Externalisation – how writing changes thinking

Alan Dix
Lancaster University

article in Interfaces, Autumn 2008

Full reference:
A. Dix (2008). Externalisation – how writing changes thinking . Interfaces, 76, pp. 18-19. Autumn 2008.
See also:
Alan's previous pieces for interfaces about writing and thinking: writing as third order experience and why examples are hard
Alan's pages on Research and Innovation
Alan's essays

How many times have you written a paper or report and after having done so thought “now I understand” or “I didn’t know that before”? One of my PhD students recently described the feeling after writing her draft thesis of being able to see the work for the first time. That is, writing does not merely record or communicate, but also transforms our thoughts

In this article we will explore how it is that we seem to learn so much during the process of writing and in understanding the phenomenon suggest how it can be used explicitly and practically as part of our own learning and research.

Writing as information transmission

Perhaps the simplest way to think about writing, or indeed public speaking, is as information transmission:

  1. Alan has an idea in his head
  2. Alan represents the idea in words (encoding)
  3. Alan writes (or speaks) the words
  4. The reader reads (or hears) the words
  5. The reader interprets the meaning
  6. The reader the idea in her head

This simple view elides many complications. The encoding in step 2 has to take into account the potential reader and his or her knowledge (see my previous Interfaces article “Writing as third order experience” [1]); and I am sure a semiotician would have much more to say!

The situation is particularly complex in more interactive situations. In face-to-face conversation the encoding is typically parsimonious relying on the listener’s reactions to see whether further amplification is necessary, and Herb Clark’s grounding theory [2] emphasises how both an utterance and the reaction to it not only convey information, but also help us determine the other party’s knowledge and understanding and help build ‘common ground’. Furthermore, when the words are intended to prompt some action in the hearer they have to convey the intention as well as their ‘propositional’ content – “it’s cold in here” means “shut the door” [3]. In previous work, I looked at the way people may adopt ‘lazy’ or ‘eager’ strategies in their messages when the pace of the channels of communication do not match the pace of the shared task.

However, when writing a paper any interaction is at a very much later point and anyway the feeling of ‘knowing more’ happens at the time of writing, not when we obtain feedback. That is, it is the act of producing the utterance (steps 1–3) that seems to help us to think more clearly or learn new things.

External representation in HCI

Writing is a form of externalisation and for more than 20 years the importance of external representations has been stressed in HCI by ethnographers and advocates of distributed cognition [4]. Field studies of collaborative situations, such as Heath and Luff’s classic study of the London Underground control room [5], have found that various forms of external representation help people coordinate their actions with one another. Distributed cognition sees the external representation as being an integral part of cognition itself: in order to add up 358287454 and 2856398562, most of us would resort to pencil and paper. Often the focus is on the role of written notes or the arrangement of artefacts as being a form of extended memory, which helps us to achieve tasks that would be impossible relying on short-term memory alone. We beat Millers 7+/– 2 by writing lists! [6] Advocates of the philosophy of the ‘embodied mind’, such as Andy Clark in Edinburgh and Michael Wheeler in Stirling, are more radical again, suggesting the mind itself is not bounded by our skulls, but includes the shifting array of materials around us [7,8].

These all emphasise the importance of external representation, however, there are two important things to note:

  • When we apply this kind of thinking in HCI it is usually to some user group we are studying. In order to understand the transformative effects of writing, we need to consider reflexively the role of externalisation in our own research practice.
  • Mostly ideas of external representations are applied to explain how humans considered as coupled with the environment solve problems or achieve goals that are in some sense ‘bigger’ than themselves. In contrast the surprising effect of writing is that we think differently afterwards even when we no longer have the writing with us. The effect is on our own minds.

The latter echoes Alison Kidd’s “The marks are on the knowledge worker” [9]. At a time (early 1990s) when there was a focus on improved information retrieval for knowledge work, she found that knowledge workers rarely consult documents once they have been filed. This to some extent runs counter to the idea of the paper record as a external memory. The marks on the paper are only important while they are being actively considered by the knowledge worker. However, in the long run, the effect that matters is the change in the way the knowledge worker thinks of her work; the significant ‘marks’ are those in her own mind.

As researchers or HCI practitioners, we are knowledge workers, so it is not surprising this applies equally to us.

How writing changes us

I suggest there are two sides to the transformative effect of writing, one connected to communication and the other to external representation:

  • writing forces processing of your ideas and reflection on them
  • writing makes your thoughts available as the object of study

Processing and argumentation

In constructive learning theory the learner is seen not as a passive receptacle of information, but as an active part of the process; learning happens when the learner constructs a meaning by relating new experiences with existing mental structures or worldviews. If a new idea, fact or experience is completely dissociated from your current understanding it can only be learnt rote, but is not an integrated part of your experience – hence Vygotsky’s zone of proximal development: the best things to learn are the ones that are new but only just beyond what you currently know or can do [10]. Any form of active processing helps to integrate knowledge and make it more sound.

