Trained musicians rated the similarity of 24 instances of set classes [0137/0467] and [0157/0267] at three different transposition levels and two different spacing types. Stringent criteria for retention of participants, to ensure greater reliability and predictive power, resulted in a final count of 30 participants. Participants' data were analyzed by using multidimensional scaling and additive tree methods. A three-dimensional multidimensional scaling solution showed clear effects for spacing type and transposition level (Tn). Additive tree analysis showed no grouping according to set-class type, but rather according to pitch height of inner voices or ordered location of semitones. The results are consistent with the hypothesis that even musicians with significant experience of atonal music do not use the equivalence relation TnI in making similarity judgments.