Lexical-Constructional Subsumption in Resultative Constructions
Lexical-Constructional Subsumption in Resultative Constructions in English
Francisco José Ruiz de Mendoza Ibáñez & Alba Luzondo Oyón
This paper is concerned with a descriptively and explanatorily adequate account of otherwise puzzling acceptability differences of the type exemplified in (1) below:
(1) a. He drove/*guided me crazy.
b. He hammered/*hit/*stroke the metal flat.
c. He carved/*cut the wood into a toy.
d. He cut the wood into boards.
e. He ran/#strolled his Nikes threadbare.
These examples are instances of what is commonly referred to in the literature as the resultative construction, i.e. a type of transitivity pattern that specifies the outcome of a change of state, property or location undergone by a person or an entity. Resultatives have been the object of a large number of studies from different perspectives: formal (e.g. Hoekstra 1988; Levin 1993), functional (e.g. Halliday 1967), cognitive and/or constructionist (e.g. Boas 2002, 2003, 2005, 2008ab; Broccias 2003, 2004; Goldberg and Jackendoff 2004; Iwata 2006). Since the amount of research is immense, we will focus our attention on what we believe—for reasons that will become apparent in our analysis below—is the most promising approach to resultatives, i.e. one where the syntactic configuration is motivated by the principled interaction between lexical and constructional structure. This is the perspective generally adopted by constructionist studies within Cognitive Linguistics, including the Lexical-Constructional Model (LCM henceforth), as developed by Ruiz de Mendoza and Mairal (2008, 2011), and Mairal and Ruiz de Mendoza (2008, 2009) (cf. Butler 2009, for a critical overview). Closely connected with the resultative constructions exemplified in (1) above are constructs or patterns such as the ones in (2):
(2)a. Pat kicked the ball across the field.
b. I want you out of my house.
c. The audience laughed the actor off the stage.
d. She loved me back into existence.
Following Goldberg (1995, 2006) these are instances of the caused-motion construction, which is often schematized as ‘x causes y to move z’: (2a) is a standard non-figurative use where caused motion is the result of physical impact; (2b) is an example of manipulative construction (Gonzálvez-García 2008), where caused motion is the result of the speaker’s ability to direct someone else’s behaviour; (2c) is a case where expected motion is the future result of psychological or emotional impact; finally, (2d) makes figurative use of the caused-motion configuration in order to express a change in an emotional state. For current purposes, the canonical meaning of this construction involves the (actual or prospective) change of location of an entity by an instigator of motion in such a way that the entity in question moves along a designated path. This canonical meaning covers cases like (2a), (2b) and (2c), while (2d) can be considered an extended use. The caused-motion construction has been discussed in the literature in relation to resultative constructions and construction grammarians like Goldberg (1995) have argued that the caused-motion construction is a metaphorical extension of resultatives on the basis of the observation that a change of state can be seen as a change of location (cf. Lakoff 1987, 1993; Lakoff and Johnson, 1999). Since, according to Goldberg, resultatives express resultant states and states can be seen as locations, then a caused change of state can be seen as caused change of location, which is captured by the caused-motion construction. We will come back to this issue in the face of our data at a later stage in this paper.
The notion of construction is another issue that has generated a vast amount of literature. The reader may be referred to Schönefeld (2006) for an overview of different treatments and developments of this notion. Following Goldberg (1995) constructions have often been described as “form-meaning pairings” where the whole exceeds the sum of the parts. However, quite recently, the notion of construction has been broadened to encompass any (highly) frequent non-compositional form-function pairing (Goldberg 2006). This downgrading of the idiosyncrasy requirement in the definition of a construction—which, as it stands now, comes fairly close to Langacker’s (1987) notion of “symbolic assemblies”—has met with criticism on the basis that the resulting definition is too unconstrained (cf. Bod 2009, Östman and Fried 2004, among others). The LCM regards constructions as form-meaning pairings where form realizes semantic (or conceptual) structure and, in so doing, cues for the activation of conceptual structure. Furthermore, conceptual structure can vary in nature: lexical structure is captured by lexical templates (cf. Mairal and Faber 2007), which in the case of verbal predicates, consists of amalgams of lexical functions and primes that have a number of external variables (for syntactic projection) within their scope; then, the LCM recognizes a number of constructional types, among them argument structure constructions, such as the caused-motion construction, which are the result of abstracting away elements that are common to lower-level verbal configurations (e.g. caused-motion verbal predicates such as kick, push, thrust, etc.). The LCM then specifies the conditions for the combination of lexical structure and argument constructions and also for the incorporation of this output into higher-level constructions of a pragmatic and discourse kind, a process that we will summarize in section 3 below. Treating argument constructions as having their own ontological status is useful in order to account for some uses of lexical predicates, which would otherwise require postulating unreasonable amounts of lexical polysemy as is evident from (2b), (2c), and (2d) above. Therefore, we concur with Goldberg (1995, 2006) that it is not reasonable to argue that want, laugh, and love have an intrinsic caused-motion sense. But if we acknowledge the independent status of the caused-motion construction, then it is only necessary to specify under what circumstances it is possible to incorporate into the construction predicates like the ones mentioned above but not others (cf. *I write you out of my house, *The audience described the actor off the stage, *She approached me out of existence). In Goldberg’s initial proposal, the construction adds (new) meaning (e.g. in Pat sneezed the napkin off the table we see the sneezing event as part of a general causal chain that can be paraphrased as ‘Pat’s sneezing caused the napkin to fall off the table’). The construction is not only capable of contributing arguments (observe that sneeze is actually an intransitive verb, and therefore the Y and Z arguments are supplied by the caused-motion construction in this example), but also of creating semantic constraints on the predicates that may fuse (Goldberg 1995: 50) with each particular construction. These findings, some of which have been supported experimentally (cf. Eddington and Ruiz de Mendoza 2010, for a review), are important to the extent that they allow the linguist to simplify the descriptive apparatus of a theory while preserving and even enhancing its explanatory power. However, there is, in our view, a more fundamental issue that still awaits a fully satisfactory response on both descriptive and explanatory grounds: there should be a fine-grained account of how lexical-constructional fusion works and what it is that licenses or puts limits to the incorporation of a lexical item into a given construction, as noted above. Usually, the notion of coercion (Goldberg 1995, 2006; Michaelis 2003) is invoked in this connection. The argument here is that lexical-constructional fusion is generally possible if there is conceptual compatibility between lexical and constructional configurations; but when there is no such compatibility, fusion may still take place if it is possible for constructional meaning to coerce lexical meaning. This is, however, a very limited explanation since it does not provide a sufficiently fine-grained specification of when coercion is possible and when it is not. In our view, the LCM has developed enough analytical tools to provide an initial solution to the problem in the form of constraints on lexical-constructional fusion (a process which in the model is referred to as subsumption, thus capturing the fact that lexical structure may be altered to be adapted to constructional structure but not the other way around).
The present paper aims to examine in detail lexical-constructional fusion processes for the resultative construction, which Goldberg (1995) characterizes as ‘x causes y to become z’, in which the Z element is an adjectival predicate (e.g. He hammered the metal flat). In our work, the caused-motion construction and the resultative are considered members of the larger family of resultative constructions (Goldberg and Jackendoff 2004). A similar view is held by Boas (2003) and Broccias (2003, 2004), who, on the basis of the states are locations metaphor, do not posit a distinction between these patterns since ultimately both denote a change in the postverbal element (whether it designates a state or a location). We also contend that there are two basic interpretation schemas underlying the resultative construction. One, which we schematize as A > A’, is what we will refer to as the “prototypical resultative construction”; this configuration, which realizes the Z element through an adjectival predicate (e.g. He painted the house red), occurs in cases in which there is a resulting state where the patient is observed as solely gaining a new property. A second schema, which we represent as A > B, obtains when, as a consequence of the action denoted by the verb, the patient is observed as changing into a new state of being. Interestingly, this schema is a resultative construction that makes use of the caused-motion configuration (e.g. He smashed his guitar to pieces). We thus examine some examples that fall into one or the other category and discuss their basic properties in terms of various amalgams of cognitive models as well as the principles that license their combination.
2. The Goldbergian approach to resultatives
When analyzing the connections that hold between verbs and constructions, Goldberg (1995, 2006) associates lexical items with rich frame semantic meanings. Hence, once constructions are decomposed into abstract predicate-argument structures such as cause-become <agt pat result-goal> for the resultative, the verb’s profiled participant roles fuse with the construction argument roles in a principled manner. According to her, it is the construction that specifies which verbs may be incorporated into the construction and which may not, applying both general and construction-specific constraints to predicate fusion. However, in such a fully constructionist theory little consideration is given to the role that lexical items play in this process. While Goldberg does not deny that verbal semantics does play a role, it is nonetheless true that the descriptive and explanatory burden revolves around constructional semantics rather than lexical semantics in her strand of Construction Grammar. In this connection, Boas (2003, 2008b) has criticized Goldberg’s account along the following lines: if the lexical entries of semantically related verbs such as talk, speak, whisper, grumble, murmur and sigh display the same structure, what is it that prevents the fusion of most of these predicates with the resultative construction?
(3) a. Miriam talked herself blue in the face.
b.*Miriam spoke/grumbled/murmured herself blue in the face.
c. ?Miriam whispered/sighed herself blue in the face.
These examples alone suffice to show that Goldberg’s account is not sufficiently refined to explain the distribution of predicates within the resultative construction, and thus both the role of constructions as well as that of predicates should be observed and explained when dealing with the complex issue of lexical-constructional subsumption. But to Boas’s examples, we can add others that express resultative meaning through the figurative use of the caused-motion construction:
(4) a.The blacksmith hammered the piece of iron into the shape of a dove.
b.Bush stunned me into silence.
c.He almost scared me out of my wits.
Examples of the type in (4) above, regardless of their varying degrees of productivity, cannot be merely discarded or assigned to the ‘periphery’. It is our position that these qualify as constructions and are to be accounted for too. We need to know (i) why they make use of a caused-motion configuration instead of an adjectival predicate, and (ii) why predicates like scare in (4c), which inherently expresses result (the predicate can be decomposed into ‘cause to become frightened’), can non-redundantly be fused into a construction that makes explicit the resultative element.
3. The Lexical-Constructional Model: a brief overview
Before presenting our analysis on lexical-constructional fusion under the scope of the resultative pattern, we shall provide the reader with a brief outline of the LCM. This model emerges as an effort to reconcile a number of opposed theoretical assumptions held by projectionist theories, on the one hand, and cognitively-oriented approaches to language, on the other. To that end, the LCM combines insights from Role and Reference Grammar (Van Valin 2005), Construction Grammar (Goldberg 1995, 2006) and Cognitive Semantics (Lakoff, 1987; Lakoff and Johnson 1999). Functional approaches usually work under the assumption that syntactic structures can be predicted on the basis of the information coded by the lexical item in conjunction with a set of linking rules, thus disregarding the indisputable strength that constructions display in predicting morphosyntactic structure. However, as has been made obvious in Goldberg (1995, 2006), constructions are capable of adding arguments whose final meaning (i.e. the caused-motion/resultative sense in the example They scorned him into a depression) is not derivable from verbal projection (i.e. the argument structure of scorn). Conversely, construction-based approaches have not fully explained what it is that allows and/or constrains the unification between a given syntactic pattern and a lexical entry. Why is it possible for us to say The bread cuts easily, The window breaks easily (middle construction), while a sentence such as *The building destroys easily is odd? Thus, the LCM has as its primary concern to develop a usage-based (i.e. focused on real attested data) model of meaning construction that is capable of explaining these facets of meaning.
