Plural-marking in L2 Korean: A feature-based approach

Abstract

This study examined the second language (L2) acquisition of the Korean plural marker -tul by native speakers of English. Seventy-seven learners at four Korean proficiency levels along with 31 native Korean-speaking controls completed five tasks designed to probe for knowledge of particular features and restrictions associated with so-called intrinsic and extrinsic plural-marking in Korean. The results suggest that knowledge of both types of plural developed with increasing proficiency. However, the features associated with the intrinsic plural, which is more similar to the English plural in terms of grammatical function, were more easily acquired than those of the extrinsic (distributive) plural, which requires recruiting the features of a completely distinct morpholexical item from the first language (L1). We also found some developmental evidence for a feature hierarchy in quantified Korean noun phrases, in which the most deeply-embedded featural co-occurrence restriction on intrinsic plural-marking was the latest acquired.

Keywords

English learners of Korean feature reassembly Korean plural L2 features L2 Korean L2 morphosyntax L2 number marking

Introduction

The acquisition of functional morphology poses widely-documented difficulties for second language learners, particularly for adult acquirers. DeKeyser (2005: 5–6) cites over two dozen studies exploring the difficulties in the second language (L2) acquisition and/or instruction of grammatical morphemes such as articles, classifiers, grammatical gender, and verbal aspect, the functions or meanings of which are ‘notoriously hard’ for learners whose native languages do not have them. It is not hard to see why acquiring functional morphology or, as Chomsky (2001: 10) puts it, ‘assembling’ the functional lexicon of a language, is difficult. The learner must acquire not only the knowledge of exactly which morpholexical forms and their allomorphic variants express which (combinations of) syntactic and semantic features, but also knowledge of potential conditioning factors that may be phonological, syntactic, lexical, semantic/pragmatic, and/or discourse-linked, including the conditions under which such forms are obligatory, optional, or prohibited. For adult L2 acquisition in particular, the constraining role of prior language knowledge (or first language transfer) should not be underestimated, especially at early stages of L2 acquisition; this is a position articulated most explicitly by Schwartz and Sprouse (1996) as their Full Transfer Full Access hypothesis, according to which the native language grammar constitutes the departure point or initial-state grammar of the L2. In other words, mature L2 learners bring to the L2 learning task an entrenched system of morphosyntactic features already assembled into lexical items; these are the morphemes associated with the functional categories of their native language(s).

In this article, we follow up on previous work by Lardiere (2009) on the acquisition of plural number marking in L2 Korean by adult native English speakers. We adopt a feature reassembly perspective: a comparative linguistic feature-based approach in which the ultimate attainment of nativelike L2 morphosyntactic knowledge depends on the extent to which learners are able to reconfigure the feature values in functional categories and lexical items, and the conditions under which these are realized, from those of the first language (L1) to the L2 in cases where these differ. We therefore rely on the Full Transfer hypothesis, in that we assume that learners initially assume that the functional categories and morpholexical items of the L2 are organized in a similar way to those of the L1 so that, say, a plural marker in the target language will be interpreted as more or less syntactically and semantically equivalent to a plural marker in the L1. The acquisition of number marking in L2 Korean presents an interesting case in which the relevant features associated with the Korean plural marker -tul overlap to some degree with those of the English plural but differ markedly in others. We present L2 acquisition data suggesting that the required feature reconfiguration is indeed attainable, at least at advanced stages of L2 development, and even for morphological phenomena that occur quite infrequently in the language environment.

The remainder of this article is organized as follows. The next section reviews some descriptive generalizations about number marking in Korean (see also Lardiere, 2009). Following this, we consider which type of theoretical framework is best suited to account for the data and adapts a feature-geometry framework similar to that used by Gebhardt (2009) to describe plural and classifier determiner phrases (DPs) in Persian, Mandarin Chinese, and English, and outlines the L2 acquisition task in terms of the required feature-reassembly. We then present our study and its findings, and the final section concludes with a summary and discussion of our findings.

Descriptive overview of number marking in Korean

Intrinsic versus extrinsic plural types

Korean plural number marking is productive yet heavily constrained. There are in fact two types of plural marking in Korean, referred to in the literature as ‘intrinsic’ and ‘extrinsic’, although both types share the same morphological -tul suffix, as shown in (1a) and (1b). The function of the intrinsic plural marker, shown in (1a), is more similar to that of English plural marking (and is glossed simply as pl in the following examples). The extrinsic plural marker (glossed as epl ) can be used not only on nouns (including non-count nouns), but also on non-nominal categories such as adverbs and locative phrases as a kind of distributive marker, as shown in (1b) (from Kim, 1994: 313); it is also much less common in the input.

One way to characterize the distinction between the intrinsic and extrinsic plural is that the intrinsic plural denotes multiple nominal referents (as in English), whereas the extrinsic plural ‘pluralizes’ entire predicates by distributing them over the individual members in the denotation of a (necessarily referentially plural) c-commanding subject. The intrinsic plural, moreover, can have either a distributive or collective reading, as in English (example from An, 2007: 2):

In the following subsections, we briefly outline some additional characteristics of both types of Korean plural marking, the acquisition of which we investigate in this study.

The Korean intrinsic plural

a Plural marking and specificity

In addition to plural number, plural marking in Korean indicates nominal specificity, as suggested by the English translations in (1a) above. In that example, the omission of the plural marker would suggest an underspecified meaning that in context would be interpreted more along the lines of (singular) ‘a/the child’, or non-specific ‘(some) children’. We can moreover observe the incompatibility of plural marking with kind-taking predicates and clearly-defined non-specific contexts as shown in (3a–3c); see example (3a), adapted from Kwon and Zribi-Hertz (2004: 147):

Plural marking is obligatory, or at least strongly preferred, for plural referents when these co-occur with demonstratives, which also render a noun specific. The bare noun in (4a) cannot be interpreted as plural when it occurs with a demonstrative; if referring to multiple entities, it must be plural-marked as shown in (4b) (from Kim, 2005: 89):

Plural marking is also strongly preferred for plural referents that have already been introduced in the discourse (and are therefore specific) (see example from Nemoto, 2005: 399):

b Plural marking with classifiers and quantifiers

Plural marking is not allowed on nouns in numeral-classifier constructions, as in (6a), unless the noun is [+human], in which case it is allowed (6b):

There is also a highly restricted case in which a numeral may appear without a classifier, namely, if the noun is a human and the numeral is small (otherwise, numerals must generally co-occur with classifiers), and in this case, the noun may also optionally be pluralized:

With non-numeric (or ‘weak’) quantifiers, however, such as manhun ‘many’, classifiers are not used and plural marking is optionally allowed, regardless of whether the noun is human or not:

Finally, numeral-classifier constructions with or without plural marking can occur in either of two equally acceptable word orders:

As shown in (9a), the pre-nominal numeral-classifier construction also requires a genitive case marker; the post-nominal numeral-classifier phrase in (9b) does not require the genitive case marker and is more common in spoken Korean.

The Korean extrinsic plural

As mentioned above, Korean also makes use of the same suffix -tul to distribute events or properties over the individual members of a plural subject; this use is often referred to as extrinsic plural marking and it optionally appears on direct or indirect objects, adverbs, locative phrases, and other types of constituents. The primary syntactic requirement is that the pluralized entity be c-commanded by a referentially plural subject in the same clause (An, 2007; Chung, 2003, 2004; Park, 2009; Yim, 2003), as illustrated in (10). (Example (1b) is repeated below as (10a); examples (10b–10c) are from Y Kim, 1994: 313–314.)²

Because a plural subject is required to license the extrinsic plural, the latter has been analyzed as a type of agreement marker (Park and Sohn, 1993; Yim, 2003). We analyze it as bearing an interpretable distributive feature ([distr]) that functions as a distributive operator on a predicate, a primary purpose of which is to disambiguate between potentially ambiguous collective versus distributive readings (Brisson, 1998). Because it introduces universal quantification over the plurality of the subject, we assume it bears an uninterpretable plural feature ([u-pl]).³ Although its use is infrequent, the extrinsic plural helps to recover the plurality of a null subject, and therefore it is more likely to be used in questions, commands, and exhortatives in colloquial speech than in declarative sentences, as shown in the examples in (11) (from Moon, 1995: 358) and (12) (from Song, 1988: 19–20):

Finally, unlike the intrinsic plural marker, the extrinsic plural is affixed outside the case marker if it is attached to a noun (example from Song, 1997: 209).

As shown in (13), the (extrinsic) plural marker affixed outside the dative case marker in (13a) yields a distributive interpretation in which there were multiple gift-giving events to one child; however, when the same (intrinsic) plural suffix precedes the dative case marker in (13b), it simply pluralizes the noun it is attached to.

The features of Korean number marking

In this section, we outline the features of Korean intrinsic and extrinsic plurals and compare them with their English counterparts. Given these features, and assuming Full Transfer, we then summarize what a native English speaker must redistribute onto new lexical items in L2 Korean.

A feature-based approach to Korean intrinsic plural marking

For the intrinsic plural, we start from the premise that there is, in fact, a plural number feature in Korean, a claim that is not universally agreed upon. Chierchia (1998), for example, has argued that there are no plurals in classifier languages (such as Korean), as nouns in classifier languages are argumental mass nouns for which plural marking is not needed. However, Chierchia’s claim has been contradicted by the findings of several subsequent studies (many of which are cited in Lardiere, 2009). One of the most detailed critiques of Chierchia to date is that by Borer (2005), who argues that in fact all nouns in all languages have a mass interpretation by default, unless they are ‘portioned out’ by some sort of ‘stuff divider’ that enables them to interact with the ‘count’ (or quantification) system. Such dividers include not only classifiers, but also plural marking. That is, Borer proposes that ‘plural inflection is classifier inflection’ (p. 93, original emphasis) and uses as her primary argument the observation that plurals and classifiers occur in complementary distribution even within languages that have both, such as Mandarin Chinese and Armenian.

Borer’s argument is in turn taken up in detail by Gebhardt (2009), who maintains that it is empirically inadequate because there are languages in which classifiers and plurals may indeed co-occur (albeit under restricted conditions for some of these), including Akatek, Armenian, Indonesian, Japanese, Korean, Persian, Tajik, and Tariana. Gebhardt instead adapts Harley and Ritter’s (2002) feature-geometry model, in which a classifier’s feature ([individuation]) only ‘individuates’ a noun as a count noun; however, plural morphology also has a [group] feature, which not only entails the presence of [individuation], but also calls for a more specific singular/plural distinction. Therefore, argues Gebhardt, independent Cl(assifier) and Number categories are both needed to accommodate these data.

For Korean intrinsic plural marking, we have observed earlier that a numeral and classifier may indeed co-occur with the plural suffix -tul if the pluralized noun is human:

Assuming this to be the most complex DP construction we need to be able to account for (for our purposes), we adopt the same sort of feature-based analysis for plural and/or classifier DPs as that used for Persian, Mandarin, and English by Gebhardt (2009). To account for the various intricacies of intrinsic plural marking in Korean as outlined above, the following features are required:

For pluralizing nouns within quantified noun phrases (npS) in Korean, there is a feature hierarchy for contextual complexity in which learners must first learn to distinguish between [q-rel] (for which nouns can optionally be pluralized) and [q-abs], and then, within [q-abs], whether the classified noun is [+human] or [–human], as illustrated in Figure 1.⁴

Figure 1.

Feature hierarchy for pluralizing quantified noun phrases in Korean.

In contrast, plural-marking in English is less complex: the only features required for English plural marking are [n] and [group], because pluralization in English is indifferent to specificity (e.g. those students, some students, any students), whether or not the pluralized noun is [human] (e.g. three students, three books), or whether a quantifier is numeric or non-numeric (e.g. many students, three students).

To summarize the discussion so far, native English speakers acquiring the Korean intrinsic plural must do the following:

acquire a Classifier Phrase (ClP) and learn which classifiers go with which nouns (e.g. [n, human], [n, animal], [n, small thing], etc.);

distinguish [q] features: [rel] versus [abs] and learn that Cl is associated with [abs] but not [rel] (e.g. chayk sey kwen = book three CL ‘three books’ versus manhun kenmul = many building ‘many buildings’);

learn that numeral+classifier NPs have two equally acceptable word orders, one of which (the one with the more English-like word order) requires a genitive case marker;

learn the [n, human] restriction (or the lexical restriction to particular human classifiers myeng and pun) for pluralizing numeral+classifier NPs;

add the uninterpretable feature [u-specific] to the functional category NumP, as English plurals are indifferent to specificity (e.g. books, some books, three books, the books, those books);

related to the specificity feature above, learn that plural-marking is required to denote plural referents in the context of demonstratives ku ‘that/those’ and i ‘this/these’;

also related to specificity, attend to discourse cues for the preferability of lural-marking to refer to a definite plural discourse referent.

Features of the extrinsic plural in Korean

As mentioned earlier, we suppose that the extrinsic plural bears an interpretable distributive feature ([distr]) that functions as a distributive operator on a predicate, as well as an uninterpretable plural feature ([u-pl]) that is licensed by the referential plurality of the subject. The closest likely equivalent lexical item in English is each (of the), which shares those same features. However, each in English is a distributive floating quantifier with complex syntactic properties, a description of which lies beyond the scope of this article; but see, for example Hoeksema (1996) and many references therein. In contrast, it is not clear to us that ‘floating’ is an accurate description of the placement of the distributive extrinsic plural -tul in Korean, because, although its position is variable, (1) it is affixal and (2) the element to which it is affixed is argued to receive focus within the event that is distributed over the individual parts of a referentially plural subject (Song, 1997), which makes it rather different from English each. The following examples are from Song (1997: 208):

Both sentences are interpreted as distributing a ‘quickly-drinking-water’ event over the set of children in the subject. However, Song suggests that (16a) should be interpreted as ‘For each of the children, it was water that s/he drank quickly’, indicating that the substance that was drunk (mul ‘water’) is focused. For (16b), on the other hand, the use of extrinsic -tul on the adverb ppalli ‘quickly’ focuses the manner in which each member of the set carried out the specified action (i.e. ‘For each of the children, it was quickly that s/he drank water’). In sum, Song suggests that the different positions of extrinsic -tul also induce different meanings, and that the use of the extrinsic plural marker in Korean is therefore, strictly speaking, not ‘optional’.⁵ However, since there are other more obvious ways to focalize elements in Korean, such as the use of topicalization or focus particles, clefting, and phonological prominence, we maintain that the use of the extrinsic plural is indeed grammatically optional, primarily used to explicitly disambiguate between a collective versus distributive interpretation in a clause with a plural (and often null) subject where such disambiguation is pragmatically preferred.

To summarize, in order to account for extrinsic plural marking in Korean as outlined above, at least the following features – some sort of distributive feature and an uninterpretable plural feature – are required:

Probably the closest morpholexical equivalent in English is each (of the), for which [distr] and [u-pl] are also required. The task for an English native speaker, then, will be to realize that the [distr] and [u-pl] features of a completely distinct morpholexical item in their native language must be recruited for reassembly onto the plural marker in Korean.

To summarize, again assuming Full Transfer, for the extrinsic plural, English native speakers acquiring L2 Korean must accomplish the following:

learn to notice the extrinsic plural marker on adverbs, locative postpositions and other non-nominal categories, and on direct or indirect object nouns that might be a mass rather than count noun (e.g. mul ‘water’), categories on which it might seem strange to a native English speaker to find a plural marker;

find (at least initially) some sort of morpholexical equivalent in English for the semantic interpretation of distributivity, such as ‘each (of the)’⁶;

learn the syntactic requirement that the pluralized element be in a clause with a plural c-commanding subject and that null subjects in a clause with the extrinsic plural must consequently be interpreted as plural;

acquire the correct affixation order with regard to case marking; that is, not confuse the affix-ordering with that of the intrinsic plural.

Learners must accomplish the above with very sparse input; extrinsic plural marking is not taught in Korean language classes, and is infrequent in the input: only 33 out of 1000 occurrences of -tul in a sample of spoken data (from the Sejong corpus) were of the extrinsic plural type, and only 5 out of 1000 in a written data sample (Hwang, in preparation; see also Kim, 2012).

Experimental study

This section describes a study designed to test whether native English speakers were able to acquire the features and conditions of Korean intrinsic and extrinsic plural marking as outlined above. Specifically, we wondered to what extent learners would assume that the plural marker -tul in Korean occurs in the same contexts and encodes the same features as plural marking in English, and at what developmental point (operationalized as L2 proficiency level) they would begin to tease apart and ‘reassemble’ the components associated with plural marking in Korean. Our feature-reassembly approach predicts that the features of Korean plural marking that diverge the most from English, and especially those that are derivationally most complex, that is embedded most deeply in the feature hierarchy shown in Figure 1, will be acquired latest (for similar suggestions, see e.g. Harley and Ritter, 2002: 498; Hawkins and Casillas, 2008; Lardiere, 2000: 122–124, 2009: 215; McCarthy, 2008, 2012). Following Lardiere (2008, 2009), determining the precise (phonological, morphosyntactic, semantic, pragmatic, and/or discourse) conditions under which a particular feature is expressed is a critical part of the feature assembly problem, and the complexity of those conditions, or the degree to which they differ from those of the L1, will contribute to the difficulty of the L2 learning task. We operationalized the notions ‘more difficult’ and ‘later-acquired’ as appearing only at higher levels of proficiency, if at all.

Predictions

In accordance with Full Transfer, we hypothesize that learners will initially assume that the Korean plural marker -tul is the morpholexical equivalent of English plural marking. For intrinsic plural marking, therefore, at initial or early stages of L2 development, we predict that English native speakers (NSs) will overuse the -tul plural marker, based on the obligatoriness of English plural-marking as well as their failure to distinguish between [q-rel] and [q-abs] features, a distinction that does not apply to plural-marking in English. There will be positive evidence in the input for pluralizing nouns with weak quantifiers [q-rel].⁷ However, negative evidence is required to ‘delearn’ that plural-marking with numeral+classifier [q-abs] constructions (for non-human nouns) is unacceptable in Korean.⁸ We also predict that early learners will fail to correctly interpret the specificity of the pluralized nouns they hear; more concretely, until they have assigned an uninterpretable [u-specific] feature to the plural marker, they will incorrectly accept the use of -tul in non-specific contexts and will produce it in both specific and non-specific contexts (as in English). Similarly, they will also likely fail to produce -tul on appropriate previous-mention plural referents.

At intermediate proficiency levels, we predict that, once learners have acquired the [q-abs] restriction on pluralization in numeral+classifier constructions, they will over-impose a blanket restriction on plural-marking in these constructions even though pluralization is allowed on numerically-quantified human nouns. In other words, they will fail to distinguish the [±human] aspect of this distinction, which prohibits plural marking on numerically-quantified non-human nouns, but allows it on numerically-quantified human nouns. At this point we also predict the appropriate co-occurrence of -tul in demonstrative ku/i contexts, as there is good positive evidence for this in the input. In other words, learners will have learned to associate plural-marking with certain co-occurring morpholexical items (demonstratives), which we interpret as the initial establishment of a feature-checking relation between the interpretable [specific] feature of D, and the assignment of an uninterpretable [u-specific] feature of the -tul plural marker. However, its felicitous production in previous-mention plural-referent discourse contexts will still be variable; previous research has suggested that conditions involving discourse-contextual information external to the grammar are acquired late and may constitute a source of persistent ‘residual optionality’ (Sorace, 2003, 2011).

Finally, at an advanced developmental stage, we predict the accurate production of plural-marking on (and restriction to) humans in numeral-classifier constructions, and the appropriate use and interpretation of the plural marker where licensed by a previous plural discourse referent.⁹

Regarding the extrinsic plural, we do not expect this to be acquired until an advanced stage of development, especially for classroom-instructed learners to whom it is rarely taught, as it requires the recruitment of a distributive feature from a completely different lexical item in English. When learners are taught it or notice it on non-nominal categories at all, we think that they will associate the -tul marker with plural number only, perhaps as a kind of plural-agreement marker.¹⁰ This view is supported by a recent preliminary study of the L1 acquisition of extrinsic plural marking, which found that even native Korean-speaking children (under the age of 8) associated the extrinsic plural marker with the feature [plural] only, whereas adults, as expected, demonstrated knowledge of both the distributive and plural interpretations; see Kim (2012). In other words, we think it possible that native English speakers will notice that the plural marker occurs in some ‘odd’ places in Korean compared to where the plural would be in English; but if so, they would treat it simply as an additional (uninterpretable) plural marker that indicates agreement with the plurality of the subject. It is only at very advanced stages that we expect L2 Korean learners to add the [distr] feature to the extrinsic plural marker.

2 Participants

The participants were 77 native-English-speaking learners of L2 Korean¹¹ and 31 native-Korean-speaking controls.¹² On the basis of their results on a Korean language proficiency C-test, the learners were divided into 4 proficiency groups: low-intermediate (LI), high-intermediate (HI), low-advanced (LA), and advanced (A) (Lee-Ellis, 2009).¹³ A one-way ANOVA confirmed that the four learner groups differed significantly from each other (Welch’s F(3, 30.439) = 154.910, p = .000). The results are presented in Table 1.

Table 1.

Scores on proficiency C-test (out of 200 points possible).

Group	Mean (Min.–Max.)	SD
Low-intermediate (n = 12)	65.7 (14–81)	20.61
High-intermediate (n = 30)	108.2 (92–123)	8.00
Low-advanced (n = 17)	134.3 (127–141)	4.42
Advanced (n = 18)	165.0 (146–185)	12.47

All 77 learners reported having formal university language instruction in Korean, and most of them were currently enrolled in Korean language classes (at various American and Korean universities) at the time of testing. Most learners had also been exposed to considerable naturalistic spoken Korean input in Korea. Participant information is summarized in Table 2.

Table 2.

Participants’ background information.

Groups	n (female/male)	Mean age at time of testing (range)	Length of stay in Korea in months^a	Mean age of first exposure to Korean^b
NS controls	31 (24/7)	27.80 (19–40)	n/a	n/a
Low-intermediate	12 (3/9)	24.75 (21–42)	18.1 (0–24)	19.7 (19–23)
High-intermediate	30 (3/27)	26.10 (20–65)	38.1 (0–420)	20.6 (17–29)
Low-advanced	17 (2/15)	24.76 (21–41)	47.5 (21–192)	19.5 (13–25)
Advanced	18 (1/17)	24.10 (21–29)	31.5 (7–72)	19.2 (16–24)

Notes: ^a Two participants in the low-intermediate group and one participant in the high-intermediate group had not resided in Korea. ^b There was one participant in the LA group whose age of first exposure was 13; the next lowest age of first exposure in this group was 18.

