The acquisition of different types of definite noun phrases in L2-English

Introduction

In recent years, there has been much research and discussion on syntax-semantics mismatches in second language (L2) acquisition (Dekydtspotter & Sprouse, 2001; Hawkins & Hattori, 2006; Marsden, 2009; Slabakova, 2003). A syntax-semantics mismatch refers to the situation when some universal meaning (e.g., definiteness, plurality, past event) is expressed differently in the native language (L1) and L2. Learners arguably have access to the whole arsenal of universal conceptual meanings (Jackendoff, 2002), but they have to learn how to express the meaning in the new language. This is a non-trivial task since these meanings often appear in different feature combinations (to be elaborated on below). Acknowledging this significant acquisition challenge, Lardiere (2008, 2009) proposes the Feature Reassembly Hypothesis which views L2 acquisition as involving the (re)assembly of the formal feature matrices of functional categories and their mapping onto new morphological exponents. The process consists of two steps: first, an initial mapping of the complete feature set of the L1 morpheme onto the target morpheme based on perceived similarities between the functional meanings of the L1 and L2 morphemes and second, ‘feature reassembly,’ where features can be added or deleted, progressively adjusting the target feature set according to input and usage. Such feature reassembly may be slow to occur or may not occur at all if the relevant evidence for the formal or semantic feature is rare or contradictory in the linguistic input. Likewise, the L1 can either facilitate or hinder feature reassembly. A number of recent studies have explored this approach to investigate learning problems related to syntax-semantics mismatches (Cho, 2012; Cho & Slabakova, 2014, 2015; Gil & Marsden, 2013; Hwang & Lardiere, 2013; Spinner, 2013; Yuan & Zhao, 2011). The present study continues to examine L2 learnability from this perspective by investigating the acquisition of the English definite article (the) by L1-Korean speakers. The goal of the present study is to investigate how L1-Korean speakers (re)assemble the appropriate values of two semantic features (definiteness and anaphoricity) that are bundled onto the definite article in English.

Second language acquisition of English articles

Recent years have seen a lot of interest and much research in the L2 acquisition of English articles by speakers whose L1s lack articles such as Russian, Korean, Japanese, or Chinese (e.g., Huebner, 1983; Ionin et al., 2004, 2012; Snape, 2008; Thomas, 1989; Trenkic, 2008). Previous research on the L2 acquisition of English articles has established that article-less L1 learners persistently omit or misuse articles even at an advanced level. A number of proposals have been made to account for such error patterns. For example, Trenkic (2008) claims that languages without overt articles lack the functional category Determiner, which licenses definiteness, thus leading speakers of article-less L1s to misanalyze English articles as adjectives. Goad and White (2004, 2008) propose that article-less L1 speakers’ errors in article production (omission, in particular) can be explained by L1 prosodic transfer. Ionin et al. (2004) propose the Article Choice Parameter with two settings (definiteness vs specificity settings) in languages with overt articles and argue that article-less L1 learners fluctuate between the two settings, that is, L2 learners’ article choice is influenced by specificity. Ionin, Baek, Kim, Ko, and Wexler (2012) argue that L1 transfer from the semantics of demonstratives prompts Korean speakers’ misinterpretation of definite descriptions as demonstrative descriptions. The present study contributes to this growing area of research in article acquisition by investigating L1-Korean learners’ interpretation of L2-English articles from another perspective, namely, the feature reassembly approach.

