A valuable perspective can be gained by examining whether claims about certain linguistic phenomena that were made largely based on observations from particular languages still hold for similar constructions in typologically different languages. This thematic issue contributes to SLA research in this respect by rethinking the putative universal influence of the noun phrase accessibility hierarchy (NPAH; Keenan & Comrie, 1977) on the acquisition of relative clause (RC) constructions in a second language (L2). The NPAH has been strongly supported by numerous studies in English and other European languages such as French, Italian, and Swedish; the articles in this special issue examine L2 RC acquisition in East Asian languages such as Japanese, Korean, and Chinese.
I read the articles in this issue with much interest because of the timeliness of the topic and because of my work on adnominal clauses in Japanese, which I have often referred to as noun-modifying constructions (NMCs) in previous work (e.g., Matsumoto, 1988, 1989, 1996, 1997, 2007). NMCs have also been referred to as Japanese-type constructions or as attributive clauses by Comrie in some of his work with broader typological elaborations of the constructions in question (e.g., Comrie, 1998a, 1998b, 2002; Comrie & Horie, 1995).
Adnominal clauses in Japanese often resemble RCs but are not satisfactorily accounted for by the standard syntactic analyses of RCs, which are mostly based on instances of RCs in English or in languages that are typologically similar to English. One common but unsatisfactory way to deal with such so-called unfitting constructions in Japanese is to exclude them as a target of study and to focus only on the instances that are amenable to existing analyses. Another way to deal with unfitting constructions is to search for explanations for the phenomena as they are found, even if that might require radically different perspectives. My work in Japanese employed this latter approach and proposed an approach that is not limited to syntactic factors but that recognizes the precedence of semantics and pragmatics in explaining constructions that go beyond syntactically defined RCs. Comrie (1998a, 1998b) showed that the characteristics of adnominal clauses that I had pointed out with respect to Japanese are shared by other languages in East Asia and Southeast Asia as well as the Dravidian languages of South Asia and some Turkic languages. I believe that this special issue is motivated by the desire to understand the constructions that are actually found in the language, instead of imposing classifications that derive from typologically dissimilar languages.
The findings of the five articles in this issue include the following: (a) It is inconclusive whether the NPAH determines SLA of RCs in Japanese, Korean, and Chinese, (b) the animacy of the referent expressed by the head noun influences a learner's ability to understand constructions, (c) the learner's first language (L1) tends to influence SLA of RCs, and (d) instruction focused on a marked type can facilitate the acquisition of other types. These findings seem realistic from the perspective that semantics and pragmatics play a determining role in the grammaticality of the adnominal constructions in question.
I will not reiterate in this commentary many of the arguments against purely syntactic treatments of Japanese NMCs, the detailed proposal based on frame-semantics (e.g., Fillmore, 1977, 1982, 1992) that I made, or the arguments made by Comrie, but I will simply focus on theoretical and practical issues that seem to be relevant to the inquiry of L2 learning and acquisition of the NMC. I will limit my examples in the following discussion to Japanese, but I hope that what I describe here can provide background for future studies on the SLA of adnominal clauses in East Asian languages and other similar languages.
THE SPECTRUM OF JAPANESE NMCs
The most crucial structural feature of Japanese NMCs, which include equivalents to English RCs, is that there is no explicit item, such as a relative pronoun and verb form encoding person or number, that indicates the grammatical and semantic relationship between the head noun and the prenominal subordinate clause. The construal of the relationship between the two main constituents cannot depend solely on one factor that clearly points to their relationship but is achieved through an aggregate of semantic and pragmatic factors. The NMC is schematically shown in (1). It should be noted that arguments of verbs can often be omitted, depending on the linguistic and nonlinguistic contexts.
Although this structural difference between Japanese and English counterparts tends to be deemphasized in purely syntactic analyses, this difference is not simply superficial and has significant implications in the construal and learners' acquisition of the constructions. For example, the NMCs in (2)–(4) present three different interpretations in terms of the semantic relationship between the head noun and the subordinate clause predicate, although the adnominal clauses are identical. The apparent absence of an expected argument of the subordinate predicate is unrevealing as to whether the clause is a main clause, a RC, or a noun complement clause. The semantics of the head noun and knowledge about the function of what it signifies in the real world play a significant role in determining the acceptability and the meaning of the construction. In contrast, the English counterparts would be different in form. Example (4) illustrates a construction that does not correspond to either relative or noun complement clauses.
