ABSTRACT
Neural plasticity, that is, the ability of the brain to change its function, is a current and fundamental issue in neuroscience (e.g., Buonomano & Merzenich, 1998; Tallal, Merzenich, Miller, & Jenkins, 1998). Work at this level of basic science has shown that the mature brain is, in principle, capable of “rewiring” itself so that new functions can be learnt by brain areas that previously performed other processes (Buonomano & Merzenich, 1998). This basic science has been extended to the applied level (e.g., Tallal et al., 1998; Wilson & Evans, 1996). This applied research suggests that a particular form of therapy/remediation programme, known as errorless learning, might have advantages over more traditional trial-and-error methods. The basic premise behind errorless learning is that learning/recovery may be limited by patients’ errors in that, not only may they correctly reinforce the link between stimulus and correct responses but they might also reinforce the association with erroneous responses. By adjusting the intervention such that the patients are much less likely to make errors, better learning arises because the patients reinforce only the correct response. In turn, computational neuroscience is beginning to provide a link between clinical application of errorless learning and basic neuroscience. This work has shown that the functioning of neuron-like processing units does not alter if the system reinforces its own errors, but change can follow in circumstances like errorless learning (McClelland, Thomas, McCandliss, & Fiez, 1999).
