Human-associated microorganisms play a vital role in human health, and microbial imbalance has been linked to a wide range of disease states. In this Review, we explore recent efforts to progress from correlative studies that identify microorganisms associated with human disease to experiments that establish causal relationships between microbial products and host phenotypes. We propose that successful efforts to uncover phenotypes often follow a chain of evidence that proceeds from (1) association studies; to (2) observations in germ-free animals and antibiotic-treated animals and humans; to (3) fecal microbiota transplants (FMTs); to (4) identification of strains; and then (5) molecules that elicit a phenotype. Using this experimental 'funnel' as our guide, we explore how the microbiota contributes to metabolic disorders and hypertension, infections, and neurological conditions. We discuss the potential to use FMTs and microbiota-inspired therapies to treat human disease as well as the limitations of these approaches.