Egocentricity: L2 Article Choice
Reconsidered
Name: Chantal Horselenberg Student Number: 10356061
Master’s Thesis
MA Linguistics of European Languages: English Supervisor: Dr. Jeannette Schaeffer
University of Amsterdam 30 June 2016
Abstract
This paper examines L2 article choice by L1 Dutch adolescents learning English as a second language. Specifically, this study investigates what these participants do when they face the choice between a definite and indefinite article. Results of an article elicitation task are interpreted in light of two existing hypotheses, which make contrasting predictions for article choice behaviour. The first hypothesis tested is the Transfer Hypothesis, as proposed by Ionin et al. (2008). This hypothesis predicts target-like L2 article choice for [+article] L1 learners of a [+article] L2 in the case of similar article semantics settings, since these settings are
believed to be transferred from the L1 to the L2. The second hypothesis tested is the Computational Egocentricity Hypothesis, as put forward by Deprez et al. (2011). This hypothesis predicts fluctuation errors when definiteness and specificity are in conflict for [+article] L1 learners of a [+article] L2, due to an increased computational load of learning a second language. Results show no clear preference for either hypothesis, but rather suggest a development influenced by proficiency. Low proficiency groups show some fluctuation in indefinite conditions, whereas high proficiency groups perform in line with the Transfer Hypothesis. Further research into the onset of second language acquisition and the
development from that point onwards should be conducted in light of these two conflicting hypotheses.
Keywords: Article Choice, Second Language Acquisition, L1 Transfer, Computational Egocentricity, Dutch, English
Abbreviations
L1 First Language
L2 Second Language
NP Noun Phrase
DP Determiner Phrase
[-article] Language without articles [+article] Language with articles [+definite] Definite
[-definite] Indefinite [+specific] Specific [-specific] Nonspecific
Acknowledgements
I would like to thank the students and teachers of the Clusius College in Hoorn, the
Netherlands and Girard High School in Girard, USA for their cooperation in the experiments. Furthermore, I would like to give thanks to the various PhD-students I consulted on the methodology of this study, Dr. Ionin for letting me use her elicitation task and Dr. Sleeman for further illustrating her paper. I am very grateful to Stefan Zuurbier and Yiwen Zou for their help on statistics and Dr. Fischer for her insights on English grammar and stylistics. I thank my family and friends for their unrelenting support and patience during this project. Finally, I would like to thank my supervisor Dr. Schaeffer for her constructive feedback and her faith in this thesis.
Table of Contents
Abstract………..2
Acknowledgements………4
1. Introduction………7
2. Theoretical Background……….9
2.1 Second Language Acquisition………..9
2.2 The English Article System………10
2.3 Previous L2 acquisition studies on article choice………...15
2.3.1 Ionin, Ko & Wexler (2004)………..15
2.3.2 Ionin, Zubizaretta and Maldonado (2008)………...17
2.3.3 Deprez, Sleeman & Guella (2011)………..19
2.4 The Present Study………..24
2.5 The Dutch Article System……….24
3. Hypotheses & Predictions………...26
4. Method………28
4.1 Participants………28
4.2 Experiment………30
4.2.1 Questionnaire……….30
4.2.2 Anglia Placement Test………...31
4.2.3 Article Elicitation Task………...32
4.2.4 Procedure………...34
4.3 Coding & Statistics………...35
5. Results & Discussion………..36
5.1 Questionnaire………36
5.2.1 Item Effect...38
5.3 Elicitation Task: Second Language Learners...41
5.3.1 Proficiency Groups...41
5.3.2 Low Proficiency...42
5.3.2.1 Item Effect...43
5.3.2.2 Transfer Hypothesis...46
5.3.2.3 Computational Egocentricity Hypothesis...47
5.3.3 Medium Proficiency...51 5.3.3.1 Item Effect...51 5.3.3.2 Transfer Hypothesis...53 5.3.3.3 Computational Egocentricity...54 5.3.4 High Proficiency...56 5.3.4.1 Item Effect...57 5.3.4.2 Transfer Hypothesis...59 5.3.4.3 Computational Egocentricity...60 5.3.5 Proficiency...62 5.4 General Discussion...64 6. Conclusion...66
7. Suggestions for Further Research...66
8. References...68
Appendices...72
Appendix A: Questionnaire...72
Appendix B: Anglia Placement Test...74
Appendix C: Article Elicitation Task...75
1. Introduction
This paper investigates article choice in English as a second language (L2) by mother-tongue (L1) adolescent speakers of Dutch. The English article system distinguishes between definite and indefinite articles, as presented in the examples below (in bold):
(1) Definite (specific) article
The lady next door was walking her dog yesterday.
(2) Indefinite (nonspecific) article
My sister wants to buy a car.
Even though article choice in English, like Dutch, depends on the semantic notion of definiteness (i.e. in the common ground of the speaker and hearer), articles can also be specific (i.e. known to the speaker) or nonspecific (i.e. unknown to the speaker). Definiteness and specificity are illustrated in example 1 and 2. Previous studies on second language acquisition show contrastive results with regard to L2 article choice, generating two different hypotheses on L2 article choice performance. Ionin et al. (2008) finds that [+article] L1 learners of a [+article] L2, in the case where both languages have article choice governed by definiteness, perform target-like on L2 article choice. This phenomenon is explained by the Transfer Hypothesis: [+article] L1 learners of a [+article] L2 transfer the definiteness setting of their mother tongue to their L2, resulting in correct article choice. Deprez et al. (2011) test [+article] Dutch L1 adolescents on [+article] French L2 on article choice; both Dutch and French are definiteness languages. Results showed that Dutch participants made errors in definite nonspecific and indefinite specific conditions, i.e. when the sematic notion of definiteness and specificity are in conflict. This pattern has been found for [-article] L1 learners of a [+article] L2 and accounted for by the Fluctuation Hypothesis: [-article] L1 speakers have no setting for article choice due to the lack of articles in their native language, thus fluctuate between the semantic notions of definiteness and specificity until input leads
them to choose a setting. However, the Fluctuation Hypothesis cannot account for the article choice behavior by the Dutch participants, as they are supposed to transfer their definiteness setting from their L1 to the L2. Instead, Deprez et al. (2011) propose a different solution: due to an increased computational load as a result of learning an L2, learners make egocentric errors, resolving to specificity settings (speaker-centered) instead of definiteness settings (based on common ground).
Considering the mixed results regarding article choice by [+article] L1 learners, this study poses the following research question: What do adolescent learners of English
(definiteness-based) with Dutch as their L1 (also definiteness-based) do when they face the choice between a definite and an indefinite article?
Based on data presented in the previous Ionin et al. (2008) and Deprez et al. (2011), article choice performance can move in two directions; either target-like, in line with the Transfer Hypothesis, or show fluctuation, in line with the Computational Egocentricity Hypothesis. In addition to testing these hypotheses, attention is paid to different levels of proficiency, as in previous studies an increase in proficiency led to a decrease in errors in article choice (Ionin et al. 2004; Zdorenko & Paradis 2008; Deprez et al. (2011)).
The paper is organized according to the following structure: Section 2 further elaborates on the English and Dutch article system and gives an overview of previous
research in the field of second language acquisition with regard to article choice. In section 3 I present the hypotheses and predictions, which are tested by the experiment discussed in section 4. Section 5 presents the results of the different tasks and discusses them in relation to the hypotheses. Concluding remarks and suggestions for further research are offered in sections 6 and 7.
2. Theoretical Background
In this chapter I first lay out the basic and relevant theoretical premises of L2
acquisition, followed by a description of the linguistic topic that the current study investigates, namely article choice. Furthermore, I discuss relevant previous L2 studies on this linguistic phenomenon, and formulate my hypotheses and predictions for the population I test: adolescent L1 speakers of Dutch learning English.
2.1 Second Language Acquisition (L2)
In the process of acquiring a second language, learners can be guided by three types of knowledge. The first source is L2 input, which can be naturalistic or classroom-based, and is vital for the end stage of target-like L2 attainment. The second source is first language (L1) transfer, where structures from the L1 are transferred to the L2, resulting in errors when these structures differ and target-like behaviour when the structures are the same in both languages (Schwartz & Sprouse, 1996). Finally, when L2 learners’ behaviour cannot be traced back to the L1 or the L2, linguists argue that learners are guided by innate linguistic knowledge (see Schwartz & Sprouse 2013, for an overview). As the first two sources have generally been accepted into the field of second language acquisition and the latter source of knowledge has caused much debate over the past years, some more specification of this innate faculty is needed. In the nineteen fifties, Chomsky pioneered with the theory of Universal Grammar, stating that humans have an innate faculty of language in the brain and that human languages share certain universal properties. Chomsky (1981) further specified this theory in the shape of principles and parameters. Principles constrain native grammars, such as the principle that all languages have subjects in sentences. Parameters are switches set to the appropriate value of a specific language based on input, for example an overt or covert subject. However,
Chomsky mostly focused on L1 acquisition. Schwartz & Sprouse (1994) combined UG theory with L1 transfer in their Full Transfer/Full Access hypothesis. This hypothesis states that the
initial state of L2 acquisition is the L1, defined as Full Transfer. In addition to L1 transfer, L2 learners have full access to UG settings through their L1, but these settings can be changed as a result of L2 input. Previous research in the field of second language acquisition has shown that the acquisition of English article choice is an aspect of language difficult to master (Robertson, 2000; Leung, 2001; Ionin et al. 2004, 2008). Article-misuse occurs most often in cases where learners of which the L1 lacks articles, causing article omission or article
substitution (eg. a for the). Less research has found article misuse by learners with an L1 with articles (Hawkins et al. 2006; Deprez et al. 2011). Before diving into English article
acquisition, a definition of the forms and meanings of the English article system is in order.
2.2 The English Article System
The definition of an article used in this paper comes from Schaeffer & Matthewson (2005), who define articles are functional items correlated to argumenthood and which are neither quantifiers nor determiners. Heim (1991) argues that article forms are mapped to two meanings: definiteness and specificity. Ionin et al. (2004:5) define definiteness and specificity as follows:
If a Determiner Phrase (DP) of the form [D NP] is . . .
A. [+definite], then the speaker and hearer presuppose the existence of a unique individual in the set denoted by the NP.
B. [+specific], then the speaker intends to refer to a unique individual in the set denoted by the NP and considers this individual to possess some noteworthy property. This definition illustrates that articles are discourse-related, where definiteness is based on common ground between speaker and hearer, while specificity depends on speaker knowledge only. English, among Dutch, French and many other languages, can be classified as having an article system based on the semantic notion of definiteness, with the as the definite article and
the is used for definite contexts and the article a is used for indefinite contexts. The sentences
(1)-(4) show examples of the four possible combinations with definiteness and specificity in English:
(1) [+definite, +specific]
I saw a girl yesterday. The girl was wearing a red dress. (2) [+definite, -specific]
I am curious who will be the winner of the tournament! (3) [-definite, +specific]
I have an uncle who could help you with that. (4) [-definite, -specific]
I would like to have a puppy for my birthday.
In (1), girl in the second sentence needs to take the definite article the in English, as the girl refers to the previously mentioned girl in the first sentence. Thus the common ground has been established between speaker and hearer and the speaker is free to use the definite article
the to refer to this entity in the common ground. The girl is also specific, as the speaker
identified a specific girl he or she saw. The definite article is also used in (2), although the semantic situation is different here; the winner is in the common ground not because this entity has been previously introduced, but because it is part of the world knowledge of the speaker and hearer. The speaker presupposes that the hearer also knows that tournaments always have winners and uses the article the. The winner is non-specific, as the description does not refer to a specific individual; the winner of the tournament is not yet known.
Example (3) takes an indefinite article, as the uncle is only known to the speaker and therefore not in the common ground of the hearer. An uncle is specific, since the speaker knows the unique individual as uncle. The indefinite article a is used again in (4), where the puppy is not in the common ground of the speaker and hearer as the speaker introduces an entity unknown
to the hearer. A puppy is not only indefinite but also non-specific since the speaker does not have a specific puppy in mind, but wishes any puppy for his or her birthday.
The following table schematises the form-meaning mapping of the English article system: Table 1: The English Article System
[+definite] [-definite]
[+specific] the a
[-specific] the a
Many languages with overt articles have an article system based on definiteness. However, researchers claim to have discovered languages in which article choice is determined by specificity instead of definiteness, such as Salish (Matthewson, 1998) and Samoan (Mosel & Hodvaugen, 1992). Since Ionin et al. (2004, 2008) discuss Samoan as a prime example of a language in which article choice is guided by specificity, the current paper will follow their lead. In Samoan, definite and indefinite specific referents take the article le, while definite and indefinite non-specific referents take the article se. The following examples from Samoan illustrate this specificity distinction (examples were taken from Ionin et al. 2008, originally from Mosel & Hodvaugen, 1992):
(5) Examples of specificity in Samoan
A. Use of le in the specific indefinite condition
C. Use of le in the specific definite condition
D. Use of se in the nonspecific definite condition
These examples illustrate that the article system of Samoan is defined by specificity, not definiteness. In 5A, the girl called Sina is an example of a specific indefinite and takes the article le, since the speaker knows the girl by name, but introduces her to a hearer, to whom the girl is unknown. The lady in 5B can be classified as a nonspecific indefinite, as she is unknown to the speaker and has not yet been introduced to the hearer, thus taking the nonspecific article se. 5C presents the man as a specific definite; he has been introduced before and is thus known to speaker and hearer, taking the specific article le. The last
condition, the nonspecific definite condition is identified in 5D as the family. Here, the family is definite, since having a family is part of shared world knowledge. However, family is nonspecific, as the speaker mentions whoever that may be, showing that he or she is not familiar with the hearer’s family, thus taking the nonspecific article se. The following table schematises the form-meaning mapping of the Samoan article system:
Table 2: The Samoan Article System
[+definite] [- definite]
[+specific] le le
[-specific] se se
Matthewson (1998), after observing differences between specificity-based article systems such as Salish and definiteness-based article systems such as English, proposes that native speakers set a parameter to govern their article choice: The Common Ground Parameter. This parameter entails that determiners may either access the common ground, such as in English, or not access the common ground, such as in Salish or Samoan. This parameter is further developed by Schaeffer & Matthewson (2005), who name it the Parameter of Article
Semantics (PAS). The PAS dictates that when languages distinguish between more than one article, this distinction relies on either speaker beliefs (Salish and Samoan) or common ground (English). Ionin, Ko & Wexler (2004) too introduce a parameter based on differences between English and Samoan, which is comparable to the PAS, but uses different terms for similar distinctions. They propose the Article Choice Parameter (ACP), which states that speakers of a language set this parameter to either definiteness (common ground) or specificity (speaker beliefs).
