Show Summary Details

Page of

PRINTED FROM the OXFORD RESEARCH ENCYCLOPEDIA, LINGUISTICS (linguistics.oxfordre.com). (c) Oxford University Press USA, 2016. All Rights Reserved. Personal use only; commercial use is strictly prohibited. Please see applicable Privacy Policy and Legal Notice (for details see Privacy Policy).

date: 21 October 2017

Argument Realization in Syntax

Summary and Keywords

Words are sensitive to syntactic context. Argument realization is the study of the relation between argument-taking words, the syntactic contexts they appear in and the interpretive properties that constrain the relation between them.

Keywords: semantic roles, predicate decomposition, argument alternations, affectedness, event structure, agent, patient, incremental theme, linking generalizations, thematic hierarchies

1. Argument Realization

Words are sensitive to syntactic context. Sentences can be syntactically well-formed, but deviant because of a mismatch between the words and syntactic structure. Nonetheless, there are different sorts of mismatches between words and their syntactic contexts, and it is the task of the linguist to characterize the precise difference between them. For example, Chomsky (1957) pointed out that example (1) is well formed syntactically, though anomalous because of, among other things, the contradiction between being green and colorless and because we do not typically attribute sleeping to ideas. The deviance of examples (2a) and (3a) is of a different sort: the demands of the verbs on their syntactic environment appear not to be met in a serious way. The verb put requires two syntactic complements, as shown in (2b). Steal allows only one bare determiner phrase (DP) complement (3b). Though the verb give allows the syntactic environment of (3c) alongside that of (3d), steal does not allow (3a) alongside (3b), under the intended interpretation that the ball was taken from Jack.

(1)

Argument Realization in Syntax

(2)

Argument Realization in Syntax

(3)

Argument Realization in Syntax

A theory of argument realization is meant to account for the distribution of argument-taking predicates—prototypically verbs1—in various syntactic contexts and the regularities of interpretation that accompany the association of predicates with syntactic contexts. In the next few paragraphs, I offer an explication of the term argument realization, focusing first on the nature of the relation between verbs and their syntactic dependents.

Verbs classify happenings in the world. The application of a verb to a slice of external reality indicates that that slice of reality is cognitively construed as an event—an event of a particular sort. In formal semantic terms, assuming an ontology that includes events, verbs are predicates of those events classifying the events they apply to as events of particular kinds. One aspect of the classification that each verb provides is a specification of the number of participants in the event. For an event to qualify as a putting event, for example, it must have three participants. The verb put classifies events as, among other things, having three participants: a putter, a thing put (or a locatum), and a location. In a linguistic description of an event, typically expressed in the core of a verb-headed clause, the linguistic units that correspond to the participants in the event are referred to as the arguments of the verb. The number of arguments of a verb can vary between zero (weather verbs2), one, two, and three. In morpho-syntactic terms, the arguments typically correspond to the subject, direct object, and obliques. The term ‘argument’ comes from the view of verbs as functions that take arguments and output a truth value or some other semantic value. Sometimes the term argument refers to the semantic argument of a predicate, and sometimes to the syntactic constituent that represents this argument in the syntax. The idea that a syntactic constituent is the expression of the argument of a predicate is behind the term argument realization.

Besides the number of participants in a particular event type, we can talk about the nature of the roles that the participants play in the event-type. The association of a DP/NP with a particular grammatical relation brings with it the interpretation associated with one of the participant roles.

(4)

Argument Realization in Syntax

Example (4) does not only convey that Martha and Dan stand in a loving relation to one another, but also that it is Martha who experiences the love (possibly unrequited), and that Dan is the target or stimulus of this love. These interpretations of Martha as the lover and Dan as the target of love is a consequence of their expression as subject and object, respectively, instantiated in English in terms of word order. In other languages, other strategies are employed for the encoding of grammatical relations of subject and object (see, for example, Croft, 1991).

There are pervasive generalizations concerning the ways in which the morpho-syntactic properties of arguments match up with their semantic interpretation as realizing the participant roles. These generalizations often fall under the rubric of ‘linking generalizations,’ a term first coined by Carter (1988) (see also Croft, 1991; Fillmore, 1968; Jackendoff, 1972; Levin & Rappaport Hovav, 1995; Ostler, 1979). For example there is no English verb—and probably cannot be—which could be used as in (5) to express the fact that there was an event of the opening of the door, and Sam was the individual who brought about this event. This is because sentences that express changes of state express the theme of the change of state as direct object and the cause of the change of state as subject.

(5)

Argument Realization in Syntax

This example illustrates that the meaning of the verb—the kind of event it classifies—has some contribution to make not only to the number of syntactic dependents it has, but also to the way the dependents are interpreted as participants in an event.

The unmarked way in which participant roles are matched up with argument positions is part of the linguistic knowledge of a speaker. Since speakers are able to coin new words and hearers are able to assign semantic interpretations to these newly coined words, it is apparent that speakers do not need to learn many aspects of these associations for each new word they learn. A theory of argument realization is a theory of how this association is represented in the minds of speakers and implemented in a theory of grammar. Research on argument realization is as old as grammatical theorizing,3 but serious work in argument realization in the framework of generative grammar begins with Fillmore (1968).

Data as shown in (1) through (3) lead to the appearance that verbs are associated with a ‘basic’ set of arguments, and therefore, each verb can be associated with a single fully articulated lexical semantic representation that determines its syntactic environment(s). The development of a theory of argument realization requires has traditionally been taken to involve the following components: isolation of the aspects of the meanings of verbs that determine (or correlate with) argument realization; determination of the nature of lexical representation; and determination of the nature of mapping between lexical representation and syntactic representation. Broad consensus is emerging among researchers regarding the semantic aspects of meaning that are relevant to argument realization, though there is still much room for fine-tuning. In contrast, there is much controversy still on the nature of lexical representations, syntactic representations and the relation between them. Despite data of the sort represented in (1) through (3), many verbs can appear with a much wider range of argument realization options, which makes deriving them all from a single lexical semantic representation challenging. The striking variability of argument realization options, and even the number of associated arguments available for verbs, appears to be a pervasive phenomenon and has lead researchers to a revised view of the division of labor between the lexicon and syntax. These issues in turn have ramifications for the nature of both lexical representation and its relation to syntax. Current theories differ significantly on these issues.

This article is organized as follows. Section 2, “Traditional Components of Theories of Argument Realization” lays out the basic ingredients of a theory of argument realization, which is focused on accounting for the intuitions behind the kind of data laid out in examples (1) through (3). A theory that elucidates the relation between verb meaning and syntactic representation must articulate the nature of a lexical representation on the basis of which appropriate generalizations can be formalized. Section 2.1, “Representing Grammatically Relevant Aspects of Meaning,” deals with the question of the nature of this representation, presenting two approaches; elements of both are shown to be necessary. A theme that runs covertly through this section is the need for a representation that can serve as the basis for an explanatory theory of mapping. Section 2.2, “Conceptual Content of Grammatically Relevant Aspects of Meaning,” deals with the conceptual content of such a representation. Section 2.3, “Theories of Linking,” shows how aspects of lexical representations are taken advantage of in theories of linking. Section 3, “Challenges,” introduces data that widen the empirical basis for theory of argument realization, and that challenges many of the traditional assumptions about what should be considered lexical properties of predicates. Section 4, “Where Do Arguments Come From? Constructionist and Neo-Constructionist approaches to Argument Realization,” introduces current approaches to the relation between the lexicon and syntax that attempt to provide solutions to the challenges laid out in Section 4. Section 5, “Summary and Prospectus,” summarizes the discussion and mentions other avenues of research and remaining challenges.

2. Traditional Components of Theories of Argument Realization

Though there is a correlation between meanings of verbs and the syntactic contexts they appear in, all work in this area has assumed that not all aspects of the meaning of verbs are relevant to argument realization. Pre-theoretically, we might then distinguish between grammatically relevant and grammatically irrelevant aspects of verb meaning, a move made by almost all current theorists4. Grammatically relevant aspects of verb meaning are seen to be elements of meaning shared by large classes of verbs. By virtue of this fact, these aspects of meaning can figure in generalizations concerning argument realization, since their wide applicability makes them potentially relevant for newly coined words. The grammatically irrelevant aspects of meaning are what distinguish between the various members of classes of verbs defined on the basis of the grammatically relevant aspects of meaning. From the point of view of linguistic theory, we need to address two partly independent issues. The first is what method to use for representing grammatically relevant aspects of meaning (section 2.1) and the second is the identification of the conceptual content of these meaning components (Section 2.2). One primary testing ground for theories of lexical representation is the mapping from lexical representations to syntax, discussed in Section 2.3. A theory of lexical representation that allows for a streamlined and explanatory description of mapping is desirable and will be preferred over a theory which does not facilitate the formulation of mapping strategies.

2.1. Representing Grammatically Relevant Aspects of Meaning

There are two broadly characterized approaches to the representation of the grammatically relevant components of meaning (see Croft, 1991, 1998; Davis, 2001; Jackendoff, 1983, 1990; Levin & Rappaport Hovav, 2005) for lengthy discussions of these approaches. The first might be called a role-centered approach and the second a predicate-centered approach.

2.1.1. Role-Centered Approaches

On the role-centered approach, the grammatically relevant semantic representation of a verb consists of a list of semantic (alternatively called, thematic, and sometimes “theta”) roles. Semantic roles classify arguments of verbs in terms of the nature of the arguments’ involvement in the events denoted by the verbs. The composition and size of the set of semantic roles is a matter of debate, but the set in (6) is typical. (“Object” is equivalent to what is known more commonly as a “patient” or “theme,” and “result” is equivalent to what is known as “effected object.”)

(6)

Argument Realization in Syntax

Each semantic role defines a natural class of arguments, with members of this natural class having a common semantic relation to their verbs and usually also shared options for their morpho-syntactic expression. Semantic roles, then, can be viewed as labels for equivalence classes of arguments, and the goal of a theory of semantic roles is the identification of a set of semantic roles that are applicable to any argument of any verb and an explication of how these roles figure in argument realization.

The use of semantic roles can be likened to the use of features in phonology. The latter distill from the wide range of phonetic detail those aspects of sounds that are phonologically relevant. The former distill from the even wider range of semantic detail those facets of meaning that are grammatically relevant. A large number of verbs, then, may have identical lexical semantic representations that abstract away from idiosyncratic components of meaning. For example, the large class of change-of-state verbs can be associated with the semantic representation in (7), in which the nature of the specific change of state is not grammatically relevant, and, hence, not represented.5

(7)

Argument Realization in Syntax

The choice of features or roles in both phonology and lexical semantics is justified to the extent that they define equivalence classes that recur in rules of phonology or morpho-syntax. Phonological features generally appear in clusters, defining natural classes of phonemes, which are shown to be relevant for the formulation of phonological rules. Semantic roles are argued to help define the argument structure of verbs and of natural classes of verbs that share grammatical properties, like the participation in various argument alternations (see section 3.1, “Multiple Argument Expression,” below).

Unstructured semantic role lists, as a grammatically relevant lexical semantic representation, predominant in the 1970s and 1980s, have received abundant criticism, and most current sophisticated theories of argument realization have moved away from attributing to them any theoretical status, though they continue to be used as a convenient shorthand for talking about the arguments of verb. The following is a summary of the types of criticisms lodged at the use of semantic roles. First, it has proven difficult to provide a closed set of universal semantic role labels relevant for all aspects of argument realization, or even the beginnings of a method to show how to do so, in the way that one can argue for a closed set of phonologically distinctive features or lexico-syntactic categories. This is not to say that there is complete consensus about the list of phonological features or lexic-syntactic categories, but there is a broad consensus on the way to go about arguing for the status of a feature or a category, and the theoretical grounding of the constructs is fairly clear to practitioners. This is not the case with regard to semantic roles. Second, it has proven to be just about impossible to exhaustively analyze all verbs in terms of the semantic roles of their arguments. In certain cases, there is no clear way to determine the exact role an argument bears. The arguments of some verbs—like confirm or require—seem not to be naturally associated with any widely recognized semantic role label. Third, because a list of semantic role labels has no internal structure, there is no way of representing what constitutes a natural list of roles. Although many verbs are associated with the semantic role lists—“agent, patient, instrument” (e.g., break, cut, mix) or “theme, source, goal” (e.g., go, come, return), a list of the form “location, experiencer, patient” seems decidedly unnatural. Yet, there is no way to distinguish between the natural groups of semantic roles and the unnatural ones. Finally, different linguistic generalizations appear to require semantic roles of different grain-size, and there seems to be no principled way to come up with a single representation composed of semantic role labels that can serve as the basis for all generalizations, which purportedly make reference to semantic role labels. For example, different generalizations make reference to more broadly or more narrowly construed notions of ‘agent,’ and different grammatical generalizations make reference to more broadly and more narrowly construed notions of ‘instrument.’ For some generalizations, ‘agent’ and ‘instrument’ can be seen as subtypes of ‘cause,’ but others need to distinguish between the categories. For extensive discussion, see Croft (1991, 1998), Davis (2001), Dowty (1991), and Levin and Rappaport Hovav (2005), Rappaport and Levin (1988).

Some of these difficulties are overcome by attempts to provide a more solid semantic grounding for the notion of a semantic role. The most coherent explication of the nature of semantic roles takes them as defining (clusters of) lexical entailments imposed by verbs on their arguments (Dowty, 1989, 1991; the approach is further developed in Ackerman & Moore, 1999, 2001; Beavers, 2010; Primus, 1999, among others). Every verb specifies certain entailments that hold of its arguments. For example the verb drink specifies a number of lexical entailments associated with the denotation of the subject argument. Use of this verb entails that the denotes of this argument initiates and undergoes a process of ingestion. A different set of entailments is associated with the denotation of the object argument (for example, that it is ingested and is of a liquid state). The semantic roles that are the basis of a theory of argument realization are not defined on a verb-by-verb basis (what Dowty, 1989 calls ‘individual thematic roles’), but represent sets of lexical entailments that recur throughout the verbal lexicon. The entailment that something is liquid is, then, probably not a likely candidate, but the entailment that something undergoes a change of state probably is. Natural classes of arguments result when arguments of a large number of verbs share certain lexical entailments, and natural classes of verbs result when the verbs are associated with arguments, all of which have shared lexical entailments. For example, a large number of verbs specify about the object argument that it undergoes a change of state. This lexical entailment distinguishes between verbs like break, melt, widen, and cut, which impose this lexical entailment and verbs like see, disregard, meet, bump, and touch, which do not. Labels that figure in argument realization generalizations (what Dowty, 1989 calls L-thematic roles) can be considered to have linguistic significance and the argument realization generalizations provide initial justification for their use. The grammatically relevant representation of verb meaning on this approach consists of a specification of a list of semantic roles for a verb, where the semantic roles represent only those (clusters of) lexical entailments that figure in grammatical generalizations.

This approach helps deal with some of the problems associated with atomic semantic roles, because different semantic roles can share subsets of lexical entailments. For example, agent will have all the lexical entailments of cause and instrument, but the latter will have only some of the entailments associated with agent. Another advantage of this approach to thematic roles is that the lexical entailments themselves may only combine naturally in some ways and not others due to how entailments work independently, and this can rule out “unnatural” combinations (see e.g., Beavers, 2010).

A similar approach is taken by researchers who suggest that semantic roles can be provided with definitions in terms of a small set of binary semantic features. This approach is adopted by Anderson (1971, 1977), Ostler (1979), Reinhart (2002), and Rozwadowska (1988, 1989), among others, and it returns us to the parallel drawn between semantic roles and phonemes, but now with both being characterized in terms of distinctive features.

Generalizations that are apparently stated with respect to semantic roles of different grain-sizes can be accommodated by referring to feature specifications of greater or lesser generality. Feature decomposition approaches mesh naturally with the approach to semantic roles developed by Dowty (1989), which take semantic roles to represent clusters of lexical entailments, if the features are understood as lexical entailments. Nonetheless, Dowty (1991) is careful to distinguish his lexical entailments from features in phonology and syntax (e.g., functional features), since the latter are parts of a coding system and are likely to draw sharp binary distinctions between categories, whereas the former classify aspects of events in the world and are less likely to have clear cut boundaries in the same way.

However, as discussed by Rappaport and Levin (1988), approaches based on feature decomposition still do not provide any insight into what constitutes a natural set of semantic roles that can be associated with an individual verb, and in many ways, neither does Dowty’s approach (see, for example Primus, 1999). Furthermore, unless the number of features is small, as in Reinhart’s and Rozwadowska’s work, the number of attested feature combinations is usually dramatically less than the number of possible combinations, so that these approaches give rise to uninstantiated roles. For example, Ostler (1979) makes use of 48 roles, which he defines in terms of eight features, yet eight features are sufficient to distinguish 256 roles—over 200 more roles than Ostler uses (Kisala, 1985). To account for these discrepancies between predicted and existing roles, a theory of the features that define semantic roles must be supplemented with a theory that predicts the possible combinations of these features. These shortcomings of semantic role approaches are rectified by predicate-centered approaches.

2.1.2. Predicate-Centered Approaches

We have seen that the move to ground thematic roles in lexical entailments helps with many of the difficulties with approaches based on atomic thematic roles, but it still leaves open the question of ‘where do thematic roles come from,’ and what can constitute a natural list of thematic roles. Predicate-centered approaches take the linguistically significant elements of verb meanings to derive from elements relevant to the classifications of events, from which participant classifications (thematic roles) naturally follow. Predicate-centered approaches represent the components of meaning that recur across significant sets of verbs in terms of one or more universally available primitive argument-taking predicates, which can be composed according to a well-defined calculus, and which are taken to be the basis for the linguistic encoding of any event. In current literature, these representations are often referred to as ‘event structures.’ The verb’s arguments are represented by the open positions associated with the predicates, and semantic roles are identified with the argument positions of particular primitive predicates, making them explicitly derived notions, following a suggestion first made by Jackendoff (1972). For example, BECOME is often posited as the element common to event structures of verbs expressing a change of state, including transitive and intransitive break, open, and dry (see example 8 below); the semantic role of patient can then be identified with the open position of this primitive predicate. Extensive explorations of verb meaning using predicate decompositions are found in Dowty (1979), Jackendoff (1983, 1990), Croft (2012), Foley and Van Valin (1984) and Van Valin and La Polla (1997).

Primitive predicates were initially introduced by generative semanticists6 (Lakoff, 1968, 1970; McCawley, 1968, 1971; Ross, 1972) to capture various entailment relations, including relations between sets of sentences containing morphologically—and, thus, semantically—related words. For example, the fact that transitive/intransitive pairs, such as Beth cooled the soup and The soup cooled, share the entailment “The soup became cool” and the fact that the same selectional restrictions hold for the object of the transitive verb and the subject of the intransitive verb are accounted for by positing a shared component of meaning, something like “become (COOL).” COOL represents the idiosyncratic aspect of this specific verb, and the predicate BECOME, with appropriately articulated truth-conditions (Dowty, 1979), is shared by all change of state verbs. COOL is, of course, also shared by the corresponding adjective. The fact that there are many triads like this in English strengthens this approach, accounting for much data across the lexicon. As Dowty (1991) points out, there is no contradiction between this view of semantic roles and the lexical entailment view, since the primitive predicates can themselves be associated with lexical entailments for their arguments.

