The third Nordic conference on Middle Eastern Studies:
Ethnic encounter and culture change
Joensuu, Finland, 19-22 June 1995

Acquisition Order in Arabic as a Foreign Language - a Cognitive Approach

Helle Lykke Nielsen
Odense University[1]

NB *This is the unedited paper as given at the Joensuu conference. An updated and edited version has been published in Sabour and Vikør, Ethnic encounter and culture change, Bergen/London 1997, 250-70. Please quote or refer only to the published article.*

1. Introduction

Research in the field of English as a second language (and, to a much lesser extent, German and French) has tried to explain the order of acquisition with regards to surface structures, such as e.g. morpheme order, as well as to markedness theory. [3] Both approaches have been met with severe critism as to their methodological approach as well as to their conclusions. [4] However, in recent years, there has been a growing focus on cognitive operations as the main factor of explaining how and in what order language learners acquire a foreign language: An important theoretical construct in this area is Manfred Pienemann's Teachability Hypothesis in the framework of which he establishes a language processing hierarchy of four levels, none of which can be jumped in the acquisition process. [5]

The fact that studies of acquisition order in second language acquisition have moved from the level of language specific surface structures towards a more cognitive approach, makes it interesting to try to apply these to Arabic as a foreign language, since cognitive structures are supposedly the same, no matter what foreign language we are dealing with, whereas language specific structures are not. Therefore, what I would like to do in this study is to see if Pienemann's language processing hierarchy can explain learner language data from students of Arabic. The question which I intend to answer can be stated as follows: Is there an order of acquisition for Arabic as a foreign language which can be explained according to cognitive operations?

The study will be structured as follows: In the following section, Pienemann's language processing hierarchy, I shall outline, in a rather sketchy manner, the premises and characteristics of the language processing hierarchy. In section 3, A language processing hierarchy for Arabic as a foreign language, I shall apply the hierarchy to Arabic, so as to see what the theory predicts to be the order of acquisition for noun phrase structures. They have been chosen because they are well suited to illustrate some of the cognitive operations implied by Pienemann's theory. In section 3, I shall also compare the predicted order of acquisition to a longitudinal study of a Danish language learner of Arabic, so as to see if the data are consistent with Pienemann's model. In section 4, the findings will be discussed, and some conclusions will be presented in section 5.

2. Pienemann's language processing hierarchy

Pienemann's theoretical outline is most often exemplified by the acquisition of word order rules in German. This does not mean, however, that the language processing hierarchy applies to this phenomenon only: Though the explanatory approach to the theoretical outline originated from the Zisa group of Meisel, Clahsen and Pienemann, [7] which focused on the order of acquisition for German word order rules, the scope was extended by Pienemann and Johnson, [8] so as to encompass a wider range of structures in morphology and syntax, which were then tested for German and English as second languages. In the words of Pienemann, the language processing hierarchy "serves as a general grid for the prediction of acquisitional chronologies for a wide range of structures in morphology and syntax". [9] Also, it should be noted that the data presented by Pienemann have been collected over the years in formal (classroom) as well as natural settings. [10]

Pienemann's language processing hierarchy is based on certain premises, four of which shall be dealt with here, because they are necessary to understand how the hierarchy works. The first one has to do with psychological complexity: According to Pienemann, a foreign language learner's acquisition of linguistic structures depends on how complex such structures are from a psychological point of view. Pienemann defines the term of psychological complexity as being the extent to which the language learner must "re-order and re-arrange linguistic material in the process of mapping underlying semantics onto surface forms". [11]

The notion of psychological complexity has been elaborated further into the following two premises both of which have to do with transfer of information on the cognitive level. The first of these has to do with perceptual saliency versus linguistic processing: Due to the cognitive structures of human beings, the language learner will initially organize his interlanguage around non-language specific features which are perceptually salient, such as for example endpoints of a sentence or a word, and from there, he will then gradually build up his language specific knowledge so as to be able to process an increasing amount of linguistic structures. [12] This premise is related to the above mentionned one in the sense that what is perceptually salient is less psychologically complex to process than structures of a pure linguistic nature.

