Prolegomenon to a Theory of Argument Structure Kenneth Hale & Samuel Jay .. soundofheaven.info~hharley/PDFs/HarleyMayfestHandoutpdf. Prolegomenon to a Theory of Argument Structure Linguistic Inquiry Monographs Samuel Jay Keyser, general editor 6. 10 Size Report. DOWNLOAD PDF. a) How do we explain the limited set of argument structure types? Now, what's wrong with our standard theory of argument structure? Hale, Ken and Jay Keyser (): Prolegomenon to a Theory of Argument Structure.
|Language:||English, Spanish, Hindi|
|ePub File Size:||19.83 MB|
|PDF File Size:||18.72 MB|
|Distribution:||Free* [*Regsitration Required]|
Prolegomenon to a theory of argument structure. By KEN HALE and SAMUEL JAY. KEYSER. (Linguistic Inquiry monograph ) Cambridge. Request PDF on ResearchGate | On Jan 1, , Kenneth L. Hale and others published Prolegomena to a Theory of Argument Structure. Request PDF on ResearchGate | On Jan 1, , Ken Hale and others published Prolegomenon to a Theory of Argument Structure.
Set up a giveaway. The double object construction contains, as its uppermost component, the a -type structure, with empty V. This is why the cognate object construction has the name that it has: Prolegomenon to a Theory of Argument Structure review For lexical items of the type represented in 10b , the sentential syntactic subject e. They smeared mud on the wall.
This bread cuts easily. He sliced the salami. Salami slices easily. She dented the fender. This fender dents easily. Transitivity Alternations d. I anger easily. The dog frightened the chicken. Chickens frighten easily. We maintain that these shortcomings are not, properly speaking, failings in our conception of argument structure.
Rather, they derive from our failure to understand fully certain implications of the system. The second problem dealt with here is just one of many similar problems that will have to be confronted, since it has to do with the interaction of subsystems, not with the basic argument structure relations themselves.
The observed asymmetries in this case relate to the connection between features of lexical meaning and the arguments of the verb—internal on the one hand, external on the other. For example, the verb laugh, we contend, is fundamentally transitive, having the structure portrayed in 1. An example is bag in bag the apples, derived as shown in 6 and 7. Our questions are the following, among others. What is the nature of the trace, if there is one?
Or is it visible at LF? There is another rather large question, however—namely, what is the proper analysis of so-called cognate argument constructions of the type represented in 8? They are dancing a Sligo jig.
He shelved the books on the windowsill. Leecil saddled old Gotch with his new Schowalter. Unless something else is involved here, this should be impossible.
In short, we must develop a theory of cognate arguments. This can be appreciated through a consideration of location and locatum verbs. While 9a,b are possible, 10a,b are not.
Leecil corralled the calves. Myrtis rosined the rope. This is what we contend, at least. Examples of the type represented by 12 are un- 52 Chapter 3 problematic, presumably, since the bare noun N is the complement, hence structural sister, of the verb. Incorporation is always permissible in cases such as this, for languages that have incorporation at all. Consider in this connection the a -examples of 13 and 14 the b -examples illustrate the case in which the nominal object remains unincorporated.
Itam tap-wari-k-na. Itam pu-t taavo-t wari-k-na. These features of the verb are displayed in 15 , where the linear order of elements itself theoretically immaterial for present purposes mimics the Hopi surface arrangement. Although it is governed by V1 -na , it bears no grammatical relation to that verb. The following sentences exemplify adjectives in their simple predicative function: HDP , this horse very completely tame: By hypothesis, then, the derived transitive of 14b has the structure shown in Abstracting away from these processes i.
Japanangka straightens spears. The north wind clears the sky. It does not really matter whether our analysis of the Hopi construction in 21 is exactly correct. This is enough to distinguish the two operations. Umu-na kaway-mu-y kuk-hep-ma. As usual, the residue is assigned Case, glossed as genitive in this instance, though it is indistinguishable from the accusative morphophonologically see Jeanne It is subject to various analyses, however.
This remains in the position corresponding to the grammatical object of the verb and is accordingly assigned accusative Case. The maximal categorial projection is a node that does not project further. For example, spear in 24 is not a sister of V, though spear is surely governed by V, demonstrating that local c-command and government are not enough. Consider, for example, the structure of the verbal projection in 25b , shown in To pursue this idea, let us consider the simplest case—for example, the verbal projection make trouble.
This is formed by selecting each of the items make and trouble, and combining them by means of the Merge operation, as shown in Select [make] b. Select [trouble] c. But here we touch on an issue that is central to our investigation, namely, the labels of lexical items and other syntactic objects. Parallel remarks apply for the noun trouble, of course. But it is not only terminal nodes that have labels; all syntactic objects do.
The label of the phrase is determined by one or the other of the two constituents, and that constituent is therefore the head of the construction. If V determines the label, then V is the head; if N determines the label, then N is the head; the choice of category labels or standard spellings is arbitrary in customary usage, and there is no need to distinguish phrasal levels. This is at least a popular working assumption.
It could be wrong, of course; it may belong rather to the category of unexamined initial impressions unsupported by any real linguistic considerations. In fact, this assumption is almost certainly wrong if, as is probably the case, the labeling of syntactic objects is automatic and simple see Chomsky This is actually quite natural, given that it is trivially true in any event: The principle formulated in 33 simply extends the automatic labeling to all nodes that are projections of X.