Before turning to L2 acquisition of English articles and the role of the ACP, a brief description of L1 acquisition of English articles is required, as problems do arise in this process. Maratsos (1974, 1976) and Schaeffer & Matthewson (2005) investigate L1 English acquisition of articles by children under 5, who overgenerate the definite article in indefinite contexts. Maratsos (1974, 1976) argues that these errors are egocentric, i.e. that the speaker assumes the hearer to have the same knowledge as himself. Schaeffer & Matthewson (2005) improve this explanation with their theory of the Concept of Non-Shared Assumptions (CNSA); young children lack this pragmatic concept due to their immature pragmatics, causing their inability to take hearer beliefs into account, which results in their overgeneration
of the definite article. This overgeneration of the definite article becomes less frequent and disappears around the age of 8, as Karmillof-Smith (1979) finds by studying L1 acquiring French children. Van Hout, Harrigan and de Villiers (2010) test the comprehension and production of articles by English children under 5 to account for their overgeneration of the definite article with Optimality Theory. Children under 5 have the same set of constraints as adults but start out with the constraints being unranked, causing incorrect production of the (non-determined and determined). These errors disappear after the age of 5, when children’s constraints become ranked and their article production target-like.
Now that I have described article choice in native child and adult speakers of English, I will turn to a discussion of some previous studies on the L2 acquisition of article choice, including different hypotheses related to input, L1 transfer, and UG, as mentioned in section 2.1.
2.3 Previous L2 Acquisition Studies on Article Choice
2.3.1 Ionin, Ko & Wexler (2004)
In 2.1 I noted that learners, especially those speaking a first language which does not mark articles, have difficulty acquiring articles in the L2. One of the studies which tested L2 learners in this area was a study done by Ionin et al. (2004). The authors tested 30 L1 Russian and 40 L1 Korean participants learning English as a second language, and 14 L1 English control subjects. Russian and Korean as first languages are interesting for L2 article
acquisition, as they do not mark definiteness and specificity with articles. Participants had to complete a forced elicitation task testing the four article conditions [+/- definiteness, +/- specificity], in which they had to choose between the, a and leaving the blank open. Results showed that participants produced a significant amount of errors in English article choice. However, these errors were not distributed at random, but displayed a specific pattern with regard to the four different conditions, as organized in the table below. Participants appeared
to overuse the definite article the in [-definite, +specific] conditions while using the correctly in [+definite, +specific] conditions. The same pattern emerged for the use of indefinite article
a; participants overgenerated this article in [+definite, -specific] contexts, while using a
target-like in [-definite, -specific] conditions.
Table 3: Performance on L2 English Article Choice by L1 Korean and Russian Speakers
Participants [+definite] [-definite]
[+specific] Correct use of the Overuse of the
Russian L1 79% 36%
Korean L1 88% 22%
[-specific] Overuse of a Correct use of a
Russian L1 33% 84%
Korean L1 14% 93%
The Russian and Korean L1 learners of English appeared to perform target-like on article choice in conditions where definiteness and specificity support each other, i.e. in [+definite, +specific] and [-definite, -specific] conditions. Only when definiteness and specificity are in conflict did these learners make errors: regardless of whether the context is definite or indefinite, learners seem to use the indefinite article in [-specific] conditions and the definite article in [+specific] conditions. This association of English articles with specificity is curious as this cannot be explained from the aspect of L1 transfer: Korean and Russian learners do not have any articles in their language. These learners are neither influenced by the L2 input, since articles in English depend on definiteness. Ionin et al. (2004) offer an explanation for this pattern based on a binary universal parameter, which they name the Article Choice Parameter (ACP). This parameter entails that humans are born with the ability to base their article choice on one of two values, definiteness or specificity, depending on the language input. Ionin et al. (2004) propose that L1 Korean or Russian speakers never set the ACP to definiteness or specificity, because this parameter is never triggered due to the lack of input with regard to articles. The absence of ACP setting entails that article-less L1 speakers have full access to both settings through Universal Grammar. The authors propose a Fluctuation
Hypothesis, which states that when learning an L2 with articles, these L1 speakers initially fluctuate between the definiteness and the specificity settings until sufficient L2 input leads them to set the parameter correctly: either to definiteness or specificity. This fluctuation can explain the overuse of the in [definite, +specific] settings and the overuse of a in [+definite, -specificity] by article-less L1 learners of an L2 with articles. Ionin et al. 2004 also
investigated the effect of proficiency on L2 article choice by dividing the L1 Russian and Korean participants into intermediate and advanced learners of English. Results revealed proficiency to have a marginal effect on the overuse of the and a for the Russian
participants: advanced learners of English were slightly more accurate with regard to article choice than intermediate learners. The Korean learners did not significantly improve on their article choice, most likely due to the fact that only 7 of the 40 participants were labelled as intermediate.
In Ionin et al.’s (2004) study, the Korean and Russian learners of English have not yet set their ACP to definiteness for English and thus sometimes use the and a for marking specificity rather than definiteness. This behaviour has been attested in other studies testing [-article] learners of a [+[-article] L2 as well, providing further support for the Fluctuation Hypothesis (Hawkins et al., 2006; Reid et al. 2006; Ionin et al. 2008).
2.3.2 Ionin, Zubizaretta and Maldonado (2008)
The Fluctuation Hypothesis predicts fluctuation for [-article] learners of a [+article] language, as these learners have not yet set their ACP. However, this hypothesis makes no predictions for learners who have already set the ACP to either definiteness or specificity in their L1 and are learning a language with articles that has the same or a different ACP setting. Besides testing 23 Russian L1 learners of English, Ionin et al. (2008) tested 24 Spanish L1 learners of English on L2 article choice. Participants were required to complete a forced elicitation task inspired by Ionin et al. (2004), but subjects had the freedom to complement
blanks in dialogues with any item they considered suitable. A control group of 6 native English speakers was tested, who reached ceiling accuracy on the elicitation task.
