Grammar vs. pragmatics – the acquisition of articles
by Polish L2 speakers of Dutch
Naomi P.C. van Eenennaam (6071775) University of Amsterdam Supervisor: dr. J.C Schaeffer Master Nederlandse taal en cultuur
14 juni 2014
Abstract
This paper examines Dutch article acquisition by Polish L2 speakers of Dutch. If languages semantically distinguish more than one article, this distinction can rely on definiteness or referentiality, following the Parameter of Article Semantics (Schaeffer & Matthewson, 2005). For Dutch, the distinction relies on definiteness (distinction on the basis of common ground). L2 acquisition of definiteness involves the acquisition of two types of knowledge: article suppliance (grammar), and article choice (pragmatics). In this study, I examined 30 adult Polish learners of Dutch with an elicited production task designed by Schaeffer and Matthewson (2005). Polish is an article-less language, and therefore does not have a preference for either definiteness or referentiality. The results of this study show that the adult Polish L2 speakers of Dutch perform well on article suppliance: the division between intermediate speakers and advanced speakers shows that article omission decreases as proficiency level increases. This study also shows that the
participants, regardless of their proficiency, perform well on article choice. Because the participants are all adults and therefore have mature pragmatics, they are able to distinguish speaker/hearer knowledge (Concept of Non-Shared
Assumptions, Schaeffer & Matthewson, 2005), and know when to use a definite article or an indefinite article.
1. Introduction
Languages of the world differ as to whether they express articles overtly or not. Some languages have more than one overt article (English, Dutch, German, French, Italian, cross-linguistic), some languages have just one overt article (Turkish: indefinite article; Hebrew: definite article), other languages do not have overt articles at all (Russian, Polish).
Furthermore, languages with one or more overt article differ regarding the reference of these articles. Some languages base their choice of article on the opposition definite-indefinite (English, Dutch, German, French, Italian) other languages rely on the contrast specific-non-specific (Salish).
In this study I investigate how adult learners of Dutch with L1 Polish acquire the Dutch article system. This involves the acquisition of (at least) two types of knowledge: a) article suppliance, and b) article choice. As Polish is an article-less language, Polish grammar is of no help in acquiring Dutch articles, which are overt, and are chosen on the basis of definiteness. My results show that Polish L1 learners of L2 Dutch initially omit articles because Polish predicates do not require overt articles to turn into arguments. However, with enough input, they become near-native like in their suppliance of articles. Moreover, as soon as they produce overt articles, they choose their articles correctly according to definiteness most of the time. The only (small) error they make is the production of an indefinite article in a definite context, which I explained on the basis of less automatically processed structures that can cause a breakdown in the performance of the L2 speakers (Sorace, 2011).
This paper is organized in the following way: section 2 provides a theoretical background on article suppliance and article choice. I define relevant terms and concepts like definiteness, referentiality and notions of speaker and hearer beliefs, followed by findings from L1 studies on these two issues. Then I lay out a L2 theory and review a number of relevant studies on L2 acquisition of article suppliance and article choice. After that I lay out the predictions and hypotheses for this experimental study. In section 3, I describe the elicited production task (Schaeffer and Matthewson, 2005) and give information about this method and the
participating Polish speakers of the study. Finally, I present the results in section 4 and include a discussion. A conclusion is drawn in section 5.
2. Background
Throughout this paper I take articles to be functional items that are correlated with
argumenthood (i.e., appear on argumental phrases and plausibly function to turn predicative common noun phrases into phrases which may occupy argument positions), and are neither demonstratives nor quantifiers.1 I do not commit myself to articles occupying a particular syntactic position, although one obvious place for them is D (cf. Longobardi, 1994 and others who tie the argument-creating function to the syntactic position of the head of DP).
As mentioned in the introduction, Dutch is an article-based language and belongs to the Germanic language family. Table (1) shows that Dutch uses the definite article de for non-neuter (common) singular and for plural nouns, the definite article het2 for neuter singular nouns and the indefinite article een with singular nouns. No article is used with indefinite plural nouns (Guella et al., 2008: 62). There is no gender distinction marked on the plural definite determiner (de) or on the singular indefinite determiner (een). According to a
dictionary-based estimate, roughly 75% of Dutch nouns are common and only 25% are neuter (Cornips & Hulk, 2008: 269). A prequisite for the correct choice between a definite and an indefinite article in an article-based language such as Dutch, is the knowledge of when an article needs to be overtly realized.
(1) Dutch article system
Definite Indefinite
plural singular plural singular
Neuter nouns de boeken ‘the books’ het boek ‘the book’ Ø boeken ‘Ø books’ een boek ‘a book’ Non-neuter nouns de tafels
‘the tabels’ de tafel ‘the table’ Ø tafels ‘Ø tables’ een tafel ‘a table’
Thus, the use of articles includes two types of knowledge, namely 1) knowledge regarding the (c)overtness of articles, determining article suppliance, and 2) knowledge of
1 Since English definite and indefinite articles are sometimes analyzed as introducing quantification (e.g. by
Russell 1905), a more precise statement of the last requirement is that articles are analyzable as being non-quantificational (as for example in Heim (1982) or Kamp (1981)), unlike canonical quantifiers such as every or
most. 2
The difference between the definite articles de and het will not be discussed whereas this difference is not relevant for this research on definiteness.
definiteness/referentiality, determining article choice. As mentioned in the introduction, the two languages under consideration in this study are Dutch and Polish. Dutch is an ‘overt-article’ language where articles are based on definiteness. Polish is language that has no articles or article-like morphemes (Ekiert, 2005: 2).
In section 2.1 and 2.2 I discuss both types of knowledge in terms of parameters and their settings in Dutch and Polish, each followed by a brief overview of relevant previous studies on the L1 acquisition of articles. In section 2.3 I lay out an L2 theory, which I take as the basis for the present study, and after that I review a number of relevant studies on the L2 acquisition of articles.
2.1 Article suppliance
2.1.1 Adult language
Chierchia (1998a, 1998b) argues that the overt use of articles, article suppliance, is a
semantic, and therefore a grammatical phenomenon. He claims that languages vary in terms of what they allow their NPs to denote, and proposes the Nominal Mapping Parameter. This parameter, in the words of Snape (2006), allows for certain semantic variation in terms of how the reference of the syntactic category NP is set. The setting of the Nominal Mapping Parameter determines how the syntax of a noun is mapped to its semantics. Chierchia makes a distinction between three kinds of languages with regard to this mapping: (1) whether all NPs can refer directly to kinds (i.e., can be used generically), and hence are possible arguments in syntactic expressions; (2) whether no NPs can refer directly to kinds, but are predicates that need to be accompanied by a determiner, and (3) whether some NPs can refer to kinds directly and some cannot. In other words, he proposes that noun denotations are subject to language-specific parameter setting of the features [±argumental] and [±predicative]. Below, (2) gives an overview of the three kinds of languages (based on Snape’s (2006) version of Chierchia, 1998a, 1998b), with examples under (3)-(5):
(2)
1. NP [+arg, -pred] 2. NP [-arg, +pred] 3. NP [+arg, +pred] Chinese, Japanese, Thai Italian, Spanish, French English, German,
Dutch - generalised bare arguments
- all nouns are mass nouns - no plural morphology - generalised classifier system
- no bare nominals in argument position
- count/mass distinction - morphological plural
- bare mass nouns and plurals in argument position
- no bare singular count nouns
- plural morphology
(3) 1. NP [+arg, -pred] – Chinese:
wò kànjiàn xióng le I see bear ASP
‘I saw (some/the/a) bear(s).’
