近畿大学学術情報リポジトリ

全文

(1)::st~. . ~1~r . ::st1~. 17~. 2 i}. 2006. 2. ON THE SOURCE-MARKING USE OF NI 'TO' IN JAPANESE Hiroyuki Takagi. 1. Introduction This article is concerned with the conceptual background that allows the Japanese goal-marking particle ni 'to' to mark an NP which is logically the source of the transfer designated by the verb of receiving. Naturally, the same situation can also be adequately described with the source marker kara 'from' for the same source NP. The following examples in (1) and (2) both describe the transfer of the book in the same direction (from Ziro to Taro) although the goal marker ni in (2) is used for Ziro, the source of the transfer of the book.. (1) Taro-wa. Taro-ToP. Ziro-kara. hon-o. moratta.. Ziro-sRc. book-ACC. got. 'Taro got a book from Ziro. (2) Taro-wa. Taro-ToP. Ziro-ni. ,. hon-o. moratta.. Ziro-GOAL book-Ace got. 'Taro got a book from Ziro.'. Here, (2) is readily interpreted as exhibiting the same logical meaning as (1) despite the apparent contradiction in the directionalities of the verb and the particle. The puzzle posed by the perfect acceptability of sentences like (2) is further underscored by the outright unacceptabil-78-. (1).

(2) ON THE SOURCE-MARKING USE OF NI. Takagi. ity of ni in (3), which differs from (2) only in that the verb. IS. katta. 'bought' instead of moratta 'got'.. (3) Taro-wa. Ziro-kara/ * ni. hon-o. katta.. Taro-ToP. Ziro-sRc/ GOAL. book-ACC. bought. 'Taro bought a book from Ziro.'. Although getting and buying both designate the transfer directed to the subject participant (i.e. from Ziro to Taro), the alternation between kara and ni is observed only when the former verb is used.. This phenomenon has been well noted in the literature, but the major proposals in the past seem to remain controversial in characterizing the nature of the alternation. In this paper, we will first examine two major lines of proposals and defend the one which maintains that the use of the goal-marker for the source NP in cases like (2) is in fact conceptually motivated. Next, we will further argue that a closer examination of the nature of the alternation requires a more elaborate characterization of the conceptual background in order to accommodate in a principled manner not only the representative examples but also some newly introduced data.. 2. The problems and possible solutions In this section, we will summarize the problems posed by the alternation, and examine two lines of approach to this problem.. (2). -77-.

(3) ::X:¥ . ~f,j\j . ::x:ft. 17~. 2 -ry-. 2006. 2. 2. 1. Problems. As partly noted in section 1, the kara/ni alternation exhibits the following characteristics.. (4) a. The particle ni 'to', which normally marks the goal NP, is adequately used to mark the NP referring to the source of the transfer, when used with certain verbs of receiving (i.e. verbs whose subject participant is the goal of the transfer), although the source marker in Japanese is normally kara 'from'. b. This alternation is not invariably available: there are verbs of receiving that do not allow for the alternation; there seem to be restrictions based on the choice of verbs. c. The alternation is limited to the source NP: there are cases where the goal marker ni adequately marks the source NP, but not vice versa; there is no case where the source marker kara adequately marks the goal NP.. In connection with (4b), verbs of receiving that allow for the alternation include morau 'get', kariru 'borrow', manabu 'learn', kiku 'hear' , and those that disallow the alternation include kau 'buy', uketoru 'receive', toru 'take', soozokusuru 'inherit'. The contrast is very. clearly felt; ni can never be used to mark the source NP with a verb of the latter class, while use of ni is perfectly acceptable in sentences using a verb of the former class. Classification of verbs into these two classes does not seem to be predictable in terms of such criteria as whether the transfer of possession is involved or whether the transfer is intentionally initiated. -76-. (3).

(4) ON THE SOURCE-MARKING USE OF NI. Takagi. As to (4c), the asymmetry is illustrated by (5).. (5) Taro-wa Taro-ToP. Ziro- nil * kara. hon-o. Ziro-SRc. book-ACC gave. yatta.. 'Taro gave Ziro a book.'. Here, unlike moratta 'got', yatta 'gave' designates a transfer of a book from the subject participant Taro, and the directionality it represents logically requires Ziro to be the goal of the transfer, hence it is to be marked by the goal marker ni. What is interesting here with respect to (4c) is that it is not possible under any conceivable circumstance for the goal NP Ziro to be marked by the source marker kara with its goal sense intact. This contrasts with examples like (2) above where the source NP is adequately marked by the goal marker ni, maintaining its source sense. We will not go into this problem in this paper, but interesting analyses of the superiority of ni over kara are found in Ikegami (1986, 1987) and Ihara and Fujita (2000).. 2. 2. Polysemous approach. One natural approach to the problem of the ka rain i alternation found in the literature is to have recourse to the polysemous nature of ni, postulating such a sense to this particle as the "secondary agent" (Ka-. bata and Rice 1997) or "agentive source" (Inoue 1976, Kumashiro 1993), and categorize this source-marking usage of ni into this "agentive source" sense. In this hypothesis, the alternation itself is no mystery at all because this ni is thought of as having the source sense. l Although this approach is not per se intended to "explain" the mechanism under(4). -75-.

(5) 'X.~. .. ~WJ. .. Jt1~. 17~. 2 -j}. 2006. 2. lying the alternation, it is in a sense a natural way to think about the use of ni in question if we notice the highly polysemous nature of this particle (i.e. location, goal, causee, time, etc.), and to capture this use of ni as one of the variants in this broad network of different but mutually related senses is certainly theoretically meaningful. Notice here that in this approach the postulated sense is not simply the "source"; it is also considered to be "agentive". This assumption is based on the observation that the alternation does not take place when the source NP is inanimate, as illustrated in the contrast between (6) and (7).2. (6) Taro-wa Taro-ToP. Ziro-kara/ni. tegami-o. Ziro-sRc/ GOAL letter-ACC. moratta. got. 'Taro got a book from Ziro.' (7) Taro-wa. gakko-kara/*ni. tegami-o. moratta. 3 (d. Kumashiro 1993). Taro-ToP school-sRcl GOAL letter-Acc got 'Taro got a book from his school.'. As we have seen, the verb maratta 'got' allows for the alternation as seen in (6). However, the alternation seems to be blocked when the source NP is inanimate as in (7), and this observation leads to the assumption that the source sense of ni has to be "agentive" (Kumashiro 1993) . We maintain, however, that to attribute the ka rain i alternation merely to the proposed "agentive source" sense misses the true significance of this phenomenon. First, it is simply a classification, and does not attempt to answer the question in which we are interested: the -74-. (5).

(6) ON THE SOURCE-MARKING USE OF NI. Takagi. mechanism that allows ni to appear to have such a source sense. The possibility that ni has indeed acquired the "agentive source" sense should not be ruled out, but even if this is the case, we will need to know what motivates it to have this separate sense. Secondly, in the empirical domain, it does not shed any light on the question as to why it is that not all verbs of receiving allow for the alternation. As mentioned above, some verbs of receiving refuse their source NP to be marked by ni and some allow it perfectly, and how to explain this distinction based. on how qualified the source NP of each verb is as the supposed "agentive source" is not apparently obvious. It indeed seems to be very difficult to show in a principled manner why the source participant of verbs like morau 'get' , kariru 'borrow' or kiku 'hear' is well qualified as one of the supposed "agentive source" while the source participant of verbs like kau 'buy' and uketoru 'receive' is to be unqualified as such. 4 (This notwithstanding, there are cases which seem to demonstrate that "agentivity" of the source participant seems be relevant in acceptability of ni as in the contrast between (6) and (7) above. We will see in 4.3 that this is only an epiphenomenon contingent to the framework we will present.). 2. 3. Source-as-the-goal view. For the reasons given in the previous subsection, we do not take the approach that postulates an independent sense to ni designed to accommodate the alternation. Instead, we will take a different tack and base our hypothesis on the conceptually motivated assumptions. We would like to show that ni in its original goal-marking sense is qualified to mark the "source NP" as it is. One attempt to pursue this line of approach to the karalni alternation is seen in Ikegami (1986); according to (6). -73-.

(7) ::st"f: . ~f,!\f. .. ::st1~. 17~. 2 -ry-. 2006. 2. him, in sentences like (1) and (2), repeated here as (8) below, the receiver of the book (i.e. the subject participant) is in a state of being obliged to the giver for the book given to him, and this asymmetric psy-. chological state which is directed from the receiver to the giver of the book motivates the giver, the source of the physical transfer of the book, to be marked by ni, the goal marking particle.. (8) Taro-wa Taro-ToP. Ziro-kara/ni. hon-o. moratta.. Ziro-sRc/ GOAL. book-ACC. got. 'Taro received a book from Ziro.'. Ikegami observes that this assumption serves to explain why the possibility of the karalni alternation is not available to all verbs that take the "goal" subject.. (9) Taro-wa. Ziro-kara/ * ni. hon-o. uketotta.. Taro-ToP. Ziro-sRc/ GOAL. book-ACC. received. 'Taro received a book from Ziro.' (10) Taro-wa. Ziro-kara/ * ni. hon-o. totta.. Taro-ToP. Ziro-sRc/ GOAL. book-ACC. took. 'Taro took a book from Ziro.'. In (9), uketotta 'received', unlike moratta 'got', only shows a change of the locus of the book, and hence Taro has no reason to be "obliged" to Ziro, nor in (10) is Taro obliged to Ziro because he "took" the book from Ziro. Note here that the anomaly of ni in (10) can also be captured by the assumption that postulates the "agentive -72-. (7).

(8) ON THE SOURCE-MARKING USE OF NI. Takagi. source" sense to ni; Ziro in (10) cannot be understood as "agentive" because his book was "taken" by Taro and hence the implication is that he did not consent to let go of the book. But (9) is not explained by this assumption because here the giver of the book, Ziro, may well be regarded as agentive in regard to the transfer of the book, but use of ni in this sentence is totally unacceptable. This line of explanation also seems to be able to explain the unacceptability of ni when used with the verb katta 'bought' as we saw in (3), repeated here as (11).. (11) Taro-wa. Ziro-karal * ni. hon-o. katta.. Taro-TOP. Ziro-sRcl GOAL. book-ACC bought. 'Taro bought a book from Ziro.'. When one buys something from somebody, one is not obliged to the seller because the implied payment of money is supposed to cancel out the feeling of obligation, in contrast to the situation shown by moratta 'got' in which there is nothing that is supposed to exhibit this. canceling function. The polysemous view does not seem to be able to handle this case either because the seller may well be regarded as an "agentive source". So far, following Ikegami's view, our working hypothesis has been that ni 'to' is indeed the goal marker even in cases where it appears to mark the source participant. In this view, the source participant is actually conceptualized as the goal as well to the effect that it codes the goal of the asymmetric relation of the receiver's "being obliged to the giver" that runs counter to the transfer designated by the verb. We have seen (8). -71-.

(9) X,,¥,-·. -iH~1f. . x1t. 17~. 2 i7. 2006. 2. that his approach provides not only a wider empirical coverage but also a motivated account of the seemingly contradictory use of ni, which could not be expected if we simply supposed that the particle has the source sense as part of its inherent semantics. We would like to take this view as our point of departure; however, we will show in the next section that in order for a wider range of data to be accommodated, we need to add some major elaboration to this approach.. 3. Decomposing the receiving event Before we present our own proposals in section 4, we need to clarify our understanding of the conceptualization of the event of receiving. As widely practiced in the cognitive linguistic literature, meaning of a linguistic expression is the profiled subpart of the base conception; thus when one speaks of the meaning of roof one has to refer to the base conception of a house on which the particular subpart is designated by this nominal expression. When one is concerned with the meaning of a verbal expression, then, our main task is to identify what kind of event conception is evoked as the base and which portion of it is being profiled by the expression under investigation. Employing the basic spirit shown in the model of the control cycle proposed by Langacker (2002), we would like to assume that the event of "receiving" is conceptualized as being composed of three phases: potential, action, and result, sequenced in this order along the conceived. time. What is significant about this model is its attempt to capture the event structure from the perspective of the "receiver" or the "goal" of the transfer, in contrast with many other models in cognitive linguistics -70-. (9).

(10) ON THE SOURCE-MARKING USE OF NI. Takagi. (e.g. "action chain" (Langacker 1991) or "causal chain" (Croft 1991)) that are designed to model the event structure from the perspective of a participant located upstream in the flow of the event. In abstract form of the control cycle model, the potential phase is a situation of "tension" on the part of the receiver which is caused by the emergence of the prospect of receiving the target entity. In the action phase, the receiver captures the target entity and incorporates it into his/her domain of control (i.e. dominion), and, in the static result phase, the target entity is part of the receiver's dominion. This cycle is thought of as part of our embodied experience. Now, which portion of the invoked phases is profiled is variable; taking Langacker's (2002: 195) example, the profile of catch in (12a) covers only the action and result phases and precludes the potential phase, as shown by the fact that the nature of the potential phase (i.e. the catcher's intention) is separately coded, but the same verb in (12b) covers all the three phases as indicated by the semantic range of the verbal modifiers.. (12) a. The center fielder intended to catch the ball. b. The cat caught the mouse through patience, alertness, silent stalking, and a lightning-quick lunge. (13). Potential. I >. Action. I b.. >. Result. I catch in I catch in. I. (12a) (12b). We would like to assume that in the kind of the transfer event under investigation the control cycle incorporates the "social" and "per( 10 ). -69-.

(11) Y.."'f. . ~f,liJ . Y..1t. 17~. 2~. 2006. 2. ceptive" components whereby interactions between the participants in these domains are reflected in the model (d. Langacker 2002: 197). Given this assumption, we contend that what is included in the potential phase is the receiver's "approach" to the potential giver (i.e. source participant), effected in such a way that the intended transfer of the target entity is initiated as the result of this approach. By the nature of the event, this approach to the potential giver could be one of requesting, encouraging, ordering or beseeching; but whatever it may be, it is understood as an intentional action on the part of the receiver of the kind to be characterized in force-dynamic terms (Talmy 1988), which is directed to the potential giver, hence is an asymmetric relation [G ---. S] (henceforth, G and S stand for the goal and the source participants of the transfer event, respectively). What happens in the action phase is naturally the actual transfer of the target entity from the dominion of the source participant to that of the goal participant, [G +- S]. Finally, in the result phase the target entity resides in the goal participant's dominion and it is in the social/perceptive components of this result phase that the goal participant is in the state of being obliged to the source participant for the transfer of the target entity (d. Ikegami 1986), the vector of which runs from the goal to the source, [G ---. S]. We will see in the next section how this framework can explain the facts we observed and the data we will add.. 4. Proposals Our working hypothesis is that instances of ni that mark the "source NP" have a corresponding conceptual underpinning to the ef-68-. ( 11 ).

(12) ON THE SOURCE-MARKING USE OF NI. Takagi. fect that the source NP is in fact qualified as the goal as well. We have further proposed that the event of receiving should be characterized sequentially based on the event schema along the line of the control cycle. Given these assumptions, we will set out to propose our analysis of the karalni alternation.. 4. 1. Acceptability of source-marking ni. In the fully inclusive form of the model, there appear to be two possible evocations of the [G - S] structure that motivates use of ni for the NP that codes S, one in the potential phase and the other in the result phase, while the [G - S] structure, which motivates use of kara, appears only in the action phase. If this much is on the right track, we are now in the position to state the condition which predicts the acceptability of ni used for the "source" NP.. (14) a. If the profiled portion of a verb of receiving includes the [GS] relations in both the potential and the result phases, use of ni for the source NP is fully motivated, hence is acceptable.. b. If the profiled portion of a verb of receiving includes only one of the [G - S] relations in the potential or the result phase, use of ni for the source NP is partly motivated, hence is awkward. c. If the profiled portion of a verb of receiving does not contain the [G - S] relation in the potential and the result phase, use of ni for the source NP is not motivated, hence is unacceptable.. First, let us show that this framework can accommodate the at( 12 ). -67-.

(13) x'#- .. ~1~f. . x1~. 17~. 2~. 2006. 2. tested data. Of the verbs that disallow the alternation, uketoru 'receive' profiles only the action phase of the cycle (ct. Ikegami 1986: 12), hence use of ni is unmotivated and therefore unacceptable because no [G --+ S] structure is included in the profile of the verb. A similar line of explanation may apply to kau 'buy'; the event of kau is a mechanical process of exchanging money and goods, not a process made possible by having effected the buyer's approach to the seller, nor is followed by the buyer's feeling obliged to the seller. Soozokusuru 'inherit' is also explicable in a similar fashion. The event of inheriting something is carried out mechanically because it is an event that takes place as legally stipulated or as decided in the will of the deceased. In short, these verbs disallow S to be marked by ni because of (14c): their semantic profiles do not include the [G. --+. S] relation at all, hence S's being marked by ni lacks. conceptual motivation.. 4. 2. Partially motivated ni. Now, a closer look at ni when used with verbs that allow for the alternation reveals an interesting behavioral characteristic. Observe the contrast between (15a) and (15b).. (15) a. Kinoo Ziro-kara/ni kono hon-o moratta yo. Yesterday I got a book from Ziro.' b. Kinoo. totuzen. Ziro-kara/?ni. kono hon-o. yesterday suddenly Ziro-sRc/ GOAL this. moratta yo.. book-ACC got. 'I suddenly got a present from Ziro yesterday.' Many speakers find. nl III. (15b) awkward, and some find it very -66-. ( 13 ).

(14) ON THE SOURCE-MARKING USE OF NI. Takagi. much so, not to say unacceptable, while kara was unanimously judged perfectly acceptable in this sentence. What effect does totuzen 'suddenly' have on the acceptability status of ni in (15b)? Observe that (16b), (17b) and (18b), where adverbials like totuzen and siranaiuchini 'without knowing' are added, exhibit a similar decline in acceptability of ni.. (16) a . Kinoo. Ziro-kara/ ni. yesterday Ziro-sRc/ GOAL. sono hanasi-o. kiita yo.. that. heard. story-Acc. 'I heard that story from Ziro yesterday.' b. Kinoo. totuzen. sono hanasi-o. kiita yo.. that. heard. Ziro-kara/??ni. yesterday suddently Ziro-SCR/ GOAL. story-ACC. 'All of a sudden 1 heard the story from Ziro yesterday.' (17) a. Sensei-kara/ni. ooku-o. manabi-masita.. Teacher-sRc/ GOAL manY-ACC. leaned. 'I learned a lot from my teacher.' b. Siranaiuchini Without knowing. sensei-kara/??ni. ooku-o. manbi-masita.. teacher-sRc/ GOAL. manY-ACC learned. 'Without knowing, 1 learned a lot from my teacher.' (18) a. Ziro-kara/ni Ziro-sRc/ GOAL. kono hon-o. kari-te kimasita.. that book-ACC. borrow came. 'I borrowed that book from Ziro.' b. Ziro-kara/??ni Ziro-SRC/ GOAL. kossori. kono hon-o. secretly that book-ACC. kari-te kimasita. borrow came. 'I secretly borrowed this book from Ziro.'. These data obviously cannot be accommodated by Ikegami's ap-. ( 14 ). -65-.

(15) x~. . ~1;Iif • xi!:. 17~. 2 % 2006. 2. proach, because whether the event takes place suddenly, without knowing or secretly is largely irrelevant to whether the goal participant is to feel obliged to the source participant, and thus is not expected to affect the acceptability of ni for the source NP. We argue, however, that these data are exactly what our framework predicts. Adverbials like totuzen 'suddenly' siranaiuchini 'without knowing' and kossori 'secretly' preclude the possibility of the [G - S] conceptualization prior to the action phase because they indicate that the event denoted by the verb happens without G's prior involvement with S. On the other hand, the possibility of the [G - S] relation in the result phase is undisturbed. Therefore, if (14b) is correct, use of ni is only partially motivated by the instance of [G - S] relation in the. result phase, hence degraded acceptability results, in contrast to the (a) examples above where use of ni is fully motivated. The awkwardness of ni in the (b) examples of (15) - (18) is clearly felt by the majority of native speakers I have consulted, but there also seem to be a certain proportion of speakers who find these sentences totally acceptable. One possible reason for this split judgment might be that in the grammar of the latter group of speakers the source sense is indeed a part of the established lexical meaning of ni, and for the majority who find these examples awkward the employment of ni for S is contingent to the conceptualization of the [G -S] relation. If this reasoning is on the right track, we are dealing with a case in which a lexical item is right in the middle of the process of incorporating a new sense into its semantics.. -64-. ( 15 ).