For present purposes, we are holding in abeyance the question of what exactly is meant by features in Without specifying too precisely the form in which this information is registered in the label, we will refer to it by the expression p-signature, representing a phonological feature set of some sort, possibly a set of feature matrices.
This is merely a notational convention, of course. This is the situation involved in zero derivation, so common in English. English denominal verbs, like unergative laugh and sneeze, exemplify the simplest case.
P-signatures are cited in square brackets. Simultaneously, we propose, the substantial p-signature of the complement is copied as in 36 into the empty psignature of the head, substantiating the latter. Consider thicken, consisting of the following items: Consider the verb bottle, as in bottle the wine. According to our view of the matter, the inner projection is headed by an empty P.
Its complement is the bare noun bottle. The only p-signature that is, so to speak, visible to a head at Merge is that of the complement, a Merge-partner. A determiner heads a D-projection, that is, a DP. This follows directly within the view being courted here, because the p-signature of the noun is inaccessible to the verb. It does not stand in the Merge relation to the verb; only DP does. The lexical head N is not the strict complement of V; rather, DP is the strict complement of V, in accordance with 28 and the Merge operation.
And this would seem to be the case, as we will show. It also has the morphophonological property that it is dependent: The adjective is the lexical head of a small clause in our sense of the term: In Hopi, by hypothesis, the adjective stands in the complement relation to a nuclear element PRED, the true head of the small clause.
It is not a Merge-partner of V. The Hopi sentence 25a illustrates possessor stranding. An item M is properly contained in a phrase XP if it is dominated by the maximal projection XP and the latter is not the label of M. But it should be said, as an aside at least, that these assumptions about phrase structure are highly theoretical and could, of course, be wrong. For one thing, in the DP examples considered here, for instance, it could well be that the label and hence the p-signature of the relevant lexical head does indeed appear at the node that dominates the strict complement and Merge-partner of the target verb.
This would be true, for example, if the phrases harboring the nouns at issue here were in fact NPs projected by those very nominal heads as assumed in Baker , following a respected traditional view, contra the DP hypothesis of Abney And the small clause structure attributed to the complement in 21 , as diagrammed in 22 , may not be the proper source of the putative incorporation.
In that case, the noun would project its label to the phrase as a whole. Consider, for example, the verb phrase bottle the wine, whose structure 45 is repeated here in The essential point we wish to make here is that the syntactic structure of the verb phrase is left intact. All information necessary for the purposes of syntax and logical form is fully present in the structure with the understanding, of course, that P and N are abbreviations of the full sets of syntactic and semantic features pertaining to those nodes.
We can assume that the p-signatures of all nodes are irrelevant, and invisible, to syntax and LF. In another sense, however, the answer is yes, trivially. This has consequences for another question asked in section 3. She slept the sleep of the just. He laughed his last laugh.
He danced a jig. He bagged the potatoes in a gunnysack. For present purposes, we will set the hyponymous argument construction aside and deal with the cognate object construction. The examples we have provided in 50a,b might ultimately prove not to be genuine examples of this construction, but we will take them to be such. Whatever the outcome in this case, we take it to be a fact that there is such a phenomenon—that is to say, a transitive verbal construction headed by a denominal verb whose object is headed by a noun that is root-identical to the verb, as in the examples cited.
Further, in the true cognate object construction, the object can only be headed by a rootidentical noun, not some random distinct noun, even a hyponym. Thus, if 50a,b are true examples, as we will assume, then the following are predicted to be ill formed. These verbs can, of course, occur in other transitive constructions, with a range of S-Structure objects e.
If verbs of the type represented by sleep and laugh are strict cognate object verbs, as we are suggesting, then not only do they not permit hyponymous objects, but they also reject pronominal versions of their cognate objects.
By contrast, verbs like dance and sing readily accept pronominal objects. John danced the tango and Bill danced it too. Robin sang the songs of the 60s and Bill sang them too. However, sentences like 53a,b need no repetition. They sound perfectly well formed immediately. To begin, we will attempt to develop an account of the strict cognate object construction. There are at least two possibilities: In English, the argument would go, both the head and the tail of the movement chain would be spelled out, as seen in 54 , an informal representation of the verb phrase of 50a.
If the strict cognate object construction arises through head movement, as suggested in 54 , it is to be expected that the point of origin i. The structure of the verb phrase here is essentially as in We cannot rule this sentence out. While it may be somewhat strange, it is perfectly grammatical; if someone could actually turn sleep into dreams by sleeping, then 57 would describe that ability. He slept the hours away. She laughed her way through life.
This is not the strict cognate object relation. Thus, in the case of the denominal verb laugh, for example, the noun that contributes its p-signature to the verb is not only the sister of the verb, but also maximally a lexical projection, not an extended projection. Similarly, in the case of bottle, as in bottle the wine see 45 , the p-signature ultimately passed on to the verb is from a lexical projection, P itself phonologically substantiated by the p-signature of its own complement bottle.
By contrast, the strict cognate object relation generally holds between the verb and an extended projection sister to it. That is to say, the nominal component of the verb is identical to the p-signature of the lexical head of the extended projection DP of the complement of the verb. This is portrayed in 54 , the structure that results under the assumption that the cognate object relation is established by incorporation.
This would involve copying the p-signature of the N into the defective p-signature of the V. If the features that project to these heights include the p-signature, in the manner shown in 59 , there is a point in the structure at which the defective verb is a structural sister to a node D in this case bearing the p-signature of the lexical head.
