We used continuous measurements of dissolved oxygen (DO) in dark bottles to characterise patterns of the dark respiration rate (R-dark) for three marine phytoplankton monocultures and in natural-water samples from two marine coastal systems. Furthermore, patterns of ecosystem community respiration rate were determined from open-water changes in DO in a fjord and in a lake. We considered two models of R-dark to describe temporal changes in DO: constant R-dark and decreasing R-dark; increasing R-dark. In addition, the effect of incubation time on R-dark was investigated in bottle incubations. Constant R-dark was observed in short-term (12-h) bottle incubations in natural-water samples from two marine coastal systems. Declining R-dark was observed in marine phytoplankton cultures and open-water measurements in a lake. Increasing R-dark was observed in open-water measurements in a fjord, particularly during summer. Long-term (120-h) bottle incubations in natural-water samples showed an increase in R-dark after 48 and 72 h. We show that the conventional expectation of constant rates of respiration in darkness is far from typical, because non-linear changes are common under both controlled experimental conditions, as well as for open-water measurements of ecosystem respiration.