An important methodological assumption sustained by the LCM has to do with the idea that all levels of linguistic description may make use of the same or at least comparable cognitive processes. This assumption has been termed the equipollence hypothesis (Mairal and Ruiz de Mendoza 2009). Metaphor and metonymy have been widely attested at the lexical level, but the two phenomena have been proved to occur at other levels of linguistic explanation, including so-called core grammar. Thus metonymy has been found to underlie such varied phenomena as categorial and subcategorial conversion processes, as well as some constructional alternations (cf. Ruiz de Mendoza and Pérez 2001; Ruiz de Mendoza and Peña 2008; Peña and Ruiz de Mendoza 2009). For example, the metonymy instrument for action motivates the categorial conversion of the noun hammer into a verb in He hammered the nail into the wall (Kövecses and Radden 1998). Similarly, the countable noun America is made uncountable in There is a lot of America in what she does, as motivated by the metonymy an entity for one of its properties. Or, think of the use of the apparently deviant complementation pattern of verbs like enjoy and begin in She enjoyed the beer (i.e. She enjoyed/began drinking/canning, etc., the beer), where the metonymy object for action is at work. The use of static verbs like live with a dynamic preposition—explained by Lakoff (1987) as involving image schema transformations—is also motivated by a metonymy, as in She lives over the hill/across the road/past the post office, all of which hinge of the metonymy action for result. Metaphor has so far only been detected in some grammatical alternations. A very clear case of metaphor is when we see one type of action as if it were another type of action (Ruiz de Mendoza and Mairal 2007). For example, in He talked me into it we treat the verb talk, which designates an activity, as if it were an effectual action (i.e. an action that has a direct physical impact on the object causing it to change location). Since metaphors and metonymies of this kind are based on generic cognitive models such as ‘action’, ‘result’, ‘entity’, ‘property’, they are usually labelled “high-level” metaphor and metonymy.
In the LCM high-level metaphor and metonymy are viewed as (external) constraining factors on lexical-constructional fusion. Other (constrained) cognitive processes such as conceptual cueing (i.e. a form of guided inferential activity) and constructional subsumption (i.e. the incorporation of lower level constructions, such as lexical templates, into higher level ones) are also contemplated in the model. For the sake of clarification, let us deal with one example. Consider again the caused-motion use of the verbal predicate laugh, as in (2c) above or in the following non-artificial example:
(5)The men who laughed him out of the St. Louis Convention when he prophesied Parker’s defeat have wondered ever since if they did not take Murphy too lightly.
Since laugh is an activity predicate (i.e. one designating a dynamic, non-telic and non-punctual state of affairs; cf. Van Valin 2005: 33), we need to address the issue of why and how this predicate—which typically governs a prepositional complement—can participate in a construction that requires a causative accomplishment predicate (which, unlike activities, is telic and involves a causal chain in its event structure specification; cf. Van Valin 2005: 33). The LCM suggests that laugh in sentences like (2c) and (5) has suffered a subcategorial conversion process from laugh-at’ (x, y) to laugh’ (x, y). It is further postulated, as in Construction Grammar, that subcategorial conversion is partially a consequence of the Override Principle (Michaelis 2003), which states that the meaning of a lexical item has to conform to the meaning of the structure in which it is placed. Thus, in order for laugh to be part of the caused-motion construction, some meaning adjustment is necessary whereby laugh ceases to be an activity predicate and is made to function as if it were a causative accomplishment predicate, capable of designating caused-motion. The difference with the Construction Grammar approach is that the LCM claims that the reason why the meaning adjustment is possible lies in a metaphorical interpretation of the verb. In harmony with the definition of metaphor largely accepted within the field of Cognitive Linguistics (i.e. a conceptual mapping between two domains), the LCM proposes the high-level metaphor an experiential action is an effectual action to account for this phenomenon. Logically, such a metaphor imposes constraints on lexical-constructional subsumption by licensing, as in (6) and (7), or blocking, as in (8), subcategorial conversion of other “experiential action” predicates: