The relationship between intelligence and personality has been of scientific interest for over 100 years. However, most contemporary estimates of these relationships are limited because they do not separate the variance due to general and narrow cognitive abilities. This study demonstrates that this methodological oversight can distort estimates of intelligence-personality associations by masking true effects and falsely showing others. To test this proposition, we examine correlations between several personality and ability scales, and then repeat the analyses using latent modeling techniques where variance due to general intelligence (g) and narrow mental abilities is appropriately separated. Our results show that estimates of specific intelligence-personality associations based on observed test scores can be both erroneously inflated or deflated.