© 2015 Acoustical Society of America. Electropalatographic and acoustic data on Catalan /Ck#C/ sequences where # is a word boundary and /k/ is preceded by /l, s, r/ and followed by /b, m, d, l, n, z, r, λ/ were collected in order to test the hypothesis that the velar stop is most prone to be reduced and deleted next to consonants involving high articulatory and aerodynamic demands. Analysis results reveal the absence of a velar stop closure in about half of the sequence tokens, mostly so when /k/ occurs after /s/ and before an oral stop presumably due to the high manner of articulation requirements involved. On the other hand, /Ck#C/ sequences where a /k/ closure period is available show a prominent realization of the velar stop mostly next to /s, z/. This scenario points to two different production mechanisms for three-consonant sequences with contextual obstruents: articulatory reduction and elision, and a slowing down and an increase in articulatory salience, of the velar stop. /Ck#C/ sequences lacking an acoustic closure for /k/ were found to show a residual velar stop articulation which was implemented through an increase in cluster duration and in dorsopalatal contact at the approximate /k/ location in comparison to identical /C#C/ sequences with no /k/.