The nominal p-signature [sleep] is therefore copied into the defective label of V. The strict cognate object relation can be no more distant than this. It is not necessary, however, to be committed to the idea that the p-signature of a lexical category actually projects to the dominating node of the extended projection, as it is shown to do in Perhaps it projects only to the maximal projection of the lexical head.
In English, stranding of determiners is not allowed. This is why the cognate object construction has the name that it has: Under certain conditions, copies are realized overtly in both positions, that is, target and source.
This will be so where the source position is dominated, within its extended projection, by certain functional categories, for example, D. In neither case is the source head contained in an extended projection dominated by a functional category. This circumstance is prominent among those in which the source copy is covert—that is, deleted or, perhaps, simply not spelled out at PF. And, in general, how do we account for the presence of the nominal component in this populous class of verbs?
There are two questions, essentially, one having to do with the relation of p-signatures to the phonological representation of terminal nodes, the other with their presence or absence in particular nodes. The evidence is against this.
Suppletion is a clear case. The English form [went], for example, cannot be inserted before the syntax has brought the verb together with the past tense, for it is the latter process that creates the environment for the suppletive element [wen]; the elements involved here are in distinct nuclear positions in the basic syntactic representations of sentences containing the form.
Assuming that arguments of this type are persuasive, we must at least reject the idea that the p-signature of a nuclear element is the phonological matrix corresponding to its actual form at PF. The p-signature of a head H must contain information that will permit H to be properly associated in syntax with the appropriate forms of vocabulary items drawn from the lexicon. And, of course, we want the right spellings to appear in the right places, that is, in the right slots in terminal nodes.
There must be some way, therefore, to relate a p-signature to the correct vocabulary item. One possibility is that the p-signature of a head H consists of the entire set of phonological matrices of H—that is to say, the entire register of the allomorphs of H and their contextual frames.
Lexical insertion would amount to discarding all allomorphs that do not satisfy the particular environment in which H appears at S-Structure, that is, at the syntactic level relevant to PF. Another possibility is that the vocabulary is entirely autonomous, consulted only at PF. Our proposal is tentative, provisional, and somewhat clumsy at this point; we adopt it as a temporary convenience.
We will simply assume that p-signatures are indices and that vocabulary items bear indices as well. The index of a particular vocabulary item must match that of the morpheme, or terminal node, into which it is inserted. So the spelling [laugh]i will appear at the terminal V-node substantiated by the p-signature index PSi , copied from the N laugh in the position of its syntactic complement.
The second question posed above can be restated as follows. Do all nonempty heads in syntax bear a p-signature? Are there categories that do not have a p-signature? Rather, what we are asking here is whether the p-signature could be entirely and systematically missing from some head or category of heads. Bittner gives a detailed argument, further developed by Bittner and Hale a , that the marked structural Cases e. If the language realizes these overtly at PF, they will acquire phonological substance in the course of Vocabulary Insertion.
We got the calves in the milkpen. Put the bandage on here. She will put the horse in there. Since a preposition has no p-signature, it cannot pass a p-signature on to the defective V; hence the impossibility of the starred forms of The overt prepositions in, on in the parenthetical examples of 62 arise directly from Vocabulary Insertion, independently of any p-signature.
Insertion can proceed without reference to phonological features—although, of course, a morphophonological register will be carried along, if one is present in the vocabulary item.
In the derivation of 61 , the preposition acquires a p-signature from its complement N, bottle, and passes it on to the verb, giving the derived verb bottle.
Put the wine in the bottle. The inserted P has the p-signature and phonological matrix corresponding to the actual preposition in, as part of its entry in the vocabulary, accounting for the preposition in 63a. Since the verb is not defective, the psignature of its complement will not be passed on to it. Here, the p-signature of the noun bottle is copied into the defective P, leaving its own copy behind in the DP complement, in accordance with the putative principles of cognate object formation.
This might be the case, for example, if surmounting the D functional projection required a special provision, say, an operation akin to Move—extraction out of the D-projection and copying into V. If so, then the following ill-formed sentence type might also be explained in this manner: This is also impossible— 66 is not a way to say Put the wine in a bottle or Put the wine in bottles. Instead, Vocabulary Insertion applies to introduce the appropriate preposition, blocking We assume that the principle involved here has to do with the licensing of an overt argument, in an argument position.
Presumably, an overt argument must be a phrase, in the traditional sense; that is, it must be an extended projection, not a bare root or stem. An overt nominal extended projection, DP, is licensed by Case and agreement; an overt bare noun, on the other hand, is licensed by incorporation. The bare noun bottle in 67 is not properly licensed. There are two possibilities, at least: Such derivations are ruled out already in the case of bare overt noun complements.
This possibility will not account for 66 , however. This may or may not be desirable. At Vocabulary Insertion, of course, prepositions acquire phonological substance. As terminal nodes in syntax, however, they are phonologically undetermined. Suppose this lack of a p-signature is true of functional categories as well—categories like tense, Case, articles, and agreement, that is, elements whose phonological constituency is often highly dependent upon morphological context.
If this is so, then, while prepositions constitute a lexical category, they share an important characteristic of the other closed class items i. Possible evidence that at least some prepositions belong with the functional categories comes from several quarters. It is also well known that adpositions often fuse with adjacent functional heads in nominal extended projections, typically with D, often in a manner that eliminates any vestige of the original, unfused shapes of the individual components.
