Dynamic Behaviour of a Spiking Model of Action Selection in the Basal Ganglia Neural Structure

A fundamental process for cognition is action selection: choosing a particular action out of the many possible actions available. This process is widely believed to involve the basal ganglia, and we present here a model of action selection that uses spiking neurons and is in accordance with the connectivity and neuron types found in this area. Since the parameters of the model are set by neurological data, we can produce timing predictions for different action selection situations without requiring parameter tweaking. Our results show that, while an action can be selected in 14 milliseconds (or longer for actions with similar utilities), it requires 34­44 milliseconds to go from one simple action to the next. For complex actions (whose effect involves routing information between cortical areas), 59­73 milliseconds are needed. This suggests a change to the standard cognitive modelling approach of requiring 50 milliseconds for all types of actions. The basal ganglia are generally believed by both neuroscientists (e.g. Redgrave et al., 1999) and cognitive scientists (e.g. Anderson et al., 2004) to be responsible for action selection. Action selection consists of choosing one action to perform out of the many actions in an organism's repertoire. Selection is done on the basis of some sort of context­dependent utility signal for each possible action. Actions that are inappropriate for the current context may have low utility, and a task of the basal ganglia is to select the action that currently has the highest utility value. Since such a mechanism forms the core of many cognitive models, including all of those based on production systems (where a single production much be chosen to fire), it is useful to develop a computational model of this process. Here, we develop a detailed spiking neuron model that takes into account a broad range of neurological details about the basal ganglia. Other spiking models of action selection exist, but tend to be organized unlike the basal ganglia (Belavkin & Huyck, 2009) and unconstrained by neural properties (Shouno et al., 2009; see Humphries et al., 2006 for an exception and alternate approach). By directly connecting our model to neuroscientific results, we constrain our parameter values. Every parameter in the model reflects neurological data from the relevant brain areas, resulting in a model that has no free parameters (that affect the results shown here). Furthermore, having a biologically realistic model allows us to make predictions about a wide range of …