The set of primitive predicates typically proposed is far smaller than the set of semantic roles proposed. That being so, the idea of coming up with a closed set of roles becomes a more manageable enterprise, partly because there is more of a consensus on the motivations for postulating primitive predicates (though see Levin & Rappaport Hovav, 2005, p. 74 for a cautionary comment). A major advantage of this kind of representation is that it immediately defines the set of allowable lists of semantic role labels for any verb. Once roles are defined in terms of positions in predicate decompositions, no verb can be associated with a set of roles that cannot be thus defined. The roles of varying grain-sizes that are needed for different generalizations can be derived by combining information from the predicates and from the entailments derived from information about the fillers of the argument positions. For example, agents and causes can in some cases be associated with the first position of a predicate CAUSE, or DO, where the agent interpretation is derived from the predicate and properties of an argument denoting a human (Holisky, 1987, Van Valin & Wilkins, 1996). See in particular Beavers (2010) and Rappaport Hovav and Levin (2007) for an approach that supports combining primitive predicates with lexical entailments.

Let us return for a moment to the suggested analysis for the verb cool by generative semanticists mentioned in the previous paragraph. The analysis in effect factors out what has traditionally been considered the meaning of a verb into two distinct types of components. One is idiosyncratic to a particular set of morphologically related words, now commonly referred to as a ‘root’ (Alexiadou, Anagnastopoulou, & Schäfer, 2015; Harley, 2014; Pesetsky, 1995). For example, an abstract element COOL can be considered the lexical core of meaning common to the adjective cool and both the transitive and intransitive variants of the verb (and other words as well, such as coolness). The second type is shared by a very large set of verbs in any language and forms the basis of linguistically encodable event types, what we have referred to as the grammatically relevant component of meaning or event structure. In more recent work, there has been an emphasis on the structural relation between the root and event structure. This can be seen in the examples (8), based on representations of the sort used in Rappaport Hovav and Levin (1998). Each example represents a causative change of state, where the element that distinguishes between them—the root—is in all cases the argument of a predicate BECOME. The same representation also shows what is common to the adjective, intransitive verb, and transitive verb of the triads mentioned above. The intransitive verbs lack the first argument of cause and the adjective lacks the predicate BECOME altogether.

(8)

Argument Realization in Syntax

The examples in (9) illustrate the representations of another set of grammatically similar verbs. In these examples, the root appears subscripted to a predicate, interpreted as a modifier of that predicate. ACT is generally assumed to underlie activity predicates, though it is sometimes represented as DO.

(9)

Argument Realization in Syntax

Rappaport Hovav and Levin (1998) argue that the roots are ontologically typed and that this typing determines how they are integrated into event structure representations, for example, as event modifiers as in (9) or arguments of primitive predicates and in (8). To the extent that there are indeed generalizations with explanatory import regarding the connection between ontological typing of roots, as suggested for example in Rappaport Hovav and Levin (2010), their interpretation and their positions in event structure, this supports both the decomposition approach and the ontological typing of roots.

Another advantage of this approach is that it introduces hierarchical structure into the lexical semantic representation. The hierarchical structure is exploited in accounting for various kinds of scopal ambiguity. For example, a verb like open is assumed to have a complex event structure similar to those in (8) above, while play is assumed to have a simplex event structure similar to that in (9). The complex event structure attributed to open can help explain the fact that (10a) has the ambiguity explicated in (b) and (c), since there are (at least) two possible attachment sites for the temporal adverbial. The simplex event structure attributed to (11), parallel to (9), provides only one attachment site for the adverbial, explaining the lack of ambiguity.

(10)

Argument Realization in Syntax

(11)

Argument Realization in Syntax

Most theories of argument realizations today admit of some kind of structured event representation and distinguish between this representation, common to large classes of predicates, and the idiosyncratic element of meaning, which distinguishes between verbs within classes defined in event structure terms. As we will see in section 4, accounts differ in the way in which the event structure relates to syntax and what aspect of these representations is considered purely lexical.

2.2. Conceptual Content of Grammatically Relevant Aspects of Meaning

The question of the conceptual content of grammatically relevant aspects of meaning is to a certain degree independent of the question of the nature of the representation. For example, most theorists assume that the notion of causer is relevant for argument realization, whether the notion is represented as a primitive (or feature-defined) role, or as the filler of a position of a predicate CAUSE. Furthermore, most theorists assume that agents should be distinguished from other kinds of causes. However, there is no universal agreement as to how to do this. Agents might be represented by a specific primitive predicate, such as DO, or by the combination of the content of a primitive predicate with inferences drawn from the nature of the argument that fill the open position of the predicate (as argued by Holisky, 1987; Van Valin & Wilkins, 1996)7. In this section, I lay out in broad terms the kinds of notions that are commonly taken to be associated with arguments realized as subject and as object, without delving into the question of how these notions are grammatically encoded.8

It is convenient to distinguish between eventive and non-eventive (stative) verbs in formulating generalizations about argument realization. For both classes of verbs we can distinguish between semantic notions that are associated with the subject position and those that are associated with object position.

For eventive predicates, the argument chosen as subject is generally a participant conceptualized as responsible in some way for bringing about the event. This category covers a range of notions including animate, inanimate, and abstract causes and instruments. Broadening the category, it appears that other entities that are not clearly causes, but whose properties and behavior are responsible for the eventuality coming into existence, are also associated with subject. Thus, glow and stink have an external argument, which can be conceived of as being ‘responsible’ by virtue of inherent properties of incandescence or smelliness; for spew, the subject argument is the source or cause of the spewing event by virtue of the fact that it has the requisite properties of kinetic energy; volitional agents have intentions and desires that lead them to initiate dynamic events; instrumental subjects are entities whose facilitating properties are presented as initiating the event because they allow it to happen. It is unclear whether all these notions should be given the same representation, and we have already seen two predicates, CAUSE and DO or ACT that can be associated with notions of this sort. However, it appears to be the case that these notions are always associated with the hierarchically highest position in event structure, and then, by some principle of prominence preservation, are associated with the syntactic subject position.

The semantic notion most commonly associated with direct objects is affectedness. However, this notion, like the notion of cause, is notoriously difficult to pin down. There are objects that do not appear to be affected in any way, though from a cross-linguistic perspective, these objects typically do not show the full range of direct object properties. Hopper and Thompson (1980) have proposed that objects can be associated with degrees of affectedness. Tenny (1987, 1994) was the first to propose that the notion of affectedness relevant for argument realization is related to the notion of affectedness (change) relevant to lexical aspect (aktionsart); Beavers (2008, 2010) extended these ideas to model the notion of degrees of affectedness. Tenny identified the direct object as being associated with a participant that ‘measures out the event.’ The notion of a measure is closely related to Dowty’s (1991) notion of an incremental theme. The measure or incremental theme is that argument whose properties can influence the aspectual properties of the predication in ways well-documented since Verkuyl (1972, 1993). The fact that the direct object is a DP with unspecified quantity in (12a) and of specified quantity in (12b), determines the atelicity of (a) and the telicity of (b), as illustrated by well-established telicity tests.

(12)

Argument Realization in Syntax

Arguments of change of state predicates also enter into determining the telicity of the predication, but in a different way. In (13a), it is not the (non)specified quantity of the direct object that determines the telicity, but rather specified degree of change.

(13)

Argument Realization in Syntax

To generalize the dependence between properties of different VP-internal constituents and the aspectual properties of the predication, the notion of scale or generalized path was developed (Beavers, 2008; Hay, Kennedy, & Levin, 1999; Krifka, 1998; Ramchand, 1997, 2008). The idea is that certain changes that can be characterized as changes in a scalar property of a predicate—change along a scale or path—form a natural class. The argument undergoing such a change is always expressed internally to the VP, and is usually expressed as the direct object. In (14a), the shoes are asserted to undergo a change in the scalar property of width, while in (14b), the suitcases are said to undergo a change in location along a path, where being situated at a point on a path is considered a scalar property and movement along path is a kind of scalar change.

(14)

Argument Realization in Syntax

Though arguments of verbs that do not specify a scalar change are often found as direct objects, as in (15), there is a certain amount of cross-linguistic variation in the expression of arguments of this sort, and even in English, verbs selecting these objects often appear with a range of unselected objects instead of these objects (Levin, 1999) as in (16).9 For example, in (16a), ‘the daylights’ is not the selected argument of ‘tickle,’ and in (16b), ‘the tiredness’ is not the argument of ‘rub.’

(15)

Argument Realization in Syntax

(16)

Argument Realization in Syntax

The identification of different semantic determinants of subject and object has ramifications for intransitive verbs. The well-known distinction between unaccusative verbs and unergative verbs (Burzio, 1986; Levin & Rappaport Hovav, 1995; Perlmutter, 1978, among many others) boils down to a distinction between intransitive verbs with a single argument with the semantic properties associated with transitive subjects and those with a single argument with the semantic properties associated with transitive objects. Syntactic approaches to unaccusativity analyze the latter as verbs that have an underlying object, which is a syntactically derived subject10. Other approaches to unaccusativity capture the shared properties between the subjects of unaccusative verbs and the objects of transitive verbs at the semantic level only (e.g., Van Valin, 1990; see Bresnan & Zaenen, 1990; Levin, 1987, for an LFG (Lexical-Functional Grammar) approach to unaccusativity).

The isolation of grammatically relevant components of meaning has not only relied on broad generalizations concerning the mapping to particular grammatical relations, such as subject and object, but has also taken into consideration phenomena like those illustrated in (3c,d) above. One can see already from these data, that some verbs have more than one way to express their arguments. Regularities in the way specified classes of verbs show more than one option for argument realization have been studied under the rubric of argument alternations. From the outset, it had been assumed that it is the grammatically relevant components of meaning—however they are represented—which determine participation in argument alternations (see in particular Fillmore, 1970; Guerssel, Hale, Laughren, Levin, & White Eagle, 1985; Hale & Keyser, 1988; Levin, 1993; Levin & Rappaport Hovav, 1995; Rappaport Hovav & Levin, 1988). For example, verbs associated with some linguistically relevant notion of possession are said to participate in the alternation illustrated in (3c,d) above. Thus, the alternation points to the notion of possession as being grammatically relevant. Further data regarding the varying syntactic contexts of verbs have challenged traditional approaches to the relation between the lexicon and syntax. These challenges are discussed in section 3. However, before we discuss these challenges, we turn to the final component in a theory of argument realization, namely the regularities in the mapping between articulated lexical representations and morpho-syntactic expression, what has fallen under the rubric of linking.

2.3. Theories of Linking

All work on linking, which assumes articulated lexical representations of the sort laid out in 2.2 and 2.3, is predicated on the assumption that the syntax preserves aspects of the lexical representation, thereby making the connection between lexical representations and syntax transparent to a certain extent. The transparency is reflected in different kinds of constraints meant to restrict the range of options available to languages for linking. See Levin and Rappaport Hovav (2005, chapters 5 and 6) for extensive discussion. I discuss here two types of constraints on mapping, each more naturally associated with one of the approaches to lexical semantic representation discussed above.

The first kind of overall constraint on mapping can be called ‘equivalence class preservation.’ The basic insight here is that each distinct semantic class of arguments is associated with a distinct morpho-syntactic expression. In its strongest form there will be a one-to-one relationship between argument type and syntactic expression. In reality, this seems very difficult to maintain given the diversity of argument realization options that do not fully align with semantic role distinctions. For example, the class of transitive verbs in English appears to be associated with direct objects with widely varying semantic roles. In (17a) the direct object seems most naturally associated with a theme role, whereas in (17b) it seems best associated with a location role.

(17)

Argument Realization in Syntax

A weaker constraint would require a many-to-one kind of mapping, and the Uniformity of Theta Assignment Hypothesis, in Baker (1988), is one implementation of such an idea.

(18)

Argument Realization in Syntax

More specifically, Baker (1997) assumes the following linking generalizations:

(19)

Argument Realization in Syntax

The many-to-one one relation between role and syntactic expression comes from the fact that Baker also assumes rather abstract underlying syntactic structures and rules of syntactic movement that allow various kinds of direct objects to originate in different syntactic positions. For example, the configuration in (20) underlies both (17a) and (17b), with object DPs originating in different positions. The object (17a) originates in the specifier position of the verb, indicated as ‘theme,’ and the object in (17b) originates in the complement position indicated as goal/path/location.’ In both cases, the verb moves from the lower VP into the V position of the higher VP, thus blurring the distinction of the underlying positions of the direct objects.

(20)

Argument Realization in Syntax

The second kind of constraint can be characterized in terms of prominence preservation. As mentioned in section 2.1, “Representing Grammatically Relevant Aspects of Meaning,” theories of predicate decomposition introduce hierarchical organization into lexical representations, often called event structures. Such theories usually assume that the mapping to syntax preserves prominence relations between arguments, such that given two arguments, the structurally higher argument in event structure will always map onto a structurally higher position in syntax. This helps lend explanatory force to the use of thematic hierarchies, which are often invoked in mapping algorithms (Belletti & Rizzi, 1988; Bresnan & Kanerva, 1989; Fillmore, 1968; Foley & Van Valin, 1984). Thematic hierarchies are composed of a set of ordered thematic roles and are coupled with mapping algorithms, which ensure that, in the mapping to syntax, the relative prominence of roles is preserved. A common criticism lodged at the use of thematic hierarchies is that it is not clear what determines the relative order of thematic roles. If thematic roles are understood as representing open positions in predicate decompositions, the ranking can be at least partially derived from depth of embedding in such representations (Kiparsky, unpublished manuscript, 1985; and see Belletti & Rizzi, 1988; Larson, 1988; Wunderlich, 1997). That being said, it is appears that the hierarchical structure inherent in these representations is not sufficient for completely determining the hierarchical syntactic relations among arguments and all details of argument realization, including the distribution of morphological case. For example, alongside causative verbs that embed a one-argument predicate as in (8), there are causative verbs with two-argument embedded predicates, as in (21).

(21)

Argument Realization in Syntax

If (21) is the appropriate representation for verbs of giving, as is usually assumed, then the fact that the possessor is expressed cross-linguistically in a position more prominent than the possessum is not derivable from the hierarchical structure of the representation in (21), since they are co-arguments of HAVE. Other factors having to do with the lexical entailments associated with the arguments need to be taken into consideration. See Rappaport Hovav and Levin (2007) and Beavers (2010) for further discussion.

3. Challenges

If most verbs had only one way in which to express their arguments, and argument alternations were clearly delimited by identifiable grammatically relevant aspects of meaning, the theory of argument realization along the lines sketched out above would be sufficient. However, this picture is complicated by a number of well-documented phenomena that show that the relation of a verb with its syntactic environment is much more variable. Accounting for the variability with formal apparatus of the type discussed in section 2 poses a serious challenge. In this section, I discuss first the problem of multiple argument realization options and then the problem of non-selected arguments. In section 4, I discuss how these challenges have been met in recent theories and how these phenomena have brought about a change in the conception of what should be considered lexical and, as a consequence, the division of labor between the lexicon and syntax.

3.1 Multiple Argument Expression

As already alluded to, many verbs have alternate ways of expressing their arguments and participate in widely recognized alternations, such as the passive (22), causative (23), dative (24), (25) locative, and conative (26) alternations, to illustrate just a few.12

(22)

Argument Realization in Syntax

(23)

Argument Realization in Syntax

(24)

Argument Realization in Syntax

(25)

Argument Realization in Syntax

(26)

Argument Realization in Syntax

If each verb is associated with a lexical representation that, along with principles of linking, determines its argument realization pattern, why should some verbs allow more than one pattern of argument realization? The relation between an active verb and its passive counterpart can be distinguished from the others in two respects. First, aside from marginal cases, the appearance of a verb in both frames is predictable from its transitive syntax.13 Second, broadly speaking, there is no truth-conditional difference between pairs of sentences that differ only in the choice of active or passive verb.14 These considerations lead to analyses in which there is a derivational relation between the active and passive forms of a verb.

For the other alternations, the syntax of neither variant determines the participation of a verb in the alternation and the different variants are often associated with sometimes subtle differences of interpretation. For example (25b) has an inference lacking from (25a) (that the amount of boxes on the truck constitutes a load). Accounting for these alternations by positing a derivational relation between the variants is then not appropriate. To account for these alternations one might suggest that alternating verbs are associated with more than one meaning, with each meaning giving rise to a distinct pattern of argument realization by independently established principles. This move allows to account for the alternations without tailoring argument realization principles to particular sets of verbs. This solution was first proposed by Rappaport and Levin (1988) for the locative alternation, and the general approach was applied more widely to argument alternations in English by Pinker (1989) and for specific alternations in Gropen, Pinker, Hollander, Goldberg, and Wilson(1989); Gropen, Pinker, Hollander, and Goldberg(1991); and Harley (2002), among many others. As mentioned in section 2.1, it was observed from the outset (Fillmore, 1970) that semantically coherent classes of verbs participate in these alternations (extensively documented for English in Levin, 1993). The challenge then is to accurately characterize which classes of verbs participate in which alternations, and then come up with an explanation of why certain verbs can be associated with more than one lexical entry and others cannot.

A drawback of this proposal is that it involves a proliferation of polysemy—many verbs are associated with more than one lexical entry. To the extent that there is regularity in these alternations, the regularity of the underlying polysemy needs to receive an explicit account. One approach to this challenge is to suggest that there are lexical rules that derive the different meanings of a verb (Aronoff, 1976; Bresnan, 1982; Jackendoff, 1975; Wasow, 1977). A theory of lexical rules does not explain why particular verbs are associated with more than one meaning, or why particular verbs are associated with a particular set of meanings, though it does give expression to the regularity of the polysemy, and then independently established linking generalizations explain the argument realization patterns.

However, the range of argument realization patterns cannot be characterized by a closed set of lexical rules making reference only to the structural aspects of meaning—those encoded in the basic predicates of event structures. Nor can one fully chart semantically coherent classes and their participation in alternations. Many verbs show a bewildering range of argument realization options, as illustrated for the verb whistle:

(27)

Argument Realization in Syntax

It is not obviously possible to derive each of the uses above via a well-established argument alternation that governs a semantically coherent class of verbs. Whistle can be classified as a verb of sound emission, but not all sound emission verbs show the same range of argument realization possibilities. Moreover, it is not clear that the notion of sound emission should be considered a grammatically relevant meaning component: it does not correspond to any primitive predicate or event structure configuration. It may of course be the case that some other way of characterizing the semantics of whistle will account for the exact range of argument realization options for this verb, but none has been forthcoming. Furthermore, it is fairly clear that the number of distinctions displayed by the wide range of verb classes cannot be captured by the information encoded in event structures. Data of this sort casts doubt on the lexical rule approach. The next section discusses another phenomenon that further challenges the lexical rule approach.

