‘doing’ in the performance of skilled activities (Sutton, 2007:776). Sutton (2007: 779) concludes that,
both developing and enacting high levels of skill require us not to cut intellect and emotion off from our embodied, grooved performances, but to achieve and then access unusual flexibility in linking thought and action, knowledge and motion, conceptual memory and procedural memory
In all the cases considered above, the two-way commerce between brain, body and world shows that the forms of our embodiment, action, and engagement are not fixed. It suggests that although conscious knowledge and bodily knowledge are distinctive ways of knowing each with its own distinctive effects, they often function in a coordinated way. Drawing can thus be understood, as Katherine Hayles (2006:9) suggests of literature, as a technology, which creates new pathways between these different types of knowing which “typically remain unevenly articulated with one another”.
represented in different formats at different levels. Her experiments indicate that human learning often starts by forming exactly the kind of interwoven knowledge structure typical of connectionist systems. A number of researchers in the connectionist fold have shown interest in models of learning and have acknowledged, even emphasised, that connectionist systems are entirely in accord with tacit knowledge, as described by Polanyi (Nelson and Nelson, 2003). A paper co-authored by Andy Clark and Karmiloff-Smith (1993a: 513), focuses on “how to achieve flexibility, manipulability, and transportability within a broadly connectionist setting”. They claim that although connectionist models explain the first phases of learning in some new area rather well, they are weak at modelling those changes in representation which distinguish the advanced and expert practitioners of a particular field from their less able peers.
3.1 From knowledge ‘in’ the system to knowledge ‘to’ the system
The subtle changes that occur in children’s performance are paralleled in the connectionist field when neural networks become more powerful, be it because of changed patterns of input or through the use of additional concealed units. However, the representations that we ascribe to such networks do not exist explicitly in some identifiable part of the network;
they are but an implied feature of how the network as a whole performs. “Whilst this is the endpoint of learning in a connectionist network, in the human case it is the starting point for generating redescriptions of implicitly defined representations” (Clark and Karmiloff- Smith, 1993: 488). Clark and Karmiloff-Smith (1993: 492) characterise the knowledge embedded in a first-order special-purpose pattern recognition/connectionist system as
“inextricably intertwined” which in effect generates a representation system that is adapted to that specific domain. At this first level procedural/implicit or tacit knowledge is so entangled in the network of connections that “it is knowledge in the system, but it is not yet knowledge to the system” (Clark and Karmiloff-Smith, 1993: 495).
Without control structures that are able to unpick parts of the web while it preserves others, such interweaving “makes it practically impossible to operate on or otherwise exploit the various dimensions of our knowledge independently of one another”. This poses a problem if we need our knowledge to generalise in an adaptable manner (1993: 495). The representational redescription approach proposes that humans are different to
connectionist networks in this respect, as they cannot but continue to develop a series of supplementary representations. We are hereby enabled to process and utilise our own
stored knowledge in ever more flexible and mutually independent ways. The following quote sums up the Representational Redescription claim:
For the genuine thinkers, we submit, are endowed with an internal organisation which is geared to the redescription of its own stored
knowledge. This organisation is one in which information already stored in an organism’s special purpose responses to the environment is subsequently made available, by the RR process, to serve a much wider variety of ends.
Thus knowledge that is initially embedded in special-purpose effective procedures subsequently becomes a data structure available to other parts of the system (Clark and Karmiloff-Smith, 1993: 487-488).
3.2 Representational change
Karmiloff-Smith’s (1979; 1986; 1990) data suggest that knowledge already represented in an implicit form spontaneously translates to explicit representations when new skills are rehearsed. As these representations progressively develop at ever-higher levels, previously acquired knowledge can be utilised in ways that were initially impossible. Skills which, though fairly sophisticated, could not be modified easily now become more flexible and adaptable. To investigate this representational change, a number of studies were conducted of how children draw. Over fifty children aged between four and eleven years old produced six drawings each. The children were first asked to draw ‘a house’, followed by further requests for ‘a house that does not exist’, ‘a pretend house’, and so on. Children were also invited to draw a man, followed by a ‘funny man’, etc. Each child was then observed closely to determine how they went about the drawing task.
The results showed that the flexibility of the child’s drawing ability increased with age. Four year olds were only able to delete elements in certain ways and make simple changes
regarding shape and size. Ten year olds can be more adventurous in their approach. They added new components, altered the way components or the entire image was positioned or orientated, and even added parts coming from other conceptual fields (Clark and
Karmiloff-Smith, 1993:501). The older children’s ability to adjust, change and integrate representations across domains thus increased.
Karmiloff-Smith hypothesises that this expanded imagination is the result of children generating explicit representations of what was before only known implicitly. Implicit knowledge can be applied, but not reviewed or modified. Explicit descriptions, on the other hand, enable modifications that are only possible thanks to the redescription of the
Because the four year olds’ drawing skill functions so predominantly at an implicit level, they cannot move beyond a rigid, “automatic” sequence of bodily operations and can barely generate any alternative versions of the things produced as part of mastering the skills in question.
A drawing skill is initially acquired as an inflexible chain of physical gestures. In this phase the mind represents the skill in terms of a rigid succession of components, which is either carried out from start to finish or not at all. At this level the skill is inflexible and does not allow for re-ordering of the parts or the insertion of extra elements into the drawings.
Drawing the first line activates a series of steps that cannot be interrupted or amended.
The level of description that follows is less constrained and allows steps to be removed one at a time and the sequence to be altered without interfering with the subsequent drawing procedure. This is possible because the skill is represented as a set of separate units, each of which can be repeated or repositioned separately to produce a variety of possible
arrangements. As representation develops further and becomes more explicit still, the arrangement of, and relations between, the second level units becomes more elastic. A ten- year-old child’s conceptual space involves many more dimensions than that of the four year old, thereby again making the range of what can be produced broader and more interesting.
According to Karmiloff-Smith “conscious self-reflection” is the consequence of multi- levelled representations of the type described here (Boden, 2004: 84).
Boden (2004: 85) argues that it is very likely that the type of representational changes shown to occur after a child has achieved fluency in drawing skills also allows adults to generate multi-dimensional conceptual spaces by redescribing previously acquired skills in successively more sophisticated ways. When a skill becomes more multi-dimensional and refined, and subject to sophisticated control, it allows all sorts of structures that were formerly merely implicit, as well as domain-specific, to become available to awareness. As implicit mental processes become supplemented by more or less explicit maps of these processes, the conscious exploration of possibilities is made easier.
Clark and Karmiloff-Smith (1993: 515) conclude that the Representational Redescription model depicts the “true cognizer” as a multi-faceted representor of its (external and internal) world” who “must somehow manage a symbiosis of different modes of
representation”. They argue that the Representational Redescription model guarantees such a symbiosis by “invoking a developmental process in which the more structured
representations arise as a result of the system’s endogenous drive towards the analysis and re-representation of its own cognitive states” (Clark and Karmiloff-Smith, 1993: 515). The Representational Redescription model thus complements the ideas of Heidegger and others discussed earlier, by implying that human cognition operates within a context of activity, through the representational actions or deeds performed in the world. The way in which active re-representation allows us to construct new and perhaps more complex
representations is a theme that will recur throughout this thesis. The suggestion that re- representation enables knowing how to become knowing that makes the environment a crucial extension of our minds.
Chapter 3