The third premise deals more specifically with linguistic processing: It is predicted that local transfer of information is easier to process than non-local transfer. [13] This has to do with the way the human brain stores information: During the planning process of a linguistic utterance, the language learner (or any native speaker, for that matter) encodes his intentions into a linguistic form which is realized as linguistic output. The information which is not immediately realized is stored in the working memory so as to be ready for use at the right moment. This storing process takes place on many different levels, often at the same time: When a language learner intents to realize an utterance like, for example, "he speaks", the use of the personal pronoun "he" is to trigger the information that a third person singular pronoun requires an "-s" at the end of the verb. Thus, while the verb is realized, the "-s" morpheme has to be kept in the working memory and realized immediately at the end of the verb. Another example of this storing process takes place during the realization of a sentence like "Before he went away, he had a cup of tea": The natural order of events are so that the person in question has a cup of tea and then goes away. When the utterance is realized, however, the speaker will have to retain, in his working memory, the utterance representing the first natural event, while realizing the second part of it. Thus, the premise here is that the language learner will be able to handle local morphology before non-local morphology, since the latter will have to be retained for a longer time in the language learner's working memory, thus leaving less space to other processing devices such as e.g. recall of vocabulary.

And finally, the fourth premise concerns the relationship between the different stages of the language processing hierarchy. According to Pienemann, there exists an implicational relation between the different levels of the hierarchy in the sense that the processes which are acquired at one level are considered a necessary prerequisite for the processing of structures at the following level. Or as Pienemann puts it: "The devices acquired at one stage are a necessary building block for the following stage". [14] Thus, no level can be jumped in the acquisition process, since learners are unable to process structures more than one step beyond their current level.

Pienemann's language processing hierarchy consists of four levels which the language learner must pass on his way towards the mastering of a foreign language. At the first level, the language learner is not able to organize the lexical material into categories - that is, into word classes - and therefore, he has no acces to the phrase structure. Thus he cannot "identify elements within the sentence from which information has to be taken or to which information has to be brought". [15] In other words: Since the learner cannot index lexical items in the foreign language, no transfer of information can take place. [16] Learner language utterances at this level typically consists of a string of morphologically independant words. Only word order plays a role, in the sense that the learner uses the words in the order which he expects them to be sequenced in the foreign language. Pienemann terms this phenomenon "the canonical word order". [17]

At the second level, the language learner is able to transfer information on the basis of features which are not language specific, that is, on the basis of perceptual saliency such as endpoints (initial or final) of sentences. However, since the language learner is still not able to organize the lexical material into categories, he cannot yet insert morphemes in the phrase structure. [18] As in level one, language learner utterances at this level typically consists of a string of morphologically independant words, but the word order can now be manipulated in such a way as to move or add words at the sentence initial or final position.

At the third level, the language learner is able to organize some of his lexical material into categories, that is, he can index some, if not all, lexical items according to word classes, which makes it possible for him to identify sentence internal constituents. [19] However, given the language processing devices available to the language learner at this point, he can only transfer information into "the computionally easier salient positions". [20] This means that the target point of the transfer will have to be the initial or the final position of a constituent or a word. In case a constituent consists of two or more words, it is predicted that the transfer of information into the endpoint position of a constituent is easier to perform than one whose insertion point is in an intraconstituent position: In the case of the former, the language learner has to identify the "insertion point by a recognition of the formal syntactic class of the element undergoing the morphological process". [21] In the latter, all the elements of the constituent must be recognized so as to enable the learner to locate the correct insertion point of the transfer. [22] At this level, we will find transfer of local morphology, e.g. inside a constituent, [23] whereas transfer of information into non-local environment, such as inter-constituent transfer (i.e. subject-verb agreement[24]) cannot take place. [25]

At the fourth level, the language learner will experience no more constraints on his transfer of information. He can now recognize all the elements in the sentence, and this gives him acces to the whole phrase structure in such a way that he can identify all possible elements within the sentence from which information has to be taken or to which information has to be brought. [26] Thus, at this level, we will find transfer of information in sentence internal positions of non-local nature, e.g. between constituents (i.e. subject-verb agreement) as well as between sentences, such as concord of tenses, subordinate clauses etc.

The four levels of the language processing hierarchy can be summerized as follows:

2.1. Predicted stages of acquisition[27]

3. A language processing hierarchy for Arabic as a foreign language

Now, what does the language processing hierarchy predicts to be the acquisition order of Arabic? In this study, I shall examine four different noun phrase structures in Arabic, namely the indefinite noun phrase consisting of a noun and an adjective, the definte variety of the same structure, the demonstrative noun phrase and the i•fa structure, or the genetive of possession. These structures have been chosen for two reasons: Firstly, they offer the possibility of comparing different cognitive operations implied by the theory. And secondly, as it is important to compare the language processing hierarchy to the input factor, so as to see if there is a correlation between the two which might eventually undermine the theory, noun phrase structures are a suitable mesuring instrument, since they are often taught at an early point in most teaching material.

