LSA Institute 2017/ July 28, 2017
class 7 notes c. The pizza was believed [ __ to have been devoured __].
1. Passive meets ECM (continuation) d. The cat was considered [ __ to have been let __ out of the bag].
... have exact counterparts with main-clause predicates that are syntactically active rather than As we noted at the end of the previous summery, when an ECM verb with an infinitival passive, and belong to a smallish (but not too tiny) class that includes the verbs seem and appear complement is passivized, the embedded subject must move: as well as adjectives such as likely and certain:
(1) Passive of verbs with ECM and Small-clause complements (4) a. Mary was believed [ ø __ to have left the room]. CP b. *It was believed [CP ø Mary to have left the room]. a. Mary seems [ __ to have left the room].
As we also noted, the embedded clause in an ECM infinitival like (1a) may itself be passive: b. The shit appears [ ___ to have hit the fan]. (2) Passive of a verb with a passive ECM complement The food is believed [TP __ to have been devoured __ (by the lion)] c. The pizza is likely [ __ to have been devoured __]. And finally, we noted that if a full-sentence idiom (i.e. one in which the subject is part of the idiom) is embedded as an infinitival complement of an ECM verb, and the ECM verb is passivized, the subject of the idiom moves for case reasons, just as in (1a) — and, unsurprisingly, d. The cat is certain [ __ to have been let __ out of the bag]. the idiomaticity is preserved: Just as in the passive sentences in (3), it looks like the subjects of the main clauses in (4) are in (3) a. Sue believed [the shit to have hit the fan on Thursday]. their surface positions for entirely syntactic reasons. The verbs seem and appear and the b. The shit was believed to have hit the fan on Thursday (by John). adjectives likely and certain take only the embedded clause as arguments. They are single- ('The shit hit the fan.' = '
LSA Institute 2017
It comes as no surprise now that if a Raising predicate subcategorizes for a finite CP (in addition to the possibility of an infinitival clause seen in (4), there is no Raising — since the subject of the But the big difference here is that in example (11a), Mary is doing something called trying, and in embedded finite clause does not move to the main clause to get case. It receives nominative case example (11b), Bill is doing something called promising or making a promise — while in Mary in its own clause. In the main clause, EPP can be satisfied by use of the dummy element it. seemed to leave the room, Mary is not doing something called seeming. Though in some sense Mary is the subject of both clauses in (11a), and Bill is the subject of both clauses in (11b), this is (7) Raising predicates with a finite CP complement not because the main-clause subject has moved from the embedded clause. Mary and Bill are not a. It seems [that Mary has left the room]. the subject of the main clauses for purely syntactic reasons in these examples. They actually b. It appears [that the shit has hit the fan]. receive a θ-role from the main-clause verb, and seem to control a silent copy of themselves, a c. It is likely [that the pizza has been devoured]. silent pronoun usually notated as PRO, in the embedded clause. For this reason, examples like d. It is certain [that the cat has been let out of the bag]. (11a-b) are called Control constructions, and verbs like try and promise are called "control predicates". Likewise it comes as now surprise that we cannot fail to raise the subject of the embedded clause when it is an infinitival, even if we satisfy EPP in the higher clause by using a dummy it — since (12) a. Mary tried [PRO to leave the room]. the embbed subject will not receive case, and will violate the case filter: b. Bill promised [PRO to read the book].
