The objective of this study was to identify behavioural adjustments leading to avoidance of hypoxia. Using the oxygen-sensitive species rainbow trout Oncorhynchus mykiss as a model, individual fish were recorded while moving freely between two sides of a test arena: one with normoxia and one with stepwise progressive hypoxia [80–30% dissolved oxygen (DO) air saturation]. The results demonstrated a gradual decrease in the total time spent in hypoxia starting at 80% DO air saturation. At this DO level, the avoidance of hypoxia could not be attributed to changes in spontaneous swimming speed, neither in normoxia nor in hypoxia. Reducing the DO level to 60% air saturation resulted in decreased spontaneous swimming speed in normoxia, yet the number of trips to the hypoxic side of the test arena remained unchanged. Moreover, data revealed increased average residence time per trip in normoxia at DO levels ≤60% air saturation and decreased average residence time per trip in hypoxia at DO levels ≤50% air saturation. Finally, the spontaneous swimming speed in hypoxia increased at DO levels ≤40% air saturation and the number of trips to hypoxia decreased at the 30% DO air saturation level. Thus, avoidance of the deepest hypoxia was connected with a reduced number of trips to hypoxia as well as decreased and increased spontaneous swimming speed in normoxia and hypoxia, respectively. Collectively, the data support the conclusions that the mechanistic basis for avoidance of hypoxia may (1) not involve changes in swimming speed during mild hypoxia and (2) depend on the severity of hypoxia.