372 | Nature | Vol 577 | 16 January 2020
Article
using direct manual observations, either by SCUBA divers or by an
observer counting the number of gridlines crossed by fish in aquaria^5 ,^26.
We filmed 582 individuals from 6 species across 3 years and quanti-
fied swimming activity in behavioural arenas using automated track-
ing software. Activity levels were assessed in adults and sub-adults
of 5 species in 2014, with 3 species showing no detectable effects of
CO 2 treatment (C. atripectoralis, P. amboinensis and P. moluccensis;
Fig. 4c–e and Extended Data Table 3; pooled across all species, n = 126,
P > 0.08). We found some evidence that activity was affected by high
CO 2 in D. aruanus, for which an interaction between CO 2 treatment and
standard length suggested that activity was elevated by approximately
59–92% in smaller individuals (<37 mm standard length) in high CO 2
levels (Fig. 4b, Extended Data Fig. 1a and Extended Data Table 3; n = 46,
P = 0.03). In A. polyacanthus, activity levels were increased by around
50% (P = 0.009) in fish acclimated to elevated CO 2 levels after control-
ling for a strong main effect of standard length (Fig. 4a, Extended Data
Fig. 1b and Extended Data Table 3; n = 16, P < 0.001).
When we extended our experiments in 2015 using captive-reared
juvenile A. polyacanthus with greater sample sizes and longer-duration
trials (Supplementary Information), the effect of CO 2 on activity disap-
peared (Extended Data Table 3; n = 66, P = 0.1). There was, however, a
weak interaction (P = 0.04); activity decreased in the high CO 2 fish (but
not controls) with increasing body size (Fig. 4a, Extended Data Fig. 1c
and Extended Data Table 3). In 2016, we conducted additional tests
of activity in wild fish across various life stages and found no effects
of CO 2 nor any interactions with body size in any of the three species
(n = 122 D. perspicillatus, n = 112 A. polyacanthus, n = 94 D. aruanus; all
CO 2 main effects P > 0.24; Fig. 4 and Extended Data Table 3).
Overall, we found that fish exposed to high CO 2 did not exhibit con-
sistently elevated activity levels compared with conspecifics under
control conditions (Fig. 4 ). Rather, we found that activity levels were
highly variable among individuals, increasing the risk of type-I errors
in experiments using small sample sizes^18 , and possibly in large-sample
experiments that rely on human observation rather than automated
video analysis^22 –^24.Behavioural lateralization
A tendency to favour the left or right side during behavioural activities
(that is, behavioural lateralization) is thought to be an expression of
functional asymmetries of the brain; this is important for tasks such as
schooling and predator avoidance^28. Elevated CO 2 has been reported
to reduce or abolish behavioural lateralization in fishes^17 ,^25 , presum-
ably as a result of brain dysfunction^17. Population-level lateralization is
present when a group of individuals collectively exhibits a side bias (the
mean number of turns to one side significantly exceeds 50%), whereas
individual-level lateralization is present when more individuals within
a tested group exhibit a side bias than expected by chance (based on
a binomial distribution with α = 0.5). Both types of lateralization are
independent of each other, but are not mutually exclusive (see Methods
and Supplementary Information for details).
Using a standard detour test in a double T-maze, we quantified
the effects of elevated CO 2 levels on behavioural lateralization using
175 fishes across four species in 2014 (C. atripectoralis, D. aruanus,
P. amboinensis and P. moluccensis). None of the species exhibited
population-level lateralization under control conditions (ExtendedLate life stagesEarly life stages
Time on one side (%)020406080100Control
High CO 2n = 37
n = 33020406080Time on one side (%)020406080100
n = 19
n = 16020406080n = 24
n = 25Time (min)020406080n = 42
Predator n =^39
cueSwitchPredator
cueSwitchPredator
cueSwitchPredator
cueSwitchPredator
cueSwitch
N=42
N=40abcden = 35
n = 24Fig. 2 | Damself ishes avoid predator chemical cues to the same degree
when exposed to present-day or end-of-century CO 2 levels irrespective
of life stage. a, b, A. polyacanthus. c, D. aruanus. d, e, D. perspicillatus.
a–e, Percentage of time (mean ± s.e.m.) that fishes from early life stages and
later life stages (including mid and late life stages) spent on one side of a two-
current choice f lume during experiments at the Lizard Island Research Station
in 2016. Control fish (maintained at approximately 520 μatm) in closed grey
circles, CO 2 -exposed fish (maintained at around 1,090 μatm) in open blue
circles. Fish were given 40 min to habituate to the f lume (during which
time their activity was quantified; see Fig. 4 ). Predator chemical cues
(C. cyanostigma) were introduced to one side of the f lume for 20 min and then
switched to the other side for another 20 min. n of biologically independent
animals are shown in the figure panels. See Extended Data Table 2 for statistics.
Fish illustrations by E. Walsh.