(8) Raising predicates with an infinitival complement but no raising of the subject It should come as no surprise now that the behavior of idioms in control constructions is a. *It seems [Mary to have left the room]. completely different from their behavior in Raising environments. The shit in (13a) cannot be b. *It appears [the shit to havr hit the fan]. understood as part of the idiom the shit hit the fan because it was never merged in the embedded c. *It is likely [the pizza to be devoured]. clause. Furthermore, unless one imagines a world in which...umm...excrement can have d. *It is certain [the cat to be let out of the bag]. intentions and perform actions, the sentence lacks even a non-idiomatic reading, since try takes an external argument to which the θ-role Agent is assigned (unlike Raising predicates like seem, Raising verbs are behaving just like the passive of ECM verbs in all these respects: which take no external argument at all). Likewise for the cat in (13b). Only in a world where cats can make promises is the sentence interpretable at all, and the idiomatic reading is (9) Passive of ECM predicates with a finite CP complement unavailable. a. It is believed [that Mary has left the room]. b. It believed [that the shit has hit the fan]. (13) a. *The shit tried to hit the fan. (no idiomatic reading) c. It is believed [that the pizza has been devoured]. b. *The cat promised to be let out of the bag. (no idiomatic reading) d. It is believed [that the cat has been let out of the bag]. In the fuller version of this class, we would also focus on verbs whose direct object controls PRO (10) Passive of ECM predicates with an infinitival complement (Mary persuaded John [PRO to leave the room]), which are the "false twins" of ECM predicates. but no raising of the subject But these are topics for another day. a. *It is believed [Mary to have left the room]. b. *It is believed [the shit to havr hit the fan]. c. *It is believed [the pizza to be devoured]. 4. VP-internal subjects once again d. *It is believed [the cat to be let out of the bag]. This is more or less a side remark, but I couldn't think of a better place to put it than this one, so here it is. 3. Raising vs. Control In Raising, a nominal subject of an infinitival complement that needs case moves to a higher In the longer version of this class... Spec,TP position where it can receive the case it needs. A very similar analysis can explain why the external argument generated as the specifier of VP must also move to Spec,TP in English: the ...we would spend lots of time on infinitival constructions that look like Raising constructions, but subject raises to receive case and satisfy EPP. are not — for example (11a-b): (14) a. The lion will devour the pizza. (11) False twins of Raising constructions b. *It will the lion devour the pizza. a. Mary tried to leave the room. b. Bill promised to read the book.
At first glance, sentences like these sound a lot like Raising constructions, swapping verbs like try and promise for verbs like seem and appear. And it is true that if Mary tried to leave the room, Mary would be the one leaving the room, if her efforts succeed — much as if Mary seemed to leave the room, Mary was the one leaving the room, if appearances do not deceive. LSA Institute 2017
take only internal arguments, and no external argument, are called unaccusative (a term due to David Perlmutter and Paul Postal).
(18)
Contrast a verb like phone in (19): In a deep sense, tense elements like will, have and silent [±Past] can be seen as Raising predicates that subcategorize a VP. That is why, for example, full-sentence idioms such as the shit hit the (19) Bill will phone. fan can have any kind of tense element intervening between the shit and hit the fan. The actual idiom is the VP [VP the shit [V' hit the fan]], from which the specifier raises for case reasons to Bill here is an agent, and therefore is plausibly an external argument (starting life as a specifier of form Spec,TP: VP), unlike the ice in (18). A verb that takes an external argument, but no internal argument, is called unergative. An unergative analysis of phone in (19) makes sense, because Bill in (19) (15)a. The shit hits the fan (on Thursdays). should occupy the same syntactic position, all things being equal as it does in sentences like Bill b. The shit will hit the fan. will phone Tom. We could draw this position as follows: c. The shit has hit the fan. d. The shit is hitting the fan. (20)1
5. Unaccusative vs. Unergative
Let us return to Raising from infinitival clauses. Recall the properties of these predicates:
(16) Raising predicates 1. do not assign accusative case; 2. do not assign an external θ-role 3. subcategorize for an infinitival clause (of a particular type).
There is no obvious reason why a predicate with properties 1 and 2 should also have to have property 3. Shouldn't we also find predicates that (a) fail to take an external argument; (b) do not Why do unaccusative and unergative verbs sound alike, i.e. why do both occur in sentences that assign accusative case; but (c) subcategorize for something other than an infinitival clause, e.g. an sound like "NP Tense V"? The answer is: case and EPP. Because, in English at least, the VP- NP? internal NP must move to Spec,TP whether it starts out as a complement or a specifier, the result will always sound like "NP Tense V". One might reasonably ask how a child acquiring language An obvious candidate for such a predicate is verbs like melt when it takes only one argument:
(17) The ice will melt. 1 If you think back on our discussion of Merge, the difference between (18) and (20) actually does not make any If a patient is always an internal argument, then the ice in (17) must be an internal argument that sense as drawn, unless we allow V in (20) to merge reflexively with itself, which we can diagram with the non- branching V'. In fact, let us assume that for the purposes of our class — though there is a better theory in which we moves to Spec,TP from the complement of V position, not from Spec,VP. There is evidence that separate the VP into two separate phrases, a vP whose head (v) takes the external argument as its specifier and a VP this analysis is correct, to which we turn shortly. Verbs like melt (or seem, for that matter) that as its complement — where V takes the internal argument as its complement. Click on http://bit.ly/1IsQ89c to see the structure I have in mind (or check the slides). But justifying this would take a class or so in itself, so for now I assume (18) and (20). LSA Institute 2017 could tell unaccusative verbs apart from unergatives, if the sentences that contain them sound the same. (23) a. We closed the door. b. The door closed. One answer that appears to be true, or shockingly close to true at the very least, is provided by what the linguist Mark Baker called the Uniformity of Theta Assignment Hypothesis (UTAH). (24) a. The waiter dropped a glass. Adapting his statement of UTAH slightly to to conform to the presentation of things in this class, b. A glass dropped. we can state it as follows: (25) a. We slid the soap into the closet. (21) The Uniformity of Theta Assignment Hypothesis (UTAH) b. The soap slid into the closet. Identical thematic relationships between items are represented by identical structural relationships at the point at which they are first merged. A few pairs are semantically similar, but show a vowel alternation: (adapted from Baker 1988) (26) a. The stagehand raised the curtain. UTAH dictates that if an argument bearing the θ-role Patient, for example, is sometimes assigned b. The curtain rose. to the complement position of V, that θ-role is always assigned to the complement position of V. Likewise, if an argument bearing the θ-role Agent is sometimes assigned to the specifier position (27) a. You must lay the object on its side. of V, that is the position where it is always assigned. Furthermore, these assignments appear to b. The object must lie on its side.2 be cross-linguistically uniform — perhaps completely identical across the languages of the world (though, as you might imagine, there are complications, questions, and lots of open UTAH- (28) a. Mary will set the lamp on the table. related puzzles that arise when we actually examine the wealth unfamiliar phenomena in the b. The lamp will sit on the table. world's languages). Others show completely different phonology, but still show the same apparent semantic identity The plausibility of UTAH coupled with the universality of θ-role assignment principles is except for the extra q-role. An example is bring and come. A curiosity of this pair is the fact that supported by much work on language acquisition, starting with Marantz's (1982) experiment in we can see the semantic identity by noting their parallel behavior in idiom. which 3, 4 and 5 year-olds readily understood that when an experimenter pounds a surface with her elbow and describes the action as moaking, the experimenter can be said to have moaked the (29) a. The mailman brought the package yesterday. surface. The surface did not moak the experimenter! b. The package came yesterday.
Let us now see how UTAH works to guide the child (and us) to a distinction between (30) a. Mary brought up the topic of linguistics at lunch. unaccusative and unergative verbs. Suppose we were to find a pair of verbs V1 and V2, that meet b. The topic of linguistics came up at lunch. the following conditions: (31) a. The war brought Bill to his senses. b. Bill came to his senses. • V1 takes both an external and an internal argument.
• V takes only one argument. 2 (32) a. I brought Sue around to my point of view. • The q-role borne by the internal argument of V1 is the same as the q-role borne by the single b. Sue came around to my point of view. argument of V2. (33) a. After John fainted, we brought him to. It would be natural to speculate under such circumstances that V2 is unaccusative: the single b. John came to. argument of V2 bears the same q-role as V1's internal argument because it too is an internal argument. It becomes a subject for case and EPP reasons, which masks its underlying status as an Still, it is quite tricky to show that the verbs of the (b) sentences in (22)-(33) are indeed internal argument. unaccusative in English. I return to this topic below.
As it happens, English has a large number of pairs that meet these specifications, where the q-role in question is patient or theme. In many cases, the members of the pair actually have the same pronunciation. These are pairs like sink/sink, melt/melt, drop/drop. The two-argument member of the pair is semantically "causative": e.g. sink in (22a) means cause to sink, and appears to be identical to the single-argument member of the pair, except for having an extra "Causer" argument:
(22) a. The navy sank the submarine. [patient is direct object] b. The submarine sank. [patient is subject] 2 Many speakers, of course, do not command this distinction. (Confusion of lay and lie is very common.) LSA Institute 2017
6. Auxiliary verb alternations: be vs. have Of course, all the tests discussed in the previous section merely tell us that the verbs one might guess to be unaccusative do indeed behave in a special manner. It is easy to show that verbs that one might plausibly identify as unaccusative, given UTAH, often show strikingly different behavior from their plausibly unergative counterparts. Great! But do they really lack an external argument and show movement from complement position to Spec,TP? This section offers some evidence that putative unaccusative verbs really do For example: many languages form the past or perfect tense3 by using an auxiliary verb followed lack an external argument. by the past participle. The auxiliary verb is typically either have or be. Some languages uniformly use one of the other. In English, the perfect tense uses have, and in Spanish the Consider the two properties of passive morphology listed below: compound past tense also uses have. In Hindi and in the South Slavic languages (e.g. Bulgarian, Slovenian and Serbo-Croatian), the past tense uses be. (37) Passive morphology: 1. ... suppresses assignment of accusative case; In other languages, however, some verbs use have and others use be (examples: Italian and 2. ... suppresses normal assignment of external θ-role Dutch). When a language employs both auxiliaries, the auxiliary be is always reserved for verbs that we would expect semantically to be unaccusative. Thus, the Italian and Dutch counterparts to • Can property 2 of passive morphology apply vacuously? That is, can morphology that the (b) verbs of (22)-(33) take be.4 tinkers with a verb's external argument be added to a verb that lacks an external argument in the first place, i.e. to an unaccusative verb? (34) Italian a. Giovanni è arrivato. `John arrived` (be) The answer, interestingly appears to be no — universally! b. La nave è affondata. 'The ship sank' (be) c. Giovanni ha telefonato. `John telephoned` (have) (38) A linguistic universal: d. Giovanni ha letto il libro. 'John read the book' (have) The "1 Advancement Exclusiveness Principle"5 A verb of the unaccusative class may not be passivized. (35) Dutch a. Jan is gevallen. 'John fell' (be) To see this fact, however, we must first observe that property 2 (in contrast to property 1) may b. Jan heeft gelachen. 'John laughed' (have) apply vacuously in some languages. These languages allow so-called impersonal passives of those verbs that we might think are unergative. The connection between (hypothesized) unaccusativity and auxiliary selection is not limited to Indo- European languages. Old Japanese, for example, appeared to make a similar distinction, using the Impersonal passives auxiliary suffix -nu for verbs that would take be in Italian (the unaccusatives!) and -tsu for verbs that would take have in Italian (the unergatives and the two-argument transitives): In an impersonal passive construction, the external argument is eliminated or reassigned to a by- phrase -- but no direct object moves to subject position. Since these are fundamentally (36) Perfective Auxiliary Selection in Old Japanese intransitive verbs (or transitive verbs optionally used intransitively), a dummy element must fill [K. Takezawa (1989 talk; citing Yoshida 1973)]: Spec,IP position. Some examples: a. -tsu: oki-tsu (`have placed`), tsuge-tsu (`have told`), kiki-tsu (`have listened`), chirash-tsu (`have scattered` trans.), shi-su (`have done`), etc. [normal transitives, (39) Impersonal passive: German unergatives] Es wurde getanzt. It was danced. b. -nu: ki-nu (`have come`), nari-nu (`have become`), he-nu (`have passed`), sugi-nu '[unspecified] danced]' — i.e. 'There was dancing going on.' (`have elapsed`), chiri-nu (`have fallen/scattered` intr.), etc. [unaccusatives] (40) Impersonal passive: Dutch a. Er wordt hier door de jonge lui veel gedanst. 7. No passivization of unaccusatives it becomes here by the young people much danced ‘There's a lot of dancing by young people here.’
b. Hier wordt (er) veel gewerkt. here becomes there much worked 3 Past tense: I lived in New York. Perfect tense: I have lived in New York. The semantics of the perfect tense is quite 'Here people work a lot.’ complex and interesting. One distinctive property is the condition of "present relevance". In I have lived in New York, the fact that I lived in New York is assumed to be relevant in some way to other current properties that characterize me. And oft-cited demonstration of this is the fact that Einstein lived in Princeton is a normal and true thing to say in 2008, but Einstein has lived in Princeton carries the odd presupposition that Einstein is still alive. 4 In some languages, such as French, certain verbs that take be in Italian take have instead (and there are rules 5 The name of this principle sounds quite odd, but reflects a rather different theory of syntax that Perlmutter & Postal governing this difference) -- but this discrepancy is one-way. No verbs that one expects semantically to have and were developing when they did their work (called "Relational Grammar"). The word "1" in the name refers to the thus to be unergative takes be instead. subject (subject was "1", direct object was "2", etc.). LSA Institute 2017
(41) Impersonal passive: Russian It appears that the P has moved to V and glued to it (a process of head movement called V gazete bylo napisano ob ètom. "incorporation"). For example, the PP may not be moved rightward -- something usually possible in paper was written about this for PPs: '[unspecified] wrote about this in the paper' (47) a. Fred sat recently on it. (42) Impersonal passive: French b. *It was sat recently on __ (by Fred). Il a été tiré sur le bateau. it was shot at the boat (48) a. Pete spoke on Friday to her. '[unspecified] shot at the boat' — i.e. 'There was shooting going on at the boat.' b. *She was spoken on Friday to __ (by Pete).
In contrast, the verbs that we would guess to be unaccusative may not be passivized at all. They Passive morphology on V appears to affect the V+P unit, so that it eliminates the case-assigning lack impersonal passives. Here are some Dutch examples, but the phenomenon is universal:6 potential of the P. Consequently, the object NP must move for case reasons.
(43) Active vs. *Passive of unaccusative: Dutch Pseudo-passives are like impersonal passives in languages like Dutch in that the verb is not an a. De lijken zijn al gerot/ ontbonden. accusative-assigner, but thanks to V+P incorporation, passive morphology is able to attach to the the corpses are already rotted/decomposed verb anyway. Thanks to this trick, it is now relevant to compare unergative verbs like the agentive b. * Door de lijken werd al gerot/ ontbonden. use of sit (see below) and speak to unaccusative verbs.
(44) Impersonal passive of unergative vs. *unaccusative: Dutch As expected, none of the verbs that would take the auxiliary be in Italian -- i.e. none of the a. In de zomer wordt er hier vaak gezwommen. unaccusatives -- have a pseudo-passive! The most famous discussion is from a 1984 paper by ‘In the summer it is swum here frequently.’ David Perlmutter and Paul Postal, from whom these examples are taken:7
b. * In de zomer wordt er hier vaak verdronken. (49) a. *The package was accumulated on by dust. ‘In the summer it is drowned here frequently.’ b. *The room was burst in by the bubble. c. *The dome was collapsed under by the model. We cannot offer the same simple demonstration in English, because English lacks impersonal d. *The bridge was existed under by trolls. passives of the Dutch and German sort. In English, in general — if a verb does not assign e. *The bed was fallen on by dust. accusative case -- be it unergative or unaccusative, it may not receive passive morphology. This f. *The hill was grown on by grass. means it is impossible to tell whether (45b) is unacceptable because of vacuous elimination of an g. *The hall was increased in by the noise. external argument, since it is already unacceptable because of vacuous elimination of accusative- h. *The oven was melted in by the ice cube. case assigning capability: i. *The woods were vanished in by Little Red Riding Hood.
(45) a. *It was swum here frequently. Minimal pairs (also from Perlmutter and Postal's paper) are provided by verbs that have both an b. *It was drowned here frequently. agentive and non-agentive use. The agentive use behaves like an unergative verb, while the non- agentive use behaves like an unaccusative:
"Pseudo-passives" in English (50) a. The table was sat on by Fred. b. *The table was sat on by the lamp. On the other hand, there is a slightly complicated way to see the effects of the ban on passivization of unaccusatives even in English. (51) a. The closet was slid into by Mary. b. *The closet was slid into by the soap. An English verb that takes a PP complement allows a passive form in which it is the object of the preposition that moves to subject position. For no good reason, this construction has come to be called "pseudo-passive" — though there is nothing "pseudo" about it! Some examples:
(46) a, The table was sat on __ (by Fred). b. Mary has been spoken to __ (by Pete).
6 Lithuanian is often cited as a counterexample, since apparent passives of unaccusative verbs are possible in this language. But research done at MIT by Olga Vaysman shows that these apparent passives are actually quite a 7 "The 1-Advancement Exclusiveness Law" in D. Perlmutter and C.G. Rosen, eds. Studies in Relational Grammar 2, different phenomenon (something called an "evidential"), so Lithuanian is probably not a counterexample after all. University of Chicago Press LSA Institute 2017
The general picture is as follows: 'Many students will phone.'
English Dutch, French, etc. b. Affonderono [VP __ molti navi ] vacuous elimination of external no (universal) no (universal) argument ok? will-sink many ships vacuous elimination of accusative- no yes 'Many ships will sink.' assigning capability ok? The arrows in (53) show verb movement, which is not relevant to our story, except insofar as we need to understand why molti studenti is post-verbal in (53a). I have also placed molti studenti in We still don't really have an argument for the unaccusative analysis of predicates that UTAH Spec,VP in (53a) and placed molti navi as the complement of V in (53b) because I had to put predicts should be unaccusative. Are some surface subjects really underlying direct objects them somewhere — though of course this contrast is what we are trying to argue for here! We (and others not)? Alas, time did not permit me to go over the following classic argument from haven't shown it yet. Italian, due to Adriana Belletti and Luigi Rizzi, but (as promised) here is a brief summary of it, So let's show it. The section after this one (section 9) picks up again with the material that we did discuss in class — albeit horrifically briefly. The key to the argument is a fact about NPs that consist of a "bare quantifier", e.g. molti without either studenti or navi, as in the English sentence Many will phone or Many will sink. Italian, unlike English, sharply restricts such bare quantifiers when they occur as a direct object 8. An argument for the analysis: ne-cliticization in Italian8 (complement) of V:
The moral of the story in advance (54) Bare quantifiers in direct object position: Italian In a language in which movement of a nominative to SPEC,TP is not forced, you see the A bare quantifier in direct object position must cooccur with a pronoun ne that cliticizes 9 argument NP of unaccusatives and passives in direct object position. (morphologically attaches) to the finite verb in I.
One might reasonably ask whether we really can tell that the sole argument of an arguably The constraint in (54) is exemplified in (55): unaccusative verb starts off as a complement to that verb — in contrast to the sole argument of an unergative verb. One way to put the proposal to the test is to find a situation in which the alleged (55) Direct-object bare quantifiers require ne direct object of an unaccusative verb does not need to raise to Spec,TP, and a test that can tell us a. Gianni inviterà molti studenti. [not a bare quantifier, no ne] whether this alleged direct object really is a direct object. Gianni will-invite many students
Italian provides us with a situation of just this type. In English (at least insofar as we've discussed b. *Gianni inviterà molti. [bare quantifier, no ne] the matter), because unaccusative verbs fail to assign accusative case — and because there is no Gianni will-invite many. other way to assign case to the sole argument of an unaccusative — the sole argument of an unaccusative must raise and become a subject (i.e. a specifier of TP). In Italian, however magic c. Gianni ne inviterà molti. [bare quantifier with ne] happens! Gianni of-them will-invite many
Finite T can assign case to an NP that remains within the verb phrase — in fact, to the closest NP Crucially, the requirement of ne as given in (54) is a property of NPs in direct object position within VP, whether it is Spec,VP or the complement of V: only. For example, pre-verbal subjects — even subjects of unaccusative and passive verbs — do not allow ne when they are bare quantifiers: (52) Magic! Finite T may license nominative on a DP that remains in VP. (56) Subject bare quantifier in Spec,TP does not allow ne. a. *Molti ne inviteranno Gianni. For this reason, though Italian is an SVO language where V raises to T, the subject of the Many of-them will-invite Gianni. sentence can be found following V (much as in Irish, which we examined several weeks ago): b. *Molti ne telefoneranno. (53) Subject remaining in VP in Italian Many of-them will-phone
a. Telefoneranno [VP molti studenti __ ].
will-phone many students
8 Work of the Italian linguists Adriana Belletti and Luigi Rizzi in the 1970s. 9 Ne is actually a pronominal form that means 'of them' or 'of it'. LSA Institute 2017
c. *Molti ne saranno invitati. Many of-them will-be invited We thus see that the "magic" case rule of Italian allows an underlying direct object to remain a direct object even when marked nominative by T, which is why it acts like a direct object with d. *Molti ne affonderono. respect to ne. Crucially, the sole arguments of unaccusative verbs (and passives) are underlying Many of-them will-sink direct objects.
All the examples in (56) become good when the ne is omitted — so it's not as though Italian completely lacks bare quantifiers.10 The constraint seen in (56) is not about linear order. When, thanks to Italian case magic (as given in (52)), the sole argument of an (arguably) unergative verb 9. Burzio's Generalization remains in the VP and thus follows the finite verb it still cannot take ne when it is a bare quantifier:11 By the time we reached this section, I had five minutes left, so you heard a very abbreviated version of this (nothing about the imaginary verb sënk, for instance). Here is a fuller (57) VP-internal bare quantifier as argument of an unergative verb does not allow ne. presentation: *Ne telefoneranno molti. [molti occupies Spec,VP; see (53)] of-them will-phone many We've assumed without comment so far that unaccusative verbs (including the raising verbs that provided our introduction to unaccusativity) fail to assign case as well as an external q-role. The Now let us look at a bare quantifier as a the sole argument of an unaccusative. If the sole properties of unaccusative verbs thus seem to be very similar to the properties of passive verbs: argument of an unaccusative really starts out as the complement of the verb, in contrast to unergative constructions, then case magic should allow it to remain in complement position. A (60) Properties of unaccusative verbs: bare quantifier in such a situation should therefore act like the direct object in (55c) and not like 1. An unaccusative verb lacks an external argument. the subjects and specifiers of VP in (56) and (57). 2. An unaccusative verb fails to assign accusative case.
The facts support our analysis of unaccusatives. A post-verbal bare quantifier with an In fact, there is a link between these two properties. The generalization in (61), called Burzio's unaccusative verb does indeed behave like a direct object, requiring ne: generalization, appears to be true (with some qualifications):
(58) VP-internal bare quantifier as argument of an unaccusative verb requires ne, like (61) Burzio's generalization direct objects. If a verb assigns accusative case, it has an external argument. a. *Affondarono molti. [molti behaves as in (55b), bad with no ne] will-sink many That is, there is no verb sënk with, say, the meaning of 'sink', that lacks an external argument but does assign accusative case. Such a verb would occur with a dummy subject:
b. Ne affondarono molti. [molti behaves as in (55c), ok with ne] ACC of-them will-sink many (62) *Itexpl sënked the ship. Not surprisingly, passives behave like unaccusatives: 'The ship sank.'
(59) VP-internal bare quantifier as argument of a passive verb requires ne, like direct Likewise, there is no verb arrüve 'arrive' with similar syntax: objects. ACC a. * Saranno invitati molti. [molti behaves as in (55b), bad with no ne] will-be invited many (63) *Itexpl arruved the package. 'The package arrived.' b. Ne saranno invitati molti. [molti behaves as in (55c), ok with ne] of-them will-be invited many For that matter, there is no raising verb (i.e. an unaccusative that subcategorizes for infinitival IP) with these properties. Such a verb, of course, would not actually require any raising: ACC
10 Thus, the following are ok, in contrast to (56a-d): (64) *Itexpl sumes [Mary to have left the room]. a. Molti inviteranno Gianni. b. Molti telefoneranno. Many will-invite Gianni. Many will-phone 'Mary seems to have left the room.'
c. Molti saranno invitati. d. Molti affondarono. We have seen that Burzio's generalization entails the absence of a verb like sënk/arruve/sume in Many will-be invited Many will-sink the lexicon. Interestingly, it also entails the absence of any morpheme that could eliminate the external argument role of a verb without also eliminating the accusative case assigning capacity 11 In this case, interestingly, omitting ne does not help. A bare quanitifier is simply bad no matter what when it of that verb. That is, it also provides a link between the two properties of passive. occupies Spec,VP. We will not attempt to explain this fact. LSA Institute 2017
Consider: if a language has a morpheme M that eliminates the external argument of a verb, the Another such construction is the so-called "Fake Reflexive" construction. Here we find a morpheme should have two other properties: reflexive pronoun coupled with a result or goal phrase. A point of interest is the fact that the (a) M may not be added to a verb that lacks an external argument in the first place. reflexive is not possible otherwise: The dog barked himself hoarse, but not *The dog barked (b) M also deprives the verb of its accusative case-assigning capacity. himself:
Property (b) is automatic, given Burzio's generalization. This means that the lexicon does not (66) Fake reflexives have to state that M deprives the verb of its accusative case-assigning capacity. It is a general a. The dog barked himself hoarse. property of grammar. b. The baby cried herself to sleep. c. John drank himself senseless. Added value from reading the summary: something I did not say in class. d. ?Mary jumped herself dizzy. An explanation for why no language allows vacuous de-external-argument-ization, but some languages allow vacuous de-accusative-case-ization. e. *The politician arrived herself to death. f. *The actress playing the Wicked Witch melted herself silly rehearsing the scene. Perhaps (a) is automatic as well. Perhaps when the lexical entry for a morpheme says "I deprive a g. *The stuntman dropped himself senseless rehearsing the scene. verb of X", the process can never happen vacuously. Also, the so-called "His/her way" construction. Once again, the accompanying direction or goal In languages that allow impersonal passives, then, the passivize morpheme does not say in its phrase seems to be obligatory (*Babe Ruth homered his way.) Frank Sinatra's I did it my way is a lexical entry "I deprive a verb of accusative case-assigning capability". The fact that the passive different construction...: morpheme has this effect is automatic. (67) The "Way" construction In languages like English that do not allow impersonal passives, the passive morpheme does say a. Babe Ruth homered his way into the hearts of America. in its lexical entry "I deprive a verb of accusative case-assigning capability". This means that the b. The dog barked his way into the room. process cannot be vacuous, and impersonal passives are ruled out. c. The baby cried her merry way across the state. d. Mary whistled her way home.
And we did not get to this at all! e. *The politician arrived her way across the state. f. *The Wicked Witch melted her way into the hearts of America.12 Notice now that Burzio's Generalization is a one-way implication (a conditional, to use the language of logic), not a two-way condition (a biconditional). We said "If a verb assigns The fact that I'm just listing constructions is in part an indication that there are mysteries accusative case, it has an external argument" — from which it does not follow that if it has an remaining about how these constructions are put together (and squib topics available). My point external argument that it must assign accusative case. On the other hand, Burzio's Generalization is merely to show that accusative case is quite generally available for verb that assign an external leaves open the possibility that any verb with an external argument may assign accusative case, if q-role, just as Burzio's Generalization (all things being equal) leads us to expect. the conditions are right.
For a verb that actually subcategorizes for a complement NP, the conditions are right whenever that NP is present. I have in mind ordinary verbs like eat as in the VP eat the pizza. But even verbs that do not normally subcategorize for an NP may find themselves in special constructions where there is an NP that might receive accusative case. In such circumstances, even intransitive unergative verbs turn out to assign accusative case. One example is the so-called "Cognate Object" phenomenon, in which a direct object morphologically connected to the verb can show up even with otherwise intransitive unergatives. Unaccusatives do not allow this construction:
(65) Cognate objects a. The dog barked a great big bark [unergatives] b. The baby yawned a tremendous yawn c. John smiled an amazing smile d. Mary jumped an astonishingly high jump
e. *Sue arrived an excellent arrival. [unaccusatives] 12 Actually, this might not be so bad, but perhaps it is significant that I have to imagine the actress who played the f. *The wicked witch melted a rapid melt. Witch going on a "melting tour" in which one way or another she melts herself repeatedly, as publicity (let's say) for g. *The careless climber dropped a terrifying drop. the film. In such a case, melt might be used as an unergative. I can almost imagine something similar for (66f) as well, actually. LSA Institute 2017
10. Appendix: Apparent exceptions to Burzio's Generalization (76) a. Bill provided Mary [with a book]. b. John appeared [to Mary] [as intelligent] I didn't even *plan* to discuss this in class, but you might find it interesting: c. Il suo nome e' sfuggito [a Teresa]. [Italian] The following construction looks like it might involve raising of the sort suggested by my arrow: the her name AUX escaped to Teresa 'Her name escaped Teresa.' (68) John struck Mary [as __ intelligent]. d. The computer comes [with a DVD drive]. John appears to get is q-role from intelligent, much as in the related construction (69): e. That dress looks good [on you]. f. We got [to Iceland]. (69) It struck Mary that John was intelligent. Furthermore, one finds alternations among the unaccusatives that resemble the provide a book to This suggests that strike (in this use, of course) lacks an external argument, much like seem. If Mary ~ provide Mary with a book alternation that we examined in the problem set as a precedent this is true, the NP Mary in (68) (and, in fact, in (69) as well) appears to be receiving case from for the proposed alternation give a book to Mary ~ give Mary G a book: strike in violation of Burzio's Generalization. (77) a. Bees were swarming in the garden. We might be tempted to reconsider the unaccusativity of strike, but strikingly (pardon the b. The garden was swarming with bees. accidental pun) the verb in this use does not allow passive. This suggests that the verb is indeed unaccusative: (78) a. Cockroaches were crawling in the kitchen. b. The kitchen was crawling with cockroaches. (70) *Mary was struck as intelligent by John. (79) a. Flies were thick in the outhouse. In fact, a number of other verbs show similar behavior: b. The outhouse was thick with flies.
(71) a. Sally's name escaped me for some reason. Both members of these pairs appear to be unaccusative: b. *I was escaped by Sally's name for some reason. (80) a. *The garden was being swarmed in by the bees. (72) a. The computer has a DVD drive. b. *The bees were being swarmed with by the garden. b. *A DVD drive is had by the computer. (81) a. *The kitchen was being crawled in by the cockroaches. (73) a. That dress really becomes you. ( = '...looks good on you') b. *The cockroaches were being crawled with by the kitchen. b. *You are really become by that dress. There is a lot to investigate about such alternations (and many others!). For a compendious list, (74) a. We reached Iceland the next morning. going far beyond the topics we will cover in this class (but related to many of them), you might b. *Iceland was reached by us the next morning. want to look at the following reference book:
All these examples, if analysed as unaccusatives, are in effect double-object constructions, which Levin, B. (1993) English Verb Classes and Alternations: A Preliminary Investigation, University suggests a possible solution to the conflict with Burzio's Generalization. It may well be that the of Chicago Press, Chicago, IL. A web-accessible index to this book is at http://www- direct objects of these verbs are introduced by a null case assigner like the "G" you posited in the personal.umich.edu/~jlawler/levin.html last assignment on verbs like give:
(75) a. Bill gave Mary [G a book].
b. John struck [G Mary] [as __ intelligent] c. Sally's name escaped __ [G me]. d. The computer has __ [G a CD drive]. e. That dress becomes __ [G you]. f. We reached __ [G Iceland]
In many cases, as it happens, we can find a similar construction that shows an overt preposition. For example, compare: