Third, the reliability of the combined scales that we used to study changes in risk-seeking andimpulsivity over time was modest, ranging from .58 (age 10) to .75 (age 20) for risk-seeking and from.46 (age 17) to .67 (age 29) for impulsivity. For both scales, this likely reflects the small number of items used to measure them and the conceptual breadth of those items. In the case of impulsivity, it may also reflect the inclusion of an item measuring both the tendency to act impulsively and the desire to avoid its repercussions. Nonetheless, modest scale reliability implies that these variables could be affected by measurement error. A related problem concerns potential measurement variance. The reliability of both risk-seeking and impulsivity scales seemed to improve with age; hence, it is possible the two scales approximate those concepts better at some ages than at others. This could mean that some of the changes we observed in risk-seeking and impulsivity over time were a result of changes in the accuracy with which we could measure the two constructs, rather than of genuine changes in behavior.