Both the attenuation paradox of traditional test theory and the assumption of local independence in person-item response theory have caused problems in interpretation. This paper demonstrates that the two are related concepts, and, through this demonstration, both are clarified. It is demonstrated that the breakdown of local independence leads to data which manifest the attenuation paradox. A series of simulation studies show the relationship between the degree of violation of local independence and the effects on the traditional KR-20 index of internal consistency. (Author)