Semantic Roles in Teaching Japanese Verbal Constructions

EAJS Conference 2008 (Lecce, Italy)

Semantic Roles in Teaching Japanese Verbal Constructions

Simone Dalla Chiesa

The University of Milano

As a Researcher of the University of Milano, I teachJapanese grammar to all five grades of the undergraduate and graduate programs in Cultural and Linguistic Mediation. My assignment consists in 40 hours a year per grade, while the remaining 80 yearly hours are taught by native speakers. The textbook we use in the undergraduate course is ICU's New Japanese for College Students (1996).

In teaching Japanese verbs and their constructions, beginning with watasi wa nihonzin dewa arimasen, I greatly evaluate the logical and heuristic power of the notion of case. As a signal of the relationship of nouns to their (verbal) heads (from Blake 2001: 1), surface case represents, as I love to say, the embodiment or morphosyntactic epiphany of the verb within the surface structure of a sentence. At the same time, however, I also feel the strong need of anchoring case assignment to a deeper substratum of meaning. This is why I also introduce my students to the notion of semantic roles. Identifying the roles of the entities syntactically realized by the phrases of a sentence makes it possibleto reveal those meaning relations that in most situations determine surface case marking, but and also to connect predicates to the events they express, and thus reach the very dimension at which verb classes are generated.

It is precisely to show the relation of verbs, cases and roles, and how semantic roles often function as a bridge between the ontology of events and the morphosyntactic marking of nominals,that I developed the graphic representations I will treatin this paper: the "balloon graph" and a two-dimension table. I use these graphic devices from "day zero", my very first lecture to the first grades, as they are simple and self explanatory enough not to require any complex theoretical framing. In my paper I will therefore touch several issues concerning argument realization, and (in the Conclusion) will also briefly consider thetheoretical implications of my approach, but will not extensively discuss linguistic theory or deal in detail with the notions of semantic role and case.

While I developed both graphs by myself, I am not the original inventor of the balloon graph. Ultimately, I owe my interest in verbal constructions and my teaching techniques to Professor Teramura Hideo and to TsukubaUniversity, where I studied in the middle of the 1980s. I was never taught by Teramura himself, but I absorbed his interests, approach and methods through my teachers, his monograph Nihongo no sintakusu to imi (1982) and his textbooks An Introduction to the Structure of Japanese (1973) and, later,Nihongo zyookyuu bunpoo kyoohon (1988). However, despite the overt influence of Fillmore (1968) case theory, Teramura's textbooks classify predicates only on the basis of their superficial case structures, without mentioningdeep cases or thematic roles. Neither they make use of any graphic aid. Students were offered some simple graphic representation, like the primitive balloon graph that follows, only on the classroom's blackboard.

Graph 1: okuru

Despite its crude form, Graph 1 has several merits. For instance, by observing the number and surface marking of the argument NPs displayed in the upper ovals, one can easily deduce the existence of a ditransitive verb class to which okuru belongs. The question words in the upper balloons also helps identifythe kind of entities participating in the event, so that this graph might be said to somehow link semantics and syntax and crudely representargument structure.Moreover, if it is read as English text, left-to-right and up-to-down, the whole balloon sequence captures the "ordinary" word order of a Japanese sentence.

The use of graphs of this kind is quite diffused in contemporary language teaching. The following is a similar but more sophisticated graph taken from an Italian language teaching manual (D'Alfonso 2001: 13-16). The verb is the same, but I changed the language to English.

Graph 2: John send flowers to Mary.

Drawing the arguments under the verbeffectively hints at the fact that the predicate can be seen as the dominating feature of a sentence, in that it controls the number and marking of its arguments.Predicate and particles are boxed exactly in order to distinguish their shared "harder" nature from that of the nominals, which, as fillers, are displayed in "softer" oval balloons. Also, while the argument NPs marked by a direct case are directly linked to the predicate box, the indirect object "Mary" connects to the predicate via an intermediate, smaller box attached under the belly of the verb, which contains the particle to. I think the logic of this is quite obvious. Finally, the double line linking the leftmost balloon to the predicate has the aim to single out that phrase as the subject. Since in Italian schools such a configurational approach is the norm, I decided to implement the double line in my graphs as well.

In its complete form,Graph 2 would appear as follows.

Graph 3: English dative verb class, complete representation

Starfish-like Graph 3 shows the complete set of phrases composing the sentence of a dative verb, but could be used to represent the construction of any verb class. Its main feature is that it does capture a hierarchy of constituents, with the most important ones for the verbal prototype displayed, alas, at the bottom. The dashedballoons sitting on the two sides of the predicate represent a second "tier" of phrases, bound to local functions, whose instantiation is required by certain classes of predicates. The upper ovals contain the other, least important adjuncts. The actual construction of a given predicate class is obtained from a matrix of this kind by pruning off the unneeded arms. But there is simply no room on the page or on the blackboard to implement such a process, the graph is too large and complicated, and, at least for English, flawed by its impossibility of representing the double object alternation.

