Parameter definitions and specifications for population modelling simulations of English past-tense acquisition

This document contains technical details to accompany several papers that employ population modelling of the acquisition of English past tense formation. These simulations are either directly applied to capturing empirical data of children learning the past tense (Thomas, Ronald & Forrester, submitted; Thomas, in preparation), or employ the past tense domain as an abstract learning problem within cognition in preparation a, in preparation b). These simulations explore a population of networks acquiring the past tense domain, where individual variability is included both in the parameters of the artificial neural networks which model the children's learning systems, and the learning environment to which they are exposed. The parameters of the artificial neural networks are encoded in an artificial genome. Population variability in parameters is created by generating and breeding populations of artificial genomes. Each genome is realised as a parameterised network. The network is exposed to an individualised learning environment, generating a trajectory of behavioural development. The inclusion of an artificial genome level in the simulations allows two avenues of exploration. First, it allows us to study the associations that can arise between values on the artificial genome and behavioural variability that is the product of an implemented developmental process (see Figure 2). Second, it allows us to generate