|
Article Excerpt
Until recently, road safety education in Britain focused primarily on classroom-based teaching of general rules intended to specify action under many circumstances (e.g. the Green Cross Code's 'stop, look, listen, think'), obviating any need for children to learn how to deal with individual situations (Thomson, Tolmie, Foot, & McLaren, 1996). In practice, however, children commonly fail to see how such rules relate to actual events (see e.g. Ampofo-Boateng & Thomson, 1990), and knowledge of rules is commonly unrelated to road crossing performance (Ryhammer & Bergland, 1980).
For many branches of developmental and cognitive theory these difficulties are not surprising. As Thomson et al. (1996) note, there is a consensus that learning proceeds from the specific to the general, from action to representation. The start point is therefore typically participation in an activity, not generalized instruction. This position can be ascribed to both Piagetian and Vygotskian theory (Piaget, 1985; Vygotsky, 1978), given their constructivist nature; and to Gibsonian theory and connectionist modelling (Gibson, 1966; Bechtel & Abrahamsen, 1991), which both characterize learning as the abstraction of invariances from specific inputs. Among contemporary approaches to developmental theory too, Karmiloff-Smith's (1992) representational redescription (RR) model proposes a general four-level sequence of developmental change, in which initially context-bound procedural knowledge (implicit or I level representations) is gradually transformed into explicit formulations (E level representations), making it available in other contexts, first to the self (E1 and E2 levels) and then to others via encoding in language (E3 level).
The implication is that, in the ordinary course of events, generalizable pedestrian skills are built up over time from specific encounters with traffic environments. To be successful, then, road safety education needs to provide experience of making actual road crossing decisions in safe and controlled contexts, affording the chance both to learn appropriate patterns of behaviour and to extend these to other situations. The efficacy of this approach has been confirmed in the Netherlands (Rothengatter, 1984) and in the UK. Thomson and Whelan (1997) examined the impact of training sessions in which adult volunteers worked with triads of 5-6-year-olds on the identification of safe places to cross roads (i.e. start points and routes which avoid dangers such as reduced visibility and spending too long on the road). Children were taken to various locations, where they were set road-crossing goals, and engaged in dialogue about the features that would make some routes unsafe, and how these might be dealt with (e.g. by moving to a place where visibility was better). The effects of training were gauged from children's unsupported road-crossing solutions at a pre-test, and on post-tests shortly after training and 3 months later. Children who had 4-6 30-minute sessions over a month showed robust improvements, making three times as many safe route decisions, and performing above the level typical of untrained 10-year-olds (Tolmie et al., 2002).
In related work, Ampofo-Boateng et al. (1993) trained 5-year-olds individually in safe place identification either at the roadside or on a table-top model with comparable road layouts and 'parked' toy cars. Both produced substantial improvements in the selection of safe crossing places, with trained children able to articulate and explain solutions of which they showed no knowledge at pre-test. Since all pre- and post-testing occurred at the roadside, these improvements must moreover have depended upon generalization across contexts amongst those trained on the model. Thus practical training methods of the type employed by Thomson and colleagues apparently have the potential to promote not just improved performance, but explicit, transferable knowledge equating with E3 level representation (conscious knowledge available in other contexts and communicable to others) in the RR model.
If there is a gap in this work, it is that the mechanisms by which training produces high-level representation are unclear. No record of the content of training sessions was kept by either Ampofo-Boateng et al. (1993) or Thomson and Whelan (1997). It is therefore difficult to determine whether improvements in children's performance depended on the exogenous influence of dialogue with trainers or (where relevant) other children; or, conversely, on endogenous processes following exposure to road-crossing problems, as the RR model assumes (Karmiloff-Smith, 1992). There are signs that cognitive engagement with the problems was influential, since untrained children employed as controls in Thomson and Whelan's research showed improvement pre- to post-test, despite the absence of feedback from testers. There are also cogent reasons, however, to suppose that it was trainer activity that was critical.
Vygotsky (1978) proposed that children learn primarily by taking over control of actions initially performed under guidance: tutors use language to help direct the child's activity, and by appropriating and then internalizing the linguistic encoding of action this involves, the child becomes able to direct him- or herself. Wood (1986) offers a similar formulation. Tutors employ physical demonstrations (e.g. putting two pieces of a puzzle together) to help children establish basic procedures (cf. I level representations). Once these procedures are in place, though, they switch to increasingly oblique verbal directions which embed procedural knowledge within explicit and generalizable strategies (e.g. 'How about starting by putting the edge pieces together?'). Both accounts point to the same mechanism as producing E3 level representations in the earlier research: the encoding in language of problem elements and problem solutions by adult trainers. Support for the effects of linguistic encoding on shifts in representational level is provided by Pine, Messer, and Godfrey (1999). Children aged 5-7 years who saw demonstrations of solutions to balance beam problems and heard explanations of these progressed in understanding more than those who worked alone, and as much as those who were guided to their own solutions, suggesting that verbal encoding was the crucial influence on learning.