2. NP [-arg, +pred] – French and Italian:
a. * Enfants sont venus chez nous. ‘Kids have come by us.’
b. * J’ai mangé biscuits dans mon lait. ‘I ate cookies with my milk.’
a. * Bambini sono venuti da noi. ‘Kids have come by us.’
b. * Ho preso biscotti con il mio latte. ‘(I) had cookies with my milk.’
French and Italian are NP[–arg, +pred] languages. In a language of this sort every noun is a predicate, and since predicates by definition cannot occur in argumental positions, such a language should disallow bare nominal arguments altogether (Chierchia, 1998: 355). Dutch is a language with the setting [+argument, +predicate], which means that some nouns are
argumental, whereas others are predicates. Bare plural and mass nouns can function as
arguments without the need for a licensing determiner. However, singular count nouns, which are predicates, require licensing by an article, otherwise they are ungrammatical (Snape, 2006). This is illustrated in (4):
(4) *(De) hond heeft honger the dog has hunger ‘The dog is hungry’
In (4), the NP hond is a singular count noun, and is thus [+ pred]. In order to occupy an argument position (in this case the subject position), it requires licensing by an article. In contrast, in (5) the bare plural honden, which is [+ arg], does not need an article to occupy the subject argument position:
(5) Honden zijn goede huisdieren dogs are good pets ‘Dogs are good pets’
Similarly, mass nouns, which are also [+ arg], can occupy argument positions without modifying them with an article, as illustrated in (6):
(6) Water is verfrissend water is refreshing ‘Water is refreshing’
Thus, in adult Dutch, articles reflect argumenthood, but argumenthood is not always expressed by articles.
As for the Slavic languages, such as Russian and Polish, Chierchia argues that they fall within the same category as English (and Dutch, for that matter): “Russian is like English, but without articles, and the bare NP arguments can have definite, indefinite or kind
interpretations.” The rationale behind this claim is that although Russian and Polish do not possess definite or indefinite determiners, they do have a plural/singular distinction. To account for bare arguments in these languages, it is assumed that covert type shifting operations allow the language to freely shift nominals from predicative to argumental (Piriyawiboon, 2010). Longobardi (2013) supports this view, and also recognizes that
languages such as Russian and Polish differ from languages such as English and Dutch in that they lack overt articles. He therefore adds that in Russian and Polish ø def = def/indef.
Thus, even though Chierchia and Longobardi argue that Dutch and Polish fall within the same category, the expression of the morpho-syntactic element mapping onto the semantic
feature argumenthood obviously varies between these languages. In Dutch, whenever a noun is a singular count noun that wants to occupy an argument position, it needs to be
accompanied by an overt article. On the contrary, in Polish, such a noun can occur without an overt article. However, the fact that these two languages belong to the same category, namely the NP [+arg, +pred] languages, may also facilitate the acquisition of the Dutch article system by Polish L1 speakers.
2.1.2 Child language
As for the acquisition of articles by monolingual children, Brown (1973) was one of the first researchers who reported that English-speaking children use articles relatively early on, but that they make errors in article use until they are at least 4 years old. Brown finds in his study that article suppliance in obligatory contexts in the children’s speech is almost adult-like between the ages of 2;8 and 3;5 and that ‘the’ is sometimes used incorrect in indefinite contexts. Since Brown (1973) it has been noted that young monolingual children often drop articles. Extensive research demonstrates that article drop by young children is a cross-linguistic phenomenon (Hulk and Van der Linden, 1996; Schaeffer, 1999; Chierchia, Guasti and Gualmini 1999; Rozendaal and Baker, 2007 among others). These studies report variation in rate of omission, for example, that children acquiring Germanic languages show higher rates of article omission than children acquiring Romance languages (Chierchia et al., 1999). Chierchia et al. (1999) explain this as follows. The task for the language learner is as in (7):
(7) Acquisition of argumenthood
(i) map the semantic feature ‘predicatehood’ to its correct morpho-syntactic counterpart;
(ii) map the semantic feature ‘argumenthood’ to its correct morpho-syntactic counterpart.
Chierchia et al. (1999) argue that article omission in L1 acquisition results from a mis-mapping between the semantic property ‘argumenthood’ and its semantic counterpart (noun, or article+noun). Misanalyzing predicative nouns (for example, dog) as argumental results in non-adultlike bare nouns. As noted before, article omission appears to be a property of the very early stages of L1 acquisition of every language, however, omission rates and duration varies between languages. As stated by the Nominal Mapping Parameter above, in the
order to occupy an argument position. In other words, articles are always obligatorily realized, and for this reason children acquiring Romance languages get plenty of evidence for the overt realization of articles in their input, which is why the stage of article omission is relatively short. In contrast, Germanic-acquiring children receive a very mixed input in terms of overt articles, because some nouns are predicates, and others are arguments, i.e., sometimes they hear nouns with overt articles (such as singular count nouns in argument positions) and sometimes they hear nouns without overt articles (such as bare plurals or mass nouns). This makes it more difficult to acquire the subtleties of when to use an overt article and when to omit it. Therefore, their article-drop stage lasts longer than that of the Romance children.
2.2 Article choice
2.2.1 Adult language
Turning now to article choice, I follow Schaeffer and Matthewson (2005) in assuming that article choice applies to languages with at least one overt article. These languages can be divided into two categories: languages that choose their articles on the basis of the definite-indefinite opposition, or languages that choose their articles on the contrast specific-non-specific. The first category includes languages such as English and Dutch.
The basics of definiteness marking can be clarified by the explanation by Lyons (1999), who uses English to explain the terminology. English and Dutch both belong to the Germanic languages, and share almost the same article system with regard to definiteness. Where English only has one definite article (the), Dutch has two (de and het). The English a and the Dutch een are indefinite articles.
The notion of definiteness by Lyons’ (1999) can be explained by making a distinction between ‘the girl’ and ‘a girl’. These different utterances can be distinguished by the
knowledge of the speaker and hearer. Lyons (1999) distinctions are exemplified in (8) (Schaeffer & Matthewson, 2005):
(8) (I) Definite referential:
‘Dit is een verhaal over een meisje. Het meisje woonde in een groot kasteel.’ This is a story about a girl. The girl lived in a big castle.
(II) Indefinite referential:
‘Ik heb een film gekeken gisteravond.’ I saw a movie last night.
(III) Indefinite non-referential:
‘Mijn moeder laat misschien een huis bouwen.’ My mother might build a house.’
The definite article de (the) is used if the speaker is referring to a particular girl, not just any girl. In the first category, (I), of (8) above, on the first mention of ‘a girl’, the girl has not been introduced yet, so the indefinite article een (a) is used (Ionin et al., 2004: 6). On the second mention, the referent of the girl is known by both speaker and hearer, due to the fact that it was mentioned in the preceding discourse (Schaeffer, 1999: 2). In this case, the definite article de can be used. As in (II), when the speaker knows which movie he is referring to but the hearer does not, the speaker will use the indefinite article een. Also when both speaker and hearer don’t know to which house is being referred, the indefinite article een is being used (see III).
