Abstract
Background: Neurological disorders may impair decisions involving reward, leading to behaviours such as pathological gambling or excessively risk-averse attitudes: this creates a need for clinical tests to assess different aspects of risky behaviour. Objective: Our aim was to create a realistic test of decisions "under risk" (i.e. when probabilities and sizes of the sums involved are known) with each chance of a win offset by the possibility of loss. We examined thresholds determining when healthy subjects are willing to engage in risky behaviour, and how confidence in a prospect changes with expected value. Methods: 17 healthy subjects were presented with a ‘roulette wheel’ scenario, with varying probabilities and reward magnitudes. The wheel depicted the probability of winning, ranging from 0.2 to 0.8, while a numeral depicted the payout, ranging from 0.8 to 3.2 times their bet. Subjects could decline to play or choose to bet 1 to 3 coins. If they did not win, they lost the sum bet. The dependent variables were the proportion of times when subjects chose to bet 1 or more, 2 or more, or 3 coins. At the end, subject received their winnings. Results: The 50% threshold for participation occurred at an expected value of -0.21 (i.e. on average subjects would lose 20% of their bet). The willingness to bet more coins was a linear function of expected value, with an increase of expected value of 0.45 per added coin. Secondary analyses showed that subjects were twice as responsive to changes in expected value generated by altered probability than to changes from altered magnitude. Conclusion: Our paradigm shows a slight risk-prone tendency in healthy subjects, a linear increase in the confidence of the gamble as a function of expected value, and greater influence of probability than magnitude information in decisions under risk.
Meeting abstract presented at VSS 2012