Interaction between the children being trained in groups in Thomson and Whelan (1997) may also have contributed to learning. For complex problems where many solutions are possible, as in road crossing, the application of knowledge to different contexts depends not just on it having been made explicit, but also on an ability to decide which elements of knowledge to accord the greatest weight in any given instance. The assessment of different viewpoints that commonly arises in discussion between peers may therefore have provided children being trained with an important experience of explicitly rehearsing this decision-making process, helping transform procedural knowledge into a more explanatory conceptual framework. Certainly, there are established effects on conceptual grasp of the evaluation of alternative ideas during peer collaboration (e.g. Howe, Tolmie, Greer, & Mackenzie, 1995; Tolmie, Howe, Mackenzie, & Greer, 1993; Williams & Tolmie, 2000). More specifically, there is evidence that such effects extend to an understanding of the rationale guiding procedural decisions (Howe, Tolmie, Duchak-Tanner, & Rattray, 2000; Miell & MacDonald, 2000; Pine & Messer, 1998; Tolmie, Howe, Duchak-Tanner, & Rattray, 1999).
These arguments are necessarily speculative in the absence of more direct evidence. The way forward would be to make a systematic comparison of the effects of different versions of the same training programme permitting (a) only adult guidance, as in Ampofo-Boateng et al. (1993); (b) only peer discussion; and (c) a combination of the two, as in Thomson and Whelan (1997). This would allow the unique and the additive or interactive effects of these potential influences to be gauged. Clear linkage between the acquisition of generalizable E3 level representations and one or other type of interaction during training would establish the influence of this as central to the outcome of practical pedestrian training, and demonstrate that individual cognition was not the key determinant of progress.
The present paper reports on two studies designed to make the proposed comparison. Study 1 contrasted the effects of adult guidance of individual children with those of peer discussion, whilst Study 2 compared the impact of adult guidance of individuals and groups. These studies had two principal aims. The first was to establish whether practical training results, as hypothesized, in the growth of explicit, communicable knowledge (i.e. E3 level representations) which assists the generalization of performance. The second was to examine whether this development, assuming it occurs, can be attributed to adult encoding in language of problem elements or peer discussion of alternative ideas. In order to provide the necessary standardization and control of training sessions, both studies used computer simulations to train 5-8-year-olds in roadside search skills (i.e. direction of attention to features relevant to road crossing, such as vehicle movements and road layout). This was chosen as the focus of training because Tolmie, Thomson, Foot, McLaren, and Whelan (1998) report that 6-8-year-olds perform poorly in this area, differentiating relevant from irrelevant features (e.g. vehicle colour, a man on a ladder against a house) at no better than chance level both on simulations and at the roadside. Children in the present research were therefore likely to possess little explicit knowledge of what to attend to prior to training, permitting development to be more readily detected. Moreover, if the outcomes for roadside search skills were consistent with those observed for safe place location, this would establish the general nature of the processes involved.
STUDY 1
Study 1 assessed the independent impact of adult guidance and peer discussion on children's roadside search skills. Identical simulation-based training exercises were carried out either by individual children working under the guidance of an adult volunteer, or else by small groups of children working collaboratively on their own. These exercises required them to view a series of animated traffic scenarios, and to decide if and when it would be safe for a figure depicted at the roadside to cross. To solve these problems, children needed to learn to attend to road layout and vehicle movements, and ignore other features. Progress was assessed via pre- and post-tests both on computer and at the roadside.
Before commencing the research, it was necessary to identify the dialogue elements characteristic of adult guidance and peer discussion, so that the incidence of these could be reliably extracted from recordings of the training sessions. The process of contingent support (Wood, 1986) is central to generic accounts of adult guidance. This involves the adult intervening to steer the child only when they falter, and then at the least directive level that achieves the desired result (e.g. prompting rather than demonstrating), since this encourages autonomy and appropriation of control of previously supported action. Recent research (e.g. Howe & Tolmie, 1998; Tolmie et al., 1999) identifies four types of dialogue as characteristic under these conditions: (a) instructions; (b) suggested actions; (c) questions requesting ideas about what to do; and (d) prompts which draw attention to factors not yet taken into account. These elements primarily constitute different forms of guidance, deployed according to the progress exhibited by the child, with a shift from instruction towards prompting occurring as children's competence improves.
Peer discussion is typified by the process of socio-cognitive conflict (Doise & Mugny, 1984; Piaget, 1932). Where individuals present incompatible interpretations of an event, this is argued to cause conceptual conflict. This in turn provokes a process of conflict reduction via individual or joint construction of new concepts that resolve the discrepancies by combining aspects of the alternative ideas available. Research (e.g. Howe et al., 1995; Tolmie & Howe, 1993; Tolmie et al., 1993) pinpoints a sequence of three types of dialogue as characteristic of this process amongst younger children: (a) suggestion of an idea; (b) disagreement with that idea; and (c) explanation justifying the original idea. Related research (see Tolmie & Howe, 1994) indicates that chairing dialogue (i.e. interactions specifying what should be done by whom) is characteristic of unproductive peer discussion, this being what children fall back on when they are struggling with the task in hand and lack the means to progress.
Using this specification of anticipated dialogue elements, the objectives of Study 1 were to examine: (a) whether the two training conditions led to different degrees of improved performance post-training; (b) whether either led to an increased incidence of E3 level representations and an associated generalization of performance from computer to roadside; and (c) whether observed changes in performance could be related...
|