English overtly expresses definiteness through articles but does not distinguish different types of definite NPs. The definite article is used to indicate [+definite] in both anaphoric and non-anaphoric definite contexts which means that there are two possible feature combinations: [+definite, +anaphoric] and [+definite, −anaphoric]. However, cross-linguistic data show that some languages differentiate between the two different kinds of feature combinations: anaphoric ([+definite, +anaphoric]) and non-anaphoric ([+definite, −anaphoric]) (strong-article definites and weak-article definites, respectively in Schwarz’s terms) by marking them with two different definite articles (e.g., Fering) or encoding only one type of definite NP while leaving the other type in bare form (e.g., Akan) (Schwarz, 2009, 2013; see also Lyons, 1999). As for languages without articles, demonstratives perform some or all of the functions of the definite article. As discussed in Chang (2009), Korean uses the demonstrative ku ‘that’ only for anaphoric definite NPs. That is, Korean differentiates between two types of definite NPs (two feature combinations) by morphologically marking only [+definite, +anaphoric], while English does not distinguish between them and marks both with the same morpheme the. This cross-linguistic variation, combined with Lardiere’s feature-based process of L2 acquisition, allows us to identify the learning task for Korean-speaking learners of English and formulate concrete predictions for Korean-speaking learners’ developmental patterns in the acquisition of L2-English articles.

As described in the previous section, the feature-based L2 acquisition process predicts that learners will do an initial mapping of the L1 morpheme’s features onto the L2 morpheme based on perceived similarities between functional meanings of the L1 and L2 morphemes. For L1-Korean speakers, the features of the demonstrative ku, [+definite, +anaphoric], are predicted to be initially mapped to the. Once some initial mapping is established, the next step involves feature re-assembly, in which the target feature set is adjusted by adding or deleting appropriate features based on evidence from the input. In the case of L1-Korean speakers learning English, this involves adding the [+definite, –anaphoric] feature setting. This study investigates L1-Korean speakers’ interpretation of different types of L2-English definite NPs that appear in the same morphological form (the + NP) by measuring learners’ interpretation of different types of definite NPs (anaphoric and non-anaphoric definite NPs). In the next section I will provide a descriptive overview of different types of definite NPs cross-linguistically, that is, types of definite NPs that carry different features that co-occur with the feature [definite].

Different types of definites

Expressions of different types of definite NPs in Korean

As illustrated in the examples (1–7), English does not distinguish between different types of definites, that is, both anaphoric and non-anaphoric definite NPs are marked with the morpheme the. In contrast, Korean is one of many languages that lack articles. However, demonstratives in article-less languages tend to take over some or all functions of the definite article (Chang, 2009; Cho, 1999; Mendoza, 2005, 2011). In Korean, ku ‘that’ signals anaphoric definiteness (Chang, 2009).

As shown in Table 1, the distribution of the demonstrative ku in Korean is similar to the English demonstrative that. Both that and ku can appear in anaphoric contexts, as in (10) and neither can appear in non-anaphoric contexts, as in (11).

OPEN IN VIEWER

Table 1.

Definiteness marking in English and Korean (adapted from Chang, 2009, p. 48).

	English	Korean
Anaphoric definite NPs [+definite, +anaphoric]	the/that	?bare NP/ku
Non-anaphoric definite NPs [+definite, −anaphoric]	the/#that	bare NP/# ku

In (10), ku is used anaphorically to refer to the book mentioned in the first sentence. Note that both the and that sound natural in English. In (11), the demonstrative NP (that bride) does not sound natural since the definite NP the bride is not anaphoric. There is no preceding linguistic expression (either explicit or implicit) that the definite NP can refer to but the bride is still more appropriate than a bride or that bride since our general understanding is that there is a unique bride at a wedding. This non-anaphroic bridging definite NP, as in (11), cannot take the demonstrative ku in Korean.

Before I further analyze whether ku is an anaphoric definite determiner (like the) or a demonstrative determiner (like that), it is crucial to first discuss whether the use of ku is obligatory or optional with anaphoric definite NPs in Korean. Article use in English often determines the grammaticality of a sentence. For instance, sentences with countable singular nouns in bare form are ungrammatical (e.g., * I bought book). However, a bare NP is the default form in Korean. Any type of noun can appear in bare form and can be interpreted as either definite or indefinite. In other words, the use of ku with anaphoric definite NPs is not necessary, nor does it affect grammaticality. However, it determines the naturalness of sentences and affects the coherence of the discourse as English articles often do. ² Bare NPs in Korean are ambiguous between definite and indefinite as well as between generic and non-generic, while ku+NP overtly signals a definite interpretation. Thus, the use of ku with anaphoric definite NPs is strongly preferred since ku unambiguously establishes a co-referential interpretation.

