Abstract In functional MRI (fMRI), dynamic functional connectivity (dFC) typically refers to fluctuations in measured functional connectivity on a time scale of seconds. This perspective piece focuses on challenges in the measurement and interpretation of functional connectivity dynamics. Sampling error, physiological artifacts, arousal level, and task state all contribute to variability in observed functional connectivity. In our view, the central challenge in the interpretation of functional connectivity dynamics is distinguishing between these sources of variability. We believe that applications of functional connectivity dynamics to track spontaneous cognition or as a biomarker of neuropsychiatric conditions must contend with these statistical issues as well as interpretative complications. In this perspective, we include a systematic survey of the recent literature, in which sliding window analysis remains the dominant methodology (79%). We identify limitations with this approach and discuss strategies for improving the analysis and interpretation of sliding window dFC by considering the time scale of measurement and appropriate experimental controls. We also highlight avenues of investigation that could help the field to move forward.