In its simple form, the above kind of graph could represent the construction of a dative Japanese verb such as osieru as follows.

Graph 4: Balloon graph of dative verb osieru, with boxed particles

I dislike Graph 4 because, with the exception of certain adverbs, in Japanese all arguments and adjuncts are realized with the "mediation" of a particle, and connecting their balloons to the verb via a "particle box" becomes superfluous.I prefer the "full balloon approach" of Graph 5.

Graph 5: Simple balloon graph of dative verb osieru

Displaying in ovals both predicate and NPs makes it easier to understand that particles and verbs are actually a whole thing, albeit of a different lexical nature. Moreover, circles are easier to trace on the blackboard and offer more line-connecting points when drawn by Microsoft Word. Finally, writing case particles within each balloon is more elegant and simple than boxing them, and better represents the logical relation that links themto the predicate.

Graph 5 can be expanded into Graph 6 in order to display adjuncts, cases and roles.

Graph 6: Balloon graph of dative osieru, with adjuncts, case and role names

This is the final form of my balloon graphs. Each balloon represents a single phrase, and displays a case name, its marker, and the label of the semantic role thus realized. Adjuncts are drawn on the extreme right, well separated from the balloons of the argument NPs. All possible adjuncts could be included in the graph, if space and time allow it, but usually that is not necessary, as students soon learn that most common adjuncts can be added to any kind of sentence.

Before proceeding further, I need to remark the fact that my balloon graphs are not well suited for representing the word order of a Japanese sentence. If faithful to the syntax, a balloon representation of a verb's sentence would appear as nothing more than a mere circling or boxing of the constituents.

Graph 7: Japanese syntax, three-argument sentence

As representations in two dimensions of a melody of phrases, balloon graphshave the purpose to reveal those logical and semantic connections that do not easily show in one dimension. They are not suitable for representing constituents' linear order. Trying to do this would make the graphs too complicatedand decrease their heuristic power. This is why I decided of not concerning myself with word order.

This is not necessary with Chinese, though. The following are the two "insect-like" balloon graphs of the alternate constructions of ditransitive predicate教jiāo.

Graph 8: Double-object construction of Chinese dative verb jiāo

(1)他教小孩子英文.

Tā jiāo xiǎoháizi yīngwén.

Graph 8represents the two-object construction of jiāo. Constituents are read left to right, following the natural word order of the sentence.

Graph 9: Alternate construction of Chinese "give" verb jiāo

(1)他给小孩子教英文.

Tā gěi xiǎoháizi jiāo yīngwén.

Graph 9shows instead the dative/benefactive alternation of jiāo, characterized bythe presence of an indirect dative phrasebetween subject and predicate, in the position required of concrete cases. I make no further comment on this chart, but just notice the boxing of the preposition给gěi.

I am introducing next some powerful applications of the balloon graph, starting with the construction of the change-of-state verb naru.

Graph 10: Balloon graph of naru

Stacked balloons represent phrases that cannot cooccur. I usually link these balloons with a vertical line. In Graph 10, the ovals stacked on the right display the realization of different lexical categories, from nouns in the two uppermost positions (the "default" category, therefore let unspecified) to an embedded sentence F' at the bottom. The e case marker is not very common with naru, and students hardly meet it in classes. I will explain later why I added it. I draw this graph when treating Unit 8 of our ICU textbook, after about 16 weeks of classes. At this early stage, students still have little knowledge of verbs' plain forms, thus I usually leave the bottom balloon empty.

Graph 10is most useful for showing case assignment and role realization, but it also helps explainthat the adverbial particle ku of i- adjectives functions as a translative case marker.[1]Moreover, it illustrates how the sentence structure of naru (and more generally of Japanese achievement verbs of change of state) is matching with that of the motion verbs of change of place.

Graph 11: Balloon graph of iku

Graphs 10 and 11 have the purpose of showing that naru/kawaruand ikuactually belong to a common "translative" prototype of "moving between points": physical places in the case of iku, abstract locations or conditions in the case ofnaruorkawaru. In order to capture this relation I use the following devices: (1) I display the eNP of naru's construction so to have a closer match between the constructions of naru and iku; (2) I assign the denomination of translative to the oblique case of both the abstract goal phrase of naruand the spatial goal phrase of iku(rather thandirectional or Martin's (1975: 41) mutative and mutative locativerespectively); (3) I use semantic role labels to point out that the parts played by the entities participating in the motionevents described by naru and iku are very similar: an entity moves from a physical or an abstract location, the source, to another location of the same kind, the goal. This moving entity is the patient, but it moves spontaneously in naruand volitionally in iku, hence its additional agent role with the latter verb.[2]I thus suggest that the matching between naru/kawaru's and iku's surface case structures originates in a corresponding "deeper" match between thebehaviors exhibited by the participants in the two event types.

