Across the United States, primary care practices are engaged in demonstration projects and quality improvement efforts aimed at integrating behavioral health and primary care. Efforts to make sustainable changes at the frontline of care have identified new research and evaluation needs. These efforts enable clinics and larger health care communities to learn from demonstration projects regarding what works and what does not when integrating mental health, substance use, and primary care under realistic circumstances. To do this, implementers need to measure their successes and failures to inform local improvement processes, including the efforts of those working on integration in separate but similar settings. We review how new research approaches, beyond the contributions of traditional controlled trials, are needed to inform integrated behavioral health. Illustrating with research examples from the field, we describe how research traditions can be extended to meet these new research and learning needs of frontline implementers. We further suggest that a shared language and set of definitions for the field (not just for a particular study) are critical for the aggregation of knowledge and learning across practices and for policymaking and business modeling.