(16) ON THE SOURCE-MARKING USE OF NI. Takagi. 4. 3. Apparent agentivity effect. In section 2, we compared what we call the polysemous approach and source-as-the-goal approach to the usage of ni in question and argued in favor of the latter, along the lines of which our own framework has been advanced. Recall that the evidence which the polysemous approach provides includes the fact that in sentences like (20), where the source NP is inanimate, ni is rendered unacceptable, although it is perfectly acceptable when the source NP is qualified as an agent, as in (19).. (19) Taro-wa Taro-ToP. Ziro-karal ni. tegami-o. Ziro-sRcl GOAL letter-Acc. moratta.. (= (6)). got. 'Taro got a book from Ziro.' (20) Taro-wa. gakko-karal * ni. tegami-o moratta.. Taro-ToP. school-sRcl GOAL. letter-Acc got. (= (7)). 'Taro got a book from his school.'. We contend that our approach not only correctly predicts that ni in (20) is unacceptable but also reveals that the apparent "agentivity effect" is only an epiphenomenon of the mechanism which we are proposing. Why ni in (20) is predicted to be unacceptable in our framework must be straightforward by now; in our normal understanding of the world a letter from one's own school does not come as the result of one's prior approach to the school (i.e. no [G --. S] in the potential phase) and we do not feel obliged to the school for the letter (i.e. no [G --. S] in the result phase). Thus according to (14c) use of ni for S is not motivated at all. One way to verify this reasoning may be to add some more contex( 16 ). -63-.

(17) tual information that helps the [G. -+. S] conceptualization to arise, and. to see how ni is going to be judged with it. For many speakers, use of ni in (21) and (22) are significantly better than that in (20).. (21) Taro-wa. yatto. Taro-ToP at last. gakko-kara/?ni. sono tegami-o. school-sRcl GOAL that. letter-Acc. moratta yo. got. 'Taro got that letter from his school at last.' (22) Taro-wa Taro-ToP. gakko-kara/?ni. kon-na ureSll tegami-o. school-sRcl GOAL. such. moratta yo.. happy letter-Acc got. 'Taro got such a happy letter from his school.'. In (21), yalta 'at last' implies Taro's approach to his school in order to get the letter, and therefore the [G -+ S] relation in the potential phase is strongly suggested. 5 In (22), the natural interpretation of the situation is that the nature of the letter makes Taro feel obliged to his school, which gives rise to the [G. -+. S] relation in the result phase.. Therefore, since (21) and (22) evokes the conceptualization of only one of the two potential [G. -+. S] relations, use of ni is predicted to be. partially motivated, hence not unacceptable, according to (14b). We are now in the position to provide our account of why "agentivity" of the source NP appears to be relevant to the acceptability of the source-marking use of ni as most typically shown in examples like (20): in our normal understanding of the world, it is not usual to conceptualize the [G. -+. S] relation, either in the potential or in the result phase,. if S is not qualified as an agentive being, when it is a location or an inanimate entity for example. Simply put, we normally do not request an inanimate entity to transfer something to us, nor do we feel obliged to -62-. ( 17 ).

(18) ON THE SOURCE-MARKING USE OF NI. Takagi. an inanimate entity for the transfer because we know that the transfer was not made to favor us but was done simply in a mechanical fashion. This is why use of ni is rendered reasonably acceptable if we set up a context in which G is likely to feel obliged to a non-agent S, though not perfect for only one [G - S] relation is evoked.. (23) Jinrui-wa Mankind-TOP. taiyo-kara/?ni. enerugu-o. moratte-imasu.. sun-sRc/ GOAL. energy-ACC. getting-is. 'Mankind is getting energy from the sun.'. Although we have shown that the apparent agentivity effect is only an epiphenomenon, the agent sense of ni as used in the passive sentence should be considered as an independent lexical entry of the particle. This is perhaps best shown by the fact that the distributional restriction imposed on the former is not necessarily observed in the latter. Thus, although ni. is rejected for the source of the transfer in a purchase-and-sale transaction as in (24), it is perfectly acceptable when used as the agent marker in a passive sentence that shows the same event type, as seen in (25).. (24) Taro-wa Ziro- *ni hon-o katta. 'Taro bought a book from Ziro.' (25) Taro-wa Taro-ToP. Ziro-ni. hon-o. ur-are-ta.. ZirO-AGT book-ACC sell-PAss-PAST. 'Taro was sold a book by Ziro.'. How is this agent sense of ni connected to other senses? To this ( 18 ). -61-.

