Definiteness Acquisition by L２−English Learners and the Feature Specification in D
Fumio Mohri＊ Remi Kawaryu＊＊
1. Introduction
The main purpose of this paper is to reconsider the universal semantics specified in the head of the determiner phrase（DP）in terms of second lan- guage acquisition. Through the discussion, we point out a pattern that English L２learners（Japanese L１）show in our experiment : overuse of the in parti- tives with disrespect for（non−）specificity. We propose the null universal hy- pothesis in the DP structure for determiner−less languages and that the feature specified in the head of a DP is not specificity/ definiteness, but is a±feature on domain restriction（Gillon２０１５）．
Ionin et al.（２００４,２００８）argue that the feature of definiteness or specificity is attached to the functional head of a determiner phrase on the basis of the cross−linguistic fact that languages can distinguish articles either on the basis of definiteness or of specificity but not both. Their analyses proceed with the assumption that adult L２learners have access to innate linguistic knowledge, which allows them to set the feature of either definiteness or specificity. How- ever, as pointed out by Mizuno（２００７）and Yusa（２００７）, even advanced Eng- lish L２learners（Japanese L１）cannot show accurate performance on definite- ness, as is the case with native speakers of English, which raises doubt about
＊ 福岡大学人文学部准教授
＊＊東京国際ビジネスカレッジ非常勤講師
adult L２learners’ UG access. Additionally, overuse and drop of English articles are said to be more prevalent in learners of determiner−less L１rather than L２ learners with determiners in their L１. Aside from the discussion of whether adult L２learners can have direct access to the universal semantics, we claim that the grammar systems in determiner−less languages have a negative trans- fer effect on the acquisition of English articles.
Through this paper, I assume that the core feature specified in the head of DPs among languages is a±feature on domain restriction. Cross−linguistically, articles vary in their semantics（Gillon２０１２,２０１５）: they encode definiteness
（English and many other languages）, deictic information（most Salish lan- guages）, and Specificity（Samoan）. I assume that the definite determiner in English is decomposed into the feature on domain restriction（as the core fea- ture in D）and the maximality feature, while the specific and deictic articles are embodied with the core feature but not with the maximality feature. As for determiner−less languages, on the other hand, I assume that bare NPs have an extended（null）projection, DP, whose head is specified with a ±feature on domain restriction. It is by no means unreasonable to say that innate linguistic knowledge provided by UG affects L２−English learners’ understanding of defi- niteness effect. The DP characterization assumed here leads learners to over- use the definite determiner for nominals that bear familiarity and also favorably judge as indefinite nominals whose description is novel, because the former case more likely makes the hearers recognize that the target item is contextu- ally restricted, but the latter is not easy to identify as a restricted item. Our analysis suggests that there exists a negative transfer in their learning process of the definite article and at the same time, may contribute to the reconsidera- tion of DP structures in natural language.
2. Backgrounds
There is a strong view on second language acquisition that L2 learners rely on three sources of linguistic knowledge :（i）L２−input ;（ii）Transfer of L1 knowledge ; and（iii）Access to universal principles. No one believes the L２ development without exposure to the input of L２. More contentious are trans- fer of L1 knowledge and UG−based knowledge. In fact it has been discussed in the literature how these two sources are interrelated.
According to Ionin（２００３）and Ionin et al.（２００４,２００８）, article languages encode their articles either on the basis of definiteness or of specificity. These features of definiteness and specificity are specified in the head of a determiner phrase（DP）. Ionin（２００３）and Ionin et al.（２００４,２００８）advocate the Fluc- tuation hypothesis in（１）and the Article Choice Parameter in（２）, giving a new explanation to L２learners’ misuse of articles.
（１）The Fluctuation Hypothesis :
１）L２−learners have full access to UG principles parameter settings.
２）L２−learners fluctuates between different parameter settings until the input leads them to set the parameter to the appropriate value.（Ionin２００３：２３）
（２）The Article Choice Parameter
A language which has two articles distinguishes them as follows :
The Definiteness Setting : Articles are distinguished on the basis of definite- ness.
The Specificity Setting : Articles are distinguished on the basis of specificity.
