Navigation – Plan du site

About the psycholinguistic models of the writing process for a didactics of written production1

What cooperation between psycholinguists and specialists in didactics?
Claudine Garcia-Debanc et Michel Fayol


This article problematises the possible areas of cooperation between psycholinguists and specialists in didactics by underlining both the interests of an interaction between them and the specific and complementary mission of both fields.
After a historic overview of how references to psycholinguistic works emerged in the research on the didactics of written production, the main models of verbal production, especially of written verbal production, published during the 1980s and1990s are presented: Hayes and Flowers (1980), Garret (1980), Levelt (1989), Van Galen (1991). Their interest and the limits of their use in didactics are investigated. What is the possible function of verbal production models in didactics? What methodological precautions are necessary before using them? More particularly, the aids to writing possibly inspired by these models and the limits of such aids are examined. In a fourth part, we compare the ways in which each of the two disciplines approaches the key issue of learning.
As a conclusion, the comparison of the methods respectively used in each of these fields, especially the part reserved to text analysis and real-time processes is useful to define an interesting cooperation area between psycholinguists, linguists and specialists in didactics, especially towards designing writing tasks and evaluation indicators.

Haut de page


Haut de page

Texte intégral

  • 1 - Originally published in Repères 26/27, 293-215

1A number of research works on the didactics of writing have drawn upon psycholinguistic research into writing processes. The present contribution, which was jointly written by a psychologist of language and a specialist in didactics offers a problematisation of possible areas of cooperation between these two disciplines by showing both the interests of an interaction between them and the specific and complementary missions of both fields. It is aimed to raise methodological questions and open up new research perspectives.

2Anglo-Saxon psycholinguistic research into writing processes was presented for the first time to the community of teacher trainers by Michel Fayol in a paper entitled “L’approche cognitive de la rédaction: une perspective nouvelle” (The cognitive approach to writing – a new perspective) published in 1984 in issue 63 of Repères. In this article, the author argued for a cognitive approach to writing to advance both psychological knowledge and teaching practice. He was aware of the limits of these analyses and of their teaching implications: “Psychologists believe that the recent study of writing processes has only led to very naive descriptive attempts and to extremely limited theorisations.” The conclusion of the article indicated the interest of cooperation between language psychologists and specialists in didactics: “if we know the world only through the initiatives we take and the ‘outcomes’ of these initiatives, then the findings of the didactics of writing will contribute to advance psychological knowledge and teaching practice simultaneously”.

3This pioneer contribution was completed in March 1986 by issue 49 of the journal Pratiques entitled “Les activités rédactionnelles” (writing activities) which also contributed to disseminating the research works conducted by Anglo-Saxon language psychologists on writing processes and published in the early 1980s. This was a landmark issue in the dissemination of these psycholinguistic works. In a long introductory article, Michel Charolles summarised the findings about the linguistic, psychological and didactic aspects of the writing processes. In addition, he requested secondary school pupils to perform constrained writing tasks – introducing paraphrastic reformulations in a scientific text or writing an argumentative text based on a particular argumentative orientation. He showed how the analysis of writing processes opened up new perspectives for the didactics of writing. All the papers he collected in this issue explored writing processes. The contribution by Claudine Garcia-Debanc sought to identify the teaching implications of the knowledge derived from the description of writing processes. She exposed the risks of a naive borrowing of a theoretical model that was not designed for didactics and investigated the risks of applicationism. To what extent can expert practice contribute to didactic progress? What is the impact of the explanation of procedures on their completion?

4In 1989, Dominique-Guy Brassart in issue 11 of the journal Recherches entitled “Du brouillon au texte” (From rough drafts to final texts) wrote a long paper to the “reviewing processes in the psycholinguistic models of writing”. He scrutinised the original models. It was all the more necessary as these works were often mentioned or quoted but not studied, which caused many misinterpretations. He also gave a detailed outline of research findings. The main specificity of this article is to refute the received idea shared among specialists in didactics and teachers that deletions were a sign of higher textual competence, which led them to value reviewing operations. Brassart argued for the crucial importance of planning operations.

5During the same period, these psycholinguistic models were reference frameworks for two research works on writing in primary school “Évaluer les écrits à l’école primaire” (Evaluating writing assignments in primary education) and “De l’évaluation à la révision” (From evaluation to reviewing). These works were followed by research reports published by INRP (French institute for educational research) and mainstream books published by Hachette Éducation. The interim findings of these works also materialised in oral contributions, especially during the DFLM (Didactics of French-Mother Tongue) symposium in Namur in 1986. These works were based mainly on textual linguistics and the models of writing processes.

6Some fifteen years later, we think it is worth evaluating the cooperation between language psychology and the didactics of writing. The didactics of French has indeed made progress and developed into a field with its own research topics and methodologies. It has become aware of the risks of applicationism in all its forms.

7Similar cooperation with the specialists of other fields that are valuable references for the didactics of writing – specialists in textual genetics who analyse writers’ manuscripts, anthropologists, and sociologists of cultural practice – would be fruitful. Such cooperation is necessary to clarify the links between the didactics of writing and some contributory disciplines and determine their specificities. Issue 118-119 of Pratiques, which was coordinated by Christine Barré de Miniac, initiated similar reflection on the research into the sociology of the cultural practices of writing.

