Neural oscillations are ubiquitous across recording methodologies and species, broadly associated with cognitive tasks, and amenable to computational modeling that investigates neural circuit generating mechanisms and mesoscale dynamics. Because of this, neural oscillations may offer an exciting potential opportunity for linking theory, physiology, and mechanisms of cognition. However, despite their prevalence, there are many concerns—new and old—about how our analysis assumptions are violated by known properties of field potential data. For investigations of neural oscillations to be properly interpreted, and ultimately developed into mechanistic theories, it is necessary to carefully consider the underlying assumptions of the methods we employ. Here, we discuss seven methodological considerations for analyzing neural oscillations. The considerations are to 1) verify the presence of oscillations, as they may be absent; 2) validate oscillation band definitions, to address variable peak frequencies; 3) account for concurrent non-oscillatory aperiodic activity, which might otherwise confound measures; measure and account for 4) temporal variability and 5) waveform shape of neural oscillations, which are often bursty and/or nonsinusoidal, potentially leading to spurious results; 6) separate spatially overlapping rhythms, which may interfere with each other; and 7) consider the required signal-to-noise ratio for obtaining reliable estimates. For each topic, we provide relevant examples, demonstrate potential errors of interpretation, and offer suggestions to address these issues. We primarily focus on univariate measures, such as power and phase estimates, though we discuss how these issues can propagate to multivariate measures. These considerations and recommendations offer a helpful guide for measuring and interpreting neural oscillations.