The Expectancy Valence Model: Can it identify individual performance deficits on the Iowa gambling task?

The Expectancy Valence Model (EVM) is currently being used in numerous studies to estimate neurological deficits in decision making on the Iowa Gambling Task (IGT). This is despite a growing body of evidence that the EVM may not be providing accurate estimates. Using Bayesian estimation techniques of the EVM with random effects, we show the EVM does not provide clear information about the neurological processes underlying deficient abilities at the individual level. Due to bi-modal and non-linear parameter distributions and due to distributions on the boundary of parameter space, parameter estimates that are obtained using the EVM are unreliable and/or have little psychological significance. A multiple run version of the IGT was also trialled in hopes of decreasing the variability of parameter estimates, but no improvement in accuracy was observed. It is therefore found that the EVM should not be used to model performance on the IGT.