Similar to English, Spanish is a language where articles are categorised according to definiteness. The Russian L1 learners of English were predicted to behave according to the Fluctuation Hypothesis with regard to English article choice, while the Spanish learners of English were expected to perform in line with a new hypothesis: The Transfer Hypothesis. This new hypothesis holds that second language learners who have an article system in their first language set to either definiteness or specificity will transfer these settings to their second language. Spanish L1 speakers, who have an ACP set to definiteness, are thus predicted to perform target-like on English article choice as English and Spanish both have an article system based on definiteness. Ionin et al. (2008) show that both the Fluctuation Hypothesis and the Transfer Hypothesis seem to be supported by the results of the Russian and Spanish learners; Russian learners fluctuate again, while Spanish learners perform near target-like on English article choice and show no fluctuation, as illustrated in the table below:
Table 4: Performance on L2 English Article Choice by L1 Spanish learners
[+definite] [-definite]
[+specific] Correct use of the 87.5%
Correct use of a 92.5%
[-specific] Correct use of the 96.7%
Correct use of a 91.7%
Ionin et al. (2008) report Spanish learners not to differ significantly from the native control group, even if accuracy rates in the definite specific conditions are considerably lower than in the other three conditions. They blame discrepancy on an item effect caused by transfer from Spanish construction. Since subjects had the freedom to supplement the blanks with any item they thought fitting, other errors were mainly to participants choosing items other than articles, such as possessive pronouns. Similar to Ionin et al. (2004), Russian and Spanish learners were also tested on proficiency with a cloze test to see whether article choice would
improve as proficiency improved. Statistical test results show Spanish participants to improve even further on English article choice as they are attuned to the discourse contexts. This proficiency effect is also found in Zdorenko & Paradis (2008), who test young (mean age 5;4) [+article] L1 learners on English article choice. In their longitudinal study of six months they find that errors in L2 article choice disappear after approximately two months. Even though child learners are tested in this study, their results do suggest an improvement in L2 article choice as influenced by further exposure of the second language.
Results supporting the Transfer Hypothesis can also be found in studies by Snape et al. (2006), Ting (2005) and Reid et al. (2006), in which Spanish learners of English perform target-like. Hawkins et al. (2006) finds support for the Transfer Hypothesis in the L2 English article choice of L1 Greek learners.
According to these previously mentioned studies and the two hypotheses these produce, L2 article choice can move in two directions. Depending on the L1 of the learner, article-less L1 learners of an L2 with articles fluctuate between specificity and definiteness until L2 input prompts them to set their ACP correctly. [+article] L1 learners of a [+article] L2, both for which articles are grouped by definiteness, will transfer their ACP setting accurately from their L1 to their L2.
2.3.3 Deprez, Sleeman & Guella (2011)
Deprez et al. (2011) conducted an experiment similar to that of Ionin et al. (2004; 2008). Their experiment was modelled after the article test done by Ionin et al. (2004), testing article acquisition of L2 French. Similar to English, French has an article system based on definiteness. Instead of testing adults, the researchers tested children and adolescents learning French as an L2 in school. Two L1 groups with definiteness-based articles took part in the experiment: 50 L1 speakers of Arabic, which has the definite article él and indefinite suffix -n and 23 L1 speakers of Dutch, which has two definite articles, de and het and an indefinite
article een. Participants in the Dutch group were aged 13-15 years and had had 200 hours of French, whereas participants in the Arabic group were divided according to age (10 and 12 years) and amount of L2 input (64 and 280 hours of French). Deprez et al. (2011) consider the age and L2 input division made in the Arabic participants as a difference in proficiency. The overall elicitation test results show a significant overuse of the definite article in [-definite, +specific] contexts and of the indefinite article in [+definite, -specific] contexts for both Dutch and Arabic participants, as shown in tables 5, 6 and 7.
Table 5: Results L1 Dutch Learners of L2 English Article Choice (Deprez et al. 2011)
[+definite] [-definite]
[+specific] Correct use of le/la 87%
Overuse of le/la 68%
[-specific] Overuse of un/une 45%
Correct use of un/une 77%
Table 6: Results L1 10-year-old Arabic learners of L2 English Article Choice (Deprez et al. 2011)
[+definite] [-definite]
[+specific] Correct use of le/la 78%
Overuse of le/la 36%
[-specific] Overuse of un/une 76%
Correct use of un/une 73%
Table 7: Results L1 12-year-old Arabic Learners of L2 English Article Choice (Deprez et al. 2011)
[+definite] [-definite]
[+specific] Correct use of le/la 80%
Overuse of le/la 25%
[-specific] Overuse of un/une 43%
Correct use of un/une 90%
Additionally, the older L1 Arabic group show less overuse of the definite and indefinite article, which suggests that as L2 input increases, proficiency in L2 article choice increases, in line with results in Ionin et al. (2008).
specificity settings of the Article Choice Parameter even if the first language has an overt article system based on definiteness (such as Dutch and Arabic). Thus, these results are not in line with the Transfer Hypothesis, which predicts that the Article Choice Parameter setting of the L1 transfers to the L2. These results could suggest support for the Fluctuation Hypothesis, but do not meet the condition of errors caused by having a [-article] L1. Deprez et al. (2011) offer an explanation for this pattern by arguing that errors in L2 article choice, in a situation in which the L1 and L2 ACP settings are identical, are caused by so-called ‘computational egocentricity’.
In the field of linguistics, egocentricity has mostly been discussed in terms of
maturational egocentricity, necessary for the L1 acquisition of definiteness-based articles. As mentioned in 2.2, Maratsos (1976, 1979) explain overuse of the definite article in indefinite conditions by L1 English children with the term ‘egocentricity’: the attribution of speaker beliefs to the hearer, resulting in overuse of the definite article. Since this behaviour is only attested by young children it can be categorised as maturational. Schaeffer & Matthewson (2005) too investigate article choice in L1 English children acquiring articles, finding overuse of the definite article in indefinite specific conditions. They argue that young children (under 5 years old) lack the pragmatic Concept of Non-Shared Assumptions, causing children as speakers to be unable to take perspectives of the hearer into account, similar to Maratsos’s (1976, 1979) account of egocentricity. This absence of pragmatics causes problems for article acquisition; child speakers are unable to consider the width of the hearer’s knowledge,
Figure 8: English article system (Schaeffer & Matthewson, 2005)
The children tested in this study were 2-4 years old. After this age errors tend to disappear as children grow more pragmatically mature. A maturational approach to egocentricity cannot account for errors in article use made by pragmatically mature participants. Still, Deprez et al. (2011) argue that the errors in L2 article use made by their L1 Dutch and Arabic participants are egocentric. As mentioned earlier, Ionin et al. (2004) defines the term definiteness as an idea shared by speaker and hearer, while specificity is defined as referring to an entity known by the speaker only. In terms of pragmatics, these definitions entail that definiteness concerns a frame shared by speaker and hearer, whereas specificity relates to an egocentric frame, namely that of the speaker. This account of definiteness and specificity would be able to explain the errors made by the [+article] L1 participants in Deprez et al. (2011)’s study. Participants overgenerated the definite article in [-definite, +specific] contexts due to their association of the definite article with specificity, caused by an inability to take the common ground into consideration. Subjects overused the indefinite article in [+definite, -specific] contexts, due to an association of the indefinite article with a lack of specificity.
