introduced by the three complementizers. It seems clear that no, koto, and to code some sort of scale relative to the directness with which the scene profiled in the complement clause is construed. However, the structural difference between no and koto, on the one hand, and to, on the other hand, needs to be recognized. The contrast between no and koto has been expressed in different terms over the years. Kuno (1973) explains it in terms of a concrete-abstract distinction. Josephs (1976) analyzes it in terms of directness versus indirectness. More recently, Horie (2000) argues that the semantic category coded by the no-koto contrast is most judiciously expressed in terms of event versus proposition. His account is certainly compatible with the earlier ones. The perception of an event is more direct and, in that sense, more concrete than the conception of a proposition, an object of thought remote from the immediacy of perception. Horie’s analysis explains why no can sometimes encode a proposition if the main verb is a cognitive one, such as siru ‘learn’, but koto does not usually code events, as illustrated in (16): (16) *Mary-wa [John-ga toori-o wataru] koto-o mi-ta. M top J nom street-acc cross nr-acc see-pst ‘Mary saw John cross the street.’ The distinction between the nominalizing no and koto and the nonnormalizing to has also been investigated. In Kuno (1973), it is expressed in terms of factivity- nonfactivity. Suzuki (1996) argues that no, koto, and to form a continuum in relation to the extent to which the speaker accepts the information presented to him or her. In later work, however, she claims that the structural difference be- tween to and the nominalizing complementizers is best expressed in terms of Frajzyngier’s (1991) terminology of de re versus de dicto. The entities coded by koto and no (events or propositions) belong to the domain of reality (de re), whereas the complements introduced by to belong to the domain of speech (de dicto) (Suzuki 2000: 34). Suzuki claims that ‘‘the framework of de re vs. de dicto was chosen because the notion of domain of speech fits well with the character of to, which was originally used only for reporting another speaker’s statement and later became a complementizer with a wider application’’ (34). As a de dicto complementizer, to’s main function is to mark ‘‘the speaker’s psychological distance from the information expressed in the complement clause’’ (Suzuki 2000: 37). This function is clearly visible in the minimal pairs in (17) and (18), where to contrasts with no (from Suzuki 2000: 36): (17) Watashi-wa kare-ga nemutte-iru to mi-ta. I-top he-nom is-sleeping comp see-pst ‘I saw (judged) that he was sleeping.’ (18) Watashi-wa kare-ga nemutte-iru no-o mi-ta. I-top he-nom is-sleeping nr-acc see-pst ‘I saw him sleeping.’ With verbs of perception, the presence of no indicates a directly perceived event, whereas to indicates that the information in the complement is inferred, that is, 790 michel achard obtained in a less direct manner. This contrast directly parallels the one illustrated in (11) and (12) for English. 3.3. French In French also, events are usually coded with infinitival complements, whereas propositions are marked with finite clauses, as illustrated in (19) and (20): (19) Jean a vu sortir Marie. John has see.pp go.out Mary ‘John saw Mary go out.’ (20) Jean a vu que Marie e ´ tait sortie. John has see.pp comp Mary is.impf go.out.pp ‘John saw that Mary had left.’ In a way consistent with the English and Japanese systems, the less verbal infinitive form in (19) codes a directly perceived event, whereas the proposition in (20)is marked with a finite clause. However, French differs from both English and Japanese in two major re- spects. First, the distinction between infinitival and finite constructions often pertains to whether or not the main and subordinate verbs have similar or different subjects. The second one concerns the presence of an indicative-subjunctive dis- tinction when the complement form is finite. The infinitival-finite contrast is presented in (21) and (22): (21) Marie aime aller au cine ´ ma. Mary likes go.inf to.the cinema ‘Mary likes to go to the cinema.’ (22) Marie aime que son fre ` re aille au cine ´ ma avec elle. Mary likes com p her brother go.subj to.the cinema with her ‘Mary likes her brother to go to the cinema with her.’ In (21), the co-reference of the subjects of the verbs aimer and aller yields an in- finitival complement. In (22), the complement verb has a conjugated form (sub- junctive), because the main and subordinate verbs have different subjects. 7 With respect to the second contrast, the verbs of perception, declaration, and propositional attitude are most often followed by complements in the indicative, whereas the verbs of volition and emotional reaction are followed by complements in the subjunctive. This distinction is illustrated in (23) and (24): (23) Jean sait que vous avez de ´ me ´ nage ´ . John knows comp you have.ind move.out.pp ‘John knows that you moved out.’ (24) Jean veut que vous ayez de ´ me ´ nage ´ avant dimanche. John wants comp you hav.subj move.out.pp before Sunday ‘John wants you to have moved out before Sunday.’ complementation 791 Although they are obviously related, the two contrasts illustrated in (21)–(24) are relatively independent from each other and are therefore analyzed separately. Building on Ruwet’s (1984) original insight that certain verbs semantically treat a self-to-self relation in a way similar to a self-to-other relation whereas other verbs treat them differently, I provide an account based on the viewing arrangement that exists between the conceptualizer and the scene conceptualized (Achard 1996, 1998). The notion of viewing arrangement needs to be understood in the sense of Langacker (1985, 1990, 1991); it refers to the vantage point from which the scene coded by a particular expression is conceptualized, as well as the precise nature of the relation that exists between the subject and object of conceptualization. I argue that the presence of a finite complement (indicative or subjunctive) reflects an objective construal on the scene coded in the complement from the vantage point of the main clause conceptualizer. Because the vantage point from which it is construed is external to its scope of predication (Langacker 1985, 1990), the whole scene, including the subject of the subordinate process, is part of the objective scene and thus profiled. The presence of an infinitival complement reflects the subjective construal of the scene coded in the complement. More precisely, that scene is conceptualized from the internal vantage point of the subordinate subject. Because the subordinate subject is construed subjectively, it is not specifically mentioned. The subordinate process alone is profiled. This configuration increases the involvement of the main subject in the scene coded in the complement, because he or she construes the latter from the vantage point of someone already involved in that process. This kind of analysis accounts for the fact that an infinitival com- plement can only occur in cases where the main and subordinate subjects are co- referential. 8 The objective-subjective construal of the scene coded in the complement is primarily a matter of speaker choice. However, as part of their semantic organi- zation, certain verbs tend to impose an inherently objective or subjective construal on their complement. The perception verbs, for example, impose an objective construal on their complement scene regardless of whether the main and subor- dinate subjects are co-referential or not (Ruwet 1984). Consequently, they are usually not found with infinitival complements, as illustrated in (25) and (26): (25)*Je remarque avoir les cheveux frise ´ s. I notice have.inf the hairs curly ‘I notice to have curly hair.’ (26) Je remarque que j’ ai les cheveux frise ´ s. I notice comp I have.ind the hairs curly ‘I notice that I have curly hair.’ Conversely, the volition verbs impose an inherently subjective construal on the complement scene, which is why a finite clause in cases of co-referentiality between the main and subordinate subjects is usually infelicitous. The examples in (27) and (28) from Ruwet (1984: 75) illustrate the well-documented phenomenon of obviation (Ruwet 1984; Farkas 1992): 792 michel achard (27) Je veux partir. I want leave.inf ‘I want to leave.’ (28)*Je veux que je parte. I want comp I leave.subj ‘I want that I leave.’ With the verbs that do not impose any inherent construal on their complement scene, an objective or subjective construal reflected by the presence of a finite or infinitival complement depends on the discourse context. The indicative-subjunctive distinction has some measure of independence from the infinitive-finite distinction. It essentially pertains to grounding, described by Langacker (1991: 440) as follows: ‘‘Grounding locates the event with respect to the speaker’s conception of reality.’’ Consequently, the base relative to which the meaning of the two inflections needs to be characterized is composed of the different Idealized Cognitive Models that articulate our conceptions of reality and possession. I argue that the propositions a given conceptualizer considers true can be manipulated just like concrete objects (Achard 2002b). Consequently, the con- ceptual control he or she exercises over those propositions can be understood in terms of abstract possession. Just as the set of objects a person owns defines his or her dominion, the set of propositions he or she considers true (i.e., his or her conception of elaborated reality) represents his or her dominion. The indicative inflection codes ‘‘the epistemic effort required to establish the conceptualized event’s putative location in elaborated reality. Importantly, this definition applies to any kind of epistemic effort that aims to establish control