（Ionin２００４：１５）
According to informal definitions given in Ionin et al.（２００４）, the difference be- tween definiteness and specificity has to do with hearers’, not speakers’ as- sumption : if a noun phrase（NP）is [+definite], both the speaker and the hearer presuppose the existence of a unique individual in the set denoted by the NP ; on the other hand, if the noun phrase is [+specific], then the speaker intends a unique individual in the set denoted by the NP and considers this in- dividual to possess noteworthy property.１
Ionin et al.（２００８）conducted tests on L１−Spanish and L１−Russian learners of English and found that L１−Spanish learners shows a higher ratio of correct use on the tests. Based on the resultant data, they reach the conclusion that L１−Spanish learners transfer their L１knowledge to the L２acquisition, while L１−Russinan learners has no choice to access to semantic universals due to the lack of L１transfer, thus showing the fluctuation between definiteness and specificity.
Note also that error patters L２learners show are not at random. According to Ionin et al.（２００４,２００８）, L２adult learners of Russian and Korean, who can- not rely on any help from their L1 due to the lack of articles, show overuse of the with specific indefinites and overuse of a with non−specific definites. This error patter is summarized in Ionin et al（２００８：５５９）as in（３）:
（３）
[+definite] : targetthe [−definite] : targeta
[+specificity] Correct use ofthe Overuse ofthe
[−specificity] Overuse ofa Correct use ofa
Their explanation of the error patter appears clear and even straightforward : L２learners has to access both semantic universals, definiteness and specificity,
but initially has no clue as to which of the semantic universals is relevant to the use ofthe or a, hence fluctuating between the two options. More specifically, some of the time they treatthe as marking definiteness, and some of the time they treattheas marking specificity. Though, things are complicated if they are in conflict. When the target expression is specific, but non−definite, i.e., specific non−definite, or when it is non−specific, but definite, i.e., non−specific definite, they show many errors of article use, as shown above in（３）. However, along the lines taken by Ionin et al.（２００４,２００８）, through trials and errors, the learners are able to recover from the fluctuation and master the target−like definiteness pattern through the L２（English）input.
Ionin et al.（２００８）discovered that L２learners whose L１（Spanish）has the same article system resulted in appropriate article use. Unlike L２learners whose L１lack an article system, they can transfer the linguistic source of their L１to the L２acquisition. This corroborates Ionin’s hypothesis that transfer overrides fluctuation’. Put differently, L２learners whose L１lack articles are not subject to L１article semantics and thus show fluctuation between specificity and definiteness on the article semantics. Their analyses proceed with the as- sumption that L２learners have access to innate linguistic knowledge, which al- lows them to finally set the feature of definiteness through L２input（intensive exposure to English）.
When it comes to another article−less language, Japanese, Ionin’s logic does not seem to apply to their L２−English learners, as it is. If L２learners can ac- cess universal semantics, as assumed by Ionin et al.（２００４,２００８）, they should finally converge on the definite option for article semantics. However, as pointed out by Mizuno（２００７）and Yusa（２００７）, even advanced Japanese L２ learners of English cannot show accurate performance on definiteness, as is
the case with native speakers of English, which raises doubt about adult L２ learners’ UG access. We will not explore Ionin’s hypothesis that L２learners draw upon universal semantics, but will contend, through the examinations of Japanese L２learners of English, that the grammar systems in determiner−less languages have a negative transfer effect on the acquisition of English articles.
Overuse and drop of English articles are more observable in learners of deter- miner−less L１rather than L２learners with determiners in their L１. In the pre- vious prevalent views, this fact has been attributed to the lack of L１transfer, but we will advocate that L２learners apply a negative linguistic knowledge to the article semantics of the target language. As a presupposition for structuring our argument, we need to reconsider the characterization of determiner phrases from a more cross−linguistic perspective.
3. Syntax and semantics of DP
Ionin et al.（２００４,２００８）have argued that articles encode either definiteness
（English and many other languages）or specificity（Samoan）but not both,
and that these features are attached to the functional head of a determiner phrase（DP）. This DP characterization lies in Ionin’s core argument, and if L２ learners cannot draw on L１knowledge on article choice, then they fluctuate be- tween the definiteness and specificity options in the process of acquiring the English articles.