The adult language article system is schematized in (9):
(9) Adult language article system
A believed by speaker and hearer part of common ground the B believed by speaker only not part of common ground a C believed by neither speaker/hearer not part of common ground a
Many linguists believe that definiteness is a universal property of human language that makes it possible to distinguish shared from non-shared knowledge, or, ‘common-ground’
knowledge from ‘non-common-ground’ knowledge (Heim, 1982). In turn, referentiality is another universal property of human language. Fodor and Sag (1982) define referentiality as in (10):
(10) Referentiality (cf. Fodor & Sag, 1982)
A nominal expression is understood to be referential if it has a fixed referent in de (model of the) world, meaning that it can be identified by the speaker or by one of the
people whose propositional attitudes are being reported.
Referentiality crucially differs from definiteness in that it concerns the speaker’s intent to refer regardless of the hearer’s knowledge status (Ionin et al., 2008: 3). Lillooet Salish, a Northern Interior Salish language spoken in Southwest mainland British Columbia, Canada, is an example of a language whose article system is organized on the basis of the referential-non-referential opposition3. Crucially, whether the object is known to speaker, hearer, both or neither is irrelevant to determiner choice in this language (Matthewson, Bryant and Roeper 2001). Rather, determiners encode whether or not the speaker is able to make an existential assertion about an individual (object) (Matthewson, 1998). If the answer is yes, the noun is referential, and one type of article is used. If the answer is no, the noun is non-referential, and another type of article is used. This opposition is illustrated in (11) (Schaeffer & Matthewson, 2005: 58):
(11) Believed by speaker:
a: wa7 ít’-em [ti smúlhats-a] IMPF sing-INTR [DET woman-DET]
‘A woman (visible to the speaker) is singing.’
Believed by speaker:
b: wa7 ít’-em [i smelhmúlhats-a] IMPF sing-INTR [DET:PL woman(PL)-DET]
‘Some women (visible to the speaker) are singing.’
Not believed by speaker: c: wa7 ít’-em [ku smúlhats-a] IMPF sing-INTR [DET woman-DET]
‘A woman (invisible to the speaker) is singing.’
The two different oppositions and the notions of definiteness and referentiality of English and Salish are schematized in (12) (Schaeffer and Matthewson, 2005: 57, 64):
3
(12) English and Salish adult language article system
English Salish
A believed by speaker and hearer the believed by speaker ti…a
B believed by speaker only a ti…a
C believed by neither speaker nor hearer a not believed by speaker ku
Languages such as English and Dutch lump contexts B and C together (indefinite) in their choice of articles, as opposed to context A (definite), whereas languages such as Salish collapse contexts A and B (referential) as opposed to C (non-referential).
In Polish, definiteness and/or referentiality are expressed without an article system:
information about the knowledge status of speaker and hearer are obtained through prosody, discourse linking and word order, for example sentence-initial DPs tend to be known to speaker and hearer while DPs that are not known to the speaker tend to be sentence final (Szwedek, 1974). The sentences in (13) and (14) are Polish and English translation equivalents and are copied from Ekiert (2005: 7).
(13) Mężczyzna wszedł do sklepu. man entered to store The man entered the store.
(14) Do sklepu wszedł mężczyzna. to store entered man A man entered the store.
The clause-final position of mężczyzna in (14) implies that this element is not known to the hearer. This example can be contrasted with sentence (13), in which man is marked by the definite article in English, because man is a specific referent, known by speaker and hearer. In the Polish equivalent the sentence starts with mężczyzna, so this element is known to the hearer. As the examples illustrate, word order in Polish takes on some functions for which articles are used in English (Ekiert, 2005).
Inspired by the idea that definiteness and referentiality are the two universal properties of human language that can determine article choice, Schaeffer & Matthewson (2005)
Article Choice Parameter and Matthewson’s (1998) Common Ground Parameter, and makes use of the definitions of definiteness and referentiality given in (12) above. The PAS is defined in (15):
(15) Parameter of Article Semantics (PAS) (Schaeffer & Matthewson, 2005) If a language semantically distinguishes more than one article,
the distinction relies on either:
I. Speaker beliefs (referentiality), or II. Common ground (definiteness)
The PAS is a parameter for two-article languages like English, Dutch, and Salish. Its setting determines which article is used in which context. It proposes three possible article systems based on different environments: environment A (believed by speaker and hearer),
environment B (believed by speaker only) and environment C (believed by neither speaker nor hearer). This creates three possibilities, shown in (16):
(16) Three environment combinations:
1. A+B
2. B+C
3. A+B+C
Grouping environment A and C together is not possible because such a language would be basing its article distinction neither on common ground (definiteness) nor on speaker beliefs (referentiality). Also, a language that maps each environment to a different article has not been found until now (Schaeffer & Matthewson, 2005: 67). English and Dutch are languages in category 2.
In order to set the PAS correctly, language learners need to be able to distinguish between speaker and hearer knowledge. Schaeffer (1999) Schaeffer & Matthewson (2005) formalize this distinction in a pragmatic concept, called the Concept of non-shared assumptions (CNSA):
(17) Concept of Non-Shared Assumptions (CNSA) (Schaeffer & Matthewson, 2005) Speaker and hearer assumptions are always independent.
The CNSA expresses an obligation for the speaker to consider the hearer’s assumptions as a separate entity and therefore as something that is in principle different from the speaker’s assumptions. Thus, the CNSA is a necessary condition for the setting of the PAS.
2.2.2 Child language
In terms of acquisition, the question arises as to how language learners acquire the CNSA and set the PAS. Schaeffer and Matthewson (2005) examined 26 monolingual English-speaking children between 2 and 4 years old with an elicited production task that contained three conditions: a definite, an indefinite referential and an indefinite non-referential condition. They predicted that the children 1) would not overgenerate a in definite contexts, 2) would not overgenerate the in indefinite non-referential contexts and 3) would overgenerate the in indefinite referential contexts. The results showed that all three predictions were borne out. The children produced 25% occurrences of the instead of a in the indefinite referential
contexts, in contrast to the native control group, who virtually never overgenerate the in these contexts. Analysis of the interaction revealed a significant difference between children and adults in the indefinite referential contexts. Both children and adults hardly ever overgenerate a in definite contexts where the is required, and also do not overgenerate the in indefinite non-referential contexts. Thus, the results showed that children up until the age of 4 sometimes fail to distinguish B contexts from A contexts and therefore group together A and B contexts, as opposed to C. This made Schaeffer and Matthewson state that young children lack the pragmatic concept CNSA (Schaeffer & Matthewson, 2005: 81). Even though the children have set the PAS correctly, based on results of research in L1 Acquisition (Hyams, 1992, 1996; Wexler, 1998), which state that parameters are set very early, the children sometimes mistakenly believe that their own beliefs are shared by the hearer, i.e., are part of the common ground (Schaeffer & Matthewson, 2005: 84). What happens is that value I of the PAS
(distinction on the basis of speaker beliefs) becomes the same as value II (distinction on the basis of common ground).
The use of articles of English monolingual children is close to adult-like by age 4, so the concept is being acquired around that age. Before that age, a child does not always distinguish environment B (believed by speaker only) from environment A (believed by speaker and hearer). In such a case, environment B becomes environment A, so the child will use the article appropriate for environment A, which is the in English and de in Dutch. Compare the article system for adults (repeated in (18)) with the article system for children (19) hypothesized by Schaeffer and Matthewson’s (2005). Here you can see that when a child
does not distinguish speaker and hearer beliefs, environment A and B are combined, as opposed to C (believed by neither speaker nor hearer). The prediction is that once children acquire the CNSA they correctly choose a definite or indefinite article, depending on the relevant referential context.
(18) Adult language article system
A believed by speaker and hearer part of common ground the B believed by speaker only not part of common ground a C believed by neither speaker/hearer not part of common ground a
(19) Developmental hypothesis child language article system
A/B believed by speaker (and therefore by hearer) the C not believed by speaker (and therefore not by hearer) a
This is in agreement with Brown’s (1972) research on article omission and article
substitution. He found, like Schaeffer and Matthewson (2005), that the definite article the was sometimes used incorrectly in indefinite contexts. From the results of his research it can be concluded that children have no difficulty with expressing definiteness, but sometimes they cannot estimate the hearer’s knowledge in the discourse. This lack of knowledge of shared assumptions can lead to incorrect use of articles (Zdorenko & Paradis, 2011: 42). Children need to acquire the subtleties of speaker/hearer knowledge before they can make the correct choice for a definite or an indefinite article. The overuse of the definite article in contexts where the indefinite article is correct, is a cross-linguistic phenomenon in child language (Maratsos, 1976; Karmiloff-Smith, 1979; Zehler and Brewer, 1982; Schaeffer & Matthewson, 2005; Van Hout et al., 2010). The explanations in the literature suggest a wide variety of causes: some postulate that children are in a different cognitive state or follow a different semantic strategy than adults; others claim that syntactic, semantic or pragmatic rules in the child’s grammar are incomplete (Van Hout et al., 2010: 1974). In this study, I follow Schaeffer and Matthewson (2005) in their claim that children lack the pragmatic concept CNSA.
In summary then, language learners need to set parameters and map morphosyntactic elements (in this case, articles) to semantic and to pragmatic knowledge. In the case of article suppliance, learners need to set the Nominal Mapping Parameter, and then map the semantic concepts of predicatehood and argumenthood to their respective morphosyntactic
counterparts, based on what they hear in the input. In the case of article choice, learners need to systematically distinguish between speaker and hearer assumptions (pragmatic CNSA), and use that knowledge in combination with the input to set the PAS.
I have briefly shown how the Nominal Mapping Parameter can account for the
different speeds and rates at which young children omit articles. Furthermore, I discussed how the lack of the pragmatic Concept of Non-Shared Assumptions could explain why young children overgenerate the definite article in indefinite contexts. Most L1 studies show that by the age of 4, most children no longer omit articles, and no longer make article choice errors. In other words, article acquisition (both in terms of suppliance and in terms of choice) in L1 takes place rapidly, and it is often difficult to disentangle grammatical and pragmatic
development, and their relative contributions to article acquisition.
One way to obtain more insight in the way grammatical and pragmatic knowledge interact in the acquisition of articles is to study the adult L2 acquisition of articles. It is plausible to assume that pragmatics is language-independent, and that all adult learners have mature pragmatics. Nonetheless, their L2 grammar may still be underdeveloped. In other words, in L2 learners pragmatics and grammar do not develop at an equal pace, as in child learners, and therefore it may be easier to tease them apart.
The question arises as to how L2 learners acquire a Dutch-like article system. Do they initially omit articles, similar to L1 learners? Do they initially overgenerate definite articles, similar to L1 learners? If so, do the same analyses hold for L2? If not, how can the L2 learners’ behaviour regarding articles be accounted for?
Before returning to these specific questions, I first briefly discuss a general theory on L2 acquisition, which I take as the basis for the present study.
2.3 L2 theory
Parameter-setting is a part of the principles and parameter framework in which the syntax of a language is described in accordance with general principles (for example abstract rules or grammars) and specific parameters that for particular languages are either turned on or off (Chomsky, 1981). Together, these principles and parameters constitute Universal Grammar (UG). The access to this UG when acquiring a second language has been discussed and researched in several studies. The Full Transfer/Full Access hypothesis (FT/FA) (Schwartz & Sprouse 1994, 1996; Schwartz 1998a) states that the L1 grammar, including L1 parameter settings, constitutes the initial state of L2 acquisition (full transfer), but that L2 learners have
full access to UG at all times during the acquisition process (full access), and thus that parameter resetting is usually possible. The FT/FA hypothesis assumes that the entire L1 grammar transfers (abstract features and functional categories) at initial state in L2 acquisition (full transfer) and learners have access to properties of UG that are not instantiated in the L1 grammar (full access). The resulting ‘interlanguage grammars’ are UG constrained even if they turn out to be non-target like and differ to those of native speakers. Thus, as Snape (2006) formulates it carefully, Full Transfer applies to the initial state of L2 acquisition and Full Access applies to the ongoing interlanguage grammar restructuring over the course of development until the end-state.
In the present study, I take Schwartz & Sprouse’s (1994; 1996) Full Transfer/Full Access Hypothesis as a basis for the formulation of my own hypotheses and predictions regarding L2 article acquisition in Dutch.
The next section discusses some previous L2 studies on article acquisition, addressing some of the questions that were formulated at the end of section 2.2.2.
2.4 L2 studies on article acquisition
Ionin, Ko and Wexler (2004) investigated the role of definiteness and specificity
(referentiality) in the acquisition of English article choice by adult speakers of Russian and Korean. They explain the use of articles based on the Article Choice Parameter (ACP), one of the parameters that are combined into Schaeffer & Matthewson’s (2005) Parameter of Article Semantics (PAS). Similar to the PAS, the ACP is a parameter for two-article languages and makes a distinction between article choice on the basis of definiteness, and article choice on the basis of specificity. Just like Polish, Russian is a Slavic language that does not have articles and therefore has no initial preference for one setting of the ACP over the other (Ionin, Ko and Wexler, 2004: 51). Their research contained a forced-choice elicitation task, which was completed by 70 English learners: 30 L1 Russian speakers and 40 L1 Korean speakers. The elicitation task consisted of 76 short English-language dialogues where an article was missing in the target sentence. The participant had to choose between a, the, and the zero article. There were four indefinite contexts and four definite context types. Two of four (in)definite contexts contained [+specific] (in)definites and the other two [-specific] (in)definites.
To start with, the results show that article omission was fairly low, rarely exceeding 15% (Ionin, Ko and Wexler, 2004: 28). However, both LI Russian and LI Korean speakers
overused the more with [+specific] than with [-specific] indefinites and overused a more with [-specific] than with [+specific] definites. It looks like L2 learners are quite good at using the with definites that are [+specific], including previous mentioned definites, but overuse a with indefinites that they consided [-specific]. Nevertheless, Ionin et al. argue that with sufficient input, the L2 learners may succeed in setting the Article Choice Parameter to the appropriate setting for English and divide articles on the basis of definiteness only. Their division in proficiency level (intermediate versus advanced) showed that the advanced L2 learners are overall more accurate on both definites and indefinites than the intermediate learners, so there is progress. (20) is an example of a context in this study where the participants often used a instead of the (Ionin, Ko and Wexler, 2004: 35):
(20) [+ definite, +specific] condition Meeting in a park
Adrew: Hi Nora. What are you doing here in Chicago? Are you here for work? Nora: No, for family reasons. I am visiting (a, the, -) father of my fiancé – he is really
nice, and he is paying for our wedding!