3.2. Non-Selected Arguments

Another problem with the lexical rule approach is that some alternations do not seem to involve anything that can plausibly be considered alternate lexical meanings of a verb (Goldberg, 1995). This is most clearly seen in examples like (28).

(28)

Argument Realization in Syntax

The constituents in the immediate syntactic environment of drink in (28) do not appear to be realizations of the arguments of drink; one does not drink a teapot and the result state of dryness is not necessarily associated with drink. While teapot as the object of drink may be thought to be licensed by a metonymy relation, as in the case of tying one’s shoes or a kettle boiling, (28b), where the teapot is licensed only in the presence of the Adjective Phrase (AP) dry (and vice-versa). argues against this.

An event structure representation that is purported to underlie a sentence such as (28a) is given in (29). However, (29) does not intuitively constitute a natural lexical entry. There is a sense in which the post-verbal DP in (28a) is not an argument of the verb; as (28b) illustrates, its appearance is licensed by the result AP. Furthermore, the causation in (28) does not seem to qualify as direct causation, a well-known constraint on lexical causatives.

(29)

Argument Realization in Syntax

It turns out that ‘non-selected’ arguments are not rare in English as the following examples illustrate. See Boas (2003); Goldberg (1995); Goldberg and Jackendoff (2004); Rappaport Hovav and Levin (2001) for extensive illustration of one kind of non-selected argument.

(30)

Argument Realization in Syntax

Non-selected arguments appear in examples in (31) as well.

(31)

Argument Realization in Syntax

Neither kick nor bake involve more than two arguments, and the inner object in each case appears to be a kind of non-selected object.

Rappaport Hovav and Levin (1998) assumed that the basic ontological type of a predicate determines its ‘basic’ association with event structure, but also recognized a process that they called Template Augmentation:

(32)

Argument Realization in Syntax

Nonetheless, it is not clear where exactly in the derivation Template Augmentation takes place. If its output does not appear to be a natural lexical entry, then perhaps Template Augmentation should not be considered a lexical operation? But if alternations involving non-selected arguments are not accounted for by lexical rules, perhaps no argument alternations should be accounted for by lexical rules. The two problems sketched in the last two sections have led to a revision in the way that theorists have approached the question of the relation between the lexicon and syntax. In particular, they have led to a fresh look at what is considered lexical and what is considered non-lexical. I review these approaches in the next section.

4. Where Do Arguments Come From? Constructionist and Neo-Constructionist Approaches to Argument Realization

As mentioned, all theories that account for restrictions on the relation between words and their syntactic contexts distinguish between what we have called structural and idiosyncratic aspects of meaning, the former in terms of an event structure and the latter commonly called a root. Traditional theories and most theories in generative grammar through the late 1980s attributed both aspects of meaning to words, in the cases under consideration, verbs. The theories of the 1980s have been termed ‘projectionist’ in the sense that they took syntactic structure to be a projection of lexical properties (Bresnan, 1982; Chomsky, 1981; Stowell, 1981). This was given expression in principles like the Projection Principle, which ensured that all syntactic structure in a given sentence is lexically licensed. In the late 1960s and early 1970s, Generative Semantics adopted a different approach. Since lexical decompositions are similar in many ways to syntactic representations including categories and hierarchical structure, they took the lexical decompositions—which represent the structural aspect of meaning—to be syntactic representations, and only the morphemes that represent the idiosyncratic aspects of meaning to be what is strictly lexically associated with the verb. Labeled brackets can, of course, be directly translated into syntactic tree representations, and generative semanticists posited structures similar to those in example (8) above, taking them to be syntactic representations, where the heads corresponding to the lexical content of words underwent a process of ‘predicate raising,’ ultimately merging the primitive predicates with these heads to produce the actual verbs we are familiar with (e.g., McCawley, 1968). (33), for example, is the representation for a sentence with the transitive variant of the verb cool. The abstract element COOL raises and merges with BECOME, which together raise and merge with CAUSE. This complex element is finally realized as the transitive verbs cool.

(33)

Argument Realization in Syntax

Since morphologically unrelated words may share both structural and idiosyncratic aspects of meaning, the morphological form of the word was taken to be determined by the syntactic context it was inserted into. In the case of the triad in (34), the idiosyncratic element of meaning can be taken to be that expressed by the word dead, and the structural aspects of meaning are the same as those in (8) above. Therefore, it is a lexical property of the abstract element DEAD that, in the syntactic context in (34c), it is pronounced ‘kill.’ The aspects of meaning expressing cause and change are abstract lexical items in their own right that can compose with many other predicates and are therefore not lexically associated with the predicate DEAD. The same goes for any other verb expressing a change of state. On this conception, then, the structure in (33), which expresses roughly the same information encoded in event structures, is not strictly projected from the transitive verb cool.

(34)

Argument Realization in Syntax

There were a variety of problems with the theory developed by Generative Semanticists, having partly to do with a lack of a deeper understanding of the nature of the interface between the lexicon and syntax. For example, the structure underlying (34c) was taken to be nearly identical to The rat caused the mouse to die, but, as numerous authors showed, the two structures have subtly different interpretations and syntactic properties (Fodor, 1970). However, the idea that aspects of meaning traditionally associated with verbs should be factored out and attributed to other elements has returned to current theorizing.

In this section, I discuss current frameworks that can be characterized broadly as “constructionist,” as opposed to the “projectionist” frameworks of the 1980s. They all differ from projectionist theories in the way they view the division between what is lexical and what is syntactic. They all share the assumption that only the idiosyncratic aspects of meaning (and their phonological expression) are lexically associated with the verbs they head. In contrast, what has been characterized as the structural aspects of meaning are syntactically encoded. We saw in section 3 that syntactic structure is taken to preserve aspects of event structure. Theories that have been called Neo-Constructionist (as opposed to Constructionist theories, to be described next) take event structure to be directly encoded in syntax, thereby making superfluous constraints that ensure that syntax reflects or preserves information in event structure. Another reason for assuming the relation between the root and event structure to be syntactic is that the interpretive integration of the root with event structure is compositional and productive, similar to the compositionality and productivity of syntax.15

The syntactic encoding of event structure dovetails with another basic assumption of Distributed Morphology (Halle & Marantz, 1993) and theories of Bare Phrase Structure in the context of the Minimalist Program (Chomsky, 1995; see also Borer, 2005; Ramchand, 2008), namely, that syntax is the sole generative engine of the grammar, in the sense of the component that builds structures from morphemes. Phonology and semantics interpret syntax. A consequence of this assumption is that there can be no rules of Template Augmentation, as in (32) above, in the lexicon, since this would be duplicating what the syntax could, and by assumption, should be performing. The appearance of certain roots in the triads discussed above is predictable from the meaning associated with the roots, which makes them compatible with three kinds of syntactic structure. The root may appear in a structure that encodes a simple predication with a state predicate—the lowest branching structure in (35); it may be further integrated into a structure that includes a “verbalizer” (little v), which brings in the notion of change of state; to this structure may be added an additional functional head (Voice), which introduces the external argument, understood as the causer of the change of state (Alexiadou, Anagnastopoulou, & Schäfer, 2015; Marantz, 2012, among others).16 The similarity between the structure in (35) and the one in (33) should be apparent.

(35)

Argument Realization in Syntax

Syntactic structure can still be seen as a projection of lexical structure, but functional heads representing notions such as causation and agency project independently from the lexical items they ultimately merge with. That is, the structure in (35) does not project from a lexical item corresponding to the transitive variant of open, but rather from the root open and the functional elements that appear in the structure. This approach does not meet with the problem of the licensing of non-selected ‘arguments’ that appear in the ‘core’ syntactic layers of the clause. If what has traditionally been taken to be core arguments of a verb are now considered arguments of functional heads, this approach can be easily extended to non-selected arguments in the more traditional sense. For example, functional heads, usually called applicative heads, can select for the variety of dative and benefactive arguments, even if these heads are not phonologically realized in a particular language, such as English. We find such heads morphologically realized in other languages, particularly Bantu languages (Pylkkänen, 2008). Example (36), would be the structure associated with Jane baked Bill a cake.

(36)

Argument Realization in Syntax

Resultative phrases can also be introduced by phonologically uninstantiated heads, and these phrases then select the arguments not selected by the head verb in constructions such as those illustrated in (30) and (31).

This constructionist approach blurs the distinction between morphology and syntax—they are both just syntax. As a result, there are no lexical entries for a derived word like break, since it is syntactically constructed. There are lexical entries for the component constituents—the root break, and the abstract functional elements.

There is a range of such Neo-Constructionist approaches that agree on many aspects of what should be considered lexical and what should be considered syntactic. Space limitations prevent us from comparing the theories mentioned with those of Borer (2005) and Ramchand (2008). The major differences between these approaches have to do with the richness of information taken to be associated with the minimal lexical unit, the root, and the way in which the restrictions on the distribution of different kinds of roots in different syntactic contexts are derived.

In contrast to this Neo-Constructionist approach, a Constructionist approach has been developing since the 1990s (Fillmore, Kay, & O’Connor, 1988; Goldberg, 1995; Kay & Fillmore, 1999). Goldberg (1995) and much subsequent work of hers is devoted to argument realization in the Constructionist framework.

