Measuring Behavior 2005: Van den Bos et al.

Towards a rodent model of the Iowa Gambling Task in humans

R. van den Bos, W. Lasthuis and B.M. Spruijt

Ethology & Welfare, Utrecht University, Utrecht, The Netherlands

The Iowa Gambling Task (IGT) is an adequate laboratory model to study real life decisionmaking and long-term efficiency of behavior in humans in a situation of uncertain outcomes and a conflict between immediate and long-term pay off of different options. Three questions are of relevance:

What external and internal factors affect long-term efficiency of behavior?
What is its underlying neuro-architecture?
Which neuro-programming rules guide or determine optimal long-term choices?

Thus far no models have been developed in animals to study in depth these questions alongside experiments in humans, although models in rodents exist that study aspects of long-term efficiency, such as delay and probability discounting paradigms. Therefore, we decided to develop a model of the IGT in rodents using its basic tenets, i.e. uncertainty of outcomes and a conflict between different options in terms of immediate and long-term pay off. In this poster we will show a mouse model as well as data of mice. We used an eightarm radial maze of which two arms were used as start arm and four arms as goal arms. In the centre of the maze mice could make their choice once they left the start arms. Internal cues helped the mice to differentiate goal arms. Goal arms differed in the extent to which they contained rewards (‘sugar pellets’) and punishments (‘quinine saturated sugar pellets’). As in the IGT two arms contained occasionally big rewards, but were not profitable in the long run, while two arms contained regular small rewards, and were profitable in the long run. The baiting of the arms was such that mice could not learn explicit rules as to when rewards or punishments would be present. Data show that as in humans C57BL/6 mice initially sample the four arms to arrive at choosing the more profitable arms more often towards the end of the test. Furthermore, C57BL/6 male mice tended to perform better than C57BL/6 female mice, and C57BL/6 females tended to perform better than DBA/2 females.

Paper presented at Measuring Behavior 2005 , 5^th International Conference on Methods and Techniques in Behavioral Research, 30 August - 2 September 2005, Wageningen, The Netherlands.