Formulaic Memorization as Barrier to Language Learning


Download Formulaic Memorization as Barrier to Language Learning


Preview text

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 11
Formulaic Memorization as Barrier to Language Learning
Amjed Saleem
Abstract
Can memorizing text in the target language facilitate L2 learning? This paper examines an extreme case—the Quran as memorized by Muslims who do not have any independent knowledge of Arabic. The hypothesis, derived from existing research literature, was that they would develop implicit sensitivities to certain grammatical patterns in Classical Arabic as a result of repeated exposure to fully correct formulations, unadulterated by any interlanguage features. Non-Arabic-speaking Quran memorizers were asked to distinguish Classical Arabic sentences with and without an introduced morphological error. Contrary to the hypothesis, they could not identify ungrammatical sentences beyond chance level, and could rarely pinpoint the error in a sentence that they did isolate. A small follow up study using Quran memorizers who spoke modern Arabic revealed the same pattern, indicating a surprising lack of extrapolation from Modern to Classical Arabic, despite sufficient points of similarity to make pattern identification feasible. The reason for the findings is considered in the light of a model of how memorization is, or is not, a support to accurate language learning.
1. Introduction
1.1 Memorization as a means of language learning For some time, there has been a debate in the research literature about whether language learners can benefit from not only being exposed to, but also deliberately memorizing, target language text (Nattinger & De Carrico 1992; Wray 2000, 2002, 2008). The rationale for supposing such benefit to learnersis that accurate memorization would somewhat protect learners from the contamination arising from extensive exposure to their own and others’ non-target-like output. The more they are exposed to material that is not only accurate but known by them to be accurate, the more they can trust it, and gradually develop intuitions about what ‘sounds right’.
Foremost in making this claim have been researchers in China, where traditional educational approaches favour a disciplined approach to internalizing knowledge in both language learning and the rest of the curriculum (Dahlin & Watkins 2000; Kennedy 2002; Liu 2002). In the L2 learning context, Ding (2007) linked success in a prestigious English language competition in China to strict memorization practices extending back through the school years. He found that students were often pressurized into committing texts to memory even before they fully understood them on the basisthat accurate target language material safely stored in the learner’s head was then reliably available for subsequent examination, to bootstrap further learning. Such students, provided they saw the tough regime of memorization through, emerged with strikingly native like capacities, perhaps due to a growing sense that the language they had memorized had become their own (p. 275). To put it another way, by reproducing texts as output that would otherwise only constitute input, they effectively bridged the gap between receptive and productive capabilities.
It goes without saying that language learners would not get far if their only recourse to was to memorized material. Bloom (1973: 17) notes, with reference to first language acquisition, that we need much more flexibility of expression than that. Research into formulaic language has

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 12

concerned understanding the boundary between novel and previously encountered forms, with notable attention to how language teaching might benefit from not assuming that everything is novel. As Nattinger & De Carrico (1992: 27) point out, “a great deal of language that people are exposed to every day is very routine and predictable, just as are the situations they encounter”. It follows, they argue (p.113ff), that provided a wordstring is suitable for the communicative job in hand, it may not matter that is not something you made up yourself. Knowing that it is 100% correct and nativelike must be a bonus that instils confidence.
Figure 1 lays out a basic model for how learning might result from the memorization of target language material. The key process depicted is the building up, through repetition of the teacher’s and one’s own faithfully reproduced models of texts, of a body of episodic memories that reinforce the particular patterns the texts contain. Port (2007) and Taylor (2012) are two theorists who recognise episodic traces to play a crucial role in pattern learning, in phonology and lexis respectively.
Auditory trace of teacher’s example

imitation

EPISODES of rehearsal and recital

imitation

Auditory trace of own
rehearsal

FFrFreFreqreqeeuquqxeuepunenoecncsyncyucyoryoefofoff eexexpxpopososusurureree

Extrapolated knowledge of patterns
Figure 1.1: Modelling language learning from memorized material
Based on this model, one might reasonably ask why a learner would not benefit from memorizing and reciting native like texts. To put it baldly, if a learner knows that he will not only learn how to say things that are accurate and native like, but also beable to extract the information necessary to construct new, native like utterances, why would he not prioritize this approach? Why would teachers not do so? Why did drilling everywhere go out of fashion?
There are of course, good reasons why. Figure 1 does not tell the whole story. In section 5 we will return to it with a critical eye, drawing on research that has demonstrated memorization not to be as effective for accurate learning as we might expect. But first, we must consider the single most likely context in which Figure 1 actuallymight hold true—one in which the learner’s extrapolation of patterns would be maximally facilitated by the absence of any competition from other, lessregulated input.

