Dissolved oxygen (DO) observations from in situ sensors show complex temporal patterns, suggesting that the balance of control by underlying processes changes across scales. At scales ranging from minutes to days, a number of physical and biological processes, such as internal waves, mixing, and ecosystem metabolism, may impart pattern on observed DO. In discriminating the control over DO variability by scale, this helps us to reduce uncertainty in estimates of important ecosystem rates, such as gross primary production and respiration. In this study, we examined DO variability over scales ranging from minutes to days and assessed the relative contributions from several physical and biotic drivers. High frequency measurements of DO, wind, temperature, and photosynthetically available radiation (PAR) were obtained over periods of approximately 4 d from 25 lakes in northern Wisconsin. Patterns in data were isolated by time scale through wavelet transforms. A suite of predictors were related to DO across time scales using artificial neural networks. At the diel scale, PAR explained most of the variability in DO signals. At sub-diel scales, temperature and wind largely explained variability in DO. However, the nature, strength, and time scale of the relationships between drivers and DO may be a function of lake size.