Undergraduates responded to an objective test in electronics and classified each item by domain (one of 14 topics covered in their text), and by type of knowledge (definition, fact, principle, or interpretation). These judgments were compared to their instructor's "standard" judgments. From these data, an index of item-domain divergence in perceived item meaning between examinees and the content specialist was computed, in addition to traditional item statistics. This new measure provides a means for detecting ambiguous items not otherwise detectable using biserial correlations. Items deemed ambiguous by this procedure were compared with items indicated as being ambiguous by students in taped interviews. Avenues for future research are discussed. (Author/CP)