Materials and procedures

In order to test our predictions for knowledge of intrinsic and extrinsic plural marking, five types of tasks were designed: an elicitation task (for intrinsic -tul), an acceptability judgment task (for intrinsic -tul), a preference task (for extrinsic -tul), a truth value judgment task (for both intrinsic and extrinsic -tul), and a multiple-choice translation task (for both intrinsic and extrinsic -tul). All participants completed a consent form and background questionnaire and then took the proficiency test; they then randomly received one of three booklets with different orderings of the five tasks. All tasks were untimed, although participants were not allowed to go back to previous items once they had completed them. The instructions for each task were given in English for the native English speakers and in Korean for the native speaker controls. All Korean items were given in Hangul orthography, although we provide only the romanized versions and their glosses in the examples here. We describe each task in turn in the following subsections.

Elicitation task (for intrinsic -tul)

This task was developed for eliciting the intrinsic plural -tul. Participants were given 12 pictures including six distracters, which depicted situations in which a tourist was traveling in a foreign country. For each picture, participants were asked to formulate a question about the objects that the tourist is pointing at, using a small set of relevant vocabulary words given below each of the pictures. For example, in one picture the tourist is pointing at a large group of white houses clustered together on a hill and the vocabulary words provided were cip ‘house’ and hayahta ‘be white’. (The vocabulary words in the pictures were provided in Korean only.) The depiction of deictic pointing by the tourist in the pictures was intended to create a specific context in which the use of the plural form would be strongly preferred.

Three of the six test pictures included a numeral in their vocabulary set, requiring participants to produce a numeral+classifier construction. The purpose of this was to see if learners could produce the correct classifier for the nouns depicted. Because this article is not specifically concerned with learners’ acquisition of the correct classifier, and due to space limitations, we do not report on these three items here. For the remaining three test items, a large set of objects (as in the example mentioned above) was shown in each picture to discourage the use of a specific cardinal numeral and test whether learners produced plural-marking as required.

Acceptability judgment task (for intrinsic -tul)

The acceptability judgment task was designed to test if learners were aware of restrictions on the use of intrinsic -tul, such as the prohibition of pluralized non-human nouns with a numeric quantifier. For this task, participants were asked to judge 26 sentences (13 test items and 13 distracters) categorically as acceptable or unacceptable by marking ‘yes’ or ‘no’ after each item. They were also asked to provide a correction for any items they rated ‘no’. Test items reflected both target and non-target uses of -tul, comprising (a) the incorrect use of a numeral + classifier with a pluralized non-human noun (n = 2), (b) the incorrect use of a numeral alone with a pluralized noun (n = 2), (c) pluralization used incorrectly on the classifier itself (n = 2), (d) pluralization used incorrectly in nouns in clearly-defined non-specific contexts (n = 3), (e) the acceptable use of a numeral+classifier with a pluralized human noun (n = 2), and (f) the acceptable use of a non-numeric ‘weak’ quantifier with a pluralized noun (n = 2). A glossed sample test item – of type incorrect condition (a) above – is shown in (18):

Preference task (for extrinsic -tul)

The preference task was designed to measure learners’ knowledge of the fact that the extrinsic plural marker can be attached to various non-nominal elements in Korean, and of other syntactic restrictions on its use. For this task, participants were provided with 12 pictures (five test items,¹⁴ seven distracters) of people doing things and a contextual description (written in English for the L2 learner groups) for each picture along with four sentences written in Korean. They were asked to respond with ‘yes’ or ‘no’ to indicate which of the four sentences a person in the picture might felicitously say, given the context, and told that they could respond ‘yes’ to more than one sentence in each group if they felt there was more than one appropriate sentence. Of the four test sentences for each situation, one contained extrinsic -tul, one did not contain extrinsic -tul, and the other two were distracters. For example, one picture depicted a woman in an apron serving a roast turkey to three members of a family seated at a table. The contextual description read ‘The family is about to have dinner. The woman who made dinner is bringing it to them. As she puts it on the table, what would she say?’. Participants chose from the options listed in (19):

Note that, because the use of the extrinsic plural is optional, both (a) and (c) responses in the example above would be correct. The test items for this task included the following types of elements marked with extrinsic -tul: adverbs (n = 1), wh-words (n = 2), oblique (e.g. locative-postposition marked) nominals (n = 1), and verbal noun (n = 1).

Truth value judgment task (for intrinsic and extrinsic -tul)

A truth value judgment task was designed to test for knowledge of the requirement for using intrinsic -tul to denote plural referents in specific contexts, and knowledge of the distributive meaning of extrinsic -tul. For this task, learners were asked to read a short paragraph in English (NS controls read it in Korean) and to decide whether a subsequent Korean sentence was an appropriate sentence based on the paragraph they read. There were 30 items: 14 test items¹⁶; seven each for intrinsic and extrinsic plural, plus 16 distracters. Each plural category included three correct ‘yes’ responses and four correct ‘no’ responses. Sample (glossed) test items for intrinsic and extrinsic plural are shown in (20a–20b) respectively:

In (20a), the sentence without intrinsic plural marking on ‘child’ does not follow felicitously from the context, because the use of -tul is not optional if a noun has been established in a previous context as plural (Kwon and Zribi-Hertz, 2004). For (20b), the use of extrinsic -tul does not match the story in which only Sam, but not Cara, bought many clothes.

Translation task (for intrinsic and extrinsic -tul)

A multiple-choice translation task was designed to test if learners had acquired interpretations of both types of -tul. For this task, participants were asked to read 18 sentences in Korean (nine test items and nine distracters), each of which was followed by four interpretations in English, and then select which interpretation was the most appropriate for the preceding sentence. Four of the test items examined weak versus numeral+classifiers with intrinsic -tul and five tested a distributive interpretation for extrinsic -tul. For this task, only 12 of the Korean NSs controls – those who self-identified as highly proficient in both Korean and English (see note 12) – were included in the control group. Sample (glossed) test items for intrinsic and extrinsic plural are shown in (21a–21b) respectively:

4 Results

a Elicitation task results

We begin with the results from the elicitation task. As mentioned above, the elicitation task comprised two types of test items: one to see whether the participants could correctly produce -tul to refer to a plural group of entities in specific contexts, and the other to investigate if they could produce the correct classifier; the latter part is omitted here.

In scoring the three test items that required the obligatory use of -tul, one point was awarded for each target-like construction of plural-marking, and no points for missing or incorrect plural-marking. (Even if other parts of the sentence were not perfect, these were not considered for our scoring purposes.) Because participants were allowed to make up their own questions (using the vocabulary provided), any responses that completely failed to address the scene depicted in the picture were eliminated from analysis. Table 3 presents the mean percentage correct scores of each group for supplying -tul in an obligatory context.

Table 3.

Suppliance of intrinsic -tul in an obligatory specific context (percent correct)

Group	Mean (Min–Max)	SD
Low-intermediate (n = 12)	68.0 (0–100)	33.68
High-intermediate (n = 30)	84.4 (0–100)	25.88
Low-advanced (n = 17)	94.1 (33.3–100)	17.63
Advanced (n = 18)	92.6 (66.6–100)	14.24
Native controls (n = 31)	96.2 (50–100)	11.95

A one-way ANOVA indicated that there were significant differences among the five groups (F(4, 103) = 4.731, p = .002). Post-hoc pairwise comparisons indicated that the low-intermediate (LI) group’s performance was significantly lower than that of the native controls (p = .005) and the low-advanced (LA) and advanced (A) groups (p = .032 and .047, respectively). On the other hand, the other three learner groups did not significantly differ from the NS controls.