As exemplified in (10–11), ku seems to behave like that. If we want to argue for Chang’s (2009) claim that ku behaves like an anaphoric definite determiner which differs from the English demonstrative that, we need to find anaphoric contexts where the and ku are felicitous but that is infelicitous. The previously discussed example in (5), repeated here as (12a), represents the type of anaphoric definite context where the and ku are appropriate but that is not.

The definite NP the novel has no explicit antecedent; however, it functions as the direct object of read, which establishes an anaphoric relation between the novel and the unmentioned indefinite NP antecedent (a novel). A demonstrative description (that novel) is inappropriate in this context in English since that requires an explicit antecedent. In Korean, as illustrated in (12b), ku soseolchayk ‘that novel’ sounds more natural than the bare NP soseolchayk ‘novel’ which is ambiguous between a definite and indefinite interpretation. In fact, the bare NP gives a strong generic interpretation in this context (as in ‘Jonathan found novels interesting but not short stories, non-fictions, etc.’). ³ This example illustrates that ku, unlike that which cannot appear in anaphoric contexts without explicit antecedents, appears in such contexts to signal definiteness like the in English. At the same time, as seen in (11), ku differs from the in that it cannot appear with non-anaphoric definite NPs, while the appears with both anaphoric and non-anaphoric definite NPs. Table 2 shows the distributions of the, that and ku, where the encodes [+definite, ±anaphoric] and ku [+definite, +anaphoric].

OPEN IN VIEWER

Table 2.

Distribution of English the and Korean ku.

	Anaphoric definite NPs [+definite, +anaphoric]		Non-anaphoric definite [+definite, −anaphoric]
	Explicit antecedent Example (10)	Implicit antecedent Example (12)	No antecedent Example (11)
English	the/that	the/#that	the/#that
Korean	#?bare NP/ku	#bare NP/ku	bare NP/#ku

This section has compared how different types of definite NPs are expressed in Korean and English. The next section describes how these cross-linguistic differences are employed to examine L1-Korean learners’ semantic judgments on different types of definite NPs in L2-English.

The study

Test design

In order to test my prediction, I categorized definite NPs into four types. First, direct anaphoric definites occur in a typical anaphoric context where new referents are introduced in the form of indefinite NPs (e.g., a microwave) and take the form of definite NPs in the second mention (e.g., the microwave), as in (13). The second type, dubbed taxonomic anaphoric definites, is similar to the first one in that the definite NP has an explicit antecedent in the text but is not identified as the same NP. The anaphoric NP and the antecedent are not identical but in a close lexical relation (synonyms, hypernym-hyponym, etc.), as illustrated in (14). The knowledge that book is a hypernym of novel leads to interpreting these two NPs as referring to the same entity. Anaphoric bridging definites represent a special type of anaphoric definite whose antecedents are implicit. In (15), the play was not mentioned previously and has no explicit antecedent. However, the play refers back to the unmentioned direct object (i.e., a play) of the verb (performing) in the previous sentence.

(13) Anaphoric definite NPs (same head noun antecedent)

  Harold got a microwave for Christmas. He put the microwave next to his new toaster.

(14) Taxonomic anaphoric definite NPs (lexically associated antecedents)

  Howard wrote a book. He sent the novel to three publishers already.

(15) Anaphoric bridging NPs (implicit antecedents)

  Hillary was performing for the first time. So, her parents attended the play.

(16) Non-anaphoric bridging definite NP (situational uniqueness)

  Brad just proposed. His fiancée accepted the ring happily.

Unlike the definite NPs in (13–15), the definite reference the ring in (16) does not have either an explicit or an implicit antecedent. Its definite interpretation is established by the fact that there is one unique ring associated with the situation mentioned. Based on general knowledge, we presume Brad gave only one ring to his fiancée when he proposed to her, thus the ring, not a ring. This type of definite NP is referred to as non-anaphoric bridging definites in this paper. Table 3 summarizes the four types of definite NPs examined in this study.

