Like many states throughout the nation, Pennsylvania is in the midst of major reforms to its teacher evaluation system. Under the new system, the state will base annual evaluations on several measures, including supervisor observations using the Framework for Teaching (FFT) and, for many teachers, their contributions to student achievement growth from a value-added model (VAM). In the past, there was concern that supervisor observations did not differentiate performance well or relate to true teacher performance. In this study, we investigate how well the new system has addressed these issues by analyzing the degree to which FFT scores differentiate performance, are internally consistent, and correlate with teachers' contributions to student achievement growth as measured by VAM scores. This report is based on data from a pilot of the new system covering 6,676 teachers from 269 districts in the state of Pennsylvania, including Pittsburgh public schools. The data include the 22 components of the FFT, each of which is designed to capture a separate teaching practice. We used these data to estimate four domain scores and one overall Professional Practice Rating (PPR) score. We also merged these scores with data on teachers' estimated contributions to student achievement growth. Based on these pilot data from the 2012--2013 school year, we estimate that, although less than 13 percent of teachers received the top rating ("distinguished") for the overall PPR score, almost 85 percent were rated in the second highest category ("proficient") (Figure S.1). Less than 0.1 percent were rated in the bottom category ("failing"). The remaining teachers (around 2.6 percent) were given "needs improvement" ratings. We found that FFT scores were internally consistent, meaning that the domains and the components within each domain appear to be measuring similar concepts. We also found that teachers with higher FFT scores tended to produce greater student achievement growth. The correlations of the FFT scores with VAM scores were all positive and generally statistically significant, ranging from 0.19 to 0.22 by domain. We compared the results based on the 2012--2013 data with results based on 2011-2012 data from a previous pilot phase. For the most part, the findings were similar. More than 90 percent of teachers were rated in the top two performance categories in both phases, although the fraction of ratings in the top two categories decreased somewhat in Pittsburgh (which contributed more teachers to the pilot than any other district). The levels of internal consistency were in the acceptable to good ranges in both phases, with the overall PPR score having higher consistency than any of the domain scores in both phases. The correlations between parts of the FFT and VAM scores were almost always positive but also below 0.30 in both phases. The lowest correlations in 2011-2012 were slightly improved in 2012-2013. In sum, although FFT scores are overwhelmingly concentrated in the top two performance categories, the positive correlations with VAM suggest that the FFT provides some meaningful differentiation and captures aspects of teacher skills related to student achievement growth.