Deprez et al. (2011) propose a hypothesis of computational egocentricity to explain the errors Dutch and Arabic adolescents make in L2 article choice. Due to the increased cognitive pressure of learning a second language, subjects fall back to an egocentric view of the world and are led by specificity as representing their own mind, rather than definiteness as representing the common ground of both hearer and speaker. Applying this to Deprez et al.’s (2011) L2 learners of French, Dutch and Arabic adolescents sometimes use the definite article
le in specific but not definite contexts, and the indefinite article un in definite but not specific
computational rather than maturational in the following studies on egocentricity and adults. Horton et al. (1996) discuss the concept of egocentricity as defined by Deprez et al, (2011), namely, when adult L2 learners do not take the common ground into account in referential descriptions. In their study, L1 English participants were tested in speaker-hearer pairs and in two conditions: one context in which both the speaker and hearer had the same visual knowledge and a second one in which the speaker had privileged visual access over the hearer. In normal circumstances, speakers took the hearer’s lack of visual access into account in the second context. However, when computational load increased due to time pressure, speakers did not recognise the hearer’s different perspective and used definite articles to introduce objects. Due to failure of considering the hearer’s perspective caused by an increased computational load, this behaviour supports the claim of computational
egocentricity. This pragmatic failure as a result of a heightened computational load in the form of time pressure is supported by results from Epley et al. (2004a, b). Keysar et al. (2000) test L1 English adult participants in a task similar to Horton et al.’s (1996), in the way that speakers have visual access to more objects than hearer’s, of which the speakers were made aware of. Eye tracker results indicated that even when speakers were specifically asked to choose the object visible to the interlocutor, their eyes first strayed to the object only visible to themselves. Keysar et al. (2000) claim that initially not taking the interlocutor’s perspective into account requires less mental effort, similar to the computational egocentricity account Horton et al. (1996) give. These studies provide support for the computational approach to egocentricity as offered by Deprez et al. (2011) and are able to explain why [+article]
pragmatically mature L2 learners fluctuate in L2 article choice. This approach will henceforth be called the Computational Egocentricity Hypothesis.
2.4 The Present Study
applied to article L1 learners of an article L2, resulting in mixed outcomes with regard to L2 article choice. The Transfer Hypothesis as proposed by Ionin et al. (2008), argues that these learners will transfer the specificity settings through the ACP to their L2, resulting in target-like article choice. The Computational Egocentricity Hypothesis, as proposed by Deprez et al. (2011), argues that these learners will overuse the definite article in [+specific] settings and the indefinite article in [-specific] settings, as a result of applying egocentric reasoning due to the additional cognitive load of learning a second language.
Ionin et al. (2004, 2008), among others, has mainly studied [-article] L1 learners of languages containing articles, such as Russian and Korean learners of English, as these learners do not perform target-like on article choice. Spanish learners of English, as shown by Ionin et al. (2008), perform target-like on English article choice and do not require further investigation. However, with the information presented by Deprez et al. (2011), [+article] L1 learners of languages with articles have been shown to make errors in L2 article choice, contrary to results found and predicted by the Transfer Hypothesis. Due to these opposing theories and outcomes, a need calls for further investigation of mother-tongue speakers of languages with articles acquiring languages with the same article settings as their L1. This study intends to look at Dutch L1 adolescents, learning English as an L2 in a formal school setting.
2.5 The Dutch Article System
Dutch, like English, has an article system based on definiteness and the same articles for definite singular nouns. Contrary to English, Dutch makes the distinction between common and neuter nouns, including a separate definite article for both the common and the neuter nouns. The indefinite article in Dutch makes no such distinction. This results in the [+definite, +neuter] article het, the [+definite, +common] article de, and finally the [-definite]
article een (Blom et al. 2008). The Dutch articles and their labels are shown in the table below:
Table 10: The Dutch Article System (Deprez et al. 2011)
Similar to English, Dutch allows for the four combinations of definiteness and specificity. Examples of these in Dutch, adapted from the English examples presented in section 2.2, are presented below.
(1) [+definite, +specific]
Ik zag gisteren een meisje. Het meisje droeg een rode jurk.
I see-PST yesterday a girl. The girl wear-PST a red dress. “I saw a girl yesterday. The girl was wearing a red dress.”
(2) [+definite, -specific]
Ik ben benieuwd wie de winnaar van het toernooi zal worden!
I be-PRS curious who the winner of the tournament will be-INF “I am curious who will be the winner of the tournament!”
(3) [-definite, +specific]
Ik heb een oom die je daarbij kan helpen.
I have-PRS an uncle who you with-that can help-INF “I have an uncle who could help you with that.”
(4) [-definite, -specific]
Ik zou graag een puppy voor mijn verjaardag willen.
I would gladly a puppy for my birthday want-INF. “I would like to have a puppy for my birthday.”
As these examples have already been discussed in light of their semantics in section 2.2, no further discussion of article semantics in these examples is necessary. In view of the Transfer Hypothesis, the similar article system in the L1 should be transferred to the L2, in this case, English. The lack of grammatical gender in English articles should not cause problems and the existence of only one definite and one indefinite article could only make the choice less complex. However, as learning a second language is believed to be computationally taxing according to the Computational Egocentricity Hypothesis, participants could fall back to an egocentric world view and be led by specificity instead of definiteness, resulting in non-target-like behaviour when the two are in conflict.
3. Hypotheses and Predictions
Previous studies in L2 article choice have generated two existing hypotheses and predictions. Based on target-like article choice by L1 Spanish learners of English, Ionin et al. (2008) hypothesize that these learners transfer their ACP settings from their native language to their second language, since Spanish has the same article semantics regarding definiteness and specificity as English. As a result of non-target-like article choice by L1 Dutch learners of English, Deprez et al. (2011) argue that this behaviour reflects egocentricity caused by the computational pressure of learning a new language. In addition to the two main hypotheses as presented above, previous studies show that L2 article choice improves as proficiency
improves (Ionin et al. (2008), Zdorenko & Paradis (2008) and Deprez et al. (2011). These three hypotheses, together with the predictions for this study, are formulated below:
Hypothesis 1 (Ionin et al. 2008). The Transfer Hypothesis: [+article] L1 learners of a [+article] L2 with an ACP set to definiteness will transfer this setting from their L1 to their L2.
Prediction 1: Dutch L1 learners of L2 English transfer their ACP setting from Dutch to English, resulting in target-like article choice in all four conditions. Specifically, English article choice by Dutch learners of English does not differ significantly from English article choice by L1 speakers.