Ionin, Ko and Wexler (2004) introduced the Fluctuation Hypothesis (see (21) below) to account for errors of L2 learners that are caused by fluctuation between two or more parameter-settings, some of which are not appropriate for the target language.
(21) The Fluctuation Hypothesis
a. L2 learners have full access to UG principles and parameter settings.
b. L2 learners fluctuate between different parameter-settings until the input leads them to set the parameter to the appropriate value.
Nearly one third of the L2 learners showed fluctuation between the two settings of the Article Choice Parameter. Ionin et al.’s study showed that L2 learners show evidence of accessing parameter-settings that are instantiated in neither their L1 nor their L2 (like Finer and Broselow (1986) originally showed). The Russian and Korean participants performed quite similarly, which provides evidence that their relatively well performance is not attributable to L1 transfer. Like the Fluctuation Hypothesis predicts, L2 learners show optional adherence to parameter-settings: their behaviour suggests that they sometimes adopt one setting of the parameter and sometimes another (definiteness and specificity).
Ionin et al. argue that their findings present evidence for direct access to universal semantic distinctions in L2 acquisition, because access to the feature [+specific] cannot be accounted for by either L1 transfer or L2 input: the participants’ L1 (Russian) does not express
specificity by means of overt articles, and their L2 (English) does not either, because it chooses its articles on the basis of the definite-indefinite opposition.
Even though Ionin, Ko and Wexler’s (2004) research is about article choice, the results showed also some information about article suppliance. In singular definite contexts, the L2 Russian participants correctly used the 75% of the time and only incorrectly used a 1% of the time. The remaining percentage, 24%, consists of article omission (Ionin, Ko &
Wexler, 2004: 46). For the L2 Korean speakers the article omission was 13%. In singular indefinite contexts, the L2 Russian and L2 Korean participants omitted articles respectively 35% and 25% of the time in [+specific] contexts, and 17% and 27% in [-specific] contexts.
The validity of the Fluctuation Hypothesis of Ionin, Ko and Wexler (2004) has been tested by Tryzna (2009), by learners of English L2 by speakers of article-less languages, namely 19 Polish L1 speakers and 17 Mandarin Chinese L1 speakers. She used a forced choice elicitation task modelled after Ionin et al. (2003)’s study: the forced-choice elicitation test. Focussing on the Polish participants, Tryzna made a distinction between intermediate and advanced L2 learners. The intermediate group is overall less accurate than the advanced one in their article use, as she predicted. The results indicate a significantly higher overuse of the with referential indefinites (38% for the intermediate learners and 25% for the advanced learners) than with non-referential indefinites (40% intermediate, 15% advanced). Also, both groups show little omission with singular definites in all contexts (less than 4%), but
relatively high omission in definite plural contexts (> 25%). For the intermediate learners, Tryzna found no more overuse of the with referential indefinites than with non-referential ones (singular indefinites: 38% referential, 40% non-referential; plural indefinites: both 20% for referential and non-referential), so the fluctuation effect is not confirmed in the group results. Also for the individual results Tryzna does not want to explain the results on the basis of the Fluctuation Hypothesis. She claims that L2 English article use is characterized by degrees of variability rather than fluctuation. The individual results potentially suggest developmental stages of article acquisition beginning with initial widespread optionality of the across all contexts, then gradually narrowing down to specific indefinite and definite contexts, and finally occurring in either definite or specific indefinite context (Tryzna, 2009: 82). Tryzna’s results also indirectly validated an assumption underlying the postulation of the
Article Choice Parameter, namely that referentiality is a semantic universal available to L2 learners regardless of whether their native language employs articles (Tryzna, 2009: 86).
Not only Tryzna (2009), but also Déprez, Guella and Sleeman (2011) questioned the
Fluctuation Hypothesis. Their investigation of the acquisition of articles did not focus on L2 learners with article-less L1 (like Polish or Russian) but on the acquisition of articles of the definiteness-based language French by L1 speakers of the definiteness-based language Dutch and Arabic. The participants were 23 Dutch learners (age 13 to 15) and 30 Arabic learners of French L2 (age 10 and 12). They all acquired French in class instruction. The task reproduced that of Ionin et al. (2004), a written forced-choice elicitation task consisting of different dialogues in French. The participants had to complete the target sentence with a missing article (definite or indefinite), or a blank. The Dutch learners of L2 French overused the definite article le significantly more in indefinite [+specific] contexts (68%) and overused the indefinite article un significantly more in definite [-specific] contexts (45%). The Arabic learners of L2 French also overused le more in indefinite [+specific] contexts (54%) and un more in [-specific] contexts (63%). Guella et al. say that these data show that Dutch and Arabic learners of French fluctuate in their choice of articles in a way comparable to learners who do not have an article-based L1. The participants acquire the French articles with a specificity bias fully comparable to L2 learners with article-less L1. This is unexpected, since both Arabic and French are definiteness-based languages. This suggests that L1 transfer does not prevent specificity-biased errors, contrary to the expectation of a parameter-based account (Deprez et al., 2011: 27). The results also show that the errors made by the participants appear to be subject to proficiency effects. Specifically, the more proficient the learner, the less frequent the errors (Deprez et al., 2011: 33).
In 2008, Ionin, Zubizarreta and Bautista-Maldonado compared 23 Russian and 26 Spanish L2 speakers of English with a test format that was modelled after the elicitation tasks Ionin et al. (2003, 2004). They made two important changes: the participants could not choose between given answers, but had to fill in the blank in the target sentence themselves, and this test included fillers targeting items other than articles.
This study found that the Russian speakers where less accurate than the Spanish speakers in their article use, despite having significantly higher proficiency and more L2-exposure. The two main error types are overuse of the with specific indefinites and overuse of a with non-specific definites. The learners are highly accurate on the other categories, specific
definites and non-specific indefinites (Ionin et al., 2008: 565). This effect is best explained by L1-transfer (Ionin et al., 2008: 569). It is argued that while L1-Spanish speakers rely on transfer, L1-Russian speakers, in the absence of transfer, have direct access to semantic universals, and fluctuate between them (Ionin et al., 2008). Speakers of Spanish can transfer their L1 (Spanish) setting of the Article Choice Parameter to their L2 English (in which the ACP setting is the same). Because of the absence of articles in Russian, speakers of Russian are not able to transfer article semantics from their L1 to their L2, so their learning proceeds through a combination of UG-access and input processing (Ionin et al., 2008: 574).
In summary, Ionin et al.’s (2004 and 2008) studies argue that a) L2 learners have direct access to parameters, and b) L2 learners with an article-less L1 initially fluctuate between the two settings of the Article Choice Parameter, namely definiteness and specificity (referentiality).
As for Polish, Ekiert (2005) examined use of articles in English by adult Polish speakers. She used a written test adapted from Goto Butler (2002), Liu and Gleason (2002) and Master (1974). The participants had to insert a/an, the, or zero article across five semantic types (generic nouns, referential definites, referential indefinites, non-referential nouns and idioms) (Ekiert, 2005: 3). 20 Polish adult L2 speakers of English participated in the research. One of the main findings of this study is for article choice. For example, the intermediate-ability speakers scored for referential definites and referential indefinites respectively 55,5% and 53,3% correct, and the high-ability speakers scored respectively 60% and 76,6%. For the overuse of the definite and indefinite articles, she gives mean proportions: for the low-ability speakers 2.3 and 3.1 for respectively overuse of the and overuse of a, and for the high-ability speakers 2.1 and 3.0. This shows the level of proficiency has no significant influence on article choice.
