当前位置:首页 >> 文学研究 >>

First Language and Second Language Writing The Role of Linguistic Knowledge, Speed of Processing, an


I

Language Learning 53:1, March 2003, pp. 165-202

First Language and Se,cond Language Writing: The Role of Linguistic Khowledge,; Speed of Processing, and Metacognitive Knowledge
Rob Schooneh and Amos van Gelderen University of Amsterdam. Kees de Glopper University of Groningen Jan Hulstijn, Annegien Simis, Patrick,Snellings, dnd , Marie Stevenson University of Amsterdam
In this study the relati've importance, of linguistic knowledge, metacognitive knowledge, and fluency or accessibility.of this linguistic knowledge,in both first language (L1; Dutch) and second language (L2; English)

Rob Schoonen and Jan Hulstijn, Department of Second Language Acquisition, Faculty of Humanities; Amos van Gelderen, Annegien Simis, Patrick Snellings, aind Marie Stevenson, SCO-Kohnstamm Institute for Educational Research, Faculty of Social and Behavioural Sciences; Kees de Glopper, Department of Language and Communication, Faculty of
i Letters. The Netherlands Organization for Scientific Research (NWO) is gratefully acknowledged for funding this research' project. The authors also wish to thank the schools and students that participated in this study as well as the test assistants who collected the data. Finally, we would like to thank the three anonymous reviewers for their comments on an earlier draft of this article. Correspondence concerning this article should be addressed to Rob Schoonen, Department of Second Language Acquisition, Faculty of Humanities, University of Amsterdam, Spuistraat 2f0, 1012 VT Amsterdam, the Netherlands. Interinet: rob.schoonen@hum.uva.nl

165

.

166

Language Learning

Vol. 53, No. 1

writing was explored! Data were collected from 281 g~rade 8students. Using structural equation modeling, the,relative importance of the three components was studied and compared across Li and L2 writing. The results showed that the fluency measures were, correlated with overall writing performance in both Li and L2. However, when, compared to linguistic knowledge resources, these fluency measures turned out to have no additional value in predicting LI or L2-writing'perfor'mance.'L2 writing proficiency turned out to be highly correlated with Li writing proficiency, miore than with either L2 linguistic knowledge or the accessibility 6f. this knowledge.

Writing in one's mother tongue is a demanding'task that calls upon several language abilities, as well as upon more general (meta)cognitive abilities. These constituent abilities are in a constant interplay. Writing in a second language (L2) is even more demanding, because several of these coAstituent abilities may be Tess well developed than in one's first language (LI). For example, linguistic knowledge of the L2 may be limited, and the 'accessibility, of this knowledge may be less rapid or automatic. The question is to what extent the decomposition of LI writing in terms of linguistic knowledge, metacognitive knowledge, and accessibility of linguistic knowledge is compara ble to the decomposition' of L2 writing and to what'extent LI and L2 writing are interrelated. I ? ' ' Because of the complexity of the wiiting process, it is 'difficult to envisage .a model for writing in terms of its "subskills" (see'Abbott & Berninger, 1993, and Grabe & Kaplan, 1996, for an attempt to arrive at such a model). Most exisfing' writing models focus 'on the writing process (Chenoweth & Hayes, 2001; Flower & Hayes, 1980, .1983; Hayes, 1996; Kellogg, 1996) or on! the development of writing proficiency (Bereiter & Scardamalia, 1987) more than on .the! characteristics of the cognitive "and linguistic resources,, needed for writing. Process models 'do, however, acknowledge that writers need to have certain

,

Schoonen et al.

.

167

resources availabl'e. Chenoweth and' Hayes (2001) 'distinguish three levels intheir description of the'writing process: a resource level, a'process level, and a control level. The 'resource level consists of 'linguistic, knowledge and general kn6wledge and is, called upon by the processes at the process level; such tas translating and revising. The control levelincludes' a.task schema consisting of the task goal and a set of productions "that govern. the interactions among the pr6cesses". (Chenoweth. & Hayes, '2001, p. 84). At the control level, other kinds of knowledge resources might be called upon,' such as knowledge -of writing strategies. The question is which 'knowledge resources'oi component skills are, of value -to a successful writing :performance (that is, to effective writing and control processes) andl ake up the cognitively complex construct of writing ability., First of all, writers, who have the intention of expressing an idea or ;message to -a reader, need to have some vocabulary knowledge of the' language 'in which they are writing (cf. Grabe &-Kaplan, 1996). Writers' lexical kno'wledge oi vocabulary size is likely to influence the qcuality of their telxts. Measures' of.lexical richness 'of texts correlate substantially with holistic ratings of these texts (Engber, 1995). Also; 'in a study by Laufer and Nation (1995),;it was shown that vocabulary size, use of words of differ-,' ent frequency bands (lexical frequency profile), 'and comnposition rating are highly intercorrelated. Limited lexical resources seem to reduce writers' possibilities for expressing their ideas. However; writers' ideas are not just expressed in single woids but need to be cast in grammatical 'structures 'that indicate the relationships between the constituents in the clauses containing those single words. Consequenti, writers need to have some grammatical knowledge at their' disposal to be -able to connect the words into ,proper clauses and sentences (pf. Grabe' &
Kaplan, 1996). i
' '

In contrast to speaking, writing also,requtires knowledge of the orthogr'aphy, 'of the language,' that'is, spelling '(Abbott & Berninger, 1993). Depending on the, language involved, the "match" between spelling and the spoken language'varies in
be. .eln .n .dgui vaiei

.

I

.

168

Language Learning

Vol. 53, No. 1

terms of transparency. The degree of transpa'rency in a particular language will affect the amount of difficulty that writers experience in encoding their ideas in written form. In producing longer stretches of text, that is, beyond clause or sentence level, writers should be aware of the organization of their texts at discourse levrel. They should also be aware of how their communicative intentions can best be expressed. In a broader, more pragmalinguistic. and sociolinguistic perspective, writers need to have. knowledge of the addressed readership and of ways texts function in their community in order to be able to write effective texts (cf. Cumming, 2001; Grabe & Kaplan, 1996). In addition to all this language-related knowledge, writers need to have (metacognitive) knowledge of what constitutes a good text and which writing strategies are likely to be successful in dealing simultaneously kwith all the constraints writing a text. poses (cf. Flower & Hayes' [1980] "juggling with constraints"). Schoonen and de Glopper (1996) showed that proficient writers have more declarative .knowledge about writing than less proficient writers and that they have a different perception of what is important for a text to be' adequate: Proficient writers -focused more on text organization compared to poor writers, who focused on mechanics and layout. In the same vein, Victori (1999) showed that successful and unsuccessful foreign language writers could be distinguished by their metacognitive knowledge in each of three domains:' knowledge one holds about oneself as cognitive processor, task knowledge, and strategy knowledge. This kind of metacognitive knowledge, which is stored. at the resource level in'the Chenoweth and Hayes (2001) model, may be considered an important knowledge source for. the task schema at the control level that orchestrates the writing process. The presence of linguistic and metacognitive knowledge resources in long-term memory is just one facet of.the writing process. At the process level writers have to access these knowledge resources when they translate their ideas into written language. At this level writers are restricted by the limitations

Schoonen et al.