over the population’’ (Achard 2002b: 212; Langacker 2002). Crucially, the indicative is the only mood that can appropriately serve to describe the elements of a con- ceptualizer’s dominion, because it is the only one that can provide events with their necessary putative address in elaborated reality. Subjunctive complements do not describe an element of a conceptualizer’s dominion because they do not make the necessary reference to elaborated reality. The complement content is merely considered with respect to a specific local mental space. For example, in Jean veut que vous partiez ‘John wants you to leave’, the complement content is merely considered with respect to the mental space of John’s desires. Viewed in this way, the meaning of the indicative and subjunctive inflections emphasize the parallel between nominal phrases and clauses. Langacker (1991) argues that a finite (indicative) clause represents a grounded instance of a process type, that is, a unique element of some conceptualizer’s dominion. In that sense, it is similar to a definite nominal. I argue that a subjunctive clause represents an arbitrary instance of a process type, that is, an instance specifically conjured up for a specific purpose (Achard 1998, 2002b). The value of an arbitrary instance is independently motivated to account for the behavior of the English indefinite article a in opaque contexts (John wants to marry a dancer) and the generic use of the indefinite article (A cat is a mammal). complementation 793 3.4. Overall Results The results obtained with respect to the meaning of the complement forms in individual languages present an interest both at the cross-linguistic and language- internal levels. Cross-linguistically, the complement systems of English, French, and Japanese are remarkably similar. First, they are all consistent with Givo ´ n’s prediction that the forms that code more independent events resemble main/ independent clauses more closely. In the three languages considered, we witnessed a striking isomorphism between the level of independence of the event in the complement and the form by which it is expressed. Second, the same kinds of events get coded as more or less independent in different languages. For example, directly perceived events are consistently coded by whatever morphosyntactic means indicate the highest possible level of conceptual integration in the language considered. The similarities are even more striking if one overlooks the specific descriptive constructs invoked in a given language because of their particular relevance to that language. To give just one example, the notion of perspective is treated system- atically in French, but it is also relevant to English. We have seen that both French and English code the way in which certain verbs code the self-to-self relation as similar or different from the self-to-other relation. However, in English, this contrast is quite restricted because it is only manifested in the choice between to and for to. In French, it is more general and systematic because it is manifested by a language-wide constraint on same-different subject distribution, as well as decisions that pertain to indicative versus subjunctive mood selection and obvi- ation. The important point, however, is that both languages code the same contrast at different levels of generality. Language-internally, the most interesting result of Cognitive Linguistics re- search is the way in which it relates the meanings of the complement forms to the global ecology of individual languages. For example, in English the sameness of time constraint the -ing-form exhibits between the main and subordinate verbs is closely related to the meaning of -ing in participial clauses ( He came into the room screaming), where simultaneity between the events expressed in the verb and the participle is also implied (Wierzbicka 1988). In a similar fashion, the future orien- tation of the complementizer to is directly related to the meaning of to as a prepo- sition, namely, to express motion toward a goal (Duffley 1999; Smith and Escobedo 2001). Finally, in Japanese, the meaning of to as expressing the speaker’s psycho- logical distance is motivated by the expression’s semantic characterization as representing remote, detached knowledge in other constructions (Suzuki 2000). This precise evaluation of the semantic role of the complement constructions in the global ecology of a given language cuts to the very core of the meaning of comple- ment systems, because it may explain why certain languages code specific contrasts as opposed to other possible ones. For example, Horie (2000) shows that two languages as closely related as Korean and Japanese code different contrasts in their complement systems because of the importance those contrasts hold elsewhere 794 michel achard in the language. He shows that Korean codes the realis-irrealis contrast as opposed to the event-proposition contrast relevant to Japanese because of a global tendency to code the realis-irrealis contrast throughout the language. 