Table 2.

Outcomes experienced in each condition (expressed as proportion of total outcomes).

	Outcome	Danger condition (95% C.I.)	Safe condition (95% C.I.)
Experiment 1	Save	0.65 (0.58–0.71)	0.57 (0.50–0.64)
	Swerve	0.06 (0.05–0.08)	0.35 (0.29–0.40)
	Killed	0.29 (0.24–0.34)	0.08 (0.07–0.10)
Experiment 2	Save	0.71 (0.62–0.81)	0.65 (0.54–0.75)
	Swerve	0.06(0.04–0.08)	0.29 (0.20–0.38)
	Killed	0.22 (0.15–0.29)	0.07 (0.43–0.09)

C.I.: 95% confidence interval. “Save” refers to the per cent of dogs actively saved by the participant in each condition; this only occurred when the participant exerted sufficient force. Otherwise, the car might swerve (more likely in the safe condition), or it might kill the dog (more likely in the danger condition). Thus, “swerve” refers to the percent of dogs neither killed by the car nor actively saved. “Killed” refers to the percent of dogs killed by the car. Note the similarity in proportions of save/swerve/killed across the long (Experiment 1) and short (Experiment 2) versions of the task. This supports the idea that people choose a ‘golden balance’ level of fatigue which they sustain across the whole block. At that level, their cost of effort matches the average disasters averted for each trial. Note also that the higher effort exerted in the danger condition (see Results section 1) only partially mitigated outcomes in the high danger condition: participants suffered about 25% ‘dog deaths’ in high threat vs. 8% in low-threat. This is what one would expect if participants already operated in the low-threat condition under considerable fatigue. As we can expect cost to be proportional both to effort but also to fatigue (and hence, as fatigue increases with effort, to a power of effort greater than one), balancing losses with fatigue will have a reduced effect in moving from low to high threat.