An investigation of the influence of internal test bias on regression slope

In educational and psychological tests, 2 forms of bias have been distingui shed: internal and external (or predictive) bias. However, the relation bet ween internal and external test bias has not been clearly defined or explic ated. Studies of predictive bias show that when regression equations across groups are compared, the most commonly occurring difference is in intercep ts rather than in slopes. Thus, does internal test bias not produce slope d ifferences? To answer this question, emphasis was given in this study to an investigation of the relation between internal test bias and slope differe nces. Monte Carlo simulation results indicate a strong relation between int ernal test bias and slope differences. However, this relation does not nece ssarily imply that an absence of internal test bias leads to slope invarian ce. Conversely, slope differences do not necessarily imply internal test bi as. Slope invariance holds only when both target ability differences and te st bias do not exist. Finally, we discuss the reasons slope differences are rarely observed in practice.