Sie sind auf Seite 1von 22

Metacognitive regulation of writing in the

classroom

Linda Allal
University of Geneva

ABSTRACT

A situated cognition approach to writing instruction provides a fruitful con-


text for studying the operations of metacognitive regulation intervening in
text roduction. Our research focuses on inferences about metacognitive regu-
lation derived from the analysis of text transformations introduced by
student writers between successive versions, of their texts (notes, draft final
version). The results of three studies conducted with sixth-grade students are
reported here. Study .1 shows that the mastery of basic written language skills
affects the students' regulation strategies in three ways.A higher level of mas-
tery allows anticipation of a wider range of transformations, encourages
monitoring of problems of text organization and entails more complex means
of adjustment. The different phases of text production appear to induce vari-
ations in regulation strategies: when students are transforming their notes
into a first draft (Study 1), they tend to use higher-level strategies (more op-
tional transformations, more complex adjustments affecting text structure)
than those used when they are revising their drafts (Studies 2 and 3). Con-
cerning the effects of instruction, the data from Studies 2 and 3 suggest that a
sociocognitive approach, emphasizing writing in a context of communica-
tion, tends to encourage strategies of regulation focused largely on ortho-
graphic and textual conventions, rather than on semantic content. In-depth
study of peer interactions in Study 3 indicates that for a given set of instruc-
tional conditions (such as dyadic co-production), there can be a wide range
of ways in which students actually function. This variation could explain in
part why the overall effect of an interactive instructional approach is weaker
than might be expected. The implications of our findings for future research
and classroom practice are discussed.

1 INTRODUCTION

The search for continuity between learning in the classroom and knowledge ac-
quisition and use in 'real-life' contexts, outside school, is a long-standing, and largely
unresolved, educational problem. More than half a century ago, Dewey (1938/
1963) called for an 'experiential continuum' linking the activities that take
place in schools with their surrounding social and cultural contexts (p. 28). It is
the absence of such a link, Dewey believed, that explains the widespread
difficulty that people encounter when attempting to use the concepts and skills
acquired in school.
146 ALLAL

If exactly the same conditions recurred as those under which it [a concept or skill]
was acquired, it would also recur and be available. But it was segregated when it
was acquired and hence is so disconnected from the rest of experience that it is
not available under the actual conditions of life....
Perhaps the greatest of all pedagogical fallacies is the notion that a person learns
only the particular thing he is studying at the time. Collateral learning may be
more important than the spelling lesson or lesson in history or geography that is
learned (p. 48).

Brown, Collins and Duguid (1989) present a similar perspective in their article on
`situated cognition'. After criticising the still prevailing tendency of schools
to treat knowledge acquisition in a compartmentalised and decontextualised man-
ner, they summarise their position as follows:

The activity in which knowledge is developed and deployed...is not separable


from or ancillary to learning and cognition. Nor is it neutral. Rather, it is an integral
part of what is learned.... Learning and cognition, it is now possible to argue, are
fundamentally situated. (p. 32)

Studies of learning activities in out-of-school settings (Rogoff, 1990; Rogoff & Lave,
1984) have helped to clarify the ways in which context becomes part of the
content of learning.As Rogoff (1982) has stated, 'context is not so much a set of
stimuli that impinge upon a person as it is a web of relations interwoven to form
the fabric of meaning' (p. 149). The learner's activity is always embedded in a
context and, at the same time, it contributes to creating the context.This phenom-
enon has been recognised for some time in sociolinguistic research, such as
that of Cook-Gumperz (1977) showing how 'situated meaning' is constructed
interactively in everyday settings (p.107).

The instructional model of `cognitive apprenticeship' has been proposed to de-


scribe school learning situations that simulate authentic, out-of-school contexts
of knowledge acquisition and use (Brown et al., 1989; Rogoff, 1990). This model
emphasises the elaboration of contextualised competencies, which incorporate
appropriate use of tools and forms of social interaction found in corresponding
cultural practices. A simplified example would he children learning to write texts
for a school newspaper using a word processor and working in collaborative
editorial groups with other students. An important theoretical foundation of the
model lies in the work by Vygotsky (1930-35/1978) on the social mediation of
learning and, in particular, in the idea that interaction can open a 'zone of proxi-
mal development' within which the child learns to accomplish new tasks. The
term 'scaffolding' (Greenfield, 1984; Wood, Bruner & Ross, 1976) is used to desig-
nate various forms of interactive guidance - teacher coaching, peer collaboration,
appropriation of tools, interaction with computer environments - that foster the
learner's progressive construction of new skills and strategies, as well as the in-
ternalisation of corresponding mechanisms of self-regulation. Certain implications
of these ideas have been developed by Collins (1991), Moll (1990) and Salomon,
Perkins and Globerson (1991), among others.
METACOGNITIVE REGULATION OF WRITING IN THE CLASSROOM 147

Although self-regulation is an important feature of all learning, it is generally sub-


ordinated to systematic external guidance in traditional classrooms, where a con-
stant stream of teacher instructions regulates the learner's activity (e.g. 'Put away
your math sheets. Get out your geography book and turn to page 25. If you've
forgotten your book, look at your neighbour's...). In out-of-school settings of knowl-
edge production and use, social interactions with peers and with mentors are
common and take on complex forms. Active participation in these interactional
processes is important for individuals to acquire the ability to regulate - initiate,
orient, assess, adjust - their own cognitive activity in a reflective, goal-oriented
manner. School activities developed in a perspective of situated cognition there-
fore need to include a concern for 'situated metacognition'. This means an ap-
prenticeship model geared toward promoting the acquisition of self-regulation
mechanisms that are adapted to the contextual constraints of different tasks, that
allow growing expertise in the integration of multiple skills needed to accom-
plish a task and that function with increasing autonomy and intentionality.

This chapter is concerned with the production of written texts in authentic con-
texts of communication and, in particular, with the processes of metacognitive
regulation intervening in text production. After a brief summary of selected re-
search on writing and metacognition, we present three studies conducted with
sixth-grade students (age 11-12) in public elementary schools of the canton of
Geneva.