The other main finding is for article suppliance, namely article omission, which she calls overuse of the zero article (or “the failure to use any article”). For the overuse of the zero article, Ekiert again shows mean proportions: for the low-ability speakers 30.1 and for the high-ability speakers 17.2. Also in this case, overuse of the zero article decreased with an increase of proficiency level. This shows that input has its influence on the performance of article suppliance of L2 learners.
One explanation of the overuse of the zero article is interpreting this omission as indication of mastery of the zero article on the initial stages of L2 acquisition (Master, 1987, as cited in Master, 1997). Parrish (1987) and Master (1997) report that in the early stages of
(article-based) L2 acquisition by article-less L1 speakers the zero article dominates in all environments for articles. Both propose that the zero article is required first by learners whose L1 lack articles, followed by the definite article and finally the indefinite article. Thomas (1989) however, adopted another approach claiming that her participants ‘failed to use any article’. Therefore, Ekiert proposed that article omission might be due to L1 transfer.
Concluding, concerning Ionin’s Fluctuation Hypothesis (Ionin et al., 2004), the studies
mentioned above showed the same: L2 learners fluctuate between different parameter-settings (definiteness and specificity) until the input leads them to set the parameter to the appropriate value (definiteness). However, these studies on L2 article acquisition provide mixed results for the transfer part of the Full Transfer/Full Access Hypothesis (Schwartz & Sprouse 1994, 1996; Schwartz 1998a). Some provide evidence against this hypothesis, like Deprez et al.’s (2011). Others in favour of this hypothesis, like Tryzna (2009). These mixed results make it difficult to draw any conclusions regarding the underlying process of article acquisition in L2. In addition, most L2 studies on article acquisition do not systematically distinguish between article suppliance and article choice. The present study investigates both article suppliance and article choice in the L2 Dutch of L1 speakers of an article-less language, namely Polish.
2.5 Hypotheses and predictions
As described in the background section, I assume that article suppliance is a grammatical phenomenon, and article choice a pragmatic phenomenon. Recall further that because of the rapid development in L1, it is not clear what the exact contributions of pragmatic and grammatical knowledge to article acquisition are, and how these interact. Monolingual
children acquire correct article choice and correct article suppliance around the same time and are therefore not the best target group to disentangle pragmatic from grammatical knowledge. The L2 studies discussed in the background section focus on article choice, rather than article suppliance, and are thus not informative on the relative contribution of grammatical and pragmatic knowledge to the acquisition of articles either. In order to be able to tease apart the grammatical and pragmatic knowledge required for correct article use, the acquisition of articles of individuals who acquire Dutch as a second language can be an interesting starting point.
As mentioned, the grammatical difference between Dutch and Polish is the fact that Polish lacks overt articles for both definites and indefinites. Adult Polish L1 learners of Dutch
need to learn that (i) Dutch has overt articles, and (ii) Dutch distinguishes between definite and indefinite articles. As argued in the background section, the first type of knowledge is grammatical in nature. Assuming with Schwartz (Schwartz & Sprouse 1994, 1996; Schwartz 1998a) that L1 transfer in adult L2 acquisition takes place mainly in the language component of grammar, I hypothesize that the lack of articles in Polish influences article suppliance in Dutch in the adult L2 acquisition of Dutch by Polish L1 speakers. I therefore predict that initially, adult Polish L1 learners of Dutch omit articles. Furthermore, I hypothesize with Schwartz (Schwartz & Sprouse 1994, 1996; Schwartz 1998a), Ionin et al. (2004; 2008; 2009) and Guella, Déprez & Sleeman, 2008) that adult L2 learners have access to parameters, and thus to the Nominal Mapping Parameter. As Chierchia (1998) and Longobardi (2013) argue, Dutch and Polish belong to the same group of languages in terms of this parameter, and are both set for NP [+arg, +pred], meaning that even though Polish lacks overt articles, there is a similarity concerning the status of nouns, i.e., in both Dutch and Polish nouns can be either arguments or predicates. If L2 learners have access to the NMP in UG and there is
grammatical transfer, Polish L1 learners of Dutch can set the NMP for the Dutch value relatively quickly and easily: no re-setting is required. What Polish learners of Dutch have to learn subsequently, is that this setting for Dutch means that nouns with the lexical status [+pred] need an overt article in order to occupy an argument position. However, since
presumably, the same set of nouns are predicates in both Polish and Dutch, this should not be too difficult a task for the Polish learners of Dutch. It is therefore predicted that adult Polish L1 learners of Dutch will initially omit articles, but will acquire overt articles relatively quickly, and that they can reach near-native levels in terms of article suppliance.
The second type of knowledge – the distinction between definite and indefinite articles – is argued to be pragmatic in nature. Hypothesizing that pragmatic knowledge is constant across languages and acquired before adulthood, it can be said that Polish adult speakers do not lack the Concept of Non-shared Assumptions. In addition, hypothesizing with Schwartz (Schwartz & Sprouse 1994, 1996; Schwartz 1998a) and Ionin et al. (2004; 2008; 2009) and Guella, Déprez & Sleeman, 2008) that adult L2 learners have access to UG parameters, they should have access to the Parameter of Article Semantics (PAS). This access needs to be direct, as the PAS is irrelevant in their L1, Polish. As argued, a necessary condition for setting the PAS is the pragmatic Concept of Non-Shared Assumptions. If speaker and hearer
assumptions are not systematically distinguished, notions such as speaker and hearer beliefs, and therefore definiteness and referentiality cannot be established. Assuming that adult Polish L2 learners of Dutch have mature pragmatics, and that they pay attention to the Dutch input, I
predict that adult Polish L2 learners of Dutch set the PAS correctly from the beginning, and thus that they do not fluctuate between the two settings. Concretely, this means that they should perform well on article choice whenever they use overt articles. These hypotheses and predictions are listed in (22):
(22) Hypothesis I:
Article suppliance is a grammatical phenomenon.
Hypothesis II:
Article choice is a pragmatic phenomenon.
Hypothesis III:
Full Access: L2 learners have access to parameters
Hypothesis IV:
The Nominal Mapping Parameter has the same setting in both Dutch and Polish, namely for [+pred, +arg]
Hypothesis V:
The Parameter of Article Semantics is irrelevant (and therefore inactive) in article-less languages such as Polish
Hypothesis VI:
Full transfer: Initially, the L1 grammar transfers to the L2
Hypothesis VII:
L2 learners have mature pragmatics, and thus have the Concept of Non-Shared Assumptions
Predictions regarding article suppliance:
(i) Initially, adult Polish L1 learners of Dutch sometimes omit articles (Polish predicates do not require overt articles to turn into arguments).
(ii) With enough input, adult Polish L1 learners of Dutch show near-native-like article suppliance (no re-setting of the NMP is required).
Prediction regarding article choice:
(iii) Adult Polish L2 learners of Dutch perform well on article choice whenever they produce overt articles (and do not fluctuate between the two PAS settings).
If the predictions are borne out, this provides evidence for the hypotheses that article suppliance is grammatical in nature, and article choice pragmatic, and that both types of knowledge are necessary to correctly use Dutch articles. It would further show that adult L2 speakers have access to parameters, and that L1 transfer mainly takes place in the component of grammar. Finally, it would provide support for the hypothesis that the relevant pragmatics is constant across languages, and that it no longer develops after childhood.