On the face of it constructivism appears to apply more to stage 5 and 6 in the information transmission view; that is to the reader not the writer. Surely as a writer you already know the things you are about to write? However, we have all had ideas that are half formulated or ‘feel right’, yet we struggle to articulate. The act of writing (or telling someone about) them is forcing us to process these partially formed ideas. Even if the ideas are internally generated they are not so unlike the externally generated experiences that we encounter as a tiny child – both need to be fully integrated into a coherent worldview. In other words writing makes you think!

However, writing is not just causing internal processing, it is forcing us to process our ideas for others. I often argue that human logic and rational thinking is an accidental side effect of social intercourse. Left to our own devices we can think freely, imaginatively, and associatively ... and then act upon those thoughts. However, if we want others to act alongside us then we need to convince them and give them a reasoned argument. Logic is not there as the way to think, but as the way to describe what you have thought. Of course we have recruited logic and use it as individuals to tackle problems too complex for more informal reasoning, but its origins are in argument. So, in presenting ideas for others we necessarily, to some extent, rationalise them and the coherence and connections in the written argument may be entirely new to us or at least make explicit the tacit reasoning that was already there.

External representation and reflexivity

Even more important than this processing itself is the fact that the ideas and reasoning behind them are on paper (or screen!) and thus available to read, skim and consider outside of your head. In terms of the information transmission model there is an extra stage, after stage 3, when you are your own reader. Indeed taking the idea of reader on board, part of the power lies in the words are slightly distanced from you, you can look at them in a way from the outside. This is the source of its power.

Andy Clark describes language as the “ultimate artefact” a tool that through “loops and circuits that run outside the head” allows us to achieve things we could not otherwise – including producing books and papers [7, p.207]. His perspective is one of ‘embodied activity’ and so his focus is on the external accomplishments of the human–language–paper system on its environment. However, I would like to also emphasise the way this permanently transforms our own thought processes; the effects inside our heads.

While you have an idea it is the thing you are thinking, but when you write it down you can think about it. Writing transforms your thinking into the object of your thinking.

In practice – writing to think

Once we understand how writing transforms thought, we can deliberately use it for the purpose.

writing for yourself

The first advice is the simplest – just write!

I often talk to students who clearly put off writing until they have something clear to say. Of course, the thrust of this article is that one of the best ways to clarify your thinking is to write about it. The writing need not be good, nor coherent, nor grammatically correct; indeed it may be bullet points, mind maps, or simply doodles. The key thing is that the thoughts are in some way represented outside your head. This writing is not for others but primarily for yourself, and often just doing this is enough to create a cycle of reflection.

Do beware of stopping at simple lists of terms. These may capture and externalise key concepts and are certainly a useful first step, but, in my experience, are often vague and lack inter-relationships. The problem is that lists, bullet points and often also mind-maps are more an aide memoire rather than communication and so often miss many of the processing advantages of writing for others. This sounds paradoxical – you are writing for yourself, but you have to do it as if it is for someone else! Two options to increase the usefulness of lists or bullet points is to either write a few sentences after them – making them more like writing, or to turn them into slides and imagine doing a talk about them. Indeed some of my best papers have been written after delivering a short presentation – reducing your ideas to a few slides forces you to bring out the key points.

However, in order to make the most of the words as an externalisation of your thoughts for you to then analyse, the opposite is the case and free text, even short notes, may be too hard to scan rapidly. One technique I’ve used when someone comes to me with a long but ill-structured paper, is to go through it and simply mark where each topic begins and ends. Arguably there should be one idea per paragraph, but often one finds ideas start in one paragraph and end in another, or even flow across section headings! Once the ideas are marked you reduce each to a short phrase and then you effectively have an outline of the ideas in the document (not the section structure). As soon as the ideas are written like this it is usually easy to restructure them, see new relationships. This can then be used to restructure the paper, or simply to help clarify your own thinking.

The reason why it is important to reduce the paper to phrases (effectively bullet points!), is that this enables you to glance at the whole. Once you start to read you get lost in the ideas, your aim is to look at them. This is also the time to remember Miller’s 7 +/– 2 – if your list of bullet points is too long, see if you can structure them so that there are fewer at each level so you can hold them in your head at the same time.

In both cases one should initially start with whatever form of writing best suits your temperament or mood, whether bulleted lists or stream-of-consciousness text – get it on paper – then worry about either making it more detailed, or abstracting out key points later.


Yes, I know this article has been about writing, so where does reading fit in! However, it is common advice to write your own notes about books and papers you read … and what are those notes? … externalisation of course! The obvious purpose of such notes is so that you can read them again later, but the act of writing them involves the same internal processing discussed earlier.

However, annotations can also become tools for thought, just like your own writing.

Have you ever read a paper where each paragraph made sense, but you lost the overall argument? If so, then simply use the same technique described above for your own writing, reduce each paragraph or idea to a short phrase and then sketch the relationships between them. The phrases should be sufficient to remind you of the content they refer to, but if you forget, you can always go back to the paragraph corresponding to a phrase. The aim here is again to reduce the argument of the paper to something that you can see at a glance.