4. Raising In addition to issues pertaining to the overall structure of complementation sys- tems, certain constructions have received individual attention in the literature be- cause they clearly illustrate relevant theoretical concerns. For example, the syntac- tic behavior of the raising verbs figures prominently in the argumentation in favor of different levels of representation in different models of Generative Grammar (Chomsky 1965; Rosenbaum 1967). Because of their heightened theoretical status, these constructions deserve to be treated separately. This section presents raising constructions as an example to show that Cognitive Grammar can provide a satis- factory account of phenomena usually regarded as strictly structural. 4.1. Description of Raising The facts about raising are well known. They are illustrated in (29)–(31). (29) Mary seems to understand. (30) Mary wants to understand. (31) It seems that Mary understands. Despite their similar surface form, the sentences in (29) and (30) are struc- turally quite different. First, seem imposes no restrictions on its subject, but want is usually only felicitous with animate subjects. Second, several structural tests were designed to show that while Mary is the real subject of want, it is not the real subject of seem, but rather that of understand. In order to capture this structural difference, generative linguists posited two separate underlying representations for raising (seem) and control (want) verbs. The raising verbs were assigned the un- derlying structure in (32), the control verbs the one in (33): 9 (32)[ NP e] seem [ S' COMP [ S Mary INFL [to understand]] (33) Mary want [ S' COMP [ S PRO INFL to understand]] The underlying form in (32) yields the surface form in ( 29 ). Mary is moved to the position of subject of seem. If the move does not occur, the empty subject position is filled by the dummy it to satisfy the structural need for a surface subject, as illustrated in (31). complementation 795 4.2. A Cognitive Grammar Solution The Cognitive Grammar solution differs from the generative account on two major points. First, it argues that the relation that exists between the raised and unraised constructions is not a structural one, but one of construal (Langacker 1991, 1995). Second, it shows that the syntactic behavior of the raising verbs directly results from their semantic organization, which directly challenges the need to assign them a specific underlying representation. Several researchers have pointed out that pairs of sentences such as (29) and (31) are not semantically equivalent (Newman 1981;Ruwet1983). Raising verbs are therefore considered polysemous, each with a raised and an unraised variant. Each variant profiles the conceptualized scene in a different way. With the unraised variant in (31), the impersonal it represents an abstract setting within which the proposition in the complement can be located (Smith 1985; Langacker 1991, 1995; Achard 1998). That setting is given focal prominence and therefore marked as the subject. The event or proposition as a whole is the landmark of the main relation. In the raised variant in (29), the main participant in the located event is chosen as the trajector of the main relation. Reality remains off profile, as part of the base. The raised variant of seem therefore profiles the apparent participation of an entity in a given process. In order for this analysis to be satisfying, we need to show that Mary can indeed be the subject of seem even though the latter is usually an event. This dif- ficulty disappears, however, if we recognize Langacker’s concept of active zones. The active zones of an entity are ‘‘those facets of an entity capable of interacting directly with a given domain or relation’’ (Langacker 1987: 485). With the raised variant, the complement process (understand) represents the subject’s (Mary) active zone with respect to its participation in the main relation (seem). It is with respect to understand that Mary can be considered the subject of seem. The choice of subject in both variants is a direct result of the raising verb’s semantic structure. 