We examine whether one of the most popular observation systems in teacher evaluation--the Framework for Teaching (FFT)--captures the range of instructional skills teachers need to be effective. We focus on the case of special educators, who are likely to use instructional approaches that, although supported by research, are de-emphasized in common observation systems. Drawing on 206 lessons from 51 teachers, we compare FFT scores to an observation system from special education. We find that FFT's psychometric properties are consistent with previous studies, but the system is limited in assessing the quality of instructional practices used in special education. We discuss implications of these findings for two practical uses of observations--supporting teacher development and informing human capital decisions.