The expectancy valence model of the Iowa Gambling Task: Can it produce reliable estimates for individuals?

The Expectancy Valence Model (EVM) of the Iowa Gambling Task (IGT) is commonly used in studies to identify the underlying psychological processes responsible for decision making deficits.We show the EVM does not provide clear information about decision making processes at the individual level by fitting the EVM, with individual random effects, to a sample of participants from various drug using populations using Bayesian techniques and to a sample of participants who complete the IGT multiple times. In particular, we show that the individual-level parameter estimates from the model may be bi-modally distributed and hence are inherently ambiguous and have little psychological significance.In an attempt to increase the validity of individual-level parameter estimates, we also considered a 2-parameter version of the EVM in which the consistency parameter was held constant. In the 2-parameter implementation of the EVM, results were clearer and more easily interpretable than when using the traditional EVM.