2 WRITING IN CONTEXT

Resnick (1990) has defined several features of school environments which favour
`literacy apprenticeships' comparable to authentic literary practices outside of
school:

Children work to produce a product that will be used by others...; they work col-
laboratively, but under conditions in which individuals are held responsible for their
work; they use tools and apparatus appropriate to the problem; they read and cri-
tique each other's writing; they are called upon to elaborate and defend their own
work until it reaches a community standard. (p.183)

These basic concerns of situated cognition are present, more or less explicitly, in
many contemporary approaches to writing instruction (e.g. Bain & Schneuwly,
1993; Englert, Raphael & Anderson, 1992; Graves, 1983; Milian Gubern, 1996;
Needels & Knapp, 1994; Slavin, Stevens & Madden, 1988).

In contextualised writing activities, rhetorical considerations which concern the


relationships linking writer, audience and topic take on an importance not found
in traditional academic tasks. Studies of writing in professional settings have shown
that the ways of dealing with rhetorical factors vary depending on the writer's
position and relationships to co-workers or supervisors (Brown & Hemdl, 1986;
Odell & Goswami, 1982). In attempting to> create authentic communication con-
texts in the classroom, it is necessaryto use instructional strategies that
i l
148 ALLAL

as convincingly as possible, the rhetorical relationships that give writing meaning


in real-life situations. Some approaches give priority to the production of texts
addressed to well-defined audiences outside the classroom, and in particular to
audiences which are likely to respond (Bain & Schneuwly, 1993; Florio-Ruane &
Lensmire, 1989). Other approaches put more emphasis on voice: Students are
encouraged to write about topics of their own choice and to refine a style of
expression reflecting their personal and cultural identity, as a member of an eth-
nic group, or as a woman, for example. Still another approach adopts the idea of
the classroom as a 'writing community'. In this case, the primary audience is com-
posed of the class members, who read and discuss one another's texts, and the
teacher or other mentor, who actively participates in writing, i.e. models
writing in front of the students, and who writes extensive reactions to their
texts (Heath & Branscombe, 1989).
In addition to rhetorical concerns, contextualised writing implies creating links
with other language skills and with activities in other subject-matter areas. Rather
than restricting writing instruction to the classical literary forms of composition
(e.g. narrative, interpretative essay, etc.), a variety of genres and types of texts,
grounded in diverse cultural realities, are dealt with, e.g. letter-to-the-editor com-
menting on a recent political event, publicity tier a book or movie intended for
young people, travel guide for visitors to the local area, leaflet for a scientific
exhibit, recipe for a multi-ethnic cookbook. Before producing any given type of
text, students read and analyse examples of 'authentic texts' which illustrate a
range of existing practices in social contexts outside school (Bain, 1987). 'Read-
ing-to-write' (Flower, Stein,Ackerman, Karatz, McCormick & Peck, 1990) is a way of
acquiring information and generating ideas to be used in writing, as well as a
means of discovering how different types of texts are constructed. In a comple-
mentary manner, 'writing-to-read', e.g. preparing an outline, or a rough draft of a
text, provides a functional reference orienting subsequent reading toward more
clearly targeted goals, such as search for in-depth information on a particular as-
pect of a topic, or expansion of the content to be covered.

3 SELF-REGULATION IN 'WRITING
There is increasing research on general processes of self-regulated learning
(Boekaerts, 1997; Schur* & Zimmerman, 1994) and on the specific types of regu-
lation that intervene in writing instruction (Schneuwly & Bain, 1993;Wegmuller,
1993). Learner self-regulation is affected by various aspects of an instructional
setting. We distinguish the structure of the activity (goals to be achieved,
content to be dealt with, task constraints, etc.), the ways in which the teacher
intervenes, the forms of peer interaction that are allowed or encouraged, the
types of tools that are provided or constructed during the activity (Allal, 1993).
Procedures of formative assessment designed to foster reciprocal peer assessment
and self-assessment also provide a framework for the development of self-
regulated learning (Allal & Michel, 1993; Nunziati, 1990).
Self-regulation in writing encompasses both cognitive and metacognitive regu-
lations.Although it is not easy to establish a clear-cut boundary between the two
levels of regulation, we find the following definitions heuristically useful. Cogni-
tive regulations are involved in the construction of the conceptual, linguistic and
METACOGNITIVE REGULATION OF WRITING IN THE CLASSROOM 149

metalinguistic knowledge required for writing and they intervene implicitly


throughout the process of text production. Metalinguistic knowledge is knowl-
edge about language, used either implicitly or explicitly during comprehension
and production activities (Gombert, 1990). For example, knowledge about rules
of noun-adjective agreement is likely to enhance the writer's ability to make
appropriate corrections when rereading a draft. Metalinguistic activity is obvi-
ously central to increasing expertise in writing, but remains (as proposed by
Gombert, p.27) a sub-domain of metacognition. Metacognitive regulations entail
the active management of cognitive resources with respect to a given goal and,
in the case of writing, the use of explicit strategies for articulating different bod-
ies of knowledge (conceptual, linguistic; metalinguistic) needed for text produc-
tion. Metacognitive regulations also assure the coordination of these bodies of
knowledge with contextual factors or with affective and social processes which
influence the way in which the writer carries out text production. For
example, in classroom writing, teacher expectations and evaluation practices
often affect the tendency of the student writer to take risks (e.g. write a long
text with a relatively complex structure) or to play it safe (e.g. avoid using
words which present orthographic difficulties). In the rest of this chapter, the
term 'self-regulation' will refer to metacognitive operations affecting writing.
Research on metacognition, and in particular the work by Ann Brown and her
colleagues (Brown, 1978; Brown 1987; Brown & Palinscar, 1982; Campione &
Brown, 1990), has led us to define metacognitive regulation as an interface which
assures the coordinated functioning of two other components of the subject's
cognitive activity: the representational network and the production processes
mobilised to accomplish the task (Allal & Saada-Robert, 1992). Metacognitive
regulations have a dual function:They orient the production processes in a man-
ner compatible with the subject's representations, and they modify his represen-
tations to take into account the outcomes of production processes.
We have identified three operations of metacognitive regulation, defined as
follows (Altai & Saada-Robert, 1992).
1. Anticipation. This operation reflects the transposition of the subject's repre-
sentations of the task and the context into goal orientations, defined with vary-
ing degrees of precision and intentionality.
2 Monitoring. This operation entails the comparison of the present state of ad-
vancement with respect to the task to an anticipated goal-state; in complex
tasks, the comparison concerns multiple aspects of the present state and a
multi-faceted goal-state.
3. Adjustment. This operation aims at reducing the discrepancy between the
present state and the goal-state. If the feedback from the monitoring
operation is negative (i.e. the goal-state is not attained, or progress in that
direction is unsatisfactory), an adjustment is introduced in the production
processes. If feedback is positive, the production processes continue without
reorientation, or cease because the goal is fully reached.
For any moderately complex task, such as writing, which requires the coordina-
tion of several production processes, the regulation of some processes is largely
automatic (i.e. with practice become automated), whereas the regulation of other
processes requires active, intentional cognitive resource management (Allal &
Saada-Robert, 1992; Iran-Nejad, 1990).
150 ALLAL

Models of writing generally include components of metacognitive regulation even


though they may not necessarily be designated by this term. For example, in the
well-known Hayes and Flower (1980) model, the components of planning, moni-
toring and reviewing regulate the central production process (translating). Re-
search has begun to show how metacognitive regulation intervenes in successive
phases of text production (Fayol, 1991). Studies of text planning and revision are
particularly useful for understanding the multiple facets of self-regulation in writ-
ing, as reflected in pre-textual adjustments of plans prior to writing, in changes
introduced while composing and in modifications carried out once a draft has
been completed (Witte, 1985). Experimental studies of text revision (Hayes, Flower,
Schriver, Stratman & Carey, 1987; Piolat & Roussey, 1991) have led to more de-
tailed models of the mechanisms (task definition, evaluation, detection, diagnosis)
regulating strategy selection. Research with school-related tasks has shown that
children's metacognitive awareness of revision strategies can be enhanced by
instructional aides (Bereiter & Scardamalia, 1987), by peer feedback (Englert, 1992;
Milian Gubern,1996) and by computer-based guidance (Daiute & Kruidenier, 1985;
Zellermayer, Salomon, Globerson & Givon, 1992).
Studies of planning and revision in writing rely frequently on methods of ver-
balisation:`think-aloud' protocols recorded during the different phases of writing,
interviews conducted before and/or after a writing task, transcriptions of peer
interactions during joint planning or revision. Since all forms of verbalisation
entail well-known biases (Ericsson & Simon, 1980), there is a need to develop
alternative methods which provide a means of triangulation with respect to
verbal self-report data.The approach presented in this chapter is to base inferences
about processes of regulation on an analysis of the transformations introduced
by writers between successive versions of their texts.Although this approach
provides no information on processes of pre-textual revision taking place in the
writer's head but leaving no traces on paper, it does allow inferences about
other aspects of revision. On-line regulations during composition can be partly
inferred from changes between an outline, a plan or a set of notes and an initial
draft of a text. Regulations aimed at improving an existing product can be largely
inferred from changes between one or more drafts and the final version of the
text.Text transformations provide an indication of only the most salient aspects of
metacognitive regulation. They obviously do not allow identification of implicit
cognitive and linguistic regulations that are fully integrated within on-going
production processes.
We will now examine three studies in which the text transformations carried
out by sixth-grade writers (age 11-12) are analysed and interpreted in terms of
underlying processes of metacognitive regulation. The presentation of the first
study includes a description and illustration of the system used for coding the
transformations.

4 STUDY 1. TEXT TRANSFORMATIONS AND MASTERY OF BASIC


SKILLS

This study examines the text transformations carried out by four sixth-grade
girls
METAGOGNITIVE REGULATION OF WRITING IN THE CLASSROOM 151

to be displayed on posters in a school exhibit on 'Childhood inAfrica'.The


school is located in an urban neighbourhood characterised by a relatively large
percentage of families of high socio-professional status.The analysis presented
here concerns the text transformations intervening between the notes taken by
the children from various reference books and the initial draft of their texts.
The study was designed as an exploratory case study of individual differences in
students' writing strategies (for more details see Allal & Michel,1993; Allal, Rouiller &
SaadaRobert, 1995).
Variations in text transformations reflecting metacognitive regulation of the
writing process are likely to be influenced by the writer's mastery of basic writ-
ten language skills. Research on individual differences in writing (McCutchen,
Covill, Hoyne & Mildes, 1994) has shown that mastery of these skills is likely to
reduce working memory load and thereby facilitate higher level processes of plan-
ning and revision. For this case study,we selected subjects with contrasting achieve-
ment levels, as indicated by their first term grades in French 'basic skills' (spelling,
grammar, conjugation, vocabulary). The two 'high-achieving' subjects had grades
of 6 (highest grade on a six-point scale); the two 'middle-achieving' subjects grades
of 4 (minimum grade required to continue secondary studies in the academic
section). We decided not to study children with lower achievement levels be-
cause of their lack of fluency in spoken French and/or other problems precluding
their implication in the proposed activity. In the presentation of the results, the
high-achieving subjects are identified as Eva and Fanny, and the middle-achieving
subjects as Maude and Sonia (fictitious names).

4.1 A system for coding text transformations

We have developed a system for coding transformations occurring between suc-


cessive drafts of a text (notes or outline, initial draft, subsequent revised drafts,
final version). The basic principles of the system remain the same in each study
presented in this chapter.The units of analysis are defined as all observable differ-
ences between two versions of a text. Each unit is identified and coded along four
main dimensions:
1. the level of language affected by the transformation: word, group,
text; t
2. the type of transformation: addition, deletion, substitution, rearrangement;
3. the object of the transformation: semantics (lexical variations, changes of mean-
ing), text organisation (primarily operations of segmentation, connection, co-
hesion), spelling (both lexical and grammatical aspects);
4. the relationship to language conventions: conventional transformation correctly
or incorrectly carried out, optional transformation not required by language
conventions.
For this fourth dimension, language convention is considered in a restrictive sense
corresponding to rules of spelling, syntax, and punctuation for which no varia-
tion is accepted by authoritative references (Le Robert dictionary, Le bon
usage de Grvisse).At the textual level, two major types of conventions are taken
into account: correct signs of segmentation between sentences (capitalisation
and final punctuation) and correct anaphoric referencing (The children...;
152 ALLAL

tional transformations therefore include changes which may be considered as


coherent with language conventions, but are not strictly dictated by rules of us-
age. Two examples will illustrate the coding system:

Example 1. A student inserts the word 'African' in the expression 'the children'
This transformation is coded as follows:
level of language: group,
type of transformation: addition,
object: semantics,
relationship to language conventions: optional.

Example 2. A student changes a verb inflection to make it agree with the


subject (in the expression it marchais rapidement, the s of marchais is changed
to a t). This transformation is coded as follows:
level of language: word,
type of transformation: substitution,
object: spelling,
relationship to language conventions: conventional and correct.

Coding of the four dimensions was based on a detailed guide which assured a
high degree of interceder reliability (83-95% depending on the data set).
The presentation of the results of Study 1 will focus on the differences in the
students' transformations that appear to be linked to their mastery of basic skills.
The differences between high-achieving and middle-achieving students arc inter-
preted in terms of the probable underlying operations of metacognitive regula-
tion. Although each transformation dimension reflects the interplay among the
three operations of regulation (anticipation, monitoring, adjustment), our inter-
pretations focus on the operation(s) that can be most directly inferred from the
available data.

42 Conventional vs. optional transformations

An analysis of the relative proportions of conventional vs. optional transforma-


tions helps us to understand the way our subjects construe their role as author
of a text. Is this role interpreted narrowly, as is the case for many beginning
writers who aim at producing a correctly written text which is essentially a
carefully executed transcription of passages copied from reference books? Or do
they interpret their role in a larger perspective, which includes concerns of
conventional correction, but emphasises the author's license to organise and
compose a text as thought best? The subject's representation of what writing is, of
what an author is supposed to do, provides a general orientation for the
metacognitive operation of anticipation that guides composing activity.
Our data show that all four subjects share a common representation of the
general requirements of the writing task. Each student produced a sizeable number
of transformations (43 to 61), and among these a substantial percentage of op-
tional transformations (at least 42%), thereby demonstrating her comprehension
that an author's role entails the selection and organisation of information, and not
METACOGNITIVE REGULATION OF WRITING IN THE CLASSROOM 153

simply the production of correct sentences.There is, however, as shown in Table


1, a significant difference between high-achieving and middle-achieving subjects
with respect to the relative frequencies of conventional vs. optional transforma-
tions (p < .0001).The high achievers carry out considerably more optional trans-
formations (81.7%), than do the middle achievers (54.7%). Moreover, the conven-
tional transformations carried out by the high achievers are rarely incorrect (3.7%),
compared to those carried Out by the middle achievers (17.9%).

Table I Conventional vs. optional transformations, by student achievement


status.

Achievement Conventional Optional Total'


Status Incorrect Correct

High-achieving 3.7 14.7 81.7 100.0


(109)
Middle-achieving 17.9 27.4 54.7 100.0
(95)
Total 10.3 20.0 (i9.1 100.0
(204)

'Percentages are calculated by line on the number of transformations given between


parentheses.
18.37, d.f. = 2, p < .0001.
)(2=

At least two interpretations of these findings are tenable, and potentially comple-
mentary. The first is linked to the problem of cognitive load during on-line process-
ing. It is likely that high-achieving students are able to correct errors fairly auto-
matically, while composing their text, and therefore devote greater attention and
regulating capacity to the formulation and execution of optional
transformations. A second explanation would be that the high-achieving students
have more cognitive resources for the task at hand, i.e. a more detailed and
differentiated representation of how informative texts can be structured, which
allows them to plan overall organisational changes, rather than simply proceed on
a sentence-by-sentence basis.

4.3 Object of transformation and level of language

Preliminary analyses showed that the joint consideration of object of transforma-


tion and language level allowed more interesting interpretations than separate
analyses of each dimension. The distribution of the objects transformed at each
language level indicates the types of units on which the writers' attention is fo-
cused. This gives an idea of the aims that guide the writer's monitoring of the
drafting process. To what extent is she concerned with improving or changing
the semantic content of the text, as compared to that of her notes? At what lan-
guage level are the changes carried out (expansion of nominal groups, addition of
explanatory material from a new source, for example)? Is she searching for ways
of organising her text that differ from the existing organisation of her notes? Is
she attentive to correct spelling when drafting her text? Although the transforma-
tion data do not provide detailed answers to these questions, they do suggest
some interesting and plausible interpretations.
In Figure 1, the transformations carried out by each subject are classified with
respect to two crossed dimensions: object of transformation and level of language
affected by the transformation. For each subject, two dominant categories, and
several secondary categories are identified by the slashed lines.

The two high-achieving students have one dominant category in common,


transformations affecting organisation at the level of the text, whereas their sec-
ond major category differs: For Eva, it is organisation at the sentence level, but for
Fanny it is the spelling of words.A similar pattern is found for the middle-achiev-
ing students.They both have one dominant concern (the spelling of words), hut
each has a second focal point: semantic content at the sentence level, for Maude;
organisation of the sentence, for Sonia.

Figure 1 Objects of transformation and language level: individual Profiles.

In summary, high and middle-achieving students are differentiated primarily by


the fact that the former share a common concern for text organisation, while the
latter have a common preoccupation with spelling. Globally, this finding is coher-
ent with the earlier analysis of conventional vs. optional transformations. It
should be stressed, however, that achievement level is not linked to systematic,
generalised differences.The transformation patterns in Figure 1 show marked
individual differences, reflecting the specific aims of each child, her
representation of what is important in the writing task, and the way in which she
monitors her drafting activity.
METACOGNITIVE REGULATION OF WRITING IN THE. CLASSROOM 155

4.4 Type of transformation

Type of transformation concerns the means used to carry out the transformation
(addition, deletion, substitution, or rearrangement). Analysis of this dimension
allows us to specify the adjustments that result from the monitoring
operations of the regulation process. How broad or narrow is the repertoire of
transformation tools used by sixth graders? With what degree of flexibility are
the tools used?
In Table 2 the data for type of transformation are grouped into two categories:
simple transformations (addition or deletion of an element) and more complex
transformations (substitution of one element for another, or rearrangement such
as transfer of elements from one location to another).These categories differ sig-
nificantly according to student achievement status (p < .0157). For high-achieving
students, complex transformations are as frequent as simple ones, whereas for
middle-achieving students, two-thirds of their transformations are simple and only
one-third complex.The individual data for the two types of transformations within
each category show that additions are more frequent than deletions (except for
Maude, for whom their frequency is equal), and that substitutions are much more
frequent than rearrangements.

Table 2 Type of transformation, by student achievement status.


Achievement Type of transformation' Total!'
Status Simple Complex

High-achieving 49.5 50.5 100.0


(109)
Middle-achieving 66.3 33.7 100.0
(95)
Total 57.4 42.6 100.0
(204)

a Simple transformations = addition, deletion.


Complex transformations = substitution,rearrangement.
bPercentages are calculated by line on the number of transformations given between
parentheses
x2= 6.53, d.f. = 1, p < .0157

In summary, our data show that all subjects make minimal use all four types of
transformations. However, despite this common repertoire of tools for making
adjustments in their drafts, the high-achieving students show greater mobility in
their deployment of these tools. This finding, combined with the results for op-
tional vs. conventional transformations, suggests the following profiles of student
functioning: High-achieving pupils vary the type of transformation so as to
attain a wide range of optional transformations,whereas middle-achieving
students carry out a relatively more conventional transformations using
predominantly simple means.The more expert functioning of the high-achieving
st dents might be e
156 ALLAL

plained by several factors: Greater automation of the simpler tools could allow for
increased use of the complex ones; greater ease in the cognitive 'management' of
the multiple task requirements would also favour flexible use of varied tools.

4.5 Conclusions drawn from Study I

The main findings regarding the text transformations carried out by sixth-grade
students when drafting short, informative texts based on notes taken from several
reference books can be summarised as follows.Analysis of their transformations
shows important similarities among the four subjects:All carry out a majority of
optional (rather than conventional) transformations, their transformations affect
all four levels of language (word, group, sentence, text), and all four means of
transformation (addition, deletion, substitution, rearrangement) are used. Never-
theless, the high-achieving students can be distinguished from the middle-achiev-
ing students in three ways: (1) they carry out a relatively larger number of op-
tional transformations; (2) they show greater concern for text organisation;
and (3) they make relatively greater use of complex means of carrying out
transformations (substitution, rearrangement), as compared to simple means
(addition, deletion).
These findings suggest that mastery of basic skills affects all three
operations of metacognitive regulation: anticipation of transformations not
dictated by convention, monitoring and adjustment at a textual level, use of more
varied means of adjustment. Our interpretation of these findings is that the high-
achieving students' transformations reflect more powerful strategies of
metacognitive regulation resulting from increased automation of sub-processes
linked to mastery of basic skills.

5 STUDY 2. EFFECTS OF INSTRUCTION ON TEXT REVISION


The second study reported here examines the effects of two instructional ap-
proaches on sixth graders' production and revision of narrative texts.' A year-long
field study was conducted in order to evaluate the effectiveness of an integrated
socio-cognitive approach to writing (IS), as compared to that of a traditional
componential skills approach (CS).The CS approach emphasises basic skill acqui-
sition through specific activities focused on the various components of writing
(vocabulary development, sentence construction, grammar, spelling, punctuation,
etc.) using the exercises proposed in the official curricular material or material
available in widely used teacher handbooks. The IS approach aims at integrating
the acquisition of basic skills - in particular the morphological aspects of spelling
which are quite difficult to master in French - within situations of text produc-
tion oriented toward a meaningful communication goal.
The IS activities which we developed were based on research in three areas:
1. The work on situated cognition (Brown, Collins & Duguid, 1989) which em-
phasizes knowledge acquisition in instructional settings that are similar to the
'real-life' contexts in which knowledge is subsequently to be used.
METACOGNITIVE REGULATION OF WRITING IN THE CLASSROOM 157

2. Models of writing instruction which draw on Vygotsky's conception of social


mediation of learning (Englert, Raphael &Anderson, 1992; Needels & Knapp,
1994).
3. Research on the `didactics' of French language instruction (Bain & Schneuwly,
1993, Pat', 1992).
The IS approach involved eight activities of text production, each of which was
designed to include two basic skill objectives. For example, in Activity 3, each
student wrote a narrative text to tell a Canadian pen pal about a Genevan
historical event (Escalade).The basic skill objectives concerned verb inflections for
the imperfect and past definite tenses (third person), and complex cases of
subject-verb agreement (e.g. intervening relative clause, subject-verb inversion).
Figure 2 shows the phases of the text production sequence that was followed
for each activity and the principles on which each phase was based. During the
drafting and revision of each text, several forms of scaffolding were proposed:
student-teacher dialogues guided by a sequence of progressively focused
questions, structures of peer interaction (e.g. reciprocal revision), use of tools
constructed in class (e.g. spelling checklist with individualised examples drawn
from the student's texts). Provision of these means of support for regulating
writing was designed to foster progressive internalisation of metacognitive
strategies of regulation.

5.1 Analysis of text transformations

The data presented here pertain to a sample of 135 students attending eight sixth-
grade classes (four of which followed the IS approach and the other four the CS
approach). The classes were located in schools characterised by relatively high
percentages of children from families of middle and lower socio-professional sta-
tus, as compared to the overall population of the Geneva public school system.
Periodic observations in class and discussions with the teachers allowed us to
characterise the degree of implementation of the proposed IS approach (Allal,
1996). Analysis showed that the teachers encouraged peer interactions during
writing, but sometimes failed to exploit the implications of writing a text for an
outside audience.The links between the writing task and the basic skill objectives
were generally well established, but the teachers did not always encourage suffi-
cient use of available tools (whether standard references, such as the dictionary,
or individualised spelling checklists). Teachers often seemed to assume that if
these tools were present, students would make use of them, without realising that
regular use needs to be fostered by active teacher intervention.
At the end of the school year, the students participated in an activity devel-
oped by Rouiller (199G,1998), which allowed the appraisal of several aspects of
their writing. Each student wrote a narrative text to accompany a set of 10 illustra-
tions showing the story of a postman whose keys are stolen by a chicken
while he and his cat are napping. After the students had revised their initial
drafts, the typed final versions of the texts, accompanied by the illustrations,
were formatted as storybooks to be given to second graders attending the same
school.The presentation here focuses on the transformations the students carried
out while revising their texts.The transformations were coded according to the
system described in Study 1.Analysis of the dimensions of transformation provides
i di i
158

whether the type of instruction students received had an effect on their strategies
of revision.

Figure 2 Phases of the text production sequence and underlying principles


of the Integrated sociocognitive approach.

Phases Prromplllal

I. Before text production

Define the communication setting: Create a functional communication situation b)


aim and audience, defining the rhetorical constraints and tilt
type of text (e.g. description, narration, characteristics of the intended text.
dialogue, cooking recipe).

Plan the content to be dealt with: By preliminary content planning, decrease the
Collective discussion enumerating ideas cognitive load required for the generation of
and lexical items likely to be used in writ- ideas during writing, and thereby facilitate the
ing the text. student's focus on text organization and or
basic skills of spelling, punctuation, etc..

Present the targeted basic skill objectives Establish reference criteria for the regulation:
and specify the corresponding reference intervening during text production.
tools.

2. During text production


Foster the development of spelling skills that
Implement different means of regulating are integrated in the processes of drafting anc
writing (during drafting and revision), text revision.
based on the student's interactions:
Enhance (across 8 successive production situ
with the teacher, ations) the progressive internalization of self
with other students, regulation strategies for drafting and revision
with instructional tools.

3. After text production

Carry out various text-based activities: Use deferred and detached follow-up activities
analysis, classification and reflection to consolidate students' basic skills; differenti
based on excerpts from student texts and ate the tasks accdifferenti-student needs.
on other supplementary material (spe-
cific exercises if needed).

Adapted from AAllal (1997).


METACOGNITIVE REGULATION OF WRITING IN THE CLASSROOM 159

On the whole, the effect of instructional approach on text transformations was


much less pronounced than we had expected. No significant differences between
the IS and CS approaches were found for the relationship of transformations to
language conventions and for type of transformation. Under both conditions, the
students carried Out a relatively small percentage of optional transformations (ap-
proximately 25%); among the large majority of transformations concerned with
the respect of language conventions (75%), approximately 65% were carried out
correctly and 10% incorrectly. Regarding the type of operations used to carry out
the transformations, the students in both IS and CS classes used relatively simple
operations (addition and substitution) for the majority of their transformations
(approximately 65%); more complex transformations consisted primarily in sub-
stitutions (around 35%) and only rarely in rearrangements.

With respect to the object of transformation a modest but statistically signifi-


cant difference was found between the two instructional approaches.The data in
Table 3 show that the integrated socio-cognitive approach tended to encourage
relatively more transformations linked to spelling and to text organisation,whereas
the componential skills approach led to a relatively larger percentage of semantic
revisions (p < .0000). It should be noted that there was no significant difference
between the degree of conventional correctness of the initial drafts produced
in IS and CS classes. The greater attention paid to spelling by IS students can not
therefore be attributed to the fact that their drafts contained more errors. It ap-
pears that the instruction provided in IS classes led students to acquire a greater
awareness of the conventions (both textual and orthographic) that enhance text
comprehension. It is also worth noting that two thirds of the IS spelling transfor-
mations concerned features of grammatical orthography (verb inflections, noun
group agreements, punctuation added to already segmented sentences) which
contribute to text organisation. This suggests that IS instruction may help stu-
dents make the transition from a 'knowledge telling' to a 'knowledge transform-
ing' strategy of text production and revision Bereiter & Scardamalia, 1987).

Table 3 Objects of transformation, by instructional approach".

Object CS IS

Semantics 18.8 11.7


Text Organization 32.7 34.0
Spelling 48.6 54.3

a Percentages calculated on total number of transformations by instructional


approsocio-cognitiveential skills) approach: n = 943 transformations.
IS (integrated sociocognitive) approach: n = 1047 transformations.
x2 = 20.204, d.f = 2, p <.0000.
160 ALLAL

5.2 Conclusions drawn from Study 2

Our analysis of the effects of instruction on text transformations allows the fol-
lowing interpretations regarding the operations of metacognitive regulation that
characterise students' revisions.
1. The integrated socio-cognitive approach to instruction, as put into practice in
our field study, did not have a discernible effect on the students' conception of
the writer's task when revising a text. Under both instructional conditions,
students seemed to consider revision primarily as a proofreading task. They
corrected their texts but made few optional transformations reflecting author's
voice and posture. Moreover, both IS and CS students used relatively simple
means to carry out their transformations (primarily additions).These findings
suggest that instruction had little effect on the metacognitive operation of
anticipation and on the ways in which adjustments were carried out.
2. The integrated socio-cognitive approach to instruction had, on the other
hand, a limited but significant effect on the selection of the components of
the text to transform. Relatively greater importance was given to
transformations of spelling and of text organisation than under the CS
approach. This suggests that IS instruction, with its emphasis on producing
texts in a communication context, enhanced students' monitoring of the features
that facilitate text comprehension.
3. Comparing the results for Studies 1 and 2, we can note one important differ-
ence.The transformation of notes into a first draft (Study 1) elicited a higher
level of metacognitive regulation (more optional transformations, more com-
plex means of transformation) than the revision of a draft (Study 2).This sug-
gests that students tend to use different strategies of metacognitive regulation
in different phases of text production.

6 STUDY 3: DYADIC INTERACTIONS AND TEXT REVISION


This study was conducted by Rouiller (1996, 1998) in order to examine the ef-
fects of a dyadic cooperative structure on students' strategies of revising narrative
texts. In contrast with the two preceding studies, Rouiller's research provides an
in-depth look at the relationship between students' verbal exchanges and the trans-
formations they carry out. Her experimentation was designed on the basis of
previous research on peer collaboration in writing, and in particular on the find-
ings showing that: (1) dyadic structures foster sustained thought and activity on
the part of each student (Higgins, Flower & Petraglia, 1992); (2) working on a
common task enhances the quality of collaboration (DiPardo & Freedman, 1988);
(3) when students individually compose different parts of a joint text, this in-
creases critical distance and favours collaborative revision (Zammuner, 1983); (4)
joint planning can increase the involvement of each student in the subsequent
phase of collaborative revision, which implies coordination of viewpoints to at-
tain a common goal (Saunders, 1989).
METACOGNITIVE REGULATION OF WRITING IN THE CLASSROOM 161

6 1 Research design

The experimentation involved two writing sequences aimed at the production of


narrative texts. The first production sequence (P1) allowed comparison of the
effects of two experimental conditions: dyadic co-production (two students pro-
duced a joint text) vs. individual production. The second production sequence
(P2) allowed verification of the persistence of the effects of dyadic co-production
when each student individually produces a new text.
The sample was composed of fifteen dyads and fifteen individuals from three
sixth-grade classes.A counter-balanced design was applied in each class to assure
an equivalent achievement level in French between the five dyads and the five
individual subjects.The classes were located in Genevan elementary schools
with a student population close to the canton average in terms of the families'
socioprofessional status.
The instructional sequences proposed an authentic communication goal: the
production of storybooks - narrative texts accompanied by illustrations - to be
read by younger students attending the same school. Each sequence included five
phases:
1. presentation of the writing task and elaboration of a production guide that
remained available to the students throughout the sequence;
2. planning the story to write based on the 10 illustrations;
3. composing a draft of the story;
4. revising the draft; and
5. transmission of the storybooks to the intended audience.
The first and last phases were conducted with all students in the class (dyads and
individual writers). The students working in dyads carried out phases 2 and 4
together, i.e. joint text planning and revision, but in phase 3 each student wrote
half of the draft, corresponding to five of the 10 pictures. Students in the indi-
vidual production condition carried out phases 2, 3 and 4 alone.

62 Two patterns of dyadic functioning

The results presented here focus on a small part of the data collected by Rouiller,
i.e. the relationships between dyadic interactions and text revisions in phase 4.
An initial analysis showed that the students of some dyads carried out a much
larger number of transformations during the collaborative sequence at P1 than
they did individually at P2, whereas other dyads made approximately the same
number of transformations at P1 as made by each individual at P2. It is worth
noting that there were no cases of fewer transformations at P1 than at P2.
In order to understand how dyadic interactions affect revision, an in-depth
analysis was conducted of the verbal exchanges and the transformations recorded
for five dyads. Dyads Alpha and Beta,whose members made an equivalent
number of transformations at P1 and at P2, are designated subsequently as DE (E for
equivalent) dyads. Dyads Gamma, Delta and Epsilon, whose members carried out a
considerably larger number of transformations together at Pl than each did
individually at P2, are designated as DS (S for superior) dyads.
162 ALLAL

Each dyad's interactions were divided into sequences of utterances defined by


the topic under discussion.The data presented here pertain to two relationships.
First, we examine the relationship between verbal fluidity, i.e. the number of ut-
terances recorded during a dyad's interaction, and the number and density of text
transformations carried out. Second, we report the relationship between the
topics of the dyads' verbalisations during text revision and the object of their trans-
formations. Interpretation of the data will be based on descriptive indicators for
each dyad rather than on the results of statistical tests grouping DE and DS data.

62 Number of verbalisations and transformations

Table 4 shows the number of transformations made by each dyad, the density of
the transformations (i.e. number of transformations expressed as a percentage of
the number of words in the draft) and the number of oral utterances recorded
during the revision activity at P1 .There is a clear relationship between all three
indicators:The DS dyads carry out more transformations (in absolute terms and
when text length is taken into account) and their interactions involve a much
larger number of utterances than those of the DE dyads.

Table 4 Number of text transformations, density of text transformation? and


number of oral utterances, by dyad at P1.

Dyad Number of Density of Number of


transformations transformations utterances
DE Alpha 17 13.8 29
Beta 20 10.6 73
DS Gamma 47 27.8 757
Delta 58 22.7 474
69 28.4 387
Epsilon

a Density of transformations = (nb. of transformations / nb. of words in the draft) x 100%.

However, as Rouiller (1998) points out, the direction of causality underlying these
results is open to question. Do students who carry out more transformations of
their text have a tendency to verbalise more about their actions? Or do dyads
with a higher level of verbalisation tend, as a result, to make more changes in their
text? These questions suggest different interpretations of the role of metacognitive
regulation in revision.The first interpretation implies that regulations-in-action
stimulate verbalisation which then increases metacognitive awareness and
reflection about revision.regulation. Althoughetation suggests that verbalisation
provokes increased metacognitive awareness of potential directions of revision
which is then translated into actions of regulation.Although it is likely that both
types of processes occur in a cyclical manner, a qualitative analysis of the dyads'
global revision strategies indicates that other factors may affect both verbalisation
and transformation.
Rouiller's analysis shows that the DS dyads reread their drafts more times than
the DE dyads, and that specific goals are often assigned to each of the successive
METACOGNITIVE REGULATION OF WRITING IN THE CLASSROOM 163

DS readings. When one member of a DS dyad is rereading, the partner is often


called upon to take part in solving the problems that are identitied.The DS dyads
also make greater use of reference materials (e.g. dictionary).The members of DE
dyads, on the other hand, tend to divide up the task of revision (e.g. each rereads
what the other wrote) and to carry out their work in a parallel manner. Their
verbalisations appear to be essentially a way of coordinating individual
activities. In summary, Rouiller concludes that multiple rereadings accompanied
by joint problem solving influence both the quantity of the utterances produced
during the interaction and the number transformations actually carried out.

6.3 Topics of verbalisation and objects of transformation

What do interacting students talk about during text revision? Table 5 presents the
distribution of each dyad's utterances (expressed as percentages of the total number
of utterances given in Table 4) for the following three categories: (1) general as-
pects of task management (e.g. suggestions such as: You read the first part, I'll
read the second.), (2) specific topics (semantics, text organization, spelling) which
correspond to objects of potential text transformations, and (3) other utterances
(reading of parts of the draft out loud and occasional remarks not linked to the
revision task).
Since the interactions of the DS dyads are considerably longer than those of
the DE dyads (see Table 4), the members of the DS dyads discuss each aspect of
text revision to a much larger extent. The distribution of the utterances in per-
centages shows, however, that the DS dyads formulate relatively more utterances
about spelling, whereas the DE dyads' verbalisations show a relatively more con-
cern for the semantic content of the text. For both types of dyads, the
percentages of utterances concerning text organisation are quite low.

Table 5 Topics of verbalization as percentage of total utterances, by dyad.


Dyad General Semantics Text Spelling Other'
task organization
DE Alpha 17.2 34.5 0.0 27.6 20.7
Beta 25.3 28.0 13.3 29.3 4.1
DS Gamma 17.7 15.1 8.6 47.7 10.9
Delta 19.5 12.1 5.5 46.7 16.2
Epsilon 12.9 11.6 11.3 45.6 18.6

a Other includes out loud reading of sentences of the draft and occasional remarks not
-

linked to the task.

A qualitative analysis of one of the most interactive DS dyads (Alan and Tony)
shows three characteristics of their exchanges.
1.They often verbalise the steps of their co-constructed verifications:
A What verb is that? To reject?
T Past indefinite tense.
A Wait.
164 ALLAL

T Yeah, wait.
A "He felt re, rejjjjj- rejected,".... He was feeling, he was feeling re-
jected.Where's that? To reject, to reject, ..
2. They remind each other of the need for joint monitoring of the revision proc-
esses:
T Benjamin...
A ...he told them the story, and... told them the whole story.That's
better, isn't it? No, but you have to say if you agree, it isn't me who
has to...
T Yeah, but I was thinking about it!
3. They confront differing viewpoints before agreeing on a change:
T We should take away a "sad".
A Why? That's the story.
T "He was sad, but so sad, so sad."
A "He was sad, but sad, so sad".... sounds stupid.
T He was sad, so sad that he wanted....
These excerpts illustrate the emergence of operations of metacognitive regula-
tion as explicit objects of peer interaction.
To further understand the relationship between verbal interaction and text
transformation, Rouiller verified the degree of correspondence between the spe-
cific topics of the utterances formulated by each dyad and the objects of the
transformations actually carried out by the dyad.
The major results can be summarised as follows.
1. Text organisation. As noted above, text organisation is the topic that gave
rise to the least amount of verbalisation. When the number of interaction
sequences concerning this topic are compared with the number of transfor-
mations carried Out, it is found that there are always fewer, sometimes far fewer,
verbal sequences than the number of actual transformations. In other words,
students in all the dyads make changes of text organisation without having
talked about the transformations. Rouiller suggests that a possible explanation
lies in the influence of the school curricular materials.At the time the study
was conducted, curricular activities concerning text organisation were gener-
ally much less developed that exercises concerning spelling and vocabulary.
Students may therefore have acquired few concepts and terms for discussing
text organisation, as compared to the references they have for talking about
spelling and lexical choice.
2. Semantics. For this topic, the results show no clear trend for either type of
dyad. Some semantic transformations occurred without having been discussed,
but some exchanges, especially about lexical choice, did not result in text trans-
formations.
3. Spelling. There is a clear difference between DS and DE dyads in this area.The
interactions of DS dyads involve a much larger number of verbal sequences
about spelling than the number of spelling transformations actually carried
out.The exchanges that did not result in transformations included a sizeable
number of verification sequences leading students to the conclusion that no
adjustment was needed. The data for the DE dyads show the opposite ten-
dency. The interactions involve fewer verbal sequences about spelling than
the number of transformations actually carried out. This result reflects
the
METACOGNITIVE REGULATION OF WRITING IN THE CLASSROOM 165

global revision strategy of the DE dyads whose members tended to have rela-
tively few verbal exchanges while they carried out largely individual proof-
reading of different parts of the text.

65 Conclusions drawn from Study 3

The research conducted by Rouiller (1996, 1998) shows that a dyadic structure of
co-operative text production is translated into different patterns of peer interac-
tion which arc associated with differences in text revision. Some dyads (DE) di-
vide up the task of revision and carry it out in a largely individual manner. They
therefore have little opportunity for confrontation of different viewpoints or for
co-construction of solutions, and they make relatively few transformations of their
drafts. This pattern of interaction implies minimal coordination of the students'
existing strategies of metacognitive regulation, but is unlikely to provoke any sig-
nificant changes in their levels of functioning.
A much more dynamic pattern of interaction is shown by other dyads (DS).
These dyads engage in lengthy exchanges organised around multiple rereadings
of their draft, leading to a large number of text transformations.Their exchanges
include numerous confrontations between the participants' viewpoints, as well
as substantial instances of joint problem solving. Exchanges concerning spelling
are particularly important occasions for the use of metalinguistic knowledge (es-
pecially about grammatical categories and rules) and for metacognitive reflection
on how to solve the problems under consideration.This synergetic form of inter-
action, which is characterised by explicit co-construction of regulation strategies,
provides a context that is likely to foster the acquisition of higher-level
strategies by each participant.

7 GENERAL CONCLUDING REMARKS

The results of the three studies presented in this chapter lead us to several con-
cluding remarks regarding the factors that affect sixth-grade students'
strategies of metacognitive regulation, as reflected in their transformations of
successive versions of their texts.
1. As shown in Study 1, the mastery of basic written language skills (linked to
knowledge of spelling, grammar, vocabulary, etc.) is a factor that affects the
students' regulation strategies in three ways. A higher level of mastery allows
anticipation of a wider range of transformations (including rewriting and re-
structuring, and not just proofreading and correcting); it encourages monitor-
ing of problems of text organisation and use of more complex means of mak-
ing adjustments (substitution and rearrangement, rather than only addition
and deletion).
2. The different phases of text production appear to induce variations in stu-
dents' strategies of regulation. When students are transforming their notes
into a first draft (Study 1), they tend to use higher-level strategies (more
optional transformations, more complex adjustments affecting text structure)
than those used when they are revising their draft (Studies 2 and 3).
3. The effects of instruction on metacognitive regulation are not easily demon-
strated.The data from Studies 2 and 3 suggest that a socio-cognitive
h
166 ALLAL

emphasising writing in a context of communication, tends to encourage strat-


egies of regulation focused largely on orthographic and textual conventions,
rather than on semantic content. But other aspects of regulation - anticipation
of optional transformations, means used to make adjustments - are generally
unaffected.The in-depth study of peer interactions in Study 3 indicates that
for a given set of instructional conditions (such as dyadic co-production),
there can be a wide range of ways in which students actually function.This
variation could explain in part why the overall effect of an instructional
approach is weaker than might be expected. In future research and
classroom practice, greater attention needs to be given to such factors as
differentiation of teacher interventions, and student appropriation of co-
operative learning techniques. These factors can increase the chances that
scaffolding and social mediation do in fact function in a manner that fosters
the progression of writing skill and the acquisition of more powerful
strategies of metacognitive regulation.
AUTHOR NOTE
Study 2 was supported in part by grant no. 4033-035811 attributed to Linda Allal,
Dominique Btrix-Kohler, Laurence Rieben, Madelon Saada-Robert and Edith
Wegmuller by the Swiss National Scientific Research Foundation. The study was
conducted in second and sixth-grade classes. Only a small portion of the sixth-
grade data is presented here. More detailed presentations of this research are found
in Allal, Btrix-Kohler, Rieben and Rouiller (1998) and in Allal, Rouiller, Saada-
Robert and Wegmuller (1999).

Das könnte Ihnen auch gefallen