3. Method
In order to test my hypotheses and predictions, I carried out an elicited production task with 30 adult Polish speakers, learning Dutch as a second language. The participants, materials and procedure are described below.
Participants
The L2-participants were 30 Polish speakers, between the ages of 22 and 65 (mean age: 44). They had learned Dutch between the ages of 10 and 57 and live in the Netherlands. 26 of the participants had learned Dutch in a formal situation: through a course at an adult education center, assimilation course or at their university in Poland. Four of them had learned Dutch through experience, without lessons or courses. Their stay in the Netherlands varied between 18 months up to 32 years.
A control group of 30 Dutch native speakers between the ages of 21 and 57 (mean age: 35) also participated in the experiment and the proficiency test. They were all born and raised in the Netherlands.
TABLE I Participants
Group Age Mean age # of females # of males Total #
Polish 22 – 65 44 25 5 30
Dutch 21 – 57 35 19 11 30
Materials and procedures Task
The actual experimental study is based on Schaeffer’s (1999; in progress) article experiment, namely an elicited production task. The participant looks at a picture or short video fragment depicting a protagonist and a(n) (potential) action. At the end of each scene, the participant is asked to explain what s/he just saw to a blindfolded teddy bear. The teddy bear has been introduced to the participant as a curious animal, but because he cannot see the screen, the participant is asked to explain the action that has been shown as clearly as possible. The experiment contained five conditions, as listed in (23):
(23) Five conditions of the elicited production task
Definite contexts Indefinite contexts
(a) Definite Non-referential
(b) Indefinite non-referential – incomplete object (c) Indefinite non-referential – non-existing object
Referential (past)
(d) Indefinite referential – past (e) Indefinite referential – partitive
Besides one definite condition, there are two indefinite non-referential conditions and two indefinite referential conditions. There is a reason for the choice for the difference between the conditions. The difference between the indefinite referential conditions has to do with the different objects. In both condition (b) indefinite non-referential – incomplete object, and (c) indefinite non-referential – non-existing object, the environment is C, which means that the
referent is neither believed by speaker nor hearer, and not part of the common ground. In (b), the object is incomplete and for adults it cannot correspond to an existing entity in the world. In this case, the participants should use the indefinite article een. However, Schaeffer (1999: 14) shows that for children it seems to be sufficient to have an incomplete object as the referent, as long as they can identify the object (Schaeffer, 1999: 14). Then it can be considered familiar to the speaker and thus it represents environment B, in which children sometimes use the definite article de. Therefore, another indefinite non-referential condition was developed.
The participants were presented with 6 items for each condition. The total number of 30 test items was interspersed with 22 other items, testing a different grammatical phenomenon, namely object-scrambling. The two types of test items served as fillers for each other. An explanation and illustration of each condition is provided in (24) (a) – (e):
(24)
a) Definite condition
In this condition a picture of a character next to an object. The participant was asked what he saw next to the character (for example a banana or a ball). Next, a short video fragment was shown where the character performed an action with the object. Because both the character and the object where introduced before, the participants should use the definite article de or het, when he is asked to explain what the character did.
Teddy: Hé, wie zie je?
Hey, who see you?
‘Hey, who do you see?’
Participant: Katrijn.
Teddy: En wat nog meer?
‘And what else?’
Participant: Een bal.
‘A ball.’
(Hand-puppet pushes the ball) Teddy: Wat deed zij daarnet?
‘What did she just do?’
Participant: Zij duwde de bal. Ze duwde *een bal. Ze duwde *bal. ‘She pushed the ball.’ ‘She pushed *a ball.’ ‘She pushed *ball.’
b) Indefinite – non-referential (incomplete)
In this condition, a picture was shown of a character that is busy completing something, for example painting a house or building a castle. The action is not
completed yet, and is only known to the speaker and not to the hearer, so the indefinite article een is correct.
Situation: Miffy building a sandcastle (not finished) Teddy: Hé, wie zie je?
Hey, who see you?
‘Hey, who do you see?’ Participant: Miffy.
Teddy: En wat is Miffy aan het doen? ‘And what is Miffy doing?’
Participant: Zij bouwt een zandkasteel/Zij is een zandkasteel aan het bouwen. Zij bouwt *het zandkasteel/Zij is *het zandkasteel aan het bouwen. Zij bouwt *zandkasteel/ Zij is *zandkasteel aan het bouwen. ‘She is building a sandcastle.’ ‘She is building *the sandcastle.’ ‘She is building *sandcastle.’
c) Indefinite – non-referential (non-existent)
In order to have a true environment C, another indefinite non-referential condition was designed, in which there is no existing direct object. The participants have to think of an object in the future, which they choose themselves. The object is therefore non-existing. In other words, this corresponding entity is not incomplete, but it is absent. It is assumed that there is no entity believed to exist by either speaker or hearer, so the correct article should be the indefinite article een.
Big Bird: Pino gaat naar het bos en daar gaat hij iets tekenen. Wat denk je dat hij gaat tekenen?
‘Big Bird is going to the forest and he is going to draw something there. What do you think he is going to paint?’ Participant: Hij gaat een boom tekenen. Hij gaat *de boom tekenen. Hij gaat *boom tekenen.
‘He is going to paint a tree.’ ‘He is going to paint *the tree.’ ‘He is going to paint * tree.’
d) Indefinite – referential (past)
In the indefinite referential conditions the difference lies in what is shown on the picture. In this condition the object is shown and is finished, the character has for example finished painting a castle. It is known to the speaker and refers to an entity in the world. However, the
referent is not known to the hearer (because the hearer in this case is blindfolded), so the indefinite article een should be used.
Situation: Jip has just finished painting a house Teddy: Hé, wie zie je?
Hey, who see you?
‘Hey, who do you see?’
Participant: Jip.
Teddy: Wat heeft Jip zojuist gedaan?
What has Jip just done?
‘What did Jip just do?’
Participant: Hij heeft een huis geschilderd. Hij heeft *het huis geschilderd. Hij heeft *huis geschilderd.
‘He has painted a house.’ ‘He has painted *the house.’ ‘He has painted *house.’
e) Indefinite – referential (partitive)
In this condition, there is being referred to a partitive, a specific item from a set. This
condition is added to the other four conditions that are designed by Schaeffer (1999). On the picture a situation is shown of a character that is for example holding a present, and on the floor there are three more presents. Referentiality can be accomplished because the subject in the picture is only holding one of the objects, so the indefinite article een is the correct article.
Situation: Cookie monster holding one cookie, 5 cookies next to him Teddy: Hé, wie zie je?
Hey, who see you?
‘Hey, who do you see?’ Participant: Cookie monster.
Teddy: Wat is Koekiemonster aan het doen? ‘What is Cookie Monster doing?’
Participant: Hij eet een koekje. He eet *het koekje.
Hij eet *koekje.
‘He is eating a cookie.’ ‘He is eating *the cookie.’ ‘He is eating *cookie.’
Each participant was tested individually in a quiet environment. The answers given were recorded by an audio device and noted on a pre-designed score-sheet by the experimenter. There were a few questions in the task that could be produced with both an article and without article. In that case the noun was interpreted as a mass noun, as in (25). This was possible in four cases, and these responses were not scored as article omission.