(19) )(l!j: . ~1f.J. . )(1t. 17~. 2 -lj}. 2006. 2. question, we would only like to suggest the possibility that the apparent agentivity effect contingent to the source-marking usage plays a mediating role between the goal sense and the agent sense. If this is the case, the semantic extension might as well be one of incorporation of the contingently inferred agentivity into the lexical meaning of the particle. We will leave this interesting problem for the future research. 6. 5. Conclusion In this paper, we have argued that use of ni, the goal marker, for the source argument of certain verbs of receiving should be understood as coding the relations in the conceptual structure which are directed from the goal to the source participant of the transfer. Ikegami (1986) characterized this relation as the feeling of obligation to the source participant. We contended that the understanding of the relations that motivate this use of ni should be further elaborated; adopting Langacker's (2002) control cycle model, we proposed that acceptability of ni for the source NP is determined by how the two possible [G -- S] relations in the potential and result phases of the control cycle are evoked. It was shown that when none is evoked ni is not acceptable, when only one is evoked use of ni is marginally acceptable, and when the two [G -- S] relations are available use of ni is fully acceptable. Notes. 1. Kabata and Rice's concept of "secondary agent" is supposed to cover the causee and the passive agent both typically marked with ni. 2. Adding an agentive role to the sense of this category of ni seems to provide a basis for understanding it in connection with the category of ni that marks the agent. -60-. ( 19 ).

(20) ON THE SOURCE-MARKING USE OF NI. 3.. 4.. S.. 6.. Takagi. in the passive sentence. I do not touch upon the question as to where this sense of ni is to be adequately placed among other senses except for a cursory suggestion at the end of section 4. I would like to emphasize in this paper that the alternation in question can be, and in fact should be, explained without recourse to supposing a separate sense that applies only to this use of the particle. Kumashiro's original sentence uses the nominal case marker ga instead of the topic marker wa. This change has been made because wa makes the sentence more natural and the change does not have any effect on the problem in which we are interested. There are verbs like soozokusuru 'inherit' and nusumu 'steal' which refuse the use of ni in question and clearly do not assign the agentive sense to its source participant. But not all verbs of this class are explained by this, as was observed in the text. In fact, for many speakers, use of ni in (21) is even perfectly acceptable. This is also a reasonable outcome of our approach because now that Taro has got a letter which he had been requesting to get, he is naturally believed to feel obliged to his school, and hence the conceptualization of the [G - S] relation in the result phase is also contextually facilitated. Our suggestion about the semantic network is different from Kumashiro's (1993) conclusion. He proposes that the goal sense is extended to the agent sense (by way of the causee sense), and the agentive source sense is a further extension from the agentive sense.. References. Croft, William (1991) Syntactic categories and grammatical relations. Chicago, University of Chicago Press. Ihara, Hiroko and Ikuyo Fujita (2000) "A cognitive approach to errors in case marking in Japanese agrammatism: the priority of the goal -ni over the source -kara," Constructions in cognitive linguistics, Ad Foolen and Frederike va der Leek (eds.) , John Benjamins, Amsterdam. Ikegami, Yoshihiko (1986) "The goal as agent," Sophia Linguistica Working Papers in Linguistics 20/21, 7-14. Ikegami, Yoshihiko (1987) "'Source vs. 'goal': a case of linguistic dissymmetry," Concepts of case, Rene Dirven and Gunter Radden (eds.), 122-146, Guter Narr Verlag, Tuingen. Inoue, Kazuko (1976) Seiseibunpo to Nihongo (Generative grammar and Japanese), Taisyukan, Tokyo. Kabata, Kaori and Sally Rice (1997) "Japanese ni: The particulars of a somewhat contradictory particle," Lexical and syntactical constructions and the construction of meaning, Marjolijn Verspoor, Keedong Lee and Eve Sweetser (eds.) , 107-127,. (20 ). -59-.

(21) xl'f.. '~1iIlJ. . xft. 17~ 2 -j}. 2006. 2. John Benjamins, Amsterdam. Kumashiro, Toshiyuki (1993) "How the goal can be the source: The semantics of the Japanese dative marker ni," LACUS 20,401-417. Langacker, Ronald W. (1991) Foundations of cognitive grammar, vo1.2: Descriptive applications, Stanford University Press, Stanford. Langacker, Ronald W. (2002) "The control cycle: why grammar is a matter of life and death," Proceedings of the Second Annual Meeting of the Japanese Cognitive Linguistics Association, 193-220.. -58-. ( 21 ).

(22)