Complex dynamic systems also predict dissociations, but they do not reduce to autonomous components

Dissociations, according to the target articles, are due to damaged autonomous phonologic (or spelling) representations. However, a damaged recurrent network model may also produce dissociations. Recurrent networks do not entail autonomous components. They are strongly nonlinear dynamic systems that self-organise through recurrent feedback. A simple model with these properties that produces both regularisation errors (PINT named to rhyme with MINT) and semantic errors (BUSH named as TREE) is described. It may also produce dissociations between ''spoken'' responses and ''written'' responses. The mathematical basis of this model is motivated by contemporary neurobiological accounts that also derive from dynamic systems theory. The mathematical basis may also predict multistability and metastability. These are indicated by hysteresis and 1/f noise, respectively, and we review recent reports of these phenomena in speech perception and word recognition. In addition, feedback has been corroborated in the feedback consistency effect. Reported generic behaviours of a complex system, the simulated dissociation of errors, and the established bidirectional nature of perception all demonstrate the utility of a cognitive systems approach to cognitive phenomena.