. That is just a matter of terminology. As linguists, we are supposed to be concerned about patterns, and identifying them and accounting for them. Larson's paper completely misses the point, because all it does is argue about whether certain expressions should be called "idioms" or "collocations." It has nothing to say about the actual patterns that we see with idioms and collocations.

Larson also argues that many if not all expressions of the form V X to NP, like send X to the showers, should not be considered "datives." Again, this is an irrelevant matter of terminology. The patterns are what is important, not what we call them. Now, Larson could claim that expressions like send X to the showers could not alternate precisely because it is not a dative; that is, it has the wrong semantics. But this does not explain the patterns: Why are there no "dative" expressions (with a caused possession semantics)? Bruening (2010) proposed an explanation. Larson does not, and does not mention the explanation in Bruening (2010). All Larson does is complain about terminology.

Larson is also quite inconsistent and imprecise. At one point (p399) he defines collocations as "strings of words used together frequently and recognized as such by speakers." He goes on to claim that expressions like give... the creeps and give... the boot are collocations, not idioms, as discussed above. This is not possible, given his definition: they are not strings. Since he has no coherent notion of what a collocation is, his terminological argument is even more beside the point.

To sum up, the patterns of fixed expressions in ditransitives were identified and analyzed in Bruening (2010). Nothing in Larson's paper alters the empirical picture outlined in Bruening (2010) or detracts from the analysis of that picture presented there. As linguists, we need to be concerned about patterns in data and accounting for them. Debates about terminology are beside the point.

References

Bruening, Benjamin (2010). Ditransitive Asymmetries and a Theory of Idiom Formation. Linguistic Inquiry 41: 519-562.

Bruening, Benjamin (2017). Syntactic Constraints on Idioms (Do Not Include Locality). In Halpert, Claire, Hadas Kotek, and Coppe van Urk (eds.), A Pesky Set: Papers for David Pesetsky. Cambridge, MA: MIT Working Papers in Linguistics, chapter 8.

Larson, Richard. (2017). On "Dative Idioms" in English. Linguistic Inquiry 48: 389-426.

O'Grady, William (1998). The Syntax of Idioms. Natural Language and Linguistic Theory 16: 279-312.

Derived Nominals

First post in a while! Blame the administrative duties that I have been forced to take up for the last couple of years. On to the post:

Borer (2014) claims that two interesting and surprising generalizations hold of derived nominals. The first is that derived nominals that are eventive and take VP-type event arguments and modifiers always have completely regular, compositional meanings. Conversely, derived nominals that have special non-compositional meanings may never take VP-type arguments and modifiers. For instance, transformation has a technical sense in the field of syntax, but it may not have this meaning when it takes VP-type arguments and modifiers like by-phrases:

(1) the transformation of the field by the linguist (regular, compositional meaning; Borer 2014, example 5b)
(2) this theory countenances transformations as a grammatical construct (special meaning)
(3) *the transformation of the structure by the linguist (cannot have special meaning; Borer 2014, example 5a)

Other examples cited by Borer include government, constitution, civilization, reading. When used with VP-type arguments and events, these only have the meaning that the verb has, not the meaning that the derived nominal used as a simple noun may have.

The second interesting and surprising generalization is that, according to Borer, derived nominals that take VP-type arguments and modifiers must embed a real, attested verb. For instance, transformation embeds the verb transform, government embeds govern, etc. In contrast, nominals like the following (from Newmeyer 2009, in Borer's footnote 10) are claimed not to allow VP-type arguments and modifiers (her judgments):

(4) the scrutiny of dubious looking tax forms (??by the IRS) (*in order to uncover tax evaders)
(5) the constant mischief by the boy (*for two hours) (*in order to get attention)
(6) the ongoing treason (*by Quisling) (*in order to support Nazi Germany)
(7) the homicide of AltaVista and AllTheWeb (*by Yahoo) (*in order to increase the value of its shares)

It appears to me, however, that Borer does not have either of these generalizations quite right. I will begin with the second one, which just seems to be incorrect, at least for English. All of the English speakers I have asked find the phrases marked as ungrammatical in 4-7 perfectly fine. I have found several possible examples on the web, as well:

(8) You are an emotional girl prone to tantrums and mischief just to get attention even if it is negative!
(9) It will start looking like a scam or mischief just to get people to waste their trade offers.
(10) You can't look at the death penalty in a vacuum. It's merely a justifiable homicide in order to protect others.
(11) Treason by Public Officials? (headline)

It appears that such nominals are perfectly able to take by-phrases and purpose clause, two of the VP-type modifiers that Borer says are limited to nominals derived from attested verbs. Less clear are modifiers like for two hours, but my own judgment is that mischief for two hours is well-formed, as is the scrutiny of tax forms for two hours.

I believe the second generalization to be incorrect, then, at least for English. I also believe that the first generalization, while correct, is actually part of a larger generalization. This larger generalization includes not just derived nominals but also derived adjectives and derived verbs. For instance, adjectival passives may have VP-type arguments and modifiers like by-phrases, but they never have special, non-compositional meanings when they do. The adjectival passive hammered can occur with a by-phrase, as in 12 (embedding hammered under looks guarantees that this is an adjectival passive and not a verbal one). Hammered can also mean heavily intoxicated, but not if a by-phrase is present:

(12) The younger spruce lining this trail look hammered by new snow that fell in fat moist flakes...
(13) She looks hammered (*by all the wine she drank).

Hung as an adjective has a special meaning when it modifies jury, but again this meaning is absent with a by-phrase:

(14) a hung jury
(15) *a jury hung by the convincing arguments of both lawyers

This indicates that Borer's generalization referring to derived nominals carries over to derived adjectives: they may not have special, non-compositional meanings when they take VP-type modifiers.

Note also that adjectival passives can take VP-type modifiers even when they are not derived from an attested verb:

(16) Even good corrections officers feel embattled by dangerous inmates who badly outnumber them.
(17) *Dangerous inmates embattled the corrections officers.

Embattle does not seem to exist as a verb, only as an adjectival passive (at least in my lexicon). This again points to the incorrectness of Borer's second generalization above.

Going back to the first generalization, I believe that derived verbs, which always have argument structure and take VP-type arguments and modifiers, are also always interpreted compositionally. I have not done a systematic investigation, but I have not been able to find any verbs derived with -ize, -ate, -ify, -en, or zero derivation that have an irregular, non-compositional meaning. (See discussion of some possibilities below.)

It therefore appears to me that the larger generalization is that no category that has argument-structure properties and takes VP-type arguments and modifiers may have special, non-compositional meanings. (One can never tell with non-derived verbs: they do not embed any morpheme that they could be compared to to determine whether they are non-compositional or not.) Note that I am being rather vague about what "argument-structure properties" and "VP-type arguments and modifiers" are; see Borer and the references she cites. Nevertheless, it does appear that there is a significant generalization here that must be accounted for.

It does not appear that Borer's own account of the incompatibility between derived nominals with argument-structure properties and special, non-compositional meanings can be extended to derived verbs. Borer explicitly allows derived verbs to have special meanings, and in fact claims that at least two do, namely civilize and naturalize. I do not know any meaning of civilize that is not also a meaning of civil, nor can I find one in the Oxford English Dictionary (although some are said to be rare). As for naturalize, it appears from Borer's footnote 22 that the special meaning she intends is that in naturalized citizen. However, for people who do not know the term natural citizen (see Borer's footnote 22), it seems more likely that this meaning is part of the adjectival passive naturalized and the derived nominal naturalization, and not the verb naturalize. Google searches return numerous hits for naturalized and naturalization in the context of citizenship, but very few for naturalize (only 5 in the first 50 hits). To the extent that naturalize can have this meaning, it is probably a back-formation from naturalized. In any case, the citizenship meaning seems to be available for natural and any word derived from it, so it does not appear to be correct to locate the special meaning in naturalize.

Borer (2013) also mentions patronize, which can have a meaning of "act condescendly toward." However, this seems to be a straightforward extension of the core meaning "act as a patron to," with the addition "act as a patron to inappropriately." A few other potential cases I have found scanning lists of -ize verbs are authorize, organize, capitalize, and weatherize, but all of these seem to have the same meaning in the stem they are derived from: authorize might be derived from authority, and author apparently has a now-obsolete meaning of one who has authority over others. Organ includes the concept of having a particular function; the "profit by" meaning of capitalize is a simple extension of the meaning of converting something into capital. Weatherize means to make weatherproof, not to make into weather, but the stem weather itself can have this meaning, as in weather the storm.