' 169

of their working memory capacity in the cogniitive processes and rhetorical.constraints they can handle simultaneously. Fluent or automatic accessing of lower-level (linguistic),'knowledge resources may take up little of writers' attention and therefore may leave sufficient cognitive capacity..for other attenition consuming, higher-level processes of writing', such as text structuring. Writers must have enough cognitive capacity in working memory at their disposal to be able to deal with'all the wvriting constraints of lexical, grammatical,! orthographical, and discourse decisions simultaneously. Automatic or fluent retrieval of lexical or grammatical chunks may contribute to an efficient use of the available, working memory capacity. 'Recently developed models of writing include w,orking memory as a critical component mediating the successful coordination of writing subprocesses (cf. Hayes, 1996; Kellogg,' 1996), and research has 'demonstrated a significant relationship between the avaiiability and efficient use of working memory capacity, on the one hand, and writing fluency, and (to a lesser extent) writing quality, on *theother (cf. Benton, Kraft, Glover, & Plake, 1984;.Bereitei & Scardamalia, 1987; Chenoweth &. Hayes, 2001; Fayol, 1999; Kellogg, 1999; Lea & Levy, 1999; ;Levy & Marek, .1999; McCutchen, 2000; McCutchen, Covill, Hoyne, .& MIildes, 1994; Ransdell & Levy, 1999;. Torrance & Jeffery, 1999). From this resea:rch it can indeed be inferred that it is n6t enough to have linguistic and metacognitive knowledge available while writing; writers must also be able to apply this knowledge efficiently and fluently. Fluent access to words and, phrases or grammatical structures in memory may lower the cognitive processing. load for a writer and may thus enhance the writing process and pos' sibly the quality of written text (cf. Chenoweth &'Hayes, 2001; Cumniiing,, 2001; Grabe & Kaplan;' 1996;. McCutchen,., 1996; Pehningroth & Rosenberg, 1995). See McCutchen (1996, 2000) for an extensive discussion of working memory and writing. However, although one may assume that fluency in lexical retrieval and in sentence building are.constituent subskills of writing at the process level-in addition to ,the linguistic and

170

Language Learning

'Vol. 53, No. 1

metacognitive knowledge, at-the resource and'control level-in (large-scale) correlational research such fluency measurements have' been largely ignored. Although the above 'sketch of writing proficiency is a simplification of a very complex construct (see, for more extensive analyses of writing and theories about writing, Grabe, 2001, and Grabe & Kaplan, 1996), it addresses three different' components of knowledge 'and skills of fundamental relevance for -writing proficiency: linguistic.knowledge, metacognitive knowledge; and fluent access to linguistic knowIedge F

Knbwledge Veisus Accessilbiity: Writing in First and Second
Language
3 t,

'When onetis.,writing in an LI, words and grammatical structures may be readily available in an automatized way, as they are in speaking. Consequently, speed of lexical and grammatical.retrieval may,not discriminate well between writers at higher levels of language proficiency. This',expectation is in-line with reading research findings that showed. low correlations between reading proficiency and word recognition speed at higher levels of reading proficiency (Stanovich, 1991). However, Benton et al. .(i984) showed that proficient and less proficient (Li) writers differ in their,"elementary information processing programs." Proficient writers can keep information in working memorywhile manipulating the content of text (e.g., reordering) far better than'weak writeis. Benton et al. suggest that proficient writers have automated certain components of'the writing process that less proficient writers have not. More recent research has also demonstrated a relationship between the efficient use of worliing memory resources (reading and writing span)' and writing fluency. The relationship between working memory resources and -writing quality, however, has been less easy to establish (Ransdell & Levy, 1999). In sum, proficient and less-proficient Li writers 'may differ.not only in their linguistic

Schoonen et al.,

'

171

I

;.

and metacognitive knowledge, but also in their, efficient use of working memory., In L21 writing, the.situation is likely to be different. Compared to,LI writers,.not only will L2 writers differ in their linguistic knowledge of the (second) language, but because of 'differences-in exposureto'the L2, they most likely will also differ in their degree of "fluency," that is, the ease with which they can access words and.grammatical structures during writing. Differences in degree of ifluency among L2 writers can also&be expected to be larger than the differences among LI Writers,- because of differences in L2 expdsure, L2 instruction, and language learn-. ing aptitude. Chenoweth and Hayes (2001).were able to deinonstrate that even two to,three semest6rs of,L2 instruction'made a difference in L2 writing fluency., Our'assumption is that'-difficulties.in fluent retrievai of words or,grammatical structures in L2 writing will burden the working memory -and thus hinder the writing process as sulch, not just with respect to writing fluency; but also with consequences for the quality of the text. Therefore, we expect that the co'ntribution of these kind of fluency variables to overall writing proficiency will be larger for.L2 writing.'than foir Li wvriting. The L2 writer may be so much,involVed in these kind of "lower-order' problems of word finding and grammatical striictures that, they .may require too much conscious attention, leaving little or no working memory capacity free! to' attend to higher-level' or strhtegic aspects of wvriting, such a's 'origanizing the text properiy or tryink to convince the reader of th5' validity. of a certain view. The'discourse and metacognitive knowledge that L2 writers are able to exploit in 'their L1 writing may remain unused, or.underused, 'in their L2 writing.2 'Whalen and M6nard (1995) reached similar 'conclhsioins;based on 'their, think-aloud, study with Canadian students learning French as anL2. 'They found that,.in comparison'to their L1.writing, in their L2 writing the stuidents planned and evaluated relafively more frequently at the linguistic level compared to -the textual and pragmatic level.' Revisions were made most frequently at the linguistic level, 'both in Li and in L2, but -these linguistic

172 17anug

Language LearnIing' Lernn

Vol. 53, N6. 1 Vo.5,N.1

revisions were at a deeper level (phrase and sentence) in Li than in L2, in which most revisions concerned morphemes and orthography. Little research has been done on the role of lower-level L2 processing skills or speed of. processing and L2 writing performance. The relevance of lower-level processing skills as predictors of L2 proficiency. (including writing) was demonstrated by Sparks et al. (1997). This study showed that a measure of phonological-orthographical decoding (i.e:; foreign-language word decoding) is highly relevant in the prediction of L2 proficiency. In a think-aloud study, Jones and Tetroe (1987) studied the interaction between'composing skills and (second) language competence in 6 native Spanish adult English as a Second Language (ESL) writers who wrote essays in both their LI and their L2 (English). The researchers also studied, among other things, to what extent the writers kept to their plans and the constraints of the tasks, and it turned out that in' ESL writing the "success xate" was much lower than in Li writing, which led the researchers to conclude, that not .only were these (elicited) plans in ESL not as fully developed as in Li, but writers also had more difficulty in keeping track of their plan. Jones and Tetroe conclude from their think-aloud protocols ,,that there is some decrease in performance simply due to the fact that it is in a second language, that workilig im an unfamiliar language does take' up cognitive capacity that would be' used for 6ther tasks, such as monitoring and revising the plan, in first-language composing. (p. 53) Sasaki and Hirose (1996) studied Li and English as,a.Foreign Language (EFL) writing performance of Japanese students and tried to model the interplay between EFL proficiency, Li writing ability, and strategic knowledge and writing experience/education. It turned out that the students' foreign language proficiency was the major predictor of their EFL writing performance (i.e.j the score on a single writing assignment), with minor roles for

Schoonen et al.

