This paper examines the phenomenon of final consonant deletion in clusters which do not agree in voicing and compares this phenomenon with clusters sharing the voicing feature. The speech studied is that of Puerto Rican and black Harlem teenagers. The data reported here refutes many of Bailey's (1972) claims. Clusters where voicing is not shared are found to simplify differently than shared voicing clusters. Clusters with "l" as the first member are found to behave differently than clusters beginning with a nasal, and nasal clusters behave differently depending on the final consonant. Certain of these clusters also behave differently from previously studied consonant cluster simplification, as well as final postvocalic consonant deletion. That all of these deletion phenomena show different frequencies of simplification, as well as different constraints on simplification, indicates that variability rules for consonant cluster simplification must be further refined and new rules added. Such refinements and new rules are proposed in the present paper, and the implications of these new rules on the ordering of all deletion phenomena are discussed. (Author/AM)