Hysteresis loops of exchange-biased permalloy/Fe50Mn50/permalloy trilayers on glass were measured as a function of Fe50Mn50 and permalloy thicknesses with the longitudinal Kerr effect employing a coherent light source. Kerr signals originate from both permalloy layers and give a superposition of hysteresis loops. In vibrating sample magnetometer or looptracer measurements the contribution of a particular layer to a major hysteresis loop cannot be identified. With the Kerr setup presented it is possible to identify the contribution of each layer individually, since the finite optical path through the trilayer gives rise to optical attenuation and interference. For an increasing total thickness of the trilayer, the signal of the buried permalloy layer will become weaker due to attenuation. Kerr measurements of trilayers up to 40 nm show a superposition of two equally oriented loops. Major loops for trilayers of thickness greater than 40 nm show a superposition of two oppositely oriented hysteresis loops. The transition is dependent only on the total thickness of the trilayer, rather than the thickness of each individual layer. This unusual effect can be explained by the phase difference of the two Kerr signals. Additional measurements performed from the glass side of the samples and measurements of a Fe50Mn50/permalloy bilayer confirm that the sense of a hysteresis loop can change for a buried layer due to optical effects. (C) 1999 American Institute of Physics. [S0021-8979(99)06903-0].