Double-subject and complex-predicate constructions

Cognitive Linguistics 14–1 (2003), 1–45 0936–5907/03/0014–0001 © Walter de Gruyter Double-subject and complex-predicate constructions* TOSHIYUKI KUMASHIRO and RONALD W. LANGACKER Abstract Double-subject constructions in Japanese are analyzed from the standpoint of cognitive grammar. Their characterization in this framework, based on reference-point relationships and other independently attested phenomena, is unproblematic. Several grammatical properties indicate that the expres- sions in question—consisting of an outer subject juxtaposed with a nuclear clause—actually represent two distinct classes of constructions: double- subject constructions (properly called) and complex-predicate constructions (with just a single subject). It is argued that both types involve the same basic conceptual configuration, from which complex predicates arise in parti- cular circumstances. When the nuclear clause has insufficient conceptual autonomy, inherently invoking the outer subject as a reference point, the nuclear predicate and the reference-point relationship profiled at the higher level of organization collapse into a single, complex predicate. Keywords: double-subject construction; complex predicate; cognitive grammar; Japanese; reference point; topic. 1. Introduction The phenomenon commonly referred to as “double-subject constructions” poses numerous descriptive and theoretical challenges. Considerable pro- gress toward meeting those challenges has been realized in research carried out from the standpoint of cognitive grammar (CG). Langacker 1999b (also to appear) proposed a basic analysis of double-subject con- structions and showed that their description in cognitive grammar is revealing and straightforward. A comprehensive treatment of the Japanese phenomenon by Kumashiro (2000) demonstrated the existence of numerous constructional variants and detailed their structural properties, 2 T. Kumashiro and R. W. Langacker The present work builds on this previous research by examining one particular matter in greater depth. There is strong evidence in Japanese for dividing the constructions into two broad classes, despite their identical form. Grammatical behavior indicates that the members of only one class are properly called double- subject constructions; members of the other are more appropriately referred to as complex-predicate constructions.1 Yet it is quite obvious that the two families of constructions are closely related, in terms of form, meaning, and function. We will try to show that their differences stem from a single conceptual factor, namely the degree of conceptual autonomy enjoyed by a clause-like nucleus. When it has a lesser degree of autonomy, this nucleus collapses into a complex predicate taking as its subject a preceding noun phrase with topic-like function. 2. The phenomenon The expressions in question have the basic form [NP1 [NP2 PREDI- CATE]]. [NP2 PREDICATE] is a clause-like nucleus; NP1 has topic-like function with respect to this nucleus; and both noun phrases have some claim to being subjects. Though prevalent in the eastern portions of Asia, suggesting an areal phenomenon, this type of construction is not unheard of elsewhere.2 Here are some representative examples:3 (1) Taroo-ga hana-ga hikui. [Japanese] Taro-S nose-S low ‘Taro has a flat nose.’ (2) Nihonjin-ga kome-ga shushoku-da. Japanese-S rice-S staple:food-be ‘The Japanese have rice as their staple food.’ (3) Na-ka pay-ka aphu-ta. [Korean] I-S stomach-S ache-ASSR ‘My stomach aches.’ (4) Na-ka ku salam-i coh-ta. I-S the man-S likable-ASSR ‘I like the man.’ (5) T dùzi è. [Mandarin] he stomach hungry ‘He is hungry.’ (6) T tóu téng. he head painful ‘He has a headache.’ (7) Chán tháaw too. [Thai] I foot big ‘I have big feet.’ Double-subject constructions 3 (8) Kháw h u yaaw. he ear long ‘He has long ears.’ (9) Wa khicaa-yaake bhugin chama du. [Newari] the dog-COM fly one exist ‘There’s a fly on the dog.’ (10) Ji-ta wa baanlaa. I-DAT she beautiful ‘I think she’s beautiful.’ (11) Noo = n no-puush konoknish. [Luiseño] I = 1s:PRES my-eye green ‘I have green eyes.’ (12) Noo = up no-te’ tiiwu-q. I = 3s:PRES my-stomach hurt-PRES ‘I have a stomach ache.’ Across languages these constructions naturally vary in specific details, as well as the kinds of evidence available to support a particular analysis. They seem nonetheless to represent a unified phenomenon, characterized by the coherence and consistent co-occurrence of a substantial cluster of properties. Note that at this juncture we are still treating “double-subject constructions” as a homogeneous class; it is not until section 6 that we will make the distinction (at least for Japanese) between “true” double-subject constructions and complex-predicate constructions. With only minor qualifications, the following properties are characteristic of both subtypes. First, of course, is the property responsible for the construction’s unoffi- cial name: NP1 and NP2 both have some claim to subjecthood. Observe that in Japanese both take the subject marker ga.4 In Luiseño, both appear in the unmarked form characteristic of subjects (e.g., noo ‘I’, no-puush ‘my eye’, as opposed to ney ‘me’, no-puuch-i ‘my eye OBJ’). Moreover, a second-position clitic that regularly agrees with the subject can, in this con- struction, agree with either NP1 or NP2. Thus the first person singular clitic = n agrees with noo ‘I’ in (11), while in (12) the third singular = up agrees with no-te’ ‘my stomach’. Luiseño (Hyde 1971; Steele 1976, 1977, 1990) also illustrates two pro- perties pertaining to clausal organization. For one thing, NP2 and the predicate can themselves constitute a clause. Thus the nucleus in (11), no- puush konoknish ‘my eyes are green’, and that in (12), no-te’ tiiwu-q ‘my stomach hurts’, can each stand alone as a clause or a sentence. At the same time, the full expression also has clausal status. One indication of this in Luiseño is that the clitic, which occurs in second position within a clause, attaches to NP1 even when it agrees with NP2. This suggests that the entire 4 T. Kumashiro and R. W. Langacker structure, [NP1 [NP2 PREDICATE]], has clausal status. Pointing in the same direction is the absence of any intonation break between NP1 and the nucleus. Another property is that, in a sense to be explicated later, NP1 functions as a topic with respect to the nucleus. We assume that the examples cited make this evident. In some languages, it may even be questionable whether NP1 should be analyzed as a subject or just as a topic.5 The examples (which we believe to be fairly representative) illustrate two additional properties, one of them being that NP1 is typically a possessor with respect to NP2. This possessive relationship is usually not morphologically marked, but is merely indicated by the juxtaposition of NP1 and NP2 in the context of the overall construction. In Luiseño, however, it has to be made explicit through a possessive prefix on NP2 (e.g., no-puush ‘my eye’). We can also observe that a particular kind of possession predominates: body-part relationships between NP1 and NP2 are prototypical. Expres- sions of this sort (e.g., [1], [3], [5] to [8], [11] and [12]) are most frequently cited and appear to constitute the core of double-subject constructions, both cross-linguistically and in a single language. A related property is that NP1 is commonly an experiencer with respect to the nucleus. This follows in cases where NP1 and NP2 are in a body-part relationship and the predicate describes a sensation (e.g., [3], [5] and [6], [12]). An experiencer role for NP1 is at least suggested as being potential when the predicate merely describes a part of its body (as in [1], [7] and [8], or [11]). In nonpossessive examples, the role of experiencer may likewise either be ascribed to NP1 explicitly (e.g., by the predicate in [4], and by the dative case in [10]) or evoked as a possibility (as in [9]). This strong experiential component goes along with the further property that these sentences are consistently stative.6 Additionally, it is worth noting that an English translation with the verb have is virtually always available for these sentences, and usually quite natural. The translation of (3) could just as well be ‘I have a stomach ache’, and for (9), ‘The dog has a fly on it’. In the case of (5), ‘He has hunger’ is awkward, but better with further elaboration, e.g., ‘He has considerable hunger’ (cf. French Il a faim). Plausible translations with have are even possible for the examples describing opinions: (4) ‘I have the man down as likable’; (10) ‘I have her as beautiful’. This translation pattern is too consistent to be coincidental. 3. Some notions of cognitive grammar As a basis for describing double-subject and complex-predicate construc- tions, we must briefly review some claims and constructs of cognitive grammar.7 Its central claim is that lexicon, morphology, and syntax form a Double-subject constructions 5 continuum consisting solely of symbolic structures (i.e., pairings between semantic and phonological structures), which form assemblies of any size. It follows that all elements validly posited in grammatical description have some kind of meaning (often quite schematic). Cognitive grammar presupposes a conceptualist semantics that accom- modates our multifaceted ability to construe the same situation in alternate ways, resulting in alternate meanings, even when the same content is evoked. A particular way of construing conceptual content is inherent in the meaning of every lexical and grammatical unit. Two general aspects of construal are especially important here: the nature and degree of prominence conferred on the various elements of a conceptualization; and dynamicity, or how a conceptualization develops through processing time, particularly the sequence in which its elements are mentally accessed. A brief example of dynamicity is the contrast in (13), where opposite paths of mental access are followed in building up to what, in static terms, is the same conceived situation. (13) a. The mustard is in the pantry, on the top shelf, behind the peanut butter. b. The mustard is behind the peanut butter, on the top shelf, in the pantry. We will return to dynamicity in section 4. For now our main concern is with prominence. Crucial for grammar is the type of prominence called profiling, illus- trated in Figure 1. An expression evokes an array of conceptual content, called its base, from which it selects some substructure as its profile (shown in bold). Defined as the entity an expression designates, or conceptually refers to, the profile is the focus of attention within its base. Expressions with the same base may nevertheless exhibit a semantic contrast by virtue of their choice of profile. For instance, both husband and wife presuppose the conception of a marital relationship (represented as a line) between a male and a female. The essential difference in their meanings resides in the Figure 1. Base and profile 6 T. Kumashiro and R. W. Langacker profile imposed on this common base—husband refers to the male, and wife designates the female. Similarly, the verb admire and the noun admirer are based on the same conceptual content, wherein some individual holds a strongly positive mental attitude (shown as a dashed arrow) toward some other entity. The expressions are nonsynonymous because the verb profiles the relationship, i.e., the process of entertaining this attitude, whereas the noun designates the individual who does so. An expression can profile either a thing or a relationship (under abstract definitions of these terms).8 The nature of its profile (not its overall con- tent) is what determines its basic grammatical category. A noun profiles a thing. A verb profiles a process, defined as a relationship whose evolution through time is rendered salient.9 Hence admire is a verb because it desig- nates a process, whereas admirer profiles a thing and is thus a noun, despite invoking the same process as its base. As further illustration, Figure 2 indicates the profiling imposed by the verb choose and two senses of the derived noun choice. The verb designates the process of selecting one option (here a thing) from a range of alternatives (shown as a double- headed arrow). On one reading, choice1, the noun profiles the thing selected (e.g., the best choice). On the other reading, choice2, the thing it designates is abstract, deriving from the process through our basic capa- city for conceptual reification (e.g., make a choice). Represented by an ellipse, this abstract entity consists of one instance of the process. In expressions that profile relationships, we have to recognize an addi- tional type of prominence, namely the varying degrees of salience accorded to the relational participants. There is usually a primary focal participant, termed the trajector (tr), describable as the entity the expression is con- cerned with locating, assessing, or characterizing. Often invoked for this purpose is a secondary focal participant, called a landmark (lm). In the case of admire, the person holding the mental attitude functions as trajector, and the object of admiration as landmark. The trajector of choose is the person making the choice, its landmark being the option selected. The trajector/landmark asymmetry naturally tends to align itself with intrinsic asymmetries based on degree of activity or other factors. Observe, Figure 2. Profiling options Double-subject constructions 7 however, that their schematic definitions are solely based on prominence, and are not tied to any specific semantic role. Metaphorically, it is helpful to conceive of trajector and landmark status as primary and secondary spotlights of focal prominence. According to need, these spotlights can be directed at anything onstage. These descriptive constructs are shown to be necessary by pairs of expressions that are nonsynonymous despite having the same conceptual base and profiling the same relationship within it. An example is before vs. after, sketched in Figure 3. Each profiles a (nonprocessual) relationship in the domain of time (t) between two events. Indeed, they profile the same relationship of temporal precedence: referentially, a before relation is also an after relation. Since they have the same content and the same profile, their semantic contrast can only be ascribed to trajector/landmark align- ment: whether the event being located (the trajector) is situated with respect to a later event or an earlier one (in either case the landmark). The difference is just a matter of making opposite choices in directing the stronger and weaker spotlights. Both linguistic expressions and the grammatical patterns they instanti- ate are characterized as assemblies of symbolic structures. In the case of expressions, these symbolic structures are specific and capable of phono- logical realization. In the case of grammatical patterns (or “rules”), some or all of the symbolic structures are schematic. A grammatical pattern, or constructional schema, is simply the reinforced, abstracted commonality inherent in a number of occurring expressions. As such, it can function as a template for constructing and understanding new expressions on the same pattern. An expression’s degree of well-formedness depends on the extent of its compatibility with the various constructional schemas evoked to categorize it. The symbolic structures comprising an assembly are linked by cor- respondences (represented as dotted lines) specifying the identity of par- ticular substructures. In Figure 4, for example, the schematic landmark of admires is identified with the profile of Bill (whose semantic specifications are abbreviated as B). Admires and Bill are two component structures which are integrated to form the composite structure, namely admires Bill. Figure 3. Trajector/landmark alignment 8 T. Kumashiro and R. W. Langacker Correspondences specify how the component structures are to be integra- ted, by indicating which of their substructures are the same, hence super- imposed (“unified”) at the composite structure level. In the case of admires Bill, the composite-structure landmark inherits the semantic specifications of Bill and is therefore specific rather than schematic. Observe that the composite structure also inherits the profile of admires, i.e., the composite expression admires Bill designates the process of admiring, not the person admired. The box enclosing admires is thus in bold, highlighting its status as profile determinant for the composite expression. A component struc- ture functioning as profile determinant, at a given level of organization, constitutes what is traditionally recognized as the grammatical head at that level. While it is typical for one component structure to be distin- guished in this manner, this is by no means always the case (many constructions are exocentric). The integration of admires and Bill to form the composite structure admires Bill represents one level of grammatical organization (one con- struction). At a higher level, admires Bill functions as a component struc- ture, combining with Alice to form the overall composite structure Alice admires Bill. In this fashion a kind of constituency is capable of emerging within a symbolic assembly.10 The higher-level construction is based on a correspondence between the schematic trajector of admires Bill and the profile of Alice (whose semantic specifications are given as A). Admires Bill functions as profile determinant at this level, since once again the com- posite expression profiles the process of admiring, not the person doing it. As described in cognitive grammar, a finite clause always profiles a process (Langacker 1991: 5.1). Figure 4. Symbolic assemblies Double-subject constructions 9 A basic claim of cognitive grammar is that all valid grammatical constructs are definable in symbolic terms, and therefore have some kind of meaning (often quite schematic). A noun, for instance, is a symbolic structure that profiles a thing, and a verb is one that profiles a process.11 The grammatical notions subject and object are defined with reference to symbolic assemblies and the semantic constructs trajector and landmark (whose adoption was justified on independent grounds). Schematically, a subject can be characterized as a nominal expression whose profile cor- responds to the trajector of a profiled relationship, and an object, one whose profile corresponds to a landmark. In Figure 4, Alice is thus the subject with respect to admires, admires Bill, and Alice admires Bill, and Bill is the object at all three levels. These are offered as universal and fully general characterizations, but not as complete descriptions of subjects and objects in any one language. A full description must further specify the subject and object prototypes (e.g., agent and patient), as well as the range of nonprototypical values the language conventionally allows. We can also make a distinction between the subjects and objects occurring in different constructions or at different levels of organization. We can distinguish, for example, between a clausal object and a prepositional object, whose pro- files respectively correspond to the landmark of a process and that of a nonprocessual relation. 4. Reference-point phenomena We often want to direct attention to something that lies completely outside our interlocutor’s scope of awareness. One useful strategy is to go about this indirectly, by first directing attention to an associated entity that is more salient or easily accessible, which can then serve as a point of ref- erence for reaching the object of interest. We can use this strategy at either the perceptual or the conceptual level: (14) a. Do you see that big rock? There’s a lizard just in front of it. b. Do you remember