It therefore appears that there really are no derived verbs with idiosyncratic, non-compositional meanings. Borer's generalization actually covers not just derived nominals, but all derived categories that have argument structure properties and occur with VP-type arguments and modifiers. I further conclude that we do not yet have a good account of this generalization, since Borer's account cannot be extended to derived verbs. I also hope that a systematic investigation of the lexicon of English and other languages can be done to verify whether the generalization is actually correct. If it is, it demands an explanation.

References

Borer, Hagit (2013). Structuring Sense: Volume III: Taking Form. New York: Oxford University Press.

Borer, Hagit (2014). Derived Nominals and the Domain of Content. Lingua 141: 71-96.

Newmeyer, Frederick J. (2009). Current Challenges to the Lexicalist Hypothesis: An Overview and a Critique. In William D. Lewis et al. (eds.), Time and Again: Theoretical Perspectives on Formal Linguistics in Honor of D. Terence Langendoen. Amsterdam: John Benjamins, 91-117.

Scope in Nominalizations

van Hout, Kamiya, and Roeper (2013) discuss a difference in possible scope readings in nominalizations. They observe that sentence (1) can mean two different things:

(1) The election of nobody surprised me.

On what they call the narrow scope reading, sentence (1) means `Nobody at all was elected, and that was surprising.' On the wide scope reading, it means `Of those elected, none of them was surprising.'

In contrast to (1), sentence (2) only has the wide scope reading:

(2) Nobody's election surprised me.

van Hout, Kamiya, and Roeper (2013) view the narrow scope reading as derived by reconstruction, and devise a theory where reconstruction is blocked in sentence (2). The details of this theory are not important here. Rather, I want to suggest that something else is going on in these examples, and that is whether or not the negative quantifier is interpreted as sentential negation. In sentence (2), nobody as the possessor of the subject is preferentially taken to negate the entire clause, such that negation actually negates the main predicate `surprise'. The wide scope reading is the result of negative quantifiers being complex: they consist of an existential quantifier and negation (e.g., Jacobs 1991). If negation is interpreted as sentential negation, what is left as the possessor is an existential quantifier. The reading is then the negation of `someone's election surprised me,' or, `it is not the case that anyone's election surprised me.' This is van Hout, Kamiya, and Roeper's wide scope reading.

The idea is that a negative quantifier as a subject or the possessor of a subject is preferentially interpreted as sentential negation. If we make the nominalization containing the negative quantifier a non-subject, then we can force it to be sentential negation or not by fronting it and either doing negative inversion, or not. A fronted negative phrase plus subject-auxiliary inversion is interpreted as sentential negation; a fronted negative phrase without subject-auxiliary inversion is not interpreted as sentential negation. Consider the following:

(3) With the election of nobody was I surprised.
(4) With nobody's election was I surprised.

The sentences in (3) and (4) only have van Hout, Kamiya, and Roeper's wide scope reading. In contrast, (5) and (6) only have the narrow scope reading:

(5) With the election of nobody, I was surprised.
(6) With nobody's election, I was surprised.

There is no contrast between the election of nobody and nobody's election once we control for sentential versus non-sentential negation. In particular, (6) has the reading that (2) is said to lack, while (5) only has one reading when it should be ambiguous.

If all of this is correct, then the scope facts described by van Hout, Kamiya, and Roeper (2013) do not reveal much about the derivation of nominalizations, and they are not about reconstruction or its lack.

References

van Hout, Angeliek, Masaaki Kamiya, and Thomas Roeper (2013), Passivization, Reconstruction and Edge Phenomena: Connecting English and Japanese Nominalizations. Natural Language and Linguistic Theory 31: 137-159.

Jacobs, Joachim (1991), Negation. In Arnim von Stechow and Dieter Wunderlich (eds.), Semantics: An International Handbook of Contemporary Research, pp 560-596. Berlin: de Gruyter.

Passives with Do So

Hallman (2013) discusses some apparently contradictory data involving English do so. The received view is that do so is incompatible with the passive (Hallman's example 5d):

(1) *These books were left in the classroom, and this cell phone was done so, too.

The typical account of this is that do so is a pro-form and does not have internal structure that can support extraction. The passive subject must move from an object position, but with do so there is no such position. What is surprising and apparently contradictory is that unaccusatives, which are also thought to involve movement from an object position, are compatible with do so (these are Hallman's examples 50a-b):

(2) The river froze solid, and the pond did so, too.
(3) The towels dripped dry, and the socks did so, too.

Unaccusatives pattern with passives in many ways, which has led to the hypothesis that the surface subject of an unaccusative, like the surface subject of a passive, starts out as an object. The fact that do so is compatible with unaccusatives but not passives seems to be problematic for this view.

However, it appears from a web search that do so is in fact compatible with passives, at least for many speakers. The following are some examples found on the web, which do not seem to me to be ungrammatical (though some are a little awkward):

(4) This means that the only the most edible meat is eaten and done so with much chewing as to liquify the food. (http://www.ehow.com/about_4740227_scorpions.html)

(5) For those who do not know Devil Fruits are extremely rare to find and the ones that are found and eaten are done so in mere happenstance unless you know what to look for. (http://shannaro.wordpress.com/2012/11/30/)

(6) I then take notice and observe when the food is brought to table that the meal is picked apart and what is eaten is done so in a controlled and seemingly not pleasurable manner. (http://www.psychologytoday.com/blog/when-food-is-family/201208/reflections-the-2012-olympics)

(7) Every photo taken and every update written is done so with the adoptive parents in mind. (http://godslittlestangelsinhaiti.org/andlifegoeson/2013/07/19/words-of-encouragement-adds-sunshine-to-our-day/)

(8) It is thrillingly written, and done so with the clarity and poignancy of a man who waited 62 years to reveal the full account of his experience, after first being approached by American prosecutors in 1947. (http://theboar.org/2013/04/19/denis-avey-believe-or-not-believe/#.UfJfWRz-nn0)

(9) And I think everyone can agree that some of the most beautiful music ever written was done so in the name of God or gods. (quote attributed to Anand Wilder, http://en.wikipedia.org/wiki/Yeasayer)

(10) The first ``Rosicrucian'' writings, the Fama Fraternitatis, Confessio Fraternitatis and the Chemical Wedding of Christian Rosenkreuz, all when written were done so anonymously and then later traced to be the works of Johannes Valentin Andreae,... (Tobias Churton, http://www.bonisteelml.org/invisible_history_of_rosicrucians.pdf)

From these examples it appears that do so is in fact compatible with any sort of VP: active transitive, unaccusative, passive, and so on. A simple account is that do so is a pro-form for a predicate that takes a subject. The actual predicate is retrieved from context and predicated of the surface subject of do so. This predicate can be a passive or unaccusative one, such that its subject will correspond to an underlying object. For instance, in example (8), done so is replaced with the predicate Lambda x.Exists y. x is written by y.

At the same time, the predicate can only be a predicate with an open subject (a one-place predicate), and no other open positions, so that extraction of other elements is impossible:

(10) *I know which book Mary read, and which book Bill didn't do so. (Hallman 2013, (5a))

As in the traditional account, the predicate has no internal structure in the syntax, and so cannot support a gap. The only gap that is possible is the subject of the predicate itself.

The fact that do so is in fact compatible with passives renders Hallman's conclusions unwarranted and his theory unnecessary. There is no reason to think that passives and unaccusatives do not involve movement in the general case.

References

Hallman, Peter (2013), Predication and Movement in Passive. Lingua 125: 76--94.

Wh-the-Hell: Pair-List Readings with Multiple Questions

Den Dikken and Giannakidou (2002) claim that there is a difference between matrix and embedded questions with wh-the-hell. They say that in a matrix multiple question, a pair-list reading is impossible:

(1) Who the hell is in love with who?

(This is den Dikken and Giannakidou's example 64a, which they give one question mark; they say that a previous publication, a UCLA MA thesis cited as Lee 1994, marked it as completely ungrammatical. I find neither judgment accurate; the sentence is perfect in the right context. See below.)

Sentence (1), according to den Dikken and Giannakidou (2002), only has a single-pair echo reading. In contrast, when it is embedded, it can have a pair-list reading (this is their example 64b):

(2) I am wondering who the hell is in love with who.

Den Dikken and Giannakidou (2002) go on to design a theory of wh-the-hell and questions generally that derives this difference. The details of this theory are not important here, since what I am concerned with is the accuracy of this claimed contrast. I believe a matrix question like (1) can have a pair-list reading, in the right context.

Imagine an Agatha Christie-type murder mystery where the detectives are called to investigate a murder at a country manor. They discover numerous love affairs, love triangles, unrequited loves, and jealousy. After interviewing multiple house guests and family members, one detective turns to the other in exasperation and says, ``Who the hell is in love with who? I can't keep track, have you been making a list?''

In such a context, the sentence in (1) easily has a pair-list reading, as indicated by the follow-up, ``have you been making a list?'' If this judgment is accurate, then the claim in den Dikken and Giannakidou (2002) is not correct, and there is no root/embedded asymmetry to account for.

To verify the accuracy of this judgment, I also looked for such questions occurring on the internet. I found a few that seem to have pair-list readings, as indicated by the context. Here are two:

• Who the Hell is Who? (title of page http://rense.com/general41/skolmdfng.htm, which goes on to list who each person is)
• But who the hell says what? (game where the reader has to guess who produced a quote, with a whole series of them; http://www.sugarscape.com/main-topics/homepage/804060/it’s-one-direction-“who-hell-said-what

Both of these are matrix questions, and both go on to give a list of pairs. Here is another one:

• Caesar starts the ape rebellion which was talked about in the first films, Caesar is the son of Cornelius and Zira who only went back in time because of Brent who went to the future because of Taylor, so who the hell started what? (http://scificolony.canaryzoo.com/Fanzine%20Movie%20Sagas/fanzine%20saga%20planet%20of%20the%20apes.htm)

It is not completely clear how many whos and how many whats there are, but the list of potential whos is quite long. In my judgment there could easily be multiple pairs, and it is quite likely that multiple pairs are intended. And again, the sentence is a matrix question, not an embedded one.

I conclude that den Dikken and Giannakidou's claimed contrast between matrix and embedded questions is not real. Multiple wh-the-hell questions can have pair-list interpretations whether they are embedded or not.

References

den Dikken, Marcel and Anastasia Giannakidou (2002), From Hell to Polarity: ``Aggressively Non-D-Linked'' Wh-Phrases as Polarity Items. Linguistic Inquiry 33: 31--61.

VP Ellipsis inside Islands

Johnson (2001) proposed that VP ellipsis in English involves a first step of VP topicalization prior to deletion. This proposal is heavily criticized by Aelbrecht and Haegeman (2012), but they do not question the data that motivated this proposal in the first place. The most important argument presented by Johnson (2001) is that VP ellipsis is ungrammatical inside islands if the clause it occurs in is non-finite (the dashes represent the elided VP):

(1) *You shouldn't play with rifles because to --- is dangerous. (subject island)
(2) *Mag Wildwood came to read Fred's story, and I also came to ---. (adjunct island)
(3) *Lulamae Barnes recounted a story to remember because Holly had also recounted a story to ---. (complex NP island)
(4) ??Ron wanted to wear a tuxedo to the party, but Caspar couldn't decide whether to ---. (wh-island)

If VP ellipsis requires a first step of topicalization, these facts are explained. Non-finite clauses do not permit topics, as in (5), and topicalization to a higher clause would cross the island boundary, which is also bad (6).

(5) *You shouldn't play with rifles because [play with rifles] to is dangerous.
(6) *[Play with rifles], I am unhappy because to is dangerous.

While the examples in (1) through (4) are indeed unacceptable, I question whether VP ellipsis is generally not permitted inside non-finite islands. The following are some examples that I found on the web and which seem acceptable to me:

(7) ...meaning, I guess, that he is not really healthy enough, because in order to be ---, he needs this surgery.
(8) Loftus, who police say started the attack, was carrying her infant at the time and attempted to hit the other woman with her fist, but in order to ---, she threw her infant to the ground.
(9) Hi, I need to go on a visa run soon and have heard a lot of people disagree as to where the best place to --- is. Any suggestions?
(10) If you're ready to start your explorations, the best place to --- is the free 5-week mini-course, How to Explore.
(11) You have to carry the fire. I don't know how to.

Examples (7) and (8) are adjunct clauses, structurally very similar to (2). Examples (9) and (10) are complex NPs, just like (3). Example (11) is a wh-island, like (4) (wh-islands are frequently quite weak, especially with non-finite clauses, so this is not surprising). Regarding subject islands like (1), I find the following (constructed) example acceptable:

(12) Being arrested once is understandable; twice even; but to have been --- so many times is just unbelievable!

Moreover, when VP ellipsis takes place in a non-finite clause that is embedded in another non-finite clause, it is perfectly acceptable:

(13) To vote Republican is bad; to have to --- is worse.
(14) To play with guns is stupid; to want to --- is just plain dumb.

Since every clause inside the subject island in these examples is non-finite, there should be no landing site for a topicalized VP, without the VP crossing the island boundary. More generally, VP ellipsis seems to be acceptable in non-finite islands, contra Johnson (2001).

If VP ellipsis is actually acceptable in non-finite islands, the main reason for thinking that VP ellipsis is preceded by VP topicalization disappears. The only other reason is that they are both similar in requiring an auxiliary verb, and there are numerous proposals that explain that fact without relating VP ellipsis to VP topicalization derivationally (for instance, the proposal in Bruening 2010).

References

Aelbrecht, Lobke and Liliane Haegeman (2012). VP-Ellipsis Is Not Licensed by VP-Topicalization. Linguistic Inquiry 43: 591--613.

Bruening, Benjamin (2010). Language-Particular Syntactic Rules and Constraints: English Locative Inversion and Do-Support. Language 86: 43--84.

Johnson, Kyle (2001). What VP Ellipsis Can Do, and What It Can't, but not Why. In The Handbook of Contemporary Syntactic Theory, edited by Mark Baltin and Chris Collins. Oxford: Blackwell, pp.439--479.

Confusion about Reciprocals

There is an obvious difference between the literal meaning of an expression and its actual usage. Consider the phrase ``scared to death,'' which everyone knows is rarely used in its literal meaning. When it is, in fact, people usually add the adverb ``literally,'' as in, ``there's no apparent cause of death; the victim seems to have been literally scared to death.'' However, if one were to simply survey attested usage of the phrase in an attempt to determine its literal meaning, one would of course be deceived.

How does one distinguish literal meaning from usage? One could simply ask a native speaker; all native speakers of English, for example, could tell you the distinction between the literal meaning of the expression ``scared to death'' and its actual usage. This does not always work, however. Another example, though, can point to some ways to get at literal meaning. Consider the phrase ``a minute.'' If one were to survey its usage, particularly when spoken to antsy children, as in ``we'll be there in a minute,'' one would conclude that it denoted a time period of quite extended range, perhaps up to one hour. But if one were to limit the context of utterance to those contexts that required greater precision in time durations, one would quickly discover that the phrase's literal meaning is a time period of 60 seconds. One such context is a race: if the TV announcer says, ``the winner beat the next runner by a minute,'' the phrase ``a minute'' means 60 seconds (give or take a few decimal points). Another context is a court of law: if the prosecutor, seeking to establish a timeline, asks me, ``How long did you leave the victim alone?'', and I answer, ``A minute,'' I would be lying if I had actually left the victim alone for two minutes or more.

Hence, it is a simple truism that people use expressions in non-literal ways. Importantly, these non-literal uses in no way require that semanticists posit truth-conditional meanings for expressions that differ from the literal meaning. The meaning of ``a minute,'' for instance, is a time period of 60 seconds; it is in actual usage that this time period extends beyond that. (If you asked someone, they would say that ``We'll be there in a minute'' is literally false if it will take more than 60 seconds to get there.)

For some reason, though, the simple distinction between literal meaning and linguistic usage is frequently ignored in studies of reciprocals. A huge literature now exists examining the possible meanings of reciprocals (e.g., Dalrymple et al. 1998, Beck 2001, Schein 2001, and now Sabato and Winter 2012). Much of this literature (Dalrymple et al. 1998 in particular) surveys attested usage of reciprocals in an effort to figure out what they mean. As with the phrase ``scared to death,'' this method runs the risk of obscuring the actual meaning by not distinguishing that from non-literal usage. This is exactly what has happened; while there is a significant amount of complexity to reciprocals, I believe that they are actually much less complex than they appear to be. The problem is exactly that described above: people use language in non-literal ways.

The kinds of meanings that reciprocals have been argued to have can be illustrated with examples like the following:

(1) The three thugs respect each other.
(2) The three thugs shot each other.

The example in (1) requires that every thug respect every other thug. This interpretation is called Strong Reciprocity, and it seems to be required by stative predicates like respect. This strong interpretation is not necessary with the eventive predicate in (2), which seems to require (minimally) that each thug shot one of the other ones, and got shot by one of the other ones. This interpretation is usually called Weak Reciprocity. I believe that these two interpretations are the only ones that reciprocals have: statives require Strong Reciprocity, while eventives require Weak Reciprocity.

This is not what most of the literature has concluded, however. Dalrymple et al. (1998) introduced a whole range of possible reciprocal meanings in order to account for attested examples like the following:

(3) The third-grade students in Mrs. Smith's class gave each other measles.

In this example, world knowledge tells us that each student can get measles only once, and one of the students must have gotten measles from a non-student, leaving at least one student with no student to give measles to. So, if Weak or Strong Reciprocity were the only meanings a reciprocal could have, sentence (3) could never be true. Yet speakers of English utter and accept sentences like (3). Dalrymple et al. (1998), and researchers after them, therefore added additional meanings for reciprocals.

The obvious alternative, though, is that sentence (3) is literally false, just like scared to death is (almost) always literally false. This is in fact my judgment: it is literally impossible for a group of children to give each other measles. But one can still say ``the children gave each other measles,'' with the meaning that measles passed from one child to another; we just allow two exceptions at either end of the chain (and possibly more, depending on the size of the group). If we switch to a context that requires more precision, though, the literal meaning of the reciprocal comes out. Consider a context where two epidemiologists are talking, and they are trying to determine the source of a measles outbreak. Dr. Q asks how the third-grade students in Mrs. Smith's class got measles. Dr. Z would never reply with the sentence in (3); if he did, he would be saying that measles was not introduced into the class from an outside source, and must have had a spontaneous genesis within the class.

This context makes it clear that the literal, precise meaning of the reciprocal expression is something other than how it is used in (3). In fact, the literal, truth-conditional meaning is Weak Reciprocity, which requires that each member of the subject set act on one other, and be acted upon by one other. That is, each child gave another child measles, and each child got measles from another child. The literal, precise meaning could never be true, which is why in a context that requires it to be, no one would use it. Outside of such contexts, however, people make general statements that have exceptions, they exaggerate, they understate, they speak metaphorically, and so on.

As another example, consider the famous example, ``The people on this island used to eat each other.'' It was probably not true that every single person ate at least one other person and in turn got eaten by someone. For some reason this is thought to be significant in discussions of reciprocals; but take a similarly general statement not involving reciprocals, like ``The people on this island used to eat dodos.'' It was probably not true that every single person ate a dodo; surely some people on the island did not like dodo, or some were vegetarians; the statement is made as a generality, which of course will have some exceptions. The reciprocal statement is no different.

Similar issues are at work in examples involving linear configurations, such as ``The children followed each other into the schoolhouse,'' or ``The acrobats stood on each other's shoulders.'' The members of the subject set at the ends of the lines (or stacks) are exceptions: they participate in the relation in only one way. That they are exceptional is shown by the fact that the size of the set matters, as pointed out by Beck 2001: people would not use ``The children followed each other into the schoolhouse'' when there are only two children, and in fact acceptability goes up with the number of students. Same with acrobats: if there are only two, they have to alternate, first one standing on the other, then they switch. This, it seems to me, points to the actual, literal truth conditions of (eventive) reciprocals being Weak Reciprocity.

Yet another publication has recently appeared in which the distinction between literal meaning and actual usage is not properly made, namely Sabato and Winter 2012. While it may be that I am wrong that there are only two possible meanings for reciprocals, the distinction between literal meaning and usage needs to be carefully kept in mind in trying to figure out what the truth-conditional meaning of an expression is. Many (possibly most) utterances that humans make are not literally true, so just surveying attested usage will not be a good guide to actual meaning.

References

Beck, Sigrid (2001). Reciprocals are Definites. Natural Language Semantics 9: 69--138.

Dalrymple, Mary, Makoto Kanazawa, Yookyung Kim, Sam Mchombo, Stanley Peters (1998). Reciprocal Expressions and the Concept of Reciprocity. Linguistics and Philosophy 21: 159--210.

Sabato, Sivan and Yoad Winter (2012). Relational Domains and the Interpretation of Reciprocals. Linguistics and Philosophy 35: 191--241.

Schein, Barry (2001). Adverbial, Descriptive Reciprocals. Proceedings of Semantics and Linguistic Theory XI, edited by Rachel Hastings, Brendan Jackson, and Zsofia Zvolenszky. Ithaca: CLC Publications.

Direct-Inverse Systems are More Common than People Think: Mandarin Chinese

Direct-inverse voice systems have been described most famously for Algonquian languages, but there are a few other cases in the world, as well. Not many, though. I would like to suggest that they are more common than people think, and linguists should start paying more attention to the grammatical properties of inverse systems.

An example of a direct-inverse voice system is Plains Cree. In Plains Cree, if there are two (or more) third-person arguments in the same clause, one is proximate (unmarked) and the other(s) obviative (marked with the suffix -ah). If the proximate NP is the subject, the direct voice is used (glossed ``3/Obv,'' meaning a proximate (3) acting on an obviative (Obv). If the proximate NP is instead the object, the inverse voice is used (glossed ``Obv/3''). The following examples, from Dahlstrom 1991 (pp64--66), illustrate the direct (1) and the inverse (2):

(1) aya:hciyiniw-ah nisto e:h=nipah-a:t awa na:pe:sis
Blackfoot-Obv three kill-3/Obv.Conj this boy
`this boy had killed three Blackfoot'

(2) ta:pwe: mac-a:yi:siyiniw e:sah nipah-ik o:hi ihkw-ah
truly bad-person kill-Obv/3 this.Obv louse-Obv
`truly the louse killed the evil man'

What characterizes the inverse is a reversal of syntactic prominence: the thematic external argument seems to be hierarchically subordinate to the thematic internal argument, regardless of surface word order. This reversal of prominence is reflected in binding, scope, and other facts; see Bruening 2009. It is important to note that the inverse is not a passive: the external argument has not been removed or demoted to an oblique. Plains Cree also has a passive (or ``indefinite subject'' form), which lacks a thematic external argument:

(3) ki-kosis nipah-a:w
your-son kill-Pass/3
`your son has been killed'

Let's now look at Mandarin Chinese. This language is usually described as basically SVO, as in the following example (from Li 2006):

(4) wo sha-le ta-le
I kill-Asp him-Asp
`I killed him.'

However, other word orders are possible, like OSV or even SOV; these are usually described as some sort of topicalization.

There are also two constructions that I would like to put side-by-side, the ba-construction and the bei-construction (5 is from Li 2006, but 6 comes from my own consultants):

(5) wo ba ta sha-le
I BA him kill-Asp
`I killed him.'

(6) ta bei wo sha-le
he BEI I kill-Asp
`He was killed by me.'

Suppose Mandarin Chinese did not have examples like (4), and all sentences were either like (5) or like (6). Linguists would then describe the voice system of Mandarin Chinese as a direct-inverse system: (5) is the direct voice, with the thematic external argument highest hierarchically, and (6) is the inverse voice, with the thematic external argument subordinate to an internal argument.

This is not how most linguists describe Mandarin; instead, (6) is referred to as a ``passive,'' while (5) is the mysterious ``ba-construction.'' However, (6) has nothing in common with Indo-European-type passives: the thematic external argument has not been demoted or removed, it retains its status as an argument (see Huang 1999). Now, the external argument can be removed, and be interpreted as an indefinite:

(7) ta bei sha-le
he BEI kill-Asp
`He was killed (by someone).'

However, as we saw above, direct-inverse languages like Plains Cree also have a passive, in addition to direct-inverse clauses. The presence of (7) therefore in no way undermines the idea that the bei-construction in (6) is an inverse.

I suggest that a more accurate terminology for Mandarin Chinese would be to call SVO sentences like (4) ``unmarked,'' and (5) and (6) ``direct'' and ``inverse,'' respectively. Linguists should then turn to the study of inverse sentences (and direct ones), both from a theoretical and from a typological standpoint. Constructions like the bei-construction in Mandarin Chinese are very common among Asian languages, and the potential to deepen our understanding of language by taking a new perspective is great.

References

Bruening, Benjamin (2009). Algonquian Languages Have A-Movement and A-Agreement. Linguistic Inquiry 40: 427--445.

Dahlstrom, Amy (1991). Plains Cree Morphosyntax. New York: Garland.

Huang, C.-T. James (1999). Chinese Passives in Comparative Perspective. Tsing Hua Journal of Chinese Studies 29: 423--509.

Li, Yen-Hui Audrey (2006). Chinese Ba. In Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell Companion to Syntax volume 1, pp 374--468. Oxford: Blackwell.

Two-Step Tough Movement?

The proper syntactic analysis of tough movement, illustrated in (2) below, has long been a matter of debate:

(1) It is difficult to get a handle on tough movement.
(2) Tough movement is difficult to get a handle on.

In (2), an object is missing within the non-finite clause, and this gap is related to the NP that appears in the matrix subject position. The NP in the matrix subject position in (2) takes the place of the expletive in (1), where there is no gap.

All analyses agree that the matrix subject in (2) needs to be related to the gap inside the non-finite clause. The point of debate is the nature of this relation. Early analyses, such as object deletion and one-step movement, have been discarded, since Chomsky (1977) showed that tough-movement has all the properties of successive-cyclic A-bar movement. Current analyses can be divided into two categories: those that posit base-generation of the matrix subject, relating it to a null operator that undergoes A-bar movement within the embedded clause (Chomsky 1977, 1981); and those that hypothesize two-step movement of the matrix subject: A-bar movement to the edge of the lower clause, followed by A-movement to the matrix subject position (e.g., Hicks 2009). This second type of analysis has to reject or reformulate the generally assumed ban on improper movement, which rules out A-bar movement followed by A-movement; or it has to hypothesize some way to get around that ban (Hicks 2009).

The two analyses make very different predictions concerning reconstruction effects and other diagnostics of movement. The null operator analysis says that the matrix subject never occupied a position in the embedded clause; it therefore predicts that it will never be able to reconstruct, and should fail diagnostics of movement. The two-step movement theory, in contrast, says that the matrix subject started out in the embedded clause, and so should be able to reconstruct, and should pass other diagnostics of movement.

Pesetsky (1987, 2012) presents data that he argues show that the matrix subject in tough movement can reconstruct for anaphor binding. The following examples are from Pesetsky (2012):

(3) [This aspect of herself] is easy [for Mary to criticize].
(4) [This side of herself] was tough [for John to get Mary to deal with].
(5)*[This aspect of herself] is easy [for [Mary's father] to criticize].
(6)*[This side of herself] was tough [for John to get [Mary's father] to deal with].

Pesetsky presents (5-6) as controls, to show that c-command is necessary for reconstruction and binding to go through.

The problem, though, is that this aspect of herself and this side of herself are picture-NPs. Anaphors inside picture-NPs are known to be exempt from the Binding Conditions, and subject instead to pragmatic phenomena like perspective (Pollard and Sag 1992, Reinhart and Reuland 1993). Pesetsky's examples in (5-6) are ungrammatical not because the intended antecedent does not c-command the anaphor, but because the animate NP father competes with Mary for perspective. Compare the following pair:

(7)*Clinton's wife carried a picture of himself in her purse.
(8) Clinton's car carried a picture of himself on the roof. (Hestvik and Philip 2001)

Wife competes with Clinton for perspective, decreasing the acceptability of the exempt anaphor taking Clinton as the perspective holder that provides its reference in (7). Since Clinton is in a less prominent position (it is only a possessor, while Clinton's wife is the matrix subject and topical), it is difficult to understand the perspective in (7) to be that of Clinton. In contrast, with an inanimate in (8), there is no other potential perspective holder, and the exempt anaphor can take Clinton as the holder of perspective to which it refers. Note that animacy makes no difference to an anaphor in argument position, which strictly requires a c-commanding antecedent:

(9)*Clinton's car backfired/collapsed/exploded behind himself. (Hestvik and Philip 2001)

To show whether c-command is necessary for reconstruction in Pesetsky's tough movement examples, then, we need to use inanimate NPs, or non-specific NPs that do not compete as perspective takers. When we do that, we see that c-command is not necessary at all:

(10) This side of herself will be easy for Sarah Palin's detractors to use against her.
(11) This aspect of herself was tough for Sarah Palin's autobiography to present in a good light.

This means that the "binding" in Pesetsky's examples is not binding, after all, and does not require that the antecedent c-command the anaphor. That being the case, these examples do not show that there is reconstruction in tough movement.

Reconstruction generally does not seem to take place. There is no reconstruction for Principle C (Pesetsky 2012), and reconstruction for variable binding also does not seem to be generally available (see Hicks 2009). Idiom chunks are not a valid diagnostic for movement, because the ones that can undergo tough movement can also antecede pronouns (and participate in control):

(12) Some strings are harder to pull than others. (Nunberg, Sag, and Wasow 1994)
(13) Kim's family pulled some strings on her behalf, but they weren't enough to get her the job. (Nunberg, Sag, and Wasow 1994)
(14) His closets would be easy to find skeletons in. (Nunberg, Sag, and Wasow 1994)
(15) There are a lot of old skeletons in his closets; there are even some new ones in there too.

Idiom chunks are compatible with the null operator theory: they would simply antecede the null operator, the same way they antecede the pronouns in (13) and (15).

The one diagnostic of movement that seems to be conclusive is expletives. Expletives cannot antecede pronouns, particularly null pronouns like PRO:

(16)*There occurred three more accidents without PRO being any medical help available on the premises. (Haegeman 1994)
(17)*It occurred to me that most people are greedy without PRO bothering me that they are.

If expletives appear displaced from the clause they are licensed in, then, they must have moved there.

It is telling, then, that expletives can never undergo tough movement:

(18)*There is hard to believe to have been a crime committed. (Chomsky 1981)
(cf. I believe there to have been a crime committed.)
(19)*It is impossible to stop from snowing in the Himalayas.
(cf. You can't stop it from snowing in the Himalayas)
(20)*It is difficult to think likely that two-step movement is involved.
(cf. I think it likely that two-step movement is involved.)

Postal (1974, page 199) claims that there is a contraint against multiple raising of expletives. Since most expletives are subjects, and tough movement cannot affect subjects, to test them in tough movement we have to first raise them to an object position in (18-20). They would then violate Postal's constraint: in all three examples, the expletive first raises to object, and then undergoes tough movement.

However, this constraint does not seem to be real. An anonymous reviewer for Language has pointed out numerous examples of multiple raising of expletives on the internet, all of which seem to be acceptable:

(21) UNICEF noted that there appear to continue to be extremely low literacy rates among the poorest 20 percent of the population.
(22) There seem to continue to be problems with the multiple ad presentation.

Since there is no such constraint, if the matrix subject started out in the embedded clause, the examples in (18-20) should all be grammatical. The fact that they are not is expected by the null operator theory, but not by the two-step movement theory.

The general lack of reconstruction effects in tough movement, and the ungrammaticality of tough movement of expletives, favors the null operator theory over the two-step movement theory. It also points to there being a ban on improper movement, since that ban rules out a two-step movement derivation of tough movement, and forces something like the null operator analysis.

References

Chomsky, Noam (1977), On WH-Movement. In Peter Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal Syntax, New York: Academic Press, pp. 71–132.

Chomsky, Noam (1981), Lectures on Government and Binding. Dordrecht: Foris.

Hestvik, Arild, and William Philip (2001), Syntactic Vs. Logophoric Binding: Evidence from Norwegian Child Language. In Peter Cole, Gabriella Hermon, and C.-T. James Huang, eds., Long-Distance Reflexives, San Diego: Academic Press, vol. 33 of Syntax and Semantics, pp. 119–139.

Hicks, Glyn (2009), Tough-Constructions and Their Derivation. Linguistic Inquiry 40: 535–566.

Nunberg, Geoffrey, Ivan A. Sag, and Thomas Wasow (1994), Idioms. Language 70: 491–538.

Pesetsky, David (1987), Binding Problems with Experiencer Verbs. Linguistic Inquiry 18: 126–140.

Pesetsky, David (2012), Phrasal Movement and Its Discontents: Diseases and Diagnostics. In Lisa Cheng and Norbert Corver, eds., Diagnostics in Syntax, Oxford: Oxford University Press, to appear.

Pollard, Carl, and Ivan Sag (1992), Anaphors in English and the Scope of the Binding Theory. Linguistic Inquiry 23: 261–303.

Postal, Paul M. (1974), On Raising: One Rule of English Grammar and Its Theoretical Implications. Cambridge, MA: MIT Press.

Reinhart, Tanya, and Eric Reuland (1993), Reflexivity. Linguistic Inquiry 24: 657–720.

Some Observations on Get Passives

Many researchers have suggested that get passives are ambiguous. For instance, Reed (2011) says that get passives are ambiguous between three different structures and interpretations:

(1) That child got hurt.
a. Verbal passive; much like a be passive.
b. Control; equivalent to That child got herself hurt, but with unpronounced PRO instead of herself.
c. Adjectival passive: hurt is an adjective (get can take adjectival complements, as in That child got sick.)

I focus on the verbal passive and the control structure here. The control structure is supposed to have the surface subject interpreted as something like an agent; as such, it can be modified by adverbs like deliberately (Lasnik and Fiengo 1974):

(2) I think that John deliberately got hit by that truck, don't you?

However, there is some reason to doubt the general availability of a control structure for get passives. First, notice that the following sequence makes sense and is not contradictory:

(3) That truck hit Marvin, but Marvin didn't get himself hit by that truck.

This is because the get passive with pronounced himself asserts more than just the corresponding active (or be passive): in addition to the truck hitting Marvin, Marvin did something to bring that hitting event about. So a truck hit Marvin can be true without Marvin got himself hit by a truck being true.

So, if any given example of a get passive could have a control analysis, we would expect the same non-contradictory pattern. This is not the case, however. The following is a contradiction, just like the corresponding sentence with a be passive:

(4) That truck hit Marvin, #but Marvin didn't get hit by that truck.
(5) That truck hit Marvin, #but Marvin wasn't hit by that truck.

Adding deliberately makes the sentence non-contradictory again:

(6) That truck hit Marvin, but Marvin didn't get hit by that truck deliberately.

From this it appears that the control (agentive) interpretation of a get passive is not generally available, but can only be brought about by the addition of something, like deliberately. Without some such element, a get passive is truth-conditionally equivalent to the corresponding active sentence, just like a be passive.

It is also worth pointing out that get does not pattern with raising verbs in truth conditional equivalence, either. Haegeman (1985), for instance, analyzed get as a raising verb like seem. Note the following contrast, however:

(7) That truck hit Marvin, #but Marvin didn't get hit by that truck.
(8) That truck hit Marvin, but Marvin didn't seem to have been hit by that truck.

The raising verb seem adds additional assertive content and so negating it does not contradict a simple active. If get is a raising verb, it is apparently a semantically contentless one.

A second observation casts doubt on the idea that get plus deliberately should be analyzed like get with an overt anaphor. It seems to me that adverbs like deliberately are degraded when there is an animate by-phrase. In (2), above, that truck is inanimate and can occur with deliberately. Contrast that example with an animate by-phrase:

(9) I think that Marvin deliberately got hit by Mike Tyson.

The example in (12) is acceptable only where Mike Tyson is not agentive, but he simply collided with Marvin. Where the by-phrase cannot be so interpreted, it is degraded with deliberately:

(10) Marvin deliberately got injured (??by his co-worker).
(11) Beckham deliberately got suspended (??by league officials) in order to attend his sister's wedding.
(12) The children deliberately got separated (??by the teacher) from their group (??by the teacher).

Note that there is no incompatibility between get plus anaphor and an agentive/animate by-phrase:

(13) Marvin got himself injured by his co-worker.
(14) Beckham got himself suspended by league officials in order to attend his sister's wedding.
(15) The children got themselves separated from their group by the teacher.

This suggests that get plus deliberately is not the same thing as get plus anaphor, as the control analysis assumes (or at least as Reed's version of it does).

I will make one last observation, which is not obviously related to the ones above. This is that sentential subjects are not very good in get passives, although they are fine with be passives and with raising verbs:

(16) *That the world is round got ignored for centuries.
(17) That the world is round was ignored for centuries.
(18) That the world is round seems to have been ignored.

(19) *That the world is round got shown conclusively in 1522.
(20) That the world is round was shown conclusively in 1522.

(21) *That these nouns behave differently got expressed/captured by this formulation of the rule.
(22) That these nouns behave differently was expressed/captured by this formulation of the rule.

This doesn't seem to be a semantic restriction, since the fact that is fine:

(23) The fact that the world is round got ignored for centuries.
(24) The fact that Columbus miscalculated the circumference of the earth got mixed up with the incorrect notion that medieval Europeans thought that the world was flat.

Sentential subjects are also bad with extraposition it:

(25) *It got expected/insisted/reasoned/predicted that the Giants would win the World Series.
(26) It was expected/insisted/reasoned/predicted that the Giants would win the World Series.

Apparent PPs are fine, but they are probably NPs; this one was found on the internet: When I vacuum, under the bed gets cleaned too.

References

Haegeman, Liliane (1985), The Get-Passive and Burzio’s Generalization. Lingua 66: 53–77.

Lasnik, Howard, and Robert Fiengo (1974), Complement Object Deletion. Linguistic Inquiry 5: 535–571.

Reed, Lisa A. (2011), Get-Passives. The Linguistic Review 28: 41–78.

Non-Question Uses of The Hell

Since Pesetsky (1987), wh-phrases with the hell, as in (1), have been extensively studied in the generative literature:

(1) Who the hell is she talking to?

None of this literature, to my knowledge (other than a footnote in Huang and Ochi 2004), ever mentions other uses of the hell. The purpose of this post is to set out some of the data. The syntactic distribution of the hell turns out to be quite limited.

There seem to be two non-question uses of the hell. The first is exemplified by the following:

(2) The hell you say!
(3) The hell I will! (responding to other person's request/command)
(4) The hell she did! (responding to other person's report)

In this use, the hell seems to attach to the left of a finite clause, and vehemently denies the validity of the proposition expressed by the clause. This use seems to be restricted to matrix clauses:

(5) *She said that the hell she will. (OK as quote: She said, ``The hell I will!'')

Note that the subject of the clause can be any person---first, second, third---, as exemplified by (2) through (4). (Plural is also possible: The hell we will! The hell they did!)

The second use has the hell to the left of a directional prepositional phrase (or particle):

(6) Get the hell out of here!
(7) I got the hell out of there.
(8) Get the hell into bed!
(9) The fox ran the hell out of the room.
(10) I drove my car the hell away from there.

Phrases other than PPs are not allowed:

(11) *I fled the hell the scene.
(12) *I got the hell lost.
(13) *I ran the hell as fast as I could.
(14) *I hope the hell (that) she's not there. (not to be confused with I hope to hell that...)
(15) *I wonder the hell where she went.

The hell cannot come before the verb, interspersed with auxiliaries:

(16) I was running the hell away when...
(17) *I was the hell running away when...
(18) *I the hell was running away when...

Although the hell seems to be attached to the prepositional phrase, it does not move as a unit with it. It cannot front with the PP in locative inversion, for instance:

(19) Out of the room ran the fox.
(20) *The hell out of the room ran the fox.

(It also can't be stranded: *Out of the room ran the fox the hell.)

It also can't front in a wh-question or relative clause, although stranding the preposition is fine:

(21) This is the person that he ran the hell away from.
(22) *This is the person the hell away from whom he ran.

Rightward shift seems to be possible:

(23) ?I drove my car on Thursday the hell away from there.

At this point I have no theory to offer of this peculiar distribution. (Few other phrases attach only to PPs; one is right, and that can move with the PP: Right out of the room ran the fox.) It is also unclear whether these three uses of the hell have anything in common, other than some kind of expressive content.

References

Huang, C.-T. James and Masao Ochi (2004). Syntax of the Hell: Two Types of Dependencies. Proceedings of NELS 34, ed. K. Moulton and M. Wolf, 279–294. Amherst, MA: GLSA Publications.

Pesetsky, David (1987). Wh-in-Situ: Movement and Unselective Binding. In Eric Reuland and Alice ter Meulen (eds.), The Representation of (In)definiteness. Cambridge, MA: MIT Press, 98–129.

Two Constraints on Fake Indexicals

(Note: I wrote this blog post in 2009, but am only posting it now. Other publications may have rendered these two observations superfluous in the meantime, but I am posting them anyway in case they are still relevant.)

Kratzer (2009) proposes a theory of fake indexicals, 1st and 2nd person pronouns used as bound variables, as in the following examples:

(1) I am the only one who takes care of my children. (bound reading: no one else takes care of their own children)

(2) I am the only one who remembers our first meeting. (bound reading: no one else remembers the meeting between them and the addressee or salient other individual)

(3) I am the only one who thinks someone criticized my paper. (bound reading: no one else thinks someone criticized their paper)

In Kratzer's theory, the pronouns in (1-2) start out as minimal pronouns, without any features, and get their features via Agree with v (the head that introduces the external argument of the verb; in (1), this would be the head that introduces the external argument of takes care of). The head v can be inserted with 1st or 2nd person features, which then get transmitted to the pronouns and spelled out. Since the subject is a relative pronoun (who), it is compatible with 1st or 2nd person features, with no clash. In the long-distance case in (3), there is a conflict between the local subject (someone) and 1st person features on v; so Kratzer hypothesizes that in this case, the pronoun has 1st person features to begin with, but a context-shifing lambda-operator can be inserted to bind 1st person to the subject of the higher verb (thinks).

The purpose of this blog post is to point out two generalizations that are not captured by Kratzer's system. In this system, there is no syntactic relation between the 1st person pronoun in the matrix clause (I am the only one...) and the fake indexical in either the local or the long-distance case. Presumably, a condition on v being inserted with 1st person features as in (1-2) is that the context must involve the speaker. However, this is not good enough. There must be an occurrence of the 1st person pronoun in the same sentence, and it is not good enough for the pragmatics to implicate the speaker, as the following examples show:

(4) You see before you the only person who can lick my eyebrows. (*bound reading)

(5) Yours truly is the only person who watches my children. (*bound reading)

(6) Yours truly is the only one who thinks someone criticized my paper. (*bound reading)

In (4), you see before you clearly evokes the speaker. However, the bound reading is impossible with a first person possessive pronoun; it is only possible with a third person pronoun. Similarly, in (5) and (6) yours truly refers to the speaker, but again the bound reading of possessive my is not possible. The generalization is that there must be an explicit first person pronoun in the sentence. Kratzer's theory does not capture this generalization. In fact, this generalization is very difficult to capture in a syntactic way at all; in most theories, there is no direct syntactic relation between the matrix pronoun in (1-3) and anything in the relative clause.

In addition, Kratzer's theory fails to capture a directional asymmetry in mismatches between singular and plural pronouns. In (2), I in the matrix clause followed by our in the embedded clause can have a bound reading, but the reverse order does not allow a bound reading:

(7) We are the only ones who watch my children. (*bound reading)

This directional asymmetry was noted briefly by Rullmann 2004 (her example 19).

Both of these generalizations will be very difficult for any syntactic theory to capture, since, as noted above, there is no apparent syntactic relation between the subject of the matrix clause and anything in the relative clause. I have no suggestions to make, but simply point out the two generalizations, as they are important ones that must be captured by an adequate theory of fake indexicals.

References

Kratzer, Angelika (2009). Making a Pronoun: Fake Indexicals as Windows into the Properties of Pronouns. Linguistic Inquiry 40: 187-237.

Rullmann, Hotze (2004). First and Second Person Pronouns as Bound Variables. Linguistic Inquiry 35: 159-168.

A-Adjectives Again: Response to Goldberg

Adele Goldberg has responded to my earlier post (3/2/2011) on A-adjectives here:

http://www.princeton.edu/~adele/Princeton_Construction_Site/Publications_files/a-adjsnotpps.pdf

She responds to the points I made to argue that A-adjectives are actually PPs, and brings up more data, some old, some new, to argue that they are adjectives, not PPs. Here I respond to all of her points, and show that, while most of the arguments are simply inconclusive, there still is positive evidence that A-adjectives are PPs and not adjectives.

First, Goldberg criticizes my earlier post for likening A-adjectives to PPs like at ease and on fire, because, according to her, "(t)hese cases are unusual, in that they necessarily involve bare Ns instead of NPs, and they pattern with adjectives" according to some of the tests given by Goldberg that are reproduced below. However, the analogy to such PPs was purposeful: if A-adjectives are PPs, they are PPs like at ease and on fire in exactly the way Goldberg says: they do not allow phrases following the a- part, but only bare nouns/stems. As we will see below, in every way A-adjectives pattern with such PPs. If one wants to argue that A-adjectives are adjectives, then one will have to also conclude that PPs like at ease are adjectives. In addition, I will also show that A-adjectives pattern with one-word PPs, what are often called "particles," like on, off, out. Again, one would have to conclude that these particle-Ps are actually adjectives, too. And yet there are still ways they pattern like PPs, and unlike adjectives.

To begin, Goldberg produces the list below of ways in which A-adjectives pattern with adjectives, and not with PPs. I reproduce each point verbatim in italics, and then respond to the point. Note that Goldberg says that PPs with a bare N (like at ease) pattern with A-Adjectives on all but point 5, but in fact they pattern the same on that point as well.

1. Semantically, they [A-adjectives] necessarily modify a property of a noun like other adjectives and unlike prepositional phrases.

It is not clear to me what this even means. Let us consider uncontroversial PPs like on the table that specify location. How is a location not a property of a noun? Or PPs like with brown hair: again, isn't this a property of a noun? Postnominal PPs, in particular, seem to predicate properties of the noun they follow, in a straightforward way (the book on the table, the girl with brown hair, the man from Tulsa). Additionally, A-adjectives, and PPs, are predicative when used predicatively, as in The light is aglow/on fire/on. That is, they predicate a property of a noun phrase. There is no semantic difference between adjectives and (at least some) PPs in how they predicate properties of nouns.

However, this does bring up another point, one related to others below. This is that A-adjectives do not like to be used attributively without being in a restrictive relative clause or modified in some other way. The first fact usually remarked upon about A-adjectives is that they do not appear prenominally. But they don't like to appear postnominally by themselves in a restrictive function, either, as Goldberg points out below:

(1) *the asleep man, ??the man asleep

Unless they are modified in a way that makes them predicative, including putting them in a relative clause:

(2) anyone still asleep, the man who was asleep

So, the question is not why A-adjectives do not like to appear prenominally, but rather, Why do they resist being used attributively (as restrictive modifiers) without syntactic help? They resist this function both prenominally and postnominally. Note that the same holds of PPs with a bare N like at ease or on fire:

(3) *the on fire man, ??the man on fire
(4) anyone still on fire, the man who was on fire

It also holds for particle-Ps:

(5) *the on light, ??the light on
(6) any light still on, the light that is on

I do not have an answer to this question, but return to it below. I think it very important that we ask the question this way, rather than focusing on the inability of A-adjectives to occur prenominally.

2. Phonologically, they are inseparable units like adjectives and unlike prepositional phrases.

To this I respond, So what? All this point indicates is that the division between morphology and syntax is not at all clear-cut (something that everyone knew anyway). Syntactically, PPs like on fire and at ease are inseparable, too: *It's fire that he was on. (vs. It's crack that he was on.)

3. The verb seem provides a classic test for adjective status, and readily occurs with the A-adjectives but not with prepositional phases (Lakoff 1970; Jackendoff 1972):
a. The child seemed alive/afraid/afloat/alone/aghast.
b. *The child seemed on the table/at two o’clock.

Everyone seems to have concluded that seem only allows adjectives, but that's simply not true (many examples here taken from the web):

(7) NPs: that seems just the thing/just the ticket/just the place to rest; seems just the opposite to me
(8) PPs with bare Ns: those people seemed at ease/on fire/at war/on target/under control
(9) PPs with phrasal NPs: even Charlie seems at a loss; tumors seem on the rise; they seem on the same page; he seems under the weather; it seems beneath the positive things you are doing; win seems within Earnhardt's grasp; post seems within the charter to me; it seems within his character; he seems on his way out; they seem on a collision course; my teen seems on the fringe; school reform seems on a roll

The examples in (9) all have fully phrasal PPs as complement to seem. Many are admittedly somewhat idiomatic, but they are still PPs, and some of them are not idiomatic at all (within the charter, within his character).

Note also that the PPs that are supposed to only allow bare Ns do admit some modification, and can still appear as the complement of seem:

(10) Home ice still seems within reasonable reach for Gophers (http://www.startribune.com/sports/blogs/81601072.html)

Also, particle-Ps can appear as the complement of seem:

(11) The light seems on/off/out.

Seem, then, does not provide a clear diagnostic for adjectivehood.

4. A-adjectives can be conjoined with uncontroversial adjectives, like other adjectives and unlike prepositional phrases,
a. The man was quiet and afraid/alone.
b. ??The man was quiet and on the table.

It is well-known that different syntactic categories can be conjoined, so conjunction shows nothing (see, e.g., Sag et al. 1985). I find Goldberg's (b) example fine, especially if the PP and adjective are reversed:

(12) the man was in the room but quiet;
(13) the man was next to me and looming;
(14) the man was near the baby and crazier-looking than ever; (and so on)
(15) Pat was healthy and of sound mind. (Sag et al. 1985, example 2c)

Again, this test is simply inconclusive.

5. Like many (but not all) other adjectives but unlike prepositional phrases, afraid (if not other A-adjectives) can occur with of phrase complements:
a. afraid of the man

PPs can take of-complements (or other PP complements):

(16) at risk of fire; on top of spaghetti; in front of the house; within reach of more people; on board with us;

Some A-adjectives even take NP complements, unlike adjectives:

(17) aboard the ship, athwart the deck

Again, particle-Ps pattern the same way: off (of the table), off the table.

6. A-adjectives do not readily appear after nouns except if they have a complement and/or an
intonation break (a), just like simple adjectives (b), but unlike prepositional phrases (c):
a. *The man asleep escaped the police. (postnominal a-adjective)
The man, asleep on the floor, escaped the police.
b.*The man short had escaped the police. (postnominal (non a-) adjective)
The man, short even with his boots on, escaped the police.
c. The man under the bed escaped the police. (postnominal PP)

I addressed this point briefly above. Again, PPs with bare Ns pattern exactly like the A-adjectives (The man at risk ??(of infecting others) eluded the authorities).

More importantly, A-adjectives and PPs with bare Ns do NOT pattern with adjectives in this respect, because the actual property here is ability to be used attributively. Adjectives can be used attributively without any modification (and when they are so used, they appear before the noun); A-adjectives and PPs with bare Ns cannot, nor can particle-Ps, as shown above. Fully phrasal PPs can, patterning with adjectives (except in their position). So, point 6 is simply inconclusive: A-adjectives (and PPs with bare Ns and particle-Ps) pattern with neither category.

Summary of the first six points: None of them indicate that A-adjectives are adjectives; all are consistent with them being PPs. All of them indicate that A-adjectives are just like PPs with bare Ns (like at ease) and particle-Ps. If A-adjectives are adjectives, then so are PPs with bare Ns and particle-Ps. But this conclusion seems silly: such PPs have an obvious P head, and in PPs with bare Ns, pretty much any P can appear in such phrases (see the range of examples above). Moreover, they all allow right-modification, which does seem to be limited to PPs (see below).

Goldberg then turns to the arguments I gave in my earlier post that A-adjectives pattern with PPs. The first one is the small clause complement of have fond memories of. Goldberg appears to be correct that this does allow adjectives (I have fond memories of him sober), so I was wrong that this context shows A-adjectives patterning with PPs.

The second, and most convincing argument in my opinion, is right-modification (also straight and clear). A-adjectives pattern with PPs in allowing right-modification, as in fall right asleep. Goldberg's response to this is to point out that not all A-adjectives allow it. The following are her judgments ("G10" means her example 10):

(G10) a. ??It became right afloat.
b. ??He became right afraid.
c. ??He was left right alone.

I personally find (G10c) reasonably good. I agree that (G10a) and (G10b) are odd, but the point remains that most (if not all) other A-adjectives DO admit right-modification. Goldberg has not indicated any way to reconcile this fact with the claim that A-adjectives are adjectives. In particular, she has not shown that anything that is uncontroversially an adjective allows right-modification. So far as I am aware, it is true that only PPs allow right-modification. ((G10a) seems odd because afloat seems to mostly be stative, not inchoative: it became afloat is itself odd. However, this one seems better: ?It bobbed right afloat. Moreover, I think right afraid could be possible, as in ?They slapped him right afraid, which contrasts sharply with *They slapped him right scared.)

Goldberg also attempts to defuse the ways in which I showed that A-adjectives do not act like adjectives. First, I claimed they could not appear with the to form kind-denoting NPs. Goldberg has found two such uses:


(G11) a. Books are written by the alone for the alone.(COCA corpus)
b. The oppressed will come to you for shelter, and the afraid will find safety with you.
books.google.com/books?isbn=0791423913

She also says that,

...the ability to be used as a kind-denoting NP is not a good test for adjectives, since not
all clear adjectives (Class I) pass it:
c. ??The full/huddled/pinkish will inherit the earth.
Thirdly, if Class III items [A-adjectives] are dispreferred relative to Classes I and II to some degree as they may
well be, it could well be because the kind-denoting NP construction prefers adjectives that can
readily appear in NPs.

I agree that her two cases sound reasonably good. However, it seems to me that some PPs with bare Ns would sound just as good, as would particle Ps:

(18) Book are written by the under control for the out of control.
(19) The on will succeed and the off will not.

Then, this test is simply not conclusive. Since I do not really understand the restrictions here, I will leave it at that.

My second point was that A-adjectives do not take -er comparatives. Goldberg says the following:

The comparative ending requires the adjective be gradient, and most a-adjectives are inchoative and therefore not gradient. Again, all adjectives obey this semantic restriction:
(G12) a. *deader, *sunker
The phonology of these adjectives may also play a role, explaining the few cases that are gradient
(such as afraid) since Class II adjectives [those that start with A but are not A-adjectives; BB] also resist the comparative ending:
b. *absurder, *acuter, *aloofer,
Thus the lack of appearance with -er cannot be taken as evidence for underlying PP status without overgeneralizing the structure to Class II adjectives.

I think that Goldberg's claim about gradience is simply false: deader than a doornail is fine, and most adjectives, including most A-adjectives, can form more comparatives, which should be impossible if they cannot be gradable: more alive, more ablaze, more alone, more aghast, etc. Nevertheless, it does seem to be true that the phonology might be playing a role, since absurd, acute, astute, and aloof also do not form -er comparatives. I will concede that this point is not conclusive, either. (However, I do note that a few hits turn up on Google for absurder and acuter, and quite a few for astuter, like "An astuter man than the French emperor would have found it difficult to resist the system of delicate flattery..." from The Living Age on Google Books. In contrast, I find nothing for *afraider, *aloner, *aghaster, *ablazer. So, I suspect that the claimed phonological restriction is nonexistent, and -er comparatives really do distinguish A-adjectives from real adjectives, but I will not push this point.)

My third point was that A-adjectives do not form adverbs with -ly. Goldberg says this:

This again is not a good test for adjective status since many [regular adjectives] would fail it:
(G13) a. *Sacredly, *fatly, *bigly, *tinily

First, sacredly does exist: there's a book called African American Satire: The Sacredly Profane Novel. However, it is true that some adjectives do not form adverbs with -ly. Nevertheless, A-adjectives fail to do so SYSTEMATICALLY. All A-adjectives are absolutely impossible with -ly, as are PPs with bare Ns and particle Ps:

(20) *at-ease-ly, *at-war-ly, *under-control-ly
(21) *on-ly, *off-ly, *outly

In contrast, the badness of *fatly, *bigly, *tinily seems to be arbitrary. Close synonyms and phonologically similar items allow -ly:

(22) synonyms: fat: stoutly, grossly, etc.; big: hugely, largely; tiny: minutely, infintesimally, diminutively
(23) phonologically similar: fat: patly ("answered patly"); big: sickly, floridly (any adjectives in -ig?); tiny: tinnily, brinily

Since there is no phonological or semantic reason for the badness of -ly adverbs with fat, big, tiny, the restriction just seems to be arbitrary. On the hypothesis that the A-adjectives are adjectives, the fact that they do not form -ly adverbs would have to be arbitrary, too. (Note that there is no phonological restriction as there might be with -er comparatives, because acutely, absurdly, astutely are fine.) That is, it would be pure coincidence that none of them happen to form -ly adverbs. I find this highly unlikely; the restriction appears to be entirely systematic, and once again the A-adjectives pattern with PPs with bare Ns and particle-Ps.

Let me summarize all of the available evidence here:

(24) Ways in which A-adjectives pattern with adjectives: None

(25) Ways in which A-adjectives pattern with PPs: right-modification, lack of -ly adverbs, (particle shift, -er comparatives)

(26) Inconclusive: semantics; phonology; seem; conjunction; of-phrase complements; small clauses; kind-denoting NPs; (-er comparatives)

(27) Ways in which A-adjectives act like neither adjectives nor PPs: inability to be used attributively without modification

Most of the arguments are simply inconclusive. However, there is NO positive evidence for the hypothesis that A-adjectives are adjectives. In contrast, there is positive evidence that A-adjectives are PPs: they allow right-modification, and they do not form -ly adverbs. (I have also added particle shift in parentheses in the table: A-adjectives participate in particle shift, like other one-word PPs. I think we should probably also include the lack of -er comparatives.)

Moreover, in every way A-adjectives pattern with PPs with a bare N and with particle-Ps. I take this to indicate that those three things have essentially the same status. Note in particular that none of these three things like to be used attributively without further modification. In this they pattern with neither adjectives nor PPs. I would take this to indicate that they are a special kind of PP, but one might also conclude that they are some other category altogether, or something with mixed status.

I will leave it at that, but I would like to reiterate that I think we should reframe the question about A-adjectives. The most striking fact about them is NOT that they cannot appear prenominally; rather, they seem to resist being used attributively at all, without further modification. Once they are modified, they pattern with all phrasal nominal modifiers and prefer to appear after the noun. So, the real question we should be addressing is the following: Why do A-adjectives, PPs with a bare N, and particle-Ps resist being used attributively without modification? Once we have answered that, we might have an answer to the learnability issue as well.

REFERENCES

Sag, Ivan A., Gerald Gazdar, Thomas Wasow, and Steven Weisler. 1985. Coordination and How to Distinguish
Categories. Natural Language and Linguistic Theory, 3:117–171.