OPEN IN VIEWER

Table 3.

The encoding of different types of definite NPs in English and Korean.

	Anaphoric definite [+definite, +anaphoric]			Non-anaphoric definite [+definite, −anaphoric]
	Direct anaphoric with explicit same NP antecedent	Taxonomic anaphoric with explicit lexically associated antecedent	Anaphoric bridging with implicit antecedent	Non-anaphoric bridging with no antecedent
English	the	the	the	the
Korean	ku	ku	ku	bare NP

Methodology

Participants (n = 63)

Twenty-six native speakers of English and 37 Korean-speaking learners of English participated in the experiment. All participants were students at a large Midwestern university in the United States at the time of testing. All participants including the native speaker controls completed a paper-based acceptability judgment task (45–60 minutes) and proficiency test (20 minutes) during a single session. The L2 participants (n = 37) were divided into two sub-groups based on their proficiency test scores (Table 4).

OPEN IN VIEWER

Table 4.

Information on participants language background and proficiency test scores.

	Proficiency Test Score (max = 40)		Years studying English		Age at Testing (years)
	Mean (SD)	Range	Mean (SD)	Range	Mean (SD)	Range
L2-English learners Intermediate (n = 22)	29.72 (2.21)	26–33	7.57 (5.2)	1–20	25.09 (3.91)	19–33
L2-English learners Advanced (n = 15)	35.46 (1.64)	34–39	9.53 (4.96)	3–20	23.73 (5.6)	19–32
Native English Controls (n = 26)	38.7 (1.18)	37–40	N/S	N/S	21.04 (3.94)	18–35

Test instruments and procedure

All participants completed a proficiency test, an acceptability judgment test, and a language background questionnaire. The proficiency test used in this study included 40 items and was based on the Common European Framework of Reference for Languages (CEFR). The acceptability judgment task (AJT) was roughly based on the task used by Evans (2005) with L1-English speakers. The AJT included 170 test items (56 target items, 18 control items, and 96 filler items) targeting four categories: 1) direct anaphoric (n = 18), 2) taxonomic anaphoric (n = 18), 3) anaphoric bridging (n = 8), and 4) non-anaphoric bridging (n = 12). Half of the items in each category were definite NPs the+NP (acceptable) and the other half were indefinite NPs a+NP (unacceptable). The control items included acceptable indefinite NPs (n = 9) and unacceptable definite NPs (n = 9) using the same target item NPs in order to balance the acceptability of the and a in target items.

As highlighted earlier, ku affects interpretation, not grammaticality, like English articles often do. This study examines L1-Korean learners’ interpretation of definite NPs in different definite contexts by measuring the learners’ felicity (acceptability) judgments rather than grammaticality judgments. Thus, all test items are grammatical. In order to maximally control for variables, the same lexical items are used in all categories, all target items are singular, specific, and referential, and all target NPs appear in the direct object position.

Participants were asked to read pairs of sentences and rate whether or not the second sentence was an acceptable continuation of the first sentence on a 4-point scale with an option for “I don’t know” (1 = unacceptable, 2 = somewhat unacceptable, 3 = somewhat acceptable, and 4 = acceptable). They were also asked to underline or circle the problem and optionally provide a correction if they selected 2 or 1. Participants who judged the target sentences based on criteria other than articles were excluded from the analyses. ⁴

Below are sample test items for each type of definite context. In the direct anaphoric context, the second-mention referent cake should be in the definite NP form (the cake). Thus, (17) should be rated higher than (18). Similarly, in the taxonomic anaphoric context, the antecedent (a dessert) is mentioned in the second sentence with a different noun (the cake); thus, (19) sounds more natural than (20). The definite NP (the cake) in the anaphoric bridging context can refer back to the unmentioned/implicit antecedent (a cake) which is direct object of the verb baked. Thus, the indefinite NP (a cake) in the second sentence, as in (22), sounds unnatural. In the non-anaphoric bridging context, the referent cake has no antecedent; however, its definite interpretation can be established through the mention of birthday in the first sentence. That is, the first sentence restricts the context to (someone’s) birthday which conjures up for the hearer many things associated with a birthday, including the general knowledge that there is usually one unique cake. This situational uniqueness makes the definite NP the cake in (23) more acceptable than the indefinite NP a cake in (24).