Note, however, that cross−linguistically, the semantics of articles are not nec- essarily dichotomous（Gillon２０１２,２０１５）in that they do not necessarily en- code either definiteness or specificity. As pointed out by Matthewson（１９９８）
and Gillon, most Salish languages, for example, have their articles encode deic- tic information. One of Salish languages, Skxwú7mesh, employskwa to refer to
individuals that are not near the speaker nor visible andti to refer to individu- als that are visible and close to the speaker. The following are the examples from Gillon（２０１５：１７８）:
（４）Skxwú７mesh a. Kw’áy’ kwaBill
get.hungry DET Bill
Bill is hungry （Bill not in room and not visible）
b. Na mi púmti−n s７átsus REALIS come swell DET−１SG face
My face is puffy/swollen. （Gillon２０１５：１７８）
They are genuine articles that encode deictic, not demonstratives, because they cannot stand on their own.
（５）Skxwú７mesh a. Kw’áy’ kwa
get.hungry DET
（Intended : Bill is hungry”）
b. Na mi púmti−n s７átsus REALIS come swell DET−1SG face
（Intended : It is puffy/swollen. （Gillon２０１５：１７８）
The other possible semantics for articles, as mentioned repetitiously, is specific- ity. It has been observed that articles in Samoan encode specificity, rather than definiteness.
（６）Samoan
a. O le povi
PRES ART cow It’s a cow. （specific）
b. O se povi lale?
PRES ART cow DEM
Is that a cow? （non−specific）（Mosel and Hodhaugen１９９２, cited in Gillon２０１５：１７９）
In Samoan the articlele is used to refer to a specific individual, andse is used to refer to a non−specific individual.
In this way, semantics of articles vary cross−linguistically in at least three ways :（in）definiteness,（non−）specificity, and deixis : they encode definite- ness（English and many other languages）, deictic information（most Salish languages）, as shown in（５）, and specificity（Samoan）as in（６）. Obviously, definiteness or specificity are not the only features that an article displays. It thus seems inadequate to assume that semantic universals learners have access to or the semantics of features specified in the head of determiner phrases are dichotomous between specificity and definiteness.
When it comes to the semantic contribution of definiteness, there seem to be roughly two camps in the literature. In the generative literature, definiteness is some uniqueness/ maximality, which is assumed to be derived by the type−
shifting operator,iota（cf. Partee１９８７, Chierchia１９９８, Dayal２００４,２０１３）:
（７）iota:λP Ps, if there exists a unique maximal entity in P, undefined other-
wise. （Chierchia１９９８：３４６）
Iota picks out the maximal entity in the extension of the NP predicate for the relevant situation, if there is one, and is undefined otherwise, as defined in（７）. With（８）below for instance, let us see howiota is at work in a concrete con- text.
（８）There are four dogs on the sofa. The dogs are sleeping.
The plural individuals, four dogs, are introduced in the context of（８）, and the dogs usually refers to the（unique）maximal entity, the four dogs, not two dogs. Put differently, the definite article that embodies iota applies to the ex- tension of the NP to return the unique and maximal entity.２
On the other hand, some linguists associate the definiteness effect with fa- miliarity（Heim１９８８）. However, the English definite article does not necessar- ily presuppose familiarity of the relevant individual, as shown in the contrast below :
（９）a. I’d like to talk to the winner of today’s race −−− she is my best friend!
b. I’d like to talk to the winner of today’s race −−− Whoever that is, I’m writing a story about this race for the newspaper.
（Ionin, Ko and Wexler２００４：８）
In（９）the noun phrase, the winner of today’s race, is definite in both sen- tences, but what differs is that the target noun in（９a）bears specificity, whereas that in（９b）does not. The former is the speaker’s best friend and, thus（s）he has a ‘noteworthy’ property. It can also be said that it has a famili- arity property since（s）he is the speaker’s best friend, but the target noun in
（９b）is obviously beyond the speaker’s knowledge.
Incidentally, we have seen that cross−linguistically articles vary in three ways, i.e.（in）definiteness, specificity, and deixis. It appears that their seman- tic properties can be characterized with either familiarity or maximality : while the definite determiner in English has the maximality feature embodied, what the specific and deictic articles have in common is that they both encode famili- arity. As shown above in（４−６）, these two articles refer to the individual that is visible or closer to the speaker and the one that has a noteworthy property, respectively −−− we do not address the question of what features deixis and specificity are decomposed into. Though it seems true that these three articles vary in their semantic properties, Gillon（２０１３,２０１５）claim that they share a common core : they all domain restriction. It is certainly true that specific and deictic items are both the ones familiar to the speakers, which are contextually restricted entities. On the other hand, it is no doubt that definite items, too, are interpreted as maximal/ unique individuals in the restricted domain.３
Let us, next, turn to the syntax of articles. Since Abney（１９８７）, it has been generally assumed that articles are part of the extended projections of a nomi- nal. It is reasonably assumed that those three types of articles occupy the same position, D. As for determiner−less languages, though still controversial, we as- sume that bare NPs have an extended（null）projection, DP, whose head is specified with a core feature. Cross−linguistically the functional head D should be distinguished as follows :
（１０）All determiners and a null D head are specified with a ±feature on do- main restriction.