8In this article, we will first seek to outline the key elements of the main models of verbal production, especially of written composition, and to question their uses in the didactics of writing. The various models of verbal production, especially of written composition, will first be examined. What is a model for a psycholinguist? What requirements must it meet? What are the assets and drawbacks of the main models of the 1980s?

9We will then investigate the uses that can be made of psycholinguistic models in the research on didactics. What is the possible function of composition models in didactics? What methodological precautions are necessary before using them? More specifically, we will focus on the aids to writing such as they may have been inspired by these models.

10In a third time, we will show how a purely didactical question, that of aids to writing, can be enlightened by references to theories of language psychology.

11We will also compare how each of the two fields addresses the key issue of learning. What are the limits to the psycholinguistic studies of development? What is specific to the contribution of didactics and how can its impact be analysed? How can longitudinal studies be conducted to account for long-term learning? Such research whose findings are expected by both researchers in language psychology and teacher trainers calls for cooperation between psycholinguists and specialists in didactics.

12The comparison of the methods used in each respective field, especially the part devoted to text analysis and to the analysis of real-time processes is useful to map a potentially fruitful cooperation area between psycholinguists, linguists and specialists in didactics, especially towards the development of writing tasks and evaluation indicators.

1. The main models of verbal production in the 1980s and their developments

13During the 1980s and 1990s, many theoretical and empirical works tackled the problems raised by written composition. These works drew upon three main categories of models that were distinguished by their topics and the methods used but, beyond these differences, were all based on a conceptualisation strongly inspired by cognitive psychology. These models were essentially descriptive. However, they contributed to regarding written composition as an activity that could be analysed as a whole and into different parts. Despite their shortcomings, they were instrumental in making assumptions in three research fields – how adult writers function; the comparison between experts and novices; the didactics of written composition.

14When we look at the various models, we realise that the models of written composition (Alamargot & Chanquoy, 2001; Fayol, 1997, 2002; Piolat, Pélissier, 1999) are all part of the model-based theory and complement each other as they focus on partially different components.

1.1. The Hayes & Flower model

15The first and most popular model from which all the other models were derived is based on Flower and Hayes’ (1980) and Hayes and Flower’s (1980) works. Following an analysis of verbal protocols designed to identify the origins of learning difficulties, the authors eventually distinguished (Figure 1):

  • Task environment (including the text itself from which the writer can build upon and writing instructions mentioning the topic, the audience and the reasons for writing);

  • Conceptual, situational (especially that relative to the audience) and rhetorical (types of texts) knowledge in the long-tem memory;

  • The writing process itself divided into three sub-processes: conceptual planning (generating, organising and goal-setting), translating and reviewing (evaluating and revising).

Figure 1. Schematic representation of the writing process (adapted from Hayes & Flower, 1980, p. 11)

  • 2 See especially the mainstream work reporting on classroom approaches: EVA group (1998) De l’évaluat (...)

16This model focuses on the conceptual aspects of writing (topical knowledge, organisation of the corresponding knowledge in the memory and translation of this knowledge according to the goal set). This model is part of a problem-solving framework and pays little attention to the analysis of language processing. It focuses instead on one box “REVISING/REVIEWING” which gave rise to considerable research. This part might prevail over all the others and the use of this tool is one of the privileged means to improve outputs. However, the potential interest of this tool for the study of learning has been little exploited, at least initially. Between 1990 and 1993 it was the research topic and the title of cooperative research “Révision des écrits2” (Reviewing compositions) coordinated by Sylvie Plane and Gilbert Turco at INRP.

17The Hayes and Flower model was extended. More specifically, other researchers raised the question of composition strategies which initially were not directly addressed. As a result, Bereiter & Scardamalia (1988), Scardamalia & Bereiter (1986) opposed two composition procedures: the strategy of knowledge enunciation which consists in writing information as it is retrieved in memory and the strategy of knowledge transformation which takes account of the constraints concerning conceptual knowledge and rhetorical organisation before text writing.

1.2. The Garrett/Levelt model

18The second, influential model does not particularly focus on written composition. Building upon the study of speaking mistakes by Garrett (1975, 1980), the initial model was enriched by Levelt (1989) and his colleagues. Unlike the model designed by Hayes & Flower, this model is set in a strictly psycholinguistic perspective. It focuses more particularly on word and sentence production and on the relations between comprehension and production (Figure 2).

19This model is based on the results from the conceptual construction of a message in order to give this message a linguistic form. The construction of this form implies a two-step approach. The model first constructs a functional representation with an abstract grammatical structure (verb/subject) and equally abstract lexical elements (lemmas). In a second time, the syntactic structure and “surface” lexical forms are retrieved and “sent” to the articulatory component. Changes were subsequently brought to this model but they do not affect the general economy, which is what matters to us here (Levelt, Roelofs & Meyer, 1999).

Figure 2

20The very conception of this model allows to account for a whole set of phenomena “words on the tip of the tongue”, naming speeds, etc. It is based on the assumption of impenetrable production mechanisms which operate on an automatic and modular basis. As a result, we only access the results of their intervention. Comprehension-based regulation can only be done “after” production. Once again, this model was extended, especially in terms of lexical access and its temporal alteration. However, the questions concerning learning and the influential factors behind it have remained little addressed (however see Levelt 1998).

1.3. The Van Galen model

21The third model is also the least influential and least popular, probably because it focuses on the last stages of written production: the graphic generation of the message (van Galen, 1991). However, this model raises the problem of the sequentiality of operations. The different operations starting at different levels are simultaneously active and likely to be in competition. They must be organised and coordinated so that only one graphic element is generated. The same reasoning holds for oral production.

22The van Galen model is instrumental in showing that a supposedly “basic” process – graphic word generation – is actually complex. This model includes the determination of the correct spelling of words, the selection of allographs, the control of the size of letters, muscular adjustment, etc. It also raises the question of the different paces to generate “units” at the different levels: the clause is a probable unit compared to the global conception of the message while the “letter” is a unit at the level of graphic generation. But given that several “units” are probably at play at middle levels, the problem is raised of their respective processing speeds and of the temporary storage waiting to be used by subsequent levels. In other words, the van Galen model – more than any other model – raises the question of the real-time coordination of the different components so that production happens in a relatively harmonious and continuous way.

1.4. Critical analysis of these models

23All the abovementioned models but also those by Beaugrande (1984) or Martlew (1983) are based on a modular conception of written or oral production characterised as follows:

  • They all isolate components, each of which receives a particular (conceptual or semantic for example) type of information and transforms it into another type of information (lexical for example). As a result, a type and a level of representation can be associated with each component.

  • There are functional relations between components. Most often, information flow is constrained by a specific architecture: the data generated by a component can be directed only to another particular data. Components function either sequentially or simultaneously. In the first case, information processing of level l (see the notion of cascade processing) must wait to start processing level l – 1. As a result, it often implies arguing for temporary memories that store intermediary representations (Van Galen, 1991). In the second case, components function in parallel which raises the question of how exits are coordinated and possible interferences controlled. The connectionist models that developed later offered interesting solutions in terms of activation/inhibition (Dell, 1988).

  • The control processes that regulate the transfer of information and evaluate final products were partially studied. Either the very structure of processes includes a control component (Hayes & Flower 1980) that tests and rejects or accepts the products (the comprehension component in the Levelt model, 1989) or the very functioning of the production system is self-regulated (Berg, 1986) and the question of a control instance is then irrelevant. Most models seek a compromise between controlled/sequential processes and automatic/self-regulated/parallel processes (Fayol, 2002).

24All the models offer relatively autonomous components that are organised according to a more or less sequential or parallel structure and whose control of productions is made by an autonomous instance or, on the contrary, by the self-regulation of mechanisms. There are three types of components: those that deal with concepts, their relations and the impact of communicative and enunciative dimensions; those that concern linguistic aspects: lexical access, syntactic production, cohesion markers, etc.; those that focus on the generation of the material (phonological or graphic) aspects of the message.

2. The use of these models in the didactics of writing

2.1. Heuristic or prescriptive model? Models in didactics: an ambiguous status

25Didactics can be defined (Garcia-Debanc, 1991) as the “study of the conditions and procedures to apply teaching contents, of their impact towards improved learning outcomes in ordinary classes”. Since the research conducted by Verret, Chevallard and Martinand, the notions of didactic transposition and of reference practices are instrumental in analysing the gaps between the contents to be taught and reference disciplines, contents to be taught and contents actually taught, and finally contents to be taught and the actual learning outcomes of pupils.

26The status of models is by nature ambiguous in didactics, probably because of the polysemy of “didactics” which, especially for the didactics of French as a foreign language, still often refers to the design of (prescriptive) teaching methods and resources and not to an epistemological analysis of actual teaching materials or practices. But while models should be heuristic, teaching experts tend to use them as norms to be reproduced, probably because they often have a praxeological aim towards effective action. Scientific models are by definition partial, provisional and heuristic. They contribute to questioning the complexity of reality and not to simplify it. They are always refutable and can historically be replaced by a more competitive but still provisional model. For example, the quinary scheme of narratives designed by Larivaille and reproduced by Jean-Michel Adam is not intended at all to identify the surface markers of a text. Instead, it is designed to organise the deep semantic structure of a text and possibly to be used as a reference when one seeks to account for the interpretation or memorisation processes of narrative texts. Similarly, the models of verbal production processes help account for the complexity of jointly requested operations and for the diversity of approaches. They are certainly not a list of instructions to be successively fulfilled or an inventory of sub-processes to be used for training.

2.2. What a model is not

A model is not a set of instructions to be sequentially performed

27This is the main distinction between the Hayes and Flower model and the works of classical rhetoric, even if many people were eager to emphasise similarities, especially the parallel between planning and inventio, translating and dispositio. While classical rhetoric lists a set of recommendations, the Hayes and Flower model focuses on the iterative nature of the different writing sub-processes and their variable weight according to individuals. In addition, Garcia-Debanc (1982) attempted to show the impact of the writing task on the writing processes applied by the same people. As a consequence, it is a major misinterpretation to regard it as a stage-based model.

A model is not an inventory of components

28Even if the abovementioned models are characterised by an analytical, even boxological dimension, they cannot be interpreted as a linear inventory of components to be successively connected.

A model is not an established truth

29As we have already pointed out above, it is the very possible reproduction of a model that guarantees its scientific quality. Therefore all models are necessarily provisional.

2.3. The functions of a psycholinguistic model in didactics

30In the research on the didactics of disciplines, the possible reference to one or several psychological models of pupils’ activities fulfils several functions. It is especially useful to:

Investigate the complexity of pupils’ activities and understanding possible difficulties

31The reference to the operations of the writing process is useful to observe the activity of writing pupils in greater detail. How much time do pupils effectively devote to translating operations? Are there any signs of planning or reviewing operations? Do some tools or work conditions facilitate these operations? How early can pupils be helped to put these operations into practice? It is the very reference to a specific model used as a grid of analysis which gives substance to these questions.

Analyse some dimensions of teaching practices

32What representation(s) do teachers have of planning? What instructions do they give to pupils to develop planning and revising/reviewing?

Adapt writing activities in class

33The reference to a set of operations also enables teachers to give a critical look at the activities performed in class. Among the writing tasks performed by pupils, what were those which contributed to improving pupils’ planning? Were aids given to improve planning? If so, what type of aids? What about translating? Teacher awareness of the various dimensions they focused on and of those they did not study enough can help them adapt the planning of their activities. For example, speaker/reader exchanges during which pupils can see the impact of their writings in terms of comprehension can contribute to realising the importance of reviewing/revising and its effective application.

Designing aids to writing and rewriting

34Even if the terms “aids” is not often used in didactical research, one might consider that aids to writing and rewriting in the form of guidance sheets and criteria lists are designed to promote planning, enable pupils to problematise their writing options (textual genre, enunciation perspectives, lexical register, etc.) and contribute to reviewing, ie foster a critical (re)reading of the written composition and focus on rewriting.

35In the didactical research on written compositions, the development of such tools is mainly based on three underlying principles:

  • Explaining the features of a written piece can guide action effectively;

  • The texts read are resources for designing criteria on how texts function and contribute to reading/writing interaction

  • The criteria identified in the tool reflect the choice of the evaluator and thus only correspond to some aspects of textual functioning. By definition, they are part of a provisional and evolutionary process.

2.4. Assets and limits of the Hayes and Flower model

36The Hayes and Flower model was the most disseminated in didactics. It contributed to indisputable advances, especially in the following points:

  • There was a shift from the linguistic analysis of final compositions to a focus on the writing process;

  • Aids to writing and rewriting were designed;

  • Writing activities in class were adapted.

37Before 1980, the only method to analyse pupils’ writing was the linguistic analysis of children’s texts whose methodologies were refined throughout the 1970s and 1980s. Bernard Schneuwly (1988) found how some linguistics markers could be interpreted as the sign of operations of the writing process, especially step-by-step planning. The merit of the Hayes and Flower model was to draw teachers’ attention to the observation of what is going on when writing. The methodology of the think-aloud protocol analysis inspired the collection of oral interactions during text writing in groups of two and three pupils (Schneuwly, 1982). These oral interactions are both the signs of negotiations between the writers and the writing processes that were used. De Gaulmyn (1999) and Garcia-Debanc (1982) applied this approach to various corpuses.

38We showed how fruitful it was to refer to the models of the writing process to design aids to writing and rewriting. The tools modelled by didactical research especially focused on planning and revising operations. It is a pity that the same work was not done on translating.

39Finally, we showed as of 1986 that these models could help teachers adapt classroom writing activities because they contributed to emphasising the operations of the writing process on which training or an aid was brought as part of writing projects.

40While the Hayes and Flower model helped make significant progress in learning text composition, some blind points remain:

The blind points of the Hayes and Flower model

The linguistic dimension in composition

41In the Hayes and Flower model, translating operations remain vague. In this respect, the Levelt model is more satisfactory but is more valid for speaking than for writing situations.

The task impact on the writing processes used

  • 3 Anna LARROUY and Laure PARADA’s master’s degree dissertation, September 2001, University Toulouse-L (...)

42Hayes and Flower especially focused on the writing of essays or argumentative texts but did not take account of the specificity of writing tasks. Operations probably differ according to whether they concern composing a written justification or writing a poetic text based on constrained verbal material. It is necessary to check that the weight and order of the operations at work in each of these two cases are identical for similar topics. Various studies3 are currently conducted with 6 to 11-year-old children in the Jacques Lordat laboratory. Such studies should contribute to explore the linguistic dimension of the writing processes used in greater detail.

3. Aids to writing

43Even if aids to writing are rarely referred to as such in didactical research (“writing tools” is more common), they drew upon the models of writing processes. They were designed as part of didactical research conducted at INRP, especially the EVA (evaluating writing in primary school) and REV (from evaluation to rewriting) programmes. Awareness of pupils’ difficulties to plan and revise/review their texts and the notion of alleviating cognitive overload mainly inspired this research.

3.1. Aids to writing in psycholinguistics

44From a psycholinguistic perspective, Fayol (1997) in his book entitled Des idées au texte: psychologie cognitive de la production verbale, orale et écrite devoted several pages of chapter V entitled “Control and regulation of composition” (pp 119 to 144) to this question. He argued that “the social demand about these topics has kept increasing over the past decade” (p. 144). The studies he summarises offer indications on the opportunities to improve written compositions and/or better train writers.

45Even if the term « aid to writing » is not explicitly used, the identification of difficulties is necessary to design and validate modelling procedures that account for the mechanisms at work. These mechanisms are then turned into procedures. Research focused on “control, regulation and revising”, a dimension specific to written composition and on planning through comparison with the research conducted on oral production. Fayol (1997) pointed out that this research targeted “higher levels” (generating, modelling and organising of concepts) and left aside the aspects relative to translating.

46For example evaluation grids intended to aid writers in the comparison/evaluation task inherent in revision activities were developed and tested in various protocols (Fayol, 1997, p. 134). Scardamalia and Bereiter also tested such grids with teenagers (Bereiter and Scardamalia 1983, Scardamalia, Bereiter and Steinbach 1984; Scardamalia and Bereiter, 1984, 1986; Daiute and Kruidenier, 1985). This aid boosts self-questioning and self-evaluation and has led to relevant evaluations. However, the improvement of written compositions is limited. Bartlett (1982) also found in another study that a procedure of reflexive instructions did not significantly improve written compositions. In the absence of solid linguistic foundations, especially from the lexical and syntactic perspective, 30% of mistakes at best are corrected.

47In the same chapter, Fayol also pointed to the insufficient number of developmental studies about childhood. In addition, mid and long-term learning outcomes were not appropriately taken into account. He finally indicated that “taking account of learning aspects seemed inevitable and indispensable in the long run”.

3.2. Aids to writing in the didactics of written composition: inventory of the parameters characteristic of aids to writing in didactics

48For several years, didactical research has focused on how pupils could be helped to write and revise a written composition. Here is below a list of parameters characteristic of aids to writing:

When to use them

  • Before writing: a grid of criteria can help pupils in CM (the final years of primary education) to determine the features of detective stories even before setting out to write their own detective short stories. This grid was developed by themselves and is based on the comparison between several detective short stories they have read (Objectif Écrire 2000 CRDP Montpellier, 2000).

  • Included in the instruction: some instructions include a list of criteria or verbal or non-verbal material that can help pupils in their writing. For example, pupils in CP (the first year of primary education) who were asked to invent instructions for math problems had at their disposal syntactic structures to develop problems and a book of pictures from which they could select ideas to add content to their problems (Anna Larrouy 2001: Analyse des processus rédactionnels d’enfants de 6-7 ans confrontés à deux tâches d’écriture complexes, DEA, Département Sciences du Langage, Université de Toulouse-Le Mirail).

  • During writing and on request: in a class of CP, pupils select elements from word banks posted in the classroom. These banks are crucial to develop autonomy in writing [Garcia-Debanc (2003) in Le manuel de CP, Observatoire de la lecture, Paris]

  • At a specific time during writing: pupils can rely on an anthology of scary descriptions that they will copy out or pastiche in order to beef up their detective short stories.

  • Between two stages of the text to improve rewriting: criteria grids are designed especially to foster rewriting.

Nature of the problem to be solved

49The levels of problems to be solved may vary according to the levels of analysis that are traditionally distinguished in language sciences:

  • spelling

  • lexical

  • syntactic: sentences

  • syntactic: textual cohesion or speech

  • writing selection, especially enunciative selection

  • contents

Linguistic nature of the tool

  • non verbal material

  • non linear (set of words, scheme, etc.) verbal material

  • lists: it is the most usual form of these aids to writing, whether it is a list of criteria or informative criteria to be found in a text

  • syntactically organised and selected verbal material

  • resource material (bank of authors’ texts)

Status of the tool

  • set of resources that are available to pupils

  • material selected by the teacher according to the difficulty to be solved

Designing procedure

  • by “industrial material”, we mean published material, whether images or a dictionary, a file or a conjugation table

  • designed by the teacher

  • designed collectively in class, especially lists of criteria

Procedures to use the aid

  • chosen by the teacher

  • provided individually by the teacher on demand

  • provided individually by the teacher according to needs

  • spontaneously used by pupils

Single use VS multiple possible uses

50Ad hoc or permanent nature of the aid to writing

Use procedures

  • the aid is always available

  • designing (process)

51This typology contributes to determining the different types of aids that can be offered to pupils. Some parameters can be experimented to evaluate the facilitating or inhibiting nature of an aid and its impact on written production.

3.3. When aids become an obstacle

52Sometimes, aids are an obstacle and have an opposite effect to that expected, as shown in the following example.

53Pupils in CE2 (third year of primary education) were invited to write a sequel to a tale entitled “Le joueur de flute de Hamelin” (Pied Piper of Hamelin). The female provided two aids to facilitate task completion:

  • An aid to planning in the form of three possible ends to choose from: “Here are three possible ends. Circle your favourite end and write the end of the tale.
    1. The four friends flee rapidly and leave for the city of Bremen.
    2. The four friends manage to make brigands flee and decide to keep living in their house.
    3. Brigands catch them after a long struggle and …them (imagine what brigands do with the four friends)

  • A draft text to be continued:
    “Then one of the brigands sneaked in through the window and was determined to kill this terrifying monster.”

54It was followed by seven blank lines to be completed.

55Here is the composition of a pupil.

« Alors, l'un des brigands se glissa à l'intérieur par la fenêtre, décidé à tuer ce monstre terrifiant. - Les quatre amis s'enfuient rapidement et partent pour la ville de Brème quant il sont arivé les quatres amis rencontrent un loup je suis tros vieux pour crier et mon mètre a desider de me tuer
L'ane lui proposa vient avec nous, tu chanteras et tu deviendras musicien de la ville de Brème tu pouras gagné ta voix pour crier les cinq amis sont à Brème. »
(The original spelling and punctuation were kept)

“Then one of the robbers, who was determined to kill this terrifying monster, slipped in through the window. – The four friends ran away and left for the city of Bremen when they arrived the four friends met a wolf I’m too old to shout and my master has decided to kill me”
The donkey said come with us, you’ll sing and you’ll become a musician of the city of Bremen you’ll earn your voice to shout the five friends are in Bremen.

56Pupils’ hesitations in the use of verbal tenses (present at the beginning of the text and simple past afterwards) are evidence that the sentences suggested by the teacher were interpreted by the pupils as translating elements. The presents, which were an option in summary-like sentences for planning, were reproduced as such by the pupil. The use of the simple past shows that the starter sentence only would probably have been more effective to prompt pupils to use this verbal tense by analogy. This example suggests that multiple aids to writing do not necessarily have a positive effect on written production.

57While the short-term effects of the writing aids were felt in pupils’ compositions, there is still a lack of more longitudinal studies of how written production is learnt by school-age children.

4. A dark continent in language psychology and didactics: the question of learning

4.1. How questions about learning are addressed in psycholinguistic research

58From our perspective in this article, the models previously mentioned focused little on the dynamics of production, ie how real-time composition happens, and on learning/instruction.

59For example, Hayes & Flower (1980) mainly studied expert behaviours and opposed them to novice behaviours (Hayes & Flower, 1986). They did not conduct research analysing the real-time behaviour of writers.

60Learning-related problems were not addressed per se. During the 1980s, only Scardamalia & Bereiter (1986) really tackled this issue. They mainly adopted a “strategic” approach inspired by the opposition expert/novice. The corresponding model is not a truly developmental model because it does not mention how the shift from the first to the second mode of text composition is made.

61These authors also developed metacognitive procedures aimed to instill among learners: (a) the acquisition and implementation of frames like those derived from textual linguistics (Fayol, 1985, 1991); (b) the use of procedural guides aimed to help subjects conduct voluntary research into in-memory information, systematic references to objectives, planning and control/revising methods (Scardamalia, Bereiter & Steinbach, 1984). The purpose is to bring subjects to shift from a strategy of knowledge enunciation to a strategy of development/transformation (Bereiter, Burtis & Scardamalia, 1988; Scardamalia & Bereiter, 1987). The suggested method to induce learning reflects the method that is used in strategic learning in general: providing examples (remediation), inducing imitation through tutoring, resorting to reflection. However, these learning procedures are subjected to the usual limits of strategic learning: importance of conditional knowledge, need for positive reinforcement, transfer difficulties, strong cognitive overload (see Fayol & Monteil, 1994 for a journal). Therefore, instruction and its procedures have drawn attention more than learning itself.

62There is little research on learning and even less concerning lexicon and the syntax of written language insofar as academics do not focus on this aspect (Levelt, 1989). By contrast, van Galen (1990) devoted a whole chapter to the development of writing. Beside the study of changing parameters during development, he mentioned the importance of evolving capacities to perform competing processing operations. Taking this aspect into account raises the problem of how written composition should be handled.

63It was not until the 1990s when real-time production was eventually theorised and studied. Initial research used VCR recording (Chanquoy, Foulin & Fayol, 1990). It soon turned to the graphic tablet for measures purposes and studied performance in dual-task situations (Olive, 2002). Theorisations are largely inspired by the works of Baddeley (1986). These works were adapted to the problems raised by Kellogg about written verbal production (1994). Interindividual differences were then taken into account in research works based on the novice/expert opposition or on the examination of development and learning-related problems.

64Development and learning-related problems are subject to a dual approach. One trend is based on the notion of cognitive resources to theorise both real-time processing of production and the difficulties faced by children when learning written composition (Bourdin & Fayol, 1994, 1996; McCutchen, 1996). Another field focuses on describing how the components of written production are implemented and the evolution of their respective dynamics during development/learning (Berninger & Swanson, 1994)

4.2. How questions about learning are addressed in didactical research

  • 4 See the reports on this research: Comment les maîtres évaluent-ils les écrits ? Comment les élèves (...)

65The question of learning was hardly addressed any better in didactical research on written production, despite favourable conditions for a longitudinal long-term observation of learning. To our knowledge, only the research conducted at INRP from 1982 to 1989 carried out an analysis of the impact of innovations tested in terms of actual learning progress made by pupils4. Those in classes engaged or not in research were requested to write a narrative according to a specific instruction: (tell a story whose last sentence will be: and it was since that day that birds have flown” or “and it was since that day that pigs have had a curly tail” or another sentence of the same kind. You will give a title to your story). This writing task was selected because the narrative skills learnt during cycle 3 of primary education had to be used but the narrative genre, the etiological narrative were not familiar to pupils as they were rarely studied at that time. Then each pupil had to evaluate a text written by one of their peers, in writing and during an oral interview afterwards. The next step for pupils was to take a critical look at their own compositions, one week after they were written, during an individual examination followed by an oral interview with the experimenter. This set of tests is useful to gauge the writing and critical analysis capacities of children as well as the nature of most used criteria. These tests highlighted a differential effect among the innovations implemented, especially in terms of average pupils’ analysis capacities and the nature of most used criteria.

66Among current research on written production, few studies collect data on actual learning outcomes. For example, in the 2001/2002 directory of current research published by the international association for the development of research on the didactics of French as mother tongue (DFLM) in September 2001, 23 research programmes on written production were inventoried or over a third of the research censed in this directory. Most of them are conducted in Canada (10) compared to 5 in France, 4 in Belgium, and 4 in Switzerland. Selected methodologies were as follows:

  • Analysis of pupils’ compositions: 12

  • Longitudinal studies: 7 (4 targeted nursery education, 2 focused on students and 1 was conducted on pupils in education priority areas

  • Questionnaires and interviews on conceptions: 6

  • Instructions towards the transformation of teaching practices and analysis of their effects: 10

  • One research conducted at the University of Sherbrooke with special needs students taking the baccalaureate collected data relative to writing protocols of tutoring and aids to writing (Recherches n° 16, p. 63)

67The summarised analysis of current research shows the very low number of longitudinal studies. It is regrettable even if we are aware of the difficulties and constraints of research with cohorts of pupils. However, in the absence of a rigorous evaluation of short and long-term effects of learning strategies applied in class, didactical initiatives on how writing is learnt might remain disputable. It is in this direction that research should now be conducted.

Cooperation perspectives between psycholinguists and specialists in didactics

68We observed that the models initially designed in the 1980s targeted expert or beginning adults and, overall, paid little attention to learning/instruction problems. During the 1990s, the joint development of the models and techniques to study composition led to adopt a functional approach. It was useful to focus both on the dynamics of adults’ verbal production and on the evolution of composition during development/learning.

69Fruitful developments are expected from further cooperation between psycholinguists and specialists in didactics. According to cases, these were the experts of either field who can help the other field make progress. Here is a provisional list of these possible cooperation areas:

  • Defining indicators of achievement. Existing indicators are still too general and insufficiently analytical from the perspective of language psychology. On the other hand, the samples of compositions in didactical research are often too limited to allow statistical processing.

  • Defining writing tasks. When pupils are faced with psycholinguistic protocols, the habits they have developed at school remain deep-seated. For example, a writing task based on sequential pictures is significant. It is essential that experimental protocols provide scientifically controllable but also not too exotic tasks compared to pupils’ school habits for fear of dramatically biased results.

  • Taking account of the possible effects of ordinary teaching practices in psycholinguistic experimentations. Detailed indicators are necessary to observe actual classroom practices as Goigoux does (2002) about the teaching of reading in CP (first year of primary education).

  • Conducting longitudinal studies of cohorts of pupils. Such research is costly and difficult to implement but they are essential to know the short, medium and long-term effects of the applied practices (Larrouy, PhD dissertation in progress).

  • Experimenting short, medium and long-term effects of types of aids to writing.

70All these enterprises are promising for sustainable cooperation between linguists, language psychologists and specialists in didactics. This is the orientation given to the programme “Cross-disciplinary approach to written verbal production” coordinated by Denis Alamargot and recently implemented by Michel Fayol and Jean-Louis Lebrave.

Haut de page


ALLAMARGOT D., CHANQUOY L. (2001) : Through the models of writing, Kluwer.

BADDELEY A. (1986) : Working memory. New York : Oxford University Press.

BEAUGRANDE R. DE (1984) : Text production : toward a science of composition. Norwood, NJ : Ablex.

BEREITER C. (1980) : Development in writing. In L.W. Gregg & E.R. Steinberg (Eds.), Cognitive processes in writing. Hillsdale, NJ : LEA

BEREITER C.& SCARDAMALIA M. (1987) : The psychology of written composition. Journal of memory and language, 27, 261-278.

BEREITER C, BURTIS P.S., & SCARDAMALIA, M. (1988) : Cognitive operations in constructing main points in written composition. Journal of memory and Language, 21, 261- 278.

BERG T. (1986) : The problems of language control : coditing, monitoring and feedback. Psychological Research, 48, 133-144.

BOURDIN B., & FAYOL M. (1994) : Is written language production really more difficult than oral language production ? International Journal of Psychology.

BRASSART D-G (1989) : « Les processus de révision dans les modèles psycholinguistiques de la composition écrite » in Recherches n° 11, AFEF Lille, novembre 1989, p. 79-114.

CACCAMISE D.J. (1987) : Idea generation in writing. In A. Matsuhashi (Éd.), Writing in real time. Norwood, NJ : Ablex.

CHAROLLES M. (1986) : « L'analyse des processus rédactionnels : aspects linguistiques, psychologiques et didactiques » Pratiques n° 49, mars 1986, Les activités rédactionnelles, p. 3-21.

DELL G. (1988) : The retrieval of phonological forms in production : tests of predictions from a connectionist model. Journal of memory and language, 27, 124-142.

FAYOL M. (1985) : Le récit et sa construction. Neuchâtel, Paris : Delachaux & Niestlé.

FAYOL M. (1991) : From sentence production to text production : Investigating fundamental processes. European Journal of Psychology of Education, 101-119.

FAYOL M., LARGY P. & LEMAIRE P. (1994) : When cognitive overload enhances subject- verb agreement errors. Quarterly Journal of Experimental Psychology, vol. , pp. 

FAYOL M. (1997) : Des idées au texte. Psychologie cognitive de la production verbale, orale et écrite, Paris, PUF.

FAYOL M., & MONTEIL J.M. (1988) : The notion of script : From general to developmental and social psychology. C.P.C./European Bulletin of Cognitive Psychology, 8, 461-475.

GALEN G.R. VAN (1991) : Phonological and motoric demands in handwriting.

Evidence for discrete transmission of information. Acta Psychologica, 74, 259-275.

GARCIA-DEBANC (1987) : La production écrite, Canada

GARCIA-DEBANC C. (1986) : « Processus rédactionnels et pédagogie de l'écriture », Pratiques n° 49, mars 1986, p. 23-49.

GARCIA-DEBANC C. (1990) : L'élève et la production d'écrits, Metz, CASUM.

GARCIA-DEBANC C. (1995) : « Le lire dans l'écrire », Pratiques n° 86, juin 1995, 71-92. Pratiques n° 105-106, La réécriture, juin 2000.

GARCIA-DEBANC C. (Éd) et alii (2001) : « Théories du texte et écriture » in Quelles grammaires enseigner à l'école et au collège ? Discours, genres, texte, phrase. Delagrave/CRDP Midi-Pyrénées, 291-346. Contributions de C. TAUVERON, M. FAYOL, C.ORIOL-BOYER, N. BIAGIOLI-BILOUS.

GARCIA-DEBANC C. (2003) : Garcia-Debanc (2003) in Le manuel de CP, Observatoire de la lecture, Paris.

GARRETT M.F. (1975) : The analysis of sentence production. In G.H. BOWER (Ed.) The psychology of Learning and motivation. New York : Academic Press.

GARRETT M.F. (1985) : Levels of processing in sentence production. In B. Butterworth (Ed.) Language production, vol. 1, Speech and talk. New York : Academic Press.

HAYES J.R., & FLOWER L.S. (1980) : Identifying the organization of writing processes. In L.W. Gregg & E.R. Steinberg (Eds.), Cognitive processes in writing (p. 3-30). Hillsdale, NJ : L.E.A.

HAYES J.R. & FLOWER L.S. (1986) : Writing research and the writer. American Psychologist, 41 , 1106-1113.

LARROUY A. (2001) : Analyse des processus rédactionnels d'enfants de 6-7 ans confrontés à deux tâches d'écriture complexes, DEA, Département Sciences du Langage, Université de Toulouse-Le Mirail.

LEVELT W.J.M. (1989) : Speaking from Intention to articulation. Cambridge, MA : M.I.T. Press.

MARTLEW M. (1983) : Problems and difficulties : Cognitive and communicative aspects of writing. In M. Martlew (Ed.), The psychology of written language. New York : J.Wiley & sons.

MAS M. et alii (1993) : Comment les élèves évaluent-ils leurs écrits ?, Paris, INRP, Rapport de recherche.

MAC CUTCHEN D. (1996) : A capacity theory of writing : Working memory in text composition. Educational Psychology Review, 8, 199-235.

PIOLAT A. et PELISSIER A. (Eds) (1998) : La rédaction de textes, approche cognitive, Delachaux et Niestlé.

Repères n° 4, Savoir écrire, évaluer, réécrire (1991), n° 10, Écrire, réécrire (1994), n° 11, Écriture et traitement de texte (1995).

SCARDAMALIA M. & BEREITER C. (1983) : The development of evaluative, diagnostic, and remedial capabilities in children’s composing. In M. Martlew (Ed.), The psychology of written language : Developmental and educational perspectives (pp. 67-95). London : John Wiley and Sons.

SCARDAMALIA M. & BEREITER C (1985) : Fostering the development of self-regulation in children's knowledge processing. In S.F. Chipman, J.W. Segal & R. Glaser (Eds.) Thinking and learning skills, Research and open questions. Hillsdale, NJ : LEA

SCARDAMALIA M. & BEREITER C (1986) : Research on written composition. In

M. Wittrock (Ed.), Handbook of research on teaching. New York : MacMillan.

SCARDAMALIA M. & BEREITER C. (1987) : Knowledge telling and knowledge transforming in written composition. In S. Rosenberg (Ed.) Reading, writing and language learning. Cambridge, MA : C.U.R

SCARDAMALIA M. & BEREITER C, & GOELMAN H. (1982) : The role of production factors in writing ability. In M. Nystrand (Ed.), What Writers Know the language, process, and structure of written discourse. New York : Academic Press.

SCHNEUWLY B. (1988) : Le langage écrit chez l'enfant. La production des textes informatifs et argumentatifs, Lausanne-Paris, Delachaux et Niestlé.

SCHNEUWLY B. (1991) : « Différence entre les processus de production de trois genres : du dialogue entre énonciateurs au texte écrit », Repères 3, Articulation oral/écrit, Paris INRP, 45 65.

Haut de page


1 - Originally published in Repères 26/27, 293-215

2 See especially the mainstream work reporting on classroom approaches: EVA group (1998) De l’évaluation à la réécriture, Hachette Éducation

3 Anna LARROUY and Laure PARADA’s master’s degree dissertation, September 2001, University Toulouse-Le Mirail

4 See the reports on this research: Comment les maîtres évaluent-ils les écrits ? Comment les élèves évaluent-ils les écrits ? INRP, collection Rapports de recherche

Haut de page

Pour citer cet article

Référence électronique

Claudine Garcia-Debanc et Michel Fayol, « About the psycholinguistic models of the writing process for a didactics of written production », Repères [En ligne], Hors-série | 2013, mis en ligne le 12 septembre 2013, consulté le 30 mai 2017. URL : ; DOI : 10.4000/reperes.505

Haut de page

Droits d’auteur

Licence Creative Commons
Les contenus de Repères sont disponibles selon les termes de la Licence Creative Commons Attribution - Pas d’Utilisation Commerciale - Pas de Modification 4.0 International.

Haut de page
  • Logo Université de Lyon
  • Logo ENS Lyon
  • Logo ENS Éditions
  • Logo Institut français de l’éducation