Direct anaphoric definite context (same head noun antecedents)

(17) Jackie made a cake for the party. She served the cake with coffee and tea.

  [ 1 2 3 ④ I don’t know ]

(18) Kevin ordered a cake from the grocery store. #He went to pick up a cake but it was not ready.

  [ ① 2 3 4 I don’t know ]

Taxonomic anaphoric definite context (different head noun antecedents)

(19) Lydia’s family purchased a dessert. They ate the cake after dinner.

  [ 1 2 3 ④ I don’t know ]

(20) Marianne and her daughters shared a dessert. #They enjoyed a cake.

  [ ① 2 3 4 I don’t know ]

Anaphoric bridging definite context (implicit antecedents)

(21) Tori baked for her office this morning. Her co-workers enjoyed the cake.

  [ 1 2 3 ④ I don’t know ]

(22) Rachel baked for her husband. #He enjoyed a cake.

  [ ① 2 3 4 I don’t know ]

Non-anaphoric bridging definite context (no antecedent)

(23) It was Sophie’s first birthday. She smashed the cake with her hands.

  [ 1 2 3 ④ I don’t know ]

(24) Patrick celebrated his birthday with his friends. #They enjoyed a cake.

  [ ① 2 3 4 I don’t know ]

Experimental results

The present study examines whether L1-Korean L2-English learners’ semantic judgments on definite and indefinite NPs across the four different types of definite contexts. As earlier, preferences between definite and indefinite NPs in test items in this study are based on acceptability (felicity), not grammaticality; thus, the mean rate-differences between acceptable and unacceptable sentences are not large even among native speakers (see Figures 1–3 for the mean ratings for acceptable definite NPs and unacceptable indefinite NPs in the four definite contexts).

OPEN IN VIEWER

Figure 1.

Native speaker controls’ mean acceptability ratings for definite and indefinite NPs in four definite contexts.

OPEN IN VIEWER

Figure 2.

Intermediate L2 learners’ mean acceptability ratings for definite and indefinite NPs in four definite contexts.

OPEN IN VIEWER

Figure 3.

Advanced L2 learners’ mean acceptability ratings for definite and indefinite NPs in four contexts.

This study does not aim to compare L2 learners’ performance to that of native speakers’ in the degree of acceptance or rejection (reflected in numerical values) for the same target items, that is, whether they rate the same sentence by 3 or 4. Moreover, this study did not set out to compare whether learners prefer certain types of definite NPs over others (e.g., rating direct anaphoric definites higher than taxonomic anaphoric definites). My a priori assumption was that definite NPs (acceptable) should be rated higher than indefinite NPs (unacceptable) in each definite context. Thus, I conducted a series of paired samples t-tests (pairwise comparisons) on the mean rate-differences for each group and for each definite context independently to detect whether learners established a contrast between acceptable definite NPs and unacceptable indefinite NPs in each definite context. Results are shown in Table 5. ⁵

OPEN IN VIEWER

Table 5.

Acceptability contrasts (pairwise comparisons) in four types of definite contexts.