In order to understand the operations which will be analyzed below, a few remarks on the linguistic structures of Arabic might be in place for readers with no prior knowledge of Arabic. Because the longitudinal study on which the following analysis is based, consists of data from a Danish learner of Arabic, a Danish translation will be provided in each case, thus making it possible to compare learner data with the learner's mother tongue.

There is no indefinite article in Arabic; the Arabic word bait thus equals "a house" ("et hus" in Danish). If the word is to be made definite, the article al is added as a prefix (and written as a part of the Arabic word), thus resulting in al-bait, "the house" ("huset" in Danish). If an adjective is to be added, it must be placed after the noun, and if the noun has the article, the adjective must also have it: Thus, bait kabžr means "a big house" ("et stort hus" in Danish), whereas al-bait al-kabžr takes the meaning of "the big house" ("det store hus" in Danish). Further, if the noun is feminine, (the only genders in Arabic being the masculine and the feminine), the adjective must agree with the noun; this is done by adding a t•ˆ marbü’ta, an "-a", at the end of the adjective, such as in bint kabžra "a big girl" or al-bint al-kabžra , "the big girl".

A definite noun can be made demonstrative by adding the pronoun of h•dh• (masc) and h•dhihi (fem) in front of the definite noun, as in h•dh• al-bait "this house" ("dette hus" in Danish) or h•dhihi al-bint "This girl" ("denne pige" in Danish). As for the i•fa structure, or the genitive of possession, it follows the English pattern of "the house of the girl" with the very important exception that a noun followed by a genitive must not take the article. "The house of the girl" thus becomes bait al-bint in Arabic (whereas in Danish, the pattern is always that of "the girl's house", "pigens hus"). For reasons of simplicity, I shall only deal with the definite i•fa in this study.

Pienemann's language processing hierarchy predicts the following order of acquisition for the four noun phrase structures:

3.1 Predicted stages of acquisition of noun phase structures in Arabic as a foreign language:

The question is now: Does this predicted order of acquisition correspond to the way language learners acquire noun phrase structures in Arabic? In order to give at least a tentative answer to this question, I have recorded the performance of a Danish language learner of Arabic on video while she was doing different oral tasks such as interviews, role plays and presentations of different subject matters in Arabic. The language learner, who is a student of Arabic at the cand. negot. programme [28] at Odense University, Denmark, was recorded 9 times during a period of 15 months, as from the end of her first year throughout her second year of studies. She was chosen for this study because she was one of the most successfull learners of the class in terms of high scores in written as well as oral tests.

Before comparing the predicted stages of acquisition with actual data, it is necessary to clairify the learning objectives of the course which the language learner took. This is important so as to establish an eventual relation between the input and the acquisition order. The language learner was taught Arabic on an average of 6 hours a week for a period of 4 semesters. She was taught by four different teachers during this period, three of whom were native speakers of Arabic (Egyptians), whereas one was Danish. The teaching was conducted exclusively in Arabic from the third month onwards, and the teaching material used was al-kit•b al-as•sž vol. 1 and 2, [29] which constituted the learning objectives of the course. Also the language learner went, for a period of 6 weeks, to a summer school in Tunisia during the semester break between the second and third semester.

As to the noun phrase structures chosen for this study, the learning objectives of the course were as follows:

3.2 Learning objectives for noun phrase structures in "al-kit•b al-as•sž" vol. 1&2

X indicates that the structure is drilled in the exercise section of the lesson, i.e. there is a conscious attempt to teach the structure;

(X) indicates that the structure is not drilled in the exercise section, but that it occurs more than 5 times in the lesson, be it in the text, the exercise or the grammar sections.

(0) indicates that the structure is not drilled in the exercise section, and that it occurs 5 times or less in the lesson

For reason of space, the lessons have been grouped according to when recordings took place. For week 1- 25 , i.e. before recordings took place, lessons are grouped according to the teaching objectives of al-kit•b al-as•sž vol. 1 [30]. Though the overall picture remains clear despite the grouping of the lessons, one should notice that there are, in fact, variations in the frequency of the use of noun phrase structures in the lessons which are grouped together, especially in the first part of the teaching material: As for the <N-adj> indef, for example, the structure is mentionned twice in lesson 2 (grammar section), once in lesson 4 (text section), twice in lesson 5 (text section) and three times in lesson 7. There is no drilling of this specific structure, but there are exercises on the conjugation of nouns in the masculine and feminine as well as conjugation of nisbe adjectives, which together constitute the building blocks of the <N-adj>structure. The same kind of pattern holds true for the <N-adj> def which is used once in lesson 2, 3 and 6, whereas in lesson 9, it is used more than 5 times. As in the <N-adj> indef, there is no drilling of the <N-adj> def in the exercise sections, but the building blocks of nouns and adjectives are there.

As for the demonstrative NP, it is not used very often in the text (e.g. twice in lesson 6, once in lesson 12, three times in lesson 23), nor is it drilled explicitely, apart from lesson 4, where the language learner, in five exercises, is asked to transform sentences from the pattern of h•dh• al-bait lž into a pattern of h•dh• baitž. I•fas, however, are very frequently used as from the first lessons onwards, and are drilled accordingly.

The following table provides information on the acquisitional order of one language learner on the chosen noun phrase structures. It should be noticed that, following the criterias used by Pienemann, the data only include productive utterances, whereas imitations and word-by- word repetitions of teacher and learner talk are excluded. [31] The numbers represent the percentage of correct answers out of the total use of the structure, whereas brackets indicate that the structure is produced less than 5 times. As for the /, it indicates that the structure is not produced once. Thus, 0.5 indicates that half of the produced structures were correct; (1.) indicates that all of the structures produced were correct, but that it was produced less than 5 times.

3.3 Acquisition of noun phrase structures by one language learner

It is important to keep in mind here that these data are the result of the language learner doing production tasks under (normal) communicative pressure. When communicating, the language learner often experiences a difference between knowing a rule and using it, i.e. between metalinguistic awareness and language use. Therefore, a low score in the use of a given structure does not, per se, indicate that the language learner does not "know" the rule underlying the correct language use. [32] A low score might as well be an indication of a wide gap between knowing and using, between competence and performance. [33]

4. Discussion

"The main purpose is not to describe the point in time during the process of language development when a structure is mastered (in terms of correct use of target norms), because this is only to pinpoint the end of the acqusition of a certain structure. Rather, the above criterion is intended to define the first systematic use of a structure, so that the point in time when the learner has - in principle - grasped the learning task can be located". [36]

In this study, I shall subscribe to Pienemann's approach to acquisition in the sense that I will focus on the acquisition process as a whole, thus not indicating any threshold level which the learner must pass in order to prove that she has acquired the given noun phrase structure. However, it should be noticed that if a structure is used, correctly or incorrectly, less than five times in any of the recordings, as indicated by brackets in 3.3, it is considered a score too low to constitute any clear evidence of the acquisition process; however, it might give some indication as to what direction the acquisition process is moving.

4.1 <N-adj> indef

The incorrect use of the <N-adj> indef has almost exclusively to do with the conjugation of the adjective: As predicted in 3.1, it is typically the feminine form which creates problems for the language learner. This can be seen from examples used by the learner such as *bint kabžr (girl big) "a big girl" or *sharika ajnabž (company foreign) "a foreign company", but also, and very interestingly so, *rajul kabžra (man big) "a big man" and *majlis urübiyya (council european) "a European council), [37] though such structures are much less frequent. Interesting also is the fact that the data show no mistakes in word order, despite the fact that the sequence <N-adj> is contrary to Danish and English word order, the latter being the language learner's first foreign language. The reason for this is, I suppose, that from a cognitive perspective, nouns are more perceptually salient than adjectives, since they contain the basic meaning of the noun phrase. This argument is in line with Pienemann's premise, as mentionned in section 2, that structures which are perceptually salient are easier to acquire than structures which are not.

We can thus conclude that despite the fact that the <N-adj> indef is not consciously drilled, the language learner acquires the structure as from week 25 (or even earlier) onwards, and that, not surprisingly, the main problem during the acquisition process seems to be centered around the conjugation of the adjective in the feminine, whereas the word order does not constitute any learning barrier. In terms of the language processing hierarchy, this indicates that the learner has passed level 2 as from week 25, and is now struggling with local morphology at level 3.

4.2 <N-adj> def

It is possible to trace an interesting pattern in the language learner's use of the <N-adj> def during the acquisition process: Up till week 29, she uses definite nouns as well as adjectives (the latters are mostly used in nominal sentences such as an• danim•rkiyya (I Danish) "I am Danish") in separate contexts, but very rarely, she combines the noun and the adjective into a definite noun phrase, the exception being expressions like al-lugha al-Þarabiyya "the Arabic language" which are learned as chunks. From week 34 up to week 50, the language learner combines the definite noun and the adjective in a <N-adj> def structure following the incorrect pattern of *<al-N adj>, e.g. *al-sayara kabžra (the car big) "the big car" instead of <al-N al-adj> which is the target form. The incorrect pattern is used even to the extent that <N-adj> def structures which were used correctly as chunks, are now submitted to this erroneous pattern. However, in week 60, the language learner seems rather suddenly to master the structure: Not only has the number of sentences in which she uses the structure risen dramatically at this point (from 12, 9, 5, 5, 3, 0 to 27 in week 34, 36, 39, 44, 50, 55 and 60 respectively); also, the number of erroneous sentences have diminished considerably and consist basically of mistakes in the conjugation of the adjective, whereas the definite article required for the adjective in the <N-adj> def is used correctly.

Only one figure blurs the overall picture: The complete lack of use of the <N-adj> def structure in week 55, which precedes the sudden mastery of the structure in week 60. Of course, the phenomenon could be caused by the nature of the task which the language learner was asked to do in week 55, though one has to admit that the <N-adj> def is a very common structure in most discourse. A closer look at the data shows, however, that the task (describing a picture) presents many occasions for the language learner to use the structure. Therefore, it seems rather likely that what happens here is an avoidance phenomenon. A possible explanation, and indeed a plausible one, is given by Pienemann as a part of his teachability hypothesis: He argues that if a learner has to perform an operation (as in our case, the transformation of an <N-adj> indef structure into a <N-adj> def) the consequences of which (adding the definite article to the adjective, which is an intraconstituent operation) the language learner is not yet able to handle due to lack of the necessary language processing prerequisites, this will often result in the learners stopping to perform this operation, [38] until it can be handled succesfully. According to Pienemann, this goes especially for learners who tend to prefer correctness over communicative efficiency:

"... correctness, however, comes at a price: namely the abandonment of a communicatively very efficient means of attention focusing (...). Thus, the premature teaching of INV (inversion) has lead to the "withdrawal" of an already developed means of communication." [39] "...we can see quite clearly that such a strategy of avoidance does nothing to promote the acquisition process, but rather limits the expressiveness of the learner's language." [40]

The conclusions to be drawn from this section are twofold: Firstly, the starting point of the acquisition process of <N-adj> def is initiated at least eleven weeks later than its indefinite variety despite the fact that from an input point of view, the two structures are presented to the student in more or less the same way and almost at the same time. Also, the <N-adj> def structure is only mastered in week 60 compared to week 36 for the indefinite variety. This is in complete agreement with the predicted order of acquisition, as was shown in table 3.1. Secondly, as we cannot decide whether the development order is due to lack of training, inter- or intralingual tranfer or to the lack of cognitively adequate language processing devices, as proposed by Pienemann, at least we can conclude that the acquisition pattern does not contradict the language processing hierarchy established by Pienemann.

4.3 Demonstrative noun phrase <dem pro al-N>

Though the language learner is frequently met, as from week 30 onwards with the demonstrative noun phrase in the texts, she produces it very infrequently, and even in week 60, she does not seem to master it (the figure (1) in week 60 representing only 2 sentences). What draws the attention, however, is the figure of 0.4 in week 36: Here, the language learner uses the demonstrative noun phrase 27 times, but a closer look at the data reveals that 22 of them are based on two nouns only (œarb , "war" and waqt "time"), the 5 remaining examples covering 4 different nouns. Not once is the demonstrative pronoun put in the feminine, though two of the nouns used require a feminine form. Furthermore, there seems to be a non-systematic variation in the correct and the incorrect use of the definite article: The target norm of (fž) h•dh• al-waqt "(at) that time" is used 4 times, whereas in 6 cases the article is omitted. The same goes for the demonstrative noun phrase based on œarb, in which the article is used 7 times, whereas it is omitted in 5 cases.

Thus, the acquisition pattern seems to be the following: Up till week 36, the language learner does not use the demonstrative noun phrase once (though the data reveal clearly that she knows the demonstrative pronoun, which she uses frequently in nominal sentences, though always in the masculine). In week 36, the data show a non-systematic variation of the target form of <dem pro al-N>, h•dh• al-waqt (this the time) "This time" alternating with the incorrect pattern of *<dem pro N>, *h•dh• waqt. The latter then seems to prevail in week 39 and 55, whereas in week 60, the target form is produced twice.

As was the case in the <N adj> def, there might be several explanations to this interlanguage development: Firstly, we should not exclude that individual variety or lack of proper training might play a part, though the counter argument as to the latter would be that the learner has seen and listened to the structure very frequently in texts from week 30 onwards. Secondly, the influence from English or Danish demonstrative noun phrases ("this war" or "denne krig") might "push" the learner towards the use of the incorrect structure of *h•dh• œarb. Thirdly, as was mentioned above, it might be a case of intra-language transfer from the nominal sentence structure (h•dh• bait (this house) "This is a house" as compared to h•dh• al-bait (this the house) "this house"), though again, there is no indication in the data that the learner seems confused about the two structures. And finally, the lack of language processing devices to perform the cognitive operation of adding not only a demonstrative pronoun but also the definite article at an intraconstituent level might be the reason for the erroneous pattern.

Once again, it is impossible to decide which explanation would be the right one on the basis of our data, since Pienemanns explanation as well as the inter- and intralingual transfer are not mutually exclusive. Rather, they all push in the same direction. However, in this case, I would dare, on the basis of my own experience as a language learner of Arabic, to attempt an answer

which makes sense, at least at an intuitive level: To me, it seems to be a combination of the interlingual transfer (from Danish "Dette hus", reinforced by English "This house" into the Arabic h•dh• bait* ) and the difficulty of performing an interconstituent operation of adding the definite article, as indicated by Pienemann. But again, this is based on introspection, not on empirical data.

Conclusively, the development pattern revealed by the data shows that the acquisition process has started in as far as the language learner uses the structure, sometimes rightly, sometimes wrongly. The non-systematic use of the article as well as the complete lack of conjugation of the demonstrative pronoun in the feminine form indicate however that the process is still at a very early stage. Thus, the acquisition of the demonstrative structure takes place much later than the <N-adj> indef and def despite the fact that it is introduced, and consciously drilled, almost at the same point in time as they are (though one has to add that the structure is less widely used in the first 29 weeks than the <N-adj> indef and def).

4.4 I•fa, or the genetive of possession <N al-N>

A closer look at the data shows that all the way through the data collection period, the language learner tends to use the incorrect form of <al-N al-N> as in e.g. *al-mak•n al-œarb (the place the war) "the place of the war" instead of the target form <N al-N> mak•n al-œarb. However, parallel to this structure, she also develops her acquisition pattern as follows: Week 25 and 29 show the correct use of chunks, such as raˆžs al-dawla (head the state) "the head of state" and Hujra(t) al-julüs [41] (room the sitting) "sitting room", which are used frequently in the teaching material. Then, from week 34 onwards, the language learner becomes more linguistically creative in the sense that she combines a definite noun with a second noun consisting of a name, according to the incorrect pattern of *<al-N N>. The target norm in most of these cases, especially when it comes to foreign names, is that the second noun does not take the definite article. Thus, the language learner produces the incorrect pattern of e.g. *al-madžna(t) Odense (the city Odense) "the city of Odense", instead of madžna(t) Odense.. Then comes a stage of correct use of this pattern as in e.g. sharika(t) SAS "the company of SAS", madžna(t) Tünis "the city of Tunis (week 44 and 50). And finally, in week 60, the target norm is used in creatively constructed i•fas, though often the language learner has to make a tentative effort as in e.g. al-bid•ya..... al-bid•ya ...... bid•ya(t) al-œa•ra "the beginning...the beginning ... the beginning of the civilization". However, the correct use of the target norm is still used in what appears clearly to be a non-systematic variation with the <al-N al-N> structure, as in *al-bid•ya al-mushkila "the beginning of the problem".

The interlanguage development can be described as follows:

Now, the most noticeable feature in the acquisition pattern of the i•fa structure is the consistent use of *<al-N al-N>. A plausible explanation to this construct might be that the learner forms a compound of two definite nouns, in the same way as the English "the beginning of the problem" (though in the learner's L1, the target form is "problemets begyndelse", literally * "the problem's beginning"). Thus, one could explain the consistent use of <al-N al-N> as being the result of the learner's linguistic awareness that a definite noun must always take the al-, also in case of compounds. The lenght of the compound does not seem to affect this awareness, as can be seen from examples like *al-raˆžs al-dawla al-yaman (The head the state the Yemen) "Yemen's head of state" or al-raˆžs al-barlam•n al-yaman (the head the parlament the Yemen) "Yemen's head of parlament",which are taken from the data collected in week 60.

In Pienemann's terms, a compound of the <al-N al-N> type is a simple operation of putting words together without adding any morphological features. Only the word order has to be reversed compared to the learner's L1, but if we apply the principle of perceptual saliency as proposed by Pienemann, it seems to be an easy operation, since it is the first noun of the i•fa which carries the meaning, the following one(s) having an attributive function. According to the language processing hierarchy, this operation will take place as early as level 2, as was the case for the <N-adj> indef. Thus, in Pienemann's framework, it is not surprising that from week 29 onwards, the language learner tends to use the <al-N al-N>

The use of <al-N al-N> alternates with another interesting development in the language learner's use of i•fa structure: The period of interest begins when the language learner starts using her linguistic creativity to form new i•fas, that is, from week 44: First, she attempts the *<al-N N> structure, a compound consisting of a definite noun and the name of a place. Such a structure might be considered an amputated form of the <al-N al-N>, in which the learner does not have to pay attention to the al- at the interconstituent level. Furthermore, this structure has a strong resemblance to the English "The city of Tunis" (though the Danish structure is "Tunis by"). Then follows the correct use of the <N N> structure, where the first al- is deleted according to the target norm. And finally, the language learner produces the target form of <N al-N>, the second part of which is not a name, in such a way as to delete the initial al- (endpoint operation) and keep the second one (at the intraconstituent level). This development corresponds to the predicted acquisition order of Pienemann in the sense that the language processing devices allow the learner to perform endpoint operations before involving intraconstituent ones. Notice also that the learner starts performing i•fas by the most easy structure, that of names of places, companies, people etc. which do, in many cases, not require an al- at the second noun.

Thus, two conclusions can be drawn from this section: Firstly, though the language learner is presented to the i•fa structure as from lesson 2 onwards, including intensive drillings, the acquisition process seems to be initiated as late as in week 44, where the language learner produces, for the first time, a correct i•fa structure of her own choice. At the end of the data collection period, the structure is not yet mastered. Secondly, the acquisition order seems to correspond to Pienemann's language processing hierarchy in the sense that the language learner uses the compound structure of <al-N al-N> which is predicted to take place at level 2, as from week 29 onwards, and parallel to this develops an acquisition order which shows a move from an endpoint operation towards a structure which requires an intraconstituent operation.

5. Conclusion:

The development pattern of each of the four structures fits into Pienemann's language processing hierarchy in the sense that reversed word order with no morphological operation preceeds transfer of local morphology to endpoints which again preceeds transfer of local morphology to intraconstituent points.

Also, it should be noted that the input does not seem to be a decisive factor in the acquisition pattern: Though the acquisition process of the i•fa is drilled more intensively and frequently than e.g. the <N-adj> def, the acquisition process of the latter is initiated 8 weeks before the i•fa, and seems to be mastered in week 60 whereas this is not the case for the i•fa.

These conclusions are consistent with Pienemann's teachability hypothesis. However, if we compare the acquisition patterns of <N-adj> def, the <dem pro al-N> and the i•fa to each other, we are left with one important question: How come that the <N-adj> def is acquired earlier than the demonstrative noun phrase and the i•fa, despite the fact that the former contains an intraconstituent operation? In other words: Why is it that the language learner can perform the intraconstituent operation successfully in the <N-adj> def whereas she cannot as of yet do the same thing in the demonstrative noun phrase and the i•fa? The theory does not provides us with an answer to that question, but a possible explanation might have to do with cognitive processing and the nature of operations which the language learner has to perform. In order to understand the point I want to make, it is necessary to consider two factors: firstly, the relationship between the different parts of the noun phrases; and secondly, the cognitive direction of the operations.

As the term "noun phrase" indicates, the noun is the core of structures under investigation here. It is the noun which is the most perceptually salient in the structures of <N-adj> indef and def, whereas the adjectives have an attributive function and can be left out, if one wishes. Also on a grammatical level, it is the properties of the noun (whether in the masculine, feminine, singular, dual or plural) which determine the form of the adjective (the definite article being invariable in Arabic). The same goes for the demonstrative noun phrase where the properties of the noun determine the form of the demonstrative pronoun. As for the i•fa structure which is composed of two or more nouns, it is the first one which constitutes the core of the noun phrase, the other(s) having an attributive function; this can be seen from the fact that what follows the first noun can be omitted without the noun phrase loosing its basic meaning whereas the reverse cannot be the case. So, perceptually as well as grammatically, the noun (and in case of the i•fa, the first noun) is the core element of these structures.

The cognitive direction of the operations concerns the storing process mentioned in section 2: There, the point was that the longer the language learner has to keep the information in his working memory, the less space there will be for other processing devices to work and the more difficult it will be for the language learner to handle the operation. As a consequence, we saw that non-local morphology was predicted to be more difficult to handle than local morphology, since in the case of the former, information has to be stored for a longer time in the working memory. What the predictions did not take into consideration, however, was the fact that, based on the same arguments, the direction of an operation might very well influence the acquisition process as well. Let me illustrate it this way: As we saw in the preceeding paragraph, it is the core, in our case the noun, which triggers the information necessary to form the grammatically correct form of the attributive, be it an adjective or a demonstrative pronoun. Now, if the core is realized before the attributive, the language learner will only have to retain the information triggered by the core for as long as it takes to realize the attributive. Graphically, this can be illustrated as

5.1 Progressive transfer

5.2 Regressive transfer

This is termed regressive transfer, because it is the second part of the noun phrase which determines the form of the first one. This kind of transfer is assumed to be more difficult for the language learner to process than the progressive one, not only because it alters the perceptual and the grammatical sequencing compared to the progressive transfer; but also because it burdens the working memory of the language learner for a longer period than the progressive transfer. Thus, the assumption made here is that, just as non-local morphology is more difficult for the language learner to handle than local morphology, so is regressive transfer compared to progressive transfer; therefore, non-local morphology and regressive transfer will be performed later than local morphology and progressive transfer.

Now, how does this apply to the three noun phrase structures in question? If we consider the <N-adj> def structure, the acquisition process can be described as follows [42]

5.3 direction of transfers for <N-adj> def

As for the demonstrative noun phrase, the acquisition order consists of a non-systematic variation of the following two structures, both of which include regressive transfer:

5.4 Direction of transfer for <dem pro al-N>

However, as was shown in 4.3, the pattern of *<dem pro N> is the most widely used of the two and prevails in week 39 and 55 (week 44 and 50 showing no examples of the structure at all). A reason for this might very well be that it only comprises one transfer whereas the <dem pro al-N> comprises two.

The transfer performed by the language learner only takes definiteness into account, leaving out the necessary conjugation of the pronoun according to the masculine-feminine distinction. Though our data collection period stops at week 60, it seems plausible that the third stage of the acquisition pattern will include this operation, in the case of which we can establish the following acquisition pattern:

5.5 Predicted acquisition pattern for <dem pro al-N>

In the i•fa structure, the first noun constitutes the core of the noun phrase, but in this very specific context, it is the attributive, that is the second noun, which requires the deletion of the al- prefix of the first noun, according to the rule that a noun followed by a genitive must never take the article. Thus, the learner has to perform a regressive operation in order to reach the target norm. However, this regressive transfer being too difficult for the language learner to perform due to lack of language processing devices, she proceeds in the following way:

5.6 Direction of transfer for i•fa structures

Now, if we compare the three structures in question from the point of view of the progressive - regressive transfer, there is an emerging pattern of structures including progressive transfer being acquired before structures including regressive transfer. Thus, the structure containing the two progressive transfers of the <al-N al-adj> is acquired before the structures of <dem pron al-N> and the i•fa both of which are based on regressive transfer.

Conclusively, it seems as if the order of acquisition of Arabic as a foreign language can be explained according to cognitive operations as they are defined by Pienemann's language processing hierarchy. Furthermore, by including the distinction of progressive and regressive transfer, we might be able to get a more detailed understanding of how the acquisition takes place at the various levels. However, one should not forget that the present conclusions are based on a longitudinal study of one language learner only. This is far from being satisfactory, if we want to build a solid ground on which we can base our understanding of acquisition order in foreign language learning. Therefore, what is needed from here is to extend the analysis of acquisition order to encompass a wider scope of learner language data and to analyze a more diversified range of structures of which the progressive - regressive distinction should be part.

Helle Lykke Nielsen