You are on page 1of 304

The Acquisition of English Articles by Arabic Speakers

Thesis submitted in partial fulfilment of the requirements for the degree of Doctor of

Philosophy

Dina Awad (BA, MA, Dip)

Supervisor: Professor A. Siewierska

English Language and Linguistics

Lancaster University

April 2011
Abstract

Mastering the English Article system is a long term challenge for L2 learners. The

difficulty originates from the fact that appropriate usage requires stacking multiple

functions into limited forms, the numerous exceptions to the rules and the mismatch

between the grammatical criteria of countability and number that determine the

appropriate supply of the indefinite article and the lexical-pragmatic values on which

definiteness depends. Differences between the first and target languages can also

cause problems even for advanced learners. We investigated the use of English

articles in the production of Arab university students by collecting data from three

different tests that varied in the degree of control and the type of knowledge they

examine. Development was followed cross sectionally after dividing the participants

into three proficiency level groups according to their scores on the Oxford Placement

Test. Statistical analyses were performed to calculate the differences across groups,

tasks and compare between learners’ use of the two articles. The results were also

compared to findings from other L2 studies to determine whether the development

map corresponds to/differs from the tendencies of learners from other L1

backgrounds.

It was found that Arabic influenced the participants’ decisions to a large

extent, especially at lower levels. In other respects error patterns paralleled those of

other L2 learners. The definite article was mastered before the indefinite while the

correct marking of non-referential bare nominals (zero article) seemed to be the most

difficult aspect of article use to master. The results suggest that task type influenced

learner’s choices considerably. Finally, faulty associations between definiteness and

linguistic notions of specificity, pre-modification and concreteness in learner

hypotheses caused variability in L2 article production.


Table of Contents

Chapter One: Introduction ......................................................................................... 1


Chapter Two: Literature Review ............................................................................... 9
2.1 The Article System in English ................................................................................. 9
2.1.1 Definiteness ................................................................................................. 11
2.1.2 Countability and Number ............................................................................ 26
2.2 Article Contexts ..................................................................................................... 31
2.2.1 Definite article environments: ..................................................................... 31
2.2.2 Indefinite article environments ................................................................... 34
2.3 The article system in Arabic .................................................................................. 35
2.3.1 Definiteness ................................................................................................. 36
2.3.2 Indefiniteness .............................................................................................. 38
2.3.3 Comparison of article use in Arabic and English ....................................... 40
2. 4 Second Language Acquisition (SLA) ................................................................... 42
2.4.1 Acquisition of Articles by L2 learners of English ...................................... 48
2.4.2 Acquisition of articles by Arab learners ..................................................... 51
2.5 Learner Hypotheses ............................................................................................... 52
2.5.1 Specificity ................................................................................................... 53
2.5.2 Premodification ........................................................................................... 56
2.5.3 Concreteness ............................................................................................... 58
Chapter Three: Methodology ................................................................................... 61
3.1 Required data ......................................................................................................... 61
3.2 Participants ............................................................................................................. 64
3.3 Task Design ........................................................................................................... 65
3.3.1 The Writing Task (T1) ................................................................................ 66
3.3.2 Stories Task (T2) ......................................................................................... 68
3.3.3 Grammaticality judgment test (T3) ............................................................. 71
3.4 Administrating Tests .............................................................................................. 77
3.4.1 Pilot ............................................................................................................. 77
3.4.2 OPT ............................................................................................................. 77
3.4.3 Tasks ........................................................................................................... 79
3.5 Data Analysis ......................................................................................................... 80
3.5.1 Accuracy ..................................................................................................... 83
3.5.2 Error types ................................................................................................... 84
3.5.3 Article Choice ............................................................................................. 87
3.5.4 L1 and IL..................................................................................................... 91
3.5.5 Learner Hypotheses .................................................................................... 94
3.6 Limitations ........................................................................................................... 108
Chapter Four: Results ............................................................................................. 110
4.1. Task 1 .................................................................................................................. 113
4.1.1 Accuracy ................................................................................................... 115
4.1.2 Errors ......................................................................................................... 117
4.1.3 Learner Hypotheses .................................................................................. 122
4.2 Task 2 ................................................................................................................... 128
4.2.1 Accuracy ................................................................................................... 129
4.2.2 Grammatical Errors ................................................................................... 130
4.2.3 Choice Errors ............................................................................................ 133
4.2.4 Replacement Errors ................................................................................... 136
4.2.5 Learner Hypotheses .................................................................................. 138
4.3 Task 3 ................................................................................................................... 143
4.3.1 Accuracy ................................................................................................... 144
4.3.2 Errors ......................................................................................................... 147
4.3.3 Learner Hypotheses .................................................................................. 152
4.4 Cross-Task results ................................................................................................ 158
4.4.1 Accuracy ................................................................................................... 158
4.4.2 Errors ......................................................................................................... 162
4.4.3 Learners Hypotheses ................................................................................. 171
Chapter Five: Discussion ......................................................................................... 177
5.1 Stages of acquisition ............................................................................................ 178
5.1.1 Accuracy ................................................................................................... 178
5.1.2 Errors ......................................................................................................... 187
5.2 Task Type Variation ............................................................................................ 202
5.2.1 Review of performance by task ................................................................ 203
5.2.2 Task effect by PL group ............................................................................ 217
5.3 Learner hypotheses .............................................................................................. 242
5.3.1 Specificity ................................................................................................. 242
5.3.2 Premodification ......................................................................................... 256
5.3.3 Concreteness ............................................................................................. 267
Chapter Six: Conclusion.......................................................................................... 273
References ................................................................................................................. 277
Appendices ................................................................................................................ 291
Appendix 1: Information Sheet and Consent Form ........................................... 291
Appendix 2: Background uniformity survey ..................................................... 293
Appendix 3: Task 1 ............................................................................................ 294
Appendix 4: Task 2 ............................................................................................ 295
Appendix 5: Task 3 ............................................................................................ 300
Chapter One: Introduction

The, a, and an are some of the most frequent words in the English language according

to COBUILD’s list of the ten most frequent words in English (Sinclair 1991). The

ranks first in this corpus of 20 million words, while a(n) comes fifth. Articles are also

among the most frequent function words in English (Celce-Murcia and Larsen-

Freeman, 1999). Some researchers consider bare nominals, or the absence of any

articles, as an article itself which they refer to as the zero article (Quirk et al. 1991). If

the zero article is identified with replacing both a(n) and the in indefinite plural and

uncountable contexts, it can be considered as the most frequently used article in

English (Master 1997).1

However, despite the high frequency of occurrence in natural discourse,

advocates of the communicative approach in second language learning (cf. Canale and

Swain 1980) hold that articles might not be essential devices to achieve

communicative success. Brown (1973) wonders whether supplying articles is really

necessary for effective communication in English as a first language. Batainah (2005)

mentions that English articles can be disposed of, since they are regularly dropped

from newspaper headings and telegraphic messages while Leech (2006b) expects

articles to eventually disappear in the expanding world of L2 English.

Pica (1983), on the other hand, maintains that incorrect article usage can

impede communication when, for example, used in reference to items in one

participant’s experience but not in the other’s. Moreover, misuse of English articles

can, at least, be perceived as a sign of ‘foreignness’ by many native speakers. This is

1
We acknowledge Master’s differentiation between zero and null articles replacing a and the
respectively. In this study however both terms will be referred to as zero.

1
evident in popular fiction where erroneous article supply gives the impression that the

speaker is not in total command of English, i.e., a foreigner who uses correct but not

target-like English. Agatha Christie’s Poirot is an example of a well known fictional

character who is made to commit numerous article errors to emphasise that he is not a

native speaker of English.

The English articles are important because of their role as functional categories

required in linguistic expressions to signal meaning associated with the propositional

level as the and a(n) denotate the definiteness status of referents, whether definite or

indefinite (±DEF). The definite article implies that the speaker/writer assumes the

hearer is able to identify the referent. It maintains relevance in discourse and creates

cohesion within a text. The indefinite article indicates that the speaker/writer thinks

the referent cannot be identified by the recipients. It also has a cardinality function

(Lyons 1999) which is restrictive to singular countable entities.

The notion of (in)definiteness is universal but its grammatical representation in

syntax varies from one language to another which can present a challenge for second

language learners (2LL) especially those whose L1s lack articles (–ART) (Goad and

White 2004, Master 1997). Notorious as one of the most difficult features of English

to be learned or taught (Kaluza 1963, Brown 1973, Dulay, Burt, and Krashen, 1982,

Pica 1983, Master 1990 inter alia), misuse of articles ranks highest among L2

learners’ errors (Covitt 1976 cited in Celce-Murcia and Larsen-Freeman 1999,

Richards and Simpson 1974). Sharma (2005) established that article errors account for

60.37% of the total number of errors committed by L1 Indian learners of English,

while Thu (2005) found that article errors constituted 31.5% of all other errors made

by L1 Vietnamese learners. Thus, it is possible that articles represent an area of

2
‘considerable prominence in any error analysis’ reflecting learners’ overall linguistic

competence (Oller and Redding, 1971: 85) (emphasis added).

Article errors in learners’ production are pervasive in the sense that they do not

necessarily decrease as proficiency increases. The misuse of English articles has been

known to persist up to advanced levels (Anderson 1984) and the mistakes that many

competent L2 learners make could become a source of frustration. The difficulties in

mastering the English articles are worth pursuing especially when errors may

undermine the L2 speaker/writer’s linguistic ability in professional circles (cf. Nagata

et al. 2005).

The difficulty that learners have with articles is often attributed to the

complicated system in which the English articles operate. As multiple functions are

stacked into one form, the speaker has to be aware of the noun’s countability, number

and definiteness at the same time (cf. Master 1990). Consider, for instance, the

example in (1):

(1). She ate an apple.

In the above sentence, choosing ‘an’ requires knowledge of:

1. Definiteness: Apple is neither uniquely identified nor known to the hearer.

2. Countability: +Countable.

3. Number: Singular

4. Phonetic rule: A singular indefinite noun beginning with a vowel requires

the supply of ‘an’ rather than ‘a’.

3
Furthermore, socio-cultural knowledge is also required to attain native-like mastery of

article use especially in expressions of figurative or idiomatic meanings. For example,

the noun love in the expression: in love is usually left unmarked. This is not done

because of the noun’s unknown definiteness status or its uncountability, but mostly

because of the pragmatic interpretation it carries by conventional usage.

In addition to the complications inherent in the English article system,

learners’ L1 might also have an impact on the L2 acquisition process (Gass 1996,

Leung 2001, Snape et al. 2006, Garcia Mayo 2008). This is particularly true when the

L1 lacks a grammatical category which is present in the L2 (Master 1997, Odlin 2003,

Hawkins and Chan 1997).

Arabic is spoken as a first language by around 200 million people in the Arab

world and as a second language by many others. It includes many regional dialects

whereby various pronunciations are found, yet the grammatical notions of

definiteness/indefiniteness, countability and number remain uninfluenced by regional

variation in the spoken register. Lipiński’s (2001) comprehensive review of Arabic

dialects confirms that in addition to a number of lexical variations, the differences

between these dialects are largely phonological. Mitchell and El-Hassan also stress

that ‘Regional differences [in Arabic] are lexical and phonological before they are

grammatical’ (1994:2). In a comparative study of Moroccan, Egyptian, Syrian, and

Kuwaiti dialects, Brustad (2000) observes that there are a couple of

expressions occasionally used in certain dialects to function as quasi-articles such as

some /ʃi:/ and one /wahid/. However, the overall system is quite stable throughout the

dialects. Other than that, the extensive chapter on definiteness/indefiniteness in those

four dialects does not list any other variation.

4
Although Arabic is a language that has grammaticalised definiteness,

indefiniteness is indirectly expressed through the absence of the definite article. In

formal, standard and classic Arabic, indefiniteness can be (optionally) represented by

small, non-morphemic accents suffixed to words. Possessing one overtly marked

article in the form of a prefixed clitic /al/, Arabic is considered as a language with

articles (+ART). Yet, the absence of an explicit or obligatory marker of indefiniteness

in Arabic can present a challenge to Arab learners in mastering the English indefinite

article.

Therefore, it is not surprising, as researchers of SLA of articles confirm, that

Arab learners too struggle with acquiring and mastering the English articles. Kharma

(1981) found that misuse of articles was one of the highest reoccurring errors in a

study on Kuwaiti students majoring in English. Diab (1996) asserts that article errors

constituted almost 28% of all grammar errors in Lebanese students’ writing. In

Yemen, Al-Samawi (2004) found that article errors represent 38.4% of all overused

expressions in students’ English writing.

A few studies have addressed the difficulties that L1 Arabic (L1Ar) learners

encounter in mastering the English articles and offered a variety of explanations to the

causes of common errors. Studies by researchers such as Kharma (1981) Maalej

(2004), Batainah, (2005) and Sarko (2008) have provided valuable insight into the

acquisition of both the form and function of articles, despite having focused on

isolated features of the English article system (EAS) rather than comprehensively

addressing their occurrence in multi-variable contexts. These studies have suffered

from several limitations in scope. Firstly, the method used to elicit data is largely

based on form-focused tasks which usually investigate accuracy rather than target-like

use (TLU). Secondly, the grouping of participants is decided by academic level (in

5
university or high school) rather than by performance tests that determine proficiency

level (PL). Finally, there seems to be a gap in the understanding of definiteness

between traditional, more grammar based approaches and modern linguistic theory.

To determine appropriate article use in the elicited data, this study will be informed by

more recent theoretical frameworks developed by linguists such as Hawkins (1991),

Lyons (1999) and Langacker (2008), whereby context, relevance and consciousness

have a larger role in deciding definiteness rather than relying merely on grammatical

aspects of accuracy.

The fact that most well-known SLA studies were predominantly administered

to learners whose L1s lacked an article system (–ART) (Hakuta, 1976, Yamada and

Matsuura 1982, Pica 1983, Tarone 1985, Master 1987; Parrish 1987, Thomas 1989,

Yoon 1993, Young 1996, Mizuno 1999, Robertson 2000, Trenkic´ 2000, Butler 2002)

while SLA studies on L1s with articles remain relatively scarce, necessitates further

research into SLA of articles by learners from languages with articles.

This study aims to investigate the acquisition of the English articles (a, an, the)

by adult female native speakers of Arabic who are studying in the United Arab

Emirates University (UAEU) in the Emirates. It seeks to examine the stages of

development learners go through as their second language improves. Therefore, a

cross-sectional approach will be followed to review the progress of three proficiency

level groups as determined by the Oxford Placement Test (OPT). This includes

finding out which article is mastered before the other and why; what kind of

difficulties are encountered at each proficiency level and which aspects of article use

are easier or more difficult to master at each proficiency level. This is achieved

through exploring learners’ ability to provide the appropriate article in different NP

6
environments. Ultimately, we hope to provide a plausible explanation of the

underlying causes of these outcomes.

Another question that this study will address is the extent to which the first

language affects learners’ performance in the L2. Our findings will be compared to

those from other studies on SLA of articles by learners from other language groups in

order to detect similarities and/or differences in the acquisition patterns. If the

outcomes are not identical, it could be suggestive of possible cross-linguistic influence

of Arabic on L2 article production. Alternatively, parallelisms in learning behaviors

should lend support to advocates of broader developmental tendencies in L2

experience, regardless of the L1.

Finally, we seek to determine whether some errors are caused by faulty

associations between definiteness and certain (non-article-related) linguistic notions of

specificity, pre-modification and concreteness. Learners’ hypotheses will be reflected

in their use of English articles under different testing conditions. A comparison

between the L1/L2 grammars will be conducted to highlight the differences in the

morpho-syntactic representations of these notions in the two languages. Contrastive

elements might prove to have an effect on learner performance especially at lower

proficiency levels where L1 influence is at its highest (Odlin 1989, 2003). The error

patterns should reveal the extent to which pragmatic, structural and lexical features of

the Arabic NP affect L2 article choice.

The following chapter will review the background literature for this study. It

includes a summary of the basic theoretical concepts that determine article use such as

definiteness and some of the well-known SLA studies on articles. It also presents a

brief account of articles in Arabic, the findings of SLA studies administered on Arab

7
learners and finally a summary of linguistic notions that might affect learner

hypotheses about the L2. Chapter 3 is concerned with the method through which data

is collected, categorised and analysed as well as the approach followed for grouping

participants according to their proficiency level. The numerical and statistical results

emerging from the study are described in the fourth chapter. In this part, quantitative

findings across tasks and groups are outlined in tables, charts and diagrams. A

comprehensive discussion of the results and their relation to other studies on L2

articles is given chapter 5. Finally, the conclusion offers a summary of the most

significant findings of this study.

8
Chapter Two: Literature Review

In this chapter, an attempt is made to review the form and function of articles in

English which entails discussing the major criteria that determine the grammaticality

and appropriateness of the supply of each article. The most important element is

definiteness, for both articles, and countability and number for the indefinite article. In

addition to these basic notions, contextual, pragmatic and communicative elements

may be involved in determining appropriate article choice. The second part of the

chapter will identify the NP environments in which the supply of each article is

considered obligatory and the contextual conditions required for their supply. The

third part of this chapter presents a brief account of the article system in Arabic. This

is followed by a basic comparison of the main elements involved in article use in the

two languages. The findings of some of the well known studies in second language

acquisition (SLA) of articles in general as well as those administered on Arab learners

in particular will be discussed in section 4. Finally, we will consider the theories that

suggest a connection between certain linguistic notions (specificity, premodification

and abstractness) and L2 article use in learner hypotheses.

2.1 The Article System in English

The Articles the and a(n) in English occur as free morphemes classified under

determiners. The definite article falls into the determiner position of the definite

phrase (DP) (Lyons 1999), while the indefinite article is projected in the determiner

position of the cardinality phrase (CardP) since, in addition to indefiniteness, it signals

number. Some researchers consider the absence of the and a(n) as an article itself

termed the zero (Ø) or null article (see Master 1997). Others accept the indefinite

9
determiner ‘some’ as another indefinite article (Quirk et al. 1991:274). Diagram (1),

based on Yule (1998), summarises the classic grammatical view of article use in

English.

Noun

Indefinite Definite

Uncountable Countable

Singular Plural

Starts with
Vowel Consonant

Ø an a Ø the

Figure (1) Illustration of article use in English

Explaining the meaning and the function of the articles requires a better understanding

of the concept of definiteness as well as the notions of countability and number that

determine the appropriateness of article choice. Therefore, there is a need to take a

closer look at these concepts and their relation to the article system in English.

10
2.1.1 Definiteness

Definiteness is a universal category of meaning found in all languages, with or

without articles. Languages may differ in the way they express definiteness or

indefiniteness; whether in the form of explicit marking (English, French, Arabic) or by

variation in sentence structure whereby the identifiable NP is shifted to the topical

position (e.g. Chinese, Serbian, etc.).

Although definiteness encompasses all types of referential expressions such as

pronouns, proper names, demonstratives as well as syntactically salient expressions in

article-less languages, in this study, the term will be used within the scope of article

use only. Definite expressions can be semantically, pragmatically or structurally

identifiable (Searle 1969), whereas the definite article represents grammatical

definiteness. The definite article results from the process whereby lexical or pragmatic

identification is realised in a discrete linguistic feature. Lyons (1999) emphasises the

need to distinguish between definiteness as a grammatical feature and definiteness as a

category of meaning. The latter is often glossed as the ‘identifiability’ of a referent. In

this sense, definite descriptions can be replaced by demonstratives, while grammatical

definites cannot. Thus, the definite article is a ‘morpho-syntactic category,

grammaticalising a pragmatic category of identifiability’ (ibid: 282). Similarly,

Lambrecht (1994) holds that definiteness is ‘the imperfect grammatical reflection of

the non-discrete pragmatic category of identifiability’ (p.92) (emphasis added).

Indefiniteness on the other hand relates to expressions that result from not using the

definite article in corresponding expressions’ (Chesterman 1991: 417). In addition to

the value of indefiniteness, the indefinite article has a cardinality function to denote

singular number.

11
In previous research on the use of articles, attempts have been made to explain

how the article system in English works by adopting various theoretical approaches.

The following sections will explore some established linguistic theories on

definiteness with regards to articles only.

2.1.1.1 Uniqueness

Uniqueness has been associated with the definite article ever since Russell (1905)

claimed that a definite expression means that its referent exists and is unique. Löbner

(1985: 282) holds that definites have a universal interpretation of ‘referring to one

object in every case’. If we accept this argument, how can definite plurals fit into this

theory? Halliday (1976) proposed that definiteness reflects the individuation of an

item or a group of items that sets it/them apart from all others belonging to the same

kind or category. This would require sufficient endophoric (textual) or exophoric

(non-textual) information. In other words, uniqueness does not imply singularity.

Rather, plural nouns can be unique if treated as one group of items set apart from

others by the virtue of being identified.

However, semantically unique referents in a certain context are not always

definite. Indefinite singular referents can be equally unique. Example (2) from Thu

(2005: 41) will help illustrate this fact.

(2) Helen sang well and won a prize

It is possible that this prize was the only one in the contest, yet the employment of the

indefinite article is plausibly felicitous. Hence, uniqueness is a trait that both definite

and indefinite expressions share. Similarly, definite NPs are not always unique. For

12
example, the definite article in (3) and (4) is used with referents that can implicate

more than one person:

(3) He was the son of a poor farmer.

(4) I do not want my daughter to marry the lover of her sister.

(Löbner (1985: 304)

In both instances, the uniqueness parameter is flouted because in (3) the son does not

necessarily mean the farmer had one son only. Similarly, in (4), the lover is by no

means unique because the sister could have had more than one lover, yet the use of

definite article is felicitous. This suggests that the uniqueness of a referent cannot

solely justify its definite status, or the successful employment of the definite article.

The disambiguation of the references in examples 3 and 4 above can be realised if the

context is restricted to a particular one which both interlocutors recognise. This is

further discussed in the following sections.

2.1.1.2 Familiarity

The use of the definite article has been traditionally related to known/given

information shared between the speaker and the listener, while indefiniteness is mostly

related to unknown information. Christopherson (1939) was the first to describe

definite references as known information. He proposed that definiteness depends on

the notion of ‘familiarity’ which entails that a noun becomes definite if it is familiar.

In this sense, definiteness represents ‘old’ knowledge as opposed to ‘new’ information

which is assumed indefinite (Chafe 1976). Heim (1982) developed the familiarity

13
hypothesis into one of the most influential theories in this respect. She puts forward

that the definite status of a given noun depends on whether or not the information is

familiar to both interlocutors. If the referent is previously mentioned in discourse then

the information is known to the hearer and the familiarity condition is fulfilled and

definiteness established. Indefinite expressions were subdivided by Prince (1992) into

brand new and unused entities. In contrast, old or re-introduced information is definite

information. Prince stresses that it is the hearer’s role to judge whether the information

provided is new or old according to available clues in the discourse.

Bickerton (1981) considers that the hearer’s knowledge is an essential element

in determining the definiteness status, and therefore the appropriate article used with a

given noun. Bickerton’s model combines both hearer’s knowledge (±HK) and

specificity of reference (±SR) as the criteria that determine article choice. These

criteria, in four combinations, were proposed to cover most known uses of the English

article system as illustrated in Table 1 below:

Table 1 Bickerton's NP environments for article use

[-SR] [+HK] the, a, Ø Generic

[+SR] [+HK] the referential definite

[+SR] [-HK] a, Ø referential indefinites, first mentions

[-SR] [-HK] a, Ø non-referential nouns: attributive indefinites, non-

specific indefinite

The above categorisation has been widely influential in SLA studies of articles. This

will be discussed in more detail in the forthcoming sections of this chapter. Although

hearer’s knowledge has been considered an indispensible criterion in determining

14
article choice in various NP environments, the appropriate use of the definite article,

according to Schaeffer and Mathewson (2005), depends mainly on the speaker’s

perception of familiarity, i.e. if the speaker believes that the information is known to

the hearer, the referent would become definite, but if the speaker thinks that the

referent is not known to the hearer, it becomes indefinite. The third possibility is when

the information is known to neither speaker nor hearer, the referent is then indefinite

and non-specific.

There are many instances in natural language whereby the definite article is

acceptably supplied with new information. I.e., newly introduced referents can

become compatible with definiteness. Accordingly, unfamiliarity does not necessarily

entail indefiniteness. Consider the underlined NP in the example below:

(5) I bought a new car but the tyres seem old.

The use of the definite article in the above example is appropriate because it is part of

most people’s epistemic knowledge that cars come with tyres and therefore the tyres

in (5) refer exclusively to the car previously mentioned. This type of use was first

termed as bridging by Clark and Haviland (1974, 1977) and then as ‘associative

anaphoric reference’ by Hawkins (1978).

As first-mention definites exist, familiarity, as a conditional prerequisite to

definiteness, fails to justify such instances. The familiarity theory also fails to explain

the use of the definite article with new information that is not within the scope of

epistemic knowledge or associative reference. The definite article can be used with

15
referents that will only be identified later in the text. This type of use is often termed

as catopharic reference, examples of which can often be found in fiction.2

2.1.1.3 Information Structure

It is more likely that the information structure of an utterance will be presented as

given-new (cf. Halliday 1967) since it is known, in most canonical structures, that the

more accessible precedes less accessible material and that the given information

precedes new information (Givón 1983, Allan 1986). This position is supported by

Anderson et. al. (1991:674) who confirm that the nature of dialogue seems to prefer

placing known information in the topic position of sentences/utterances while new/

unknown information is mostly presented in the comment position. Logically, this

theory would lead to one possible outcome which is that given information is definite

while new information is indefinite. This is illustrated in the diagram below.

Initial position subject topic given information definite

While given information is typically definite and new indefinite, this need not be

the case. Lyons (1999, following Hawkins 1978)3 argues that new information can be

definite if it has been introduced to be explained and made salient later in discourse.

2
See Emslie and Stevenson’s (1981:326) analysis of first-mention definites in literature. Epstein (2002)
also provides ample examples from classic literature with referents known to the writer, not the reader,
being definite despite first mention. For a full review see Walter (1975).
3
Hawkins’ (1978) Accommodation Theory: The recipient is invited [by the speaker] to locate the
referent through previous or forthcoming information, then becomes more willing to
accept/accommodate a definite NP of the intended referent.

16
i.e., the position of a definite expression in an utterance can occur before it is

explained. The following example further illustrates this:

(6) We are looking for the vandals who broke into the office yesterday (Lyons

1999:10).

It is most unlikely that the addressee knows the identity of the vandals being

introduced at the beginning of an utterance as new information that is unfamiliar to the

hearer, yet a definite article is appropriately used. Hawkins (1978) suggests that it is

the ‘inclusiveness’ of forthcoming explanation which justifies this type of usage.

Based on a larger contextual derivation and degrees of coherence and relevance

(Sperber and Wilson 1986, 1995, Zegarac 2004), a referent can be identified through

further explanation, linguistically termed as cataphoric reference. Furthermore, Poesio

and Vieira (1998) examined the use of the definite article in newspaper corpus and

found that 50% of definite descriptions are newly introduced referents in discourse,

while 18% are bridged (associative anaphoric) references. In other words, almost 70%

of definite descriptions are not associated with a prior entity occurring in natural

discourse. Therefore, the structure of information in discourse does not relate to

definiteness in a straightforward way (Chafe 1976) since, regardless of the NP’s

position in an utterance, a referent can be identified by the recipient, if enough clues

are provided.

2.1.1.4 Pragmatic

Historically, the definite article is known to have evolved from the deictic terms this,

that. Hence pragmatic values of definiteness are embedded in the available literature

17
on the definite article. In this sense, the identifiability of a referent is largely

dependent on the immediate context in which the utterance is used, rather than its

semantic value or the grammatical conditions of its use. The listener, in a particular

situation, can then identify the referent that is seen or heard as it is immediately

accessible through physical presence. Searle’s (1969) Speech Act Theory proposes

that conversation is composed of implied messages (Conversational Implicatures)

dependent on and relevant to the situation in which the utterance was used. The

illocutionary force (ibid) of these utterances transcends the literal meaning of the

individual components in order to deliver the speaker’s intended message to the

listener. The amount of information the speaker provides has to be precisely adequate.

This is what Grice (1975) termed the Quantity maxim (Q) according to which the

speaker’s contribution in a conversation is successful if it is as informative as

required, not more and not less. Having the appropriate quantity of information will

enable the listener to correctly identify the referent through the Implicatures (I)

provided in context (ibid). Similarly, pragmatic knowledge enables the reader to

correctly identify the referent as it occurs within limited setting of a textual context.

To fulfil the conditions that require the supply of the definite article, Roberts (2003)

asserts, the referent needs to be unique in a context, textual or spoken.

The concepts of quantity and identifiability (Searle 1969) were first put

together by Kempson (1975) as truth conditionals for definiteness. The Q and I

principles were analysed and re-presented by Levinson (1987) as the key criteria

responsible for the derivation of conversational implicatures. Hawkins (1991) later

applied these concepts to his description of definiteness, whereby the pragmatic

setting (P-set) is paramount in determining definiteness. Hence the quantity of

information implied in the conversational context should be sufficient to help the

18
listener identify the referent, i.e. the supply of the definite article becomes obligatory

if there is enough information in the situational discourse.

In terms of article use, Zigarac holds that the definite article is used when the

speaker wishes to indicate that it is ‘sufficiently clear from the situation which

particular thing ….the speaker has in mind’ (2004: 200) (emphasis added). In this

case, the hearer may actually have little information about the noun that the speaker

has in mind and might ask for further clarification. The Location theory, advocated by

Hawkins (1978:167), suggests that when the speaker/writer introduces a definite

referent in discourse s/he is instructing the listener/reader to locate the relevant

information that contributed to establishing its definiteness status within the shared

pragmatic set in which the referring expression occurs. This is all based on the

hypothesis that the speaker/writer assumes that what little information is available is

adequate to form an idea of the referent. In other words, the speaker would have used

the definite article, being the stronger of the two articles (Hawkins, 1991: 417) if at all

possible. Indefinite expressions on the other hand are used when the speaker is not

sure if the knowledge presented is enough to identify the referent or not.

2.1.1.5 Specificity

There has been constant debate in the literature about the relation between

definiteness, specificity and article use. Quirk et al. (1985), for example, distinguish

between specific and generic references in their description of article use. Bickerton

(1981) found that NPs are divided according to notions of specificity in Creole

languages, and proposed that specificity of reference is one of the two main elements

of definiteness while the other is the information status of the referent, being either

19
known or unknown to the hearer. Bickerton presented the well-known Semantic

Wheel which combines both specificity of reference and knowledge as preconditions

to achieving definiteness. The model was later developed by Huebner (1983) who

reformulated the criteria of specificity (±SR) and hearer knowledge (±HK) into four

NP environments that determine article use. It follows that if a referent was both

specific and known (+SR +HK), the supply of the would become obligatory. It is

worth mentioning that, according to this approach, the definite status of a referent is

traditionally dependent on the hearer’s knowledge, while specificity on the other hand

is decided by the speaker.

The notion of specificity had originally evolved from Frege’s (1892) analysis

of natural language semantics (for a full review, see Horn 2007). Fodor and Sag

(1982) proposed that specificity lies in the speaker’s intention to refer to an entity of a

noteworthy property. This influential description was later confirmed by Givón (1984)

who stressed that what makes an expression refer is the speaker’s intent. If the

intention of the speaker is not to refer to a particular entity, then the reference

becomes non-specific or attributive (Donnellan 1966) and vice-versa. A specific NP

means that its description refers to a particular entity in the real world (Brinton 2000:

292) rather than to any item/person with general prototypical attributes that fit all

members of that kind.

Unlike some other languages (e.g. Samoan) which encode specificity, English

actually does not. Lyons suggests that although both definite and indefinite NPs are

‘potentially referring’ (1999: 166), the distinction of specificity/ non-specificity

applies equally to both articles. In other words, indefinite NPs can bear either specific

or non-specific readings, as in examples (7a) and (7b) below. Similarly, definite

20
descriptions can have both specific or non-specific readings as in examples (8a) and

(8b).

(7) a. Peter intends to marry a bank manager (even though he doesn’t get on at all

with her)

b. Peter intends to marry a bank manager (even though he hasn’t met one yet)

(from Lyons 1999)

(8) a. I want to buy presents for the children (whom I teach in my class)

b. I want to buy presents for the children (who might attend the party)

The bank manager in (7a) is specific because she is a particular individual whom Peter

knows, whereas in (7b) the referent could be any one who fits the description of a

bank manager, and therefore non-specific. Similarly, the definite referent in (8a) is

specific because the speaker knows these children, their names, age, number as they

are a particular group that stand out from all others (see Langacker 1997), whereas in

(8.b) the reference is not specific because there could be an unlimited number of

children whom the speaker does not know anything about (their names, age group,

number, etc.).

The specificity of a given referent in discourse does not always fall clearly into

either specific/ non-specific dichotomy; rather it varies in degrees. Givón (1993)

observes that NPs exhibit a continuum of referentiality, ranging from non-referring to

fully referring expressions depending on the ease/difficulty of access. However, given

the scope and limitations of this study, the specificity status of a reference will have to

be considered as either specific (+SP) or non-specific (-SP) to simplify task design,

data calculations and subsequent analysis.

21
Non-specificity, on the other hand, is related to the non-individuation of

referents. Non-specific or generic expressions are kind-referring lexical items that

stand for a whole class of analogous units. Langacker (1997) asserts that with

generics, the generalisation is global rather than local, i.e., instead of being applied to

a limited number of occurrences, the referent equally characterises a group.

Grammatically, it is possible to supply any article form to deliver a generic reference

without appreciable difference in meaning, as in examples (9), (10) and (11) below.

(9) A lion is a fierce animal

(10) The lion is a fierce animal

(11) Lions are fierce animals

The expression fierce animal equally applies to the singular and the plural, the definite

and the indefinite as it refers to a kind with typically similar qualities which people

recognise through their epistemic knowledge of the world (cf. Goldsmith and

Woisetschlaeger 1982). It is clear from the above examples that distinctions of

in/definiteness become neutralised in generic contexts, because the focus is on a whole

set rather than on a specific member of a set (Lyons 1999). In other words, these

nouns become semantically neither definite nor indefinite, but could be better

described as being class identifiable (ibid: 236).

However, there are certain expressions where one article is preferred because

of custom or convention, while using another article becomes less preferable or at

least non-native like within the same context. For example, the relatively low

occurrence of generic a(n) in corpora suggests that the indefinite article does not have

a strong generalising power (Master 1987) yet, in certain regional dialects in the US, it

22
was the more preferable option in delivering generic meanings (Hong and Seonkyung

2001). Furthermore, despite the neutrality of articles in generic referencing, there is a

formality element that affects appropriate article choice. Celce-Maurcia and Larsen-

Freeman (1999) consider the indefinite article as the least formal while generic the is

believed to be more formal and bare plurals were presented as the most formal.

However, a study done by Master (1987) confirms that generic the was more frequent

in scientific journals than other article(s). Finally, it has been widely acknowledged

that the bare plurals are inherently ambiguous and therefore mostly non-referential

(Weir 1986, Krifka 1995, 2003, Behrens 2005). This quality makes bare plurals the

most popular form of generics in written discourse.

2.1.1.6 Cognitive

Salmon-Alt and Romary (2000), following Langacker (1991), suggest that reference

can be established cognitively across domains rather than being dependent on the

individual descriptions of the entities themselves. Therefore, both definite and

indefinite articles can be grammatically used with both new and old information,

unique and non unique NPs, specific and non-specific referents. It is hence necessary

to introduce a more accurate description of the linguistic criteria that determine the

conditions of appropriate article use.

Strawson’s (1950) notion of presupposition was later incorporated into

cognitive literature and became essential in later interpretations of definiteness. Thus,

it is when the speaker presupposes, rather than knows, that the hearer is familiar with

the information in an utterance that the referent becomes definite. According to the

cognitive approach, definiteness depends mostly on the status of a referent in the mind

23
of the speaker or his supposition about the mind of the addressee. By knowing ‘other

minds’ (Givón 1989: 207), cognitive theory does not approach definiteness on the

basis of knowing the identity of the referent, but addresses it from the angle of mental

representations of the referent in question (Lambrecht 1994:335) in the interlocutors’

minds. If the entities have similar representations in the interlocutors’ mental spaces

(Fauconnier, 1994), and the interlocutors achieve mental contact the referent becomes

definite. The task of a speaker is to assess the hearer’s cognitive environment, and to

choose which article to use to accordingly.

Chafe (1976) posits that givenness is not a category of knowledge but of

consciousness where information is active, inactive or re-activated in discourse.

According to Gundel et al. (1993) determiners restrict possible referents by

conventionally signalling different cognitive statuses within a hierarchy of givenness.

In definite constructions, the speaker assumes that these signals are available in the

present memory and attention state of the addressee which would enable them to

identify the intended referent. Lambrecht (1994) presented an argument which

combined the traditional view of familiarity with the more cognitive notion of

accessibility. He classified information types as categories of either consciousness or

knowledge. If the category is consciousness-related, definite expressions will result

from ‘activated’ information, believed by the speaker to be currently available in the

consciousness of the hearer (Chafe 1976). ‘New’ information, on the other hand, is the

information which is either accessible (familiar) to the hearer, but not presently

available in their consciousness, or inaccessible and therefore indefinite. As categories

of knowledge, given and new mark the distinction between the entities believed by the

speaker to be either active (ibid) or accessible, and therefore identifiable, or

24
inaccessible and therefore unidentifiable. The division can be schematically

represented as in Figure 2:

Category of Consciousness: GIVEN NEW

The status of the referent: Activated Accessible Inaccessible

Category of knowledge: GIVEN NEW

Figure 2. Information Status and accessibility of referents

Epstein et al. (1996) maintain that the function of the definite article is to signal that a

referent is ‘accessible from the point of view of both speaker and hearer’ (p. 109).

Therefore, felicitous use of the definite article is contingent not only on the hearer’s

knowledge (familiarity theory), but also on the speaker’s assessment of the hearer’s

knowledge (ibid: 100), i.e., it is up to the speaker to assume whether the hearer is able

to identify the information referred to in discourse or not.

A cognitive explanation of the use of the definite article with new information,

otherwise inaccessible in the traditional view of familiarity (see example 6) is that the

definite article triggers a frame. Within this frame (context) the referent is evoked

(activated) as the speaker invites the listener to search for a route or a path (Epstein

2002, Langacker 2008) to follow in order to identify the referent through further

information in discourse.

The degree of accessibility of a referent then depends on the relevance of such

information and its spatial/temporal distance to the referent (Ariel 1988) as well as its

25
prominence (Epstein 2002). This forms a continuum of levels of activation, ranging

from highly active (accessible) to non-active (inaccessible) information. If the referent

represents a prominent idea, then it can be easily recalled by using the definite article

to re-activate this idea. If the information cannot be accessed, then indefiniteness

arises as a result of lacking information on the part of the listener, or an

underestimation on the part of the speaker about the hearer’s knowledge.

In the previous sections, several theories that describe definiteness have been

presented. The concepts in this overview approach definiteness from different angles

providing a wide spectrum of possible explanations which should help analyse learner

production in the Discussion chapter.

2.1.2 Countability and Number

The countability status of a noun plays a crucial role in determining the

grammaticality of article use in indefinite NPs and, combined with the number value,

whether or not the indefinite article is to be obligatorily supplied. Countable entities

are defined as those that can stand independently as individual units (e.g. chair, pen,

idea, etc) but can also be pluralised (e.g. chairs, pens, ideas). On the other hand,

uncountable entities represent ‘conceptual information about substances and

abstractions’ (Yule 1998:31). They include mass substances (e.g. rice, sugar, water)

and some collective nouns (e.g. humanity, furniture) and abstract notions. 4

4
Certain collectives are compatible with singular verbs (e.g. government, family, crowd) under
restrictive contextual conditions. For a detailed account see Depraetere (2003)

26
To capture the distributional characteristics of countability, Langacker (1991,

1997) proposes that nouns can be categorised according to their boundedness. A

bounded entity is separate, enclosed and delimited and can be thus conceived as a

single item or group. Unbounded entities on the other hand, cannot be individuated

into separate items/units. The noun water, for instance, is uncountable because it

cannot be cut and divided into smaller units. Another way to conceptualise the

count/non-count distinction is based on the fact that the two classes of nouns in

English are drawn from two separate semantic domains (Chierchia 1994) whereby a

countable noun denotes a set of individuals or items while an uncountable noun

denotes portions of substances or fusions of materials whose denotational domain

does not consist of individuals. Sensitivity, for example being a human emotion,

cannot be individuated or calculated and is therefore uncountable.

Yet Countability is neither categorically polarised nor consistently

dichotomous. There are numerous instances where nouns, formally categorised into

one countability class, occur with characteristics of the other, as in the examples (12,

13) below.

(12) Can I have a coffee?

(13) It was a sudden death

It is known that mass nouns (coffee) and abstract nouns (death) are uncountable which

usually prohibits the supply of the indefinite article, yet both (12) and (13) are

grammatically correct as well as semantically logical. One explanation, offered by

Quirk et al. (1985), suggests that some nouns possess Dual class memberships (p: 247),

27
i.e., the same noun can be countable in one context but uncountable in the other.

According to Langacker, a formally-uncountable noun can be conceived as bounded

(countable) if it occupies a ‘circumscribed portion of quality space’ (1991: 18) which

distinguishes from other substances such as the cup or the mug in which the coffee, in

(12), is put in.

Another countability shift is noticed in contexts where proper nouns, typically

uncountable, exhibit qualities of countable nouns in enumeration by permitting the

presence of the plural suffix (14.a) and the indefinite article in (14.b):

(14) a. All the Johns that I know are intelligent

b. A Mr. John called this morning.

Traditionally, proper nouns are considered uncountable, yet the singular and the plural

forms in the examples above are grammatical and the expressions are commonly

acceptable. Furthermore, with regards to articles, the use of the both definite and

indefinite articles is felicitous.

Chesterman (1991) discerns article choice in overlapping cases of countability

on the basis of typical use. For example, nouns such as Furniture typically reject the

indefinite article, while others, such as ‘humanity’ conventionally reject the presence

of the definite article. However, the examples in 14.a and 14.b do not represent the

typical uses of proper nouns. Instead, John is grammatically used as a common noun

because it represents a name borne by more than one person and thus permits the use

of articles.

28
Abstract nouns, known to represent attitudes, statuses and emotions (e.g.

wealth, happiness, hatred) are largely uncountable, yet there are many abstract nouns

that are countable (e.g. feeling, joy, thought, etc.). In certain occurrences, abstract

nouns -perceived as uncountable- can shift in class into countable entities (Quirk et al.

1985).5 Consider the following example from Master (1997:225)

(15) It has a wonderful rhythm.

Master explains that it is the evaluative modifier wonderful that causes the shift in the

countability status of the word rhythm, because, in this context, the expression is

reduced from the original: It has a wonderful [kind of] rhythm. Therefore, the

indefinite article refers to the missing words rather than the uncountable abstract noun

rhythm. Similarly, the use of the indefinite article with a generically uncountable noun

in (13): It was a sudden death is appropriate because the writer -probably a journalist-

assumes that the readers would be able to guess that there is a missing (countable)

noun, which bears the meaning of accident/incident/case and therefore requires the

supply of the a(n). Cognitively, the elliptic expression signifies the speaker/writer’s

confidence that the recipients will easily fill-in the information gap by drawing on

their epistemic knowledge of the world.

In this sense, the argument that formulaic expressions (or set phrases) require a

special use of articles6 (Nettle and Hopkins 2003) because they seemingly violate

traditional classifications of countability, will collapse since article use can be

explained by the hypotheses of countability shifts, dual-class memberships (Quirk et

al. 1985) and the range of a countability continuum determined by context (Allan

5
For a range of comparative phrases with countability shifts of nouns, see Master (1997).
6
Many Grammar textbooks have a separate section for special or exceptional cases of article use. See
for instance Swan and Walter (2001).

29
1981). For instance, a noun that is originally classified as countable, occurring in the

singular form, can become uncountable in certain contexts with a generic

interpretation, referring to a kind rather than an individual instance/item/person

(Chesterman 1991, Behrens 2005). As a result, the obligatory indefinite article

becomes redundant. On the other hand, the supply of the indefinite article can become

obligatory in certain contexts with uncountable nouns as in example (15) below.

(16) A person without a past is someone without a future.

Past and future are conventionally uncountable, yet the use of the indefinite article is

equally acceptable in both NPs. Hence, instances of countability shifts, as in the

examples above, should not be conceived as exceptions to the rules. Rather, the

context becomes the parameter against which the appropriateness of such occurrences

is measured. The interlocutors in a given context are supposed to possess situational

knowledge that enables them to deduce the intended the referent even in cases of

ellipses. Countability, therefore, describes the context, not the head noun itself. This

view explains all exceptions, shifts and dual-memberships. For instance, in example

(12) the speaker, probably a customer in a cafe bar or a guest at a friend’s house,

realises that the quantity of information is enough for the listener, perhaps the waiter

or the host, to realise that the utterance is contracted and that the indefinite article is

referring to a cup of coffee, rather than the amount of coffee. Contextual knowledge

can also justify the shifts in the countability statuses of proper nouns. In (14.a), Johns

is a contraction of the longer version of the phrase: all the men that are called John

into this shorter form. In (14.b) the indefinite article is referring to the missing word:

person: A person, whose name is John called this morning.

30
Finally, one of the most influential views on countability presented by Allan

(1981, 2001) suggests that countability should be construed as a continuum rather than

a dichotomy. Some words are inherently more countable while other expressions are

less susceptible to countability although in some contexts they might assume a

countable position. In this study, however, we shall have to dichotomise NPs into

either Countable or non-Countable categories based on context in which they occur to

facilitate the analysis of test results and the processing of statistical calculations.

2.2 Article Contexts

2.2.1 Definite article environments:

It seems from the above that definiteness is based on two basic notions of

identifiability by both speakers and uniqueness within a particular context. Therefore,

a referent becomes definite when it is uniquely identifiable by interlocutors in a

mutually manifest pragmatic (P) set online. Hence, NP environments, in which the

articles are known to occur, are considered obligatory when there are certain

conditions which necessitate the suppliance of either article to ensure grammatical

correctness. Failing to do so could result in grammatically unacceptable constructions

or at least ambiguous and misleading references. These contexts may vary in certain

language communities. A study by Hollmann and Siewierska (2011) for example,

examined the phenomenon of definite article reduction (DAR) in the Lancashire

dialect, a variety spoken in the North West of England. In this study, however, the

criteria which determine obligatory environments of definite article will be defined

according to broader terms that transcend the limitations of particular regional

31
conditions. The approach we follow is derived from Liu and Gleason’s (2002)

classification of (non-generic) contexts of definite article use, which, in turn is based

on the theory of definiteness advanced by Hawkins (1978) and Lyons (1999).

Accordingly, the supply of the definite article is necessary if the referent is introduced

to the addressee either textually, pragmatically or is part of the interlocutors’ common

knowledge. The following is a summary of NP environments in which the supply of

the definite article becomes ‘obligatory’ in verbal and textual communication.

1. If the head noun is previously introduced in the textual context

(second/subsequent mention)

I bought a book. The book was expensive.

2. If the head noun is conventionally associated with a previously mentioned noun

(Associated anaphoric reference, bridging)

I went to a wedding. The bride was beautiful.

3. If the head noun is semantically unique in a universal sense/ larger situational

context:

The sun is shining.

The world is a stage.

4. If the head noun is unique in a smaller situational context of a certain community

or group:

Where is the teacher? (A student asked another in the same lecture hall)

32
5. If the head noun is physically/immediately present (seen, heard). Deictic

reference.

Can you please pass the salt?

6. If the head noun is identified by explanatory modifiers such as defining/

explanatory relative clauses. Forward identification/cataphoric reference

The car that she bought is new.

7. If the head noun is adjectivally premodified with a superlative adjective:

He is the fastest runner in the world.

8. If the head noun is preceded by semantically unique expressions: the only; the

same

33
2.2.2 Indefinite article environments

Indefinite interpretations are analysed as implicatures that result from the absence of

the definite article. i.e., a NP becomes indefinite when it does not meet the conditions

required for definiteness. The I-principle, represents implicatures of not using the

definite article. In other words, the speaker uses a(n) because it was not possible to

have used the.

Heim (1991) ascribes this lack of definiteness to either one of the following

possibilities: the referent exists, but is not unique (17.a); or it does not exist in the

speaker’s knowledge of the hearer’s mind (17.b).7

(17) a. I bought a new bag (any one bag in the world)

b. I bought a new bag (which you have never seen/heard of before).

Although (a) and (b) are exactly identical, the reasons for not using the definite

article are different because in each case one of the conditions for definiteness is not

fulfilled. While indefiniteness is only indirectly implied -by the absence of the definite

article- it is the cardinal property of a(n) which explicitly and directly refers to the

singular status of the noun. This is why a singular indefinite NP has been branded by

linguists as a cardinal phrase (CardP) (Lyons 1999). Since the indefinite article had

originated from the numeral one, both are mutually exclusive as they both indicate

singularity and occupy a similar position (Specifier) in the NP. Possessing the

numerical value disqualifies the indefinite article from being the equivalent

7
While Heim’s (1982) theory on definiteness depended mostly on the familiarity condition, her later
work (1991) shows an incorporation of pragmatic and cognitive concepts into the previous proposals.

34
counterpart of the definite article. Rather, as is the case with the definite article, the

polar opposite of the indefinite article is no article.

The Absence of the indefinite article signals that the noun is indefinite and that

it is not singular, i.e. either plural or uncountable. NPs without articles are known as

bare nominals, yet in many SLA studies on articles, nouns in bare constructions are

frequently described as being modified by the zero (Ø) article. The zero article is not

the only marker of indefinite plurals. Another indefinite determiner is the quantifier

some which is considered by many researchers as the plural form of the indefinite

article. Yet unlike the indefinite article, its use is not obligatory and therefore its

absence does not make the utterance ungrammatical. Furthermore, while bare plurals

are largely non-referential, the indefinite plural determiner some is often used

referentially.

2.3 The article system in Arabic

In this section, we will review the usage of articles in Arabic and briefly outline their

form and function in the Arabic NP structure. A basic contrastive analysis with

corresponding notions from English will also be presented in an attempt to highlight

potential sources of learner errors.

Arabic is a Semitic language formally categorised as a language with article(s)

(+ART) because definiteness is grammaticalised. However, there is no explicit and

obligatory marking of indefiniteness. In that sense, it is similar to Hebrew

and Bulgarian which are two other languages that possess a definite article but do not

require overt marking of indefinite nouns. This fact does not necessarily entail

35
deficiency in functional categories because the semantic notion of indefiniteness is

available in these languages but its morphological representation is not compulsory.

While definite descriptions can be both specific and non-specific (±Specific)

in Arabic, indefinite expressions are mostly used in specific references. Specificity is

largely indefinite; generic reference is all definite (Hawas 1989).

2.3.1 Definiteness

Definiteness in Arabic is grammaticalised as well as obligatory. The definite article is

transcribed as a bound affixal morpheme /al/, which occurs as a proclitic prefix to the

head noun as is shown in example 18.

(18) ‫ﺃﻠﺑﻴﺖ‬

Al-bait

The house

Similar to English, the definite article in Arabic occurs before nouns and adjectives. It

is similarly supplied in all the obligatory contexts of the English definite article (see

2.1.3). However, unlike the English articles, the definite article in Arabic is not a free

morpheme. Another difference lies in the function of the definite article. In addition to

identification, the definite article in Arabic is largely used in most generic references

regardless of the level of formality or convention. Arabic is comparable to French and

Greek in the sense that definite descriptions can carry both specific and generic

readings.

36
With regards to information structure, known, given, information mainly

occurs in the subject position of most canonical structures and is considered as topic,

whereas the complement position is typically reserved for new, indefinite information.

