Artificial Light at Night Affects Emergence from a Refuge and Space Use in Guppies

The use of artificial light at night (ALAN) is a major form of anthropogenic pollution affecting a wide range of environmental processes^1–9. The two main sources of light pollution are direct artificial light and skyglow. Direct light is usually high in intensity as it originates directly from street lighting, domestic and commercial light sources, or is reflected via surrounding surfaces¹. Skyglow, on the other hand, is the portion of artificial light that is scattered and reflected back to Earth within the atmosphere^10,11. Skyglow is a growing global phenomenon that is highly dynamic^12–14. Compared to direct light, it is dim and spatially homogenous on a small scale, but can be bright compared to natural celestial light sources, such as stars or even moonlight¹⁵ and can alter nightscapes located far from urban areas^16,17.

In terms of behavioural processes, ALAN can have distinct effects on the behaviour of animals during the nighttime¹. Many species use light for orientation during nighttime navigation. ALAN can thus alter the movement trajectories of individuals, either because individuals are directly attracted to light (‘flight-to-light’) or because natural light sources used for navigation are masked (e.g., by skyglow). These effects have been observed in insects^18,19, birds²⁰, amphibians^21,22, reptiles²³ and fish²⁴. Next to direct attraction, ALAN can alter the trade-offs animals face during the night. Especially, foraging benefits and predation risk can be strongly influenced by light conditions^1,25. Approximately 30% of all vertebrates and more than 60% of all invertebrates are nocturnal, and for these organisms, their temporally differentiated niche has been promoted by highly developed senses². Consequently, ALAN has different effects on diurnal and nocturnal species. Diurnal (and crepuscular) animals often become more night active with increasing ALAN (‘exploiting the night light niche’)^26–28, whereas nocturnal animals often become less night active^29–32.

Next to affecting behavioural processes during nighttime, ALAN can also alter behavioural processes during daytime. This has received especially attention in the context of activity patterns: ALAN has been shown to advance the onset of daily activity in European blackbirds, Turdus merula^33,34, great tits, Parus major²⁶ and to increase overall daily activity in zebra finches, Taeniopygia guttata³⁵. Moreover, several studies, predominantly in rodents, have shown that ALAN can impair cognitive abilities such as spatial learning and memory. Rats and mice exposed to ALAN performed poorly in the Morris water maze^36–38. In the diurnal rodent, the Nile grass rat, Arvicanthis niloticus, individuals exposed to three weeks of dim light during night showed impaired learning and memory in the Barnes maze³⁹. In zebra finches, constant ALAN also resulted in impaired learning and cognition³⁵. In contrast, in peafowls, Pavo cristatus, one-night light exposure did not affect problem-solving success⁴⁰.

Although the effects of ALAN on daytime behaviour have received some attention, most notably in the context of (the onset of) activity and cognition, the extent to which ALAN affects other important behavioural traits such as risk-taking or sociability, has received very little attention (but see³⁵). Such personality traits that are known to differ consistently among individuals can be linked to important life history traits of individuals, including survival and reproductive success^41–43. Thus, to understand how ALAN affects a broader range of personality traits can be important to further increase our understanding of the potential impact of ALAN on animal populations. Here we studied how ALAN affects several such personality traits, including emergence, activity, space use, sociability and memory ability in the Trinidadian guppy, Poecilia reticulata.

Animals living in aquatic habitats, including fish, are by no means exempt from light pollution^6,24,44. Skyglow, streetlights, floodlights at harbours and (fishing) boats can emit light onto the water surface that reaches higher intensities than the light of the natural full moon light⁴⁵ and can be detected even off-shore in marine ecosystems above^46,47 and under water⁴⁸. ALAN might not only alter fish communities⁴⁹ and physiology⁵⁰, but also fish behaviour⁵¹. Several studies have shown that fish become disoriented when swimming near lights^24,52. In Atlantic salmon, Salmon salar, ALAN disrupted the timing of migration (and thereby the social synchrony)^53,54. ALAN may also affect schooling behaviour because vision is often crucial for communication between school members^55,56. In walleye pollocks, Theragra chalcogramma, neighbour distance increased with decreasing light levels until it became too dark and schooling stopped⁵⁷.

