Cannabis use correlates with numerous adverse brain and behavioral outcomes, including increased risk of mental health problems. However, disentangling cannabis’s causal effects from confounding environmental and genetic factors remains challenging due to ethical constraints on human experimental studies and the fact that cannabis use does not occur randomly in the population. This chapter examines the magnitude of this issue, focusing on residual confounding—the persistent bias from unmeasured variables that correlate with both cannabis use and outcomes. We present the “co-twin control” or “twin difference” study design as a powerful approach for mitigating residual confounding and improving effect estimation. We review findings from both non-twin and twin studies on cannabis-related brain and behavioral outcomes, highlighting how twin studies elucidate and extend results from conventional observational research. We conclude by discussing the strengths and limitations of the twin comparison approach and propose future research directions. This overview provides researchers and clinicians with a critical framework for interpreting cannabis-outcome associations as well as a guide to one observational framework for more accurately estimating cannabis’s true effects.