Therefore, subjects are largely definite while predicates are indefinite in canonical

structures. The verb usually precedes the subject in classic Arabic, but colloquial

dialects apply the SVO order more often. Adjectives occur post-nominally and agree

with nouns in number, definiteness and gender. The definite article /al/ does not

necessarily render the adjective uniquely identifiable or a superlative (as is the case in

English). Rather, the vacuous definite article (Lyons 1999) serves as a sign that links

the definite noun to its adjectives: Definite article + N + definite article + adjective:

)19( ‫ﺃﻠﺑﻴﺖ ﺍﻠﮐﺒﻴﺭ‬

Al-bait al-kabeer

def-house def-big

The big house

It is worth mentioning that in generic contexts, the definite singular form is often

preferred to the plural (cf. Kremers 2003, Maalej 2004, Schulz 2004). Hence, a

complement to the NP could be: The house is usually larger than the flat. In Arabic,

this would be interpreted generically, whereby the scope of reference is extended

beyond a certain, uniquely identifiable, house, to the larger group that includes all

houses in the world.

An expression can be equally definite in both Arabic and English in genitive

constructions, but the syntactic position of the definite article in a canonical NP is

different. Syntactically, Arabic is a determiner-genitive (DG) language which allows

37
possessives to occur in the determiner position (Hawas 1989). In this respect it is

similar to English, but the difference between the two languages lies in the structure of

the genitive constructions. In Arabic the definite article is part of the complement

rather than the topic. This topic, although unmarked, is semantically definite through

being post-modified with a grammatically definite noun. Price in (20) and plan in (21)

are uniquely identified by being limited to the context of the following definite noun.

(20) *Price of the oil has gone up (from task 3).

(21) *Plan of the government is to build new schools.

2.3.2 Indefiniteness

Indefiniteness in Arabic is grammatically expressed by the absence of the prefix /al/ in

a given NP. It is enough indication of the indefinite status of a given NP. Thus, unlike

languages with two or more articles such as French, English, Spanish or Italian in

which the marking of indefinite singular nouns is obligatory, indefinite marking is

optional in Arabic. Therefore, its absence does not render a given utterance

ungrammatical. Furthermore, indefiniteness markers are exclusively limited to the

formal, written classic register while in the spoken, regional dialects of Arabic there is

no audible indefinite article (Schulz 2004).

To mark indefiniteness in classic Arabic, a small sign or accent termed

‘tenween’ in Arabic or nunation in English (Smith 2001) is suffixed to nouns. There

are three types of nunation: /in/, /un/ and /en/. The choice depends on the function of

the noun; A Subject suffix is pronounced /un/ and is transcribed as a small sign above

the last letter. Object suffixes on the other hand are pronounced /en/ and are

38
transcribed above an /a:/ sound added to the last letter, while accents for nouns that

fall within a prepositional phrase are pronounced /in/ (Smith 2001) and are transcribed

under the last letter. The indefinite markers are almost unnoticeable because the

suffixed noun remains identical in form to the un-suffixed one because indefiniteness

markers are represented in small accents written over or under the last letter of the

word.

Structurally, both SV/VS sentence structures are permissible in Arabic.

Moreover, unlike English, indefinite nouns are premodified with indefinite adjectives

(Adj + N) as in example (22) below

(22) ٌ‫ﺑﻴﺖٌ ﮐﺒﺑﺭ‬ ‫ﻫﺫﺍ‬

Haatha bait-un Kabeer-un

Dem:prox house-N-Indef-Sg big-Adj-Indef-Sg

This is a big house

Countability in Arabic is very similar to English in the sense that there is a difference

between singular and plural noun forms and that most mass and abstract nouns are

uncountable. Nonetheless, there are a few instances where the two languages do not

share the exact perceptions of noun countability. Some nouns which are largely

uncountable in English are countable in Arabic. These include many abstract concepts

(e.g. information, advice, work, time experience, luck, news, help and knowledge),

some mass nouns (e.g. equipment, paint, chalk, wood,) and a few collective nouns

(e.g. sheep). The opposite is also true but extremely limited. In other words, there are

only a few cases where nouns which are countable in English, (e.g. life) are

39
uncountable in Arabic. Such lexical disparities in classification across languages

(Quirk et al. 1985, Celce-Murcia and Larsen-Freeman, 1999) may lead to misuse of

articles by L2 learners (Yoon 1993). With regards to number, the indefinite markers

in Arabic do not have a cardinality function as accent suffixes occur equally with

singular, plural or uncountable nouns.

As the Arabic speaking media is moving towards accepting vernacular

versions of spoken Arabic, indefiniteness is becoming less explicitly marked even in

formal contexts. This discrepancy between the formal and informal, or spoken and

written discourse, causes most Arabic speakers to neglect marking indefiniteness

especially when the absence of indefinite markers does not render utterances

ungrammatical. These facts, in addition to some lexical disparity in countability, could

turn into constraints that limit L2 acquisition because of the possible negative transfer

of these features into the target language.

2.3.3 Comparison of article use in Arabic and English

Hawas (1989) offers a comprehensive, cross linguistic analysis of the major

differences in the article systems of Arabic and English as well as a review of the

main structural and semantic elements that characterise representations of definiteness

in these languages, in addition to other influential factors such as countability and

number. A more modest attempt is presented in Table 2 below outlining some of the

basic factors that distinguish article use in Arabic and English.

40
Table 2 Comparison of articles between Arabic and English

Linguistic
English Arabic
criteria
Obligatory supply Obligatory supply
Definite article
To uniquely identify To identify and generalise

Obligatory Optional

Indefinite article
To classify one item as non- To classify one or a group of

uniquely identifiable items as non-uniquely

identifiable

Prenominal Postnominal

Uncountable Countable
Adjectives
Agrees in (in)definiteness with
No definiteness value
the head noun

Abstract nouns Largely indefinite Always definite

Generic Mostly indefinite plural Always definite

Mostly singular

Genitive case Definite subjects Definite predicates

41
2. 4 Second Language Acquisition (SLA)

The term acquisition refers to the collection of processes that lead to production, use

and ultimately, full attainment of another language. It therefore encompasses the

different ways of internalising language, ‘be it learning a system or picking up useful

expressions’ (Crystal 1987). A distinction between acquisition and learning was

originally made by Krashen who maintained that acquisition ‘is very similar to the

process children go through in acquiring first language’ (1981:1). Accordingly,

language is attained naturally through exposure and produced communicatively with

meaning as the target. Learning on the other hand refers to the knowledge gained

through classroom instruction, expressed mostly in formal contexts with large focus

on accuracy. However, since it is difficult to be absolutely certain of the source of the

participants’ L2 knowledge, whether it came solely from instruction, or was a result of

individual interest in reading novels or watching films in English, in this study the two

terms will be used interchangeably regardless of the source, means and targets.

Traditionally, SLA studies have approached L2 data from the perspective of

Universal Grammar (UG). Supporters of UG believe that humans are endowed with

an innate language faculty that enables them to acquire (any) language through

accessing subconscious knowledge of universal concepts (principles). The forms of

these abstract concepts are represented in actual languages, each of which has its own

set of descriptors (parameters) that vary from one language to the other. In the

process of acquiring a second language, learners will have to adjust the settings of

their L1 parameters to new L2 values. Some very well known studies in SLA support

the UG theory (e.g. Dulay and Burt 1972, Epstein et al. 1996 inter alia). However, UG

theorists hold different positions with regard to the degree of accessibility learners

have to UG principles and the extent to which L1 parameters are operative in the L2

42
process. Some hold that L1 parameters affect production but to a certain stage when

they switch to the L2 grammars (The Full Access Full Transfer hypothesis. Schwartz

and Sprouse 1996). Others propose that it is the UG, not the L1 parameters, that

influence initial analyses of the L2 (Full Access hypothesis. Epstein et al. 1996).

Partial availability suggests that only some UG principles are accessed directly

without going through the L1 parameters (Schachter 1990, Johnson and Newport

1991, among others). In this sense, the role of UG is often perceived as the opposite to

that of the L1. However, the argument that UG and L1 provide opposite explanations

to second language production/error is not always valid, since the UG theory does not

exclude the role of the L1 in the learning process. Rather, the power of the first

language has often been considered influential by UG theorists.8 Therefore, in this

study, the term developmental will be used to refer to the similarities in L2 production

when error patterns, orders of acquisition and learning strategies are common among

learners of divergent L1 backgrounds. On the other hand, variation can become

evidence of transfer in the production of learners of a particular language group once

erroneous usage patterns are traced to the systems found in learners’ L1s.

However, the question of whether or not the strategies employed in the

processes of acquiring second language knowledge are innately generated or UG

governed is beyond the scope of this study. In other words, this is not an attempt to

find evidence for or against the presence of UG principles and parameters in learner

data or the extent to which they are operative. Nevertheless, some UG-based SLA

theories that address the L2 acquisition of articles (e.g. the Missing Surface Inflection

hypothesis (MSIH) Prévost and White 2000, the Failed Functional Features

8
For a full review, see White (1998).

43
hypothesis (FFFH) Hawkins and Chan 1997; the Representation Deficit hypothesis

(RDH) Hawkins 2004) will be considered as possible interpretations of our findings.

Usage-based studies on the other hand view L2 acquisition as part of a larger

cognitive process whereby language ability is not separable from epistemic knowledge

and the acquisition of L1 and L2 is not identical. According to this approach, learners

abstract regularities from the input and construct new forms in the L2. Usage based

models better account for L2 errors since they accommodate a multiplicity of

interacting factors such as the type of input, level of awareness and the attentional

demands of various tasks (see also Skehan 1989). The processabilty of L2 knowledge

in learners’ minds is dependent on the syntactic category of the form,

lexical/grammatical complexity of the item learned and its perceptual salience in

context (Goldschneider and DeKeyser, 2001) among other factors. Furthermore, the

frequency of input plays a determinant role in the way L2 progresses (cf. Ellis 2002,

Gass 2002). Usage-based theories, such as Goldberg’s (2009) Constructionist

Approach offer a more comprehensive view of SLA since cognitive, pragmatic and

processing constraints are equally involved in measuring L2 achievement. More

importantly, UG is not excluded as a language acquisition device (LAD) while

learners’ first language remains one of the most influential elements in the L2 process.

Incongruent SLA development patterns were reported in studies involving

learners from different L1 backgrounds as the semantic and syntactic properties of the

L1 are believed to affect the L2 production. The effect of L1 semantics is reported in

a study by Snape et al (2006) on Spanish and Japanese learners. Incompatible results

were found in the two L1 groups’ use of the definite article in specific and non-

specific contexts. Similarly, Ionin et al. (2008) compared the performance of L1

Russian with L1 Spanish learners and found discrepancies in their interpretations of

44
specificity-definiteness with regards to article use. The syntactic properties of

learners’ L1 have also been noticed in L2 production as was recorded by Jarvis (2002)

in a study on Finnish and Swedish learners. Transfer is believed to be manifest in the

overextension of analogy of L1 rules into L2 structures (Ellis 1994: 59).

Advocates of the Contrastive Analysis (Lado 1957) believe that the presence

of a feature in L1 facilitates the acquisition of the same feature in L2 (Ard and

Hombarg 1983, cited in Mclaughlin 1987). Brown (1994) maintains that ‘interference

can actually be greater when items to be learned are more similar to existing items

than when items are entirely new and unrelated to existing items’ (p. 201). Ringbom

(1987), among others, holds that similar L1/L2 representations of functional,

grammatical and lexical categories can lead to positive transfer evident in earlier

emergence and faster mastery. Similarly, Goad and White (2004) propose that if

certain features of the L2 functional morphology, such as the articles, are not available

in the learners’ L1, then negative transfer of the L1 prosody (prosodic transfer) is

expected to occur in learners’ production.

If we assume that only syntactic features available in L1 are easily acquired,

while others -absent from the L1- are not, then learners’ performance will always

remain non-target-like (Smith and Tsimpli 1995, Hawkins and Chan 1997,

Franceschina 2001, Hawkins 2001) because there is a shortage in corresponding

representation.

In SLA literature it is typical that in acquiring a second language, learners go

through a series of phases through which production deviates from the target language

(TL) but is not necessarily attributable to influence of the learners’ first language.

Selinker (1972) proposed that such errors can be explained if the language produced

45
by learners was perceived as being an Interlanguage (IL). This notion

comprehensively covers L2 production during all stages of second language

development from the initial point of emergence, or the onset of production (Ellis

1994), until full attainment, which indicates the learners’ mastery of the target

language. Richards (1971) identifies five distinct processes of the IL phenomena:

1. Hypercorrection.

2. Over-generalisation.

3. Ignorance of rules and restrictions.

4. Incomplete applications of rules.

5. Falsely hypothesised concepts.

Richards’ study analysed non-target like use of articles and ascribed learner errors to

the universal language learning tendencies listed above. Towell and Hawkins (1994)

suggest that in addition to the known IL causes, L2 errors can be ascribed to a

subconscious transfer of the L1 mental grammar into the IL production, followed by a

staged development of the IL towards more target like L2. Many SLA researchers

have assigned great significance to the learners’ first language (L1) in influencing the

L2 learning process and have ascribed learners’ production to transfer whether

positive or negative (Gass 1996, Goad and White 2004, Ellis 2006). The argument for

and against first language effect on SLA is based on the extent of its influence and the

type of influence. Most researchers agree that the first language is most influential at

earlier learning stages or at lower proficiency levels (Odlin 1989, 2003, Schwartz and

Sprouse 1996, Slabakova 2000, Haznedar 2001, inter alia) since it is the primary

source from which newly acquired knowledge is built. Haznedar stresses that there is

46
‘compelling evidence for L1 transfer in … early L2 English’ (2001: 24 emphasis

added).

Research in SLA has also been concerned with the way L2 notions evolve in

learners’ minds through examining written or spoken output administered under

various testing conditions. It has been reported that learners draw upon different types

of knowledge, depending on the manner through which knowledge is received, how

the information is stored and the way it is accessed in production. The distinction

between the two types of knowledge originated in the philosophical literature (see

Ryle 1949). Linguists who adopted this idea distinguished between learned and

acquired competence (Krashen 1982); controlled and automated production

(Anderson 1983); explicit and implicit knowledge (Bialystok 1987, 1988); declarative

and procedural ability (Anderson 1980, Levelt 1989), analysed and automatic forms

of representation (Ellis 1984), integrated linguistic competence and conscious

intellectual understanding (Hale 1988) and performative and propositional knowledge

(Crystal 2000). Recently, the terms conscious/subconscious became more popular in

referring to the two types of knowledge (see Roehr 2005 for example). In this study,

both types of knowledge were considered in the design of the tests and data analysis.

The next section will include a brief review of literature on the acquisition of

articles with special attention to L1 Arabic (L1Ar) learners. A summary of certain

linguistic notions L2 learners have been known to hypothetically associate with article

use will also be presented.

47
2.4.1 L2 Acquisition of Articles

The two articles in English have not been reported to be acquired at the same time nor

follow the same route of development in SLA. Studies show that each article is

produced and mastered at variable IL stages and to incur different error types (e.g.

omission, overuse, etc.) at different PLs. Several criteria, such as learners’ L1, PL and

task type, in addition to various NP contexts, to name but a few variables, determine

the L2 development map of each article separately.

Many researchers ascribe L2 errors to the influence of their first language

(Huebner 1983, Master 1987, Ringbom 1987, Odlin 2003, Young 1996, Goad and

White 2004 inter alia). L1 induced errors are supposed to account for up to one third

of all 2L errors (George 1971, Abisamra 2003) or more (cf. Habash 1982). Most

researchers agree that syntactic and semantic differences between the L1 and the L2

result in ‘negative transfer’, while it is known that similarities lead to positive transfer

(Ringbom 1987) which, in effect, facilitates learning (Lado 1957).

In the field of article acquisition, transfer depends on the presence or absence

of a feature in the L1. There is a common view in SLA that learners with –ART L1s

have more difficulty in acquiring the EAS than learners with +ART L1s. Masters

(1987) contends that learners whose L1 possess articles are at an advantage in their

second language acquisition of English articles. L2 studies have since confirmed

Master’s theory. For example, Spanish and Greek (+ART L1) learners in Hawkins et

al.’s (2006) study performed better than their Japanese counterparts. The proposal that

there is an advantage to possessing articles in learners’ L1 is evident in the delayed

emergence of articles and the higher inclination to omit articles in the production of

learners from –ART L1s. The mastery of the bare-plural/uncountable nouns and the

48
higher overuse rates of the zero (Ø) article by learners from article-less backgrounds

further substantiates the former hypothesis. A study by Young (1996) on Czech and

Slovak learners yields that the two articles only emerge at later IL stages, while the

failure to supply articles (omission) might persist onto more advanced stages. Similar

findings were reported by Thomas 1989, Master, 1997, Trenkic´ 2002 and Ekiert

2004). Similarly, Paradis and Zdorenko (2008) recorded more omissions in the L2

production of Korean, Chinese and Japanese (–ART) learners of English than in the

production of Spanish, Arabic and Romanian (+ART) learners.

It was also found that the omission stage is usually followed by a phase in

which the definite article, rather than the indefinite, is overused in indefinite contexts.

This universal phenomenon in article acquisition is known as the ‘the-flooding’ stage.

It is worth mentioning that the flooding of the definite article is reported to occur after

initial omission stages, but well before mastery of both English articles (Huebner

1985, Parrish, 1987, Thomas 1989, Chaudron and Parker, 1990 and Master, 1987).

This pattern seems to be reflected in most L2 production regardless of L1.

Most studies report that the mastery of the definite article precedes that of the

indefinite (Huebner 1983, Master 1997, Parrish 1987, Thomas 1989) while mastering

a(n), according to Thomas (1989) is at least significantly delayed. Parrish’s student

did not master indefinite contexts until Time 9, which is a relatively late stage in the

12 set phases of her L2 learning time. The relative delay in acquiring the indefinite

article has been attributed to its multiple functions (cardinality, indefiniteness) by

many SLA researchers (Master 1997). Others, such as Haznedar and Schwartz (1997),

Lardiere (1998), and Robertson (2000) hold that the variability in L2 production is

caused by the difficulty in mapping surface morphological forms, i.e., articles, onto

49
abstract notions (e.g. definiteness, countability, number). This position is known as the

Missing Surface Inflection Hypothesis (MSIH) (Prévost and White 2000) which was

followed by Slabakova (2000) in her study on Bulgarian learners of English.

From a developmental perspective, the RDH (Kuribara 1999, Hawkins 2004)

offers an explanation for the difficulty in mastering the indefinite article. It suggests

that the delay is caused by the absence of formal representation of certain L2 features

from learners’ IL. This hypothesis is supported by many researchers in the field such

as Snape (2008). It is also noticeable that the indefinite article was found to be better

supplied in existential (There + is + a(n) + N) and have constructions (Huebner 1983,

Butler 2002, Yule 1998) as most indefinite singular phrases are learned as formulaic

chunks, especially in formal education (Trenkic´ 2000).

It is worth mentioning that this progression in L2 article acquisition

corresponds to the sequence reported in studies on the L1 acquisition of English. It has

been frequently reported that bare nominals (zero article) temporally precede the

emergence of articles (cf. Zehler and Brewer 1981, Cziko 1986, Roeper 2003).9

Researchers ascribe early absence of articles to delay in the emergence of functional

categories while overuse of the definite article, according to studies in

psycholinguistics, is ascribed to child ego-centricism (cf. Maratsos 1976). Thus a child

(speaker) would consider a referent as definite if it is identifiable to them disregarding

the perspective of the hearer.

9
See also Clahsen et al. 1994 and Chierchia et. al. 1999

50
2.4.2 L2 Acquisition of articles by Arab learners

All studies on L1Ar learners’ acquisition of English articles agree that the definite

article is the first to be acquired, while producing appropriate bare plurals was found

to be the most difficult. In the well-known study by Kharma (1981) on Kuwaiti high-

school and university students, it was revealed that the last article to be acquired and

mastered was the zero (Ø) article. Al-Fotih also found that the highest percentage of

errors in the cloze tests he administered on Yemeni students was the misuse of the

zero (Ø) article. These results are divergent from findings obtained from other studies

on learners with –ART L1s.

However, researchers also provided evidence for the tendency of Arab learners

to overuse the definite article (Kharma 1981, Batainah 2005, Maalej 2004) across

indefinite contexts. Yet researchers such as Habash (1982), Diab (1996), Al-Fotih

(2003) and Kharma and Hajjaj (1989) attribute the error to the negative transfer of the

generic definite in Arabic while others, such as Abi Samra (2003) and Batainah (2005)

believe that overuse is caused by the universal L2 phenomenon of the-flooding.

According to the MSIH Arab learners are expected to experience greater difficulty in

producing new L2 features, such as the indefinite article, because there is no

equivalent formal representation available in their L1.

On the other hand, another group of studies argue that errors made by Arab

learners are compatible with those made by learners from other L1s. In this sense, the

variability in article production is developmental and experienced by all L2 learners.10

The impact of Jordanian learners’ first language was considered ‘minimal’ (2005:13)

10
As L2 development occurs during the various stages of Interlanguage (IL), the terms developmental
and IL will be used interchangeably to describe common tendencies in SLA. Despite the subtle
differences between the two terms, the universal learning processes that these expressions convey is in
contrast with variability caused by the influence of a particular L1.

51
by Batainah. Along the same lines, Abi Samra (2003) asserts that L1 cannot account

for more than one third of the errors made by Lebanese learners. Similar to results

obtained from learners with –ART L1s, Arab learners have managed to better supply

the indefinite article in existential and have constructions (Maalej 2004, Batainah

2005, Abi Samra 2003).

The effect of other linguistic notions such as NP structure, ontology and

referentiality on Arab learners’ L2-production of articles will be reviewed in the

following section.

2.5 Learner Hypotheses

Learning a second language involves a continuous process of hypothesis making,

testing and experimenting. During the various IL stages, hypotheses are either

sustained by positive evidence or dropped because of negative feedback. It is possible

that learners’ interpretations of the L2 input and the variability in production are

derived from associations between certain forms and functions of features in the L2

with the syntax and lexis of the L1. Previous knowledge thus provides the basis on

which hypotheses of the L2 rules are formulated, but L1 can also constrain the

hypotheses learners make on L2 grammars (Schachter 1983).

To find out whether there learners hypothetically associate certain linguistic

notions with definiteness, several studies examined L2 error patterns and compared

between the error rates in two contrasting NP environments: one with the linguistic

notion present and another without. If the overuse rate of one article, for example, was

found to be significantly higher in the context with the linguistic notion present, then

52
there is definitely an association between article use and that particular criterion and

vice-versa.

In this section, we will review three linguistic notions which have been

considered in earlier studies to affect 2LL’s use of the EAS. They include the notion

of specificity, the structural aspect of adjectival premodification, and the ontological

concept of concreteness/ abstractness.

2.5.1 Specificity

Errors made in L2 production of English articles have been attributed to difficulties

learners encounter with the specific / non-specific distinction (Hakuta 1976).

Traditionally, it was conceived that learners associated definiteness with hearer’s

knowledge (±HK) (Huebner 1983, Young 1996, Master 1997). i.e., the overuse of the

definite article was considered to be the result of overestimating the hearer’s

knowledge and vice-versa. Huebner (1983) postulates that after the initial the-flooding

stage, the definite article gradually recedes from non-specific (indefinite) contexts.

However, many others regard the overuse of the in indefinite contexts as being due to

the encoding of specific readings as definite. Thomas (1989) observed that most

instances of overuse were in referential (indefinite) contexts and therefore contends

that the definite article is associated with its referential function -as opposed to HK- in

learner hypothesis. This view is corroborated by other studies such as those of

Trenkic´ (2002) and Butler (2002) where it was found that learners attributed

referential meanings to the English articles, i.e. the definite article is associated with

[+SP] referents while the indefinite article was associated with [–Sp] reference.

53
Recently, there has been a debate in the SLA literature about the effects of

specificity on L2 article use brought about by Ionin, Ko and Wexler (henceforth IKW)

(2003) who presented a hypothesis suggesting that 2LL of English go through a stage

of fluctuation (FH) between the values definiteness (±Def) and specificity (±Sp) when

using articles, until, after being exposed to sufficient input, target-like accuracy is

attained. During the fluctuation phases, articles in L2 production encode either

specificity or definiteness. Using articles to represent specificity (rather than

definiteness) is similar to Creole languages which divide NPs according to notions of

specific/non-specific meanings (Bickerton 1981). The fact these studies were all

administered on learners of –ART L1s must be taken into consideration. Hawkins et

al. (2005) as well as Snape et al (2006) compared the outcomes of two L1 groups of

learners, one group with +ART L1 background and the other from a –ART L1

background. Both studies noticed that learners from +ART L1s did not fluctuate (i.e.,

associate definiteness with specificity) while learners whose L1s lacked articles did.

On the other hand, a group of researchers rejected the L1 effect in the

definiteness-specificity fluctuation. Paradis and Zdorenko (2008), for example, claim

that there is fluctuation but it is a developmental error that overrides transfer. Their

study was based on two groups of young L2 learners; One group with –ART L1s

(Chinese, Korean and Japanese) and the other was from a +ART L1s (Arabic,

Romanian and Spanish). They found that children from both language groups

substituted a(n) with the in [–def +Sp] contexts; thus associating specificity with

definiteness. This is in line with the proposal forwarded by Matthewson et al. (2001)

which claims that (all) learners of English go through a stage in which specificity is

associated with definiteness. Their study was administered on 25 American children

54
from the Northeast region of the US where the influence of a Salish11 language is

predominant.

However, this hypothesis might not apply to L1Ar learners, especially if we

assume a strong L1 influence since in Arabic, as in Greek and Hungarian, generic

reference is grammatically definite (Kremers 2003, Behrens 2005). The example

below is a literal translation of a popular phrase in Arabic.

(23) *In choosing a flight, the most important thing is the safety.

(From task 3)

Such occurrences of generic-the are probably common in L1Ar learners’ production

and frequently observed by English language teachers across the Arab world. In a

study on Tunisian university students, for example, Maalej (2004) confirms that one

of the main impediments to learners’ mastery of English articles was the overuse of

the with bare plurals in generic contexts. In other words, L1Ar learners do not

randomly flood the definite article across indefinite contexts; rather, the tendency to

overuse is limited to –specific (generic) contexts (ibid.). In a study on Yemeni learners

of English, Al-Fotih (2003) reported similar findings. Conversely, Sarko (2008) found

that Syrian learners predominantly selected the definite article in [–def +Sp] contexts.

It is difficult to confirm this view because, despite the random occurrences of non-

referential the in L2 production, not many studies have addressed the alternative

possibility of a [+def] [–Sp] association especially in the production of learners whose

L1s employ the definite article in generic referencing.

11 Salish refers to a group of endangered languages including St’át’imcets (also known as Lillooet)
spoken in the southwest region of British Columbia, Canada. In these languages, specificity, rather than
definite determiners, encodes identifiability.

55
2.5.2 Premodification

SLA Researchers have noticed that nouns modified by adjectives seem to be treated

differently by L2 learners than unmodified nouns (cf. Young 1996, Jarvis 2002). In

general, learners have been frequently reported to better supply the indefinite article in

non-premodified constructions [Art + N] than in adjectivally premodified ones [Art +

Adj + N]. Trenkic´ (2000) reports higher omission rates in premodified contexts. She

explains that articles and adjectives compete for the same determiner position in the

IL syntax (in –ART languages),12 once that position is filled with another element

(i.e., adjectives) articles will be omitted. Similarly, Jarvis (2002) found that Finnish (–

Art) learners were more reluctant to mark nouns -modified by adjectives- with articles

than Swedish (+ART) learners. Hawkins et al. (2005) justify higher omission rates by

Japanese (–ART) learners by proposing that learners’ mental activation levels [of

referents], dependent on the strength of connection between head noun and article,

decrease when the article and the head noun are syntactically separated by an

adjective.

Another view is that negative transfer accounts better for errors made in cases

where the basic word order of the two languages differs (cf. Odlin 1989). i.e., there is

a relationship between the acquisition of rules governing L2 article use and NP

structure of the L1 (Zobl 1986): if the L1 and the L2 differ in the sequence of nouns

and adjectives, there will be a delay in mastering the articles caused by restructuring

the L1 rules to comply with those of the TL (see also Givón 1984, Bickerton 1981).

Parodi et al (1997) also confirm that there is a strong influence of L1 DP syntactic

properties on the L2 production. Goad and White (2004) suggest that Prosodic

12
Hypothetically, Arabic can also fall into the category of –ART L1s where indefiniteness is concerned
since it lacks an overt morphological representation of indefiniteness.

56
Transfer from the L1 (PTH) is responsible for the higher omission rates of articles in

modified contexts (Art + Adj + N) made by SD, (see also Goad et al. 2003, White

2003) whose L1 (Turkish) lacks a morphological representation of definiteness.

Although Turkish is the exact opposite of Arabic where articles are concerned,

i.e., indefiniteness is explicitly marked while there is no grammatical representation of

definiteness, in principle, PT could be equally operative in the L2 production of L1Ar

learners. In other words, the absence of an indefinite marker in Arabic would provide

an opportunity for higher indefinite article omissions in modified constructions.

However, the opposite was found to be true in a study by Hamdallah (1988) as Arab

learners overused, rather than omitted, the indefinite article before adjectivally

modified nouns. Hamdallah maintains that ‘learners over-generalised the rule that

requires the use of a before adjectives followed by singular nouns’ (p.218) and applied

it to plural nouns premodified by adjectives because L1 Arabic learners were ‘unable

to recognise that a refers to the noun that follows the adjectives’. It is not surprising

that L1Ar learners dealt with adjectives as singular nouns (ibid) since in Arabic,

adjectival modification largely is post-nominal (cf. Fassi-Fehri 1999). Therefore,

syntactically, modification in Arabic is the mirror image of that in English (Kremers

2003).13 It is expected that the different syntactic structure of Arabic and English as

well as the absence of an indefinite marker will present a challenge to L1Ar learners

in the process of acquiring the EAS.

13
Shlonsky (2004) describes modification in all Semitic languages as being the opposite of that in
English.

57
2.5.3 Concreteness

In theory, concrete nouns refer to tangible objects, largely perceived as solid and

discrete items with limited boundaries occupying a certain space. They are first order

entities (Lyons 1977) that can be individuated into separate items or grouped as

homogenous units bounded together (Langacker 1997, 2008)). Therefore, concrete

denotations can be singular, plural (countable) or mass (uncountable) nouns. Abstract

notions, on the other hand, are categorised as second order entities (Lyons (1977) and

usually refer to non-solid, amorphous substances such as attitudes, feelings, events or

attributes which are largely intangible. Expressions such as wealth, happiness, hatred

sadness (see Countability 2.1.2) cannot be separated into individual units neither

could descriptions such as ‘cumulative, distributive, or homogeneous’ possibly apply

to abstract nouns (Joosten 2003:220). Thus, while concrete nouns can be ±COUNT

and ±Singular, abstract nouns are largely uncountable (ibid).

In contrast, abstract notions can be either countable or uncountable in Arabic.

For example, nouns such as wealth and sadness are countable whereas happiness and

hatred are not. In terms of (in)definiteness, although there is no grammatical relation

between the ontological status of a referent and its definiteness status, abstract nouns

notably occur in indefinite contexts (Larsen-Freeman and Murcia 1999). However, in

genitive constructions abstract references can be definite since they become the

possession of a certain person, group, or object. Notice, for example, the difference

between ‘geography’ in (24) and (25):

(24) You must study geography (INDEF).

58
(25) Tom studied the geography (DEF) of England. 14

In Arabic, however, abstract denotations are systematically definite (Kremers 2003,

Schulz 2004).

Richards (1976) postulates that it is not only the grammatical complexity that

influences the order of second language acquisition but also the difficulty of the

lexical concepts expressed by that category. Vainikka and Young-Scholten (1994,

1996) maintain that lexical categories can transfer from learners’ L1 to L2. Hence, it

is tenable to expect that variability in article production might be influenced by the

ontological properties of the head noun and its definiteness status in the first language.

Van Patten et al. (2004) concede that there is a form-meaning relationship in L2

production. In this sense, learners might associate the meaning of the head noun with

the form of article chosen in that context.

Trenkic´ (2002) reported that Serbian learners made form-meaning

connections in their L2 use of articles as the overuse of the definite article was

significantly higher in concrete than abstract contexts. Conversely, the indefinite

article was produced equally well with abstract and concrete nouns. This pattern

suggests that there is an inclination to associate definiteness with concreteness; and,

according to Trenkic´, abstract notions as sometimes perceived as individuated

discrete items. Similarly, Hua and Lee (2005) found that the ontological category of

the NP affects the sequence of L2 acquisition by Chinese learners of English as their

grasp of singular countable noun rules appeared earlier in concrete, rather than

14
In Arabic genitive constructions the predicate becomes obligatorily definite (see example 16 above)

59
abstract contexts; and therefore, their awareness of the rules governing the use of the

indefinite article was also higher when the (singular) head noun was concrete.

Yet, results reported from studies on L1Ar learners do not yield identical

association patterns. The influence of L1/L2 variation is evident in the results of a

study by Diab (1996) who examined the writing of Lebanese students and recorded

significant overuse instances of the definite article in abstract contexts. Equally,

Tunisian students overused the definite, rather than the indefinite, article with abstract

nouns (cf. Maalej 2004). Despite the previous proposals which suggest association in

learner hypothesis between concreteness and definiteness, results emerging from L1

Arabic studies suggest that errors can be triggered by associating definiteness with

abstractness. In this study, we aim to find out whether the ontological status of the

head noun causes variability in L1Ar learners’ choice of articles.

60
Chapter Three: Methodology

Chapter three presents the approach followed in conducting this research. In section

3.1, the necessary information required to form the database of this study is identified.

Section 3.2 describes the participants who volunteered to take the tests, their

background and language levels. The design of each of the three tasks is explained in

Section 3.3. Task Design examines the rationale for implementing different tests in

terms of knowledge type, focus and degree of control while details of data elicitation

and test administration are reported in section 3. 4. The fifth part explains the

techniques followed in analysing the data, and offers a basic account of how accuracy

rates and error types were calculated. In addition, this part provides the grounds for

investigating hypotheses underlying article use in learner production in relation to

certain linguistic notions. Section 5 of this chapter also reviews the peculiarities of

evaluating input from various tests. The final part of the chapter points out the

practical limitations of this study.

3.1 Required data

The data required for this study should cover a wide range of L2 production including

a variety of forms, structures and lexemes in order to act as a sample corpus of article

use by Arabic speakers. For example, the data should include definite and indefinite

descriptions of countable and uncountable nouns in singular and plural forms. This

necessitates a good number of participants to provide statistically reliable data.

There is also a need for the elicited data to reflect various degrees of overall

L2 knowledge in order to follow the effect of the second language development on

article production. This was achieved through a cross-sectional analysis of learner

61
input to capture the stages that learners go through as their English improves. The

information needed was taken from participants within a succession of proficiency

levels in an attempt to follow, synchronically, the L2 progress which might otherwise

be operationalised longitudinally. It is accepted by theorists and applied by field

researchers (e.g. Raymond and Fisher 2002) that the variation across different

proficiency groups can reflect, to a large extent, some aspects observed in diachronic

change. Accordingly, preliminary tests had to be conducted to ensure appropriate

banding of L2 levels by an internationally recognised examining authority. Tests such

as The Oxford Quick Placement Test (OPT 2001) have been used in similar studies to

determine L2 learners’ proficiency levels in English (Hawkins et. al. 2005, Snape et al

2006, Al-Thubaiti 2007, Pongpairoj 2007 and Garcia Mayo 2008 inter alia).

Furthermore, there was a need for participants to be of similar L1 background

in order to enable the researcher to measure the extent of L1 transfer in learner

hypotheses by applying linguistically contrastive criteria (see Table 2 in 2.2.3). A

similar L1 background also helped distinguish between L1 influenced errors from

more developmental tendencies in L2 article use.

Expected variance in performance is not only tied to proficiency level but is

also dependent on the type of test from which the data originates. Many researchers

have confirmed that different task types yield different results (Tarone 1983, 1985;

Tarone and Parrish 1988, Krashen 1982, Skehan, 1998), i.e., a one group of learners

might demonstrate different tendencies, accuracy rates and error patterns depending

on the form of the test. Hence, it was necessary to explore the performance of learners

on a range of tasks to ensure obtaining balanced and objective results.

62
Tests vary in the type of knowledge they explore (Bialystok 1982), the degree

of control exercised over learner production, and in whether the focus of learners is

directed to meaning or form (see Long 1991). Therefore, there is a need for three

types of tests: a free production task, a cloze test and a grammaticality judgement test.

Deployment of these types is necessary to obtain data that is most reflective of

students’ knowledge produced in different testing environments. Data collected from

free production helps explore learners’ spontaneous use of the L2. This type of test

taps into learners’ implicit (automated) knowledge of language (Krashen 1982,

Paradis 1994) which is derived mostly from hypotheses structured on the basis of

target features noticed in the input (Ellis 1994). With minimum control by the

examiner in the directions, such tests would entice participants to communicate their

ideas and feelings, i.e. the main focus of such tests is on meaning.

On the other hand, the collected data should also provide evidence for

learners’ explicit knowledge of the L2, gained mostly from formal pedagogical

instruction. Cloze tests, such as multiple choice exams, generally reflect learners’

meta-linguistic knowledge about language. The information obtained from this type of

tests helps assess learners’ awareness of the formal properties of articles, such as

countability or number. This is usually achieved through a higher degree of

production control by means of precise, largely restrictive, prompts that drive

attention maximally towards form and accuracy.

The data collected from either type of test should also reflect learners’

hypotheses about article usage. In the context of this particular study, NPs were

analysed in terms of specificity, premodification and their ontological status to find

out whether or not article use was affected by specific and non-specific reference,

63
adjectivally premodified and non-adjectivally premodified structures as well as

abstract and concrete meanings. The results should enable us to verify if L1Ar learners

associate L2 articles with linguistic notions of referentiality, premodification and

ontology, respectively.

3.2 Participants

The participants in this study are a group of female students from the United Arab

Emirates University (UAEU) studying in different colleges and departments including

English (e.g. Literature, Translation and TESOL) and non-English majors. The study

was conducted on the women’s campus in Al-Ain city.

After having acquired ethical clearance from the university administration to

perform the tests, volunteer participants were asked to sign a consent form stating

their agreement to take the tests in question. Both Arabic and English versions of the

consent forms were provided.

The students were then asked to answer a few questions aimed at ensuring

background uniformity (see Appendix 2). The purpose of the questionnaire was to

confirm that the participants were homogeneous in terms of age, first language,

parents’ language, length of residence in English speaking countries and number of

years of formal study of English and finally type of school attended.

To be eligible to take part in the study, participants should not have lived in an

English-speaking country for more than three months nor been to an English medium

school. All participants had to be native speakers of Arabic, regardless of nationality,

with native Arabic-speaking parents. Most of the participants were Emirati but there

were also a few other Arab nationals. Those who qualified to take the tests had all

64
taken English language classes under the same state-curriculum for 8 years,

comprising two years in primary school and six years in high school, with an average

of five hours per week. In addition, students of the UAE University have all had a one

year intensive course of English in the University General Requirements Unit

(UGRU) before starting their majors in various colleges and departments in the UAE

University.

Also included in this study is a group of twelve native speakers of English who

work as lecturers/professors in different disciplines in the UAE University. They

kindly volunteered to act as a control group by answering the questions in the cloze

test and the GJT.

3.3 Task Design

The aim of the tasks designed for this study was to thoroughly investigate learners’

acquisition of English articles through examining their competence and performance

under various testing environments. To achieve this purpose, three tasks were

designed to elicit various kinds of information based on the type of knowledge

involved, focus and degree of control. The three tasks were: a free composition task, a

multiple-choice blanks test and a grammaticality judgment test (GJT). Ultimately,

these tests should help find how production patterns vary from one test to the other.

The three tasks vary in the priority of attentional resources (see Skehan 2003).

It has been proposed that performance goals in L2 production include communication

on the one hand and grammatical accuracy on the other (see Crookes 1989, Mehnert

1998, Ortega 1999, Skehan 1998). Focus in the three tasks is scaled from full attention

65
to meaning in the writing test (T1) to full attention to form in (T3), with T2 possessing

variable degrees of both elements.

The variety in test types also provides the researcher with multiple

perspectives on learners’ types of knowledge. The writing test mainly taps learners’

communicative competence (Hymes 1972) derived from their subconscious

understanding of articles, while controlled tests with explicit questions that prompt

article production tap the learners’ conscious metalinguistic knowledge of the L2 (see

chapter 2 section 3). In terms of the control exercised by the examiner, the three tests

range from very little in free production to full control in T3.

3.3.1 The Writing Task (T1)

The writing task is a test designed to elicit learners’ unrestrained L2 production with

attentional configurations directed towards communication (Hu 2002). As a free

production task, this test could help investigate learners’ implicit knowledge (Krashen

1982, Paradis 1994, Muranoi 2000, Ellis 2004) acquired through exposure and

positive evidence.

In the prompts, the learners were asked to describe their hometowns in essays

of 350- 500 words in length. The question is not entirely unfamiliar as UAE

University students come from different emirates and they are often asked to introduce

themselves in terms of the emirates they come from. Learners were also directed to

include information about the location, tradition, life style and tourist attractions of

their hometowns, but no reference to article use was mentioned. The topic provides

students with an opportunity to express themselves freely and creatively, and a variety

of forms and structures would be produced to cover the NP descriptions required for

66
this study. For example, when participants described some noticeable features of their

hometowns, new information was introduced. This ensured the availability of

indefinite constructions. Definite expressions, on the other hand would result from the

repetition of or elaboration on the former mentioned information.

The data required to examine learner hypotheses and assumptions with regards

to associations between article use and certain linguistic notions (specificity,

adjectival premodification and concreteness/abstractness), were obtained from the

produced text without the learners’ conscious introspection of the underlying

grammatical rules. The wide scope of the topic should facilitate the inclusion of

specific and generic references, while the suggestions in the rubric to discuss culture,

traditions and values should help produce some abstract notions needed for the

dataset. Furthermore, the instruction to describe hometowns was intended to help

furnish the data with sufficient adjectives.

Free production tests are known to accrue higher accuracy rates (Tarone 1983,

Tarone and Parrish 1988, Foster and Skehan, 1996) typically attributed to learner

strategies such as avoidance, common to free production tasks, which would

eventually lead to a decrease of grammatical errors. Nonetheless, such tasks are also

likely to incur high erroneous instances since learners, unaware of the purpose of the

test, would be inclined to omit articles. In this case, accuracy outcomes could be

negatively affected (Hakuta 1976) as the erroneous occurrences might balance the

overall high accuracy rates accrued in such tests causing results to become more in

line with those emerging from more controlled tests.

67
3.3.2 Stories Task (T2)

The second task is a fill-in-the-blanks test comprising six short stories of diverse

themes but comparable lengths. With the articles removed from the original text, the

stories were used as a forced-elicitation test. In the prompts, the participants were

directed to fill in the blanks with one of the articles (a, an, the) or to leave blanks

empty (Ø) if no article was required. When the data was analysed, the quantifier

‘some’ was also accepted as a grammatically correct marker of indefinite plurals,

though not mentioned in the prompts.

The text used in this study was first used by Trenkic´ (2000) on L1 Serbian

learners of English, but many changes to the original version had to be implemented

in order to make it more suitable for the needs of this study. One such change was to

add a few explanatory modifiers such as relative pronouns, in order to make the

definiteness status of the entities in question unambiguous and the supply of the

definite article more obligatory. A case in point is illustrated in (1).

(1) Original: _____ party they had last night.

Altered: _____ party which they had last night.

In addition, some events and contexts in the stories were adapted to be more culturally

suitable for the Emirati environment. For example, the break-up between a girl and

her boyfriend was turned into a friends’ row, and, in another story, the drink bottles

left after a party were crossed out. Furthermore, the number of blanks was reduced

after removing NPs in which appropriate use of articles requires knowledge of

formulaic or idiomatic expressions which might carry meanings other than those

68
corresponding to the sum of their individual constituents (Cowie and Mackin 1975,

Seidl and McMordie 1992). For cultural suitability and avoidance of set-phrase

structures, the blank number (135) in story (4) of the original test (He might be in

____ love) was removed. The final alteration was to simplify some vocabulary that

might have constituted an additional challenge to the participants.

The criteria required to build the dataset for this study are naturally available

in the texts of the stories. There are 120 definite NPs and 98 indefinite NPs, 166

countable and 55 uncountable; 125 singular nouns and 93 plurals. The text also

includes 55 premodified nouns, 23 of which are definite and 30 indefinite. In addition,

the stories comprised 57 abstract notions, 64 generic and 110 specific references.

Therefore, T2 covers all possible NP contexts to elicit the necessary information with

frequent occurrences of the listed variables. The results were calculated in

percentages to rationalise comparisons across contexts.

The goal of introducing this type of task is that it represents a mixture of focus,

control and knowledge types. Firstly, although the learners’ focus was driven mainly

towards form, the content of this task was utilised as a source of meaning.

Accordingly, the opportunity of forward and backward referencing should have

helped participants to better realise which article is more suitable for the available

contexts. Secondly, in terms of control, T2 is traditionally considered a cloze test

because the specific directions given in the prompts limit the items participants are

allowed to insert in the slots and thus exclude other options. Participants also have no

choice with regards to the structural position of blanks as this is pre-determined by the

examiner.

69
Due to the fact that the text is authentic, it was not possible to ensure that all

variables were equally represented. For example, the number of concrete nouns, in the

total dataset of six stories, was 161, while only 57 occurrences were classified as

abstract. This type of task is therefore often referred to as semi-controlled (see

Trenkic´ 2007, Swan 2005) because of contextual effects on the outcome. In terms of

knowledge type, the results that emerge from this test are expected to be largely

attributable to metalinguistic knowledge since learners would have to consciously

decide on which article to employ, drawing on rules they have been formally taught at

school. Proponents of the communicative approach believe that the interpretation of

the cloze test results does not accurately reflect learners’ understanding of a certain

feature in the L2 because, unlike natural communication where the focus is on

meaning, such tests raise awareness of form and limit the language to the options

available in the prompts. Therefore, the results will reflect only a part of learners’

knowledge giving ‘an inadequate view of how learners actually use articles’ (Thomas

1989:339) [emphasis added]. However, Oller and Conrad (1971) asserted that data

collected via cloze-tests can differentiate levels of proficiency more adequately.

Furthermore, in more recent SLA studies, researchers hold that cloze, content-based

tests, such as T2 can reflect, to certain extents, both implicit as well explicit types of

knowledge (see Skehan 1998, 2003, Hu 2002, Roehr 2004).

Thus, on the basis of the criteria according to which the three tasks were

chosen, T2 can be placed midway between T1 and T3 in all the respects discussed.

70
3.3.3 Grammaticality judgment test (T3)15

The term Grammaticality Judgement refers to a kind of test carried out in many well-

known studies on second language acquisition. Its purpose is to examine learners’

ability to accept the usage of a certain feature in the L2 as correct or to reject it as

wrong (cf. Chaudron 1983, Birdsong 1989, Ellis 1991). What determines the

acceptability of one utterance and the unacceptability in such tests is the NS’s

judgements. When the test was piloted, only two responses were allowed from the CG

despite the fact that many utterances included may vary in their degree of

acceptability (Schütze 1996:70) in the target language. However, no inconsistencies

were found in the NS’ data.

It is the third test in this study, introduced as a complementary method of data

elicitation to supplement the results obtained in the stories and the writing tasks. This

task was originally made up of 40 sentences which comprised grammatically correct

and incorrect sentences. The number was reduced to 37 after it was piloted on a group

of native speakers (see section 3.4.1). Most of the incorrect instances were taken from

student essays in the college writing courses. In the prompts, participants were asked

to judge each underlined NP16 as either ‘correct’ (√) or ‘wrong’ (X) in the target

language.

The NPs chosen in the design of T3 covered the basic variables that determine

article choice, namely, definiteness, countability, number. In addition, to meet the data

requirements, both specific and non-specific references were included, adjectivally

premodified and non-premodified constructions as well as abstract and concrete

15
A full discussion and review of the design features of grammaticality judgement tests can be found in
Mandell (1999) and Ellis (1991).
16
The instruction to judge underlined NPs rather than the full sentences was done to emphasise the
investigated head noun/phrase, thus avoiding possible confusion with other nouns in the same sentence

71
contexts were incorporated. This resulted in multiple combinations of variables. For

example, the underlined NP in (2) a wish was categorised as

–Def +Count +Singular +Spec –Premodified –Concrete:

(2) I made a wish and it became true.

An attempt was made to ensure reasonable representation of these variables in the test

which would result in multiple combinations. The number of various constructions is

adequate for the purposes of the test, but having an equal number of multiple

combinations would result in a number too large for the scope of this study.17 Instead,

an approximate number of contrastive values for each variable were selected in order

to provide a balanced representation of selected items. For example, out of 36

countable nouns, 17 were singular while 19 were plural. Similarly, out of 22 specific

references, 9 were definite while the others were indefinites. There were 9 indefinite

singular nouns, 3 of which were adjectivally premodified.

In addition to a methodical combination of the major variables that determine

article use, on a micro-level, definite descriptions were designed to cover the most

typical definite environments (see section 2.2.1). To avoid ambiguity, the indications

to the necessity of supplying the definite article were emphasised by, for example,

adding an identifying relative pronoun in the immediately following context

(examples 3, 5) or using superlatives (example 4).18

17
It would be practically unfeasible if there was an equal representation of all the criteria in multiple
combinations of variables, the total number of instances would be the result of 6 NP descriptors and
their negative counterparts (12) in and correct and erroneous instances of both articles (omission,
overuse, replacement) in addition to the zero article. The total number would at least be 14 to the power
of 14.
18
The examples include both correct and incorrect instances.

72
(3) *I bought a same cheap computer that my friend has. (semantically unique)

(4) Emirates mall is the biggest mall in the UAE (superlative)

(5) I found the book that you gave me very interesting (post-

modification/cataphoric reference)

(6) *Aisha went to a wedding. She said a bride was beautiful (bridging/associated

reference)

Furthermore, with contexts limited to sentential scope in T3, the referents were in

close proximity to the syntactic position of the articles which, in effect, could have

facilitated access to the available information (see Ariel 1988) and emphasised the

necessity of the presence of articles. The above examples include two erroneous

instances in a randomly ordered list of correct and incorrect NPs. Most incorrect

examples in T3 were selected from students’ essays in the Writing course. For

example, a couple of sentences in the test show the overuse of the indefinite article

without a head noun. The erroneous sentences were taken from students’ essays in the

Writing course at UAE University. The structure [a/n + adjective + missing noun] is

exemplified in (7) and (8) from T3.

(7) *Fatima is hardworking because she wants to be a successful in her job.

(8) *My ambition is to become a fluent in English

To investigate learners’ ability to recognise the ungrammatical overuse of the

indefinite article, it was supplied with uncountable or plural nouns (e.g. a gold, and a

dresses) as well as with nouns of contrastive L1/L2 countability status, such as advice

and information which are countable in Arabic. Similarly, awareness of definite article

73
omissions was examined via instances where the structural position of definiteness in

genitive constructions is different in the two languages, such as is shown in the

following sentences:

(9) *Damascus is capital of Syria

(10) *Price of oil has gone up

In addition to errors of omission, overuse instances of the definite article were

included to test learners’ ability to detect a more subtle form of misuse. For example,

the noun milk in ‘the milk is necessary for babies’ is expected to be marked incorrect

if interpreted generically as it fails to denote kind, i.e., all the milk in the world.

Rather, a more preferable, target-like, option would be a bare nominal, as no article

(zero Ø) is required to deliver similar meanings. However, the same NP could be more

acceptable if interpreted within a different context, when, for example, the speaker is

referring to a particular bottle of milk. The same is true for the NP in example (6)

Aisha went to a wedding. She said a bride was beautiful if the utterance referred to an

occasion where many couples are getting married and one bride was beautiful. In this

sense, article use becomes a matter of choice that depends largely on the particulars of

certain contexts, relevance to other pieces of information in the setting, the speaker’s

intent and the listener’s interpretation of the message. If this further information

relating the particulars of each setting were to be provided, extra text will be required

to clarify various interpretations which allow degrees of acceptability. This in turn

will change the format of this test to content-based, similar to the other two tests,

instead of being decontextualised and discrete. A cross task comparison of

performance under different conditions would thus not be possible. Moreover, since

74
the items that remained in the test after the pilot were only those with 100%

agreement by the CG, it was necessary to find out whether learners were able to infer

the same prototypical meaning of each item as understood by the NS. In other words,

students were forced to process the limited information available to them within each

discrete sentence, and base their judgements on the face value of each utterance.

Therefore, no more than one answer was accepted for each item.

In designing T3, we implemented maximum control in the sense that the

participants had no part in the input, were forced to make conscious decisions with

respect to constructions they might have otherwise avoided and within a context

where information cannot be inferred as it does not extend further than the sentential

level. Generally, GJTs force learners to compare existing structures and forms to rules

they learned (Richards 1976, Gass 1996, Alderson 1997) through periods of EFL

instruction. Thus, in all likelihood, such tasks reveal more of learners’ explicit

understanding about language (Robinson 1996, Trenkic´ 2000) than how the L2 might

be produced for communicative purposes. i.e., it is a different cognitive demand that

this type of test places on learner production.

The focus on form and accuracy rather than communication in GJT (cf.

Granfeldt 2000) has been considered a shortcoming by many researchers (Bialystok

1982, Birdsong 1989, Ellis 1991, Norris and Ortega, 2000). In addition, the reliability

of the results incurred from sentence-based, decontextualised tests is often questioned

(Gass 1994, Yule 1998). However, GJTs have been defended as being one of the best

methods of measuring L2 competence without distractive elements (Ellis 2004,

Mandell 1999). Besides, these tests can yield results that help the researcher study

learners’ purely receptive type of knowledge, produced with the least processing

effort. Finally, despite their limitations, form- focused tests offer an alternative

75
perspective to results emerging from communicative tests so that outcomes can be

compared across the two types of tests, providing a more comprehensive

understanding of learners’ IL. Researchers such as Bialystok (1979), Sorace (1985)

and Alderson (1997) used GJT which required identification of the errors as well as

corrections and stating of the violated grammatical rules. One of the most popular and

frequently cited studies on SLA of articles is a paper by Butler (2002) where he

depended on a GJT for his data collection from Japanese learners of English.

The argument between the proponents and critics of GJTs is not entirely

polarised. Some SLA researchers have suggested that there could be a more balanced

stand. Sorace (1996), for instance, proposed that when learners are asked to judge the

grammaticality of an utterance, it is more likely that they would rely on their

subconscious knowledge, but if more time was available to them, learners would be

able to access their explicit knowledge more easily. Therefore, rather than dispensing

with tests of this kind as entirely unreliable, GJTs can certainly serve as rough

indicators of learners’ knowledge and can, under certain conditions, reflect a part of

learners implicit understanding of the L2 (see Roehr 2004). Hence, it becomes

possible to assume that the results this test yields are valid and reliable, especially if

some were compatible with those emerging from more communicative tests (such as

T1) and more content-based tests (such as T2). In other words, results from GJTs,

cautiously interpreted, could serve as supplementary evidence to consolidate findings

deduced via other means.

76
3.4 Administrating Tests

3.4.1 Pilot

Both cloze test and GJT were first answered by native speakers. The responses were

collected, entered and analysed to be used as key in deciding correct/appropriate

article supply in the learner data collected from T2 and T3. One blank in T2 had to be

cancelled because the three choices mentioned in the prompts (a/n, the, zero) were all

provided for the same blank by various NS participants (see 3.5.4.2). Three sentences

were eliminated from T3 based on feedback from the CG who thought that

definiteness in those instances was ambiguous because of the possibility of multiple

interpretations of the same utterance. Only the undisputed NPs with 100% agreement

by NS remained in the test. This action brought down the total number of questions

the GJT to 37.

We piloted the tests with nine students, assigning three participants to each

level group of lower, intermediate and advanced proficiency. The tests were run and

the information was entered, categorised and analysed. Based on the results, it was

decided to keep T1 and T2 as originally designed, and to change the prompts in T3 by

removing the request to justify the responses. This was done because the explanations

provided in the pilot were largely subjective and proved difficult to analyse

statistically.

3.4.2 OPT

A total of 80 participants signed the consent forms and answered the background

uniformity survey. 21 students were eliminated because their answers did not meet our

requirements. For example, a few students had studied in English speaking countries.
77
The remaining 59 students took the Quick Oxford Placement Test (2001). Observed

by teachers, the potential participants were given 50 minutes, the duration of one class

session, to answer the questions in the test.

After the tests were marked, the learners were divided into three major groups

according to their scores in the OPT. The system we applied for banding was not

exactly identical to the one in the OPT descriptions but was more suitable to our

needs. This was because it was impossible to find Beginners (0-10) or Breakthrough

participants (scoring below 20/60) since passing the IELTS entry exam with a

minimum band of 4.5 (equivalent to lower intermediate level) is a prerequisite to

entering the UAE university. Similarly, it was impossible to find Very Advanced

participants (55/60), as the highest score was 53/60. This would have placed all our

students in the intermediate group. Instead, participants with OPT scores of

Elementary level (30 out of 60) were placed into the weakest group (G1). G2 included

participants whose scores ranged between 33 and 43. This range covers OPT Lower

Intermediate and average Intermediate levels. Finally, the highest group (G3) included

students with 45 points and above, covering both Upper Intermediate and Advanced

learners by the OPT definitions.

To further enhance the distinction between the three sections and to facilitate

ensuing cross-group analysis of the data, we increased the gaps between the PLs. Ten

subjects were eliminated from the tests for scoring totals of 31, 32 and 44 in the OPT

because they were considered borderline marks between the three groups. Thus the

difference between the three groups became more distinct. The number of participants

in each group averaged 19-20 students.

78
To ensure the consistency of responses at each level, an SD test was run. It

was found that the data were normally distributed within groups and significantly

different when compared across groups (see chapter 4).

3.4.3 Tasks

T1 was administered before the other tasks to keep learners unaware of its purpose

and their production least influenced by the explicit instructions in T2 and T3. The

amount of time allocated to perform this task was a maximum of 50 minutes under the

supervision of teachers. With time pressure, participants had little chance to revise

their writing and reflect on accuracy. The pressure was intended to the increase the

opportunity of communicative production (see Sorace 1996 for time effects).

After having completed T1 without prior knowledge of the purpose of the test,

learners were given/emailed the stories test with clear instructions to provide articles

in the slots. Thus, they became aware of what they were being tested on. The

participants were allowed to return/email their completed sheets of T2 within a

maximum of one week. The reason for having plenty of time allocated for responding

to this test was to induce learners to carefully consider the questions and give them

more chance to focus on form and accuracy. This time should enable students to take

into account the consciously learned rules of L2 while revising their answers. Thus,

the probability of spontaneous and immediate response was minimised. Accordingly,

the results should better reflect learners’ metalinguistic knowledge about language

than their competence or implicit knowledge of the L2.

After answering the stories task, the learners were asked to take the

grammaticality judgement test (T3). Most responses were completed immediately as

79
little effort was required in marking NPs as right or wrong. However, as in the case of

the administration of T2, the learners were given plenty of time to answer the GJT to

ensure that the extra time provide the opportunity to reflect on their original responses

and make the necessary changes before handing back the question sheets.

The disparity between the time pressure in administering T1 and the more

flexible time limits of the blanks test and the GJT was aimed at controlling the type of

information each type of task conveys and to set the two types of knowledge further

apart causing the division between the two test types more distinctly dichotomous for

data analysis purposes. As the production in T1 was immediate while learners were

under time pressure to communicate their ideas, little attention was paid to form and

accuracy, as rapid judgements are known to reflect implicit knowledge. With extra

time allowed for responding to the cloze test and the GJT, the results should reflect

learners’ explicit knowledge since they had the opportunity to compare the NPs in the

tests and the formally learned rules of the L2. Robinson (1996), for example, allowed

participants to make the judgements in their own time to elicit data that reflects

conscious understanding of the L2 grammar.

Unfortunately, because of time limitation and practical restrictions, we were

unable to re-run the tests with the stories (T2) and the sentences (GJT) presented in a

different order. Doing so might have yielded dissimilar results (see Cowart 1997) to

the ones we obtained.

3.5 Data Analysis

Describing the data collected from the tests tasks and the three groups necessitates

categorising the data entries, which include both NP descriptions and the analysis of

80
article use. The NPs occurring in the tasks were first categorised according to the

criteria that determine article use, i.e. their definiteness, countability and number

(Quirk et al 1991, Celce-Murcia and Larsen-Freeman 1999). Hence the NPs in the

task design and data analysis were described as possessing or lacking these features:

a. Definiteness (±def)

b. Countability (±Count)

c. Number (±Singular)

In addition to the above descriptions, to find out if article use is affected by other

linguistic notions, the NPs were described in terms of specificity of reference,

modification and ontological status.

d. Referentiality (±Specific)

e. Adjectival premodification (±Premodified)

f. Ontological status (±Concrete)

The second type of analysis involves article use. The main division was into either

correct or incorrect. Erroneous article use was further subdivided into errors of

omission, overuse and replacement (substitution). The NPs in 11 and 12 are examples

of omission and overuse errors (respectively) taken from students’ essays submitted

for the writing task.19

19
Each participant was given a sequence number and a code for proficiency level (A, B and C)
followed by a letter to refer to the type of task taken while NPs were numbered by order of appearance
in the essay. For example, 14CW4 refers to student number 14 whose PL is C (highest group), W is the
sign of the writing task and 4 is the number of the NP in one essay.

81
(11) *It’s quite big town (14cw4)

(12) *He told us an interesting stories (1aw22)

According to the above categories, the NP in the first sentence was described as [–def]

[+Count] and [+Sing] while in the article use section it was categorised as incorrect

and the type of error was that of omission. The NP in the second sentence was [–def]

[+Count] [–Sing]. The article use was incorrect and the type of error was that of

overuse.

However, classifying article errors is not always as straight forward as the

above examples suggest. Responses from the native speakers in T2 for example,

varied in some contexts although no grammatical errors were made. This case calls for

introducing a third type of analysis which does not solely depend on grammaticality,

but rather on how favourable or frequent one choice is. A detailed account of what

constitutes choice errors is explained in sections (3.5.3).

Results emerging from all tests were calculated in percentages and compared

across PL groups and test types to find out if there is any statistically significant

difference in certain error types or accurate/appropriate rates of either article. The aim

of comparing results across groups was to follow learner progress with regards to

article use, while cross-test comparison should help reveal the effect of task type

variation on outcomes.

Finally, there is a need to define what constitutes accuracy, what makes one

use correct and the other incorrect, and to clearly distinguish between various error

types in order to establish a rationale for considering one response more acceptable

82
than the other. The following section will provide a detailed account of the

parameters used to determine the appropriateness of articles used in each NP context.

3.5.1 Accuracy

In the data analysis, correct article use includes the grammatical and appropriate use

of the, a, an and correct unmarking of indefinite plural and uncountable nouns (zero

article). To find out the rates of correct or target-like usage (TLU), accuracy rates

were calculated by dividing the number of correct occurrences supplied in obligatory

contexts (SOC) relative to the sum of all the NP environments in which articles should

have appeared multiplied by the number of participants in one group.

Number of observed occurrences (SOC)

TLU =

Total number of obligatory contexts X number of students

The calculation above was used to measure the degree of accuracy for both articles as

well as correct marking of indefinite uncountable and plural nouns (accurate zero) in

different NP environments. Since the number of obligatory contexts varied across

tasks, outcomes of the above calculations had to be measured in percentages to allow

comparisons across tasks and across groups of varying numbers of participants.

Correct article supply primarily implies accurate use which largely refers to

grammatical correctness, but it also includes appropriate or suitable responses. While

the former entails rules without which an expression would be totally unacceptable,

83
such as the use of the indefinite article with a plural noun, the latter relates more to

target-like norms and common preferences with varying degrees of suitability, such as

the overuse of the definite article in certain contexts. The distinction between the two

uses is further described in section 3.5.3. Details relating to minor discrepancies in

analysing types of correctness in each task are described in section (3.5.6).

3.5.2 Error types

The second perspective in data analysis is concerned with error types. It has been

customary in SLA studies to examine learner errors with the aim of establishing areas

of difficulties, IL tendencies and find possible indicators to the stages of acquisition.

The development patterns which emerge from these results should help the researcher

to follow the progression of learners’ IL. Cross-group comparisons of error rates can

reflect the shifts in learners’ performance as their proficiency level improves while

cross-task analysis of the same error type can show the effect of task variation on the

development map. Article errors can be divided into three main categories: omission,

overuse (or commission Hakuta 1976:340) and replacement errors.

84
3.5.2.1 Overuse

An error of overuse refers to instances where articles should not have appeared (Pica

1983). In principle, the formula used in calculating errors was similar to the one used

for accuracy, i.e., the observed instances were compared against the total number of

contexts where such occurrences were expected to appear. For example, to calculate

the percentage of overuse errors of a(n) with uncountable nouns in one group, the

following equation was used:

Number of incorrect supply of a(n)


Overuse of a = %
Total number of [–Def –Count] NPs X Number of students

To apply the above formula, NP types and article use were described and summed. A

similar method was used to examine the overuse of the indefinite article with plural

nouns simply by changing the [–count] contexts into [–Sing] ones. The overuse rates

of the definite article were calculated by dividing the total number of overuse

instances in learner data by the total number of indefinite NPs and multiplied by the

number of students in one PL group.

3.5.2.2 Omission

The ‘Omission’ of articles is an inclusive term widely used in SLA to refer to errors

resulting from the failure to supply either article in obligatory contexts (T1 and T2) or

not being able to recognise that an article is missing (in T3). If the definite article was

not supplied in uniquely identifiable contexts or a/n was not placed before indefinite

countable singular nouns, an error of omission is thereby committed.

85
To calculate omission rates numerically, the sum of omission instances in

the responses of each PL group was divided by the total number of obligatory contexts

relevant to each article in each task multiplied by the number of participants in each

group. The figure resulting from such formula was then converted into a percentage to

facilitate cross group and cross task comparisons,20 as was the case with overuse

errors explained above. For example, to calculate the indefinite article omissions made

by the weaker group in task 2, we added up the omission instances, then divided them

by the total number of [–Def] [+Count] [+ Sing] NPs in that test, multiplied by the

number of students in G1.

3.5.2.3 Replacement errors

Replacement (substitution) errors refer to the employment of one article instead of the

other. In other words, to substitute a-for-the or the-for-a, but such errors do not

include a/the for zero or vice-versa.

Given the fact that countability and number are decisive variables in the use of

the indefinite article but not the definite, replacement errors had to be calculated in a

manner that would make the two NP contexts more comparable. Hence, only

+Singular +Countable nouns were selected to examine replacement errors while

±Definiteness value remained as the only common denominator that determines the

use of either article. Practically, countability and number became constants while

(in)definiteness was considered a dependent variable. Replacement errors were

calculated by applying the following formula to database entries:

20
This was done because the number of obligatory contexts in each task was not identical.

86
Total number of erroneous use of a
a-for-the = %
Total number of [+def] NPs X no. of participants in one group

This method differs from the one used for calculating overuse errors in that it does not

include the overuse of the indefinite article in [–def] [–Count] or [–Sing] NPs. A

similar calculation was used to examine the error of replacing the indefinite article

with the definite. The latter type of error is expected to be more frequent than the for a

errors because of the influence of learners’ L1.

3.5.3 Article Choice

The types of article use described in the previous section were further subdivided

according to whether the use is the native speaker’s first choice or not. This preference

might vary from one community to the other according to well-known social (non-

linguistic) factors such as region, gender and age (cf. Hollmann and Siewierska (2011)

for the suggestion that in definite article reduction in Lancashire dialect identity may

play a role as well). Therefore what might be considered as the NS’s favourable use in

one community might not be the same option in another.

In addition to being grammatically correct or incorrect, article use was also

described according to NS preference or choice determined by custom and

convention. In other words, an expression might be grammatically correct but not

native-like and can be therefore considered as an IL error of choice. Grammatically

inaccurate descriptions largely involve violations of the rules governing the use of the

indefinite article which can be summarised into the following points:

87
a. Using the indefinite article with indefinite plural nouns.

b. Using the indefinite article with indefinite uncountable nouns.

c. Failure to supply the indefinite article with indefinite singular nouns.

The types of misuse listed above include errors of overuse (a) and (b), and omission

(c). On the other hand, instances of grammatically incorrect the, are rarer because

definiteness depends mostly on the semantic value of an utterance (see Literature

Review 2.1.1). In other words, people’s assumptions about the availability of a

mutually manifest P-set in which the referent would exist and is uniquely identifiable

are not limited to one interpretation, therefore, in the one context, with the same

distribution of information, there can be more than one possible answer. This

necessitated finding a new category to include such use. In this study, following

Trenkic´ (2000), this type of use is termed as choice.

The erroneous choice of articles is a common problem in L2 English (Huebner

1983; Tarone 1985; Parrish 1987; Tarone and Parrish 1988; Thomas 1989; Master

1990; Trenkic´ 2002; Leung 2001; Ionin and Wexler 2003; Ionin et al. 2004).

However, this type of error is more difficult to determine than grammatical errors

since it is usually caused by varying perceptions of definiteness, a semantic quality,

rather than the more concrete values of countability and number.

SLA studies report that when choice errors occur in learners’ data, they are not

random (Enç 1991) but rather caused by the lexical meanings attributed to forms (for a

discussion of form-meaning connections see VanPatten et al. 2004). For example, if

learners associate the with the notion of specificity, they would overuse the definite

article in [–def +Sp] contexts (IKW 2003), possibly without affecting the

grammaticality of the utterance. However, such non target-like use could lead to

88
ambiguity and misunderstanding, prompting the listener/reader to ask for further

clarification as in this hypothetical dialogue:

(13) a. I bought the bag.

b. Which bag?

a. The one we both saw at the shops yesterday!

The speaker in (13a) assumes that the listener (13b) can uniquely identify the referent

in a limited context; whereas it seems that the referent is not known, judging by the

response in (13b). This ambiguity could have been avoided if speaker A used the

indefinite article. It is also clear from the example above that choice errors can be

made by some NS.

There were difficulties in determining the criteria that make the use of one

article less favourable than the other. In other words, there was a need to define what

constitutes a choice error. One suggested parameter is the frequency of use (see Ellis,

N 2002). The reoccurrence of a particular noun, or what Huebner (1985), following

Givón (1983), terms as persistence in discourse can be an indicator of the more

popular forms in certain contexts. For example, the indefinite plural/uncountable form

(zero article) is preferable in expressing generic meanings to the use of the definite or

indefinite articles (Langacker 1991, Onishi and Murphy 2002, Behrens 2005).

Another way to measure suitability or appropriateness is to examine the

contextual effect (Sperber and Wilson 1986) of relevance which influences the degree

of importance (Huebner 1985). The distance between referent and article calculated by

the number of clauses to the left of the noun can also determine whether a pronoun or

the repetition of the head noun would be more preferable than the use of an article.

89
Ariel (1988) found that other definite descriptions are preferred to the if the distance

to the referent was shorter. For example, pronouns are preferred for second mention

expressions within the same sentence, while demonstratives are more likely to appear

within the same paragraph. The definite article is typically used for less-accessible

referents within a paragraph or across paragraphs (ibid).21 In short, despite the

grammatical correctness of various structures that express similar meanings, there is a

stereotypical use that is preferred in natural dialogue and authentic text. Accordingly,

in this study, an error of choice in article use was set according to the following

factors:

a. The NP does not violate any grammatical rules of countability and number.

b. It is not the most widely used form by NS to deliver the same meaning in

similar contexts.

c. Other expressions might be more favourable under certain structural settings.

Hence, erroneous article use was categorised as either grammatically incorrect or

permissible but as an error of choice. For example, the overuse of the indefinite article

can be a grammatical error in some contexts and a choice error in others.

a. If ‘a/n’ is used with uncountable or plural nouns then it falls under

grammatical error type

b. If ‘a/n’ replaces ‘the’ with singular countable nouns, then it is an error of

choice.

21
Ariel 1990, 2001 used the term recency to describe the relevant distance that determines the choice
definite descriptions.

90
In this sense all replacement errors in this study are choice errors since only singular

countable NPs were selected to measure both types of replacement, i.e., only

definiteness remained a variable. Cases where the definite article replaces the

indefinite with newly introduced information, as is used in fiction, are not perceived

as wrong or non-native-like because these do not violate rules of countability and

number. However, such use makes the retrieval of the explanatory signals less

felicitous as the reader continues to search for relevant cues to identify the referent,

cataphorically.

3.5.4 L1 and IL

As we aim to analyse learner data qualitatively as well as quantitatively, it was

necessary to address the source of errors in L2 production. Most SLA researchers

agree that errors committed by L2 learners are usually ascribed to either transfer from

learners’ first language (L1) or to known norms of behaviour during the various stages

of L2 acquisition, better known as Interlanguage, regardless of their mother tongue.

Although error types overlap in many ways and are sometimes difficult to

discriminate between, what distinguishes those attributable to L1 influence from those

that result from common L2 processes is that the first type might be traceable to the

rules, structures and morphology of the learners’ L1 while the latter can be found in

the production of 2LL with various L1s.

Recall from section 2.3 that English and Arabic are not analogous in their

representation of indefiniteness, as English possesses an obligatory and explicit

marker of indefiniteness, while the indefinite marker in Arabic is covert and optional.

The two languages also differ in the function of the definite article because the non-

specific reference is always definite in Arabic. These discrepancies, in addition to the

differences mentioned in section 2.3.3, might lead to negative transfer to the TL. For

91
example, the absence of an indefinite article in Arabic can cause higher omission rates

than those of the definite. Furthermore, dissimilar lexical settings that underlie the use

of the definite article might cause overuse in generic references. However, despite the

fact that these errors can be attributed to transfer from Arabic, they can also be a part

of a more common pattern found in IL, regardless of the L1.

In order to identify possible sources of errors, certain features that affect

article use, as well as other features under investigation, were compared between the

two languages. If the error rates in instances where there is a divergence between the

two languages proved to be significantly higher than their correct counterparts, then

this could strongly suggest the presence of cross linguistic influence. Conversely,

similarities with tendencies and error patterns recorded in findings of studies on other

L1s would point towards common tendencies which are part of the L2 acquisition

process. For example, if higher error rates were recorded in contexts where there is a

disparity in L1/L2 countability status, it is probable that there is a certain degree of L1

transfer, since dissimilarity in countability notions between the first and the target

language is known to be one of the major causes for article errors (Yamada and

Matsura 1982, Yoon 1993, Butler 2002, Al-Fotih 2003, Hua and Lee 2005). On the

other hand, if errors were found to be higher in all uncountable than in plural contexts,

and similar findings were observed in other studies on L2 acquisition of articles, this

could provide grounds for supposing that L2 learners find it more difficult to

recognise the irregular morphological structure of uncountable nouns (being without

s) than plurals, regardless of the L1.

Table (3) demonstrates possible article errors resulting from transferring L1

patterns onto the TL in NP environments of contrastive L1/L2 natures, with examples

92
from learners’ production.22 The linguistic variables include those that determine

article use, in addition to other notions that learners might wrongly associate with

article use.

22
All examples are from T3

93
Table 3 Grammatical versus choice types of errors

Linguistic Evidence of negative L1


Example
features transfer

Using a(n) with uncountable/


My brother did a good
Countability plural Ns that are countable in
housework.
Arabic.

Overusing the in indefinite non- Reading the books fills


Specificity
referential contexts people’s time.

Overuse of the definite article The most important thing is


Ontology
with abstract Ns the safety.

Omission of obligatory
Fatma became successful
Syntax indefinite article with
business woman.
premodified NPs

Using the in the complement

Structure position of partitive ‘of’ Price of the oil.

constructions

3.5.5 Learner Hypotheses

In this study, the theories underlying L2 use of articles by L1Ar learners were

examined from three different angles: The pragmatic notion of specificity; the

structural aspect of adjectival premodification, and the semantic value of the

abstractness/concreteness of the head noun. The effects of these three different factors

on article use were studied to help gain some insights into learners’ internal

94
hypotheses about the L2, i.e., whether or not faulty associations exist between

definiteness and these notions (see chapter 2 section 2.3). The findings were compared

to those from other studies concerned with the effect of similar concepts on L2 article

production. The following sections will explain how the presence of such associations

was observed and calculated in our database.

3.5.5.1 Specificity

To investigate the relation between specificity and definiteness in learner hypotheses

(IKW 2004), instances in which the definite article was overused in [-def] [+SP]

contexts were compared against those made in [-def] [–Sp] references. If systematic

oversupply in either (indefinite) context was found, it would point towards a faulty IL

connection between the interpretation of definiteness and the pragmatically-

determined referentiality status of the NP. To further validate our findings, counter-

hypothesis tests were run (see Skehan 1989, 1998) to find out whether a parallel

connection exists between the indefinite article and non-specificity.

95
3.5.5.2 Premodification

Previous research (Parodi et al. 1997, White 2003, Goad and White 2004) has

indicated that the prosodic structure in learners’ L1 negatively influences their L2

production. In this study, this claim was investigated by examining article use in

adjectivally premodified structures in Arab learners’ production. Omission patterns of

the indefinite article in [+Premod] NPs were observed and contrasted against the rates

in [–Premod] contexts. If the results yielded significant differences between the two

environments, it would be possible to postulate that the position of the noun in the NP

affects the L2 supply of articles by Arab learners. Counter tests were performed to

examine learners’ predisposition to overuse of the definite article in similar NP

contexts.

3.5.5.3 Concreteness

The third aspect of learner hypotheses with regards to L2 article use examined in this

study was the IL association between definiteness and abstractness/concreteness. In

order to achieve this, error rates were contrasted in both lexical contexts (cf. Trenkic´

2000). Critically different results would mean that learners’ choices of articles could

be influenced by the ontological status of the head noun, possibly originating from the

transfer of L1 lexical notions (Hamdallah 1988, Diab 1996, Maalej 2004). If this was

the case, higher overuse occurrences of the definite article would be expected to

appear in abstract NP environments. Accordingly, definiteness could therefore be

more associated with abstractness, rather than concreteness. As a confirmatory

measure, we also tried to find out if the L2 meanings of the indefinite article were

associated with abstract or concrete concepts.

96
3.5.6 Task variation

Every NP produced by learners (in the free production test) or prepared by the

examiner (in the cloze and GJ tests) was analysed according to the following criteria:

identifiability (definite/indefinite), countability (countable/uncountable), and number

(singular/plural). Entries were also categorised in terms of premodification, specificity

and concreteness.

NP values were considered as constants; whereas the Article Use section

represented the variables examined. L2 use of a/n, the and zero was analysed with

respect to accuracy and error (overuse; omission; replacement). A further subdivision

divided article use into grammatically incorrect and choice errors. This table illustrates

how article use was determined by matching the type of article usage, be it correct or

incorrect, with certain criteria in the NP description accordingly filtered from the

database.

97
Table 4 Article use calculation based on NP description

Article Use Noun Phrase description

Correct the Def=1

Correct a/n [Def=0] + [Count=1] + [Sing=1]

Correct 0 [uncountable] [Def=0] + [Count=0]

Correct 0 [plural] [Def=0] + [count=1] + [Sing=0]

Omission ‘the’ Def=1

Omission a/n [Def=0] + [Sing=1]

Overuse a/n [–count] Count=0

Overuse a/n Plural [Count=1] + [Singular=0]

Overuse ‘the’ Def=0

a/n for the [Def=1] + [Singular=1]

The for a/n [Def=0] + [Singular =1]

Overuse the [–Sp] [Def=0] + [Specific=0]

Omission a/n premod [Def=0] + [Sing=1] + [Premod=1]

Overuse the abstract [Def=0] + [Concrete=0]

98
In all three tests, we calculated article use based on the description of NPs as filtered

in the Excel data sheets. However, because these tasks vary in length, focus and

degree of control, there were some minor differences in the methods of data analysis

for each test. For example, the overuse of the definite article was considered an error

of choice in T1 because its supply in indefinite contexts, although not target-like, does

not violate any grammatical rules. On the other hand, the overuse of the in T2 was

regarded as a grammatical error in instances where the answers did not match any of

the NS responses in a particular blank, and a choice error in cases that matched the

exceptional few from the control group. Finally, due to the nature of T3, the inability

to recognise overuse instances as incorrect was considered a grammatical error of

judgement. Further details of task-specific data analysis are discussed in the following

sections.

3.5.6.1 T1

When data from this test was analysed, only instances where the presence of the

definite article is deemed necessary were categorised under [+def] (please refer to

obligatory contexts in the Literature Review 2.2.1). This renders the definite article as

the sole representative of definiteness while other semantically definite descriptions,

such as proper nouns and pronouns were not coded as [+def]. The reason is that if

such NPs were described as [+def], the supply of the definite article in electronic data

would be automatically required once the NP description is compared against article

use, leading to confusion in subsequent calculations.

Expressions which exclude articles were exempted from being analysed in the

database. They include:

99
a. Quantifiers: many, lots of, much

b. Determiners: every, any, each, no, none, another

c. Demonstratives: this, that , those, these

d. Possessive pronouns

e. The numeral one.

However, the determiner some was accepted as a correct indefinite plural marker in

this test only.

Other expressions that were not categorised in the descriptions were those

without head nouns. In these cases, article use was described and calculated, but it was

practically impossible to analyse the NPs, which could include both correct as well as

incorrect instances of oversupply. These examples from participants illustrate how

such occurrences can be wrong in one context but completely acceptable in another:

(14) They consider him the best. (Correct the).

(15) *He wanted to be a successful. (Incorrect a).

In both correct and incorrect instances, only article type of use was entered.23

There were cases in this task in which head nouns were preceded by Arabic

proper nouns that have the definite article (al) as part of the name, (e.g. Alain

museum). Such instances might lead to the omission of the obligatory the because

students rely on the definiteness of the Arabic counterpart. For example, many

students preferred to use ‘Alain city’ instead of the genitive structure (The city of

Alain) which led to dropping the English definite article. In both cases ‘city’ bears a

23
Cases without NP descriptors were not compared against article use in data calculations. Instead,
article use in such instances was counted separately.

100
definite value either grammatically through the morphological presence of the Arabic

definite article (al), or semantically as a part of a proper noun. However, the problem

becomes more complicated when the NP structurally requires grammatical

definiteness as in the following example:

(16) Alain museum is interesting to new comers in Alain. Part of (the) Alain

museum shows traditional weapons.

Practically, we depended on capitalisation or quotation marks to distinguish between

what learners intended to be considered proper nouns and those that were treated as

common nouns.

(17) It was known as an oasis. common (Ref: 5BW)

(18) It means ‘oyster’. common (Ref: 10AW)

(19) The meaning of ‘Saad’ is big hole. proper (Ref: 8aw)24

In the article-use section of the data analysis, all of these instances (17, 18 and 19)

were categorised under ‘correct zero’.

Correct use of the definite article was decided according to the contexts of

obligatory supply (2.2.1) but if the deployment of the definite article was extended

beyond these contexts, such as to non-referential expressions,25 or was not

24
These are samples from students in groups 1 and 2.
25
According to published research, including corpus-based studies, bare plurals seem to best represent
non-referential expressions as the most widely applied form to deliver generic, kind-referring meanings,
as is observed by a number of theorists (e.g Chesterman 1991, Onishi and Murphy 2002, Partee 2005,
Behrens 2005) as well as SLA linguists (e.g. Huebner 1983, Trenkic´ 2000 and Thu 2005). Thus, it is
possible to consider that alternative options constitute errors of choice, although they are not incorrect.

101
appropriately supplied in obligatory contexts, the misuse was treated as a choice error

since, unlike the overuse and omission of the indefinite article, no grammar rule was

violated, rather, definite article errors are perceived as errors of meaning (cf. Trenkic´

2000).

The relationship between the definite article and referentiality has always been

complicated. It was sometimes difficult to justly categorise the specificity status of

references in learner data since the intention to refer is entirely subjectively

determined. In this test, the parameter used to distinguish ±specific contexts depended

on whether the expression referred to particular entities within the context of the

student’s essay, or to all similar entities of prototypical descriptions in the world (see

Abbott 2000, 2003). In this sense, we followed Salmon-Alt and Romary’s (2000)

approach whereby referentiality can be deduced from the domain in which the referent

operates.

Finally, unlike T2 and T3 where idiomatic and culturally-bound formulaic

expressions were deliberately avoided, the appearance of such NPs was inevitable in

the data obtained from T1. Each of these instances was judged individually according

to the context in which they occurred. The fact that more native-like structures and

expressions would have been preferred in certain instances was ignored. The only

variable considered and analysed was learners’ choice of articles. 26

For example, a student from the weak group wrote that her town did not have electric power supply.
Later, she said the town is naturally beautiful but We suffer from this lack in the service (Student
reference 2aw).
26
Other errors concerning parts of speech, vocabulary or structure which might have had an indirect effect on
learners’ choices of articles were not taken into consideration. Instead, NPs were analysed on article use only.

102
Learner data Native like choice

(20) As a conclusion in conclusion

(21) The houses of people People’s houses

3.5.6.2 T2

For the purpose of the analysis of the results, the 6 stories were treated as a single set

of data. The set contained a total of 217 full NPs, identified as target contexts. Each

target context was classified according to criteria that determine article use

(definiteness, countability, number), in addition to specificity, premodification and

concreteness. Correctness and suitability depended on the degree to which the native

speaker group provided consistent answers.

Accuracy and error were determined by NS responses. Most blanks in the

stories task were unanimously supplied with the same article by all NS participants.

Agreement on one article per blank (100% consistency) dictates the obligatory supply

of a similar response from learner data. Therefore, in such cases, there was only one

acceptable (correct) answer in learner data, which made the other two remaining

options impermissible in data analysis, although they would not necessarily be

considered grammatically incorrect per se.

In cases where responses from the CG were not consistent, the appropriateness

of learners’ answers was determined by the level of agreement among the native

speakers. If the majority of NS (more than 6 out of 12) agreed on one article for one

particular blank, the corresponding answer from learners was considered correct. On

the other hand, if the answers from the learner database matched the responses

provided by the minority of NS for a given blank (fewer than 6 out of 12), the answer

103
was categorised under choice error, leaving one remaining article which was

automatically excluded as an incorrect option.

Finally, if the CG participants produced all three possible answers (a/n, the, zero)

for the same blank, it was omitted from the test because the purpose of the question

was defeated as no incorrect option was left available. This case happened in one

blank in story 4 which led to its cancellation.

(22) The position of ______ top politician.

Variability in NS responses represents different individual readings of definiteness,

possibly because of textual distance between article and referent, recurrence in

discourse (persistence) (Huebner 1985) or the prominence of the referent in the

consciousness of the participants (Chafe 1987). Alternatively, disparate responses can

be the result of dissimilar perceptions of entities in the world, as in example below:

(23) _____window was open (story 2).

Most NS filled in the blank with the indefinite article, assuming that the average size

house has more than one window, with any one of them open while the few who

chose to use the instead, might have assumed the speaker to be looking at a certain

window (referent physically present) or the only window in the limited P-set of the

kitchen mentioned in the preceding discourse.

104
Some respondents from the CG asked if supplying possessive pronouns would be

more suitable in certain blanks, but they were instructed to limit their answers to the

options available in the prompts, namely a(n), the and zero.

3.5.6.3 T3

Since the description of the NPs was pre-designed in the controlled tasks, article use

in the GJT was directly analysed according to whether it corresponded to or differed

from the answer key decided by NS responses. Thus, accepting only one correct

answer implies that expressions that are most unlikely to be used by NS in natural

discourse, though grammatically correct, had to be regarded as incorrect. This means

that there were no choice errors in this test. The remaining techniques of analysis

followed in categorising data entries from T3 were congruous with the approach

followed in the other two tests (T1 and T2).

Unlike the grammatically incorrect instances where the indefinite article is

supplied before uncountable or plural nouns such as a + gold (6) and a + dresses (7),

it would be difficult to rule out the possibility of the definite article occurring with

mass and abstract nouns in certain occasions. However, since no choice-error options

were permitted in this test, non-native-like occurrences of the definite article had to be

marked wrong in the key. For example, the underlined nouns in (24) and (25) were

expected to be marked incorrect by the CG despite the fact that the constructions do

not violate any obvious grammatical rules. NSs did consider them erroneous because

the definite article is deemed redundant with abstract and mass nouns in most non-

referential readings.

105
(24) In choosing a flight, the most important thing is the safety.

(25) ‘The milk is necessary for babies’

As the occurrence of the definite article in a similar situation of example (24) is

implausible, learners were expected to reject this utterance for its improbability.

Similarly, the example in (25) is expected to be marked wrong if it was referring to a

kind of substance with indistinguishable universally-common attributes. In this case,

the zero article would be the more appropriate choice (Chesterman 1991, Behrens

2005, Onishi and Murphy 2002, Partee 2005). However, the same expressions can be

accepted if interpreted within a limited context in which the speaker is referring to a

certain safety measure or a particular bottle of milk which the listener might be able to

recognise.

In addition to the dichotomous categorisation of responses, another constraint

in analysing data from this task was the limited amount of context available for each

instance. The meaning of referents in discontinued expressions had to be drawn from

what the narrow sentential scope of the NP permitted the reader to understand. Thus,

without sufficient information available in the text, responses are expected to largely

reflect the students’ epistemic knowledge of the world. The lack of content restricts

the denotation of NPs in T3 to common generalisations and does not allow further

interpretations that could have been possible in alternative environments.

Recall example (6), given in the description of T3: Aisha went to a wedding.

She said a bride was beautiful. On impulse, the first reading establishes a logical

connection between the two sentences through ‘bridging’ or ‘associated anaphora’

106
which should lead to automatic rejection of the indefinite article as the reader

recognises that the definite article would have better expressed the thought. However,

if there were a context in which the bride was one of many, such as on occasions in

which several couples celebrate their weddings at the same place/time, the supply of

an indefinite singular form could have been feasible. Nevertheless, the expression

remains awkward, since, in a similar situation, a NS would have probably expressed

the same idea differently (e.g. One of the brides was beautiful; or There was a bride

who was beautiful). The lack of prior/further explanation does not encourage

presuming alternative interpretations beyond the typical usage (Chesterman 1991) in

the context of a wedding ceremony where there is usually one bride only. Therefore,

despite the limited context in this task, the conventional 27uniqueness (of one bride) in

a limited setting (of a wedding) helps make definiteness obligatory whereas choice to

use the indefinite article becomes less acceptable.

Another example from T3 illustrates how the use of the definite article could

have been permitted under different contextual conditions in which more detailed

information about the referent was available.

(26) When I needed the pen in the last exam, Fatma gave me hers.

If the NP in (26) was part of a story in which the pen had been mentioned in former

discourse with some distinctive quality, such as being the speaker’s favourite or has

been given to her as a present, the use of the definite article would have been

accepted. However, given the limitation of the individual sentence, the interpretation

would have to bear a [–def] value because the referent cannot be uniquely identified.
27
Definiteness in Conventional contexts is listed among Liu and Gleason’s (2002) list of non-generic
uses of the based on Hawkins’ (1987) uniqueness theory.

107
3.6 Limitations

Because of practical restrictions of time and space as well as the empirical nature of

this study that combines a multiplicity of proficiency levels, testing tools and criteria

investigated, certain theoretical and methodological elements could not be included.

This section points out these limitations and describes the alternative measures

followed in conducting the tests and analysing the results.

1. As was confirmed by the pre-testing data, all participants even of the lowest

PL, did produce articles, therefore emergence, as the primary stage of

production in SLA, was not included in this study.

2. In analysing the data, the two forms of the indefinite article (a, an) were treated

similarly in terms of obligatory suppliance or erroneous use (omission,

overuse). However, the replacement of one form with the other was simply

considered as an incorrect choice of article. This case happened only once with

one student in T2 and was therefore ignored.

3. Despite the possibility of countability shifts of nouns (from countable to

uncountable and vice versa) depending on NP context (Allan 1981), or the

notion that some nouns have a ‘dual class membership’ (Quirk et. al. 1985,

Celce-Murcia 1999), a basic taxonomy of countable/uncountable was applied

in the data analysis in order to facilitate NP description and categorisation of

108
article use in T2 and T3. Yet, the countability status of nouns in the free-

production task was decided on according to the context of each essay.

4. Although an attempt was made to provide equal instances of variable

combinations when designing the cloze test and the GJT, it was extremely

difficult to have exactly identical numbers of each variable as this would

require too large a number of examples for the scope of this study.

Nevertheless, we made every effort to provide a balanced representation of

contrasting variables.

5. The number of participants in each PL group was very close but not exactly

equal. There were participants who took one test but not the other. Therefore,

in addition to the numerical outcomes, the results were calculated in

percentages.

6. The expressions low, intermediate and advanced used to describe proficiency

level groups 1, 2 and 3 respectively were used according to the parameters set

for this study (see 3.4.2) although learners’ band of proficiency level might not

be termed as such under a different language testing apparatus.

7. Despite the subtle theoretical differences between subdivisions of specificity

readings such as wide/narrow-scope, or de re/de dicto (see Jaszczolt 1997), a

basic dichotomy of [–/+Sp] references was followed in categorising NPs to

facilitate data description and statistical analysis.

109
Chapter Four: Results

This chapter presents the results obtained from analysing the data elicited from the

three tests taken by participants of the three PL groups. This will include the

performance of each group in terms of accuracy and error rates. Learners’ hypotheses

are also examined to find out whether or not faulty associations exist between certain

linguistic features and article use, and if so, to what extent. The first three parts will

cover the results from each test separately; while a cross-task analysis of performance

rates will be reported in the fourth part.

In order to ascertain the validity of grouping the participants into three separate

sections according to proficiency level, entries from each student within a PL group

were calculated separately. Once the responses of each student were summed up

and measured against the key, they were compared to the totals of other participants

within the same group. The results showed the data to be normally distributed within

each group.

While the placement test was used to allocate participants into groups

according to their proficiency level, statistical analysis was performed to ensure that

the difference between the three groups was significant to justify the separate

categorisation. This was calculated using non-paired two-tailed t-test assuming equal

variance between the three groups with 95% confidence, comparing two groups at a

time. The higher means emphasise the better performance by the upper groups. Tables

5 and 6 below show the homogeneity of responses from participants within one group

on the one hand, and the statistically significant difference in the performances of

participants across the three groups (G1 v G2: p<0.0001, G1 v G3: p<0.0001 and G2 v

110
G3:p=0.0057) on the other. The figures reflect the results obtained from both T2 and

T3.

Table 5 Similarities within each PL and cross group differences in T2.

n M CI at 95% p

G1 15 124.07 ±7.89

G1 v G2

<0.0001

G1 v G3
G2 20 159.1 ±8.26
<0.0001

G2 v G3

0.0057

G3 16 177.57 ±8.57

Table 6 Similarities within each PL and cross group differences in T3

n M CI at 95% p

G1 15 23.13 ±1.73

G1 v G2

<0.0001

G1 v G3
G2 19 30.05 ±1.71
<0.0001

G2 v G3

0.0477

G3 18 32.35 ±1.26

111
In categorising data entries from the three tests, the criteria used for measuring

accuracy/error rates in learners’ production depended primarily on grammatical

correctness, and secondly on correctness of choice. Recall from Section 3.5.1 that

correct article use meant the use of the definite article with uniquely identifiable NPs

in obligatory contexts, the indefinite article with non-identifiable, countable, singular

NPs and no article (or some) with unidentifiable plural/uncountable NPs in non-

referential contexts.

The information is presented in percentages because the number of criteria

compared was not identical. For example, the number of obligatory definite contexts

is not exactly equal to that of obligatory indefinite contexts. This is especially true in

the free composition where it was up to the individual participant to decide upon the

form and type of each NP.

In addition to numerical and percentage-based calculations, the data was

statistically analysed. We performed a Z-test to find the equality of proportions across

groups, article use and tasks based on the null hypothesis.28 We tested our hypothesis

at 5% level of significance.

28
This is based on the supposition that there is no statistical difference between the proportions of any
two groups.

112
4.1. Task 1

The selected sentences from the free writing test contain 1800 NPs, excluding those

preceded by possessive S, possessive pronouns and most proper nouns (see

Methodology 3.5.6.1). The remaining number of selected NPs ranged between 35-45

NPs per student across groups, with G2 producing the highest number of NPs per

student.

The essays produced by the three PL groups were first broken down into

individual NPs and entered into Excel spreadsheets. Each NP was broken down

according to values of definiteness, countability, singularity, specificity, pre-

modification and concreteness. Next, the article use in each NP was classified as

either correct or incorrect. Incorrect article use is further dissected into omission,

overuse (grammatical) and choice errors. Table 7 provides an example of how NPs

and article use were categorised and analysed in the database.

113
Table 7 Datasheet sample from one of the participants in task (1).
Ref Article Use NP description
.
No. Student name The A 0

Overuse

Overuse

Concrete
Omission

Omission

Countable

Premod
Specific
Correct

Correct

Correct
Definite

Singular
11
aw

1 a special place to live 1 1 1 1 1


2 place to work 1 1 1 1
3 has relatives 1 1 1
4 I live in small town 1 1 1 1 1 1
5 the most beautiful town 1 1 1 1 1 1 1
6 in the world 1 1 1 1 1
It has many places which
7 1 1 1
attract tourists
8 from all over the world 1 1 1 1 1
9 It has mountains 1 1 1
beaches where you
0 1 1 1
breathe
11 fresh air 1 1 1
12 and farms where you find 1 1 1
13 different kinds of 1 1 1
14 fruit and 1 1
15 Vegetable 1 1 1
It has the most important
16 1 1 1 1 1 1
factor which are
17 Safety 1
18 Quietness 1
19 and purity 1
whenever I have a
20 1 1 1
problem
21 my place in the society 1
22 small simple houses 1 1 1 1
23 to villas 1 1 1
24 and from simple work 1 1
25 to complicated one 1 1 1 1
26 the person without 1 1 1 1 1
27 Past 1 1 1
28 will be without future. 1 1 1
29 in the same time 1 1 1 1 1 1

114
4.1.1 Accuracy

The results emerging from the weaker and intermediate groups show significantly

higher accuracy rates of the definite than the indefinite article (p=0.0079). G1,

employed the definite article correctly 150 times in 199 obligatory contexts (75%),

while the indefinite article a(n) was correctly supplied 64 times in 106 indefinite

singular contexts (59%). This suggests that L1Ar learners initially perform better on

the definite than the indefinite article. G2’s performance also showed significantly

higher accuracy rates of the definite than the indefinite article (p=0.0395). The

advanced group, however, performed better on the indefinite article, though not

significantly so. Cross group analysis reflects sustained improvement in performance

on both articles, as is illustrated in Table 8.

Table 8 Accuracy rates in Task 1 compared across groups

Correct Correct
n % p % p
the a/an

G1 19 150/199 75 G1 v G2 64/106 60 G1 v G2

0.0276 0.0277

G2 17 245/293 84 G2 v G3 86/116 73 G2 v G3

0.3742 0.0080

G3 12 191/221 86 G1 v G3 57/64 89 G1 v G3
0.0039 0.0003

It seems that learners gradually improve their performance on both articles starting

with higher rates on the definite than on the indefinite articles. This is evident in the

significant difference rates across groups. However, while the improvement on the

115
indefinite article was noticeable, the difference between the performance of G2 and

G3 on the definite article was not significant (p=0.3742).

In this task, the development pattern of the indefinite article diverges from the

definite. Although the two lower groups did not perform on the indefinite article as

well as they did on the definite, i.e., the accuracy rates of the definite article were

higher, the progress was more consistent correlating positively with PL. In other

words, there was less progression on the definite article between the intermediate and

the advanced groups.

With regards to indefinite contexts, NPs in the singular did not constitute more

than 25%-27% of learners’ production, whereas the use of bare plurals was more

preferable across the board averaging around 75% of all indefinite NPs. Participants in

G1 produced 106 indefinite singular NPs which counted for 28% of all indefinite NPs.

The remaining 72% of NPs were bare plurals. This choice is not random. G1

appropriately used bare nominals 226 out of 279 instances, which is 80% of the time.

The accuracy rate of the zero article improved to 85% in the production of the

intermediate group and to 91% in that of the advanced. Out of all correct indefinite

plural/uncountable nouns, accuracy rates were higher in indefinite plural NPs than in

uncountable ones all across PLs. The difference is noticeable in Table 9.

116
Table 9 Correct marking of indefinite plural and uncountable nouns in T1

Group n Plural % Uncountable %

G1 19 177/213 83 49/66 74

G2 17 208/238 87 86/107 80

G3 12 124/131 95 32/41 78

4.1.2 Errors

Erroneous occurrences in article use formed 26% of all the produced NPs by the

weaker group, 19% by G2, and 12% of all G3’s production. It is worth mentioning

that Choice errors were higher than grammatical errors for all PLs.

4.1.2.1 Omission

The omission of the indefinite article was the highest of all grammatical errors

recorded in G1’s production (44%). It constituted 34% of all grammatical errors made

by G2, and 18% of all grammatical errors made by G3. G1 omitted the indefinite

article 42 times in 106 contexts. In percentages: weaker learners failed to supply a(n)

40% of the time with singular indefinite NPs. The intermediate group omitted the

indefinite article 30 times in 116 indefinite NP contexts (26%) while the advanced

group omitted a(n) in 11% of obligatory contexts. Omission rates of the indefinite

article seem to decrease regularly and significantly as PLs improve.

117
In analysing learner data from T1, omissions of the definite article were

considered as choice -rather than grammatical- errors since it is the semantic, rather

than the grammatical, value that is being flouted. These omissions were generally

lower than those of the indefinite article across groups perhaps because the learners do

not feel the same semantic pressure to additionally mark indefiniteness by an article.

G1 omitted the definite article in 48 obligatory instances, which is 24% of all definite

contexts. The rate dropped down to 17% in G2’s production and further down to 12%

by G3. Although the omission rates of the definite article are generally low in this

task, as PLs improved, they dropped significantly down as learners’ PL progressed

from G1 to G2 (p=0.0476)

As the accuracy rates of the definite article were found to be higher than the

indefinite, the omissions of the definite article in the two lower groups were

considerably rarer than omissions of the indefinite article. In other words, G1 and G2

omitted the indefinite article significantly more than the definite (G1 p=0.0059, G2 p=

0.0476). The gap between the omissions of the two articles became smaller as overall

proficiency improved (see Figure 1). More advanced learners (G3) omitted the

definite article slightly more (1%) than the indefinite.

Figure 1 Omission rates of the definite and the indefinite articles across groups in Task 1

118
4.1.2.2 Overuse

The results obtained from T1 reveal fewer overuse errors of the definite article than

originally expected. In G1, non-referential indefinite plural/uncountable nouns were

unconventionally preceded by the definite form 52 times out of 385 obligatory

contexts (14%). This is relatively low. Yet the overuse rate dropped down to 10% in

G2’s (47 out of 461) production and further down to 6% only by G3 (14 of 236)

which means that learners gradually drop the unnecessary definite article from

indefinite contexts as their PL improves. It is worth mentioning that the overuse of the

definite article was not considered ungrammatical in T1 because it does not violate

any grammatical rules of countability or number; rather, its presence depends on the

speaker/writer’s assumptions about the state of the referent in the recipient’s mind,

whether it can be recognised as uniquely identifiable or not. Therefore, all overuse

instances of the definite article in the free production task were considered as errors of

choice.

Not many NPs in the singular were produced by our participants at any level.

Singular indefinite NPs did not represent more than 25-28% of all indefinite contexts

produced in T1. This relatively low production rate may undermine the reliability of

results relating to error or accuracy rates, yet it is important to notice learners’

preferences which certainly did not include many indefinite singular constructions. A

comparison with results that emerge from similar contexts in other tasks will be

carried out in the section reviewing Cross-Task results to help validate these

observations.

119
The total number of errors resulting from the overuse of the indefinite article,

whether grammatically or by choice, was no more than 25 instances across PLs, most

of which were choice errors. The ungrammatical supply of a(n) [with plural and

uncountable nouns] totalled (7%) with an average of 2.3% in each PL group.

Although the error rate is low, it is worth mentioning that G3 overused the indefinite

article more often than the two weaker groups. The positive correlation between error

rate and proficiency level in this exceptional case will be discussed in the following

chapter.

When the overuse of the definite article was compared to that of the indefinite

in the writing of the two weaker groups, the difference in the error rates was

significant. We found that choice errors caused by the overuse of the definite article

were considerably more frequent than the ungrammatical supply of the indefinite

article. The results of the advanced group, however, reveal a much smaller difference

between the overuse rates of a(n) and the. Table 10 illustrates the sharp contrast in

overuse patterns between the two articles.

Table 10 Overuse rates of a/n and the in Task 1

P
n= the % a(n) %
the : a

G1 19 52/385 13.51 6/279 2.15 p<0.0001

G2 17 47/461 10.20 8/345 2.32 p<0.0001

G3 12 14/236 5.93 4/172 2.33 0.0603

120
It is noticeable that the overuse of a(n) with plural nouns did not exceed 1% in the

outcome of any PL group, whereas the overuse of a(n) with uncountable nouns was

more recurrent (5%). However, both types of indefinite article errors are relatively low

compared to the rates of the overuse of the definite article.

4.1.2.3 Replacement Errors

There are two possible replacements errors: the supply of the indefinite article with

uniquely identifiable referents (a/n for the); and the supply of the definite article in

indefinite contexts (the for a/n). Since it has always been grammatically acceptable for

the definite article to replace the indefinite while the reverse is not always possible

(e.g. the head noun was plural/uncountable), only +Count + Sing contexts were

selected for both error types. Thus the probability of occurrences of both articles is

relatively approximated. With grammatical incorrectness excluded, the

appropriateness of article use becomes a matter of preference. Therefore, in this study,

all replacement errors are considered as choice errors (see chapter 3).

Choice errors form 59% of all errors committed in T1; a considerably higher

rate than that of grammatical errors (41%). In analysing data entries for this task, it

was evident that despite the low rates of replacement errors, the definite article was

the preferred option, especially for weaker learners, as it replaced the indefinite at

least three times as many in +Count + Sing contexts (see Table 11). It was also

interesting to see that the inclination to substitute a with the was reduced with

improved PLs while the production of the indefinite article in definite contexts

increased steadily.

121
Table 11 Replacement errors across PL groups in Task 1
a for
Groups n= the for a % %
the

G1 19 5/106 4.7 1/111 0.9

G2 17 4/116 3.4 2/161 1.2

G3 12 2/64 3.1 4/116 3.4

4.1.3 Learner Hypotheses

4.1.3.1. Specificity

Results from the free writing task indicate that there is a clear association between

specificity and article use in learner production especially at lower PLs. Confirming

earlier predictions, learners from all groups overused the definite article more

frequently in non–specific contexts. Out of a total of 52 occurrences, G1 overused the

definite article 39 times in–Sp (indefinite) contexts, compared to 13 only in [–def

+Sp] contexts. In percentages 75% of all overuse instances were non-specific while

25% only were in indefinite specific NPs. Fewer errors were made in the higher group

as G2 overused the 29 times in [–Sp] contexts out of a total of 47 overuse instances,

while not more than 18 occurrences were observed in specific indefinite contexts. In

percentages, 62% of all overuse errors by G2 were in non-specific contexts while

overuse in specific contexts was 38%. The association between non-specificity and

definiteness became significantly weaker in the production of the higher group, but

the preference remained unchanged as 64% of all overuse errors were noted in [–Sp]

122
NPs, and only 35% was recorded in indefinite referential contexts, as is illustrated in

Table 12.

Table 12 The overuse of the in T1 in ±Sp contexts compared to total overuse errors

n Non-Specific % Specific %

G1
19 39/52 75 13/52 25

G2
17 29/47 62 18/47 38

G3
12 9/14 46 5/14 35

The difference between the overuse rates in +/–Sp contexts was more significant at

lower levels (p=0.0335) whereas the difference in overuse rates in the production of

the two upper levels was not significant.

When the overuse errors in the two –/+SP contexts were calculated separately,

i.e. without comparing them to the total overuse errors, a different error map emerged

(see Table 13).

Table 13 Overuse of the in specific/non-specific contexts

G1 G2 G3
n=19 n=17 n=12
Non-
16% 11% 7%
Specific

Specific 9% 9% 5%

123
It is evident from Table 13 that there was a steady decline of overuse rates in non-

specific contexts as learners advanced while a smaller reduction in errors was noticed

across groups when the NP environments were specific.

There was little difference between the overuse rates of the lower and

intermediate groups in specific contexts, but what was evident was that these overuse

rates were both lower than those in –specific contexts, which could be an apparent

indication of L1 influence in the weaker groups.

An opposite pattern was observed with regards to the indefinite article. The

majority of learners’ production shows that the indefinite article was better supplied in

specific contexts, while almost no generic meanings were expressed using the

indefinite article. What is interesting is that the overuse of the indefinite article in both

+/–Sp contexts increased as learners’ English improved. This will be further discussed

in the following chapter.

4.1.3.2. Premodification

The results emerging from the weaker group show that the indefinite article was

omitted more in non-premodified contexts than in adjectivally premodified ones. G1

learners omitted the indefinite article 17 times in 41 non-premodified indefinite

singular NPs. In percentages, they failed to supply a(n) in 41% of all obligatory

contexts; while in premodified contexts there were only 25 omissions in 65 cases

(38.5%). As learners’ PLs improved, this pattern was reversed as higher omissions

were recorded when the head noun was adjectivally premodified. The intermediate

group omitted the indefinite article more often in Adj + N constructions (31%) than in

non-premodified ones (16%). The rate of omission before modified nouns was almost

double the rate of omissions in non-premodified contexts. In this respect, the

124
performance of G2 was in line with that of the advanced group as the errors made by

G3 in premodified contexts were 71% of all indefinite article omissions, while only

28% of omission errors were observed in nouns not preceded by an adjective.

Learners in this group omitted a(n) in 13% of all obligatory contexts where the head

nouns were premodified, but error rates were not higher than 7% with non-

premodified singular nouns.

Figure 2 Omission of the indefinite article in ±Premodified constructions in T2

It is clear from Figure 2 that the improvement across PLs was noticeable, i.e.,

omission rates were significantly reduced as learners progressed from G1 to G2. The

errors continued to fall in G3’s production with growing association between the

article with the head noun, rather than the adjective.

Similar outcomes were recorded with respect to the definite article. Omission

rates were consistently higher in premodified constructions in the production of

125
learners from all PLs, but the difference between omission rates in the two contexts

was more substantial in the weaker group’s production (p=0.0399). Despite the

regular decrease in omission rates with the overall L2 improvement, the cross group

comparison reveals that the gaps between PLs was not as noticeable as in the case of

the indefinite article. This is evident in the non-significant P values in statistical

calculations as shown in Table 14.

Table 14 Omission of the in ±Premodified constructions in T1

Non-
n= % Premod % p
Premod
G1 19 22/117 18.80 26/82 31.71 G1 v G2

0.1167
G2 17 29/199 14.57 20/94 21.28 G2 v G3

0.3692

G3 12 13/134 9.70 14/87 16.09

4.1.3.3 Concreteness

The results show a significant correlation between article use and the semantic value

of concreteness in learners’ writing, especially at lower PLs. Learners of all

proficiency levels supplied the definite article more readily when the head noun is

abstract.

126
Table 15 Overuse of the definite article in abstract and concrete contexts

Groups concrete % Abstract %

G1 33/285 11.58 19/100 19.00

G2 28/303 9.21 19/158 12.03

G3 7/165 4.24 7/71 9.86

The figures and percentages in Table (15) confirm that the overuse of the definite

article in both semantic contexts decreased as PLs improved but the reduction in error

rates was most noticeable between G2 and G3 in concrete contexts (p<0.0288).

Despite the fact that the majority of nouns produced in students’ writing were

concrete, abstract NP environments attracted higher overuse rates of both articles, i.e.,

overuse rates of the indefinite article were also higher in abstract contexts. For

example, six instances of a(n) overuse were recorded in G1’s production four of

which were with abstract nouns but only two were found in concrete contexts. In other

words, the overuse of the indefinite article in abstract environments was double the

rate of its overuse in concrete contexts. In the production of the two upper groups,

however, the difference was smaller and not as significant, but the preference was

similar.

It is worth mentioning that in both ontological contexts, the definite article was

overused more frequently than the indefinite. For example, in concrete contexts, G2’s

overuse rate of the was 9%, while the overuse rate of the indefinite article was 1.2%.

127
A comparison between the overuse rates of the two articles in abstract contexts is

illustrated in Figure (3).

0.25

0.2

0.15
a
0.1 the

0.05

0
G1 G2 G3

Figure 3 Overuse proportions of both articles in abstract contexts

4.2 Task 2

The second test is a cloze test, made up of six short stories with some articles removed

and replaced by blanks. The results emerging from the stories task (T2) represent the

outcome of comparing students’ responses in each blank to those of native speakers in

the control group. Therefore, what constitutes a grammatical error is the unavailability

of a similar response in the CG data. Similarly, an answer is considered correct if an

identical form is produced by a NS in that particular blank. However, since not all

responses of the CG were similar, learners’ responses were considered ‘correct’ if

they matched the form produced by the majority of NS, and as errors of choice if they

matched the answers provided by the minority of NS (see section 3.5.3.2).

128
4.2.1 Accuracy

A regular pattern of development is detected in the use of both articles across various

NP environments as PLs improve, i.e. a steady rise in accuracy rates was found to be

significant across groups as presented in the figures of Table 16 below.

Table 16 Accuracy rates across groups in T2

Groups n the % p a(n) % p

G1 15 1214/1890 64.23 G1 v G2 468/690 67.83 G1 v G2

<0.0001 <0.0001

G2 20 1906/2520 75.63 G2 v G3 753/920 81.85 G2 v G3

<0.0001 <0.0001

G3 16 1752/2016 86.90 G3 v G1 629/736 85.46 G3 v G1

<0.0001 0.0467

Unlike the results obtained from T1, the difference between the accuracy rates of the

definite and indefinite articles was not significant in G1’s and G3’s production, but

the intermediate level students performed significantly better on the indefinite than the

definite article (p<0.0001).

With regards to appropriate article use in indefinite plural/uncountable

contexts, the rates were low, but there was significant progression across PL groups.

The majority of the CG left 45 NPs unmarked while only 26% of the responses from

the weaker group were correct. The accuracy rates in G2’s results were higher (56%)

129
and further improvement was evident in G3’s answers as the accuracy rate rose to

64%. Finally, learners from all PL groups performed best on the definite article and

worst on the zero article in this task.

4.2.2 Grammatical Errors

In this task, grammatical errors refer to all responses that do not match any of the CG

data (see Methodology 3.5.6.2). Therefore, unacceptable learner responses include

omission, overuse and replacement errors of either article.

4.2.2.1 Omission

In the 45 indefinite singular NPs included in this task, learners in the weakest group

failed to supply the indefinite article 78 times out of 690 (45 blanks X 15 participants)

obligatory contexts. The omission rate for G1 was therefore 11% only. It fell to 4.4%

(41 out of 920) in G2’s production as the intermediate group supplied the indefinite

article better in obligatory contexts. With improved PLs, only 2.9% (22 out of 736)

total blanks were missed by G3.29 The gradual decrease in omission rates of the

indefinite article was most significant as learners advanced from lower to

intermediate PL (G1: G2 p<0.0001).

29
The number of total obligatory contexts varies according the number of students taking the test in
each group.

130
Figure 4 Omission errors of the definite and indefinite articles in obligatory contexts

The omissions of the definite article were lower than those of the indefinite in G1’s

production. The weak learners failed to supply the in only 8% of the blanks that were

unanimously filled with the definite article by the CG. The omission rates were

slightly reduced in G2’s production (7.3%) but the results from G3 seem to suggest

that the omission of the was considerably reduced to 4% which signals significantly

better performance (G2: G3 p<0.0001).

Apparently, the omission errors of both articles decreased with progress

towards the TL but a closer look at the error map shows that the two articles were

divergent in their development: while participants in G1 omitted the indefinite article

significantly more frequently than the definite (p=0.0116), the image was reversed in

the results of G2 as learners supplied the indefinite article better than the definite in

obligatory contexts (p=0.0007). The gap in omission rates drew closer to improved

131
L2 knowledge as no significant difference was found in the omission rates of the two

articles in G3’s results.

4.2.2.2 Overuse

Because of the nature of this task, there were far more instances of overuse than

omission errors. Overuse in this task is defined as the supply of any form of article in

instances where none of the NS did so.

Since none of the NS control group committed ungrammatical errors of

indefinite article overuse, the criteria that describe NPs and the ‘key’ answers (against

which learners’ responses are measured) were one and the same. The results show that

there is a regular and systematic decline in the overuse of a(n) as PLs improve.

G1overused the indefinite article in 27% of the impermissible contexts, whereas the

overuse rate in G2 was 15%. G3 overused the indefinite article 169 times across 2688

uncountable/plural contexts (6%). Therefore the progress was significant across PL

groups (p<0.0001).

It is worth mentioning that learners in the weaker group had more difficulty in

recognising the unsuitability of supplying the indefinite article in uncountable contexts

while better performance was recorded in plural contexts. G1 participants incorrectly

oversupplied the indefinite article within uncountable contexts, which is significantly

more than they did with plural NPs (p<0.0032) while the difference in overuse rates of

uncountable and plural contexts in groups 2 and 3 was not significant. However, what

was interesting to find is that the development across groups was significant in both

plural and uncountable NP environments (see Table 17).

132
Table 17 Overuse of the indefinite article in uncountable and plural contexts

uncount % p plural % p

G1 G1 v G2 G1 v G2
174 /450 39 76/270 28
n=15 <0.0001 0.0145

G2 G2 v G3 G2 v G3
108/600 18 71/360 20
n=20 <0.0001 0.0002

G3 G3 v G1 G3 v G1
42/480 8 28/288 10
n=16 <0.0001 <0.0001

The definite article was overused in G1’s responses 355 times in 1365 indefinite NP

contexts. In percentages, it is 26% of all blanks marked as indefinite by NS. This rate

was reduced to 18% in G2 production as learners in the intermediate group over

supplied the definite article 319 times out of 1820 possible NPs which NS did not

mark as definite. Fewer still overuse errors were made by more advanced learners in

G3 (15%). Finally it is worth mentioning that in the blanks test, there was little

disparity in the overuse rates between a(n) and the in the G1’s results, whereas the two

upper groups overused the definite article significantly more frequently than the

indefinite (G2: p=0.0097, G3: p<0.0001).

4.2.3 Choice Errors

Choice errors are grammatically permissible options of article use, yet they are not the

options preferred by the majority of NS. In this section, we will review the results

133
from the three groups of learners in relation to the exceptional responses provided by

the few NS in the CG with instances of grammatical30 overuse excluded.

Recall from chapter 3 section 5.6.2 that not all the CG responses were

categorised into either analogous or semi-analogous key, as there was one case

(example 45. top politician) where the three possible options (a(n), the, zero) had all

been supplied for one blank, which consequently led to the blank being dropped from

our calculations.

4.2.3.1 The

Participants from G1 overused the definite article in 25% of the NS choices while the

better group’s error rate was 15.75%, which is significantly lower than that of G1

(p<0.0017). Choice error rates of the were 12.3% in G3’s production, but the

difference between the performance of the intermediate and advanced groups was not

significant.

4.2.3.2 a(n)

The choice to use the indefinite article in contexts which most NS considered

definite31 was less favourable than applying the definite article in blanks which the NS

considered indefinite. However, the decrease in the error rates of indefinite article

choices was categorically more regular in pattern across the PL groups. This too is not

clear. The weaker group supplied the indefinite article in almost 11% of the available

choice contexts whereas the error rate in the performance of the intermediate group
30
In T2, Grammatical errors refer to all forms of responses in learner data that are not found in any of
the CG choices.
31
This type of use denotes the supply of the indefinite article in countable singular NP contexts which
only few NS assigned it indefinite meanings.

134
fell to 3.5% which is statistically lower than that of G1(p= 0.0010). The G3 data also

revealed a substantial decrease in choice error rates (G2: G3 p< 0.0001) as participants

chose to use the indefinite article in only 0.8% of possible choice contexts.

4.2.3.3 Zero

What learners seemed to be mostly struggling with is recognising that indefinite

plural/countable nouns should not be marked by articles. Surprisingly, it was G2’s

answers (rather than the weaker group) that approximated the responses made by the

NS consulted. It was also interesting to note that little cross group improvement was

noticed as PLs improved.

Table 18 Choice errors of the zero article

Groups n Ø %

15 8/60 13.33
1

2 20 21/80 26.25

3 16 9/64 14.06

With regards to overuse patterns in the blanks test, the preference to use the definite

article in NPs which most NS considered indefinite, was statistically stronger

(p<0.0001) than the use of the indefinite article by all PL groups. Although all

participants overused the definite article more frequently, the preference was more

135
notable in the weaker groups’ production, while the difference between the error rates

of the two articles became smaller as PL improved.

Figure 5 illustrates how error rates decreased with progress in PLs. It also

shows that learners had consistently overused the definite article more frequently (in

singular indefinite contexts) than the indefinite article in (singular) definite contexts.

Figure 5 Comparison between choice errors of a(n) and the in T2

4.2.4 Replacement Errors

Replacement errors refer to instances of suppliance of one article form in blanks

where the majority of NS in the CG chose to supply another. In order to approximate

the number of possible occurrences of both articles, only singular NPs were selected.

Thus, by excluding countability and number, the variables were restricted to those

involving definiteness alone.

136
In the results emerging from G1, the definite article was used as a substitute

for the indefinite in 24% of all singular indefinite constructions. The replacement error

rate dropped significantly down to 17% in G2’s production (p=0.0003). The decrease

in substitution errors rate was greater between G1 and G2’s performance, but the

difference between the two upper groups was not significant. G3 used the to replace

a(n) in 111 NPs in a total of 736 indefinite article contexts (15%). The distinction

between the two error types is shown in Figure 6.

30.00%
26%
25%
25.00%

20.00%
17%
15% 15%
15.00% the for a
a for the
10.00%
6%
5.00%

0.00%
G1 G2 G3

Figure 6 Replacement errors in T2 compared across groups

On the other hand, the indefinite article replaced the definite article more frequently in

the production of the weaker group. 29% of all definite singular contexts were

oversupplied with a(n), while G2 preferred to supply the indefinite article in 18% of

all definite singular contexts which is statistically significant (G1:G2 p<0.0001). The

advanced group used a(n) in 7.7% of all contexts that were considered definite by the

137
majority of NS which also signals a considerable improvement from the intermediate

group (G2:G3 p<0.0001).

The replacement rates of both articles fell gradually as PLs improved, but it

was more noticeable in contexts where the definite article was replaced by the

indefinite (a for the) as the gaps between groups were larger and the differences in the

error rates more significant. This tendency is in line with overuse/omission patterns in

which error rates concerning the indefinite article show a sharper and more regular

improvement than those of the definite.

4.2.5 Learner Hypotheses

4.2.5.1 Specificity

To find out if article use was influenced by the specificity status of the referent, nouns

were divided into [–def +Sp] on one hand and [–def –Sp] on the other. Overuse

instances of the definite article were then compared across the two indefinite contexts.

We found that the definite article was employed more frequently in (indefinite) non-

specific contexts than in (indefinite) specific contexts by participants from all the PL

groups.

The difference between overuse instances of the in +/–Sp contexts was

statistically significant in G1’s production, with higher percentages recorded in non-

specific NP environments (p=0.0335). G2 also overused the definite article more

frequently with [–def] [–Sp] nouns, but the gap between the two contexts was not as

large as the one observed in the weaker learners’ performance. Similarly, the higher

group applied the slightly more often in non-specific contexts.

138
Table 19 Overuse of the definite article in indefinite -/+Specific contexts

Groups n= –Sp % +Sp %

G1 15 39/241 16.18 13/144 9.03

G2 20 29/269 10.78 18/192 9.38

G3 16 9/131 6.87 5/105 4.76

From the figures in Table 19, it is evident that overuse of the definite article in [–def]

[–Sp] noun phrases decreased as PLs rose, yet the records do not reveal statistically

substantial improvement rates across groups.

We also investigated the effect of specificity on learners’ use of the indefinite

article. The results indicate very few occurrences of unnecessary a(n), circa 1%

average per PL group. But it is worth acknowledging that despite the low overuse

rates, most instances of supply were recorded in [+SP] contexts across all groups.

4.2.5.2 Premodification

The effect of premodification on article choice was measured by comparing learners’

ability to supply the indefinite article in premodified (+Premod) and non-premodified

(–Premod) NP contexts. The obtained results confirm that lower and intermediate PL

learners omitted the indefinite article more frequently in instances where the head

noun was not premodified. In fact, when occurrences of omission in the two contexts

were compared, it was found that out of the 78 singular indefinite nouns, G1 omitted

the indefinite article 50 times before bare nouns and 28 times only before modified

nouns. In percentages: 64% of omission instances were made in –Premd contexts

139
while 35.8% of omissions were in +Premd constructions. A similar pattern was

observed in the performance of G2. There were 41 instances of indefinite article

omissions; 27 of which were noted in –Premd contexts (66%), while the remaining 14

were in +Premd NP environments (34%). The reverse of these choices was observed

in G3’s performance, since more omissions of the indefinite article were committed

before adjectivally premodified nouns.

Another test was conducted to examine participants’ responses against those in

the CG data. The results were similar to those found in the first test: the two weaker

groups supplied the indefinite article better with adjectives, while the more advanced

group made significantly more omissions in +Premd environments (p=0.0438).

Further details are presented in Table 20.

Table 20 Omission rates of the indefinite article in ±premodified contexts in T2.

Group n +Premd % p –Premd % p

G1 15 27/258 10.47 50/405 12.35


G1 v G2 G1 v G2

0.0015 p<0.0001
G2 20 14/380 3.68 27/540 5.00

G3 16 14/304 4.61 G2 v G3 8/432 1.85 G2 v G3

0.5503 0.0057

Table 20 shows that the decrease in error rates across PL groups in premodified NP

contexts is somewhat erratic: Although participants made significantly fewer

omissions as they progressed from G1 to G2, the sudden slight rise in errors by G3 in

+Premd contexts disrupts the expected pattern of progression in performance. On the

140
other hand, the cross-group change in –Premd environments is evidently more steady

and regular, perhaps because of the absence of the distraction of the modifying

adjective.

To further investigate learner trends in article use with regards to premodified

nouns, the use of the definite article was also examined. The results show very low

omission rates and very small differences between the two NP contexts in the

production of all the groups. However, it was interesting to find that the omission

pattern was the mirror image to that of the indefinite article since the weak group

supplied the better in +Premd and made more omissions in –Premd constructions

while better PL learners omitted the definite article more frequently in +Premd

contexts. This is illustrated in Table 21.

Table 21 Omission of the definite article in ±Premodified contexts

Group Nonmodified % Modified %

G1
113 / 1475 9.60 36/375 7.69
n=15

G2
131 / 1960 5.40 27/500 6.68
n=20

G3
70 / 1568 3.00 12 / 400 4.46
n=16

141
4.2.5.3 Concreteness

The results indicate that the supply of the definite article was more frequent in

instances where the head noun bears an abstract (–Concrete) semantic value. The

weaker group overused the definite article in 31.5% of the NPs which NS considered

indefinite, while fewer overuse rates were observed in concrete ones (25.3%). The

difference between misuses by G1 in the two environments (+/ –Concrete) was

statistically significant (p=0.0340). The intermediate group overused the definite

article 113 times out of 440 abstract indefinite contexts (26%) compared to 206 times

in 1320 concrete contexts (18%) which was also a significant difference (p<0.0001).

Fewer overuse rates were recorded in G3’s results, yet learners’ preferences remained

the same. More advanced participants over-supplied the definite article in 23% of

(indefinite) abstract contexts while only 15% of indefinite concrete nouns were

preceded by the. Again, the difference between overuse rates in the two ontological

environments was considerable (p<0.0001).

Figure 7 Overuse of the definite article in Abstract and Concrete contexts

142
Figure 7 illustrates how the ontological status of a referent affected L1Ar learners’

choices of the definite article. Moreover, the concreteness/abstractness status also

influenced, although to a lesser degree, learners’ choices of the indefinite article.

Evidently, participants from all PL groups marked abstract nouns more often than

concrete nouns with the indefinite article. In the production of participants from all

groups, significantly higher overuse errors were recorded in abstract domains (G1

p<0.0001, G2 p<0.0001, G3p<0.0001). As learners’ L2 knowledge became better,

fewer overuse errors were made in both environments (G1: G2 p<0.0001, G2:G3

p<0.0001). This is an indication that the association between articles (form) and

concreteness/abstractness (meaning) weakens as TL proficiency develops.

Proportionately higher overuse rates remained within definite descriptions in

both abstract and concrete contexts. Therefore, the supposition that there is a stronger

inclination to associate definiteness, rather than indefiniteness, with abstract notions

still holds.

4.3 Task 3

The third test (GJT) was designed to find out learners’ ability to decide on the

grammaticality of underlined NPs within discrete decontextualised sentences. The

results were based on correctly identifying accurate or erroneous article use and

marking them as either correct or wrong, respectively. Therefore, no choice errors

were included.

143
The effect of the extreme directness and high control of this task was evident

in the sharper contrasts observed in the results and the distinct differences of the

performances between groups.

4.3.1 Accuracy

The results emerging from this test show a positive correlation between accuracy rates

and L2 knowledge. This was true throughout the results from the three PL groups

whereby the higher the PL, the higher the accuracy rates. Learners’ recognition of

accurate definite descriptions was particularly high; 88% of all definite NPs were

correctly marked in G1’s production. G2 managed to correctly mark 92% of contexts

where the was used while G3 scored a very high (98%) accuracy rate.

On the other hand, the accuracy rates of the indefinite article were much lower

than those of the definite in the production of weaker learners where the difference

between the accuracy rates of the two articles was found to be statistically significant

(p=0.0079). The more advanced group, however, scored slightly higher rates in their

responses to singular indefinite than to definite NPs. In percentages, the weaker

participants recognised 66.6% of correct singular indefinite NPs. This rate improved

sharply in G2’s performance to 91%, while participants from G3 were able to

acknowledge all correctly marked singular indefinite NPs. The improvement on the

indefinite article across PLs was substantial in this task as significant gaps were found

between groups (G1:G2 p=0.0020, G2:G3 p=0.0192). The difference in the

performance of all three groups in T3 is evident from the percentages in Table 22.

144
Table 22 Correct marking of definite and singular indefinite NPs in T3

Group n The % a(n) %

1 15 53/60 88.33 30/45 66.67

2 19 70/76 92.11 52/57 91.23

3 17 67/68 98.53 51/51 100.00

In contrast to the higher accuracy rates on the two articles, participants seem to have

had difficulty in realising that non-referential bare nouns do not usually require the

presence of articles. This is evident in the fact that participants from all groups judged

most correctly unmarked indefinite plural/uncountable NPs as incorrect. Therefore,

the lowest accuracy rates in this test were found in learners’ reaction to unmarked

nouns. G1 achieved a 69% accuracy rate. In numbers, there were 80 correctly marked

bare nouns in 120 contexts. A slight improvement was observed in the performance of

G2 as they were able to identify correct zero-determiner nouns 112 times in 152

instances of similar contexts; in percentages, 74% of indefinite plurals were properly

marked. G3 recognised correct occurrences 90% of the time as advanced participants

were capable of distinguishing122 correct zero article instances out of 136 NPs. It is

worth mentioning that although learners’ accuracy rates increased with higher

proficiency levels, the scale of improvement across groups is not notable. This is

illustrated in figure (8).

145
120

99 100
100 92 91 91
88 90

80
67 69
Correct the
60
Correct a

40 correct 0

20

0
G1 G2 G3

Figure 8 Percentages of accurate marking of NPs in T3

As Figure 8 demonstrates, the gap in performance between articles diminishes

gradually as learners L2 knowledge improves. The chart also illustrates the systematic

rise of indefinite article accuracy rates while negligible variation is observed in the

rates of the definite article. Finally, it appears that learners were more aware of

appropriately unmarked plural nouns than uncountable nouns in indefinite contexts.

Performance patterns did not vary considerably with improved proficiency levels.

Table (23) clearly shows how accuracy rates were higher in plural NP contexts across

PLs.

Table 23 Correct marking of indefinite NPs in plural and uncountable contexts

Group n Uncountable % Plural %

G1 15 38/60 63.33 45/60 75.00

G2 19 61/76 80.26 68/76 89.47

G3 17 60/68 88.24 62/68 91.18

146
4.3.2 Errors

4.3.2.1 Omission

In this task, definite descriptions were purposely designed to cover various contexts of

obligatory supply (see definiteness chapter 2 section 1.1) such as physically present or

universally unique referents. We will present the omission maps of the definite and the

indefinite articles respectively. Learners in G1 could not identify 30.6% of NPs with

definite article omissions. G2 failed to recognise 18% of similar instances. The error

rates fell sharply with improved PLs as G3 participants were able to identify most

definite article omissions missing only 5.8% of the total NPs. Hence, the difference in

the error rates between the two upper groups was significant (G2:G3 p=0.0060).

Unsurprisingly, learners were not as able to detect instances of indefinite

article omissions as well as they were of the definite. In the production of the weaker

group, NPs with omitted a(n) were mistaken as correct 37.7% of the time. This

relatively high error rate dropped considerably down to 16.7% in G2’s results. Thus

the improvement scale between G1 and G2 was statistically considerable (p=0.0006).

With better awareness of the indefinite article supply, G3 accepted only 10.7% of

omission instances. The difference between learners’ performance on the two articles

as well as the gradual decrease in error rate is shown in Table 24.

147
Table 24 Omission rates of both articles in T3

Group n= the % a(n) %

G1 15 23/75 30.67 34/90 37.78

G2 19 18/95 18.95 19/114 16.67

G3 17 5/85 5.88 11/102 10.78

In comparing the performances on both articles, it seems that G1 were better at

recognising omissions of the definite article than the indefinite. In G2’s results, little

difference is noticed between participants’ behaviour on both articles, but they were

slightly more able to identify indefinite article omissions. G3 performed better on the

definite than on the indefinite article as they accepted twice as many instances of

indefinite article omissions as they did of the definite article. However, cross article

evaluations did not yield statistically significant results.

4.3.2.2 Overuse

There was apparent progress in error-recognition with learners advancing from

beginner to intermediate levels. As participants’ L2 proficiency improved, they were

more capable of noticing the unnecessary supply of the definite article in (–def) non-

referential contexts. G1 marked 42% of NPs where the was overused as correct, while

learners in the two upper groups viewed overuse instances as correct in only around

14% of the time. It is clear that this error rate is significantly lower than that of the

weaker group (G1:G2 p<0.0001) while no great difference was observed in the

achievement of learners from G2 and G3.

148
The overuse pattern of the indefinite article was predictably more regular than

that of the definite article. Error rates decreased proportionately maintaining a steady

pace between the PL groups. G1 participants accepted more than half of the

ungrammatical NPs as correct while participants from G2 marked 47% of these

instances as correct. More advanced learners’ recognition of indefinite article overuse

in this test was 37% erroneous.

Table 25 Overuse of the definite and indefinite articles in T3


p
Group n the % a(n) %
the: a

G1 15 58/135 42.96 43/75 57.33 0.0437

G2 19 24/171 14.04 45/95 47.37 <0.0001

G3 17 23/153 15.03 32/85 37.65 <0.0001

As is shown in Table 25, the gap between the overuse rates of the two articles was

substantial across PLs because participants from all groups accepted the overuse of the

indefinite article more readily than the definite, causing higher overuse rates of the

indefinite article.

The ungrammatical supply of a(n) was much better recognised in NPs where

the indefinite article preceded plural nouns than it was in uncountable contexts. In

other words, participants from all PL groups were not as able to notice overuse

instances in mass environments. The difference between the two contexts was

statistically significant (G1:G2 p=0.0105, G2:G3 p<0.0001) and visibly noticeable in

Figure 9.

149
Figure 9 Overuse of the indefinite article in uncountable and plural contexts in T3

4.3.2.3 Replacement

Despite the fact that only singular nouns were selected in assessing replacement

errors, in this test both types of replacement errors are considered grammatical errors.

Thus, unlike the approach followed in analysing similar errors in T1 and T2 in which

definiteness is the only variable, there are no choice errors in the GJT (see chapter 3,

section 5.6.3).

The results show that the definite article replaced the indefinite in 31% of

possible instances in the production of the weaker group. In other words, G1

participants failed to recognise a third of the instances with replacement errors. The

two upper groups performed distinctively better (p=0.0395) in identifying the-for-a

cases as the error rates in their responses dropped down to 15% and 14%.

150
Similarly, the replacement errors were also reduced with improved PL in

contexts where the indefinite article substituted the definite. While G1 participants

mistook 27% of a-for-the occurrences as correct, G2 failed to acknowledge only 5.2%

of replacement NPs. Furthermore, it was interesting to find that none of the students in

G3 accepted any of the erroneous instances. It is easier to track the progress of groups

and compare their performance on the two articles by considering the figures in Table

(26).

Table 26 Replacement errors in T3

Group n the for a % a for the %

G1 15 14/45 31.11 4/15 26.67

G2 19 8/57 14.04 1/19 5.26

G3 17 8/51 15.69 0/17 0.00

It is apparent from the comparison of replacement errors in definite and indefinite NP

environments that learners tended to replace the indefinite article with the definite

more readily than the opposite. This was more evidently true in G3’s results since a

wider gap was observed between the replacement rates of the two articles (p=0.0020).

151
4.3.3 Learner Hypotheses

4.3.3.1 Specificity

The results emerging from T3 confirm that learners were better able to recognise

overuse instances of the definite article32 when the reference was specific, while non-

referential overuse of the was largely accepted as correct. As PL improved, the error

rates decreased steadily and the differences between error rates in +/–Sp domains

gradually diminished.

NPs with the definite article overused in [–Sp] contexts were marked as correct

75% of the time by G1, whereas only 25% of the total number of erroneous the was

mistaken as correct when the referent was specific. In other words, G1 accepted non-

specific use of the three times more often than the specific, causing differences

between the two contexts to be significant (p<0.0001). The results from G2 suggest

increased levels of error recognition, i.e., fewer overuse NPs were marked as ‘True’ in

both +/–Sp contexts. Although error rates dropped significantly down from G1’s

responses (G1:G2 p<0.0001), acceptability of overuse remained considerably higher

in [–Sp] contexts (p=0.0474). The largest proportion of errors (62.5%) made by

intermediate level learners were in [–Sp] contexts while only 37% were in specific

ones. Thus, G2 participants misjudged the non-referential presence of the definite

article twice as often as they did when the referent was specific. Similarly, higher PL

learners overused instances of the more often when the supply was in (indefinite) [–

Sp] environments. The improvement in G3’s responses however can be noticed in the

fewer number of errors, in addition to the smaller gaps between the two specificity

contexts. The overuse rate in [–Sp] contexts was 65.5% compared to a lesser 43.4% in

[+SP] environments.

32
The term Overuse refers to the employment of the definite article in indefinite contexts, regardless of
the specificity status of the NP.

152
Another type of calculation was conducted to examine learners’ responses to

overuse instances by comparing rates in the two [-Sp/+Sp] NP environments. The

results obtained from this method confirm those that emerged from the former, i.e., all

instances of overuse of the definite article, regardless of context, are reduced as PL

improves, yet the inability to recognise overuse errors in [–Sp] contexts is more

prevalent across PLs. The breakdown of findings in numbers and percentages is

presented in Table 27 below.

Table 27 Overuse of the definite article in -/+Premodified constructions in T3

Group n +Sp % –Sp %

1 15 14/45 31.11 44/90 48.89

2 19 9/57 15.79 15/114 13.16

3 17 10/51 19.61 13/102 12.75

4.3.3.2 Pre-modification

A comparison of the omission errors of the indefinite article in premodified (+Prmd)

and non-premodified (–Prmd) contexts reveals that the weaker learners were better

able to notice omissions in adjectivally premodified constructions. In terms of error

analysis, this means that omission rates were higher in bare nouns. In contrast,

participants from the highest group performed better in –Premd setting.

Participants from G1 marked 6 out of 15 singular indefinite NPs with omitted

a(n) as correct in premodified contexts, while they were less able to notice its absence

directly before a singular noun (14/30), which caused omission rates to become

153
higher. The distribution of errors in contexts found in G2 responses echoed the results

of G1, i.e., learners better noticed the omitted article when the head noun was

preceded by an adjective. Intermediate PL participants failed to recognise the missing

article a few times (3/15) in modified compositions, but the error rate was higher

(8/38) in NPs without adjectives. G3 participants, on the other hand made fewer

mistakes in both +/–Prmd contexts, and their preferences were quite the opposite of

those in G1 and G2 since they were better at identifying omissions when there were

no adjectives preceding the head noun (1/34) than in cases of premodification (2/17).

Figure 10 Omission of the indefinite article in +/-Premodified constructions in T3

It is worth mentioning that the progress across groups was statistically distinctive in –

Premod NP environments (G1:G2 p=0.0228, G2:G3 p=0.0121).

Another calculation was performed to verify the distribution of total omission

errors across the two NP structures. The results confirmed the pattern found in the first

calculation across PLs. It proved that out of all omission errors made by the weaker
154
group, 30% were in premodified contexts as compared to 70% in non-premodified

ones. The rates observed in G2’s responses closely approximated G1’s proportions.

However, the error pattern was remarkably reversed in G3’s production with better

recognition of omissions in non-premodified constructions., i.e. lower error rates were

observed in –Prmd NPs (33.33%) than in premodified ones (66.67%).

With regards to the definite article, a corresponding tendency was found in the

results emerging from learners’ responses to premodification. Weaker learners marked

omission NPs as correct more recurrently in –Prmd environments while more

advanced group mistook overuse in +Prmd NPs more often. Finally, it is worth

mentioning that significant improvement in performance was noticed through larger

gaps across PL groups in non-premodified rather than in modified contexts (G1:G2

p=0.0102, G2:G3 p=0.0121).

Figure 11 Omission of the definite article in +/-Premodified contexts in T3

155
4.3.3.3 Concreteness

The difference between the overuse of the definite article in abstract and concrete

contexts is the most significant of all other learner-associations in this task. As the

proportions of overuse were compared across abstract and concrete concepts, it was

found that G1 accepted the overuse of the definite article with abstract nouns in 53.3%

of the cases (24/45) compared to 37% in concrete contexts (34 errors in 90 possible

NPs). G2 participants made fewer mistakes than the weaker group in both contexts,

but they still considered the overuse of the definite article more acceptable when the

meanings were abstract (26.3%) than when they were concrete (7.8%). Hence the gap

between overuse in +/–Concrete environments was significant (p=0.0037). The results

emerging from G3 confirm the previous hypothesis as learners misjudged 35% of

overuse instances in abstract NPs as correct while less than 5% of erroneous responses

were detected in concrete contexts. This indicates that the difference between errors in

the two ontological environments remained significant (p<0.0001). To compare

learners’ performance in the two contexts, see the percentages in Table 28.

Table 28 Overuse rates of the definite article in -/+ concrete contexts in T3

Group N Abstract % Concrete %

1 15 24/45 53.33 34/90 37.78

2 19 15/57 26.32 9/114 7.89

3 17 18/51 35.29 5/102 4.90

156
The cross-sectional map indicates that overuse rates were significantly reduced as

groups improved from beginner to intermediate levels in both abstract (G1:G2

p=0.0042) and concrete (G1:G2 p<0.0001) domains. However, error rates remain

invariably higher in abstract NPs in the responses across PLs.

Finally, investigating the overuse of the indefinite article across ontological

contexts yielded similar findings as we found that the indefinite article was overused

significantly more frequently in abstract, rather than concrete contexts by all groups,

as is explained in Table 29.

Table 29 Overuse of the indefinite article in Abstract and concrete contexts in T3

p
Group n Abstract % Concrete %
Abs: Conc

G1 15 23/30 76.67 24/60 40.00 0.0002

G2 19 33/38 86.84 13/76 17.11 <0.0001

G3 17 28/34 82.35 4/68 5.88 <0.0001

157
4.4 Cross-Task results

In cross task analysis, we calculate d values by comparing the responses of each group

across the three tests. However, this cannot be directly achieved because the number

of dependent variables differs from one test to the other. For example, when the

supply of a(n) was evaluated, the responses entered were measured against the total

number of obligatory (indefinite countable singular) contexts, which, in turn, vary

from one test to another. In T1, for instance, there were 385 indefinite singular

countable nouns produced by students, while in T2 46 blanks had to be filled with a(n)

compared to 9 similar NPs in the GJT. Therefore the only way to examine the effect

on task type on accuracy, errors and learner hypotheses is weigh the figures in

percentages.

4.4.1 Accuracy

In general, learners’ performance on the indefinite article was considerably less

accurate than their performance on the definite across tasks. The highest accuracy

rates of the definite article were achieved in T3 while the highest accuracy rates of the

indefinite article were achieved in T2. Conversely, the lowest rates of both articles

were recorded in T1 as shown in Figure 12.

158
Figure 12 Accuracy of the definite and indefinite articles by G1 across tasks

Task 2 yielded the least differences in accuracy rates of both articles while the gap

was considerably wider in T1 and T3.

4.4.1.1. the

The highest accuracy rates of the definite article were observed in T3, reaching

statistically significant rates than the rates in the other two tasks. This was true for all

groups except when comparing T1 and T2 rates in G3’s results. On the other hand, the

lowest rates for the two weaker groups were scored in T2 while G3’s results were

weakest in T1.

159
Table 30 Accuracy rates of the definite article

T1 T2 T3
T1 v T2 T2 v T3 T1 v T3
% % %
75.38 64.23 88.33
G1
0.0005 p<0.0001 0.0118
83.62 75.63 92.11
G2
0.0005 p<0.0001 p<0.0001
86.43 86.90 98.53
G3
0.8431 p<0.0001 p<0.0001

From the results reported in Table 30, it seems that cross-group improvement is more

noticeable in T2 than in the other two tasks because the difference in accuracy rates

across the three PL groups is more systematic and regular.

4.4.1.2 A(n)

As far as the accuracy of the indefinite article is concerned, little statistical variation

was recorded between tasks 1 and 2 in the results of all PL groups, yet the cross-group

progression was more noticeable in T1 than in the other two tasks, as the percentages

in table 31 illustrate.

160
Table 31 Correct indefinite singular marking across tasks

T1 T2 T3
T1 v T2 T2 v T3 T1 v T3
% % %
60.38 67.83 66.67
G1
p=0.1419 p=0.4584 p=0.8729

74.14 81.85 91.23


G2
p=0.0702 p<0.0001 p<0.0001

89.06 85.46 100.00


G3
p=0.3812 p<0.0001 p=0.0050

G1 scored the highest accuracy rates of a(n) in T2, but their results did not differ

greatly across tasks, with accuracy rates ranging between 60-67%. The more advanced

groups performed considerably better in T3 than in the free production task and the

stories task.

It is worth mentioning that the delay in production of the indefinite article is

not as noticeable in certain syntactic formulae such as memorised chunks learned at

earlier stages of L2 acquisition. Existential and have constructions show higher

accuracy rates of the indefinite article especially in T1.

4.4.1.3 Zero

Learners’ ability to leave indefinite plural/uncountable nouns unmarked depended to a

large extent on task type. All PL groups scored invariably higher accuracy rates in the

161
free writing test than they did in the controlled tests whereby the lowest accuracy rates

of the zero article were observed in T2.

Table 32 Correct indefinite marking of plural and uncountable nouns

T1 T2 T3
T1 v T2 T2 v T3 T1 v T3
% % %

G1 81.00 p<0.0001 26.81 p<0.0001 69.17 p=0.0141

G2 85.22 p<0.0001 56.35 p<0.0001 73.68 p=0.1648

G3 90.70 p<0.0001 63.67 p<0.0001 89.71 p=0.7718

There were significant differences in achievement across tests by G1, while closer

rates were observed in the performance of the two upper groups as T3 rates

approximated the rates recorded in T1, without much statistical variation, as is evident

in Table 32 above.

Test type however did not influence the variation observed in learners’

reaction with regards to plural and uncountable contexts. In all the three tests,

participants from all PL groups were better able to unmark/notice indefiniteness in

plural rather than uncountable NPs.

4.4.2 Errors

In this section, article errors will be compared across the three tests and presented in

three sections of overuse, omission and replacement. The first two types of article

misuse cover both grammatical and choice error while the latter is exclusive to choice

errors (see chapter 3 section 5.3). For example, overuse of the indefinite article is a
162
grammatical error if it is used before plural/uncountable nouns but it can also be a

grammatically permissible error when the referent is singular but uniquely

identifiable. Similarly the omission of the indefinite article is a grammatical error if it

precedes an indefinite singular noun, but it is regarded as an error of choice if it is

frequently used for generic reference, or it does not sound native-like. Errors

concerning the definite article are largely semantic, but in T2 if responses did not

agree with the majority it was regarded as less acceptable than the same error when

committed in T1. The overuse of the definite article, for example, was regarded as a

choice error in the free composition task when it was employed with referents that are

not uniquely identifiable. Therefore, to avoid inconsistency, the following section will

review article misuse in general regardless of the particular parameters employed in

each test.

4.4.2.1 Overuse
4.4.2.1.1 The

The was overused more frequently in T2 and T3 than in the free production task. For

example, G1 overused the definite article 52 times out of 385 indefinite NPs in T1, but

in T2 the overuse rates were much higher (355 out of 1365). The highest overuse rates

were recorded in T3 where G1 participants accepted 58 out of 135 instances of

overuse as correct. Similarly, the intermediate group made relatively fewer errors of

overuse in T1 (47/461) than in the stories task (319/1820) and the GJT (24/171). In

line with the performances of G1 and G2, the advanced group overused the few times

in T1 (14/236) but doubled the rate in the stories test (219/1456) and the GJT

(23/153). Hence all overuse rates proved to be significantly different from one task to

another, as Table 33 demonstrates.

163
Table 33 Overuse rates of the definite article across tests

T1 T1 v T2 T2 T2 v T3 T3 T1 v T3

G1 13.51 p<0.0001 26.01 p<0.0001 42.96 p<0.0001

G2 10.20 p<0.0001 17.53 p=0.4093 14.04 p=0.1323

G3 5.93 p<0.0001 15.04 p=0.9977 15.03 p=0.0054

4.4.2.1.2 A(n)

Similar to the findings on the overuse of the definite article, the results indicate that all

participants oversupplied the indefinite article more frequently in T2 than in T1 while

exceedingly higher rates were observed in the GJT.

Table 34 Overuse of the indefinite article across tasks

T1 T2 T3
T1 v T2 T2 v T3 T1 v T3
% % %

G1 2.15 p<0.0001 27.02 p<0.0001 57.33 p<0.0001

G2 2.32 p<0.0001 14.73 p=0.0015 47.37 p<0.0001

G3 2.33 p=0.0014 6.29 p<0.0001 37.65 p<0.0001

164
In Table 34, significantly diverse rates are listed across tests in the results obtained

from all PL groups.33 However, it was interesting to notice the dissimilar routes of

cross-group progression in T1 and the two other tests: while the production of

indefinite article increased as learners’ overall English improved in the writing task,

the error rates correlated negatively with PL in the cloze test and the GJT, i.e., fewer

overuse errors were made as competence levels rose.

Finally, it is worth mentioning that task type did not influence learners’

choices in oversupplying the indefinite article more frequently with uncountable NPs

whereas error rates in countable plural contexts remained consistently lower.

4.4.2.2 Omission

4.4.2.2.1 The

The omission of the definite article was considered an error of choice in T1, a possible

grammatical or choice error in T2, depending on the number of responses by CG

members, and a grammatical error in T3 (see chapter 3 section 5.6). Nevertheless, in

reviewing omission patterns across tests, these differences in categorisation were

temporarily ignored. It was found that T2 accrued the lowest error rates, significantly

below the rates observed in T1 and T3. The highest omission errors however were

committed in T1, except for the intermediate group whose highest rates were recorded

in T3.

33
The percentages shown in table (34) represent overuse instances with both plural and uncountable
nouns.

165
Table 35 The omission of the definite article across the three tests

T1 T2 T3
T1 v T2 T2 v T3 T1 v T3
% % %

G1 24.12 p<0.0001 7.88 p<0.0001 30.67 p=0.2853

G2 16.72 p<0.0001 7.34 p=0.5753 18.95 p=0.0012

G3 12.22 p=0.0002 4.07 p=0.4834 5.88 p=0.0602

The p values in Table (35) indicate that while the differences between T2 and the

other tests were significant, the number of errors in learners’ composition (T1) and the

GJT did not vary greatly. In all tests, occurrences of definite article omissions were

reduced with improved PLs.

4.4.2.2.2 A(n)

The failure to supply the indefinite article in obligatory (indefinite singular) contexts

was greatly affected by task type. Similar to the pattern of definite article omissions,

the highest omission rates of a(n) were incurred in T1 by all participants. On the other

hand, the lowest omission rates were recorded in T2. In other words, a(n) was better

supplied in the blanks test while it was more difficult for learners to supply a(n) where

necessary in their written composition.

166
In this regard, the figures prove that all PL groups performed significantly

better in T2 than in T1 and T3 as fewer omission errors were made in the blanks test

than in the other two tests, as Table (36) shows.

Table 36 Omission of the indefinite article across tasks

T1 T2 T3
T1 v T2 T2 v T3 T1 v T3
% % %

G1 39.62 p<0.0001 11.30 p<0.0001 37.78 p=0.7915

G2 25.86 p<0.0001 4.46 p=0.0443 16.67 p=0.0030

G3 10.94 p=0.0442 2.99 p=0.0129 10.78 p=0.9754

Despite the significant cross-task differences in omission errors, the results from all

tests confirm that rates were consistent with PLs, decreasing regularly and

systematically as learners’ overall English competence improved.

In most cases, more omissions of the indefinite article were made than the

definite across tests except for a slightly higher overuse of a(n) by the intermediate

group in the cloze test. It is worth mentioning that the delay in production of the

indefinite article is not as noticeable in certain syntactic formulae such as memorised

chunks learned at earlier stages of L2 acquisition. Existential and have constructions

show higher accuracy rates of the indefinite article especially in T1.

167
4.4.2.3 Replacement errors

The purpose of studying learners’ patterns of replacement of one article by the other is

to find out if learners’ preferences of article use were affected by task type. The results

in this section do not include bare nominals or what is termed as the zero article.

Instead, only countable singular nouns were included in this comparison keeping

definiteness as the only variable to balance the number of possible replacement

occurrences between the two articles. Thus, the results provided in this section (the-

for-a and vice versa) reflect choice, rather than grammatical errors.

The lowest the-for-a errors were recorded in T1. The rates were significantly

lower than those observed in T2 and T3. This is true for all PL groups. The higher

rates of replacement in the cloze test and the GJT were not considerably different from

one another (see the percentages in Table 37) and therefore no significant differences

were incurred between T2 and T3.

Table 37 Replacement of the indefinite article with the definite (the-for-a) across tasks

T1 T2 T3
T1 v T2 T2 v T3 T3 v T1
% % %

G1 4.72 p<0.0001 24.78 p=0.3724 31.11 p=0.0002

G2 3.45 p<0.0001 17.39 p=0.7450 14.04 p=0.0226

G3 3.13 p<0.0001 15.08 p=0.9085 15.69 p=0.0233

A-for-the replacement errors were significantly higher in T2 and T3 while not many

learners chose to supply the indefinite article in definite contexts in their writing. The

168
substitution of the by a(n)34 in T1 was minimal, not exceeding 0.9% by the weaker

group, 1.2% by the intermediate and 3.4% in the production of more advanced

learners. Although these rates are certainly low, there was a gradual increase in error

rates with improved English levels. In this sense, G3 learners made more errors of

overuse than the two weaker groups. This however was not the case in the blanks test,

since error rates dropped with PL progress. Participants of lower and intermediate PLs

accepted replacement errors more readily in T3 than they did in the writing test,

scoring significantly higher rates but it was interesting to find out that the advanced

group did not mark any erroneous a-for-the NP as correct in T3. Therefore, task

affected the results when the rates of the free writing were compared against those in

T2 and 3, but no statistical variation was proven between the GJT and the stories task

in the results of all PLs. A breakdown of cross-task significance rates is available in

Table (38).

Table 38 Errors of Substituting definite by the indefinite article (a-for-the) across tasks

T1 T2 T3
T1 v T2 T2 v T3 T3 v T1
% % %

G1 0.90 p<0.0001 25.62 p=0.9275 26.67 p=0.0245

G2 1.24 p<0.0001 14.74 p<0.0001 5.26 p=0.1547

G3 3.45 p=0.2041 5.71 p<0.0001 0.00 p=0.0418

34
These are Choice errors as the indefinite article is supplied with singular countable but possibly
definite nouns.

169
By comparing the replacement rates of the two articles, most results indicated that the

definite article was the more favourable option, while a-for-the was the exception to

the rule. Even in cases where the indefinite article was preferred, the rates remained

extremely low. In tasks 1 and 3, for example, weaker and intermediate participants

preferred the definite article to the indefinite, but in T2, instances where a(n) replaced

the were slightly more frequent (1%). The error map of more advanced learners was

different to some extent. Although participants in this group chose the-for-a more

often in the cloze and GJTs, in T1, instances where the definite was substituted by the

indefinite were a fraction higher (0.4 %) than the rates of using the definite article to

replace the indefinite. Both types of replacement errors in T1 were statistically lower

than their counterparts in the two other tasks as is shown in the Tables 37 and 38.

Finally, it is worth mentioning that, with regards to definiteness, overuse errors

were generally higher than replacement errors.35 In other words, the supply of the in

(indefinite) plural/uncountable contexts was significantly more frequent than

occurrences where it replaced the indefinite article. This fact was more evident in the

weaker group’s production across tasks with significant difference rates between the

two errors. The results from G2 echoed the choices of the weaker group but the

difference between the two errors was less noticeable in T2 and. In T3, overuse and

replacement errors were almost equal, with a small difference of 1.7%. G3 in T1

behaved like the other groups, but the difference between replacement and overuse

errors was insignificant.

35
Both types of errors refer to the unnecessary supply of the definite article in indefinite contexts.

170
4.4.3 Learners Hypotheses

In this section we report learners’ responses to notions of specificity, premodification

and ontology in three tests to find out the effect of task type on their performance.

4.4.4.1 Specificity

Results emerging from all three tasks show that participants, especially the two lower

PLs, overused the definite article more frequently in (indefinite) non-referential

contexts than in (indefinite) specific ones. The overuse rates of the definite article in

Specific contexts were the lowest in T1. This is no surprise as all overuse errors in the

writing task were significantly fewer than in the other two tests. A breakdown of the

percentages and P values across tests is presented in Table (39).

Table 39 Overuse of the definite article in specific contexts

T1 T2 T3
T1 v T2 T2 v T3 T1 v T3
% % %

G1 9.03 p<0.0001 31.40 p=0.9676 31.11 p=0.0025

G2 10.78 p<0.0001 21.67 p=0.0160 13.16 p=0.6057

G3 6.87 p<0.0001 18.96 p=0.0980 12.75 p=0.1392

T1 yielded significant differences between the overuse rates of the in (indefinite) +/–

Sp contexts. In T2, however, the difference between the errors made in the two

contexts was not statistically significant although overuse rates were higher in [–Sp]

environments. In T3, the most significant difference in error rates of specificity was

evident in the results obtained from G1 while little variation was noticed in G2’s

171
responses. What is most interesting is the shift noticed in the performance of the

advanced group in both cloze test and GJT as they overused the definite article more

frequently in specific, rather than non-specific contexts. However, the difference

between the ‘higher’ overuse rates in [–Sp] contexts is not significant in both tasks.

There were sporadic instances when learners opted to use the indefinite article

in non-specific NP environments. This type of overuse was minimal in the free writing

test, but with a progression map opposite to that of the definite. In other words, while

overuse rates of definite decreased with improving PLs, the proportions of a(n)

production grew as PLs advanced (see section 4.1.3.1). Non-referential employment

of a(n) increased in T2 with significantly higher rates than those observed in T1

(p<0.00001), while the highest [–Sp] recurrences were recorded in T3, as all

participants accepted the uncommon use of generic a(n) more readily than in the other

two tests (G1 p<0.0001, G2 p<0.0001, G3 p<0.0001).

4.4.4.2 Premodification

The results indicate that the supply or the failure to supply the indefinite article was

affected by the presence/absence of a premodifying adjective in the NPs. As discussed

in the previous sections, the omission of a(n) was more frequent in non-premodified

contexts in the results of weaker learners whereas more advanced learners omitted the

indefinite article (or accepted its omission) more frequently in modified constructions.

Task variation affected omission patterns in –/+Premd NP environments.

Participants of all proficiency levels omitted the indefinite article in both NP

constructions the least in T2. The highest rates of omissions by the weak group were

scored n T3 while the two upper groups omitted a(n) most frequently in T1. Cross-

172
task variation was more noticeable in the results of weaker learners as statistically

significant differences were found between T2 and T1 in G1 and G2’s production, but

test type had a smaller effect on more advanced learners.

Figure 13 The omission of a(n) in +/-Premodified constructions across tests

Figure (13) illustrates the decrease in errors as PLs improve, as well as the reduction

in the gaps between the results of different tests by each group. More importantly, the

chart demonstrates how much lower the omission rates are in T2 than in the other two

tests, while the proportions in T1 approximated those reported in T3 in responses from

the weaker and the more advanced groups.

It is also interesting to observe how the choices of each group alternate between –

/+Premd NPs in accordance with task type especially by G2 participants. In T1, the

intermediate group for example made more omissions in Premd contexts while in T2

and T3, the same group omitted the indefinite article more frequently in non-

173
premodified contexts. The effect of task type was less influential on the more

advanced learners. In all three tasks, G3 made more omissions before premodified

nouns than non-premodified.

4.4.4.3 Concreteness

Task-type did not have a large affect on learners’ inclination to oversupply the definite

article more often with abstract than concrete nouns without exception. In this regard,

there were no inconsistencies throughout PLs. However, the influence of test variation

can be noticed in the increase of errors in more controlled tests. The highest overuse

rates were recorded in T3 where learners mistook the incorrect instances of overuse as

correct especially with abstract nouns. The second highest overuse rates were scored

in T2 although the difference between the errors in the two ontological contexts was

not statistically significant. On the other hand, T1 yielded the lowest overuse rates

among all tasks, as evident in the cross-task significance rates presented in Table 40.

Table 40 The overuse of the definite article in abstract contexts across tests

T1 T2 T3
T1 v T2 T2 v T3 T1 v T3
% % %

G1 19.00 p=0.0075 31.52 p=0.0055 53.33 p<0.0001

G2 12.03 p<0.0001 25.68 p=0.1702 26.32 p=0.0012

G3 9.86 p=0.0026 22.44 p=0.0684 35.29 p=0.0008

174
Despite the differences in error rates across tasks, participants’ choices remained

unchanged in the sense that the overuse of the remained significantly more recurrent

in abstract than in concrete domains (see Figures 14 and 15).

Figure 14 Overuse of the definite article by G1 in abstract and concrete contexts across tasks.

Figure 14 shows that the overuse instances in the weaker group’s results were more

frequent in abstract environments. The chart also demonstrates that error rates

increased steadily with the increase in the degree of control in each test, reaching its

peak in T3.

Advanced learners’ choices were similar to those of the weaker group. In fact,

the gap between overuse errors in abstract and concrete contexts increased with

improved PLs. Furthermore, the tendency to oversupply in T2 and T3 remained

unchanged, as is shown in Figure 15.

175
Figure 15 Overuse of the definite article by G3 in abstract and concrete contexts across tasks

Finally, this chapter presented the findings of this study in numbers and percentages

and, where the differences were significant, P values. The results were compared

across groups and tasks. These included correct and erroneous instances of both

articles as well as the zero article in learner data. The following chapter offers a

detailed explanation of possible interpretations of these findings.

176
Chapter Five: Discussion

This chapter discusses the findings reported in the Results chapter and attempts to

present possible explanations of the causes that underlie learners’ L2 article use. The

section considers the stages of article acquisition as learners progress from one PL to

the other; the effect of task variation on learner performance within each group and

possible learner hypotheses in which an association is made between certain linguistic

notions and definiteness/indefiniteness.

The occurrence and patterning of responses from participants will be presented

within the larger SLA perspective and compared with findings obtained from other

studies on L2 acquisition of English articles in particular. This will cover both types of

L1 backgrounds: those which possess and those which lack an article system.

The content of this section is divided into three main parts. The first part

addresses the progress of acquisition of English articles by Arabic speakers. In this

part, the results are closely examined and analysed cross-sectionally based on the

broad changes noticed in article use and preference by different proficiency level

groups, without the particular digressions that result from task type variation. In the

second part, the effect of task type on learners’ results is discussed by looking at the

changes in the performance of each PL group across the three tests. These effects will

be analysed qualitatively according to the type of knowledge that each task examines

as well as quantitatively in numbers, percentages and significance rates. The third and

final part reviews the possible hypotheses that might be underlying learners’

performance and affecting their choices of L2 articles. This will be presented with

177
regard to the associations which learners make between certain linguistic notions on

the one hand and definiteness/individuation on the other.

5.1 Stages of acquisition

In this section, we follow the participants’ varying patterns of article use as learners

progress from beginner levels to higher stages of upper-intermediate and advanced

proficiency levels, by considering the correct/incorrect responses of article use across

tasks. Learners’ performance will be reviewed in general as is common in each group,

excluding the atypical outcomes caused by test type.

Since this study is a cross-sectional one, it could be easily assumed that the

higher the PL of the group is, the fewer the errors learners will make, while accuracy

rates will rise. This principle is largely true but does not categorically apply to all

findings. Several investigators, such as Sang et al. (1986) and Milanovic (1988 cited

by Bachman 1991) concede that the structure of production changes as the level of

proficiency changes. In this sense, it is possible for learners’ of a lower PL to perform

better on a certain aspect of the second language. This is so because the learning

process is a series of continuously changing assumptions that learners subconsciously

make on particular domains of IL.

5.1.1 Accuracy

In this study instances of appropriate supply of the definite article were higher than

overuse or replacement error rates across tasks and across PL groups.

178
5.1.1.1 G1

Generally, G1’s performance on the definite article was better than on the indefinite in

all three tests, but this was especially true for participants in the two lower PL groups.

On the surface, the average accuracy rate of the definite article in all three tasks (75%)

appears higher than originally estimated for participants of similar PLs, and indeed

higher than definite article accuracy rates reported in other studies (Ekiert 2004, Thu

2005, inter alia). However, if the flooding stage, the dramatic rise in the use of the

definite article, (Huebner 1983, see also section 2.3.1) is taken into account, it casts

doubt on how controlled the EFL learners’ production of the definite article is, even

when it approaches native-like standards. It is likely that at least some of the correct

uses are caused by uncontrolled overuse of the form, rather than being a result of

careful consideration on the learners’ part.

The results of this study suggest that the stage of flooding in the L2

development of the definite article is concurrent with the stage of lower supply/higher

omission of the indefinite article. This was most noticeable in the weaker group’s

performance, which suggests that this study on university students started at the same

stage where Kharma and Hajjaj’s (1989) finished. The results from G1 are similar to

the description of article use by the high-school-graduates in Kharma’s study.

Therefore, it seems possible that by the end of high school/beginning of university

education, L1Ar learners find the definite article by far the easier to master, with

learners scoring the highest accuracy rates, while the indefinite article is the most

difficult, with learners scoring the lowest accuracy rates.

179
One possible explanation for this phenomenon is the different functional

categories that need to be acquired in the case of the two articles: cardinality for a(n)

and definiteness for the (Lyons 1999). It would be easier to acquire the definite article

because it directly indicates definiteness and can occur with both singular and plural

nouns. However, a(n) only indirectly indicates indefiniteness and can only occur with

singular nouns which is a grammatical requirement. Number is not the only decisive

element, the mass vs. count noun distinction is also implicated. Therefore, it is not

surprising that the stage of encoding definiteness precedes that of indefiniteness

(Chaudron and Parker 1990) and mastering the functions of the indefinite article

seems to be more difficult than those of the definite (Avery and Radišić 2007).

Another possible explanation of the higher accuracy rates of the definite article

is the type of error involved in using each article. Errors concerning the definite

article, whether of omission or overuse, do not contravene the grammaticality of an

utterance while misuse of the indefinite article, i.e. its presence with plural or

uncountable nouns (overuse) or its absence in singular indefinite contexts (omission)

renders the NP ungrammatical. Hence, it is more likely that the definite article would

accrue lower error rates where grammaticality is concerned. However, despite being

grammatically permissible, misuse of the definite article can be unacceptable or non-

native like if employed with non-uniquely identifiable referents because the referent

might become ambiguous to the recipient.

From an L1 transfer perspective, the fact that indefinite NPs in Arabic are

morphologically unmarked can present a valid justification for the late emergence and

mastering of the indefinite article, especially in meaningful interaction. It is generally

accepted that while the ease of acquisition depends on the presence of a marker in

180
both first and second languages (Odlin 2003), features unavailable in learners L1 take

longer to acquire (Master 1997) because the process of mapping a feature that is not

present in learners’ L1 onto their IL structures requires more effort and would

therefore be more difficult (Goad and White 2004). Consistent with our expectations,

the accuracy rates of the indefinite article remain low until proficiency levels improve.

5.1.1.2 G2

The accuracy of the definite article remained higher than that of the indefinite in the

intermediate learners’ production. This result is not exclusively caused by the

presence of the definite article in learners’ L1. Except for two known studies (Leung

2001, Young 1996), most researchers tend to agree that mastering the definite article

precedes that of the indefinite regardless of the learners’ L1 background. This

supposition is corroborated by findings from many studies administered on L1Ar

learners (Kharma and Hajjaj 1989, Maalej 2004, Hamdallah 1988) as well as studies

on other L1s including –ART languages (Hakuta 1976, Huebner 1983, Master 1987,

Thomas 1989, Zdorenko and Paradis 2008, Yamada and Matsuura 1982) which hold

that definiteness, as a semantic concept, is at least encoded before indefiniteness

(Chaudron and Parker 1990). Therefore, it seems that the higher accuracy rates of the

definite article compared to those of the indefinite, especially at earlier stages, is a

common feature in SLA studies regardless of the L1.

With PL improvement, the production of the indefinite article increased, which

led to an increase in the accuracy rates. This was not only noticed in the production of

participants in this study, but also in studies on both ±ART L1s (Huebner 1983,

Yamada and Matsuura 1982, Kharma 1981 inter alia). Due to overall language

181
improvement, the overuse of the definite article with non-uniquely identifiable

referents was reduced. In T3, for instance, the accuracy rates of both articles were

almost identical which suggests that under different conditions, such as higher focus

on form, the use of the indefinite article may improve. This was true in T2 as well

whereby the indefinite article was better supplied than the definite by the same PL

group. It is not surprising that G2’s understanding of the indefinite article is better

than that of G1, but the accuracy rates are also largely affected by task-type variation.

This will be further elaborated on in the following section.

Many researchers in the field of L2 article acquisition corroborate the

supposition that mastering the indefinite article follows that of the definite in IL

development (Master 1987, Thomas 1989, Snape and Velasco-Zarate 2005, Trenkic´

2007) regardless whether the L1 possesses or lacks articles. Avery and Radišić (2007)

ascribes the incompatible rate of acquisition to the fact that more time is required to

better understand the cardinality function of the indefinite article in addition to its

semantic value of indefiniteness. Cardinality involves awareness of singular/plural as

well countable/uncountable distinctions.

5.1.1.3 G3

As learners proficiency levels improve, so do their accuracy rates on the indefinite

article, approaching those of the definite, and perhaps higher, such as in the results

from T1 and T3. However, unlike the results from the two lower groups, there was

little difference in the accuracy rates of G3’s performance on the definite and

indefinite articles. The non-significant rates in the results of this group are interesting

because they mark a shift in production tendencies. The lower group’s accuracy rates

182
on the definite article were higher than on the indefinite because of the reasons

discussed above, yet not significantly better. Therefore, the learning curve, according

to available accuracy rates, seems to start with higher awareness and a better supply of

the definite article, but this tendency however seems to gradually change, as PLs

improve, towards a better conceptualisation of the indefinite article. Thus the gap

noticed in correct use of the two articles at earlier stages becomes smaller as the L1

influence decreases and the newly acquired indefinite marker is incorporated into

learners’ IL.

The delayed production and mastery of the indefinite article as compared to

the definite in L2 English has been often reported on in regard to learners from

various L1 backgrounds (e.g. Huebner 1983, Parrish 1987, Thomas 1989, Master 1997

Chaudron and Parker 1990, Garcia Mayo 2008). Thus, it is conceivable that the

earlier command of the definite article/ the later command of the indefinite article are

attributable to developmental trends. This view is shared by Arab researchers such as

Abi-Samra (2003) and Batainah (2005) who hold that L1Ar learners’ performance is

largely in line with that of learners from other L1s whilst the negative L1 transfer does

not account for more than a small proportion of their errors. Nonetheless, according to

a report for the UNRWA programme in Jerusalem, Habash (1982) postulates that

most errors committed by Arab learners of English are caused by negative interference

from L1Arabic rather than by any other learning difficulties.

5.1.1.4 Correct zero marking across groups

Judging by the accuracy rates of the marking of bare nominals, or the zero article, our

results go to prove that it was most difficult for L1Ar learners to leave indefinite

183
plural and countable NPs in non-referential contexts unmarked. The low accuracy

rates of the zero article was one of the most prevalent and observable facts across

tasks, especially for lower PL participants. The tendency to overuse articles in these

contexts confirms the findings of well-known studies on L1Ar learners such as

Hamdallah (1988), Maalej (2004), Sarko (2008) and Batainah (2005). However, such

error is rarely established as common in many studies on SLA of English articles,

since researchers have often found that the failure to supply articles, rather than

overuse them, whether in spoken or written discourse, was the most noticeable error

that persisted well into the production of advanced 2LL (Jarvis 2002, Yamada and

Matsuura 1982, Master 1987, Thomas 1989, Parrish 1987, Ekiert 2004). However, the

fact that these studies were administered on –ART L1 learners must be taken into

consideration because presence or absence of an article system in the first language

has been known to affect L2 article use (Schachter 1983, Master 1997, Parodi et al

1997, Goad and White 2003) and, in effect, learners’ order of acquisition.

If the discrepancy between our results and those from other studies is analysed

contrastively (CA), the lower accuracy rates of the zero article, or learners’ inability to

recognise that non-referential indefinite plural/countable NPs should be left unmarked,

are probably due to the difference in function between the definite articles in the

native and target languages. So while the main function of the definite article in

English is to identify, the function of the definite article in Arabic is to generalise

(Kremers 2003) as well as to uniquely identify (see Table 2 in 2.3.3). The overuse of

the definite article in non-referential contexts where no article should have appeared

can be therefore drawn from lexical associations of definiteness carried over from the

L1 (Snape et al 2006).

184
The weaker group produced very few bare nominals in the free writing task,

overused both articles in T2 and failed to recognise overuse instances in T3. Therefore

their accuracy rates of the zero article were quite low. With improved PLs, more bare

(plural and uncountable) NPs were produced (T1), overuse errors of both articles

dropped (T2) and there was a better recognition of overuse instances (T3). Thus the

accuracy rates of the zero article improved.

Accordingly, it seems that learners at more advanced PLs, start approaching

native like standards in their preference to employ the zero article in generic contexts

despite the grammatical acceptability of both articles to fulfil the same function.

Dropping the definite article from non-referential contexts seems to be one of the most

difficult aspects of L2 article acquisition for L1Ar learners to master and is not likely

to take place until later IL stages. Finally, the results seem to suggest that learners

responded better to number, a feature available in both their first and TL, than they did

to countability which bears some divergent qualities. This is extended to the

performance of advanced learners who performed better on plural than non-countable

nouns.

5.1.1.5 Conclusion to Accuracy

Based on the results obtained in this study, it is evident that learners found it easier to

accurately produce the definite article than the indefinite. Therefore, in line with the

findings of most studies of articles in SLA , regardless of the L1, it is clear that the

acquisition of definiteness precedes that of indefiniteness. This is perhaps the case

because of the different functions involved in the correct use of each article. The

disparity between the results of this study and others remains in the order of

185
acquisition of the zero article; whether it precedes or follows that of the indefinite. In

other words, which concept is easier or more difficult for learners to command? Is it

the fact that most bare nominals, in plural or mass form, used generically should be

left unmarked in English, or that of the need to place a(n) before a singular countable

noun regardless of the type of reference? Our results show that learners found it easier

to master the indefinite article than the zero. Therefore, according to the accuracy

rates in this study, the order of acquisition of articles is: The, a(n), zero. On the other

hand, some researchers have reported that the order of acquisition of L2 English

articles was zero, the, a(n), (Master 1987, Parrish 1987, Thomas 1989, Liu and

Gleason 2002) while others such as Lightfoot (1998) hold that the order of acquisition

is the, zero, a(n).

These findings should only be considered as an indicator for future research

since variable approaches in data collection and/or methods of analysis were followed.

For instance, the higher accuracy rates of the definite article could have also been

caused by the different methods through which the correct use of each article was

measured. The correct use of the definite depended on the writer’s assumptions about

the reader’s knowledge of the referent or their ability to find that referent. Therefore,

most uses could be judged as appropriate, or at least grammatically acceptable,

although they might not be what the NS would have preferred. Such productions of

[+def] NPs were categorised under choice errors rather than absolute errors. On the

other hand, the correct use of the indefinite article depends on multiple grammatical

elements in addition to the semantic parameter; a fact which narrows the range of

acceptability according to function and would result in lower accuracy rates.

186
5.1.2 Errors

This section attempts to provide a possible rationale for learners’ errors in article use.

Misuse of English articles generally includes errors of omission where an article

should have been supplied in obligatory contexts; errors of overuse, where no article

should have been used, and finally replacement errors when one article was

erroneously substituted by the other. The results discussed in this part reflect the

general error patterns as observed in the results of each PL group, regardless of the

exceptions in the outcomes that resulted from task type variation.

5.1.2.1 Omission

Omission is the technical term for the failure to voluntarily supply articles in

obligatory contexts (T1), the incorrect article choice (T2) or the inability to recognise

that an article is missing (T3). In general, this type of error can be both grammatical,

such as the absence of the indefinite article with indefinite countable singular nouns,

or an error of choice, such as the omission of the definite article when its presence or

absence does not affect grammaticality. As discussed earlier, definiteness depends on

the speaker’s assumptions of the hearer’s ability to uniquely identify a referent in a

pragmatic set, which could vary from one situation to the other. However, if the

referent is felicitously accessible to the hearer, providing the definite article can

become obligatory under certain conditions (see 2.2.1). The criteria that we used to

determine the obligatory supply of a given referent in T1 were based on Hawkins’

(1978) description of definiteness as uniquely identifiable and Liu and Gleason’s

(2002) categories of obligatory contexts in which non-generic the operates, namely

physical and textual availability as well as situational or universal uniqueness. In the

187
controlled tests (T2 and T3), obligatory supply was determined by responses from the

majority of the NS control group (For further detail, see 3.5).

Our results show that in all three tests, the weaker group invariably omitted the

indefinite article more frequently than the definite. The results are similar to

Trenkic´’s (2000) as the indefinite article was omitted significantly more often by L1

Serbian learners, while the definite article was generally better supplied. This is an

expected behaviour because of the stronger L1 influence in earlier learning stages

(Berman and Olshtain 1983, Slabakova 2000, Mahmoud 2002, Sharma 2005 inter

alia) drives the non-obligatory markedness of indefiniteness in Arabic to be negatively

transferred into the TL. Eckman (1977) posits that L1 structures that are different from

L2 structures, especially if (typologically) less marked will be transferred to the target

language. Therefore it is not surprising that with both criteria present in Arabic, L1’s

effect is perhaps stronger.

Where indefiniteness is concerned, Arabic, similar to Hebrew and Bulgarian,

does not categorically fall under either +ART or –ART languages because it possesses

an obligatory definite article but does not enforce the supply of its indefinite

marker(s). In this sense, it is possible to attribute the higher omission rates of a(n) to

the absence of a overt and compulsory marking of indefiniteness in the L1. This

proposal corroborates Prévost and White’s (2000) MSIH36 whereby learners

experience difficulty in mapping abstract syntax onto surface morphological forms

(cf. Haznedar and Schwartz 1997, Lardiere, 1998, 2000, White 2003).

High omission of the indefinite article is often reported by researchers on SLA

of English articles by L1 Arabic learners, such as Batainah (2005), Maalej (2004) and

36
Missing Surface Inflection Hypothesis

188
Kharma (1981). The omission of the indefinite article can be an indication of the way

learners’ demonstrate their conceptualisation of indefiniteness by the absence of the

definite article.

In computerised corpus analysis, indefiniteness can be negatively defined since

‘the absence of the itself is a mark of indefiniteness’ (Leech 2006:13). This argument

stands as another possible explanation of the failure to supply the indefinite article in

obligatory contexts as its presence might be considered by some students as

(communicatively) redundant. In this sense, leaving a referent unmarked, such as the

case in Arabic and article-less languages, signals the existence of the ‘primary

counterpart’ of definiteness, i.e. indefiniteness (Arabski 1979). Goad and White

(2004) propose that ‘failure to consistently produce overt morphology cannot be

attributed to failure to represent the relevant morphosyntactic features’ (p. 125

emphasis added) because it may appear under different conditions, such as in forced

elicitation tasks. Hence, failure to produce the indefinite article does not necessarily

imply that learners are not aware of indefiniteness as a concept. Jarvis (2002) believes

that the higher omission of articles by Finnish learners does not necessarily represent a

‘simplified register’, rather, ‘it represents the L1 Finnish convention of avoiding…

redundant markers … when these properties of a NP are already salient in a given

discourse context’ (416).

It is possible that the difficulty L2 learners have with supplying the indefinite

article has pragmatic origins. According to the recoverability principle proposed by

Givón (1983), the most continuous referents need the least marking, i.e. there is lesser

need to overtly mark NPs for (in)definiteness if the information encoded in the NP is

recoverable from the context. The information being that of [–def] represented by the

absence of a definite form and +Sing represented by the absence of a plural marker

189
(s). In learners’ assumption, the provision of the indefinite article becomes redundant

because the underlying assumption is that the speaker is being ‘as informative as

required’ (cf. Searle’s 1969 Speech Act Theory and Grice’s 1975 quantity maxim).

Thus, higher omission rates of the indefinite article can be seen to follow. This line of

argumentation is also adopted by Robertson (2000) in analysing L2 article use/misuse.

Robertson (2000) holds that if the encoded information is recoverable from context,

the article may be dropped when it is within the scope of an immediately preceding

co-referential NP. He terms this as the Pro drop tendency in SLA.

The non-contrastive approach to error analysis is followed by many

researchers in SLA such as Richards (1971), George (1971), Abisamra (2003),

Batainah (2005) and Zdorenko and Paradis (2008) who ascribed most L2 article errors

to factors beyond L1 influence. Thus, an alternative explanation for the high omission

rates of the indefinite article in L2 production can be the multiple functions that this

article entails (Avery and Radišić 2007). This is supported by a good number of

studies on –ART L1s which report higher omission rates of the indefinite than the

definite articles. In a study on Japanese learners, for example, Yoon (1993) observed

an overall low percentage of indefinite article supply in obligatory contexts. Hence,

higher omissions of the indefinite article in the primary stages of IL becomes

comprehensible once the added difficulty of the cardinality function is acknowledged.

L2 Learners, regardless of L1 background, should be aware of the intricate relation

between the semantics of a given noun and the morpho-syntactic representation of its

singularity/ plurality on the one hand and its countability status on the other. The

problem becomes more intricate when we consider the fact that most nouns cannot be

simply categorised into either countable or uncountable. Rather, the learner has to be

familiar with the range of NP environments in which a shift in the countability status

190
can occur as result of semantic, rather than structural effects of the contexts (Allan

2001). The higher processing effort that the indefinite article places on learners

inevitably leads to reluctance in production. If given the choice, learners might resort

to economy of effort (Richards and Sampson 1974) and simply prefer the definite

article to the indefinite in their production.

With improved overall proficiency of English, the learners’ production of the

indefinite article increased and thus fewer omissions were committed. As a result, the

gap between the omission errors of the two articles became smaller with learners

beginning to realise the necessity to employ the indefinite marker. At a later learning

stage, as is evident from the results of G3, there was an increase in the omission rates

of the definite article. So while the omissions of the indefinite article were

considerably reduced in G3’s results, there was an increase in the omission rates of the

definite article in obligatory contexts. This is certainly an irregular pattern, since the

error rates and PLs correlate positively. Yet such tendencies are not out of the

ordinary if the accumulative effects of factors such as over-generalisation of new rules

and repeated classroom instruction (cf. Selinker 1972) are taken into account.

Learners might have been repeatedly instructed by their EFL teachers not to overuse

the definite article. Having learned that the definite article in English does not refer

attributively quite as often as it does referentially, learners tend to extend this rule to

unsuitable contexts (Ellis 1985) causing higher omission rates. According to Ellis, this

is a form of hypercorrection.

The omission lines of the two articles seem to evolve in almost opposite

directions. Learners start with a low production / high omission stage of the indefinite

article while the omission rates of the definite are quite low. This stage is followed by

191
a decrease in the omission rates of the indefinite article, thus closing the gap between

the omission rates of the two articles. The higher group’s performance is characterised

by a drop in the omission rates of the indefinite article while the error rates of the

definite article rise. However, even if there were more errors made by higher PL

learners and the overall pattern of development of the two articles is reversed, it is a

sign of progress in the learning continuum because these changes reflect the

development of learners’ hypotheses (Selinker 1972, Ellis 2004) with regards to the

functions of the English articles.

5.1.2.2 Overuse

The other type of error known to be committed by L2 learners in article use is the

suppliance of articles in NP environments where they are not required. The term

‘overuse’ covers the misuse of either article with NPs that are usually left bare (Ø

article) by most native speakers. This definition distinguishes overuse errors, where

either article is overused with bare nouns; replacing the zero article, from replacement

errors, where either article replaces the other. The latter will be discussed in the

following section.

As discussed in chapter 3 section 5.3, overuse instances can be either

grammatical or choice errors, depending on the article that is overused. Firstly, the

overuse of the indefinite article with uncountable or plural nouns is ungrammatical in

any context and is an unacceptable option regardless of meaning and communicative

intent. On the other hand, the overuse of the definite article does not affect

grammaticality, as definite descriptions depend largely on the speaker’s perception of

the hearer’ ability to uniquely identify a referent and therefore this type of article

192
misuse is often considered as a lexical error of choice by many L2 researchers such as

Trenkic´ (2000), Nagata et. al. (2005) and Thu (2005). We followed this rule in

categorising learners’ errors in T1, applied it when only a minority of the CG chose to

use the definite article in a given blank (T2) or accepted it as a correct option (in T3).

For a full overview of the criteria used for categorising error types, please refer to

Section3.5.2.

Learners in the weaker group seem to have overused the definite article

indiscriminately across indefinite contexts. Most of these instances include flooding

the definite article into non-specific contexts regardless of the countability or number

descriptions of the head noun. Although learners’ awareness of particular aspects of

the procedural meaning of the vary (Zegarac 2004), associating the definite article

with non-referential meaning was predominant in the lower group’s production and

continued, to a lesser degree, well into the responses from the intermediate group.

Since ‘the most prominent and most frequent type of device for encoding generic

meaning’ is the indefinite plural (Behrens 2005:283, Partee 2005), this type of use was

marked as an error.

Most known studies on L1Ar learners reported the tendency to overuse the

definite article in indefinite contexts (Maalej 2004, Kharma 1981, Kharma and Hajjaj

1989, Hamdallah 1988, Batainah 2005). In a study on Chinese and Arabic speakers,

Garcia Mayo (2008) found the overuse of the is strongly influenced by learners’ L1s

but improved significantly with better proficiency. The predisposition to overuse the

definite article by Arabic speakers originates from the fact that generic reference can

only be referred to by the definite article /al/ in Arabic.

193
However, the overuse of the definite article at lower PLs is not by any means

exclusive to L1Ar learners. Similar findings were reported in SLA studies on other

L1s, including languages that possess or lack a formal representation of articles

(Thomas 1989, Young 1996, Huebner 1983, 1985, Andersen 1977, Nagata et al 2005).

In a study by Master (1987) on Japanese learners of English, for example, the definite

article was flooded into indefinite contexts although Japanese does not possess an

article system. Few researchers, however, believe that overuse of the definite article is

not of much consequence. Lightfoot (1998), for example, believes that the overuse use

of the in generic meanings was not one of the most noticeable errors in Japanese

learners’ L2 article acquisition. This is probably because overusing the definite article

does not affect grammaticality on the one hand, and can be regarded as a sign of

progress compared to the higher omission rates in the production of learners of –ART

L1s. In our results, the main cause of the low accuracy rates of the zero article was the

constant overuse of the definite article in indefinite contexts, and if this is not regarded

as sufficiently problematic, the overuse of the definite article is at least non-native-

like. Furthermore, what distinguishes the overuse by L1ArL and other learners from –

ART L1s, is that the overuse instances of the definite article by our participants are

non-referential whereas findings from –ART L1 studies suggest that the definite

article was reported to be overused mostly in referential (+SR)37 contexts as learners

apply it to deliver specific meanings in indefinite contexts (Thomas 1989, IKW 2004).

This issue will be discussed further in Section 5.3.1.

As learners progressed, the overuse rates of the definite article were reduced.

The initial stage of the flooding at lower PLs is followed by a more careful use as

learners begin to change their hypotheses towards the target language and feel there is

37
Specific Reference. See Literature review for Huebner’s environments for article use.

194
a lesser need to employ the definite form in non-referential contexts. G3 learners seem

to realise that it is the uniqueness element that lexically distinguishes the English

definite article from its Arabic counterpart, and that its presence in English is

unnecessary unless the referent is uniquely identifiable to the hearer/reader.

It is worth mentioning that overuse of the in generic contexts occurred much

more frequently than generic a(n). This was especially true in the weaker group’s

production. This result confirms that the presence of a feature in the L1 (the definite

article) facilitates the second language acquisition of the same feature (Master 1997,

Ard and Hombarg (1983) cited in McLaughlin 1987, Odlin 2003) while the

acquisition of features that are unavailable in the L1, in this case the indefinite article,

are more difficult to master and are usually acquired at later stages (Master 1987).

Although the overuse instances of the indefinite article were considerably

lower than those of the definite, their effect on the grammatical correctness of the

sentence is more serious. It is interesting to observe the dissimilarity in learners’

tendencies of overuse. Unsurprisingly, the most preferred choice was the definite

article, followed by the indefinite, while the least preferred option was leaving

indefinite (plural and uncountable) nouns unmarked. This pattern of acquisition runs

counter to findings of other studies especially those on learners with –ART L1s

whereby omitting both articles was found to precede the stage of overuse (Master

1997, Huebner 1983). However, in line with many studies, cloze tests are known to

trigger higher overuse than omission errors. It is also noticeable that learners’

inclination to overuse the definite article in indefinite contexts is stronger than the

propensity to do the opposite, which is an established phenomenon in SLA literature

on articles regardless of L1 (Trenkic´ 2000, Thomas 1989).

195
It is also evident that learners from all three PL groups and in all three tasks

overused the indefinite article more frequently with uncountable than with plural

nouns. Countability has been signalled as one of the most common difficulties that L2

learners of English encounter, regardless of the L1. Researchers on –ART L1s such as

Hiki (1991), Yoon (1993) and Snape (2008) have reported similar findings.

One of the possible causes of this problem is that determining the countability

status of a given noun depends critically on the context in which it occurs rather than

on the class of the head noun itself. In other words, nouns can shift from countable to

uncountable and vice-versa depending on parameters other than the simple

categorisation of countable/uncountable. Some nouns, such as money, paper,

experience, shift their countability status (Chesterman 1991) or are of a dual

countability membership (Larsen-Freeman and Celce-Murcia 1999). Allan (1980)

suggests that nouns fall into a continuum of countability contexts, from the extremely

countable to the extremely uncountable (e.g. chair and fear, respectively). Therefore,

when learners are unable to detect the countability status of a given noun, they have

difficulty determining what Langacker (1991) terms as the degree of boundedness of a

given entity; i.e. they are unable to perceive whether or not the referent represents a

separate entity or not. For example, in a study on Japanese (–ART L1) learners of

English, Butler (2002) proposes that the problem of correctly marking uncountable

nouns stems from the fact that ‘the countability detection required for the proper usage

of articles is often found at the whole-NP level’ (p. 467). Countability can be

challenging to the 2LL which impedes correct article use.

A second possibility for the higher overuse rates of the indefinite article with

uncountable nouns is that in the absence of the plural marker, nouns are

196
morphologically similar to indefinite singular nouns. Learners might be more

conscious of the unsuitability of adding the indefinite article before a noun in the

presence of the plural suffix (s/es) than they are with uncountable nouns where no

feature is present to indicate that the insertion of the indefiniteness marker is

ungrammatical. Moreover, the fewer overuse errors of indefinite article before plural

nouns by L1Ar learners can be a result of a higher awareness of plurality marking.

The structural presence of plural markers in the form of suffixal clitics in Arabic

makes the plural marker in English more prosodically salient and easier to identify

than uncountable nouns.

Although there is no explicit indefiniteness marker in Arabic, overuse of the

indefinite article can be L1 induced as higher overuse rates were noticed in cases

where the head noun was of a contrasting L1/L2 countability status; i.e. countable in

one language, but uncountable in the other. For example, higher overuse rates were

observed in the GJT with nouns such as information, advice and work which are

countable in Arabic. On the other hand, there might be a non-contrastive explanation

to this error. Batainah (2005), in a study on Jordanian university students proposes

that the overuse of the indefinite article is due to the IL strategy of avoidance in the

sense that by overusing the indefinite article, learners are trying to avoid making

omission errors. In other words, learners think it is better to supply the article than to

risk making an error of omission. However, I think that this strategy is also indirectly

caused by L1 negative transfer since Batainah’s explanation implies that the stage of

overuse is preceded by a stage of omission during which learners might have received

feedback from their EFL teachers instructing them to supply the indefinite article.

197
Finally, it is possible that learners’ overuse of the indefinite article results from

learners’ perception of number in generic reference. Once learners recognise the

generic reference of a given NP, they apply their own conceptualisation of the non-

referential function which is often represented in the singular form (Hawas 1989,

Fehri 2004) instead of the bare plural/uncountable which is more common in generic

referencing in English (Behrens 2005, Partee 2005).

With L2 improvement, the former tendency is reversed. The higher PL group

overused the indefinite article more frequently than the definite. According to well

known studies in SLA (Selinker 1972, Ellis 1994, Towell and Hawkins 1994, Krashen

1982) introducing a new element, in this case the indefiniteness cardinality marker,

into the learners’ IL can result in over-generalising the newly acquired grammatical

rule at later PL stages. Abi Samra (2003) also postulates that most errors are caused by

an over-application of a new rule.

The irregular developmental pattern whereby overuse of the definite article

decreases with the advancement of PLs while the production of the indefinite

gradually increases is probably caused by a combination of factors. Deciding whether

this outcome is a result of L1 transfer or IL developmental strategies is beyond the

scope of this study. So far the results suggest that elements of both contribute to

learners’ decisions on article use.

If we exclude learners’ correct production of the indefinite article learned

produced as formulaic chunks, especially in have and existential (to be) constructions,

which are usually learned in earlier stages of 2LL, it becomes clear that the emergence

stage of the indefinite article in meaningful communication is delayed at least until the

overall L2 proficiency reaches intermediate levels (cf. Kharma and Hajjaj 1989).

198
From our results, it seems that the very low production/high omission stage of the

indefinite article coincides with the flooding stage of the definite. The definite article

is then gradually dropped from non-referential contexts, while the indefinite begins to

appear in written production, but occurs more frequently in form-focused tests.

Therefore, we may speculate that by the time L1Ar learners start dropping the definite

article from non-uniquely identifiable contexts, they are experiencing a higher

production rate of the indefinite article, which could reach a flooding stage at more

advanced stages. This phenomenon has been recorded by studies on L1Ar learners

such as those of Hamdallah (1988) and Batainah (2005).

The difference in the overuse rates of the two articles was more noticeable in

the production of the weak and advanced groups than in the production of the

intermediate group. While the definite article was overused significantly more

frequently than the indefinite by G1, the advanced group overused the indefinite

article more frequently than the definite. One may conclude that this shift of

tendencies is due to the shift from L1 influenced production at lower levels where the

function of the definite article is extended to include non-referential use, to the over-

application of the newly acquired feature in the TL which is the indefinite article. The

difference in overuse rates was least noticeable in the production of the intermediate

group as their IL stage marked the transitional period whereby both influences were

equally present.

It is clear that the processes of acquiring the two articles develop

incongruently. Kellerman (1977) as well as Chaudron and Parker 1990 have

confirmed that a and the develop independently of each other in learners’ IL. It is

possible to propose that in L1Ar learners’ production, the developmental map suggests

that the two articles progress in inverse directions. In other words, while the overuse

199
of the definite subsides, the production of the indefinite article increases and fewer

omission errors are made.

5.1.2.3 Replacement

Replacement refers to the misuse of one article in a position where the other article

would have been chosen as a more preferable option by a native speaker of English. If

learners are able to assign ‘correct meanings’ to article forms, there should be no

substitution errors, but if they are not able to do so, then substitution errors should

occur, reflecting the meanings that learners have principally associated with article

forms at different proficiency levels. The results of this study indicate that

replacement errors did not occur quite as frequently as expected. In fact this error type

was the lowest occurring of all errors across tasks.

In this section, grammatical correctness is excluded since only countable and

singular forms are considered. Hence, article choice becomes a matter of preference

where the only remaining variable for a given NP is its definiteness status (see

3.5.2.3). Appropriateness or suitability of one article rather than another depends on

the presence or absence of certain criteria that determine whether the definite article is

required or not (see Definite Contexts in section 2.2.1). 38 Therefore, in this study,

replacement or substitution errors are all errors of ‘choice’ rather than grammatical

errors.

Learners of all PL groups preferred to use the definite article in indefinite

singular NP constructions significantly more frequently than the employment of the

indefinite article in singular definite contexts across tasks (except for the advanced

38
These criteria may vary from one test to another.

200
group in the free writing task which will be discussed in detail in the following

section). The preference of the definite article over the indefinite is common in the

production of L1Ar learners of English and has also been reported in various studies

on acquisition of English articles, including those on learners from –article L1

backgrounds (Hakuta 1976, Yamada and Matsuura 1982, Thomas 1989, Huebner

1983, Master 1987, Parrish 1987) while the substitution of the definite article by the

indefinite was rarely reported as the norm. It is most likely that this distribution of

replacement errors is not random; rather, it may be a consequence of an uncontrolled

use of definite constructions while learners go through ‘the’ flooding stage. The

following phase witnesses a gradual decrease of the-for-a replacement errors as

proficiency levels improve.

On the other hand, a couple of studies have reported that the overuse of the

indefinite article relative to the definite was more common. Leung (2001) reported

that Vietnamese learners of English preferred to substitute the definite article with the

indefinite. Similar observations were made by Young (1996) who studied SLA of

articles by Czech and Slovak learners of English. These exceptional outcomes might

be due to certain methods of data elicitation and/or analysis, or particular test types

with different focus and demands. To ascertain which of the two replacement norms is

more common than the other in IL, a larger scale of investigation is required to

provide a solid proposition. However, it is not one of the immediate concerns of this

study.

Higher replacement errors caused by substituting the indefinite article with the

definite can also be L1 induced as the indefinite marking is optional in MSA while the

definite article is more likely to replace the indefinite as it is used non-referentially. In

201
a study on Indian learners, Sharma (2005) attributed the replacement errors to the

influence of the learners’ L1 as the learners made more errors replacing the definite

article with the indefinite. This he attributes to the fact that Hindi has an obligatory

indefinite marker but no definite, as it conveys definiteness, like other Indo-Aryan

languages, via word order and in the case of objects, the use of case marking

(McGregor 1995). Although the nature of the difficulty is the reverse of what it is in

Arabic, the main principle underlying the cause of replacement errors is similar.

So far, the results suggest that L1s appears to have a stronger influence on

learners’ use of the indefinite article while their competence on the definite article

approaches native-like accuracy. Similarities with learners from other L1

backgrounds can be observed in learning strategies, such as avoidance, hyper-

correction and order of acquisition, but error patterns and preferences at each learning

stage seem to be more related to learners’ own L1 influence.

5.2 Task Type Variation

Recall that the task types used in this study were a free production test, represented by

the writing task (T1) and two close tests: contextualised information gaps (T2) and

grammaticality judgement test (T3). The major differences between the tasks lie in the

focus and goal, degree of control, and the kind of knowledge that learners draw upon

to provide the required information.

The results of this study show that the type of task affected the performance of

each PL group to different extents. By observing the differences in accuracy, overuse

202
and omission rates within each PL group, the influence of task variation can be

detected through the divergent patterns of article use.

5.2.1 Review of performance by task

Different tasks tap different types of knowledge (Bialystok 1982, Tarone and Parish

1988, Krashen 1982) and affect the way learners produce the TL. It has been found

that task type could drive production to the extent that one group would project

incompatible, and sometimes contrastive, error maps if tasks were of different focus.

Tarone and Parish (1988) and Skehan (1998) refer to ‘Task induced L2 variation’. The

common view is that form focused tasks generally reflect explicit analysed knowledge

whereas more spontaneous, communicative tasks involve L2 learners’ implicit

knowledge (Krashen 1982, Paradis 1994, Muranoi 2000, Ellis 2004, inter alia).

In a study on L1 Serbian learners of English, Trenkic´ (2000) found that the

weakest group performed better on the task that required least attention to form while

the highest PL group performed better in the task that required the highest attention to

form. Ingram (1985) studied the language of ESL students in Australia and EFL

students in China and reported radically different correlations between communicative

and structure-based tests. Similarly, our results show that participants’ performance

varied according to task type. In addition, diversity in article use was also noticed

across tests. While the definite article was felicitously employed in the task with more

attention to meaning, the indefinite article was supplied more appropriately in tasks

with more attention to form.

203
5.2.1.1 T1

Researchers have found that more accurate production is observed in communicative

tasks than form-focused tasks. In other words, free production tests are known to yield

lower error rates (Bialystok 1982, Tarone 1985, Foster and Skehan 1996, Skehan and

Foster 1999). In general, researchers have attributed better L2 performance to

avoidance strategies that L2 learners are known to resort to (Mizuno 1985, Tarone and

Parrish 1988, Ellis 1994). Through free production, learners have the opportunity to

replace articles with other determiners, pronouns or simply by choosing different

sentence structures can rely on plural forms. In this test, most learners opted for

producing plural/uncountable nouns three times more often than singular nouns in

indefinite contexts.

The highest accuracy rate of the zero article was scored in this test. Learners’

production of bare nominals was most appropriate in T1 than any other test.

Moreover, better performance was observed in the production of indefinite plural

nouns than uncountable nouns across PL groups. From a contrastive approach to error

analysis, plurality marking is morphologically salient in both languages while many

nouns belong to different L1/L2 countability categories (cf. Hawas 1989, Thubaiti

2007).

The argument that meaning-based tasks yield lower error rates does not apply

to omission errors in this test as the highest omission rates of the indefinite article

were made in T1 by all PL groups. The highest omission errors of the definite article

were also made in T1 by the advanced group. With less attention to accuracy and

more focus on expressing thoughts and feelings, it is expected that learners would pay

less attention to form. When compared with the lowest rates of omission in the cloze

204
test and, to a lesser extent, the GJT, T1 seems to have accrued the highest instances of

omission. This lends support to Granfeldt’s (2000) observation that accuracy will

decrease if learners’ attentional resources39 are channelled towards goals other than

accuracy. Furthermore, learners might have found it redundant to mark a nominal

overtly for indefiniteness when its semantic indefinite status has already been inferred

by the absence of a definite marker.

Beside task focus, another factor that contributed to bringing about these

results is the time pressure that learners were under when this task was administrated.

Time pressure adds a processing constraint on participants preventing them from

contemplating the forms they produced (cf. Sorace 1996, Robinson 1996).

With regards to overuse of both articles, the percentages of errors were

radically lower in T1 than in the cloze test and the GJT. The low overuse rate of the

indefinite article with plural and uncountable indefinites does not necessarily imply

better L2 aptitude, competence or awareness of relevant grammatical rules. Instead,

the outwardly correct use of ‘zero’ might well be the outcome of learners’ reluctance

to produce indefinite singular NPs which require the supply of the indefinite article.

The lowest replacement error rates were also recorded in T1 despite a clear

preference of the definite article by the weaker groups and the indefinite by the higher

PL participants. In general, the weaker group seemed to prefer the definite article over

the indefinite as they overused it to replace the indefinite four times as often as they

did the opposite. The ratio was reduced with progressing PLs since the intermediate

group only used the-for-a twice as often as the indefinite to replace the definite. At a

later learning stage, the higher group’s replacement rates became very close, i.e. the

39
For Attentional Resources and task variation see Bialystok and Ryan (1985)

205
difference between the rates of replacing the-for-a were almost equal to those of

replacing a-for-the, with the indefinite article preferred. This is probably a result of

learners’ recently increased awareness of the importance of supplying the indefinite

article. Moreover, this result could have been equally influenced by the receding

influence of L1 represented by the drop in the overuse rates of the definite article

before singular indefinite contexts as the generic use of the definite article with

singular nouns is substantially common in Arabic. Although acceptable in certain

expressions in English (I can’t imagine how life was before the invention of the

telephone; She plays the piano) it is not likely that learners have been sufficiently

exposed to authentic material which would enable them to detect similar uses and

employ them unprompted. The singular NP is certainly not the most widely used form

to deliver generic reference in English; rather ‘generically introduced entities are

conceptually plural’ (Onishi and Murphy 2002:97) and largely unmarked (Behrens

2005). This case could be a typical example of what Smith and Tsimpli (1995) refer to

as misanalysis of L2 material based on the argument that features instantiated in the

L1 could lead to negative transfer.

Oblivious to the purpose of the task, the results from production tests better

reflect learners’ communicative competence (Hymes 1972) in the TL as they would

use it in real-life situations (Lightbown and Spada 1999, Power 2003) and provides

the researcher with a sample of the language used in non-test situations (Skehan

1989).

206
5.2.1.2 T2

Many SLA researchers believe that form-focused tests could, partially, reflect

learners’ proceduralised competence, as the ability to provide the required information

in gaps or judge correctness does not depend entirely and exclusively on the rules that

participants learned through classroom instruction. There are also researchers who

hold that cloze tests can also reflect, to a certain extent, learners’ communicative

ability and implicit knowledge (Ellis 2004, Skehan and Foster 1997, Roehr 2004).

Richards (1976), for example, contends that the ability to supply a missing word in a

sentence depends on the learners’ overall knowledge of the syntax and word

distribution in the TL, and therefore controlled tests are as valid a tool in predicting

competence as any other. In this light, the results from T2 and T3 could also be

employed to measure participants’ knowledge of language use and not only

knowledge about the second language (Roehr 2004). We believe that this is true only

if the cloze test was contextually based on a meaningful content rather than based on

contextless, discrete sentences.

It is hardly surprising that the improvement rate in the supply of the indefinite

article, a feature not available in L1, is regular and systematic in T2, while the

development map of the definite article seems more erratic. It is probable that learners

found it more difficult to dispose of their transferred assumptions about the meaning

and function of a form already established in their L1 (Ringbom 1987).

All PL groups overused both articles more in T2 than in T1. Since learners of

all PLs seem to have been keen on supplying items to fill in the blanks in T2, this

error can only be ascribed to the task’s layout/design. What might have contributed to

higher overuse rates in this task is the direct prompting to the purpose of the task to

207
supply articles (or to leave blanks unmarked) which actually limits the choices in the

determiner position to articles only while no other determiners or pronouns that occur

in more authentic text can be used. As mentioned previously, some researchers

concede that with blanks tests there is tendency for learners to provide rather than to

refrain from providing articles. Liu and Gleason (2002) have argued that learners,

especially of low ability, are inclined to fill in every blank with a or the if they had the

chance to do so. Snape and Velasco-Zarate (2005) found that articles were supplied by

both Japanese (–art) and Spanish (+ART) learners more frequently in the blanks test

than in the production test. In this study, the urgency of supplying the blanks with

articles being more attractive than leaving them unmarked, led to inappropriate over-

marking of nominals with non-referential readings which, in effect, caused the

accuracy rates of the ‘zero’ article in this test (26%) to be the lowest of all tests.

Some overuse errors might be caused by overgeneralisation on the learners’

part as a result of a transfer of previous training, since most textbooks and classroom

instructions attribute definiteness to the sequence of occurrence of a referent in a

given text. Hence, it is very important to note that it is not the first/second mention

distinction that decides article choice, but rather the reader’s belief that the referent is

uniquely identifiable in a particular context.

The tendency for overuse in blanks tests has been firmly established in the

testing literature which means that some researchers, such as Ekiert (2004) who did a

study on Polish learners of English articles, refrained from using this type of task in

eliciting data. Our findings comply with earlier findings and this has been taken into

account in assessing when the results point towards higher accuracy or overuse rates.

Because the frequency of each error type differs according to the task performed

208
(Kharma 1981, Mizuno 1985, Tarone 1985, Tarone and Parrish 1988), it is possible to

argue that gap tests can give learners a false impression that an item is missing and

thus prompt them to oversupply in blanks. Another factor that may have contributed to

this result is the participants’ awareness of the subject matter of the test through the

direct prompts, which might have intensified the need to supply articles.

The replacement error pattern in T2 was interesting since the weaker learners’

performance exhibited a very small difference between the two types of substitution

errors. This behaviour is similar to that of the advanced group in T1. This

phenomenon offers some insights into the question: how far do testing types

constitute a valid tool for examining learners’ knowledge? The weak group’s error

pattern resembled that of the higher PL group when the test was examining their

explicit knowledge of article use. Both groups opted to replace the definite article by

the indefinite more frequently than the expected the-for-a option. The intermediate

group (G2) remained in favour of using the definite article to replace the indefinite,

yet, statistically, there was no significant difference between the two replacement

errors. This could be a sign of a stage of hesitation and altering hypotheses; a typical

attribute of intermediate PL learners.

The responses of G3 in the blanks test (T2) were in line with their performance

in the GJT (T3) as the replacement of the indefinite article with the definite (the-for-a)

was the more favourable choice, significantly higher than the a-for-the error. On the

other hand, the indefinite article replaced the definite more frequently in the free

composition test (T1). Hence it is possible to propose that the replacement error map

in the cloze and the GJT is a mirror image of the one that emerges from T1.

209
It seems that the inclination of weaker learners to consistently choose the

definite article over the indefinite proceeded gradually towards a better supply of both

articles with improved L2 levels to reach a stronger preference for the indefinite by

the G3 learners. In T2 however, the substitution by lower ability participants was

more in favour of a-for-the, but as PLs improved, the definite article became more

favourable by G2 and significantly more frequent than the-for-a errors in the

production of G3. The contrastive developmental trends in the two tasks reflect the

variation in task focus, degree of control and knowledge type. The higher the control

of the task and its focus on form, and the more the reliance on consciously learned L2

grammar rules, the better learners perform on the indefinite article. Conversely, with

focus on meaning, learners depend more on their implicit knowledge, in which the L1

influence is stronger, and produce the definite article more frequently.

5.2.1.3 T3

It is clear from the above that the highest accuracy rates for both articles were

achieved in T3. This is perhaps due to a few task-type related factors. Firstly, the

participants in this study are all formally instructed EFL learners with very limited

exposure to the target language, so it is not surprising that their performance would be

better in tests which focus on form (Thu 2005). Furthermore, with the extra time

allowed to students in this test, it was expected that learners would have a better

chance to reflect on the questions and draw more upon the abstract rules they formally

learned before consciously deciding on one answer. Thus, it is no wonder that this

task had incurred higher accuracy rates.

210
The other factor that has contributed to the sharp contrasts in accuracy rates

between the results of this task and T1 is the fact that the underlined NPs fall within

the sentential focus, rather than the extended paragraph /cross paragraph of the

content-based framework of the other two tests. The shorter structural distance could

have helped learners become more aware of the lexical connection between referent

and article and has thus enhanced the relevance between the two to optimal levels

(Sperber and Wilson 1986, 1995). Tracing referents, their countability status and

number is therefore facilitated through higher importance, repetition (persistence) and

shorter distance between article and referent (Huebner 1985) 40 leading to better

judgement. This diverges from the more contextualised text found in T1 and T2, and

indeed in authentic material, where the distance is longer, and where, in several

instances, other determiners, demonstratives or pronouns could be more preferable

(Ariel 1988).

Finally, despite the attempt to cover as many variable combinations as possible

in a balanced representation of grammatically correct/incorrect instances, the number

of underlined NPs remains limited. Being the shortest of all the three tests entails that

the GJT required the least processing effort whereby factors such as stress, exhaustion

and fatigue, known to negatively affect learners’ performance (Thomas 1989,

Batainah 2005, Trenkic´ 2007, Sarko 2008) were at minimal levels.

A closer inspection into the results of T3 reveals that there are large gaps and

sharp contrasts in the accuracy/error rates as well as between PL groups. For example,

while the difference between replacement errors across PL groups’ in T1 was not

40
Later, Epstein (2002) used the term Prominence to describe similar notions.

211
significant, in T3, the difference between the performance of learners from G1 and G2

on the one hand, and G1 and G3 on the other was highly significant.

Other differences were noticed in the error maps of task 3. The excessive

overuse of the definite article evident in the results of T1 and T2 was not recorded in

T3. Rather, it was the overuse of the indefinite article which participants from all PL

groups failed to recognise. Since participants’ responses, from all PLs, were uniformly

similar in this regard, it is highly probable that this is a product of the design of the T3

itself rather than any other cause such as PL or the nature of free/controlled tests in

general.

High overuse of the indefinite article in this test is rarely found in the

production of lower ability learners; This is confirmed by the results of tasks 1 and 2,

as well as evidence from personal experience in the EFL classroom. Therefore, this

particular error pattern, noticed exclusively in the GJT, could undermine the validity

of the results emerging from this test. However, such errors can be the effect of

extreme focus on form and accuracy exercised in this task, coercing learners into the

hyper-correction stage whereby omissions at previous IL stages are replaced with

overuse. This will be further discussed in the following section (5.2.2).

Similarly, learners were less able to notice the absence of the indefinite than

the definite article. It is possible that they assumed that the absence of the definite

article is sufficient to indicate that the NP is indefinite. Failure to recognise both

overuse and omission of the indefinite article is perhaps due to the form of the

indefinite article itself. Being, in most instances, a single letter makes it easier to miss

than the morphologically salient definite.

212
In the GJT learners of all PLs unanimously recognised the substitution of the

indefinite article by the definite as erroneous more frequently than the replacement of

the definite by the indefinite. With maximum attention to form, learned mainly

through explicit instruction, learners performance on the indefinite article was better

than on the definite in this task. In other words, learners were better able to recognise

the misuse of the indefinite article than the misuse of the definite by relying mostly on

their metalinguistic understanding of indefiniteness and cardinality. This result

confirms the fact that T3 is a form-focused test which explores explicit knowledge.

Yet when this result (preference of a over the) is compared with the results from T2

where the definite article replaced the indefinite more frequently by the two upper

groups, the relation between task type and knowledge type becomes questionable. Our

results support Dienes and Perner’s (1999) claim that the division between explicit

and implicit types of knowledge is more of a continuum than a dichotomy.

Furthermore, we propose that the difference between the design/format of T2 and T3

plays a decisive role in determining substitution patterns.

In T3, the weaker group performed significantly better on the definite article

while the accuracy rates from the intermediate group showed no difference in the

supply of both articles. The stronger group performed better on the indefinite article.

The replacement error pattern complies with the general findings on learners’

performance in the other tests and lends support to the argument that the results from

the GJT are in line with the overall results from other tests. It exemplifies the relation

between article use and PL as presented earlier (5.1.1) in the broader scale of

acquisition norms. In short, T3 results suggest that as PLs are low, the accuracy rates

of the definite article are relatively high, but, as PLs improve, the indefinite article is

better supplied and its omission rates are reduced, while the definite article is dropped

213
from many contexts, including obligatory ones. Thus, the gap between the accuracy

rates of the two articles becomes smaller.

However, the high accuracy results of T3 cannot be totally reliable a means of

measuring learners’ L2 ability, since the accuracy rates of the definite article in this

test for example range between 88% by G1 and 98% by G3, which are well above the

levels determined by the scores in the OPT. Moreover, the accuracy rates in the GJT

are higher than the average results found in the other two tasks, despite the fact that

T2 and T3 are both controlled tasks. The accuracy rates of the definite article in T2 for

example (G1: 64%, G2: 75%, G3: 86%) better reflect participants’ proficiency level

by being more compatible with the scores in the OPT. In all likelihood, the disparity

between most of the results from T1 and T2 on the one hand and T3 on the other can

be ascribed to the only difference between T3 and the other two tests which is the fact

that the GJT is presented in a de-contextualised format of discrete sentences, whereas

article choice in both T1 and T2 is largely based on information available in the

context. Furthermore, despite the attempt to provide as many variable combinations as

possible, the number of instances for each variable-combination set remains limited in

T3. This fact might have affected the results.

In addition to the above, there are certain problems inherent to GJTs in

general. Having the highest degree of control on learner production, allowing

participants only to mark questions as right or wrong, makes it difficult to determine

whether the learner is guessing the answer or making conscious decisions. For

example, if a learner chooses to mark all instances as correct (or incorrect)

indiscriminately, without actually reading or considering any of the questions, a 50%

accuracy result could still be achieved.

214
In previous SLA studies, the reliability of the results that emerge from GJTs

has often been questioned. Birdsong (1989) for example argued that metalinguistic

knowledge of L2 examined in highly controlled tests that focus on form does not

reflect linguistic competence. He also stresses that in GJTs learners are more likely to

judge a sentence as incorrect rather than correct [emphasis added]. Research by Ellis

(2004) substantiates Birdsong’s propositions as he found that among 19 SLA studies

published between 1997 and 2001, which depended on GJT as the data elicitation tool,

only two reported some measure of reliability.

5.2.1.4 Appropriate zero marking across tasks

There was no difference across groups in their performance on the zero article.

Participants from all PLs scored the highest accuracy rates in the free production task

(T1) while the lowest accuracy rates of correctly (un)marked indefinite plural and

uncountable NPs were recorded in T2. This is perhaps self explanatory since

appropriately produced bare nominals might not only be the result of learners mastery

of the L2 generic function; rather this could be a direct effect of the ‘simplification’

technique known in SLA. In other words, it is difficult to determine whether leaving

indefinite plural/countable nouns unmarked is based on actual knowledge of how to

employ generic references in English or on a mere lack of awareness of the

importance of supplying overt marking without being directed to do so in a meaning-

focused task such as T1. On the other hand, the deliberate prompting in T2 raises

learners’ awareness of the importance of supplying articles in the available blanks. In

addition, learners were more readily inclined to fill in the blanks rather than leave

them empty, although this was one of the options in the prompts. The effect of

215
learners’ response to blanks (see 5.2.1.2) was that the participants from all groups

scored the lowest accuracy rates of the zero-article in task 2. This result confirms that

the accurate outcome of T1 cannot be traced learners’ familiarity with the fact that

non-referential NPs are preferably left unmarked by NS of English (Behrens 2005,

Partee 2005). It is more likely that appropriate use of the zero article is tied to the

attributes of T1 discussed earlier (5.2.1) namely, higher omission/ lower overuse rates.

In other words, learners’ refrained from supplying words they conceived as

unnecessary, i.e., or that the absence of articles would not stop them from delivering

their ideas.

By comparison, participants from all PL groups performed better on the plural

than on the uncountable nouns in tasks 1 and 3, while the difference was less evident

in T2 where the accurate occurrences were almost identical due to comparable overuse

instances. It is worth mentioning that when non-countable nouns were selected for T3,

only a few were L1/L2 contrastive while the majority were uncountable in both

languages. Yet many learners failed to realise that the use of the indefinite article in

these cases is grammatically unacceptable. It is therefore conceivable that number is

acquired and mastered before countability (Butler 2002, Thu 2005).

5.2.1.5 Conclusion to performance by task

A comprehensive comparison of performance across tasks reveals that the progress

from discourse-driven towards grammatically-driven production, from the ‘least

obvious’ to the ‘most obvious’ contexts, parallels the diachronic development in L2,

(cf. Robertson, 2000). In other words, the PL of the EFL learner is reflected in the

focus of each task; the higher focus on accuracy, the better the performance.

216
5.2.2 Task effect by PL group

5.2.2.1 G1
5.2.2.1.1 Accuracy

It is clear that even at very low PL, (G1) learners did produce both articles, which

negates the supposition that learners do not produce enough articles at beginner levels.

This fact also negates the theory that presumes that the zero article is acquired first.

The results emerging from all these tests suggest higher accuracy rates for the definite

article than what many studies on –ART L1s (e.g. Trenkic´ 2002, Master 1997, Ekiert

2004, Butler 2002, inter alia) seem to suggest.

The weaker group performed best in T3 where the highest accuracy rates of

the definite article were achieved, while their weakest performance was recorded in

T2. Although the accuracy rates in T1 were lower than those in T3, they were higher

than originally expected. We believe that two reasons underlie this performance.

Firstly, as many researchers (e.g. Mizuno 1985, Kharma and Hajjaj 1989) concede,

the higher accuracy rates observed in production-based tasks may be attributed to

avoidance strategies that are unavailable to learners in more controlled tests.

Secondly, the relatively better performance might be due to the fact that the presence

of features in the L1 that are similar to those in the L2 lead to positive transfer which

ultimately facilitates the process of L2 acquisition (Ringbom 1987).

However, the accuracy rates of the definite article in T1 were significantly

lower than the scores obtained from T3. It is possible that the higher rates in T3 do not

imply learners’ sensitivity to the importance of the hearer’s perspective or, according

to Heim (1982), the common background parameter, since, by definition, most

grammaticality judgement tests lack the communicative element on which sharedness

217
of information is based. Our results are also contrary to Trenkic´’s (2000) claim that

the correct use of the definite article -by L1 Serbian learners of English- is attributable

to the speaker’s assumptions about the recipient’s knowledge. Rather, we propose

that with respect to the high accuracy rates of the definite article , the stronger

influence of the L1 at lower PLs (Snape 2006, Paradis and Zdorenko 2008, Hawkins

and Chan 1997) could be interpreted as a sign of positive transfer of L1 semantic

properties into L2 contexts since both Arabic and English have almost identical

conditions for obligatory suppliance (cf. Liu and Gleason’s (2002) list of non-generic

the contexts as an interpretation of J. Hawkins (1991) theory of definiteness).

That the accuracy rates of the definite article are significantly lower in T2 may

be the result of variation in the method through which the data was analysed. In task 2,

the learners’ answers had to strictly match the control group’s responses to be

considered accurate. If the learners’ answers did not match the CG, they were

categorised incorrect. On the other hand, accuracy in T1 data was judged on the basis

of grammaticality, which renders contextually unsuitable responses errors of choice

(see section 3.5.3).

Learners from this group showed little disparity in their performance on the

indefinite article across the three tasks since, consistent with our expectations, the

accuracy scores were consistently low with no cross-task significance recorded.

Hence, the effect of task type on the weaker group’s performance on the indefinite

article was quite minimal. Therefore, we may assume that G1 learners lack the

competence to appropriately supply the indefinite article felicitously where it is

deemed obligatory whether the indefinite article was expected to be produced

218
automatically (T1), supplied in obligatory contexts (T2) or its absence recognised

(T3).

Despite the insignificant range of variation across tasks, the best suppliance of

the indefinite article was observed in the blanks test (T2) while the poorest results

were found in T1. We may speculate that the failure to supply the required indefinite

marker in obligatory contexts in T1 was probably most affected by the lack of

prompting in the rubric to the purpose of this test, which, if used, could have alerted

learners to the necessity of any obligatory supply. The better supply in the blanks test

is in turn the effect of direct instructions. It is also possible, since the focus in this task

(T1) is to negotiate meaning, the L1 could have exerted a stronger influence on

weaker learners’ production. i.e., the participants could have carried the semantic

notions from Arabic in which the absence of a definite marker in a noun phrase can be

enough indication of its indefiniteness. In principle, however, this concept is not

entirely limited to Arabic. For example, Leech (1992) contends that ‘it is convenient,

from many points of view, to regard an initial determiner as obligatory for English

noun phrases, so that the absence of an article is itself a mark of indefiniteness.’

(1992:15).

The failure to provide the indefinite article in T1 can also be driven by

learners’ assumption that its absence does not constitute a hindrance to successful

communication of ideas, which is the goal of the free writing test. It is likely that

weaker learners have subconsciously applied the Economy Principle (Poulisse 1997)

whereby maximal comprehensibility is achieved while exerting minimal processing

effort.

219
Supposing that the above hypothesis were true, T2 would have been expected

to yield higher accuracy rates of the definite article since learners were alerted to

supply articles. Instead, there was higher overuse of the indefinite article in positions

where the definite article should have appeared (see the section on Replacement

Errors below). Consequently, the accuracy scores of the definite article plunged to

their lowest levels in all tasks. The use of the indefinite article confirms that learners

rely more on their explicit knowledge in cloze tests which is known to precede the

implicit, at least for adult EFL learners (Dulay et al. 1982, Ellis 2004).

Finally, from a non-contrastive perspective, the accuracy rates being higher in

T1 than in T2 could have resulted from the dissimilar methods through which its use

was analysed in the two tasks. In T1 the correct use of the definite article depended on

the learner’s own assumptions of the hearer’s knowledge of a referent; a subjective

decision with a wider range of acceptability, whereas the correct use in T2 was set

against the responses from the majority of NS in the control group, which limited the

options of acceptability.

The highest accuracy rates of correct marking of bare nominals (zero article)

were achieved in T1. Cautiously interpreted, high accuracy rates do not necessarily

imply that lower PL learners are able to discern generic from specific (or attributive

from referential) meanings in their choices of articles. Rather, it is more likely the case

that the absence of overt marking is a trait noticed throughout T1, evident through the

highest omission rates scored in this task. Perhaps article supply was not perceived as

an essential hindrance to communicating meaning (Leech 2006b). Contrary to the

common belief regarding the primary acquisition of the zero article (Master 1997,

Young 1996, Parrish 1987, Thomas 1989), we propose that the order of acquisition, if

220
based on high accuracy rates, might not be a result of learners’ consciously associating

non-referential meanings with bare plural /mass NPs; rather, it could well be a matter

of negligence or ignorance of the importance of article supply. Once the cognitive

environment changes into declarative and the learners’ awareness is raised, such as is

the case in T2, overuse errors are committed and the accuracy of bare nominals

plunges to the lowest levels. Therefore, it is possible to propose that the high accuracy

rates of bare nominals in T1 are not an indication of early acquisition or that the zero

article is acquired before the other two articles.

Accuracy rates show that the weaker group performed significantly better on

the definite than the indefinite article in T1. This is consistent with our predictions for

several reasons. When learners are given the choice, the definite article presents a

safer option of the two articles since it collapses the more complicated aspects of

countability and number which affect the grammatical accuracy of a NP. In addition,

most studies on learners from both ±art L1s confirm that the definite article is

mastered at an earlier IL stage than the indefinite (Huebner 1983, Master 1987,

Parrish 1987, Thomas 1989, Zdorenko and Paradis 2008).

Similarly, the members of G1 recognised the correct use of the definite article

in the GJT better than they did the indefinite. Therefore, it would seem what learners

are consciously aware of, regarding the correct usage of the indefinite article, does not

extend to recognition of correct instances. The similarity in G1’s behaviour on the

meaning-focused task and the form-focused task might undermine the supposition of

task type effect on learners’ performance. In other words, the free-versus-cloze task

distinction in this respect is a not a crucial element in determining article choice.

However the task-type effect is more noticeable in the variation of accuracy rates that

221
emerged from T2 as there was no significant difference between the accuracy rates of

the two articles in the stories task. Thus, in T2, the weaker learners’ performance

approximates the advanced group’s closer accuracy rates. This reveals that learner’s

explicit understanding of the rules that govern the use of the indefinite article is better

than their implicit knowledge. However, since T2 is the only task that presented a

different accuracy pattern, we propose that this variation is perhaps more due to the

higher overuse rates of both articles that this type of task has yielded rather than to

actual awareness of obligatory supply of the indefinite article.

If we accept that low accuracy rates of the indefinite article across tasks is

caused by a stronger influence of the L1 at this stage, the accuracy rates of the definite

article would be expected to be higher across the board regardless of task type. In fact,

the results suggest that the two lower PL groups performed slightly better on the

indefinite article than the definite in T2, with accuracy rates matching those of the

definite by G1, and significantly higher than those of the definite by G2. Therefore, it

is not possible to exclude other factors from being as effective in determining the

current map of L2 articles.

5.2.2.1.2 Errors

Earlier studies on SLA of articles suggest that production tasks yield lower error rates

than objective tasks because of the avoidance strategy (cf. Tarone and Parrish 1988,

Mizuno 1985, and Maalej 2004); i.e. learners resort to using other determiners such as

quantifiers and demonstratives, freely, to reduce the risks of committing errors caused

by using the wrong article. Judging by the results found in this study, the above

proposal is restricted to the accuracy rates of the definite article, since the accuracy

222
rates of the indefinite achieved by the two lower groups are higher in T2 than in T1

and, similar to their performance in T1, the weaker learners also performed

significantly better on the definite article than on the indefinite in the GJT.

Furthermore, the error map in T3 does not match the one in T2 because the two upper

groups performed better on the indefinite than the definite in the blanks test.

Therefore, the proposition that free tests result in lower error rates is debatable since

various factors such as the presence/absence of an L2 feature in the L1 as well as

learners’ PL are also crucial in determining accuracy/error rates.

5.2.2.1.2.1 Omission

The failure to supply a(n) with indefinite singular countable nouns was the most

noticeable difficulty in the lower group’s performance across tasks as the omission of

the indefinite article was the higher than all other errors. Furthermore, G1 participants’

omission rates of the indefinite article were significantly higher than the definite

across tasks. Tsimpli’s (2003) conviction that the absence of features in the L1 causes

syntactic representations in L2 production to become defective applies to the

difficulties which L1Ar learners experience since the language lacks a dedicated

marker of indefiniteness.

That little variation was found in the omission rates of the same group across

tasks indicates that this group of learners possesses incompatible levels of explicit and

implicit types of knowledge or indeed lack thereof. Despite the small differences in

omission rates across tasks, the weaker learners omitted the indefinite article more

frequently in T1 than in the cloze test and the GJT. From a transfer perspective, this

was mainly because the grammaticalised form of definiteness is already available in

the learners’ subconscious and thus easily automated in free production whereas the
223
indefinite article is mostly learned through explicit instruction and is more accessible

in tasks that draw on metalinguistic information such as T2 and T3. G1’s higher

omission rates of the indefinite article in T1 implies that this newly acquired feature

has not yet reached the point when the rules become internalised enough to be

produced in communicative discourse. The shift from knowing the abstract rules to

automatically exercising them in meaningful contexts has been proposed by several

SLA researchers (cf. Ellis 2004, Roehr 2004. Accordingly, learners’ theory-building

skills (Anderson 1983) allow declarative knowledge to evolve into communicative

usage given time and sufficient practice (see also DeKeyser 1998, 2003).

While this group scored the highest omission rates of the indefinite article in

T1, the highest omission rates of the definite article were observed in T3. The

availability of content in T1 and T2 offers a better opportunity to grasp meaning

through forward and backward referencing which enables learners to establish the

connection between the referent and the article and realise its definiteness. On the

other hand, the de-contextualised format of T3 puts the communicative burden on the

reader whose role is to resolve the ‘referential puzzle’ (Jaszczolt 1997) without

sufficient information about the referents. Within the sentential focus of meaning,

students might presume that the information provided is adequate and clear and no

definite marking is required to establish relevance between article and noun.

The contrast between this group’s performance in T1 and T3 stems from their

understanding of the roles of each article. The rules which determine the appropriate

supply of the definite article are tacitly available in the learners’ subconscious, despite

the low PL. On the other hand, their understanding of the conditions that necessitate

providing the indefinite article is better reflected in tasks that investigate declarative,

224
metalinguistic knowledge about the second language, and would perhaps need to be

evoked by deliberate prompting.

5.2.2.1.2.2 Overuse

The significant differences in the overuse rates across tasks are strongly suggestive of

learners’ errors of overuse being dependent on task type. The highest overuse rates for

both articles were observed in the results from T3, while the lowest overuse rates were

in T1. Given that these rates were scored by the same participants within one PL

group, the significantly lower overuse rates of the indefinite article in T1 may not be

entirely due to learners’ developed awareness of article use. Instead, it could well be

attributed to the absence of direct prompts which might have alerted learners to supply

articles. Furthermore, a stronger L1 influence at this IL stage dictates that the absence

of an overt indefinite article implies that the NP is of [–def] semantic value. Learners,

regardless of the first language, are not expected to overuse a feature that is non-

existent in their L1. However, when prompted, the weaker learners’ production of the

indefinite article increased to rates approximating those achieved in T2 and T3. The

higher production rates of the indefinite article in T2 and T3 do not necessarily imply

that learners would be capable of producing this article automatically in obligatory

contexts because in the production task, the highest error rate was that of failing to

supply the indefinite article rather than overusing it.

In principle, our participants’ notions of definiteness should be close to the TL

since the conditions that render the supply of the definite article obligatory are quite

similar in both Arabic and English (i.e. the referent being available textually;

225
physically or semantically unique) (2.2.3). Therefore, despite the inappropriate supply

of the definite form in non referential contexts, in T1, with meaning in focus, the

overuse rates did not reach the extent observed under the influence of prompting in the

cloze test or the underlining of NPs in T3. Yet, the definite article in the writing task

was overused significantly more frequently than the indefinite, while the same group

overused the indefinite article more than the definite in T3.

In T2, the overuse rate of both articles was almost identical. Hence it becomes

safe to assume that task type is a decisive factor in determining learners’ overuse

patterns.

5.2.2.1.2.3 Replacement

Task effect was most noticeable in replacement errors as the production/answers of

G2 varied from one task to another according to the focus of each test. In T1, weaker

participants chose the definite article in place of the indefinite four times more

frequently than vice versa. The reasons underlying the mastering of the definite

article before that of the indefinite are developmental, namely the cardinality function

that the indefinite article involves, and L1 influence, i.e. the absence of an explicit

marker of indefiniteness in Arabic. A similar error pattern was observed in the results

from the GJT (T3) where learners’ ability to recognise the- for- a instances as

incorrect was better than recognising a- for- the replacements. However, the

difference in the reaction to T1 and T3 was noticeable, as the gap between recognising

the two types of replacement errors (a-for-the vs. The-for-a) in T3 was not as large as

it was in T1. This may be due to the varying length and requirements of each task. In

T1 learners were expected to write essays with a minimum of 350 words which would
226
accrue more errors while T3 was more limited. A more plausible explanation for this

variation in performance is that learners are better able to accurately employ L2

information (of the indefinite article) that they were taught when the focus of the task

is on form and the requirements are explicitly stated, as in the case of T3.

However, this proposal does not apply to the results from T2 where the

replacement rates of the two articles were very close. Moreover, instances where the

indefinite article was used in place of the definite were slightly more frequent than

those where the definite replaced the indefinite. Although the difference is small, G1’s

preference of a-for-the to the-for-a is indicative of the amount of influence that task

focus exerts on learners’ choices. It could have been the result of an attempt to provide

an equal number of both articles throughout the blanks. In other words, the preference

of a- over- the is due to the design of T2 rather than the learners’ awareness of

indefinite marking. An argument in favour of this possibility is that G1 chose to

replace the indefinite article by the definite in the other two tests (T1 and T3) which

have different focus and examine different abilities. Therefore, G1’s performance in

T2 becomes the exception, not the norm. In this sense, it seems that the division of

task type into free vs. close is not the only source of variation in L2 production.

Rather, the format and the design of the task is a critical element that led to incurring

the present results.

The pattern observed in T3 is similar to the one in T1 as the weaker learners

preferred the definite to the indefinite article. This is in line with G1’s better

performance on the definite article in general. The weaker group had supplied the

definite article better than the indefinite, achieving higher accuracy and lower

omission rates across tasks. However, the difference between the two replacement

227
errors in T3 was not significant as is known of results obtained from the GJT (see

section 4.3.2.3). This is perhaps due to the different type of knowledge that T3 taps

into. Although learners had found it more difficult to mark instances where the

indefinite article was replaced with the definite as incorrect, their recognition of the

necessity to supply the indefinite article was better than it was in T1. This is because

the task’s requirements were clearly stated and the focus on form was maximised. In

effect, this helped reduce the difference between the performance maps of the two

articles. Therefore, the gap between the two replacement errors was insignificant.

Hence, it is reasonable to argue that weaker L1Ar learners’ formal knowledge of the

indefinite article is better than their ability to supply it communicatively.

5.2.2.2 G2
5.2.2.2.1 Accuracy

When comparing the accuracy rates of a(n) and the across tasks, it is noticeable that

the gap between the accuracy rates of the definite and indefinite articles in both T1

and T2 decreased as learners PL improved. Therefore, it seems safe to assume that the

levels of learners’ implicit and explicit levels of understanding article use become

more compatible with overall progress in L2.

Similar to the weaker group, the intermediate group also scored the highest

accuracy rates for the definite article in T3. The accuracy rates of the indefinite article

were also highest in T3. Therefore the highest accuracy rates for both articles were

achieved in T3. However, the difference between G2’s performance on the two

articles can be noticed in the fact that the lowest rates were scored in different tests.

While the weakest performance on the definite article was found in T2, the lowest

228
accuracy rates of the indefinite article were recorded in T1. The variance in the error

pattern is apparently attributable to the requirements and focus of each task.

Since the definite article has a semantic equivalent in the participants’ L1, the

grammatical representation of definiteness is more accessible in L2 production.

Therefore, it is not surprising that G2 did better on the definite in the test that

examines their implicit knowledge (T1), as is expected in communicative tasks. On

the other hand, since the morphological representation of indefiniteness is not readily

available in learners’ subconscious, it is likely that accurate use would be more

evident in tests that depend mainly on explicitly learned information and assess

metalinguistic knowledge of the L2. Conversely, this entails that lower accuracy

should be expected in tasks that do not provide directions that activate the newly

learned L2 form.

Furthermore, the design of T2 could have contributed to raising the accuracy

rates of the indefinite article as the blanks, being in the syntactic position of the

determiner, may have triggered learners’ awareness of the necessity of supplying the

indefinite article and directed them to search for a singular noun in the following NP.

However, the significant difference between the accuracy rates of the indefinite article

in tasks 1 and 2 suggests that the higher accuracy rates in T2 cannot be considered as a

true reflection of the learners’ actual competence in correctly supplying the indefinite

article wherever its presence is deemed obligatory in meaningful discourse. Rather,

the low rates in T1 suggest that the grammatical rules that control article use in

English have not yet become internalised enough to be spontaneously produced in

communicative settings without prompting. Dulay et al. (1982) argue that errors in

form-focused tests occur when formally learned rules have not yet become part of the

229
learners’ linguistic competence, i.e. learners need time to practice their explicitly

learned L2 rules in order to be able to produce grammatically appropriate forms. In

addition to being the cause for errors, the proposal submitted by Dulay et al. applies

equally to accuracy rates of the indefinite article in T2 and T3 as an indication of

learners’ awareness of the conditions that necessitate the supply of the indefinite

article.

In T3, the intermediate group’s results diverged from those observed in T1 and

T2 in regard to accuracy. Thus, G2’s performance resembled that of the higher, rather

than the lower, PL group. The disparity between the two patterns confirms that G2’s

awareness of the grammatical rules of article use, examined in T3, is one step ahead of

their communicative abilities (competence) in T1 where their knowledge of how

articles operate is reflected. It seems that learners from G2 have improved their

declarative knowledge of the grammaticality of the indefinite article and have

therefore approached the rates of the advanced group. This is not merely an effect of

moving further away from L1 influence since improved performance on the indefinite

article is considered as a broader IL tendency having been recorded in studies on –

ART L1 learners’. For example, Chaudron and Parker (1990) found that Japanese

learners’ control of the indefinite form increased with better L2 proficiency.

5.2.2.2.2 Errors

5.2.2.2.2.1 Omission

SLA researchers, such as Hawkins and Chan (1997) and Prévost and White (2000),

ascribed the difficulty 2LL have in the employment of a feature that does not exist in

their L1 to a failure in mapping functional features present in the L2 onto their

production of the target language. In other words, learners rely on L1 syntax in their
230
L2 production especially at lower PLs. These theories apply mostly to the high

omission rates of the indefinite article observed in the responses from the lower group.

It is a typical occurrence of what Eckman (1977) describes as the most difficult aspect

to acquire in the target language, namely the production of a feature which is

unmarked in the L1 but marked in the L2. This was what happened in T1.

The performance of this group was considerably affected by task type. We

noticed that their average rates of omission were analogous to the weaker group’s

results in the free production task while being closer to the advanced group in the

controlled tests. In the free production test, although there were fewer omissions by

this group than by the weaker group, G2’s performance was similar as intermediate

PL participants omitted the indefinite article more frequently than the definite.

Learners might have found it redundant to mark a nominal overtly for indefiniteness if

its semantic indefiniteness is already inferred by the absence of the definite marker.

However, the lowest omission errors of both articles were in the cloze test. This lends

support to the theory that the presence of blanks urges learners to supply rather than

omit articles (Trenkic´ 2000, Snape and Velasco-Zarate 2005). Furthermore, learners

from G2 made more omissions of the definite than the indefinite article. By scoring

higher omission errors of the, G2’s performance approximated that of the advanced

group.

The shift is caused by the types of knowledge that each task triggers. Through

classroom instruction, intermediate PL learners have become more aware of the

conditions of indefinite article employment and at the same time they are beginning to

realise that it is not common practice by NS or in authentic text to supply the definite

article with indefinite plural/uncountable nouns.

231
Advocates of teaching articles (e.g. Master 1997) in the EFL/ESL classroom

propose that explicit teaching of rules can lead to automated use, i.e., for a learner to

know how a feature operates precedes the eventual ability to voluntarily apply these

rules in communicative settings (VanPatten 1994, De Keyser 2003; Doughty 2003;

Ellis 2001; Norris and Ortega 2001). Yet more time is needed for this declarative

knowledge to become internally proceduralised and consequently produced in

meaningful output such as in T1. Evidently, the two types of knowledge, implicit and

explicit, do not evolve simultaneously and as the above mentioned studies confirm,

the intermediate group’s performance in free production was one stage behind their

performance in T2 and T3.

5.2.2.2.2.2 Overuse

There was a smaller disparity between the overuse rates of the two articles in the

production of the intermediate group as compared to the weaker group’s performance,

i.e. while the weaker group overwhelmingly preferred to replace the indefinite by the

definite article, this preference was less noticeable in G2’s production. The decreased

difference in the overuse rates may be seen as a mark of change in learners’

hypotheses and understanding of article use and as a fluctuation typical of the stage in

IL in question.

The intermediate group produced the highest overuse errors of the definite

article in T2 and the lowest in T1. This is self explanatory as in T2, there were

awareness raising prompts towards supplying, rather than omitting articles whereas

there were none in T1. Similarly, the lowest overuse rates of the indefinite article were

recorded in T1. This is attributable not only to the just mentioned absence of prompts

232
in T1, but also to the fact that that intermediate PL EFL learners in general and L1Ar

learners in particular, do not voluntarily overuse forms that are unavailable in their L1.

However, unlike the overuse errors of the definite article, the highest overuse

rates of the indefinite article were observed in T3. In judging the appropriateness of

article use in T3, G2 learners rely maximally on their conscious knowledge in

interpreting the grammatical system of the indefinite article. That the highest overuse

rates of the indefinite article were scored in T3 is tied to it being the task with the

highest focus on form and analysed knowledge learned through explicit instruction. In

this sense, G2 approaches the more advanced group most closely in T3 in

overproducing the indefinite article as it examines the strongest aspect of their

knowledge. The newly acquired L2 feature is overused (Selinker 1972) because it

might have been misanalysed as a general form that signals indefiniteness, regardless

of countability or number, i.e., a counterpart of the definite article. Thus this type of

overuse is an example of what is referred to in SLA literature as partial understanding

of TL features (Richards 1976) that leads to a flooding stage similar to the one

observed before mastering the definite

5.2.2.2.2.3 Replacement

Despite the higher PL, intermediate learners still preferred to substitute the indefinite

article with the definite across tasks, but the cross-group development is noticed in the

decrease in the gap between the two replacement errors. In T1, the replacement of the

definite article for the indefinite was four times higher than the replacement of the

indefinite for the definite. In T2, G2 learners replaced the indefinite article with the

definite more frequently than replacing the definite article with the indefinite, but by

being more aware of the requirements of the task, they managed to produce the
233
indefinite article more frequently than in T1 and thus the gap between the two types of

replacement errors was reduced to the degree of no significance. It seems that not

only the prompts in T2 had activated learners’ awareness of the indefinite article, but

also the nature of the test itself, i.e. it being a gaps test encouraged learners to supply

both articles, including the indefinite. Therefore, it was not surprising that the definite

article replaced the indefinite only twice as many times as the indefinite article

replaced the definite in T1. In comparison with G1’s fourfold replacement rate in T1,

this is considered an improvement.

In T3, although form-focused and with prompts, G2 learners preferred the

definite article to the indefinite in their choices. Nevertheless, the difference in the

replacement rates in T3 is more extreme, a predictable effect of GJTs, as the

distinction was significantly varied. (For a full discussion of GJT see section 5.2.3).

The results indicate that G2 learners preferred the definite article to the indefinite in

all three tests.

5.2.2.3 G3

Task type had a smaller effect on G3’s performance with regards to accuracy rates

between the two articles in all three tests. However, the advanced group were more

susceptible to task type variation in their use of the indefinite article.

234
5.2.2.3.1 Accuracy

Similar to the performance of the two lower PL groups, the highest accuracy rates of

the definite article remained in T3 but with a significantly smaller gap of variation

across tasks.

The highest accuracy rates of this group’s performance on the indefinite article

were also observed in T3. Because of their increased ability in mastering the indefinite

article, G3 participants managed to identify all the correct instances of indefinite

article occurrences in the GJT, scoring the highest accuracy rates possible. With

slightly lower accuracy rates than T3, more advanced participants also managed to

appropriately supply the indefinite article in most obligatory contexts in T1, obtaining

higher accuracy rates on the indefinite than on the definite article. The only test where

the accuracy rates of both articles were almost equal was T2. It is worth mentioning

that G3’s highest error rates of the indefinite article were not observed in T1, as is the

case with the two lower groups, but rather in T2. This low error rate was not caused

by the omission of a, but rather the overuse of a(n).

Although the prompting in T2 raised learners’ awareness to the necessity of

supplying articles, it may have encouraged them to overuse the indefinite article

which consequently contributed to lowering the accuracy rates. The difference

between correct the and its overuse was not significant in T2. However, in T3 the

same participants were better able to recognise correct definite contexts and identify

overuse instances. As learners had the most limited opportunity to contribute freely to

the task, their responses were most accurate.

235
Since the difference in the performance of G3 participants on the two articles

was not significant in any task, we can assume that with stronger L2 ability, learners’

mastery of the two articles becomes more compatible and their choices less

susceptible to task variation.

5.2.2.3.2 Errors

5.2.2.3.2.1 Omission

The disparity in the omission rates of the two articles was insignificant in G3’s

production. While the omission rates of the two lower groups showed significant

difference between the two articles, the gap in the omission rates decreased with

improved PLs. Weaker learners from G1 and G2 omitted the indefinite article more

frequently than the definite in the writing task (T1), but G3 participants made more

omissions of the definite than the indefinite article in T1. Since task (T1) reflects

subconscious knowledge whether acquired through exposure to L2 materials or learnt

through formal instruction, learners’ command of the indefinite article has become

more internalised to be produced in communicative output. On the other hand, the

definite article is dropped from many contexts, including obligatory ones, because

after increased awareness of overuse errors in non-uniquely identifiable contexts,

learners might be experiencing a stage of hyper-correction which results in higher

omission errors of the definite article.

Despite the fact that T2 is a cloze test mainly examining explicit knowledge,

the results from T2 are quite similar to those from T1 in the sense that the omission

rates of the definite article, although significantly fewer than in T1, were also higher

than the omissions of the indefinite article.

236
However, results from T3 reveal that the same group of participants accepted

the instances of indefinite article omissions as grammatically correct more often than

instances where the definite article was omitted. Statistically, this was interpreted as

more omissions of the indefinite than the definite article.

Hence, the error map in the omission rates here does not seem to be divided by

task type since similar results were observed in two tasks of different focus while

results from the other cloze test do not follow the same line. Rather, this phenomenon

calls for finding a common denominator between the form and meaning-based tasks

(T1, T2) that sets them apart from the divergent pattern observed in T3. One element

that presents itself as the most probable factor in determining this divergence is that of

content since both the free composition and stories tasks are content-based while T3 is

composed of discrete sentences. Within the focal constituent of the sentence, learners

had a better opportunity to recognise the absence of the definite marker because of the

structural proximity between the determiner position and the referent, giving the

participant a ‘very small amount of … information .. to focus on, or be ‘active’ at any

one time.’ (Chafe 1987: 22). Thus referents become more prominent (Huebner, 1985)

and highly accessible. On the other hand, in more authentic discourse a pronoun is

usually the more preferable option for shorter distance referencing (Ariel 1988, 1990).

To follow Ariel’s criteria for definiteness, there is a smaller number of antecedent

competitors in a sentence-based test and a higher importance given to the NP since it

is underlined in T3. Hence, the conditions of unique identifiability become more

accentuated and the omissions of the definite article become less acceptable and its

absence more visible. Thus, a referent in the preceding phrase can be felicitously

retrieved since all conditions of optimal relevance (cf. Sperber and Wilson 1995) are

satisfied within the frame of the immediate context.

237
The results from T1 and T2 solidly confirm previous findings that with

increasing proficiency levels, there is increased production and fewer omissions of the

indefinite article. Simultaneously, the definite article is systematically dropped from

non referential contexts (Cziko 1986, Chaudron and Parker 1990, Habuto 2000, Ekiert

2004) and sometimes more omissions of the definite article are committed before a

more target like production is reached.

5.2.2.3.2.2 Overuse

In the first part of this chapter, we mentioned that the most noticeable improvement in

learners’ production was the significant and systematic drop in the overuse rate of the

definite article with progressing proficiency standards. Yet, the error pattern relating

to task variation was not quite divergent. Similar to the performance of the weakest

group, the highest overuse rates of both articles were recorded in T3 and the lowest

overuse rates of both articles were recorded in T1. It therefore seems that these error

norms are a result of task variation, while the error rates are governed by PLs.

The overuse rates of the definite article were identical in T2 and T3, while T1

showed significantly lower rates. The results of overuse errors are clearly divided

according to task type, i.e. the cloze test with attention to form yielded higher overuse

rates than the free production task. Furthermore, as the error rates of overusing the

indefinite article fell regularly and significantly with the improvement in the PL in

both T2 and T3 in line with expectations of developmental progress, this was not the

case in the results obtained from T1. Although the overuse of both articles was

particularly lower in T1 than in the controlled tests, the total level of overuse by the

advanced group was slightly higher than that of the two weaker groups in the free

238
composition task. At first sight, this could be interpreted as a form of regression

because this irregular developmental pattern suggests that the overuse rates increased

as PLs improved.

To make more mistakes at higher PLs contradicts Oller and Redding’s (1971)

claim that overall second language competence and performance on articles correlate

positively. On the other hand, some researchers such as Lightfoot (1998) found that

learners’ performance on articles does not necessarily reflect their overall PL.

Batainah (2005) also found that senior Jordanian learners overused the indefinite

article more frequently than lower ability learners.

The overuse of the indefinite article could be the result of hyper-correction due

to classroom experience since it is known that ArL1 learners are constantly instructed

to mark singular indefinite NPs with the indefinite article, which could result in over

application of these instructions into uncountable and plural contexts, i.e., this error is

caused by a ‘transfer of training’ (Selinker 1972) from experienced at previous

learning stages. Richards (1971) maintains that failure to observe restrictions of

countability and number in article use may be due to analogy. In our case, the analogy

can be based on formulaic expressions learned as chunks such as existential [there is +

a(n)] or [have/has + a(n)] acquired at former IL stages and rationalised into a deviant

usage with uncountable / plural nouns. G3 equally accepted the erroneous overuse

instances of a twice as often as the definite in T3.

Although the advanced learners made more overuse errors of the indefinite

article than the weaker groups in T1 and T3, it can be an indication of progress. The

tendency to overproduce a newly acquired feature in IL is common and it usually

follows the initial stage of emergence (Thomas 1989 add, Ekiert 2004, Garcia Mayo

239
2008). It can also be the result of hyper-correction as learners try to avoid omission

errors committed during their previous learning experience at lower PLs.

L1Ar learners experience a mapping problem of the L2 feature (a/n) into IL

grammar, more similar to the Japanese rather than the Spanish learners in Snape and

Velasco-Zarate’s (2005) study, or learners from –ART than +ART L1s in the study by

Zdorenko and Paradis (2008). Therefore one may expect even the G3 to overuse the

definite article more than the indefinite in tasks that examine explicit knowledge or

even in any form-focused test which has content. This group overused the definite

article twice as many times as the indefinite in T1 and T2 where the lowest accuracy

rates of the indefinite article were recorded.

If the stronger inclination to fill in the blanks (rather than to leave them empty)

in T2 is added to the stage of flooding of a(n) that the advanced group is already

experiencing, the higher overuse errors would become inevitable and would

eventually contribute to lowering the accuracy rates in this task. Therefore, we believe

that the exceptional pattern of close accuracy rates observed in T2 is more relevant to

the design of this test than the usual performance observed in the production of G3.

5.2.2.3.2.3 Replacement

As in the case of the choices of the two lower groups, the participants from G3 scored

the lowest replacement errors in the free production task (T1) while the highest

replacement errors were observed in T3.

G3’s awareness and their improved use of the indefinite article bridged the gap

and balanced the rates of the two replacement errors in T1, with a slight preference for

240
the indefinite article. The shift is in accordance with the development map in article

use discussed earlier whereby representational deficit (Hawkins 2004) of a cardinality

marker in the L1Ar leads to a low production at the beginning followed by a stage of

over-application of the newly acquired feature, before a more target-like use is

reached. If we suppose that, in marking indefiniteness, Arabic is an –ART language,

G3’s implicit understanding of the indefinite article expressed in their choices in T1,

corresponds to findings from Leung (2001)’s study on Japanese (–art) learners who

preferred to substitute the by a more often than a by the.

However, this development map was reversed in the cloze test and the GJT.

The higher PL learners’ preference was for the definite article to replace the indefinite.

In fact, the difference between the two types of replacement errors was more

significant in G3’s production than in that of the lower ability learners as none of G3

participants accepted a for the as correct. It is clear from the above that the most

decisive element in learners’ choices of articles was task focus. G3’s mastery of the

indefinite article has obviously progressed into their implicit competence as expressed

in the free composition task, while their awareness of conditions determining

obligatory supply of the definite article is revived in tasks that focus on form. It is

worth mentioning that this trend is the exact opposite of the one noticed in the

performance of the weakest group.

241
5.3 Learner hypotheses

In the SLA literature, it is widely acknowledged that learners go through a process

called ‘hypothesis formation’ (Schachter 1983, 1992) in which the learner

incorporates all the imperfect conceptualisations that she or he has of the L2 to

construct a hypothesis about a given feature in the L2. The sources of information that

form these hypotheses could originate from the learner’s L1, but they could also be

caused by incomplete understanding of the TL rules (Richards 1971) or from over-

generalising newly acquired L2 rules to other contexts in which they do not apply

(Selinker 1972, Gass 1996, Ellis 1994). Throughout the IL process learners’

hypotheses change and affect their L2 production.

There have been several studies that addressed the effect of associating certain

linguistic concepts and article use in learners’ hypotheses (see section 2. 5). In this

section we will attempt to examine the hypotheses that underlie L1Ar learners’

choices of articles by careful consideration of the error patterns obtained from the

three tests. We will also try to present a possible rationale for these assumptions,

whether induced by L1 interference or developmental, cross-linguistic tendencies. The

role of task focus and task design will also be discussed as influential variables.

The L2 hypotheses examined in this study involve three linguistic notions of

specificity, premodification and the abstractness/ concreteness of the head noun.

5.3.1 Specificity

One of the most influential hypotheses regarding learner associations with respect to

L2 acquisition of articles is that of definiteness and specificity ( IKW 2004). In order

242
to determine whether or not there is association in the learner hypotheses between

specificity and definiteness and whether the type of task affected learner choices, the

overuse of the definite article in two indefinite NP environments [+/–Sp] was

compared.

5.3.1.1 G1

Weak learners used the definite article to refer specifically as well as non-specifically

in all tests. Their accurate supply of the definite article in [+def] contexts was

significantly higher than erroneous occurrences in indefinite contexts (overuse and

replacement).

In the free production (T1) the definite article was overused significantly more

frequently in [–def] [–Sp] contexts than in [–def] [+Sp] contexts by G1 learners. In

other words, there were more overuse instances in non-referential plural/mass NPs

referring mostly to kinds (cf. Chierchia 1998) than in specific NPs referring to a

particular thing or group of things (Zegarac 2004).

The above information implies that overuse (the supply of the in zero-article

contexts) instances were greater than replacement errors (the-for-a) throughout PL

groups and across tests since most [–def] [–Sp] NPs do not require marking while

almost all [+SP] [–def] NPs are singular and therefore necessitate the supply of the

indefinite article. The stronger tendency to overuse rather than to replace was

certainly more evident in the production of the weaker learners in T1.

The higher error rates in T1 are in contrast to Tarone and Parrish’s (1988)

observation that the accuracy of articles supplied in referential contexts was higher in

243
communicative tasks. The outcome also differs from those of Trenkic´’s (2000) study

on Serbian speakers where the weaker group performed better, in this respect, in the

task that required least attention to form.

As in the case of general overuse errors, the lowest rates of overusing the

definite article in [–def] [±Sp] contexts were scored in T1 due to the same grounds as

those discussed in the previous section (5.2. 2). In T2 the weaker learners made more

overuse errors, yet they maintained the preferences of T1 as they continued to supply

the definite article in [–def] [–Sp] contexts more frequently than in [–def +Sp] ones

but not significantly so. The role of blanks cannot be ruled out as a potential factor

that evidently yields higher frequency (cf. Paradis 1994, Ellis 1994 Trenkic´ 2000,

Snape and Velasco-Zarate 2005) of both articles. Therefore it seems that the increased

production of the indefinite article minimalised the supply of the in the limited

number of blanks in T2, causing the overuse and replacement errors to be almost

identical. The effect of the design of T2 as well as its higher focus on explicit

knowledge known to attract better accuracy rates, is evident in the smaller difference

between replacement and overuse errors of the.

Again, in T3, the overuse rates of the definite article in [–Sp] contexts was

higher than in (indefinite) [+SP] ones, but unlike the results of T2, the difference

between the two errors was significant. It is hardly surprising as T3 has proven to

yield higher significance rates (see 4. 4. 2. 3). The consistency in the error pattern

across tasks suggests that variation in task focus and knowledge type have little effect

on G1’s choices where specificity is concerned. The finding corroborates Dienes and

Perner’s (1999) argument that representations of knowledge are more of a spectrum

244
rather than a dichotomy of extreme opposites. Hence, learners’ responses varied in

number and significance rates but not in pattern.

However, in this regard, our results contradict findings from many studies on

L2 learners of –ART L1s (Trenkic´ 2007, Kaku 2006, Thomas 1989, Parrish 1987

inter alia) in which specificity was found to be associated with definiteness at earlier

IL stages. On the other hand, studies on +ART languages in this regard indicate that

there is no significant overuse of the in [–def +Sp] contexts (Snape et al 2006,

Hawkins et al. 2005). However, the question of whether definiteness is associated

with non-specificity in +ART languages in which the definite form is used generically

(e.g. Greek, French) has not been often addressed nor thoroughly investigated. It is

also worth mentioning that in the most widely cited studies on L2 specificity-

definiteness confusions are based on data collected through controlled tests with focus

on form, and not many have analysed communicative output from free production.

Thus, when the findings of this study are compared to other findings, the criterion of

task type/focus must be taken into consideration.

In short, statistical outcomes suggest that learners’ initial tendency is to

associate the grammaticalised form of definiteness with the semantic value of non-

specificity since erroneous overuse before (indefinite) [–Sp] referents was higher than

overuse in (indefinite) [+SP] contexts, especially in the task that focuses more on

meaning and examines learners’ implicit communicative competence. This clearly

indicates that the influence of the Arabic (see 2.3.3) in which the definite article is

used to generalise as well as to specify (see also Hawas 1989, Maleej 2004, Kremers

2003) is stronger at lower stages and more evident in less controlled production.

245
5.3.1.2 G2

There was a marginal difference between the overuse rates of the definite article in

non-specific indefinite contexts and in [–def +Sp] contexts in the intermediate group’s

performance in the writing task. In that sense, G2’ performance paralleled the pattern

observed in the performance of the lower group (in T1) since these participants

associated the with non-specificity more frequently than with specific references,

although significance rates between the two overuse errors was much lower than G1.

In T2, learners’ choices continued to evince a preference for the definite

article in [–Sp] contexts, yet, with more focus on accuracy and awareness of task

demands the error rates were closer, i.e., the number of overuse instances in specific

indefinite contexts approached the number of overuse instances in non-specific

references. This means that G2 learners’ performance in T2 was closer to that of the

advanced group. The approximating number of overuse of the definite article in ±Sp

environments suggests a state of hesitation and uncertainty typical of intermediate

stage of IL in which hypotheses formulated in earlier experience are questioned.

Since our results do not tally with findings of known L2 studies in this respect,

the possibility arises of a strong L1 influence especially at lower PLs. G2 learners

seem to be having difficulty in breaking away from the Arabic notions of specificity in

which the definite article /al/ is the least explicit marker (Aziz 1993:132) that serves to

denote generalised conceptions as well as specific (cf. Kremers 2003:53, 169).

Behrens (2005) postulates that the use of the in Arabic is expected in generic contexts

because it is a discourse referent-marking language in which non-referential

expressions are conceived as definite since they are perceived as part of people’s

established epistemic knowledge. In this sense, weaker L1Ar learners are possibly

246
trying to de-link definiteness from non-specificity triggered by the L1 settings in the

same way that Chinese learners go through a mental process of de-linking definiteness

from plural number marking as part of their IL process towards TL attainment

(Lardiere 2007).

In T3, the intermediate learners overused the definite article more in specific

contexts than in non-specific; i.e., they were better at recognising overuse instances

when the reference was non-specific. Although this constituted an error of overuse in

both cases, it is indicative of a shift in learner hypothesis from associating definiteness

with non-specificity to associating definiteness with specificity, which, in a way, is a

form of progress as learners are progressively adjusting their conceptualisations

towards a more target like setting.

It is not surprising that learners’ preference to overuse the definite article in [–

def +Sp] contexts was more evident in T2 than in T1, since, as a cloze test, it can

reveal more explicit knowledge of L2 rules which participants could have learned

through negative evidence in the EFL classroom. However, the results in T2 are less

extreme than those that emerge from T3 in the sense that although overuse instances

of the were more frequent in [–def +Sp] contexts, no significant differences were

found between the errors across the two NP environments in T2. This is possibly

attributable to the design of T2 where the availability of context is a common

denominator with T1, which helps drive the participants’ attention more in the

direction of meaning as a goal (Richards, J 1976, Dulay et al. 1982, Krashen 1982,

Hulstijn and Hulstijn 1984, Granfeldt 2000) and eventually balances the focus of this

test between the two extremes of T1 and T3. Since the L1 transfer is more operative in

meaning-based environments, non-referential overuse of the definite form is more

247
likely to reoccur, reducing the contrast between the two specificity contexts. On the

other hand, being more direct and form-focused than T1, T2 raised the participants’

awareness of specificity as a precondition to most L2 meanings of the definite form.

This set participants to try and retrieve particular information about referents directly

available in the preceding/following NPs. Once this particular information is

recovered, it lowers the probability of ambiguity, inherent in genericity (Krifka 2003)

and thus specificity becomes a more favourable option.

Hence, it is conceivable that the learning curve with respect to specificity

seems to go beyond the cross-sectional level to include task type variation. The way in

which G2’s performance changed across tasks is an indicator of progress towards TL

production in the same way the improvement in the PL affects the results. While the

production of the intermediate group clearly paralleled the choices of weaker learners

in T1, overuse rates in [+/-Sp] contexts were almost identical in T2. In T3 however,

the responses of group 2 participants were more in line with the advanced group in

associating definiteness more with specificity. Therefore, it would be safe to propose

that G2’s performance on specificity is weakest in the most communicative task and

strongest in the most controlled, form-focused test.

5.3.1.3 G3

The advanced group performance also showed a similar pattern to that of G2 across

tasks in overusing the definite article more frequently in non-specific than specific

contexts. The oversupply of the definite article in [–Sp] contexts was more evident in

T1, while the two controlled tasks yielded a reduced rate of overuse in indefinite [–Sp]

contexts and an increased tendency to associate the definite article, rightly or wrongly,

248
with specific referents. This pattern goes to prove that the TL terms of specificity were

possibly drawn from learners’ conscious knowledge about the L2 rather than their

automated ability to supply the article. Nonetheless, advanced learners made fewer

overuse errors than the two lower groups (see 4. 4. 2. 1), and did not associate

definiteness with non-specificity as often as the weaker learners did. This confirms

Huebner’s (1983) proposal that dropping the definite article from non-referential

contexts is not mastered until later stages (Tarone and Parrish 1988, Ekiert 2004).

Rather, G3 participants’ overuse rates of the definite article in [–def +Sp] NP

environments in T1 almost matched those in [–def] [–Sp] contexts. Thus, G3

participants were ahead of the intermediate group in their implicit ability to

appropriately associate the definite article with specificity.

In T2 most overuse instances that this group made were in specific contexts,

but in the GJT, specific overuse was significantly higher than generic overuse. In

short, G3 started with a slight prevalence of the generic the in T1 and moved towards

specificity in T2 but with higher focus on explicit knowledge in T3. In short this

group exhibited a better understanding of the relation between specificity and

definiteness in English. Therefore, it seems that the more controlled the task is, the

more specific the references become. Although what is being compared is two

contexts of the same error, relating definiteness to specificity can be regarded as a step

towards the TL.

Although the difference between overuse of the in specific and non-specific

contexts in T1 was not statistically significant, the fact that the rates were higher in [–

Sp] environments constitutes a crucial argument against the FH (see section 2.5.1).

SLA research has shown that the overuse of the definite article in [–def] [–Sp]

249
contexts occurs indiscriminately at the flooding stage which takes place at earlier IL

levels (Huebner 1983, Master 1987). Conversely, overuse in these contexts persisted

in the production of L1Ar to higher PLs when this stage was supposed to have been

passed. However, fluctuation between definiteness and specificity for learners from –

ART L1s is challenged by other researchers. According to Kaku (2006) advanced

Japanese learners were not sensitive to reference because specificity was not

associated with the definite article. It is not clear whether Kaku’s finding is based on

lack of understanding of specificity and definiteness or perhaps on the extended

influence of the L1 into advanced IL stages.

Since the function of the Arabic definite article (al) is to generalise as well as

to specify (Kremers 2003), an alternative reading of L1 influence presents itself as an

explanation to G3’s performance in the writing task, which better reflects learners’

implied understanding of the definite form. The argument that the L1 has a stronger

role in causing this error, namely the overuse of the definite article in [–Sp] contexts,

is supported by findings from other studies on L1Ar learners such as those by Maalej

(2004), Hamdallah (1988) and Kharma and Hajjaj (1989).

5.3.1.4 Conclusion to Specificity

The results of this study indicate that an overwhelming majority of participants

employed the indefinite article only referentially, while most overuse instances of the

definite article were in non-specific (indefinite) contexts. Therefore it seems that

participants have employed articles in a manner that corresponds to the settings of

specificity-definiteness in Arabic whereby referential meaning is largely indefinite

250
while generic reference is entirely definite (Maalej 2004, Kremers 2003). In other

words, indefinite NPs are mostly specific but definite descriptions can carry both

specific and non-specific readings.

As the definite article does not appear to be often overused in (indefinite)

specific contexts by weaker and intermediate learners, and its overuse in [–def +Sp]

contexts is limited to form-focused tests, it is possible to corroborate the proposal

forwarded by Hawkins et. al. (2005) on the existence of a dual ACP41 whereby

learners from +ART L1s do not fluctuate between specificity and definiteness (see

section 2.5.1). However, this does not by any means prove that L1Ar learners

hypothesise that definiteness, per se, is non-specific. Rather, we propose that the

misuse of the definite article in non-referential contexts by Arab learners is caused by

L1 processing constraints (Zobl and Liceras 1994) whereby the grammaticalised form

of definiteness is employed in non-referential meanings as well as referential. The

generic definite article in Arabic functions as a meaningless filler (Lyons 1999:290)

and thus, similar to Tsimpli’s (2003) description of the generic definite article in

Greek, it lacks the semantic value while possessing a purely grammatical function.

Hence, it is possible to confirm the existence of a fluctuation phase in learner

hypothesis between the notion of specificity and definiteness marking but with

outcomes reversed to those originally proposed by IKW (2003, 2004).

Developmentally, results indicate that learners start with the assumption that

most generic references, not necessarily ‘uniquely identifiable’, are marked with the

definite article. This is probably due to the attributive function of the definite article in

Arabic. The overuse starts to decrease with improved PL as learners begin to doubt

41
Article Choice Parameter (IKW 2004). See Literature Review.

251
their former suppositions and drop the definite marking from NPs that bear a –

Specific value. The development is a possible indicator of the gradual abandonment of

the hypothesis that associates article use with specificity. In other words, the

definiteness-specificity map suggests that learners start with a stronger association

between non-specificity and definiteness, which is later weakened by better overall L2

abilities. The shift in preferences is evident in G3’s higher overuse of the in

[+SP](indefinite) contexts. It signals the constant change in learners’ understanding of

the meaning and use of the definite article and that their hypotheses are under

continual revision. Although the form of the definite article is acquired at early PLs, it

takes learners time to realise the subtle differences between its function in the L1 and

the target language (cf. Ellis 1985).

It is possible that L1Ar learners are experiencing a Salish stage42 of language

acquisition (Mathewson and Schaeffer 2000) during which the definite article is used

without its presuppositional value i.e. without the familiarity/common ground

condition (Heim 1982). The progress towards associating the definite article more

with specificity and less with genericity is attained at higher PL and visibly more

evident in controlled tasks. This could be interpreted in terms of increased sensitivity

to the uniqueness element in the meaning of the English definite article which restricts

its referential domain to identifiability of a particular person/thing/situation etc.

(Hawkins 1987, Lyons 1999).

Other studies on L1 Arabic confirm that there is higher overuse of the definite

article in non-specific contexts (Kharma and Hajjaj 1989). A comprehensive study

42
Mathewson and Schaeffer (2000, 2005) propose that language learners go through a Salish stage
during which their determiner system parallels that of Salish languages (e.g. the language spoken in
British Columbia) where no determiners are required to represent the familiarity component of
definiteness, while the ±Sp distinction is formally represented.

252
carried out by Maalej (2004) on the use of determiners in Tunisian students’ L2

compositionssuggests that what impedes progress most is the use of the with indefinite

plurals; which confirms a distinct association between non-specificity and

definiteness. Batainah (2005) also found that Jordanian learners of English never used

the indefinite article non-referentially. However, not all researchers agree that L1Ar

learners overuse the definite article more frequently in non-referential contexts. By

comparing the performance of two sets of learners; L1 French and L1 Arabic, Sarko

(2008) found that Arab learners selected the in [–def +Sp] NP environments more

frequently than in [–def –Sp]. These results differ from the results found in T1 where

non-referential overuse of the definite article exceeded the error rates in [–def +Sp]

contexts. We propose several factors that might have contributed to the discrepancy

between our results and Sarko’s. Firstly and most importantly, the method of data

collection that Sarko employed was a forced-choice elicitation task while most

instances of non-referential overuse in this study are recorded in the free production

test. However, Sarko’s findings correspond more closely with the results emerging

from our cloze and GJ tests. Secondly, there is the range of the participants’ PL. The

weakest group in Sarko’s study was of a lower intermediate level, while our weakest

group included participants who tested ‘elementary’ by the OPT bands. This affects

the results since the higher the PL, the more specific the choices were. It is worth

mentioning that the Oxford Placement Test was used to determine proficiency levels

in both studies. Finally, the presence of directions in Sarko’s study primed participants

to insert the articles (a, an, the, Ø) in the gaps, while our participants were unaware of

the exact purpose of taking T1. These reasons may fully account for the difference in

the results and the more target-like performance of the Syrian participants. Therefore,

in order to compare our results with other studies, it is essential that the variables of

253
method, task focus and selected sample of participants be taken into consideration to

ensure reliability and validity of the samples before drawing on similarities and/or

differences (Skehan 1989).

As a confirmatory measure of substantiating the validity of our findings, the

FH was tested on the indefinite article as well as the definite to determine whether

learners assumed that it was associated with non-specificity or not. We found that the

generic use of the indefinite article was almost non-existent in the written production

task and extremely limited to just a few instances in the cloze tests. Rather, the

majority of our participants, especially weaker learners, opted for the definite singular

form as a favoured option to deliver generic meanings. Contrary to the FH, the use of

the indefinite article by most learners of variable PLs was mainly in [–def +Sp]

contexts.

This further confirms that the settings of the ACP as proposed by IKW are

reversed for L1Ar learners, as it is indefiniteness, rather than definiteness, which is

more associated with specificity in learners’ hypothesis on specificity.

Non-specific use of the indefinite article in general has been reported as one of

the common errors in SLA studies on –ART L1s (Cziko 1986 (cited in Habuto 2000),

Bickerton 1981, IKW 2003/4). However, Trenkic´ (2000, 2002) posits that L1 Serbian

learners of English consistently attributed referential meanings to the indefinite article.

Our results in this regard are in line with Trenkic´’s as they reveal only few

occurrences where the generic function was expressed by the indefinite article.

Moreover, these instances were limited solely to responses from G1 in the controlled

tasks. We believe that these responses were incidental because none of the participants

produced such expressions in the free writing task. Therefore, there is not enough

254
evidence to substantiate the claim that non-specificity is associated with indefiniteness

in learners’ hypothesis, apart from the very few task-induced occurrences where the

indefinite article is used non-referentially. Rather, the fact that the vast majority of

indefinite article use was noticed in specific referencing refutes the FH as originally

advocated by IKW (2003).

More recently, some SLA studies compared between two groups of learners

from ±ART L1s. Hawkins et al. (2005) and Snape et al (2006) found that learners

whose L1s lacked articles showed patterns of fluctuation but the groups from +ART

L1s correctly associated definiteness, rather than specificity, with article use. In a

study on Mexican and Russian learners of English Ionin et al. (2008) found that L1

Russian learners fluctuated between the concepts of specificity and definiteness

whereas the Spanish (+ART) L1 learners did not.

The argument for L1 influence fails if the scarcity of non-specific occurrences

of a(n) in T1 is taken into consideration. Attributive employment of the indefinite

article was exclusively limited to T2 and T3, while none of the participants,

volunteered to use singular indefinite constructions non-referentially in their free

composition. Furthermore, it is the singular form, rather than indefiniteness, that is

often used to deliver generic meaning in Arabic (see Section 2.3.3).

The generic use of the singular indefinite is an example of an error that is not

caused by negative transfer. Such errors exist in IL (Richards 1971, Nemser 1971,

Gass and Selinker 2001) and are considered part of the L2 acquisition process. The

supply of the indefinite article in [–Sp] environments is neither L1 induced as there is

no explicit indefinite marker in Arabic, nor is it common practice in English to refer

generically by means of the indefinite article, except in certain language communities

255
(cf. Maurcia and Freeman 1999, Abbott 2001, Hong and Seonkyung 2001). Thus, a

more probable explanation for this use is that erroneous L2 production which cannot

be justifiably traced to the learners’ L1 or the TL reflects arbitrary attempts frequently

observed during the developmental stages of interlanguage (Richards 1971, Selinker

1972, Gass 1996, Odlin 2003).

As participants from all PL groups systematically used the indefinite article in

specific contexts in T1 and T2, the exceptional preference to use the indefinite article

to deliver generic meaning in T3 was logically and primarily a result of task variation

in both focus and form/design. The direct prompts in the cloze test and the underlining

of NPs in T3 alerted students to the importance of supplying articles. However, while

the blanks in T2 caused random oversupply (see 5.2.1.2) the underlined NPs in T3

yielded maximum attention to form. Furthermore, its de-contextualised design

increased the focus on accuracy rather than meaning and the reliance on explicit

knowledge where the indefinite article is more operative in the L2.

5.3.2 Premodification

When L1Ar learners fail to provide the indefinite article before singular unidentified

NP contexts, they might be assuming that the absence of the definite article indirectly

signals that the NP bears a [–def] semantic value (see section 5.1.2.1). This was

evident through the higher omission errors of the indefinite article across tasks.

However, when adjectival premodification, or the lack of it, affects omission patterns,

the question of word-order effect arises as it might be the cause underlying the

variability in learner performance (cf. Odlin 1989). We assume that the difference in

the syntactic position of adjectival modification in the two languages presents learners

256
with an additional constraint to accommodate L2 syntactic structures. Various

researchers have proposed that the L1 structure can be transferred to L2 production

(Odlin 1989, Brown 1991, Parodi et al. 1997). SLA studies have also found that the

syntactic properties of the L1 affect the use of articles in the L2 production, (Young

1996, Jarvis 2002, Prévost and White 2003, Trenkic´ 2008).

In investigating L1Ar learners’ hypothesis regarding the use of the indefinite

article in premodified constructions, there are three aspects of the L1/L2 contrast

which need to be taken into account: the absence of an explicit indefiniteness marker;

the fact that adjectives agree with nouns in number and (in)definiteness, and finally

that the syntactic position of adjectival modifiers is post nominal in Arabic. In this

section, we will review how learners of different proficiency levels responded to

modification across the three tasks.

5.3.2.1 G1

Unlike studies on –ART languages (Trenkic´ 2008, Jarvis 2002, Young 1996) which

reported greater difficulty in supplying articles before premodified nouns, the results

of this study confirm that L1Ar learners, especially at lower PLs, had more difficulty

in providing a(n) when the head noun was not premodified than when it was in

premodified environments.

The higher overuse of the indefinite article with adjectivally premodified

nouns in all tasks and by all groups invalidates the proposition that articles and

adjectives compete for the determiner position (Trenkic 2007). Furthermore, it is not

possible that even the lower PL learners assume that the adjective has occupied the

syntactic position of the determiner, since adjectives post-modify nouns in Arabic and

257
the definite article is already present and grammatically realised in Arabic as a

prefixed clitic to adjectives as well as nouns. Hence, there were few occurrences in

learners’ free composition where the indefinite article was graphically part of the

adjective as shown in the following example.

1. *Al-Ain is anice town

This is an obvious case of prosodic transfer of the L1 syntax whereby the newly

acquired article is prosodified as a prefixed clitic in the same manner that its

counterpart is presented in Arabic.43 Similar examples have been observed by EFL

teachers of introductory writing courses in the UAE University. This error has also

been reported by other researchers as one of the most common errors in L1Ar

learners’ written production (see also Batainah 2005).

G1 omitted the indefinite article more frequently in non-premodified nouns

across tasks and uniformly overused it in premodified constructions with no distinct

variation caused by task type. Similar findings were reported in studies on L1Ar

learners such as Hamdallah (1988) and Maalej (2004). However, the effect of task

type was noticed in the higher occurrences of omission errors in T1 and the lowest

being observed in T2, which is line with our overall findings of task effects (see

5.2.1).

Finally, a few instances were recorded whereby weaker participants used the

indefinite article with adjectives without a noun in the free production test. This

phenomenon was also noticed by Batainah (2005) as Jordanian learners also used the

indefinite article with adjectives without a noun. Hence, based on this distinct error
43
Selkirk (1996) forwards that functional material can be prosodified in the form of affixal clitics (such
as the case in Arabic) or free clitics (English).

258
pattern, it is possible to propose that the indefinite article is possibly associated with

the adjective rather than the noun in Arab learners’ IL hypothesis. This supports

Hawkins and Chan’s (1997) theory of failed functional features hypothesis (FFFH) as

they proposed that the difference in L1/L2 syntax constrains the representation of the

indefinite article in L2 production.

5.3.2.2 G2
With regards to premodification, the intermediate group was the only group whose

omission patterns were sensitive to task type while the weaker and the more advanced

groups were unanimous in their choices and preferences.

In G2’s free composition, the indefinite article was better supplied with bare

nouns and omitted more often before adjectives. Therefore, G2’s choices were closer

to those of the advanced group. On the other hand, in T2 and T3, the indefinite article

was frequently omitted before bare nouns while better supply was recorded in

premodified constructions, a performance which renders the responses of G2 to be

more in line with those of the weaker group. This is another instance whereby G2

results approximated those of G1 in the free production task (see section 4.4.4.2). This

distinct error pattern is incongruous with the usual patterns observed in G2’s

production in the controlled tests where the results of the intermediate group was more

compatible with the performance of the advanced (G3). On a closer look, this

behaviour would not seem unusual since learners’ ability to use the articles in

premodified contexts in English is mostly derived from their conscious knowledge

about the L2 which form-focused tasks examine. This ability exceeds their skill to use

articles properly in communicative tasks in which they rely more on the subconscious

knowledge from the first language to express meaning. In other words, as the cloze

259
and GJ tests examined learners’ explicit knowledge, it is more likely that learners

would be more aware of the L2 grammatical rule which relates the function of the

article with the head noun, rather than the adjective that it precedes. The blanks could

have alerted learners to the necessity of supplying the indefinite article by encouraging

them to actively seek for the singular noun that follows the adjective in the NP before

supplying the required article. Similarly, in T3 the focus being limited to the sentential

scope and the inclusion of the head noun in the underlined part of the sentence could

have contributed to better recognition of the relevance between article and noun,

without being misled by the adjective that separates the two.

More probably, the basis for the incompatibility of the error pattern recorded

in this respect with other observations noted in the responses of learners to the three

tasks is that what is being considered here is not a comparison of accuracy or error

rates across tasks, but rather a comparison of learners’ reaction to two different NP

environments in which the same error could possibly occur.

However, in all the tasks there was little difference between error rates in

premodified and non-premodified contexts. This is perhaps because uncertainty and

indecisiveness are typically characteristic of the intermediate stages of language

learning. Nevertheless, the tendency to appropriately employ the indefinite article in

modified contexts was more apparent in the form-focused tests. These results are

incompatible with Trenkic´’s (2000, 2008) study in which Serbian speakers of all PLs

were statistically more likely to omit articles before nouns modified by adjectives than

with non-modified nouns (pp. 232. emphasis added) .

The map of overuse in premodified contexts entails an association between

article and adjective. However, this was limited to the indefinite article, while the

260
Missing Surface Inflectional Hypothesis (Prévost and White 2000) cannot account for

L1Ar learners’ errors with regards to the definite article.

5.3.2.3 G3

The advanced group’s performance with regards to modification was unanimous

across tasks. Participants from G3 supplied the indefinite article significantly better in

non-premodified contexts while higher omission rates were recorded before pre-

modified nouns. These choices mark a total shift from the model in the results by the

weak group. It means that with better L2 proficiency, learners’ conceptualisation of

cardinality/individuation on the one hand and the English sentence structure on the

other has possibly led to a disassociation of the adjective from the indefinite marker in

learners’ IL hypothesis. Thus, this group’s behaviour is more in line with findings of

other SLA studies.

Although the advanced group’s performance was consistent in all three tasks,

the participants supplied the indefinite article better in T2 than in the other two tests,

while their weakest performance was in T1. These results are consistent with the

overall omission patterns according to task type (see 5.2.1.1) as T1 revealed the

highest omission rates while the lowest were in T2.

According to Trenkic´ (2000, 2008), higher omission occurs in adjectivally

premodified constructions because the determiner position is occupied by the

adjective which renders the article supply unnecessary. In other words, omission

occurs because adjectives and articles are competing for the same modifier position in

learners’ IL grammars. This explanation, however, does not apply to L1Ar learners’

hypothesis as the syntactic position of a determiner is already established in their

261
subconscious through the existence of a definite form in Arabic. Goad and White

(2004) maintain that the failure to supply the indefinite article in premodified contexts

is a result of adjectives acting as distracters from the head noun. The latter explanation

is more plausible with respect to Arab learners because of the postmodifying position

of adjectives in Arabic syntax, and could therefore stand as a valid rationale to the

existing difficulty.

The apparent similitude of this study with findings from other studies that

address the effect of premodification on L2 article production (e.g. Parodi et al. 1997,

Trenkic´ 2000, Prévost and White 2000, Sharma 2005, Pongpairoj, 2007,) is put into

question once the PL is taken into consideration. In the aforementioned studies, this

inclination was noticed at lower PLs while in this study, it was the advanced learners

who supplied the indefinite article better when the head noun was – Premd and

omitted it more when the head nouns were adjectivally premodified. Hence, although

the error pattern is similar, the development of learners’ hypothesis is inconsistent.

Through the progression observed in the error patterns of the three PL groups,

it is more plausible that the cause of frequent omission of a(n) in modified

constructions by G3 is a form of hyper-correction rather than a competition between

article and adjective for the determiner position. Once learners begin to internalise the

grammatical rule that relates the indefinite article to the noun rather than the adjective,

the supply in obligatory contexts improves. However, at later IL levels, the newly

acquired rule becomes overgeneralised to the extent that makes learners supply the

article appropriately before indefinite singular nouns but refrain from doing the same

when they are presented with adjectives. Although both responses, regardless of

premodification, are errors of omission, we assume that the shift to omission in

262
premodified constructions signals a more advanced learning stage of internal

processing of L2 rules.

SLA research has confirmed that switch of error types is a significant marker

of L2 progress as learners are not known to attain TL levels before going through

various stages of different types of errors (Corder 1967, 1981, Selinker 1972, Ellis

1985, Prévost and White 2000, Towell and Hawkins 1994). Therefore, when omission

instances occur more often before bare nouns, this suggests an association of the

indefinite marker exclusively with the adjective, whereas omitting the indefinite

article in [+Premod] environments indicates a more target-like assumption based on

the realisation that the indefinite article modifies nouns, not adjectives.

5.3.2.4 Conclusion to Premodificiation

Our findings corroborate the PTH44 proposed by Goad and White (2004) whereby

learners’ appropriate use of articles in premodified constructions is influenced by

constraints on prosodic structures carried over from the L1. In Arabic, the constraints

are represented by the absence of an overt indefiniteness marker and the post-nominal

position of adjectival modification that contribute to the low suppliance of functional

morphology (ibid) in L2 production. A combination of the aforementioned factors has

restricted the weaker learners from supplying the indefinite article before singular

indefinite nouns.

The structural proximity between determiner and head noun as well as the

post-modifying position of adjectives in Arabic makes adjectivally premodified

structures in English more challenging especially to weaker L1Ar learners. G1 tended

44
Prosodic Transfer Hypothesis

263
to supply the indefinite article better in adjectivally pre-modified constructions as

premodified singular constructions (a(n) + Adj + N) are likely to have been learned as

chunks in elementary stages, while non-modified head nouns are left unmarked,

ensuing a higher omission rate.

In a study on Swedish learners of French, Granfeldt (2000) found that the

presence of an adjective did not affect determiner production negatively. Actually,

higher omissions of determiners (13.5%) were recorded in non-modified constructions

while only (2.8%) of determiners were omitted in adjectivally premodified contexts.

The lack of structural similarity between the L1 and L2 in presence/absence of

explicit marking of singular indefinite NPs on the one hand and the licensing of

grammatical definiteness of adjectives in [the + adj + the + noun] constructions in

Arabic on the other, might have led L1Ar learners to the misconception that the

indefinite article can be attributed to the adjective in the same way as the definite

article is affixed to the adjective in the L1. The fact that adjectives are grammatically

definite in Arabic could have led learners to assume that it is the adjective, rather than

the noun, which the articles modify (Hamdallah 1988). The free composition test

exhibits multiple examples of how learners have graphically attached the indefinite

article to the adjective which corresponds to the affixal definite marking of adjectives

in the L1. Batainah (2005) concedes that one of the trends she recorded in the

production of Jordanian learners of English was ‘writing the indefinite article as part

of the noun following it’ (P.65).

An opposite pattern was observed in the performance of the two upper groups

who produced a(n) better with non-modified head nouns. This marks a shift in

learners’ conceptualisation of L2 structure. With more input and instruction, more

264
advanced learners’ awareness of the necessity to supply the indefinite article before

singular indefinite nouns increased. It is a step farther from the effects of L1 and more

in line with IL tendencies observed in other studies. The higher omissions in

premodified contexts are in line with findings from other studies including those on –

ART L1s. This error is explained by researchers as a structural representation deficit

where adjectives can become distracters, replacing the determiner position in learners’

cognition, misleading them away from the necessity of obligatory suppliance in that

particular structure.

It is not only the omission errors that suggest an association between the

indefinite article and the adjective in learners’ hypothesis, but the hypothesis is further

consolidated by the statistical records of overuse errors. The indefinite article was

consistently overused by all learners across tasks more frequently in premodified

environments. These tendencies were more evident in the communicative test and in

the production of the weaker group when a stronger L1 influence is operative (Jarvis

2000, Haznedar 2001, Siegel 2003, Odlin 1989, 2003). On the other hand, the

performance of the intermediate and advanced level groups was not quite as

discriminate. Parallel to the decreasing rates of omission errors before bare nouns, G2

and G3 learners overused the indefinite article less frequently in premodified contexts.

In their composition, better learners became more reluctant to overuse the indefinite

article with premodified nouns, but due to their improved ‘analysed’ knowledge about

language, even fewer overuse errors were observed in the cloze and GJ tests. This

means that as PLs improve, learners begin to disassociate the indefinite article from

the adjective.

265
In general, the L2 acquisition process of the indefinite article observed in this

study (see section 4.4.1.2) suggest that there was a gradual and systematic

improvement from failure to supply to a stage of a better supply and finally a tendency

to overuse. However, where premodification is concerned, this tendency is reversed;

i.e., learners start with a better supply of the indefinite article when a noun is

adjectivally premodified and shift gradually, as their PLs improve, to omitting the

indefinite article more often in such constructions. This shift in learner performance

signals a newer stage in learner hypothesis regarding the structure of L2 in which

learners begin to dissociate the article from the adjective and increasingly link it to the

head noun itself.

Variability in L2 article use is proposed to be caused by the syntactic structure

of English where adjectives are placed between the determiner and the head noun, a

position that can be misrepresented, according to Trenkic´ 2007, for the determiner

itself leading to omissions in premodified constructions. SD, the Turkish learner of

English, in White’s (2003) study was reported to have omitted the indefinite article

significantly more frequently in premodified constructions (70%).45 However, what is

happening with the L1Ar learners is quite the contrary, whereby higher overuse of

both articles is observed at lower PLs while more omissions are evident as learners’

L2 improves.

Adjectival premodification exerts a negative influence on the production of

articles caused by the difference in syntactic representations in the L1 and the TL.

Learners seem to apply L1 syntax to L2 lexical items (Hawkins and Chan 1997).

Therefore, a dissimilar error pattern emerges from the one reported by studies on

45
. Turkish has the same prosodic structure of Arabic regarding the position of articles as affixal clitics.

266
learners from –ART languages. The variation is noticeable in two main trends. Firstly,

L1Ar learners overused, rather than omitted the indefinite article in premodified

contexts. Secondly, L1Ar learners failed to supply the indefinite article only, rather

than both articles in non-premodified constructions. Therefore, contrary to Trenkic´’s

(2007) explanation, we propose that in weaker L1Ar learners’ hypothesis, the errors

are caused by the association of the indefinite article with the adjective, and not

because the article and the adjective are competing for the determiner position.

5.3.3 Concreteness

Similarities and/or differences in lexical connotations of nouns between the L1 and the

L2 have been reported to affect L2 production (Trenkic´ 2000, Kuribara 2003). Some

went as far as to propose that it is the lexical -rather than functional- categories that

are solely transferable from the L1 to the L2 grammars (Vainikka and Young-

Scholten1994, 1996). Since no significant differences were found in the outcomes of

the three PL groups, the discussion of the results will be confined to the variation

found across tasks.

Previous studies on SLA articles provide evidence for the existence of a

misconception in L2 learners’ IL grammars in which the ontological status of an NP

affects the choice of articles. Articles have been reported to be used in a more native-

like manner in concrete environments before being mastered in abstract contexts (Hua

and Lee 2005, Richards 1976). Trenkic´ (2000) reported overuse instances of the

definite article occurring more frequently with concrete nouns, because, according to

Trenkic´, Serbian learners perceived the definite article as discrete and concrete

267
(ibid:293). On the other hand, in research on a newspaper corpus published in English

but written by L1 Japanese journalists, Nagata et al. (2005) found that many abstract

nouns were needlessly preceded by the definite article.

There are variable accounts of the extent of L1 influence on L2 article

production. Supporters of the non-contrastive approach to error analysis propose that

learners from different L1 backgrounds experience similar difficulties in SLA

(Richards 1971, Nemser 1971) and that the effect of the L1 does not account for more

than a third of learners’ errors (Abi Samra 2003). Proponents of this approach in SLA

have observed similarities in the performance of learners from various L1

backgrounds. Paradis and Zdoenko (2008) did a study on learners from ±art L1s and

found that L1 influence was limited in the sense that learners from both –ART and

+ART L1 backgrounds committed the same errors. On the other hand, many

researchers have acknowledged the effect of the first language on learners’

performance.

However, studies on L1Ar learners’ contradict findings of most studies on –

ART L1 learners. Higher overuse rates of the definite article were found in abstract,

rather than concrete contexts. Researchers such as Maalej (2004) and Diab (1996)

have reported a faulty association in learner hypothesis between definiteness and

abstractness. Abstract nouns seem to have attracted higher overuse rates of both

articles. Maalej for example, maintains that the use of the with abstract nouns is one of

the major impediments to L2 progress.

Our results, in line with findings from studies on L1Ar learners, show a

stronger tendency for marking abstract nouns more often with articles than concrete

ones. The tendency to overuse articles in abstract contexts was evident in the

268
production of participants of different PL groups and in all three tasks. The uniformity

of the responses and the similarity with findings from studies on L1Ar learners leads

us to advance a proposal that the misassociation between the ontological status of the

NP and its definiteness is a result of negative transfer from the L1. Similar to reports

by Diab (1996) on Lebanese students and Maalej (2004) on Tunisian students, in this

study, the influence of the grammatical definiteness of abstract nouns in Arabic is

reflected in L2 article production.

Despite the unified response to mark abstract rather than concrete nouns with

articles by all PLs, task type appears to have affected learners’ choices of articles. A

comparison of the overuse patterns of each article reveals that in T1, learners of all

PLs overused the definite article significantly more frequently than the indefinite to

mark abstract contexts. T2 results exhibited similar outcomes with the exception of

results from the weaker group which indicated slightly higher overuse rates of a. In T3

however, this preference was reversed. There was a shift towards marking abstract

nouns significantly more frequently with the indefinite rather than the definite article.

The switch from one extreme to the other can only be explained in terms of task type

and task focus since there was no discernible variance in the reaction of participants of

different PLs.

In T1, learners produced almost twice as many concrete as abstract nouns. This

could be the outcome of the prompts to describe hometowns which entails writing

about places, tourist attractions and people. All of these ‘things’ or first order entities

(Lyons 1999) tend to be concrete (ibid). More importantly, learners have shown a

strong preference to construe abstract notions as grammatically definite in T1 ensuing

higher overuse of the definite article. This confirms earlier proposals that unprompted

269
free production tests make L2 learners rely mostly on hypotheses subconsciously

influenced by their L1 –in which case all abstract notions are grammatically definite–

to communicate meaning.

On the other hand, the indefinite article was the more favourable option in

marking abstract contexts in T3. This choice is in line with learners’ general

performance in T3 where higher awareness, recognition and acceptability of the

indefinite article were noticed. Therefore, despite the prominent preference to mark

abstract rather than concrete nouns with articles, it is obvious that L2 article choice

was strongly affected by the shift in the focus of each task.

The odd phenomenon of G1 in T2 can be ascribed to the focus on conscious

knowledge in which the indefinite article is known to be better employed. In addition,

the design of the task with gaps seemed to have encouraged arbitrary overuse of either

article. Both of these factors had a stronger affect on weaker learners (see 5.2.1). It

would also be possible to adopt Trenkic´’s interpretation of this exceptional outcome

especially if we accept that Arabic could be categorised as a –ART language with

regards to overt indefinite marking. However, since similar behaviour was not

observed in the free production of G1, this interpretation cannot account for the close

overuse rates of the two articles in T2.

Judging by the results from T1 and T2, Trenkic´’s proposal of association

between the definite article and discreteness/concreteness in learners IL cannot be

plausibly accommodated; neither could findings from studies on Russian and Chinese

learners whereby article use was deemed more felicitous in concrete environments (cf.

Richards 1976, Hua and Lee 2005). Rather, learners have more probably resorted to

270
grammatical representations of ontological notions from their L1 and mapped them

onto the L2 syntax.

Similar to previous findings, the development in learner-hypothesis with

regards to the effect of the ontological environment on article use can certainly be

followed through the effect of task types –rather than PLs- on article choice. The three

tasks ranged in the type of knowledge they examined from the implicit (in T1) to the

most explicit (T3); in the degree of control from the minimum (where no directions to

article use were made in the free composition task) to the maximum control in T3

where learners were only allowed to judge statements without any contribution to the

test and finally, in terms of focus, as tasks ranged from the highest attention to

communicating meaning (T1) to the form-driven, underlined NPs in T3 (see 3.3)

Learners’ responses followed these criteria as they have all marked abstract nouns

with the definite article in T1; with both articles in T2 but comprehensively preferred

the indefinite article in T3.

Although what is being analysed in the data is the same type of error, namely

overuse, the shift from choosing the in T1 towards a(n) in T3 can be construed as

some form of progress. In other words, the early meanings of (in)definiteness linked

with the L1 norms of presenting abstract notions were most evident in T1. Learners

gradually moved towards more declarative use in T2 whereby the interpretation of the

definite form is gradually divorced from its L1 representation in abstract

environments. The performance in the last test the participants took (T3) reflects

higher awareness of the grammatical rules governing article use regardless of the

ontological category the NPs belong to. The responses also incorporate a widely

adopted L2 developmental strategy of over-applying a newly acquired feature, namely

271
the indefinite article. Therefore, it seems that more control and higher focus on form

and accuracy help trigger a disassociation process in L1Ar learners’ hypotheses.

5.3.3.4 Conclusion to Concreteness

In learner hypothesis, the ontological category of the NP affects article choice; yet the

association was consistently directed towards abstract rather than concrete contexts.

Learners from all PLs produced almost identical responses, whereby abstract, rather

than concrete contexts were more readily marked with articles. The significant

differences between overuse rates in abstract and concrete contexts demonstrate that

learners from all PL groups associated definiteness more readily with abstract, rather

than concrete notions.

To verify our findings, we investigated the omission rates of the indefinite

article in +/–Concrete NP environments and found that they were remarkably close.

This result confirms the supposition that abstractness is more associated with

definiteness, rather than indefiniteness in learners’ hypothesis.

Definiteness is therefore not associated with concreteness in L1Ar learners’

assumptions; neither is indefiniteness associated with abstractness. Instead, we may

propose that it is the vacuous the (Lyons 1999) that is over applied onto abstract

domains by L1Ar learners due to constraints from the first language most evident in

communicative production.

272
Chapter Six: Conclusion

In this study, we found that L1Ar learners acquire and master the definite article well

before the indefinite. The most difficult aspect of article use to master was the zero

article. Based on our findings, the acquisition order for Arab learners is therefore the

a(n) Ø, which echoes findings from SLA studies on Arab as well as other

learners from various L1 backgrounds.

Larger discrepancy in accuracy rates between a(n) and the was noticeable at

lower proficiency levels. This lends support to the claim that syntactic features

unavailable in learners’ L1 do not emerge until later in SLA, since Arabic is a

language that has an overt definite article but lacks overt marking of indefiniteness.

The acquisition patterns of the two articles were found to be divergent, almost

contrastively so. As the definite article was initially flooded (overused in indefinite

contexts), hyper-correction in later stages led to more omissions in obligatory

contexts. On the other hand the indefinite article failed to be supplied where necessary

by weaker learners, but ended with being overused by G3. Although primarily

attributable to the different functions of articles, dissimilar properties of

morphological representation in L1/L2 NPs could have also had a bearing on delaying

the mastery of the indefinite article. It is also possible that the progression of the

indefinite article in L2 follows the developmental map of the definite article but with

deferred attainment. The process of acquisition could be therefore summarised into

these consecutive stages:

Omission Emergence Overuse Mastery

273
The results were significantly affected by task-type variation. Communication

oriented free production (T1) yielded lower overuse rates of both articles, but higher

omission rates than those found in the cloze and GJ tests. Unprompted, learners

appear to think that failing to supply articles does not hinder communication, or is at

least preferable to the use of a structure which they consider difficult. On the other

hand, higher accuracy rates were observed in T2 and T3. It leads us to believe that

learners’ awareness of L2 rules, formally taught and learnt, is better than their

competence abilities in communicative settings. The influence of learners’ L1 was

more easily noticeable in the free production task.

The design in T2 seems to have enticed learners to randomly oversupply

articles in blanks which led to higher overuse errors and lower omission rates than

recorded in the other two tests. Generally, the accuracy rates in the stories test were

higher than those in T1 as the focus shifted more towards explicit knowledge of L2.

With focus channelled maximally towards grammatically correct performance,

T3 yielded higher accuracy and significantly sharper differences among PL groups.

However, the better achievement observed in the GJT must be carefully interpreted

because, with content limited to the sentential level, the results do not practically

reflect learners’ communicative abilities.

With regards to learner hypotheses, the outcomes of this study suggest that

there is a relation between article choice and specificity. Learners associated the

definite article more often with non-specificity while the deployment of the indefinite

article remained predominantly referential across PLs and tasks.

274
Unlike findings in other studies, premodified nouns accrued higher overuse

errors of both articles especially at lower PLs, while higher omission rates were found

in non-premodified NPs in learners’ free writing. Similarly, the ontological status of

the head noun affected L2 choice of articles. Yet unlike Serbian learners of English

(Trenkic´ 2000, 2002) who associated definiteness with concrete substances, L1Ar

learners overused the definite article more often in abstract NP contexts. Similar

preferences were made with regards to the indefinite article.

The effect of Arabic was most perceptible in learner hypotheses underlying L2

article use, since our results were incompatible with those found in similar studies. We

support the claim that errors, especially at lower PLs, are partly driven by negative

transfer from the L1. Yet many developmental tendencies parallel those reported in

research conducted on learners of various L1s such as the over application of TL

goals, hyper-correction of previous mistakes, simplification and omission. In this

sense, we agree with Sharma (2005) that in error analysis, common developmental

strategies, complement, rather than contradict, L1 influence in L2 production. In our

view, both accounts are reliable at different PLs.

Some findings in this study can provide evidence for UG-based theories. The

L1 transfer being more prevalent in the production of weaker learners -while the

production of higher PLs is in line with that of learners from other L1s- lends support

to the Full Access/ Full Transfer Hypothesis proposed by Schwartz & Sprouse (1994,

1996). Similarly, the delayed mastery the indefinite article can be ascribed to the

Failed Functional Features Hypothesis (Hawkins and Chan 1997) or the

Representation Deficit Hypothesis (Hawkins 2004) since both theories claim that the

representation of an L2 feature which is absent from the L1 constrains second

language acquisition.

275
However, usage-based theories can better account for the available results

since they admit a hierarchy of learnability (Johnson and Newport 1991) in which the

sequence of acquisition is controlled by the lexical-grammatical complexity of the

target component in addition to other elements including learners’ cognitive

development as well as the frequency of input (Ellis 2002). The Processability

hypothesis (Levelt 1989, Pienemann and Håkansson 1999), for example, can justify

the earlier mastery of the definite article as the representation of the concept of

definiteness in the form of the requires less processing effort than that required for

a(n) because the latter lacks the one-to-one relationship between form and meaning

while the notions of countability and number are decisively involved.

SLA studies should consider the integration of a wider range of variants that

affect the L2 process as there is no single interpretation that can account for all the

complexities of language acquisition.

Finally, it is worth mentioning that despite variability observed in L2

production, we believe that the errors participants made were principled rather than

random. The shifts in error types and patterns, as learners’ L2 improved, may be

indicative of progress in IL. Generally, with higher PLs, fewer errors were made and

production became more target-like.

276
References

Abbott, B. (2001): Definiteness and indefiniteness. In L.R. Horn and Greory G. Ward
(eds.), Handbook of Pragmatics, Oxford: Blackwell.
Abbott, B., (2000). Definiteness and identification in English. In Németh T. Enikö, (ed.)
Pragmatics in 2000: Selected papers from the 7th International Pragmatics
Conference (2). Antwerp: International Pragmatics Association. pp 1-15.
Abbott, B., (2003). Specificity and referentiality. Abstract for the workshop: Direct
Reference and Specificity. ESSLLI (15). Vienna.
Abi Samra, N. (2003). An Analysis of Errors in Arabic Speakers’ English Writings.
American University of Beirut. Retrieved from:
http://abisamra03.tripod.com/nada/languageacq-erroranalysis.html
acquisition, accuracy methodology. Working Papers on Bilingualism 14: 47-82.
Toronto: Ontario Institute for Studies in Education.
Alderson, J., Clapham, C., and Steel, D. (1997). Metalinguistic knowledge, language
aptitude and language proficiency. Language Teaching Research 1: 93–121.
Al-Fotih, T. A. (2003), Acquisition of the English articles by Arabic-speaking students.
Indian Linguistics, 64: 157-174.
Allan, K (1980). Nouns and Countability. Language 56: 541-67.
Allan, K (2001). From a to the. Australian Journal of Linguistics, 21 (1).
Allan, K. (1986). Linguistic Meaning. vol. 1, London: Routledge.
Al-Samawi, A. (2004). Redundant terms in Yemeni students’ FL writing: A case study
showing another evidence of language transfer. Journal of Humanities and Social
Sciences, UAE University 20 (1):55-80. United Arab Emirates.
Al-Thubaiti, K. (2007). Age Effects on the Acquisition of Uninterpretable Features by
Proficient Saudi Arabic Speakers of English. Proceedings of the Fifth University of
Cambridge Postgraduate Conference in Language Research (CAMLING).
Andersen, R. W. (1977). The impoverished state of cross-sectional morpheme
Anderson, A. H., Clark, A. and Mullin J. (1991). Introducing information in
dialogues: forms of introduction chosen by young speakers and the responses
elicited from young listeners. Journal of Child Language 18: 663–87.
Anderson, J.R. (1983). The Architecture of Cognition. Cambridge, Harvard University
Press.
Arabski, J (1979). Errors as Indications of the Development of Interlanguage. Katowice.
Polland. Uniwersytet Slasky.
Ard, J. and Hombarg, T. (1983). Verification of language transfer. In Gass. S and
Selinker, L (eds.) Language Transfer in Language Learning. Rowley MA; Newbury
House.
Ariel, M. (1988). Referring and accessibility. Journal of Linguistics 24: 65-87.
Ariel, M. (1990). Accessing noun-phrase antecedents. London: Routledge.
Avery, P. and M. Radišić (2007). Accounting for Variability in the Acquisition of
English Articles. In Alyona Belikova et al. (eds.), Proceedings of the 2nd Conference
on Generative Approaches to Language Acquisition North America (GALANA).
Somerville, MA: Cascadilla Proceedings Project, pp 1-11.
Aziz, Y. Y. (1993). Explicit and Implicit Reference in Arabic-English Translation. Babel
39: 129-150.
Bachman, L., F. (1991). What Does Language Testing Have to Offer. TESOL Quarterly.
25 (4): 671-704.