173

"metakAowledge" and Li writing performanice. These predictors of EFL writing explained 54.5% of the total amount of variance in EFL writing performance; unique contributions of EFL proficiency, metaknowledge, and L1 writing performance accounted for 32.6%, 0.3%,, and 1.5%, respectively,; of the variance. Furthermore, Sasaki and Hirose found that good (EFL) writers paid more attention to the overall organization of the text. They also seemed to be more fluent, that is, they wrote'longer texts, but there were no differences'in reported pausing. As an explanation for the relatively small contribution. of Li writing to EFL writing performance, Sasaki and Hirose suggested that weak writers may be "tied up with word- or sehtence-level processing (in other words, the 'what next strategy' cited in Bereiter & Scardamalia, 1987), and could not afford to think about overall organization much" (p. 158). Difficulties with the lower-order processing of text may account for the low correlation between *students''EFL and Li writing scores in the Sasaki and Hirose study,, as lower-order skills are more language specific than metacognitive knowledge of texts and writing strategies, w,hich may be applicable across languages. If metacognitive knowledge were a major discriminating factor in EFL writing, as it often seems to be in LI writing, then higher'correlatidns between Li and EFL writing could be expected. Consequently,' it is of interest to study further'with new data the interrelations betw'een Li and EFL writing proficiency with the effects of lower-order EFL linguistic knowledge and skills partialed out. ,Suggestions about the interrelations';between, lower-order linguistic knowledge and fluency and higher-order textual and pragmatic skills in L2 writing (Jones & Tetroe, 1987; Sasaki & Hirose, 1996; Whalen & Menard, 1995) are in accordance with recent findings of Chenowefth and Hayes (2001). Chenoweth and Hayes showed in a timed think-aloud'study that writing performance in Li and foreign language (FL) differs remarkably not onlyin fluency (i.e., in words per minute), but also in the length of bursts of text production (i.e., in the number of proposed new words between two pauses) and in the number of acceptable

174

Language Learning

Vol. 53,'No. 1

word retrievals (i.e., in the percentage of proposed new, words that eventually show up in the written text). ,These Li-FL differences seem to be mediated by amount of experience in the FL, as students with three semesters of FL,instruction turned out to be less fluent than students with five or six semesters of FL instruction. According to the authors, the "increase in blurst size reflects an increase in the capacity of the translator to handle complex language str:uctures" (p. 94). ~~~~~~~~~~~~ I Research Questions; Within the context of possible differences in the roles of different knowledge and processing resources in Li and L2 writing, the present study aims to unravel to some extent the contribution of the different comp'onent skills to the overall quality of-the written text, be it in LI or L2. The-primary focus is on the hypothesized importance of fluency in Li and'L2 writing (questions-1, 2, and 3' below). Furihermore, the relationship between Ll and -L2 writing proficiency (question 4) will be. explored.' The following questioins guided 6 ur analyses: 1. WVhat is the relationship,between speed of language *processing (i.e., lexical retrieval and sentence building) and writing proficiency in Dutch as an,LI and in English as an L2? 2. Is this relationship betweeii speed of language processing skiils and writing proficiency the same for Li and L2? 3. Howkr does the contribution of speed of language, processing to Li and L2 writing proficiency compare to-the contribution made by Li and L2 linguisti6 knowledge and metacognitive'knowledge of writing and written texts? 4. Does Li writing proficiency contribute to L2 writing proficiency beyond the,contribution of L2'1inguistic knowledge, processing speed, and metacognitive knowledge- of writing and written texts?

al

,

'Schoonen et al.

:.

175.

'We expect that, in L2 writing; writers will experience problems with both linguistic knowledge (vocabulary, grammar, and orthography) and fluency or accessibility of linguistic knowledge (lexical retrievaI and sentence building). -Individual differences in'these component skills should show up in individual differences in, the L2 writing, performance.. Differences in metacognitive knowledge are expected, to be less important for L2 writing performance than for L1 .writing perfo'rmmnce. In Li writing, we expect' a somewhat different picture': Differences in metacognitive and linguistic knowledge are expected to' constitute the main determinants'of Li writing performance, *heieas the role of access'ibility of linguistic knowledge will be less decisive.
I

|
'.

!

%

~- ~
i. ,

* Method , i>1,
' .' ,'

~

~ ~ ~ ~ ~ ~ ~ ~ ~~~
&,

I

1,

;"

'

',

'

i

Participants: The data from this study were collected from' 281 grade 8 students,(i.e., second year of secondary.education) in the Netherlands. These students are a subsample from, the 397 students participating in a longitudinal study of the development of reading and writing proficiency in Duitch 'as an Li or'L2,and in English as an FL (project NELSON).3 We have confined ourselves to students who participated in all tests relevant for the following analyses and who identified themselves as speakers of Dutch as 'an Li (N=281), that is, students who reported 'both speaking Dutchwith their parents anid having acquired Dutch as their Li. The students were sampled from both higher and lower streams of education, ranging from vocational to preacademic, in eight secondary schools in urban areas. These 'students had received' on average 3.5 years of education' in EFL: 1.5 years'at` the lev'el of'secondary' education and 2 years at the level'of primary 'education. In the primary'education years, attention' is paid during instruction (about 1 hr per week) only to basic oral communication, skills.

176

Language Learning

Vol. 53, No. 1

Instruments The testing instruments used in the study will be described briefly.1The Appendix provides sample items. Writing proficiency. Students wrote three texts in each language. As -far as possible, writing prompts were matched across the two languages. The prompts specified the rhetorical situation and intended audience. Texts were rated by six raters working in panels of two. Every panel rated a Dutch text and the "matched" English assignment. The raters rated according to a primary trait instruction and used benchmark texts as a reference. The five benchmarks represented a very weak (10th percentile), a'weak (25th), an average (50th), a good (75th), and a very good (90th) text. The benchmarks were selected and scaled in a separate analysis, according to a procedure described by Blok (1986). Panel reliability (k = 2, Cronbach's alpha) ranged from .76 to .82 for Dutch and from .87 to .90 for English compositions. The following (abbreviated) prompts were given for the
Dutch compositions:,
' , I

D1 Write to an emigrated friend what you plan to do when he visits yoA and stays at your place. D2 -Write to a television -network to complain about their pulling the plug on your favorite soap. D3 Describe your Dutch language 'lessons for a 'school magazine. The following (abbreviated) prompts were given for the English
compositions:'

El Write to an English pen pal who is moving to your hometown.' E2 Write to an English music magazine to complain about their ignoring your favorite group. E3 Describe your English language lessons for an English school magazine. I I

Schoonen et al.