Individual analyses were also carried out. Since there were only three test items, participants who scored 0% (zero out of three correct) were classified as incorrect responders, those who scored 100% (three out of three correct) as correct responders, and those who provided one or two correct answers out of three as optional responders. Table 4 presents the individual data for each group categorized according to the response pattern. It shows the number of participants per category type and the percentage of the group representing that category type (in parentheses).

Table 4.

Number of participants per response pattern for obligatory use of -tul in elicitation task (percentages in parentheses).

Group	Incorrect (0%) (percentage of group)	Optional (33%–66%) (percentage of group)	Correct (100%) (percentage of group)	Total
Low-intermediate	1 (8.3)	6 (50.0)	5 (41.7)	12
High-intermediate	1 (3.3)	9 (30.0)	20 (66.7)	30
Low-advanced	0 (0)	2 (11.8)	15 (88.2)	17
Advanced	0 (0)	4 (22.2)	14 (77.8)	18
Native controls	0 (0)	3 (9.7)	28 (90.3)	31

The individual data appear to support the group data in that only the low-intermediate (LI) participants are split quite evenly between optional and correct response types; in contrast, most participants in the other three learner groups, along with the native speakers, fall into the ‘correct’ response type.

Acceptability judgment task results

In scoring the acceptability judgment task, one point each was awarded for grammatical items correctly judged as ‘yes’ and for ungrammatical sentences correctly judged as ‘no’, provided that the correction given by the participant to correct the ungrammaticality was relevant and appropriate. For example, for the ungrammatical test item shown in (22a), one participant provided the correction shown in (22b); however, although plural-marking was addressed, the revision itself was incorrect, thus no points were awarded for this particular item.

For (22), even though plural-marking was addressed in the correction, the provided revision ‘sikyey sey kay-tul’ was incorrect because -tul is not allowed on classifiers.

Items that were correctly judged as ungrammatical but for which the corrections provided were completely irrelevant to pluralization were excluded from analysis. In short, one point was given only if the participant judged a grammatical sentence as ‘yes’ or an ungrammatical one as ‘no’ and also fixed it in a correct way.

Group results for the acceptability judgment task are presented in Table 5, shown as the means of participants’ percent correct answers out of total test items. The acceptability judgment data present a clear developmental pattern with increasing proficiency: the high-intermediate (HI) group performed better than the low-intermediate group (LI), and the advanced (A) learners were better than the intermediate (LI and HI) and low-advanced (LA) learners. All learner groups differed significantly from the native controls. A one-way ANOVA indicated significant differences among the groups (Welch’s F (4, 41.442) = 99.482, p = .000). Post-hoc pairwise comparisons indicated that scores for the LI group were significantly lower than those for the HI group (p = .018), the group (p = .000), and the native controls (p = .000). The HI group scores were significantly lower than those of the A group (p = .007) and the native controls (p = .000). The scores for the LA group were significantly lower than those for A group (p = .017) and the native controls (p = .000), and the scores for the A group were significantly lower than those for the native controls (p = .001).

Table 5.

Acceptability judgment task (percent correct).

Group	Mean (Min.–Max.)	SD
Low-intermediate (n = 12)	32.5 (25.0–55.5)	8.18
High-intermediate (n = 30)	44.9 (25.0–83.3)	16.34
Low-advanced (n = 17)	44.6 (18.2–76.9)	16.58
Advanced (n = 18)	63.9 (33.3–100)	17.74
Native controls (n = 31)	85.2 (61.5–100)	8.89

Although all learner group means differed significantly from the native controls, the group results obscure the possibility that some individual learners performed in a native-like way. Therefore, individual analyses were conducted. Similar to the individual analysis for the elicitation task, participants were categorized according to whether they fell into one of three response-type categories: incorrect (less than 30% correct), optional (30%–70% correct), and correct (more than 70% correct). Table 6 summarizes the distribution of response patterns for each group.

Table 6.

Number of participants per response pattern (acceptability judgment task) (percentages in parentheses).

Group	Incorrect (<30%) (percentage of group)	Optional (30%–70%) (percentage of group)	Correct (>70%) (percentage of group)	Total
Low-intermediate	5 (41.7)	7 (58.3)	0 (0)	12
High-intermediate	3 (10.0)	24 (80.0)	3 (10.0)	30
Low-advanced	4 (23.5)	12 (70.6)	1 (5.9)	17
Advanced	0 (0)	11 (61.1)	7 (38.9)	18
Native controls	0 (0)	1 (3.2)	30 (96.8)	31

The individual data show that only the LI group failed to produce any learners within the ‘correct’ response pattern. Seven of the advanced learners fell into the ‘correct’ response type, and indeed one of these learners provided 100% correct responses, suggesting that eventual native-like attainment is at least possible.

Preference task results

Participants were awarded one point for each appropriate selection of extrinsic plural marking. However, for any test item in which a learner marked ‘yes’ on all four sentences, including the distracters, those items were excluded from analysis, because this might have been the result of a tendency to simply respond ‘yes’, or carelessness in reading the sentences, rather than a reflection of having acquired knowledge regarding extrinsic -tul.¹⁷ Group mean scores for the preference task are presented in Table 7, as percent correct.

Table 7.

Preference task (percent correct).

Group	Mean (Min.–Max.)	SD
Low-intermediate (n = 12)	35.0 (0–80)	30.89
High-intermediate (n = 30)	39.3 (0–100)	36.94
Low-advanced (n = 17)	34.4 (0–100)	36.48
Advanced (n = 18)	58.6 (0–100)	37.09
Native controls (n = 31)	95.5 (60–100)	11.20

A one-way ANOVA revealed that there were significant differences among the five groups (Welch’s F(4, 35.403) = 34.592, p = .000). Post hoc pairwise comparisons indicated that all learner groups significantly differed from the native controls (for LI, p = .000; for HI, p = .000; for LA, p = .000; for A, p = .005), suggesting that the learners had trouble recognizing the use of extrinsic plural marking. None of the learner groups differed significantly from each other; although the advanced group performed better than the other three learner groups, that difference did not achieve statistical significance.

Similar to the other tasks, individual results were analyzed. As in the Acceptability Judgment Task, three categories were used (incorrect for below 30%, optional for between 30% and 70%, and correct for over 70%), and the participants were classified into one of the three categories. The distribution of response types is provided in Table 8. As Table 8 illustrates, the general tendency is for the rate of correct responses to increase with proficiency, while the rate of incorrect and optional response patterns decreases. In the A group, half of learners fall into the correct response pattern, while in the LI group only around 16% of learners are classified into the correct response pattern. This suggests that the learners became more target-like with proficiency in performance with respect to extrinsic -tul, indicating that successful acquisition is not impossible. This suggestion is supported by the observation that four learners in the HI group, two learners in the LA group, and five learners in the A group provided 100% correct responses for the task.

Table 8.

Number of participants per response pattern (preference task) (percentages in parentheses).

Groups	Incorrect (<30%)	Optional (30–70%)	Correct (>70%)	Total
Low-intermediate	5 (41.7)	5 (41.7)	2 (16.6)	12
High-intermediate	15 (50.0)	6 (20.0)	9 (30.0)	30
Low-advanced	10 (58.8)	3 (17.6)	4 (23.5)	17
Advanced	4 (22.2)	6 (27.8)	9 (50.0)	18
Natives	0 (0)	2 (6.5)	29 (93.5)	31

Truth value judgment task results

For the truth value judgment task, one point was awarded for correctly accepting or rejecting a test sentence. We distinguish here between the results for intrinsic and extrinsic plural marking, both of which were tested in this task. Table 9 presents the mean scores for each group for the intrinsic plural. Since none of the data were excluded from analysis and the total of test items was the same for each participant, we report the mean score for number of correctly answered items, out of seven test items.

Table 9.

Intrinsic plural group mean scores on truth value judgment task.

Group	Mean (Min.–Max. = 7)	SD
Low-intermediate (n = 12)	3.25 (2–4)	0.622
High-intermediate (n = 30)	3.70 (2–6)	0.988
Low-advanced (n = 17)	4.24 (2–7)	1.437
Advanced (n = 18)	4.17 (2–7)	1.339
Native controls (n = 31)	5.19 (3–7)	1.195

A one-way ANOVA indicated that there were significant differences among the groups (F(4, 103) = 9.042, p = .000). Post hoc pairwise comparisons indicated that both intermediate groups (LI and HI) differed significantly from the native controls (p = .000 and p = .000, respectively). The LA and A groups did not differ significantly from any other groups, including the natives.

As before, we analysed the individual responses, dividing them into three response-types: correct, incorrect, and optional. Participants who answered at least five out of seven items correctly were classified as ‘correct’ responders, those who answered three or four items correctly were classified as ‘optional’ responders, and those who answered fewer than three items correctly were classified as ‘incorrect’ responders. These results are shown in Table 10.

Table 10.

Number of participants per response pattern (intrinsic -tul, truth value judgment task) (percentages in parentheses).

Group	Incorrect (0–2) (percentage of group)	Optional (3–4) (percentage of group)	Correct (5–7) (percentage of group)	Total
Low-intermediate	1 (8.3)	11 (91.7)	0 (0)	12
High-intermediate	2 (6.7)	21 (70.0)	7 (23.3)	30
Low-advanced	2 (11.8)	9 (52.9)	6 (35.3)	17
Advanced	2 (11.1)	11 (61.1)	5 (27.8)	18
Native controls	0 (0)	11 (35.5)	20 (64.5)	31

The individual results appear consistent with the group means, in that the LA group showed more correct responses than the other learner groups, and most participants for all four learner groups appear to fall into the ‘optional’ category. However, it should be noted that the responses of one learner in the LA group, and two learners in the A group were 100% accurate.

Table 11 presents the group results for the test items involving extrinsic -tul. A one-way ANOVA indicated a significant difference among the groups (F(4, 103) = 4.162, p = .004); post hoc Scheffé analysis revealed that the HI learners differed significantly from the native controls (p = .009); otherwise, there were no significant differences. A post hoc Tukey analysis indicated a marginally significant difference between the LI group and the natives (p = .092) as well as between the HI group and the natives (p < .01).

Table 11.

Extrinsic plural group mean scores on truth value judgment task.

Group	Mean (Min.–Max. = 7)	SD
Low-intermediate (n = 12)	3.92 (3–6)	0.900
High-intermediate (n = 30)	3.80 (2–6)	1.064
Low-advanced (n = 17)	4.29 (3–6)	1.213
Advanced (n = 18)	4.00 (2–7)	1.328
Native controls (n = 31)	4.87 (3–7)	1.024

An analysis of individual results was conducted similar to that for the intrinsic plural. The individual extrinsic plural results are presented in Table 12. The individual results indicate that the rate of correct response types increases with proficiency, from about 17% for the LI group to 41% for the LA group and 39% for the A group. (We note that the LA group had more correct responders than the A group, though we have no good explanation for it; the same result was obtained for the group means, although these groups did not significantly differ.) Only one learner, in the A group, scored 100% accurately. These results suggest that it is ultimately possible, although likely very difficult, for native English speakers to attain knowledge of the distributive feature of extrinsic -tul.

Table 12.

Number of participants per response pattern (extrinsic -tul, truth value judgment task) (percentages in parentheses).

Group	Incorrect (0–2) (percentage of group)	Optional (3–4) (percentage of group)	Correct (5–7) (percentage of group)	Total
Low-intermediate	0 (0)	10 (83.3)	2 (16.7)	12
High-intermediate	4 (13.3)	20 (66.7)	6 (20.0)	30
Low-advanced	0 (0)	10 (58.8)	7 (41.2)	17
Advanced	3 (16.7)	8 (44.4)	7 (38.9)	18
Native controls	0 (0)	10 (32.3)	21 (67.7)	31

e Multiple-choice translation task results

The translation task results for the four intrinsic plural test items were discarded after it became clear that the task was too easy and all participants scored at ceiling. Therefore, only the results for the extrinsic plural test items are reported here. One point was awarded if the correct answer was selected. Table 13 presents the mean scores for each group. Since none of the extrinsic plural translation task data were excluded from analysis and the total of test items was the same for each participant, we report the mean score for number of correctly answered items, out of five test items.

Table 13.

Extrinsic plural group mean scores on translation task.

Group	Mean (Min.–Max. = 5)	SD
Low-intermediate (n = 12)	3.50 (1–5)	1.446
High-intermediate (n = 30)	3.07 (0–5)	1.741
Low-advanced (n = 17)	3.24 (0–5)	1.562
Advanced (n = 18)	4.22 (2–5)	0.943
Native controls (n = 12)	4.67 (3–5)	0.651

A one-way ANOVA revealed significant differences among the groups (Welch’s F(4, 36.979) = 6.454, p = .000). Post hoc pairwise comparisons indicated that the HI group performed significantly worse than the A group and the NS controls (p = .035 and p = .001, respectively). The LA group performed significantly worse than the NS controls (p = .02). The A group did not significantly differ from the natives. Surprisingly, however, there was also no difference between the LI group and the natives. To check individual performance, we again categorized the groups according to whether their answers were consistently correct (= four or five out of five correct), incorrect (= zero or one out of five correct), or indicated optionality (= two or three out of five correct). Those results are presented in Table 14.

Table 14.

Number of participants per response pattern (extrinsic -tul, translation task) (percentages in parentheses).

Group	Incorrect (0–1)	Optional (2–3)	Correct (4–5)	Total
Low-intermediate	1 (8.3)	4 (33.0)	7 (58.3)	12
High-intermediate	7 (23.3)	9 (30.0)	14 (46.7)	30
Low-advanced	2 (11.8)	7 (41.2)	8 (47.0)	17
Advanced	0 (0)	4 (22.2)	14 (77.8)	18
Native controls	0 (0)	1 (8.3)	11 (91.7)	12

The individual results show that there is quite a large jump in accuracy (i.e. in the percentage of individuals who place in the ‘correct’ response pattern) among the advanced group.

Discussion

For Korean intrinsic plural marking, the results generally show increasingly target-like performance as proficiency increases. We predicted that lower-proficiency learners would not yet have acquired knowledge of the uninterpretable [u-specific] feature required for the intrinsic plural, but that this feature would be acquired as proficiency increased. Results from the elicitation task showed that only the lowest-proficiency (LI) learner group statistically differed from the native controls. The truth value judgment task also tested for knowledge of specificity, and, in this task, neither the LA nor A groups differed from the NS controls, again showing development with increasing proficiency. Results from those test items on the acceptability judgment task that tested knowledge of specificity similarly showed development with increasing proficiency; the A group did not significantly differ from the NS controls on the three items testing for knowledge of specificity on this task. These findings suggest to us that English native speakers can eventually successfully incorporate the [u-specific] feature into their representation of intrinsic plural-marking in Korean.

For quantified NPs, we hypothesized that learners at lower proficiency levels would overgeneralize plural marking on quantified NPs. This prediction appears to be supported. Table 15 presents a breakdown of the quantified NP test items on the acceptability judgment task. We see that the lowest-proficiency (LI) learners overwhelmingly accepted all pluralized nouns, accepting those that were grammatically pluralized, but failing to reject those that were ungrammatically pluralized. The rate of correct rejection of unacceptable plural items rises gradually with proficiency, and it appears that only the A group more reliably distinguishes [q-rel] from [q-abs]. Most interesting, however, is the dip in the A group’s correct acceptance of grammatical number+classifier pluralized human nouns, which exactly matches its (correct) rejection rate of ungrammatical number+classifier pluralized non-human nouns (as shown in boldface type in Table 15). This result suggests to us that, as predicted, some A group learners have acquired the [q-rel] versus [q-abs] distinction but have also adopted an overgeneral blanket ‘avoidance’ strategy prohibiting pluralization for all numerically-quantified nouns, regardless of whether they are human or not. That is, they have apparently not yet acquired the next-lower division of the feature-restriction hierarchy for pluralizing [q-abs] classified NPs: [+human] versus [–human], thereby supporting the developmental predictions of the hierarchy presented earlier in Figure 1, repeated as Figure 2. Nonetheless, recall from Table 6 that seven A participants scored in the ‘correct’ response range for this task (and one of these scored 100% accurately), suggesting that eventual native-like attainment is at least possible for some learners.

Table 15.

Mean percentage correct for quantified NPs in acceptability judgment task.

Group	Ungrammatical		Grammatical
Group	Num+CL+tul/–hum	Num+tul/–human	Num+CL+tul/+hum	WQ+tul
Low-intermediate	12.5	12.5	87.5	100.0
High-intermediate	36.7	38.3	84.5	98.2
Low-advanced	44.1	47.1	85.3	86.7
Advanced	69.4	72.2	69.4	91.7
Native controls	88.7	87.1	98.4	100.0

Figure 2.

Feature hierarchy for pluralizing quantified noun phrases in Korean.

We next turn to whether native English speakers acquiring Korean are aware of and able to correctly interpret the extrinsic plural. Our main objective was to test for knowledge of the [distr] interpretation associated with extrinsic -tul. The preference task, truth-value judgment task, and multiple-choice translation task all tested this. As predicted, the extrinsic plural proved to be much more difficult for native English speakers, and it is not hard to see why. Its use is optional, it occurs only rarely in the input, and it occurs on a suffix that native English speakers (and apparently even L1 Korean children, at least up until the age of eight; see Kim, 2012) associate with a completely different grammatical function, namely, plural marking on referentially plural nouns. Its distributive function is absent from plural marking in English although, if we assume full transfer, is available via the recruitment of the [distr] feature from other morpholexical sources in the L1, such as the lexical item each. In this respect, the results from the multiple-choice translation task, which was admittedly likely to be the easiest of the three tasks mentioned above, are nonetheless informative because participants who performed well on this task necessarily selected an appropriate English translation that included the lexical item each (of the).¹⁸ The advanced group did not significantly differ from the native controls on this task, suggesting that such feature-recruitment is possible. The finding that only a relatively small number of individual learners (only one learner on the truth-value judgment task and 11 learners on the preference task) scored completely accurately attests to the difficulty of the construction for English NSs.¹⁹ Nonetheless, similar to the findings for the intrinsic plural, we found that learners’ performance improved with increasing proficiency, suggesting that eventual native-like attainment is possible in principle.

In comparing the acquisition of intrinsic versus extrinsic -tul, our results suggest that learners do search for the closest morpholexical counterpart from their L1 (or other prior language knowledge). Our participants apparently recognized the intrinsic plural as having essentially the same grammatical function as plural-marking in English, albeit with more complex featural co-occurrence restrictions. These restrictions were eventually added as required by the most advanced proficiency group. Acquiring the extrinsic plural, on the other hand, requires extending the use of the ‘plural’ morpheme in Korean to categories such as adverbs and postpositional phrases that could never be pluralized in the L1 and associating it with a grammatical function that is situated on a completely different morpholexical item (or set of items) in the L1. The features themselves are present in both languages; however, their grammaticalized distribution on lexical items in each language is strikingly different.

Here we might also point out that theories that propose that acquiring new uninterpretable features is impossible in adult second language acquisition, such as the Interpretability Hypothesis (e.g. Franceschina, 2005; Hawkins and Hattori, 2006; Tsimpli and Dimitrakopoulou, 2007; Tsimpli and Mastropavlou, 2008), would fail to account for our findings. For example, the uninterpretable [u-specific] feature associated with Korean but not English plural marking is not only successfully acquired, but its acquisition also generally occurs at intermediate to advanced proficiency levels, or at any rate earlier than the acquisition of the interpretable distributive property associated with extrinsic plural marking. The irrelevance of (un)interpretability to ultimate attainment comports with findings of other studies, such as that of White et al. (2004) – who found that L1 English and French advanced proficiency learners of L2 Spanish were able to acquire number and gender agreement features to native-like levels, and Papadopoulou et al. (2011) – who investigated the acquisition of case and verbal agreement marking in L2 Turkish by native Greek speakers, as well as the interaction between Turkish accusative case marking, specificity, and object scrambling. (All the relevant features in question were uninterpretable.) The latter study found that verbal agreement features were reliably acquired by higher-proficiency learners, whereas the requirement for accusative case marking on scrambled specific objects was not. They attribute their findings to the particular morphological differences between Greek and Turkish with respect to the greater complexity of the case system of the latter, including the specificity feature associated with accusative marking in Turkish but not Greek, concluding that features that are grammaticalized or encoded differently in the L1 than the L2 will be more difficult to acquire. Such findings thus suggest the usefulness and plausibility of framing the acquisition of functional morphology in terms of the feature-assembly of lexical items, especially as a way to further refine the construct of L1 influence (and the Full Transfer/Full Access hypothesis) in second language acquisition.

Footnotes

Acknowledgements

The order of authors is alphabetical. We are deeply grateful for helpful feedback on this research and/or assistance in recruiting study participants to the following people: the editor and three anonymous reviewers, Héctor Campos, Jung Yoon Choi, Wonil Chung, Lewis Gebhardt, Chae-Eun Kim, Hyun Jeong Kim, Young-Key Kim-Renaud, Ruth Kramer, Eunji Lee, Danica MacDonald, Yoonju Nah, Jong Un Park, Kamil Ud Deen, and audiences at the University of Calgary (especially Susanne Carroll and Betsy Ritter), and the 2011 BUCLD, where earlier versions of this talk were presented by the second author. Special thanks to Mark A Peterson and all the faculty and study participants in the Korean language programs at Brigham Young University, USA and Sogang University, Korea. All remaining errors are our own.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

References

(2007) Korean tul and English all. LSO Working Papers in Linguistics 7: Proceedings of WIGL, pp. 1–16.

Borer

(2005) In Name Only: Structuring Sense: vol 1. Oxford: Oxford University Press.

Brisson

(1998) Distributivity, maximality and floating quantifiers. Unpublished doctoral dissertation, Rutgers University, New Brunswick, NJ, USA.

Carroll

(2001) Input and evidence: The Raw Material of Second Language Acquisition. Amsterdam: John Benjamins.

Chierchia

(1998) Reference to kinds across languages. Natural Language Semantics 6: 339–405.

Cho

Y-M

Lee

Schulz

Sohn

H-M

Sohn

S-O

(2001) Integrated Korean: Intermediate 2. Honolulu, HI: University of Hawaii Press.

Cho

Y-M

Lee

Schulz

Sohn

H-M

Sohn

S-O

(2010) Integrated Korean: Beginning 1. 2nd ed. Honolulu, HI: University of Hawaii Press.

Chomsky

(2001) Derivation by phase. In: Kenstowicz

(ed.) Ken Hale: A Life in Language. Cambridge, MA: MIT Press, pp. 1–52.

Chung

(2003) Dummy -tul in gapping construction. Proceedings of the Spring Korean Generative Grammar Circle. Daejeon: Chungnam National University, pp. 75–83.

10.

Chung

(2004) Multiple dominance analysis of right node raising constructions. Language Research 40: 791–812.

11.

DeKeyser

(2005) What makes learning second-language grammar difficult? A review of issues. Language Learning 55 (supplement 1): 1–25.

12.

Franceschina

(2005) Fossilized Second Language Grammars: The Acquisition of Grammatical Gender. Amsterdam: John Benjamins.

13.

Gebhardt

(2009) Numeral classifiers and the structure of DP. Unpublished doctoral dissertation, Northwestern University, Evanston, IL, USA.

14.

Harley

Ritter

(2002) Person and number in pronouns: A feature-geometric analysis. Language 78: 482–526.

15.

Hawkins

Casillas

(2008) Explaining frequency of verb morphology in early L2 speech. Lingua 118: 595–612.

16.

Hawkins

Hattori

(2006) Interpretation of English multiple wh- questions by Japanese speakers: A missing uninterpretable feature account. Second Language Research 22: 269–301.

17.

Hoeksema

(1996) Floating quantifiers, partitives and distributivity. In: Hoeksema

(ed.) Partitives: Studies on the Syntax and Semantics of Partitive and Related Constructions. Berlin: Mouton de Gruyter, pp. 57–106.

18.

Hwang

(in preparation) The acquisition of Korean plural-marking by native English speakers. Unpublished doctoral dissertation, Georgetown University, Washington, DC.

19.

Joh

Y-K

(2005) A semantic analysis of the Korean plural marker tul. In: Maier

Bary

Huitink

(eds) Proceedings of SuB9. Nijmegen: Nijmegen Centre for Semantics, pp. 170–82.

20.

Kim

C-E

(2012) Acquisition of the distributive meaning of EPM in Korean. Working Papers in Linguistics: University of Hawai’i at Manoa 43: 1–10.

21.

Kim

(2005) The Korean plural marker tul and its implications. Unpublished PhD dissertation, University of Delaware, Newark, DE, USA.

22.

Kim

(1994) A non-spurious account of ‘spurious’ Korean plurals. In: Kim-Renaud

Y-K

(ed.) Theoretical issues in Korean linguistics. Stanford, CA: Center for the Study of Language and Information, pp. 303–323.

23.

Kim-Marshall

(2008) Step by Step: Korean 1 through 15 Action Verbs. Seoul: Hankwuk.

24.

Kwon

Zribi-Hertz

(2004) Number from a syntactic perspective: Why plural marking looks ‘truer’ in French than in Korean. In: Bonami

Cabredo Hofherr

(eds) Empirical Issues in formal syntax and semantics 5: 133–158.

25.

Lardiere

(2000) Mapping features to forms in second language acquisition. In: Archibald

(ed.) Second language Acquisition and Linguistic Theory. Malden, MA: Blackwell, 102–29.

26.

Lardiere

(2008) Feature-assembly in second language acquisition. In: Liceras

Zobl

Goodluck

(eds) The Role of Formal Features in Second Language Acquisition. Mahwah, NJ: Lawrence Erlbaum, pp. 106–140.

27.

Lardiere

(2009) Some thoughts on a contrastive analysis of features in second language acquisition. Second Language Research 25: 173–227.

28.

Lee-Ellis

(2009) The development and validation of a Korean c-test using Rasch analysis. Language Testing 26: 245–274.

29.

McCarthy

(2008) Morphological variability in the comprehension of agreement: An argument for representation over computation. Second Language Research 24: 459–486.

30.

McCarthy

(2012) Modeling morphological variation and development: Person and number in L2 Spanish. Linguistic Approaches to Bilingualism 2: 25–53.

31.

Moon

(1995) Plural marker -tul is subject to Principle A. Harvard Studies in Korean Linguistics 6: 355–369.

32.

Nemoto

(2005) On mass denotations of bare nouns in Japanese and Korean. Linguistics 43: 383–413.

33.

Noh

(2008) A study on the meaning of Korean zero-marked and ‘tul’-marked nouns. Discourse and Cognition 15: 43–62.

34.

Papadopoulou

Varlokosta

Spyropoulos V (2011) Case morphology and word order in second language Turkish: Evidence from Greek learners. Second Language Research 27: 173–205.

35.

Park

(2009) Distributive effects of the plural marker -tul in Korean. In: Leow

Campos

Lardiere

(eds) Little words: Their History, Phonology, Syntax, Semantics, Pragmatics and Acquisition. Washington, DC: Georgetown University Press.

36.

Park

M-K

Sohn

K-W

(1993) A minimalist approach to plural marker licensing in Korean. MIT Working Papers in Linguistics 20: 193–208. Cambridge, MA: MITWPL.

37.

Schmitt

Miller

(2010) Using comprehension methods in language acquisition research. In: Unsworth

Blom

(eds) Experimental Methods in Language Acquisition Methodology. Amsterdam: John Benjamins, pp. 35–56.

38.

Schwartz

(1993) On explicit and negative data effecting and affecting ‘competence’ and ‘linguistic behavior’. Studies in Second Language Acquisition 15: 147–163.

39.

Schwartz

Sprouse

(1996) L2 cognitive states and the Full Transfer/Full Access model. Second Language Research 12: 40–72.

40.

Song

(1997) The so-called plural copy in Korean as a marker of distribution and focus. Journal of Pragmatics 27: 203–224.

41.

Song

(1988) A ubiquitous plural marker. Korea Research Monograph, no. 14: Explorations in Korean syntax and semantics. University of California at Berkeley, Institute of East Asian Studies and Center for Korean Studies.

42.

Sorace

(2003) Near-nativeness. In: Doughty

Long

(eds) The Handbook of Second Language Acquisition. Malden, MA: Blackwell, pp. 130–151.

43.

Sorace

(2011) Pinning down the concept of ‘interface’ in bilingualism. Linguistic Approaches to Bilingualism 1: 1–33.

44.

Tsimpli

Dimitrakopoulou

(2007) The interpretability hypothesis: Evidence from wh- interrogatives in second language acquisition. Second Language Research 23: 215–42.

45.

Tsimpli

Mastropavlou

(2008) Feature interpretability in L2 acquisition and SLI: Greek clitics and determiners. In: Liceras

Zobl

Goodluck

(eds) The Role of Formal Features in Second Language Acquisition. New York: Lawrence Erlbaum, 142–183.

46.

White

Valenzuela

Kozlowska-Macgregor

Leung

Y-KI

(2004) Gender and number agreement in nonnative Spanish. Applied Psycholinguistics 25: 105–133.

47.

Wilder

(1997) Some properties of ellipsis in coordination. In: Alexiadou

Hall

(eds) Studies on Universal Grammar and Typological Variation. Amsterdam: John Benjamins, pp. 59–107.

48.

Wilder

(1999) Right node raising and the LCA. In: Brid

Carmie

Jaugen

Norquest

(eds) WCCFL 18 Proceedings. Somerville, MA: Cascadilla Press, pp. 586–598.

49.

Yim

(2002) The semantics of non-nominal tul in Korean. Journal of the International Circle of Korean Linguistics 11: 183–202.

50.

Yim

(2003) Subject agreement in Korean: Move F, Attract F, or Agree? In: Griffin

(ed.) The Role of Agreement in Natural Languages: TLS 5 Proceedings. Texas Linguistic Forum, 53, pp. 147–156.