Table 11: Predictions L2 Article Choice per Condition According to Hypothesis 1
[+definite] [-definite]
[+specific] Correct use of the Correct use of a [-specific] Correct use of the Correct use of a
Hypothesis 2 (Deprez et al. 2011). The Computational Egocentricity Hypothesis: Regardless of the L1, due to the additional computational load of learning a new language, learners will take an egocentric perspective in using articles, where they will be led by specificity instead of definiteness.
Prediction 2: Dutch L1 learners of L2 English will significantly overuse the definite article
the in [-definite, +specific] conditions and significantly overuse the indefinite article a in
[+definite, -specific] conditions. When definiteness and specificity are not in conflict, errors are not expected.
Table 12: Predictions L2 Article Choice per Condition According to Hypothesis 2
[+definite] [-definite]
[+specific] Correct use of the Overuse of the
[-specific] Overuse of a Correct use of a
Hypothesis 3 (Ionin et al. (2008), Zdorenko & Paradis (2008) and Deprez et al. (2011)): Proficiency influences correct article choice.
Prediction 3: L2 article choice will become more native-like as proficiency improves. Specifically, this entails that the high proficiency group will perform better on English article choice than the medium and low proficiency group. In turn, the medium proficiency group will perform better on English article choice than the low proficiency group.
4. Method
4.1 Participants
The participants in this study tested on their article choice were 104 Dutch L1 speakers, varying between the ages of 12-16 (mean age: 14.2). Participants were recruited from a secondary school in the northwest of the Netherlands, where they were learning English in a formal classroom setting. All students were enrolled in a VMBO track, respectively pre-vocational training in the first, second and third grade of a four-year
secondary school programme (grade 7, 8 and 9). Six classes participated in the experiments, of which specifics can be found in the table below:
Table 13: Dutch L1 Participants Class Date of Testing Number of Students Number of Males Number of Females Mean Age (years; months) 2C 08-04-2016 14 5 9 13;11 1F 08-04-2016 21 8 13 12;7 2A 08-04-2016 16 8 8 13;11 3B3 09-05-2016 16 10 6 14;11 3B4 09-05-2016 13 4 9 14;8 3K1 09-05-2016 24 13 11 15;0 Total 104 48 56 14;01
None of the students tested were reported to have language disorders such as SLI and all of the students were typically developing. Students differed in their age of onset with regard to English, as not all primary schools start with teaching English in the same grades. These differences among subjects were recorded with a questionnaire. From the first grade of secondary school onwards, all students were exposed to the same amount of English in school. Participants were also checked for their exposure to English outside of the classroom and their socio-economic status (SES), which was measured by the parents’/caretakers’ level of education. For more information, see section 4.2.1 on the questionnaire.
Besides the target group, this study also included a control group of 40 native English speaking adolescents aged 14-17. Control group participants were recruited from a secondary school in Kansas, US. Students were reported to come from lower to middle class
Table 14: Native English Control Group Grade Date of Testing Number of Students Number of Males Number of Females Mean Age (years; months) Standard Deviation 9 08-04-2016 20 6 7 15;4 0.8 10 08-04-2016 20 8 6
Even though Ionin et al. (2008) tested her elicitation task on six native English speaking university students from California, these university students might have a different command of English than adolescent native speakers of English in middle and secondary schools. The control group participated in an online elicitation task, where data was collected via the survey programme SurveyMonkey.
4.2 Experiment
Participants took part in a test composed of three parts: an elicitation task, an Anglia Placement proficiency test and a questionnaire regarding language background and SES. The control group only took part in the elicitation task.
4.2.1 Questionnaire
As mentioned in the participant section, not all Dutch students in secondary school have the same background regarding English language exposure. In addition to formal exposure, exposure to English outside the classroom can vary from student to student. To control for these factors, students were asked to fill out a questionnaire. This questionnaire required subjects to list some information regarding language exposure, such as their first language, other languages they spoke and their age of onset and use of English.
Previous research has shown that a low SES has a negative influence on (second) language acquisition, especially in children (Ginsborg 2014, among others). One way of measuring the SES of students is their parents’ educational level (Hauser 1994, among
others). To ensure that the students tested would not be disadvantaged and thus all had a similar indication of SES, participants were asked to list their parents’ level of education and occupation. Other personal information asked was the subject’s date of birth and gender. For the full questionnaire, please see the appendix.
4.2.2 Anglia Placement Test
Zdorenko & Paradis (2008), Ionin et al. (2008) and Deprez et al. (2011) reported in their results that as age and/or language proficiency increased, participants’ number of target-like article choice responses increased as well. These three studies show that language proficiency appears to correlate with article choice but do not measure second language proficiency (in the same way). Zdorenko & Paradis (2008) do measure second language proficiency for the children they test but assume in their longitudinal study that their subjects become more proficient in their L2 every month. Ionin et al. (2008) measure second language proficiency with a Rutherford cloze test. Deprez et al. (2011) do not measure the proficiency of their Arabic learners of French, but note the hours of French classes and age as a means of measuring proficiency.
The current study chose not to assume English proficiency from the different grades students were in, as the age of onset and exposure to English varied outside of the classroom. Instead, students were tested on their English proficiency with a standardized English
placement test frequently used to test secondary school students, namely an Anglia Placement Test offered by the Anglia Examinations England Network. The test is composed of multiple choice questions where partakers are asked to give the correct words, verb conjugations, synonyms, among others. Test results place the student in one of ten levels, each of which are linked to the Common European Framework of Referencing (CEFR). The CEFR is a
different levels of language proficiency. These levels are compared to indices of English proficiency by the Anglia Network in the figure below:
Figure 15: Level of English Proficiency According to the Anglia Placement Test (http://www. anglianetwork.eu)
The Anglia Placement Test is composed of a 100 questions. Students must receive a certain number of points on each page of questions for the corrector to move on with marking the test, since the test increases in difficulty. For instance, when a student receives 10/22 points on page 3, where a minimum of 16/22 points is required for the student to move on, points received on following pages do not count. The Anglia Placement Tests have until now been mainly used for teachers and schools to grade their students, but researchers in the field of second language acquisition in educational settings have used the test as well to assess language proficiency (Kopeckova, 2011).
4.2.3 Article Elicitation Task
Ionin et al. (2008) created an article elicitation task to test Spanish and Russian L1 learners of English, based on the test designed by Ionin et al. (2004). The 2008 article test was composed of multiple short passages containing a blank, where participants were told to fill in
the missing element, instead of choosing either a, the, or ---. These elements targeted definite and indefinite articles, along with different fillers. Deprez et al. (2011) designed an article elicitation task much like the one Ionin et al. (2008) composed, only without fillers. A pilot study of the 2008 article test proved too difficult for participants with regard to free choice of completing the blanks, indicated by the random answers given. Thus the present study
employed a multiple choice variant, where students could choose between the two articles. In the case of the indefinite article a/an, only nouns starting with a consonant were included, so students would only need to use a and not have a third article choice variation (an). Unlike Deprez et al. (2011), this test included two fillers Ionin et al’s article test offered, as to draw away the attention of what was being tested. Of the four different categories with regard to specificity and definiteness, each category contained five items. Only singular nouns were tested. Besides the twenty target items, the test included ten filler items, targeting either a blank or the negative not. Examples of each condition and the two fillers are shown below:
Example 1: [+definite, +specific]
Carrie: Did your funny uncle Reuben visit you for Thanksgiving?