,177 177

Vocabulary knowledge. Vocabulary tests comprised 75 (Dutch) or 65 (English) multiple-choice items. Only -nouns, verbs, adjectives, and adverbs were included in the tests. In the butch test we followed the format of Hazenberg and Hulstijn (1996): Each item consists of a neutral carrier sentence'with a target word in boldfac'e type, and the students must choose among four synonyms or paraphrases, of the' stimulus word. In the English test there were four Dutch translations, of the stimulus word (see Appendix). The Dutch word selection -was based, on Hazenberg and Hulstijn (1996), which spans the 20,000 most frequent wbrds of Dutch. The English word selection was based on' four frequency bands of the COBUILD corpus: the first 2,000 words; 2,000-3,000, 3,000-4,000, and 4,000-5,000. Because our students, unlike native speakers, are primarily exposed to English in a classroom setting, we also based our target word selection on an analysis of EFL textbooks. We avoided Dutch-English cognates as target words. The words used for the tests were not!specificqlly related to the topics of the writing -assignments. Orthographic knowledge. In the orthographic knowledge tests, students had to choose, between two or three concurrent options for the spelling of the missing part of a word. The target words were presented-in a sentence that provided'.a strong cue to the intended word. The Dutch test consisted of 100 items covering well-known spelling problems (single or double consonants, homophones, etc.),;including the spelling oftconjugated verbs, which is notoriously problematic for students. The English test consisted of 89 items covering typical spelling problems in English (Castley, 1998). Some items had a multiple-choice format, requiring students to choose either of two options (e.g., single or double consonant), and other items -w'ere of the fill-in-the-blank type, requiring students to decide which consonant(s) or vowel(s) had to be inserted. To ensure that students knew which word they needed to spell in the English items of the latter'type, the Dutch translation of the intendedwords was also provided. For example: , ' ' I_e is very cold ("IJs").
I

178

Language Learning
I~~~~~~~~~~~~~~~

Vol. 53, No. 1

Grammatical knowledge. In the grammatical knowledge test, students had to fill in the correct forms of verbs, adjectives; anaphora, comparatives, pronouns, and articles or had to put randomly presented words or phrases into the right order. For example, students had to choose between contrasted elements like any(thing).and some(thing), many and much, or,little and' few. The word-order itemns required students to take giammatical number, tense,, aspect, position of adverbs, and agreement into account, as in, for example, [He], [take]2 [the bus]3 [not]4 [does]5 ; "The best order is: 1, .. " The Dutch test consisted of 69,items. The English test consisted of 80 items. Metacognitive knowledge. For the measurement of metacognitive knowledge, a questionnaire consisting of 80 statements about text characteristics and reading and writing strategies was used. Students had to check whether they, agreed or disagreed with the. statements. All statements were in Dutch. Statements concerned knowledge of texts, and knowledge of reading and writing strategies. 'Some of the statements were general, and some applied specifically to xeading and writing in EFL. Note that, whereas all other constructs were measured for Dutch and English separately, metacognitive knowledge.was assessed with-a single test. Speed of lexical retrieval. In the test for speed of lexical retrieval, students were presented with pictures on a.laptop screen and weie requested to type the first letter of the noun referring to the person or object depicted. Responses were coded for 'accuracy and reaction times (RTs). RTs faster than 550ms or slower than three standard deviations of the mean RT on' a particular stimulus were considered invalid and were replaced (see "Scoring"). RTs were corrected for "typing fluency," that is, the speed with which, a particular student is able to find a certain letter on the keyboard. This typing fluency was assessed separately, and students' scores for typing fluency (i.e., mean RTs), were partialed out of their mean RTs in speed of lexical retrieval. The testing format was the same for Dutch and English. Each test started with 10 trial items. Only items with

Schoonen et al.

179.

I

high hit rates (see "Scoring") were 'used to compute the test score (i.e., mean.RT). Foi Dutch, 38 out,of 39 items had high enough hit.rates' to be used.in scoring tests; for English, 18 out of 38 items met this criterion. Speed of sentencAe building.. In the test for.. speed of sentence building, students were presented with the begiinning of a sentence on a laptop screen. Then they had' to choose as quickly as possible, by' pressing a' corresponding key, which -of two constituents presented to them should continue the sentence. Each test started with 10 (final items; 'RTs faster than 050ms or slower than 'three standard. deviations of.the nmeah' RT for the. item were considered invalid and. -were replaced (see' '"Scoring'?). Only items with high hit rates (see, "Scoring") were .used to compute the test score. (i,e.,, mean RT). Thirty-two out of 43 Dutch items and 24 out of'44 English items.',met this criterion.

Procedure

'

.

Data were collected in 1999qas.the first phase of:a longtudinal study. The tests were administered in a classroom setting by trained test assistants during school hours. All the tests were administered within a period of about 10 weeks, ranging from 8.5 weeks to 11.3 weeks between the first and-the last'test' Tests were assigned in 'a quasi-.random order, taking into account the availability. of the laptop computers needed for"the : speed tests and the timetables of the schools.'.

Scoring

''

Skipped items from' the knowledge tests were scored as incorrect. Missing more than half of the number of items on a participant test resultedin the student's having.a missing value onthi6test as awhole. . ' . '. .

180

Language Learning

Vol. 53, No. 1

In speed -tests, two scores are usually relevant: accuracy and RT. Because we were solely interested in the speed of processing and wanted to avoid the interference of linguistic know: ledge, we used RTs only of responses to items with sufficiently high hit rates (i.e., the percentage of students providing a correct response to a word). This allowed us to assume that the right answers were known to (almost) all students. For lexical retrieval items, for which there is only a negligible chance of guessing the right answer, .the hit rate on a particular item had to be .75 or higher' (i.e., 75% or more studeAts had to have recognized the word as a word) for the item to be included in scoring. For sentence building items (with a 50% chance level for guessing), the hit rate had to be .875 (=..75 + [.25 x .501) or higher. Responses to all other items were removed. RTs of (incidental) wrong answers and missing RTs were replaced by estimates according to algorithm an expectation maximization (cf. Acock, 1997; Hox, 1999). The EM estimates involved 9.4% (Dutch) and 17.0% (English) of the lexical retrieval scores and 6.7% (Dutch) and 7.5% (English) of the sentence building scores. Missing test scores (due, for example, to absence of students on one of the testing. days) were also estimated according to the EM procedure. In our data set,4 12.1% of test scores were missing. Estimating these test scores had hardly any effect on the test means or standard deviations. In terms of Cohen's (1988) effect size for means, the largest difference (.06) is still small according to C6hen's rule of thumb that effects of .10 are considered to be small effects. Analyses Means and standard deviations were computed for all tests. All tests, except for writing proficiency, were split into two parallel parts according to the Gulliksen (1950) procedure. For writing proficiency, the three scores (one for each assignment) were kept separate. Using the "observed" scores per variable,

Schoonien et al.

181

latent variables were estimated in order to perform structural equation modeling (SEM) in LISREL (Joreskog & Sorbom,, 1996a; Raykov & Marcoulides, 2000). According'to the procedure described in Fleishman and Benson (1987), reliabilities of the sums of the two test halves and the three writing assignments were estimated within the SEM analyses. Variables were normalized in PRELIS (J6reskog & Sorbom, 1996b), and the covariance matrix of the observed (parts of the) tests was computed as input for SEM. We will report on the bivariate correlations between, the latent component variables and the corresponding, dependent (latent) variable Writing Proficiency in Li and L2. These correlations are often referred to as "true" correlations, because measurement error is partialed out. These correlations inMdicate the relationship between individual 'component variables and Writing Proficiency. Furtherniore, we will fit a model in-which all component variables, are simultaneously relatdd to WVriting Proficiency. Such a "regression analysis" will show the relative importance of component variables in th,e contexct of the other component variables. Regression weights of the component variables will be compared across Li and L2 writing.
* .

