a, To obtain the growth curve, we sampled cells from the original culture (see Methods) at the specified time points (t = 0, 36, 68, 90, 114, 126, 140, 164, 288, 360 h). Cells were counted by imaging them inside the flip chamber, in the middle plane. Red dots represent the mean number of cells over time (between 76 and 93 frames were analysed for each data point). Error bars represent ± 1 s.d. from the mean. The growth curve of HA452 is shown in linear and semi-log scale (inset). The population density at carrying capacity was 3 × 105 cells ml−1, reached after ~2 weeks of incubation (= 360 h). The population’s intrinsic growth rate was found to be 0.4 day−1, as measured by fitting a logistic curve to the data (black line). In the inset, the shaded orange region shows the growth stage at which cells were harvested for experiments with exponential-phase cells (most experiments), while the shaded magenta region denotes the growth stage at which cells were harvested for experiments with starved cells. b, Upward bias, r, of HA452 cells over the course of a day, with time measured from midnight. For each data point the equilibrium vertical distribution was measured 30 min after termination of the overturning treatment, for both 10 flips (n = 560 cells) and 100 flips (n = 723) (for the control: 30 min after introduction of cells in the flipping chamber, n = 674). A positive upward bias denotes negatively gravitactic cells (that is, preferentially up-swimming). Gravitaxis can be seen to follow a diel cycle, even though the culture was kept under constant illumination. Flipping experiments consistently show a population split, leading to a reduction in the upward bias of the 10 flips and 100 flips treatments compared to the control treatment. The experiments were all conducted between 09:00 and 12:00, where the upward bias measured for the control cells presents the maximum stability.