	Direct Anaphoric		Taxonomic Anaphoric		Anaphoric bridging		Non-anaphoric bridging
	t	p	t	p	t	p	t	p
NS (n = 26)	10.706	< .0001*	11.557	< .0001*	7.437	< .0001*	10.822	< .0001*
Adv L2 (n = 15)	3.205	.006*	2.906	.012*	1.562	.14	1.942	.072
Intm L2 (n = 22)	2.745	.012*	2.769	.012*	2.905	.008*	.156	.877

As expected, the L1-English speakers rated definite NPs significantly higher than indefinite NPs in all definite contexts (p < .0001*). Confirming my prediction, the intermediate L2 group showed a statistically significant contrast between definite and indefinite NPs in the three anaphoric definite contexts but not in the non-anaphoric/unique context, as seen in Table 5. Unexpected patterns of contrast were found among the advanced L2 learners. They differentiated between definite and indefinite NPs in their acceptability ratings in two anaphoric definite contexts (direct anaphoric and taxonomic anaphoric) but not in the anaphoric bridging context (p = .14) or the non-anaphoric bridging context (p = .072), suggesting that anaphoricity may not be a main factor in their semantic judgments about definite NPs. I will return to this issue in the discussion section.

Finally, I report learners’ judgments on the control items where indefinite NPs are acceptable. Control items include nine indefinite NPs and nine definite NPs in indefinite contexts, as in (25).

Acceptable indefinite NP/unacceptable definite NP

Both intermediate and advanced learners correctly rated indefinite NPs significantly higher than definite NPs in indefinite contexts (p = .04). In the next section I discuss what these findings, that is, the accurate ratings in indefinite contexts (which is always non-anaphoric since anaphoric contexts are necessarily definite) indicate with respect to the learners’ judgments in the two definite contexts (anaphoric and non-anaphoric definite).

Discussion and conclusions

This study explores L1-Korean L2-English learners’ acceptability judgments for definite and indefinite NPs in four definite contexts: 1) direct anaphoric, 2) taxonomic anaphoric, 3) anaphoric bridging, and 4) non-anaphoric bridging. The definite article the in English marks definiteness in both anaphoric and non-anaphoric contexts (i.e., in all four contexts), while the demonstrative ku in Korean encodes definiteness only in anaphoric contexts, that is, the first three contexts, as shown in Table 3. In other words, the feature configuration for the is [+definite, ±anaphoric] and for ku it is [+definite, +anaphoric]. Under the feature re-assembly model, Korean-speaking learners of English are predicted to initially map the onto ku and correctly rate the acceptability of definite NPs higher than that of indefinite NPs in anaphoric definite contexts but not in non-anaphoric definite contexts. This prediction was confirmed by the intermediate-level L2 learners. The intermediate group demonstrated a contrast between acceptable (definites) and unacceptable (indefinites) sentences in all three types of anaphoric contexts (direct, taxonomic, and anaphoric bridging) but not in non-anaphoric contexts. The advanced L2 learners showed significant distinctions between acceptable definite NPs and unacceptable indefinite NPs in two anaphoric contexts (direct and taxonomic) but they did not exhibit such contrasts in anaphoric bridging (with implicit antecedents) or in non-anaphoric bridging contexts (with no antecedents). In both the anaphoric and the non-anaphoric bridging contexts, the advanced L2 group rated the acceptability of unacceptable indefinite NPs as high as that of acceptable definite NPs, as seen in Table 5 and Figure 3. It is premature to draw any affirmative conclusion at this point about the advanced learners’ judgment patterns without a further investigation; however, I would like to consider one possibility. ⁶ Although anaphoric bridging NPs are considered anaphoric and non-anaphoric bridging NPs are considered non-anaphoric, they both represent the same special category of definite NPs, so-called bridging definites. Unlike definite NPs with explicit antecedents such as direct and taxonomic anaphoric definite NPs whose values are largely (if not entirely) determined in the domain of semantics, interpretations of bridging definite NPs (with implicit or no antecedents) require additional pragmatic accommodation, referred to as presupposition accommodation. As discussed earlier, licensing bridging definite NPs require hearers’ accommodation of the implied (and potentially arbitrarily complex) link between the bridging description and its anchor (the element that the bridging description is related to). Let us compare two types of bridging definites examined in this study: anaphoric bridging definites, as in (26), and non-anaphoric bridging definites, as in (27).

In (26), the cake is anaphoric to the (unmentioned) entity that Tori baked and would be translated into Korean as ku keik ‘that cake,’ not the bare NP keik ‘cake.’ In (27), the cake is not anaphoric as it has no element (either explicit or implicit) to which it refers. Using ku keik ‘that cake’ in this context is inappropriate. If L1-Korean learners are influenced by L1 knowledge, they will be more accurate in interpreting the definite NP the cake in (26) than the cake in (27). However, licensing definite NPs in both (26) and (27) requires pragmatic accommodation. That is, hearing the cake in the second sentence in (26), the hearer should be willing to accommodate the implied presupposition that Tori baked a cake, not cookies and she baked only one cake. In (27), the hearer should also be willing to presuppose that there was only one cake at Sophie’s birthday although it is possible that there was more than one (or none at all) at her birthday party. I further examined advanced L2 learners’ judgment patterns (incorrect acceptance or incorrect rejection) in bridging contexts. Advanced learners rated both definite (acceptable) and indefinite NPs (unacceptable) high in both anaphoric and non-anaphoric bridging contexts, that is, they accepted unacceptable indefinite NPs (incorrect acceptance). This suggests that advanced learners were not only able to accommodate unmentioned propositions to license bridging definite NPs (e.g., birthday – the cake) but also able to accommodate to license indefinite NPs (e.g., birthday – a cake) in bridging definite contexts by coming up with situations where indefinite NPs would sound natural. For example, in (27), the indefinite NP (a cake) is acceptable if one assumes there was more than one cake at Sophie’s birthday. Advanced learners might have been ‘overcompensating’ based on their ability and confidence and tried hard to come up with alternative interpretations that might make unacceptable sentences acceptable, ⁷ while intermediate learners were probably unable to do so since their understanding of the contrast between the vs a is anaphoric vs non-anaphoric at their proficiency level (assuming they have mapped the onto ku [+definite, +anaphoric], as I predicted).

To sum up, the intermediate group’s judgments between definite and indefinite NPs were more accurate in anaphoric contexts than non-anaphoric contexts, indicating that learners made use of their L1 knowledge and associated the with [+anaphoric, +definite]. This was also supported by the intermediate (as well as advanced) learners’ accurate judgments on definite vs indefinite NPs in indefinite (always non-anaphoric) contexts (control test items). Since indefinite NPs are always non-anaphoric, learners correctly rejected definite NPs and accepted indefinite NPs in [−definite, −anaphoric] contexts. As for the advanced learners, their judgments did not correlate with anaphoricity, suggesting the absence of L1 transfer. However, they still differed from the native control group in their preference between definite and indefinite NPs in bridging definite contexts but the overall patterns suggest that they have successfully re-assembled the feature bundle by adding [−anaphoric] and have acquired that the may appear either in anaphoric or in non-anaphoric contexts. That is, the advanced learners did not rely on anaphoricity to interpret definite NPs, suggesting that they have overcome L1 transfer and re-assembled features associated with the: [+definite, ±anaphoric]. What they have not yet acquired may be the full understanding of ‘definiteness’ which involves various types of presupposition accommodation (e.g., presupposition of familiarity, presupposition of uniqueness, presupposition of inclusiveness, presupposition of existence) and types of information (both mentioned or unmentioned) in discourse that can accommodate presupposition related to definiteness. Learnability issues related to presupposition accommodation in article acquisition certainly warrant further investigation.

The findings of this study as well as the cross-linguistic data (discussed in Schwarz, 2009, 2013) have shown that the feature [definite] co-occurs with different features cross-linguistically; thus, feature-reassembly is a necessary process in the L2 acquisition of articles. The advanced learners, however, showed non-target-like presupposition accommodation strategies which result in comprehension of articles that differed from that of native speakers. These findings imply that there is another hurdle in the article acquisition process beyond feature re-assembly – presupposition accommodation.

By focusing on the acquisition of the semantics of definite NPs, exclusively, this study provides new data and information which enable us to come to a more precise and fine-grained understanding of learnability in article acquisition. Thus, the results of the present study bring out new and insightful conceptual issues that open up new directions for future research on article acquisition and the acquisition of definiteness.