Direct vs.

Translated Writing: What Students Do and the Strategies They Use1

Andrew D. Cohen

Department of English as a Second Language

University of Minnesota{ HYPERLINK mailto:adcohen@tc.umn.edu}

Amanda Brooks-Carson

Department of Foreign Languages and Literatures

University of Miami

Melody Jacobs-Cassuto

Department of Psychology
University of Minnesota

June 2000

1
Research report submitted to the Center for Interdisciplinary Studies of Writing (CISW)
and to the Center for Advanced Research on Language Acquisition (CARLA), University
of Minnesota, June 2000. The third author of the final report for this study, Melody
Jacobs-Cassuto, was responsible for conducting the statistical analyses. Funding for
the study was provided by the Center for Interdisciplinary Studies in Writing and by the
Center for Advanced Research on Language Acquisition, University of Minnesota, and
by the Max Orovitz Summer Award in the Humanities, University of Miami. We
gratefully acknowledge the assistance of Anne Fretheim in revising this report.

1
 Abstract

This study explored an alternative approach to short essay writing on language

assessment tasks. Thirty-nine intermediate learners of French performed two essay writing

tasks: writing directly into French as well as writing in L1 and then translating into French.

Two-thirds of the students did better on the direct writing task across all rating scales; one-third,

better on the translated task. While raters found no significant differences in the grammatical

scales across the two types of writing, differences did emerge in the scales for expression,

transitions, and clauses. Retrospective verbal report data from the students indicated that they

were often thinking through English when writing in French, suggesting that the writing tasks

were not necessarily distinct in nature. Since the study was intended to simulate writing

situations that students encounter in typical classroom assessments, the findings suggest that

direct writing may be the most effective choice for some learners when under time pressure.

2
 Introduction

This report explores the issue of nonnative writers attempting to think directly through

the second or foreign language while composing text. The question is whether, contrary to the

common-sense view held by many language teachers and students alike, thinking through the

first language (L1) and even writing out a text first in the L1 may actually enhance the

production of good writing, at least at the beginning and intermediate levels. The report will first

briefly review the literature on writing strategies as it relates to this issue of translation and will

then present a study which explored the use of translation from the L1 or a dominant language

while composing text in yet another language.

Studies Contrasting L1 and L2 Strategies

Most studies that have compared first-language (L1) and second-language (L2)2 writing

have found that there are similarities among the strategies used for the two processes. Both ESL

and FL studies point to a transfer of writing strategies from L1 to L2 writing, particularly for

planning and revision strategies. In one such study on the use of writing strategies, verbal report

data collected from eight ESL students showed that the students used strategies for L2 writing

2
The study reported in this paper looked exclusively at a foreign language

context, namely one in which the language being learned was not that of a community

of speakers in which the students lived nor did they have any regular contact with such

a community of French speakers. For the sake of simplicity, this paper will generally

refer to the foreign language situation under the same L2 label that is often reserved for

second language environments in which contact with native speakers takes place.

3
that were similar to those of many L1 writers: engaging in limited prewriting and planning, and

using much rescanning (Raimes, 1987). However, the ESL students did tend to edit and correct

their work more than did L1 writers. Of the eight students, those who were in non-remedial ESL

classes also tended to revise and edit more and to be more attuned to issues of audience that

those ESL students who were in remedial ESL courses. Background experience with the target

language and with writing instruction in general seemed to have more relevance than did

language proficiency scores on the kinds of composing strategies used.

L1-L2 Strategy Studies Dealing with Thinking About Writing in the L1

Among studies contrasting L1 and L2 strategies, there have been those that have dealt

with how students revert to the L1 in order to think about the writing task. In one of the early

and often cited studies on the transfer of writing strategies, it was found that six Spanish-

speaking adults in an ESL program used the same writing strategies for writing in L1 as for

writing in the target language (Jones & Tetroe, 1987). Those who normally planned L1 writing

in terms of abstract goals or who engaged in global planning for the essay planned in similar

ways in the L2. Moreover, for the L2 writing tasks, those who planned in L1 were able to do so

in more detail than those who planned in L2. Language proficiency was found to play a role in

how much planning was used but not in determining the kind of planning used.

In another small-scale study, the writing processes in Japanese and English of four

intermediate learners of Japanese as a foreign language were compared (Uzawa and Cumming,

1989). The students wrote expository essays in which they had to include some given historical

information, producing two individual essays, one in Japanese and one in English on the same

4
topic (where the two essays were not translation equivalents). All four of the students had

initially reported that they generally used the L1 (English) extensively for generating ideas,

searching for topics, developing concepts, and organizing information. Writers 1 and 2 provided

little verbal report about their writing processes in the two languages but performed similarly on

the essays: providing essentially the same content information in the two essays but simplifying

the target language (TL) Japanese version semantically, syntactically, and lexically. Writer 3,

with beginning level proficiency in Japanese, relied heavily on the L1 essay, attempting to retain

the L1 organization and information while simplifying the Japanese essay. Writer 4 was unable

to produce an essay in Japanese and reported generally translating in order to complete

homework assignments. When these writers attempted to compose in the FL, they reported

limiting the amount of the information in the essay, simplifying the syntax and the vocabulary,

and neglecting questions of audience. The authors suggested that having students try to think

through the FL at this level may actually result in weaker writing.

One final study in the category of L1 planning involved responses to two letters by 28

Chinese-speaking ESL students, whereby their responses were planned in either Chinese or

English and then written in English (Friedlander, 1990). The analysis revealed that when writers

planned in Chinese on a Chinese culture topic and in English on an English culture topic, their

plans and texts were rated significantly better than when they did the reverse. Subjects also

wrote longer plans and essays when there was a language and culture match. Being allowed to

plan in the L1 increased the number and kinds of ideas the students could generate for topics that

subjects had dealt with primarily or exclusively in their L1.

5
Translating a Given Text from L1 to L2

There have also been studies that not only have combined elements of all the above

factors but have also included translation of a given text from the L1 into the L2. An example of

such a study is one by Uzawa (1996), that looked at both writing and translating processes,

attention patterns, and at quality of language use. Twenty-two Japanese ESL students studying

at a Canadian college performed three tasks individually in randomized order, after receiving

training in thinking aloud. One task was to write a first draft of an essay in Japanese L1. The

second task was to write an ESL essay on a different topic. The third task was to translate a

short journal article from Japanese to English. The investigator sat alongside, taking

observational notes and recording the sessions. The protocols were coded according to

metacognitive level attention, discourse level attention, linguistic level attention, and personal

comments. These think-aloud protocols were analyzed and then supplemented by the

observational notes and interviews, and the writing samples were evaluated. The data analysis

was performed by two trained judges for each language with attention to theories of composing

processes.

It was found that (1) most students used a "what-next" approach both in the L1 and L2

writing tasks and a "sentence-by-sentence" approach in the translation task, (2) attention patterns

in the L1 and L2 writing tasks were very similar, but quite different in the translation task --

attention to language use in the translation task was significantly higher than in the L1 and L2

writing tasks and, (3) scores on language use in the translation task were significantly higher than

on the L1 and L2 writing tasks, which were not significantly different from each other.

Consistent with the findings from the previously mentioned studies, the author noted that it was

6
lower proficiency students who benefited most from the translation task. The frequent attention

to language use during the translation process seems to have assisted the writers in being more

accurate. They were "forced" to use words and expressions slightly beyond their levels when

they translated -- consistent with Swain's "pushed" output hypothesis (1985). Uzawa noted that

L2 writing may be conducted at the "i - 1" level, wherein students use only words and

expressions readily accessible to them in their interlanguage (i) and thus avoid expressions and

syntax that they feel are too difficult. Likewise, they may try to write at the "i + 2" level, ending

up writing an incomprehensible text or giving up entirely. The translation approach was also

seen to help at the presentation and organization levels especially if the material has been

prepared by a skilled or professional writer. It can, according to the investigator, constitute a

learning experience.

Another study involving translation of a given text was a case study of a female adult

Chinese-English bilingual, which explored the extent and nature of translation from L1 in ESL

writing and problem-solving (Qi, 1998). The subject performed two tasks in each of the

following categories: 1) writing an essay in ESL, 2) translating a text from Chinese (L1) into

English (the texts presented had been originally in English and then translated into Chinese), and

3) completing a mathematical problem-solving task in English. In each set, one task required a

high level of knowledge and another a low level of knowledge in order to be completed. On the

basis of think-aloud protocols and retrospective interviews, it was determined that the subject

switched to the L1 when capturing the beginning of an idea, when developing a thought, when

verifying lexical meaning, and when working memory was overloaded. In addition, those tasks

requiring a high level of knowledge were also associated with language switches, and according

7
to the author such tasks may even have provoked the language switches.

Studies Involving Mental Translation from the L1

There have also been several studies that have involved mental translation from L1 in the

actual L2 composing process. A case study involving think-aloud verbal report data from four

Chinese ESL students revealed that when the subjects found it difficult to write in the L2, they

switched to the L1 and relied on their L1 writing strategies to complete the writing tasks (Lay,

1982). In particular, they reported translating key vocabulary words from the L1 in order to

generate ideas for the essay, especially for essays on topics associated with their native cultures

or on new topics that were very unfamiliar. So in this study, the use of L1 seemed to be

beneficial in generating ideas to include in L2 writing.

A case study of twenty-three adult ESL learners at a bilingual English-French university

program in Ontario found that the writers reportedly used the L1 to search out and assess

appropriate wording, compare cross-linguistic equivalents, and, though less frequently, reason

about linguistic choices in the L2 (Cumming, 1990). The frequency of their thinking in the L1

was found to be related to the learners’ writing expertise in L1, not to their levels of L2

proficiency. In another paper analyzing data from the same study, Cumming (1989) categorized

the students into three levels of writing expertise and two levels of ESL proficiency. The expert

writers in the study were found to use their L1 strategies when writing in the L2, while the

weaker writers demonstrated problems in several areas related to planning and maintaining

coherence when writing in the L2. Proficiency in the L2 did not seem to affect the type of

writing strategies used by the writers; instead higher L2 proficiency levels allowed the writers to

8
exploit their levels of writing expertise when in the L2. The investigator suggested that L1

writing expertise and L2 proficiency combined to influence L2 writing performance.

Finally, Chelala (1982) found that for two native speakers of Spanish, reverting to the L1

in L2 writing produced both positive and negative effects. Chelala's study sought to explore the

composing process and the coherence of the resulting L2 texts of the two Spanish-speaking

professional women. Chelala identified seven successful and ten unsuccessful strategies

employed by the two subjects, and concluded that using the L1 to compose in the L2 was in the

case of these two subjects somewhat more counter-productive than productive.

Translating an L1 Essay into the L2

While the bulk of the published studies include strategies for L1 thinking about L2

writing and about mental translation from the L1 at the time of L2 writing, there are fewer

studies which have explored strategies involved in producing an L2 essay by full translation of a

draft written in the L1. These studies have suggested that translating may bring some benefits in

terms of organization and complexity to the TL essay, especially for students at the lower

proficiency levels. One such study involved translated versus direct EFL writing with 48 fourth-

year Japanese university students at two proficiency levels (as determined by an oral interview

and a standardized grammar test) (Kobayashi & Rinnert, 1992). One group was instructed to

write their first essay in Japanese L1 and then translate it into the foreign language, English,

while a second group wrote directly in English first. The next day the groups reversed tasks and

wrote their second essay on another topic. The compositions written in the translation mode

demonstrated higher levels of syntactic complexity, showed benefits in the areas of content, style

9
and organization, and had more clearly stated theses. Students at lower levels of proficiency

benefited from translation whereas those at higher levels did not. When the students were asked

for their writing preference, 77% reported preferring direct composition to translation. They

based their view on the difficulty of conveying subtle nuances of meaning when translating, and

on the tendency to use familiar words and structures and simpler ideas when writing directly. In

addition, several indicated preferring the direct approach because they wanted to think in

English. As for the advantages of translating, the students felt that the ideas were easier to

develop, thoughts and opinions could be expressed more clearly, and words could be more easily

found through the use of a dictionary. The students reported being able to think more deeply in

their native language and better express their thoughts and opinions. Translating was also

viewed by some as helping in vocabulary acquisition.

The investigators asked for retrospective self-report from the students as to "how much

Japanese they thought they were using in their minds while they were writing directly in

English." Since 55% of the higher-proficiency students and 87% of the lower-proficiency

students reported using Japanese half the time or more when supposedly writing directly in

English, the direct writing treatment was actually somewhat less direct than the label would

imply. In any event, the researchers suggested on the basis of their results that at least for

students at a lower proficiency level a translation strategy in writing might be beneficial, and that

as their proficiency improves, they would switch more to direct foreign language writing,

depending on what they were writing.

At the same time that the Kobayashi and Rinnert study was being conducted, another

study of writing through translation was being conducted, examining the effect of using

10
translation as a writing strategy in writing for French as a foreign language (Brooks, 1996).

Thirty-one intermediate-level students participated by preparing similar pre-writing activities and

two compositions on similar topics. One composition was written directly in the target language

(French), while the other was translated from a native language rough draft (in all cases,

English). The findings showed that participants received higher overall scores in the translation

mode than in the direct writing mode. When ratings of performance were broken down by

categories (accuracy, cohesion/coherence, and argument), the translated versions of the essays

were rated significantly higher on the cohesion/coherence dimension (i.e., the extent of cohesion

as a contributor to coherence). These results may be explained by an analysis of syntactic

complexity in the writing, where higher levels of subordination and coordination were viewed as

indicators of "good writing."

Studies such as those by Kobayashi and Rinnert and by Brooks might lead to speculation

that for some lower proficiency nonnative writers, trying to think directly in the L2 while writing

may actually result in a lowered standard of writing than that which can be produced by both

thinking and writing out the text in the L1 first, and then translating it into the L2. For certain L2

writers, then, might the attempt to think directly through the L2 actually detract from the

production of good writing? If so, this would be an indication that initial thinking about the topic

and writing it up in the L1 may actually support the production of the FL despite the admonition

that such cognitive behavior encourages negative transfer and is thus counter-productive.

Of course, any such conclusions need to be tempered by considering the linguistic and

sociolinguistic relationship between the two languages involved (e.g., is it an L2 or FL

situation?), the learners' control of their L1 and the TL (e.g., are they lower or higher proficiency

11
writers?), the learners' motivation to write in the language, and the genre of the writing. An

effort at replicating the Kobayashi and Rinnert study with Arab learners of English as a second

language (Ali, 1996) produced results that favored direct writing in English rather than

translation from Arabic. Sixty native Arabic-speaking university students wrote 60 essays in

Arabic, 60 in English, and also translated 60 essays from Arabic into English on four topics

during regular class hours. On the basis of holistic ratings of writing ability, direct writing in

English as a second language was rated higher than writing translated from the L1, Arabic. So,

for this configuration of students in this given context, direct writing in English was rated more

positively.

In the translation approach to writing in the TL, past research would suggest that

cohesion (e.g., through markers of transition) and syntactic complexity (e.g., clause variety)

would be enhanced. Likewise, breadth of expression might benefit as well since writers would

be attempting to use a broader vocabulary and set of phrases, consistent with L1 expression.

Although not substantiated by the research literature so far, it may be the case that grammar,

especially syntax, would suffer in translated writing because writers would no longer be using

only those "safe" grammatical forms that they know how to use in order to avoid what they do

not fully control. Always in such studies it is important to determine the writer's ability to be

cohesive, use syntactic complexity, demonstrate breadth of expression, and be grammatical in

the L1 since any lack of these abilities in the L2 may simply reflect a similar lack in the L1.

Another area of interest is that of cross-language comparisons. It might be that those engaging in

direct vs. translated writing whose L1 is more similar to the TL would have an easier time.

12
The Application of Translated Writing to Assessment of Written Language

It has been suggested that language testers make every effort to "bias for best" in their

assessment practices (Swain, 1984). Since language behavior is complex and since any

assessment of it is just a snapshot in time, the onus is on the tester to construct the instrument in

such a way that students have an opportunity to perform at their very best. One way to do that

has been to use more than one task in order to assess a given modality (e.g., speaking, listening)

(Shohamy, 1984). Thus if respondents perform poorly on one task, especially if it is one where

the stakes are high, such as in a testing situation, they still have at least one more opportunity to

demonstrate their proficiency.

Although language assessment has over the years included translation tasks, there has

been little or no systematic effort to utilize translated and direct written essays as a means for

obtaining a more complete assessment of writing. As the research reviewed above has shown,

while a translated version of an L1 essay may in some ways be a reduced version of the L1

essay, there is also an indication that this approach to FL writing may nonetheless enhance

certain aspects of the writing, such as the cohesion and coherence of the piece, the breadth of

vocabulary, the sophistication of clause use, and control of certain grammatical forms. Until

now, no research has sought to include direct vs. translated essay writing within assessment

measures.

The study to be reported on below investigated in a simulated assessment context the

value of including translated writing as a means of obtaining a more complete indication of a

learner's written language ability. In the spirit of biasing for best, an effort was thus made to

determine the extent to which students might benefit from translated writing – whether, for

13
example, the translation mode would encourage the use of more complex language than would

the direct mode, assuming that the students were working from a presumably sophisticated L1

text. Another purpose of the study was to include Spanish-English bilingual college students to

determine if they had different learning experiences with French given the proximity between

these two Romance languages, the numerous true cognates and translation equivalents, and other

ways in which the languages parallel each other. It was posited that the Spanish-English

bilinguals would have certain advantages and that their strategy patterns would be different from

those of the English speakers. There is growing documentation of the advantage that accrues to

multilinguals (with three languages or more) when they choose to think through a similar

language similar to the target language while learning it (Cohen, 1998: 166-169).

Research Questions

1. How does writing a short essay directly in a second language compare to writing the essay in

the first or dominant language and then translating it?

2. What do students consider to be the relative advantages and disadvantages of the direct and

the translated writing modes?

3. How might the strategies of students writing in a second language compare to those writing in

a third language in direct vs. translated writing tasks?

4. How do student strategies and writing mode preferences interrelate?

5. What strategies do Spanish-English and English-Spanish bilinguals share and how are their

strategies different?

6. How do Spanish-English and English-Spanish bilingual students’ ratings compare in the two

14
written modes and on language background variables?

Research Design

Pilot Study

A pilot study involving direct and translated writing in French was conducted with eleven

learners of French at the University of Miami, at the beginning of the Spring semester, January

1999. They were all enrolled in one section of an intermediate-level course. Of the eleven, two

were native speakers of English who wrote their L1 essay in English and then translated it into

French. Two were Spanish-speakers who chose to write the L1 essay in English. Four were

Spanish-speakers who wrote their L1 essay in Spanish, and three had native languages other than

French or Spanish. Two of these chose to write the L1 essay in English and one wrote the L1

essay in Portuguese. All eleven students wrote short one-page essays in class in response to four

writing prompts taken from the Test of Written English. All participants completed a Participant

Questionnaire and three Writing Strategies Checklists, and completed a Comments Sheet.

Finally, they participated in an in-class, recorded discussion in English (French or Spanish, as

they needed) about the experience, their opinions of the prompts, and about their comments from

the Comments Sheet. From this feedback, the researchers decided to use for the main study the

Participant Questionnaire, the Writing Strategies Checklists, and the Comments Sheet as in the

pilot study and chose the two writing prompts that had elicited the best responses.

Sample for the Main Study

Out of an initial sample of sixty-seven learners of French from six intact class sections of

15
an intermediate-level, third semester French course at the University of Miami, thirty-nine

participated in the study. Students who were absent for one of the data collection sessions were

eliminated from the study. Twenty-five of the students in the study were native English

speakers, ten Spanish-English bilinguals, and four native speakers of other languages. The

average age was 20.6, with 30 females and 9 males. This four-skills course met three times a

week, for 50 minutes at a time. French was taught in the TL from a communicative approach,

where writing received similar emphasis as did the other skills.

Students at this intermediate level were expected to write essays in French (4-5 in the

semester of about 1-2 pages in length). All compositions were written outside of class, without

time constraints and with the support of reference materials. Only three writing tasks were given

in class, to be written during an exam (two or three paragraphs long). Topics covered different

modes of writing including narrative, description and argumentation, so that the task posed in

this activity and the amount of writing requested was consistent with the students’ prior

experience. Such in-class writing assignments were used primarily for students to demonstrate

linguistic control and to help them in language learning and in developing writing strategies,

rather than primarily to connect with a reader.

Instrumentation3

Questionnaires

A Background Questionnaire was used to obtain demographic information on the

students' age, gender, and year at the university, as well as language background information

3
All instruments are available from the authors upon request.
16
regarding the students’ dominant language of literacy and self-assessed proficiency in Spanish

and English writing (see Appendix A).

This information was supplemented by a brief Follow-Up Background Questionnaire

which was developed to obtain information not gathered in the first round. The questionnaire

asked for additional language background information about the students’ past course work in

French (i.e., where they had studied, for how long, and when) and about any formal schooling

they may have had in a language other than English.

Writing Prompts

There were two topics for the writing prompts administered to all study participants,

taken from a list of sample essay topics for the Test of Written English (Educational Testing

Service). These topics were chosen because they were expected to provoke personal opinion

essays requiring a certain level of TL complexity. They were translated into French and Spanish

by native speakers and then back-translated into English to verify the accuracy of the

translations. One of the co-investigators, who was not a regular instructor of this course,

administered all of the instructions, questionnaires, writing prompts, and checklists during two

normal class meetings for each course section. The investigator instructed the participants to

write a short though formal, in-class essay of about one-page length on each of the following

topics:

17
 Writing Prompt A:

(French version) Êtes-vous d'accord ou non avec la déclaration suivante? Il n'y a rien que les

jeunes peuvent enseigner à ceux qui sont plus âgés.4 Soyez précis dans votre réponse et justifiez

votre argument par des exemples concrets.