The study of certain language impairments also provides evidence that P can be viewed as a functional category. It is nevertheless arguably true that caselike adpositions form a restricted and closed class, even in languages with a large and potentially extendable inventory of nominal adpositions like those of the Athabaskan language Navajo, its relatives, and the Misumalpan languages Miskitu and Sumu, for example.
There is, therefore, an element of contextual dependency in their semantics. An adposition is interpreted by virtue of the construction in which it appears and, in that respect, it shares a feature with the functional category Case. Some subset of the properties listed in 70 could well render the psignature at least redundant, or even impossible, for functional categories and adpositions. But adpositions do not possess these characteristics uniformly.
The property that might be attributed to English prepositions as a group is 70c , membership in a closed class. Although the class of prepositions is rather large in English, exceeding in number the inventory of true verbs in some languages of the world e. It is reasonable to suppose, it seems to us, that the members of a closed class would lack p-signatures.
It is possible as well that 65 and 66 are to be explained in relation to these considerations. The idea would be that Vocabulary Insertion is necessary in the P-position in these cases, and it fails because the psignature of the P-node in syntax i. Vocabulary Insertion would be necessary, presumably, because P is otherwise not properly licensed in syntax.
In 64 , of course, P is licensed by Vocabulary Insertion.
This is an impossible notion, not only for general theoretical reasons. She is playing a jig. He put the books on the windowsill. For our purposes, Vocabulary Insertion is the relevant operation, applying to the verbal projection contained in 73a , for example, to give the structure in But if the lexicon contains a verb dance, complete with phonological matrix, what prevents us from saying that that item itself is inserted in the V-position, giving 78?
The phonological matrices that give the verbs in question their alleged denominal appearance must already be present in the lexicon, as seen in the case of 75 and It follows that they take an external subject and consequently cannot transitivize in the simple manner of inchoative or unaccusative verbs of the type represented by break and clear.
The association of a particular phonological matrix with the terminal V-node would result from insertion of a vocabulary item already supplied with a phonological matrix, as is necessary in We might simply generalize it to all cases, including In our more recent, label-copying conception of the process, the empty heads are presumably licensed through the connection established between a complement and its governing host by label copying itself. The verb is presumably licensed by Vocabulary Insertion.
It is the empty nominal complement N that is now at issue. Is it generally true that an empty N complement must be licensed? If so, then N in 78 must be licensed.
Evidently, transitive light verbs cannot take an empty object. He cleared the screen. Leecil tightened the cinch. The log split. She split the log. Intuitively, what is wrong here is that there is not enough information around to posit an object.
The baby slept. Isadora danced. The colt sneezed. But we have evidence, from many sides, that the nominal component is in the verb in any event—movement is redundant, and arguably impossible. Suppose we say, then, that the relevant ingredients here are these: The second of these is in a sense nothing more than classical semantic selection. In the case of the unergatives illustrated in 81 , the verbs identify the nonovert complements in an intuitively more or less obvious manner, as eventualities or entities corresponding to the English nouns sleep, dance, sneeze.
With this background, we can say something now about the verbs of But the verbs of 79 lack the lexical semantic component present in the verbs of 81 ; it is this component, represented by the braced index in 83 , that licenses the nonovert nominal complement.
Here we have appealed to the second part of 82 , on the assumption that semantic features of lexical items are relevant to the licensing of nonovert complements. This structural consideration is the most likely factor involved in the verbs of But the hypothetical nonovert nominal—the argument omitted in the ill-formed sentences—is not selected by the verb and hence does not enter into the right structural relation with it.
Thus, it is not 94 Chapter 3 eligible to be linked to the upper verb in the manner expressed informally by means of the brace coindexation; hence, that position must be overt, as it is in the parenthetical examples cited in Another contrast we must deal with is found in the domain of location and locatum verbs, as in 84 and John put the books on the top shelf.
John shelved the books. John shelved the books on the top shelf. Leecil saddled the horse. Leecil saddled the horse with a new Schowalter. By contrast, the verbs of 85 freely permit omission of this constituent. Evidently, then, verbs like shelve and saddle, and other members of the location and locatum types, have the lexical semantic features required to license a nonovert [P N] that our hypothesis assumes to be structurally present in the sentences of Thus, by hypothesis, the verb shelve must be brace-coindexed with the nominal complement of the preposition, as shown in Assuming that something of this sort is correct, or at least reasonable, there is an additional detail that must be dealt with.
Strictly speaking, the verb in 86 does not bear a direct relation to the object of the preposition. The preposition intervenes—from the structural point of view, it is P that selects N in So how is it that the verb can license the nonovert N?
Thus, there is a chain of selection extending from the verb to the object of the preposition, and one of the links, the P, might be especially porous in relation to semantic selection. It can be seen in sentences like those in 87 and I corralled the calves in the milkpen.
He boxed the apples in a gunnysack. They armed the trap with teeth. They saddled us with responsibilities. We armed the priest with a lawyer. We armed the lawyer with a priest. This congruity is lacking in 87b ; at least, it is lacking in the ordinary understanding of the meanings of the verb box and the noun gunnysack.
We conclude from this that a location or locatum verb bears a selectional relation to the nominal object of its prepositional complement. If the latter is nonovert, it must be licensed; it will be licensed if it can be brace-coindexed with the verb as in The verbs of 84 stand in the proper structural relation to the nominal at issue, but the required coindexing cannot take place, because the verb is semantically empty in the relevant sense.