Results Descriptive and Psychometric Statistics Table 1 presents the descriptive and psychometric statistics per test, for Dutch Li and L2. Table 1 shows that the language proficiency tests are, of average difficulty and show a fair amount of variance in scores. On the basis of the difference in,performance on the speed measures ,in' Li and L2, one might be inclined to conclude that students are less "fluent" in L2 than in Li. However, one should bear in'miiid that no attempt was made to equate the Dutch and English stimuli in terms of difficulty or accessibility. In general, it seems that students differ

182

Language Learning

Vol. 53, No. 1

Table 1
Descriptivesfor the Li Dutch and L2 English tests: Mean, standard deviation, reliability, and number of items
Means (standard deviation)' Dutch Writing Proficiency (scale mean is 300) Vocabulary. Knowledge Grammatical Knowledge, Orthographic Knowledge , Metacognitive Knowledge (not language specific) Speed of Lexical Retrievala (RTs in, ms) Speed of Sentence Building (RTs in ms) 298.8 (19.5) "English 295.7, (24.7) Reliability '(number of items) Dutch English

.

.65 .80 (3 assignments) (3 assignments)
. a:

1.0 (9.7)

51.8 (9:8).

.89 ,(75) .88 (69) ' . .89 . (100)

,

.90 (65) .95 . (80). I . .,.93 (89)

-,

.49.6 ' . 43.7' (8.7) . (16.2) 70.8 .,. (11.2) ,' i
l

48.8 (13.1)

61.6 I I (7.3)

1 '

:81' (80) .

1826 '(287) 1938

2205 (425) 2344

.77 (38) .95 .

.74 (18) .93

(318)' (464) ' (32) (24) Note. N=281. Maximum possible score equals the number of items, except for Writing Proficiency and the speed measures. The use of different scales prevents comparison of the means across the.languages. Reliabilities are estimated under Model 1 (see Table 2 and text). Some of the reliabilities are slightly different from the ones reported in Schoonen et al. (2002), because those reliabilities were estimateId under a different'model. But generally speaking, the estimates are quitl stable. aln the subsequent SEM analyses, the,Speed of Lexical Retrieval response times were corrected for typing fluency; typing fluency was partialed, and the residuals were used as indicators of Spee'd of Lexical Retrieval. The mean residual is (by definition) zero.
:, . . , I

~Ip

.

..

Schoonen et al. `83

.183

I

more in their L2 skills and proficiencies than in their LI skills and proficiencies, as can be inferred from the standard ,deviation-mean ratios.; The right'panel, df the table shows that all. tests have a satisfactory reliability. Both measures of Speed of Lexical Retrieval have reliability scores just below .80, perhaps'because these measures are residuals (i.e., RTs-with the RTs for typing fluency partialed out). Dutch Writing Proficiency is clearly below .80 in'reliabilty, which is not that surprising given the known effects of topic and assignment in writing assessment (Cooper, 1984; Schoonen & de Glopper, 1999). These effects make it difficult to reach a satisfactory level of`score reliability. Howkrever, in the subsequent SEM 1 analyses we will be dealing -with the 'latent variables from whi6h measurement error is partialed out, and, in that sense, measurement error should not influence'the size of correlations or regresIsion weights. Nevertheless, one should be a bit' cautious in interpreting the dependem,t variable Dutch Writing . ' . ' Proficiency. I
I~~~~~~~~~~~~~~~~~~~~~~~I

A Baisic Model of Ll and L2 Writing .
, , ,. i

, .;

.i

In this section we describe the selection of a`satisfactory model ofL1 and L2 writing. Model fit is evaluated by chi-square (x,) and the corresponding degrees'offreedom (df). However, it is well-known that this statistical measure,is very sensitive to sample size (Raykov & Marcoulides, 2000),. in the sense that uminor deviations from the model lead to a significant chi-square and thus to rejection' of the model.' Therefore,'more' descriptive measures-of fit should be reported as well, for example; the' x 2/df ratio, which preferably is lower than' 2; the root' mean square error of approximation (RMSEA), which preferably is lower than .05; and the non-7normed fit indei (NNFI?, which should be in the upper :90s (Raykov',& Marcoulides, 2000)., For the expeczted cross-validation index (ECVI),, there is no absolute criterion, but it can be'used to, compare different models for the same ;, data;,.the lower,.the ECVI the better.

Schoonen et al.

191 '191

WritingProficiency (mainly'.at the cost of the predictive value of . Metacognitive Knowledge). Discussion To summarize our findings, we return to our.four r'esearch questions. The first question concerned the relationship between speed of language processing, (i.e., lexical retrieval and sentence building) and writing proficiency in.Dutch as Li and English as L2. Table 3 shows that the speed. measures are correlated with Writing Proficiency: -. 10 and -. 48, respectiVely, in Dutch and -. 38 and ;-.60,.respectively, in English. These correlations (except for Dutch Lexical Retrieval) can be considered substantial, but we should also bear in mind that these are "true" correlations, which are not attenuated by measurement error. The correlations, between the speed measures and Writing Proficiency are generally. lower than the correlations between the.knowledge measures and Writing Proficiency. The correlations between the speed measures and CWriting Proficiency are higher in English than they aie "in Dutch (research question 2). Thus speed of access to linguistic knowledge may be more important in the L2 than in the Li. In subsequent analyses we showed that the regression: of English .Writing Proficiency 6n the component skills differs statistically from the corresponding 'regression of Dutch Writing Proficiency onits comnponent skills. However, the major differences in the regression equations jertain to the knowledge tests. This finding need not.-surprise us if we consider the third question, that is, the unique. contribution of the speed measures in the context of the linguistic and mefacognitive knowledge tests. It turned out that the speed.measures made virtu6ally no unique contribution to the "prediction" of.writing'proficiency, either in L2 or in L1, and in this respect,the'firdings for L2 and'Li are similai. Finally, we improved our. basic model by regrepsing'L2 Writing Proficiency not only on its language-specific component skills and general Metacognitive Knowledge, bui also on Li

192

Language Learning

Vol. 53, No. 1

Writing Proficiency. Students acquired their Li writing proficiency prior to their L2 writing proficiency, and therefore, it may predict L2 writing proficiency. It turned out that the model improved and that Li Writing Proficiency indeed contributes to the prediction of L2 Writing Proficiency, as could be expected given the high correlation between the two variables. The additional value of Li Writing Proficiency also implies that the correlation between the writing measures is not fully captured by the Metacognitive Knowledge measure (cf: Schoonen et al., 1998, for a similar pattern of results in the case of L2 reading). Apparently there are typical aspects of writing proficiency that are captured in both writing assessments (Li and L2), but not in our component variables. This finding underscores the idea that our models for writing, despite the number of variables involved, are not all-encompassing (as this was not dur primary goal). More general variables might account for the high correlation between LI and L2 writing (see also below), such as general "world knowledge," certain writing strategies, or general cognitive fluency as indicated by, for example, working memory span (see the article's introductory section). However, we do not have separate measures for such variables and thus can only speculate about the nature of these other variables that are common to Li and L2 writing. It would also be interesting to explore further the Telation between metacognitive knowledge and writing proficiency. Our operationalization of metacognitive knowledge; though based on eailier research, was, for practical reasons, rather straightforward. knowledge. However, the construct of metacognitive knowledge could be extended to the actual writing strategies the writers use (cf. Whalen & Menard, 1995) or consider to be most important in writing (cf. Schoonen & de Glopper, 1996), in either Li or L2 writing. Surprisingly, vocabulary knowledge makes little contribution in L2 and an unexpected negative contribution in Li. The regression weight of the latter must be an artifact of the procedure; we are dealing with a so-called suppress6r effect (Tabachnick & Fidell, 1996, p. 165). This suppressor effect is

Schoonen et al.