In the section to follow, we will structure our argument based upon the DP characterization assumed here. Though it is a controversial issue whether arti- cle−less languages have covert articles or not, we suggest that assuming the functional DP projection cross−linguistically will help us more successfully cap- ture some overuse patter oftheby L２−English learners.
4. Hypothesis and Preliminary Tests
We have suggested that some semantic knowledge provided by UG leads L２ learners to overuse the definite determiner for familiar nominals and also favor- ably judge the description that is novel in the relevant context as indefinite nominals. We assume that the UG knowledge is traceable to the universal DP hypothesis. Our DP hypothesis is different from Ionin et al.（２００４,２００８）and their followers in that the functional D head is characterized with a ±feature on domain restriction（as the core feature）.
We reconsider the DP semantics in terms of second language acquisition.
Through the experiment we point out a pattern that English L２learners（Japa- nese L１）show : overuse ofthe in partitives with disrespect for（non−）specific- ity. We propose the null universal hypothesis in the DP structure for deter- miner−less languages and that the feature specified in the head of a DP is not specificity/ definiteness, but a±feature on domain restriction.
At present, our experiment is still at a preliminary stage and is a stepping stone to a full−scale investigation. A group of twenty college students partici- pated in this test as subjects. We focused only upon the specific definite in
（１１）, the non−specific definite（invoking functional reading）in（１２）and the partitive cases in（１３−１５）. In the task we worked out, participants were asked to write the definite or indefinite article in each space or leave it blank when
appropriate not to write anything.
What we particularly focused upon is their article choice in partitive cases, shown in（１３−１５）, where the referents are set in context and thus are treated as familiar individuals. We investigate these cases with reference to the well−at- tested patterns that L２learners show : proper use ofthe for specific definites and fluctuation between the definite and indefinite choices for non−specific defi- nites. Examples（１１）and（１２）are classified into the [+definite, +specific] and [+definite, −specific] categories, respectively. The remaining examples from
（１３）to（１５）are all partitive cases : we have classified（１３）as a [−definite,
−specific] case, and（１４）and（１５）as [−definite, −specific] cases.
（１１）[+definite, +specific]
（ The ）students, namely, John and Bill, are shouting.
（１２）[+definite, −specific]
Yesterday we paid a visit to a textile factory. When we arrived at the gate, we were greeted by（ the ）manager.
（１３）[−definite, −specific]
My daughter called me this morning, saying that she received a lot of ap- ples from her uncle. She was also delighted to tell me that she peeled and ate（ φ ）two apples for breakfast.
（１４）[−definite, −specific]
There are a lot of talented soccer players in our team. Please name
（ φ ）two talented players−−− whoever you know.
（１５）[−definite, +specific]
When I visited John’s house, I found that he had five dogs. Among them
（ φ ）two dogs look very much alike. They were pretty cute!
With respect to the target entities in（１１）and（１２）, as expected, the partici- pants showed similar patterns to the studies of Ionin et al.（２００４,２００８）. They performed nearly perfectly on the [+definite, −specific] target item in（１１）but poorly on the [+definite, −specific] target.４Our main concern is their perform- ance in the partitive cases. The antecedents for the target items in（１３−１５）
are all presupposed as supersets and, in turn, the target items are treated as their subsets. These target items are interpreted as familiar entities and, of course, are also considered contextually restricted entities. Note also that the target item in Example（１５）, unlike those in（１３）and（１４）, are specific indi- viduals because the noteworthy dogs are described as cute by the speaker ; on the other hand, those in（１４）and（１５）are obviously given non−specific inter- pretation because any item in the domain holds true of their truth conditions.