It is possible to argue that (a) once we understand that Japanese is a pro-drop language, the distinction between RCs and noun complements can be syntactically maintained and (b) the fact that there are NMCs in Japanese that do not correspond to RC and noun complement constructions is theoretically insignificant. However, the first point still needs to account for how an interpretation is determined—for example, how (2) and (3) get different interpretations although no arguments are bound with the head noun. The second point either ignores the significance of grammatical constructions or counts them as relevant only when they also occur in more favored languages.
It is true that there are Japanese NMCs that present a close parallel to English RC and noun complement constructions and that would satisfy all requirements of semantic and pragmatic naturalness (e.g., animacy, worldview) as well as syntactic explanations, as in (5) and (6).
As long as we limit the target of observations to such well-behaved constructions, in which syntax, semantics, and pragmatics all conspire to provide well-formed constructions that would parallel their English counterparts, Japanese NMCs would appear to behave in the same manner.
As mentioned earlier, however, there are a great number of attested adnominal constructions in Japanese that fall outside of (or that fall in both of) the conventional domains of RC and noun complement constructions, as exemplified in (7)–(11).1
Examples (7)–(11) and (16) are attested examples.
In these examples, none of the subordinate clauses contain a clear syntactic gap that is bound to the head noun, although many exhibit missing arguments. For example, in (7), there is no syntactic gap in the clause atama ga yoku-naru “head gets better” because the predicate is an intransitive verb. Nokori “remainder” in (10) is not a likely candidate as a noun that instantiates a missing subcategorized argument of the verb tabeta “ate.” Furthermore, unlike the noun complement clause in (6), the proposition given in the clause does not express the content of what is denoted by the head noun. In (11), kimuti hazimeta “started making kimchee” is not the propositional content of the head noun dooki “motive,” but the action that resulted from a motive that is not mentioned in the construction. The money in (8), on the other hand, is something that resulted from the work of translating and is neither a content noun nor a noun that could be extracted from the clause.
Because the construal of NMCs relies crucially on the construer's ability to infer a relationship between the two constituents, the speaker's production of acceptable constructions depends on the plausibility of the relationship. The example in (12) is not an easily construable construction even though it could be a subject RC (Matsumoto, 1997).
As the English translation shows, the English counterpart of (12) has one indisputable reading—namely the reading in which the tomato is the eater and Tokyo is the meal. Given this syntactically imposed interpretation, the construer is invited to imagine a world in which these conditions are satisfied—most likely, a fantasy world. The interpretation of the tomato as the agent and of Tokyo as the patient of the eating is dictated by the grammar of the sentence: It is not influenced by whatever difficulties the construer might have in envisioning the necessary situation.
The construal of (12) in Japanese presents a very different picture. Because there is no explicit marking to indicate how the head noun is linked to the predicate of the modifying clause, the construer needs to infer the semantic relationship between the two elements. As described in Matsumoto (1997), I presented example (12) to 10 native speakers of Japanese in informal interviews. None of the participants gave the reading of the tomato as the eater as their first response. Because the interviews were conducted informally and on a very small scale, the results have little statistical validity; yet, the responses will illustrate how the Japanese deal with examples such as (12). Upon hearing (12), all hesitated before responding. About half then said that they did not know what (12) meant, and the rest asked whether it was meant to be tookyoo de “in Tokyo” rather than tookyoo o “Tokyo-OBJ.” If it were tookyoo de, as some of the interviewees wished, (12) could provide the plausible reading of referring to a tomato that someone (presumably the speaker, although it is not explicitly given in the phrase) ate in Tokyo. After being assured that this was not the intended meaning, about half of the participants decided that (12) was uninterpretable, although the rest finally reached the intended reading.
It is interesting to contrast the construal process in English and Japanese in light of this example. In English, one is first aware of the reading that the syntax of the construction imposes, although some might feel that one's worldview does not support that interpretation. In Japanese, on the other hand, hearers construe the relationship between the two constituents only after they first discard their ordinary or default worldview and adopt one of fiction or fantasy. Unless the situation activated by the construction is plausible, the interpretation will be at best hesitant or, often, impossible. Judgments of grammatical acceptability, then, rely on appropriateness in terms of the construer's worldview as much as, if not more than, his or her knowledge (unconscious though it might be) of the valency description (or predicate-argument structure) of a predicate. In other words, the well-formedness of NMCs depends on pragmatic, semantic, and syntactic aspects of the construction as a unified whole, not on only one of these aspects.