This system is not quite right, however. Licensing of an empty argument evidently requires something more. In this respect, the relation between the semantic features of the verb and the empty nominal made visible by them has the character of a strictly local binding relation.
In the simplest case, that of a verb and its nominal complement e. In the more complex case of a verb and the object of its prepositional complement, the relation is not local if an actual P intervenes, in the sense that it is closer to V than N is i. Let us suppose that the relation holding between the coindexed verb and noun in 83 is antecedent government, continuing the thought that this is a binding relation and that an empty N must be antecedentgoverned. In 86 , let us assume, the relation between the verb and the empty noun is also antecedent government.
If so, the structure projected by an empty preposition is not a barrier for this relation. In 89 , however, a contentful preposition evidently projects a barrier to antecedent government; hence, the empty N is not licensed.
In 87 and 88 , however, the post-prepositional overt nominal projection is licensed, since it does not need to be antecedent-governed. The binding requirement of an empty nominal complement is reminiscent of the requirement that an anaphor must be locally bound.
Our conclusion is this. The noun can therefore be nonovert i. Observationally, it is not enough to say that a verb capable of antecedent-binding a nonovert noun has certain semantic content. This idea comes, no doubt, from the fact that most of the verbs at issue here are morphologically related to nouns, typically, but not always, to the extent of being phonologically identical to them: Moreover, this relationship, to a greater or lesser extent, carries over to the meaning: At this point, we suggest that the question is not one of a noun being related to a verb; rather, it is one of an indeterminate item, a root, as it were, which is not inherently either a noun or a verb.
And the alleged cross-categorial relation, accounting for such terms as denominal and deverbal, is simply an appearance, a consequence of categorial indeterminacy. There is, however, one issue that remains to be discussed. But there is another verbal type that must be considered in this discussion. This is the class of deadjectival verbs of the type represented by clear, as in The sky cleared.
The wind cleared the sky. The question here is whether these verbs should also be analyzed as cases of selection and coindexation, like the denominal verbs we have been considering so far.
The selection alternative would maintain that, in the case of clear, there is a root element in the vocabulary that, in addition to appearing in the head position of an adjectional A projection, also appears in the position heading a verbal extended projection. As a verbal head, it selects a complement, as depicted in This follows, because of the transitivity alternation that it freely enters into, namely, the one in There are at least two reasons to doubt that the account just sketched is correct.
One of these is, so to speak, a negative reason. Denominal verbs of the type represented by dance, and denominal location and locatum verbs like shelve and saddle likewise, enter productively into cognate object and hyponymous argument relations, discussed in section 3. By contrast, deadjectival verbs do not enter into this relation productively. Thus, clear it very clear, redden the cloth bright red, lengthen the road two miles long, and the like, do not seem to us to be grammatical.
If this is true, then it is a fact that does not follow from the selection theory sketched above and embodied in the structure modeled by The unacceptability of cognate and hyponymous complements in the case of deadjectival verbs might be explained, on the other hand, by a theory in which the process involved in their derivation is in fact incorporation, the very process ruled out for denominal verbs.
According to the incorporation theory of deadjectival verbs, the adjectival complement would move from its base position into the verb adjoining to it in accordance with the principles constraining head movement. Two additional and generally accepted assumptions will account now for the failure of the cognate object construction with deadjectival verbs: Another observation that casts doubt on the selection theory of deadjectival verbs is the manifestly composite morphological makeup of most of them, for example, the type represented by redden, widen, lengthen, strengthen, tighten, and darken.
So far as we can see, there is no natural account of these in which the composite verb e. Without additional machinery, the features associated with red in redden cannot antecedent-bind the adjective, since red in redden does not c-command the empty A in complement position, assuming redden to be in fact composite. The adjectival complement moves to V in conformity with the Head Movement Constraint, adjoining to the left of V, in conformity with the prevailing word-internal headedness arrangement of the languages of the world.
Accordingly, 94 is the correct structure for the sky redden, if empty A is the trace of head movement and -en is the verb hosting the incorporated adjective.
The setting sun reddened the sky. The fall in prices narrows our options. According to our general assumptions, the verb phrase of 95a is abstractly as depicted in A by V2 and V2 by V1. In the derived structure, the DP the sky is governed and Case-marked by the complex verbal head V1 redden. Our explanation is this. Rather, the phenomenon we have been calling by that name is in reality merely the binding relation that holds between the semantic features of a verb phonologically overt now and features of the nominal head of its complement.
This in turn is a result of the selectional relation between the verb and its complement. Our purpose here is merely to lend a certain degree of crosslinguistic perspective to this work. The DP e. This scenario for English is familiar from chapter 1. In this chapter, a similar scenario emerges in the Native American examples, which could have been chosen arbitrarily from any number of other languages of the world.
The four languages we discuss are. Hopi, also Uto-Aztecan, spoken in northern Arizona section 4. However, this failing should not interfere in any way with the principal points of data and theory inherent in this chapter.
Navajo In this section, we confront our basic theory of argument structure with certain aspects of the grammar of the Navajo verb, limiting our discussion primarily to the simple transitivity alternation reviewed in section 4. Categorially, however, it is a light verb, symbolized v, suggesting voice or valence see Rice Here we are concerned with its productive use as a light verb with the value transitive, capable of assigning abstract Case to the object of a verb.