277
Batainah, R. (2005). Jordanian Undergraduate EFL Students’ Errors in the Use of the
Indefinite Article. Asian EFL Journal.
Behrens, L. (2005). Genericity from a cross-linguistic perspective Linguistics 43–2: 275–
344.
Berman, R. and E. Olshtain. (1983). Features of first-language transfer in second
language attrition. Applied Linguistics (4) 2: 222-234.
Bialystok, E. (1979). Explicit and implicit judgments of L2 grammaticality. Language
Learning. 29: 81–103.
Bialystok, E. (1982). On the relationship between knowing and using forms. Applied
Linguistics. 3: 181–206.
Bialystok, E., and Ryan, E. B. (1985). A metacognitive framework for the development
of first and second language skills. In D. L. Forrest-Pressley, G. E. Mackinnon, and T.
G. Waller (Eds.), Metacognition,cognition, and human performance: Vol. 1.
Theoretical perspectives. San Diego, CA: Academic Press. pp. 207–252.
Bickerton, D. (1981). Roots of Language. Ann Arbor, MI: Karoma Press.
Birdsong, D. (1989). Metalinguistic Performance and Interlinguistic Competence.
Berlin, Germany: Springer-Verlag.
Brinton, L., J. (2000): The Structure of Modern English. Amsterdam/Philadelphia: John
Benjamins.
Brown, H. D. (1994). Principles of Language Learning and Teaching. Englewood Cliffs:
Prentice Hall Regents.
Brown, R. (1973). Development of the first language in the human species. American
Psychologist, 28: 97-106.
Brown, R. (1991). Group work, task difference and second language acquisition. Applied
Linguistics 12 (1) OUP.
Brustad, K. (2000). The Syntax of Spoken Arabic. Georgetown University Press.
Butler, Y.G (2002). Second language learners’ theories on the use of English article: An
analysis of the metalinguistic knowledge used by Japanese students in acquiring the
English article system. Studies in Second Language Acquisition 24: 451-480.
Canale, M. and Swain, M. (1980). Theoretical bases of communicative approaches to
second language teaching and testing. Applied Linguistics, 1: 1-47.
Celce-Murcia, M., and Larsen-Freeman, D. (1999). The Grammar book. Los Gatos, Sky
Oaks Production (2).
Chafe, W. (1976). Givenness, contrastiveness, subjects, topics, and point of view. In C.N.
Li, (Ed.) Subject and Topic. Academic Press: New York. pp. 23–55.
Chafe, W. (1987). Cognitive constraints on information flow. In: Tomlin, Russell (Ed.),
Coherence and Grounding in Discourse. Typological Studies in Language. (6): 21–51.
Benjamins, Amsterdam.
Chaudron, C. (1983). Research on metalinguistic judgments: A review of theory,
methods and results. Language Learning 33: 343–377.
Chaudron, C., and Parker, K. (1990). Discourse markedness and structural markedness:
The acquisition of English noun phrases. Studies in Second Language Acquisition, 12:
43–64.
Chesterman, A. (1991). On Definiteness. A Study with Special Reference to English and
Finnish. Cambridge: Cambridge University Press.
Chierchia, G. (1994). Syntactic bootstrapping and the acquisition of noun meanings: the
mass-count issue. In B. Lust, M. Suner, and J. Whitman (eds.) Syntactic Theory and
First Language Acquisition: Cross-linguistic Perspectives (1). Hillsdale, NJ:
Lawrence Erlbaum.

278
Chierchia, G. (1998). Reference to kinds across languages. Natural Language Semantics
6:339-405.
Chierchia, G., M. Guasti and A. Gualmini. (1999). Nouns and articles in child grammar
and the syntax/semantics map. GALA Conference Proceedings. Potsdam.
Christopherson, P. (1939). The Articles: A study of their Theory and Use in English.
Munksgaard. Copenhagen.
Clahsen, H., Eisenbeiss, S., and Vainikka, A. (1994). The seeds of structure. In T.
Hoekstra and B. D. Schwartz (Eds.), Language acquisition studies in generative
grammar. Amsterdam: Benjamins. pp. 85–118.
Collin’s Birmingham University International Language Database (COBUILD).
Corder, S. P. (1967). The significance of learners’ errors. International Review of
Applied Linguistics 5 (2):161-170.
Cowart, W. (1997). Experimental Syntax: Applying Objective Methods to Sentence
Judgments. Thousand Oaks, California. Sage.
Cowie , A. and Mackin , R. (1975). Oxford Dictionary of Current Idiomatic English.
Oxford: Oxford University Press.
Crookes, G. (1989). Planning and Interlanguage Variation. Studies in Second Language
Acquisition. 11:367-383.
Crystal, D. (1987). The Cambridge Encyclopaedia of Language. CUP.
Crystal, D. (2000). On competence and performance and related notions. In F. Brown, K.
Malmkjaer, & J. Williams (Eds.), Performance and competence in second language
acquisition (pp. 11–34). Cambridge: Cambridge University Press.
Cziko, G. (1986). Testing the language hypothesis: A review of children’s acquisition of
articles. Language 62: 878-898.
DeKeyser, R. M. (2003). Implicit and Explicit Learning. In C. J. Doughty and M. H.
Depraetere, I. (2003). On Verbal Concord with Collective Nouns in British English.
English Language and Linguistics 7: 85-127 Cambridge University Press.
Diab, N. (1996). The Transfer of Arabic in the English Writings of Lebanese Students
The ESP, São Paulo, 18 (1): 71-83 editors, European Second Language Association
Yearbook: Volume 2.
Dienes, Z., and Perner, J. (1999). A theory of implicit and explicit knowledge.
Behavioural and Brain Sciences, 22: 735-755.
Donnellan, K. (1966). Reference and Definite Descriptions. The Philosophical Review
75: 281–304.
Doughty, C. J. (2003). Instructed SLA: Constraints, Compensation, and Enhancement. In
C. J. Doughty and M. H. Long (Eds.), The Handbook of Second Language
Acquisition. Malden, MA: Blackwell. pp. 256-310.
Dulay, H., Burt, M. and Krashen, S. (1982). Language Two. New York: Oxford
University Press.
Dulay, H.C. and Burt, M.K. (1972). Goofing: an Indicator of Children’s Second
Language Learning Strategies. Language Learning 22, 2. 235-252.
Eckman, F. (1977). Markedness and the contrastive analysis hypothesis. Language
Learning. 27 (2): 315-330.
Ekiert, M. (2004). Acquisition of the English Article System by Speakers of Polish in
ESL and EFL Settings Teachers College, Columbia University Working Papers in
TESOL and Applied Linguistics 4 (1): 1-23.
Ellis N. C. (2006). Language Acquisition as Rational Contingency Learning. Applied
Linguistics (27) 1: 1–24. Oxford University Press.

279
Ellis, N. C. (2002). Frequency effects in language processing: A Review with
Implications for Theories of Implicit and Explicit Language Acquisition. Studies in
Second Language Acquisition (24): 143–188.
Ellis, R. (1985). Understanding Second Language Acquisition. Oxford: OUP.
Ellis, R. (1991). Grammaticality judgments and learner variability. In H. Burmeister and
P. Rounds (Eds.) Variability in second language acquisition. Proceedings of the tenth
meeting of the Second Language Acquisition Forum. Eugene: University of Oregon.
pp. 25–60.
Ellis, R. (1994). The Study of Second Language Acquisition. Oxford. OUP.
Ellis, R. (2001). Investigating Form-Focused Instruction. Language Learning 51(1): 1–
46.
Ellis, R. (2004). The Definition and Measurement of L2 Explicit Knowledge. Language
Learning (54) 2: 227–275.
Emslie, A. and R. Stevenson (1981). Pre-school children's use of the articles in definite
and indefinite referring expressions. Journal of Child Language 8: 313-28.
Enç, M. (1991). The Semantics of Specificity. Linguistic Inquiry 22: 1-25.
Epstein, R. (2002). The definite article, accessibility, and the construction of discourse
referents. Cognitive Linguistics (12), 4: 333-378.
Epstein, S., S. Flynn, and G. Martohardjono (1996). Second Language Acquisition:
Theoretical and Experimental Issues in Contemporary Research. Behavioral and
Brain Sciences 19: 677–758.
Fassi Fehri, A. (1999). Arabic modifying adjectives and DP structures. Studia Linguistica
53(2): 105-154.
Fassi Fehri, A. (2004). Nominal classes, reference, and functional parameters, with
particular reference to Arabic. Linguistic Variation Yearbook 4: 41 –108.
Fauconnier, G. (1994). Mental Spaces: Aspects of Meaning Construction in Natural
Language. Cambridge University Press.
Festschrift for Tracy David Terrell. New York: McGraw-Hill.
Fodor, J. and I. Sag (1982). Referential and quantificational indefinites. Linguistics and
Philosophy, 5: 355–398.
Foster, P. and P. Skehan (1996). The influence of planning and task type on second
language performance. Studies in Second Language Acquisition. 18: 299-323.
Franceschina, F. (2001). Morphological or syntactic deficits in near-native speakers? An
assessment of some current proposals. Second Language Research 17, 213-47.
Frege, G. (1892). On sense and reference. Translation by Geach and Black, reprinted in
A. W. Moore (Ed.), Meaning and Reference. Oxford: Oxford University Press, (1993).
pp. 23-42.
Garcia Mayo, M. P. (2008). The acquisition of four nongeneric uses of the article the by
Spanish EFL learners. System 36: 550–565.
Gass, S. (1994). The reliability of second-language grammaticality judgments. In E.
Tarone, S. Gass, and A. Cohen (Eds.) Research Methodology in Second Language
Acquisition. Hillsdale, NJ: Erlbaum. pp. 303–322.
Gass, S. (1996). Second language acquisition and linguistic theory: the role of language
transfer. In Handbook of Second Language Acquisition, W. Ritchie and T. Bhatia
(eds.), 317–345. San Diego: Academic Press.
Gass, S. and Selinker, R. (2001). Second Language Acquisition. An Introductory Course
(2nd Edition). Hillsdale, NJ. Lawrence Erlbaum. pp:112-135.
George, H. V. (1971). English for Asian learners: Are we on the right road? English
Language Teaching 25: 270-277.

280
Givón, T. (1983). Topic Continuity in Spoken English. In Givón, T. (ed) Typological
Studies in Language. John Benjamins. Amsterdam. pp. 343-63.
Givón, T. (1984). Syntax: a functional-typological introduction Vol. I. Amsterdam:
Benjamins.
Givón, T. (1989). Mind, Code and Context. Lawrance Erlbaum Associates.
Givón, T. (1993). English grammar: A functional introduction. Amsterdam: Benjamins.
Goad, H and L. White (2004). Ultimate attainment of L2 inflection effects of L1 prosodic
structure. EUROSLA Yearbook 119-145.
Goldberg, A. (2009) The nature of generalization in language. Cognitive Linguistics (20)
1: 93–127.
Goldschneider, J.M. and R. DeKeyser (2001). Explaining the natural order of L2
morpheme acquisition in English: A meta-analysis of multiple determinants.
Language Learning 51: 1–50.
Goldsmith, J. and E. Woisetschlaeger (1982). The logic of the English progressive.
Linguistic Inquiry 13: 79–89.
Granfeldt, J. (2000) The acquisition of the determiner phrase in bilingual and second
language French. Bilingualism: Language and Cognition 3: 263-280.
Grice, H. P. (1975) Logic and conversation. In Peter Cole and Jerry Morgan (eds.) Syntax
and Semantics 3. Academic Press, New York, 41-58.
Gundel, J. K., N. Hedberg and R. Zacharski (1993) Cognitive Status and the Form of
Referring Expressions in Discourse. Language (69), 2: 274-307.
Habash, Z. (1982). Common Errors in the Use of English Prepositions in the Written
Work of UNRWA Students at the End of the Preparatory Cycle in the Jerusalem Area.
Retrieved from http://www.zeinab-habash.ws/education/books/master.pdf
Habuto, J. (2000) Comprehensible Output Hypothesis. Study of Japanese ESL students
and the Acquisition of the English Article System. Classroom Second Language
Acquisition.
Hakuta, K. (1976). A case study of a Japanese child learning English as a second
language. Language Learning (26): 321-351.
Halliday, M. A. (1967). Notes on transitivity and theme in English: Part 2. Journal of
Linguistics, 3: 199–244.
Halliday, M.A.K. and R. Hasan (1976). Cohesion in English. England: Longman Group.
Hamdallah, R. (1988). Syntactic Errors in Written English: Study of errors made by Arab
students of English. Unpublished PhD Thesis. Lancaster University. UK.
Haviland, S. and Clark, H. (1974). What’s new? Acquiring new information as a
process in comprehension. Journal of Verbal Learning and Verbal Behaviour 13:
512–21.
Hawas, H.M. (1989). The articles in English and Arabic: A contrastive study. Indian
Journal of Applied Linguistics 15 (2): 23-52.
Hawkins, J. A. (1978). Definiteness and Indefiniteness. London: Croom Helm.
Hawkins, J. A. (1991). On (in)definite articles: implicatures and (un)grammaticality
predictions. Journal of Linguistics (27): 405-442.
Hawkins, R. (2001). The theoretical significance of universal grammar in second
language acquisition. Second Language Research 17 (4): 345-367.
Hawkins, R. (2004). Explaining full and partial success in the acquisition of second
language grammatical properties. Paper presented at J-SLA, Gunma Prefectural
Women’s University, Gunma, Japan.
Hawkins, R. and Y. Chan (1997). The partial availability of Universal Grammar in
second language acquisition: the Failed Functional Features Hypothesis. Second
Language Research 13 (3): 187–226.

281
Hawkins, R., S. Al-Eid, I. Almahboob, P. Athanasopoulos, R. Chaengchenkit, J. Hu, M.
Rezai, C. Jaensch, Y. Jeon, A. Jiang, I. Leung, K. Matsunaga, M. Ortega, G. Sarko, N.
Snape, K. Velasco-Zarate (2005). Non-target-like article use in L2 English;
implications for current UG-based theories of SLA. EUROSLA.
Hawkins, R., Al-Eid, S., Almahboob, I., Athanasopoulos, P., Chaengchenkit, R., Hu, J.,
Rezai, M., Jaensch, C., Jeon, Y., Jiang, A., Leung, I., Mtsunaga, K., Ortega, M.,
Sarko, G., Snape, N. and Velasco-Zarate, K. et al (2006). Accounting for English
article interpretation by L2 speakers. In S. Foster-Cohen, M. Medved Krajnovic and J.
Mihaljevic Djigunovic (eds.), EUROSLA Yearbook (6): 7-25. Amsterdam: John
Benjamins.
Haznedar, B. (2001). The acquisition of the IP system in child L2 English. Studies In
Second Language Acquisition, (23): 1–39. The United States of America.
Haznedar, B., and Schwartz, B.D. (1997). Are there optional infinitives in child L2
acquisition? In E. Hughes, M. Hughes, and A. Greenhill (Eds.), Proceedings of the
21st Annual Boston University Conference on Language Development 21: 257–268.
Heim, I. (1982). The Semantics of Definite and Indefinite Noun Phrases, Ph.D.
dissertation, University of Massachusetts, Amherst, published (1989); New York.
Garland.
Heim, I. (1983). File change semantics and the familiarity theory of definiteness. In
Meaning, Use and the Interpretation of Language, 164-190. Berlin: Walter de
Gruyter.
Heim, I. (1991). Articles and definiteness. In A. Stechow and D. Wunderlich, (Eds.),
Handbook of Semantics. Aninternational handbook of contemporary research. Berlin:
W. de Gruyter. pp. 487-535.
Hiki (1991). A Study of Learners' Judgement of Noun Countability. Ph.D. dissertation,
Indiana University. In Snape, (2008). Bilingualism: Language and Cognition,
Cambridge University Press.
Hollmann, W. and A. Siewierska (2011). The status of frequency, schemas, and identity
in Cognitive Sociolinguistics: A case study on definite article reduction. Cognitive
Linguistics 22:25-54.
Hong, R. and J. Seonkyung (2001). The L2 Acquisition of Articles in English: The
Effects of the Age of Arrival and the Length of Residency in the United States. Paper
delivered at the University of Southern California USC. Retrieved in (2007). from:
http://www.hichumanities.org/AHproceedings/R%20G.%20P.%20Hong%20and%20S
%20Jeon.pdf
Horn, L. (2007). Toward a Fregean pragmatics in Kecskes and L. Horn (eds.),
Explorations in Pragmatics: Linguistic, Cognitive and Intercultural Aspects. Mouton.
Hu, G. (2002). Psychological constraints on the utility of metalinguistic knowledge in
second language production. Studies in Second Language Acquisition 24:347–86.
Hua, D. and T. Hun-Tak Lee (2005). Chinese ESL Learners’ Understanding of the
English Count-Mass Distinction. In Laurent Dekydtspotter et al. (eds.) Proceedings of
the 7th Generative Approaches to Second Language Acquisition Conference
(GASLA), Somerville, MA: Cascadilla Proceedings Project. 138-149.
Huebner, T. (1983). A Longitudinal Analysis of the Acquisition of English. Ann Arbor
Huebner, T. (1985). System and variability in Interlanguage Syntax. Language Learning
35:141-63.
Hulstijn, J. H. and W. Hulstijn (1984). Grammatical errors as a function of processing
constraints and explicit knowledge. Language Learning 34 (1):23-43.
Hymes, D. (1972). On communicative competence. In J. B. Pride and J. Holmes (Eds.),
Sociolinguisitcs, London: Penguin. pp. 269-293.

282
Ingram, D. E. (1985). Australian second language proficiency ratings (ASLPR). In John
H.A.L. De Jong (Ed.) Standardization in Language Testing (1990). Aila Review (7).
Ionin, T., and Wexler, K. (2003). The certain uses of 'the' in L2 English. In J. M.
Liceras, H. Zobl and H. Goodluck. Somerville, (Eds.) Proceedings of the 6th
Generative Approaches to Second Language Acquisition (GASLA 6) Conference: L2
Links, Cascadilla Press, pp. 150-160.
Ionin, T., H. Ko and K. Wexler (2004). Article semantics in L2-acquisition: the role of
specificity. Language Acquisition 12:3–69.
Ionin, T., Ko, H., and Wexler, K. (2003). Specificity as a grammatical notion: Evidence
from L2 English article use. In G. Garding and M. Tsujimura (Eds.) WCCFL 22
Proceedings. Somerville, MA: Cascadilla Press. 245–258.
Ionin, T., M. Zubizarreta and S. B. Maldonado (2008). Sources of linguistic knowledge
in the second language acquisition of English articles. Lingua 118:554–576.
Jarvis, S. (2002). Topic continuity in L2 English article use. Studies in Second Language
Acquisition, 24:387-418.
Jaszczolt, K. (1997). The default de re principle for the interpretation of belief utterances.
Journal of Pragmatics 28:315-336.
Johnson, J. and E. Newport (1991). Critical period effects on universal properties of
language: the status of subjacency in the acquisition of a second language. Cognition
39: 215–58.
Joosten, F (2003). Accounts of the Count-Mass Distinction: A Critical Survey.
Proceedings of the 19th Scandinavian Conference of Linguistics 31 (1).
Kaku, K. (2006). Second Language Learners’ Use of English Articles: A Case Study of
Native Speakers of Japanese. CLO/OPL 34:63-74.
Kaluza, H. (1963). Teaching the English articles to speakers of Slavic. Language
Learning, 13:113-124.
Kellerman, E. (1977). Towards a characterization of the strategies of transfer in second
language learning. Interlanguage Studies Bulletin 2:58-145.
Kempson, R. (1975). Presupposition and the Delimitation of Semantics. Cambridge:
Cambridge University Press.
Kharma, N. (1981). Analysis of the errors committed by Arab university students in the
use of the English definite/indefinite articles. International Review of Applied
Linguistics 19:331-345.
Kharma, N. and A. Hajjaj (1989). Errors in English among Arabic Speakers: Analysis
and Remedy. London: Longman Group UK Limited.
Krashen, S. D. (1982). Second language Acquisition and Second language learning.
Prentice Hall International.
Kremers, J. M (2003). The Arabic noun phrase. LOT. The Netherlands.
Krifka, M. (1995). Common nouns: a contrastive analysis of Chinese and English. In G.
N. Carlson and F. J. Pelletier (Eds.), The Generic Book. Chicago: University of
Chicago Press, pp.398–411.
Krifka, M. (2003). Kinds of Kind reference: Bare Plurals –Ambiguous or not? Semantics
and Linguistic Theory (13). University of Washington at Seattle.
Kuribara, C. (1999). Resettability of a syntactic parameter: acquisition of category D in
English by adult Japanese speakers. In Robinson, P., (ed.), Representation and
Process: Proceedings of the 3rd Pacific second language research forum (1): 13-22.
Tokyo.
Kuribara, C. (1999). Resettability of a syntactic parameter: acquisition of category D in
English by adult Japanese speakers. In Robinson, P., (Ed.), Representation and

283
process: Proceedings of the Third Pacific second language research forum. (1): 13-
22. Tokyo.
Kuribara, C. (2003). Subjects and Verbal Inflections in SLA: In defence of full transfer /
limited access model. Journal of Foreign Language Education and Research (5): 17-
40. Kansai University. Osaka.
Lado, R. (1957). The Necessity for a Systematic Comparison of Languages and Cultures.
Linguistics across Cultures. Applied Linguistics for Language Teachers, Ann Arbor.
University of Michigan Press, pp. 1-8.
Lambrecht, K. (1994). Information Structure and Sentence Form: topic, focus, and the
mental representations of discourse referents. Cambridge University Press Cambridge.
Langacker, R.W (2008). Cognitive Grammar. A basic Introduction. OUP
Langacker, R.W. (1991). Foundations of Cognitive Grammar: Descriptive Application.
Vol.2. California: Stanford University Press.
Lardiere, D. (1998). Case and tense in the “fossilized” steady state. Second Language
Research, 14: 1–26.
Lardiere, D. (2000). Mapping features and forms in second language acquisition. In
Archibald, J. (ed.) Second Language Acquisition and Linguistic Theory, Malden, MA:
Blackwell, pp. 102-129.
Lardiere, D. (2007). Acquiring (or Assembling) Functional Categories in Second
Language Acquisition. In Belikova et al. (Eds.) Proceedings of the 2nd Conference
(GALANA). Somerville, pp. 233-244.
Leech, G. (1992). Introducing English Grammar. London, Penguin.
Leech, G. (2006a). A glossary of English grammar. Edinburgh University Press.
Leech, G. (2006b). English as a Lingua Franca. LAEL Conference Proceedings,
Lancaster University.
Leung, Y. K. (2001). The Initial State of L3A: Full Transfer and Failed Features. In X.
Bonch-Bruevich, W. J. Crawford, J. Hellermann, C. Higgins, and H. Nguyen, (Eds.)
The Past, Present, and Future of Second Language Research: Selected Proceedings of
the 2000 Second Language Research Forum, Cascadilla Press, Massachusetts.
Levelt, W.J.M. (1989). Speaking: from intention to articulation. Cambridge, MA: MIT
Levinson, S. C. (1987). Pragmatics and the grammar of anaphora: a practical pragmatic
reduction of binding and control phenomena. Journal of Linguistics 23: 379–434.
Lightbown, P.M. and Spada, N. (1999). How Languages are Learned (Revised Edition)
Lightfoot, A. R. (1998). Japanese Second-Language Learners and the English Article
System: A study in Error Analysis. University of Leeds. UK. Retrieved from:
http://ardle.net/linguistics.html.
Lipiński, E. (2001). Semitic languages: outline of a comparative grammar. Peters.
Liu, D., and Gleason, J. I. (2002). Acquisition of the article the by nonnative speakers of
English: An analysis of four nongeneric uses. Studies in Second Language
Acquisition, 24: 1-26.
Löbner, S. (1985). Definites. Journal of Semantics 4: 279–326.
Long (Eds.) The Handbook of Second Language Acquisition. Malden, MA: Blackwell.
pp. 313-348.
Long, M. (1991). Focus on form: A design feature in language teaching methodology. In
K. De Bot, D. Coste, R. Ginsberg and C. Kramsch (Eds.) Foreign language research
in cross-cultural perspective. Amsterdam. John Benjamins, pp.39-52.
Lyons, C. (1999). Definiteness. Cambridge Textbooks in Linguistics. Cambridge: CUP.
Lyons, J. (1977). Semantics. Cambridge University Press.

284
Maalej, Z. (2004). On The Misuse of Determination in Arab Students’ Writing.
Unpublished paper submitted to the University of Manouba-Tunis, TUNISIA.
Retrieved from: www.executivetranslators.com.
Mahmoud, A. (2002). Interlingual Transfer of Idioms by Arab Learners of English. The
Internet TESL Journal 8 (12). Sultan Qaboos University. Oman.
Mandell, O. (1999). On the reliability of grammaticality judgment tests in second
language acquisition research. Second Language Research, 15:73–99.
Maratsos, M. (1976). The use of definite and indefinite reference in young children.
Cambridge: University Press.
Master, P (1987). A Cross-linguistic Interlanguage Analysis of the Acquisition of
Articles. Unpublished PhD Thesis. University of California. Los Angeles.
Master, P (1987). Generic the in Scientific American. English for Specific Purposes 6
(3):165-186.
Master, P. (1990). Teaching the English articles as a binary system. TESOL Quarterly
24:461–478.
Master, Peter (1997). The English article system: Acquisition, function, and pedagogy.
System, 25: 215-232.
Matthewson, L. and J. Schaeffer. (2000). Grammar and pragmatics in the acquisition of
article systems. in J. Gilkerson, M. Becker, and N. Hyams (eds.) UCLA Working
Papers in Linguistics: Language Development and Breakdown, University of
California: Los Angeles, pp: 1-39.
Matthewson, L., Bryant, T. and Roeper, T. (2001). A Salish stage in the acquisition of
English determiners: Unfamiliar ‘definites’. Proceedings of SULA: The Semantics of
Under-represented Languages in the Americas. University of Massachusetts
Occasional Papers in Linguistics (25) Amherst, pp. 63–71.
McGregor, R. S. (1995) Outline of Hindi grammar with exercises. Oxford: Oxford
University Press.
Mclaughlin, B. (1987). Theories of Second-language Learning. Arnold
Mehnert, U. (1998). The Effects of different lengths of time for planning on second
language performance. Sudies in Second Language Acquisition, 20:52-83.
Milanovic (1988). The construction and validation of a performance-based battery of
English language progress tests. Unpublished doctoral dissertation, London
University, Institute of Education.
Mitchell, T.F. and S.A. El-Hassan (1994). Modality Mood and Aspect in Spoken Arabic
Kegan Paul International, London, New York.
Mizuno, H. (1985). A psycholinguistic approach to the Article System in English. JACET
Bulletin 16:1-29.
Mizuno, M. (1999). Interlanguage analysis of the English article system: Some cognitive
constraints facing the Japanese adult learners. International Review of Applied
Linguistics 37: 127-152.
Muranoi, H. (2000). Focus on Form Through Interaction Enhancement: Integrating
Formal Instruction Into a Communicative Task in EFL Classrooms. Language
Learning 50 (4): 617-673.
Nagata, R. Tatsuya Iguchi, Kenta Wakidera, Fumito Masui, and Atsuo Kawai (2005).
Recognizing Article Errors in the Writing of Japanese Learners of English. Systems
and Computers in Japan 36 (7):54-62.
Nemser, W. (1971). Approximate systems of foreign language learners. IRAL 9: 115-23.
Nettle, Mark and Dianna Hopkins (2003). Developing Grammar in Context. CUP
Norris, J. M. and L. Ortega. (2000). Effectiveness of L2 Instruction: A Research
Synthesis and Quantitative Meta-analysis. Language Learning 50 (3): 417-528.

285
Odlin, T. (1989). Language Transfer. Cambridge: Cambridge University Press.
Odlin, T. (2003). Cross-Linguistic Influence, in Catherine J. Doughty, Michael H. Long
(Eds.) The Handbook of Second Language Acquisition, pp. 436-486.
Oller, J. W. and E. Z. Redding (1971). Article usage and other language skills. Language
Learning 21 (1): 85-95.
Oller, J., and C. Conrad. (1971). The cloze technique and ESL proficiency. Language
Learning 21:183-195.
Onishi K. H. and G. L. Murphy (2002). Discourse model representation of referential and
attributive descriptions. Language and Cognitive Processes 17 (2): 97–123.
Ortega, L. (1999). Planning and focus on form in L2 oral performance. Studies in Second
Language Acquisition. 21: 108-48.
Oxford: Oxford University Press.
Paradis, M. (1994). Neurolinguistic aspects of implicit and explicit memory: Implications
for bilingualism and SLA. In N. Ellis (Ed.) Implicit and explicit learning of
languages. San Diego, CA: Academic Press. pp. 393–419.
Parodi, T., B. D. Schwartz, and H. Clahsen (1997). On the L2 Acquisition of the
Morphosyntax of German Nominals. Essex Research Reports in Linguistics 15: 1–43.
Parrish, B. (1987). A new look at methodologies in the study of article acquisition for
learners of ESL. Language Learning 37: 361–383.
Partee, B. H. (2005). Semantic Typology of Indefinites. Course handout of the Research
Seminar on Topics in Formal Semantics (MGU). Retrieved from
http://people.umass.edu/partee/MGU_2005/MGU055.pdf.
Pica, T. (1983). Adult Acquisition of English as a second language under different
conditions of exposure. Language Learning 33: 465-97.
Pienemann M. and Håkansson, G. (1999). A unified approach towards a theory of the
development of Swedish as L2.A processability account. Studies in Second Language
Acquisition. 21: 383–420.
Poesio, M. and R. Vieira (1998). A corpus-based investigation of definite description
use. Computational Linguistics 24: 183–216.
Pongpairoj, N (2007). Are L2 English Article Choices UG-regulated? Cambridge
University Conference Proceedings in Language Research CAMLING. pp: 213-220.
Poulisse, N. (1997). Some Words in Defense of the Psycholinguistic Approach: A
Response to Firth and Wagner. The Modern Language Journal 81 (3): 324-328.
Power, T. (2003). Communicative Language Teaching: The appeal and poverty of
Communicative Language Teaching. Retrieved from
http://www.btinternet.com/~ted.power/esl0404.html
Press.
Prévost, P. and White, L. (2000). Missing surface inflection or impairment in second
language acquisition? Evidence from tense and agreement. Second Language
Research 16: 103-133.
Prince, E (1992). Subjects, Definiteness and Information Status. In Mann and Thompson
(eds.) Discourse Description. John Benjamins.
Quick Placement Test (2001). Oxford University Press, Oxford.
Quirk R. S., Greenbaum, G. Leech, and G. Svartik (1985). A Comprehensive Grammar
of the English Language. London and New York: Longman.
Quirk R. S., Greenbaum, G. Leech, and G. Svartik (1991). A Comprehensive Grammar
of the English Language. Longman (2nd Edition).
Raymond, W., J. A. Fisher, and A. F. Healy (2002). Linguistic knowledge and language
performance in English article variant preference. Language and Cognitive Processes.
17 (6): 613–662.

286
Richards, J. (1971). A non-contrastive approach to error analysis. English Language
Teaching Journal 25: 204-19.
Richards, J.C. (1976). The Role of Vocabulary Teaching. TESOL Quarterly.10 (1): 77-
89.
Richards, J.C. and G.P. Sampson (1974). The study of learner English. In J. Richards
(Ed.) Error Analysis. Perspectives on second language acquisition. Essex: Longman
pp. 3-18.
Ringbom, H. (1987). The Role of The First Language In Foreign Language Learning.
Clevedon, UK: Multilingual Matters.
Roberts, G. (2003). Uniqueness in Definite Noun Phrases. Linguistics and Philosophy
26: 278-350.
Robertson, D. (2000). Variability in the use of the English article system by Chinese
learners of English. Second Language Research (16) 2: 135-172.
Robinson, P. (1996). Learning simple and complex second language rules under implicit,
incidental, rule-search, and instructed conditions. Studies in Second Language
Acquisition, 18: 27–67.
Roehr, K. (2004). Exploring the role of explicit knowledge in adult second language
learning: language proficiency, pedagogical grammar and language learning
strategies. Centre for Research in Language Education (CRILE) Working Papers (59).
Lancaster University, UK.
Roehr, K. (2005). Metalinguistic Knowledge in Second Language Learning: An
Emergentist Perspective. Unpublished PhD Thesis. Lancaster University, UK.
Roeper, T. (2003). Watching Nounphrases Emerge: Seeking compositionality. Kluwer
Academic Publishers. The Netherlands.
Russell, B. (1905). On Denoting. Mind 14: 479-593.
Ryle, G. (1949). The Concept of Mind. London. Huchinson.
Salmon-Alt, S. and S. Romary (2000). Reference Resolution within the Framework of
Cognitive Grammar. Cognitive Science Quarterly (1).
Sang, B., Schmitz, H.J., Vollmer, J. and Roeder P.M. (1986). Models of second language
competence: a structural equation approach. Language Testing 3: 54–79.
Sarko, G. (2008). Morphophonological or Syntactic Transfer in the Acquisition of
English Articles by L1 speakers of Syrian Arabic? Proceedings of the 9th Generative
Approaches to Second Language Acquisition. In Roumyana Slabakova et al. (Eds.),
GASLA Conference Proceedings. Cascadilla Project. Somerville, MA. pp. 206-217
Schachter, J. (1983). A new account of language Transfer. In S. Gass and L. Selinker
(Eds.) Language Transfer in Language Learning. Rowley, MA: Newbury House
Schachter, J. (1990). On the issue of completeness in second language acquisition.
Second Language Research. 6:93–124.
Schaeffer, J. and Matthewson, L. (2005).Grammar and pragmatics in the acquisition of
article systems. Natural Language and Linguistic Theory, 23, 53–101.
Schütze, C. T. (1996). The Empirical Base of Linguistics: Grammaticality Judgments and
Linguistic Methodology. Chicago: University of Chicago Press.
Schulz, E. (2004). A Student Grammar of Modern Standard Arabic. Cambridge
University Press.
Schwartz, B.D. and R. Sprouse. 1996 L2 cognitive states and the full transfer/full access
model. Second Language Research 12: 40-72.
Searle, J. R. (1969). Speech Acts. Cambridge: Cambridge University Press.
Seidl, J. and W. McMordie (1992). Oxford Pocket English Idioms. Oxford University
Press.

287
Selinker, L. (1972). Interlanguage. International Review of Applied Linguistics 10: 209-
231.
Selkirk, E. O. (1996). The prosodic structure of function words. In J.L. Morgan and K.
Demuth (eds.), Signal to Syntax. Hillsdale, NJ: Lawrence Erlbaum. pp: 187-213.
Sharma D. (2005). Transfer and Universals in Indian English Article Use. Studies in
Second Language Acquisition. 27: 535–566.
Sharma, V. A. (1981). Syntactic Errors as Indices of Developing Language Proficiency
in Arabic Speakers Writing English at the Intermediate and Advanced Levels of
English as a Second Language. Ph.D. Thesis. Indiana University.
Shlonsky, U. (2004). The form of Semitic noun phrases. Lingua 114: 1465–1526.
Siegel, Jeff (2003). Substrate influence in creoles and the role of transfer in second
language acquisition. Studies in Second Language Acquisition, 25 (2):185-209.
Sinclair, J.M. (1991). Corpus, Concordance, Collocation. Oxford University Press, UK.
Skehan, P. (1989). Language Testing. Language Teaching. 22:1-13 Cambridge
University Press.
Skehan, P. (2003). Task-based Instruction. Language Teaching. 36:1–14.
Skehan, P. and Foster, P. (1999). The influence of task structure and processing
conditions on narrative retellings. Language Learning, 49 (1): 93−120.
Skehan, Peter (1998). A cognitive approach to language learning. Oxford, England:
Oxford University Press.
Slabakova, R. (2000). L1 transfer revisited: the L2 acquisition of telicity marking in
English by Spanish and Bulgarian native speakers. Linguistics 38 (4): 739–770.
Smith, B. (2001). Learner English. A Teacher’s guide to Interference and other
problems. OUP.
Smith, N. and Tsimpli, I.-M. (1995). The mind of a savant. Language Learning and
Modularity. Oxford: Blackwell.
Snape, N. (2005). The uses of Articles in L2 English by Japanese and Spanish Speakers.
Paper submitted to The Annual Conference on Language Acquisition, University of
Cambridge. Essex Graduate Student Papers in Language and Linguistics, 7: 1–23.
Snape, N. (2008). Resetting the Nominal Mapping Parameter in L2 English: Definite
article use and the count–mass Distinction. Bilingualism: Language and Cognition 11
(1). Cambridge University Press.
Snape, N., Y.-k. Ingrid Leung, and H-C. Ting (2006). Comparing Chinese, Japanese and
Spanish Speakers in L2 English Article Acquisition: Evidence against the Fluctuation
Hypothesis? In Mary Grantham O’Brien, Christine Shea, and John Archibald, (Eds.)
Proceedings of the 8th Generative Approaches to Second Language Acquisition
Conference (GASLA). pp: 132-139.
Sorace, A. (1985). Metalinguistic knowledge and language use in acquisition-poor
environments. Applied Linguistics 6: 239–254.
Sorace, A. (1996). The use of acceptability judgments in second language acquisition
research. In W. Ritchie and T. Bhatia (Eds.) Handbook of Second Language
Acquisition. San Diego, CA: Academic Press. pp. 375–409.
Sperber, D. and Wilson, D. (1986). Relevance. Blackwell.
Strawson, P. F. (1950). On Referring. Mind 59: 320–344.
Swan, M. (2005). Legislation by Hypothesis: The Case of Task-Based Instruction.
Applied Linguistics. 26 (3): 376-401.
Swan, M. and C. Walter (2001). The Good Grammar Book. Oxford University Press.
Tarone, E. (1983). Some thoughts on the notion of "communication strategy." In C.
Faerch and G. Kasper (Eds.) Strategies in Interlanguage Communication. London:
Longman, pp. 61-74.

288
Tarone, E. (1985). Variability in interlanguage use: a study of style-shifting in
morphology and syntax. Language Learning. 35: 373-404.
Tarone, E. and Parrish, B. (1988). Task-related variation in interlanguage: the case of
articles. Language Learning 38: 21-44.
Thomas, M (1989). The Acquisition of articles by native and non-native speakers of first
and second language learners. Applied Psycholinguistics 10, 335-355.
Thu, H. N. (2005). Vietnamese Learners Mastering English Articles. Doctoral thesis.
University of Groningen. Retrieved from:
http://dissertations.ub.rug.nl/FILES/faculties/ppsw/2005/h.n.thu/thesis.pdf
Towell, R. and Hawkins, R. (1994): Approaches to second language acquisition.
Clevedon: Multilingual Matters, pp: 317–61.
Trenkic´, D. (2000). The acquisition of English articles by Serbian speakers.
Unpublished PhD thesis. Cambridge University.
Trenkic´, D. (2002). Form-meaning connections in the acquisition of English articles. In
Foster-Cohen, S., Ruthenberg, T. and Poschen, M. (eds.) European Second Language
Association Yearbook 2. John Benjamins, pp:115-33.
Trenkic´, D. (2007). Variability in L2 article production - beyond the representational
deficit vs. processing constraints debate. Second Language Research, 23 (3): 289-327.
Trenkic´, D. (2008). The representation of English articles in second language grammars:
determiners or adjectives? Bilingualism: Language and cognition, Cambridge
University Press, 11: 1-18.
Tsimpli, I. M. (2003). Clitics and determiners in L2 Greek. In Juana M. Liceras et al.
(eds.), Proceedingsof the 6th Generative Approaches to Second Language Acquisition
Conference (GASLA 2002). Somerville, MA: Cascadilla Proceedings Project, pp 331-
339.
Vainikka, A., and Young-Scholten, M. (1996). Gradual development of L2 phrase
structure. Second Language Research 12: 7–39.
VanPatten, B. (1994). Cognitive aspects of input processing in second language
acquisition. In P. Heshemipour, I. Maldonado, and M. van Naerssen (Eds.),
Van Patten, B., Williams, J., and Rott, S. (2004). Form-meaning Connections in Second
Language Acquisition. Mahwah, NJ. Lawrence Erlbaum.
Walter, J. (1975) The Writer’s Audience is Always a Fiction. Modern Language
Association 90 (1): 9-21
Weir, Cyril (1986). Communicative Language Testing. University of Exeter.
White, L. (2003). Fossilization in steady state L2 grammars: Persistent problems with
inflectional morphology. Bilingualism: Language and Cognition 6 (2):129-141.
Yamada, J. and N. Matsuura, (1982). The use of the English article among Japanese
students. RELC Journal 13: 50-63.
Yoon, K.K. (1993). Challenging prototype descriptions: Perception of noun countability
and indefinite vs. Ø article use. International Review of Applied Linguistics 31: 269-
289.
Young, R. (1996). Form-function relations in articles in English interlanguage. In R.
Bayley and D. R. Preston (Eds.) Second language acquisition and linguistic variation.
Amsterdam: John Benjamins, pp. 135-175.
Yule, G (1998). Explaining English Grammar, OUP.
Zdorenko, T. and Paradis, J. (2008). The acquisition of articles in child second language
English: Fluctuation, transfer or both? Second Language Research 24 (2): 227-250.
Zegarac, V. (2004). Relevance theory and the in second language acquisition. Second
language Research, University of Luton, 20 (3): 193-211.

289
Zehler, Annette M. and William F. Brewer. (1982). Sequence and Principles in Article
System Use: An Examination of A, The, and Null Acquisition, Child Development 53:
1268–1274.
Zobl, H. (1980). Developmental and Transfer Errors: Their Common Bases and
(Possibly) Differential Effects on Subsequent Learning. Tesol Quarterly. 14 (4).
Zobl, H. (1986). A functional approach to the attainability of typological targets in L2
acquisition. Second Language Research, 2 (1):16-32.
Zobl, H. and Liceras, J. (1994). Functional Categories and Acquisition Orders. Language
Learning 44 (1):159-180.

290
Appendices

Appendix 1: Information Sheet and Consent Form

INFORMATION SHEET

As part of my studies in the Department of Linguistics in Lancaster University, UK, I


have been asked to carry out a study involving a translation task, free writing and a
multiple choice cloze test. Later, I am going to analyse the test results and will look
for particular features that will help me to find out Arab learners’ difficulties and
stages in acquisition of the English article system.

I have approached you because I am interested in learning about the way Arabic
speakers learn English articles. The session will take about 30 minutes.
I would be very grateful if you would agree to take part.

You are free to withdraw from the study at any time. At every stage, your name will
remain confidential. The data will be kept securely and will be used for academic
purposes only.

If you have any queries about this project, please feel free to contact me on
dina.awad@uaeu.ac.ae

Dina Awad

291
CONSENT FORM

Project title: Acquisition of English articles by Arabic speakers

1. I have read and had explained to me by Ms. Dina Awad the Information Sheet
relating to this project.

2. I have had explained to me the purposes of the project and what will be
required of me, and any questions have been answered to my satisfaction. I
agree to the arrangements for my participation as described in the Information
Sheet.

3. I understand that my participation is entirely voluntary and that I have the right
to withdraw from the project any time.

4. I have received a copy of this Consent Form and the accompanying


Information Sheet.

Name:

Signed:

Date:

292
Appendix 2: Background uniformity survey

Name: ________________________ ID ____________________

Background uniformity survey

Please answer the following questions:

1. What is your date of birth? (yyyy)

2. Is Arabic your first language?

3. What is your nationality?

4. Is Arabic the first language of your parents?

5. Do you speak English at home?

6. What are you studying/ what is your major?

7. Have you stayed in an English speaking country for more than three

months?

8. Have you attended an English medium / international school?

9. Have you ever taken any extra courses in English outside

school/university?

10. How many years have you been studying English?

Thank you very much

Dina Awad

293
Appendix 3: Task 1

Name:________________________ ID number: ____________________

Writing Task

Write an essay describing your hometown in detail. You can include information
about its location, tourist attractions, history, life style and people. Try to write a
minimum of 350 to 500 words maximum in one hour.

294
Appendix 4: Task 2

Fill in the gaps with ‘a’, ‘an’, ‘the’ or leave the blank empty (Ø) if you think no article
is needed.
Story 1

1____ two mountaineers were walking across some hills when they met 2____ old
hunter. 3____ worried hunter warned them to be very cautious since several wolves
and 4____ bear were roaming 5____ hills, causing 6____ serious problems to 7____
peasants of 8____ nearby villages.

Suddenly, 9____ rifle-shots were heard in the distance, and 10____ three men spotted
11____ two figures running along 12____ stream at the foot of 13____ hill. They
were wearing 14____prison uniforms and were carrying 15____ guns and 16____
bag.
Feeling very afraid, 17____ mountaineers and 18____ hunter hid behind 19____ bush.
From there, they could see (20)____ runaway prisoners slowing down, looking
around, taking off (21)____ prison uniforms and putting on 22____ clothes that they
took out from 23____ bag. Then they hid 24____ guns under 25____ big tree.

26____ moment later, 27____ flock of 28____ sheep followed by 29____ dog and
30____ shepherd appeared from around 31____ hill.
32____ prisoners went to 33____ shepherd and asked him something. 34____ dog
started to bark furiously at 35____ strangers, but 36____ owner calmed it down.
Then he left 37____ dog to watch 38____ sheep and led 39____ prisoners over
40____ small bridge into 41____ forest on 42____ other side of 43____ stream.

44____ hunter and 45____ mountaineers emerged from behind 46____ bush and ran
after them. Before they reached 47_____ bridge, they saw 48____ poor shepherd and
49____ prisoners running back quickly, followed by 50____ bear. 51____ hunter
aimed at 52____ angry beast and fired. He fulfilled 53____ promise which he gave to
54____ villagers that he will soon free them from 55____ fear which 56____ bear was

295
spreading. Now he only had to take care of 57____ wolves. At that moment, one of
58____ mountaineers took 59____ blanket out of 60____ backpack which he was
carrying, and threw it over 61____ prisoners when they crossed 62____ bridge. He
jumped on them and tied them down. 63____ escape attempt did not succeed.

Story 2

64____ fact that Ahmed had organised 65____ graduation party while Saeed was
away annoyed Saeed very much. But when Saeed entered 66____ small kitchen that
they shared, he became very angry! Although 67____ fan was working and 68____
window was open, he could still smell 69____ stale smoke of 70____ cigarettes.
71____ table was turned upside down and 72____ chair was broken. 73____ fridge
was left open and now water was dripping out of it.
74____ milk and 75____ flour were spilt over 76____ cooker and 77____ dirty
saucepan was standing on 78____ top of it all. 79____ filthy sink was full of 80____
dirty plates and 81____ glasses and 82____ broken cup was thrown in 83____ bin.
Someone had taken 84____ sugar and 85____ coffee out of 86____ cupboard, and left
87____ slice of 88____ bread with 89____ butter and 90____ honey on it uneaten.
91____ knife was lying on 92____ dirty floor.

Story 3

Stephan had 93____ exciting day yesterday. He had finally become 94____ member
of 95____ Parliament. He was always 96____ dreamer, and since 97____ time when
he was in 98____ navy, he hoped to rise to 99____ position of 100____ top politician.

In the past, while he was sailing 101____ seas from 102____ Equator to 103____
North Pole, he would spend 104____ long night hours on 105____ deck staring into
106____ dark water of 107____ sea or looking at 108____ stars and 109____ moon in
110____ sky, dreaming of days to come.

Yesterday, while he was crossing 111____ bridge over the river Thames on his way to
112____ Parliament, he noticed 113____ strange light which 114____ sun was casting

296
in 115____ water. Then, as he was approaching, he spotted 116____ Prime Minister
getting out of 117____ big black car, waving to 118____ reporters who were waiting
at 119____ entrance.

120____ masses of 121____ people were in 122____ streets hoping to see 123____
Queen who was coming to give 124____ traditional opening speech to 125____
members of Parliament.

Story 4

When Amna asked Mariam several months ago whether she can help her with her
project, Mariam refused and said that she did not have time and did not agree to take
126______ responsibility of doing other students’ homework.
127_____ fact that Amna asked 128____ question annoyed Mariam. Several days later
they had 129____ argument and Mariam informed her that 130____ friendship was
over.
After 131____ incident, Mariam changed her room in the hostel and Amna felt lonely
and tried to absorb 132____ shock. She was still living in 133____ room that they had
been sharing together since the beginning of the semester.
134____ idea that Mariam might now have new friends in her new hostel, made Amna
jealous. She spent 136____ days reading 137____ magazines and browsing138____
websites. When she was told by one of 139____ students in the hostel that joining
140____ club reduces 141____ stress, Amna agreed.
But 142____ club couldn’t erase 143____ memory of 144____ loss.

Then, one day, 145____ friend took her out for 146____ picnic, and Amna told her
147____ news. 148____ friend asked her to make 149____ wish. Amna thought and
came to 150____ conclusion that she wanted 151____ new roommate. She wanted
someone whom she could have 152____ sincere friendship with.

153____ sympathetic friend told her that she must be full of 154____ confidence to get
what she wanted and to expect 155____ something good to happen. That was 156____
recipe for 157____ success in 158____ life.

297
First, she decided to get rid of 159____ jealousy. Looking into herself, she realised
that 160____ source of her jealousy was 161____ deep feeling of 162____ insecurity.
She decided to go to 163____ students’ party. 164____ thought of going to new places
and meeting 165____ new people cheered her up. Thus, she made 166____ first step
towards 167____ happiness. She liked 168____ way she was dealing with 169____
problem now. She was feeling much stronger and was regaining 170____ faith in
herself.

Story 5

When the Gold Rush started, Jim went to 171____ West. He stopped at 172____ little
mountain river where, 173____ people were saying, 174____ gold had recently been
discovered. All Jim could see was 175____ water and 176____ sand at 177____
bottom of 178____ river. But since he had 179____ patience and 180____ courage, he
decided to stay and never regret 181____ choice.

Since 182____ river often flooded 183____ banks, leaving 184____ mud behind, Jim
decided to build himself 185____ little house, way up in 186____ mountain where
187____ tall grass was growing.

Every day he would cross 188____ mud, come to 189____ water and sieve 190____
sand until darkness. One day, something shone bright at 191____ bottom of 192____
sieve! He found 193____ gold!

In 194____ evening, he hid it carefully in 195____ tall grass behind 196____ little
house, but 197____ next morning 198____ unfortunate man found out that 199____
gold had disappeared.

298
Story 6

200____ woman is sitting at 201____ table in 202____ small dirty kitchen and she is
reading 203____ book. It is about 204____ beautiful young woman in 205____ love.
She is standing in front of 206____ mirror in 207____ spacious, nicely decorated
room, and is combing her hair. There is 208____ knock at 209____ door. She opens it
and 210____ handsome young man enters 211____ room.
At that moment, 212____ woman in 213____ kitchen also hears 214____ knock at her
door. She puts down 215____ book and opens 216____ door.
217____ stout old man enters 218____ kitchen and 219____ poor old lady bursts into
tears.

299
Appendix 5: Task 3

Some of these sentences are correct, others are wrong. Put (√) next to sentences if
you think the underlined part is correct, and (x) next to the sentence in which the
underlined part is incorrect.

1. An old man walked into a hospital to see his son. The first thing the man did was
ask for the room number.
2. At a dinner table: “Can you please pass me salt?”
3. Reading the books fills people’s free time.
4. I asked the shopkeeper for bottle of water.
5. You can ask a friend to help you answer this question.
6. At the jewellers: ‘I have a gold that I would like to sell.’
7. My sister bought a dresses for the wedding party.
8. No thank you. I don’t take the sugar in my tea.
9. Damascus is capital of Syria
10. I made a wish and it came true.
11. When I needed the pen in the last exam, Fatma gave me hers.
12. Aisha went to a wedding. She said a bride was beautiful.
13. I found the book that you gave me very interesting.
14. A sign at the cinema says: ‘the children are not allowed to see the film unless
accompanied by parents.’
15. The team has plans to win this match.
16. My friend has degree in Science.
17. All plants need sunlight, oxygen and water to live and grow.
18. Fatima became successful business woman
19. We can’t go up that mountain. My father has the fear of heights.
20. Our small company succeeded in making the relationships with other companies.
21. Our teacher gave us an advice.
22. A wife reminding her husband: ‘Remember to buy bread on your way back.’
23. Emirates is the biggest mall in the UAE
24. There should be a balance between economic development and respect for the
environment
25. In choosing a flight, the most important thing is the safety.
26. Teacher to student: ‘You must find a better book to read.
27. Research has proved that women make better managers than men.
28. I must warn you: There are mice in this flat.
29. The milk is necessary for babies.
30. People should have clean drinking water.
31. New manager in our company is very serious.
32. My brother did a good house work
33. Ali always takes a very nice pictures.
34. In the past, there was the old tree in our garden, but they cut it down.
35. Layla bought detective novels from the bookshop yesterday.
36. I bought a same cheap computer that my friend has.
37. Their hard work finally led to success.

300

You might also like