10 Verbs such as seem code the way in which specific facets of reality reveal themselves to some conceptualizers, without any epistemic effort on their part. The only entities available for subject selection are therefore part of the conceptualized scene. If no participant is particularly salient, the abstract location within which the scene can be observed (i.e., the relevant part of reality) is chosen. This configuration corresponds to the unraised variant. Any participant in the complement scene can be chosen as the main subject because of its salience, and that choice corresponds to the raised variant. The selection of the raised or un- raised variant therefore depends on the specific way in which the speaker chooses to structure his or her conceptualization for expressive purposes. This analysis makes an interesting prediction concerning the discourse dis- tribution of the raised and unraised variants: 11 It is claimed that, in Don is likely to leave, Don functions as a reference point with respect to the process of his leaving: the notion of leaving is accessed via the conception of Don and conceived in relation to that individual. The reference- point relationship is absent in the corresponding sentence That Don will leave 796 michel achard is likely , which consequently has a slightly different meaning. The ‘raised’ NP can be thought of as a kind of local topic, i.e. a topic for purposes of ascertain- ing the actual (or direct) participant in the profiled main-clause relationship (Don calls to mind a process involving Don, and such a process can be accessed for likelihood). It makes the prediction that raised NPs should tend to ex- hibit greater ‘topicality’ than their unraised counterparts. (Langacker 1995: 37–38) I argue that this prediction is indeed borne out for French subject-to-subject raising constructions (Achard 2000). Using a corpus of journalistic prose, I show that the overwhelming majority of the subjects of the raised variant of sembler ‘seem’ are indeed cognitively accessible (topical or inferable), while the subjects of the unraised variant tend to be less accessible. The semantic structure of the raising verbs directly accounts for the absence of restrictions placed on their subjects. It also explains their other syntactic behaviors. For example, in French, the distribution of the clitic en ‘of it’ is different with raising and control verbs. 12 This difference is illustrated in (35) and (36) (from Ruwet 1983: 17 ): (35)a.L’ auteur de ce livre semble e ˆ tre ge ´ nial. the author of this book seems be.inf brilliant ‘The author of this book seems to be brilliant.’ b. L’ auteur de ce livre pre ´ tend e ˆ tre ge ´ nial. the author of this book pretends be. inf brilliant ‘The author of this book pretends to be brilliant.’ (36) a. L’ auteur semble en e ˆ tre ge ´ nial. the author seems of.it be.inf brilliant ‘The author of it seems to be brilliant.’ b. *L’ auteur pre ´ tend en e ˆ tre ge ´ nial. the author pretends of.it be.inf brilliant ‘The author of it pretends to be brilliant.’ The examples in (35) and ( 36) show that the clitic en ‘of it’, replacing the prepo- sitional phrase de-NP (de ce livre), part of the subject NP, can only be cliticized on the subordinate verbs when the main verb is a raising verb. In the case of a control verb such as (36b), the cliticization of en on the subordinate verb is impossible. I show that the distribution of en results from the interplay between the respective semantics of the raising and control verbs and that of the clitic en in the specific context of the test situation (Achard 2001). 5. Conclusion This chapter has provided an overview of Cognitive Linguistics research on com- plementation. It presented the notion of conceptual subordination and showed the strong correlation that exists between the form of the complement and the level complementation 797 of independence of the event it codes. Several contrasts were introduced, coded by different constructions in English, Japanese, and French. These contrasts were shown to be relevant both language-internally and cross-linguistically. Finally, the Cognitive Grammar analysis of raising constructions was presented in order to illustrate the fact that a semantically based account can explain complex syntactic behavior in a satisfactory fashion. This overview is far from being exhaustive, but it should suffice to show that the area of complementation perfectly illustrates the isomorphism between form and function that stands at the core of Cognitive Linguistics. Despite its promising beginning, however, cognitive research on complemen- tation is far from being completed. In particular, the notion of conceptual sub- ordination needs to be defined more specifically. The broad range of phenomena it covers has raised questions as to the legitimacy of considering the grammar of complementation as a single category. For example, Thompson (2002, 155) argues that in everyday English conversation, the grammar of complementation ‘‘may be better understood as a combination of an epistemic/evidential phrase together with a declarative or interrogative clause’’ (see also Englebretson 2003). This alternative analysis raises interesting issues. For example, is conceptual subordination as it was presented in this chapter predominantly a written-language phenomenon? Or is it more strongly relevant in particular languages? In order to precisely delineate the scope of complement research, we need to precisely describe constructions from a larger number of languages. Furthermore, the diachronic evolution of these con- structions needs to be carefully investigated, both language-specifically and cross- linguistically. NOTES 1. This definition is also valid for other subordinate structures, such as adverbial constructions and relative clauses (Langacker 1991; Cristofaro 1998, 2003). 2. Other researchers argue that to’s future orientation only represents one instance of a more abstract meaning of the complementizer, namely, to code the conceptual distance that exists between the main and subordinate clauses (Smith and Escobedo 2001). 3. These two contrasts are obviously related. The speaker can reasonably be more confident in the outcome of a process he or she is experiencing than in some other conceptualizer’s experience. 4. In a compatible analysis, that-complements are characterized as belonging to the domain of discourse (de dicto), as opposed to the domain of reality (de re) in Frajzyngier and Jasperson (1991). 5. Other complementizers include tokoro, used to present a visually witnessed event (Horie 2000), tte, toka, and nante, which are argued to belong to the domain of discourse (de dicto) in Suzuki (2000). 6. This structural difference explains the glossing differences between no and koto on the one hand, and to on the other hand. 798 michel achard 7. In certain cases, the infinitive is preceded by the prepositions  aa ‘to’ or de ‘of’. The preposition obviously lends its own semantic import to the construction in which it occurs. Kemmer and Bat-Zeev Shyldkrot (1995) provide a semantic analysis of these prepositions. Achard (2002a) illustrates how their semantics contribute to the meaning of causative constructions. 8. The only case where the complement is an infinitive clause and the main and subordinate subjects are different is the perception/causation constructions illustrated in (19). These structures have their own individual history, and they are often consid- ered separately in the literature. For a cognitive account of those constructions, see Talmy (1975, 1976, 1983, 1988), Kemmer and Verhagen (1994), Achard (1998), and Stefanowitsch (2001). 9. The rules in (32) and (33) are presented in the spirit of Chomsky’s (1981) Gov- ernment and Binding theory. They are simply intended to capture the spirit of the gen- erative tradition, still very much alive in the current, more sophisticated models. 10. The difference in the semantic structure of raising and control verbs has received a fair amount of attention in the literature. Langacker (1995: 41) argues that a control predicate, such as persuade, designates ‘‘a complex direct interaction between the tra- jector and the landmark.’’ Conversely, the trajector of a raising predicate, such as expect, ‘‘does not directly interact with the landmark per se.’’ In a compatible analysis, I argue that the subjects of French control verbs have a conceptualizing role with respect to the scene profiled in the complement whereas the subjects of the raising verbs do not (Achard 1998, 2001). 11. For an in-depth discussion of the notion of reference point, see Langacker (1993). 12. Among the battery of structural tests designed to differentiate between the rais- ing and control verbs, the one based on en’s behavior represents in the words of Ruwet (1991: 56) ‘‘the most spectacular and most strictly syntactic.’’ REFERENCES Achard, Michel. 1996. Perspective and syntactic realization. Linguistics 34: 1159–98. Achard, Michel. 1998. Representation of cognitive structures: Syntax and semantics of French complements. Berlin: Mouton de Gruyter. Achard, Michel. 2000. The distribution of French raising constructions. Berkeley Linguistics Society 26: 1–15. Achard, Michel. 2001. The syntax of French raising verbs. In Alan Cienki, Barbara Luka, and Michael B. Smith, eds., Conceptual and discourse factors in linguistic structure 1–26. Stanford, CA: CSLI Publications. Achard, Michel. 2002a. Causation, constructions, and language ecology: An example from French. In Masayoshi Shibatani, ed., The grammar of causation and interpersonal manipulation 127–55. Amsterdam: John Benjamins. Achard, Michel. 2002b. The meaning and distribution of French mood inflections. In Frank Brisard, ed., Grounding: The epistemic footing of deixis and reference 197–249. Berlin: Mouton de Gruyter. Beck, David. 2000. Nominalization as complementation in Bella Coola and Lushootseed. In Kaoru Horie, ed., Complementation: Cognitive and functional perspectives 121–47. Amsterdam: John Benjamins. complementation 799 . event is chosen as the trajector of the main relation. Reality remains off profile, as part of the base. The raised variant of seem therefore profiles the apparent participation of an entity in a. 1990), the whole scene, including the subject of the subordinate process, is part of the objective scene and thus profiled. The presence of an infinitival complement reflects the subjective construal of. prominence and therefore marked as the subject. The event or proposition as a whole is the landmark of the main relation. In the raised variant in (29), the main participant in the located event