| Summary: | The quantification of choice has been a major area of research for behavioural
scientists for several decades. This is, in part, due to the discovery of the matching law (Herrnstein, 1961) which stipulates that relative response rates on concurrently available alternatives “match” the available relative reinforcement rates. This
theoretical construct has been developed to describe response allocation in more complex situations, such as concurrent chains, and successfully describes both human
and non-human behaviour. Typically, this phenomenon becomes evident when
behaviour settles at an asymptote after several sessions of training where
contingencies are held constant, and is often called “steady-state” behaviour.
However, a fundamental question still remains: what causes matching – that is, what are the underlying momentary process(es) that produce matching? Researchers have suggested that what is necessary to answer this question is to take a molecular approach to the analysis of choice behaviour, thereby assessing choice in transition (Grace, 2002a). Recently, a new model of choice acquisition has been developed that
appears to offer promise. It combines two separate mechanisms; a “winner-takes-all” categorical discrimination, and a linear-operator acquisition process (Grace & McLean, 2006). The initial results suggest this model could provide an alternative explanation for what underlies matching – that two separate processes are cooccurring in the acquisition of choice behaviour – allowing response allocation to be either linear or non-linear. This thesis extends the Grace and McLean model to include the situation of response strength ‘carrying-over’ from session to session to describe the process of acquisition gradually accumulating with experience. Moreover, additional assumptions have been added to describe temporal phenomena 2 and presumed discounting of previous experience on current choice behaviour. A steady-state version of the extended model was derived and, when fitted to published
data sets, describes choice behaviour equally well when compared to existing models of steady-state choice. As a consequence of these additions, the Extended Decision Model (EDM) predicts a unique response allocation pattern – choice behaviour follows a bitonic function when initial-link durations were increased and the terminallink
delays were held constant. The results from experiments presented in this thesis support this prediction, whilst steady-state analyses found the EDM was parameter invariant – differences between parameters from two schedule types across several archival data sets were non-significant, while existing steady-state models had significant differences. These findings provide further support for the claim that the
EDM and the Decision Model (DM) mechanisms provide unique and accurate
descriptions of the molecular processes governing choice behaviour. Moreover, the implication from these results is that the underlying assumption of the EDM and DM – that choice is determined by the propensity to respond rather than conditioned reinforcement – appears to have further foundation. This challenges the assumptions
of existing models of choice behaviour and presents the possibility that probabilistic approaches are perhaps more appropriate for describing response allocations than discrete estimates of relative value when contingencies change.
|
|---|
