Studies have found that some teacher professional development programs that are based on Cognitively Guided Instruction (CGI) can increase student mathematics achievement. The mechanism through which those effects are realized has been theorized, but more empirical study is needed. In service of this need, we designed a novel measure of instructional practice to assess the extent to which observable features of mathematics instruction are consistent with the principles of CGI. We describe the conceptual foundations and first use of the instrument, which we call M-CLIPS. We found that teachers involved in the first 2 years of a CGI program were using methods consistent with the principles. In contrast, instructional practice in the comparison condition was mostly inconsistent with those principles.