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 13

1.2 Quran memorization by non-Arabic speakers The Quran is the primary religious text of Muslims. It is written in Classical Arabic and is read and studied by millions of Muslims all over the world in its original version. (Later, it will become relevant to discuss the similarities and differences between Classical and Modern Arabic – see 4.2 below). Memorization of sections of the Quran is part of almost every Muslim’s early education. But in addition, a good number of Muslims memorize the whole of it, and memorization of the Quran in full has always been an important tradition in the Islamic world. It is a point of honour to have a Quran memorizer, or Hafiz, in the family; and across the world Muslim boys and girls are encouraged to attend classes after school and at weekends to commit the full text to memory.
The word Hafiz (plural, Huffaz) literally means ‘guardian’ and this is important. For it is imperative that memorizers apply the highest standards of accuracy to their reproduction of the text. Changing the Quran in any way is considered blasphemous. Yet the holy Quran is a text of considerable length. It consists of 114 chapters in 30 sections. The total number of verses in the Quran is sometimes taken to be 6666, and sometimes 6236, depending on what is counted as a complete verse. The length of time it takes to read/recite the Quran from cover to cover can be 15 to 20 hours (though it can be recited much faster by fluent Huffaz during Ramadan).
Of particularinterest to us here is how the Quran is memorized with full accuracy by Muslims who do not know any modern Arabic. If one listens to non-Arabic speaking memorizers’ recitation of the Quran, one is struck by their fluency and ease of production, complete with a convincingly near-native pronunciation and lilt of the language. This is achieved according to the pattern laid out in Figure 1, though with some additional supports, which are added in Figure 2. They reflect the particularprocesses by which Quran learning is achieved, at least by sighted, literate memorizers: the position of text on the page, use of the configurations of alphabetic symbols as a cue, and personal images including episodic memories and associations (Saleem 2015).

Auditory trace of teacher’s example

reading

Alphabetic cues

Visual image of text on page

imitation

EPISODES of rehearsal and recital

reading

imitation

Auditory trace of own
rehearsal

Personal associative
images

FFrFreFreqreqeeuquqxeuepunenoecncsnycyucyoryoefofoff eexexpxpopososusurureree

Extrapolated knowledge of patterns
Figure 1.2: Potential learning processes for Quran memorizers

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 14
In the closed system of Quran memorization, we see a potentially perfect context for examining the capacity for a dedicated learner to develop implicit knowledge of an otherwise unknown L2. The study described below was undertaken to ascertain whether one sample of non-Arabicspeaking memorizers were in possession of such knowledge.
3. Investigating the implicit grammatical knowledge of Classical Arabic in non-Arabic-speaking Quran learners
The research question for this study was: Can non-Arabic-speaking Quran memorizers distinguish ungrammatical from grammatical sentences in Classical Arabic? As outlined above (and discussed in more detail later), there is a rationale, in theory at least, for why they might be able to. Thus, the hypothesis under test was that they would be able to distinguish the stimulus types at above chance level.
3.1 Participants The participants were eleven1male memorizers drawn from three different mosques in a UK city. Nine out of the eleven participants had memorized the whole of the Quran. They were all aged over 18, were born in the UK, with parents born in Pakistan and from the Pashtun community. None of them had learned any Arabic, modern or Classical, other than that entailed in reading and writing the Quran, which they undertook without understanding the text (Saleem 2015). It is not possible to say that they knew no Arabic vocabulary at all, because Pashto and also Urdu (one of the official languages of Pakistan), being languages associated with Muslim cultures, have imported many Arabic words. In addition, Arabic itself contains Persian words that have independently entered both Pashto and Urdu. All the same, it was accurate to state that these memorizers did not have any specific knowledge of Arabic.
3.2 Stimuli For the experiment, a set of Classical Arabic sentences was required, some of which were grammatical and some of which were not. Two important considerations had to be kept in mind. Firstly, they needed to be sentences that the participants had never seen before. Secondly, it would be extremely unacceptable to introduce errors into sentences from the Quran. For this reason, texts contemporary with the Quran were used as the source for the test items. With the assistance of a Classical Arabic scholari thirty sentences were selected that were similar in overall construction and length to verses from the Quran. Half were left as they were, and the other half were changed, to introduce a morphological error.
The errors so-introduced adhered to strict criteria. Firstly, they were associated with one of three marker types, number, gender or case, in equal measure. Second, it had to be possible to identify the error without knowing the meaning of words in the sentence. That is, the agreement needed to be at odds with another morphological marker, since these closed class forms would be much more frequent in the Quran, and carry their ‘meaning’ in part through their colligation with other markers. To exemplify this using English, it would be unreasonable to expect memorizers who did
1The use of only eleven informants was a result of finding remarkable similarity in their responses, which suggested no additional value would be gained from further data collection. In lieu of a larger sample from this population, data were also collected from representatives of another population of memorizers, for the purpose of comparison, as described in section 4.2.