(25) Indefinite referential (past): een (‘a’) required Situation: Ernie is baking a pizza
Teddy: Hé, wie zie je?
Hey, who see you?
‘Hey, who do you see?’
Participant: Ernie.
Teddy: Wat heeft Ernie zojuist gedaan?
What have Ernie just done?
‘What did Ernie just do?’
Participant: Hij heeft een pizza gebakken/ Hij heeft pizza gebakken. He have a pizza baked / He have pizza baked ‘He baked a pizza / He baked pizza’
Proficiency test
In order to get a picture of their general proficiency in Dutch, the Polish participants had to complete Keijzer’s (2007) C-Test. The C-Test is an instrument to measure general language proficiency based on their vocabulary. The original C-test consists of five texts, but time limitations forced us to use only three out of these five texts. Each text contained 20 gapped items and only the begin letter(s) of the missing word was/were given. The participants had to fill in the word that they thought was correct. Based on the number of correct filled in words, the participants were assigned different proficiency scores. An example of one paragraph of text one is shown in (26), the other two texts can be found in the Appendix.
(26) Ik houd van Nederland en niet zo’n beetje ook. Waarom ik van het land houd is niet alleen omdat velen van wie ik houd hier leven, nee, het is me__________ dan
d__________ De groo__________ reden v__________ mijn lie__________ voor het land ko__________ voort u__________ het feit dat al__________ zo geor__________ en syste__________ is.
Translation:
I love the Netherlands and not too little. Why I love the country is not only because many of the people I love live here, no, it’s mo__________ than t__________. The most important reason f__________ my lo__________ for the country co__________ from the fact that eve__________ is so org__________ and syste__________.
4. Results & discussion Proficiency task
The proficiency task was used to determine the levels of proficiency of the Polish
participants. Only five participants had a score lower than 30 out of 60. The native Dutch control group also had to fill in the proficiency test so the results of both groups could be compared. The control group had an average score of 95% correct answers; the Polish group 72%. As shown above in the results of the task, a distinction between the levels of proficiency of the Polish participants could be made on the basis of this test. None of the Dutch
participants scored beneath 78%. 16 of the Polish participants had a score higher than 78%. The thirty participants were divided into one group of 16 ‘advanced’ L2 learners and another group of 14 ‘intermediate’ L2 learners.
TABLE II
Polish participants: proficiency task results
Participant Score Participant Score
1 96,66% (58/60) 16 78,33% (47/60) 2 95% (57/60) 17 76,66% (46/60) 3 95%(57/60) 18 76,66% (46/60) 4 90% (54/60) 19 71,66% (43/60) 5 90% (54/60) 20 68,33% (41/60) 6 90% (54/60) 21 66,66% (40/60) 7 88,33% (53/60) 22 65% (39/60) 8 86,66% (52/60) 23 61,66% (37/60) 9 86,66% (52/60) 24 58,33% (35/60) 10 86,66% (52/60) 25 53,33% (32/60) 11 85% (51/60) 26 46,66% (28/60) 12 83,33% (50/60) 27 38,33% (23/60) 13 80% (48/60) 28 30% (18/60) 14 80% (48/60) 29 20% (12/60) 15 78,33% (47/60) 30 10% (6/60)
Elicited Production Task
I will discuss the results of the elicited production task on the basis of the predictions
formulated in section 2.5. The first prediction concerning article suppliance is that (i) initially, adult Polish L1 learners of Dutch will sometimes omit articles. Table II shows the overall results on article omission for the Polish L2 speakers of Dutch, compared to the control group.
TABLE III
Article suppliance: omission across conditions
L2 intermediate L2 advanced Controls
35% (145/420) 10% (49/480) 0% (0/900)
As we can see, all Polish L2 learners of Dutch omit articles. This is illustrated by the following responses:
(27) (a) Definite
Katrijn deed ø balletje rollen. Katrijn did ø ball roll ‘Katrijn let ø ball roll.’
(b) Indefinite non-referential (incomplete) Nijntje is ø kasteel aan het bouwen. Nijntje is ø castle building
‘Nijntje is building ø castle.’
(c) Indefinite non-referential (non-existing) Koekiemonster gaat ø konijn schilderen. Koekiemonster goes ø bunny to paint ‘Koekiemonster is going to paint ø bunny.’
(d) Indefinite referential (past) ‘Jip heeft ø huis getekend’
(e) Indefinite referential (partitive) Janneke speelt met ø bal. Janneke plays with ø ball ‘Janneke is playing with ø ball.’
As indicated in Table III the intermediate proficiency group displays a much higher
percentage of article omission than the advanced proficiency group. The difference between the article omission of the advanced L2-speakers and the intermediate L2-speakers is significant, according to a Kruskal Wallis test (U = 54, p = .015)4. Also the difference between the advanced L2-speakers and the Dutch control group is significant (U = 60, p = .000). This confirms my first prediction on article suppliance that adult Polish learners of Dutch initially omit articles. I argue that since Polish does not require [+pred] nouns to have an overt determiner to be realized as an argument, and because of the Full Transfer
Hypothesis, the Polish L2 learners of Dutch initially transfer this system to Dutch. The intermediate (rather than the beginner) level of the first Polish L2 group (according to their proficiency scores) explains why articles are not always omitted by this group, but ‘only’ 35% of the time.
4 Instead of a one-way ANOVA, a Kruskall Wallis test was used. The data does not meet the requirements for
Furthermore, the results indicate that Polish L2 learners of Dutch can reach near-native levels of article suppliance once they become highly proficient in the Dutch language, confirming my second prediction. No re-setting of the NMP is required, because as hypothesis IV says: the parameter setting of the NMP for Dutch and Polish is the same and because of Full Transfer, the Polish learners of Dutch transfer the setting of the NMP.
TABLE IV
Article suppliance: article omission per condition
L2 total L2 intermediate L2 advanced Controls
Definite 29% (52/180) 45% (38/84) 15% (14/96) 0% (0/180) Indefinite non-referential (incomplete) 27% (48/180) (36/84) 43% (12/96) 13% (0/180) 0% Indefinite non-referential (non-existing) 16% (29/180) 27% (23/84) 6% (6/96) 0% (0/180) Indefinite referential (past) 15% (27/180) 27% (23/84) 4% (4/96) 0% (0/180) Indefinite referential (partitive) 21% (38/180) 30% (25/84) 14% (13/96) 0% (0/180)
When we look at the percentages in Table IV, we see that the advanced Polish speakers all performed better compared to the intermediate speakers. This is just like Hypothesis IV predicted. At first, L1 grammar transfers to the L2, which causes that less advanced speakers will perform less accurate than more advanced speakers.
Of all conditions, the Polish participants omit articles mostly in the definite condition, followed by the indefinite non-referential (incomplete) condition. The 29% omission in the definite condition resembles the findings of Ionin, Ko and Wexler: they found 24% article omission in their definite condition (Ionin, Ko & Wexler, 2004: 46). When we combine the percentages of both indefinite referential conditions and both indefinite non-referential conditions, we see that for the indefinite referential condition in this study the percentage is 18% and for the indefinite non-referential condition 21%. The results of Ionin, Ko and Wexler, were respectively 35% and 17%. The results of the indefinite non-referential
conditions bear resemblance to each other. Unfortunately, they do not give an explanation for these results, because they focus more on article choice instead of article suppliance. A