-193

likely to 'be due to the high intercorrelation between Metacognitive Knowledge and Vocabulary Knbwledge and the lower correlation between Vocabulary Knowledge and WVriting Proficiency. Although, it may not be readily clear from the regression weights, it seems that L2 writing is more dependent on L2 linguistic knowledge' and processing speed than is Li writing. Li writing is harder to explain in terms of linguistic variables, and this may imply' that in Li writing, other factors (which we did not measure) such as topic'knowledge may play a role. In c6ntrast, in L2 writing, linguistic variables seem to play a major role in writing, suggesting that students are more concerned with tackling language problems when writing in the L2 than when writing in the Li. In Li writing, metacognitive knowledge plays a more important role. When we' consider possible transfer from Ll (highei-order) writing skills to L2 writing or the existence of a threshold either in terms, of linguistic knowledge or'speed of language processing, asposited by Sasaki and Hirose (1996), which inhibits the transfer of higher-order skills, our correlatioiial data can give only circumstantial evidence. As mentioned above, project NELSON. is part of a longitudinal study, and so far we have data from just a single round of measurements. The important question of how, in each language (i.e., Li and L2) the relationship develops between, on the one hand, language proficiency and language processing skills, and, on the other, writing proficiency will remain unanswered until the longitudinal data have been analyzed. The same applies',to the development of 'the relationship 'between Li and L2 writing proficiency. At this stage, our conclusions cin be no more than preliminary. Bearing these lirmiitations in mind, we see that 'lexical retrieval correlates far less with writing proficiency in Li than in L2, which is treconcilable with the notion of a linguistic (speed) threshold. It seems plausible that 14-year-old students writing in their mother tongue have already surpassed the level beyond which differences in speed matter in writing. It may also

194

Language Learning

Vol. 53, N6',l

be the case that students have reached a ceiling in Ll lexical retrieval. Whether this is actually the case should become clear when we have analyzed more,longitudinal data in the coming years. With respect to the relation between Li and L2' writing proficiency, our results differ remarkably from those reported by Sasaki and Hirose (1996). In our case, metacognitive knowledge and LI writing do make a (unique) contribution to the prediction of the L2 writing scores in Model 1 and Model 5, respectively 1(see Table, 4; in Model. 5, the contribution of metacognitive knowledge to L2 writing proficiency is completely suppressed by Li writing proficiency). At least three explanations come to.mind for this difference in outcomes. First of all, this difference may be explained by differences between the two studies in assessment and analyses. Sasaki and Hirose used one writing assignment per language.. Scores on single writing assignments, usually show low intercorrelations and are therefore considered of low reliability. Our approach used three writing assignments per language, and in our analyses the common variance of those three assignments is-considered the "true score" variance. This means that task-specific variance, such as topic-related variance, is "filtered out." Estimates of correlations are not atteiiuated by error (or task-specific) variance. Furthermore, we tried to match writing tasks in Li and L2 as well as possible, because writing performance is very much dependent on the features of the writing assignment. Another (methodological) difference is that our sample of participants is relatively heterogenous. We did not aim at a nationally representative sample but tried to cover the range of language proficiency found in grade 8 of Dutch secondary education. Colisequently, we can expect a great deal of variance in LI.and L2 writing proficiency, and in such a context correlations are more likely to show up than in contexts with a restriction of;range. To'what extent prior (Li and L2) language instruction influenced the writing proficiencies of our students we do not know, but this information may not be very relevant in view of.our goals.

Schoonen et al.

195 9

Our primary goal was to ;find out what constitutes proficient'Ll and/or L2 writing, Tather than to determine what.kind -of language instruction.mrost' enhances writingi proficiency. The second, more theory-based explanration for the, differences between the findings of Sasaki,and,Hirose and'our findings w,ould be that most students in our sample already have surpassed the threshold level (in both.linguistic knowledge and fluency) that inhibits a. substantial correlation between LI and 12 writing! whereas Sasaki and Hirose's students may have not surpassedithis level. It is well known that Japanese learners of English continue,for a .'long time during their L2 learning careers to process written English words primarily. as pictures,' because of the fuindamental differencei betweenf the nonphonological :nature of their'LI writing system in comparison to the phonological nature of' the alphabetic system of -English (Koda, 1999). 1This imeans that' Japanese students may profit much less from phonological cues that facilitate access 'to written word forms in the mental lexicon than do'L2 learners with an alphabetical Li. If this is the case, the students in our study may be,less occupied with the spelling.of the English.words than the'Japan6se students in Sasaki and Hirose's study, thus saving more cognitive' capacity, to use' their .metacognitive knowledge in L2 -writing. Nonetheless, it seeriis very unlikely that English word prodiuction. and sentence construction pr6cesses .are already fully automated ',in- the 'Dutch students aft6r such a relatively shortbperiod 'of not very intensive instruction. An indication of the Dutch students' not being fully automated in English lexical retrieval and sentence building is that they -were less successful and slower. in retrieval of English lexical'items'and English sentence building than,in simila'r tasks i a . in Dutch. The third possible explanation is that the threshold hypothesis may be invalid or applies to L2' writing to. a-far lesser extent than to:L2 reading. Although reading and writing.ma~y partly use the same knowledge resources, the nature of the cognitive processes involved and the flow. of inforniation may be, quite different'. Readers have little control over the linguistic and,
i

196

Language Learning

Vol. 53, No. 1

cognitive difficulty of the texts to which they are exposed, whereas writers are likely to 'avoid writing texts of linguistic and cognitive complexity. beyond their knowledge and skills. Lack of (some) language iknowledge and slow processing in L2, then, does not necessa'rily prevent successful application of higher-order strategies in L2 writing. It is conceivable that students profit from their metacognitive knowledge in spite of very limited linguistic resources, thereby even compensating for some of these limitations. Which one of these (latter two) explanations is more acceptable will be one of the main focuses of our research project in the near future. In order to be able to choose the more acceptable explanation, we will need to gather more evidence on the role of linguistic and metacognitive knowledge in Li and L2 composition. As our data are correlational, strictly speaking, 'causal interpretations are not warranted. We cannot infer, for example, that increased speed of L2 sentence building for our students would actually help them to become better L2 writers. The same holds, of course, for the role of metacognitive knowledge. It cannot be concluded from the.above results that improving metacognitive knowledge about texts and writing strategies helps students to become better writers. Our longitudinal study may show that when' students acquire more speed in language processing, the-importance of this variable in explaining L2 writing proficiency will diminish. The pattern of regression weights of L2 writing proficiency, and its constituent variables may then become more similar to that of Li writing proficiency, with a bigger role for metacognitive knowledge. Such findings would point to a kind of (fluency) threshold. Students might have experienced problems in applying their metacognitive knowledge in L2 writing because of (among other things) slow language processing. Additional data might corroborate this hypothesis. Another way to get more insight regarding this important issue is to take a closer, look at cognitive processes in Li and L2 writing. Stevenson (2003) is conducting a qualitative study .in which think-aloud data are combined with RT data.

Schoonen et al.

197

Of couise, to conduct an experimental intervention, study to test the effect of increased speed of lexical and grammatical fluency on L2 writing is yet another way to approach this issue. Such a study has been conducted by Snellings and his colleagues, and preliminary findings show that lexical retrieval in writing can be improved through training .(Snellings, van Gelderen, & de Glopper, 2002). Whether this will improve the quality' of the written texts is currently being iiivestigated (Snellings, van Gelderen, de Glopper, 2003). Revised version accepted 28 July 2002 Notes -We use the term second language in a general sense, referring both to situations in which the L2 is the language of the majority or of'the community in which the language user lives, and to situations in which the L2 is not the community language but is a so-called foreign language. In the present article, our participants' L2 is English,, which is a foreign language in their linguistic context. 2A similar line of reasoning is used in the threshold hypothesis, in which it is stated that lack of linguistic knowMldge of the second or foreign language (instead of lack of fluency) hinders application of strategic knowledge in'FL reading: Poor foreign language reading is due to reading strategies .in the first language not being employed in the foreign language, due to inadequate knowledge of the foreign language. Good first-ianguage readers will read well in the foreign language once they have passed a threshold of foreign language ability. (Alderson, 1984, p. 4 ) However, one could 'aiso argue that L2 readers experiencing language difficulties tiy' to compensate for this lack of language comprehension by using certain content-oriented' reading strategies., See, for an extensive discussion, StevensoA, Schloonen, and de Glopper (2002). 3General information about project NELSON is available at http:// www.sco-kohnstamminstituut.uva.nllnelson/index.htm. ' 4Our data set also included reading scores. These reading data were included in the missing data estimation procedure in order to 'ptimize the accuracy of the estimates. 5 The,amount of unexplained variance is estimated to be -5%;'however, zero was in its 68% confidence interval, and therefore the amount of unexplained variance was considered to be zero, which is the theoretical minimum.

198

Language Learning

Vol..53,No. 1

References
Abbott,-R. D., &.Berninger, V. W. (1993). Structural equation modeling of relationships among developmental skills and writing skills in primaryand intermediate-grade writers. Jouihal of Educational Psychology, 85(3); 478-508.

Acock, A. C. (1997). Working with missing values. Family Science Review,
10(1), 76-102:

Alderson, J. C. (1984). Reading inj a foreign 'language: A reading or la language problem? In J. C. Alderson & A. H. Urquhart (Eds.), Reading
in a foreign language (pp. 1-24). London: Longman.

Benton; S. L., Kraft, R., G., Gl6ver, J. A., & Plake, B. S. (1984). Cognitive
capacity differences among writers. JournalQf EducationalPsychology,

76(5), 820-834.

'

Bereiter, C., & Scardamalia, M. (1987). The psychology of written comp6-

sition. Hillsdale, NJ: Lawrence Erlbaum. Blok, H. (1986). Essay rating by the comparison method. Tijdschrift voor
Onderwijsresearch [Journalfor EducationalResearch], 11, 168-176.

Castley, A. (1998). Practicalspelling. New York: Learning Express. Cheno~weth, N.-A., & Hayes, J. R. (2001). Fluency in writing. Generating
text in Li and L2. Written Communication; 18(1), 80-98. Cohen, J. (1988). Statistical power analysis'for the behavioral sciences

(2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.
Cooper, P. L. (1984). The assessment of writinig ability: A review of research

(GRE Board Research Report No. GREB 82-15RIETS Research Report No. 84-12). Princeton, NJ: Educational Testing Serivce. Cumming, A. (2001). Learning to write in a second language: Two decades
of research. InternationalJournalof English Studies, 1(2), 1-23.

Engber, C. A. (1995). The relationship of lexical proficiency to the quality of
ESL compositions.,Journal of Second Language Writing, 4(2), 139-155.

Fayol, M. (1999). From on-line management problems to strategies in written composition. In M. Torrance & G. Jeffery (Eds.), The cognitive
demands of writing. Processing capacity and working memory in text

,production (pp. .13-23). Amsterdam: Amsterdam University Press.' Fleishman, J., & Benson, J. (1987). Using LISREL to evaluate measurement model and scale reliablity. Educationaland PsychologicalMeasurement, ,47(4), 925-939. Flower, L. S., & Hayes', J. R. (1980). The dynamics.of composing: Making plans and juggling constraints. In L. W. Gregg & E. R. Steinberg (Eds.), Cognitive processes in writing (pp. 31-50). Hillsdale, NJ: Lawrence Erlbaum.

I

Schoonen et al.

.

199

Flower, L. S., & Hayes, J. R. (1983).A cognitie niodel of the writing process

in adults (Final Report): Pittsburgh,. PA: Carnegie. Mellon University. (ERIC Document Reproductio'n Service No. ED240608) Grabe, W: (2001). Notes towardsi a theory of -second language writing. In T. Silva & P. K. Matsuda.(Eds.), On second language writing (pp. 39-57).
. "' Mahwah, NJ: Lawrence Erlbaum. Theory anid pr`actice of writing. An Grabe, W., & Kaplan, R. B. (1996). applied linguisticperspective. London and New York: Longman.. Gulliksen, H. 01950). Theory of mental. tests. New York-: Wiley.

Haye, .J. R. (1996). A new framework for understanding cognition and affect in writing. In CU M. Levy & S. Ransdell (Eds.), The science of
writing: Theories,, methods, individual differences and applications

(pp. 1-,2 7). Mahwah, NJ: Lawrence.Erlbaum. minimal'receptive Hazenberg, S.,. & 'Hulstijn, J. H. (1996). Defining for non-native university students: An *'second-language vocabulary
.a

empirical investigation. Applied,Linguistics, 17(2), 145-163.

Hox, J. J. (1999). A review of.current software for handling missing data.
KwantitatieveMethoden [QuantativeMethods], 62, 123+138.

'Jones, S:, & ,Tetroe,' J. (1987). Composing. in' a second language. 'In
A. Matsuhashi (Ed.), Writing in real time. Modelling productionprocesses

(pp. 34-57). Norwood, NJ: Ablex. J6reskog, K', & Sorbom, D. (1996a). Chicago. Scientific Software International.

,S'-

User's reference guide. UISRELg8:

6Joreskog, K.? & Sorbom, D. (1996b). PRELIS 2: User's reference guide.

Chicago: Scientific Software International. Kellogg, .R. T. (1996). A model of worliing memory in writing. In C. M. Levy
& S. Ransdell (Eds.), The science of writing: Theories, methods, individual differences and applications (pp. 57-71). Mahwah, NJ' Law'rence
Erlbaum. ,
,,
,.

,

.

Kellogg,j. T. (1999). Components of working memory in text production. In
M. Torrance & G. C. Jeffery (Eds:), The cognitive demands of writing. Processingcapacity and working memory in text production (pp. 43-61).

' Amsterdam: Amsterdain University Press. Koda, K. (1999). Development of.L2 intraword orthographic sensitivity and
decoding skills. Modern.Language Journal, 83(1); 51-64.

Laufer, B.

,&Nation,

P. (1995). Vocabulary size-and use: Lexicalrkchness in

L2 written production. Applied Linguistics, 16(3), 307-322.

Lea, J., & Levy, C. M! (1999). Working.memory as a resource in the writing
process. In M. Torrance & G. Jeffery (Eds.), The cognitive demands of writing. Processing capacity and working ,memory in text production

(pp. 63- 82) Amsterdam: Amsterdam University Press.

.

200

Language Learning

Vol. 53, No. 1

Levy, C. M., & Marek, P. (1999). Testing components of Kellogg's multicomponent model of working memory in writing: The role of the phonological loop. In M. Torrance & G. Jeffery (Eds.), The cognitive demands
of writing. Processing capacity and working memory in text production

(pp. 25-41). Amsterdam: Amsterdam University Press. McCutchen, D. (1996). A capacity theory of writing: Working memory in
composition. EducationalPsychology Review, 8(3), 299-325.

McCutchen, D. (2000). Knowledge, processing, and working memory:
Implications.for a theory of writing. Educdtional Psychologist, 35(1),

13-23. McCutchen, D., Covill, A., Hoyne, S. H., & Mildes, K. (1994). Individual differences in writing: Implications of translating fluency. Journal of
EducationalPsychology, 86(2), 256-266.

Penningroth, S. L., & Rosenberg, S. (1995). Effects of a high information processing load on the writing process and the story written. Applied
Psycholinguistics, 16, 189-210.

Ransdell, S., &.Levy, C. M. (1999). Writing, reading, and speaking memory spans and the importance of resource flexibility. In M. Torrance &
G. C. Jeffery (Eds.), The cognitive demands of writing.Processingcapacity and working memory in text production (pp. 99-113). Amsterdam:

Amsterdam University Press.
Raykov, T., & Marcoulides, G. A. (2000). A first course in structuralequa-

tion modeling. Mahwah, NJ, and London: Lawrence Erlbaum. Sasaki, M., & Hirose, K. (1996). Explanatory variables for EFL students' expository writing. LanguageLearning,'46(1), 137-174. Schoonen, R', & de Glopper, K. (1996). Writing performance and knowledge *about writing. In G. Rijlaarsdam, H. van den Bergh & M. Couzijn (Eds.),
Theories, models and methodology in writing research (pp. 87-107).

Amsterdam: Amsterdam University Press. Schoonen, R., & de Glopper, K. (1999). The assessment of writing proficiency: Problems and possibilities. In W. Decoo, M. Simons, &
A. Stepanyan (Eds.),.L'enseignement curriculairedes langues etrangeres [Curriculum-bound foreign language education] .(pp.l91-2 12). Lier:

VanIn. I I Schoonen, R., Hulstijn, J., & Bossers, B. (1998). Language-dependent and language-independent knowledge in native and foreign language reading comprehension. An empirical study among Dutch students in grades
6, 8 and 10. Language Learning, 48(1), '71-106.

Schoonen, R., van Gelderen, A., de Glopper, K., Hulstijn, J., Snellings, P., Simis, A., & Stevenson, M. (2002). Linguistic knowledge, metacognitive knowledge and retrieval speed in LI, L2 -and EFL writing. A structural

Schoonen et al.

.

201

equation: modelling approach. In S. Ransdell & M.-L. Barbier (Eds.),
New directions for research in L2 writing (pp. 101-122). Dordrecht,

Netherlands: KMuwer Academic. Snellings, P., van Gelderen, A., & de Glopper, K. (2002). Lexical retrieval: An aspect of fluent second language production that can be enhanced.
Language Learning, 52(4), 723-754.

Snellings, P., van Gelderen, A., & de Glopper, K. (2003). The effect
of enhanced lexical retrieval on L2 writing. Manuscript submitted for

publication. i , Sparks, R. L., Ganschow, L., Patton, J., Artzer, M., Siebenhar, D., & Plageman, M. (1997). Prediction of foreign language proficiency. Journal
of EducationalPsychology, 89(3), 549-561:

Stanovich, K. E. (1991). Word recognition: Changing perspectives. In R. Barr, M.'L. Kamil, P. B. Mosenthal, & P. D. Peairson (Eds.), Han;dbook of readingresearch (Vol. 2, pp. 418-452). New York and London: Longman.
Stevenson, M. (2003). Writing strategies, writing fluency and text quality in Li (Dutch) and EFL. Manuscript in preparation.

Stevenson, .M., Schoonen, R., & de Glopper, K. (2002). Inhibition or compensdtion? A multi-dimensional comparison of reading processes in Dutch and English. Manuscript submitted for publication. ' Tabachnick, B. G., & Fidell, L. S. (1996). Using rmultivariate statistics

(3rd ed.). New York: HarperCollins College. Torrance, M., & Jeffery, G. (1999). Writing processes- and cognitive demands. In M. Torrance & G. Jeffery (Eds.), The cognitive demands of
writing. Processing capacity and working memory in text production

(pp. 1-11). Amsterdam: Amsterdam University Press. Victori, M. (1999). An analysis ofwriting knowledge in EFL composing: A case study of two effective and two less effective writers. System,' 27, 537-555. Whalen, K., & Menard, N. (1995). Li and L2 writers' strategic and linguistic knowledge: A model of multiple-level discourse processing. Language Learning, 44(3), 381-418.

Appendix: Sample Test Items
Vocabulary (English Test) '

1. That's important news. important is: [literal translation of the-Dutch alternatives] X a. belangrijk [significant] .b. interessant [interesting]

202

Language Learning

Vol. 53, No. 1

-c. opwindend
d. saai

[exciting]
l[boring]

OrthographicKnowledge (English Test) He likes e_s. (g)
f

gg

GrammaticalKnowledge (English Test)

She never,

_

(listen) to me.

1istens l

Metacognitive Knowledge (LiteralYTranslation From Dutch) 28. The order in which you present'the information in y6ur text is usually not relevant. agree ? don't agree 55. It is wise to keep the organization of the text in mind while
wrting. % agree

don't agree, Speed of Lexical Retrieval (RT) , ,

C [cheese]
Speed of Sentence Building (RT) We think that ... 1.is alright > 2. it
.~~~~~~~~~~ I

COPYRIGHT INFORMATION

TITLE: First Language and Second Language Writing: The Role of Linguistic Knowledge, Speed of Processing, and Metacognitive Knowledge SOURCE: Lang Learn 53 no1 Mr 2003 WN: 0306001954005 The magazine publisher is the copyright holder of this article and it is reproduced with permission. Further reproduction of this article in violation of the copyright is prohibited. To contact the publisher: http://www.blackwellpub.com/

Copyright 1982-2003 The H.W. Wilson Company.

All rights reserved.


相关文章:
Chapter11 Linguistics and Language Teaching
to take a more balanced view on the role of grammar in language learning...second language acquisition researchers because knowledge of linguistic universals...
The role of first language
“Learning is a cumulative process. The more knowledge and skills an ...So now I will talk about the role of first language in second language ...
母语的负迁移和中国式英语
(10) 1. Introduction The role of our mother tongue in second language ...linguistic knowledge or skill.[5] And the language transfer refers to the ...
Literature Review on the Output Hypothesis for SLA
role in second language acquisition because the ...linguistic knowledge which is new for them, or ...first language production to the target language ...
14 Characteristics of a Good Language Learner
Good language learners use linguistic knowledge, ... first language, in learning a second language. ...writing to learn to vary their language according...
语言学简答题
internalized linguistic knowledge of a language ...They may learn a second language in order to ...When a pidgin begins to play the role of a ...
chapter11 语法
view on the role of grammar in language learning...–Linguistic knowledge helps teachers to better ...?In the case of foreign/second language learning...
论文目录2--Second Language Research
acquisition of second language knowledge Susanne E....first language and second language lexicon: a ... and Shoba Bandi-Rao Linguistic structure with ...
First Language Acquisition
linguistic environment that provides sufficient ...knowledge of a second language usually obtained in...the role of first language in second language ...
...Metacognitive Growth of ESL Student Writers
Possessing a strong metacognitive knowledge base is...second language and first language basic writers ...role than linguistic competence in successful second...
更多相关标签: