This paper describes the cognitive methods used and the findings derived from a test of new forms for the Current Employment Statistics (CES) Survey conducted by the Bureau of Labor Statistics. CES currently collects five data elements from a multimode sample of about 400,000 businesses. The new forms add employment, hours, and earnings for all employees to the currently collected data on production, construction, or nonsupervisory workers, thereby doubling the number of data elements. The revised forms also introduce a new data element: gross monthly earnings (GME)—-a complex concept with a different reference period and definition than the other data items. A sample of current CES respondents was used to test the new forms with a range of establishments varying in size, industry, response mode, pay schedule, and commission payments. Respondents were assigned randomly one form version (out of four variants) to complete prior to the cognitive interview. In addition, two reference period definitions for GME were tested. This allowed us to explore the optimal form layout while examining the substantive interpretations of all new data items.