Before proceeding further, I believe it necessary to define now the most important semantic roles I am mentioning in this paper. The following is my own, customized role list, which I composed by eliminating many of those too fine-grained or didactically unnecessary roles that usually find a place into more detailed treatments of this topic. It contains all the roles I use in my graphs and throughout this paper.

Graph 12: List of semantic roles

agent / The entity initiating an event. I conflate in the agent roleboth the animate entities acting volitionally and the inanimate entities initiating events as mechanical or natural causes. If necessary, the latter can be labelled as force or as stimulus (if causing an event of perception).
beneficiary / The entity advantaged or disadvantaged by an event. So labelled to distinguish it from the benefactive case. When thebeneficiary and thegoal (dative) roles are borne by a same entity, its NP is marked by a dative and the two roles cannot be syntactically or semantically distinguished. A distinct benefactive marking appears only when the benefactive and goal roles are borne by different entities.
experiencer / The animate entity receiving a sensory input but not undergoing a change of state or acting volitionally.
goal / The physical or abstract location at which a patient locatum ends its movement or toward which it is directed. Following a weak form of localist approach, I usually consider the role of goal to alsoencompassthe final state of a change-of-state event (factitive role) and the "animate goal in an event of physical or abstract transfer of possession" (recipient) (Levin and Rappaport 2005: 159), also defined as a consciously affected goal (dative).
instrument / The entity used by an agent to carry out an action. An effector or immediate cause.
manner / The entity representing the condition under which the process of a dynamic event is carried out or a stative event takes place.
path / The physical or abstract space through which a patient locatum moves. This is the standard definition; sometimes all spatial dynamic roles such as source, goaland path proper could be usefully conflated into a generalized path role.
patient / The undergoer of a change of physical or abstract state. I consider to be a patient the entity representing the resultative object of accomplishment verbs of creation, sometimes associated with the role of result.
patient
locatum / The undergoer of a change of place. I keep this role distinct from the more general patient role in order to distinguish between the two different kinds of affectedness of patients in a two-patientevent, in which either patient is a p.locatum.
place / The physical location at which a dynamic or stative event takes place. Often location; here place to distinguish it from the grammatical case.
source / The physical or abstract location at which a p.locatum initiates its movement. It includes the giver of dative verbs.
theme / The entity located in a place or condition but not acting or undergoing a change of state. This narrow definition is only possible because I gave a broad definition of agent, experiencer and patient.

I am introducing now a series of constructions that the balloon graphs manage to illustrate quite powerfully. The first one is the -te ageru benefactive construction.

Graph 13: Balloon graph of benefactivepredicateof giving ageru

The dashed lines connecting two balloons indicate that the markersthere displayed are typically used jointly.The first two horizontally joined balloons show how arguments are realized when the predicate refers to the transfer of possession of a physical object between two sentient entities. I represent this event as follows. The entity initiating the dative event bears both agent and sourceroles, as it acts volitionally and is also the location where the transferred object begins its movement. This object bears the role of p.locatum, since its change of possessorentailsa movement between two locations. The third participant is the entity where the p.locatum ends its movement, and I assign to it the role of goal. However, agerualso entails that this entity beneficiates from the transfer. To indicate this I assign it thebeneficiary role.Finally, since ageru implies that this third participant acts volitionally in receiving the object (this point is important because it makes all the difference between the verbs of giving on the one hand and the verbs of sending and putting on the other, and also helps to explain the conversivity and case structures of the predicates of giving and receiving.) I assign to it the role of a secondary agent – secondary because this entity is not the initiator of the event but her willied participation is crucial to its successful completion.

However, the object does not need to be an affected physical entity, but can be an act that the agentive source effects in behalf of the beneficiary. As a performance, this object must be expressed by a verb, and there ought tobe a way of substituting the accusative NP with a verb or a sentence. The balloon stacked under the o NP balloon shows precisely that the –te form is the morphologic device used to assign a clausal argument to ageru. I consider this argument to be associated with the role of patient rather than p.locatum because it does not represent a transferred physical object.

Our textbook also makes an extensive use of tameni phrases, which are not easily represented by balloon graphs. I explain out their occurrence by saying that a tameni benefactive is only needed when the entity inthebeneficiary role is not referred to within the –te sentence. The problem is that, as adjuncts, tameniphrases should be drawn on the extreme right of the graph, but this would mean putting the balloons relating to the one beneficiaryin two separated places. Therefore, for the sake of clarity I usually stack the two beneficiary phrases on the left of the object phrases.

Semantic roles labels are particularly useful in Graph 13as they make help to underscore the fact that these constructions of giving are akin to those of sending and putting and, most important, show that the same set of entities participating in a "giving" event is also involved in the matching "receiving" event described by a conversive predicate of the morau class.

Graph 14: Balloon graph of benefactive predicate of receivingmorau