Turning to the results for the target items, we divided the participants’ re- sponses in each category into definiteness and indefiniteness : use of the defi- nite article for the former and use of the indefinite article or leaving a blank for the latter. The table below reports the resultant percentage :
（１６）L１−Japanese L２English learners（N＝２０）
Definite（the） Indefinite（aor blank）
[+definite, +specific] ９５％ ５％
[+definite, −specific] ５０％ ５０％
Partitive [−definite, −specific] ３０％ ７０％
Partitive [−definite, −specific] ２５％ ７５％
Partitive [−definite, +specific] ２０％ ８０％
In overview, almost all the participants show accurate use of the definite article for the specific definite : they are obviously more accurate than any other case.
When it comes to non−specific definiteness, they do indeed fluctuate between definite and indefinite interpretations but, more accurately, they wrongly over- use the indefinite article. Though we cannot spell out the details, we predict that the way the target item in（１２）was introduced may have caused some difficulty for the participants. Unlike the non−specific definite in（９b）above, the target item in（１２）is introduced without any prepositional phrase or rela- tive clause as a modifier. The hearers are required to identify the target as the unique entity only by associating it with the presupposed relevant nominal.
They may not yet have acquired this process of association for definiteness and judged the target item that was introduced out of the blue as an（unfamiliar）
contextually unrestricted item. However, the argument here remains specula- tive ; it will be buttressed by a future full−fledged experiment.
Finally we turn to the results of the partitive cases from（１３）to（１５）. The table above reports that the participants certainly show fluctuation on partitives between definite and indefinite interpretations. Though limited to the learners committed to this experiment, it is highly obvious that fluctuation occur with disregard to（non−）specificity. The partitive targets have all been introduced first as part of their supersets in the context, so they are treated as familiar in- dividuals and also as a contextually restricted set of individuals. We have as- sumed that based upon the universal DP semantics, determiner−less languages have the（covert）D head specified only with the core feature, namely, a±fea- ture on domain specification. Thus, whether a target item is contextually re- stricted or not can be a crucial factor on the learners’ judgement of definiteness effect. The superset including the partitive item is a specific definite, which
means that the D head is set with +feature on domain restriction. This feature setting may have affected the learners’ understanding of definiteness effect in English. Japanese grammar does not have the counterpart to theiota operator thattheembodies, defined above in（７）, so the participants may be more likely to judge definiteness effect in terms of domain restriction or familiarity, not necessarily associating it with the maximal property ofthe. This reasoning may have reflected the result in the table（１６）.
5. Final Remarks
As we have repeated numerous times, this was only a preliminary experiment intended to help us create our hypothesis. Through this experiment, we have confirmed that we need to reconsider the characterization of the functional de- terminer phrase（DP）. Since Ionin et al.（２００４,２００８）, the research on L２ learners’ acquisition of articles has been developed on the basis of a firm as- sumption that the semantic universals that L２learners draw on are definiteness and specificity. Certainly it has been reported through numerous experiments that for L２learners (precisely, L２learners whose L１s are article−less lan- guages), both definiteness and specificity have highly significant effects on arti- cle use. Cross−linguistically, however, the semantics of articles are not neces- sarily dichotomous（Gillon, ２０１２, ２０１５）: they do not necessarily encode either definiteness or specificity. As pointed out by Matthewson（１９９８）and Gillon（２０１２,２０１５）, semantics of articles vary cross−linguistically in at least three ways :（in）definiteness,（non−）specificity, and deixis. I have assumed, along with Gillon（２０１５）, that all determiners are specified with a ±feature on domain restriction, and also that they are decomposed into the feature of do- main restriction（as a core feature）plus their own distinctive features. The
definite determiner embodiesiota, which plays a role in picking up the maxi- mal（unique）entity in a contextual domain. Our hypothesis is that the delay or lack of understanding of this DP characterization (maximality) leads L２ learners to overuse the definite article in some (indefinite) cases. We tested our participants on some partitives and obtained the results above in（１６）. It is obvious that their responses fluctuated between the definiteness and indefinite- ness options −−− though they are slightly more accurate on the specific indefi- nite in（１５）.
However, we cannot say precisely that the fluctuation observed in the parti- tive cases is in fact statistically significant because we do not calculate p−val- ues. It is also true that we have not examined the role that proficiency plays in the degree of fluctuation. We leave these issues for the next investigation.
Notes
The second author is a part−time lecturer at Tokyo International Business College. We would like to thank Robert Cvitkovic and Stephen Howe for his comments on an early version of this paper. We are also grateful to Yuko Hayashi and Shota Kawano for their instruction of statistic approach, though, unfortunately, we could not fully make use of it.
Our thanks also go to Tim Cross for his helpful suggestion. Finally we should add that a part of this research is supported by the Ministry of Education, Culture, Sports, Science, and Technology（Grant−in−Aid for Scientific Research（C）, No.２５３７０５６８）.
１ Many works, including the ones from a psychological view, states that children’s er- rors（omission or overuse of the）are related to their inability to take the listener’s knowledge. It seems that specificity represents a more basic article meaning than defi- niteness.
２ Given the type−shifting framework in Chierchia（１９９８）and Dayal（２００４,２０１３）, in
article−less languages, too,iotawould be available in semantics in definite contexts. This suggests that article−less languages have theiota operator in their grammar system. If so, their English L２learners could rely on（positive）L１transfer in acquisition of the definite article, successfully showing good performance on article choice, contrary to fact.
Thus, Chierchia’s strategy, also called Last Resort’ mechanism, is not easy to accept in terms of L２research.
３ Gillon（２０１５:１７９）states that ‘（d）omain restriction is essentially the context that a nominal must be interpreted within.’
４ In the case with backward reference（１１）, the drop of the article was hardly ob- served among our participants. We presume that since the antecedent is made explicit and even easy to identify, the participants may have successfully chosen the definite arti- cle（cf. Jackendoff and Wittenberg２０１２）.
References
Abney（１９８７）The English Noun Phrase in Its Sentential Aspect. Ph.D. Thesis. Massa- chusetts Institute of Technology.
Chierchia, G.（１９９８）Reference to Kinds across Languages.Natural Language Semantics ６,３３９−４０５.
Dayal, V.（２００４）Number Marking and（In）definiteness in Kind Terms.Linguistics and Philosophy２７,３９３−４５０.
Dayal, V.（２０１３）On the Existential Force of Bare Plurals across Languages.From Gram- mar to Meaning, ed. by I. Caponigro and C. Cecchetto,４９−８０, UK : Cambridge University Press.
Heim, I.（１９８８）The Semantics of Definite and Indefinite Noun Phrases.New York : Gar- land.
Ionin, C.（２００３）Article Semantics in Second Language Acquisition. Ph.D. Thesis. Mas- sachusetts Instutute of Technology.
Ionin, T., Ko, H., Wexler, K.（２００４）Article Semantics in L２−acquisition : the Role of
Specificity.Language of Acquisition１２,３−６９.
Ionin, T., Zubizarreta, L.M., Maldonado, B. S.（２００８）Sources of Linguistic Knowledge in the Second Language Acquisition of English Articles.Lingua１１８,５５４−５７６. Gillon, C.（２０１５）Investing D in Languages With and Without Articles.Methodologies in
Semantic Fieldwork,ed. M. R. Bochnan and L. Matthewson,１７５−２０６, NY : Oxford University Press.
Gillon, C. and Armoskaite S.（２０１２）The Semantic Import of（C）overt D. Proceedings on Formal Linguistics, ed. Jaehoon Choi et al.,３３７−３４５. Somerville, MA : Casca- dillla Proceedings Project.
Jackendoff, R. and E. Wittenberg（２０１２）What you can Say without Syntax : A Hierar- chy of Grammatical Complexity. Measuring Linguistic Complexity, ed. by F. New- meyer and L. Perdue,６５−８２, Oxford : Oxford University Press.
Matthewson, L.（１９９８）Determiner System and Quantificational Strategies : Evidence from Salish.The Hague : Holland Academic Graphics.
Mizuno, M.（２０００）Tyukangengobunseki−−− Eigokansi−no Kiseki（The analysis of Inter- language −−− the Process of English Article Acquisition）, Tokyo : Kaitakusya.
Partee, B.（１９８７）Noun Phrase Interpretation and Type−shifting Principles. Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers,ed. by T.
Janssen and M. Stokhof,１１５−１４３, Dordrecht : Foris.
Yusa, N.（２００７）Dainigengo−niokeru Kasi−no kobetusei−to Fuhensei（Individualism and Universality of Articles in Second Language Acqusition）Eigo−to Bunpo（English and Grammar）, Tokyo : Kaitakusya.