Native American Perspective b. We will refer to these as oblique object pronominals. Returning to the simple transitivity alternation, it is a fact that the set of Navajo verbs participating in it is rather large. The following list is a sample, in addition to the verbs illustrated in 3 — 6 , extracted from Young, Morgan, and Midgette ; page numbers are cited in brackets see that source for details: Setting aside verbs that are obviously transitive, verbs of dubious transitivity lacking a simple transitive counterpart are probably as numerous as those that enter into the simple transitivity alternation.
We will refer to these as unergatives, following current linguistic usage cf. The verbs upon which the causatives of 7 — 9 are based represent this type. Additional examples are cited in It is not known whether this, in and of itself, inhibits those verbs from participating in the transitivity alternation. The principal point here is that there are verbs that participate in the simple transitivity alternation and there are verbs that do not. The verb of 8 belongs to the class whose members cannot alternate, the class represented by the verbs in Thus, while it is possible to form a causative of that verb, as seen in 8b , the simple transitive is not possible.
It is nevertheless a fact that this verb cannot form a simple transitive, of the type represented in 3 — 6. The same can be said of the verbs of 11 and the general class of verbs they represent. Why do they resist simple transitivization? Why is there this asymmetry among the verbs of Navajo?
This asymmetry is not accidental. Even in the absence of a theory, the systematic nature of this asymmetry is evident immediately when the Navajo facts are compared with the corresponding phenomenon in other languages. Consider, for example, the following sample verbs from English, an Indo-European language; Miskitu, a Misumalpan language of eastern Nicaragua and Honduras; and Navajo.
These details aside, the verb classes distinguished by these three languages are identical, for all intents and purposes. It is virtually impossible for this to be an accident.
Something fundamental underlies this coincidence in lexical behavior. This is the question, then: What is behind the asymmetry of the transitivity alternation?
What factor permits the alternation in the case of the verbs in 3 — 6 , 10 , and 13 , and what factor blocks it in the case of the verbs in 7 — 9 , 11 , and 14? We know that alternating verbs share the feature that the subject of the intransitive alternant appears as the object of the transitive.
This suggests that it is internal to the lexical structure. And this, if true, leads us to assume that this verb, and other alternating verbs like it, must realize one of the lexical structures that contains a Native American Perspective predicate-like subconstituent, and therefore a subject, internal to the lexical structure itself—so these verbs must realize the structure depicted in 1. The nonalternating verbs have only an external subject, one appearing in sentential syntax but not in the lexical argument structure itself.
The subject is symbolized DP in 16 , since it will appear in sentential syntax as a full nominal argument. The former appears as the complement of the latter. This appears in the uppermost head position, where it governs the internal subject, DP. This argument will, of course, function now as the sentential syntactic object, not the subject.
The structure portrayed in 18 , we claim, has a nonpredicative root R , therefore giving rise to an unergative, nonalternating verb. This cannot be transitivized in the simple manner of an alternating verb i. Be this as it may, it all follows from the essential properties of the elements involved: The nature of R in 18 is clear.
It is therefore not surprising, perhaps, that nonalternating verbs in many languages are based on roots belonging morphosyntactically to the category N. This idea is supported not only by the syntactic behavior of these verbs but also crosslinguistically by the fact that in many languages, unergatives are verb-noun compounds i. The following verb themes exemplify this type with bracketed numbers corresponding to Young and Morgan , dictionary section: The causative is built upon the full verbal projection of the unergative and, therefore, includes the subject of the latter.
This surfaces as an oblique object b- glossed 3 above , in this case , held by Athabaskanists to be an incorporated postpositional phrase with null postposition.
We assign the abstract structure 25 to the verb of 24 see Hale a, , for remarks on the causative and the surface ordering of the morphemes internal to the Navajo verb. In summary, Navajo, like English, possesses a class of labile verbs that enter freely into the simple transitivity alternation, and it possesses as well a class of verbs that do not participate in the simple alternation, requiring use of the true causative construction.
Ulwa A pervasive feature of the Misumalpan languages is the existence of transitivity alternations marked by corresponding alternations in verbal morphology. Kuring abuk-d-ida. Kuring abuk-pa-h. Kuring batirh-da-rang yataihdaram laih. Turum ya waya batirh-p-am was ya utuhdangh. It is the latter, we must assume, that functions as the head of the lexical projection in which it appears. It is the true verb, so to speak, like the nonovert verbal head postulated for the English deadjectival verb clear in 1.
It is not surprising—and not an accident, presumably—that the root elements in some of the alternating da-themes of Ulwa also enter into the formation of adjectives in the language. The root is morphologically nominal, but it functions as a stative predicator in the derived form to which we have applied the term adjective.
The verbs of 26 — 28 are based on roots that participate in this adjectival use, as shown in 29 , where -ka is the construct morphology. The root element, R, corresponding here to abuk-, batirh-, and sang-, is perhaps of indeterminate or neutral category.
But it has a lexical property of consequence: We claim that the root element in these structures forces the head V i. And it is this property that accounts for the transitivity alternation. This is appropriately situated in relation to V1 , its governor and potential Case assigner in sentential syntax. The structures in 30 and 31 are abstractions, representing just the syntactic relations involved, not the morphology. The following is a sample listing of Ulwa da-theme verbs alternating with pa-theme transitives.