Older sister: No, he and his wife went to visit her family instead. They went to ________ capital of North Dakota – I can’t remember what its name is. It’s probably a
very cold place!
Example 2: [+definite, -specific]
Mother: What are you reading in the newspaper?
Daughter: I’m reading a poem about baby lions – I really like it. I would like to write a letter to ________ writer of that poem – unfortunately, I have no idea who it is… The poem isn’t signed!
Example 3: [-definite, +specific]
In an airport, in a crowd of people
Man: Excuse me, do you work here? Security guard: Yes. Can I help you?
Man: Yes, please. I am trying to find ________ red-haired girl; I think that
Example 4: [-definite, -specific]
In a school
Child: It’s my birthday next week!
Teacher: That’s great. Are you going to have a party?
Child: Yes! A big party! I am hoping to get ________ new dog! I love animals! Example 5: Filler (---)
At the bus stop
Mike: Hello, this is my first time seeing you here. When did you start taking the city bus? Chris: I started taking the bus when I started school ________last week.
Example 6: Filler (not)
At the bus station
Mildred: Where is the bus? It was supposed to come five minutes ago! Station Attendant: I’m sorry. The schedule has changed. The bus will ________ come today.
4.2.4 Procedure
The testing of the participants took place at a VMBO secondary school in the northwest of the Netherlands. Due to the large amount of students, two testing days were needed, respectively April 8th, 2016 and May 9th, 2016. Parents were informed of the
experiment through a passive consent e-mail, to which none replied negatively. The entire test was completed in fifty minutes during their usual English classes. The usual English teacher was present during the testing to keep order and to introduce the researcher. Instructions were put on a screen and repeated by the researcher prior to testing, to ensure all students
understood the test procedure. Students were allowed to ask questions during the test and they made use of this privilege. After instructions, participants received a booklet in which they started with the article test. Subjects were required to fill in one of the four options, which were projected onto the screen during the session. After finishing the article test, students moved onto the Anglia test, where they were required to circle the correct option. When they
finished the Anglia test, students completed the test by filling out the questionnaire. During the session, most students were quiet. One class displayed disruptive behaviour, but only the students who had already finished the test. The other students did not seem very distracted by their behaviour.
4.3 Coding and Statistics
After finishing the experiments, data was fed into an Excel spreadsheet and coded. Correct article choice was coded with a “1”, substitution errors (eg. a instead of the) were coded with a “0” and filler insertion (eg. not instead of an article) was coded with a “2”. Anglia tests were scored numerically and assigned a proficiency level (see 4.2.2)
The Socio-economic status of the parents was coded according to the following schema:
Table 16: Coding of Educational Background for SES
Level of Education parent 1 Level of Education parent 2 Code MBO (vocational training) MBO (vocational training) 0 MBO (vocational training) HBO (applied university) 1 HBO (applied university) HBO (applied university) 2
HBO (applied university) WO (university) 3
WO (university) WO (university) 4
Fillers in the article elicitation task served to ensure students understood the task and did not answer at random. After coding was completed, statistics were applied to results of the article tests by using SPSS. The Transfer Hypothesis required L2 article choice performance to match the native controls and therefore an independent samples t-test was used to compare mean scores of L2 subjects with the L1 controls on each condition. As the Computational Egocentricity Hypothesis predicts that L2 learners produce errors in [+definite, -specific] and [-definite, +specific] conditions, a one-way ANOVA and Tukey post-hoc test were used to see if L2 learners produce significantly more errors in these conditions than in the [+definite,
+specific] and [-definite, -specific], in which they are predicted to perform target-like. A significant p-value for both the independent samples t-test and the ANOVA is determined at 0.05 or lower.
5. Results & Discussion
In the following section I will discuss the results of the elicitation task per hypothesis. However, before looking at article choice, I will consider the results of the questionnaire and the background of the participants.
5.1 Questionnaire
This section reports the results of the questionnaire. As mentioned in the method section, students differed in their age of onset with regard to the acquisition of English. Participants had a mean age of onset of 9 years and 8 months, with a standard deviation of 1.9. Most subjects started acquiring English as a second language between the ages of 9 and 11, but reported that the hours of English classes they received before the first form of secondary school was minimal. One student was excluded because her first language was Polish, but all the others reported Dutch as their first language. Students indicated that they came in contact with English through films, series, internet and games, but usually not through reading books or contact with English family members.
The socioeconomic status of students was also examined by way of
parents’/caretakers’ level of education. Coding these results produced a mean SES of 0.5 with a standard deviation of 0.9, which are both quite low. The low SES shows that at least one parent, if not both, had a vocational educational level, sometimes paired with the other parent having completed applied university. The low standard deviation indicates a large
that of the native English speakers, as they were reported to mostly have lower middle class to middle class backgrounds.
5.2 Elicitation Task: Control Group
The following table shows the scores of the native speakers on the elicitation task with regard to correct article choice, incorrect article choice (substitution errors) and other errors (filler insertion).
Table 17: First Results Article Elicitation Task Native Speakers (N=40)
[+definite] [-definite] [+specific] 98.5% the 0% a 1.5% other 95.5% a 4% the 0.5% other [-specific] 94.5% the 3.5% a 1% other 92.5% a 7% the 0.5% other
On the surface, native speakers appear not to reach the 100% correct article choice as
expected. Nevertheless, percentages of correct article use are all above 90% and when taking into account a standard error margin of 10%, the control group performs at ceiling level. Still, some differences between the different conditions surface, such as the 6% difference in accuracy rates between the [+definite, +specific] condition and the [-definite, -specific] condition. An accuracy rate of 98.8% on filler performance indicates that participants did understand the task; thus this behaviour cannot explain the irregularities presented in table 17. No subjects showed peculiar behaviour compared to other participants with regard to article choice; thus, a subject effect can be excluded. However, a closer look at the items in the different conditions reveals an item effect, which will be discussed in more detail in the following paragraph.
5.2.1 Item Effect
The table below presents performance on items identified as deviating sufficiently from the cut-off norm, set at 90% correct:
Table 18: Item Effect
Condition [+definite, +specific] [-definite, -specific] [+definite, -specific] [-definite, +specific]
Item 12 13 21 23 27 3 11 14 20 22 2 7 8 18 25 5 9 16 24 30 Correct (%) 98 98 98 100 100 10 0 83 88 95 98 98 98 83 98 98 85 98 98 98 10 0 Incorrect (%) 0 0 0 0 0 0 17 12 3 2 2 2 12 2 2 15 2 2 0 0 Other (%) 2 2 2 0 0 0 0 0 2 0 0 0 5 0 0 0 0 0 2 0
Table 18 shows that items 11, 14, 8 and 5 do not meet this norm. Lower accuracy rates in the definite nonspecific condition reveals that participants overused the indefinite article (13%) in dialogue 8:
8. Marcus: Can you and your friend Rick come over this week-end?
Jim: I’ll come over, but Rick isn’t here. He went to ________ house of his uncle George… I have no idea where that is. But Rick was very excited about going!