SEMANTIC AND PRAGMATIC FACTORS IN CONSTRUAL AND ACCEPTABILITY
The observations on (12) suggest that the semantics of the elements of a NMC as well as an understanding of real-world relations, such as the hierarchy of animacy, significantly influence the acceptability of Japanese NMCs. This is in fact what should be expected, given that the animacy hierarchy generally parallels the agent-patient (or subject-object) hierarchy and also given the fact that many languages encode animacy linguistically (see Comrie, 1989, for details on how the concept of an animacy hierarchy impinges on grammars of languages). In this special issue, Ozeki and Shirai as well as Kanno touched upon the question of animacy as a factor in RC acquisition. Their findings are, in this sense, in line with the observation that semantics and pragmatics are important factors in NMCs.2
Ozeki and Shirai (this issue) labeled a bus and a truck in their examples as inanimate objects. As extensions of their drivers, however, they can be understood as pseudoanimates (see also Shibatani, 1990).
The articles in the special issue present inconclusive results as to whether the NPAH determines SLA of RCs in Japanese, Korean, and Chinese. This is likely a reflection of the fact that the NPAH might be one of the factors that influences the comprehension of some (possibly prototypical) constructions, but it is not the single determining factor, and it might not be a relevant factor for other constructions in Japanese.
In my earlier studies (Matsumoto, 1996, 1997), I considered interactions of various factors that affect the construal of Japanese constructions, including (a) core roles of the predicate in the modifying clause, (b) the semantics of the head noun, (c) aspectual, empathy, and other information given in the predicate and other elements in the clause,3
The elements in the clause that I discussed include an adverb issyoni “together,” which enables the coactor of the action described in the subordinate predicate to be expressed as the head noun and no hoo “GEN side” for the head noun to be the object of comparison.
Those earlier studies (Matsumoto, 1996, 1997) cautioned against using only so-called default examples with semantic and contextual paucity. For example, if uttered in a situation of little contextual knowledge, we might conjecture that the most likely interpretation for the example given in (13) would be the one given in (a), the English translation, in which the denotatum of the head noun is the buyer, who would be the subject of the subordinate clause verb. However, other interpretations such as (b) and (c) are not precluded.
The interpretation given in (13a), however, is plausible only when the construction is considered with little contextual background or when the context of the utterance supports the interpretation. If the denotatum of the head noun does not match the construer's worldview or if the semantics of the head noun clearly indicates the instantiation of a noncore role (i.e., a less crucial participant in the scene of the described action), a core role interpretation will not be selected. For example, (13b) is preferred in linguistic contexts such as (14); similarly, (13c) would be the preferred interpretation in the context of (15).4
See Sheldon (1974) for similar discussions of English RCs.
It should be pointed out that interpretations based on core roles—often the only interpretations discussed in the literature—are merely default readings that are preferred only where the context is maximally generalized. Any such preference can be overridden by factors that depend on more specific information provided within or outside of the NMCs.
Because various factors are relevant in the construal and production of Japanese NMCs by native speakers, it is reasonable to assume that multiple factors also influence conditions for the acquisition of Japanese NMCs by L2 learners. How much information is provided through linguistic means or nonlinguistic factors in speech as well as how much knowledge of such information is shared by interlocutors could make a crucial difference in construal. The main characters of an action or event (e.g., agent, experiencer, patient) that are expressed as (core) arguments of a predicate might tend to be prominent in most contexts. Other situations might require more specific information to be provided through linguistic means, or perhaps a familiarity with a broader cultural context is necessary. Instances of the construction that do not rely upon specialized knowledge are easier to construe and easier for L2 learners to acquire.
The construction in (7), repeated as (16), is one example that can easily be understood by native speakers, young and old, but that presents difficulty for learners of Japanese even at the advanced level.
For L2 learners of Japanese to acquire the full spectrum of adnominal constructions, it would be to their advantage to be exposed to such marked constructions as (16). In this regard, Yabuki-Soh's study in this special issue provides an interesting direction for further study. It would be beneficial to explore further the relationship between the range of constructions to which Japanese L2 learners are exposed and the extent and degree of their acquisition.