To investigate the effects of ALAN on a range of important personality traits, we exposed guppies to different ALAN treatments: bright light (24 hrs bright light, ~5,000 lx), dim light (12 hrs bright light; 12 hrs dim light, ~0.5 lx) and control (12hrs bright light; 12 hrs dark). Light treatments took place for 10 weeks after which we repeatedly tested fish during the day on emergence, activity, space use, sociability and their ability to memorize the location of companion fish in a T-maze. We expected that ALAN would lead to reduced memory-based performance in the T-maze. In terms of behavioural traits, we had no a priori expectations since research linking ALAN to personality traits is largely absent.

We did not observe a significant interaction between light treatment and trial number on activity (both P > 0.2) nor a significant main effect of light treatment (both P > 0.2; Fig. 4C). There was a significant effect of sex on activity (−0.64 [−1.09, −0.22], P = 0.004; Suppl. Fig. 1C) with females showing higher average speed than males and a negative effect of trial number on activity (−0.04 [−0.07, −0.02], P < 0.001). There was no significant effect of body size (P > 0.2) or the side of the companions (P = 0.094) on activity. Activity was significantly repeatable (adj r [CI_lower, CI_upper] = 0. 45 [0.35, 0.55]).

Likewise, there was no significant interaction between light treatment and trial number on decision speed (both P > 0.2) nor a significant main effect of light treatment (both P > 0.15; Fig. 4D). Females made faster decisions than males (0.42 [0.14, 0.68], P = 0.006; Suppl. Fig. 1D). There was no effect of body size, trial number or side of the companions on decision speed (all P > 0.1). Decision speed was significantly repeatable (adj r [CI_lower, CI_upper] = 0. 27 [0.19, 0.37]).

Also for social time, there was no significant interaction between light treatment and trial number (both P > 0.2) nor a significant main effect of light treatment (both P > 0.1, Fig. 4E). There was an overall negative effect of trial number on social time (−10.15 [−17.55, −3.35], P < 0.001) implying that fish shoaled less with companion fish over the course of the experiment. Males were more likely to shoal with the companion fish after discovery than females (54.38 [1.81, 111.17], P = 0.028; Suppl. Fig. 1E). There was no effect of body size or side of the companions on social time (all P > 0.2). Social time was significantly, albeit lowly, repeatable (adj r [CI_lower, CI_upper] = 0. 21 [0.11, 0.34]).

The light at night treatments affected the diurnal behaviours of our focal fish in two ways. First, fish exposed to ALAN emerged quicker from the refuge than control fish. And whereas control fish increased their emergence time with repeated exposure, fish exposed to ALAN did not change their emergence time over the course of the experiment. Second, fish from the bright light treatment spent less time close to the wall of the arena and more time in the open (riskier) part of the arena than control fish. Importantly, the dim light treatment had a relatively low level of light intensity (~0.5 lx), only slightly above full moon light conditions and in the range of urban skyglow for overcast nights¹⁵. Fish from the bright light treatment showed a stronger reduction in emergence time than fish from the dim light treatment (Fig. 3). Most studies on ALAN compare only two ALAN levels (ALAN versus no ALAN), but the effects of ALAN on behavioural and ecological processes are in most cases not binary, but gradual, and recent studies call for more research on such dose dependent effects by considering a wider range of (realistic) ALAN levels⁹. Although our study only included three levels of ALAN and firm conclusions on such dose dependent effects may require a larger range of ALAN levels (e.g.²⁶), our results nonetheless suggest that increased levels of brightness may be associated with reduced emergence time in a dose dependent manner.

Faster emergence and increased use of open areas, most likely, would increase the risk of an individual. When startled, guppies quickly flee into hiding, searching the cover of floating leaves or small rock crevices. In our experiment, the cover of the start box is likely to serve a similar purpose, and quickly leaving the start box can thus be seen as an increased propensity to take risks, in line with previous studies using emergence time to quantify risk taking in this species⁶⁶ and other poeciliids^67,68. It is possible that ALAN increased activity levels at night and associated metabolic costs thereby altering the trade-off between foraging and risk during daytime, forcing hungrier animals to prefer food over safety. Harris et al. showed that guppies from high predation sites emerged sooner from shelter than guppies from low predation sites⁶⁶. Similar results were found in other poeciliids species^67,68, and repeatedly chasing fish with a net resulted in increased risk taking. Interestingly, more active, bold, and exploratory guppies survived longer when exposed to a piscivorous predator in a laboratory experiment⁶⁹. Higher risk taking in guppies thus seems, at least partly, to be an adaptive response to elevated dangerous and stressful environments. Our light treatments may have induced similar increased stress levels, resulting in increased risk taking. However, an alternative explanation is that fish in the (dark) control treatment, may have developed a preference for darker regions (as compared to fish from the light treatments), which would also result in longer emergence time and more time close to the wall. To disentangle both explanations, further risk taking tests independent of light conditions (e.g., startle test) would be required. Nonetheless, in both scenario’s, ALAN would lead to an increase in risk taking.

Activity showed significant repeatability, but was not affected by the light treatments. Previous studies have shown that ALAN can advance the onset of daily activity^26,33,34 but the effects on average activity during the day are less clear. Earlier work has reported that ALAN either increased³⁵ or decreased daytime activity²⁶. However, comparing the effects of ALAN on activity patterns during day- and nighttime separately suggests that the effects on nighttime activity are substantially larger²⁶ so we might expect guppies to show stronger activity responses to ALAN during nighttime. Likewise, we did not find an effect of ALAN on daytime sociability. We had no a-priori expectations regarding sociability as little to no studies have investigated this relationship. In fact, the consequences of ALAN on social and group processes is largely unknown⁷⁰. Future studies quantifying the degree (and onset) of activity and sociability both during day and nighttime and over longer periods, would be required to obtain a more complete picture of how ALAN affects such important behavioural processes.

Over the course of the experiment, guppies did not develop a preference for the choice arm containing the companion fish, implying that fish were either unable to learn the location of the companion fish, or they simply lacked the motivation to shoal with conspecifics. The latter is suggested, at least partly, by an overall decrease in sociability over the course of the experiment. Burns and Rodd found that guppies were able to perform above chance level in an arena containing an empty arm and an arm with conspecifics⁶⁰. However, they used only males as focal fish and females as conspecifics exploiting the well-known drive of male guppies to search for mating opportunities⁷¹. In our experiment, we also found that, once discovered, males were more likely to shoal with their companions than females (Suppl. Fig. 1E), suggesting that this experimental design is indeed better suited for male than female guppies. Given the overall absence of any learning, we were thus not able to test the effect of ALAN on learning ability.

Finally, there was a slight difference in average daytime light levels between the three treatments, with the light level of the bright light treatment (±5,202) being slightly higher than the dim (±4,652) and control (±4,607) treatment. Though we cannot exclude that this may have had some effect, the relative difference in light levels between the treatments during the day is very small (±10%) as compared to the relative difference in light levels during the night. It thus seems reasonable to assume that our observed behavioural differences are the result of nighttime differences in light regime.

To conclude, we have shown that ALAN reduced daytime emergence time and increased the time spent in the open in guppies. Moreover, whereas control fish increased their emergence time over repeated exposures; fish from the light treatments did not adjust their emergence time over repeated exposure. ALAN did not affect daytime activity, sociability or decision speed. Our results demonstrate that ALAN, next to affecting nocturnal behaviours, can also affect daytime behavioural processes, associated with risk-taking.