The association of the house of his uncle George with indefiniteness could have multiple causes. First of all, participants might not have processed the house as being a unique entity, i.e. belonging to his uncle George. Thus the house could have been interpreted as any house and the addition of the speaker having no idea where it is may have added to the house being considered unknown and requiring an indefinite article for a first mention. Secondly, though less probable, a reader might have understood that uncle George owned multiple houses and therefore chosen an indefinite article as meaning one of many. However, native speakers would usually express owning multiple houses and visiting one more explicitly, in the form of going to one of his uncle’s houses. For item 5, native speakers incorrectly used the definite article in 15% of the cases, contra the minimal overuse of the in other items in the indefinite nonspecific conditions. The dialogue containing item 5 is presented below:
5. In an airport, in a crowd of people Man: Excuse me, do you work here? Security guard: Yes. Can I help you?
Man: Yes, please. I am trying to find ________ red-haired girl; I think that she flew in on Flight 239.
In this dialogue, the red-haired girl is known to the speaker and therefore specific. However, since the girl is unknown to the security guard, an indefinite article is required. The use of the definite article here could be the result of an association of the definite article with specificity, behaviour previously observed for pragmatically immature L1 English children (Schaeffer & Matthewson, 2005).
Overuse of the definite article in indefinite nonspecific condition also occurs, which is mainly caused by two items: 11 and 14, as presented below:
11. Mother comes home
Mother: How did Peter spend the day at his grandmother’s?
Father: He had a good time. He did his homework for tomorrow. Then he went outside and played with ________ little girl – I don’t know who it was. Then he came back inside; and then I came and took him home.
14. Jessie: I baby-sat yesterday for the first time ever. Lesley: How was it?
Jessie: Fine. I baby-sat a little boy named Niles. I played a monopoly game with him. Then I did my homework, and Niles read ________ short story – I don’t know what it was about. And then I put him to bed.
The indefinite article a is needed for these conditions, since neither the hearer nor the speaker has knowledge of the item discussed. This lack of knowledge is expressed by I don’t know
who it was in 11 and I don’t know what is was about in 14. A possible explanation for errors
in these items is that participants did not read the sentences after the blank, which contain important information for specificity, and interpreted the little girl and short story as being
[-definite, +specific]. The same explanation for item 5 can then be used, namely that
participants associated the definite article with specificity. Native English speakers in other papers on L2 article choice do not show this behaviour (Snape et al. 2006; Ionin et al. 2004, 2008; Hawkins et al. 2006). Since these studies tested adult university speakers of English, subjects in the present study could be more related to L1 English child behavior related to article choice. As Schaeffer & Matthewson (2005) show, children are prone to pragmatic egocentricity errors and could thus associate specificity with definiteness. When subjects interpret the indefinite nonspecific condition as indefinite specific, they might choose the definite article instead of the indefinite article in this condition. Nevertheless, these errors are explained by an account of pragmatic immaturity and participants in this study have mature pragmatics.
Another possible explanation for the low scores in the indefinite nonspecific condition can be found in Lardiere (2004), who finds that L2 learners have more trouble with indefinite articles than definite articles. The explanation for this pattern is that indefinite article
semantics are more complex and thus more difficult to acquire than definite articles. The same preference for definites has been found in L1 English article choice (Maratsos, 1976).
Considering that participants in this study have not yet reached adulthood and are still acquiring their L1, traces of their childhood struggles with articles might remain. Since this L1 age group particular age group has not been tested as control group for L1 article choice, these explanations will remain speculations until further research into this matter is
conducted. For now, the items discussed above will be removed from the data, resulting in ceiling accuracy for article choice performance in all conditions:
Table 19: Results Native Speakers Article Elicitation Task (N=40) [+definite] [-definite] [+specific] 98.5% the 0% a 1.5% other 98.1% a 1.3% the 0.6% other [-specific] 97.5% the 2.5% a 0% other 97.5% a 1.7% the 0.8% other
Even though accuracy rates are not a 100% and slight overuse is still present, these data can be interpreted as ceiling accuracy and are comparable to native speaker data and reports (Ionin et al. 2004, Snape et al. 2006, Hawkins et al. 2006, Ionin et al. 2008). Furthermore, these data will form the baseline of target-like article choice performance to which the L2 learners will be compared in the following paragraph.
5.3 Elicitation Task: Second Language Learners
The following section will present the results of the elicitation task as carried out by the Dutch L1 learners of English as their second language. These findings will be discussed per proficiency group and in relation to the Transfer and Computational Egocentricity Hypothesis. Distinguishing participants on the basis of proficiency is necessary, since
previous studies have shown article choice to be influenced by proficiency in the L2 (Ionin et al. 2004, Ionin et al. 2008, Zdorenko & Paradis 2008, Deprez et al. 2011). Before discussing the findings per proficiency group, the distribution of participants over the proficiency groups will be justified.
5.3.1 Proficiency Groups
All participants completed the Anglia Placement Test to assess their level of proficiency in English. Overall, students received a mean score of 29 out of a 100, with a standard deviation of 14. Due to this high standard deviation, which suggests large differences in proficiency, students were divided into three different proficiency levels:
Table 20: Participants per Level of Proficiency Level of Proficiency
(Anglia)
Number of Students Level of Proficiency (current paper) Number of students Primary (13-20) 38 Low 38 Preliminary (21-36) 49 Medium 49 Elementary (37-46) 1 High 13 Pre-Intermediate (47-61) 5 Intermediate (62-71) 7
Instead of the initial 104 students tested, 4 students had to be excluded from the data: one student reported her native language to be Polish, two students appeared distracted during the experimental procedure, and one student did not use any fillers, which was interpreted as a lack of task comprehension. In accordance with the different proficiencies as proposed by the Anglia Placement Test, the current paper created three proficiency groups: low, medium and high proficiency. Since the three highest Anglia proficiency levels students reached contained so few students, these three levels were merged into the current high proficiency group. Even though differences in proficiency scores appear high (37-71), these differences had little effect on article choice in this group, because students did not differ very much in their scores on the article elicitation task.
5.3.2 Low Proficiency
The following table shows performance on the article elicitation task by the 38 participants in the low proficiency group:
Table 21: First Results Elicitation Task Low Proficiency (N=38) [+definite] [-definite] [+specific] 53.7% the 31.7% a 9.7% other 65.8% a 22.1% the 12.1% other [-specific] 55.3% the 29.5% a 15.3% other 71.1% a 18.4% the 10.5% other
Ostensibly, accuracy rates on article performance appear lower than those presented in table 21. This finding can first of all be accounted for with a lack of task comprehension, since the