The verbs are given in their bare theme forms, with -da in the intransitive, -pa in the intransitive. Accordingly, their intransitive alternant is of the form shown in 30 , and their transitive alternant takes the form shown in While this is a fundamental characteristic of adjectives, given their attributive and predicative functions, the root elements in the verbs of 32 are not always attested independently in an adjectival use.
At this point, we do not know in which cases the missing use is principled and in which cases it is simply a gap in the record. In fact, this illustrates one of the reasons why the sort of theoretical speculation Chapter 4 we are engaging in here is appropriate even at this relatively adolescent stage of dictionary making.
In this instance, our theoretical speculations tell us that we must, at some point, determine for every verb the full range of lexical projections in which the root R may appear. For example, we must know whether the root element in all of the verbs of 32 appears independently in the adjectival form and partakes of the corresponding adjectival syntax? If not, why not? The dictionary must, it seems to us, be a resource that, to the extent possible, purports to answer questions of this nature.
We will return to this topic at a later point. Given the generality of the grammatical and lexical principles involved here, it is therefore not surprising that many of these Ulwa verbs translate into English as verbs that alternate in that language as well e. And we expect the principles observed in Ulwa to be replicated to a degree in the other Misumalpan languages.
The da-theme alternating verbs of Ulwa are not always paired with patheme transitives. Some of the latter are listed in This is another among many matters that remain to be dealt with properly.
Not all Ulwa labile verbs have intransitive themes based on -da. This element is of some historical interest for Misumalpan, given that it has an apparent cognate in the related language Miskitu Hale and Salamanca Alas baka-ka yam-ka ala-t-ang. The unifying feature of both types of verbal themes considered here is presumably to be found in the lexical character of the root R. Part of the theoretical interest in labile, or alternating, verbs lies in the contrast between these and another large class of verbs, namely, the nonalternating verbs.
As shown above, many Ulwa intransitives in -da have Native American Perspective transitive partners. But many do not. Like these Ulwa verbs, their English translations also fail to alternate, as a rule, permitting only the intransitive use in sentential syntax. For example: I made the child cry. The book made me yawn. The teacher has the children playing. The ideas intended here are perfectly easy to express in Ulwa, using the productive causative construction e.
The same is true in English. What is the reason for this? Given the striking meaning correlation between English and Ulwa, it is tempting to lay the entire business at the feet of semantics. And at some deep, as yet largely inaccessible level of linguistic form this is quite probably where the matter resides. But at the level at which we are now able to operate, semantics is too unreliable, partly because we simply cannot say what the meanings of words are.
Good reason for being cautious here comes from crosslinguistic considerations, ironically the very area that inspires optimism much of the time. In Hopi, the verbs that translate many of the unergatives of English and Ulwa do indeed participate in the very transitivity alternation we have been examining here Jeanne and Hale ; and see below.
We are stuck then with what is observable, namely, the syntactic behavior: Assuming that we are correct in assigning the structures 30 and 31 to Ulwa alternating verbs, we can express the phenomenon of nonalternation in a simple and straightforward manner. The root elements R of nonalternating i. Thus, the argument structures of the verbs of 39 have fundamentally the form shown in The subject of an unergative, like that of a transitive e.
We have given a partial account of the alternating and nonalternating verbs of Ulwa. As in the previous sections, our primary focus will be the standard transitivity alternation, though some attention will be given to the so-called applicative construction. Our interest here, however, is essentially syntactic. The object of the derived transitive corresponds to the single argument i. I AUX.
English expresses this alternation with so-called zero derivation: Through derivation, these yield a class of intransitive verbs, which in turn enter into the so-called causative-inchoative alternation. In languages that recognize adjectives as a distinct morphosyntactic category, that category is a common source for verbs participating in the causative-inchoative alternation. However, this property is directly relevant to the larger issue that interests us, namely, the constraints on possible argument structures.
In this light, let us consider the nouns and corresponding derived denominal verbs illustrated in Hopi, section 4. Baker In addition, a quantifying preverb may appear. And similarly for the others. In general, why are some derived transitives causative while others are applicative?
Is this simply random? Or is there some principled basis for it? The contrast noted here is the same, in essence, as a contrast that is virtually universal among the languages of the world. While both deadjectival and denominal verbs participate in processes of transitivization, denominal transitives are regularly applicative, not causative.
It is therefore clear why ki: In short, as 54 shows, there is no position for an argument DP corresponding to the grammatical object in the hypothetical causative. Therefore, there is no position for the DP argument that is necessary in the derived transitive. In and of Chapter 4 itself, 54 is impossible, so far as we know. These have applicative, rather than causative, function. How do these arise? The invariant fact about this element is that it forms transitive verbs.
Suppose it is semantically empty, basically. The inner structure associated with the applicative relation is the dyadic type: Analytic forms also exist in the case of verbs that we take to be adjectivebased.
In Miskitu, for example, some adjectives primarily borrowings from Northern Sumu form synthetic verbs, alternating between transitive and intransitive in the expected way. These verbs, like most of the alternating and nonalternating verbs of English cited in the course of this discussion, give no direct evidence of an adjectival or nominal source. Nonetheless, the verbs of 47 behave like the overtly deadjectival verbs of 43 and 45 , and the verbs of 54 behave like the denominals of 48 in lacking a causative derivative.
It is the behavior of a verb, not its form, that gives evidence of its argument structure. Certain details must be left underdetermined, such as the morphosyntactic category of the overt element, but this is not strictly speaking relevant to the argument structure and corresponding syntactic behavior of the derived verb.
Verbs of this type are the so-called unergatives of current linguistic literature. Since 58 is a monadic structure, it follows that it cannot be further transitivized as a causative. It is, therefore, a dyadic structure embedded in a monadic one, as shown in The central theme of this discussion has been the idea that certain aspects of predicate argument structure are invariant.
The claim is that the abstract monadic and dyadic structures involved in the derivations suggested here are basic to all languages.
Neither of these themes could be investigated without the linguistic diversity which exists in the world and which, sadly, is seriously imperiled in these times. We would not be able to learn as much about these matters if, say, English were the only language. Eventual understanding of unaccusativity in English did not have its origins primarily in research on that language; rather, it was due largely to foundational research done on Dutch and Italian, much enhanced by subsequent work on a variety of languages, and with important feedback from English see Levin and Rappaport Hovav If linguistic diversity is valuable in identifying problems i.
Thus, the idea that the nonalternating verbs are noun-based is suggested, somewhat, by the English data, but it is strongly supported by the existence of languages, like Basque, that use the hypothetical [V V N] overtly in the sentential syntax projected by lexical items corresponding to verbs of the type in question. And it is supported as well by the Tanoan languages, for example, where verbs corresponding to work, speak, whistle, laugh, cry, sing, and others, are likewise overtly noun-based, taking the form of N-V compounds.
In the absence of diversity, there is little one can say about particular linguistic problems in many cases, perhaps in most cases; and, as noted earlier, the very existence of a linguistic problem is itself something that often goes completely undetected without the backdrop of diversity. This function of diversity is well known, and it is exploited with energy and enthusiasm by linguists in all frameworks. In the following section, these remarks will be further supported by material from Hopi, another member of the Uto-Aztecan family.
This element deviates from the straightforward causative meaning under certain conditions. The present discussion is a preliminary and highly tentative introduction to the study of Hopi transitivization from a comparative and theoretical perspective, with primary attention to -na and its interaction with other elements. The following examples of -na are taken from Jeanne Pokyaya wake. I Pokyaya-ACC wake. Pokyaya cry-LAWU. With a class of exceptions to be noted, virtually any monadic verb in Hopi can be transitivized in this manner.
The issue we wish to address here is an extremely small and narrow one, but it has implications for a general study of argument structure relations in Hopi, in Uto-Aztecan languages, and in general.
Um yan-wat kii-ta-ni. ACC little. Pam piiki-t nitkya-ta. Pam koongya-y piiki-t nitkya-toy-na. Pam pas-ta. Pas pu-ma nu-y qa qeni-toy-na-ya. Unlike in the standard causative-inchoative alternation, in the alternation seen here the object of the derived transitive is, so to speak, an introduced argument, in the sense that it does not correspond to the subject of the corresponding underived verb; in fact, it corresponds to no argument of the underived verb.
The problem we wish to address here can be stated in terms of this vague semantic label, as in There are two issues here, in fact. First, why do the verbs of 77 have the benefactive meaning? For example, why must 79 have the benefactive meaning and not the causative meaning?
We are concerned in particular with the grammatical, or structural, aspects of the problem and, accordingly, we are especially interested in accounting for the syntactic observation embodied in In particular, the object of the verbs of 77 does not correspond to the subject of the underived verb.
This is in contrast with the situation represented in the canonical causative-inchoative of the verbs of 65 — With this background, we can begin to consider answers to the question posed in 78 and, correspondingly, an explanation for the structural observation in We emphasize that this is a mere beginning, since we are investigating only a small part of a large domain. We seek answers that are consonant with general principles of Universal Grammar. The relevant observations are these: These formulations are not exceptionless, but they point in a familiar direction, namely, that expressed in That is to say, whether a verb undergoes simple transitivization depends upon its makeup.
The logical structure of syntactic theory. Columbus, OH: Brandeis University, MS. On the syntax of verb-particle, triadic, and causative constructions. Oxford University Press. Modularity of mind. Where cognitive science went wrong. Impossible words? Linguistic Inquiry Argument structure.
Prolific domains. On the anti-locality of movement dependencies. John Benjamins. On argument structure and the lexical expression of grammatical relations. The view from Building Essays in honor of Sylvain Bromberger, ed.
The basic elements of argument structure. Cambridge MA: Bound features, merge, and transitivity alternations. Instituto Universitario Ortega y Gasset. Distributed morphology and the pieces of inflection. The view from building A minimal ish linking theory. University of Maryland, May 16, Merge, conflation, and head movement: The First Sister Principle revisited.
State University of New York, November , A minimalist theory of construal. To appear. Understanding minimalism. An introduction to minimalist syntax. Kenneth Locke Hale. Language Remarks on denominal verbs. Complex Predicates, ed.
Palo Alto, CA: CSLI Publications. On the double object construction. Chains and phono-logical form. Proceedings of the 22nd Annual Penn Linguistics Colloquium. University of Pennsylvania Working Papers in Linguistics 5. A minimalist approach to A- movement and control. Lingua The raising of predicates. Category-less roots in syntax and word learning: Objections to Barner and Bale University of Maryland, MS.
Related Papers. Two Argument Restriction. By Eva Juarros Daussa. A layered derivation approach to conflation. Verb Movement and Minimality. Small Clause Results Revisited.
By Mateus Ritter. Hyponymous objects and Late Insertion. By Jason D Haugen. Download pdf. Remember me on this computer. Enter the email address you signed up with and we'll email you a reset link.