Writing models


This paper documents my failed attempt to use reflections of my writing experiences to find a critical fault in the Flower and Hayes model of the writing process. First I chose two extreme writing situations (writing a song and answering a maths question) that do not fit Flower and Hayes' model as described in Module 3 of the MATA course [Williams and Radjewsca (1999)] and by Flower and Hayes themselves [Flower and Hayes (1980)]. Then I modified the structure of the Flower and Hayes model (shown in Figure 1) so that it could account for these extreme situations and noted the limitations in the original model that called for the changes (it gives no account of the relationship between the author's mental representation of the text and the physical representation of the text). Finally I looked for a writing situation from my work as a technical author (maintaining a global Intranet site) in which these limitations could be critical. I found that, in spite of the limitations, the Flower and Hayes model could account for the real work situation.

Figure 1: The structure Flower and Hayes writing model [Flower and Hayes (1980) p.11]

Figure 1: The structure Flower and Hayes writing model [Flower and Hayes (1980) p.11]

Exploring the model: Writing a song

My first writing situation is one particular experience of song writing. Cycling home one night I was thinking about the common theme for blues songs, "if you leave me it would be terrible", and I decided to twist it into "if you leave me it would be wonderful". During the course of the cycle ride, which took about an hour, I wrote Riverside Blues (the lyrics can be found below in Appendix A).

Structure of the writing model

There is a fundamental difference between this scenario and the ones studied by Flower and Hayes (1980), I could not write down my ideas as I rode so the whole writing process was done mentally. There was no "TEXT PRODUCED SO FAR" in my task environment.

Figure 2 shows a writing model for this type of writing, the differences between this and Figure 1 are: "TEXT PRODUCED SO FAR" has become "TEXT COMPOSED SO FAR" and moved into "THE WRITER'S LONG TERM MEMORY" and "READING" in the reviewing writing subprocess has been replaced with "RECALLING".

Figure 2 - A model for writing Riverside Blues

Figure 2 - A model for writing "Riverside Blues"

Task environment - Writing assignment

The external factors that formed the writing assignment included: blues songs and my own experiences (topic); the Penny Theatre open stage, and other musicians that I know (motivating cues); and the people who go to listen to the open stages (audience).

The writer's long term memory - Background knowledge

This is equivalent to the whole of "The writer's long term memory" in Flower and Hayes' model, it included: my knowledge of life and of what constitutes a typical blues song and blues clichés (knowledge of topic); my knowledge of what people will expect from a blues song and will find humorous (knowledge of audience); and my familiarity with the form of twelve bar blues songs (stored writing plans).

The writer's long term memory - Text composed so far

This is the lyrics of the songs which I memorised as I wrote.

Planning - Generating

I looked for ideas that would fit together to tell a story of someone in an unhappy relationship, for pairs of rhyming words and for phrases with the right metre.

Planning - Organising

Even though the song is so short, I arranged it into a story with a definite beginning, middle and end.

Planning - Goal setting

As I wrote the song I set myself goals of adding features that would add variety to it without breaking away from the typical structure of a blues song.


I translated the ideas, rhyming words and phrases into a series of propositions express in rhyming pairs of lines.

Reviewing - Recalling

The lyrics that I revised were recalled from my memory, as opposed to being read from a physical representation of the text.

Reviewing - Editing

Once the lyrics had been recalled from my memory I was able to decide how they needed to be changed, and then to commit the required changes to my memory.

Exploring the model: Answering an undergraduate maths question

The second writing situation is answering an undergraduate maths question. I, like many of my fellow students, found it impossible to pursue even a moderately difficult line of mathematical reasoning without writing it down as I went along. One of my essential skills as a mathematician was manipulating complex mathematical expressions by breaking the tasks down into a series of simple abstract symbol manipulations. The process would typically be something like:

  1. Write the first expression by copying or deducing it from the question;
  2. Start a new line with either the symbol for "equals" ("=") or for "implies" ("");
  3. Identify a component of first expression and either copy it as is or change it and write the new version to the new line;
  4. Repeat step 3 until a new version of the statement has been written down;
  5. Repeat steps 2 to 4 until a expression that answers the question has been written down.

It is worth noting that this technique uses a writing process that is very close to knowledge telling typically used by children to write simple stories [described by Carl Bereiter and Marlene Scadamalia (1987) and by Mike Sharples (1999 pp. 22-23)].

Structure of the writing model

The one minor change I needed to make to Flower and Hayes' model was to add a feed from "TEXT PRODUCED SO FAR" into "GENERATING" to take into account they way that each written expression was based on the one before.

Figure 3 - A model for answering a maths question

Figure 3 - A model for answering a maths question

The writer's long term memory

This included my knowledge of mathematical methods and theorems (knowledge of topic), my knowledge of what examiners would look for in my work (knowledge of audience) and my knowledge of how to write a mathematical proof (stored writing plans).

Task environment - Writing assignment

The external factors that formed the writing assignment were the subject of mathematics (topic), the examiners who would be marking my work (audience) and the degree coursework, lectures and exams (motivating cues).

Task environment - Text produced so far

This was the mathematical expressions I wrote down.

Planning - Generating

The ideas for each expression were generated from the expression written immediately before or from the question being answered.

Planning - Organising

I applied a standard structure that I was familiar with through my experience in maths, so very little effort was need to decide how to organise any given question.

Planning - Goal setting

My goals were to answer the question quickly and accurately.


The content of each expression was translated into a new expression until the question was answered.

Reviewing - Reading and editing

The complex nature of the expressions made reading them for reviewing purposes difficult. When I got stuck in the writing process or wrote an expression that could not be right (for example, one that was inconsistent with the question) I would read through my work looking for mistakes and, if I found one, go on to editing. Often, though, I'd fail to find the mistakes and start the writing process again from scratch.

Reviewing - Editing

Sometimes I would be able to correct the mistake that I found in the existing text, but more often I would correct its first occurrence and then completely rewrite the rest of the work.

Testing the model: Maintaining a global Intranet site

The first writing situation shows that a writer has a mental representation of the text and that that representation can form an essential part of the writing process. The second writing situation shows that the physical representation of the text can be similarly essential. I will now apply these findings to a normal writing situation taken from my work as a technical author.

My company has an Intranet website that can be accessed by employees from all our offices worldwide and I am a member of the team responsible for maintaining it. We use a content management tool to let designated people outside the Intranet team create and update Intranet pages. As a result sometimes I have to update pages that I am familiar with but which have been modified without my knowledge, so the physical representation of the text will have changed but my mental representation won't have. Given that Flower and Hayes' model makes no reference to a mental representation of the text, can it account for this scenario? What happens when I try to edit text, which differs in its mental and physical representations?

When this happens, no matter how familiar I am with the page, I will still need to read the text before I can edit it (if only to locate it) and then I will spot that the text has been updated by somebody else and update my mental representation accordingly. So when I come to type the new text it will always be based on the same text as in the physical representation.

Insofar as I always read the existing text before typing revised text I can say that I always read the text before editing it, so I can apply Flower and Hayes' model in its original form. However, interpreting the model in this way means that it does not distinguish between occasions when I know what I want to change the text to before I read it and occasions when I have to read the text before I can decide how to change it.

Conclusion - Evaluation of the Flower and Hayes model

On the basis of my introspections and taking Flower and Hayes' model as "a first approximate description of normal composition than can guide research and afford a valuable starting point in the search for a more refined model" (Flower and Hayes 1980 p.10) I cannot fault the model. The model needed some refinement to explain the feature of my two unusual writing situations, but when I looked at my normal work as a technical author Flower and Hayes' original version of the model sufficed.

The "more refined model" will need to give a much fuller account of how physical and mental representations of text are used in writing processes, probably as part of a more general account of how the writer uses long term memory (including the mental representation of the text) and external sources of information (including the physical representation of the text). This is becoming increasingly important as technology makes the physical representation of the text more dynamic. For example, the AutoCorrect tool in Microsoft® Works Word Processor corrects certain spelling, typing and grammatical errors as you type. The text you are reading now is different from the text I typed, and I will not know exactly what has been transcribed to my computer screen until I read it myself. Where do situations like this fit into the writing model?

Deborah McCutchen argues that the fluent sentence generating processes of skilled writers allows them to make more extensive use of their long-term memory resources, including "earlier text choices (stored in [a long-term memory] text representation)", than novice writers [McCutchen (2000) p. 21]. Accepting this, I would expect that the more skilled a writer is the more she will recall her text from memory instead of reading it, so the greater the effect of any differences between the mental and physical representations will be. One way to investigate this would be to look at skilled and novice writers working in an unfamiliar writing environment with a dynamic physical representation of the text (for example, people using word processors with AutoCorrect for the first time), if I am right then the skilled writers will be more aware of the changes in the text than the novice writers because it will a greater impact on their writing processes.


Bereiter, C. and Scardamalia, M. (1987) The psychology of written composition. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc

Flower, L. S., and Hayes, J. R. (1980) "Identifying the organization of writing processes", in Gregg and Steinberg (eds)(1980), pp. 3-30

Gregg, L. W. and Steinberg, E. R. (eds) (1980) Cognitive processes in writing, Hillsdale, NJ: Lawrence Erlbaum Associates, Inc

McCutchen, D. (2000) "Knowledge, Processing, and Working Memory: Implications for a Theory of Writing", Educational Psychologist, 35(1), pp. 13-33

Radjewsca, R and Williams, N. (1999) MA Technical Authorship module 2.1: Approaches to communication and linguistics, Sheffield Hallam University

Sharples, M. (1999) How we write: Writing as creative design, Routledge

Appendix A: Riverside Blues

I've been thinking 'bout you baby, you're always sitting on my mind.
I've been thinking 'bout you baby, you're always sitting on my mind.
If you leave me now woman, that would be just so kind.

I've been paddling in the river wondering how I got mixed up with you
I've been paddling in the river wondering how I got mixed up with you
I've been here so long my feet have turned cold and blue

I've been scaring all the fish,
Throwing stones out at a frog,
There's a terrified otter hiding underneath the log.
I've been thinking 'bout you baby, your picture's burned into my head.
I'm gonna leave you now woman and make do with a stone cold bed.