(English version) Do you agree or disagree with the following statement? There is nothing that

young people can teach older people. Use specific reasons and details to support your answer.

(Spanish version) ¿Está de acuerdo o no con la declaración siguiente? No hay nada que los

mayores puedan aprender de los jóvenes. Sea preciso/a en su respuesta y justifique su

argumento con ejemplos concretos.

Writing Prompt B:

(French version) Êtes-vous d'accord ou non avec la déclaration suivante? Quelquefois il est

préférable de ne pas dire la vérité. Soyez précis dans votre réponse et justifiez votre argument

par des exemples concrets.

(English version) Do you agree or disagree with the following statement? Sometimes it is better

not to tell the truth. Use specific reasons and details to support your answer.

(Spanish version) ¿Está de acuerdo o no con la declaración siguiente? Algunas veces es mejor

no decir la verdad. Sea preciso/a en su respuesta y justifique su argumento con ejemplos

concretos.

4
The subjunctive construction here was avoided because the students had not

yet studied this usage.

18
 Each set of topics was provided in a bilingual format, in French/English or in

French/Spanish, in order to make the topics as clear to the writers as possible. For each writing

session, the participants were allowed to choose between the two presentations of the same

prompt according to their language preference. Native speakers of English and of languages

other than Spanish chose the French-English version. Eight self-reported Spanish-English

bilinguals chose the French-Spanish presentation. Two of the Spanish-English bilinguals,

however, preferred the English-French presentation since many of these participants had

completed formal schooling in English and were at the time studying in an English-language

university.

During this mode of writing, there was no supervision or correction by the

administrator, only explicit instructions to write directly in French. While students left no

notes or marks in English on their direct writing French essays, there was no way to

control for any mental translation back into English. In the translated writing format,

they were asked to write the essay in their native or dominant language (allowing for

revision if necessary) and then write a French version, which was a product of

translation. In the direct French writing format, they were asked to write a draft and then

edit it. No dictionaries were allowed in either writing session because of time

constraints (i.e., all data collection was limited to a 50-minute class period; see Data

Collection Procedures below) and because their introduction would have constituted a

new, extraneous variable in the writing process.

19
The Strategies Checklist and Evaluative Feedback

For each of the three essay tasks, a Strategies Checklist was constructed, based on

empirical evidence of strategy use in similar previous research and on reports of strategy use

from the pilot data collected for this study (Appendix B). The checklists contained possible

processing strategies on different phases of each specific set of writing tasks. For the direct

writing mode, there was a checklist for writers to indicate the extent to which they used the listed

strategies in direct writing in French. For the translated writing mode, there were two separate

lists: one for indicating the extent to which listed strategies were used while preparing a draft of

the essay in the L1/dominant language (English or Spanish), and then a different checklist of

strategies specifically related to the translation of the English or Spanish-language essay into

French.5 Participants were asked to provide retrospective verbal report regarding the extent to

which they made use of the various strategies while performing the task.

In addition to the Strategies Checklists, an open-ended Comments Sheet was designed on the

basis of feedback gathered during the pilot study, and it contained a series of open-ended

questions asking the students to sum up their reactions to the study (Appendix B).

A content analysis of the open-ended responses to the Comments Sheet led to the

construction of a new checklist based on the most common responses (Appendix C). This

instrument, known as the Follow-Up Strategy Checklist, constituted a retrospective measure of

strategy use and attitudes. Students were requested to indicate on a 5-1 frequency scale the

degree to which they agreed with statements about the effectiveness of translation as a strategy,

5 It is somewhat simplistic to refer to “dominant language” since it was not so much

dominance as preference of language for writing the L1 essay which suggested
dominance in this case. For this reason, we tend to refrain from using the word
dominance, but rather just indicate the direction of language preference (e.g., S-E or E-
20
and to supply additional feedback about the experiences which they had initially reported on

during the completion of the essay tasks.

Data Collection Procedures

Writing Sessions

The order in which the participants received both the writing mode (direct or translated)

and the writing prompt (A or B) was scrambled to control for an order effect. In other words,

roughly equal numbers of students had the same ordering of modes or of prompts. Of the six

intermediate sections, two small ones had to be combined in order to balance out the numbers.

Directions were given to participants to include an introduction, transitions, and a

conclusion, and to use paragraph structure in their writing. The students were told how long they

would have to write for each task. Before the timed writing session was started, they were given

two to three minutes to think about the writing prompt and to make written notes.

When the group was writing in French, one of the co-investigators read the prompt aloud,

in French, and orchestrated a three-minute spoken activity among paired participants in which

they discussed, in French, two arguments for and against the statement given. This activity was

designed to give the participants an opportunity to warm-up in French before beginning to write.

This brainstorming session lasted about three minutes and was designed to balance out the

perceived warm-up advantage of the translated writing mode, where students had an opportunity

to think through their ideas in the L1 before writing in French. In essence, the students treated

the L1/dominant language version of the essay as a working draft and not as a final product.

S). 21
 Study participants were asked on two separate days to write on each prompt in a 30-

minute timed session, according to each of two modes. In one sitting, they wrote first in the

L1/dominant language of choice (12 minutes) and then translated the essay into French (18

minutes). While this mode did not provide time for revision of the French essay, the production

of a translation was seen as a reworking of the L1 draft. In the other sitting, they wrote on the

topic exclusively in French (for 20 minutes), with announced time for revision (10 minutes).6

The rationale for providing revision time in the direct French writing mode was to offset the

advantage that may have accrued in the translated writing mode from writing the essay twice,

once in the L1 and the other time in the target language. The time limits for the various activities

were determined on the basis of the pilot results. The students wrote longhand, since

keyboarding and writing seemed to be different skills, with keyboarding introducing the

variables of FL accents and speed. No effort was made to control for handwriting speed

variations.

Strategies Checklist, Comments Sheet, Follow-Up Questionnaire and Checklist

After each writing session, all participants completed the Strategies Checklist specific to

the writing mode they had just used: the Strategies Checklist for direct writing in French, the

checklist for writing in the L1/dominant language, and the checklist for translating into French.

Participants were asked to indicate the frequency with which they used a series of strategies

during the writing task that they had just completed. After the second of the two writing

6
An analysis of the essay pages revealed that the students did not write out any

notes or text in the L1/dominant language while preparing and writing the French

essays.
22
sessions, the participants were asked to complete a Comments Sheet. One month after the essays

were written, the Follow-Up Background Questionnaire and the Follow-Up Strategy Checklist

were administered, again during regular class periods.7 One of the co-investigators distributed

the two and asked participants to complete them during the last ten-minutes of class time.

Data Analysis Procedures

Rating of Essays

The Raters

The six raters selected for this study were all native speakers of the language that they

were to rate so that they would be the most able to perceive native-like expression. The two

native-speaking raters for French underwent a day of training to ensure that they would use

appropriately the multitrait rating scales specifically designed for this essay writing task, and the

two English and two Spanish native speakers were trained to use multitrait rating scales

appropriate for assessing essays in English and Spanish respectively. Their ratings were

calibrated to assure that there was consistency. Both of the raters had over five years of prior

teaching, with experience in grading essays in language courses.

The Multitrait Scales

The multitrait rating scales for the French essays were designed to assess four multitrait

aspects of writing that focused more on the form and function of the writing than on the content

of the ideas: expression (freedom from translation effect, variety in vocabulary, and sense of the

7
It took a month to content analyze the open-ended responses and prepare the

new checklist and other follow-up questions.

23
language8), transitions (organizational structure, clarity of point, and smoothness of connectors),

clauses (use of subordination and use of relative pronouns), and grammar (prepositions/partitive

articles, noun/adjective agreement, and verbs). These rating scales were based in broad terms on

scales used in previous research studies to reflect the language elements in writing likely to be

influenced by a given topic and set of specific writing conditions (Cohen, 1994, Ch. 9). Hence,

while some attention was paid to matters of style, this study did not attend to rhetorical

dimensions of students’ writing, given that the writing period was only 20 minutes or less.

The specific primary traits that went into the chosen multitraits were generated

empirically from close scrutiny of the essays written by students in the pilot. Thus, for example,

problems with French partitive articles (e.g., du, de la, des) were frequently observed in the pilot

data so this category was included under “grammar,” while other more traditional categories

were omitted. The categories were rated on a 5-to-1 point scale, where 5 equaled "excellent" and

1 equaled "poor."

The multitrait ratings for the English and Spanish essays were designed to examine three

of the four categories: expression, transitions, and clauses. The category of grammar with its

three scales was eliminated, but space was provided for raters to make open-ended comments.

Again, each scale had a point range from 5 to 1, where 5 equaled "excellent" and 1 represented

"poor" (see Appendix D). In the course of the interrater reliability training, the Spanish raters

determined that some of the Spanish essays that they were rating reflected negative transfer from

8
Sense of the language referred to the writer’s control over native-like expression.

This included idiomatic expressions and stylistic turns of phrase – in other words, the

extent to which the otherwise “correct language” was what the native writer would write

in that context. 24
the language of the university (English). In addition, six non-natives of English wrote essays in

English, two of them Spanish-English bilinguals. Hence, the two scales in the expression

category, freedom from translation effect and sense of the language, were kept in the rating

instrument.

Raters' Feedback on the Rating Process

In order to increase the reliability and validity of the essay ratings, the raters were given a

set of rater feedback questions to answer in order to provide information beyond interrater

reliability coefficients. They were asked to send an e-mail message responding to the following

questions:

1. How well did the rating criteria/descriptors characterize the evaluated essays?

2. Was it clear how to score the essays on the rating sheets?

3. To what extent were they measuring what the researchers had intended them to

measure?

4. How well did the rating approach fit with the raters’ own personal assessments of the

essays and the extent to which there were differences between the evaluation they would

normally have given and the one that they gave according to the criteria?

Statistical Procedures

Inter-rater reliability coefficients and one-way ANOVA comparisons were calculated for

the three sets of raters (French, English, Spanish) by topic (A or B) and by mode (direct vs.

translated). Because the English and Spanish essays had only 8 rating scales and the French

25
essays had 11 scales (the first 8 + 3 grammar-related scales), three summed scores were

computed: 8 scales, 3 scales, and 11 scales. These sums were averaged across raters by

language. The average summed scores were then compared within and across groups. Paired t-

tests were run comparing direct vs. translated writing. Regarding translated writing, the analyses

looked at the thirty-one students who wrote the L1/dominant language essay in English, those

eight who wrote it in Spanish, and the whole group together. Independent t-tests were also run

comparing performance of students writing in English and those writing in Spanish. Since there

were two Spanish-English bilinguals who chose to write in English, and there were four students

whose native language was neither English nor Spanish and had to write in English, three

different permutations of groupings were analyzed.

In order to relate reported strategy use and task preferences to rated performance,

individual scale ratings were correlated with reported strategies on the direct and translated

writing tasks (considering as significant all correlations of .40 or better, p<.01). In addition,

individual rating scale criteria were cross-tabulated with the strategy and task-preference items

from the Follow-Up Questionnaire, introducing into the analysis the variable of whether the

students had scored higher on the direct or the translated writing task. Finally, self-assessed

French ability (ability to speak informally with friends, comprehend a fast-paced movie, write a

term paper for a course, and give a class talk in French) was correlated with students’ perception

of the helpfulness of the direct vs. the translated writing modes.

26
 Results

The Reliability of the Ratings

Before answering the research questions, it is necessary to report on the reliability of the

essay ratings for the six raters who rated in pairs across the three languages (see Table 1). For

the French raters using the 11 rating scales on the 39 essays written in the direct mode, the

reliability was high, .92. For the essays written on the first topic in the direct mode (“youth

teaching their elders”), the reliability was even higher, .97, while for the second topic (“telling

the truth”), it was lower, .87. For the translated mode, the overall reliability was .89, with a

higher reliability for the first topic, .94, and a lower one for the second topic, .86. The interrater

reliability coefficients for the two raters of the 31 English essays were lower: .80 overall (31

essays), .56 on the first topic and .94 on the second. The ratings of the eight Spanish essays had

still lower overall interrater reliability, .74, with .40 on topic one (3 essays) and .73 on topic two

(5 essays), attributed largely to the small number of essays. Nonetheless, one-way ANOVA

confirmed that none of the pairs of raters were significantly different from one another in their

ratings overall or by topic. The reliability of the essay ratings for the two raters (see Table 1) was

excellent, with an overall interrater reliability coefficient of .92 (see Table 1).

Table 1

Interrater Reliability for the Two Raters

Topic Direct Writing Mode Translated Writing Mode

1. “Youth
has nothing to .97 .94
teach their
elders”

27
 2. “A lie is
sometimes .87 .86
preferable to
the truth”

Overall .92 .89

Rating

With regard to raters’ feedback on the rating process, the first Spanish rater felt that the

scales for freedom from translation effect and sense of the language were most relevant to her

ratings because she found the main problem in the Spanish essays to be transfer from English.

Given what she saw as the writers’ problems with grammar, she would have wanted a scale for

general control of L1 grammar. This problem had not been foreseen and had not emerged as an

issue with the native English writers. It appears to be an issue for Spanish-English bilingual

students studying in an English-medium university.

The other Spanish rater had concerns about reliability across languages, in terms of the

descriptors for each rating scale check point:

I do not feel that the rating criteria were very clear. The researchers basically left it

up to the raters to decide what Excellent, Good, Fair, Weak and No Control

constituted. This is very subjective, and even when a pair of raters may more or less

agree on what these criteria mean, there is nothing to say that there exists any

agreement across the languages.

In addition, he felt that the English and Spanish language raters were being asked to do

two things at the same time -- not only assess the writing skills of the students, but also judge the
28
native language ability of the writers:

These are different abilities that are not necessarily associated. With some of the

Spanish essays, it was clear that the writer never had any formal training in Spanish.

Rather, their writing resembled what one would produce in speech.

Research Question #1: Direct Writing of a Short Essay in the Second Language vs.
Writing the Essay in the First Language and then Translating It

For all 39 students, the mean performance on the direct method was significantly better

than performance on the translated method (using t-tests for paired samples) (see Table 2). For

the 25 native English speakers and six nonnatives who wrote the translated essay task in English,

the direct version was also rated significantly higher (p < .05). In the case of those eight

students who started the translated essay task in Spanish, the difference between the two modes

was not significantly different, most likely due to the small sample size. However, the

differences were all in the same direction and were of the same magnitude. We note that the

students writing in Spanish did perform better in French in both modes than those writing in

English. This finding then seems to confirm one of the hypotheses, namely that it would be

easier for Spanish speakers to write in French than for English speakers.

(see table 2 - on next page)

29
 Table 2

Performance on the Direct vs. the Translated Writing Modes

(using t-tests for paired samples)

Direct Writing Mode Translated Writing Mode

Mean S.D Mean S.D. t

All Rating Scales:

All students
(N = 39) 34.23 9.45 31.41 8.42 2.67**

English essay writers

(N = 31) 33.39 8.58 30.52 8.20 2.36*

Spanish essay writers

(N = 8) 37.50 12.39 34.88 8.91 1.16 n.s.

Grammar Scales (prepositions/partitive articles, noun/adjective and pronoun agreement,

verbs taken together):

All students 9.00 2.40 8.72 2.67 0.88 n.s.

English essay writers 8.81 2.15 8.48 2.64 0.88 n.s.

Spanish essay writers 9.75 3.28 9.62 2.76 0.18 n.s.

Other Scales with Statistically Significant Differences:

sense of the language

All students 2.51 1.04 2.20 0.91 2.25*

English essay writers 2.40 0.90 2.05 0.76 2.28*

Spanish essay writers 2.94 1.47 2.81 1.22 0.42 n.s.

organizational structure
All students 3.27 1.16 2.76 1.14 2.76**

English essay writers 3.19 1.14 2.76 1.12 2.14*

30
 Table 2 (Contd.)

Direct Writing Mode Translated Writing Mode

Mean S.D. Mean S.D. t

Spanish essay writers 3.56 1.27 2.75 0.72 1.76 n.s.

smoothness of connectors
All students 3.12 1.09 2.54 0.98 3.52***

English essay writers 3.11 1.02 2.47 0.97 3.42**

Spanish essay writers 3.12 1.41 2.81 1.00 0.96 n.s.

clarity of points
All students 3.52 1.04 3.13 0.92 2.50*

English essay writers 3.44 1.04 3.13 0.97 1.61 n.s.

Spanish essay writers 3.88 1.03 3.12 0.79 3.55**

_________

Key: *p<.05, **p<.01, ***p<.001,

n.s. = not significant

As it turned out, twenty-five did better in the direct mode, thirteen students did better in

the translation mode than in the direct one, and one did the same on both. A cross-tabulation

analysis found that there was no significant pattern between whether a student reported one mode

or the other being more helpful to them and how they were actually rated by outside raters. One

reason for the lack of significant patterns was that for four of the thirteen students rated better on

the translated writing mode, the mean difference was small (within two points or less) and

moreover the sample size was small.

31
 When looking at just the three rating scales for grammar, we see that there were no

significant differences between the direct vs. translated modes (see Table 2). With regard to the

other rating scales, when all students were included in the analysis, the following scales showed

higher ratings for the direct than for the translated mode: one scale from the expression

category, sense of the language, and all three of the scales from the transitions category –

organizational structure, clarity of points, and smoothness of connectors.

In addition, those students with a higher self-assessed index of proficiency in French

(“speaking informally with friends,” “understanding a fast-paced movie,” “writing a term paper,”

and “giving an academic talk”) rated the direct mode of writing as more helpful to them than the

translated mode. For the overall index, the F ratio for the one-way ANOVA was 4.93 (p < .05)

and for “speaking informally with friends” alone the F was higher still (9.94, p < .01). Thus,

those students who rated themselves as more proficient in French also tended to prefer the direct

method of writing.

Although comparisons of ratings were conducted across languages, it is important to bear

in mind that the pairs of raters across languages were different, and so the criteria for rating were

interpreted somewhat differently. With this caveat in mind, then, the 31 L1/dominant-language

essays were rated significantly higher on the eight scale criteria that were common to the two

tasks than were the French essays written in the direct mode (see Table 2). The mean for the

translated writing task was also significantly lower. The comparison between rated performance

on the Spanish language essays and the French essays written in the direct mode was not

significant on the eight criteria. The mean for French essays in the translated mode was

significantly lower at the .01 level.

32
Research Question #2: Student Appraisal as to the Advantages and Disadvantages of the
Direct and the Translated Modes

A key finding resulting from the retrospective verbal report data gathered by means of the

checklists was that the direct writing mode was not as “direct” as might have been expected.

Although the student writers were instructed to write one essay directly in French without

translation, just as was the case with Japanese writers supposedly writing directly in English in

the Kobayashi and Rinnert (1992) study, the vast majority (80%) of English L1 students (N=25)

reported thinking in English “often” or “always” while doing the French essay in the direct

writing mode. Contrary to the researchers’ expectations, the Spanish-English bilinguals (N=10)

also reported thinking at least “some of the time” in English when writing the French essay in the

direct mode, and in addition reported less thinking in Spanish than in English, most likely an

artifact of being students in an English-medium university.

As indicated above, one month after performing the essay tasks, the students were given a

Follow-Up Strategy Checklist based on a content analysis of their open-ended responses from

the Comments Sheet which they filled out at the time they completed the second writing tasks.

For this follow-up checklist they were asked to indicate the extent to which they agreed with a

series of statements regarding their performance on the direct and translated writing tasks and

their attitudes towards them. The picture that emerged from the responses of the student group

as a whole was as follows: at the same time that the virtues of direct writing in French were

endorsed, the advantages of translated writing were also noted (see Table 3).

33
 Table 3

Student Retrospective Appraisal of Behavior and Attitudes

Towards the Direct vs. the Translated Modes of Writing

Scale: 5 = always 4 = often 3 = sometimes 2 = rarely 1 = never

Direct Writing Mode

Mean response:

4.15 -- felt that writing in French helped them learn the language.

3.83 -- felt that writing in French helped to focus on French expression.

3.36 -- felt that thinking in French during the whole process was better than translating.

3.22 -- felt that writing in French was faster than translating.

3.10 -- found it difficult to write directly in French.

2.90 -- found it easier to write directly in French than to translate.

Translated Writing Mode

Mean response:

4.44 -- had better vocabulary in the dominant language essay for the French essay.

4.22 – found it easier to write in the dominant language than in the essay directly written French.

3.97 -- had a greater number of ideas for the dominant language essay than for the French essay.

3.86 -- thought through their ideas more clearly in the dominant language essay than in the

French essay written directly in French.

3.58 -- translating the essay from their dominant language brought better organization to the

34
 Table 3. (Cont)

French essay than if they had written it directly in French.

3.47 -- changed the organization of the dominant language essay somewhat to fit the French

language.

3.35 -- purposely simplified their dominant language text in order to translate it into French.

3.33 -- felt that they had time pressures to complete the translation into French.

3.12 -- found it difficult translating into the foreign language.

2.82 -- disliked being required to use translation as a strategy.

The student group as a whole reported that writing directly in French “often” helped them

learn the language and helped them to focus on French expression. Given that many of them

reverted to English while supposedly writing directly in French, it is not surprising that they

averaged at the middle of the scale with regard to whether thinking in French during the whole

process was better than translating. They indicated that even less often did they find it faster or

easier to write directly in French than to translate. So, while it took them some effort to write

directly in French, they could see the benefits.

With regard to the translated writing mode, the student group as a whole reported “almost

always” having better vocabulary in their L1/dominant language, “often” thinking through their

ideas more clearly in the L1/dominant language essay, and having a greater number of ideas,

which they had expressed in their L1 essay. In addition, translating the essay from their

L1/dominant language reportedly brought better organization to the French essay more often

than if they had written it directly in French. They reported “sometimes” changing the

35
organization of the L1 essay to fit the French language and also reported that “sometimes” they

purposely simplified their L1 text in order to translate it into French.

The eight Spanish-English bilinguals who wrote their L1 essay in Spanish reported

“often” or “always” avoiding word-for-word translation from Spanish in the translated writing

mode. Half of them also reported “often” or “always” having difficulty finding translation

equivalents (Table 3).

Research Question #3: Strategies Used by Students Writing in a Second Language

Compared to those Writing in a Third Language in Direct vs. Translated Writing Tasks

Direct Writing Mode

When students’ self-reported frequency of strategy use was correlated with rated

performance, those who reported thinking less in English while writing directly in French were

also rated higher across all scales and by individual scales as well, the highest being vocabulary

and the lowest being organizational structure (see Table 4). Purposely simplifying the text in

order to translate it had a negative correlation with variety in vocabulary. In other words, those

who reported that they did not simplify were rated higher in vocabulary by the outside raters.

Finally, those students who reported that they found direct writing faster were also rated higher

for variety in vocabulary, clarity of points, and prepositions/partitive articles (see Table 4). So

the picture that emerges is one where those who found direct writing more effective were also

those who were rated higher on various writing proficiency scales.

(see Table 4 - on next page)

36
 Table 4

Strategies Used by Students Writing in Their L2 or L3 Related to

Their Rated Performance in Direct vs. Translated Writing Tasks

(all correlations significant at the .01 level)

All Writers in the Direct Writing Mode:

__reported thinking in English while writing in French with overall rating (-.56), variety in

vocabulary (-.56), use of relative pronouns (-.55), prepositions/partitive articles (-.52), clarity of

points (-.51), smoothness of connectors (-.50), sense of the language (-. 46), use of subordination

(-.45), organizational structure (-.44).

_ purposely simplifying the text in order to translate it and variety in vocabulary rating (-.42).

_ finding direct writing faster and variety in vocabulary rating (.42), clarity of points (.44),

prepositions/partitive articles (.45).

All Writers in the Translated Writing Mode:

_ the use of simpler words and phrases in French and overall ratings (-.43), variety in

vocabulary (.42), noun/adjective and pronoun agreement (-.41), prepositions/partitive articles (-

.48).

_ finding it not really easier to write in their L1 and higher overall on the translation mode

essay (-.48), use of subordination (-.40), prepositions/partitive articles scale (-.42).

_ being less likely to find more ideas when writing in the L1 and higher overall ratings (-.53),

sense of the language (-.44), use of subordination (-.42), prepositions/partitive articles (-.48).

_ not necessarily finding better vocabulary in the L1 than when writing in French and higher

37
 Table 4 (cont.)

ratings overall (-.42), higher ratings in use of subordination (-.44).

Spanish-English Bilinguals in the Translated Writing Mode:

_ thinking through their ideas in the L1 before translating and higher ratings overall

(-.42),

_ thinking through their ideas in the L1 before translating with variety in vocabulary

(-.51),

_ finding it difficult to translate into French with variety in vocabulary (-.41),

_ finding it easier to write in the L1 with variety in vocabulary (-.52),

_ having more ideas in the L1 with variety in vocabulary (-.53).

Translated Writing Mode

With respect to the translation mode, 64% of the English L1/dominant students indicated

that they engaged in word-for-word translation at least some of the time, with only 36%

reporting that they “often” or “always” avoided it. Two-thirds of the English L1/dominant group

reported “often” or “always” having difficulty finding translation equivalents. Among other

things, it was found that those who did not report simplifying their French were rated higher on

the translation task.

For the Spanish-English bilinguals in the translation writing mode, thinking through their

ideas in the L1 before translating correlated negatively with both overall ratings and with variety

38
of vocabulary, suggesting that those bilinguals who were rated higher overall and in vocabulary

range were, by self-account, less likely to think through their ideas in Spanish before producing

the text in their third language, French (Table 4). In addition, not surprisingly, those benefitting

from strategies consistent with the translation mode were also seen to be rated higher on the

various scales in the translation mode (Table 4).

The students in the full group (N=39) who reported that it was not really easier for them

to write in their L1, were also rated higher overall on the translation mode essay and on several

individual rating scales – use of subordination and on the prepositions/partitive articles scale.

Likewise, being less likely to find more ideas when writing in the L1 correlated with higher

overall ratings in the translation mode, and with three individual rating scales – sense of the

language, use of subordination, and prepositions/partitive articles. In addition, not necessarily

finding better vocabulary in the L1 than when writing in French correlated with higher ratings

overall and with higher ratings in use of subordination as well.

Finally, those rated higher on noun-adjective and pronoun agreement were also those

having less difficulty translating into French and those who found it faster to write directly in

French. Those who reported not feeling that they thought better in their L1 were also those rated

higher on the prepositions/partitive articles scale.

Question #4: Interrelationships among Strategies and Mode Preferences

The following results are based on responses to the retrospective strategy and attitude

questionnaire given to the students one month after they had completed the two writing tasks.

(See Table 5 for a complete list of findings.)

39
Direct Writing Mode

If students were more likely to think that direct writing was better, they also found direct

writing to be easier than the translated writing approach, more likely not to find writing directly

in French to be difficult, reported writing faster in the direct mode than in the translated one, and

were more likely to think that direct writing helped them learn the language. In addition, writing

faster in the direct mode correlated negatively with difficulty in direct writing, which meant that

those writing faster in that mode also encountered difficulties in direct writing less frequently.

Those students who felt most strongly that direct writing helped them to focus on French

expression were also those who reported that direct thinking in French was better than thinking

through the L1 first and also felt that direct writing in French helped them in learning French.

Finally, those who reported that direct writing was easier for them than translated writing were

also those who tended not to have difficulty translating into French, tended not to purposely

simplifying in order to translate, could write faster in direct writing.

Translated Writing Mode

Students who held the view that they thought through ideas more clearly in the dominant

language than in the direct writing mode were also those who felt the following: it was easier to

write in the L1, they had more ideas in the L1, it was difficult to write directly in French, it was

not faster for them to write in the direct mode, and they had better vocabulary in the L1.

Those who reported it being easier to write in the dominant language (English or

Spanish) also reported that they could think through their ideas more clearly in the L1, purposely

40
simplified their L1 text in order to translate it into French, tended not to find direct writing in

French easier, tended to find it difficult to write directly, tended not to find direct writing faster,

had a better vocabulary in the L1.

Those who reported having more ideas when writing in the dominant language also

reported that they had a better vocabulary in the L1, simplified their L1 text in order to translate

it into French, tended not to find direct writing faster. Those who reported having more

vocabulary when writing in the dominant language also reported that they did not tend to write

faster in the direct mode. Those who reported difficulty in translating also were those more

concerned about the time pressure and less likely to find direct writing in French faster. Those

who simplified their L1 text in order to translate it into French also reported encountering

difficulty in translating and tended not to find direct writing faster.

Research Question #5: Comparison of Strategies Reportedly Used by Spanish-English and

English-Spanish Bilinguals

Direct Writing Mode: Comparative Strategy Frequencies

For the purposes of in-depth analysis of Hispanic bilingual students of French, we will

focus just on those eight who were more Spanish-dominant bilinguals and the eight who were

more English-dominant1 (see also, Brooks-Carson & Cohen, forthcoming). Regarding strategy

use for these bilingual students writing in the direct mode, both the Spanish-English (S-E) and

the English-Spanish (E-S) groups had similar frequencies of use for most strategies (see Table

6). In terms of planning strategies, both groups reported engaging in more on-line planning, that

is they reported planning as they wrote rather than planning in advance. The Bilingual S-E

41
group reported “sometimes” to “often” (3.6) planning as they wrote and sometimes (3.3)

planning in advance. The Bilingual E-S group reported using much more often the strategy of

planning as they went along (E-S 4.1) and only “rarely” to “sometimes” planning in advance (E-

S 2.6).

For the strategies relating to grammatical accuracy, the two groups tended to report

similar frequencies: “sometimes” to “often” verifying correct negative forms (S-E 3.9 and E-S

4.4), checking for adjective agreement (S-E 3.8 and E-S 4.0), and verifying subject-verb

agreement (S-E 3.4 and E-S 3.6). However, for the strategies concerning syntactic complexity,

the two groups reported different uses: the Bilingual S-E group reported “rarely” connecting

short sentences into longer, complex sentences (2.6) and “rarely” purposely using complex

grammatical structures (2.4) although they “often” used connecting words purposely to link ideas

(S-E 4.0). The Bilingual E-S group reported using these strategies more consistently, reporting

always using them at “sometimes”: connecting short sentences into longer, complex sentences

(E-S 3.4) purposely using complex grammatical structures (E-S 3.0), using connecting words

purposely to link ideas (E-S 3.6). Both groups reported “sometimes” attempting to use a wide

variety of vocabulary, though the Bilingual S-E group tended to do so more often (S-E 3.5 and

E-S 2.9).

In terms of the strategy use for the language of thinking about the task, both groups

reported thinking in languages other than French during the direct French writing task. In both

cases, they reported thinking more often in English than in Spanish. Those in the Bilingual S-E

group reported “rarely” thinking in English (2.6) or in Spanish (2.3), and those in the Bilingual

E-S group reported “often” thinking in English (4.0) but “rarely” thinking in Spanish (2.6).

42
Translated Writing Mode: Strategy Frequencies

For those strategies relating to grammatical and syntactic accuracy that applied to both

writing modes, students in the Bilingual S-E group and the Bilingual E-S group tended to report

different frequencies. The Bilingual E-S group reported using the following three strategies

more often than did the Bilingual S-E group: using complex grammatical forms (E-S 2.6, S-E

2.1), using adjective agreement (E-S 3.8, S-E 3.0), checking negative forms (E-S 3.8, S-E 3.4).

The two groups had the same frequencies for checking subject-verb agreement (E-S 3.4, S-E

3.4) and for using connecting words (E-S 3.9, S-E 3.9). See Table 6 for the frequencies reports.

Both groups reported “sometimes” attempting to use a wide variety of vocabulary (E-S

2.8, S-E 3.0). For the strategies that were specific to the translated mode, both groups reported

using the strategies in similar ways: “often” making an effort to think about how to best express

the ideas in French (S-E 4.3, E-S 3.9), “sometimes” avoiding word-for-word translation (S-E 3.3,

E-S 3.5) and “sometimes” having difficulty finding translation equivalents (S-E 3.0, E-S 3.8). In

the planning strategies, both groups reported using the two strategies: changing the organization

(S-E 3.6, E-S 3.5) and keeping the original organization from the L1 essay (S-E 3.4) about as

often. Interestingly however, the Bilingual S-E group reported “often” to “always” using the

original organization in the translated French essay (E-S 4.4).

Research Question #6: Comparison of Two Groups of Bilinguals on Rated Performance in

the Two Writing Modes and on Language Background Variables

Performance on the Written Tasks

The raters provided scores on the three essays from each student. The scores for each

43
essay within a category (direct French, translated French, and L1) were summed, and the means

and standard deviations were calculated. These results are shown in Table 2. Consistent with

earlier findings, both the Bilingual S-E and Bilingual E-S groups were rated higher on the direct

French writing (S-E 37.50 and E-S 38.06) than on the translated French essays (S-E 34.88 and E-

S 31.94). It may be surprising that the Bilingual E-S group was rated higher than the Bilingual

S-E group on the direct French writing, in spite of the presumed advantage of the Spanish-

dominant bilinguals, given the relative closeness between Spanish and French. However, this

advantage did bear itself out in the scores for the translated French writing, where the Bilingual

S-E group was rated higher than the Bilingual E-S group. While it may be interesting to note

that for the L1 essays, the Bilingual E-S (writing in English) were rated higher on their L1 essays

(E-S 34.69) than were the Bilingual S-E group (writing in Spanish) on their L1 essays (S-E

31.50), it is important to remember that the two L1 groups had different raters reading the essays.

Given the issues raised by the raters themselves about possible differences in rating across

languages, the meaningful comparisons must be limited to those between the direct French

essays and the direct Translated essays. To recap, the Bilingual E-S group was rated higher on

their direct French essays than on their translated French writing. The Bilingual S-E group was

also rated higher on their direct French essays than on their translated French writing.

By way of reminder, the rating scales that were used for the three essays had four criteria

in common: variety in vocabulary, sense of the language, organizational structure, clarity of

points (see Table 3 for the reports by criterion). Freedom from translation effect was a category

for the direct French writing and for the L1 writings but not for the translated writing.

Smoothness of connectors appeared as a criterion for translated writing and for L1 essays. In

44
comparisons of the scores for the direct French writing, the Bilingual E-S group was rated higher

than the Bilingual S-E group in most categories (freedom E-S 3.2 > S-E 2.9; vocabulary E-S 3.9

> S-E 3.6; sense of the language E-S 3.1 > 2.9; organization E-S 3.6 = S-E 3.6; clarity E-S 4.0 >

S-E 3.9). In comparisons of the scores for the translated writing, the results were more mixed.

The Bilingual S-E group was rated higher in vocabulary (S-E 3.6 > E-S 3.3), sense of the

language (S-E 2.8 > E-S 2.0) and in smoothness of connectors (S-E 2.8 > E-S 2.2). The

Bilingual E-S group was rated higher in organization (E-S 3.1 > S-E 2.8) and in clarity (E-S 3.4

> S-E 3.1).

Language Background

Students in the Bilingual S-E group all rated themselves as “excellent” or near

“excellent” in their abilities to use Spanish to complete certain more informal tasks such as,

writing an informal letter to a friend (4.9), speaking informally with friends (5.0), and listening

to a movie with fast-moving and sometimes technical vocabulary (5.0).

Predictably, they indicated that they used Spanish “all or most of the time” in completing

tasks in more informal settings such as speaking at home (4.9), speaking with members of their

extended families (5.0), and “often” speaking with friends (4.4). For formal or academic tasks,

they felt less confident than when completing informal tasks; however, they still rated

themselves between “very good” and “excellent” in their abilities in Spanish to write a term

paper (4.6), and to give an academic talk (4.5). Understandably, they reported using Spanish only

sometimes for the following tasks: when talking at the university (3.5),when speaking to

neighbors (3.8: only two students reported “never” or “rarely”), and “rarely” when speaking with

45
teachers (1.1).

The Bilingual E-S group reported slightly lower proficiency levels and frequencies, but

the patterns were largely the same. They too reported feeling most confident using Spanish to

write an informal letter to a friend (3.9), to speak informally with friends (4.5), and to listen to a

movie with fast-moving and sometimes technical vocabulary (4.6). They also reported using

Spanish most often: to speak at home (4.6), to speak with members of their extended families

(4.9), and sometimes to speak with friends (3.5). Again, they were less confident using Spanish

to complete academic tasks: to write a term paper (3.3), or to give an academic talk (3.5).

Following the same pattern, they also reported using Spanish least often to speak with teachers

(1.1) and rarely to “sometimes” to talk at the university (2.6) and “sometimes” to talk with

neighbors (3.4).

Discussion and Conclusion

This study represented an effort to experiment with an alternative approach to short essay

writing on language assessment tasks. The results showed that two-thirds of the students did

better on the direct writing task across all rating scales, one-third on the translated writing task,

and one student did the same on both. When considering the various multitrait scales, it was

found that grammar ratings were not significantly different across modes. It would appear that

the kinds of choices made in selecting prepositions and partitive articles in French, in

determining adjective-noun agreement, and in processing verb inflections were not so influenced

by the mode. Since these decisions are tied so directly to the production of French text, then the

same considerations would need to made whether the text were written directly in French or in

46
the L1 first and then translated. This would explain the lack of significant difference.

Regardless of how well organized the essay was, how rich the vocabulary range, or how rich in

ideas it was, the writer would still have to focus on the same grammatical needs in producing the

French text.

Interestingly, the students reported thinking through English much of the time that they

were supposedly engaged in the direct French writing task. This report suggests that, while for

the translated writing task they were engaged in written translation on paper, they were

nonetheless engaged in mental translation during the direct writing task. The point is that the

two tasks, then, were not necessarily distinct in nature, but rather overlapping. The extent to

which these two forms of translation overlap is an interesting research question and one that was

simply identified and not probed in this study.

As noted at the outset of this paper, studies by both Brooks (1996) with English-speaking

writers of French and Kobayashi and Rinnert (1992) with Japanese writers of EFL had found

distinct advantages for the translated writing mode. In the Brooks study, the students had been

asked to prepare the two essays out of class, first in draft form and then in revised form, using

software for French word-processing. So they had spent a considerable amount of time

preparing the essays in both the direct and translated writing modes. In the Kobayashi and

Rinnert study, the students were given an hour of class time to write each of two essays on two

successive days, one in Japanese L1 and then one in the FL, English. Then, on a third day they

were given another hour to translate one of the essays into English. This procedure gave them

ample time to assure that their translated essays were of the highest quality they could produce.

The current study was intended to simulate typical writing situations that students face when they

47
participate in classroom assessment. Perhaps it is a better measure of the impact of a given

writing mode since it is possible that in the other two studies, the extra time factor could have

neutralized differences or perhaps even given the advantage to one of the modes, such as to

translate writing, which is not the most effective choice for some or many learners when under

time pressure.

Because giving only one mode of response on an assessment instrument may bias

against students, the findings from this study would underscore the value of multiple measures in

assessment, something that experts in the field of language assessment have been asserting for

some time. Furthermore, this study would suggest that what is ostensibly one task, namely

writing a short essay in the TL, can be enhanced by allowing for alternate means of “warm up”

or task preparation, such as through an initial writing of the essay in the L1 or the dominant

language of instruction at the institution.

Collecting strategy use data along with performance ratings can help investigators obtain

reasons for observed outcomes. For example, these data show that those who reported that they

did not simplify their French in order to produce the French essay in the translation mode tended

to receive higher overall ratings in written performance on the task. In other words,

simplification was detrimental to rated performance because the French raters were trained to

give a higher rating for the use of subordination and relative pronouns, and simplification usually

meant the reduction or elimination of subordinate clauses.

The finding that sometimes the students had difficulty writing directly in French and

sometimes in using the translation mode reflects, in part, differences across the students, since it

is likely that they encountered difficulties when they were asked to work in a non-preferred

48
mode. Partly it could indicate that the tasks were at a comparable level of difficulty and that

both were demanding in their own ways.

As stated at the outset, a purpose for including Spanish-English bilinguals in the study

was to determine if they had different learning experiences with French given the proximity

between these two Romance languages, the numerous true cognates and translation equivalents,

and other ways in which the languages parallel each other. Consequently, it was thought that the

Spanish-English bilinguals would have certain advantages and that their strategy patterns would

be different from those of the English speakers. For this reason, it was surprising to the current

investigators to find that the Spanish-English bilinguals in the study reported thinking in Spanish

less often than they thought in English while writing the French essay in the supposedly direct

mode. It is true that two of them actually used English as their L1 for the translation mode

writing task, but the other eight reported here used Spanish. The predominance of reported

thinking or mental translation into English during French writing may be an immediate

consequence of the fact that their American French textbooks use English, not Spanish, as the

“explaining” language. In addition, these students needed to function in an English-language

university within a bilingual community. It would also indicate a measure of their bilingual

dominance, with the scale weighted towards English dominance, but no language measure was

administered to determine the actual language dominance configuration of the bilinguals.

Several findings underscored this reality of Hispanic bilinguals in an English-language

institution. The Spanish-English bilinguals reported that they “often” or “always” avoided word-

for-word translation from Spanish in the translated writing mode. Half of them also reported

“often” or “always” having difficulty finding translation equivalents. Perhaps these students

49
were actually avoiding thinking in Spanish to some extent, or if not doing it consciously, at least

feeling the effects of being lost in translation or in a language limbo. In any event, their reported

strategy behavior was quite similar to that of the English speakers.

The comparison of the Spanish-English and the English-Spanish bilinguals allowed for

exploration of what it means to be bilingual in a monolingual university, studying a third

language. Both bilingual groups in this study were rated higher on the direct French essays than

on the translated French essays. Such a finding may invite speculation that the bilingual students

were able to perform better in the direct writing mode because of the relative proximity between

the Spanish and French languages. While this conclusion may at first seem plausible, it does not

account for the success on the same task of two-thirds of the participants in the overall study

(where 25 of the 39 were native speakers of English). Furthermore, it is not supported by the

finding that the bilingual students who were dominant in English (rather than in Spanish) were

rated the highest on the direct writing. Instead, an explanation for the success of all students on

the direct writing task may be better linked to the nature of the writing task itself and to the fact

that students had an extended period of writing time to complete the direct writing essay while

the time for the translated task was divided into two distinct parts: writing in the L1 and then

translating.

The advantage of the E-S over the S-E bilinguals on direct writing may be surprising

given the presumed advantage of Spanish speakers over English speakers for working in another

Romance language. But the assumption did seem to hold true in terms of the translated French

writing. A closer look at the scores by category confirms the presumed advantage, which may be

attributable to the connections between Spanish and French for the purposes of translating. In

50
fact those who were dominant in Spanish were rated higher in the categories of variety of

vocabulary, sense of the language, and smoothness of connectors. The English dominant

bilinguals were rated higher in the areas of organization and clarity, which recalls findings from

earlier studies findings about the benefits of using the L1. Studies mentioned in the Review of

Literature described benefits in planning, idea generation, and coherence from allowing L1 use

during the L2 writing process, benefits that may be present here as well.

It is also interesting to note the finding that the English-Spanish bilinguals were rated

higher on their L1 essays than were the Spanish-English bilinguals on their L1 writing. It is

important to remember that the bilinguals were identified as dominant in one language over the

other based on their responses to the Language Background Questionnaires whereas regardless

of the L1 reported, students were free to choose the language in which they would write the L1

essay. It was possible, and it was the case, then that a student could report Spanish as the L1 but

then choose to write the L1 essay in English. It may be that the categorizations, as the students

identified themselves, are not good measures of the students’ actual bilingualism or of their

relative experience with academic writing in the two languages. However, it should also be noted

that the raters for the Spanish and for English essays were different people. Therefore, it is

possible that the raters, while reliable within groups and across languages in general, may have

rated the L1 essays differently. The lower ratings on the L1 essays may be explained by higher

expectations for essays by the raters given that students were to be writing in the L1, or by

possible differences in the styles between the English and Spanish raters.

A third explanation may be the fact that the students knew that they would translate the

L1 essay into French (that it was essentially a rough draft) and that it was the French essay that

51
received the most attention. Moreover, the Spanish Raters reported to the researchers that some

of the Spanish-English bilinguals showed examples of English interference in their Spanish

writing. For this reason, the category of freedom from translation was retained in the L1 essay

rating scale. However, without some verbal report about what the students were planning and

doing, it is not possible to know why the L1 essays were rated lower than most of the TL essays.

Part of the explanation of the performance ratings may come from differences between

the two groups of bilinguals in terms of their experience with the type of writing task given. It is

possible that the English-Spanish bilinguals had had more experience writing personal opinion

essays in English than the Spanish-English group. On the Follow-Up Background

Questionnaire, seven of the Bilingual E-S students reported having most if not all of their formal

schooling in English (the other did not reply). Five of the eight Bilingual S-E students reported

having undergone their formal schooling in Spanish, two did not reply, and one described her

experience in a bilingual school in the Dominican Republic where the language of instruction

was Spanish until the fourth grade and then English from that point on. More information about

the specifics of their formal schooling would have been perhaps helpful in understanding if this

potential difference could account for the performance ratings. In any case, this background

information serves to reinforce the fact that the division between the two groups was not

absolute; instead the cross-over nature of the two bilingual groups’ backgrounds reflects the

complexity of the bilingual experience.

As for the language of thought while writing the essays, the English-Spanish group

reported more often thinking in English than in Spanish while writing in French, and under the

checklist for L1 writing, they reported “rarely” to “never” thinking in French while writing in the

52
L1. This finding suggests that the English-Spanish bilinguals seemed to be much more

comfortable thinking through these types of writing tasks in English, regardless of the bilingual

component of their linguistic background. Moreover, their reported use of thinking strategies

points to some differences from the Bilingual S-E group whose report of “rarely” thinking in

English or Spanish when writing in the direct mode differed from the findings for the overall

sample where, at the intermediate-level, most students engaged in some thinking in their L1 or in

this case, in the dominant language of the academic environment (here, English).

In the translated writing mode, both subgroups tended to report using strategies of

grammatical/syntactic accuracy in similar ways. For the strategies specific to the translated

mode, the Spanish-English bilinguals reported using about equally often the strategies for

keeping and changing the original organization of the essay. This finding may suggest a

flexibility in style and/or a perceived difference between the appropriateness of the organization

for the TL essay. The English-Spanish group reported more often sticking with the original

organization in the translated French essay and less often changing the organization. Again, this

pattern reflected that of the rest of the sample who also reported keeping more often than not the

original organization of the essay. While this finding is consistent with that of earlier studies

suggesting that the L1 can bring a more coherent organization to the TL essay, it is not clear to

what extent this continued use of L1 organization is a result of the time pressure several students

reported feeling. Because they had completed an in-class, timed task, not rethinking the

organization of an essay to be translated may have been a successful strategy for producing a

target language essay within the study’s constraints.

53
 As for the strategies relating to actual translation, those in the Bilingual S-E group

reported only “sometimes” avoiding word-for-word translation and “sometimes” having

difficulty finding translation equivalents while those in the Bilingual E-S group reported slightly

higher frequencies in trying to avoid it and in having difficulty finding equivalents. It may be

that the variation lies in their different Spanish proficiency levels -- that the Bilingual S-E group,

having a firmer knowledge of Spanish, had a stronger sense of real and false cognates and

translation possibilities between the languages. But without data on why individuals chose their

strategies, it is impossible to explain the differences here. In order to elucidate this finding,

further research needs to be conducted using verbal report during the writing sessions.

Limitations of the Present Study

The major limitation of the study was the sample size. It was originally intended that

there be more robust groups of students, especially more Hispanic participants. The actual

numbers made it necessary to consider some of the findings as suggestive rather than definitive.

Secondly, the writing samples themselves were rather short due to the limited amount of time

available for writing in class. While an in-class writing task does reflect in-class essays, the

samples themselves may not have been long enough to bring out some potential differences that

emerged, for example, in the Brooks (1996) study when writers did their essays as homework.

Particularly given the finding that the students tended to think through English when supposedly

writing directly into French, the two tasks may not have been so different in practice, again

possibly due to the limited amount of time available to the students for writing in class.

It would have been ideal to have raters that were balanced bilinguals in French, Spanish,

54
and English in order to make rigorous comparisons of essay results across languages. While

efforts were made to calibrate the ratings of the pairs of raters for each language, no efforts were

made to calibrate across languages. So for this study, the question remains how exactly the

various scales were defined and utilized by the English and Spanish raters in comparison to the

French raters. The gathering of feedback from the raters during the rating process did help to

address this issue but by no means resolved it.

No effort was made to assess the learning style preferences of the students participating

in this study. Such data may have helped to interpret the results for the direct and the translated

writing modes. For example, it may be that in this sample of nonnative writers in French, those

with a more concrete, sequential, analytic, and closure-oriented learning style did better in the

translation mode than those who with a more abstract (intuitive), non-sequential, synthetic, and

open style.

It needs to be noted that subjects were responding to a pre-established set of strategies.

The open-ended questions dealt with likes and dislikes, not with specific processes. So while the

strategies checklists were based both on strategies described in previous empirical findings in the

research literature and on results from pilot data in this study, there was no open-ended probe of

strategies in this study, and it is possible that such a probe would have furnished insights into

other noteworthy strategies that students use in these two modes of writing.

In terms of the students’ use of the L1, there was no mechanism in this study for

discovering what students meant when they reported thinking in the L1. This label was intended

to mean two things: both thinking about the content and the organization of the ideas as well as

using the L1 to think through the arguments for making their case. Hence, “thinking in the L1”

55
could refer to a complex mix of ideas about content, about the organization of ideas, about the

selection of language material, and about the conversion of that material into French.

Finally, no effort was made to deal with rhetorical issues in this study with students

composing for 20 minutes or less. Given the rich literature on contrastive rhetoric to refer to,

this avenue for research may be fruitful for future investigation, especially when essays are

written out of class without time constraints. Then it could be insightful to see whether, for

example, rhetorical patterns would be more native-like when the text has been written directly in

the TL as opposed to when it is translated from an L1/dominant language version.

Suggestions for Future Research

In terms of future research, one project might be to plan an interventionist study with

larger numbers of learners and more options. The study could vary the amount of writing time

allowed. Different groups could get training in different sets of strategies -- some sets favoring a

direct approach to writing and some favoring the translated mode. The training would be

informed by the insights gained in this current study regarding students’ strategy use preferences.

Students could, for example, be assessed for their learning style and writing strategy preferences

before performing the writing tasks. An immediate type of follow up using data from this study

would be to look closely at the fourteen students who did not do better in the direct mode, even

returning to their actual essays to see if any other patterns emerge that might characterize their

behavior.

At the same time that it would pay to enlarge the study, it can also be beneficial to do

more small-scale and case-study work to better understand the issues at play in the current study.

56
For example, it may be beneficial to explore further the issue of whether bilinguals resist

thinking in their L1 if their studies at a university are through another language, even if the

university is located in a bilingual community. This situation could be compared to that of a

university where instruction is bilingual. Other work has established that the language of

instruction at the university level can have a decided impact on the language of thought of the

students. A study by Cohen and Allison (1998; in press), for example, found by means of

retrospective verbal report that Spanish, French, and German university-level immersion students

engaged in less mental translation and more cognitive processing directly through the immersion

language than did non-immersion program students who were also studying in the immersion

classes.

Likewise, verbal report data could be collected from subjects while they are planning and

writing their essays to identify and describe factors contributing to their relative success in one

writing mode in contrast to another. In this study, for example, it would have been instructive to

get insights as to why thirteen of the students did better in the translation mode than in the direct

writing mode. Correlating the features in a given mode of the writing and the students’ reported

strategy use may provide some understanding of how translation as a writing strategy might

benefit certain students.

Finally, it could be insightful to investigate the results of direct vs. translated writing

modes across a wider range of languages, including, say, Chinese and Arabic. The study would

consider the larger sociocultural context of writing, since how translation is viewed by teachers

and by students in different educational cultures may vary.

57
 Pedagogical Implications

This study has called attention to the various ways that students may use their L1 or

dominant language when preparing and writing a TL essay. First, student writers may think

about the topic in the L1, as the literature review indicated. Secondly, the writers may engage in

mental translation, where the translation is in their minds and not on paper. Thirdly, they may

actually write out an L1/dominant language version and then translate it into the TL. Earlier

research (Jones & Tetroe, 1987; Friedlander, 1990; Kobayashi & Rinnert, 1992) had suggested

that the L1 or dominant language may be best used to plan and organize the writing while the TL

is best for writing on the sentential level. Results from this study would support the notion that

thinking through the L1 or dominant language may be a benefit for some students in certain

writing tasks, and therefore the study seems to lend some support to the value of generating

training or coaching materials dealing with translation strategies.

For example, the materials could provide suggestions for how the systematic use of

translation might serve effectively as a means for organizing ideas and for expressing them in a

concise, lexically acceptable, and grammatically appropriate manner. Given the significant

relationship between ratings for variety in vocabulary and for both ease in direct writing and

ease in translating, it would seem that particularly at an intermediate level, training could focus

on how students go about selecting appropriate vocabulary for a given writing context and the

role that mental and written translation might play in the selection process.

Needless to say, training materials would need to include a series of caveats, indicating

that translation strategies may have a differential effect, depending on the nature of the task (e.g.,

58
an in-class essay with/without time pressure, an in-class essay exam, or an essay as homework),

the topic, the learning style preferences of the writer, and so forth.

59
REFERENCES:

Ali, N. H. (1996). Translation versus direct composition: Effects of first language on second
language Arab writers. Unpublished M.A. thesis. University of Illinois, Urbana-Champaign.

Brooks, A. W. (1996). An examination of native language processing in foreign language

writing. Unpublished doctoral dissertation. Vanderbilt University, Nashville.

Chelala, S. I. (1981). The composing process of two Spanish speakers and the coherence of their
texts: A case study. Unpublished doctoral dissertation. New York University, New York.

Cohen, A. D. (1994). Assessing language ability in the classroom. Boston: Heinle & Heinle.

Cohen, A. D. (1998). Strategies in learning and using a second language. Harlow, Essex:
Longman.

Cohen, A. D., & Allison, K. (1998). Bilingual processing strategies in a university-level

immersion program. Ilha do Desterro, 35, 187-201.

Cohen, A. D., & Allison, K. (In press). Bilingual processing strategies in the social context of an
undergraduate immersion program. In E. Shohamy & J. Walters (Eds.), Perspectives and issues
in educational linguistics. Amsterdam: John Benjamins.

Cohen, A. D., Brooks-Carson, A. W., & Jacobs-Cassuto, M. (In preparation). Direct vs.
translated writing: What students do and the strategies they use. Research report to be submitted
to the Center for Interdisciplinary Studies of Writing, University of Minnesota.
Cumming, A. (1989). Writing expertise and second-language proficiency. Language Learning,
39, 81-141.

Cumming, A. (1990). Metalinguistic and ideational thinking in second language composing.

Written Communication, 7, 482-511.

Friedlander, A. (1990). Composing in English: effects of a first language on writing in English as

a second language. In B. Kroll (Ed.), Second language writing: research insights for the
classroom (pp. 109-125). Cambridge: Cambridge University Press.

Jones, S., & Tetroe, J. (1987). Composing in a second language. In A. Matsuhashi (Ed.), Writing
in real time (pp. 34-57). Norwood, NJ: Ablex.

Kobayashi, H., & Rinnert, C. (1992). Effects of first language on second language writing:
Translation versus direct composition. Language Learning, 42, 183-215.

Lay, N. D. S. (1988). The comforts of the first language in learning to write. Kaleidoscope,
4, 15-18.

60
Qi, D. S. (1998). An inquiry into language-switching in second language composing processes.
The Canadian Modern Language Review, 54, 413-435.

Shohamy, E. (1984). Does the testing method make a difference? The case of reading
comprehension. Language Testing, 1, 147-170.

Raimes, A. (1987). Language proficiency, writing ability, and composing strategies: A study of
ESL college student writers. Language Learning, 37, 439-467.

Swain, M. (1984). Large-scale communicative language testing: A case study. In S. J. Savignon

& M. S. Berns (Eds.), Initiatives in communicative language teaching (pp. 185-201). Reading,
MA: Addison-Wesley.

Swain, M. (1985). Communicative competence: Some roles of comprehensible input and

comprehensive output in its development. In S. Gass & C. Madden (Eds.), Input in second
language acquisition (pp. 235-253). Rowley, MA: Newbury House.

Uzawa, K. (1996). Second language learners' processes of L1 writing, L2 writing, and

translation from L1 to L2. Journal of Second Language Writing, 5, 271-294.

Uzawa, K., & Cumming, A. (1989). Writing strategies in Japanese as a foreign language:
lowering or keeping up the standards. The Canadian Modern Language Review, 46, 178-194.

61
 Appendix A

Participant Questionnaire

1. Student Research Number: _________ 2. Age: _____ 3. Gender: M ___ F ___

4. Number of semesters of university study completed: ____

5. Indicate the language that you consider to be your native language: ____________

6. In the languages that you know, please rate your ability to write a term paper for a course:

1= poor 2=fair 3= good 4=very good 5= excellent

English _____ Spanish (if applicable)________French ____Other languages ______

7. In the languages that you know, please rate your ability to speak informally with friends:

1= poor 2=fair 3= good 4=very good 5= excellent

English______ Spanish (if applicable) ___ French ____ Other languages _________

8. In the languages that you know, please rate your ability to give a talk on an academic topic:

1= poor 2=fair 3= good 4=very good 5= excellent

English_____ Spanish (if applicable) ___French ____ Other languages _________

9. In the languages that you know, please rate your ability to understand a movie with fast-

moving and sometimes technical vocabulary:

1= poor 2=fair 3= good 4=very good 5= excellent

English_____ Spanish (if applicable) ___French ____ Other languages _______

10. If you are willing, please indicate the letter grade you received in English composition:____

11. Please indicate any grades you have received in past French courses:

62
 Appendix B

Writing Strategies Checklist: French Direct Writing

Please rate your responses to the following questions on a scale of 1-5, and please note any

questions or comments you have about the questions or about how you interpret the question

alongside of the item itself.

As you prepared and then wrote the essay directly in French, to what extent did you:

5 = always 4 = often 3 = sometimes 2 = rarely 1 = never

_____ plan out the organization of the essay in advance

_____ plan out the organization of the essay as you went along

_____ pay attention to the connecting words between ideas

_____ attempt to use a wide variety of vocabulary

_____ find yourself thinking in English and translating

_____ find yourself thinking in your dominate language (if not English) and translating

_____ purposely make use of complex grammatical structures

_____ purposely connect shorter sentences into longer, complex sentences

_____ check for subject-verb agreement in the essay

_____ make sure that the negative forms were used properly

_____ check for adjective agreement in gender and number

63
 Writing Strategies Checklist: L1 Writing

Please rate your responses to the following questions on a scale of 1-5, and please note any

questions and/or comments you have about the questions or about how you interpret the

question.

As you prepared and then wrote the version of the essay in your dominant language, to

what extent did you:

5 = always 4 = often 3 = sometimes 2 = rarely 1 = never

____ plan out the organization of the essay in advance

_____ plan out the organization of the essay as you went along

_____ pay attention to the connecting words you chose to link ideas together

_____ attempt to use a wide variety of vocabulary

_____ find yourself thinking in French

_____ find yourself thinking in another language (which? ___________)

_____ purposely make use of complex grammatical structures

_____ purposely connect shorter sentences into longer, complex sentences

_____ check for subject-verb agreement in the essay

_____ make sure that the negative forms were used properly

_____ check for adjective agreement in gender and number

64
 Writing Strategies Checklist: Translated Writing

Please rate your responses to the following questions on a scale of 1-5, and please note any

questions and/or comments you have about the questions or about how you interpret the

question.

When you translated the essay from your dominant language to French, to what extent did

you:

5 = always 4 = often 3 = sometimes 2 = rarely 1 = never

_____ stick to the organization used in the first essay

_____ change the organization somewhat to fit the French language

_____ attempt to find the best connecting words used to link ideas together

_____ make an effort to think how best to express the ideas in French

_____ avoid translating word-for-word

_____ have difficulty finding translation equivalents in French for words in your essay

_____ find yourself using simpler words and structures in French

_____ attempt to use a wide variety of vocabulary

_____ purposely make use of complex grammatical structures

_____ check for subject-verb agreement in the essay

_____ make sure that the negative forms were used properly

_____ check for adjective agreement in gender and number

65
 Writing Strategies Checklist: Comments Sheet

Please compare your experience of writing in your dominant language and then translating

the essay to French vs. writing directly in French:

1. What aspects did you like about writing in your dominant language first and then translating?

What advantages do you see to this approach to producing a well-written essay in French?

2. What aspects did you not like?

3. What aspects did you like about writing directly in French? What do you see as the

advantages of attempting to write directly in French?

4. Do you enjoy writing the kinds of essays you wrote for the study?

Thank you so much for your participation in this study. Do you have any other points you would

like to make about the writing tasks we gave you? (You may write on the back if you need to.)

66
 Appendix C

Follow-Up Strategy Checklist

Based on your responses to the Comments questionnaire, we have put together a summary list of

the possible responses. Please rate your responses to the following questions on a scale of 1-5,

and please note any questions and/or comments you have about the questions or about how you

interpret the question.

When you translated the essay from your dominant language to French, to what extent did

you:

5 = always 4 = often 3 = sometimes 2 = rarely 1 =never

_____ bring BETTER organization to the FRENCH ESSAY than IF you had written it directly

in French.

_____ change the organization of the dominant language essay somewhat to fit the French

language.

_____ think through your ideas more clearly in the dominant language essay than in the French

essay written directly in French.

____ find it easier to write in the dominant language than in the essay directly written in

French.

_____ have a greater number of ideas for the dominant language essay than for the French essay.

_____ have a better vocabulary in the dominant language essay than you did for the French

essay.

67
_____ feel that you had time pressures to complete the translation into French.

_____ find it difficult to translate into the FL.

_____ dislike being required to use translation as a strategy.

_____ purposely simplify your dominant language text in order to translate it into French.

_____ find it easier to write directly in French than to translate.

_____ find it difficult to write directly in French.

_____ feel that thinking in French during the whole process is better than translating.

I feel that writing directly IN French:

5 = always 4 = often 3 = sometimes 2 = rarely 1 =never

____is faster than translating.

____helps to focus on French expression

____helps you to learn the language.

68
 Appendix D

Rating Scales for French Essays

A. EXPRESSION:

freedom from translation effect variety in vocabulary sense of the language

5. There is no negative Vocabulary variety Excellent control

transfer from the is excellent over native-like expression

dominant language

4. There is only slight Vocabulary variety Good control

negative transfer is good over native-like expression

3. There is moderate Vocabulary variety There is fair control

negative transfer is fair over native-like expression

2. There is extensive Vocabulary variety There is weak control

negative transfer is weak over native-like expression

1. There is very extensive Vocabulary variety There is little or no control

negative transfer is poor over native-like expression

B. TRANSITIONS

organizational structure clarity of points smoothness of

Excellent statement Excellent connections

structure of points across points

4. Good organizational Good statement Good connections

69
 structure of points across points

3. Acceptable Fair statement Fair use of connectors

organizational structure of points

2. Weak organizational Weak statement Weak use of connectors

structure of points

1. Poor organizational Poor statement Poor use of connectors

structure of points

C. CLAUSES

use of subordination use of relative pronouns

5. Excellent control Excellent use of

over subordination relative pronouns

4. Good control Good use of

over subordination relative pronouns

3. Fair control Fair use of

over subordination relative pronouns

2. Weak ability Weak facility with

to subordinate relative pronouns

1. Poor control over Poor control over

subordination relative pronouns

70
D. GRAMMAR

Prepositions/ Noun/ Verbs

Partitive Articles Adjective Agreement

5. Excellent control Excellent control Excellent control

4. Good control Good control Good control

3. Fair control Fair control Fair control

2. Weak control Weak control Weak control

1. Poor control Poor control Poor control

71