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 15

not understand the language to know that the word ‘duke’ refers to a male person, and thus inappropriate to expect them to spot the error in *The duke washed herself. On the other hand, it would be reasonable to test their capacity to note the error in *He washed herself because exposure to texts, even without much knowledge of the language, would presumably result in more exposure to the patterns She … herself and He… himself than *He … herself.

It is difficultto fully illustrate using English the opportunities for agreement available in Classical Arabic, which is much more morphologically inflected. Thus, we illustrate now the sortsof sentences used in the experiment. One of the five for each of the three types is given in Table 1. The correct version shows the morphological agreement across the sentence. The version with the error shows the morpheme that was changed, while the rest remained the same.

Table 3.1: Example stimuli

Type

Correct version

Gender Number Case

Wayla-kum l-ajra

ta’khuz-ūna wa-l-

a’mal-a tufsid-ūna

sawfatalq-awna mā

tahzar-ūna.

La tastazilla l-faqīr-a

wa-la

taghbiti-l-

ghanniya wa-kun

enda zikr-ī Khāsih-

an.

Inna l-Allāh-kariha la-

kum l-a’bas-a fī l-

salāt-I wa-l-rafas-a fīl-

seyām-i.

Translation

Woe unto you! You take wages but perform bad work. You shall encounter what you are afraid of. Do not despise the poor and do not envy the rich, and be humble when mentioning Me.

God does not want you

to be frivolous during

ritual prayer or to

behave

indecently

during fasting.

Version with introduced error Wayla-kunnal-ajra ta’khuzūna wa-l-a’mal-a tufsid-ūna sawfatalq-awna mā tahzarūna.
La tastazilla l-faqīr-a wa-la taghbiti-l-ghanniya wa-kun enda zikr-ī Khāsih- īna.
Inna l-Allāh-kariha la-kum l-a’bas-a fī l-salāt-a wa-lrafas-a fīl-seyām-i.

In these examples, we see that, in accordance with the design criteria, correctness or incorrectness is morphologically marked and can be derived from the other morphological markers. For example, in the gender example, the interjection waylakunna (“woe unto you!”) that opens the sentence suggests that women are being addressed (the pronominal suffix -kunna is 2nd person feminine plural). The rest of the sentence contains four verbal forms marked as masculine plural by the ending -ūna or -awna. Thus, either waylakunna is an error for waylakum (“woe unto you”, masc. plur.), or the four verbs are all wrong (instead of the masculine plural forms taʾkhudhūna “you take”, tufsidūna “you spoil, corrupt”, talqawna “you encounter”, and taḥdharūna “you are afraid”, one should expect the feminine plural forms taʾkhudhna, tufsidna, talqayna, and taḥdharna, respectively.)

In the number example, the imperative kun (“be”) is masculine singular and according to the rules of Arabic grammar the predicate is correctly given in the accusative. However, kun is singular and the predicate should also be in the singular: khāshiʿan (“humble”); khāshiʿīna, with the plural marker -īna, is incorrect. Khāshiʿīna, would be correct if the imperative was plural, that is, kūnū

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 16

instead of the singular kun. The plural of tastazilla and taghbiti, the other two imperative singular verbs, would then be tastazillū and taghbitū.

As far as the case sentence is concerned, in Classical Arabic every preposition must be followed by a noun or pronoun in the genitive. In this sentence, fī l-ṣalāta is incorrect because the preposition fī (“in”) is followed by a noun that has the marker of the accusative, -a.

In addition to the two sets of test sentences just described, fifteen sentences were selected from the portion of the Quran that the participants hadmemorized, for use in task 1 (see below).

3.3 Design The participants were asked to undertake three tasks, the first two of which involved sorting cards on which individual stimulus sentences had been written. In task 1, they were given forty-five individual cards featuring, respectively, the 15 sentences from the Quran, the 15 correct sentences of Classical Arabic, and the 15 incorrect sentences. The cards were shuffled into random order, and they were asked to separate them into two piles: sentences from the Quran and sentences not from the Quran. The purpose of this first task was to establish that the participants were able to examine text in Arabic script and do something with it. It was predicted that they would have no difficulty with this task, since they were accustomed to seeing, and reading, the Quran sentences.

The second task involved a fresh set of just the non-Quranic sentences, which they were asked to separate into correct and incorrect. In both the first and second task they were asked to read the sentence aloud, since this was anticipated to help them judge its status. In the third task, the pile of cards that they had separated out as incorrect was used. For each sentence, they were asked to read it aloud and then say where they believed the error was and, if possible, correct it. Note that the pile used in task 3 contained only those sentences with errors in that they had picked out as incorrect, and it contained also any correct sentences that they had identified as containing an error.

3.4 Analysis Task 1 was easily accomplished by the participants. The combined results are shown in Table 2. In only nine instances out of 165 (15 x 11) was a Quranic verse judged to be non-Quranic. Rather more non-Quranic verses were judged to be Quranic, but still it was less than 10% (30 out of 330, i.e. 30 x 11).

Table 3.2: Combined results for task 1

Judged as Quranic

Quranic

156

Non-Quranic

30

Total

186

Judged as non-Quranic 09 300 309

Total 165 330 495

Fisher’s exact test was used2 to compare the actual distributions with those predicted by chance. For the combined group, the difference was highly significant (p < 0.0001, one-tailed). Next, each
2 The sample was too small for a linear regression. Use of both Fisher’s exact test and chi squared.

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 17

participant’s individual performance was analysed. The statistical values ranged from 11.5 to 45 and all results had a p value of less than 0.01. Participants performed at this level even though the Quranic sentences were presented out of context: they were isolated sentences picked up from different parts of the Quran at random (though excluding parts that had not yet been memorized). The participants’ behaviour while performing the task was consistent with having a strong mental representation of the Quran verses as compared to the non-Quran texts. Although their speed of response was not measured, there was a noticeable tendency for the Quranic verses to be categorised more quickly than the nonQuranic ones. Moreover, the participants were not fluent in reading sentences from unseen Classical Arabic texts and, at times, found it difficult to pronounce them correctly.

In task 2, it was a different story, with little evidence that the two stimulus types could be distinguished (see Table 3). A Fisher’s exact test gave a non-significant p value of 0.2158 (onetailed).

Table 3.3: Combined results for task 2 Judged as grammatical
Actually grammatical 95

Actually

103

ungrammatical

Total

198

Judged as ungrammatical 70 62
132

Total 165 165
330

Examined individually, the Fisher’s exact test values ranged from 0 to 3.393, and, with all p values greater than 0.05, none was significant. Thus, the hypothesis that the Quran learners would display sensitivity to morphological patterns in unseen Classical Arabic sentences was not supported. Task 3 sought to understand the basis on which sentences had been judged ungrammatical. Each participant was asked to read aloud and then comment on the sentences in their ‘ungrammatical’ ile. Table 4, which reports only those items in the ‘ungrammatical’ pile that actually were ungrammatical, shows low rates for locating the error, and even lower ones for correcting it.

Raises a potential problem, in that the stimuli themselves are a potential source of variance (for discussion of how individual sentences were responded to in tasks 2 and 3, see [reference withheld]). However, a statistician confirmed, after simulating a larger dataset, that these tests offer a reasonable account of the patterns

Table 3.4: Identification of error location and cause in task 3

Participant Correctly identified as Location of error correctly

ungrammatical (out of identified (out of n correctly

15)

identified)

1

8

1

2

6

0

3

2

0

4

4

0

5

8

0

6

6

1

Correct explanation of error given (out of n correctly identified) 1 0 0 0 0 1

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 18

7

5

4

0

8

5

5

2

9

7

2

0

10

5

4

1

11

6

3

1

However, these low figures are still meaningful. While it is easy to understand how a random allocation of sentences to piles could give many correct answers by chance, it is much less obvious how chance could lead participants correctly to pinpoint what the error was, let alone come up with a correction. In addition, it is useful to examine whether errors in all three types of morphological marker (gender, number and case) were equally likely to be identified.

Dealing with the latter issue first, ungrammaticality in gender agreement was the most likely to be correctly categorised as such, with 29 instances out of 55. However, this was only a little over 50%, which is chance level. Number agreement errors were only correctly categorised 19 times out of 55 (34.5%) and case ones 14 times out of 55 (25%). These low levels reflect the overall conservatism of the participants, who tended to judge sentences grammatical more often than ungrammatical (see Table 3), and it seems that the gender type was elevated by a greater genuine awareness of errors in that type.

Of the 29 instances of correct categorization for gender, 11 were accompanied by a correct location of the error, though only three participants were able to indicate what the word-ending should have been (and no two participants did so for the same sentence). In contrast, sentences with an ungrammatical number agreement only attracted two correct locations of the error (out of 19). Only in one case could the participant correct the error (replacing Qālū with Qāla), and he was not able to say why it was correct. Case agreement errors were located only four times (out of 14), and only one of these errors was corrected. It should be noted that across the entire sample, there were no instances at all of an error being located by more than one person. In sum, we see from these results a low level of sensitivity to errors. It falls below the threshold for significance in statistical terms but we cannot completely discard it as meaningless. The explanation is difficult to adduce from the group as a whole, but looking at the individual performances, it is a little easier to understand.

Participant 7 correctly located errors in four sentences (Gender 3, Case 1), participant 8 in five sentences (Gender 3, Number 1, Case 1), participant 10 in four sentences (Gender 3, Case 1), and participant 11 in three sentences (Number 2, Case 1). Participant 8 provided correct replacements for errors in two sentences and participant 11 in one. These four participants were also more confident in their decisions than the others. The performances are still at a low level, but they do suggest that individual variation may be a factor in how well linguistic information is taken in. Later, we return to this matter, since it will be proposed that Quran learners have a particular approach to learning that some may find easier to adopt than others.

4. Discussion: why did the Quran memorizers not perform better?
Figure 2, presented earlier, laid out how memorization might lead to the extrapolation of implicit knowledge of an L2 grammar. Yet we found little evidence of it. This finding resonates with

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 19
observations by Boyle (2006) in relation to memorization and learning in Islamic schools in Morocco, Yemen, and Nigeria:
On a basic level memorization of the Qur’an is associated with knowledge of the Qur’an, although not in the Western sense of being able to understand and explain it, but in the sense of being able to recite it. …[T]he various groups and ages of learners I observed in the three countries were generally unable to explain what they had memorized (p. 487).
The achievements of non-Arabic-speaking Quran memorizers in terms of execution are particularly striking given that they are not able to use semantics to assist their recall. Where actors would identify features in the narrative structure to help anchor their memorization (Noice & Noice 1996), these memorizers cannot do so. Certainly, they deploy mnemonic devices (see Figure 2) but necessarily rather shallow ones, compared to what would be available if they understood the text.
4.1 The relationship between memorization and understanding
If you understand something, really understand it, you will have very strong impression and can memorise it without much effort (Marton et al 1993: 73).
As outlined below, it has been argued by some researchers that successful memorization is only possible with understanding. However, the context of that research is substantially different, and we must examine carefully the extent to which the claims are relevant to Quran learners. Firstly, there is a difference in what the memorization is for. Previous studies identify memorization as a means to achieve effective mastery, or temporary recall, either of content (Cooper 2004; Dahlin & Watkins 2000; Marton et al 1993) or of linguistic form (Ting & Qi 2001; Ding 2007; Wray & Pegg 2009; Dai & Ding 2010).
Regarding content, Cooper (2004: 289) identified in Chinese accountancy students a specific “tradition of memorization through repetition [that] can be used to deepen understanding and achieve high levels of academic performance”. He found this technique more effective than the surface memorization approach used by Australian students. Dahlin & Watkins (2000), similarly, found that while Western students focussed on the process of memorization, their Hong Kong Chinese counterparts focussed on the content. Furthermore, nearly twice as many Chinese (60%, n = 48) as German-Swiss students (33%, n = 18) agreed with the statement “Repetition plus ‘attentive effort’ can lead to new meaning” (p. 68). Thus, as Kennedy (2002: 433) suggests, in the Chinese tradition,
Memorization has never been seen as an end in itself but as a prelude to deeper understanding—mentally ‘photocopying’ texts, committing them to memory, enabled the ‘learner’ to savour and reflect on them later, and, finally, to integrate them with his/her prior learning and experience.

KASHMIR JOURNAL OF LANGUAGE RESEARCH, VOL. 21 NO. 1 (2018) 20
Of course, Quran memorization is not ‘an end in itself’. However, its purpose is not to enhance understanding of the content. The memorizers do not consider that they need to understand it as part of their recall.
When it comes to language learning, memorization is again recognized as potentially playing a role, though the caricature of Chinese learners engaging in shallow, rote learning is vociferously challenged. For instance, Liu (2002) states:
It can be argued that memorizing words and phrases does not necessarily mean rotelearning long lists of vocabulary. If it is based on the process of understanding the materials first by means of elaborate study of the items, it may result in better and longer retention in students’ memory.
As noted earlier, Ding (2007) found that memorization could be the gateway to highly accurate command of a second language. Such a command can even over-state the learner’s knowledge. Wray and Pegg (2009) showed how extensive verbatim learning could enable IELTS test-takers to outwit the protocols of the marking system, resulting in high scores for accuracy and appropriacy that belied their actual knowledge. It was achieved by memorizing lengthy expressions that could be used generically across topics, e.g. In this essay, I will be presenting my opinion on why I believe that...; There are three reasons for claiming that …; To sum up, it is possible to conclude that…. Wray and Pegg observe that since this practice is consistent with what native speakers learn to do under the auspices of stylistic choice for genre, it is difficult to judge it ‘wrong’, even though assessors recognize its contribution to an overestimation of the candidate’s linguistic capabilities.
One could argue that the Quran learners are like the IELTS test takers in Wray & Pegg’s study. They display accurate output based on memorization and recall, without the understanding to match. However, Wray & Pegg do not suggest that these IELTS test performances would be possible without any understanding of the internal content. Although they observed a marked reduction in linguistic accuracy in the novel material between the accurate generic clauses, it is not plausible that the learners had no knowledge of the formal and semantic features of what they had memorized.
Can, we, though, speculate that the Quran learners would learn faster, or better, if they did understand the text? We will return to that question later. For now, we turn to the question of whether their failure to identify the ungrammatical sentences of Classical Arabic was because they did not understand the text. In the light of the existing research literature, it seems entirely reasonable to hypothesize that this is so. And the hypothesis was tested by giving the same task to memorizers who had a much greater potential to gain semantic access to the text.
4.2 Does increased capacity to understand the Quran text result in greater sensitivity to its linguistic patterns?
Four Quran learners who were native speakers of Palestinian Arabic were given tasks 2 and 3 as described earlier. They were all male, aged 18 or over, and born in the UK. None had studied Arabic at school. This meant that they had no literacy in standard modern Arabic, only native

Preparing to load PDF file. please wait...

0 of 0
100%
Formulaic Memorization as Barrier to Language Learning