This is particularly valuable if English is not your first language. By reducing the paper to a bullet list you are effectively separating the job of reading the paper into two activities: (i) reading single paragraphs and (ii) understanding the overall message. This separation can be helpful for everyone, but if the act of reading is more complex because it is not your first language, then you are having to expend more cognitive effort at the level of reading individual words and sentences and so do not have as much ‘spare’ to understand the big picture.

The same principles apply to a whole literature review or bibliography. I have always told students to write their own mini-abstracts while reading papers, but used to think of this merely to aid memory, not realising the full power of it. The writing of an abstract that summarises what is important in the paper for you is of course encouraging the internal processing that is valuable for understanding now, not just when you revisit the paper. I also tell students to always produce a annotated bibliography before producing a literature review, mainly to avoid the literature review being an annotated bibliography: “paper X says blah, paper Y says blah blah ...”. However, I have only slowly understood the power of the annotated bibliography as external representation. As with one’s own papers writing the mini-abstract as a series of bullet points makes it so much easier to scan a range of material and see patterns and themes.

Final words

Imagine a cat and then imagine a trip to the dentist. They are totally different in so many ways, but, once they are words, “cat” and “trip to the dentist” become tokens we can manipulate and move around almost without regard for what they refer to. So I can say “imagine a [cat]” and “imagine a [trip to the dentist]” and these two, so different, things are being manipulated and thought about in similar ways.

Once written it becomes easy to notice anomalies: it is sensible to say “stroke a cat” but not “stroke a trip to the dentist”. How do “cat” and “trip to the dentist” differ so that one makes sense and the other doesn’t? And how do “imagine” and “stroke” differ? Notice we are now thinking not about cats or trips to the dentist, but about the idea of a cat, the idea of a trip to the dentist and even in this very sentence the idea of the idea of a cat! Language enables us to think both at higher levels of abstraction but also at higher meta-cognitive levels.

It is not necessary to externalise this language in written words in order to use it in this way. However, most people find it hard to think about their own thinking, or to examine the strength of their own arguments. Externalisation in writing or other ways makes this easier.

While the practical techniques given are focused on the processes of writing and reading itself, they can also be applied when working with other forms of data: transcripts, ethnographic notes, experimental results. In general get used to listening to the words you use when you speak to people about your work – so often there is more there than you think you know, especially in the verbs and linking phrases.

And what about this article, has externalising made a difference to my own thinking? I had, in fact, not expected it to do so as it is a topic I have discussed so many times with students and colleagues and so had already externalised verbally and in so doing performed much of the mental processing and argument construction I expect from writing. That is I expected this writing to be like the initial simplistic information transmission model. But I was wrong. For example, while I had often discussed with students the importance of mental processing, the full relationship with communication and ‘processing for others’ was new.

I didn’t know that before.

It works!


  1. A. Dix (2006). writing as third order experience. Interfaces, 68, pp. 19-20. Autumn 2006. http://www.hcibook.com/alan/papers/writing-third-order-2006/
  2. The orginal reference for gronding theory is Clark and Brennan
    Clark, H. H. and S. E. Brennan. Grounding in communication. In L. B. Resnick, J. Levine, and S. D. Behreno, editors, Socially Shared Cognition. American Psychological Association, Washington DC, 1991.
    However, Andrew Monk did a great introduction to grounding theory:
    Monk, A. F. (2003) Common ground in electronically mediated communication: Clark's theory of language use, In Carroll, J.M. HCI models, theories and frameworks: towards a multidisciplinary science, San Francisco: Morgan Kaufmann Publishers, pp. 265-289.
  3. Speech-act theory has its origins in the work on Searle and Austin and other lingusits (see Wikipedia article), but I first read about it in Winograd and Flores' Understanding Computers and Cognition (Addison-Wesley, 1987)
  4. J. Hollan, E. Hutchins and D. Kirsh: Distributed Cognition: Towards a New Foundation for Human–Computer Interaction Research. Chapter 4 In: Carroll, J. (ed.) Human–Computer Interaction in the New Millennium. pp. 75–94. Addison-Wesley Professional, Boston (2002).
  5. C. Heath and P. Luff (1991). Collaborative activity and technological design: Task coordination in London Underground control rooms. Proceedings of ECSCW 91, 65-80. Dordrecht, The Netherlands: Kluwer Academic Publishers. (PDF)
  6. G. A. Miller. The magical number seven, plus or minus two: some limits on our capacity to process information. Psychological Review, 63(2):81-97, 1956.
  7. A Clark. Being There: Putting Brain, Body And World Together Again, By Cambridge MA: MIT Press, 1997
  8. M. Wheeler. Minds, Things, and Materiality. To appear in Renfrew C. and Malafouris L. (eds.), The Cognitive Life of Things: Recasting the Boundaries of the Mind, McDonald Institute for Archaeological Research Publications, Cambridge.
  9. A. Kidd, A. 1994. The marks are on the knowledge worker. In Proceedings of CHI'94. ACM, New York, NY, pp. 186-191. DOI=http://doi.acm.org/10.1145/191666.191740
  10. L. Vygotsky (1978). Mind and society: The development of higher mental processes. Cambridge, MA: Harvard University Press.

Alan Dix 16/7/2008