Constructionist approaches also challenge the idea that there is a strict division between syntax and the lexicon. But instead of stressing the ‘generative’ and compositional nature of the lexicon, Constructionists stress the idea that constructions, like lexical items, should be viewed as independently stored units. It may be appropriate to say that the main difference between Constructionist and Neo-Constructionist approaches is that the latter places most emphasis on the compositional aspects of linguistic constructions, while the former places most emphasis on the non-compositional. Neo-constructionist approaches do not recognize constructions as such. Linguistic units of all sizes are constructed from lexical and functional heads that compose according to the principles of syntax, and the phonology and semantics are read off of these structures. While linguistic units of any size can be associated with non-compositional meaning, the compositional meaning cannot be suppressed. A VP such as ‘kick the bucket’ may have a special meaning, but there is no way to suppress its regular meaning. Much theorizing in Neo-constructionist approaches concerns the ways in which special meanings can be associated with the lexical heads that make up structures of different sizes, but the special meanings are always contextually specified for the components making up any structure. The Constructionist approach takes constructions to be Saussurean signs, in the sense that they are form-meaning pairs for which not all properties are predictable from the component parts. In this way, constructions are likened to words. Constructionist theories posit constructions such as a range of resultative constructions and double object constructions, each of which has properties that need to be specified. It employs inheritance hierarchies to capture generalizations across constructions, making it possible to state relevant generalizations once.

5. Summary and Prospectus

It is common to characterize languages as being composed of words and rules (Pinker, 2000), or lexicon and syntax. It turns out that the demarcation in grammar between lexicon and syntax is much more challenging than is ordinarily thought. We stated at the outset that a theory of argument realization must isolate the aspects of the meanings of verbs which determine (or correlate with) argument realization, it must determine the nature of lexical representation, and it must determine the nature of mapping between lexical representation and syntactic representation. As we have seen, there appears to be a broad consensus across all theories as to the elements of meaning which figure into argument realization, but the conception of what is to be taken as strictly lexical has undergone radical changes over the years. It should be stressed that a full account of argument realization will not apportion all properties of argument realization phenomena to the syntax and the lexicon. Argument realization properties in specific cases can be attributed to a wide range of factors which are just beginning to be teased apart. These include, in addition to properties attributed to lexical head and functional heads and their compositional properties, factors such as information structure and discourse principles, as demonstrated, for the causative alternation in Rappaport Hovav (2014) and the dative alternation in Arnold et al. (2000), Bresnan et al (2007) and Rappaport Hovav and Levin (2008).

Finally, it should be pointed out that there is a rich literature on cross-linguistic variation in argument realization, a topic not addressed here because of space limitations. Much of this literature focuses on what has come to be called ‘lexicalization patterns,’—patterns of how conceptual components of event descriptions are distributed across morpho-syntactic elements—beginning with Talmy (1985); see also the work of Slobin (1996, 1997); and analyses in a generative framework of these data in Acedo-Matellán and Mateu (2013) This work has uncovered interesting cross-linguistic patterns and has developed a typology of languages that display different lexicalization patterns. In addition, there is a rich literature on cross-linguistic patterns of valency, as documented, for example in Mal’chukov and Comrie (2015).

In all areas of research having to do with the relation of words and their morpho-syntactic environments, the solutions to certain problems bring into focus new ones and the area of argument realization remains a fertile for research.

Acknowledgments

This article was written when I was a fellow at the Mandel Scholion Interdisciplinary Research Center in the Humanities and Jewish Studies. I am thankful to an anonymous reviewer for very helpful input.

Further Reading

Beavers, J. (2010). The structure of lexical meaning: Why semantics really matters. Language, 86(4), 821–864.Find this resource:

Borer, H. (2005). Structuring sense: Vol. 20. The normal course of events. New York: Oxford University Press.Find this resource:

Croft, W. (1998). Event structure in argument linking. In M. Butt & W. Geuder (Eds.), The projection of arguments: Lexical and compositional factors (pp. 21–63). Stanford, CA: CSLI.Find this resource:

Davis, A. (2001). Linking by types in the hierarchical lexicon. Stanford, CA: CSLI.Find this resource:

Dowty, D. (1991). Thematic proto-roles and argument selection. Language, 67, 547–619.Find this resource:

Goldberg, A. (1995). Constructions: A construction grammar approach to argument structure. Chicago: University of Chicago Press.Find this resource:

Grimshaw, J. (1990). Argument structure. Cambridge, MA: MIT Press.Find this resource:

Hale, K., & Keyser, S. J. (1993). On argument structure and the lexical representation of syntactic relations. In K. Hale and S. J. Keyser (Eds.), The view from building 20 (pp. 111–176). Cambridge, MA: MIT Press.Find this resource:

Hale, K., & Keyser, S. J. (2002). Prolegomenon to a theory of argument structure. Cambridge, MA: MIT Press.Find this resource:

Jackendoff, R. (1990). Semantic structures. Cambridge, MA: MIT Press.Find this resource:

Levin, B. & Rappaport Hovav, M. (1995). Unaccusativity: At the syntax-lexical semantics interface. Cambridge, MA: MIT Press.Find this resource:

Levin, B., & Rappaport Hovav, M. (2005). Argument realization. Cambridge, U.K.: Cambridge University Press.Find this resource:

Marantz, Alec. (2012). Verbal argument structure: Events and participants. Lingua, 130, 152–168.Find this resource:

Pylkkänen, L. (2008). Introducing Arguments. Cambridge, MA: MIT Press.Find this resource:

Pinker, S. (1989). Learnability and cognition. Cambridge, MA: MIT Press.Find this resource:

Ramchand, G. (2008). Verb meaning and the lexicon: A first phase syntax. Cambridge, U.K.: Cambridge University Press.Find this resource:

Rappaport Hovav, M., & Levin, B. (2010). Reflections on manner result complementarity. In M. Rappaport Hovav, E. Doron, & I. Sichel (Eds.), Syntax, Lexical Semantics and Event Structure (pp. 21–38). Oxford: Oxford University Press.Find this resource:

Rappaport Hovav, M., & Levin, B. (2001). An Event Structure Account of English Resultatives. Language, 77, 766–797.Find this resource:

Kiparsky, P. (1985). Morphology and Grammatical Relations, Unpublished Manuscript. Stanford, CA: Stanford University.Find this resource:

Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.Find this resource:

References

Acedo-Matellán, Víctor & Mateu, Jaume. (2013). “Satellite-framed Latin vs. verb-framed Romance: A syntactic approach.” Probus. International Journal of Latin and Romance Linguistics, 25, 227–265.Find this resource:

Ackerman, F., & Moore, J. (1999). Syntagmatic and paradigmatic dimensions of cause encodings. Linguistics and Philosophy, 22, 1–44.Find this resource:

Ackerman, F., & Moore, J. (2001). Proto-properties and grammatical encoding: A correspondence theory of argument selection. Stanford, CA: CSLI Publications.Find this resource:

Alexiadou, A., Anagnostopoulou, E. & Schäfer F. (2015). External arguments in transitivity alternations: A layering approach. New York: Oxford University Press.Find this resource:

Alexiadou, A., & Doron, E. (2012). The syntactic construction of two non-active voices: passive and middle. Journal of Linguistics, 48(1), 1–34.Find this resource:

Alexiadou, A., & Schäfer, F. (2011). An unaccusativity diagnostic at the syntax-semantics interface: There-insertion, indefinites and restitutive again. In I. Reich & G. für Semantik (Eds.), Proceedings of Sinn & Bedeutung 15 (pp. 101–115). Saarbrücken, Germany: Saarland University Press.Find this resource:

Anderson, J. M. (1971). The grammar of case. Cambridge, U.K.: Cambridge University Press.Find this resource:

Anderson, J. M. (1977). On case grammar. Atlantic Highlands, NJ: Humanities Press.Find this resource:

Arnold, J. E., Wasow, T., Losongco, A., & Ginstrom, R. (2000). Heaviness vs. newness: The effects of structural complexity and discourse status on constituent ordering. Language, 76, 28–55.Find this resource:

Aronoff, M. (1976). Word formation in generative grammar. Cambridge, MA: MIT Press.Find this resource:

Baker, M. (1988). Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press.Find this resource:

Baker, M. (1997). Thematic roles and syntactic structure. In L. Haegeman (Ed.), Elements of Grammar (pp. 73–137). Dordrecht: Kluwer.Find this resource:

Belletti, A., & L. Rizzi. (1988). Psych-verbs and –Theory. Natural Language and Linguistic Theory, 6, 291–352.Find this resource:

Beavers, J. (2008). Scalar complexity and the structure of events. In J. Dölling, T. Heyde-Zybatow, & M. Schäfer (Eds.), Event structures in linguistic form and interpretation (pp. 245–265). Berlin: Mouton de Gruyter.Find this resource:

Beavers, J. (2010). The structure of lexical meaning: Why semantics really matters. Language, 86(4), 821–864.Find this resource:

Beavers, J. (2011). On affectedness. Natural Language and Linguistic Theory, 29(2), 335–370.Find this resource:

Boas, H. (2003). A constructional approach to resultatives. Stanford, CA: CSLI.Find this resource:

Borer, H. (2005). Structuring sense: Vol 2. The normal course of events. New York: Oxford University Press.Find this resource:

Bresnan, J. (1982). The mental representation of grammatical relations. Cambridge, MA: MIT Press.Find this resource:

Bresnan, J., Cueni, A., Nikitina, T., & Baayen, H. (2007). Predicting the dative alternation. In G. Boume, I. Kraemer, & J. Zwarts (Eds.), Cognitive foundations of interpretation (pp. 69–94). Amsterdam: Royal Netherlands Academy of Science.Find this resource:

Bresnan, J., & Kanerva, J. (1989). Locative inversion in Chichewa: A case study of factorization in grammar. Linguistic Inquiry, 20(1), 1–50.Find this resource:

Bresnan, J., & Zaenen, A. (1990). Deep unaccusativity in LFG. In K. Dziwirek, P. Farrell, & E. Mejias-Bikandi (Eds.), Grammatical relations: A cross-theoretical perspective (pp. 45–57). Stanford, CA: Stanford University Center for the Study of Language and Information.Find this resource:

Burzio, L. (1986). Italian syntax: A government-binding approach. Dordrecht, Netherlands: Reidel.Find this resource:

Butt, M. (2006). Theories of Case. Cambridge, U.K.: Cambridge University Press.Find this resource:

Carter, R. (1988). Towards a linking grammar of English. In B. Levin & C. Tenny (Eds.), On Linking: Papers by Richard Carter (pp. 93–108). Cambridge, MA: MIT Press.Find this resource:

Chomsky, N. (1957). Syntactic structures. The Hague, Netherlands: Mouton.Find this resource:

Chomsky, N. (1970). Remarks on Nominalization. In R. Jacobs and P. Rosenbaum (Eds.), Readings in English Transformational Grammar (pp. 184–221). Waltham MA: Ginn and Co.Find this resource:

Chomsky, N. (1981). Lectures on government and binding: The Pisa lectures. Dordrecht, Netherlands: Foris.Find this resource:

Chomsky, N. (1995). The minimalist program. Cambridge, MA: MIT Press.Find this resource:

Croft, W. (1991). Syntactic categories and grammatical relations: The cognitive organization of information. Chicago: University of Chicago Press.Find this resource:

Croft, W. (1998). Event structure in argument linking. In M. Butt & W. Geuder (Eds.), The projection of arguments: Lexical and compositional factors (pp. 21–63). Stanford, CA: CSLI.Find this resource:

Croft, W. (2012). Verbs: Aspect and causal structure. Oxford: Oxford University Press.Find this resource:

Davis, A. (2001). Linking by types in the hierarchical lexicon. Stanford, CA: CSLI Publications.Find this resource:

Dowty, D. (1979). Word meaning and Montague grammar. Dordrecht, Netherlands: Reidel.Find this resource:

Dowty, D. (1989). On the semantic content of the notion “thematic role.” In G. Chierchia, B. H. Partee, & R. Turner (Eds.), Properties, types, and meaning (pp. 69–130). Dordrecht, Netherlands: Kluwer.Find this resource:

Dowty, D. (1991). Thematic proto-roles and argument selection. Language, 67(3), 547–619.Find this resource:

Fillmore, C. (1968). The case for case. In E. Bach & R. T. Harms (Eds.), Universals in linguistic theory (pp. 1–90). New York: Holt.Find this resource:

Fillmore, C. (1970). The grammar of hitting and breaking. In R. Jacobs & P. S. Rosenbaum, (Eds.), Readings in English transformational grammar (pp. 120–133). Waltham, MA: Ginn.Find this resource:

Fillmore, C. (1971). Types of lexical information. In D. Steinberg & L. Jakobovits (Eds.), Semantics (pp. 370–392). Cambridge, U.K.: Cambridge University Press.Find this resource:

Fillmore, C. (1977). The case for case reopened. In P. Cole & J. M. Sadock (Eds.), Grammatical relations (pp. 59–82). New York: Academic Press.Find this resource:

Fillmore, C., Kay, P. & O’Connor, M. C. (1988). Regularity and idiomaticity in grammatical constructions: the case of “let alone.” Language, 64, 501–538.Find this resource:

Fodor, J. A. (1970). Three reasons for not deriving “kill” from “cause to die.” Linguistic Inquiry, 1(4), 429–438.Find this resource:

Foley, W. A. & Van Valin, R. D., Jr. (1984). Functional syntax and universal grammar. Cambridge, U.K.: Cambridge University PressFind this resource:

Goldberg, A. (1995). Constructions: A construction grammar approach to argument structure. Chicago: University of Chicago Press.Find this resource:

Goldberg, A. & Jackendoff, R. (2004). The English resultative as a family of constructions. Language, 80(3), 532–568.Find this resource:

Grimshaw, J. (1990). Argument structure. Cambridge, MA: MIT Press.Find this resource:

Grimshaw, J. (2005). Word and structure. Stanford, CA: CSLI Publications.Find this resource:

Gropen, J., Pinker, S., Hollander, M., & Goldberg, R. (1991). Affectedness and direct objects: The role of lexical semantics in the acquisition of verb argument structure. Cognition, 41, 153–195.Find this resource:

Gropen, J., Pinker, S., Hollander, M., Goldberg, R., & Wilson, R. (1989). The learnability and acquisition of the dative alternation in English. Language, 65(2), 203–257.Find this resource:

Guerssel, M., Hale, K., Laughren, M., Levin, B., & White Eagle, J. (1985). A cross-linguistic study of transitivity alternations. In W. H. Eilfort, P. D. Kroeber, & K. L. Peterson (Eds.), Papers from the Parasession on Causatives and Agentivity (pp. 48–63). Chicago: Chicago Linguistic Society.Find this resource:

Gropen, J., Pinker, S., Hollander, M., Goldberg, R., & Wilson, R. (1989). The learnability and acquisition of the dative alternation in English. Language, 65(2), 203–257.Find this resource:

Hale, K., & Keyser, S. J. (1988). Explaining and constraining the English middle. In C. Tenny, Lexicon Project Working Papers, 24. Center for Cognitive Science, MIT.Find this resource:

Hale, K., & Keyser, S. J. (1993). On argument structure and the lexical representation of syntactic relations. In K. Hale & S. J. Keyser (Eds.), The view from building 20 (pp. 50–110). Cambridge, MA: MIT Press.Find this resource:

Hale, K., & Keyser, S. J. (2002). Prolegomenon to a theory of argument structure. Cambridge, MA: MIT Press.Find this resource:

Halle, M., & Marantz, A. (1993). Distributed morphology and the pieces of inflection. In K. Hale & S. J. Keyser (Eds.), The view from building 20 (pp. 111–176). Cambridge, MA: MIT Press.Find this resource:

Harley, H. (2002). “Possession and the double object construction.” Yearbook of Linguistic Variation, 2, pp. 29–68.Find this resource:

Harley, H. (2014). “On the identity of roots.” Theoretical Linguistics, 40(3–4), 225–276.Find this resource:

Hay, J., Kennedy, C., & Levin, B. (1999). Scalar structure underlies telicity in degree achievements. Proceedings of Semantics and Linguistic Theory (SALT), 9, 127–144.Find this resource:

Holisky, D. (1987). The case of the intransitive subject in Tsova-Tush (Bats). Lingua, 71(1–4), 103–132.Find this resource:

Hopper, P., & Thompson S. (1980). Transitivity in grammar and discourse. Language, 56(2), 251–299.Find this resource:

Jackendoff, R. (1972). Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press.Find this resource:

Jackendoff, R. (1975). Morphological and semantic regularities in the lexicon, Language, 51(3), 639–671.Find this resource:

Jackendoff, R. (1976). Toward an explanatory semantic representation. Linguistic Inquiry, 7(1), 89–150.Find this resource:

Jackendoff, T. (1983). Semantics and Cognition. Cambridge, MA: MIT Press.Find this resource:

Jackendoff, R. (1987). The status of thematic relations in linguistic theory. Linguistic Inquiry, 18(3), 369–412.Find this resource:

Jackendoff, R. (1990). Semantic structures. Cambridge, MA: MIT Press.Find this resource:

Kay, P. & Fillmore, C. (1999). Grammatical constructions and linguistic generalizations: The “What’s X doing Y?” construction. Language, 75, 1–33Find this resource:

Kisala, J. (1985). Review of chapter 1 of Ostler’s Thesis, Case-linking: A theory of case and verb diathesis applied to classical Sanskrit,” In B. Levin (Ed.), Lexical semantics in review. Lexicon Project Working Papers 1. Cambridge, MA: Center for Cognitive Science, MIT.Find this resource:

Krifka, M. (1998). The origins of telicity. In S. Rothstein (Ed.), Events and grammar (pp. 197–235). Dordrecht, Netherlands: Kluwer.Find this resource:

Lakoff, G. (1968). Some verbs of change and causation. In S. Kuno (Ed.), Mathematical linguistics and automatic translation. Report NSF-20, Aiken Computation Laboratory. Cambridge, MA: Harvard University.Find this resource:

Lakoff, G. (1970). Irregularity in syntax. New York: Holt, Rinehart, and Winston.Find this resource:

Landau, I. (2009). Saturation and reification in adjectival diathesis. Journal of Linguistics, 35(2), 315–361.Find this resource:

Larson, R. K. (1988). On the double object construction. Linguistic Inquiry, 19, 335–391.Find this resource:

Levin, B. (1993). English verb classes and alternations: A preliminary investigation. Chicago: University of Chicago Press.Find this resource:

Levin, B. (1999). Objecthood: An event structure perspective. CLS 35, Part 1: Papers from the Main Session (pp. 223–247). Chicago: Chicago Linguistic Society.Find this resource:

Levin, B., & Rappaport Hovav, M. (1995). Unaccusativity: At the syntax-lexical semantics interface. Cambridge, MA: MIT Press.Find this resource:

Levin, B., & Rappaport Hovav, M. (2005). Argument realization. Cambridge, U.K.: Cambridge University Press.Find this resource:

Levin, L. (1987). Towards a linking theory of relation changing rules in LFG. Report CSLI (pp. 87–115). Stanford, CA: Center for the Study of Language and Information, Stanford University.Find this resource:

Mal’chukov, A., & Comrie, B. (Eds.). (2015). Valency alternations in the world’s languages (2 vols.). Berlin, Germany: De Gruyter Mouton.Find this resource:

Marantz, A. (1997). No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. University of Pennsylvania Working Papers in Linguistics, 4(2), 201–225.Find this resource:

Marantz, A. (2012). Verbal argument structure: Events and participants. Lingua, 130, 152–168.Find this resource:

McCawley, J. (1968). Lexical insertion in a transformational grammar without deep structure (pp. 71–80). Chicago: Chicago Linguistic Society.Find this resource:

McCawley, J. (1971). Prelexical syntax. Report of the 22nd Annual Roundtable Meeting on Linguistics and Language Studies (pp. 19–33). Washington, DC: Georgetown University Press.Find this resource:

Ostler, N. (1979). Case-linking: A theory of case and verb diathesis applied to classical Sanskrit. Doctoral thesis, MIT, Cambridge, MA.Find this resource:

Perlmutter, D. (1978). Impersonal passives and the unaccusative hypothesis. Proceedings of the 4th Annual Meeting of the Berkeley Linguistics Society, 4 (pp. 157–189). Berkeley, CA: Berkeley Linguistics Society.Find this resource:

Pesetsky, D. (1995). Zero syntax: Experiencers and cascades. Cambridge, MA: MIT Press.Find this resource:

Pylkkänen, L. (2008). Introducing Arguments. Cambridge, MA: MIT Press.Find this resource:

Pinker, S. (1989). Learnability and cognition. Cambridge, MA: MIT Press.Find this resource:

Pinker, S. (2000). Words and rules: The ingredients of language. London: Phoenix.Find this resource:

Primus, B. (1999). Cases and thematic roles: Ergative, accusative and active. Tübingen, Germany: Niemeyer.Find this resource:

Ramchand, G. (1997). Aspect and predication. Oxford: Clarendon Press.Find this resource:

Ramchand, G. (2008). Verb meaning and the lexicon: A first phase syntax. Cambridge, U.K.: Cambridge University Press.Find this resource:

Rappaport Hovav, M. (2008). Lexicalized meaning and the internal structure of events. In S. Rothstein (Ed.), Theoretical and crosslinguistic approaches to the semantics of aspect (pp. 13–42). Amsterdam: John Benjamins.Find this resource:

Rappaport Hovav, M. (2014). Lexical content and context: The causative alternation in English revisited. Lingua, 141, 8–29.Find this resource:

Rappaport Hovav, M., & Levin, B. (1988). What to do with θ-roles. In W. Wilkins, (Ed.), Syntax and Semantics 21: Thematic Relations (pp. 7–36). New York: Academic Press.Find this resource:

Rappaport Hovav, M., & Levin, B. (1998). Building verb meanings. In M. Butt & W. Geuder, The projection of arguments: Lexical and compositional factors (pp. 97–133). Stanford, CA: CSLI.Find this resource:

Rappaport Hovav, M., & Levin, B. (2007). Deconstructing thematic hierarchies. In J. Grimshaw, J. Maling, C. Manning, J. Simpson, & A. Zaenen (Eds.), Architectures, rules, and preferences: A festschrift for Joan Bresnan. Stanford, CA: CSLI.Find this resource:

Rappaport Hovav, M., & Levin, B. (2008). The English dative alternation: The case for verb sensitivity. Journal of Linguistics, 44, 129–167.Find this resource:

Reinhart, T. (2002). “The Theta System: An overview.”. Theoretical Linguistics, 28, pp. 229–290.Find this resource:

Ross, J. R. (1972). Act. In D. Davidson and G. Harman (Eds.), Semantics of Natural Languages (pp. 70–126). Dordrecht: D. Reidel and Company.Find this resource:

Rozwadowska, B. (1988). Thematic restrictions on derived nominals. In W. Wilkins (Ed.), Syntax and Semantics 21: Thematic Relations (pp. 147–165). San Diego: Academic Press.Find this resource:

Rozwadowska, B. (1989). Are thematic relations discrete? In R. Corrigan, F. Eckman, & M. Noonan (Eds.), Linguistic categorization (pp. 115–130). Philadelphia, PA: John Benjamins.Find this resource:

Slobin, D. (1996). “Two ways to travel: verbs of motion in English and Spanish,” In M. Shibatani & S. A. Thompson (Eds.), Grammatical Constructions (pp. 195–219). Oxford: Clarendon.Find this resource:

Slobin, D. (1997). “Mind, code, and text,” In J. Bybee, J. Haiman, & S. A. Thompson (Eds.), Essays on Language Function and Language Type (pp. 437–467). Amsterdam: John Benjamins.Find this resource:

Stowell, T. (1981). Origins of phrase structure. Doctoral thesis, MIT, Cambridge, MA.Find this resource:

Talmy, Leonard. (1985). “Lexicalization patterns: semantic structure in lexical forms.” In T. Shopen (Ed.), Language Typology and Syntactic Description 3: Grammatical Categories and the Lexicon (pp. 57–149). Cambridge, U.K.: Cambridge University Press.Find this resource:

Taylor, J. R. (1996). “On running and jogging,” Cognitive Linguistics, 7, 21–34.Find this resource:

Tenny, C. (1987). Grammaticalizing aspect and affectedness. Doctoral thesis, MIT, Cambridge, MA.Find this resource:

Tenny, C. (1992). The aspectual interface hypothesis. In I. A. Sag & A. Szabolcsi (Eds.), Lexical matters (pp. 1–30). Stanford, CA: CLSI.Find this resource:

Tenny, C. (1994). Aspectual roles and the syntax-semantic interface. Dordrecht, Netherlands: Kluwer.Find this resource:

Van Valin, R. (1990). Semantic parameters of split intransitivity. Language, 66, 221–260.Find this resource:

Van Valin, R., & Wilkins, D. (1996). The case for “effector”: Case roles, agents, and agency revisited. In M. Shibatani & S. A. Thompson (Eds.), Grammatical constructions: Their form and meaning (pp. 289–322). New York: Oxford University Press.Find this resource:

Van Valin, R. & La Polla, R. (1997). Syntax: Structure, Meaning, and Function. Cambridge, U.K.: Cambridge University Press.Find this resource:

Verkuyl, H. (1972). On the compositional nature of the aspects. Dordrecht, Netherlands: ReidelFind this resource:

Verkuyl, H. (1993). A theory of aspectuality: The interaction between temporal and atemporal structure. Cambridge, U.K.: Cambridge University Press.Find this resource:

Wasow, T. (1977). Transformations and the lexicon. In P. Culicover, T. Wasow, & A. Akmajian (Eds.), Formal syntax (pp. 327–360). New York: Academic Press.Find this resource:

Wasow, T. (2002). Postverbal behavior. Stanford, CA: CLSI.Find this resource:

Wunderlich, D. (1997). Cause and the structure of verbs. Linguistic Inquiry, 28, 27–68.Find this resource:

Notes:

(1.) Nouns and adjectives can also take arguments, though verbs are the prototypical predicators. For this reason, the examples in the text make reference to verbs. There is a rich literature of the effects of category on argument realization (e.g., Chomsky, 1970; Grimshaw 1990; Landau, 2009).

(2.) The idea that weather verbs have no arguments (in English) is a matter of controversy (e.g., Chomsky, 1981).

(3.) See Butt (2006)

(4.) For an alternative minority view, see Taylor (1996).

(5.) Most change of state verbs participate in the causative alternation, and so in principle, we might want to put the cause argument in parentheses).

(6.) The generative semanticists introduced these primitive predicates as abstract heads in syntax. We return to this issue in section 4 “Where Do Arguments Come From?”

(7.) It is likely in fact that some agents are represented one way and others another way.

(8.) Of course, a complete account of argument realization must cover more than just subject and object selection, but space limitation prevents us from widening the scope of the discussion.

(9.) Rappaport Hovav (2008) suggests that certain incremental themes do not show core direct object properties because the scalar change expressed in sentences they appear in are not directly lexicalized by the verb.

(10.) This is of course a simplification, since not all unaccusative verbs have only one argument (see Alexiadou & Schäfer, 2011; Levin & Rappaport Hovav, 1995)

(12.) All of these alternations can be morphologically marked in some language or another, but the role of morphology in argument realization is beyond the scope of this paper.

(13.) This is true for English and many other languages, but not true for other languages, such as Greek and Hebrew (Alexiadou & Doron, 2012). The predictability of the passive from the active was one of the major motivations for the postulation of deep structure and transformations in early generative grammar (Chomsky, 1957, 1965).

(14.) There are well-known exceptions for sentences with quantifiers (Chomsky, 1957). Another alternation which has no truth-conditional effects is locative inversion. The dative alternation sometimes, but not always, involves truth-conditional differences (Rappaport Hovav & Levin, 2008).

(15.) Hale & Keyser (1993) and Baker (1997) have also attempted to show that the integration of roots with event structure should be considered cases of head movement as in syntax, since it is governed by the same constraints as head-movement in syntax.

(16.) These newer approaches avoid the problems faced by the generative semanticists in a number of ways, for example by taking into consideration the fact that in periphrastic causatives such as ‘cause to die’ there are two tense nodes.